ABSTRACT
Polarity in the Drosophila eye is manifested as a dorsoventral reflection of two chiral forms of the individual unit eyes, or ommatidia. These forms fall on opposite sides of a dorsoventral midline of mirror symmetry known as the equator. Polarity is established in the eye imaginal disc as cells adopt their fates and as the ommatidial precursors undergo coordinated rotation within the epithelium; the mechanisms that coordinate these early patterning events remain poorly understood. We have identified a novel gene, strabismus (stbm), which is required to establish polarity in the eye, legs and bristles of Drosophila. Many stbm ommatidia are reversed anteroposteriorly and/or dorsoventrally. In stbm eye discs, ommatidial rotation is delayed and some ommatidial precursors initiate rotation in the wrong direction. Mosaic analysis indicates that stbm is ommatidium autonomous and required in most, if not all, photoreceptors within an ommatidium to establish normal polarity. stbm also appears to play an instructive role during the establishment of the fates of photoreceptors R3 and R4. stbm encodes a novel protein with a potential PDZ domain-binding motif and two possible transmembrane domains. Sequence analysis of both vertebrate and invertebrate homologs indicates that stbm has been highly conserved throughout evolution.
INTRODUCTION
The establishment of polarity occurs throughout the development of unicellular and multicellular organisms, both prokaryotic and eukaryotic, and is fundamental for the generation of cellular diversity and tissue specialization. In single cells, the asymmetric distribution of regulatory molecules is required to establish distinct fates in daughter cells (Rhyu et al., 1994; Kraut et al., 1996) and underlies the establishment of specialized anterior and posterior structures, such as dendrites and axons in a neuron, and the specialized apical and basal domains in many cell types. In animal tissues, the polarity of groups of cells is coordinately regulated. This tissue polarity, or polarity within the plane of an epithelium, organizes epithelia into functional units. We are using the developing Drosophila eye, a polarized epithelium, as a system to understand the molecular mechanisms governing the establishment of developmental symmetries. We have identified a novel gene, strabismus (stbm), which plays a critical role in setting up polarity in multiple tissues in the fly.
Polarity is manifested in various ways, depending upon the tissue in which it is expressed. Many epithelia are covered in a coat of bristles and hairs, and these elaborated structures are uniformly oriented in space. In addition, appendages elaborate unique distal and proximal surfaces. In the Drosophila eye, the dorsal and ventral halves of the eye are mirror image reflections of one another.
The Drosophila compound eye comprises approximately 800 hexagonal unit eyes, or ommatidia, packed in a smooth array. Each ommatidium is a precise assembly of 20 cells: a central core of 8 photoreceptor cells (R1-R8) and 12 non-neuronal support cells. The rhabdomeres, or light-sensitive organelles of the photoreceptors, are arranged in a characteristic asymmetric trapezoid and are identified by their positions within the trapezoid; R3’s rhabdomere occupies the point of the trapezoid (Figs 1C, 2C, and Dietrich, 1909). Each eye contains two chiral forms (i.e. mirror image reflections) of the trapezoid which fall on opposite sides of a midline of mirror symmetry known as the equator (Figs 1C, 2E).
Assembly of cells into ommatidial precursors, or preclusters, begins in the eye imaginal disc, the primordium of the adult eye. Recruitment of cells begins posterior to the morphogenetic furrow, a dynamic front of differentiation which progresses from posterior to anterior across the epithelium. Photoreceptor recruitment follows a characteristic sequence, starting with R8 and followed by the pairwise addition of R2/5, R3/4, R1/6 and finally by R7. The remaining cell types are recruited between late larval and mid-pupal development (for review see Wolff and Ready, 1993).
The group of 8 photoreceptor cells is initially bilaterally symmetrical across the anterior-posterior (A/P) axis. As development proceeds, morphological movements break the symmetry so that the ommatidium becomes polarized across this axis. These morphological movements are associated with the specification of distinct fates in two of the photoreceptor cells: R3 and R4. Within each R3/R4 pair, the cell that is located closest to the midline of the eye disc, the equatorial cell, adopts the R3 fate, and the more lateral cell, the polar cell, adopts the R4 fate (Fig. 1A). Evidence that the R3 and R4 cells differ first becomes apparent in the third instar eye disc when the bilateral symmetry of the photoreceptor precluster breaks down: R4 loses contact with R8 and its cell body becomes displaced relative to that of R3 (Fig. 1B; and Tomlinson, 1985). The chirality of an ommatidium is therefore associated with the adoption of the R3 and R4 fates.
