muscleblind, a gene required for photoreceptor differentiation in Drosophila, encodes novel nuclear Cys3His-type zinc-finger-containing proteins

Inductive signalling plays a crucial role in the patterning of many tissues and organs in vertebrates and invertebrates. The assembly and patterning of the Drosophila eye has served as a good model system to study such inductions and signalling pathways (for reviews see Wolff and Ready, 1993; Dickson and Hafen, 1993). Ommatidial assembly follows an invariant sequence of cell fate inductions. The photoreceptor R8 cell is the first to differentiate, followed by the sequential and pairwise addition of photoreceptors R2/5 and R3/4. Following a new round of mitosis, the remaining cells are being recruited to the growing ommatidial cluster: photoreceptors R1/6, R7, the non-neuronal cone cells and the other accessory cells (reviewed in Tomlinson, 1988; Ready, 1989). Over the past few years, a wealth of information has been accumulated leading to a fairly good understanding of the signalling pathways during ommatidial assembly and photoreceptor development. Activation of the Ras/EGF-receptor pathway causes induction of precursor cells as photoreceptor neurons (reviewed by e.g. Zipursky and Rubin, 1994; Freeman, 1997). This is followed by the establishment of distinct photoreceptor subtypes, probably achieved by the expression of different specific factors (e.g. Chang et al., 1995; Dickson et al., 1995). Much less is known about the molecular basis of subsequent differ-entiation that leads to the establishment of functional photoreceptor neurons.

The differentiation of each photoreceptor subtype is accomplished by a series of intercellular signals between differentiating cells and uncommitted precursor cells. Pivotal to this process is the activation of the Ras-mediated signalling pathway via the activation of at least two distinct receptor tyrosine kinases, the Drosophila EGF-receptor (DER; Xu and Rubin, 1993; Freeman, 1996), and Sevenless (Sev; Hafen et al., 1987; Simon et al., 1991; Fortini et al., 1992). DER-mediated activation of the Ras pathway is required for the initial determination of all cells in the ommatidium, including cone and pigment cells (Freeman, 1996). This is reflected in the transient activation or inactivation of the nuclear proteins Pointed (Pnt) and Jun (Brunner et al., 1994; O’Neill et al., 1994; Bohmann et al., 1994; Treier et al., 1995), and Yan (Rebay and Rubin, 1995), respectively (reviewed in Dickson, 1995). The Sevenless kinase is activated specifically in the R7 cell by its ligand Boss (reviewed in Zipursky and Rubin, 1994). It appears that Sevenless enables the R7 precursor to continue differentiation by a second burst of Ras signalling, since a number of genes marking an earlier differentiation are already expressed in this cell prior to Sevenless activation (e.g. klingon (H214) and prospero ; Mlodzik et al., 1992; Kauffmann et al., 1996; Butler et al., 1997).

Precursor cells of the outer photoreceptors, R1-6, express specific transcription factors that determine their developmental fate. One of these is Seven-up (svp), a member of the nuclear hormone receptor superfamily, which is required for R3/4 and R1/6 subtype identity (Mlodzik et al., 1990). Svp acts cell autonomously and, in the absence of its function, these precursors do not realise their developmental program, but develop as R7 cells instead. Misexpression experiments have shown that Svp is able to confer neuronal development upon the non-neuronal cone cell precursors, acting as a positive regulator of neuronal fate rather than inhibiting R7 development (Begemann et al., 1995; Kramer et al., 1995). Expression of svp is negatively regulated by the AML/runt family transcription factor Lozenge (lz) (Daga et al., 1996). Lz restricts svp expression to the four outer photoreceptors, as eye discs mutant for a null allele of lz ectopically express svp in precursors of R7 and the cone cells. Lz and svp are also required for the R1/R6-specific expression of BarH1/H2, and svp controls the R3/R4 expression of pipsqueak (Higashijima et al., 1992; Weber et al., 1995; Daga et al., 1996).

Once a precursor cell fate has been established, an as yet poorly characterised set of genes ensures the subsequent cellular morphogenesis of photoreceptors, which concludes with the differentiation of the highly specialised light harvesting structures, the rhabdomeres. A well-studied gene in this context is the zinc-finger-containing transcription factor Glass, which was shown to directly activate transcription of rhodopsin 1 (Moses and Rubin, 1991) and chaoptin, another late marker of photoreceptor differentiation (Moses et al., 1989). Glass (Gl) is generally required for photoreceptor differentiation. This is exemplified in weak gl alleles, where photoreceptor cells die in mid-pupal stages when they normally would produce the photosensitive organelles (Ready et al., 1976). Mutations in elav (Homyk et al., 1985), ultraspiracle (Oro et al., 1992) and orthodenticle (Vandendries et al., 1996) show defects in photoreceptor cell morphology, such as disorganisation or disruption of the rhabdomeres. The role of these genes in eye development remains, however, unclear.

