Abstract
The wave of differentiation that traverses the Drosophila eye disc requires rapid transitions in gene expression that are controlled by a number of signaling molecules also required in other developmental processes. We have used a mosaic genetic screen to systematically identify autosomal genes required for the normal pattern of photoreceptor differentiation, independent of their requirements for viability. In addition to genes known to be important for eye development and to known and novel components of the Hedgehog, Decapentaplegic, Wingless, Epidermal growth factor receptor, and Notch signaling pathways, we identified several members of the Polycomb and trithorax classes of genes encoding general transcriptional regulators. Mutations in these genes disrupt the transitions between zones along the anterior-posterior axis of the eye disc that express different combinations of transcription factors. Different trithorax group genes have very different mutant phenotypes, indicating that target genes differ in their requirements for chromatin remodeling, histone modification, and coactivation factors.
THE Drosophila eye is an excellent model system in which to study developmental processes such as specification of a tissue, propagation of a signal, or cell-cell interactions leading to cell fate determination. The eye imaginal disc is formed in the embryo and specified in the second larval instar by a hierarchy of transcription factors: the two Pax-6 homologs Twin of Eyeless (Toy) and Eyeless (Ey), the compound transcription factor formed by Eyes absent (Eya) and the homeodomain protein Sine oculis (So), and the Ski-related protein Dachshund (Dac; Boniniet al. 1993; Mardonet al. 1994; Quiringet al. 1994; Chenet al. 1997; Pignoniet al. 1997; Halderet al. 1998; Czernyet al. 1999). In the third instar, a wave of photoreceptor differentiation marked by an indentation called the morphogenetic furrow propagates across the eye disc. This wave is driven by Hedgehog (Hh), a secreted protein first expressed at the posterior margin of the disc and subsequently in the photoreceptors (Readyet al. 1976; Heberleinet al. 1993; Maet al. 1993). The eye disc differs from the wing and leg discs in that Hh indirectly activates its own expression in target cells, causing a dynamic expansion of the Hh expression domain. When cells differentiate as photoreceptors, they are no longer able to respond to Hh; thus Hh target genes are only transiently activated and must undergo rapid changes in their expression state (Schwartzet al. 1995; Dominguezet al. 1996; Strutt and Mlodzik 1996).
One target of Hh is atonal (ato), a proneural gene responsible for specifying the first photoreceptor, R8 (Jarmanet al. 1994; Dominguez 1999). The Notch pathway also contributes both to the initial activation of ato and to its restriction to a single cell (Cagan and Ready 1989; Baker and Yu 1997; Baonza and Freeman 2001). Once R8 has been determined, it recruits the remaining seven photoreceptors and four cone cells to the ommatidium by secreting Spitz, a ligand for the epidermal growth factor receptor (EGFR; Freeman 1996; Tio and Moses 1997). Another gene activated by Hh is decapentaplegic (dpp), which encodes a bone morphogenetic protein-related signaling molecule (Heberleinet al. 1993; Maet al. 1993). Dpp activates the expression of hairy (h), an inhibitor of Atonal function expressed in a region anterior to the morphogenetic furrow, known as the preproneural zone (Greenwood and Struhl 1999). In addition, Dpp represses homothorax (hth), which encodes a homeodomain protein that acts in concert with Ey and the zinc-finger protein Teashirt (Tsh) to repress expression of Eya, So, and Dac in the anterior region of the eye disc (Bessaet al. 2002). Dpp thus indirectly allows expression of these molecules in the preproneural zone. Marginal regions of the eye disc contribute to head cuticle rather than to the eye itself; their development is controlled by wingless (wg), which antagonizes photoreceptor differentiation (Ma and Moses 1995; Treisman and Rubin 1995; Royet and Finkelstein 1997).
Sequence-specific transcription factors play a critical role in directing the expression of their target genes; however, transcription is also regulated by more general factors that control chromatin structure or recruitment of the basal transcription machinery. Genes of the Polycomb group encode proteins that contribute to the repression of homeotic and other genes (Orlando 2003). Two major complexes of Polycomb group proteins have been identified. One contains Enhancer of zeste [E(z)], a SET domain protein that has recently been shown to have histone methyltransferase activity, with a preference for lysine 27 of histone H3 (Caoet al. 2002; Czerminet al. 2002; Kuzmichevet al. 2002; Mulleret al. 2002). The second complex, Polycomb repressive complex 1 (PRC1), can block nucleosome remodeling by SWI/SNF-related complexes (Franciset al. 2001). PRC1 contains Polycomb (Pc), a chromodomain protein, which binds to H3 methylated at K27 (Caoet al. 2002; Fischleet al. 2003; Minet al. 2003), allowing the complex to be recruited by E(z) activity. The complex also includes Polyhomeotic (Ph), Posterior sex combs (Psc), Sex combs on midleg (Scm), and dRING1, which have not yet been assigned enzymatic activities.
