Activation and repression activities of ash2 in Drosophila wing imaginal discs

Differential gene expression results in cell diversity, although how different cell identities are established early in development and maintained throughout life is still poorly understood. Most of the transcription factors required for early developmental decisions are expressed transiently, but the gene expression patterns they trigger are maintained during cell division and inherited by daughter cells. Actively dividing cells must preserve individual genes in an on or off expression state after an initial commitment is made,especially given that some regulators disassemble from promoters during DNA replication or mitosis. Thus, developmental decisions may be maintained by the ability to deposit epigenetic marks involving chromatin-modifying complexes to control the cellular memory of gene activity states(Francis and Kingston, 2001; Simon and Tamkun, 2002).

Genes of the Polycomb (PcG) and trithorax group (trxG) encode proteins that are engaged in the regulation of cellular memory(Orlando, 2003). In early Drosophila embryonic development, Hox gene expression is controlled by a genetic cascade that includes the segmentation genes(Simon, 1995). When the segmentation proteins decay, Hox expression is maintained in the correct spatiotemporal pattern by the action of PcG and trxG genes, which often act as transcriptional repressor- and activator-chromatin complexes(Francis and Kingston, 2001; Simon and Tamkun, 2002). In addition to Hox genes, PcG/trxG also act on other target genes(Beltran et al., 2003; Francis and Kingston, 2001). In a genome-wide prediction of PcG/trxG response elements (PRE/TRE) in Drosophila, more than 100 elements were identified that mapped to genes involved in development and cell proliferation(Ringrose et al., 2003). However, epigenetic marks are not only restricted to embryonic stages. At later stages, developmental fates are also frozen and inherited by the repressor and activator activities of PcG/trxG complexes. The Drosophila wing imaginal disc has proven to be a useful model with which to study how these complexes act to maintain cell identities, as shown for wingless (wg) and hedgehog (hh)pathways (Collins and Treisman,2000; Maurange and Paro,2002).

The non-neural tissues of the Drosophila wing are organised into two types: the A-E intervein regions and the L1-L6 veins(Fig. 1A). The specification of veins in the wing imaginal disc occurs during larval and pupal stages, and is controlled by a network of cell-to-cell interactions, including the Egfr signalling pathway(Diaz-Benjumea and Hafen,1994). The rho gene, which encodes a seven-pass transmembrane serine protease, is an activator of Egfr(Bier et al., 1990; Sturtevant et al., 1993). rho is expressed in rows of cells coinciding with vein primordia and is required for vein formation, as indicated by the observation that the loss-of-function allele rho^ve displays truncated veins(Diaz-Benjumea and Garcia-Bellido,1990; Sturtevant et al.,1993). Localised expression of rho and vein(vn), which encodes a diffusible neuregulin class of ligands,activates the Ras/MAPK signalling cascade necessary for vein differentiation(Sturtevant et al., 1993; Schnepp et al., 1996). By contrast, inhibition of Egfr signalling by the transcription factors blistered (bs) and net is responsible for intervein specification. bs, the Drosophila homologue of the Serum Response Factor, is expressed in a pattern associated with intervein regions and is required for the organisation and differentiation of intervein cells(Fristrom et al., 1994; Montagne et al., 1996). During disc proliferation, bs expression is independent of rho, but during the pupal period bs and rho expression become mutually exclusive (Roch et al.,1998). The net gene, which encodes a basic helix-loop-helix (bHLH) protein, is also expressed in the intervein regions(Brentrup et al., 2000). In contrast to bs, net and rho expression is mutually exclusive in the wing discs of third instar larvae. Lack of net activity causes rho expression to expand, and vice versa. Furthermore, ectopic rho expression results in repression of net, thus generating wings with ectopic vein tissue (Brentrup et al., 2000).

The refined localisation of the L2-L5 veins in the wing depends on positional cues established along the AP axis of the wing imaginal disc(Biehs et al., 1998; de Celis et al., 1996; Sturtevant et al., 1997; Sturtevant and Bier, 1995). The posterior compartment cells express engrailed (en),which activates hh. Hh diffuses through a few cell rows of the anterior compartment, activating vn, knot (kn) and decapentaplegic (dpp), which form the AP organising centre(Bier, 2000). Vn activates Egfr in the borders of the organiser, giving rise to the L3 and L4 vein primordia, while Kn prevents the domain between L3 and L4 responding to vein differentiation, ensuring the intervein fate of this region(Crozatier et al., 2002; Mohler et al., 2000). Dpp diffuses from the AP boundary and activates target genes in a threshold-dependent fashion (Lawrence and Struhl, 1996). The development of L5 is dependent upon the two abutting Dpp target genes, optomotor-blind (omb) and brinker (brk) (Cook et al., 2004). Another Dpp target is the spalt-major(salm)/spalt-related (salr) complex(sal-C) of zinc-finger transcription factors(de Celis et al., 1996; Lecuit et al., 1996; Nellen et al., 1996), which is expressed in the central domain of the wing. The anterior low sal-C-expressing domain promotes the activation of the knirps (kni)/knirps-related (knrl) complex(kni-C), which results in L2 specification(Lunde et al., 1998).

