Ommatidial polarity in the Drosophila eye is

0 downloads 0 Views 3MB Size Report
ventral halves of the disc rotate in opposite directions. This leads to the creation of an equator running along the dorsoven- tral midline, which corresponds to an ...
Development 121, 4247-4256 (1995) Printed in Great Britain © The Company of Biologists Limited 1995 DEV8284

4247

Ommatidial polarity in the Drosophila eye is determined by the direction of furrow progression and local interactions David I. Strutt and Marek Mlodzik Differentiation Programme, EMBL, Meyerhofstrasse 1, D-69117 Heidelberg, Germany

SUMMARY The adult eye of Drosophila is a highly ordered structure. It is composed of about 800 ommatidia, each displaying precise polarity. The ommatidia are arranged about an axis of mirror image symmetry, the equator, which lies along the dorsoventral midline of the eye. We use hedgehog pathway mutants to induce ectopic morphogenetic furrows and use these as a tool to investigate the establishment of ommatidial polarity. Our results show that ommatidial

clusters are self-organising units whose polarity in one axis is determined by the direction of furrow progression, and which can independently define the position of an equator without reference to the global coordinates of the eye disc.

INTRODUCTION

different terms, relative to the direction of furrow progression and on which side of the equator they lie (summarised in Fig. 1). Little is known of the mechanisms underlying ommatidial rotation and chirality. However, the known mutations disrupting the process affect ommatidia as discrete units, the individual ommatidia themselves being correctly assembled. This suggests that assembly is independent from the establishment of polarity. In the nemo mutant, ommatidia initially rotate 45 degrees, but then fail to complete rotation to 90 degrees, whilst in the roulette mutant ommatidia rotate further than 90 degrees (Choi and Benzer, 1994). As chirality and orientation relative to the equator remain normal in these two mutations, they appear to be involved in the actual mechanics of rotation, or in the reception or interpretation of signals required to determine the degree of rotation, rather than in the establishment of polarity. A second class of genes, the tissue polarity mutants (Gubb and García-Bellido, 1982), affect ommatidial polarity, producing defects in both rotation and chirality (Gubb, 1993; Theisen et al., 1994). These genes are good candidates for being involved in a signalling pathway required to transmit polarity information. In particular, one of them (frizzled) encodes a protein with structural similarity to a G-protein linked receptor (Vinson and Adler, 1987; Vinson et al., 1989). A major question is whether ommatidia respond to global signals (throughout the disc epithelium) or to local signals (e.g. between individual ommatidia), or both, when determining their polarity. A number of different models can be envisaged for how correct ommatidial polarity could be achieved. In an extreme ‘global coordinate’ model, each individual ommatidial cluster would determine its position relative to global positional information present in the eye disc, and then adopt the appropriate degree of rotation (in the wild-type eye, always 90 degrees) and chirality. In fact, as

In the Drosophila eye imaginal disc, pattern formation and associated differentiation begin during the third larval instar stage, and proceed in a wave starting at the posterior edge of the disc and progressing anteriorly. This wave is marked by the passage of the morphogenetic furrow, an indentation in the disc epithelium produced by transient cell shape changes. The furrow takes about 2 days to traverse from posterior to anterior. Ahead of it (anteriorly), cells are unpatterned and freely dividing, whilst in and behind it (posteriorly) cells are recruited into clusters of photoreceptors in a precise sequence, ultimately giving rise to the geometrically arranged mature ommatidia that constitute the adult eye (Ready et al., 1976; Tomlinson and Ready, 1987; Wolff and Ready, 1991). As the furrow advances in the anteroposterior axis, nascent ommatidial clusters emerge in rows (row 0 being defined as lying in the furrow itself). All clusters initially have a single axis of symmetry and face in the same direction. However, by row 6 the clusters have rotated by approximately 45 degrees away from the anteroposterior axis. They maintain this angle of rotation for about 10 rows, before rotating a further 45 degrees, bringing them 90 degrees from their original axis. The clusters then remain in this orientation throughout the rest of development and in the adult eye. Ommatidia in the dorsal and ventral halves of the disc rotate in opposite directions. This leads to the creation of an equator running along the dorsoventral midline, which corresponds to an axis of mirror-image symmetry. Concommitant with their rotation, the ommatidial clusters also lose their symmetry, and opposite chiral forms are established in the dorsal and ventral halves of the eye (Dietrich, 1909; Tomlinson and Ready, 1987; Wolff and Ready, 1991). Thus the final polarity of an ommatidium is defined relative to both the dorsoventral and the anteroposterior axes, or in

Key words: Drosophila, furrow progression, tissue polarity, ommatidial rotation, equator

4248 D. I. Strutt and M. Mlodzik clusters are normally born with anteroposterior polarity (as a result of the furrow moving in a defined direction), this model can be simplified to postulate only the existence of a system of positional information on the dorsoventral axis (Tomlinson, 1988; Baker and Rubin, 1992; Ma and Moses, 1995). A prediction of this model would be that the position of the equator is fixed relative to the global coordinates of the disc. Alternatively, in a ‘self-organising’ model, ommatidial clusters would determine their polarity by communicating with each other, for instance in a manner that might be considered to be analogous to crystallisation around a single nucleation centre (Gubb, 1993). Such a nucleation centre would be provided by an initial asymmetry, from which pattern would then propagate outwards. Traditionally, models invoking positional information have been favoured. For instance, classic grafting experiments on the retina of the insect Oncopeltus suggested that at least anteroposterior ommatidial polarity was determined by a gradient of positional information in the eye-disc epithelium (Lawrence and Shelton, 1975). In this study, we use mutations that induce ectopic morphogenetic furrows and neuronal differentiation to investigate the establishment of ommatidial polarity, and to test the predictions of these models. Our results indicate that the final polarity of ommatidia in the adult eye is dependent both upon the direction of furrow progression and upon the ability of ommatidial units to organise themselves independently to form an equator. We find no evidence for the involvement of a global positional information system in directly determining the polarity of ommatidia or the position of the equator.