Two events contribute to the origin of the equator in the third instar eye disc. First, chiral forms are created as a consequence of R3 and R4 adopting their appropriate fates with respect to their dorsal or ventral location in the eye (Fig. 1B). Second, the ommatidial precursors in the dorsal and ventral halves of the eye rotate as units, 90° in opposite directions to one another (Fig. 1A,B; and Ready et al., 1976). As a result of these patterning events, the adult eye displays global mirror symmetry (Figs 1C, 2C,E).
A number of genes are involved in setting up this polarity. frizzled (fz) encodes a protein with 7 transmembrane domains and is a member of the serpentine class of receptors (Vinson et al., 1989). fz ommatidia display an assortment of disruptions in ommatidial polarity, including partial rotations, reversals on either their A/P, dorsal/ventral (D/V) or both A/P and D/V axes (Zheng et al., 1995). dishevelled (dsh) mutant eyes resemble fz eyes (Theisen et al., 1994). Dsh contains a PDZ domain (Klingensmith et al., 1994; Theisen et al., 1994), suggesting it may be localized at sites of cell-cell contact (for review see Fanning and Anderson, 1996). The Drosophila homolog of the p21 GTPase RhoA has recently been shown to be required for the generation of tissue polarity. The RhoA eye phenotype is similar to that of fz and dsh and genetic interactions suggest RhoA is a component of the signaling pathway mediated by Fz and Dsh (Strutt et al., 1997). In addition to their eye phenotypes, fz, dsh and RhoA mutant flies also show a variety of polarity defects in other epithelia and cell types. Finally, there are eye-specific genes, such as nemo and roulette, which carry out the rotation program (Choi and Benzer, 1994).
We have identified a novel gene, strabismus (stbm), which affects the planar polarity of multiple structures in Drosophila. stbm ommatidia display anteroposterior, dorsoventral and both anteroposterior and dorsoventral reversals. Our results suggest these phenotypes are likely to be caused by a defect in specification of the R4 cell fate. Mosaic analysis indicates that stbm is ommatidium-autonomous and is required in most, if not all, photoreceptor cells for normal orientation. The stbm locus encodes a novel protein that has been highly conserved through evolution and contains two putative transmembrane domains as well as a potential PDZ domain-binding motif.
MATERIALS AND METHODS
Genetics
Fly culture and crosses were carried out according to standard procedures. X-irradiation (4000 rads) was used as the mutagen in the FLP/FRT screening strategy (Xu and Rubin, 1993) to identify genes that play a role in eye development. A total of 104,000 flies were screened and 39 mutants required for eye development were identified. Of these, 12 were mutations in the stbm locus, and both genetically null and hypomorphic alleles were isolated. 11 additional alleles, also X-irradiation-induced, were identified in a complementation screen. The FLP/FRT system was used to generate clones for mosaic analysis in alleles stbm6cn, stbm15cn and stbm10cn. Phenotypic analysis was performed on stbm6cn, a molecular null allele. Standard genetic crosses were conducted to generate H123; stbm and double mutant stocks. Alleles used for genetic interaction studies include fz1, dsh1, nmoP1, pk1 and sple1.
Histology
Scanning electron microscopy was performed as described by Kimmel et al. (1990). Fixation and sectioning of adult eyes were performed as described by Tomlinson and Ready (1987). Lead sulfide staining was performed according to Wolff and Ready, (1991) and cobalt sulfide staining according to Melamed and Trujillo-Cenoz (1975). Elav and lead sulfide staining and mitotic labeling (bromodeoxyuridine (BrdU)-pulsed eye discs) were used to demonstrate that the center-lateral growth of rows is normal in stbm eyes. BrdU labeling was performed as described by Wolff and Ready (1991), except that pepsin digestion was omitted. Antibody and X-gal staining were performed according to standard methods. The full length stbm cDNA was used as a template to generate the RNA probes used for tissue in situ hybridizations, which were performed as described by Dougan and DiNardo (1992).
