mirror, a Drosophila homeobox gene in the iroquois complex, is required for sensory organ and alula formation

The pattern of bristles on the Drosophila thorax is remarkably precise and has been used extensively as a model to study the genetic basis of pattern formation. The thorax is covered with 22 macrobristles, which can be easily identified by their unique positions. The macrobristles are sensory organs of the adult periphery nervous system. Each organ consists of one neuron and three non-neuronal cells, all of which originate from a single sensory organ precursor (SOP) cell (Hartenstein and Posakony, 1989). Thoracic structures are derived from the wing imaginal disc, which is compartmentalized into several subregions (Garcia-Bellido et al., 1973; Bryant, 1975; Williams et al., 1993). These specific regions give rise to the wing as well as other thoracic structures, such as the notum and pleura, which correspond to the dorsal and ventral body walls of the thorax, respectively. An interesting developmental question is how a cell in the notum region of the wing disc is defined to form a bristle sensory organ at a specific position, leading to a precise two-dimensional pattern of macrobristles in the adult thorax.

A number of mutations in Drosophila have been characterized that alter the pattern or number of sensory organs (Jan and Jan, 1990). These studies have suggested that the formation of sensory organs occurs in sequential steps (de la Concha et al., 1988; Hartenstein and Posakony, 1989; Huang et al., 1991). Selective groups of cells in the wing disc express the proneural genes achaete-scute (ac-sc), which commit these cells to become proneural clusters (Garcia-Bellido, 1978; Villares and Cabrera, 1987; Ghysen and Dambly-Chaudiere, 1989; Jan and Jan, 1990). One cell is singled out from each proneural cluster to become a sensory organ precursor (SOP). The SOP then prevents neighboring cells from becoming SOPs by lateral inhibition (Simpson, 1990). The ac-sc genes are involved in the initial decision to differentiate a bristle, but not in the differentiation process itself. ac and sc are required for different subsets of bristles (Ghysen and Dambly-Chaudier, 1989), suggesting that the field of bristle formation might be further subdivided based on a prepattern as proposed by Stern (1954). Therefore, it is important to understand how the expression of proneural genes is controlled by a prepattern of transcriptional regulators in different regions of the epidermis.

Several genes that regulate ac-sc gene expression have been characterized. For example, hairy (h) and extramacrochaete (emc) act as negative regulators of ac-sc since mutations in these genes result in the generation of ectopic SOPs (Skeath and Carroll, 1991). Proteins encoded by these genes, as well as AC-SC, contain basic helix-loop-helix domains that were found in a number of proteins involved in transcriptional regulation (Villares and Cabrera, 1987; Murre et al., 1989; Rushlow et al., 1989; Ellis et al., 1990; Garrell and Modolell, 1990). Hairy was shown to be a direct transcriptional regulator of ac-sc, while Emc appears to down-regulate ac-sc indirectly by interacting with other factors (Gomez-Skarmaeta et al., 1995). Pannier (Pnr), a zinc finger protein with homology to the vertebrate transcription factor GATA-1, also acts as a negative regulator of ac-sc (Ramain et al., 1993; Heitzler et al., 1996). The u-shaped (ush) gene is involved in transregulation of ac-sc in the dorsal region of the notum (Cubadda et al., 1997). Ush, a zinc finger protein, heterodimerizes with Pnr as a cofactor and negatively regulates the transcriptional activity of Pnr (Haelin et al., 1997).

iroquois (iro) has been recently identified as a candidate for prepattern genes (Dambly-Chaudiere and Leyns, 1992; Leyns et al., 1996; Gomez-Skarmeta et al., 1996; Gomez-Skarmeta and Modolell, 1996). In contrast to pan and ush, which affect the dorsal notum, iro mutations specifically eliminate bristles in the most lateral domain in the notum without affecting bristles in the dorsal region (Dambly-Chaudiere and Leyns, 1992; Leyns et al., 1996; Calleja et al., 1996). The lack of lateral bristles in these mutants is due to the loss of sc expression, resulting in the failure of SOP formation (Leyns et al., 1996). The specific effects of iro mutations in the lateral notum suggest that the notum is divided into subdomains that can be genetically distinguishable. Indeed, studies on the pattern of gene expression in adult structures have provided evidence that the thorax is divided into previously unsuspected genetic subdomains (Calleja et al., 1996). The notum can be subdivided into at least three longitudinal dorsal bands along the body length: a central region, a lateral region and a region that falls between the central and lateral domains. Genes expressed in these longitudinal domains may be responsible for the establishment of the prepattern of individual pattern elements, such as bristles (Calleja et al., 1996).