MATERIALS AND METHODS Fly strains and generation of clones The pka-C1 allele used was the lethal P-insertion pka-C1l(2)01272 (Lepage et al., 1995; Strutt et al., 1995). The ptc allele used was ptc98/29, a lethal P-insertion identified in this laboratory, which behaves as a strong allele by all criteria tested and has been confirmed to carry no other mutations on the chromosome by reversion analysis (D. S. and M. M., unpublished data). Disc clones of pka-C1l(2)01272 and ptc98/29 were marked using arm-lacZ reporters inserted at 28A and 51A, respectively (Vincent et al., 1994). dpp expression was monitored using an insertion of the BS3.0 dpp-lacZ reporter gene on the third chromosome (Blackman et al., 1991). The svpl(3)7842 enhancer trap allele was used to reveal svp expression (Mlodzik et al., 1990). Clones were generated using the FLP/FRT system (Xu and Rubin, 1993). For adult clones, larvae of genotype w hsFLP1; pka-C1l(2)01272, P[ry+; hs-neo; FRT]40A / P[mini-w+; hs-πM]21C, 36F, P[ry+; hsneo; FRT]40A or w hsFLP1; P[ry+; hs-neo; FRT]43D, ptc98/29 / P[ry+; hs-neo; FRT]43D were heat-shocked at 38˚C for 120 minutes at 24-48 hours after egg-laying; clones were identified by the altered dosage of the white mini-genes carried by the P[mini-w+; hs-πM] or the ptc98/29 P-element insertions. For clones in discs marked by lack of arm-lacZ expression and carrying the svp enhancer trap, larvae of genotype w hsFLP1; pka-C1l(2)01272, P[ry+; hs-neo; FRT]40A / P[w+; arm-lacZ]28A, P[ry+; hs-neo; FRT]40A; svpl(3)7842 / + or w hsFLP1; P[ry+; hs-neo; FRT]43D, ptc98/29 / P[ry+; hs-neo; FRT]43D, P[w+; arm-lacZ]51A; svpl(3)7842 / + were used. Hh misexpression was carried out using flies of genotype y hsFLP1; P[ry+, Tub>y+>hh] / + (Basler and Struhl, 1994), again heat-shocking for 120 minutes in the first instar stage.

Histology Antibody stainings of eye imaginal discs were carried out by standard methods (Tomlinson and Ready, 1987). β-galactosidase staining and immunohistochemical double staining was as previously described (Strutt et al., 1995). Elav was detected using a rat monoclonal antibody (gift of G. Rubin), and β-galactosidase was detected using either a mouse monoclonal antibody (Promega) or a rabbit polyclonal antibody (Cappell). Secondary antibodies conjugated to peroxidase, FITC or Texas Red were used. Standard histological methods were used for sections of adult eyes (Tomlinson and Ready, 1987).

RESULTS

hh pathway mutations in the eye disc induce ectopic morphogenetic furrows Morphogenetic furrow progression is regulated by the hedgehog (hh) signalling pathway (Heberlein et al., 1993; Ma et al., 1993). Recently it has been shown that either misexpression of Hh protein or loss-of-function mutations in the catalytic subunit of cAMP-dependent protein kinase A (pkaC1, another component of the pathway) can induce ectopic expression of the TGF-β homologue decapentaplegic (dpp) and ectopic morphogenetic furrows anteriorly to the endogenous furrow (Heberlein et al., 1995; Pan and Rubin, 1995; Strutt et al., 1995). The segment-polarity gene patched (ptc) (Nüsslein-Volhard and Wieschaus, 1980) is also involved in the hh signalling pathway in the embryo, wing and leg imaginal discs. ptc encodes a multipass transmembrane protein, that acts downstream of hh, and has been proposed as a possible candidate for the Hh receptor (Hooper and Scott, 1989; Nakano et al., 1989; Ingham et al., 1991). In the anterior compartment of the wing disc, ectopic dpp expression can be induced by either ectopic Hh expression (Basler and Struhl, 1994), loss-offunction pka-C1 mutations (Jiang and Struhl, 1995; Lepage et al., 1995; Li et al., 1995; Pan and Rubin, 1995), or loss-offunction ptc mutations (Capdevila et al., 1994). Given the similarity of the pka-C1 and ptc mutant phenotypes in the wing disc, it seemed likely that ptc clones in the eye disc would also be capable of inducing ectopic morphogenetic furrows, associated with ectopic dpp expression and neuronal differentiation. This is indeed the case: in discs in which homozygous mutant ptc clones have been induced, ectopic dpp expression is seen anterior to the advancing furrow (Fig. 2A). Close to the furrow itself, ectopic ommatidial clusters can be seen surrounded by ectopic dpp expression (Fig. 2B). Thus as for pka-C1 mutant clones, ptc clones are capable of inducing ectopic photoreceptor differentiation and an associated outward moving morphogenetic wave. Again, as for pka-C1, there is a limited region in the zone anterior to the moving furrow that is able to sustain such ectopic differentiation, which we refer to as the ‘zone of competence’ (Strutt et al., 1995). The ability to induce ectopic morphogenetic furrows provides a powerful tool for investigating the relationship between furrow progression and the establishment of pattern and polarity in the eye disc. We have used Hh misexpression, and loss-of-function clones of pka-C1 and ptc to analyse the rotational behaviour of ommatidial clusters born from morphogenetic furrows that are not moving in the anteroposterior axis.

