The MYST-Containing Protein Chameau Is Required for Proper Sensory Organ Specification during Drosophila Thorax Morphogenesis

The adult thorax of Drosophila melanogaster is covered by a stereotyped pattern of mechanosensory bristles called macrochaetes. Here, we report that the MYST containing protein Chameau (Chm) contributes to the establishment of this pattern in the most dorsal part of the thorax. Chm mutant pupae present extra-dorsocentral (DC) and scutellar (SC) macrochaetes, but a normal number of the other macrochaetes. We provide evidences that chm restricts the singling out of sensory organ precursors from proneural clusters and genetically interacts with transcriptional regulators involved in the regulation of achaete and scute in the DC and SC proneural cluster. This function of chm likely relies on chromatin structure regulation since a protein with a mutation in the conserved catalytic site fails to rescue the formation of supernumerary DC and SC bristles in chm mutant flies. This is further supported by the finding that mutations in genes encoding chromatin modifiers and remodeling factors, including Polycomb group (PcG) and Trithorax group (TrxG) members, dominantly modulate the penetrance of chm extra bristle phenotype. These data support a critical role for chromatin structure modulation in the establishment of the stereotyped sensory bristle pattern in the fly thorax.


Introduction
Twenty-six large sensory bristles (or macrochaetes) are arranged in a stereotyped pattern on the dorsal thorax of Drosophila melanogaster. The regulation of macrochaete number and position has been widely studied as a paradigm to decipher genetic and molecular mechanisms involved in epithelium regionalization and differentiation (reviewed in [1][2][3][4][5][6][7]). The macrochaete is composed of five cells issued from the asymmetric division of a unique progenitor, the sensory organ precursor or SOP [8]. Each SOP is selected during larval development from a proneural territory comprising 10 to 30 cells of the wing imaginal disc epithelium that are provided with the potential to develop a neural fate by bHLH (basic helix-loop-helix) transcriptional activators encoded by proneural genes of the achaete-scute (ac-sc) complex [9][10][11]. Proneural genes are controlled by the products of so-called prepattern genes, whose regulatory activities are integrated by multiple, independent cis-regulatory elements scattered in the ac-sc locus [3,4]. Within the proneural cluster, SOP singling out results from coordinated processes of Notch-mediated lateral inhibition and autoregulatory activation loop, leading to increased accumulation of Ac-Sc factors in the SOP. Conversely, different classes of transcriptional repressors that prevent ac-sc expression promote the epidermal fate of non-SOP cells [12][13][14][15][16][17][18][19]. Once selected, SOPs enter differentiation and neuronal identity programs under combinatorial controls involving prepattern, proneural and tissue-specific gene products (reviewed in [4,7]).
The pannier (pnr) gene encodes a transcription factor of the GATA family, which is a key morphogenetic factor for dorsal identity specification in flies (reviewed in [3]). Pnr noteworthily promotes the development of dorsocentral (DC) and scutellar (SC) sensory organs in the mesothorax, which are part of the stereotyped bristle pattern that decorates the adult thorax [20,21]. The determination of DC bristles is widely studied in particular because an enhancer specifically recapitulating ac-sc expression in the DC proneural cluster has been cloned and dissected. For instance, Wingless (Wg) signaling provides a permissive environment for ac-sc expression [22]. Several transcription factors including Pnr, Chip (Chi), dLMO, Daughterless and possibly Ac (see however ref. 4), are required for the activation of the DC enhancer. It has been proposed that these factors function within a multimeric complex where Chi acts as a bridge between Pnr and Ac-Daughterless heterodimer to allow enhancer/ promoter communication and transcriptional initiation [23][24][25]. A Iswi/Toutatis chromatin complex is also recruited to facilitate enhancer/promoter communication and proneural activity [26]. Several directly interacting co-repressors down-regulate Pnrmediated transcriptional activation of ac-sc in the DC proneural cluster, such as U-shaped (Ush) [27], dCtBP [28] and the member of the Brahma (Brm) chromatin remodeling complex Osa [29].