Two clustered genes isolated from the iroquois region, araucan (ara) and caupolican (caup), encode similar homeobox proteins. These genes also show similar spatial expression and function in the wing. Because of their structural and functional similarities, these loci were named the iroquois complex (IROC) (Gomez-Skarmeta et al., 1996). Confirming the function of iro genes in the lateral notum, both ara and caup are expressed in the presumptive lateral heminotum of the wing disc (Gomez-Skarmeta et al., 1996). The ARA protein was shown to directly bind to an ac-sc enhancer element. Therefore, the iro genes fulfill the characteristics of prepattern genes that direct sensory organ formation in the notum.

Such regional control of gene expression can also be found in other tissues. Enhancer trap strains containing a P-lacW element have been isolated that show specific silencing of the white gene expression in the ventral domain of the eye (Sun et al., 1995; Brodsky and Steller, 1996; Choi et al., 1996; McNeil et al., 1997). Such strains are attractive since genes nearby the P-lacW insertions might also be expressed in either dorsal- or ventral-specific regions and therefore may be involved in dorsoventral patterning of the eye. Interestingly, most of these strains contained the P-lacW insertions in the vicinity of the iroquois genes. One of these genes, mirror (mirr), which is expressed in the dorsal half of the eye, was isolated independently by McNeil et al. (1997) as well as our group (this report). This gene encodes a homeodomain protein that is most similar to the ARA and CAUP proteins.

In addition to the eye, mirr is expressed in the wing disc in a similar, but not identical, pattern as the iro genes. In this report, we focus our study on the role of the mirr gene in the formation of alula and sensory organs in the wing. Our results suggest that mirr is a member of the iro complex and acts together with other iroquois genes in the prepatterning of sensory organs and alula development.

B1-12, a P-lacW (Bier et al., 1989) insertion mutation, was isolated from an enhancer trap screen (Choi et al., 1996). B1-12 is allelic to the previously identified mutations, Sail¹ (Sai¹) and Sail² (Sai²) (Carpenter, 1994) (see Results). B1-12 homozygotes are semilethal with rare escapers (<1 %). B1-12^g14 was isolated from a gamma-ray mutagenesis as described below. Sai¹ and Sai², which are both X-ray induced, were provided by A. T. C. Carpenter (Cambridge, England). Both alleles cause dominant outheld wings and recessive lethality while failing to complement each other for the recessive lethality. Sai¹ is associated with In(3LR)69D2-6; 84E12-F3. Sai² is cytologically normal, but the dominant phenotype maps to 3-37.9 relative to hairy (h) and thread (th), consistent with its cytological location at 69D (A. T. C. Carpenter, personal communication). Since B1-12 and Sai mutations are allelic to mirr (see Results), we will use the following nomenclature recommended by FlyBase: B1-12 to mirr^B1-12, B1-12^g14 to mirr^g14, Sai¹ to mirr^Sai1, Sai² to mirr^Sai2. mirror^e48 (mirr^e48) was obtained from H. McNeil and M. Simon (Stanford University). McNeil et al. (1997) used ‘mrr’ as the symbol for mirror. The FlyBase uses ‘mirr’ since mrr is the symbol for anther gene, myosin rod-related. Tp(3,3)iro¹, Df(3L)iro² and iro^DFM1 were provided by Gomez-Skarmeta and Modolell (Madrid, Spain). The recessive lethality of D¹ is due to a chromosomal inversion break at 69D (Russell et al., 1996) and is labeled in the FlyBase as mirr^Dichaete-1. Other mutations are listed in Lindsley and Zimm (1992).

To obtain additional alleles of mirr^B1-12, the P-lacW element was mobilized as described (Choi and Benzer, 1994). We also mutagenized mirr^B1-12 chromosomes with gamma irradiation (4000 rad) and screened for flies that have lost w^+mC. mirr^g14, one of the strains isolated from this screen, is recessive lethal and fails to complement all mirr mutations.

To isolate mutations that modify the dorsal-specific w^+mC expression pattern in the eye of mirr^B1-12, mirr^B1-12 chromosomes were mutagenized with ethylmethane sulfonate (EMS). y w; mirr^B1-12/TM3 Sb males were fed with 25 mM EMS as described (Lewis and Bacher, 1968) and mated to y w; D/TM3 Sb. y w; (^*) mirr^B1-12/TM3 Sb flies containing dominant mutations (marked ‘^*’) that alter the mirr^B1-12 pattern were isolated. Mob¹ (Modifier of B1-12) was isolated in this mutagenesis.

The P-lacW element in mirr^B1-12 was localized to 69D by in situ hybridization to polytene chromosomes. Genomic DNA adjacent to the P-lacW insertion was isolated by plasmid rescue (Pirrotta, 1986). A P1 clone in the mirr region (obtained from K. Zinn) and phage clones isolated from a library (a gift of J. Tamkun) were used for physical mapping of the region. A 5 kb HindIII-EcoRI genomic DNA fragment was used as a probe to isolate cDNA clones from an embryonic library (provided by K. Zinn). Preparation of DNA/RNA, blotting and hybridization were carried out using standard method (Sambrook et al., 1989).

