pannier acts upstream of wingless to direct dorsal eye disc development in Drosophila

Developmental fields are frequently organized by a boundary formed between two differently determined regions, which subsequently serves as a source of patterning molecules (Vincent, 1998). The best studied example is the anterior-posterior compartment boundary of the Drosophila wing disc, which is formed at the border between posterior cells expressing the homeodomain protein Engrailed (En), and anterior cells that do not express En (Tabata et al., 1995). En controls the expression of Hedgehog (Hh) protein, which is secreted from posterior cells and activates the expression of Decapentaplegic (Dpp) on the anterior side of the boundary (Basler and Struhl, 1994; Tabata and Kornberg, 1994; Zecca et al., 1995). Secreted Dpp then acts over a long range to pattern the entire wing disc (Lecuit et al., 1996; Nellen et al., 1996). In the Drosophila eye disc, the morphogenetic furrow, where dpp is expressed under the control of Hh secreted by posterior retinal cells, resembles the wing anterior-posterior compartment boundary, but is a moving border that only transiently separates anterior and posterior regions of the disc (Heberlein and Moses, 1995). However, the dorsoventral midline of the eye disc has been shown to act as a lineage restriction boundary and a patterning center, which both promotes growth of the eye field and organizes its polarity (Cavodeassi et al., 1999; Cho and Choi, 1998; Dominguez and de Celis, 1998; Papayannopoulos et al., 1998; Zheng et al., 1995).

The adult compound eye of Drosophila comprises approximately 800 ommatidial units, each composed of eight photoreceptors organized in a characteristic and asymmetric trapezoid, as well as cone and pigment accessory cells. Two chiral forms of these ommatidial clusters are arranged in a mirror-symmetric pattern around the dorsoventral midline, also known as the equator (Ready et al., 1976). This global mirror symmetry originates during the third larval instar. The ommatidial preclusters are initially symmetric, but as they mature, cell movements introduce asymmetry within each precluster. In addition, preclusters on the dorsal and ventral sides of the disc undergo a 90˚ rotation in opposite directions (Ready et al., 1976; Tomlinson, 1985). The direction of this rotation appears to be controlled by tissue polarity genes including wingless (wg; Ma and Moses, 1995; Wehrli and Tomlinson, 1998), frizzled (fz; Zheng et al., 1995), Notch (N; Cooper and Bray, 1999; Fanto and Mlodzik, 1999), prickle/spiny legs (Gubb, 1993), strabismus (Wolff and Rubin, 1998) and RhoA (Strutt et al., 1997), while its extent depends on the genes nemo and roulette (Choi and Benzer, 1994).

The subdivision of the eye disc into dorsal and ventral territories takes place prior to ommatidial differentiation and is essential to define the site at which differentiation will initiate and to promote the growth of the eye field. The critical outcome of this subdivision appears to be the activation of the Notch (N) receptor in cells at the dorsoventral midline. This is due to the restricted expression of fringe (fng) in the ventral half of the eye disc (Papayannopoulos et al., 1998; Cho and Choi, 1998; Dominguez and de Celis, 1998). fng encodes a secreted molecule that mediates the interactions between N and its ligands Delta (Dl) and Serrate (Ser), promoting the response of N to Dl and inhibiting its response to Ser (Irvine and Wieschaus, 1994; Panin et al., 1997). In the wing disc, fng is coexpressed with Ser in the dorsal compartment, while Dl is expressed in the ventral compartment; this results in N activation only at the dorsoventral compartment boundary (Kim et al., 1995; Panin et al., 1997). In the eye disc these interactions are very similar, except that fng and Ser are expressed ventrally and Dl dorsally (Cho and Choi, 1998; Dominguez and de Celis, 1998; Papayannopoulos et al., 1998). Either loss of fng activity or misexpression of fng throughout the disc causes a failure of the eye disc to grow and differentiate (Cho and Choi, 1998; Dominguez and de Celis, 1998; Papayannopoulos et al., 1998). This failure of eye development can be rescued by a constitutively activated form of N, which leads to overgrowth of the eye field (Dominguez and de Celis, 1998).