The molecular regulation of SC bristles determination, on the other hand, is poorly understood. In particular, the regulatory elements controlling ac-sc expression in the SC proneural cluster are yet to be described. Nevertheless, mutations of several, but not all of the factors involved in DC determination also alter the pattern of SC bristles: Pnr, Ush and the Brm complex, are involved in DC and SC determination.
The MYST containing protein encoded by chameau (chm) has been initially identified as an epigenetic regulator of transcription involved in distinct mechanisms: modulation of Hox regulatory functions of Polycomb group (PcG) [30] and Trithorax group (TrxG) proteins (unpublished); control of JNK/AP-1 signaling during metamorphosis and thoracic closure [31]; regulation of the assembly and recruitment of the replication machinery at the vicinity of a replication origin [32]. In addition, chm has also been associated to the dendritic patterning of sensory neurons during embryogenesis [33], to the formation of the dorso-ventral boundary in wing imaginal discs [34] and to odor-guided behavior in adult flies [35]. Here, we report on the function of Chm in the control of macrochaete patterning in the Drosophila melanogaster thorax.

chm mutants display extra bristle phenotype
Homozygous chm 14 and chm 221 mutants carry deletions spanning part of the chm coding sequence and are zygotically null [30]. These animals develop up to the pharate adult stage but fail to emerge from the pupal cage. They present morphological defects of external cuticle including a thoracic cleft and an increased number of mechanosensory organs ( Fig. 1A; [31]). Supernumerary bristles in the chm mutant are only observed for SC and DC machrochaetes. While heterozygous animals, as wild-type flies, display four SC and four DC macrochaetes, chm pharate adults exhibit supernumerary anterior SC (aSC, an average of 5.73 per animal) and anterior DC (aDC, 4.24 per animal) (Fig. 1B). Compared to the normal, the extra bristles are thinner and shorter, but contain socket cells of apparently normal morphology (Fig. 1A). Depleting the maternal contribution of chm slightly aggravates the penetrance of this phenotype. Thus, 11% of chm zygotic mutants present extra aDC and 88% extra aSC, compared to 30% of chm mutants deprived of both maternal and zygotic contributions that exhibit supernumerary aDC and 100% supernumerary aSC, with an average of 4.42 aDC and 6.39 aSC per animal (Fig. 1B). However, zygotic or maternal plus zygotic loss of Chm results in qualitatively similar sensory bristle phenotypes, which only consist in the appearance of supernumerary DC and SC macrochaetes.
Because chm loss-of-function animals present extra aSC and aDC bristles, we investigated the impact of a gain-of-function in the wing imaginal disc on mechanosensory organ formation. UASchm transgenic flies were crossed with various Gal4 drivers (heatshock-Gal4 with a heat pulse during the third larval stage, 69B-Gal4, arm-Gal4; pnr-Gal4). In all combinations we could follow expression of the Myc-tagged Chm protein (data not shown), but failed to notice any alteration of the WT phenotype (Fig. 1B), indicating that an excess of Chm does not affect thorax development. Constitutive chm overexpression in a chm mutant background perfectly rescues thoracic defects (Fig. 1B [31]), establishing that the UAS-Chm transgene is functional and that the phenotypes originate from the lack of chm activity. Driving chm under the control of pnr-Gal4 in the most dorsal part of wing discs similarly allows normal thorax morphogenesis and proper number of SC and DC bristles to be specified. Thus, Chm plays a crucial role during thorax closure and macrochaete development.
We previously reported that chm controls the activity of the JNK signaling pathway in the proximal part of the wing imaginal disc epithelium during thoracic closure [31]. In order to check whether Chm also controls JNK signalling during bristle specification, we performed genetic interaction assays between chm and genes encoding key factors of the JNK pathway. Reducing the gene dosage of kayak, hemipterous or jra/djun, which aggravates chm thorax closure defect [31], does not significantly modify the frequency and number of supernumerary SC and DC macrochaetes present in chm mutant pharate adults (Table 1). This result strongly suggests that the control of sensory bristle formation by Chm occurs in a manner that does not depend on its ability to modulate JNK signaling.

