The domino gene of Drosophila encodes novel members of the SWI2/SNF2 family of DNA-dependent ATPases, which contribute to the silencing of homeotic genes

In Drosophila, cell proliferation mutants are often characterized by late larval or early pupal lethality (Gatti and Goldberg, 1991). It has been hypothesized that most, if not all, divisions needed to form a larva are ensured by maternal products stored in the egg. Indeed, the first 13 rounds of embryonic mitoses occur in the almost complete absence of zygotic transcription. Subsequent to the 17-18 embryonic replication cycles, larval growth is due uniquely to an increase in cell size, which is accompanied by chromosome polytenization in several tissues. In larvae, the only tissues that maintain mitotic capacity are the imaginal discs and histoblasts, the neuroblasts, the germ cells, and the blood cells. All of them are dispensable for larval viability (Shearn et al., 1971; Szabad and Bryant, 1982). Depending on the severity of the proliferation mutation, mutant larvae should exhibit defects in these structures. A corrolary to the assumption that early embryonic divisions are dependent on maternal products is that cell proliferation genes should have female sterile alleles. Zygotes homozygous for weak mutations in proliferation genes might have sufficient gene activity to develop to adulthood, but resultant homozygous females would not be able to provide sufficient product to the eggs to ensure embryonic mitosis.

Many genes necessary for cell proliferation have been characterized and, strikingly, a number of them correspond to members of the Polycomb (PcG) or Trithorax groups (TrxG) of homeotic regulators (Santamaria and Randsholt, 1998). In Drosophila, body segment identities along the anteroposterior axis are set up by the coordinated action of homeotic genes of the Antennapedia (ANT-C) and Bithorax (BX-C) complexes (Lewis, 1978; McGinnis and Krumlauf, 1992). The pattern of homeotic gene expression is initiated early in the embryo by the segmentation gene products (Harding and Levine, 1988). As these transcriptional regulators are expressed only transiently, a maintenance mechanism has to ensure correct homeotic gene expression throughout development. This mechanism is based on the activity of two groups of regulators: members of the TrxG, which are required to keep homeotic genes active in the appropriate cells, and members of the PcG, which keep them repressed in the cells in which they were not initially activated (reviewed by Orlando and Paro, 1995; Simon, 1995).

Members of the trxG have mostly been identified as suppressors of phenotypes caused by derepression of homeotic genes (Kennison and Tamkun, 1988). Loss-of-function mutations in PcG genes lead to ectopic expression of homeotic genes, and therefore to body segment transformations in the embryo and/or in the adult. Genetic analysis suggests the existence of 30 to 40 members of this group (Jürgens, 1985). A number of PcG proteins have been found associated in multiprotein complexes (Franke et al., 1992; Shao et al., 1999). A proposed role for such complexes is to drive local chromatin packaging and silencing of the targeted loci by preventing the binding of transcriptional activators (McCall and Bender, 1996; Zink and Paro, 1995).

We have recently described a mutation named domino (dom) that presents all the characteristics typical for loss-of-function mutations in proliferation genes (Braun et al., 1997). Generation of new alleles at the dom locus produced weak mutations that resulted in female sterility. Cloning of the dom gene showed that it encodes two novel members of the SWI2/SNF2 family of proteins, which are characterized by a highly conserved DNA-dependent ATPase domain (Eisen et al., 1995). This domain in Domino is most closely related to that of another member of this family in Drosophila, Brahma (Tamkun et al., 1992), which belongs to the TrxG. We show here that dom loss-of-function mutations enhance PcG and counteract TrxG mutations. We also show that dom can act as a weak suppressor of variegation. We propose that dom products function as global transcriptional regulators via interactions with chromatin.

The domino P-element strain was as described (Braun et al., 1997). Transposon excisions were induced with the w¹¹¹⁸; Sp/Cyo; μ2-3 Sb,ry⁵⁰⁶/TM6 stock. white− progeny were tested for reversion and lethality. Fly stocks are described in FlyBase (http://flybase.bio.indiana.edu/) and obtained from the Bloomington Stock Center, if not otherwise specified. The double mutants ash¹⁷trx^B11 and brm²trx^E2 were kindly provided by Allan Shearn, the 39C4 reporter fly line and the In(1)w^m4h line used to test PEV (Wallrath and Elgin, 1995; Reuter et al., 1982) by Laurent Perrin and the first chromosome 5F24 25.2 transgenic line (Zink and Paro, 1995) by Giacomo Cavalli. Oregon-R was used as a wild-type strain. All flies were raised at 25°C on standard cornmeal sucrose medium. Eggs were collected from cages on agar plates spread with yeast paste to obtain staged embryos for RNA preparation, and from egg-laying chambers for antibody staining.