Isolation of genomic DNA and cDNA
A P1 phage insert (Berkeley Drosophila Genome Project clone DS00300) was used to probe genomic blots of X-ray-induced stbm alleles according to standard procedures (Sambrook et al., 1989). A polymorphic 10 kb EcoRI fragment was identified in stbm10cn and was used to screen both a Drosophila eye-antennal imaginal disc library (prepared by A. Cowman) and a cosmid library of D. melanogaster genomic DNA (Tamkun et al., 1992). A human EST (I.M.A.G.E. Consortium (LLNL) cDNA Clone ID no. 120749; Accession no. T95333; Lennon et al., 1995) was used as a probe to screen a mouse PCC4 teratocarcinoma cDNA library (Stratagene). Transformation rescue experiments were conducted using the pGMR expression vector (Hay et al., 1994) containing the 2.1 kb EcoRI fragment of the stbm cDNA.
DNA sequencing
DNA sequence was determined using the dideoxy chain termination procedure (Sanger et al., 1977) using Automated Laser Fluorescence (Pharmacia) and Applied Biosystem, Inc. 373 sequencers. Templates were prepared by sonication and insertion of plasmid DNA into the M13mp10 vector. The entire coding regions as well as the genomic regions that correspond to the stbm locus were sequenced on both strands. Sequences were analyzed using the Staden (R. Staden, Medical Research Council of Molecular Biology, Cambridge, England) and the Genetics Computer Group, Inc. software packages. Genomic libraries of X-ray-induced and cytologically normal stbm alleles were constructed using the lambda ZAP expression vector (Stratagene).
RESULTS
Identification of stbm
stbm was identified by virtue of its rough eye phenotype in mosaic clones in the Drosophila eye. The FLP/FRT system was used to generate clones of X-irradiation-induced mutant tissue in a heterozygous background (Golic and Lindquist, 1989; Xu and Rubin, 1993). Externally, mutations in the stbm locus are manifested as a mild roughening of the normally smooth ommatidial lattice (Fig. 2B). stbm is a recessive, homozygous viable mutation, although null alleles display reduced viability.
Ommatidial polarity is disrupted in stbm mutant eyes
In stbm mutant eyes the normally parallel rows of ommatidia are not properly aligned, giving the eyes a rough appearance (Fig. 2B; compare to wild type, 2A). Tangential sections through adult eyes indicate that the vast majority of stbm ommatidia are correctly assembled (Fig. 2D; compare to wild type, 2C) and that the misaligned eye lattice is a consequence of aberrant orientation of ommatidial units. stbm ommatidia show several orientations: they are either normal, reversed dorsoventrally, anteroposteriorly, or both dorsoventrally and anteroposteriorly. In addition, while most ommatidia rotate the normal 90°, many undergo either partial or no rotation. Finally, in a small percentage of ommatidia, photoreceptors R3 and R4 are bilaterally symmetrical so the chirality of the ommatidium is abolished. The arrangement of rhabdomeres in these achiral ommatidia is rectangular rather than trapezoidal (Fig. 2D-asterisk, F-yellow). We have named this gene strabismus, which is formally defined as “a disorder of vision due to a deviation from normal orientation of one or both eyes” and is the medical term for ‘cross-eyed’.
To determine the relative proportions of the various ommatidial forms, polarity of genetically mutant ommatidia was quantitated in clones of homozygous stbm6cn tissue, a molecular null allele. The equator in the surrounding wild-type tissue provided a reference for a mutant ommatidium’s dorsal vs. ventral position in the eye. Of 181 ommatidia scored, 46% showed normal orientation, 35% showed D/V pattern reversals, 7% displayed A/P reversals, 7% were reversed both dorsoventrally and anteroposteriorly, 3% failed to rotate and 2% were achiral.
As a consequence of ommatidial misorientation, the pigment cell lattice is disrupted. Cobalt sulfide-stained pupal eyes reveal that stbm eyes contain a normal complement of primary pigment cells, bristles and cone cells, although occasional ommatidia are missing one cone cell. Secondary and tertiary pigment cell numbers are often correct, except at the vertices of partially or unrotated ommatidia where there is an excess of these cell types (data not shown).
Ommatidial rotation is delayed in stbm eye discs
To assess ommatidial rotation, eye discs were stained with an antibody that recognizes the nuclear neuron-specific protein Elav (Fig. 3A,B) (Robinow and White, 1991), or cobalt and lead sulfide, which highlight the apical surfaces of cells (Fig. 3C,D). The Elav-stained discs demonstrate that stbm does not play a role in recruiting photoreceptors into the assembling ommatidium, as photoreceptors are recruited in the normal sequence and with normal timing in mutant discs (Fig. 3B, compare to wild type, Fig. 3A).