Genes of the trithorax group were identified as suppressors of Polycomb phenotypes and have therefore been implicated in activation of homeotic genes (Kennison and Tamkun 1988; Simon 1995); however, their functions are quite heterogeneous. Three members of this group, brahma (brm), moira (mor), and osa, encode components of the SWI/SNF-related Brahma chromatin-remodeling complex (Tamkunet al. 1992; Papoulaset al. 1998; Collinset al. 1999; Crosbyet al. 1999). This complex alters the positions of nucleosomes or their interactions with DNA to modulate transcription factor accessibility (Narlikaret al. 2002) and is likely to be involved in gene repression as well as activation (Holstegeet al. 1998; Collinset al. 1999; Collins and Treisman 2000; Martens and Winston 2002). Two other trithorax group genes, skuld (skd) and kohtalo (kto), encode subunits of an accessory submodule of the Drosophila mediator complex (Boube et al. 2000, 2002; Treisman 2001; Janodyet al. 2003). The mediator complex is required even in the absence of nucleosomes to link transcriptional activators or sometimes repressors to the basal transcriptional machinery (Rachez and Freedman 2001). Homologs of the Trithorax (Trx) protein, and probably also Trx itself, act as histone methyltransferases for lysine 4 of H3 (Roguevet al. 2001; Czerminet al. 2002; Milneet al. 2002; Nagyet al. 2002; Nakamuraet al. 2002); unlike K9 and K27, methylation of K4 is associated with transcriptional activation (Wanget al. 2001; Santos-Rosaet al. 2002; Nget al. 2003). Another SET domain trithorax group protein, Absent, small and homeotic discs 1 (Ash1), similarly methylates K4 and K9 of H3, as well as K20 of H4, and appears to recruit the Brm complex (Beiselet al. 2002). The function of Ash2, a PHD protein, is not known, although its yeast homolog Bre2 associates with a Trx homolog, Set1 (Adamson and Shearn 1996; Roguevet al. 2001). Other members of the group include kismet, which encodes several large chromodomain proteins (Daubresseet al. 1999; Therrienet al. 2000), Trithorax-like, which encodes GAGA factor (Farkaset al. 1994), and additional, less well-characterized genes (Kennison and Tamkun 1988; Gildeaet al. 2000; Calgaroet al. 2002; Gutierrezet al. 2003).
Almost all genes known to act in eye development are also required for embryonic survival; thus additional genes important for eye development may remain unidentified because they cause early lethality when mutated. Establishment of the FLP-FLP recognition target (FRT) system for mitotic recombination in Drosophila (Xu and Rubin 1993) and use of FLP recombinase driven by the eye-specific ey enhancer to promote recombination at high efficiency in the eye disc (Newsomeet al. 2000) have made it possible to systematically identify genes required for photoreceptor differentiation. A mosaic screen of this nature should also allow identification of novel components of the Hh, Dpp, Wg, and EGFR signaling pathways; those that are maternally contributed to the embryo and required for oogenesis would be particularly difficult to find by other methods. We have used this technique to screen the autosomes for genes required for normal photoreceptor differentiation. In addition to known and novel components of the signaling pathways that contribute to patterning the eye disc, we identified several trithorax and Polycomb group genes. The markedly different mutant phenotypes of different members of the trithorax group suggest that they are used to regulate different target genes in vivo. The rapid gene expression transitions that occur during morphogenetic furrow progression may require a variety of transcriptional control mechanisms.
MATERIALS AND METHODS
Fly stocks and genetics: For the screen, w flies carrying the FRT40, FRT42, FRT80, or FRT82 insertions were isogenized for the corresponding chromosome; males were then mutagenized with 25–35 mm ethyl methanesulfonate (EMS) and crossed to y, w, eyFLP1; FRT40 (or 42, 80, or 82), P(w+, arm-lacZ) females. Flies were allowed to lay eggs for 5 days and then discarded. F1 progeny were screened for reduced eyes containing no visible white tissue. Such flies were mated to the appropriate balancer stock (w; CyO/Sco or w; TM3/TM6B). In the next generation, three white males were mated individually to y, w, eyFLP1; FRT40, P(w+, armlacZ)/CyO (or the analogous stock for the other chromosome arms). If the reduced-eye phenotype was observed in flies carrying both FRT chromosomes, the balancer flies were used to generate a stock carrying the mutant chromosome. Complementation tests were performed with alleles of the following candidate genes: smoothened, thick veins, dpp, Mothers against dpp, wg, Protein kinase A, spitz, Star, son of sevenless, eya, dac (2L); patched, costal-2, tout velu, Epidermal growth factor receptor, downstream of receptor kinases, leonardo, so (2R); vein, daughter of sevenless, fringe, eyegone, naked (3L); supernumerary limbs, punt, Medea, pointed, ras1, ato, glass (3R). The remaining complementation groups were first mapped by crossing to the Bloomington deficiency kit for the appropriate chromosome arm. Testing likely genes in a region defined in this way allowed us to identify additional groups as lines, arrow (arr), hyperplastic discs, schnurri (shn), axin, kuzbanian, nicastrin, scribbled (scrib), brm, trx, E(z), Pc, and belle. Our connector enhancer of ksr mutations complemented the entire deficiency kit, but were identified on the basis of a phenotype resembling that of EGFR pathway mutations and were confirmed by complementation testing. Further fine-scale mapping and cloning were required to identify sightless (Lee and Treisman 2001), Myosin binding subunit, act up (Benlaliet al. 2000), kto, and skd (Treisman 2001).