The effect of dom mutation on white variegation in 39C4 and 5F24 25.2 flies was quantified by measuring the relative eye pigment content as previously described (Lyko et al., 1997; Reuter and Wolff, 1981).

General molecular procedures were carried out as described (Sambrook et al., 1989). Plasmid rescue was used to obtain the DNA that flanks the P-element insertion. A 3.8 kb genomic fragment from the rescue plasmid was used as a probe to screen a ‘random primed’ larval cDNA library. Subsequently, the positive cDNA fragments were used as probes to successively screen the same library and an embryonic oligo(dT) cDNA library (both in μZAPII, kindly provided by Carl Thummel) to recover a total of 30 cDNA clones corresponding to the two transcription units of the dom gene. Sequencing was carried out on both strands of selected cDNA clones, of the genomic DNA flanking the P-element insertion, and of PCR fragments produced to determine the size of the dom¹⁴ deletion.

RNA was isolated using Trizol^TM reagent (Gibco BRL), and poly(A)⁺ RNA was isolated by chromatography on oligo(dT) cellulose. RNA was fractionated in 1% agarose gels containing formaldehyde, transferred to nylon filters and probed with labelled DNA fragments. The 5μ RACE technique was applied to larval poly(A)⁺ RNA using the Marathon^TM cDNA Amplification Kit (Clontech).

Sequence similarity searches were performed at NCBI and BDGP, using Blastp and tBlastn programs on nr and dbest databases. The nucleotide sequence of dom transcripts is available from GenBank (Accession Numbers AF076776 and AF254373).

Third instar larvae were fixed in Carnoy’s fixative and embedded in paraffin. Subsequent histological sections were stained in Hansen’s Hematoxylin and Erythrosin.

To generate somatic clones of homozygous dom mutant cells, FRT (42D) pwn P[y⁺, ry⁺]44B (gift from P. Heitzler) was recombined onto dom chromosomes. yw P[ry⁺, hsFLP]; FRT (42D) pwn P[y⁺, ry⁺]44B, dom /CyO y⁺ males were crossed to yw; FRT (42D) virgin females, and their progeny were exposed to two successive 1 hour heat shocks at 38°C at the following stages: 20-hour-old embryos; first instar larvae; second instar larvae; third instar larvae; and pupae. Body parts of experimental (yw P[ry⁺, hsFLP] / yw; FRT (42D) pwn P[y⁺, ry⁺]44B, dom / FRT (42D)) and control (yw P[ry⁺, hsFLP] / yw; FRT (42D) pwn P[y⁺, ry⁺]44B/FRT (42D)) females were mounted and examined by bright-field microscopy. To generate somatic clones in a Minute context, yw P[ry⁺, hsFLP]; FRT (42D) M(2)S7 / CyO y⁺ males were crossed to yw; FRT (42D) pwn P[y⁺, ry⁺]44B, dom / CyO y⁺ virgin females. To generate germline mosaics, FRT (42B) was recombined onto dom chromosomes. yw P[ry⁺, hsFLP]; P[w⁺, FRT]^2RG13P[w⁺, ovoD1]^2R32X9/ CyO y⁺ males were crossed to yw; P[w⁺, FRT]^2RG13dom / CyO y⁺ virgin females (Chou and Perrimon, 1996), and their progeny were exposed to heat shock conditions as described above. Female progeny of the genotype yw P[ry⁺, hsFLP] / yw; P[w⁺, FRT]^2RG13P[w⁺, ovoD1]^2R32X9/ P[w⁺, FRT]^2RG13dom were mated to either yw; P[w⁺, FRT]^2RG13dom / CyO y⁺ males or for the control to yw; P[w⁺, FRT]^2RG13/ CyO y⁺ males. Females were scored for their ability to lay eggs. Ovaries were dissected and stained with DAPI for observation.

Polyclonal antibodies were raised in rats, and in rabbits, using either the DOM-A-specific fragment containing nucleotides 7417 to 8175, subcloned as a SacI/EcoRI fragment into the BamHI/EcoRI site of pGEX 3T (Pharmacia) or the DOM-B-specific fragment containing nucleotides 7145 to 7663 subcloned as a HindIII/XhoI fragment into the HindIII/XhoI sites of pGEX 3T, to generate GST-DOM fusions. The fusion products were expressed in Escherichia coli BL21, affinity purified over glutathione-agarose (Pharmacia) and injected into animals (four rats and two rabbits for each construct). After four injections, sera were tested for DOM-binding activity by ELISA and their specificity was tested by western blotting of DOM fusion proteins. The rat antibodies that were used in immunostaining experiments did not label dom hypomorphic larvae, which indicates that they do not crossreact with other proteins (data not shown).

