Toutatis, a TIP5-related protein, positively regulates Pannier function during Drosophila neural development

Transcriptional activation of many developmentally regulated genes is mediated by proteins binding to enhancers scattered over the genome, raising the question on how long-range activation is restricted to the relevant target promoter (Bulger and Groudine,1999; Dorsett,1999). Numerous studies have highlighted the essential role of boundaries, which maintain domains independent of their surrounding(West et al., 2002).

The patterning of the large sensory bristles (macrochaetae) on the thorax of Drosophila melanogaster is a powerful model to study how enhancers communicate with promoters during regulation of gene expression. Each macrochaeta derives from a precursor cell selected from a group of equivalent ac-sc-expressing cells, the proneural cluster(Gomez-Skarmeta et al., 2003). ac and sc encode basic helix-loop-helix proteins (bHLH) that heterodimerize with Daughterless (Da) to activate expression of downstream genes required for neural fate. Transcription of ac and scin the different sites of the imaginal disc is initiated by enhancers of the ac-sc complex (Gomez-Skarmeta et al., 1995; Garcia-Garcia et al., 1999) and the expression is maintained throughout development by autoregulation mediated by the (Ac-Sc)-Da heterodimers binding to E boxes within the ac-sc promoters(Martinez and Modolell, 1991; Van Doren et al., 1991; Van Doren et al., 1992). Each enhancer interacts with specific transcription factors that are expressed in broader domains than the proneural clusters and define the bristle prepattern(Gomez-Skarmeta et al., 2003). Thus, the GATA factor Pannier (Pnr)(Ramain et al., 1993; Heitzler et al., 1996) binds to the dorsocentral (DC) enhancer(Garcia-Garcia et al., 1999)and activates proneural expression to promote development of DC sensory organs. The Drosophila LIM-domain-binding protein 1 (Ldb1), Chip(Morcillo et al., 1997)physically interacts both with Pnr and the (Ac-Sc)-Da heterodimer to give a multiprotein proneural complex which facilitates the enhancer-promoter communication (Dorsett, 1999; Ramain et al., 2000).

Chromatin plays a crucial role in control of eukaryotic gene expression and is a highly dynamic structure at promoters(Näär et al., 2001). In Drosophila, the polycomb (Pc) group and the trithorax (Trx) group proteins are chromatin components that maintain stable states of gene expression and are involved in various complexes(Simon and Tamkun, 2002). The Pc group proteins are required to maintain repression of homeotic genes such as Ultrabithorax, presumably by inducing a repressive chromatin structure. Members of the Trx group were identified by their ability to suppress dominant Polycomb phenotypes. We recently provided evidence that enhancer-promoter communication during Pnr-driven proneural development is negatively regulated by the Brahma (Brm) chromatin remodelling complex(Heitzler et al., 2003; Treisman et al., 1997; Collins et al., 1999),homologous to the yeast SWI/SNF complex.

Here, we present Toutatis (Tou), a protein that associates both with Pnr and Chip and that positively regulates activity of the proneural complex encompassing Pnr and Chip during enhancer-promoter communication. Tou has been previously identified in a genetic screen for dominant modifiers of the extra-sex-combs phenotype displayed by mutant of polyhomeotic(ph), a member of the Pc group in Drosophila(Fauvarque et al., 2001). Tou shares functional domains with Acf1, a subunit of both the human and Drosophila ACF (ATP-utilizing chromatin assembly and remodelling factor) and CHRAC (chromatin accessibility complex)(Ito et al., 1999), and with TIP5 of NoRC (nucleolar remodelling complex)(Strohner et al., 2001). Hence, Tou regulates activity of the proneural complex during enhancer-promoter communication, possibly through chromatin remodelling. Moreover, we show that Iswi (Deuring et al., 2000), a highly conserved member of the SWI2/SNF2 family of ATPases, is also necessary for activation of ac-sc and neural development. Since Iswi is shown to physically interact with Tou, Pnr and Chip, we suggest that a complex encompassing Tou and Iswi directly regulates activity of the proneural complex during enhancer-promoter communication,possibly through chromatin remodelling.

Sequences encoding the DNA binding domain (DBD) of Pnr (Arg141-Thr280) were PCR amplified and inserted in frame with the LexA DBD(LexA^DBDPnr^DBD) into pBTM116. The resulting fusion protein was used as bait to screen an embryonic cDNA expression library made into pASV4 carrying the VP16 activation domain (VP16^AD) (Beackstead et al., 2001). Yeast two-hybrid screening followed the method of Le Douarin et al. (Le Douarin et al.,2001).

β-Galactosidase assays on transformants of the L40 yeast strain were carried out as described previously(Seipel et al., 1992). VP16^AD and LexA^DBD fusion proteins expressed throughout the current study did not activate the β-galactosidase reporter in L40 cells. Expression of LexA and VP16 fusion proteins were analysed by western blot using anti-LexA- or anti-VP16-specific antibodies.