chm controls SOP specification
To approach the role of chm in SOP determination, we first used a LacZ enhancer trap in neuralized (neur), neur A101 , which labels SOPs and their progeny [36]. As chm mutation more frequently induces the formation of supernumerary SC macrochaetes, we focused on SOPs emerging from the SC proneural cluster in the notum of imaginal wing discs. While two SOPs (one aSC and one posterior SC) are generated in the SC cluster of wild type wing discs, the LacZ pattern of neur A101 reveals three SOPs in chm mutant discs ( Fig. 2A). The ectopic SOP localizes in the presumptive region of the wing imaginal disc where the ectopic bristle will form. The extra SOP is determined later than normal SOPs, at a time where aSC and posterior SC have already completed the first asymmetric division, and emerges at a distance of the normal aSC with unlabelled cells in between. Therefore, additional SC bristles in chm mutant do not result from a defect in the bristle differentiation lineage and/or in SOP asymmetric division.
To assess whether Chm could act early in proneural clusters to influence SOP singling out, we performed phenotypic rescue experiments using various drivers: scabrous(sca)-Gal4, pnr-Gal4, neur-Gal4 and MZ980-Gal4. Expression of UAS-chm under sca-Gal4 in DC and SC proneural clusters or under pnr-Gal4 in the dorsal part of a prospective thoracic region that includes DC and SC proneural clusters, perfectly rescues chm supernumerary bristle phenotype. As a control, using MZ980-Gal4, that drives expression specifically in a wing disc proximal region distinct from DC and SC clusters and later along the prospective junction of the contralateral discs [37], perfectly rescues the thoracic cleft of chm mutants [31] but not the extra bristle phenotype. Interestingly, UAS-chm driven by neur-Gal4, which promotes expression later than sca-Gal4, when SOPs have already been singled out, does not allow supernumerary bristle rescue (Fig. 2B). Therefore, chm function in DC and SC proneural clusters is required early in the process of SOP specification prior to SOP specification.
Df(1) sc10-1 /Y living males are null for ac-sc and display a complete absence of thoracic sensory organs [38]. chm mutation does not restore SC and DC bristle development in the Df(1) sc10-1 / Y genetic background (Fig. 2C). This genetic epistatic relationship indicates that chm cannot provide proneural activity in the absence of ac-sc. Thus, chm may act upstream of ac-sc to repress or restrict their expression in the proneural cluster. Alternatively, Chm may reduce Ac and Sc proneural activities in the proneural cluster to prevent excess of SOP specification. chm genetically interacts with the GATA factor Pnr encoding gene during SOP specification Phenotypes observed in gain-of-function mutants of pnr as well as in combinations of hypomorphic loss-of-function pnr alleles indicate that Pnr is required for thoracic closure and for DC and SC bristle specification as well [20,21]. The resemblance of chm loss-of-function phenotypes and pnr gain-of-function phenotypes prompted us to test for genetic interactions between null alleles of chm and well-characterized mutations in pnr: the pnr D1 allele that produces a truncated protein no longer able to bind the Ush protein, known to act as an antagonist of Pnr activity in specific contexts [27,39]; the pnr VX1 and pnr VX6 loss-of-function hypomorphic and pnr VX2 null alleles [21].
Scoring DC macrochaetes, we observed that pnr VX mutations dominantly suppress and that pnr D1 dominantly enhances chm extra DC bristle phenotype (Table 2). Of note, pnr D1 heterozygous flies present ectopic DC bristles, which is enhanced further upon chm mutation, indicating that the two genes have opposite functions in macrochaete specification originating from the DC proneural cluster.
Scoring SC macrochaetes, we observed that pnr VX mutants dominantly suppress the chm phenotype (Table 2), again consistent with opposite roles for the two genes. Surprisingly, pnr D1 was found to also suppress chm extra SC bristle phenotype, which was not expected from above-mentioned genetic interactions during DC macrochaete formation. This suggests that unlike for DC bristle specification, pnr D1 does not behave as a gain-of-function allele for macrochaete singling out from the SC proneural cluster. This also highlights that despite sharing common regulators, SOP specification likely also relies on mechanisms specific to DC and SC proneural clusters.