Furrow progression and ommatidial polarity 4249 Direction of furrow progression provides the first axis of polarity We analysed first the initial orientation of ommatidial clusters when they emerge from an ectopic furrow. In the wild-type disc, the direction of furrow movement is always along the anteroposterior axis. All ommatidia are then born lying in this axis, pointing posteriorly. By inducing loss-of-function ptc or pka-C1 clones, it is possible to produce furrows progressing in any direction. In this case, the initial orientation of the clusters is always along an axis coincident with the direction of furrow progression (Fig. 3A). The critical rôle of furrow movement in the initial determination of polarity can also be demonstrated in a second way: when a dpp-expressing ptc or pka-C1 clone enters the ‘zone of competence’ as a result of anterior furrow progression, neuronal differentiation occurs within the clone, and ommatidial clusters are born de novo (at least in the posterior of the clone), without emerging from a moving furrow. In many cases, these ommatidia are missing photoreceptors, apparently due to problems in recruitment caused by the nascent clusters being irregularly spaced, making it difficult to assess their orientation. However, in ommatidia that are normally constructed, orientation is found to be random. This can be seen both in clones in discs (Fig. 3B), and then later in the adult after differentiation and patterning is complete (Fig. 3C). Whilst it is possible that some normally constructed clusters are prevented from rotating by being too closely positioned to their neighbours, we frequently see well-spaced normally constructed clusters with random orientation. This suggests that clusters that are not born from a moving furrow are unable to adopt a correct orientation relative to either each other or their position in the disc. A second indication that clusters born within ptc or pka-C1 clones do not subsequently adopt polarity consistent with their position within the disc is the presence of mixed chiral forms of ommatidia in the adult eye. Normally all ommatidia in the dorsal half of the eye have the same chiral form, which is the mirror image of that adopted by ommatidia in the ventral part of the eye (i.e. on the other side of the equator that runs along the dorsoventral midline, see Fig. 1D). As both chiral forms are seen within clones on either side of the equator (Figs 3C, 5A-C), we conclude that chirality cannot be determined by absolute position within the disc. Ommatidia born from ectopic furrows do not always rotate towards the equator Although ommatidia are normally born facing along the anteroposterior axis, before photoreceptor recruitment is complete they rotate 45 degrees towards the equator of the disc (which runs along the dorsoventral midline) (Fig. 1B,C). Still later, they turn a further 45 degrees, bringing them 90 degrees from their original orientation, a configuration that is maintained in the adult eye (Fig. 1D). We have analysed the rotational behaviour of ectopic ommatidial clusters born from ectopic furrows progressing outwards from ptc and pka-C1 clones. In these cases, it is generally possible to predict the initial orientation of the clusters (i.e. along an axis determined by the direction of furrow progression), and thus to infer in which direction they have rotated. Clones were analysed in discs, using markers that permit the

Table 1. Summary of ptc clones* analysed by confocal microscopy in discs Total number of ptc clones examined† Scorable ommatidial polarity defects‡ Fields of clockwise and anticlockwise rotated ommatidia Opposing fields of rotation defining equators

41 29 12 10

*Consistent results were obtained for pka-C1 clones. †Approximately 2000 discs were inspected by fluorescence microscopy, of which 1 in 8 were the correct genotype. Discs with sufficiently large clones in suitable positions that disrupted patterning as revealed by Elav staining were selected for confocal imaging. ‡About 25% of the discs were not informative due to distortions in the epithelium caused by overgrowth as often associated with ectopic dpp expression (Strutt et al., 1995; Heberlein et al., 1995).

easy visualisation of the boundary of the clone and of ommatidial rotation, at a time when clusters would normally have turned 45 degree towards the equator of the disc (Table 1). A number of different observations were made (Fig. 4, see also Fig. 6). Firstly, in clones on either the dorsal or the ventral side of the equator, clusters are seen which rotate both clockwise and counterclockwise, and either towards or away from the equator of the disc. Thus, the direction in which clusters rotate cannot be an intrinsic property of their position in the disc. Moreover, neither is it obligatory that they rotate towards the normal equator of the disc. Secondly, in some cases rotation appears to be retarded, that is, clusters can be still facing along the axis of inferred furrow progression, even though a cluster of this maturity would normally have rotated 45 degrees. Thirdly, we do not find clusters rotated more than 45 degrees at this stage in their development. Thus, rotation does not always occur at a fixed point in the maturational sequence of ommatidial development, but if rotation has occurred then it is limited in degree. Ectopic ommatidia can organise themselves around ectopic equators Despite the fact that ectopic clusters do not always rotate towards the equator, their organisation nevertheless does not appear to be random. Instead, groups of clusters all rotating in the same direction are often observed. Strikingly, when two such groups of clusters are apposed, they can reveal the presence of an ‘ectopic’ equator, i.e a line of symmetry about which polarity is inverted. As the degree of rotation around disc clones is limited, and also in some cases retarded, ommatidia surrounding clones were analysed in the adult eye, after patterning is complete (Table 2 and Fig. 5). These experiments confirmed the results that were obtained by looking at rotation in discs. Firstly, ommatidia around clones are seen to have both chiral forms and to be able to rotate either clockwise or counterclockwise, regardless of where they lie with respect to the endogenous equator of the eye. Secondly, these ommatidia are very often organised around clearly defined ectopic equators. By analysing a large number of ptc and pka-C1 clones, we have found that ectopic equators can apparently run any direction out of a clone. The amount of non-autonomy (i.e. ectopic furrow progression) that can be produced by inducing clones is limited by the size of the zone of competence. Nevertheless, ectopic equators are commonly seen extending for 5 or more ommatidial rows with dozens of ommatidia arranged around