Ash2 is a trxG protein that belongs to a 0.5 MDa complex thought to be involved in chromatin remodelling(Papoulas et al., 1998). Ash2 accumulates uniformly in imaginal discs, fat body cells and salivary glands(Adamson and Shearn, 1996). Loss-of-function alleles of this gene cause homeotic transformations(LaJeunesse and Shearn, 1995; Shearn, 1974; Shearn et al., 1987; Shearn, 1989; Shearn et al., 1971) and downregulation of Hox genes (Beltran et al., 2003; LaJeunesse and Shearn, 1995), in addition to severe abnormalities in the wing,such as reduction of intervein and enhancement of vein tissues(Adamson and Shearn, 1996; Amorós et al., 2002). To gain more insight into the function of ash2, we examined whether vein- and intervein-specific genes and vein positioning genes act as putative targets of ash2 function. We found that ash2 is involved in activating intervein-promoting genes and downregulating the Egfr pathway. Moreover, ash2 also acts as a kni repressor independently of sal-C. These results strongly support a role for ash2 in maintaining vein/intervein developmental decisions and vein patterning in the developing wing.

We used the following genetic strains as ash2 alleles: yw;ash2¹¹²⁴¹¹/TM6C (Deak et al., 1997; Amorós et al., 2002) and yw; ash2^I1/TM6C(Amorós et al., 2002; Beltran et al., 2003,). Canton S was used as a wild-type strain. To study interactions between ash2 and Egfr signalling pathway elements, we tested the hypomorphic combination rho^vevn¹, top¹/top^3C81(Clifford and Schupbach, 1989; Diaz-Benjumea and Garcia-Bellido,1990), the gain-of-function allele of Egfr Elp^B1/CyO and the rolled gain-of-function mutation D-Raf^C110; rl^Sem (provided by A. García-Bellido). In order to eliminate the D-Raf^C110 allele, crosses were designed to score the male +/Y; rl^Sem/+; ash2^I1/+ progeny. The alleles bs⁰³²⁶⁷/CyO (provided by M. Affolter), E(spl)mβ-lacZ (provided by S. Bray) and net¹ were used as intervein markers; the stock w; h kni^ri–1 was used to analyse L2 development. To study the effects of ash2 on kni expression, the minimal L2-enhancer element EX-lacZ, a 1.4 kb fragment that contains an activation and repression domain of the kni gene(Lunde et al., 2003) was provided by E. Bier. For ectopic expression of sal-C, we used the UAS-sal^64d transgenic, on the first chromosome, the UAS-salr⁸ transgenic, on the second chromosome, and the nubbin-Gal4 insertion line (provided by J. F. de Celis). We also analysed whether ash2 regulates brk and scalloped(sd) expression by using the brk^X47-lacZtransgene (provided by G. Morata) and the sd^EXT4 stock. The stocks Elp^B1/Cyo, net¹, w; h kni^ri–1 and sd^EXT4 mentioned above were obtained from the Bloomington Stock Center.

Clones mutant for ash2^I1 were obtained by mitotic recombination using the FLP/FRT technique(Xu and Rubin, 1993). yw;FRT82Bash2^I1/TM6C crossed with ywhsflp;FRT82BGFP/TM6B and wing imaginal discs from third instar Tubby⁺ larvae and pupae were dissected. Heat shock was carried out for 30 minutes at 37°C [52±4 hours after egg laying (AEL)] to induce clone formation. To monitor EX-lacZ(Lunde et al., 2003)expression, a yw; EX-lacZ; FRT82Bash2^I1/TM6C stock was created and clones were induced as above.

Overexpression of sal-C was obtained by crossing UAS-Sal^64d; FRT82Bash2^I1/+ males flies with were ywhsflp; nubbin-Gal4; FRT82BGFP/TM6B females. Tubby⁺female larvae were dissected. To monitor brk and sdexpression, brk^X47-lacZ; FRT82Bash2^I1/+ males and sd^EXT4-lacZ; FRT82Bash2^I1/+ males were crossed to ywhsflp;FRT82BGFP/TM6B and Tubby⁺ female larvae were dissected. In both cases, only 50% of the progeny contained ash2^I1 clones.

