Genome-wide identification of in vivo DrosophilaEngrailed-binding DNA fragments and related target genes

Identification of target genes that are directly regulated by transcription factors is a key issue in developmental biology, and has been the purpose of several recent studies. Indeed, the genome-wide location of DNA-binding proteins using genomic microarrays has been performed in yeast(Iyer et al., 2001;Lieb et al., 2001;Ren et al., 2000). In mammalian cells, CpG island microarrays have allowed the identification of promoter regions capable of binding to the E2F transcription factor(Weinmann et al., 2002). Recently, whole-genome microarray assays associated with bioinformatic methods have also been successfully performed to identify direct target genes of the Dorsal transcription factor in Drosophila(Markstein et al., 2002;Stathopoulos et al., 2002). Identifying the genes that are directly regulated by transcription factors,rather than merely in the downstream pathways, remains essential for understanding gene function (Liang and Biggin, 1998; Mannervik,1999; Furlong et al.,2001; Egger et al.,2002). Homeodomain transcription factors play key roles during development by coordinating the behavior of most cells within their domains of expression (Garcia-Bellido,1975; Lawrence and Morata,1992), and identifying their target genes is challenging(Biggin and McGinnis, 1997). Interestingly, whereas homeodomain proteins recognize closely related binding sites, they are involved in specific genetic pathways and their absence produces very specific phenotypic effects(Wolberger, 1996). Therefore,to understand how homeodomain proteins function, it is essential to learn how they are acting in vivo and the identity of the downstream genes that they directly regulate.

To address this issue, we decided to identify genomic fragments that are bound in vivo by the homeoprotein Engrailed in Drosophila embryos.engrailed is involved in posterior cell identity, as well as in the maintenance of the anteroposterior (AP) boundary throughout development(Lawrence and Morata, 1976;Kornberg, 1981;Maschat et al., 1998). Furthermore, engrailed is expressed in a subset of neuroblasts and neurons in the developing central nervous system and in the brain. This suggests a role for engrailed during neurogenesis that is believed to be highly conserved during evolution (Bhat and Schedl, 1997; Hanks et al., 1998; Siegler and Jia,1999; Simon et al.,2001).

We have previously used UV crosslinking and in vivo chromatin immunoprecipitation to analyze Engrailed-binding sequences associated with particular genomic regions such as polyhomeotic orβ3-tubulin genes (Serrano et al., 1995; Serrano et al.,1997). We present a similar approach, performed on a larger scale,leading to a genome-wide view of direct Engrailed-binding loci in embryos. UV light or formaldehyde crosslinking are currently used to purify protein-DNA complexes, and to isolate specific binding fragments(Graba et al., 1992;Serrano et al., 1995;Serrano et al., 1997;Liang and Biggin, 1998;Cavalli et al., 1999;Toth and Biggin, 2000;Weinmann et al., 2001;Weinmann et al., 2002). However, UV light is believed to be more efficient in fixing proteins that are directly bound to DNA (Toth and Biggin,2000).

In the present report, we constructed a library enriched in genomic sequences that bind Engrailed protein in Drosophila embryos, by using UV crosslinking and chromatin immunoprecipitation (UV-X-ChIP). Systematic sequencing of the recovered clones led to the identification of 203 potential direct targets of engrailed and evidence is presented to show that some of them represent bona fide engrailed targets.

Reference wild-type flies are Oregon R. engrailed null mutant strain is Df(2R)en^E, deleted in both engrailedand invected (Gustavson et al.,1996). Heterozygous tramtrack mutation (ttk804)has been used to express engrailed ectopically in salivary glands(Fauvarque and Dura, 1993). The following transgenic stocks were used: UAS-Engrailed (referred as to UAS-En) (Tabata et al., 1995);UAS-VP16-HA tagged Engrailed (referred as to UAS-VP16-En)(Lecourtois et al., 2001;Alexandre and Vincent, 2002);(hs-Gal4); (MS1096-Gal4)(Milan and Cohen, 2000); and(en-Gal4) (Tabata et al.,1995).