4250 D. I. Strutt and M. Mlodzik Fig. 1. The arrangement of ommatidial units in the wild-type eye. Anterior is to the left and dorsal is up, in this and in all subsequent figures. (A) An eye disc from a midthird instar stage larva. The position of the morphogenetic furrow as it progresses from posterior to anterior is revealed by expression of a dpplacZ reporter gene (blue). Behind the furrow, the arrangement of the assembling ommatidial clusters is revealed by expression of the neuronal specific antigen Elav (brown). The equator is visible as the axis of symmetry running along the dorsoventral midline (dotted line). (B) Confocal image of the equatorial region of an eye disc, stained for expression of a svp-lacZ enhancer trap (red). Expression is initially seen at highest levels in photoreceptor neurones R3 and R4, and then later at lower levels in R1 and R6 (Mlodzik et al., 1990). In row 4, when expression is first seen, clusters are only negligibly rotated from their original axis of symmetry along the anteroposterior axis. By row 6, clusters are rotated 45 degrees towards the equator (dotted line). (C) Schematic drawing of an eye disc, illustrating the rotational events that occur during third instar larval life. When ommatidial clusters emerge from the furrow, their axis of symmetry is in the anteroposterior axis (grey arrows), and they point backwards (if the first photoreceptor to mature, R8, is defined as initially lying at the apex of the cluster). By row 6, clusters have rotated 45 degrees towards the equator (black arrows indicate clockwise rotating clusters, red arrows anticlockwise), and maintain this orientation for about 10 rows, before rotating a further 45 degrees. Thus they finally lie 90 degrees from their original axis, pointing towards the equator. During rotation, the ommatidia acquire asymmetries, leading to the establishment on each side of the equator of opposite chiral forms (i.e. forms that no longer share rotational symmetry, indicated by tails on arrows). This arrangement is then maintained throughout pupal and adult life. (D) Section of an adult eye (left) at the R7 level and schematic drawing (right), showing ommatidial arrangement around the equator (yellow line in left panel, green line in right). Note that ommatidia are all facing towards the equator, and that opposite chiral forms (which do not share rotational symmetry) are present on each side (represented by black arrows for dorsal identity and red arrows for ventral identity). (E) Magnifications of single R7-level and R8-level ommatidia, illustrating relationship of arrows used in schematic drawings to actual photoreceptor arrangement. Numbering indicates identity of individual photoreceptors.

Fig. 2. Homozygous mutant ptc clones induce ectopic morphogenetic furrows. Third instar eye discs are shown in which unmarked ptc clones have been induced (see Figs 3-6 for marked clones). Ommatidial clusters are revealed by staining for the Elav nuclear antigen (brown), whilst dpp expression (normally in the furrow) is revealed by blue staining. (A) Disc showing ectopic dpp expression well anterior of the advancing furrow, and thus outside the ‘zone of competence’ in which ectopic morphogenetic furrow progression can occur (see text). (B) Disc showing ectopic dpp expression and differentiation of ommatidial clusters in the zone of competence anterior to the advancing endogenous furrow. In an example such as this, ectopic dpp expression would initially have been in a ring moving outwards from the clusters of ectopic photoreceptors, before merging with the advancing furrow. The behaviour of ptc clones thus observed is identical to the phenotypes of Hh misexpression and pka-C1 loss-of-function clones (Heberlein et al., 1995; Pan and Rubin, 1995; Strutt et al., 1995).

Furrow progression and ommatidial polarity 4251

Fig. 3. The direction of furrow progression determines initial ommatidial polarity. (A) Disc containing an ectopic furrow produced by induction of a ptc clone. Photoreceptors in the ommatidial clusters are stained brown, dpp expression in the furrow is blue. At the top of the panel, the endogenous furrow can be seen, progressing posterior to anterior. This is merged with a curved ectopic furrow. Emerging photoreceptor clusters have an axis of symmetry which is coincident with the direction of furrow progression. (B) Confocal image of a disc containing a ptc clone just merging with the endogenous furrow. The clone is marked by lack of cytoplasmic arm-lacZ staining (red). svplacZ nuclear staining (red) is superimposed with Elav nuclear staining (green). Ommatidial clusters which have arisen de novo inside the clone have random orientation. (C) Section of an adult eye containing a ptc clone (left) and schematic drawing (right). Clone is marked by increased levels of pigment (left) or grey shading (right). Arrows in schematic drawing are as in Fig. 1, circles indicate incomplete ommatidia which could not be scored. Within the clone, randomly oriented ommatidia of both chiral forms are observed. At the edges of the clone and outside, greater order is observed, but both chiral forms are nevertheless present.

Fig. 4. Organisation of ommatidial pattern around a ptc clone in the disc. Upper panel shows a confocal image of a ptc clone marked by lack of cytoplasmic arm-lacZ staining (red). svp-lacZ nuclear staining in a subset of photoreceptors (red) is superimposed with Elav in all photoreceptors (green). The sensitivity of the nuclear svp-lacZ staining is somewhat reduced by being used in a triple label with cytoplasmic arm-lacZ and Elav: thus strong svp staining is only seen in clusters mature enough to have completed the first 45 degree turn (compare Fig. 1B). Lower panel is a schematic drawing, with clone shown by grey shading. Inset indicates position of clone in disc. Arrows are as in Fig. 1, with their colour indicating deduced direction of rotation. Circles mark clusters outside the clone which could not be scored. Within the posterior part of the clone, ommatidial orientation appears random, but many clusters are incomplete or have distorted shape, which makes precise scoring difficult. The clone is sufficiently elongated in the anteroposterior axis, that differentiation in the anterior part (i.e. furthest from the zone of competence) is occurring as a result of an anteriorly moving furrow, rather than de novo. A number of clusters on the dorsal edge are unrotated or only neglibly rotated (grey arrows in lower panel), compared to more mature neighbours. More dorsally, ventrally and anteriorly, fields of ommatidia are seen which have rotated in opposite directions. In some cases, field of oppositely rotated ommatidia are opposed, providing the first evidence of equator formation (yellow lines in upper, green in lower).