chm genetically interacts with genes encoding transcriptional coregulators of Pnr
Several transcription factors and signaling pathways together with Pnr control the location of DC and SC proneural clusters in the epithelium and thus, the cells where ac-sc will be activated. Pnr cooperates, among others, with Wingless (Wg) and Decapentaplegic (Dpp) signaling pathways in the patterning of the dorsal thorax. This prompted us to test for interactions between alleles of the dpp and wg pathways and chm. Scoring SC and DC macrochaetes, we did not observe a qualitative alteration of the chm phenotype by mutation of dpp and wg signaling effectors armadillo (arm) and disheveled (dsh) (Fig. 3A). These results suggest that chm may not regulate the function of Pnr as a pre-pattern gene, consistent with the rescue experiment with the sca-Gal4 driver.
On the other hand Pnr activates the expression of ac and sc in the DC and SC proneural clusters [22,23]. Extensive molecular and genetic studies have identified multiple partners of Pnr in the regulation of DC and SC bristle specification. In particular, Chi, Dlmo, the bHLH proteins Ac/Sc and Daughterless may participate in the transcriptional activation of proneural genes in the DC cluster. Several additional factors are known to repress Pnr-mediated transactivation in the DC cluster, including Ush and TrxG group members of the Brm complex. Accordingly, mutations in the encoding genes result, as for chm and pnr, in the formation of extra-DC and/or SC macrochaetes or, on the contrary, the lack of DC and/or SC [23][24][25]. This prompted us to test for interactions between these genes and chm. Significant genetic interactions were observed with ush, Chi, osa and brm (Fig. 3B). The mutation of one copy of Chi, which cooperates with pnr in ac-sc transcriptional activation, suppresses the extra DC bristle phenotype of chm mutants without affecting the extra SC bristle phenotype. Thus, Chi antagonizes chm activity in a contextdependent manner, for DC macrochaete specification. The null allele of ush, ush VX22 , dominantly enhances the penetrance of extra DC and dominantly suppresses that of extra SC bristles seen in chm mutants. Note that heterozygosis for ush VX22 or for pnr D1 (Table 1) similarly modifies chm bristle phenotypes. Lastly, mutating one copy of one of the TrxG genes osa and moira (mor) enhances both DC and SC chm phenotypes (Fig. 3B), in agreement with their products being part of the Brm complex. Surprisingly, no genetic interaction was detected with brm that encodes the third TrxG member of the Brm complex (Fig. 3B), suggesting that in the absence of Chm one dose of the brm gene is sufficient to maintain Brm complex function. Together these results indicate that chm genetically interacts with pnr and several of its co-factors in DC and SC specification. In addition, the context-specific effects of several genetic combinations suggest that the control of ac-sc by these factors differ in its requirement for chromatin remodeling, histone modifications and co-factors in the DC and SC proneural clusters. As a corollary, chm does not repress SOP emergence through a unique and shared mechanism in the SC and DC proneural clusters. This context specific effect was further supported by looking at genetic interactions with two additional genes involved in the prepatterning of the thorax. Extra macrochaete (Emc), a helix-loophelix factor lacking the basic DNA binding domain, sequesters Ac and Sc in heterodimers unable of binding to DNA [40][41][42]. The mutation of one copy of emc dominantly enhances the chm supernumerary SC bristle phenotype without altering the extra DC bristle phenotype (Fig. 3A). Thus, emc synergizes with chm activity in a context-dependent manner for SC macrochaete specification. Finally, hairy shows extensive similarity with transcription factor N-myc and is also involved in bristle patterning [43]. The mutation of one copy of hairy does not modify the chm supernumerary SC or DC phenotype (Fig. 3A).
chm bristle phenotype is not altered by mutants of the lateral inhibition process Lateral inhibition is important to restrict the neural fate to the SOP in the proneural cluster. This is achieved by cell-cell communication where signals emanating from the SOP will repress the neural fate of neighboring cells. An important player in this process is the Notch signalization [44]. We thus tested whether mutations of Enhancer of split (E(spl)), Hairless (H) and Notch (N), which mediate Notch pathway activity, will genetically interact with chm (Table 3). We found no genetic interaction between chm and these mutants. These data are consistent with a genome-wide RNAi screen in Drosophila cell culture aiming at the identification of genes that control Notch-dependent transcription. In the screen, depletion of chm does not significantly affect Notch-dependent transcription [45]. Thus, extra-SOP present in chm mutants may not result from a defect in lateral inhibition.