Transgenic Drosophila lines that ectopically express Svp in R7 and cone cell precursors (sev-svp2) have a dosage-sensitive rough eye phenotype due to the transformation of an increas-ing number of cone cells into photoreceptors (Hiromi et al., 1993; Begemann et al., 1995). To isolate new genes that are required in photoreceptor development or differentiation, we have taken advantage of the dosage sensitivity of the sev-svp2 phenotype and screened for mutations that dominantly modify the sev-svp2 phenotype. This approach has previously identi-fied components of the DER/Ras pathway as essential factors for Svp function (Begemann et al., 1995; Kramer et al., 1995), and also new genes required for photoreceptor development, patterning and differentiation (G. B., Caryl A. Mayes and M.M., unpublished).

Here, we report the isolation of the embryonic lethal mutation muscleblind (mbl). We show that mbl acts as a dominant suppressor of the sev-svp2 phenotype. Clones mutant for mbl indicate that it is autonomously required for photore-ceptor differentiation. Mutant cells are recruited into develop-ing ommatidia and initiate neural development, but they fail to properly differentiate, e.g. to form rhabdomeres. Molecular analysis reveals that the mbl locus is large (>100 kb) and complex, giving rise to different proteins with common 5′ sequences but different lengths and carboxy termini. Mbl proteins are nuclear and share a Cys3His zinc-finger motif, which is also found in the TIS11/NUP475/TTP family of proteins and is highly conserved in vertebrates and invertebrates. Functional analysis of mbl and the observation that it also dominantly suppresses the sE-JunAsp gain-of-function phenotype suggest that it is a gene required for photoreceptor differentiation in general.

The 2xsev-svp2line used in the screen (Begemann et al., 1995) carries two insertions of the sev-svp2 construct on chromosome III (Hiromi et al., 1993). The sE-Jun^Asp line used for the isolation of mbl²⁵⁶³ is as described (Treier et al., 1995). Several P-element lines were used for the characterisation of the mbl locus: PP2 was a gift of A. MartinezArias (University of Cambridge, UK), l(2)k05507 was originally isolated in the screen of the P-element collection (Török et al., 1993) and l(2)2563 was isolated in the sE-Jun^Asp screen from a P-element collection covering both the second and third chromosome (Erdelyi et al., 1995). To confirm that the observed suppression and phenotypes were linked to these P-insertions, the l(2)k05507 and l(2)2563 P-lines were mobilised by providing an external source of transposase (Robertson et al., 1988) and their progeny was screened for flies that had lost the P[white⁺] marker. For both P-elements, several excisions resulted in homozygous viable flies that lost their ability to suppress sev-svp or sE-JunAsp. In addition, several imprecise lethal excisions were recovered in these experiments. Some of these were used in the present study: mbl^E27 and mbl^E127 are derived from l(2)k05507, and mbl^E2 from l(2)2563.

cDNA isoforms A and B were cloned into the transformation vector pKB267PL (Basler et al., 1991; Zeidler et al., 1996), which carries the eye-disc-specific sev-enhancer expression module driving expression in the precursors for R3/R4, R7 and the cone cells, and weaker in the mistery cells and R1/6 during ommatidial assembly. Glass-driven overexpression was achieved by cloning the cDNA of isoform mblA into the pGMR vector (Hay et al., 1994). Germline transformation was performed by standard procedures (Spradling and Rubin, 1982).

Somatic clones were induced by X-ray with 1000 rads in 1st instar larvae, using either a P[white+ ] element at 49D as autonomous clonal marker for excision alleles or a w− background for the P insertion lines. Homozygous clones of mbl^E27 and mbl^E127 mutants were generated in imaginal discs with the FLP/FRT system (Xu and Rubin, 1993). Larvae of the genotype w hsFLP1; P[ry +; hs-neo; FRT] 43D, mbl^E27 or E127 /P[ry+; hs-neo; FRT] 43D, P[mini-w+; armLacZ] 51A were heat shocked at 38°C for 120 minutes at 24-48 and 48-72 hours after egg laying. Clones were marked by loss of arm-LacZ staining (Vincent et al., 1994).