4252 D. I. Strutt and M. Mlodzik Table 2. Summary of ptc and pka-C1 clones analysed in sections of adult eyes ptc clones

Position

No. No. showing analysed No. showing No. showing ‘hijacking’ of in detail* ectopic equator(s) multiple equators normal equator

Dorsal Central Ventral Total

13 12 7 32

8 10 3 21

Dorsal Central Ventral Total

13 25 9 47

5 18 4 27

4 7 − 11

n.a. 6 n.a. 6

1 9 1 11

n.a. 7 n.a. 7

pka-C1 clones

*Total number of clones sectioned was >100. Only the ones with good morphology and angle of section are listed. n.a. = not applicable.

them and, in exceptional cases, large proportions of the eye field are repatterned. Interestingly, in the adult, we do not see large numbers of unrotated ommatidia at the edge of clones, where we can predict their initial polarity. Instead 90 degree rotation is seen, as for ommatidia in a wild-type eye. This is in contrast to our observation in the disc, where clusters are often seen to be unrotated at the edge of clones. This is consistent with a view that all ommatidia rotate 90 degrees during their development, but that the time of rotation is not strictly fixed. The position of the equator is not fixed along the dorsoventral midline In certain cases, when a clone is induced close to the normal equator of the disc, clusters anterior to the clone can be arranged around an ectopic equator which is to one side of the original equator (Fig. 6A). This can lead to the endogenous equator being ‘hijacked’ by the ectopic equator, leading to a single equator positioned away from the dorsoventral midline, and not necessarily running precisely along the anteroposterior axis. Thus, the position of the endogenous equator can not be fixed by a global mechanism. A similar result can be produced by perturbing the even progression of the endogenous furrow throughout its movement across the disc, leading to a furrow that has abnormal topology. Such an effect can be produced by activating the hh signalling pathway throughout the disc, either by using hypomorphic viable alleles of ptc (D. S. and M. M., unpublished data), or by using the tub>y+>hh construct that permits ubiquitous expression of Hh protein throughout a tissue (Basler and Struhl, 1994). In such a situation, the furrow does not move evenly across the disc in the anteroposterior axis. This can lead to not only the creation of more than one ‘leading edge’ of furrow progression, but also to the formation of multiple equators in the disc, around which large fields of ommatidia are arranged (Fig. 6B). In summary, these results show that there is not a fixed equator in the eye disc around which ommatidia organise themselves, but that an equator may occur in any position.

DISCUSSION Within the highly ordered geometric array of the adult eye, ommatidia display precise polarity, being arranged in mirrorsymmetric fashion about an equator that runs along the dorsoventral midline (Dietrich, 1909; Tomlinson and Ready, 1987). Our results provide insights into what mechanisms underlie the establishment of this polarity. Direction of furrow progression is a direct determinant of ommatidial polarity Our results indicate a central rôle for the furrow as an organiser of ommatidial polarity. This is reflected in the fact that the direction of furrow progression determines the initial orientation of each ommatidial cluster. Furthermore, it is likely that ommatidial units are limited to rotating no more than 90 degrees, as to a first approximation most ommatidia surrounding ptc or pka-C1 clones always show this degree of rotation (in the adult). Clusters that are not born from a moving furrow have random orientation, which they seem to maintain throughout development. This supports the hypothesis that ommatidia are not able to freely rotate in order to achieve a polarity that is consistent either with that of their neighbours or with their position in the disc. Thus the first component of polarity information that an ommatidium receives by virtue of being born from a moving furrow is a direct determinant of its final polarity. We propose therefore that there is no requirement for an anteroposterior positional information system in the eye imaginal disc to which ommatidia are able to respond in order to determine their polarity. How is the direction of ommatidial rotation determined? The second component of polarity information that an ommatidium responds to is apparent in the direction that it rotates and in the chiral form that it subsequently adopts. Normally, all ommatidia in the dorsal half of the eye will rotate in the same direction and adopt the same chiral form, whilst all ommatidia in the ventral half will rotate in the opposite direction and adopt the opposite chiral form. One possible explanation for this behaviour would be that ommatidia were aware of their position in either the dorsal or the ventral half of the disc. This information could be conveyed by a system of positional information and the developing clusters would then be ‘pre-programmed’ to rotate in the appropriate direction and adopt the correct chiral form. A simpler form of such a model would be that an ommatidium would merely sense in which direction the equator (i.e. the dorsoventral midline) lay and rotate towards it, as has recently been proposed (Ma and Moses, 1995). Following induction of ptc and pka-C1 clones, we see aberrant behaviour of ectopic ommatidial clusters. Firstly, both within and around clones, we observe ommatidia of both chiral forms. Secondly, around clones, we observe ommatidia that are predicted to have rotated in either of the possible directions. Thirdly, rotation is not always in a consistent direction relative to the dorsoventral midline for an ommatidium of given starting polarity. Thus, chiral form and direction of rotation cannot be directly linked to absolute dorsoventral position in the disc. Neither can their rotation be correlated with the direction in which the dorsoventral midline lies. Therefore, it