To obtain Minute⁺ clones the stock used was yw;FRT82B arm-lacZ M(3)/TM6C and heat shock was carried out for 7 minutes at 34°C (110±4 hours AEL). Adult ash2 mutant FLP/FRT M⁺ clones marked with the forked mutation were analysed in males with the following genotype: ywhsflpf^36a;FRT82BP[f⁺]87DM(3)w[124]/FRT82Bash2^I1. The heat shock was carried out for 10 minutes at 37°C (80±12 hours AEL).

Larvae and pupae of the appropriate genotypes were cultured at 25°C and timed in hours AEL or after puparium formation (APF).

Immunohistochemistry was performed according to standard protocols. Primary antibodies used were: guinea pig anti-Kni (1/50), provided by J. Reinitz;rabbit anti-Salm (1/500), provided by R. Barrio; rat anti-Bs (1/200), provided by M. Affolter; rabbit anti-Plexus (1/1000), provided by H. Matakatsu; rabbit anti-Vestigial (1/20), provided by S. Carroll; mouse anti-En (1/25) from the Developmental Studies Hybridoma Bank of the University of Iowa; and rabbit anti-β-galactosidase (Cappel) (1/1000). Kni, Salm and Vestigial antibodies were pre-absorbed before use.

Secondary antibodies were obtained from Jackson Immuno Research and include: donkey anti-rat-Rhodamine Red (1/200), donkey anti-guinea pig-Cy5(1/400), goat anti-mouse-FITC (1/200) and donkey anti-rabbit-Rhodamine Red(1/200). Propidium Iodide (Molecular Probes) was used as a nuclear marker after RNAase treatment. Fluorescence was visualised with a Leica TCS confocal microscope.

In situ hybridisation using digoxigenin-labelled antisense RNA probes was carried out essentially as described previously(Sturtevant et al., 1993; Brentrup et al., 2000).

DIG-labelled riboprobes for net RNA were synthesised using a 2.2 kb insert of netcelΔ922 (gift of M. Noll) linearised with EcoR1, and for rho from a rho cDNA clone (gift of E. Bier) linearised with HindIII. Sense RNA probes for netand rho did not show detectable signal.

Total RNA from wild-type and ash2^I1 homozygous larvae was extracted using Trizol (GibcoBRL) and a poly(dT)-24 primer was used for cDNA synthesis. The reaction was carried out in a final volume of 25 μl with five units of avian myeloblastosis virus-RT (Promega) and 200 units of Moloney murine leukaemia virus RT (Gibco). One microlitre of the RT reaction was used for PCR. The specific primers used[5′-cgccgccctgcccttcttc-3′ (forward) and 5′gggctgctgctagtcggagtggt 3′ (reverse)] were designed to amplify a 369 bp product of the kni gene.

The ash2¹¹²⁴¹¹ mutation is a single PlacWinsertion (Deak et al., 1997)in the fourth intron of the ash2 gene(Amorós et al., 2002)that causes pharate lethality. Homozygous flies that reach the adult stage(12% at 25°C) are sterile and show reduced wing size, crooked L2 and ectopic vein tissue, mainly extra crossveins and thicker veins(Fig. 1A,B). The ash2^I1 allele was generated after an imprecise excision of the ash2¹¹²⁴¹¹ insertion and is lethal in late third instar/early pupae. Molecular analysis of the ash2^I1mutation has shown that it comprises a 2 bp deletion and a 5 bp insertion that result in the absence of the full-length 2 kb transcript(Beltran et al., 2003). Imaginal discs of both alleles are reduced in size, ash2^I1being smaller than ash2¹¹²⁴¹¹(Fig. 1C-E). The smaller size and abnormal shape suggests that the ash2^I1 mutation alters proliferation and patterning. Clones homozygous for ash2^I1 exhibit impaired proliferation, intervein reduction and extra vein tissue, preferentially close to the normal veins, which appear thickened (Amoros et al., 2002)(Fig. 1F,G). This phenotype is a consequence of intervein cells acquiring morphological features of vein cells, which are typically smaller, more pigmented and with shorter and thicker trichomes than wild-type intervein cells. This extra vein phenotype led us to question whether ash2 functions as a negative regulator of vein differentiation in intervein territories. To test this hypothesis, we assessed genetic interactions with alleles of genes involved in vein/intervein development.