Views of the apical surface of eye discs indicate that ommatidial rotation normally begins approximately 5 rows posterior to the furrow and is complete by row 12 (Fig. 3A,C). In stbm eye discs, the majority of preclusters are delayed in initiating rotation (Figs 3B,D – asterisks and 4F). As the photoreceptor preclusters mature, their tips become effaced from the apical surface of the disc so their orientation can no longer be evaluated at the apical surface. However, the cone cells, which lie on top of the photoreceptor cells in stereotypic positions, provide a useful compass by which to measure ommatidial orientation at the posterior of the third instar eye disc. Observations of cone cells in lead and cobalt sulfide-stained mutant eye discs reveal that the majority of ommatidial preclusters at the posterior of stbm eye discs have rotated no more than approximately 45° whereas wild-type precursors would have rotated 90° at an equivalent point in development (not shown). Of those ommatidia that do begin to rotate on schedule, most appear to initiate rotation in the correct direction with respect to their dorsal or ventral location within the epithelium, but some initiate rotation in the wrong direction (Fig. 3B).
These preparations demonstrate that ommatidial orientation is disrupted in two ways during the initial stages of pattern formation. First, ommatidial rotation is significantly delayed in stbm eye discs and second, some ommatidial precursors initiate rotation in the wrong direction.
Equator-lateral growth of a row is not disrupted in stbm eye discs
The position of the future equator is evident in the third instar eye disc prior to the start of ommatidial rotation. New rows emerging from the furrow are initiated at the D/V midline (the site of the future equator) and grow laterally as ommatidia are added to each end in a symmetrical manner (Wolff and Ready, 1991). This ‘center-lateral’ growth of a row and ommatidial rotation are both affected in fz mutants (Zheng et al., 1995), suggesting these two events may be linked. However, while the adult phenotypes of fz and stbm are similar, the center-lateral growth of a row is not altered in stbm mutants (see Methods, data not shown). This observation indicates that the two events are genetically separable and that stbm may act downstream of fz, or in a parallel pathway, in its role in orientation but not in its role in the center-lateral growth of new ommatidial rows.
The R3 and R4 fates are mis-specified in stbm
The aberrant ommatidial forms seen in stbm mutant eyes (Fig. 2F) can be accounted for by invoking defects in either chirality, rotational direction or both (Fig. 4A,B; see legend for details). To distinguish between these possibilities we used the enhancer trap line H123, which is differentially expressed in R3 and R4 and therefore provides a molecular marker to differentiate between these two cell fates (Fig. 4C,E). We examined H123 expression in a null stbm allele and found both reversals in specification of the R3 and R4 cell fates and misrotations (Fig. 4D,F ; see legend for details). Four classes of preclusters were seen in stbm discs and these correlate with the predicted classes described in Fig. 4A. Ommatidia were seen in which the R3 and R4 fates were correctly specified and rotation occurred in either the correct (Fig. 4D,F: no. 1) or incorrect (no. 4) direction with respect to the R3 and R4 fates. In addition, ommatidia were seen in which the R3 and R4 fates were reversed and rotation occurred in either the correct (no. 2) or incorrect (no. 3) direction with respect to the R3 and R4 fates. This analysis of H123 expression in stbm discs suggests that stbm may participate in imparting the R3 and R4 fates and it therefore follows that the phenotype results from a failure in fate specification rather than later patterning events involving placement of the R3 and R4 rhabdomeres.
We have also examined lead sulfide-stained stbm eye discs to assess cell contacts within the photoreceptor preclusters. In wild-type eye discs, the symmetry in the 8 cell precluster is lost when R4, the polar cell of the R3/R4 pair, loses contact with the central photoreceptor, R8 (Fig. 1B). In some preclusters in stbm eye discs the equatorial cell loses contact with R8 (data not shown). These findings are consistent with the H123 results described above and suggest that stbm participates in the specification of the R3 and/or R4 fates.