A 2.8 kb cDNA was sequentially deleted using the Erase-A-Base (Promega) kit and sequenced by dideoxy chain termination reaction using Sequenase (United States Biochemical Corp.). MIRR protein sequence was compared to the databases using BLAST and FASTA programs. The evolutionary tree for homeoproteins was constructed using Pedro’s BioMolecular Research Tools (Computational Biochemical Research Group).

For northern blot analysis, total RNA was extracted from flies in different developmental stages using the procedure described by Cathala et al. (1983). Poly(A)⁺ mRNA was purified using oligo(dT) columns (Pharmacia), separated on a 1% formaldehyde agarose gel and transferred to a Nylon membrane (Amersham). Radioactive mirr cDNA probe was generated using the random prime labeling kit (Boehringer Mannheim).

mRNA in situ hybridization to imaginal discs was carried out as described (Choi and Benzer, 1994) using digoxigenin-labeled probes (Tautz and Pfeifle, 1989). Sense and anti-sense DNA probes were generated from a cDNA subclone by polymerase chain reaction.

Immunohistological detection of lacZ expression in imaginal discs was carried out using mouse anti-β-gal antibody (Promega) (1:250 dilution) as described (Choi and Benzer, 1994). For double labeling, imaginal discs were incubated with rabbit anti-β-gal antibody (1:100 dilution) and mouse anti-Achaete (AC) antibody (1:1 dilution, from J. Skeath). Secondary antibodies were rabbit Cy3-(1:500 dilution) or mouse FITC-conjugated IgG (1:100 dilution). Immunofluorescence was analyzed using confocal microscopy.

The enhancer trap strain B1-12 was initially identified by the dorsal-specific white (w^+mC) expression in the eye (Choi et al., 1996). The P-lacW element in B1-12 was localized to 69D by in situ hybridization to polytene chromosomes. Dominant mutations in the sail (sai) locus were also mapped to the 69D region (Carpenter, 1994; Materials and methods). Unlike normal wings, which are oriented posteriorly in parallel to the body axis, wings of +/Sai¹ and +/Sai² mutants protrude approximately 45° from the body axis. We have found that mutations in B1-12 and sai affect the sensory bristles in the notum and are allelic (see next sections). Because of these observations, a gene nearby the P-lacW insertion in B1-12 was isolated in order to study its expression and role in the development of sensory organs in the notum. Genomic DNA adjacent to the P-lacW insertion was isolated by plasmid rescue (Pirrotta, 1986). A P1 and phage genomic clones in the mirr region were used for physical mapping of the region. In order to isolate cDNA clones, a 5 kb HindIII-EcoRI genomic DNA fragment (Fig. 1A) was used to probe an embryonic library.

A 2.8 kb cDNA was sequentially deleted and sequenced by dideoxy chain termination reaction. B1-12 cDNA contained a single long open reading frame encoding a polypeptide of 641 amino acid residues (The nucleotide sequence of B1-12 cDNA was deposited in the GenBank with the accession number AF004710). This protein sequence was identical to that reported recently by McNeil et al. (1997) (GenBank accession number U95021). Therefore, B1-12 is renamed hereafter as mirrB1-12.

Two genes encoding similar homeoproteins, ara and caup, are clustered in the 69D chromosomal region and are located within 50 kb from the mirr locus (Gomez-Skarmeta et al., 1996; Fig. 1A). Comparison of the predicted MIRR protein sequence with known proteins in the databases revealed the presence of a homeobox domain signature (Fig. 1B) (Gehring et al., 1994a). Seven amino acids, including three residues involved in DNA binding, are conserved in 95% of homeoproteins (Gehring et al., 1994a). Six of these seven residues are conserved in MIRR (Fig. 1B). The predicted secondary structure of the MIRR protein shows helix-turn-helix motif characteristic of most homeodomains (Gehring et al., 1994b) (Fig. 1B). The most closely related proteins are two homeobox proteins of the iroquois complex (IROC), Araucan (ARA) (94% identity) and Caupolican (CAUP) (92% identity).

The next homologous group of homeoproteins includes Extradenticle (Exd) (Flegel et al., 1993), a Drosophila homolog of the human proto-oncogene product PBX1. The PBX homeodomains are atypical since they contain three extra amino acid residues that are located in the turn between helix1 and helix2 (Fig. 1B). These three residues are involved in cooperative interactions with other homeoproteins to modulate DNA-binding specificity as well as affinity (Chan et al., 1994; van Dijk and Murre, 1994). Although MIRR also contains these extra residues (Fig. 1B), sequence identity between the homeodomains of MIRR and the PBX-family proteins is only about 35%. Analysis of evolutionary distances between the homeodomains shows that the IRO class of homeodomains can be grouped into a distinct subfamily (Fig. 1C).