Proteins were extracted from Drosophila embryos and analyzed by SDS-polyacrylamide electrophoresis and western blotting as described (Elfring et al., 1998). Gel filtration chromatography was performed on a Superose 6 10/30 FPLC column (Pharmacia; Elfring et al.,1998).

For immunostaining, dechorionated embryos, larvae or ovaries were fixed in formaldehyde. Larvae were subjected to 30 minutes of methanol treatment after fixation to permeabilize the brain. Rat anti-DOM-A and rat anti-DOM-B were used at a 1: 1000 dilution. Mouse anti-UBX (Ultrabithorax) and rabbit anti-SCR (Sex combs reduced) were used at 1:200 and 1:500 dilutions respectively. Histochemical detection was carried out using biotinylated rabbit anti-rat, horse anti-mouse or goat anti-rabbit immunoglobulins (IgG), and avidin-horseradish peroxidase (Vectastain Elite kit, Vector labs). Antibody staining of polytene chromosomes was done as described by G. Cavalli (http://www.igh.cnrs.fr/equip/cavalli) with a fixation time of 15 seconds. Rat anti-DOM-B was used at a 1:500 dilution, rabbit anti-PH (kindly provided by G. Cavalli) was used at a 1:300 dilution, and both were detected with Alexa fluor^TM 546 goat anti-rat IgG and Alexa fluor^TM 488 goat anti-rabbit IgG respectively at a 1:250 dilution.

The original dom mutation l(2)k08108 is due to the insertion of a P{lacW} element in the 57D11-12 region. It is uncovered by Df(2R)AA21 (56F9-17; 57D11-12) and by Df(2R)Pu-D17 (57B4; 58B). It was recovered in a screen for hematopoietic disorders (Braun et al., 1997) and causes a phenotype of prolonged larval development followed by lethality at pupariation. Homozygous larvae are devoid of imaginal discs and histoblasts, exhibit a reduced neuroblast region in the brain and, most strikingly, a melanized hematopoietic organ (the lymph glands), which is due to cell death within the organ. As a result there are no circulating hemocytes in mutant larvae (Braun et al., 1997). This original mutation was renamed dom¹.

By P-element excision, we obtained several independent phenotypic revertant lines, which demonstrates that the dom¹ mutation is due to the insertion of the transposon. To better understand the function of the dom gene, in the same experiment we generated a series of new alleles by imprecise P-element excision. Fourteen new non-complementing mutants were obtained that were classified according to the strength of the mutation, as defined in interallelic complementation assays (Table 1). The strongest allele, dom¹⁴, exhibits early larval lethality (first and second instar) and in mutant larvae, black lymph glands can occasionally be observed. Analysis of mutants that are transheterozygous for different dom alleles and for the deficiencies mentioned above showed that dom¹⁴ may represent an amorphic allele, as the phenotype of dom¹⁴/dom¹⁴ mutants is similar to that of dom¹⁴/Df(2R)AA21. The next class of mutations causes early pupal lethality after a prolonged third larval instar. The most severe defects observed in this class of alleles are similar to those described in dom¹ mutants, and are observed in dom², dom⁶ and dom¹¹. Weaker alleles (dom³, dom¹³ and dom¹⁵) present more extreme hematopoietic disorders as all lymph gland lobes undergo a massive overgrowth and reach a size more than 10 times that of wild-type lymph glands (Fig. 1A,B). The lymph gland overgrowth is probably due to the fact that proliferating hemocytes are unable to cross the basement membrane to reach the circulatory compartment. Consistently, no circulating hemocytes are found in these mutants. The cells within the lymph glands undergo blackening to various degrees, depending on the allele. There are no obvious brain defects and imaginal discs are occasionally reduced in size.

The remaining alleles are semi-lethals with less obvious larval defects (dom⁷, dom⁹, dom¹⁰, dom¹², dom¹⁸). Imaginal wing discs are occasionally mis-shapen. Lethality can occur at all pupal stages, but adult escapers are frequent. They often display wing defects such as crumpled or held-out phenotypes (Fig. 1C). Homozygous males are fertile, but female fertility is significantly reduced or abolished. The ovaries appear normal, but the eggs that are laid by dom⁹/dom⁹ homozygous females do not develop if fertilized by dom⁹/dom⁹ males. However, fertilization by wild-type males ensures partial embryonic development. In the case of dom¹² and dom¹⁸ alleles, a small percentage of such embryos hatch and can survive to adulthood.