Sequences encoding the Tou domains (TouA: Thr546-Met1415, TouB:Met1-Ala600, TouC: Thr546-Glu1120, TouD: Ala1089-Met1415, TouE:Pro1303-Iso1866, TouF: Pro1826-Val2086, TouG: Asp2059-Asn2615, TouH:Pro2446-Leu2857, TouI: His2831-Ser3116, TouJ: Thr546-Glu953, TouK:Leu954-Glu1120, TouL: Ala1089-Pro1238, TouM: Gly1239-Asn1307 and TouN:Pro1308-Met1415) were PCR amplified and inserted in frame with VP16^AD into pASV4. The fragment encoding TouA was also inserted into pXJB, which allows expression in transfected Cos cells of tagged proteins carrying the B epitope of the oestrogen receptor at their N terminus(Ramain et al., 2000). The expression vector pXJPnr, encoding the truncated wild-type Pnr has been described previously (Haenlin et al.,1997). The expression vector encoding the full-length Chip carrying a N-terminal Flag epitope (F-Chip) has also been described previously(Ramain et al., 2000). pXJF-Iswi encodes a full-length Iswi carrying an N-terminal Flag epitope. The expression vectors encoding the N-terminal domain of Chip(B10-N^TChip) and the C-terminal domain of Chip(B10-C^TChip) carrying at their N terminus the B epitope of the oestrogen receptor were described by Ramain et al.(Ramain et al., 2000). Sequences encoding the N^T Chip, the C^T Chip and Iswi were also inserted in frame with the LexA^DBD into pBTM116.

The following mutant stocks were used: pnr^D1,pnr^VX1 (Ramain et al.,1993), Chip^E(Ramain et al., 2000) and osa⁶¹⁶ (Treisman et al., 1997). tou¹ and tou²(Fauvarque et al., 2001). tou^E44.1 was generated by imprecise excision of EY08961 (Bellen et al.,2004). Misexpression of full-length transgenes was achieved using the UAS/Gal4 system, using the following stocks: EP622, an EP target element from the collection of P. Rorth (Rorth,1996) allowing overexpression of tou,UAS-Iswi^K159R (Deuring et al., 2000). Thirty hemithoraces were examined for each of the genotypes presented in Fig. 2. For each genotype, the phenotype was found to be remarkably similar from fly to fly.

Cell transfections, protein extracts preparations, immunoprecipitations and western blot analysis follow the method of Haenlin et al.(Haenlin et al., 1997).

The method of Ramain et al. (Ramain et al., 2000) was used for GST pull-down assays.

Wing discs were stained using the method of Cubadda et al.(Cubadda et al., 1997). A101 contains a lacZ gene inserted into the neuralized locus(Boulianne et al., 1991) and is found by specific staining in all sensory organs precursors(Huang et al., 1991). The transgenic strain DC-aclacZ has been described previously(Garcia-Garcia et al., 1999). In Fig. 6, each staining was for one hour at 22°C and a representative imaginal disc is shown for each genotype. The stainings were done at different stages of the study but they were performed using the same batch of reagents. In addition, control stainings of imaginal discs from wild-type larvae were included in each experiment to allow comparisons between the different genotypes.

Previous experiments in vertebrates(Tsang et al., 1997) and in Drosophila (Haenlin et al.,1997; Ramain et al.,2000) have demonstrated that the function of GATA factors during development is regulated by dimerization of their DNA binding domain (DBD)with cofactors. Hence, we used the yeast two-hybrid system to identify Pnr-interacting proteins and we used a bait in which the DBD of Pnr was fused to the DBD of LexA. A screen of 10⁶ primary transformants of a Drosophila embryonic cDNA expression library yielded 35 potential positive clones. Among the different clones, we identified a domain of Tou,designated as TouA (Fig. 1B). Tou is predicted to encode a 3116 amino acids sequence protein that contains a MBD domain, a DDT domain, a WAKZ motif, two PHD zinc fingers and a C-terminal bromodomain (Fig. 1). Tou shares these functional domains with members of the WAL family of chromosomal proteins, including human TIP-5 (Strohner et al., 2001), Drosophila Acf1(Ito et al., 1999) and human WSTF (Williams syndrome transcription factor)(Lu et al., 1998). This observation suggests that these proteins may have related functions.