The iroquois complex genes also share similar regions in addition to the homeodomains. A region of highly acidic amino acid residues, which are involved in transcriptional activation (Klemsz and Maki, 1996; Roberts and Green, 1994), can also be found (not shown, see McNeil et al., 1997). MIRR and IRO also contain an EGF (Epidermal Growth Factor)-like motif of Notch, which has been implicated in protein-protein interaction (Lieber et al., 1992). Both ARA and CAUP proteins contain opa repeats, stretches rich in glutamine, toward the C-terminal end. However, the presence of an opa repeat (Wharton et al., 1985) in MIRR is not as obvious as in ARA and CAUP, although MIRR has some scattered glutamine residues in the C-terminal region (not shown).

A search for similar genes using the EST (Expressed Sequence Tag) databases revealed the presence of highly conserved sequences (over 80% identical) in several species, including the human and mouse, suggesting there may be a family of IRO-like homeoproteins in vertebrates as well.

Northern blot analysis revealed a single mirr transcript of approximately 3.9 kb. The transcript was expressed in all developmental stages (data not shown).

We examined mirr expression in the imaginal discs by in situ mRNA hybridization and immunohistological detection of the lacZ reporter expression. In the wing disc, mirr mRNA is expressed in several regions, including the notum and pleura. mirr expression was also detected in the alula region (Fig. 2), an accessory basal structure of the wing (Fig. 3). This expression pattern is very similar to that of ara and caup (Gomez-Skarmeta et al., 1996). However, mirr is not expressed in the precursors for L3/L5 wing veins, tegula and dorsal radius while ara and caup are expressed in these regions. The tegula is the most proximal part of the anterior wing margin. lacZ expression in the wing disc was similar to the mRNA expression pattern (Fig. 2), suggesting that the lacZ reporter reflects the pattern of mirr expression.

One of the interesting characteristics of mirr^B1-12 is that the mini white⁺ gene (w^+mC) in the P-lacW enhancer trap element is expressed specifically in the dorsal region of the eye. mirr mRNA expression is also restricted to the dorsal region of the eye disc (Fig. 2A). The dorsal-specific expression was also detected in undifferentiated second instar eye discs, suggesting that dorsoventral specification occurs prior to retinal morphogenesis (data not shown). We also examined the pattern of lacZ expression in the mirr^B1-12 enhancer trap strain. Eye discs from mirr^B1-12 third instar larvae also showed lacZ expression restricted to the dorsal region. lacZ was more abundantly expressed in the polar region of the dorsal half and anterior to the morphogenetic furrow (Fig. 2B).

mirr^B1-12 is semi-lethal over both mirr^Sai1 and mirr^Sai2 alleles. In rare cases, mirr^B1-12/mirr^Sai heterozygous adult flies were found. All mirr^B1-12/mirr^Sai adults showed outheld wings similar to +/mirr^Sai wings. In addition, alulae are completely absent in mirr^B1-12/mirr^Sai (Fig. 3C). The lack of alulae is also a characteristic of Dichaete mutants (Lindsley and Zimm, 1992) (Fig. 3D), although mirr and D are different genes. Dominant phenotypes of mirr^Sai and D¹ are not identical, since alulae are missing in +/D¹ wings but not in +/mirr^Sai wings (Fig. 3B, D). Wings found in mirr^B1-12/mirr^g14 and mirr^B1-12/mirr^e48 flies are also heldout and deficient in alulae, identical to those found in mirr^B1-12/mirr^Sai1 flies. Therefore, the loss of alula is a recessive phenotype of mirr mutations. Since some mirr^B1-12/mirr^g14 and mirr^B1-12/mirr^e48 can survive to the adult stage, while mirr^g14/mirr^e48 is 100% lethal, mirr^B1-12 appears to be a weaker allele than mirr^g14 and mirr^e48.

mirr expression is greatly reduced in the notum region of the wing disc of mirr^B1-12 homozygous flies, but there is significant expression in the alula region (Fig. 2E), consistent with the relatively normal development of the alula in these flies. In contrast, mirr^B1-12/mirr^Sai1 showed a severe reduction of mirr expression in the alula region as well as the notum, suggesting that this heterozygote represents a strong hypomorph or null in these tissues. These results are consistent with the lack of alula in mirr^B1-12/mirr^Sai1 flies (Figs 2F, 3C) and the requirement of mirr⁺ for the formation of alula and normal positioning of the wing.