Furrow progression and ommatidial polarity 4253 can be excluded that the rotational behaviour of these ectopic ommatidial clusters is a result of their ability to sense their position in the disc relative to the dorsoventral axis. Nevertheless, despite the fact that there is no consistent relationship between direction of rotation and the dorsoventral axis, ommatidial behaviour is not random. Fields of ommatidia rotate in the same direction and adopt the same chiral form. Strikingly these fields are often apposed in such a way as to give rise to ectopic equators. Furthermore, the ommatidia arranged around such ectopic equators have the correct chiral form relative to the arrangement around the endogenous equator. Such ectopic equators can be oriented in all directions within the disc, and exhibit identical behaviour whether they lie in the dorsal or the ventral half of the disc. The failure of ectopic ommatidia to respect the dorsoventral coordinates of the eye disc could have two possible explanations. The first is that there is no system of dorsoventral positional information, but instead the cells in the disc epithelium are essentially naive about their position. In this case, they would gain information about polarity by communication with their neighbours. As the ectopic ommatidia are isolated from the ‘organised’ ommatidia behind the furrow, they are unable to adopt a polarity consistent with their absolute position. The second explanation is that ectopic activation of the hh pathway by induction of ptc or pka-C1 clones is in some way interfering with or redefining the dorsoventral coordinates of the disc epithelium. This might imply a central rôle for the hh pathway in defining coordinates on two axes. Although we cannot definitely exclude either possibility, we favour the first view as being sufficient to account for our results. It is interesting to note that quite small clones can permanently shift the position of the equator: if these clones were causing only a local perturbation in positional values, then it might be expected that the equator would eventually shift back to its original axis. The fact that this does not happen, again supports the hypothesis that dorsoventral positional values are not globally determined.

as a result of ubiquitous hh pathway activation) then ommatidia throughout the disc can be organised around multiple equators. These ‘extra’ equators originate at the leading edges of the curved furrow (see Fig. 6). Taking these results together, we conclude that equators are defined by the leading edge of the furrow and that furrow topology is the critical factor. With respect to the normal equator in a wild-type disc, this hypothesis fits well with the previous experimental observation that ommatidia in the centre of a row emerge sooner than those further towards the edges of the disc (Wolff and Ready, 1991). These most mature ommatidia at the dorsoventral midline could then constitute an organising centre or node which defines the position of the equator. This information would then be propagated outwards to less mature ommatidial clusters (Fig. 7). This model of an outward sweep of patterning from a node has already been proposed as the most likely mode of patterning in the eye imaginal disc (Gubb, 1993). It is a corollary of this model that ommatidial units can organise themselves, with no input from external coordinate systems. The formation of ectopic equators by ptc and pka-C1 clones can be easily explained by this model. Such clones give rise to ectopic furrows that are normally curved and also, when they merge with the endogenous furrow usually produce convex distortions. This curvature produces leading edges similar to those that would occur at the posterior edge of the eye disc when furrow progression first begins (see Fig. 7). The clusters that emerge from these ectopic leading edges then form nodes around which ectopic equators are organised. To a large extent, whether and where a node forms will depend on the shape of the clone and where it intersects with the normal furrow. Furthermore, it is easy to envisage that a clone could give rise to multiple nodes which might compete with each other and with the endogenous equator in attempting to organise adjacent ommatidia. These factors would account for the lack of regularity with which ectopic equators are formed, and the fact that not all ommatidia around clones can be clearly seen to be organised around an equator.

How is the position of an equator determined? Although we observe equators moving in all directions in the disc, and sometimes with curved rather than straight paths, they always begin at the edge of a clone and move outwards. Therefore, their initial direction of travel corresponds to the inferred direction of furrow progression. This is consistent with the observation that ommatidia apparently can only rotate 90 degrees from their initial axis (which is itself determined by the direction of furrow progression). Thus an equator must always travel in the same direction as the furrow itself. This intimate relationship between the axis of equator formation and the direction of furrow progression is an important observation, because it suggests that equator formation may actually be dependent upon furrow progression. Two experimental results are particularly informative in this context. Firstly, the fact that the endogenous equator can be shifted off its normal course by perturbing furrow progression (with a small ptc or pka-C1 clone), indicates that there is nothing special about this equator and that it too is influenced by furrow progression. Secondly, if the endogenous furrow does not move evenly throughout development (for instance,

How is polarity information propagated from nodes? An interesting question is how a node is formed and how information is subsequently propagated outwards. There is clearly no requirement for a special ‘node-building’ mechanism at the posterior edge of the disc (where the normal equator begins). Rather, it is observed that cells anywhere in the eye field are capable of participating in the interactions that can give rise to a node (and hence an equator). Our favoured hypothesis is that when ommatidial clusters are born from the furrow, they are receptive to signals from their more mature neighbours which normally would determine in which direction they should rotate. If their neighbours (lying either posteriorly or laterally) are already organised relative to an equator, then they also will rotate appropriately and similarly signal this decision to their neighbours. If no clear signals are received, then a group (or at least a pair) of adjacent clusters communicate and define an equator between them. As the clusters already have defined polarity in one axis (as a consequence of the furrow moving in a defined direction), this definition of an equator provides them with the second axis. The clusters then rotate and adopt the correct chiral form. This group (or pair) of clusters now provides the

4254 D. I. Strutt and M. Mlodzik node about which other clusters can be recruited into a regular array. The definition of the node is obviously the critical stage, but to a large extent this may occur by default, with the most mature clusters performing this function. Thus equators are formed by the leading edge of the furrow because this gives rise to the most mature clusters. It is interesting to note that a node can also be ‘broken’. This is the case when a ptc or pka-C1 clone perturbs or stops the progress of the endogenous equator, as a result of an ectopic furrow intersecting with the endogenous furrow close to the dorsoventral midline (Fig. 6A). This again emphasises the point that the endogenous equator has no special properties, but itself is born from a self-organising node. Further support for the model is provided by the observation that ommatidial clusters born from ectopic furrows often seem slow to rotate, relative to clusters born from the normal furrow. This delay might well represent the time needed for ommatidia to communicate together in order to define a node. It should also be noted that clusters on the edges of clones are lying adjacent to randomly oriented clusters inside the clone, which may themselves be sending out conflicting signals, a factor that again might delay rotation. Of course, delays in rotation around a clone might also represent a conflict between signals from a global positional information system and a pertubation Fig. 5. Ectopic equators in the adult eye induced by ptc clones. Panels on left show sections, the area of the clone being marked by increased levels of pigment and ectopic equators being indicated in yellow. Panels on right are schematic drawings in which clones are marked by grey shading and ectopic equators by green lines; arrows are as in Fig. 1 with colour indicating chirality, circles mark incomplete ommatidia which could not be scored. Insets show position of clone in eye. None of these sections include the endogenous equator, which lies ventrally in A, posteriorly in B and dorsally in C. Note that ommatidia of both chiral forms are seen in all the sections, and that rotation both towards and away from the endogenous equator must have occurred. (A) Clone touching dorsal edge of eye. An ectopic equator runs out dorsally, organising five rows of ommatidia. Ventrally a short equator can also be defined extending for 2 rows. (B) Large clone lying centrally just below the dorsoventral midline. Ectopic equators running largely in the dorsoventral axis are present both above and below the clone. Within the section shown, about 5 rows of ommatidia are organised around each of these equators, but in fact the equators extend further. The endogenous equator stops at the posterior edge of the clone (outside section shown). (C) Ventral clone, giving rise to an ectopic equator running approximately in the anteroposterior axis, parallel to the endogenous equator.