Chm epigenetically represses SOP formation
Having established that Chm is an important regulator of SOP specification, we asked whether its putative acetyltransferase activity was required in this process. We used a Chm derivative (Chm G680 ) bearing a mutation at an invariant position of the acetyl-CoA binding site. A similar mutation has been reported to impair the enzymatic activity of MYST histone acetyl transferases (HATs; i.e. Sas3, Esa1 and Mof) [46][47][48] and we previously demonstrated that this mutation impairs Chm ability to specifically complement Sas2-mediated yeast telomeric position effect [30]. We thus tested the ability of Chm G680 [31], to rescue chm extra DC and SC bristle phenotypes. In contrast to what previously observed with wild-type Chm (Fig. 1B), providing the Chm G680 variant in the pnr expression domain (pnr-Gal4) or in the proneural clusters (sca-Gal4) does not allow rescuing chm bristle phenotypes (Fig. 4A). Since Chm and Chm G680 Myc-tagged constructs are expressed at similar levels [31], we concluded that Chm represses neural fate in DC and SC proneural clusters in an acetyltransferase activitydependent manner.
Since chromatin structure regulation often requires a complex interplay between multiple chromatin remodeling complexes, we next investigated whether already-identified chromatin modifiers could be involved with Chm in the determination of SOPs in proneural clusters. Using the loss-of-function chm mutant as a sensitized background, we screened for mutations in genes encoding chromatin proteins (i.e. 16 genes tested) able to dominantly modify chm DC and SC bristle phenotypes. In addition to already-mentioned mutations in the TrxG members osa and mor (Fig. 3B), we found that mutations in kismet (kis), domino (dom), taranis (tara), absent, small, or homeotic discs 1 (ash1), nejire (nej)/ CBP and Enhancer of Zeste (E(z)) dominantly modify chm sensory bristle phenotypes (Fig. 4B). Among these genes, dom, ash1 and E(z) affect both DC and SC phenotypes, while kis, tara and nej/CBP affect the SC bristle phenotype only. Of note, and validating our approach, several of these genes (kis, tara and nej/CBP) were found to play a role during SOP determination in the thoracic expression domain of Pnr [49].  Discussion chm is required to restrict sensory organ formation This study uncovers a novel role of Chm during adult thorax development. We report that, in addition to a defect in thoracic closure, homozygous null pharate adults exhibit supernumerary dorsal macrochaetes. Contrasting however with thoracic closure, where chm modulates JNK signaling during wing disc migration and fusion along the midline [31], its function during macrochaete development is genetically independent of JNK signaling activity (Table 1). Furthermore, the histone deacetylase Rpd3 counteracts Chm-mediated epigenetic control of JNK target genes during thorax closure, as mutating one copy of rpd3 rescues the thoracic cleft of chm mutants [31]. Conversely, rpd3 has been found here not to genetically interact with chm during sensory bristle formation (Fig. 4B). Thus, Chm is likely recruited in distinct molecular pathways during thorax closure and dorsal macrochaete specification.