fng expression is restricted to the ventral half of the eye disc, apparently due to dorsal repression by genes of the iroquois Complex (iro-C) (Cavodeassi et al., 1999; Cho and Choi, 1998; Dominguez and de Celis, 1998). The iro-C contains three genes, araucan (ara), caupolican (caup) and mirror (mirr), encoding highly related homeoproteins that are expressed in the dorsal half of the eye (Cho and Choi, 1998; Dominguez and de Celis, 1998; Gomez-Skarmeta et al., 1996; McNeill et al., 1997). Ectopic expression of either caup or mirr has been shown to repress fng expression (Cho and Choi, 1998; Dominguez and de Celis, 1998), while loss of iro-C function leads to ectopic fng expression (Cavodeassi et al., 1999). Furthermore, it has been shown that the dorsal expression of iro-C genes requires the secreted proteins Wingless (Wg) and Hh (Cavodeassi et al., 1999; Heberlein et al., 1998). Early Wg expression appears to be restricted to the dorsal side of the eye disc (Cavodeassi et al., 1999), although it is also present at the ventral margin of the disc later in development (Baker, 1988). The JAK/STAT pathway ligand Unpaired (Upd), present at the midline of the posterior margin (Zeidler et al., 1999), and genes of the Polycomb group of chromatin repressors (Netter et al., 1998) have been implicated in the ventral repression of mirr. These genes thus define a signaling pathway that functions to position the equator. The most downstream component of this cascade is the activation of N at the dorsoventral midline, and so far, the most upstream is wg expressed at the dorsal margin of the eye disc.

The expression of wg is restricted to the more anterior regions of the lateral margins by the activity of Dpp present at the posterior margin (Wiersdorff et al., 1996); however, little is known about which positive regulators are responsible for its expression in the eye disc. In this paper we report that pannier (pnr), which encodes a transcription factor homologous to vertebrate GATA-1 (Ramain et al., 1993), activates the dorsal expression of wg. pnr and the iro-C genes act as prepattern genes in the presumptive notal region of the wing disc, where they are required to delimit the expression of the proneural genes achaete (ac) and scute (sc) (Diez del Corral et al., 1999; Garcia-Garcia et al., 1999; Gomez-Skarmeta et al., 1996; Kehl et al., 1998; Leyns et al., 1996; Ramain et al., 1993). Pnr also acts as a regulator of wg in the notum (Calleja et al., 1996; Garcia-Garcia et al., 1999). The expression of pnr at the dorsal margin of the eye disc led us to investigate its function during eye development using mosaic analysis and ectopic expression. We show that the boundary between cells expressing and not expressing pnr promotes eye growth and organizes ommatidial polarity. We find that pnr acts upstream of wg by positively regulating its dorsal expression. Consequently, pnr is also required to activate dorsal mirr expression and repress ventral fng expression, leading to the activation of N at the dorsoventral midline. These results implicate Pnr as the most upstream component of the signaling cascade that leads to the normal position of the equator. The presence of pnr in the dorsalmost cells of the embryonic eye disc primordium suggests that dorsoventral patterning of the embryo directly establishes the dorsoventral axis of the eye disc.

Alleles used were pnr^VX6 (Heitzler et al., 1996), ush² (Nüsslein- Volhard et al., 1984), and wg^CX2 (Baker, 1987). Transgenic flies used were UAS-pnr^D4 (Haenlin et al., 1997), UAS-ush (Cubadda et al., 1997), UAS-ara (Gomez-Skarmeta et al., 1996), UAS-wg (Azpiazu et al., 1996), UAS-fluΔarm (Zecca et al., 1996), UAS-FLP (Duffy et al., 1998), UAS-N^intra (Doherty et al., 1996), eyFLP1 (a generous gift from

B. Dickson), pnr-GAL4 (Calleja et al., 1996), and ey-GAL4 (Hazelett et al., 1998). The lacZ reporters were wg^P (Kassis et al., 1992), mrr^cre2 (Hartenstein and Jan, 1992), iro^rF209 (Gomez-Skarmeta et al., 1996), ey-lacZ (Quiring et al., 1994) and the 35UZ-1 P insertion in fng (Irvine and Wieschaus, 1994).