We perturbed the Ras/MAPK signalling pathway in the wing using mutants of genes required for Egfr activation. We first analysed loss-of-function mutants of the pathway. In flies mutant for the hypomorphic viable combination rho^ve vn¹, activation of the MAPK pathway in presumptive vein cells is prevented and, as a consequence, veins fail to differentiate. By contrast, the triple mutant rho^vevn¹ ash2¹¹²⁴¹¹ develops veins(Fig. 2A,B). We observed varying degrees of rescue, ranging from wings that develop only L2 to wings that develop veins almost completely, even with extra crossvein tissue or proximal vein fusions between L2-L3 and L4-L5. Rescue of L2 and L5 is more pronounced than L3 and L4, which are never distally complete, and many wings show notches in the posterior wing margin (77% of cases, n=75 wings; Fig. 2B). In 25% of these cases, wings show a tube-like shape, possibly owing to detachment of the dorsal and ventral cell layers. This variety of phenotypes is probably due to the variable expressivity found in the ash2¹¹²⁴¹¹ allele. The top¹/top^3C81 allelic combination is a hypomorphic mutation of Egfr (top) in which wings lack the anterior crossvein (a-cv) and a segment of vein L4(Fig. 2C). Rescue of missing vein tissues is observed in top¹/top^3C81;ash2^I1/+ flies, as shown by the presence of a-cv (58% of the cases, n=93), a complete L4 (4%), or restoration of both a-cv and L4(28%; Fig. 2D).

We also tested whether ash2 alleles enhance gain-of-function phenotypes of the Egfr pathway. The Ellipse^B1(Elp^B1) allele is an activated form of Egfr. In addition to other phenotypes, Elp^B1 individuals consistently develop wings with ectopic vein tissue in L2(Fig. 2E). In Elp^B1; ash2^I1/+ no enhancement was found (data not shown), but Elp^B1/+;ash2¹¹²⁴¹¹/ash2¹¹²⁴¹¹ individuals had reduced wings with ectopic vein tissue (46%, n=50; Fig. 2F), including vein fusion of the proximal regions and extra crossveins between L2 and L3, and between L4 and L5, but never between L3 and L4. In this case, a graded enhancement was also found, ranging from a wing similar to that of ash2¹¹²⁴¹¹ to ectopic vein tissue covering most of the wing. Similarly, the MAPK gain-of-function allele, rl^sem,generates ectopic vein tissue, a phenotype that is enhanced in ash2^I1 combinations (61%, n=70; Fig. 2G,H).

As these results indicate that ash2 antagonises the Egfr pathway,we tested whether rho expression is affected. In situ hybridisation for rho mRNA in third instar ash2^I1 discs showed either no expression at all or expression in only a few scattered cells(Fig. 3A,B), possibly owing to a strong perturbation of patterning in these discs. However, in ash2¹¹²⁴¹¹ homozygous pupal wings, rho is expressed and organised in veins, but the domains of rho expression are larger (Fig. 3C-F).

As several genes that promote intervein specification are antagonists of the Egfr signalling pathway, we tested whether the phenotypes described above are associated with loss of intervein gene activity. During larval stages, net is the key gene involved in intervein development and acts as an antagonist of rho (Brentrup et al., 2000). In third instar wing discs, net transcripts are confined to broad domains corresponding to prospective interveins, in a pattern complementary to rho(Brentrup et al., 2000). net RNA expression is considerably reduced in ash2 mutants,although in ash2¹¹²⁴¹¹, some expression is still found in the central domain of the wing pouch (Fig. 4A-C). Likewise, homozygous flies for the net¹/net¹ allele develop extra vein tissue,which is preferentially associated with transverse connections between L2 and L3 and between L4 and L5 (Fig. 4D), resulting in wing blades with an increased number of cells(Diaz-Benjumea and Garcia-Bellido,1990; Garcia-Bellido and de Celis, 1992). The mutant combination net¹/net¹; ash2^I1/+ results in extra vein connections along the proximal and distal L2-L3 region, thickening of veins, blistering and a lanceolated shape to the wing. In addition, the intervein tissue is strongly reduced between veins L2-L3 and L4-L5, and to a lesser extent in A and E intervein regions(Fig. 4E). A more extreme phenotype (75%, n=20 wings; Fig. 4F) was obtained in net¹/net¹;ash2¹¹²⁴¹¹/ash2¹¹²⁴¹¹ wings, in which the intervein areas between L2-L3 and L4-L5 are totally missing and result in thicker and fused veins (Fig. 4G). The region between L3 and L4 did not show extra vein tissue and was much less affected.