The phenotypes associated with the dom loss-of-function mutations suggest that dom is necessary for correct cell proliferation. As neuroblast squashes of dom¹ larval brain did not reveal mitosis defects, we suspected an indirect effect on proliferation. To characterize the proliferative capacity of dom mutant cells, clonal analysis was performed using the FLP/FRT mitotic recombination system (Xu and Rubin, 1993). We induced mutant clones (for dom¹⁴, dom¹, dom³ and dom⁹) at all stages from embryo to pupa, and scored for dom clones in adult wings, legs, thorax and abdomen. Regardless of the time of induction, we were unable to recover homozygous dom¹⁴, dom¹ or dom³ mutant cells, even though wild-type twin clones of the expected size were produced in all experiments (Fig. 1D). Similar experiments performed with the weak dom⁹ allele yielded different results. dom⁹ clones were always found; however, the number of cells in these clones was inferior to that of the wild-type clones.

To discriminate between growth retardation and cell viability defects, we induced clones in a heterozygous Minute M(2)S7 genetic background (Theodosiou and Xu, 1998). Homozygous dom¹⁴, dom¹ and dom³ cells, induced at various developmental stages, were again unable to survive. dom⁹ clones could be observed in all experiments. Thus, even in a context where mutant clones are provided a growth-rate advantage, cells with strong loss-of-function dom mutations cannot survive.

Germline mosaic experiments indicated that dom is also required for oogenesis. By using the autosomal FLP-DFS technique (Chou and Perrimon, 1996), we generated germline clones carrying various dom mutations. With the strong alleles dom¹⁴, dom¹ and dom³, we did not observe egg-laying. Dissection of ovaries showed that oogenesis never proceeded beyond stage 5: development was arrested prior to vitellogenesis (not shown). With the weak allele dom⁹, few eggs were laid that exhibited a broad range of defects and were not viable.

The P-element responsible for the dom¹ mutation maps at position 57D11-12. We cloned the flanking genomic regions by plasmid rescue and used them to probe poly(A) RNA northern blots. 10.4 kb of genomic sequences 5μ to the insertion and 3.8 kb of 3μ sequence both revealed a transcript doublet around 10 kb (Fig. 2). We hypothesized that the P-element was inserted within the coding sequence that produces this transcript doublet. To further examine this possibility, we probed RNA of ovaries dissected from wild-type flies and from dom⁹/dom⁹ flies, with the 3.8 kb genomic fragment (Fig. 2). We confirmed that the partially excised transposon (see below) indeed impaired the transcription of the 10 kb doublet, as it is nearly abolished in dom⁹/dom⁹ ovaries. The same genomic fragment served as a probe to screen cDNA libraries in order to identify the full-length cDNAs. A total of 30 clones were recovered from both random-primed larval and oligo(dT)-primed embryonic cDNA libraries. The analysis of the 30 clones confirmed the existence of two dom transcripts (Fig. 3A) that share a common 6212bp 5μ portion and totally diverge in their 3μ sequence. The cDNA sequences that were obtained covered 10524bp for dominoA, the longest transcript, and 8682bp for the shorter form dominoB. The common transcription start site was mapped by a 5μ RACE experiment.

The comparison of the dom cDNA sequences with the corresponding genomic sequence from the Drosophila BDGP database allowed us to determine the gene structure (Fig. 3A). The analysis of the plasmid rescue fragments showed that the P-element responsible for the dom¹ mutation is located 1 bp downstream of the splice site for the first intron. No canonical TATA-box was found upstream of the transcription start site. The proposed start codon in exon 2 is immediately preceded by a Drosophila consensus sequence for translation start sites (C/AAAAC/A, Cavener, 1987). dominoA is encoded by 14 exons and dominoB by 11 exons. An alternative splicing takes place in exon 11, which generates the two forms of transcripts.

A developmental analysis of the domino transcripts (Fig. 2) shows that both forms are present throughout embryogenesis but are less abundant later in larvae. They are strongly re-expressed at the pupal stage. In adults they are not detected, except for a strong expression in ovaries.

Southern blotting and PCR experiments were used to analyze the different lines created by P-element excision. All mutations except dom¹⁴ result from internal transposon excisions inactivating the white gene. A 1649bp deletion of genomic DNA is found 3μ to the excised transposon in dom¹⁴ and removes exons 2 and 3 (Fig. 3A).

The open reading frames predict two proteins of 3202 (DOM-A) and 2498 (DOM-B) amino acids (Fig. 3B). The N-terminal common portion contains a Pro-rich (9%) domain and an acidic domain (42% Glu and Asp). Searching the available databases, we have found no significant similarity to known proteins, except for a 500 amino acid DNA-dependent ATPase domain that is highly similar to that of the SWI2/SNF2 family of helicases (Fig. 3C). In DOM, this domain is bi-partite and both portions are separated by a 451 amino acid gap. It is most closely related to the ATPase domains of the yeast putative YGPO helicase (37.8% similarity, James et al., 1995; Voet et al., 1997), of the Drosophila Brahma (BRM) protein (45.3%, Tamkun et al., 1992), and of the human BRM homolog BRG1 (SMARCA4 – Human Gene Nomenclature Database; 44.9%, Khavari et al., 1993). The C-terminal divergent part of DOM-

A bears a bipartite nuclear localisation signal and a long domain which contains numerous poly-Gln repeats. The C-terminal domain of DOM-B does not share such features but contains an additional acidic domain (43% Glu and Asp). Several putative PEST sequences are found in the N-terminal common domain.