β-galactosidase was detected by either activity or antibody staining as previously described (Tomlinson and Ready, 1987). Anti-Elav and anti-β-Gal double stainings were done in 0.1 M phosphate buffer, 0.2% saponin, 0.3% deoxycholate, 0.3% Triton X-100 and 10% normal donkey serum. The same conditions were used in anti-β-Gal double stainings with anti-Chp and anti-Bar. Elav was detected with a rat monoclonal antibody (gift of G. Rubin), Chp with the mouse monoclonal antibody mAb24B10 (Van Vactor et al., 1988), Bar with a rabbit polyclonal serum (Higashijima et al., 1992), and β-galactosi-dase with a mouse monoclonal antibody (Promega). Secondary anti-bodies were purchased from the Jackson Labs. Standard histological methods were used for sections of adult eyes (Tomlinson and Ready, 1987). The enhancer trap line H214 (Mlodzik et al., 1992) was used to detect R7 cells in 3rd instar eye discs. To monitor Rhodopsin expression, we used the Rh3-lacZ and Rh4-lacZ strains (Fortini and Rubin, 1990); detection in whole-mount retinas was performed as described earlier (Weber et al., 1995).

DNA sequences flanking the P-element insertion points were recovered using the plasmid rescue method (Wilson et al., 1989). All rescued plasmids map to polytene bands 54A1-3.

Lambda and cosmid clones from the 54A1-3 genomic region were isolated from an Oregon R DNA library in λEMBL3 (from J. Tamkun) and from a Canton S DNA library constructed in cosmids (Hoheisel et al., 1991). mbl cDNAs were isolated from several libraries by using different genomic fragments as probes. This led to the identification of four cDNA isoforms: A and B derived from the disc-specific library, and C and D from the embryonic libraries. Genomic and cDNA fragments subcloned in pBluescript or pUC were sequenced by standard procedures. DNA and protein sequences were analysed using the GCG software package (Wisconsin Genetics Computer Group). Similarity searches were carried out using the BLAST network service located at http://www.ncbi.nlm.nih.gov/BLAST/.

Mbl ORF A was PCR-amplified and cloned in the expression vector pFP98 as an NdeI-EcoRI fragment (gift of F. Peverali). This vector is derived from the pGEX-2T bacterial expression vector and contains a histidine tag. A fusion protein GST-MblA-His of 48 kDa was obtained and injected into rabbits using standard procedures. The serum was used at a 1:100 dilution in 0.1 M phosphate buffer, 0.2% saponin, 0.3% deoxycholate, 0.3% Triton X-100 and 10% normal donkey serum. No staining was detected with the preimmune serum.

A collection of lethal P-element insertions on the second chromosome (Török et al., 1993) was screened for dominant modifiers of the sev-svp2 phenotype. One of the lines identified as a suppressor carried two P-element insertions at polytene chromosome bands 42D4-5 and 54A1-3. After their separation by meiotic recombination, the insertion at 54A1-3 (subsequently referred to as l(2)k05507) was found to interact with sev-svp2 (Fig. 1). The chromosome carrying this insertion was further associated with embryonic lethality and defects in homozygous mutant eye clones (see below). The second insertion from the original line did not interact with sev-svp2. To confirm that the observed suppression and clonal phenotypes (see below) were linked to the insertion, we remobilised the P-element l(2)k05507 (Robertson et al., 1988) and screened the progeny for flies that had lost the P[white+] marker. Several excisions of the P-element resulted in homozygous viable flies that lost their ability to suppress sev-svp2. Adult eyes of eight such independent lines were sectioned and did not show any aberrant ommatidia, indicating that no second site mutation was responsible for the lethality or the eye phenotype (see below).

The l(2)k05507 insertion acts also as an enhancer-detector with expression in the eye disc in cells posterior to the mor-phogenetic furrow (MF, Fig. 2). Independently, we have identified another P-element insertion at 54A1-3, PP2, which is homozygous viable with a wild-type appearance, that had an identical expression pattern in the eye disc (Fig. 2). Moreover, both lines, l(2)k05507 and PP2, showed expression in embryonic muscle precursor cells and developing muscles (not shown), accounting for the embryonic lethality of embryos homozygous for the l(2)k05507 insertion (R. A., unpublished data). Reflecting the phenotypes of the associated gene in the development of the visual system (see below) and the muscles, we named this gene muscleblind (mbl; the P-allele will subsequently be referred to as mbl⁰⁵⁵⁰⁷).