The 1A4-GFP strain was made for the purpose of this study, by inserting 1A4 DNA fragment (either as a monomer or a trimer) into the WH.GFP (obtained from B. Bello, MRC, Mill Hill, UK). This vector corresponds to a mini white-based P-element vector with multiple cloning sites upstream of the hsp70minimal promoter from RHT vector (Bello et al., 1998) and the GFP F64L/65T-drosomycin poly A signal(Levashina et al., 1998). Different transgenic lines were obtained with both 1 A4 monomer and trimer and showed the same GFP expression pattern as described inFig. 5.

A Schneider two-cell line, transformed with a gene fusion where theengrailed cDNA was placed under the control of the hsp70promoter, was grown at 25°C (Gay et al., 1988). To induce engrailed expression, cells were placed at 37°C for 45 minutes, followed by 2 hours at 24°C. Soluble nuclear extract (referred as to HS-EN) was prepared as described(Gay et al., 1988). A typical extract contained 2.5 mg/ml of proteins, of which 2% is Engrailed (estimated from western analysis in comparison with FPLC Engrailed purified proteins). HS-EN protein has been used in in vitro immunoprecipitation and in gel shift assays.

Anti-Frizzled 2 antibody is a mouse monoclonal antibody(Cadigan et al., 1998). The following anti-Engrailed antibodies were used: monoclonal antibody 4F11 for immunoprecipitation and super-shift assays, and rabbit polyclonal antibody raised against the entire Engrailed protein for immunostaining(Serrano et al., 1995). Anti-HA 12CA5 is a mouse monoclonal (Roche) and is monitored using a biotinylated secondary antibody, prior to HRP detection (Vector, Vectastain). Cy3-conjugated anti rabbit is from Immunotech.

cDNAs used in this study were obtained from different laboratories:frizzled 2 (Cadigan et al.,1998); branchless(Ribeiro et al., 2002);frazzled (Kolodziej et al.,1996) and hibris(Artero et al., 2001).

A detailed protocol of X-ChIP is available athttp://www.igh.cnrs.fr/equip/WebFM/. Briefly, nuclei were prepared from a 0-16 hour Oregon R embryo collection. Purified nuclei were irradiated with UV light Stratalinker (254 nm at 10 cm for 10 minutes) in order to freeze DNA-protein interactions. Chromatin was recovered from these nuclei by using detergents. Free proteins were separated from the rest of the chromatin by CsCl ultra centrifugation. The supernatant was dialyzed against TE buffer overnight. Chromatin was sonicated in order to shear DNA to 0.1-3.0 kb average length. We usually obtained 100 μg of chromatin from 1 g of embryos. Around 40 μg of chromatin were used for each immunoprecipitation experiment. Chromatin was first incubated for 1 hour at 4°C in the presence of 100 μl of protein A sepharose CL-4B resin(Amersham Pharmacia). The chromatin was then incubated overnight at 4°C,with either 50 μl of resin coupled to 4F11 anti-Engrailed antibody (which corresponds to `EN probe') or 50 μl of resin with no antibody (`background probe'). After several washes, chromatin bound on the resin was eluted in the presence of 4% Sarkosyl. Samples were dialyzed, and DNA ends were repaired by Klenow before further purification. Samples were treated with RNase,proteinase K and phenol/chloroform extracted before precipitation in the presence of 20 μg glycogen. Linkers containing cloning sites were added and DNA was amplified by PCR, using a primer that covers the linker. Typically,10-20 μg of amplified DNA were recovered. At that stage, DNA can be either labeled by random priming to be used as a probe on a Southern blot or processed for further purification. Indeed, to construct the library, we performed an additional in vitro immunopurification using a quarter of the PCR amplification after the `in vivo' step, under the conditions described by Serrano et al. (Serrano et al.,1995). DNA was incubated in the presence of 50 μl resin where 4F11 anti-Engrailed antibody was bound, as well as Engrailed protein isolated from 1 mg of HS-EN cell culture nuclear extracts (containing ∼20 μg of Engrailed protein). After overnight incubation at 4°C, resin was washed and DNA was eluted in the presence of 1 M KCl. After phenol/chloroform extraction and precipitation in the presence of glycogen, DNA was PCR amplified using a primer that covers the linker. DNA was then phenol/chloroform extracted and precipitated, digested using HindIII and cloned into pBluescript KS⁺. Each clone was sequenced.