We have raised polyclonal rabbit and rat antibodies against C-terminal portions specific for each DOM isoform (see Fig. 3B). Endogenous DOM proteins are detected in 0-12 hour embryonic extracts (Fig. 4) and their size is greater than 250 kDa, which is in agreement with calculated masses (275 kDa for DOM-B and 350 kDa for DOM-A). To determine whether DOM proteins are present in embryos as monomers or integrated in a complex, as is the case for the BRM and the yeast SWI2/SNF2 proteins, we examined the native molecular weight of DOM-A and DOM-B in embryos using gel filtration chromatography. Native protein extracts from Drosophila embryos were fractionated on a Superose 6 FPLC column, and the eluted fractions were assayed for DOM proteins by western blotting (Fig. 4). The native molecular weights of DOM-A and DOM-B are >2 MDa, indicating that both proteins are incorporated into large complexes. It is not, however, possible from these data to conclude whether both DOM proteins belong to the same or to different complexes. We observed with anti-DOM-B antibodies a labeled band with lower molecular weight (∼160 kDa), which was always associated with the 275 kDa protein and may correspond either to a degradation product of DOM-B, or to a cleaved form that is also present in the complex.

We used the anti-DOM polysera to examine the developmental expression pattern of DOM-A and DOM-B proteins (Figs 5,6). DOM-A staining only appears at the end of germ band extension (Fig. 5B), in a pattern restricted to the developing central nervous system. The protein accumulates in the nuclei during the whole development of the embryonic nervous system. DOM-A staining also appears in the peripheral nervous system at stage 13 (Fig. 5C,D). Strong DOM-B staining was detected in the nuclei during early embryogenesis, at syncytial and cellular blastoderm stages (Fig. 5E). From gastrulation to later embryonic stages, DOM-B is present in all nuclei (Fig. 5F,G,H). Wholemount in situ hybridization experiments of embryos with RNA probes specific for domA or domB confirmed these observations (data not shown).

During post-embryonic development, DOM-B was found in the brain (Fig. 6B), the imaginal discs (Fig. 6D), the lymph glands (Fig. 6F) and the salivary glands, whereas DOM-A expression is restricted to some brain regions (Fig. 6A) and to the photoreceptor precursor cells posterior to the morphogenetic furrow in the eye disc (Fig. 6C). It was also detected in the sensory organ precursors associated with imaginal discs (Fig. 6A). Both proteins are nuclear. Immunolocalization experiments run in parallel with both antibodies on dom³ larvae did not yield significant staining compared to Oregon-R larvae (data not shown), which confirms the specificity of the polysera.

Finally at the adult stage, DOM-B was found strongly expressed in the ovary, in follicle cells, nurse cells (Fig. 6G) and the oocyte. DOM-A is not expressed in the ovary even though the transcript is there.

Owing to its DNA-dependent ATPase domain, DOM belongs to the SWI2/SNF2 family of proteins, which are involved in functions as diverse as transcriptional activation, transcriptional repression, DNA repair, recombination and chromosome segregation. All these functions are achieved via interaction with chromatin. To test a possible effect of domino on chromatin structure, we analyzed its influence on position effect variegation (PEV). We examined the effect of the strong alleles dom² and dom¹⁴ on PEV associated with the chromosomal rearrangement In(1)w^m4h, which positions the white⁺ (w) gene close to the X-chromosome pericentric heterochromatin (Reuter et al., 1982), and with the nonclassical 39C4 PEV line P{white⁺} insertion where the w reporter gene is inserted near the second chromosome centromere (Wallrath and Elgin, 1995). As shown in Fig. 7, In(1)w^m4h and 39C4 variegations leads to a few clones of phenotypically wild-type cells in a w^- mutant background. This phenotype is moderately suppressed in dom¹⁴ and dom² transheterozygotes: the number of wild-type clones is increased. We conclude that domino is a weak suppressor of variegation and functions to maintain condensed chromatin structure.