in this system being caused by ectopic activation of the hh pathway. However, if this were the case, it is surprising how often ectopic equators are observed (Table 2), rather than ommatidia adopting an orientation appropriate to their position in the disc. At present the molecular nature of the signals involved in interommatidial communication is unknown. However, there is increasing evidence that the tissue polarity class of genes might constitute a signalling system and, as their mutant phenotypes affect ommatidial polarity, these are excellent can-

Furrow progression and ommatidial polarity 4255 Fig. 6. The position of the endogenous equator is not fixed. (A) Upper panel shows a confocal image of a ptc clone in the disc, lower panel shows schematic drawing. Staining and labelling is as for Fig. 4. The endogenous equator (running from the posterior edge) is disrupted when it comes into contact with clusters born from an ectopic furrow around the ptc clone. A single equator continues anteriorly and slightly dorsally out of the clone, around which all of the clusters anterior to the clone are organised. Two other possible small equators are marked as extending out of the clone in the dorsoventral axis, one of almost intersects with the endogenous equator. (B) Disc from a larva carrying the tub>y+>hh transgene, which has been used to induce ubiquitous Hh expression during first instar larval life. This treatment gives rise to a furrow (marked by dpp expression in blue) which progresses unevenly in the anteroposterior axis. The ommatidial arrangement is revealed by Elav staining (brown). At least two lines of symmetry can be seen in the arrangement of the ommatidia, representing two equators (black lines). These equators originate at the leading edges of the furrow, which we regard as acting as organising centres.

Fig. 7. Model for the formation of the equator. (A) The mature ommatidial clusters at the posterior edge of a wild-type third instar eye disc, as revealed by Elav staining. The rows of ommatidia are curved, and more clusters are added in successive rows, leading to centre out development. Arrow indicates position of equator. (B) Schematic drawing, representing the formation of a node at the posterior edge of the disc. Three consecutive stages of development are shown, from least mature on the left to most mature on the right. Rotational events have been accelerated for clarity. Initially ommatidial clusters face towards the furrow from which they have just emerged. By row 5-6, the most mature clusters are rotating away from each other to form an equator, following inter-ommatidial communication. Less mature clusters can then receive polarity information both by communicating with their neighbours and with more mature clusters lying posteriorly. Note that polarity information essentially travels from the equator outwards, ensuring that precise packing can be achieved.

4256 D. I. Strutt and M. Mlodzik didates to be involved (Gubb and García-Bellido, 1982; Vinson and Adler, 1987; Vinson et al., 1989; Gubb, 1993; Theisen et al., 1994). A recent analysis of the function of one of this class of genes (frizzled) in eye development suggests that polarity information is transmitted outwards from the equator (Zheng et al., 1995), in a fashion consistent with our view of an outward sweep of polarity information from a node at the leading edge of the furrow. An independent analysis of the effects of ectopic furrows on ommatidial pattern formation also supports our conclusion that the leading edge of the furrow serves as an organising centre for equator formation (Chanut and Heberlein, 1995). Multiple equators can also be produced in the Drosophila eye by localised cell death (caused by induction of eyeless clones) dividing the eye field into two separate regions (Campos-Ortega, 1980). Although it is not clear what the mechanism underlying this phenomenon might be, it is nevertheless compatible with the notion that isolated ommatidia are able to self-organise independently of the global coordinates of the disc. CONCLUSIONS In summary, we propose that ommatidial polarity is defined by two factors: firstly, the direction of furrow progression and, secondly, the ability of ommatidia to communicate in a polarised manner and autonomously organise themselves to produce an equator. This then defines a node or nucleation centre about which subsequent pattern is built. We thank Konrad Basler for flies, Gerry Rubin for providing the Elav antibody, Steve Cohen and Bill Brook for reading the manuscript, and Ulrike Heberlein, Kevin Moses and Richard Carthew for sharing unpublished results. D. S. was supported by a Wellcome Trust Travelling Research Fellowship and an EU Human Capital and Mobility Fellowship.