—Design of the screen. (A) eyFLP was used to induce exchange of a chromosome arm between chromatids on homologous chromosomes after replication specifically in the eye-antennal disc. Separation of the chromatids at mitosis produced one daughter cell homozygous for the chromosome arm carrying an EMS-induced mutation. These mutant cells lost the P(w+) element on the wild-type chromosome arm and thus produced white clones in the adult eye. (B and C) Adult eyes with clones generated as described in A. (B) Phenotypically normal white clones. (C) shn56M clones, which produce scars in the eye with no visible white tissue. Mutations like this were selected in the screen.
The alleles of arr, lines, shn, scrib, skd, kto, brm, E(z), and Pc used here were identified in the above screen. Several of our alleles of skd and kto have been shown to introduce early stop codons (Treisman 2001). PcT181 was sequenced and shown to change K92 to a stop codon, truncating the protein shortly after the chromodomain. E(z)E4G.2 was sequenced and shown to change R622 to a stop codon, truncating the protein at the beginning of the SET domain. Thus these are likely null alleles. The other alleles we isolated have not been sequenced. The null alleles trxE2 and mor1 are described in Gindhart and Kaufman (1995), and osa308 is described in Treisman et al. (1997). Other stocks used were dpp-lacZ (Blackmanet al. 1991), UAS-Ubx (Castelli-Gairet al. 1994); y, w, eyFLP1; FRT82, M(3)96C, arm-lacZ; y, w, eyFLP1; FRT82, RpS31, Ubi-GFP; y, w, eyFLP1; FRT80, M(3)67C, Ubi-GFP; y, w, eyFLP1; FRT82, Ubi-GFP; and y, w, eyFLP1; FRT80, Ubi-GFP. To generate Ubx-expressing clones, FRT82, UAS-Ubx males were crossed to eyFLP, UAS-GFP; tub-GAL4, FRT82, tub-GAL80/TM6B females.
Immunohistochemistry: Third instar eye discs were stained as described (Hazelettet al. 1998). Antibodies used were rat anti-Elav (1:5; Robinow and White 1991), rabbit anti-β-galactosidase (Cappel, 1:5000), rabbit anti-Hth (1:500; Kurantet al. 1998), rabbit anti-Tsh (1:2000; Wu and Cohen 2000), rabbit anti-Ey (1:1000; Halderet al. 1998), mouse anti-Eya (1:1; Boniniet al. 1993), mouse anti-Dac (1:200; Mardonet al. 1994), mouse anti-H (1:5; Carrollet al. 1988), and mouse anti-Ubx (1:10; White and Wilcox 1984). Fluorescent images were obtained using a Leica TCS NT confocal microscope.
RESULTS
A mosaic genetic screen for genes required for photoreceptor differentiation: Most of the genes previously shown to play a role in early eye development were identified either through eye-specific alleles or by testing the function in the eye of genes known to be required for embryogenesis. To systematically identify genes controlling the pattern of photoreceptor differentiation only on the basis of their phenotype in the eye and regardless of earlier requirements for viability, we used a mosaic approach (Xu and Rubin 1993). Male flies isogenic for a chromosome carrying an FRT insertion close to the centromere were mutagenized with ethyl methanesulfonate (EMS) and mated to females that had both the same FRT linked to a P element carrying the white gene [P(w+)] and FLP recombinase driven by the eye-specific eyeless enhancer (eyFLP; Newsomeet al. 2000; Figure 1A). This resulted in F1 progeny with clones in the eye that were homozygous for mutations on the chromosome arm carrying the FRT and lacked w+ expression (Figure 1B). Adult flies were screened for reductions in eye size or scars in the eye accompanied by a lack of white tissue, indicating clones that failed to differentiate into photoreceptors but persisted long enough to prevent their replacement by wild-type cells (Figure 1C). Mutant chromosomes were balanced and retested by crossing to the eyFLP stock. As a secondary screen, mutant clones in the third instar eye disc were stained for the neuron-specific nuclear protein Elav (Robinow and White 1991), and wild-type tissue was marked by X-gal staining of an armadillo (arm)-lacZ insertion present on the nonmutagenized chromosome (Vincentet al. 1994). Mutations were discarded if the clones appeared wild type at this stage or if cell proliferation within the clone was severely reduced. Mutations showing defects in the pattern of photoreceptor differentiation were placed into complementation groups and tested for their ability to complement candidate genes on the same arm that were known to affect eye development (see materials and methods).