We also generated double mutant flies to test whether ash2interacts with the intervein-promoting gene bs. The loss-of-function allele bs⁰³²⁶⁷, a PlacZ insertion within the bs gene, is not viable in homozygosis(Karpen and Spradling, 1992). In pharate bs⁰³²⁶⁷ homozygotes, the whole wing blade is transformed into corrugated vein tissue, whilst the hinge region is almost intact. Heterozygous bs⁰³²⁶⁷ flies display small patches of ectopic veins in the proximity of L2 and L5(Roch et al., 1998)(Fig. 5A). However, flies that are homozygous for ash2¹¹²⁴¹¹ and heterozygous for bs⁰³²⁶⁷ exhibit blistering(Fig. 5B), reduced wing size associated with localised reduction of the intervein tissue(Fig. 5B,C) and development of extra crossveins (Fig. 5C,D). Veins L2 and L3 are thicker and totally or partially fused in the proximal region. Similarly, the D-intervein region between veins L4 and L5 is reduced and these veins are partially fused and thicker. Moreover, in pupal bs⁰³²⁶⁷/+; ash2¹¹²⁴¹¹/ash2¹¹²⁴¹¹wings, rho is expressed in broader domains(Fig. 5E,F). Interestingly, the C-intervein region is much less affected. This region is smaller than the corresponding wild type, probably owing to the overall wing reduction. However, as for net, neither ectopic vein tissue in the C-intervein region nor fusion of L3-L4 was found (Fig. 5C).

To determine whether ash2 function is necessary for transcriptional activation of bs, we used genetic mosaics. In wild-type third instar imaginal wings, bs is expressed in stripes corresponding to the intervein regions(Fig. 5G). When ash2^I1 mutant clones fall within intervein tissue, bs expression is reduced. The most severe cases correspond to those clones located in B or D regions, where Bs expression is completely eliminated. Downregulated expression of bs is weaker in clones located in A, C and E regions (Fig. 5H-J). bs and rho expression is only mutually exclusive during pupal development, when bs promotes intervein differentiation and is expressed in intervein regions with abrupt borders with veins (Roch et al., 1998). In ash2¹¹²⁴¹¹ pupal wings, bs is also expressed in interveins, although the size of the expression domains is smaller than the corresponding wild type. These bs-expressing domains are separated by wide stripes, corresponding to veins (Fig. 5K-M). The pattern of bs expression in ash2¹¹²⁴¹¹ interveins is non-uniform, as indicated by the presence of negatively stained cells spread throughout the wing. Accordingly,pupal ash2^I1 clones result in a clearer downregulation of bs (Fig. 5N-S).

Enhancer of split [E(spl)mβ], a gene downstream of Notch, is also expressed in the wing pouch in broad domains that correspond to most interveins (de Celis et al.,1997). However, when we generated clones of ash2^I1, we observed no effect on either E(spl)mβ or plexus (px), another intervein-associated gene (Matakatsu et al., 1999) (data not shown).

As the phenotypes of double mutants in ash2 and either net or bs are reminiscent of mutants with reduced Dpp signalling (de Celis et al.,1996), we investigated whether the expression pattern of the Dpp target gene salm is altered in these flies or in ash2 single mutants. We found that in ash2^I1/ash2^I1 discs salm is slightly downregulated(Fig. 6A,B). In addition, ash2^I1 clones resulted in weak cell-autonomous downregulation of salm (Fig. 6F-H). However, homozygous ash2¹¹²⁴¹¹ discs express salm in a central domain(Fig. 6C), as in net¹/net¹;ash2¹¹²⁴¹¹/ash2¹¹²⁴¹¹ and bs⁰³²⁶⁷/+;ash2¹¹²⁴¹¹/ash2¹¹²⁴¹¹ discs(Fig. 6D,E). In all these cases, the expression pattern resembles wild type. We also tested for possible alterations of brk, an antagonist of the Dpp signalling pathway that is expressed in peripheral cells of the wing disc in a pattern complementary to the sal-C domain (Campbell and Tomlinson, 1999; Jazwinska et al., 1999). However, ash2^I1 clones in the wing did not show any perturbation of brk expression(Fig. 6I-L).

We next examined whether the function of ash2 is exclusive to vein/intervein differentiation or whether vein positioning genes are also targeted. As ash2¹¹²⁴¹¹ shows a crooked L2 and the expression of kni-C is required to organise this vein, we studied a possible interaction between the two genes. Analysis with anti-Kni reveals kni expression in L2 and in some cells of the wing pouch margin in wild-type discs (Fig. 7A),while in ash2^I1 homozygous mutant discs its expression is expanded throughout the whole disc (Fig. 7B). Some expansion of kni expression is also found in ash2¹¹²⁴¹¹ discs, which in addition show discrete Kni-staining of L2 (Fig. 7C). We also tested the effects in mitotic clones and found that kni is upregulated in ash2^I1 clones(Fig. 7D-L). This effect is cell-autonomous. Activation of kni was observed in twin and Minute⁺ clones generated in early developing discs, as well as in small clones generated late in development, suggesting that the activation of kni by ash2^I1 can occur at any stage of larval development.