For each of the 315 clones of the library, 1 μg of plasmid DNA was digested by HindIII, separated on 1% agarose gel in 0.5× TBE and transferred onto GeneScreen Plus membranes (NEN Life Science). Membranes were hybridized and washed following manufacturer instructions. Probes were prepared by ³²P labeling of 75 ng PCR amplified DNA (isolated after`in vivo' immunoprecipitation), using the Rediprime kit labeling system(Amersham Pharmacia). Signals have been quantified using a phosphoimager.

DNA probes were synthesized by PCR amplification, using specific primers that were ³²P end labeled using T4 kinase. After gel purification,binding assays were performed in the presence of 1-5 ng labeled DNA,corresponding to 2000 cpm. Different quantities of HS-EN protein, isolated from cell culture nuclear extracts, were incubated with DNA probe for 30 minutes at 4°C in 10 μl of 25 mM HEPES (pH 7.6), 10% glycerol, 100 mM KCl, 1 mM DTT, 1% PVA, 1% NP40, 0.1% BSA and 200 ng of poly(dI:dC). DNA-protein complexes were resolved on 6% native polyacrylamide minigels in 0.5× TBE buffer (pH 8.3). For competition experiments, 750 ng unlabelled competitor DNA were added to the mix, and incubated with the protein and labeled DNA for 30 minutes at 4°C, before loading on gel. For supershift experiments, 4F11 antibody was incubated together with the labeled DNA for 30 minutes at 4°C, in the absence (-) or in the presence (+) of HS-EN protein, before loading on gel.

Embryo fixation and in situ hybridization using DIG labeled antisense RNA probes were performed as described previously(Alexandre et al., 1996). Dissected larvae were fixed for 20 minutes in fixation buffer [30 mM PIPES (pH 7.4), 160 mM KCl, 40 mM NaCl, 4 mM EGTA, 1 mM spermidine, 0.4 mM spermine,0.2% β-mercaptoethanol, 0.1% Triton X-100 and 4% paraformaldehyde], then washed four times in PBS + 0.1% Tween. The conditions for in situ hybridization were the same as for embryos. For double staining in embryos, in situ hybridization was performed first with an alkaline phosphatase-based detection system (Roche), followed by incubation with polyclonal anti-Engrailed antibody (dilution 1:300) overnight at 4°C. Detection of the immune signal was carried out with biotinylated secondary antibody, prior to HRP detection (Vectastain). Embryos were dehydrated and mounted in Canada Balsam for observation.

Squashes and hybridization were performed in tramtrack mutant background (ttk804), allowing engrailed expression in the salivary glands, as described by Serrano et al.(Serrano et al., 1995), with the following modifications: the 1A4-GFP DNA, labeled using Bionick labeling system (Invitrogen), was used as a probe and detected using fluorescein anti-biotin antibody (Vector) (1:200). Polyclonal anti-Engrailed antibody(1:200), secondary detected by Cy3 anti-rabbit antibody (1:200) has been used to identify Engrailed-binding sites. Chromosomal banding was detected with DAPI.

A series of Perl programs were specifically written to analyze UV-X-ChIP datasets (D. Martin, F. M. and B. J., unpublished). The sequences of all immunoprecipitated fragments were automatically compared with theDrosophila genome sequence (BDGP), using the blastn program(Altschul et al., 1990). The output was automatically treated using scripts in order to remove contaminating vector or non-Drosophila sequences and to detect repeated elements or chimeric fragments. Using this approach, 357 distinct Engrailed-binding loci have been identified, from 542 clones sequenced. Functional assignments for potential target genes were automatically performed through a script querying Gene Ontology (GO) terms associated to them in FlyBase (Ashburner et al.,2000).