Genes identified in the screen
All four major autosomal chromosome arms, 2L, 2R, 3L, and 3R, were screened in this way. A total of 302,040 F1 flies were screened, ranging from 45,000 to 104,000 per chromosome arm. We found 2559 mutants, of which 1391 were fertile. We were able to recover and balance 613 of these; loss at this stage could have been due to mosaicism of the F1 mutants or to recombination between the mutation and the P(w+) element in female mutants. Following the secondary screen, 301 lines were retained. Their distribution between known genes, unidentified complementation groups, and single hits is given in Table 1. We identified most of the expected genes, including many components of the Hh, Dpp, Wg, and EGFR pathways. Two novel regulators of Hh signaling, sightless and hyperplastic discs, were isolated in this screen (Lee and Treisman 2001; Leeet al. 2002). Some components of these pathways may have been missed because they act nonautonomously (e.g., hh, dispatched, and spitz), are redundant (e.g., rhomboid and roughoid), are on the X or fourth chromosome or are proximal to the FRT used (e.g., shaggy, cubitus interruptus, and rolled), or are encoded by small genes. A pilot screen that would have allowed identification of nonautonomous mutations because clones were generated in a Minute background (Morata and Ripoll 1975) was unsuccessful due to high levels of lethality, presumably caused by leaky eyFLP expression in other tissues. We note that although arrow and lines are both required to mediate Wg signaling in the embryo (Hatiniet al. 2000; Wehrliet al. 2000), their phenotypes in the eye were distinct. Unlike dishevelled and arrow, lines did not induce the ectopic lateral furrows associated with loss of Wg signaling (Ma and Moses 1995; Treisman and Rubin 1995), but instead produced small clones that induced overgrowth of the surrounding tissue (Figure 2, A–C). In addition, the phenotype of schnurri clones in the eye was indistinguishable from Mad and Medea (Wiersdorffet al. 1996; Daset al. 1998), suggesting that Dpp signaling in eye development is mediated by repression of brinker (Martyet al. 2000; Figure 2, D and E).
—Phenotypes of known genes identified in the screen. In this and subsequent figures third instar eye discs are shown with posterior to the right. Photoreceptors are stained with Elav in brown in A–F, and blue X-gal staining marks wild-type tissue expressing arm-lacZ. (A) dshv26 clones. (B) arr63D clones. (C) lines13B clones. While arr and dsh both induce ectopic dorsal furrows, lines has a different phenotype. (D) MadB1 clones. (E) shn56M clones. shn has the same effect as Mad, blocking furrow initiation in posterior margin cells. (F) scribA128 clones. Mutant cells overproliferate and show reduced and disorganized differentiation.
Three components of the Notch pathway, fringe, kuzbanian, and nicastrin, were found in our screen, but did not cause very strong adult phenotypes, perhaps explaining why other components were missed. We isolated alleles of the three eye specification genes present on the chromosome arms we screened, eyes absent, dachshund, and sine oculis. In addition, we found mutations in the proneural gene atonal (Jarmanet al. 1994) and in the glass gene, which is required for normal photoreceptor differentiation and survival (Moses and Rubin 1991).
We identified several additional complementation groups as genes encoding components of the cytoskeleton. act up/capulet encodes an inhibitor of actin filament polymerization that appears to retard Hh protein transport, perhaps by promoting apical constriction in the morphogenetic furrow (Benlaliet al. 2000). Myosin binding subunit (Mbs) encodes the myosin binding subunit of myosin light chain phosphatase (Mizunoet al. 2002; Tanet al. 2003), which appears to be required to maintain photoreceptors within the eye disc epithelium (A. Lee and J. E. Treisman, unpublished results). scribbled, which encodes a component of septate junctions (Bilder and Perrimon 2000), is required to restrict proliferation and maintain normal photoreceptor differentiation (Figure 2F). As we have not systematically tested for failure-to-complement genes encoding cytoskeletal proteins, it is possible that more members of this class are present among our unidentified complementation groups.
trithorax and Polycomb group genes are required for normal photoreceptor differentiation: The largest unanticipated class of genes found in our screen was a set of general transcriptional regulators of the trithorax group (skd, kto, brm, and trx) and the Polycomb group [E(z) and Pc]. The identification of brm, which encodes the SWI2/SNF2-related ATPase subunit of the Brm chromatin-remodeling complex, is consistent with our previous observation that photoreceptor differentiation requires Osa, another subunit of this complex (Treismanet al. 1997; Collinset al. 1999). However, the phenotype of brm mutations is distinct from that of osa. In osa308 mutant clones, a reduced number of photoreceptors was still able to differentiate (Figure 3, B and E), while no Elav-expressing cells were present in brmT485 mutant clones (Figure 3A). Most brm mutant clones were extremely small, but some of the alleles we isolated (brmT485 and brmT808) allowed some growth of the mutant cells (Figure 3A). When brmT362 clones were generated in a Minute background to give the mutant cells a growth advantage, the resulting discs were very small and contained photoreceptors only within the remaining wild-type tissue (Figure 3D). Clones mutant for moira1, which encodes another essential Brm complex subunit, had a very similar effect (data not shown). The milder phenotype of osa mutant clones is consistent with our finding that Osa is present in only a subset of Brm complexes (Collinset al. 1999).
skd (previously named blind spot; Gutierrezet al. 2003) and kto encode subunits of the Drosophila mediator complex, homologs of TRAP240 and TRAP230, respectively (Treisman 2001; Janodyet al. 2003). Their mutant phenotypes in the eye disc, while identical to each other, were distinct from those of brm, mor, and osa. skd and kto mutant clones showed normal growth, but had a stronger effect than osa on photoreceptor differentiation; only a few Elav-positive cells appeared at the posterior of the mutant clones (Figure 3C; Treisman 2001). Multiple alleles, including molecular nulls, were analyzed and shown to have the same phenotype. In a Minute background, loss of skd or kto reduced the size of the eye disc, but this may have been an indirect effect due to the greatly reduced numbers of photoreceptors and therefore the reduced expression of the mitogen Hh (Figure 3F).