We observed that although ash2 homozygous cells within the Minute⁺ clone show a strong activation of kni,the heterozygous cells show residual activity in the whole wing disc(Fig. 7E). To test whether this residual activity is due to the lack of one copy of the ash2 gene, we generated twin clones (Fig. 7G-I) and found that the residual kni expression seen in heterozygous cells was completely missing in cells homozygous for the wild-type allele (Fig. 7J-L). This result supports kni de-repression not only in ash2homozygous cells but also in ash2 heterozygous cells. However,whereas Kni in the L2 vein is nuclear, in homozygous ash2 clones, the localisation is both cytoplasmic and nuclear(Fig. 7M,N), suggesting that although there is an increase in Kni protein, it may not be fully functional. To gain insight into the interaction between ash2 and kni we analysed mutant combinations of ash2 and the hypomorphic allele kni^ri–1. In kni^ri–1 wings,the L2 vein fails to develop and the rest of the veins develop normally(Fig. 7O), whereas in kni^ri–1 ash2¹¹²⁴¹¹ double homozygous mutants, the L2 vein is partially (25%, n=58) or almost completely restored (75%, n=58; Fig. 7P) and rho is expressed in the L2 primordium(Fig. 7Q,R). Further evidence in favour of a role for ash2 as a repressor of kni was gained by semi-quantitative RT-PCR. We found that RNA from ash2^I1 homozygous larvae contains significantly increased levels of kni, compared with wild-type larvae(Fig. 7S).

We also examined whether repression of kni by ash2 is mediated by regulators of kni expression in the wing. A well-known regulator of kni in L2 is sal-C. Low levels of sal-C activate kni expression in the presumptive L2 region,whereas higher levels repress kni-C expression(de Celis and Barrio, 2000; Lunde et al., 1998). As we observed some perturbation of salm expression in ash2^I1 tissues, we investigated whether the low levels of salm could be responsible for the ectopic kni expression. To achieve this, we generated ash2^I1 clones in UAS-salm or UAS-salr backgrounds, using a nubbin-Gal4 driver. Adult wings overexpressing salm or salr lose L2 and L5 and show severe size reduction(de Celis et al., 1996)(Fig. 8A). Surprisingly, we observed that ash2 clones generated in those flies exhibit de-repression of kni, even when high levels of Salm are maintained in the clone (Fig. 8B-E). Therefore, we conclude that de-repression of kni induced by loss of function of ash2 is independent of sal-C. This interpretation is strengthened by the observation that in mitotic clones and in homozygous ash2^I1 mutants the ectopic expression of kni is not exclusive to the sal-C domain(Fig. 7B,G). In regions outside the sal-C central domain, loss of ash2 also activates kni, in anterior as well as posterior cells.

None of the other possible activators or inhibitors of the L2 enhancer,such as sd, vg or en, showed significant alteration of their expression in ash2^I1 clones (data not shown). Moreover,using a lacZ reporter construct that includes a minimal L2 enhancer element containing only the activator and repressor sites (EX-lacZ)(Lunde et al., 2003), we observed no β-galactosidase expression in ash2^I1homozygous discs or in mitotic clones (data not shown). Taken together, these results indicate that the ash2-induced repression of kni is independent of sal-C, and suggest that ash2 could be interacting with a kni enhancer other than L2.

An important step towards understanding how cell determination is maintained is the identification of targets of the PcG and trxG genes. Although ash2 has been considered to be a member of the trxG(Adamson and Shearn, 1996), its biological and molecular function is still unknown. In this work, we show evidence that ash2 is required for normal vein/intervein patterning,and demonstrate that it plays a role in two major biological events –determination of intervein identity and maintenance of kni repression beyond L2.

Loss of ash2 function causes differentiation of ectopic vein tissue, indicating that ash2 is required for intervein development,where it functions as an activator of the intervein-promoting genes net and bs, restricting rho expression to vein regions. In addition, the loss-of-function phenotypes of Egfr alleles are rescued in ash2 mutants, while the gain-of-function phenotypes are enhanced. Furthermore, rho mRNA exhibits an expanded expression pattern in ash2 mutant tissues. Thus, ash2 promotes the maintenance of intervein fate, either by activation of net and bs or by repression of the Egfr pathway. As rho and bs/net expression is mutually exclusive, we cannot determine whether the Ash2 complex interacts directly with one or all of them. However,as bs expression is inhibited by the loss-of-function of ash2 during larval and pupal stages, we can propose that ash2 acts as a long-term chromatin imprint of bs that is stable throughout development.