DOM contains a DNA-dependent ATPase domain that is very homologous to that of the BRM protein, a member of the TrxG. dom mutations do not cause homeotic phenotypes by themselves. The cuticle of mutant embryos appears normal, possibly owing to the important maternal contribution. No misexpression of UBX and SCR could be observed in these embryos (data not shown). When the maternal supply is reduced (in eggs laid by dom⁹/dom⁹ females fertilized by wild-type males), the embryos exhibit such a broad array of defects that clear homeotic defects would be difficult to detect. Finally, we did not observe homeotic transformations in weak allele adult escapers. The occasional crumpled wing phenotype is not associated with ectopic UBX expression in imaginal discs and therefore does not represent a partial transformation of wing towards haltere (data not shown).

Given the features of the DOM proteins, their nuclear localization and the effect of dom mutations on PEV, we asked whether dom mutations could enhance or suppress homeotic transformations in PcG mutants. We tested alleles of several PcG members in combination with dom alleles (Table 2). All transheterozygous dom/+; PcG/+ adults are viable. Clear enhancement of PcG phenotypes was observed with dom/+ flies carrying one copy of E(z)⁶⁰ (Wu et al., 1989), mxc^G43 (Santamaria and Randsholt, 1995), Pc^K (Zink and Paro, 1989) or ph⁴¹⁰ (Dura et al., 1987) mutations. The transheterozygotes exhibit an increased number of ectopic sex comb teeth on second and third legs and/or an increased number of transformed individuals compared to single PcG mutants (Table 2). In the case of the ph⁴¹⁰ allele, the addition of dom¹⁴ significantly increased the number of individuals with partial A6 to A7 tergite transformation in males. Whenever it was assayed, we always found a good correlation between the PcG synergistic effect and the strength of the dom allele (not shown). Interaction was weak or non-significant with Psc and Scm alleles (Adler et al., 1989; Breen and Duncan, 1986).

PcG proteins have been shown to exert their repressive action through interaction with DNA regulatory sequences termed PcG response elements (PREs), which were identified in the ph gene (Fauvarque and Dura, 1993), in BX-C and ANT-C (Chan et al., 1994; Simon et al., 1993; Zink et al., 1991). To test if silencing imposed by Fab-7 on w is dependent on dom, as is the case for several PcG genes (Cavalli and Paro, 1999), we used the first chromosome 5F24 25,2 transgenic line where the PRE-containing Fab-7 sequence from the BX-C is fused to the w gene (Zink and Paro, 1995). The eye color of flies carrying this transgene was compared with that of flies carrying the same transgene and heterozygous for dom² or dom¹⁴ mutations. The darker eye color in dom mutant background (Fig. 8) is interpreted as derepression and confirms a role of dom in creating a repressed chromatin structure.

We further tested whether dom is able to counteract the effect of TrxG mutations. For this we crossed dom¹⁴ in trans to the single mutants brm², ash-1^B1 or trx^E2 and to the double mutants ash¹⁷trx^B11 or brm²trx^E2 (Gindhart and Kaufman, 1995; Kennison and Tamkun, 1988; Tamkun et al., 1992; Zorin et al., 1999). The single mutants display homeotic transformations, such as transformation of abdominal segment A6 to the more anterior segment A5, of haltere to wing, and partial transformation of first to second legs. The double TrxG mutants exhibit an enhanced penetrance of these phenotypes. We could not detect a variation in the number of sex comb teeth in mutant first legs when we added a copy of dom¹⁴. However, the transformation of A6 into A5 was significantly decreased by the addition of a copy of dom¹⁴ (Table 2). Similarly, the haltere to wing transformation of brm trx flies was totally abolished in dom¹⁴ heterozygotes.

We examined the distribution of DOM proteins on salivary gland polytene chromosomes in third instar larvae by immunofluorescence microscopy. DOM-B binds strongly to a number of discrete sites and consistently stains more diffusely stretches of undercondensed chromatin, i.e. interband regions where DAPI staining is reduced (Fig. 9A-C). The chromocenter region is only weakly labeled. The same pattern was observed with different anti-DOM-B rat or rabbit polysera, but not with preimmune sera. DOM-A antibodies did not stain chromosomes, which reflects the fact that this isoform is not expressed in salivary glands.

To address whether DOM-B and PH co-localize on salivary gland chromosomes, we examined simultaneously their binding sites in double immunostaining experiments (Fig. 9D-F). PH binds to some 100 sites which perfectly overlap those labeled with PC and PCL antibodies (Franke et al., 1992; Lonie et al., 1994). We show that there is only limited overlap between PH and DOM-B strong binding sites (some 10 sites per chromosome set), and that PH can also co-localize with weak DOM-B sites.