In order to discover over-represented motifs in these sequences, only clones without internal HindIII restriction site were considered. These genomic sequences might contain all the information necessary for Engrailed recognition. Out of the 203 positive clones, 107 sequences agreed with this criteria and were analyzed using the MEME algorithm, according to Bailey and Elkan (Bailey and Elkan,1995). This led to the identification of 49 related motifs present in 204 hits and compiled in a position weight matrix presented onFig. 2A.

Chromatin immunoprecipitation on UV-treated Drosophila embryonic nuclei was performed in parallel with either anti-Engrailed monoclonal antibody or with no antibody (Fig. 1A). This step is referred as to `in vivo' because the source of Engrailed protein corresponds to the endogenous embryonic protein (see Materials and Methods). We then performed an additional immunoprecipitation cycle with an exogenous source of Engrailed protein (referred as to `in vitro'IP, in Fig. 1A). We assumed that this step reduces background and purifies immunoprecipitated DNAs that bind Engrailed directly, as previous studies have show that Engrailed is able to bind DNA with high affinity in vitro, even though co-factors may be required in vivo for full activity(Saenz-Robles et al., 1995;Serrano and Maschat, 1998). We have also previously shown that Engrailed-binding fragments that had been specifically immunoprecipitated in vivo after UV crosslinking were highly purified during further in vitro immunoprecipitation(Serrano et al., 1995;Serrano et al., 1997). After the `in vitro' step and a final PCR amplification, the fragments were cloned,leading to a library enriched in genomic Engrailed-bound sequences. From this library, 542 individual clones were isolated and sequenced(Fig. 1A). Sequence analysis revealed a total of 357 distinct genomic loci (see Materials and Methods).

To determine if these fragments were specifically enriched through immunoprecipitation, and do not simply represent nonspecific binding to the protein A sepharose resin, we performed Southern blots on 323 clones. We compared hybridization signal intensity with EN probe and a Background probe(see Materials and Methods) (Fig. 1A). With this assay, only the fragments for which the ratio EN probe/Background probe was higher than 2 were considered for further analysis. This was the case for 65% of the 315 intronic and intergenic fragments,whereas the eight DNA fragments lying in exonic sequences showed signals comparable with both probes (data not shown). As an example,Fig. 1B shows the results for 14 positive clones. Thus, 203 clones isolated in the library correspond to DNA that was specifically enriched by UV-X ChIP.

In order to localize precisely the 203 Engrailed-binding fragments, the corresponding sequences were compared with the published Drosophilagenome sequence (Adams et al.,2000). This allowed the identification of genes in the same location, which were categorized according to Gene Ontology (GO) annotation(see Materials and Methods, Table 1 andhttp://www.igh.cnrs.fr/equip/WebFM/).

Forty-seven percent of the fragments were localized within gene introns. In this case, we assumed that the corresponding intron is a part of theengrailed regulated target gene. Fifty-three percent were present in intergenic regions. In this case, we restricted our analysis to the nearest transcription unit, whatever its orientation and its distance with respect to the Engrailed-binding fragment. Half of the intergenic fragments are localized at less than 5 kb upstream of the genes, suggesting that the Engrailed-binding fragment may be a part of their promoter region. In rare cases (5%), when the Engrailed-binding fragment lies between two transcription units among which only one encodes a known function, we considered the latter as the putative target.

In 55% of the cases, the Engrailed-binding fragments could be associated with a gene whose function is known or that contains a recognizable protein domain. In all the other cases (45%), the binding fragments were associated to genes with an unknown function, which is approximately the ratio of this category in the Drosophila genome(Adams et al., 2000). For an overview, 81 genes of known or predicted function are presented here and have been grouped into functional classes (Table 1). The other 30 genes encode proteins with recognizable domains,the function of which is unknown, and have not been listed here.