Our results in adult clones and from analysis of genetic interactions suggest that ash2 acts principally by maintaining B and D intervein regions, as the C intervein remains unaltered in ash2 mutants. This region is under the control of organising genes that respond to the Hh signal(Tabata and Kornberg, 1994; Zecca et al., 1995). One of these genes is kn, which prevents vein differentiation in the C intervein (Crozatier et al.,2002; Mohler et al.,2000) and is required for the expression of bs in this domain (Vervoort et al.,1999). bs expression is regulated by two enhancer elements: the boundary enhancer, which is dependent on hh and controls bs expression in the C intervein region through kn;and another enhancer dependent on Dpp activity, which controls bsexpression in B and D intervein domains(Nussbaumer et al., 2000). Thus, the role of ash2 as a positive regulator of bs is mainly restricted to regions beyond the kn domain where the Dpp dependent bs enhancer is active.

It has been found that some combinations of dpp alleles and mosaic clones of sal-C result in elimination of B and D intervein regions,along with fusion of their flanking veins(de Celis et al., 1996). Although the genetic interactions between ash2 and either bsor net could be the result of a synergistic failure to activate genes downstream of Dpp, our results indicate that this may not be the case because salm is expressed in the central domain of the wing pouch of those mutant combinations.

It has been recently shown that another trxG complex, the Brm complex, is involved in regulating wing vein development(Marenda et al., 2004). The authors found that components of that complex interact genetically with net and bs at pupal stages to regulate the expression of rho, and that the complex is specifically required in cells within and bordering L5 to mediate proper signalling. There are some key differences between the Brm complex and Ash2: (1) Ash2 maintains bs expression from the third instar stage; (2) the Ash2 complex is mainly required for interveins B and D; and (3) the enhancement or suppression phenotypes of the genetic interactions with Egfr and intervein-promoting alleles are much stronger for ash2 than for the Brm complex. Taken together, these results suggest that ash2 plays a crucial role in intervein identity and that each trxG complex acts in a specific spatiotemporal program to maintain organ identity.

The positioning of vein tissues depends on the sal-C patterning dictated by the Dpp signalling pathway(Sturtevant et al., 1997; Sturtevant and Bier, 1995). Low levels of sal-C in the anterior compartment are required for the expression of kni-C, which triggers the differentiation of L2(de Celis and Barrio, 2000; Lunde et al., 1998). We have shown that lack of ash2 activity results in downregulation of salm and upregulation of kni. Thus, it is possible that within the sal-C domain, the ectopic expression of kni is a result of low levels of salm. However, when high levels of salm or salr are maintained by ectopic activation, lack of ash2nevertheless results in de-repression of kni. Moreover, kniis also cell-autonomously de-repressed by loss-of-function of ash2 in cells outside of the sal-C expression domain. Thus, the repression state in the whole wing must be maintained by factors other than sal-C. The kni/knirl L2-enhancer is subdivided into activation binding sites for Brk, En and the Sd/Vg complex, and repression binding sites for Sd/Vg, En, Salr and Brk(Lunde et al., 2003). We did not observe changes either in β-gal expression from the EX-lacZenhancer or in sd, vg, brk or en expression in clones lacking ash2. Therefore the de-repression of kni in ash2 mutant cells must be accounted for by a mechanism entirely different from that of the signal-dependent induction of L2, perhaps through another enhancer more global than that of L2.

The low levels of salm expression associated with ash2^I1 clones may also be explained by de-repression of kni. In dorsal tracheal cells, kni/knrl activity represses salm transcription, and this repression is essential for branch formation. Similarly the establishment of the border between cells acquiring dorsal branch and dorsal trunk identity entails a direct interaction of Knirps with a salm cis-regulatory element(Chen et al., 1998). Also in the wing, kni and knrl are likely to refine the L2 position by positive auto-regulation of their own expression and by providing negative feedback to repress salm expression(Lunde et al., 1998).

It is possible that the de-repression of kni, intervein inhibition and appearance of extra vein tissues are linked events. The kni-Ccomplex organises the development of the L2 vein by activating rhoand inhibiting bs (Lunde et al.,1998; Montagne et al.,1996). Thus, kni-C participates in L2 morphogenesis by functioning downstream of salm and upstream of vein-intervein genes. The ectopic activation of kni by lack of ash2 could trigger intervein repression and vein activation. Indeed, ectopic activation of UAS-kni results in broad expression of rho and elimination of Bs expression in pupal wings, leading to the production of solid vein material (Lunde et al., 1998). However, in adult clones not all ash2 mutant cells develop vein tissue. This raises the possibility that de-repressed kni may not be fully functional, as ectopic kni is often localised to the cytoplasm rather than the nucleus. Alternatively, ash2 could have independent functions in the wing, maintenance of the repressed state of knialongside maintenance of the intervein condition, by acting on different targets.