Genes that are required during photoreceptor differentiation or are acting together with Svp during this process should affect photoreceptor development in general or have eye phenotypes either similar to or overlapping with that of svp. In order to elucidate if mbl⁰⁵⁵⁰⁷ gives rise to such a mutant phenotype, mutant clones of mbl⁰⁵⁵⁰⁷ were analyzed in adult eyes. They occurred at a normal frequency of about 1/50 eyes, suggesting that mbl is not required for cell proliferation or survival. Although such clones appeared externally wild type, tangential sections revealed phenotypic defects in many mutant ommatidia: the rhabdomeres of up to 4 outer photoreceptors (mostly R2/5 and R3/4) had a smaller diameter resembling morphologically the small rhabdomeres of the central R7 photoreceptor (Fig. 3A). More basal sections of the same ommatidia revealed that all photoreceptors could be affected in mbl⁰⁵⁵⁰⁷ clones, with malformed rhabdomeres that frequently did not extend into basal regions of the retina (not shown).

Following plasmid rescue from mbl⁰⁵⁵⁰⁷ flies of flanking genomic DNA, we isolated overlapping genomic clones and 10 different cDNAs (Materials and Methods). All cDNAs were shown by in situ hybridisation to be expressed in eye imaginal discs (data not shown) and represented two different splice variants, mblA and mblB. Further screens in embryonic cDNA libraries have identified at least two additional transcripts, mblC and mblD that share common 5′ exons (Fig. 4) indicating that this locus is subject to alternative splicing. Sequencing of the rescued plasmids confirmed that the P-element had inserted within the common 5′ untranslated region (UTR) of exon 1, 6 bp upstream of the exon/intron border (Fig. 4). The PP2 P-element insertion maps approximately 500 bp upstream of mbl⁰⁵⁵⁰⁷ and upstream of the common 5′ end of the respective cDNAs isolated.

In order to address the question of whether the isolated cDNAs were indeed linked to the mbl⁰⁵⁵⁰⁷ phenotype, we asked whether additional copies of the putative mbl cDNAs are capable of reverting the mbl⁰⁵⁵⁰⁷-caused suppression of the sev-svp2 phenotype. To this end, type A and B cDNAs were cloned into a sevenless-enhancer expression vector, subsequently yielding sE-mblAand sE-mblB flies that were over-expressing the mbl A or B open reading frames (see below) in photoreceptors R3/4, R1/6, R7 and in the cone cells.

Although the transgenic lines did not show an eye phenotype with one or two copies of the construct on their own, when they were combined with the sev-svp2 tester strain and the mbl⁰⁵⁵⁰⁷ P-element, these constructs reverted the suppression of sev-svp2 normally observed in the presence of mbl⁰⁵⁵⁰⁷ (data not shown). Moreover, when crossed to sev-svp2 in an otherwise wild-type background, the rough eye phenotype of sev-svp2 was dramatically enhanced. While only approxi-mately 15% of ommatidia in a sev-svp2 background have additional outer photoreceptors, many ommatidia in sevsvp2; sE-mblA or sev-svp2; sE-mblB flies contain up to 11 outer photoreceptors (Fig. 1D). This phenotype is reminis-cent of 4 copies sev-svp2 or of genotypes, in which sev-svp2 was combined with activated components of the Ras-signalling pathway (Begemann et al., 1995). Taken together with the molecular evidence, we conclude that the isolated gene is disrupted in mbl05507 and is responsible for the interaction with sev-svp2.

Homozygous clones of P-element mbl⁰⁵⁵⁰⁷ show phenotypes only in a subset of mutant ommatidia. This relatively variable phenotype and the molecular nature of the allele (insertion in the 5′-UTR, Fig. 4) suggested that this original mbl allele was a hypomorphic allele. To isolate additional, possibly stronger, alleles of mbl, we remobilised the P-element and selected for imprecise excisions that affect the respective gene. We obtained several lethal excisions that were allelic to mbl⁰⁵⁵⁰⁷, suppressed sev-svp2 and were mapped by Southern analysis to the genomic map (Fig. 4 and not shown). Excisions E27 and E127 removed (at least partially) exons 1 and 2 or just exon 1, respectively. In particular, excision E27 takes out 2 regions containing the splice-donor site and acceptor sites of exons 1 and 2 and parts of the second exon that encodes the N terminus (Fig. 4). Independently, a new strong allele, mbl²⁵⁶³, was identified in a collection of lethal P-element insertions (Erdelyi et al., 1995). Genetic and molecular analysis of mbl²⁵⁶³ revealed that it was revertible and had inserted in exon 4 within the open reading frame of the gene (Fig. 4). Remobilisation of mbl²⁵⁶³ generated the excision allele mbl^E2.