—trithorax and Polycomb group mutations have different effects on eye development. (A–C and G–I) Mutant clones. Wild-type tissue is marked by blue X-gal staining revealing arm-lacZ expression and photoreceptors are stained with anti-Elav in brown. (A) brmT485. (B) osa308. (C) skdT773.(G) trxE2. (H) E(z)T643. (I) PcT181. brm affects growth strongly, osa and trx affect growth more weakly, and skd does not affect growth. brm is required for photoreceptor differentiation and skd strongly affects differentiation, while osa and trx have weaker effects. E(z) and Pc affect differentiation only at the morphogenetic furrow and the posterior margin. (D–F and J–L) Clones generated in a Minute background. (D) brmT362. (E) osa308. Anti-Elav is shown in red and green fluorescent protein (GFP) in green marks wild-type tissue. (F) skdT413. (J) trxE2. Anti-Elav is shown in red, anti-β-galactosidase reveals dpp-lacZ expression in blue, and GFP marks wild-type tissue. (K) E(z)T643. (L) PcT181. Elav is stained in brown and blue X-gal staining shows dpp-lacZ expression. brm affects growth more strongly than the other mutations. skd shows reduced growth that is probably due to the lack of nonautonomous growth factors, as it is not seen in smaller clones. The effects of trx, Pc, and E(z) are restricted to particular regions of the disc.
Mutations in trx, homologs of which encode histone methyltransferases specific for H3 lysine 4 (Roguevet al. 2001; Milneet al. 2002; Nagyet al. 2002; Nakamuraet al. 2002), showed a third distinct phenotype, affecting predominantly marginal regions of the eye disc. The allele found in our screen, trx67c, behaved similarly to the amorphic allele trxE2, which we used for further analysis. trx mutant clones in the central region of the eye disc caused only a delay in photoreceptor differentiation, but clones at the posterior or lateral margins of the disc showed a strong loss of photoreceptor differentiation accompanied by overgrowth (Figure 3G). Interestingly, clones generated in a Minute background and occupying almost the entire eye disc showed normal differentiation in a central posterior region, which was surrounded by regions lacking photoreceptors and expressing dpp-lacZ (Figure 3J). Misexpression of dpp was also observed in smaller trx clones close to the morphogenetic furrow or at the posterior margin (data not shown).
We also identified mutations in two members of the Polycomb group of genes, Pc itself and E(z). The Pc and E(z) proteins are components of separate complexes required for the repression of homeotic genes. The E(z) complex has histone methyltransferase activity for lysines 9 and 27 of H3 (Caoet al. 2002; Czerminet al. 2002; Kuzmichevet al. 2002; Mulleret al. 2002). Pc is a chromodomain protein that binds to methylated H3 K27 (Caoet al. 2002; Fischleet al. 2003; Minet al. 2003). Loss of either gene in the eye disc showed a very similar phenotype: photoreceptor development was largely unaffected, but differentiation failed in some mutant clones centered on the morphogenetic furrow or at the posterior margin (Figure 3, H and I). Generating very large clones lacking Pc or E(z) function in a Minute background likewise produced loss of photoreceptor differentiation and dpp expression only in a region of the disc close to the furrow (Figure 3, K and L). Both complexes are thus likely to act on the same genes during eye development.
—Hth, Tsh, and Ey show different patterns of regulation by trithorax and Polycomb group genes. (A–F) trxE2 mutant clones. (G and H) ktoT241 mutant clones. (I–L) ktoT555 mutant clones. (M–P) E(z)E4G.2 mutant clones. (Q and R) E(z)T643 mutant clones. Wild-type tissue is marked with GFP in B, D, F, H, J, L, N, P, and R. Hth is white in A, G, and M and magenta in B, H, and N. Ectopic Hth is indicated by an arrow in B, and loss of Hth is indicated by an arrow in H. Tsh is white in C, I, and O and magenta in D, J, and P. Ey is white in E, K, and Q and magenta in F, L, and R. Hth is repressed by trx and E(z) but activated by kto. Tsh is repressed by E(z) but activated by trx and kto. Ey is repressed by kto but activated by trx and E(z).
Transitions in gene expression are differently regulated by trithorax and Polycomb group genes: Eye development requires a rapid series of transitions in gene expression as the morphogenetic furrow traverses the eye disc. The trithorax and Polycomb group genes are thought to be involved in the maintenance, respectively, of activated and repressed states of gene expression. To explain their effects on photoreceptor differentiation, we examined how transitions between different gene expression domains were affected in the absence of these genes. The most anterior domain of the disc gives rise to head cuticle and expresses the homeobox gene hth, while the adjacent domain of the eye disc proper expresses the transcription factors encoded by ey and tsh, in addition to hth (Bessaet al. 2002). We did not observe significant changes in the expression of any of these genes in brmT362 or osa308 mutants (data not shown). In anterior trxE2 mutant clones, expression of ey and tsh was lost, while in clones at the posterior or lateral margins, hth was misexpressed (Figure 4, A–F). skd or kto mutant clones caused a reduction in hth and tsh expression, as well as autonomous maintenance of ey expression in all clones posterior to its normal domain (Figure 4, G–L; Treisman 2001). E(z)T643, E(z)E4G.2,PcT181, and PcT351 clones showed a variable reduction in ey expression, misexpression of hth in some posterior clones, and strong posterior misexpression of tsh (Figure 4, M–R). tsh expression was often associated with overgrowth at the posterior margin, and mutant cells tended to sort out to the basal region of the disc in internal clones (data not shown). This shows that the three genes are under independent regulatory control. In addition, trithorax and Polycomb group genes do not always have opposing functions: ey and hth are regulated similarly by trx and E(z)/Pc and oppositely by trx and kto/skd.