The fact that the R3 and R4 fates can be mis-specified in stbm ommatidia raises the possibility that the fates of the remaining symmetrical photoreceptor pairs, R2/R5 and R1/R6, are also reversed, such that the anterior and posterior faces of the trapezoid are inverted. Because molecular markers that distinguish between the members of these photoreceptor pairs do not exist, it is not possible to address this question directly. However, in wild-type ommatidia, R8’s rhabdomere extends between R1 and R2 (Dietrich, 1909), and only rarely between R5 and R6 (R. Cagan, personal communication), suggesting R8 can recognize a difference between cells to the anterior, R1 and R2, and those to the posterior, R5 and R6. If we assign the R3 fate to the cell occupying the point of the trapezoid, then in stbm ommatidia, R8 extended its rhabdomere between R1 and R2 in 95 out of 96 cases examined. These results are consistent with the notion that the anterior (R1, R2 and R3) and posterior (R4, R5, and R6) cell fates may be switched in stbm ommatidia, and that the entire ommatidium is ‘backwards’.
stbm is ommatidium autonomous and is required to specify the R4 cell fate
stbm could be directing ommatidial orientation by one of several mechanisms. It could act locally to coordinate the process within an ommatidium or between neighboring ommatidia, or it could provide a diffusible polarity signal across the entire disc. To distinguish between these possibilities, we analyzed ommatidia located in and near clones of homozygous mutant stbm tissue. First, we found that clones anywhere in the disc are autonomous. Consistent with this observation, we have found stbm RNA throughout the eye disc using stbm as a probe for in situ analysis (not shown).
Second, we assessed the phenotype of genotypically mutant ommatidia lying along the border of the clone. We found that genotypically mutant (as well as mosaic) ommatidia are not rescued by their wild-type neighbors (Fig. 5A, asterisks) – they misrotate even when located next to genotypically wild-type ommatidia. Third, genetically wild-type ommatidia are not affected by neighboring mutant tissue (Fig. 5B, dark blue ommatidia, see legend for details). Therefore, stbm acts autonomously within ommatidia; in other words, the presence or lack of Stbm function in one ommatidium does not affect the polarity of neighboring ommatidia.
To establish if stbm acts in one specific cell to establish orientation of the entire ommatidium, we carried out a mosaic analysis of ommatidia displaying either mutant (Fig. 5C) or normal (Fig. 5D) orientation. We randomly chose 169 misoriented and 119 normally oriented mosaic ommatidia at clonal boundaries and scored the genotype of photoreceptors within these ommatidia. Our overall observation from both phenotypically mutant and wild-type mosaic ommatidia is that misorientation or normal orientation does not correlate absolutely with the mutant genotype of any specific photoreceptor or group of photoreceptor cells. Specifically, the combination of stbm mutant and wild-type photoreceptors does not predict the orientation of the ommatidium. Rather, many of the photoreceptors make a contribution.
However, we have also noticed some interesting biases in the frequency with which certain photoreceptors were mutant. First, there is an over-representation of stbm− R3 cells in misoriented mosaic ommatidia: 87% of misoriented ommatidia have a mutant R3 cell. Second, stbm− R4 cells are under-represented: 26% of misoriented ommatidia and 21% of properly oriented mosaic ommatidia have a mutant R4 cell (Table 1). This suggested two possibilities: (1) stbm is required in R3 for proper orientation or (2) stbm is required to specify the R4 cell fate. We therefore analyzed mosaic ommatidia in which the R3/R4 pair is mosaic (R3+/R4− or R3− /R4+). We found a total of 144 such cases, 141 of which were of the R3− /R4+ type. This implies that if one of the R3/R4 pair is mutant it will become an R3. Thus, Stbm is required for the R4 cell to establish its fate.
Furthermore, the choice of the R3/R4 cell fate appears to in turn influence the direction of ommatidial rotation. Consistent with this is the observation that of these 141 R3−/R4+ ommatidia, 74% acquire an incorrect orientation while only 26% adopt the correct orientation. If both cells are mutant, 59% are incorrectly oriented and 41% are correctly oriented, as would be expected if orientation becomes random. However, while having R3+ and R4+ increases the probability of correct orientation, it is not sufficient as 73% are correctly oriented and 27% are incorrectly oriented. These data are consistent with the results obtained using the H123 marker.