chm mutants express a mild neurogenic phenotype, restricted to one supernumerary aSC and occasionally one aDC per hemithorax, emerging from the SC and DC proneural territories. Of note, restoring in a chm mutant background chm expression in proneural clusters (rescue by sca-Gal4 or pnr-Gal4) impairs the formation of extra but not of normal sensory bristles. Thus, Chm does not play an essential role in neural fate repression, but definitively acts to restrict neural fate commitment within DC and SC proneural clusters. Another feature of the phenotype lies in the systematic occurrence of epidermal cells between the ectopic and the normal macrochaete, indicating that Notch-mediated lateral inhibition is still operating, consistent with the fact that genetic alteration of the Notch-Delta regulatory loop does not modify the penetrance of chm bristle phenotype (Table 3). Likewise, we failed to detect any genetic interaction between chm and genes acting in Wg and Dpp signaling pathways (Fig. 3A) known to be involved in bristle patterning on the notum [50]. Altogether, these data suggest that Chm restricts neural fate in the DC and SC proneural clusters without influencing lateral inhibition and prepattern establishment.
Driven after SOPs are selected (neur-Gal4) in phenotypic rescue experiments, Chm is no longer able to restore appropriate macrochaete number, indicating that its function of neural fate restriction is required prior SOP singling out. In addition, the epistatic status of ac-sc to chm suggests that Chm may act upstream of ac-sc genes to repress ectopic bristle formation from SC and DC proneural clusters; that Chm may regulate Ac and Sc transcriptional activity to prevent extra-SOP emergence; or that Chm restricts SOP emergence through a mechanism yet to be described.

An epigenetic network controls SOP determination
Increasing evidence involves chromatin dynamics as an important regulatory level during sensory organ development, and several chromatin factors or complexes have been proposed to function in the control of proneural gene expression [26,28,29,51]. Here, we have shown that Chm is required within the DC and SC proneural clusters to restrict the neural fate to two SOPs only. Phenotypic rescue experiments furthermore revealed that Chm fulfills this function in a manner dependent of its putative HAT domain (Fig. 4A). Indeed, a mutation in the acetyl-CoA binding domain (Chm G680 ) prevents the rescue of chm extra DC and SC phenotypes.
HATs are often involved in defining landscapes of open chromatin that are permissive for transcription initiation [52]. Thus, Chm function as a negative regulator of transcription is counter-intuitive. In contrast, Chm may enhance the expression of a transcriptional repressor that would in turn repress the neural fate. Alternatively Chm may acetylate non-histone substrates to regulate their stability, their localization, their activity and eventually their function. Both Pannier and Scute are negatively regulated by phosphorylation in the wing epithelium [53]. Whether acetylation also controls their function should be tested.
Our candidate genetic screen identified several chromatin factor encoding genes that interact genetically with chm, either positively such as mor and osa TrxG members and the HAT encoding gene nej/CBP, or negatively such as three other TrxG genes, kis, tara and ash1, and two members of the PcG, E(Z) and dom (Fig. 4B). It is interesting to note that chm and TrxG/PcG genes act together in other developmental processes: osa, tara and chm have been recently simultaneously found in a suppressor screen of wing dorso-ventral boundary defect [34]; chm mutation enhances homeotic phenotypes of TrxG (unpublished) and of PcG genes as well [30], indicating a function in both PcG-mediated silencing and TrxGmediated maintenance of Hox gene activity. Mor and Osa are TrxG subunits of the Brm complex, thus the enhancement of the chm bristle phenotype resulting from the removal of one copy of mor or osa genes is in agreement with Chm synergizing with the Brm complex to counteract the neural fate. Kis, Tara and Ash1 TrxG members do not belong to the Brm complex, and have not yet been assigned with a function in sensory organ formation. The dominant suppression of the bristle phenotype in chm animals heterozygous for the encoding genes suggests that Kis, Tara and Ash1 play a role opposite to that of