We have described the cloning and characterization of the dom gene of Drosophila, which encodes two protein isoforms with an ATPase domain typical for the SWI2/SNF2 family (Eisen et al., 1995; Pazin and Kadonaga, 1997; Tamkun, 1995; Vignali et al., 2000). The sequencing of the Drosophila genome (Adams et al., 2000) has highlighted the existence in this model organism of 17 members of this family, only eight of which have been described so far: brm (Tamkun et al., 1992); Iswi (Elfring et al., 1994); 89B helicase (Hel89B – FlyBase; Goldman-Levi et al., 1996); kismet (Daubresse et al., 1999); lodestar (Girdham and Glover, 1991); dMi-2 (Mi-2 – FlyBase; Kehle et al., 1998); Chd1 (Stokes et al., 1996); and dmRAD54 (okra – FlyBase ; Kooistra et al., 1997). The SNF2/SWI2-related family, extensively investigated in yeast, flies and mammals, functions in various aspects of DNA maintenance, chromosome segregation or transcriptional regulation, and is generally active through chromatin remodeling. Two of the best studied members in Drosophila are BRM and ISWI: both were shown to be components of large chromatin-remodeling complexes (Ito et al., 1997; Papoulas et al., 1998; Tsukiyama et al., 1995; Varga-Weisz et al., 1997).

The initial dom mutation is due to the presence of a P-element in the first intron of the gene, which is likely to interfere with transcription efficiency. The reduction of zygotic transcription in mutant embryos could not be tested, however, owing to the significant maternal supply and the lethality of dom germline clones. The maternal contribution of DOM is sufficient for embryonic development as individuals homozygous for the null allele dom¹⁴ survive until first or second larval instar. All hypomorphic dom alleles, which we have analyzed either have defects in larval proliferating structures/tissues, or produce homozygous females with reduced fertility. Altogether these phenotypes are reminiscent of mutations in a proliferation gene (Gatti and Goldberg, 1991). Clonal analysis showed that in addition, dom is required for cell viability, and that this requirement is constant during development. dom is also necessary to ensure proper oogenesis. A subset of SWI2/SNF2 genes such as ISWI (Deuring et al., 2000), brm (Elfring et al., 1998) and dMi-2 (Kehle et al., 1998) are similarly required for proliferation, cell viability and oogenesis.

The ATPase domain clearly relates DOM to the SWI2/SNF2 family of proteins. In this domain, the strongest similarity is found with that of BRM; however in DOM there is an unusually large insertion between motifs IV and V of the domain (see Fig. 3). Such a gap does not exist in BRM, but is found in the yeast YGPO helicase, the function of which has not yet been investigated (James et al., 1995; Voet et al., 1997). Strikingly, it also exists in a human molecule recently identified: SRCAP (Snf2-related CBP activator protein, Ghosh et al., 2000; Johnston et al., 1999), which, in addition to the bipartite ATPase domain, contains highly charged regions. We propose that the presence of such a gap within the ATPase domain defines a new subfamily of SNF2-related helicases, and it should be investigated whether this structural peculiarity is associated with comparable functions.

Outside the ATPase domain, DOM proteins present features that are found in a number of PcG and TrxG proteins. For example, Gln-repeats are reported in PH (DeCamillis et al., 1992), ASX (Sinclair et al., 1998), CCF (Kodjabachian et al., 1998), E(PC) (Stankunas et al., 1998), PCL (Lonie et al., 1994), PC (Paro and Hogness, 1991) and the GAGA factor (Farkas et al., 1994), and it was proposed that their role is to mediate protein-protein interactions (Emili et al., 1994; Pinto and Lobe, 1996). In addition, acidic-rich and Pro-rich regions are often encountered in transcriptional activators (Mitchell and Tjian, 1989).

The analysis of the distribution of both DOM isoforms shows that they do not co-localize. Both are nuclear, but whereas DOM-B is rather ubiquitous, DOM-A is restricted to the developing nervous system. The specific function of each isoform is not yet understood.

While most of the known PcG mutants exhibit phenotypes consistent with misexpression of homeotic genes, a few members are included in the group only because of their ability to enhance homeotic phenotypes when placed in trans of other PcG mutations. Su(z)2, E(Pc) and dMi-2 share such characteristics (Adler et al., 1989; Kehle et al., 1998; Sato et al., 1983; Wu and Howe, 1995). When dom mutations are in trans with a number of PcG mutations, the homeotic phenotypes are enhanced: this interaction suggests that dom also exerts a repressive function on homeotic genes. Consistent with this, dom counteracts the effects of brm, trx and ash1 mutations, all belonging to the TrxG genes, which are homeotic gene activators.

We show that the repressive function of dom can be mediated via the Fab-7 element of cis-regulatory DNA of the BX-C (Zink and Paro, 1995), an element that is the target of either PcG or TrxG proteins for the maintenance of repressed or active gene expression, respectively. An additional connection between dom and PREs has recently been provided by independent studies in J. M. Dura’s laboratory (A. Boivin and J. M. Dura, personal communication). They recovered dom mutations in a genetic screen designed to identify strong interactors with the ph PRE.