Pnr promotes development of DC sensory organs(Garcia-Garcia et al., 1999; Ramain et al., 2000). Tou physically interacts with Pnr in yeast and tou mRNA was found to be ubiquitously expressed in the wing imaginal disc (data not shown). tou mRNA is expressed in the wing pouch, in agreement with the wing defects associated with tou alleles(Fauvarque et al., 2001) and in the dorsal-most region of the thorax, covering the site of appearance of the DC bristles and the domain of pnr expression. Hence, we asked whether Tou is also required during development of DC sensory organs.

tou¹ corresponds to a PlacW transposon insertion within the large intron of tou at 836 bases pairs (bp) upstream of the second exon of tou that contains the ATG translational start(Fig. 1A)(Fauvarque et al., 2001). tou² was generated by imprecise excision of the tou¹ insertion, in which 200 bp of the intron was replaced by a 7 kb sequence from the P-element. Oregon wild-type flies have 2.00 DC bristles per heminotum (Fig. 2A) We found that both tou¹ and tou² give rise to homozygous escapers lacking DC bristles(Fig. 2B) (1.92±0.05 and 1.4±0.05 DC bristles/heminotum, respectively). It suggests that tou is necessary for DC neural development. tou¹and tou² are associated with molecular lesions affecting intronic sequences and display an abnormal bristle phenotype, suggesting that tou¹ and tou² disrupt functioning of an important regulatory element. Accordingly, the level of touexpression is strongly reduced in tou² thoracic discs(data not shown), suggesting that tou² is likely to be a loss of function allele. This hypothesis is reinforced by the fact that the loss of DC bristles is more severe for tou¹/Df(2R)en-SFX31and tou²/Df(2R)en-SFX31 (1.23±0.09 and 1.15±0.06 DC bristles/heminotum, respectively) than for tou¹ and tou² (1.92±0.05 and 1.4±0.05 DC bristles/heminotum, respectively). Df(2R)en-SFX31is a deficiency that spans tou and other genes between 48A1 and 48B5. Moreover, the requirement for tou to promote DC neural development is further demonstrated by analysis of tou^E44.1. tou^E44.1 was generated by imprecise excision of the transposon insertion in EY08961 (Fig. 1A) (Bellen et al.,2004) and encodes a mutant protein lacking the C terminus,containing the PHD fingers and the bromodomain. Since the tou^E44.1 flies lack DC bristles (1.57±0.13 DC bristles per heminotum) and because PHD fingers and the bromodomain are believed to play important functions, we conclude that tou^E44.1 is a loss-of-function allele and that Tou is required to promote neural development.

We next investigated genetic interactions between the pnr and tou alleles. We made use of tou¹ and tou², which are loss-of-function alleles(Fig. 1) and addressed whether they interact with pnr^D1. pnr^D1 encodes a mutant protein carrying a single point mutation in the DBD that disrupts interaction with the Ush antagonist(Cubadda et al., 1997; Haenlin et al., 1997). Hence,Pnr^D1 constitutively activates ac-sc at the DC site,leading to excess DC sensory organs(Ramain et al., 1993; Haenlin et al., 1997). pnr^D1/+ flies have 3.34±0.1 DC bristles per heminotum (Fig. 2C) and we observed that this excess is suppressed when tou function is simultaneously reduced. For example, tou²;pnr^D1/+ flies have 2.17±0.08 DC bristles per heminotum(Fig. 2D). Conversely, pnr^VX1 encodes a truncated protein, lacking the domain of interaction with Chip, which is required for enhancer-promoter communication during ac-sc expression and development of DC sensory organs(Ramain et al., 2000). Hence, pnr^VX1 displays a loss of DC bristles (1.86±0.03 DC bristles/heminotum) and this loss is aggravated when tou function is simultaneously reduced. tou²/+; pnr^VX1/+ flies display 1.57±0.05 DC bristles/heminotum. Chip^Ecarries a single point mutation that disrupts the enhancer-promoter communication (Ramain et al.,2000) and is associated with reduced proneural expression at the DC site and loss of sensory organs (1.6±0.07 DC bristles/heminotum)(Fig. 2E). The loss of DC bristles was accentuated when tou function was simultaneously reduced as in tou²Chip^E flies, which display 1.11±0.05 DC organs/heminotum (Fig. 2F). These observations indicate that Tou probably cooperates with Pnr and its Chip cofactor during development of the DC bristles.

We also analysed the effects of overexpressed tou on neural development. We made use of the UAS/Gal4 system and tou was overexpressed in the dorsal compartment of the disc Gal4-ap^MD544(Rincon-Limas et al., 1999). Gal4-ap^MD544/EP622 flies have excess DC sensory organs (2.78±0.12 DC bristles/heminotum)(Fig. 2G). Moreover, the loss of DC bristles associated with Chip^E is suppressed when Tou is overexpressed (Fig. 2H). Indeed, Chip^E ap^Gal4/Chip^E EP622flies have 2.23±0.08 DC bristles/heminotum.

The development of extra DC sensory organs, revealed by phenotypic analysis, was compared with the positions of DC bristle precursors detected with a lacZ insert, A101, in the neuralized gene that exhibits staining in all sensory organs. In EP622/Gal4-ap^MD544, additional DC precursors are seen that lead to excess DC bristles (Fig. 2I,J). Hence, we conclude that Tou cooperates with Pnr and Chip during neural development.