To determine the role of mbl in ommatidial differentiation, we induced mitotic clones (Materials & Methods) of the mblE27, mblE127, mbl2563 and mblE2 alleles (Fig. 3), which were selected because their molecular nature indicated that they were strong, possibly null alleles. The mutant tissue was analysed both in adult eyes and 3rd instar imaginal discs. Although all these alleles showed stronger phenotypes than the original P-insertion, they did not reveal an external phenotype in adult eyes (Fig. 3D). However, mblE27, mbl2563 and mblE2 revealed an absence of normally differentiated photoreceptors within mutant tissue (Fig. 3B,C). Clones of the mblE127 allele gave similar but weaker phenotypes consistent with its less severe molecular lesion, which does not affect coding sequences (Fig. 4). Mosaic analysis revealed that at the borders of clones all photoreceptors with normal appearance were genotypically wild type (arrowheads in Fig. 3B,C) arguing for a cell-autonomous function of the gene product. Neither pigment cells nor cone cells were affected as judged from the external appearance of the mutant area and their analysis in sections (Fig. 3D and not shown).

To analyse at what stage photoreceptor development was affected in mutant mbl tissue, mosaic 3rd instar larval eye discs for the strong mblE27 allele were stained with different neuronal and photoreceptor-specific markers. The Elav protein, which is expressed in nuclei of all developing neurons, was expressed normally in mutant mbl tissue indicating that mbl− photoreceptor cells initiate neural development (Fig. 3E). As markers for specific photoreceptor subtypes, we used the enhancer trap line H214, which in wild-type eye discs is expressed in R7, and Bar, which is specific for R1 and R6 photoreceptor cells. mbl mutant cells showed normal expression of both these photoreceptor subtype markers (Fig. 3E and not shown), indicating that the putative transformation of outer photoreceptors to R7 as seen in clones of mbl05507 is due to differentiation defects and not to a mis-specification of photoreceptor precursors. In addition, we analysed the expression of the photoreceptor-specific membrane protein Chaoptin in mbl mutant clones in eye discs. Anti-Chaoptin stainings revealed no significant difference between wild-type and mutant tissue (not shown).

To further refine the analysis of mbl mutant photoreceptors, we have looked at their axonal projections to the optic lobes and the R7-specific expression of the Rh3 and Rh4 rhodopsins. Although axonal behavior was largely unaffected with normal projections to the lamina and medulla, the analysis of mutant clones by vertical thin sections revealed that the organization of the neural tissue underlying the retinal clone was affected as compared to the corresponding surrounding wild-type tissue (Fig. 3F,G and not shown). This might suggest that some properties of the photoreceptor axons (e.g. expression of cell adhesion molecules) is not normal in mbl mutant cells. The expression of the rhodopsins was analyzed in whole-mount adult retinas, in which mbl mutant clones were induced, using the R7-specific Rh3-lacZ and Rh4-lacZ strains (Fortini and Rubin, 1990). The expression of these R7 markers was not affected in adult mbl05507 clones, indicating that no cells other than wild-type R7 have differentiated as R7 cells (not shown). This is in agreement with the analysis of mbl mutant clones in eye discs, which has suggested that all photoreceptor cell fate decisions were normal in mbl mutant clones (e.g. Fig. 3E).

Taken together, these data suggested that photoreceptor induction, subtype specification and the initial stages of differentiation were not affected in mbl mutants, but that some aspects of later differentiation did not proceed normally.

Analysis of the different cDNAs with respect to the genomic walk revealed a complex genomic organisation of the mbl locus with at least 9 exons spanning over 150 kb of genomic DNA with several large introns (>80 kb). Our data indicate that the mbl transcription unit has a single initiation site and that the primary transcript follows a complex splicing pattern giving rise to alternatively spliced mRNAs. All four mbl splice forms shared common 5′ sequences and the first 2 exons (including the ATG and amino terminal part of the putative peptides) and differ at the 3′-ends giving rise to four open reading frames (ORFs) of different lengths and carboxy termini. The sequences of the cDNAs mblA, mblB and mblC revealed three ORFs of 203, 316 and 243 amino-acids, respectively, sharing the first 179 amino terminal residues (Fig. 5A). The transcript represented by mblD also contains an ORF that shares the first 63 amino-acids with the other Mbl isoforms, but due to the use of exon 3, which contains several stop codons, only gives rise to a short peptide of 84 amino-acids (Figs 4, 5A).

Mbl isoforms A, B and C are novel proteins that contain two copies of a putative Cys3His zinc-finger structure with a typical spacing of CX7CX6CX3H (Fig. 5A,B). This is similar to the TIS11/NUP475/TTP family of proteins, of which the mouse NUP475 has been shown to bind zinc (DuBois et al., 1990; Lai et al., 1990; Varnum et al., 1991). The short MblD protein contains only one complete finger. Several other proteins contain very similar finger structures and are exemplified by the Drosophila proteins Unkempt and Clipper, the zebrafish Clipper homologue No arches and the C. elegans protein PIE-1 (Fig. 5B, for references see figure legend). Apart from Clipper, which can act as a endoribonuclease, no specific function has been assigned to proteins containing such finger motives as yet.