Chm in the control of SOP specification, and act at a level different from the Brm complex to facilitate the neural fate. Recent studies shed new light on Kis/ Ash1 and Brm complex interplay during transcriptional activation [54,55]. The Brm remodelling complex is proposed to act in first for chromatin opening and coactivator recruitment, and the Kis-L protein to function afterwards, as a monomer required for the recruitment of the histone methyltransferases Ash1 and Trx that lay down a methylation mark promoting early steps of transcriptional elongation. Although this sequence has been described in contexts where the Brm complex acts as an epigenetic activator, a similar mechanism might also account for a putative repressive activity, assuming for instance that chromatin remodeling by the Brm complex allow corepressor recruitment. In this case, we speculate that Chm would favor this negative effect of the Brm complex on Pnr function and impair Kis/Ash1-mediated activation. Transcriptional repression by PcG proteins is also likely to take place during SOP determination, in a manner opposite to Chm, as suggested by negative genetic interactions of chm with E(Z) and dom (Fig. 4B), A role for PcG factors in proneural gene control is for instance supported by the observation that Polycomb mutation enhances the extra bristle pnr D1 phenotype [28]. Finally, the synergistic effect of chm and nej/ CBP mutations in extra bristle formation (Fig. 4B) supports further that the chromatin structure and histone tail acetylation play an important role in the neural vs epidermal fate decision. Further work will be necessary to approach at the molecular level whether and how Chm interferes with achaete/scute genes, transcription factor Pnr and TrxG/PcG factors in chromatin-mediated control of sensory organ development. Chm is conserved in mammals, where it is known as HBO1, KAT7 or MYST2 [56]. Studies in human cancer cells have mostly focused on its function in DNA replication initiation and its involvement in inhibiton of growth (ING) protein-based complexes [57][58][59][60]. KAT7 behaves as a negative or positive regulator of gene expression. For instance, in gene reporter assays, it represses androgen receptor and NF-kappaB mediated transcription [61,62] while it enhances AP-1 and p53 mediated transcription [63,64]. KAT7 molecular role(s) in gene expression is still barely understood and therefore the direct or indirect effects of KAT7 in each study remain to be assesed. The phenotype of KAT7 knock-out mice was recently reported and KAT7 is essential for early mice embryogenesis [65]. In contrast to studies in human cancer cells, KAT7 appears dispensable for DNA replication in mice embryos while being required for gene regulation. Whether, as Chm in Drosophila, KAT7 also function in parntenrship with GATA factors to control neural develoment remains to be investigated.

Hypothetic function of Chm in SOP determination
Several lines of evidence suggest that Chm may affect the function of Pnr. Firstly, mutants loss-of-function of pnr present similar phenotypes as Chm, including extra-SC and DC bristles. Secondly, pnr and chm genetically interact in a manner consistent with their products playing opposite roles during SOP cell fate commitment, since loss-of-function of pnr (pnr VX alleles) dominantly rescues chm bristle phenotypes, and pnr D1 , commonly presented as a gain-of-function allele producing a constitutive activator unable to recruit the corepressor Ush, dominantly aggravates the extra DC chm phenotype ( Table 2). The dominant rescue of the extra SC phenotype rather suggests that pnr D1 conversely behaves as a lossof-function allele in the SC cluster. In support of this, previous reports have proposed that the SC (although not molecularly identified) and DC enhancers display different topographies in binding sites for Pnr and for Pnr cofactors [23,66,67]. Thirdly, chm genetically interacts with four of the genes known to interfere with pnr in the transcriptional control of ac-sc: ush, Chi, mor and osa. Of particular interest is the dual function of ush, which synergizes or antagonizes chm activity during DC or SC macrochaete formation, respectively. These opposite effects are reminiscent of the genetic interactions discussed above between chm and pnr D1 , supporting further that the partnership between Pnr and Ush may be different in the SC and DC cluster.