TouA physically interacts with the DBD of Pnr in yeast. Since the C terminus of Pnr also mediates physical interactions with cofactors(Ramain et al., 2000), we then investigated whether TouA interacts with the C terminus of Pnr in yeast(Fig. 3A). Expression vectors encoding either the C terminus of Pnr fused to the LexA DBD(LexA^DBDPnr^CT) or the unfused LexA DBD(LexA^DBD) were introduced into the L40 yeast strain together with the vector for the unfused VP16 activation domain (VP16^AD) or for TouA fused to VP16^AD (VP16^ADTouA). L40 cells contain a lacZ reporter driven by eight LexA binding sites. Protein extracts were prepared from L40 transformants grown in liquid medium and were assayed for β-galactosidase activity (Fig. 3B). No reporter activity was observed in extracts made from yeast expressing LexA^DBDPnr^CT/VP16^AD,LexA^DBD/VP16^ADTouA or LexA^DBD/VP16^AD. In contrast, a robust activation of the reporter was detected with yeast expressing LexA^DBDPnr^CT/VP16^ADTouA, indicating that TouA associates with the C terminus of Pnr (Fig. 3B). Then, both the DBD and the C terminus of Pnr physically interact with TouA. Accordingly, we found that TouA interacts with wild-type Pnr (Pnr^WT) in which both the DBD and the C terminus of Pnr are present and with Pnr^VX1, a truncated Pnr lacking the C terminus(Fig. 3B). TouA also associates with Pnr^D1, a mutated Pnr in which the structure of the DBD is probably disrupted since a coordinating cysteine of the N terminal zinc finger has been replaced by a tyrosine (Fig. 3B).

We then analysed whether TouA interacts with Pnr in a cultured cell line. We performed immunoprecipitations of protein extracts made from Cos cells transfected with expression vectors for TouA and either Pnr(Fig. 3C) or GST fusion proteins containing the DBD of Pnr (GSTPnr^DBD) or the C terminus of Pnr (GSTPnr^CT) (Fig. 3D). We found that Pnr, GSTPnr^DBD and GSTPnr^CT coimmunoprecipitate with TouA(Fig. 3C,D). Finally, we also tested the abilities of in vitro-translated ³⁵S-labelled TouA to bind to GSTPnr^DBD or GSTPnr^CT attached to glutathione-bearing beads. TouA interacts both with the DBD of Pnr and the C terminus of Pnr (Fig. 3E). As expected, TouA did not bind GST control beads, whereas GSTPnr^DBDand GSTPnr^CT did not bind the negative luciferase input.

To investigate the interactions between Tou and Pnr in more detail, we constructed several expression vectors encoding in yeast contiguous domains of Tou (Fig. 3A,F; TouB to TouK). The various segments of Tou were fused with VP16^AD and introduced into yeast together with unfused LexA^DBD or LexA^DBDPnr^WT. We observed activation of the reporter only in yeast expressing LexA^DBDPnr^WT/VP16^ADTouC and in yeast expressing LexA^DBDPnr^WT/VP16^ADTouJ, showing that the amino terminus of the MBD domain of Tou mediates physical interactions with Pnr in yeast.

Chip is an essential cofactor of Pnr(Ramain et al., 2000) since it facilitates enhancer/promoter communication necessary for ac-scexpression during Pnr-driven neural development. Since TouA associates with Pnr in yeast, we made use of the yeast two-hybrid system to ask whether TouA also associates with Chip. Chip was fused to LexA^DBD and assayed for interaction with VP16^ADTouA. We observed a strong increase ofβ-galactosidase activity in extracts made from yeast expressing LexA^DBDChip/VP16^ADTouA above the background level seen with extracts made from yeast expressing LexA^DBD/VP16^ADTouA or LexA^DBDChip/VP16^AD(Fig. 4B; data not shown),indicating that TouA associates with Chip(Fig. 4B).

Sequence comparison between the Ldb proteins from various species reveals two conserved functional domains, involved in protein-protein interactions(Matthews and Visvader, 2003). Thus, Chip (Fig. 4A) contains a N-terminal homodimerization domain, also involved in interactions with Pnr(Ramain et al., 2000) and Osa(Heitzler et al., 2003) and a C-terminal LIM interacting domain (LID) mediating heterodimerization with LIM-homeodomain proteins (LIM-HD)(Fernandez-Funez et al., 1998)and basic helix-loop-helix proteins (bHLH)(Ramain et al., 2000).