According to GO annotation, the majority of these genes are involved in cell communication and developmental processes(Table 1). As expected from previous work, potential Engrailed targets identified using this approach include genes that are involved in the establishment and the maintenance of the AP axis body (Kornberg,1981; Vincent and O'Farrell,1992; Serrano et al.,1995; Alexandre and Vincent,2002). We also identified several genes involved in wing development (Hidalgo, 1994;Maschat et al., 1998),tracheal development, muscle development(Serrano et al., 1997) and axon guidance (Siegler and Jia,1999). Furthermore, different categories of genes encoding proteins involved in signal transduction were found (signal proteins,receptors, protein kinases, protein phosphatases, transcription factors and cell adhesion protein). Interestingly, cell adhesion proteins and receptors were particularly well represented (Table 1;http://www.igh.cnrs.fr/equip/WebFM/). This suggests that engrailed could act at different molecular levels in several developmental processes.

A motif research analysis was performed on a subset of 107 sequences from the 203 clones selected in the UV-X-ChIP library (see Materials and Methods). It revealed that the most frequent motifs were a group of 49 related octanucleotides, compiled in a position weight matrix presented inFig. 2A and resolved as a`YAATYANB' consensus. This consensus sequence largely overlaps those already described for Engrailed (Desplan et al.,1988; Kissinger et al.,1990).

In order to verify that Engrailed binds to this consensus in vitro, we performed a gel shift assay on the most represented motif `CAATTAGC', used as a pentamer. Several retarded complexes are formed in the presence of HS-EN protein with an affinity close to 10<sup>-9</sup> M(Fig. 2B). The formation of these complexes is competed in the presence of either a single double strand unlabelled motif `CAATTAGC' (referred as to C) or with a known specific Engrailed binding fragment D2 (Serrano et al., 1995). Furthermore, super-shifts of the complexes are observed in the presence of a specific anti-Engrailed antibody 4F11(Fig. 2B). By contrast,addition of either a mutated form of the motif `CAGCCGGC' (referred as to Cm)or polyhomeotic N fragment, which does not bind Engrailed(Serrano et al., 1995), had no effect on the formation of the complexes. Together, these data show that Engrailed binds specifically to this motif.

Gel shift assays were then performed on 14 Engrailed-binding fragments isolated from the library, and the results are shown here in four cases where the associated target genes are involved in different signaling pathways(Fig. 3). 1A4 clone corresponds to a genomic fragment lying 5 kb downstream of frizzled 2(fz2), which encodes one of the wingless (wg)receptors (Cadigan et al.,1998). 2H10 clone corresponds to a genomic fragment lying withinhibris (hbs), which encodes a member of the immunoglobulin superfamily involved in muscle guidance(Artero et al., 2001;Dworak et al., 2001). 1B12 clone corresponds to a genomic DNA fragment lying within the first intronof branchless (bnl), encoding the Drosophilahomologue of the Fibroblast Growth Factor (FGF) involved in tracheal morphogenesis (Sutherland et al.,1996). 2C5 clone corresponds to a genomic fragment lying in the first intron of frazzled (fra), which encodes a netrin receptor involved in motor axon guidance(Kolodziej et al., 1996).

In each case, two sets of experiments were performed, either with the entire immunoprecipitated fragment (150 bp to 350 bp), or with a shorter 100 bp fragment, surrounding the YAATYANB motifs, previously defined inFig. 2A. Similar results(number of complexes, affinity) were obtained for both types of probes, and the data presented in Fig. 3correspond to the shorter 100 bp fragments.

All these DNA fragments formed retarded complexes in the presence of HS-EN protein (Fig. 3). Addition of D2 DNA was able to compete the formation of the complexes, while addition of N DNA, as expected, had no effect. Moreover, addition of 4F11 antibody super-shifted the complexes. These data show the specificity of Engrailed binding, which was also confirmed using purified Engrailed protein (data not shown). The addition of the cold DNA fragment itself allowed us to compare the affinity of Engrailed on this fragment to the affinity of the strong Engrailed-binding fragment D2 (compare lanes D2 and A inFig. 3). The affinities are at least 10^-9 M and are closely related to the affinity of the motif`CAATTAGC' (compare Fig. 2Bwith Fig. 3).

In conclusion, we have shown that Engrailed is able to bind specifically to these four in vivo immunoprecipitated DNA fragments, which lie close to genes involved in different developmental processes, most probably via the`YAATYANB' consensus sequence that we have identified.