To generate pnr^VX6 mutant clones in the eye, FRT82, pnr^VX6/TM6B males were crossed with eyFLP1;FRT82, P(arm-lacZ)/TM6B females. To generate clones in a Minute background, the same males were crossed with eyFLP1;FRT82, P(arm-lacZ), M(3)96C/ TM6B females. To generate wg^CX2 or ush² mutant clones in the eye in a Minute background, FRT40,wg^CX2 or ush²/SM6;TM6B males were crossed with eyFLP1;FRT40, M(2)24F, P(arm-lacZ)/CyO females. To rescue pnr^VX6 clones by overexpressing wg or ara, ey-GAL4, UAS- FLP;FRT82, P(arm-lacZ)/SM6;TM6B females were crossed to UAS- wg or UAS-ara, FRT82, pnr^VX6/ TM6B males.

Eye imaginal discs were dissected from third instar larvae into 0.1 M phosphate buffer, pH 7.2, fixed for 30 minutes on ice in 4% formaldehyde in PEM (0.1 M Pipes, pH 7.0, 2 mM MgSO₄, 1 mM EGTA) and washed in 0.1 M phosphate buffer, 0.2% Triton X-100. Eye imaginal discs were stained with rat anti-Elav (Robinow and White, 1991; diluted 1:5) as described by Hazelett et al. (1998), followed by X-gal staining at 37˚C in staining buffer (0.2% X-gal, 10 mM phosphate buffer, pH 7.2, 150 mM NaCl, 1 mM MgCl₂, 3 mM K₄[FeII(CN)₆], 3 mM K₃[FeIII(CN)₆], 0.2% Triton X-100) to reveal the pattern of arm-lacZ expression. Tangential sections of adult eyes were prepared as described by Tomlinson and Ready (1987).

In situ hybridization to whole-mount eye-antennal imaginal discs was performed using an antisense RNA probe labeled with digoxigenin- UTP (Roche Molecular Biochemichals) and synthesized from a 2 kb fragment of genomic DNA containing the GATA domain of pnr (a gift from M. Frasch). Eye-antennal discs were dissected in PBS, fixed for 20 minutes on ice in 4% formaldehyde in PBS, and fixed again at room temperature for 20 minutes in 4% formaldehyde, 0.1% sodium deoxycholate in PBT (PBS + 0.1% Tween 20). After five washes in PBT, they were dehydrated through an ethanol series and stored, if required, in 70% ethanol/PBT at −20˚C. On the day of hybridization, discs were rehydrated through an ethanol series and washed three times in PBT. They were then fixed in 10% formaldehyde in PBT for 20 minutes and rinsed five times in PBT. Prehybridization, hybridization, anti-DIG incubation and signal detection were performed as described for in situ hybridization to whole-mount embryos by Ronchi et al. (1993). In situ hybridization to embryos was also performed as described by Ronchi et al. (1993) and was followed by dehydration, rehydration and staining with anti-β-galactosidase (Cappel; 1:5000).

The pnr gene is expressed in the dorsalmost embryonic cells, in a domain of the notum surrounding the dorsal midline, and at the dorsal anterior margin of the eye disc (Heitzler et al., 1996; Ramain et al., 1993; Winick et al., 1993; see Fig. 3A,I). We chose to investigate the role of its expression in the eye disc.

To address this, we used the FLP-FRT system to generate clones of cells mutant for pnr^VX6, a null allele caused by a deletion of all but 9 amino acids of the coding region (Ramain et al., 1993; Heitzler et al., 1996; Xu and Rubin, 1993). Mutant clones were produced in the eye disc using the yeast FLP recombinase expressed under the control of the eye-specific enhancer of eyeless (ey; Quiring et al., 1994; B. Dickson, unpublished data). As expected, considering the dorsal expression of pnr, only clones produced at the dorsal margin of the eye disc gave rise to a phenotype. In such discs an ectopic field of differentiating photoreceptors appeared anterior to the main eye field (Fig. 1D). In adult flies this resulted in the formation of an ectopic eye field in the dorsal head cuticle, which could be either separate from or fused with the normal eye (Fig. 1C and data not shown). Interestingly, these ectopic eye fields did not arise exclusively from the pnr mutant cells within the clone itself, but also contained a domain of wild-type cells (marked by red pigment in Fig. 1C and by blue X-gal staining in Fig. 1D). These observations suggested that the new boundary of pnr expression present at the edge of the clone could be responsible for the induction of this new eye field. We frequently also observed a duplication of the antenna (Fig. 1C,E,F), probably reflecting the function of pnr expressed dorsally in the antennal disc (Heitzler et al., 1996).