We next examined the interaction between TouA and Chip. Assays forβ-galactosidase activity revealed that the interaction between TouA and Chip is mediated by the N-terminal homodimerization domain(Fig. 4B). We also investigated whether TouA can interact with Chip in a cultured cell line by immunoprecipitating proteins extracts of Cos cells transfected with expression vectors encoding TouA and Chip, the N-terminal domain of Chip (N^TChip) or the C-terminal domain of Chip (C^T Chip)(Fig. 4C). We found that TouA physically interacts with Chip and the interaction is mediated by the N-terminal domain of Chip. Finally, we further found that in vitro translated ³⁵S-labelled TouA interacted with GST Chip attached to glutathione-bearing beads (Fig. 4D) but did not bind GST control beads, whereas GST Chip did not bind the negative luciferase input.

We next investigated the interaction between Tou and Chip in more detail. The segments of Tou (Fig. 3A:TouB to TouI; Fig. 4E: TouL to TouN) were fused with the VP16^AD and introduced into yeast together with unfused LexA^DBD or LexA^DBDChip. We found that only TouD containing the DDT domain associates with Chip(Fig. 4E). Moreover, the physical interaction was shown to be mediated by the DDT domain itself(Fig. 4E; TouM).

Tou functionally cooperates with Pnr and Chip during neural development. Since Tou physically interacts with Pnr and Chip, we then asked whether a ternary complex containing Tou, Pnr and Chip can exist in living cells. We performed double immunoprecipitations of extracts made from Cos cells containing Pnr, Tou and Chip (Fig. 5). The extract was first immunoprecipitated with the M2 antibody recognizing the flagged TouA (Fig. 5B; IP 1) and the precipitated proteins were recovered by elution with the Flag peptide. The eluate was then immunoprecipitated with the B10 antibody (Fig. 5B: IP 2)directed against the tagged Chip. We observed that Pnr coimmunoprecipitates with Chip and TouA (Fig. 5B),indicating that the trimer Chip-Tou-Pnr can be formed. Moreover, this suggests that Chip and Pnr act together to recruit Tou and to target its activity to the ac-sc promoter sequences. This observation is reminiscent to what was previously observed on the role of Osa during neural development (Heitzler et al.,2003). Osa belongs to the Brm complex and it was suggested that Pnr and Chip cooperate to recruit Osa and to target activity of the Brm complex to the ac-sc promoter sequences, leading to negative regulation of enhancer-promoter communication. However, Tou and Osa display antagonistic activities during regulation of ac-sc and may probably define distinct chromatin remodelling complexes since the loss of DC bristles associated with reduced tou function (tou² flies have 1.4±0.05 DC bristles/heminotum) is suppressed by lowering the dosage of osa. Indeed, tou²; osa⁶¹⁶/+flies have 1.97±0.07 DC bristles/heminotum(Fig. 2K).

To investigate whether the interactions between Pnr, Chip and Tou function in vivo, we examined effects of both loss-o-function and gain-of-function of tou on expression of a lacZ reporter driven by a minimal ac promoter fused to the DC enhancer [transgenic line DC:ac-lacZ (Garcia-Garcia et al.,1999)]. We found that expression is strongly impaired in tou² flies and in tou^E44.1 flies(Fig. 6B,C), in agreement with the lack of DC bristles. In contrast, overexpressed Tou(ap^Gal4/EP622) leads to increased lacZ expression at the DC site (Fig. 6G,H),consistent with the excess of DC bristles(Fig. 2D). The excess of DC bristles in pnr^D1 correlates with expanded ac-scexpression at the DC site (Fig. 6D). Since loss of tou function suppresses the excess DC bristles associated with pnr^D1, we addressed the consequence of loss of tou function on expression of the lacZ reporter in pnr^D1/+ flies. We found that expanded lacZ expression associated with pnr^D1 is decreased when tou function is simultaneously reduced(Fig. 6A,D,E). We conclude that Pnr and Tou cooperate during neural development and regulate ac-scexpression through the DC enhancer.

The primary structure of Tou reveals that, among the members of the WAL family of chromatin remodelling factors, it is most closely related to TIP5(Fig. 1). The fact that TIP5 associates in vivo with SNF2h, the mammalian homologue of DrosophilaIswi (Strohner et al., 2001)prompted us to investigate whether Iswi associates with Tou and plays a role during neural development.