—Establishment of the preproneural domain requires both trithorax and Polycomb group genes. (A–F) trxE2 mutant clones. (G–J) ktoT241 mutant clones. (K and L) skdT342 mutant clones. (M and N) PcT181 mutant clones. (O–R) E(z)T643 mutant clones. Wild-type tissue is marked with GFP in B, D, F, H, J, L, N, P, and R. Dac is white in A, G, and M and magenta in B, H, and N. Eya is white in C, I, and O and magenta in D, J, and P. Hairy is white in E, K, and Q and magenta in F, L, and R. High-level expression at the anterior edge of the expression domain of Eya, Dac, and H requires trx, skd, kto, Pc, and E(z).
The morphogenetic furrow is preceded by a region known as the preproneural domain, in which eya, so, dac, and h are expressed (Greenwood and Struhl 1999; Bessaet al. 2002). The initiation of this domain appeared to be delayed in trx mutant cells; h expression was reduced, and dac and eya were lost from mutant clones in the anterior region of their expression domains (Figure 5, A–F). skd and kto were also required for normal levels of h and eya expression, but in their absence dac was initiated normally and inappropriately maintained in posterior regions of the disc (Figure 5, G–L; Treisman 2001). Interestingly, Pc and E(z) mutant clones resembled trx mutant clones in this domain; h was reduced, while eya and dac were lost from anterior regions of their expression domains (Figure 5, M–R). Loss of osa had no apparent effect on genes expressed in this domain, but we observed reduced and posteriorly shifted dac expression when brmT362 clones were generated in a Minute background (data not shown).
Ultrabithorax is not the only target of E(z) and Pc in the eye disc: Many of the effects of Polycomb group mutations have been attributed to the derepression of homeotic genes, and misexpression of homeotic genes in the eye disc can prevent eye development by inhibiting Ey function (Plazaet al. 2001; Benassayaget al. 2003). We found that both Pc and E(z) mutant clones in the eye disc strongly misexpressed the homeotic gene Ultrabithorax (Ubx), but not Abdominal-B (Abd-B) or Antennapedia (Antp; Figure 6, A, C, and H, and data not shown). However, the ectopic Ubx did not appear to be the cause of hth misexpression, as Ubx was excluded from regions of the clones that expressed Hth (Figure 6, C–E). To test whether Ubx misexpression could be responsible for other aspects of the Pc and E(z) phenotypes, we generated clones ectopically expressing UAS-Ubx from a tubulin-GAL4 driver. Such clones were able to induce tsh expression and block photoreceptor differentiation in posterior regions of the eye disc (Figure 6, F and G). However, in discs largely lacking Pc or E(z) function, Tsh was misexpressed in some cells that did not express Ubx (Figure 6, H–J). We conclude that Pc and E(z) are likely to have targets in addition to Ubx in the eye disc and might themselves directly regulate hth and/or tsh. Misexpression of Hth and Tsh could be responsible for the downregulation of h, eya, and dac in Pc and E(z) mutant clones (Bessaet al. 2002).
—Ubx is not the only target of Polycomb group genes in the eye disc. (A and B) PcT181 mutant clones. Ubx is white in A and magenta in B, and wild-type tissue is marked with GFP in B. (C–E) E(z)T643 mutant clones. Ubx is white in C and blue in E, Hth is white in D and red in E, and wild-type tissue is marked with GFP in E. The arrow in E indicates mutant cells that express Hth but not Ubx. (F and G) Clones expressing UAS-Ubx under the control of tubulin-GAL4. Elav is stained red, Tsh is stained blue, and Ubx is stained green in F. (H–J) E(z)E4G.2 clones made in a Minute background. Ubx is white in H and red in J, Tsh is white in I and blue in J, and wild-type tissue is marked with GFP in J. The arrow in J indicates mutant cells that express Tsh but not Ubx. Ectopic Ubx can activate Tsh expression, but within E(z) or Pc mutant tissue Tsh and Hth can be misexpressed in cells that do not express Ubx.
DISCUSSION
Genes required for photoreceptor differentiation: Using a systematic genetic screen, we have attempted to define the set of genes required for photoreceptors to differentiate in their normal numbers. The known genes that we identified include components of all the pathways previously shown to be involved in photoreceptor differentiation, as well as genes encoding the transcription factors Eya, So, Dac, Ato, and Glass. We have shown that two additional genes, sightless and hyperplastic discs, encode novel components of the Hh pathway (Lee and Treisman 2001; Leeet al. 2002). It is possible that other components of these pathways will be found among the complementation groups that are still unidentified. We have also found evidence that differentiation can be affected by defects in the cytoskeleton that may alter cell shape or cell motility (Benlaliet al. 2000; A. Lee and J. E. Treisman, unpublished results). In addition, several members of the trithorax and Polycomb groups of transcriptional regulators had striking effects on photoreceptor differentiation and were identified in our screen.