To analyse the subcellular localisation of Mbl proteins, anti-bodies against the complete MblA isoform were raised (Material & Methods). As all Mbl isoforms share the amino termini (see above), this polyclonal antiserum should detect all four protein isoforms. Mbl proteins were specifically detected in nuclei of ommatidial precluster cells and at lower levels in all nuclei of imaginal disc cells (Fig. 5C and not shown). The specificity of the antibody was confirmed in eye imaginal discs of flies expressing the mblA cDNA under the control of the glass response element, which revealed the same nuclear staining in photoreceptor cells, but at higher levels than seen in wild-type eye discs. The nuclear localisation of Mbl proteins is consistent with their molecular nature as members of a zinc-finger family of proteins. Similarly, the analysis of the embryonic expression pattern revealed nuclear localisation in developing muscle cells (not shown). A description of the embryonic expression pattern and phenotype will be presented elsewhere (R. A., unpublished data).

Extensive database searches have identified one potential mbl-homologue in C. elegans and several closely related genes in ver-tebrates that share the paired zinc fingers and their C-terminal flanking region. Conceptual translation of the C. elegans cosmid K02H8 sequence has revealed the existence of a gene with 79% amino acid identity within a region of 89 residues that contains the two zinc-finger domains of mblC (Fig. 6). This remarkable sequence conservation, together with the positioning of the finger at the N terminus of the protein, indicates that the K02H8-encoded gene is a true homologue of mbl.

At least three different human genes, characterised as expressed sequence tags (ESTs), share regions of extensive homology with mbl, all of them encompassing the zinc fingers: four ESTs from human cDNAs (R35479; H69637; H47129; N76537) were found to align and result in a protein of at least 250 amino acids in length and possibly containing two zinc fingers with a spacing of 108 amino acids. Two further human ESTs (Z19309; W16832) share the same overall organisation with two pairs of zinc fingers spaced at an interval of 111 amino acids. A third human gene is characterised by four ESTs (F00133; D31587; D30846; AA186683) and contains one pair of zinc fingers. The highest homology to the human genes is found in a chick EST (Z29350) (Fig. 6) and a mouse-derived EST (AA108023) (data not shown).

The sequence of Mbl proteins with the zinc-finger domain and their nuclear localisation (Fig. 5) suggests that Mbl could act as transcription factor or a co-factor in this context. Since mbl suppresses the Svp-induced phenotype, we tested whether Mbl isoforms can directly interact with Svp. The experiments addressing this question, far western (Guichet et al., 1997) and GST-pull down assays (Peverali et al., 1996), however, did not show any molecular interaction (not shown), indicating that Mbl and Svp do not directly interact and that mbl might be required in a more general term during photoreceptor differentiation.

In support of this notion, an ongoing parallel screen for modifiers of the eye phenotype induced by activated Jun, sE-JunAsp (Treier et al., 1995), has identified the mbl2563 allele as a suppressor of this genotype (U. Weber, D. Jackson, D. Bohmann and M.M., unpublished). The JunAsp isoform behaves like a constitutively activated Jun protein (Papavas-siliou et al., 1995) and, expressed under the control of the sevenless-enhancer during eye development, it causes a similar phenotype to sev-svp2 or activated components of the Ras pathway: transformation of cone cell precursors to photoreceptor neurons (Fig. 7A; Treier et al., 1995). By screening a collection of lethal P-elements (Erdelyi et al., 1995) for modifiers of this phenotype (U. Weber, D. Jackson, D. Bohmann and M.M., unpublished), one of the isolated suppressors was mbl2563. As in the case of sev-svp2, mbl2563 dominantly suppressed the sE-JunAsp eye phenotype, reducing the number of extra photoreceptors (Fig. 7). However, the mbl allele specifically and only suppressed the extra photoreceptor phenotype but not the polarity defects associated with sE-JunAsp (U. Weber and M. M., unpublished). The original mbl05507 allele and excisions derived thereof also suppressed to a similar extent the sE-JunAsp phenotype (not shown). Taken together with the molecular analysis this suggests that mbl serves a general faction during photoreceptor differentiation.

We have identified the muscleblind (mbl) gene as a dominant suppressor of the gain-of-function phenotypes of Svp and Jun. Consistent with these genetic interactions, the subsequent phenotypic analysis revealed an essential and autonomous requirement for mbl in photoreceptor differentiation. Mbl codes for nuclear proteins that contain Cys3His-zinc-finger domains and are conserved in C. elegans and vertebrates.