To test the hypothesis that the boundary of pnr expression, rather than the absence of pnr, could be important for promoting eye growth, we removed all pnr function in the eye. For this purpose, we placed a Minute mutation, causing slow growth and recessive cell lethality (Morata and Ripoll, 1975), on the wild-type chromosome. In combination with eyFLP, this results in an eye composed almost entirely of homozygous mutant tissue. Again, we examined adult eyes and eye discs from flies containing very large pnr mutant clones. In some cases, we observed a dramatic loss of the eye and an absence of differentiating photoreceptors in the eye disc, resulting from the loss of all pnr function (Fig. 1E,F). In other cases eye overgrowth was observed (data not shown); probably these eye discs retained some pnr-expressing cells, allowing the establishment of an ectopic pnr expression boundary. Only a small percentage of adults with large pnr clones were recovered; most animals died as late pupae and their heads were sometimes entirely missing, probably due to loss of all tissues deriving from the eye-antennal disc. A similar phenotype has been reported for some hypomorphic combinations of pnr alleles (Heitzler et al., 1996).

In the notum, the activity of Pnr as a transcriptional activator is inhibited by binding to the zinc finger protein U-shaped (Ush), which is expressed in an adjacent domain (Cubadda et al., 1997; Haenlin et al., 1997). Ush does not appear to be required in the eye disc, as clones mutant for ush developed normally even when they were very large (data not shown). However, ectopically expressed ush was able to inhibit the function of Pnr in the eye disc; expression of ush with a pnr- GAL4 driver resulted in phenotypes similar to those induced by pnr mutant clones (data not shown). Thus Pnr is likely to act by activating the transcription of target genes in the eye as well as in the notum.

We next eliminated the pnr boundary using the opposite approach, by inducing ubiquitous pnr expression using the UAS/GAL4 system (Brand and Perrimon, 1993). We used an ey-GAL4 driver (Hazelett et al., 1998) to ectopically express pnr throughout the early eye disc. We used a form of Pnr (Pnr^D4) that is resistant to inhibition by Ush, but appears to behave like wild-type Pnr in the absence of Ush function (Haenlin et al., 1997). As a consequence, we observed a complete loss of the eye (Fig. 1G,H), confirming the importance of the border of pnr expression.

From these experiments, we conclude that the dorsally restricted expression of pnr is critical for eye development. A boundary between pnr-expressing cells and pnr-non-expressing cells appears to be necessary to induce growth and differentiation of the eye field.

Recently, several studies have established that N activation along the dorsoventral midline of the eye disc is critical for eye growth as well as for positioning the equator. This local activation is the consequence of the ventrally restricted expression of fng, which is negatively controlled by the iro-C homeobox genes expressed in the dorsal half of the eye disc (Papayannopoulos et al., 1998; Cho and Choi, 1998; Dominguez and de Celis, 1998). Either loss of fng function, or ubiquitous expression of fng, caup or mirr, abolishes eye growth. The iro-C genes appear to act redundantly, as both ara and caup must be removed from clones of cells to promote the formation of ectopic dorsal eyes similar to those we report for pnr (Cavodeassi et al., 1999; McNeill et al., 1997; F. Pichaud and F. Casares, personal communication). The similar effects we observed for gain or loss of pnr function suggested strongly that pnr might act in the same pathway as the iro-C and fng. To confirm this and to order pnr with respect to these genes, we looked at the expression of mirr and fng in eye discs mutant for pnr or misexpressing pnr. To follow mirr and fng expression, we used enhancer-trap insertions in these genes. In eye discs in which pnr function had been removed by producing mutant clones in a Minute background, we found that mirr expression was greatly reduced (Fig. 2B), whereas fng was derepressed dorsally (Fig. 2E). In eye discs expressing pnr^D4 ubiquitously from an ey-GAL4 driver, mirr expression was expanded ventrally, shifting the point of morphogenetic furrow initiation to the ventral side (Fig. 2C). The enhancer trap insertion disrupts one copy of the mirr gene, perhaps explaining the weaker effect on photoreceptor differentiation than we observed when we overexpressed pnr in the absence of mirr-lacZ (Fig. 1H). pnr overexpression had the same effect on the expression of another enhancer trap insertion, rF209, thought to reflect the expression of ara and caup (Gomez-Skarmeta et al., 1996; data not shown). fng expression was dramatically reduced in discs overexpressing pnr^D4 (Fig. 2F).