The screen was probably not fully saturating. Although we were able to find multiple alleles even of quite small genes such as act up [424 amino acids (aa)], so (416 aa), fringe (412 aa), downstream of receptor kinases (211 aa), and ras1 (189 aa), many of the genes we found were identified only by single alleles. In addition to size, the probability of our finding mutations in a gene depended on the severity of its phenotype and was reduced if only clones in a particular region of the eye disc had a visible phenotype; for instance, dac clones cause photoreceptor loss only if they occur at the posterior margin. In addition, we could not find redundant genes, such as rhomboid and roughoid (Wassermanet al. 2000). Finally, we were unable to screen the regions proximal to the FRT sites or the fourth chromosome and have not yet screened the X chromosome. However, it may be possible to draw some conclusions about the prevalence of certain classes of genes. Among our unidentified mutations, none have a phenotype resembling that of the eye specification genes eya, so, and dac. Unless additional genes of this class are clustered in a region that was not screened, they are unlikely to be very numerous. It has been proposed that optix also encodes an eye specification gene (Seimiya and Gehring 2000). However, although we found five alleles of so, which encodes a smaller protein than optix does, we did not find any gene with a similar phenotype mapping to the same region as optix (44A).
—Effects of trithorax and Polycomb group genes on domains of transcription factor expression in the eye disc. Anterior-posterior domains of gene expression are depicted as boxes, with anterior to the left. Dashed lines delimit the preproneural domain and thick arrows mark the position of the morphogenetic furrow. (A) The effects of trithorax group genes. Trx activates ey and tsh and may therefore indirectly activate genes in the preproneural domain, since Ey and Tsh in combination can activate dac and eya (Bessaet al. 2002). Skd and Kto activate hth in the anterior of the disc and repress ey, dac, and dpp in posterior regions. They also contribute to the activation of h and eya, perhaps by promoting the effects of Dpp on transcription. (B) The effects of Polycomb group genes. Pc and E(z) repress hth and tsh posteriorly to their normal expression domains, as well as repress inappropriate expression of Ubx; Ubx may contribute to the activation of tsh. Ectopic Hth and Tsh may be responsible for the repression of eya and dac (Bessaet al. 2002). Arrows in this diagram do not imply that any of the interactions are direct.
Polycomb group genes repress Ubx, tsh, and hth: We observed very similar phenotypes in clones mutant for Pc or E(z), which encode components of two distinct complexes implicated in transcriptional repression. Although we used likely null alleles for both genes (see materials and methods), the phenotype of E(z) clones appeared slightly stronger, with a greater likelihood of derepressing hth in posterior regions of the eye disc. The E(z) protein has been shown to act as a histone methyltransferase for H3 K27 within a complex that also includes Extra sex combs (Esc), Suppressor of zeste 12 [Su(z)12], and the histone-binding protein NURF-55 (Caoet al. 2002; Czerminet al. 2002; Kuzmichevet al. 2002; Mulleret al. 2002). esc appears to act only early in embryonic development (Struhl and Brower 1982; Sathe and Harte 1995; Simonet al. 1995), while E(z) and Su(z)12 are continuously required to repress inappropriate homeotic gene expression in wing imaginal discs (Jones and Gelbart 1990; LaJeunesse and Shearn 1996; Birveet al. 2001; Czerminet al. 2002; Mulleret al. 2002). The core PRC1 complex contains Pc, as well as Ph, Psc, and dRing1, and can prevent SWI/SNF complexes from binding to a chromatin template (Franciset al. 2001). Pc, Psc, and ph are all required to prevent homeotic gene misexpression in wing discs; however, Psc and ph act redundantly with closely related adjacent genes (Beuchleet al. 2001). The two complexes are thought to be linked through binding of the Pc chromodomain to K27-methylated H3 (Caoet al. 2002; Fischleet al. 2003; Minet al. 2003). The stronger phenotype of E(z) mutations in the eye disc might suggest that methylation of H3 K27 can recruit other proteins in addition to Pc.
In the eye disc, loss of E(z) or Pc leads to misexpression of the homeotic gene Ubx, but this does not seem to account for the entire phenotype. Although Ubx is sufficient to turn on tsh ectopically, misexpression of hth and tsh can occur in E(z) or Pc clones in which Ubx is not misexpressed. This suggests that hth and tsh are either direct targets of Pc/E(z)-mediated repression or targets of a downstream gene other than Ubx, possibly one of the homeotic genes that we did not examine (Figure 7B). Tsh misexpression would be sufficient to explain the suppression of photoreceptor differentiation in clones close to the morphogenetic furrow, since it is able to maintain expression of Hth and Ey and, in combination with them, to repress eya (Bessaet al. 2002). Misexpression of Tsh can also account for overgrowth of Pc or E(z) mutant cells at the posterior margin of the eye disc (Bessaet al. 2002).