In order to discriminate among the list of putative targets, the ones responding to engrailed regulation in vivo, we used a simple screen. We monitored by in situ hybridization, the expression of several potential target genes, after ectopic expression of Engrailed using the UAS-GAL4 system. Because Engrailed can act as a repressor or an activator(Serrano et al., 1997;Serrano and Maschat, 1998), we overexpressed either the wild-type Engrailed protein (UAS-En) or a chimeric activator form (UAS-VP16-En) (Lecourtois et al., 2001; Alexandre and Vincent, 2002), under the control of MS1096-Gal4, in third instar wing imaginal discs (Fig. 4). We first tested this approach on β3-tubulin,which we have previously shown to be directly repressed by engrailed(Serrano et al., 1997). As expected, overexpression of wild-type Engrailed protein led to a repression of endogenous β3-tubulin in the wing disc(Fig. 4F), whereas overexpression of the activator form of Engrailed had no detectable effect,probably because of the strong expression of endogenous β3 tubulin in the discs (Fig. 4D,E).

Using this assay, we studied the expression of 14 genes that are localized close to the genomic DNA fragments isolated in the library and tested previously for their Engrailed-specific binding ability(Fig. 1B). The results are shown for four genes (frizzled2, hibris, branchless, frazzled) that are representative of the different pathways where engrailed seems to be involved (Table 1;Fig. 3). frizzled 2expression is activated in the presence of (VP16-En)(Fig. 4H) and repressed in the presence of En (Fig. 4I) (see wild-type expression in the wing pouch for comparison,Fig. 4G). This suggests thatengrailed might act as a repressor on fz2 expression.hibris is expressed along the wing margin and in the presumptive region of wing vein L3 and L4 in wild type(Fig. 4J). This expression is slightly activated in the presence of (VP16-En)(Fig. 4K), but strongly repressed when En is overexpressed (Fig. 4L), suggesting that hbs expression is regulated byengrailed in vivo. branchless is essentially expressed in a dorsal/posterior territory surrounding the wing pouch in wild type(Fig. 4M). In the presence of(VP16-En), several additional patches of bnl expression are detected within the wing pouch (Fig. 4N), whereas no activation of bnl is observed after wild type En overexpression (Fig. 4O). As expected, because MS1096 drives Gal4 expression only in the wing pouch (Fig. 4B), endogenous bnl expression outside the wing pouch is not affected (Fig. 4M-O),showing the specificity of the experiment. Finally, frazzled is slightly expressed in wild-type wing disc(Fig. 4P). This expression is activated when (VP16-En) is overexpressed(Fig. 4Q), and repressed upon En overexpression (Fig. 4R).

In conclusion, these data demonstrate that the expression of several potential target genes identified via UV-X-ChIP is modulated whenengrailed is misexpressed. This test has been successfully performed on 12 genes over 14 tested (Fig. 4 and data not shown).