The phenotypic analysis has revealed that mbl is required for the differentiation of all photoreceptors. The initial steps of photoreceptor induction and determination appear normal as judged from the expression of neural- and photoreceptor-specific markers in eye imaginal discs. Photoreceptors mutant for mbl, however, fail to differentiate as seen in adult eyes. Nevertheless, even in large clones, the regular ommatidial array is undisturbed and radial sections of mbl clones show that (at least partially) photoreceptor cell bodies and their axons are present.

These observations suggest two possible explanations for the mbl phenotype. The lack of differentiated photoreceptors in mbl mutant tissue could be caused by: (i) failure of the terminal differentiation of the precursor cells or (ii) degeneration of the mutant photoreceptors (in particular their rhabdomeres). The latter explanation can be largely excluded: degeneration of photoreceptors in known mutants is light induced and slow (reviewed in (Zuker, 1996) and, in mbl mutant tissue, we see no difference between eyes of flies treated normally (usually aged for a few days before analysis) and very young or old flies that have been raised in the dark (data not shown).

The defects observed in mutant clones of the strong mbl alleles resemble the phenotype of the photoreceptor-specific transcription factor Glass (Moses et al., 1989; Moses and Rubin, 1991). Glass null alleles develop with hardly any eye tissue except for a sea of pigment cells, which is thought to be a secondary defect caused by the lack of a supporting ‘structure’ due to the absence of all the photoreceptors (Moses et al., 1989). Nevertheless, mutant clones of both glass and mbl appear externally normal and affect only photoreceptor cells, but not the cone or pigment cells. In both cases, the initial neural induction of photoreceptors is normal as judged by the extension of axons and the expression of neural markers such as Elav, 22C10 or anti-HRP. A significant difference between glass and mbl is reflected in the expression of the photoreceptor-specific protein Chaoptin (Chp) (Van Vactor et al., 1988): Chp is not detected in glass− eye discs (Moses et al., 1989), but it is expressed in mbl mutant clones. Furthermore, the expression of Rhodopsins appears unaffected in mbl mutant tissue (data not shown). Taken together, this suggests that mbl is required for the terminal differentiation of photoreceptor cells and that it acts at a later stage in this process than glass, affecting a subset of the aspects that are required in this process.

The expression patterns of glass and mbl posterior to the morphogenetic furrow in the eye disc are very similar. Although both genes are exclusively required for photoreceptor development, their nuclear expression is found in all cell types behind the furrow. Moreover, during embryogenesis both genes are expressed in the developing Bolwig organ, the larval visual system (Moses and Rubin, 1991; R. A., unpublished). To test whether mbl expression depends on glass function, we analysed its expression in discs of the glass3 allele. Mbl expression in glass− eye imaginal discs is indistinguishable from the wild-type expression pattern (not shown), indicating that mbl expression is controlled by a mechanism independent of glass-mediated photoreceptor development.

Although many modifiers of sev-svp2 are part of the Sev/Ras/MAPK pathway or interact with it genetically (Begemann et al., 1995), mbl does not show any genetic interactions with either sevS11 (Basler et al., 1991) or sE-raftorY9 (Dickson et al., 1992), suggesting that mbl is not a component of this signalling pathway. Mutations in mbl suppress the eye phenotypes induced by the nuclear proteins Svp and Jun, which are thought to directly trigger photoreceptor induction and differentiation. Moreover, overexpression of Mbl isoforms from the sev promoter or glass response elements has no effect on eye development itself unless combined with these nuclear proteins (Fig. 1 and not shown). It is also worth noting that mbl suppresses only the ectopic photoreceptor aspect of the JunAsp-mediated eye phenotype and not the polarity and rotation defects (Fig. 7). Taken together with the similarity to the glass phenotype, all the above data suggest that mbl is required downstream of Svp and JunAsp during the terminal differentiation of photoreceptors.