It thus appears that pnr acts upstream of the iro-C genes, activating their expression dorsally. Consistent with this, we found that ubiquitous expression of ara abolished photoreceptor differentiation, and that removal of pnr function either by making pnr mutant clones or by misexpressing ush did not restore photoreceptor formation (data not shown). If pnr were downstream of ara, blocking its function should have induced ectopic eye development even in the presence of ara. In the absence of pnr, iro-C expression is lost and fng is thus ectopically activated. We expect that this would lead to an ectopic region of N activation at the new boundary of fng expression, promoting morphogenetic furrow initiation from an ectopic position. Overexpression of pnr leads to expansion of iro-C expression and loss of fng, which should prevent N activation. As predicted, we could rescue the loss of eye development caused by ectopic pnr by co-expressing a constitutively activated form of N, N^intra (Fig. 2G).

The results above show that Pnr acts upstream of the iro-C genes to regulate dorsal eye development. Another molecule that has been shown to act upstream of the iro-C in this context is Wg (Cavodeassi et al., 1999; Heberlein et al., 1998). wg is required to inhibit the initiation of the morphogenetic furrow at the lateral margins of the eye disc, preventing ectopic eye differentiation there (Ma and Moses, 1995; Treisman and Rubin, 1995). The dorsal ectopic eyes induced by removing pnr function thus suggested that the functions of pnr and wg may be related. Consistent with this idea, the block in morphogenetic furrow initiation caused by expressing wg throughout the eye disc (Hazelett et al., 1998), like the block caused by expressing pnr^D4, could be rescued by co-expressing an activated form of N (Fig. 2H). pnr and wg may thus act in the same cascade to prevent eye differentiation.

We used in situ hybridization to show that pnr mRNA was restricted to the dorsal margin of the eye disc, anterior to and overlapping the morphogenetic furrow. Its domain of expression was broader than was previously reported (Ramain et al., 1993; Fig. 3A), though not as broad as the expression of lacZ driven by a GAL4 insertion in the pnr locus (Heitzler et al., 1996). wg is expressed at the dorsal and ventral edges of the eye disc with stronger expression dorsally (Baker, 1988), and its dorsal domain of expression resembles that of pnr (compare Fig. 3A with E).

To test the epistatic relationship between wg and pnr, we looked at pnr mRNA expression in eye discs from which we had removed wg function by producing clones of a null allele (wg^CX2) in a Minute background, as described above for pnr. Adult flies carrying such clones show a transformation of the dorsal head cuticle into ectopic eye tissue, as well as missing antennae (data not shown). Eye-antennal discs carrying such clones were identified by a severe reduction in the size of the antennal disc. In these eye discs pnr mRNA expression was wild type (Fig. 3B), showing that wg is not required for pnr expression. Overexpression of either Wg or an activated form of Armadillo (Arm), a downstream component of the Wg pathway (Peifer and Wieschaus, 1990; Zecca et al., 1996), using the ey-GAL4 driver also had no effect on pnr expression (Fig. 3C,D). Thus, wg is neither necessary nor sufficient for pnr expression.

We next tested whether pnr might be responsible for the dorsal expression of wg. Indeed, when pnr mutant clones were produced in a Minute background, dorsal wg expression was lost (Fig. 3F). Conversely, when we overexpressed pnr^D4, although the size of the eye disc was dramatically reduced, we observed a derepression of wg expression in both the eye and the antennal discs (Fig. 3G; compare wg expression in the antennal disc in Fig. 3E). We conclude that pnr indeed activates wg expression at the dorsal margin.