Interactions between engrailed and the wingless signaling pathway have been extensively described(DiNardo et al., 1988;Martinez-Arias and White,1988; Bejsovec and Martinez Arias, 1991; Heemskerk et al.,1991; Hatini and DiNardo,2001). A direct regulation of frizzled receptor expression by engrailed has been shown(Lecourtois et al., 2001). In this study, we found that the other wingless receptor gene,frizzled2 (fz2), might also be directly regulated byengrailed. A high-affinity Engrailed-binding fragment (1A4) was detected in the close vicinity of the fz2 transcription unit(Fig. 3). In wild-type embryos,fz2 expression becomes segmentally repeated around stage 9, in two or three rows of cells just anterior to engrailed. In stage 9engrailed mutant embryos, fz2 expression is extended posteriorly, being detected in 4 rows of cells(Fig. 5A). This shows that Engrailed acts as a repressor of fz2 expression in embryos, as has been suggested with the previous test in the wing disc(Fig. 4). We verified whether the 1A4 Engrailed-binding fragment was able to drive the expression of a reporter gene in vivo and whether it was responding to engrailedregulation. For this purpose, this 170 bp fragment, either as a monomer or a trimer, was cloned upstream of a GFP reporter gene and hsp70 minimal promoter and introduced into the Drosophila genome by P element-mediated transposition (Fig. 5B). In these transgenic lines, GFP expression was essentially detected in the embryonic hindgut (Fig. 5C9) and in half of the larval hindgut(Fig. 5C1). GFP is expressed in the ventral cells of the larval hindgut that do not expressengrailed, which mimics endogenous fz2 expression(Fig. 5D). This demonstrates that the 1A4 DNA fragment might be a part of endogenous fz2regulatory sequences. Overexpression of (VP16-En) fusion protein driven byhs-Gal4 leads to ectopic GFP expression in the entire hindgut(Fig. 5C2), but also in tissues that do not express the transgene in wild type, such as the midgut(Fig. 5C3,C4), the salivary glands (Fig. 5C5,C6), and the wing disc (Fig. 5C7,C8). Overexpression of (VP16-En) fusion protein driven by en-Gal4 in embryos leads to ectopic GFP expression in a striped pattern(Fig. 5C9,C10). Such activation does not occur with overexpression of wild-type Engrailed, confirming a repressor role of Engrailed on fz2 expression through this 1A4 fragment (data not shown). These results show that 1A4 is able to respond toengrailed regulation in vivo.

Finally, in order to verify that the Engrailed-binding on 1A4 was direct,we analyzed the pattern of Engrailed immunostaining on polytene chromosomes,in the transgenic 1A4-GFP line. An additional Engrailed-binding site was detected on polytene chromosomes in the locus of the transgene (25C), clearly showing a direct fixation of Engrailed on the 1A4 fragment, in vivo(Fig. 6).

Altogether, these data show that the 1A4 fragment that was isolated by UV-X-ChIP is a part of the fz2 regulatory regions and is able to directly respond to engrailed regulation in vivo.

In this report, we used UV-X-ChIP to prepare a genomic library enriched in DNA fragments that bind the Engrailed transcription factor in vivo. A systematic sequencing of the clones led to the isolation of 203 Engrailed-binding fragments that can be assigned to potential targets, because they lie either in an intron (in 47% of the cases) or in the close vicinity of a gene. The gel shift assay analysis, presented here in four examples but verified in 12 cases (over 14 tested), confirmed that these DNA fragments bind Engrailed with high affinity in vitro. Furthermore, in vivo tests on a subset of potential target genes showed that in most cases (12 of 14 tested), the expression of the selected genes was sensitive to engrailedmisexpression. The results obtained on this sample indicate that X-ChIP may indeed be an efficient method with which to isolate direct targets. The identification of in vivo Engrailed binding fragments, and of the related target genes, constitutes a first step for the further analysis ofengrailed direct targets.

Using this approach, we found that potential engrailed target genes are involved in different developmental processes, such as AP patterning, neurogenesis, wing or tracheal development, and muscle development. In these different pathways, we identified a number of target genes encoding cell adhesion molecules and all types of receptors, which is in agreement with engrailed involvement in cell-cell contact events(Dahmann and Basler, 2000). The identification of targets that also encode signal proteins, enzymes such as protein kinases or phosphatases and, to a lesser extent, transcription factors, suggest that engrailed can act at different levels on a regulatory cascade.

Using correct markers, we notice that engrailed mutants exhibit severe pleiotropic phenotypes. This includes defects in axon migration(Siegler and Jia, 1999), but also in the attachment of the muscles along the AP axis of the body(Serrano et al., 1997) and in the pathfinding of the tracheal network (data not shown). These different phenotypes could result from the abnormal setting of the compartments inengrailed mutant embryos. The identification of potential targets involved in these processes (Table 1) suggests rather that engrailed may be more directly implicated in these phenotypes. This confirms the contribution ofengrailed in the orderly assembly and migration of cells during morphogenesis and pattern formation, leading to the normal positioning of the tissues along the AP axis.