The mutant phenotypes of clones for ultraspiracle (usp) and orthodenticle (otd) share some similarities with those of weak mbl alleles (Oro et al., 1992; Vandendries et al., 1996). Nevertheless, it is unlikely that usp and otd have common functions to or act in concert with mbl during eye development. Although, usp mutant clones show abnormal photoreceptor morphology as analyzed in adult eyes (Oro et al., 1992), it was shown recently that the main role of usp in eye development is during the morphogenetic furrow (MF) movement. Moreover, a pGMR-Usp transgene (expressing Usp in all cells behind the MF) is not capable of rescuing the defects of usp− clones, suggesting that usp is not required for the differentiation of photoreceptors but rather for early eye patterning (Zelhof et al., 1997). Similarly, in otd mutant ommatidia, rhabdomeres also have aberrant morphology (Vandendries et al., 1996) comparable in appearance to those within mbd05507 clones. However, other defects observed in otd mutant clones (e.g. misplacement of R7 and R8 photoreceptors and gaps/holes between photoreceptors; Vandendries et al., 1996) indicate that otd is required more generally for eye development and not just for the terminal differentiation of photoreceptor neurons. Our results do not suggest a close link between mbl and otd.

The molecular analysis of the mbl gene has revealed that it encodes for at least four protein isoforms containing two Cys3His-type zinc fingers (DuBois et al., 1990; Lai et al., 1990; Varnum et al., 1991). This motif has been found in many proteins from distantly related animals, although no biochemical function has been assigned to it yet. Some of these proteins have been proposed to be involved in transcriptional regulation, as is the case of the mouse NUP475 protein, which contains two Cys3His regions (DuBois et al., 1990). Its nuclear localisation and the fact that its expression is induced by stimulation with various mitogens and growth factors suggested that it might be a nucleic-acid-binding protein involved in regulating the response to those factors. This is supported by the finding that it binds Zn2+ (DuBois et al., 1990). Another protein containing this motif is the C. elegans PIE-1 protein, which is expressed in the totipotent germline blastomere after each division in the early embryo. It has been proposed to act as a general repressor of somatic cell fates (Mello et al., 1996). Other Cys3His finger-containing proteins, such as U2AF35 in mammals (Zhang et al., 1992) and Suppressor of Sable in Drosophila (Voelker et al., 1991), have been implicated in pre-messenger RNA splicing. Nevertheless, the overall structure of these peptides is different from the proteins mentioned above since they contain up to five Cys3His repeats. An endoribonuclease activity has been demonstrated for the Drosophila protein Clipper (Bai and Tolias, 1996). Thus, although it is intriguing to speculate that Mbl might control the transcription of genes required during photoreceptor differentiation, a different role involving binding of RNA or affecting RNA metabolism in this process cannot be excluded. In this context, it is also possible that Mbl acts as co-factor for several transcription factors required in photoreceptor development.

The mbl locus is very complex, giving rise to several protein isoforms through alternative splicing. Sequence analyses of all these proteins reveal that they share the amino terminus, but contain different carboxy terminal regions. Interestingly, one of the Mbl isoforms (Mbl D) contains only one Cys3His region.

It is thus possible that different isoforms have different properties, specificities or functions as has been proposed in other similar cases. For example, the homeotic gene Ubx gives rise to multiple isoforms through alternative splicing and it has been suggested that, depending on the assay, they can display functional differences (Busturia et al., 1990; Subramaniam et al., 1994). Nevertheless, in the case of mbl, the co-expression assay with 2xsev-svp2and MblA or MblB (Fig. 1) shows indistinguishable properties of these two isoforms.

The embryonic phenotype of mbl affects mainly muscle differentiation (R. A., unpublished data) and it is also expressed in the Bolwig organ and in specific cells of the central nervous system. Thus, although mbl shares many similarities with glass (see above), its function is not solely restricted to the development of the visual system. The existence of direct Mbl homologues in C. elegans, and also in humans and other vertebrates suggests a very general function for this family of proteins. The mbl mutants will permit the unravelling of the function of this gene family in a genetically amenable organism.

We are grateful to U. Weber, D. Jackson and D. Bohmann for identifying the mbl²⁵⁶³ allele and A. Martinez-Arias for the generous gift of the PP2 line. We thank A. Cyrklaff and U. Weber for chromosome in situ hybridization, A.-M. Voie for help with sequencing and generation of transgenic fly strains, J. Hoheisel for providing the cosmid clones and J. Terol for his initial participation in the molecular analysis. We are most grateful to U. Weber for the generation of the mbl^E2 allele and her patient and skilled help with the β-galactosidase detecion in whole-mount retinas. We thank the Bloomington Stock Center, the Berkeley Drosophila Genome Project, K. Basler, M. Noll, G. Rubin and L. Zipursky for fly strains and reagents, and D. Brunner for comments on drafts of the manuscript. This work was supported by the EMBL and grant PB95-1084 from DGICYT to M. P.-A.; G.B. was in part supported by a fellowship from the Fonds der Chemischen Industrie, N. P. by fellowships from EMBO and the EEC, and R. A. by a fellowship from Conselleria d’Educació i Cultura de la Generalitat Valenciana.