An individual bristle on the notum can be easily identified by its specific position. mirr mutations caused specific loss of macrobristles only in the lateral domain of the notum. mirr^Sai1/TM3 Sb or mirr^Sai1/TM6 heterozygotes showed dominant bristle phenotype: deletion of presutural (PS) and/or posterior supraalar (pSA) bristles (Fig. 4B, Table 1). In many cases (31%), both PS and pSA bristles were absent. The chromosomes that were used to balance mirr^Sai1 mutation (TM3 Sb or TM6 Tb) did not show any significant effects on the bristle pattern (Fig. 4A, Table 1). mirr^B1-12 homozygous flies also showed consistent bristle defects. Approximately 90% of mirr^B1-12/mirr^B1-12 flies were missing one bristle, although occasionally two bristles were deleted. The deletion of bristles was specifically restricted to two of the seven macrobristles in the region; the pSA and anterior postalar bristles (aPA) (Fig. 4C, Table 1). Similar bristle defects were also found in mirr^B1-12/mirr^g14 and mirr^B1-12/mirr^E48 (data not shown). Another dominant allele, +/mirr^Sai2 displayed normal macrobristle formation, indicating that the effects on the bristles are allele-specific. However, the PA bristles were selectively removed in mirr^B1-12/ mirr^Sai2, as seen in mirr^B1-12/mirr^B1-12 (Table 1). The strongest phenotype observed was found in mirr^Sai1/mirr^B1-12, in which up to four lateral bristles were missing (Table 1). This is consistent with our observation that mirr mRNA level is greatly reduced in the notum of this mutant wing disc (Fig. 2F). Our results indicate that the elimination of lateral macrobristles by mirr mutations are allele-specific and are mainly restricted to four bristles; the PS, pSA, aPA and pPA. This is consistent with the expression of mirr in this subset of sensory organ precursors (SOPs) (see Fig. 5 in the next section). The loss of only 1 to 3 macrobristles in weaker allelic combinations might be due to residual mirr⁺ activity or partial compensation by iro⁺ expression in the heminotum region. It is important to note that the PA bristles and alula are affected in mirr^Sai/mirr^B1-12 but not in +/mirr^Sai or +/mirr^B1-12. Therefore, the removal of these. bristles and alula is a recessive phenotype of mirr^Sai and mirr^B1-12. On the contrary, the PS bristle is missing in +/mirr^Sai1 but not in mirr^Sai1/mirr^B1-12, indicating that the lack of PS is a dominant phenotype of mirr^Sai1.

Clonal analysis has shown that iro gene function is required for alula formation (Gomez-Skarmeta et al., 1996). Since mirr and iro genes are similar in structure and function, we tested whether mirr mutations show genetic interactions with iro mutations.

iro^DFM1 is a deletion that causes embryonic or early larval lethality (Gomez-Skarmeta et al., 1996). iro^DFM1/mirr^B1-12 is semi-lethal with rare escapers (less than 4% of progeny from a cross between mirr^B1-12/TM3 and iro^DFM1/TM3, see Table 2). iro^DFM1/mirr^B1-12 flies that escaped lethality showed outheld wings and lack alulae, which is identical to wings seen in mirr^Sai/mirr^B1-12(Fig. 3C, Table 2) and mirr^g14/mirr^B1-12 (data not shown). The failure of the iro^DFM1 deletion to complement mirr^B1-12 might be explained if the deletion uncovers the neighboring mirr locus. However, this is not the case since a normal level of mirr transcript can be detected in iro^DFM1 homozygous embryos (data not shown). Therefore, the lack of alula and the abnormal wing positioning in iro^DFM1/mirr^B1-12 flies appears to be due to genetic interaction between the iro and mirr genes. 50% reduction of mirr and iro activity together might be sufficient to cause the wing and alula phenotypes. Alternatively, it is also possible that the iro^DFM1 deletion affects wing-specific enhancers shared by both mirr and iro, thereby causing wing and alula defects seen in iro^DFM1/mirr^B1-12.

The iro genes are also required for the formation of macrobristle sensory organs in the lateral notum (Dambly-Chaudiere and Leyns, 1992; Gomez-Skarmeta et al., 1996). Both +/mirr^B1-12and +/iro^DFM1 heterozygotes do not show significant defects in the macrobristles. However, iro^DFM1/mirr^B1-12trans-heterozygotes showed the deletion of the pSA and/or aPA bristles (Fig. 4D, Table 2), similar to the defects seen in mirr^B1-12/mirr^B1-12 (Fig. 4C, Table 1). This suggests that the mirr and iro genes interact in the formation of sensory organs as well as alula.