The ash2¹¹²⁴¹¹ mutation can partially rescue the loss of L2 in kni^ri–1 mutants. This is in contrast to our observation that the L2 enhancer appears not to mediate the effect of ash2. The kni^ri–1 allele is a 252 bp deletion in the enhancer of L2 (Lunde et al., 2003) that results in lack of kni expression in L2(Lunde et al., 1998). It has been shown, however, that it is possible to rescue the vein-loss phenotype of kni^ri–1 by expressing a UAS-rho transgene in L2 (Lunde et al., 2003). In addition, double mutant flies for kni^ri–1 and net partially rescue L2(Diaz-Benjumea and Garcia-Bellido,1990). It is therefore likely that the antagonistic effect of ash2 on rho could account for the partial rescue of L2 in kni^ri–1 ash2¹¹²⁴¹¹ wings, as rhomRNA is expressed in the rescued L2.

Some PcG genes are known to be required for the maintenance of kniexpression domains in the embryo (McKeon et al., 1994; Pelegri and Lehmann, 1994; Saget et al.,1998). It is also likely that some trxG genes or other complexes of trxG proteins, such as the Ash2 complex(Papoulas et al., 1998), may interact with repressor sequences necessary to keep kni expression in an off state beyond L2. Moreover, in a genome wide prediction screen it has been shown that kni contains PRE/TREs(Ringrose et al., 2003). Thus,we propose here that ash2 acts as regulator of kniexpression in the wing through an epigenetic mechanism of cellular memory similar to the trx-G regulation of homeotic genes, albeit that it remains to be seen whether kni is a direct or indirect target of ash2.

A well-studied mechanism through which to induce and preserve cell identities in wing imaginal discs is the response to gradients of the morphogen Dpp. This raises questions about the extent to which the response to Dpp occurs through concentration-dependent mechanisms or cellular memory. There is compelling evidence in favour of the existence of Dpp gradients that organise the pattern and growth of the wing imaginal disc(Podos and Ferguson, 1999; Strigini and Cohen, 1999). Dpp signalling causes a graded transcriptional regulation of brk by an interaction between the Dpp transducers and a brk morphogen-regulated silencer (Muller et al.,2003). Thus, brk appears to respond to direct morphogenetic signalling rather than remembering the inputs of previous developmental events. However, whereas activation of salm requires continuous signalling through the Dpp pathway(Lecuit et al., 1996; Nellen et al., 1996), other targets of Dpp, such as omb, remember exposure to the signal(Lecuit et al., 1996). We have shown here that stable regulation of other genes involved in wing development,such as kni repression, and net and bs activation,would also respond to the cellular memory conferred by epigenetic marks of the Ash2 complex. Thus, both mechanisms – morphogen-dependant, which will be required for growth and patterning, and epigenetic, which will keep specific genes in an off or on state – are likely to act simultaneously to maintain cellular identities within the wing.

Because many developmental regulators are only expressed transiently during development, the function of epigenetic complexes is likely to be very dynamic. The developmental events required for the construction of the wing,as with many other morphogenetic events, cannot only rely on an on or off state of gene expression. Instead, morphogenesis is rather malleable and epigenetic marks could act as a means to facilitate, rather that fix, the preservation of developmental fates. It may well be that the epigenetic marks of the Ash2 complex allow changes in chromatin structure to assist the access of proteins that activate or repress gene expression. From an epigenetic point of view, the ultimate refinement of morphogenesis and maintenance of cellular memory will depend upon the interaction of these chromatin remodelling complexes with the factors that trigger or inhibit transcription.

We are grateful to J. F. de Celis, F. Roch and E. Moreno for advice and constructive criticism. We thank A. García Bellido, J. F. de Celis, M. Affolter, E. Bier, E. Moreno and G. Morata for providing stocks; J. Reinitz,M. Affolter, H. Matakatsu, S. Carroll and R. Barrio for providing antibodies;and M. Noll, E. Bier and J. F. de Celis for the constructs used to make riboprobes. We also acknowledge the Bloomington Stocks Center and the Developmental Studies Hybridoma Bank for sending all the stocks and antibodies requested. We especially thank E. Moreno and S. Beltran for helpful comments on the manuscript, and M. Milán and M. Marsal for advice on pupal wings and in situ hybridisation. M.A. is the recipient of a fellowship from Universitat de Barcelona (UB). This work was supported by the Ministerio de Ciencía y Tecnologia (Spain), grants BMC2000-0766, BMC2003-05018 and HI2001-0009.