We first investigated genetic interactions between Iswi and pnr. We made use of Iswi¹ and Iswi², which are both characterized by a point mutation that introduces a premature stop codon and are consequently predicted to encode truncated Iswi. However, Iswi¹ and Iswi² behave as null alleles since truncated Iswi¹ and Iswi² are not detected in western analyses,suggesting that the C terminus is required to stabilize Iswi(Deuring et al., 2000). The Iswi²/+ flies are wild type (2.00 DC bristle/heminotum)whereas the pnr^D1/+ flies have excess DC bristles(3.34±0.1 DC bristle/heminotum). However, this excess is lowered when Iswi function is simultaneously reduced[Iswi²/+; pnr^D1/+ flies display 2.84±0.18 DC bristles/heminotum](Fig. 2C,L. Conversely, the lack of DC bristles associated with pnr^VX1(1.86±0.03 DC bristles/heminotum) is aggravated when Iswifunction is simultaneously reduced (1.61±0.07 DC bristles/heminotum). We also observed that the loss of DC bristles associated with Chip^E (1.6±0.07 DC bristles/heminotum) is accentuated when Iswi function is simultaneously reduced (1.29±0.07 DC bristles/heminotum) (data not shown). Thus, loss-of-function Iswialleles behave like loss-of-function tou alleles, implying that Iswi is also required during neural development and suggesting that Iswi and Tou may act as subunits of a multiprotein complex.

It is not possible to generate flies lacking both maternal and zygotic Iswi function (Deuring et al.,2000). Hence, to assess the consequences of the loss of Iswi function on neural development we performed overexpression experiments of a dominant negative Iswi. We used the UAS-Iswi^K159Rline (Deuring et al., 2000)which allows overexpression of a dominant negative Iswi through the UAS/Gal4 system. A conserved lysine in the ATP-binding site of the ATPase domain is replaced with an arginine. This K159R substitution eliminates ATPase and chromatin-remodelling activities of Iswi in vitro, but does not affect the stability of Iswi and its incorporation into high molecular mass complexes. Hence Iswi^K159R behaves as a strong dominant negative protein.

Overexpression of Iswi^K159R in the precursor cells using the sca^Gal4 driver provokes a loss of multiple sensory bristles (Deuring et al.,2000) and it has been proposed that Iswi has a late function during neural development, essential for either viability or division of the precursor cells. We then induced widespread expression of Iswi^K159Rin the dorsal compartment of the imaginal wing disc using Gal4-ap^MD544. Overexpression of Iswi^K159R is therefore induced early during development and results in a lack of sensory organs, including a frequent loss of DC bristles(Fig. 2M).

Tou and Iswi promote DC development and could be subunits of a multiprotein complex. However, we can also hypothesize that Tou can substitute for Iswi function and vice versa. To address this issue, we overexpressed Iswi^K159R in either a wild-type genetic background or in conditions of loss of tou function. Overexpressed Iswi^K159R in the domain of pnr expression reduces DC bristles (1.56±0.11 bristle/heminotum) (Fig. 2N). The loss of DC bristles is aggravated when the tou function is simultaneously reduced. Indeed, tou²/+;pnr^Gal4/UASIswi^K159R and tou²;pnr^Gal4/UASIswi^K159R flies have 1.14±0.12 bristle/heminotum (data not shown) and 0.75±0.11 bristle/heminotum(Fig. 2O), respectively. This observation reinforces the hypothesis that Tou and Iswi could be subunits of a complex during neural development.

We next investigated whether overexpressed Iswi^K159R affects the activity of the DC enhancer. We found that overexpression of Iswi^K159R leads to (Fig. 6I) reduced expression of the lacZ reporter driven by the ac minimal promoter fused to the DC enhancer (line DC:aclacZ) (Garcia-Garcia et al.,1999). We next examined the consequences of loss of Iswifunction associated with the Iswi¹/Iswi²transheterozygous combination, which dies during late larval stages(Deuring et al., 2000). We observed that loss of Iswi function leads to decreased lacZexpression (Fig. 6F). These observations indicate that Iswi is also necessary for activation of ac-sc expression at the DC site, although we cannot rule out the possibility that this interaction is indirect.

Since Tou, Pnr, Chip and Iswi are required to activate proneural expression at the DC site, we next investigated whether Iswi can interact with Tou, Pnr and Chip. We found that Iswi can interact with Tou in both yeast and Cos cells, in agreement with previous reports showing that Iswi can interact with Tou-related proteins in insect and in mouse cells (Fyodorov et al., 2002; Strohner et al., 2001). An expression vector encoding LexA^DBDIswi was introduced in yeast together with a vector for the unfused VP16^AD or VP16^ADfused to one of the various segments of Tou(Fig. 3A, TouA to TouI). Measurement of β-galactosidase activity revealed that only protein extracts made from yeast expressing LexA^DBDIswi and either VP16^ADTouA or VP16^ADTouC display activity above background levels (Fig. 7A). Hence, TouA associates in yeast with Pnr, Chip and Iswi. Finally,immunoprecipitation of protein extracts revealed that TouA and Iswi also associate in transfected Cos cells (Fig. 7B).

We next addressed whether Iswi can interact with Chip and Pnr, by testing the abilities of in vitro translated ³⁵S-labelled Iswi to bind to GST-Chip attached to glutathione-bearing beads. We found that Iswi associates with full-length Chip. Similarly, we observed that Iswi can also associate both with the DBD and the C terminus of Pnr(Fig. 7C,D). Since Pnr and Chip directly regulate ac-sc expression at the DC site, our findings suggest that Iswi and Tou may belong to a complex, which, in vivo, regulates the activity of the proneural complex during enhancer-promoter communication,possibly through chromatin remodelling.