iro¹, a recessive mutation associated with a transposition Tp(3,3)iro¹, is probably a null allele for caup but not for ara (Skarmata-Gomez et al., 1996). iro¹/iro¹ and iro¹/iro² (iro² is a deletion uncovering the iro locus) eliminates all macrobristles in the lateral notum (Fig. 4F). iro¹/mirr^B1-12 flies were viable with no obvious defects in the alula. However, they showed a deletion of PS and aPA, as seen in mirr^B1-12/mirr^B1-12 (Fig. 4C), or a duplication of pSA (Table 2). We have isolated a mutation Mob¹ (for Modifier ofB1-12) that enhances mirr^B1-12 phenotypes (see Materials and methods and next section for the origin of Mob¹ mutation). The Mob¹ mutation was tested for enhancement of the mirr^B1-12 phenotype in the lateral bristles. Strikingly, iro¹/Mob¹mirr^B1-12 resulted in the complete removal of the lateral macrobristles (Fig. 4G). These results also show a correlation of the severity of mirr mutant alleles and the strength of phenotypic interaction with iro mutations. Mob¹ also showed a specific dominant effect on a lateral bristle. The pSA and aPA bristles were deleted and duplicated, respectively, in Mob¹mirr^B1-12/TM6 Tb (Table 1).

ARA and CAUP are expressed in SOPs, acting as positive transcriptional regulators of ac-sc in the wing disc epithelium (Gomez-Skarmeta et al., 1996). We examined whether MIRR is also expressed in the SOPs in the notum. Wing discs from mirr^B1-12 double stained with anti-β-gal and anti-Achaete antibody showed that MIRR and AC expression overlap only in a subset of bristle SOPs in the lateral heminotum, including the PS and PA bristles, which are affected by mirr mutations (Fig. 5C). In contrast, SOPs for the notopleural and anterior supraalar bristles are stained with anti-AC but not with anti-β-gal antibody, suggesting that MIRR is either not expressed or is expressed at low levels in the notopleural (NP) and aSA bristles. This is consistent with our findings that these bristles form normally in mirr mutations. These results suggest that MIRR as well as ARA and CAUP might control ac-sc expression, and the loss of a subset of bristles in different mirr alleles might result from the loss of the corresponding SOPs rather than the degeneration of bristles. To confirm the absence of SOPs responsible for the formation of the PS, pSA and/or PA bristles, we examined mirr mutants for expression of AC or A101 lacZ reporter, a marker for SOPs in imaginal discs (Huang et al., 1991). Although there was a consistent reduction in the number of anti-AC (or anti-β-gal)-positive SOPs in the mutant discs, it was difficult to interpret the results since wild-type wing disc did not show all 7 lateral SOPs in most discs examined. This is probably due to low levels of AC expression or transient expression. Therefore, we examined iro¹/Mob¹mirr^B1-12 heterozygote since the entire lateral bristles are removed in the presence of Mob¹, a dominant modifier mutation that enhances mirr^B1-12 phenotype (Fig. 4G). In iro¹/Mob¹mirr^B1-12 wing discs, all SOPs in the lateral heminotum were eliminated as expected (Fig. 5B). This is consistent with our speculation that the loss of bristles in mirr mutations is due to the loss of corresponding SOPs.

In an attempt to identify genes that might regulate the dorsal-specific expression of w^+mC gene, we searched for dominant mutations that alter the pattern of dorsal w^+mC expression. Mob¹, one of these mutations, disrupts the dorsal specificity of w^+mC gene expression in the eye of mirr^B1-12 flies. The Mob¹ mutation slightly reduces the level of dorsal w^+mC expression from red, as seen in mirr^B1-12, to orange. Mob¹ also causes a dramatic derepression of the ventrally silenced w^+mC, resulting in pigmentation in the ventral region of the eye (Fig. 6D).

The Mob¹ mutation is homozygous lethal. It is also lethal over all tested mirr mutations, including the dominant alleles mirr^Sai1, mirr^Sai2 and mirr^Dichaete-1, as well as the recessive alleles mirr^g14 and mirr^e48. Furthermore, the Mob¹ mutation enhances weak mirr phenotypes. As mentioned earlier, mirr^B1-12 homozygotes that escaped from lethality show mild bristle defects but have normal alula. However, Mob¹mirr^B1-12/ mirr^B1-12 results in the complete elimination of the alula (data not shown). Mob¹ appears to be located very close to mirr^B1-12 since we have screened over 11,000 progeny but have failed to separate the two mutations by meiotic recombination (data not shown).

Since Mob¹ derepresses the ventrally silenced w^+mC in the mirr^B1-12 eye, we examined whether Mob¹ also alters mirr expression using the lacZ reporter in the mirr^B1-12 enhancer trap. In Mob¹ mutants, lacZ expression was greatly reduced in the eye as well as in the wing disc (Fig. 6E,F). Interestingly, Mob¹ induced ectopic lacZ expression in the L3 wing vein precursor, tegula and dorsal radius areas, where mirr is not normally expressed (Figs 5B, 6F). These results suggest that Mob⁺ plays dual roles in silencing or enhancing mirr expression in different regions of tissue. In the eye, Mob⁺ is required for normal level of mirr expression in the dorsal half and for the silencing of w^+mC in the ventral half. In the wing, Mob⁺ enhances mirr expression in most regions but silences it in the L3 vein, tegula and dorsal radius.