In summary, two conclusions can be drawn from the mosaic analysis. First, Stbm is ommatidium autonomous and the presence of the normal gene product in any one cell is not sufficient to establish correct orientation; rather, it is required in many, if not all, photoreceptors. This result suggests that each photoreceptor makes some contribution to the orientation process which must be a highly coordinated effort between cells within an ommatidium. Second, Stbm appears to be important in specifying the fate of the R4 cell.
stbm affects the polarity of multiple structures
stbm affects the polarity of multiple tissues in Drosophila. In stbm, the polarity of many of the bristles and hairs is abnormal. In wild-type animals, the thoracic bristles and hairs are aligned in parallel rows with each bristle pointing toward the posterior of the animal (Fig. 6A) while the leg bristles lie flat against the leg surface and point distally. In stbm homozygotes, many thoracic bristles have aberrant polarity (Fig. 6B) and the leg bristles are oriented perpendicular to the leg. Hair polarity is disrupted throughout the body, for example surrounding the eye and on the wings and thorax (not shown). stbm also plays a role in setting up the cuticular polarity of the legs. In stbm legs, extra tarsal segments or partial duplications (Fig. 6D, arrow and arrowhead, respectively; compare to Fig. 6C) frequently occur in tarsal segments 3 and 4 and occasionally in tarsal segment 2 (see Held et al., 1986 for detailed description of this phenotype). Finally, the wings of stbm flies are held out (not shown). This widespread requirement for stbm in various tissues is not uncommon among tissue polarity mutants; for example, fz, dsh, prickle (pk) and spiny legs (sple) affect the polarity of various structures, including bristles, hairs and the leg (see Gubb, 1993 for review). These apparently very distinct types of polarity are clearly being regulated by the same basic processes, yet the cellular basis and common theme underlying these various phenotypes remains to be established. stbm also plays a role in early embryonic development, as many homozygous embryos die between 0-4 hours following egg laying. We have not yet determined the role for stbm during these early developmental events.
The stbm gene encodes a novel protein with a putative PDZ domain-binding motif
The stbm locus was meiotically mapped to 2-61 and cytologically mapped to 45A7-10 based on polytene chromosome analysis of X-ray-induced stbm alleles. A P1 phage insert (see Methods) covering this locus was used to screen genomic DNA blots prepared from X-ray-induced stbm alleles. A polymorphism was identified in a 10 kb EcoRI fragment in allele stbm10cn. This genomic fragment was used to screen an eye-antennal imaginal disc cDNA library which identified a single class of cDNAs. The longest cDNA (3.4 kb) was sequenced and found to contain a single ORF, 1.9 kb in length; this ORF predicts a protein of 584 amino acids. Sequence analysis of the corresponding genomic region revealed that the gene contains no introns.
BLAST searches of protein and nucleotide databases indicate that stbm encodes a novel protein. Amino acid sequence analysis of the putative Stbm protein reveals 4 hydrophobic stretches in the N-terminal third of the protein, the longest of which spans 29 amino acids, suggesting a possible transmembrane localization for the protein (Fig. 7). The four C-terminal amino acids of the protein, ETSV, are of particular interest as they match a consensus sequence for PDZ domain-binding motifs (X S/T X V; Songyang et al., 1997).
Confirmation that this gene corresponds to the stbm locus was provided by sequence analysis of 4 mutant alleles. Alleles stbm6cn, stbm153 and stbm15cn contain small deletions and stbm60 is a duplication of adjacent bases; all four mutations lie within the ORF and cause frame shift mutations in the ORF (Fig. 7). Further confirmation comes from transformation rescue experiments in which the stbm cDNA was expressed in cells posterior to the morphogenetic furrow in the eye imaginal disc using a glass responsive promoter (Hay et al., 1994). While the eye phenotype was rescued in these flies, other stbm phenotypes were not rescued (not shown).
Stbm has been highly conserved through evolution
A search of translated nucleotide databases using the TBLASTN program (Altschul et al., 1990) identified a human EST (I.M.A.G.E. cDNA Clone no. 120749-see Methods) with homology to Stbm. Complete sequence of this partial cDNA revealed that the C-terminal half is 55% identical and 72% similar to the fly homolog (Fig. 7). The human EST was used to screen a mouse teratocarcinoma cDNA library. The longest cDNA identified (3.7 kb) was sequenced; sequence comparison showed that the mouse protein is 50% identical to the fly protein with the C terminus showing a greater degree of conservation. We also identified a C. elegans homolog through a search of the C. elegans database (Accession no. U52000); the Drosophila and C. elegans proteins are 33% identical. The four hydrophobic domains and the PDZ domain-binding motif are conserved among the four species (Fig. 7).
DISCUSSION
The mechanisms that regulate the establishment of tissue polarity during development are poorly understood and are likely to vary depending on the epithelial type. We report the role of a novel gene, stbm, which is involved in setting up polarity in the Drosophila eye. stbm plays a role in coordinating the fate specification of photoreceptors R3 and R4 in the dorsal and ventral halves of the eye. The rotation machinery subsequently acts to induce rotation, thereby establishing global mirror symmetry in the eye.