Transcriptional activity of GATA factors during development is regulated by dimerization with cofactors. This feature is illustrated in vertebrates by the function of GATA-1 during haematopoiesis(Cantor and Orkin, 2002) and in Drosophila by the role of Pnr during neural development. GATA factors carry a DBD containing a C-terminal zinc finger responsible for DNA-binding and a N-terminal zinc finger that stabilizes the interaction. In addition to binding DNA, the DBD makes critical interactions with proteins that modulate activity of the GATA factors (Cantor and Orkin, 2002). There is considerable evidence that these factors function as multiprotein complexes. Thus, Wadman and colleagues(Wadman et al., 1997) have identified an erythroid complex formed between the bHLH SCL/E2A, LMO2, GATA1 and Ldb1. Functional analyses of the role of Drosophila Pnr during neural development have shown that Pnr also contains a C-terminal domain mediating interaction with the Drosophila Ldb, Chip(Ramain et al., 2000). Chip assists in the formation of a proneural complex containing Pnr, Chip and the(Ac-Sc)-Da heterodimer bound to the ac-sc promoters and facilitates enhancer-promoter communication (Bulger and Groudine, 1999; Dorsett,1999) required for activation of ac-sc and neural development. The Drosophila complex and the vertebrate complex(Wadman et al., 1997) contain related transcription factors, illustrating how the mechanisms of regulation of GATA factors by dimerization are remarkably conserved in a variety of species.

The activity of the Drosophila complex is antagonized by dimerization with various transcription factors. Pnr associates with U-shaped(Ush), a protein structurally related to Friend of Gata, since they both contain multitype zinc fingers (Cubbada et al., 1997; Haenlin et al., 1997; Tsang et al., 1997). Moreover,regulation of GATA factors by Ush and Fog appears remarkably conserved during evolution since they contain nine zinc fingers, four of which (zfs1, 5, 6 and 9) mediate interaction with the N-terminal zinc finger of the DBD of GATA factors (Fox et al., 1999). The activity of the proneural complex is also antagonized by dimerization of Chip with the LIM-HD Apterous and by dimerization of the proneural bHLH Ac-Sc with the bHLH Extra macrochaetae (Emc). Chip is an essential cofactor for Pnr and Apterous and neural development relies on a competition between Pnr and Ap to associate with the common Chip cofactor(Ramain et al., 2000). Emc lacks the basic domain required for DNA binding and it is believed that it sequesters Ac and Sc and prevents them from associating with Da to give the(Ac-Sc)-Da heterodimer required for activation of proneural expression(Ellis et al., 1990; Garrell and Modolell,1990).

Our recent study in Drosophila has shown that the function of the proneural complex is also regulated by the Brm chromatin remodelling complex. Thus, Osa dimerizes with both Pnr and Chip, leading to recruitment of the Brm complex that regulates enhancer-promoter communication(Heitzler et al., 2003). Our present study describes Tou, a protein that associates with Iswi and that regulates activity of the proneural complex during enhancer-promoter communication, possibly through chromatin remodelling.

Enhancer-promoter communication is of fundamental importance, as in most cases the activities of enhancers play a determining role in turning on and off specific genes in a temporally and spatially regulated manner. Models to explain long range activation of gene expression invoke enhancer-promoter communication either through protein-protein interactions resulting in formation of loops, the free sliding of proteins recruited by a remote enhancer along DNA or the establishment of modified chromatin structure by facilitator factors which generate a chain of higher order complexes along the chromatin fibre (Bulger and Goudine,1999; Dorsett,1999).

In Drosophila, Chip was postulated to be a such facilitator required both for activity of a wing specific enhancer of the cutlocus (Morcillo et al., 1996; Morcillo et al., 1997) and for activity of the DC enhancer of the ac-sc complex(Ramain et al., 2000). Enhancer-promoter communication at the ac-sc complex is negatively regulated by the Brm complex whose activity is targeted to the ac-scpromoter sequences through dimerization of the Osa subunit with both Pnr and Chip (Heitzler et al., 2003). The Brm complex is thought to remodel chromatin in a way that represses transcription.