ARA and CAUP can functionally replace each other for activation of the proneural genes ac-sc (Gomez-Skarmeta et al., 1996). The coexpression of mirr and iro genes in some regions of the wing disc suggests that these genes may share cis-regulatory sequences. However, mirr and the iro genes show different expression patterns in the wing vein area. The iro genes are expressed and required for normal development of wing veins L1, L3 and L5. However, mirr is not apparently involved in wing vein formation since it is not expressed at significant levels in this area. Correspondingly, mirr mutants do not show any wing vein defects.

The alula is a basal structure of the wing located in the hinge region between the notum and the base of the wing veins (Bryant, 1975). iro genes are required for alula formation (Gomez-Skarmata et al., 1996). However, since iro mutations used in this study were large deletions affecting both ara and caup, it is unclear whether both ara and caup are essential for alula formation. On the contrary, mirr alone is essential for alula formation (Fig. 3). Our results on genetic interactions between the mirr and iro genes suggest that they act in the same or a parallel pathway for alula formation.

The D¹ mutation causes similar dominant wing phenotypes as mirr mutations, that is, outheld wings with no alula (Fig. 3D). The D¹ chromosome contains an inversion that has breakpoints at 70D1-2 and 69D, where D and mirr are located, respectively. The lethality of the D¹ homozygote is due to a chromosomal break at 69D, named mirr^Dichaete-1 (Materials and methods). The dominant D phenotype is caused by misexpression of the D protein (or SOX70D) in the wing, which contains the SOX DNA-binding domain (Russell et al., 1996). mirr mRNA expression is significantly reduced in the +/D¹ wing disc (unpublished results). It is possible that SOX70D misexpression in the wing might interfere with transcription of mirr, resulting in the failure of alula development. The SOX domain, found in proteins of the HMG (High Mobility Group) superfamily, has the capability of DNA bending upon binding (Ferrari et al., 1992; Giese et al., 1992). DNA bending in the regulatory region of mirr by misexpressed SOX70D and subsequent reduction of mirr expression might be a mechanism for the dominant D wing phenotype.

Different mirr mutations affect different lateral bristles (Table 1). Such allele specificity was also found in the ac-sc mutations (Lindsley and Zimm, 1992). Different alleles of ac-sc mutations eliminate specific proneural clusters and the corresponding sensory organs (Cubas et al., 1991; Gomez-Skarmeta et al., 1995). Proneural clusters in the wing disc are constructed by spatiotemporally specific enhancer elements of ac-sc. Each enhancer drives ac-sc expression in only one or a few proneural clusters (Ruiz-Gomez and Modolell, 1987; Gomez-Skarmeta et al., 1995). Similarly, it is possible that different mirr alleles might have mutations in different regulatory regions that control the development of different lateral bristles. It was also noted that mirr mutations affect specifically four of the seven macrobristles in the lateral heminotum: presutural (PS), posterior supraalar (pSA) and two postalar (PA) bristles. This is strongly correlated with high levels of MIRR expression in this subset of SOPs (Fig. 5). MIRR appears to be involved in the formation of the PS, pSA and PA bristles while IRO or a combination of IRO and MIRR are required in the other bristles of the lateral notum.

Since MIRR is structurally and functionally similar to ARA and CAUP, it might also be involved in the direct transcriptional activation of ac-sc. Therefore, bristle defects found in mirr mutants might be caused by reduced transcription of ac-sc. This is consistent with our observation that the bristles as well as AC expression in the lateral notum in the wing disc are eliminated in iro¹/Mob¹mirr^B1-12. Alternatively, MIRR might also act as an upstream activator of the iro genes. Since mirr is expressed normally in embryos that are homozygous for the iro^DFM1 deletion, mirr is not likely to be downstream of the iro genes (unpublished results).