The identification of Engrailed-binding fragments close to connectin,18 Wheeler or eagle, which have already been defined as geneticengrailed targets involved in neurogenesis, confirms a direct implication of engrailed in axon guidance(Dittrich et al., 1997;Eldon et al., 1994;Siegler and Jia, 1999). Furthermore, we identified that the netrin frazzled receptor gene,which is involved in motor axon guidance, might also be a direct target ofengrailed.

We isolated several Engrailed-binding fragments closely related to genes involved in muscle development, and, in particular, genes that ensure connections between the epidermis or PNS and the muscles. Indeed, our data suggest strongly that the involvement of engrailed in myotube guidance (Serrano et al.,1997) might result, in part, from the direct regulation ofhibris (Dworak et al.,2001).

We also report a direct link between engrailed and tracheal development. We indeed isolated several target genes that are involved in directing tracheal cell migration, such as the FGF-like secreted molecule,bnl. The tracheal system originates from placodes that consist in part of ectodermal cells. Interestingly, bnl is not expressed in the trachea, but in the ectodermal cells that overlie the migrating and branching trachea, thus acting as a guidance molecule that controls tracheal cell migration (Sutherland et al.,1996). This result suggests a direct effect of engrailedon tracheal guidance, which correlates with the engrailed mutant phenotype (data not shown), although engrailed involvement in tracheal migration has never been previously suggested.

Finally, one important and well-conserved function of engrailedconcerns its relationship with the wingless signaling pathway. The spreading of the Wingless (Wg) signal is crucial for establishing the pattern of differentiated cell types within tissues and organs(DiNardo et al., 1988;Martinez-Arias and White,1988; Bejsovec and Martinez Arias, 1991; Heemskerk et al.,1991; Lecourtois et al.,2001; Hatini and DiNardo,2001). This might depend in part on the level of expression of Wg receptors. Our screen strongly suggests that the Wg receptor frizzled 2, is likely to be directly regulated by engrailed. During hindgut morphogenesis, the morphogen dpp is repressed in the dorsalengrailed expressing cells, but is activated in the opposite ventral cells and it has been suggested that wg might be responsible for this activation (Takashima and Murakami,2001). We show that fz2 is only expressed in the ventral cells of the larval hindgut. Furthermore, we show that the 1A4 genomic fragment, isolated in the library, was directly bound by Engrailed in vivo and was effective in driving the expression of a GFP reporter gene in thesefz2-expressing cells. Our results strongly suggest that 1A4 is responsible for the repression of fz2 by engrailed in the dorsal cells of the hindgut, which validates the criteria of proximity to assign a gene to one immunoprecipitated DNA fragment. The results obtained for 1A4 and in two other cases (data not shown) demonstrate that this approach allowed us to identify genomic fragments that are functional Engrailed-binding sites in vivo.

As shown in this report, the UV-X-ChIP technique, when associated with sequencing, provides a means to enable the rapid collection of a large data set of high-affinity binding sites used by a transcription factor during development. This method is general enough to be used to identify binding sites and targets for other transcription factors (F. G. and F. M.,unpublished). Among the 203 sequences that localize within unique genomic regions, only 40 were found two or three times, indicating a low level of redundancy. This suggests that Engrailed is able to bind to a large number of genomic sites. Therefore, a more exhaustive genome-wide localization analysis should rather combine the same chromatin immunoprecipitation procedure with genomic DNA microarrays, which are not yet available inDrosophila.

We thank Karène Mahtouk and Sophie Vidal for their technical support, and Andrew Goldsborough for his comments concerning the manuscript. Thanks to G. Cavalli, G. Chanas and N. Nègre for helpful discussions during this work; C. Alexandre and J. P. Vincent for (UAS-VP16-En) fly strain;B. Bello for WH.GFP vector; R. Nusse for providing anti-Frizzled 2 antibody and frizzled 2 cDNA; M. Affolter for branchless cDNA; C. Goodman for frazzled cDNA; and M. K. Baylies for hibriscDNA. Funding for the research was supported by CNRS, FRM and by ARC grant 5341 to F. M.; a Singer-Polignac fellowship to P. S.; and a MENRT fellowship to D. M.