Despite significant genetic interactions with the PcG gene ph, DOM-B and PH do not share many binding sites on larval polytene chromosomes. We have shown that DOM proteins are present in large complexes at the embryonic stage, raising the possibility that one of them may correspond to the PRC1 complex, a similarly large complex of PcG proteins that includes PH and PC (Shao et al., 1999). However, all high molecular weight components of the PRC1 complex have now been sequenced and none of them corresponds to DOM (A. Sourin and R. Kingston, personal communication). The DOM proteins are thus not stably associated with PRC1 but this does not exclude the possibility that they are able to interact transiently with the PRC1 complex. The molecular features and the chromosomal localization of DOM-B suggest that its role could be to alter locally the chromatin conformation, for example at PREs, and subsequently facilitate interactions between different components necessary for efficient gene silencing.

Our data suggest a functional convergence between dom and PcG members, however dom cannot be defined as a new typical PcG gene. The effect of dom on homeotic regulation could be indirect if dom is a regulator of the expression of PcG genes, which would be in agreement with its action on the ph PRE. In this case, the observed genetic interactions could be explained by the dom transcriptional regulation of genes encoding homeotic repressors or activators. A second possibility to consider is that dom could also exert an effect directly on only a subset of PcG targets.

dom was shown to be a weak suppressor of PEV, which suggests its role in maintaining extended heterochromatization. This effect is however not as robust as that on the Fab-7 element, suggesting that the primary target of DOM is not centromeric heterochromatin. This hypothesis is supported by the fact that DOM-B protein does not extensively localize at the centromere of polytene chromosomes.

As dom acts both on homeotic regulation and on PEV, we conclude that this gene exerts multiple repressive functions on gene expression that may all be linked to a structural effect on chromatin organization.

The proposition that dom somehow participates in chromatin modification is based on genetic data and on its similarity to known SWI2/SNF2 proteins for which interactions with chromatin have been biochemically established. A direct interaction of DOM with chromatin, and DNA, remains to be formally demonstrated. Chromatin remodeling complexes have generally been studied in vitro as factors that promote gene activation; however, recent reports indicate that ATP-driven chromatin remodeling factors facilitate not only transcriptional activation, but also repression (reviewed by Tyler and Kadonaga, 1999). In addition, a given complex can exert both activating and repressive functions. Indeed, a DNA microarray analysis of global gene expression in yeast has revealed that in SWI2/SNF2 mutants, among the 6% of yeast genes with altered expression, the majority is upregulated in mutants (Holstege et al., 1998; Sudarsanam et al., 2000). In our study we show that dom has a repressive effect on homeotic genes via interactions with PcG and TrxG members. This negative regulation does not exclude that dom could also exert activating functions on other genes that remain to be identified.

DOM-B binds to a fairly large number of euchromatic sites on polytene chromosomes, which is consistent with a regulatory role on transcription. The weaker but reproducible binding of DOM to less condensed chromatin may reflect a more general effect on chromosome regions that might be transcriptionally active. This hypothesis has to be investigated. As a rule, chromatin remodeling complexes contain only one ATPase subunit that provides energy for the nucleosome rearrangement machinery. Thus BRM belongs to the SWI/SNF complex, ISWI to NURF, CHRAC and ACF, and dMi-2 to NuRD (Brehm et al., 2000; Vignali et al., 2000). A challenge for the future will be the identification of the partners of DOM proteins in their respective complexes. Eventually the knowledge of the developmental and tissue-specific composition of the various chromatin remodeling complexes will help us understand the specific functions and phenotypes related to the corresponding genes. Likewise, the identification of dom target genes should provide clues as to how we can link the chromatin modification concept with the effect of loss-of-function dom mutations on cell proliferation.

The authors thank Jules Hoffmann for his support and interest in this study. They are very grateful to Roger Karess for neuroblast squashes, to René Lanot for histological preparations, to Marie Paschaki for help with antibody tests, and to Pedro Santamaria, Pascal Heitzler, Laurent Perrin, Maria Capovilla, Julien Royet, Dominique Ferrandon and Giacomo Cavalli for valuable discussion, advice and fly stocks. They thank Giacomo Cavalli, Maria Capovilla and ThomasC. Kaufman for the kind gift of anti-PH, anti-UBX and anti-SCR antibodies, Allen Shearn for the double trxG mutant stocks, and Carl Thummel for cDNA libraries. The research of M. L. R., A. B. and M. M. was funded by the Centre National de la Recherche Scientifique, the Association pour la Recherche sur le Cancer and the Ligue Nationale et Régionale contre le Cancer. J. W. T. and O. P. were supported by a grant from the National Institutes of Health (GM49883) to J. W. T.