The role of stbm in establishing polarity in the retina
Ommatidia in stbm mutant eyes display an array of orientation defects, including dorsoventral, anteroposterior, or both dorsoventral and anteroposterior reversals. In addition, many ommatidia fail to complete the full 90° of rotation and some fail to rotate altogether. Finally, in a small proportion of ommatidia, the R3 and R4 rhabdomeres are symmetrically positioned, forming rectangles rather than trapezoids. Defects in both fate specification and ommatidial rotation must be invoked to account for these various ommatidial forms. Our studies implicate stbm in playing a primary role in fate specification and possibly a minor role in rotation.
Two lines of evidence illustrate stbm’s role in specifying the fate of the R4 cell. First, enhancer trap line H123, which is expressed more intensely in the polar cell than the equatorial cell in wild-type discs, is expressed more intensely in the equatorial than the polar cell in some stbm− preclusters. This result suggests that these fates are often reversed in the mutant. Second, the mosaic analysis indicates stbm function is absolutely required to assign the R4 fate: in mosaic ommatidia in which one of the R3/R4 pair is mutant, the stbm+ cell becomes an R4 cell in 98% of cases.
While stbm is required for the R4 cell identity, the R3 and R4 cells can assume these fates in the absence of Stbm function (Fig. 2D), yet in these cases, the choice to become an R3 vs. an R4 cell appears to become randomized. This suggests that Stbm tips the balance such that the polar cell adopts the R4 fate, thus enabling chirality to be coordinated in the dorsal and ventral halves of the eye. In rare cases, R3 and R4 are symmetrical, possibly reflecting situations in which R3 and R4 remain undecided as to which fate to adopt.
Even though the R3 and R4 fates are often reversed in stbm−ommatidia, the direction in which these mutant ommatidia rotate is usually consistent with the fates adopted by the R3 and R4 cells (the R3 fate is defined here as the cell that occupies the point of the trapezoid or has low H123 expression). In other words, in wild type, the equatorial cell always becomes an R3 and ommatidia always rotate such that the R1, R2 and R3 cells end up along the anterior side of the eye. When the R3 and R4 fates are reversed in the mutant,?rotation subsequently becomes reversed such that R1, R2 and R3 still end up along the anterior face of the eye: in stbm− clones, the large majority (81%) of genotypically mutant ommatidia rotate in a direction that is consistent with the identity of the R3 and R4 fates (46% normal orientation plus 35% D/V inversions). Therefore, rotation respects the identity of the R3/R4 cells in the majority of mutant ommatidia.
A minority of genotypically mutant ommatidia rotated ‘backwards’ with respect to the R3/R4 fates (7% A/P and 7% A/P and D/V inversions) and a few (5%) either failed to rotate or displayed rectangular trapezoids. These misrotated ommatidia can be explained as follows. The mosaic analysis shows that many, if not all, photoreceptors contribute to ommatidial orientation; for example, ommatidia in which only R1 is stbm− can be misoriented. While these cells play a less important role than that of R3 and R4, their contribution is clearly not insignificant. Perhaps a ‘disagreement’ or competition between R3/R4 and the remaining cell types leads to the delay in onset of ommatidial rotation in the eye disc.
An alternative possibility is that the ommatidia that are misrotated could represent ommatidia in which the future R3 and R4 cells were either undecided or switched fates following a key point when the rotation decision was made. A delay in the R3/R4 fate choice could also explain the delay in initiation of rotation in mutant eye discs.
The data from our analysis of mosaic ommatidia in which one cell of the R3/R4 pair is wild type and the other stbm−/−/R4+ genotype acquire an incorrect orientation in 74% of the cases. Furthermore, when both cells are mutant, their fates may be established randomly and thus their orientations more closely approximate a 50:50 distribution of phenotypically mutant and wild-type ommatidial forms.
We cannot rule out the possibility that stbm contributes directly to choosing the direction in which an ommatidium rotates. However, since ommatidia that rotate in a direction that is inconsistent with the R3/R4 cell fates are uncommon, it is likely that additional genes are needed for this decision.