To confirm that the function of pnr in the eye is mediated by wg, we attempted to rescue the phenotype of pnr mutant clones by overexpressing wg. We used in combination the FLP-FRT system to produce pnr clones and the UAS/GAL4 system to ectopically express wg throughout the eye disc using ey-GAL4 as the driver (see Materials and Methods). Misexpressing wg with ey-GAL4 in a wild-type background prevents photoreceptor differentiation (Hazelett et al., 1998). If wg acts downstream of pnr, loss of pnr should not alter the effect of wg overexpression, while if wg produces its effects by misregulating pnr, removal of pnr function should restore normal photoreceptor differentiation. As predicted, even large dorsal clones of pnr mutant cells, which would be likely to produce an ectopic eye field in a wild-type background (Fig. 1D), showed no photoreceptor differentiation in discs overexpressing wg (Fig. 3H). Taken together, these results allowed us to conclude that pnr acts upstream of wg by activating its expression at the dorsal margin.

It is not known what regulates the dorsal expression of pnr in the eye disc. pnr is expressed in a dorsal stripe along the length of the embryo, where its expression seems to be established by embryonic dorsal-ventral patterning genes (Heitzler et al., 1996; Winick et al., 1993). Its later dorsal expression in the dorsal imaginal discs could simply reflect maintenance of expression in the derivatives of embryonically expressing cells. To test this, we looked for an overlap between the embryonic stripe of pnr and the eye disc primordium, which we identified by its expression of β-galactosidase driven by the ey promoter (Quiring et al., 1994). Indeed, we found that the dorsalmost cells of the eye primordium contained pnr RNA in late embryos (Fig. 3I). We have also observed pnr expression at the dorsal margin of second instar eye discs (data not shown), suggesting that its expression is continuous. Dorsal-ventral patterning of the eye disc may thus be a direct consequence of the establishment of the dorsal-ventral axis in the embryo.

Both wg, acting at the dorsal and ventral margins of the eye disc, and fng, controlling N activation at the equator, have been shown to regulate the polarity signals that direct ommatidial rotation (Cho and Choi, 1998; Dominguez and de Celis, 1998; Papayannopoulos et al., 1998; Reifegerste et al., 1997; Wehrli and Tomlinson, 1998). We therefore tested whether pnr was also involved in the control of ommatidial polarity. We analyzed ommatidial polarity in tangential sections of adult eyes containing pnr mutant clones. Consistent with the dorsal expression of pnr, ventral clones and dorsal clones lying close to the equator did not show any polarity defects (Fig. 4B,B′). Only the largest and most dorsal clones, up to eight ommatidial rows from the equator, were abnormal. Ommatidial clusters in the equatorial regions of these clones adopted a ventral polarity and chirality, and more dorsally, the formation of an ectopic equator was observed (Fig. 4C,C′,D,D′). This new equator formed within the clone rather than at its boundary, and the polarity inversion did not strictly follow the borders of the clone. On one hand, some mutant clusters near the margins of the clone showed normal dorsal polarity; on the other, wild-type ommatidia adjacent to a mutant clone sometimes showed chirality changes (Fig. 4C,C′,D,D′).

These findings, together with the observation that only large mutant clones induced a polarity reversal, suggest that the effect of pnr on polarity is mediated by non-autonomous components. Indeed, the ommatidial repolarizations seen in pnr^VX6 clones are very similar to those induced by clones mutant for downstream components of the wg signaling pathway (Wehrli and Tomlinson, 1998). The loss of wg expression caused by removing pnr would be expected to alter the polarity of ommatidia receiving insufficient levels of Wg. As Wg is a secreted protein that is thought to act by regulating the expression of a second non-autonomous signal, it is not surprising that the effects of loss of pnr are not autonomous.