These genetic interactions, dominant phenotypic suppression by mutation in genes encoding Pnr coactivators (Ush in the SC cluster, Chi) and dominant phenotypic enhancement by mutation in genes encoding Pnr corepressors (Ush in the DC cluster, the Brm complex members Mor and Osa), are suggestive, yet not demonstrative, of a role for Chm in counteracting ac-sc activation by Pnr. Given the network of genetic interactions observed with chromatin regulators, this could be achieved by regulation of the higher order chromatin structure or in the spatial organization of the ac-sc locus, which harbors multiple regulatory elements scattered throughout ,100 kilo-bases of the X chromosome [4]. Also consistent with a role connected to chromatin proteins, Chm may be part of a PcG/TrxG maintenance mechanism that prevents ac and sc re-expression once cells are committed to the epidermal fate. In support of this model, PcG/TrxG response elements have been described in the ac-sc locus and chm was first identified has a PcG/TrxG genetic modifier [30,68]. However, our data do not exclude that Chm controls SOP emergence at a step different than transcription of the ac-sc genes. Alternative models for Chm function in SOP emergence are discussed below. First, Chm may be involved in the activation of neurogenic downstream targets of Ac and Sc [69][70][71]. Second, chm mutation may alter mechanisms involved in the elimination of cells still expressing high levels of Ac and Sc after SOP differentiation to prevent excess SOP determination. Yet to be described for macrochaete development, SOP-like cells, expressing high level of Ac and Sc but not selected as SOPs, are eliminated by programmed cell death during microchaete development [72]. Thus, extra-SOPs in chm mutants may originate from cells normally eliminated. Intriguingly, we previously observed that chm activity is necessary for programmed cell death following distortion of proximo-distal information in the wing imaginal disc [31]. Third, Chm may be required in the SOP to repress the neural fate of cells not adjacent to the SOP. In addition to lateral inhibition, it has been shown that SOP inhibits cells over several cellular diameters by extending long filopodia, and that these filopodia remain after SOP selection [73][74][75]. This later model seems less likely, since supernumerary SOPs in chm mutants appear after the completion of the asymmetric division of the extant SOPs and thus chm extra-SC and DC bristles should be rescue by expression of UAS-chm in 'normal' neuralized-positive cells. Finally, Chm may interfere with a recently-described Daughterless/Emc regulatory loop, which defines proneural territories independently of Ac and Sc [76].
To conclude, our work identifies a new player in the control of macrochaete development on the notum of Drosophila melanogaster and suggests that epigenetic mechanisms are important in this process. Future work will nevertheless be necessary to unravel the molecular mechanisms underlying the selection of DC and SC SOPs by Chm.

Fly lines and manipulations
Flies were grown at 22uC on a molasse-yeast-gelatin standard medium. pnr-Gal4, pnr D1 , pnr VX1 , pnr VX2 , pnr VX6 and ush VX22 lines were kindly provided by P. Heitzler (Strasbourg, France); MZ980-Gal4 by E. Martin-Blanco (Barcelona, Spain); hep 1 and hep 75 by S. Noselli (Nice, France); mof 1 by A. Akhtar (Heidelberg, Germany). All other fly stocks were obtained from the Drosophila stock centers in Bloomington and Szeged. Canton flies were used as wild type.

Cloning and generation of pUAS-Chm WT and pUAS-Chm G680E
The UAS-T-Chm WT transgene encodes a Chm protein tagged in its N-terminus with 6 Myc-tags. It was generated by sub-cloning a Myc-tagged Chm fusion in the pUAS-T plasmid. To generate the pUAS-T-Chm G680E plasmid, an internal fragment encompassing the HAT domain was removed from pUAS-Chm WT by digestion with restriction enzymes AgeI and XbaI and replaced by the same AgeI/XbaI fragment from the Chm G680E variant. The Chm G680E variant was generated by site-directed mutagenesis [31]. The resulting pUAS-T-Chm G680E and pUAS-T-Chm WT are the same plasmids but the point mutation in the invariant Glycine. Different transgenic flies were generated by random insertion of the transgene.

Phenotype analysis
Mutant chromosomes used for genetic interactions assays were maintained over balanced chromosomes labeled by GFP (chromosome II) or Tb or Ser dominant markers (chromosome III) to allow genotypic selection in the progeny. 32 to 133 pharate adults of each genotype were dissected out of pupal cages and thoracic morphogenetic defects were scored under stereomicroscope.

Statistical analyses
Each experiment was conducted in triplicate but the hep 75 ; chm 14 genotype for which pupae from different crosses were pooled to obtain over 30 individuals. Standard error of the mean is reported in the different figures and tables (6 s.e.m.). To determine statistical significance between two data sets, the Student t-test was performed: ns, no statistical difference; *, statistically significant difference P,0.05.