While a detailed morphological analysis has not been conducted to define the order in which rotation and the specification of the R3/R4 occur, it is clear that the two events are virtually coincident. Our studies on stbm imply that the R3 and R4 fates are specified prior to activation of the ‘rotation machinery’. One interpretation for the relationship between these two events is that the rotation machinery recognizes the R3/R4 fates and instructs the precluster to rotate in a direction consistent with this information (this has also been proposed by Zheng et al., 1995 and Choi et al., 1996). The mosaic analysis indicates that R3 and R4 are not the only photoreceptor precursors important for rotation; perhaps there are asymmetries in the other pairs of cells which can also provide supplementary information regarding the chirality of an ommatidium.
Comparison of stbm to other tissue polarity mutants
stbm mutant flies display many of the polarity phenotypes seen in fz, including ommatidial orientations in the eye, bristle and hair polarity in numerous epithelia, and duplications of leg segments. In addition, as in stbm mutants, the rotation of some ommatidia in fz eye discs is delayed (Zheng et al., 1995).
There are also a number of intriguing differences between the two genes. First, fz shows directional non-autonomy (it acts non-autonomously on the polar side of clones) (Vinson and Adler, 1987; Zheng et al., 1995) whereas stbm is ommatidium autonomous. This difference suggests that the two genes play distinct roles in the signaling pathway that sets up polarity. Second, the mosaic analyses of stbm and fz indicate that while fz is required to impart R3 identity on a cell, stbm is required for R4 cell identity. In fz ommatidia mosaic for the R3/R4 pair, the fz+ cell adopts the R3 cell fate in 96% of cases (Zheng et al., 1995); a similar analysis of stbm showed that the stbm+ cell adopted the R4 fate in 98% of all cases. A third significant difference is that stbm does not affect the center-lateral growth of new ommatidial rows whereas fz does. In their analysis of fz, Zheng et al. (1995) noted that preclusters at the ends of a leading row are often more mature than those close to the midline, suggesting fz mediates an equatorial signal for neuronal differentiation.
In addition to its similarity to the fz phenotype, stbm also bears a remarkable resemblance to tissue polarity phenotypes exhibited by dsh and RhoA, both in the eye and elsewhere. The similarity between the stbm, dsh, fz and RhoA phenotypes suggests that their gene products may function in the same or parallel pathways. In fact, genetic interactions have been demonstrated between fz and dsh (Krasnow et al., 1995; Strutt et al., 1997), fz and RhoA, and RhoA and dsh (Strutt et al., 1997).
Because of the phenotypic similarities between stbm and these other tissue polarity mutants, we have begun to look for genetic interactions between these genes. We were unable to identify obvious enhancement or suppression of the stbm phenotype by the removal of one copy of pk, sple, fz, dsh or nmo in a stbm mutant background. Furthermore, animals with mutations in both stbm and one of the above tissue polarity genes did not exhibit phenotypes that differed from those seen in stbm alone. The possible genetic and molecular relationship between these molecules and stbm will require further genetic and biochemical analysis.
Dfz2, a Drosophila Frizzled homologs, and Dsh have been implicated to act in the Wingless (Wg) signaling pathway. Dfz2 is the putative receptor for Wg (Bhanot et al., 1996), so the equatorial signal mediated by fz may be a Wnt. Dsh is required to transduce the Wg signal (see Klingensmith and Nusse, 1994 for review). Furthermore, Dsh contains a PDZ domain (Klingensmith et al., 1994; Theisen et al., 1994) and Stbm contains a PDZ domain-binding motif.
Conserved domains in the Stbm protein
Stbm shares no homology with known proteins, and with the exception of the putative PDZ domain-binding motif, there are no domains that suggest a possible role for Stbm. The existence of mammalian and worm homologs indicate the function of this protein is conserved. A structure-function analysis of this protein should help determine which regions are functionally important.
ACKNOWLEDGEMENTS
T. W. would like to express her gratitude to Marc Therrien and David Wassarman for their extensive expertise and patience in teaching her molecular biology; the cloning of stbm would not have been possible without their input. We are indebted to Todd Laverty for polytene chromosome analysis and Noah Solomon and Chris Suh for assistance with sequencing. We thank Bruce Hay for examining embryos and Kwang-Wook Choi and the Bloomington Stock Center for fly stocks. We are grateful to Mike Brodsky, Henry Chang, Ulrike Heberlein, Don Ready and David Wassarman for enlightening discussions and critical comments on the manuscript. We also thank John Tamkun for providing an aliquot of his cosmid library. T. W. is supported by an American Cancer Society fellowship and G. M. R. is a Howard Hughes Medical Institute Investigator.