We describe, here, the role of Tou and Iswi which may act together as subunits of a multiprotein complex to positively regulate activity of Pnr and Chip during enhancer-promoter communication. Tou and Iswi therefore display opposite activity to that of the Brm complex, raising questions about their molecular function during neural development. Tou shares essential functional domains with members of the WAL family of chromatin remodelling proteins,including Acf1 of ACF and CHRAC (Ito et al., 1999) and also TIP5 of NoRC (nucleolar remodelling complex),involved in repression of the rDNA(Santoro et al., 2002). Importantly, Acf1 and TIP5 associate in vivo with Iswi(Ito et al., 1999; Strohner et al., 2001),showing that Iswi can mediate both activation and repression of gene expression. Tou positively regulates Pnr/Chip function during the period of ac-sc expression in neural development, and it associates with Iswi. Since Iswi also positively regulates Pnr/Chip function, we hypothesize that a complex encompassing Tou and Iswi acts during long-range activation of proneural expression, possibly through chromatin remodelling. Further studies will help to resolve this issue.

Interestingly, Chip and Pnr seem to play similar roles both during recruitment of the Brm complex and recruitment of Tou and Iswi, since they dimerize with Osa, Tou and Iswi. In addition, Pnr and Chip apparently cooperate to strengthen the physical association with Osa and Tou. However,Osa, on the one hand, and Tou and Iswi, on the other, display antagonistic activities during neural development. Since they are ubiquitously expressed,accurate regulation of ac-sc expression would require a strict control of the stoichiometry between Osa, Tou and Iswi. It remains to be investigated whether the functional antagonism between Osa and Tou/Iswi relies on a molecular competition for association with Pnr and Chip. Determination of this would require a complete molecular definition of the putative complex encompassing Tou and Iswi, together with a full understanding of how this complex and the Brm complex molecularly interact with the proneural complex to regulate enhancer-promoter communication during development.

Biochemical analysis of Iswi and Iswi-containing complexes, together with genetic studies of Iswi and associated proteins in flies and in budding yeast,has revealed roles for Iswi in a wide variety of nuclear processes, including transcriptional regulation, chromosome organization and DNA replication(Corona and Tamkun, 2004). Accordingly, Iswi was found to be a subunit of various complexes, including NURF (nucleosome remodelling factor)(Badenhorst et al., 2002), ACF and CHRAC. Iswi-containing complexes were primarily recognized as factors that facilitate in vitro transcription from chromatin templates(Mizuguchi et al., 1997; Mizuguchi et al., 2001; Okada et al., 1998). However, genetic analysis in Drosophila and in Saccharomyces cerevisiae have provided evidence that Iswi-containing complexes are involved in both transcriptional activation and repression in vivo. For example, immunostaining of Drosophila polytene chromosomes of salivary glands showed that Iswi is associated with hundreds of euchromatic sites in a pattern that is non-overlapping with RNA polymerase II(Deuring et al., 2000). It suggests that Iswi may play a general role in transcriptional repression. In contrast, it was also demonstrated that expression of engrailed and Ultrabithorax are severely compromised in Iswi-mutant Drosophila larvae (Deuring et al., 2000). Recent studies have also shown that a mouse Iswi-containing complex, NoRC, plays an essential role during repression of transcription of the rDNA locus by RNA polymerase I(Strohner et al., 2001; Santoro et al., 2002; Zhou et al., 2002). Here, we present Tou, a protein that is structurally related to the TIP5 subunit of NoRC (Santoro et al., 2002). Tou positively regulates enhancer-promoter communication during Pnr-driven proneural development and its activity is targeted to the ac-scpromoter sequences through dimerization with Pnr and Chip. We also provide evidence that Iswi is required during neural development. Overexpression of Iswi^K159R in the precursor cells of the sensory organs using the sca^Gal4 driver(Deuring et al., 2000) leads to flies lacking multiple bristles, suggesting that Iswi functions late during neural development, essential for either cell viability or division of the precursor cell. Using the Iswi¹/Iswi²transheterozygous combination and individuals overexpressing Iswi^K159R in earlier stages of development and in less restricted patterns, we show that Iswi also regulates ac-scexpression. Interestingly, the regulation is probably direct since Iswi associates with the transcription factors Pnr and Chip, known to promote ac-sc expression at the DC site(Garcia-Garcia et al., 1999; Ramain et al., 2000). Since Iswi interacts with Tou, we propose that Tou and Iswi may positively regulate activity of Pnr and Chip during enhancer-promoter communication, possibly as subunits of a multiprotein complex involved in chromatin remodelling.

We thank Irwin Davidson for critical reading of the manuscript; Pascal Heitzler, Juan Modolell, John Tamkun, Carl Wu and the Bloomington stock center for reagents; Jean-Maurice Dura, Jean-Luc Vonesch and Didier Hentsch for helpful suggestions; Marie-Louise Nullans for excellent technical assistance;and the sequencing, oligonucleotides synthesis and cell culture services of the IGBMC. L.V. is supported by a fellowship from the Luxembourg Ministère de la Culture, de l'Enseignement Supérieur et de la Recherche. This work was supported by grants from the CNRS, INSERM, the Université Louis Pasteur, the association pour la recherche contre le cancer (ARC) and the Ministère de la Recherche et de la Technologie.