The notum can be divided into at least three longitudinal genetic subdomains (Calleja et al., 1996). The central domain is marked by the expression of pannier (pnr) (Ramain et al., 1993). The pnr domain is laterally bounded by em462 gene expression. It was suggested that the most lateral domain may be defined by the iro genes (Calleja et al., 1996). This is consistent with our results that mutations in mirr specifically reduce the number of macrobristles in the lateral domain but do not affect bristles located more medially (Fig. 5D). A critical boundary between the pnr and em462 domain appears to be the longitudinal line connecting the two dorsocentral bristles (dotted lines in Fig. 4B), which coincides with the stripe of wingless (wg) expression (Calleja et al., 1996). The boundary defined by mirr and the other iro genes is probably located more lateral than the wg domain since the dorsocentral bristles are not affected by mirr mutations. Unlike the A/P and D/V compartment boundaries that are established early by cell lineage, the longitudinal borders between these notum domains are not lineage boundaries (Calleja et al., 1996). It was suggested, therefore, that secreted WG protein might play an inductive role for the formation of the boundary between the pnr and em462 domains. Candidate molecules involved in the induction of the iro domain boundary have not yet been identified.

It has been proposed that specific ac-sc enhancers interact with site-specific transcriptional regulators of ac-sc (Ghysen and Dambly-Chaudiere, 1988, 1989), which function as prepattern genes. Hairy, a basic-helix-loop-helix protein, is one such regulator that acts as a transcriptional repressor by binding to an ac regulatory sequence (Rushlow et al., 1989; Skeath and Carroll, 1991; Ohsako et al., 1994; Van Doren et al., 1994). ARA and CAUP are positive regulators of ac-sc and bind directly to the ac-sc enhancers (Gomez-Skarmeta et al., 1996). The striking similarities between MIRR, ARA and CAUP suggest that at least three iro complex genes play roles in prepatterning the sensory organs of the lateral notum by controlling ac-sc expression.

Photoreceptor clusters are arranged in opposite orientations in the dorsal and ventral halves of the compound eye, resulting in mirror symmetry about the equator (Ready et al., 1976). mirr^B1-12 as well as similar enhancer trap lines have been isolated that show w^+mC expression restricted to the dorsal region (Sun et al., 1995; Brodsky et al., 1996; Choi et al., 1996; McNeil et al., 1997). These lines suggest that there are genetic distinctions between the dorsal and ventral domain of the eye, although the equator is not a cell-lineage boundary (Ready et al., 1976). Since mirr is expressed in a dorsal-specific pattern in undifferentiated second instar eye discs (data not shown), the D/V boundary appears to be established prior to the initiation of retinal differentiation. It is an interesting possibility that the dorsoventral boundary defined by mirr gene expression might play a role in the formation of the equator. Some defects in dorsoventral polarity were found in mirr mutants (McNeil et al., 1997). However, mirr may not be the sole determinant of the equator since mirr mutations are not always sufficient to cause severe polarity defects (McNeil et al., 1997).

The dorsal-specific expression of mirr mRNA and w^+mC suggests that there is a ventrally expressed gene that acts on the regulatory region of mirr to silence its expression. Out results suggest that Mob⁺ is required for the silencing of w^+mC and mirr in the ventral domain of the eye since w^+mC is derepressed in the Mob mutant. Mob⁺ is also necessary for normal level of mirr expression in the dorsal domain of the eye. It appears, therefore, that Mob⁺ is necessary for silencing and enhancing in the ventral and dorsal domain, respectively. Interestingly, mirr is ectopically expressed in the L3 vein in Mob¹ wing disc. Therefore, Mob⁺ is required to repress mirr in the L3 vein precursor as well as in the ventral eye domain.

Mob might be a new gene nearby the mirr locus encoding a protein necessary for silencing mirr expression in the L3 vein and the ventral eye, or it may be a regulatory element within the mirr locus. We prefer the second possibility since Mob appears to be a cis-acting element. For instance, if Mob is a separate gene, Mob¹mirr^B1-12/Mob⁺mirr^B1-12 is expected to have the effect of ventral silencing because MOB protein will transact on mirr^B1-12, unless it is haploinsufficient. However, w^+mC was expressed in the ventral eye of these flies, indicating that the silencing effect occurs in cis. Furthermore, Mob fails to complement all mirr mutations. Therefore, Mob is likely to be an intragenic mutation in a regulatory region of the mirr locus. This is also consistent with the failure to find meiotic recombination between Mob and mirr^B1-12 despite our extensive effort (unpublished results). However, the possibility that Mob might be another gene nearby the mirr locus is not excluded.

We thank J. Botas and S. Izaddoost for comments and suggestions. We are grateful to J.-L. Gomez-Skarmeta and J. Modolell for iro mutants and cDNA, to H. McNeil and M. Simon for mirr^e48 and unpublished data, to J. Skeath for anti-AC antibody, to J. Tamkun for a phage genomic library, and to K. Zinn for the embryo cDNA library and P1 clone. We extend our gratitude to G. Roman for help with construction of the evolutionary tree as well as S. Izaddoost and R. Atkinson for abetment with confocal analysis. We are most grateful to A. Carpenter, for kindly providing us with mirr^Sai mutants prior to publication and for numerous constructive discussions. This work was supported by a grant from the National Institute of Health (R29 EY11110).