trithorax group genes have a variety of distinct functions: trithorax group genes were initially identified as suppressors of Polycomb phenotypes (Kennison and Tamkun 1988) and are therefore thought to contribute to the activation of homeotic gene expression. Some members of the group encode components of the Brahma chromatin-remodeling complex (Tamkunet al. 1992; Papoulaset al. 1998; Collinset al. 1999; Crosbyet al. 1999), others encode components of the mediator coactivation complex (Boubeet al. 2000; Treisman 2001), and still others encode histone methyltransferases (Roguevet al. 2001; Beiselet al. 2002; Czerminet al. 2002; Milneet al. 2002; Nagyet al. 2002; Nakamuraet al. 2002) or less-well-characterized proteins (Farkaset al. 1994; Adamson and Shearn 1996; Daubresseet al. 1999; Gildeaet al. 2000; Therrienet al. 2000; Calgaroet al. 2002; Gutierrezet al. 2003). Our analysis shows that in addition to their distinct biochemical functions, members of the trithorax group act on different sets of target genes during eye development and can also have different effects on the same target genes. Components of the Brahma complex are strongly required for cell growth and/or survival; brm and mor, but not osa, are also absolutely required for photoreceptor differentiation. However, these three genes do not seem to be required for the restricted expression in anterior-posterior domains of the eye disc of the transcription factors we examined. In contrast, the mediator complex subunits encoded by skd and kto are not required for cell proliferation, although they are strongly required for photoreceptor differentiation (Treisman 2001). trx, which encodes a histone methyltransferase (Czerminet al. 2002), is required primarily for the normal development of marginal regions of the disc. We have not seen any significant effect on photoreceptor differentiation in clones mutant for kismet1, which encodes chromodomain proteins (Daubresseet al. 1999; Therrienet al. 2000), or ash21, which encodes a PHD protein (Adamson and Shearn 1996; data not shown). These differences are unlikely to be due to different expression patterns of the trithorax group genes, as Trx, Skd, Kto, and Osa are ubiquitously expressed in the eye disc (Kuzinet al. 1994; Treismanet al. 1997; Janodyet al. 2003).
The effects of these genes on the rapid transitions between domains of expression of different transcription factors are of particular interest (Figure 7A). In the most anterior region of the eye disc, hth expression is enhanced by skd and kto. The domain just posterior to this also expresses tsh and ey, and activation of both of these genes requires trx. However, skd and kto have opposite effects on the two genes, enhancing tsh expression and preventing the maintenance of ey expression in posterior cells. Since Hth and Tsh can positively regulate each other's expression (Bessaet al. 2002), it is possible that only one of these genes is directly dependent on skd and kto. Next, dac and h are activated transiently and eya is activated and sustained. The establishment of both dac and eya is delayed in trx mutant clones, and h expression is reduced. This delay in establishing the preproneural domain may be due to the failure to activate ey and tsh earlier in development, since Ey and Tsh combine to activate eya (Bessaet al. 2002). The effect of Pc or E(z) mutations on eya, dac, and h appears very similar to the effect of trx mutations. However, in Pc or E(z) clones, the delay in eya and dac expression is likely to be caused by the failure to repress tsh and hth, since the combination of these two proteins has been shown to repress genes expressed in the preproneural domain (Bessaet al. 2002). skd and kto clones also show a reduction in h and anterior eya expression, but an inappropriate maintenance of dac and dpp (Treisman 2001). These mediator complex components may thus contribute both to the activation of genes such as h in the preproneural domain and to the activation of unknown genes that shut off the expression of ey, dac, and dpp. Alternatively, skd and kto could be directly involved in the repression of these genes. Finally, trx is important to prevent misexpression of hth in cells near the posterior and lateral margins. Although Dpp normally represses hth (Bessaet al. 2002), in trx mutant clones dpp and hth are both inappropriately expressed in marginal cells. This may reflect a role for trx in the process of morphogenetic furrow initiation, perhaps contributing to the ability of dpp to control gene expression.
Further study will be needed to determine which genes are direct targets of each trithorax group protein. However, our results point to a strong specificity of these general transcriptional regulators, suggesting that they may be specialized to mediate the effects of particular signaling pathways (Collins and Treisman 2000; Janodyet al. 2003) or to control specific subsets of downstream genes.
Acknowledgments
We thank Tetsu Akiyama, Nick Baker, Sean Carroll, Steve Cohen, Mark Fortini, Iswar Hariharan, Ulrike Heberlein, Ken Irvine, GraemeMardon, Richard Mann, Marek Mlodzik, Francesca Pignoni, Laurel Raftery, Adi Salzberg, Allen Shearn, Mike Simon, Nicole Theodosiou, Uwe Walldorf, the Bloomington Drosophila Stock Center, and the Developmental Studies Hybridoma Bank for fly stocks and reagents. We particularly thank Barry Dickson for allowing us to use eyFLP for our screen prior to its publication. We are grateful to Ruth Lehmann for the use of her confocal microscope. The manuscript was improved by the critical comments of Inés Carrera, Kerstin Hofmeyer, and Arnold Lee. This work was supported by National Institutes of Health grants EY-13777 and GM-56131, National Science Foundation grant IBN-9728140, and the Irma T. Hirschl/Monique Weill-Caulier Charitable Trust. F.J. is the recipient of an Association pour la Recherche sur le Cancer postdoctoral fellowship and J.D.L. and G.I.M. are National Institutes of Health training grant recipients.
Footnotes
-
Communicating editor: S. Henikoff
- Received July 25, 2003.
- Accepted September 9, 2003.
- Copyright © 2004 by the Genetics Society of America