REFERENCES Baker, N. E. and Rubin, G. M. (1992) Ellipse mutations in the Drosophila homologue of the EGF receptor affect pattern formation, cell division and cell death in eye imaginal discs. Dev. Biol. 150, 381-396. Basler, K. and Struhl, G. (1994) Compartment boundaries and the control of Drosophila limb pattern by hedgehog protein. Nature 368, 208-214. Blackman, R. K., Sanicola, M., Raftery, L. A., Gillevet, T. and Gelbart, W. M. (1991) An extensive 3′ cis-regulatory region directs the imaginal disc expression of decapentaplegic, a member of the TGF-β family in Drosophila. Development 111, 657-666. Campos-Ortega, J. A. (1980) On compound eye development in Drosophila melanogaster. In Current Topics in Developmental Biology (ed. R. K. Hunt), pp. 347-371. London: Academic Press. Capdevila, J., Estrada, M. P., Sánchez-Herrero, E. and Guerrero, I. (1994) The Drosophila segment polarity gene patched interacts with decapentaplegic in wing development. EMBO J. 13, 71-82. Chanut, F. and Heberlein, U. (1995). Role of the morphogenetic furrow in establishing polarity in the Drosophila eye. Development 121, 4085-4094. Choi, K.-W. and Benzer, S. (1994) Rotation of photoreceptor clusters in the developing Drosophila eye requires the nemo gene. Cell 78, 125136. Dietrich, W. (1909) Die Facettenaugen der Dipteren. Z. Wiss. Zool. 92, 465539. Gubb, D. (1993) Genes controlling cellular polarity in Drosophila. Development Suppl. 1993, 269-277.

Gubb, D. and García-Bellido, A. (1982) A genetic analysis of the determination of cuticular polarity during development in Drosophila melanogaster. J. Embryol. Exp. Morphol. 68, 37-57. Heberlein, U., Singh, C. M., Luk, A. Y. and Donohue, T. J. (1995) Growth and differentiation in the Drosophila eye coordinated by hedgehog. Nature 373, 709-711. Heberlein, U., Wolff, T. and Rubin, G. M. (1993) The TGFβ homolog dpp and the segment polarity gene hedgehog are required for the propagation of a morphogenetic wave in the Drosophila retina. Cell 75, 913-926. Hooper, J. E. and Scott, M. P. (1989) The Drosophila patched gene encodes a putative membrane protein required for segmental patterning. Cell 59, 751765. Ingham, P. W., Taylor, A. M. and Nakano, Y. (1991) Rôle of the Drosophila patched gene in positional signalling. Nature 353, 184-187. Jiang, J. and Struhl, G. (1995) Protein kinase A and hedgehog signalling in Drosophila limb development. Cell 80, 563-572. Lawrence, P. A. and Shelton, P. M. J. (1975) The determination of polarity in the developing insect retina. J. Embryol. Exp. Morph. 33, 471-486. Lepage, T., Cohen, S. M., Diaz-Benjumea, F. J. and Parkhurst, S. M. (1995) Signal transduction by cAMP-dependent protein kinase A in Drosophila limb pattern. Nature 373, 711-715. Li, W., Ohlmeyer, J. T., Lane, M. E. and Kalderon, D. (1995) Function of protein kinase A in hedgehog signal transduction and Drosophila imaginal disc development. Cell 80, 553-562. Ma, C. and Moses, K. (1995) wingless and patched are negative regulators of the morphogenetic furrow and can affect tissue polarity in the developing Drosophila compound eye. Development 121, 2279-2289. Ma, C., Zhou, Y., Beachy, P. A. and Moses, K. (1993) The segment polarity gene hedgehog is required for progression of the morphogenetic furrow in the developing Drosophila eye. Cell 75, 927-938. Mlodzik, M., Hiromi, Y., Weber, U., Goodman, C. S. and Rubin, G. M. (1990) The Drosophila seven-up gene, a member of the steroid receptor gene superfamily, controls photoreceptor cell fates. Cell 60, 211-224. Nakano, Y., Guerrero, I., Hidalgo, A., Taylor, A., Whittle, J. R. S. and Ingham, P. W. (1989) A protein with several possible membrane-spanning domains encoded by the Drosophila segment polarity gene patched. Nature 341, 508-513. Nüsslein-Volhard, C. and Wieschaus, E. (1980) Mutations affecting segment number and polarity in Drosophila. Nature 287, 795-801. Pan, D. and Rubin, G. M. (1995) cAMP-dependent protein kinase and hedgehog act antagonistically in regulating decapentaplegic transcription in Drosophila imaginal discs. Cell 80, 543-552. Ready, D. F., Hanson, T. E. and Benzer, S. (1976) Development of the Drosophila retina, a neurocrystalline lattice. Dev. Biol. 53, 217-240. Strutt, D. I., Wiersdorff, V. and Mlodzik, M. (1995) Regulation of furrow progression in the Drosophila eye by cAMP-dependent protein kinase A. Nature 373, 705-709. Theisen, H., Purcell, J., Bennett, M., Kansagara, D., Syed, A. and Marsh, J. L. (1994) dishevelled is required during wingless signalling to establish both cell polarity and cell identity. Development 120, 347-360. Tomlinson, A. (1988) Cellular interactions in the developing Drosophila eye. Development 104, 183-193. Tomlinson, A. and Ready, D. F. (1987) Neuronal differentiation in the Drosophila ommatidium. Dev. Biol. 120, 366-376. Vincent, J.-P., Girdham, C. H. and O’Farrell, P. H. (1994) A cellautonomous, ubiquitous marker for the analysis of Drosophila genetic mosaics. Dev. Biol. 164, 328-331. Vinson, C. R. and Adler, P. N. (1987) Directional non-cell autonomy and the transmission of polarity information by the frizzled gene of Drosophila. Nature 329, 549-551. Vinson, C. R., Conover, S. and Adler, P. N. (1989) A Drosophila tissue polarity locus encodes a protein containing seven potential transmembrane domains. Nature 338, 263-264. Wolff, T. and Ready, D. F. (1991) The beginning of pattern formation in the Drosophila compound eye: the morphogenetic furrow and the second mitotic wave. Development 113, 841-850. Xu, T. and Rubin, G. M. (1993) Analysis of genetic mosaics in developing and adult Drosophila tissues. Development 117, 1223-1237. Zheng, L., Zhang, J. and Carthew, R. W. (1995) frizzled regulates mirrorsymmetric pattern formation in the Drosophila eye. Development 121, 30453055. (Accepted 17 August 1995)