We have demonstrated that pnr acts to specify the dorsal margin of the eye disc to form head cuticle rather than initiating photoreceptor development. The same effect has previously been reported for wg (Ma and Moses, 1995; Treisman and Rubin, 1995). The genes of the iro-C were originally implicated in dorsal eye development by gain-of-function studies (Cho and Choi, 1998; Dominguez and de Celis, 1998). The phenotype of removal of all three genes has recently been shown to resemble that of pnr (Cavodeassi et al., 1999; F. Pichaud and F. Casares, personal communication), although loss of the single gene mirr has a much weaker phenotype (McNeill et al., 1997). Our results show that pnr acts upstream of these components; the expression patterns of wg, the wg target mirr and the mirr target fng are altered in conditions in which pnr is removed or misexpressed, while pnr expression is unaffected by changes in Wg activity. In addition, the pnr mutant phenotype can be suppressed by expression of wg from an exogenous promoter. Finally, pnr is present in the dorsal eye disc primordium at late embryonic stages, suggesting that it reflects a very early step in dorsoventral patterning of the eye disc. The downstream response to this cascade is the activation of N at the dorsoventral midline (Fig. 5).

The role of wg in directing dorsal development is unexpected because wg is also expressed at the ventral anterior margin of the eye disc, although at a lower level than at the dorsal margin (Baker, 1988); this expression must have an upstream regulator other than pnr. However, the effects of loss of wg are more robust on the dorsal than the ventral side of the eye disc (Ma and Moses, 1995; Treisman and Rubin, 1995), and misexpression of wg symmetrically at both lateral margins dorsalizes the eye disc (Heberlein et al., 1998). These observations may be explained by the finding that at early stages wg is limited to the dorsal side of the eye disc and may exert its dorsalizing effect at this time (Cavodeassi et al., 1999). It is interesting that clones of cells mutant for components of the dpp pathway, such as Mothers against dpp (Mad), often induce an ectopic morphogenetic furrow initiating at the ventral margin adjacent to them (Burke and Basler, 1996; Hazelett et al., 1998; Wiersdorff et al., 1996). Mad mutant cells at the posterior margin have been shown to ectopically express wg (Wiersdorff et al., 1996); this ectopic wg might activate the iro-C genes to induce a new dorsoventral boundary. Indeed, ectopic furrows are very rarely induced by clones doubly mutant for both Mad and wg (J.E.T., unpublished data).

It is possible that another factor regulated by pnr cooperates with wg to regulate the iro-C genes. Since the concentration of Wg is likely to be very low near the midline of the eye disc, the sharp boundary of iro-C expression at the equator (Cavodeassi et al., 1999; Cho and Choi, 1998; Dominguez and de Celis, 1998; McNeill et al., 1997) is difficult to explain without invoking other regulators. The existence of another factor downstream of Pnr could also explain why diffusion of Wg from adjacent cells does not rescue the ectopic differentiation phenotype of pnr mutant clones. Hh may be a possible candidate for this factor (Cavodeassi et al., 1999). It is not clear whether a counterpart to pnr exists on the ventral side of the eye disc; no ventrally restricted molecules other than fng and Ser, which is a downstream target of fng (Cho and Choi, 1998; Dominguez and de Celis, 1998; Papayannopoulos et al., 1998) have yet been described. One possible candidate is the homothorax (hth) homeobox gene, which is expressed at the anterior margin but seems to function to repress photoreceptor differentiation specifically on the ventral side (Pai et al., 1998). However, the effects of loss of hth are cell-autonomous, suggesting that it does not function in dorsoventral boundary induction.

Establishment of the dorsoventral midline of the eye is linked to the opposite polarities of ommatidia on each side of the midline, in a process involving a series of diffusible signals. Wg appears to act on both the dorsal and ventral sides of the eye disc to allow ommatidia to measure their distance from the polar regions (Ma and Moses, 1995; Wehrli and Tomlinson, 1998). Because cells mutant for intracellular components of the Wg signaling pathway show non-autonomous repolarizations of adjacent tissue, it has been proposed that Wg has this effect by regulating the levels of a second signal, factor X (Wehrli and Tomlinson, 1998). Although the JAK/STAT pathway ligand Upd is a polarity-altering signal secreted from the equator, its expression is not affected by Wg and it is therefore unlikely to be factor X (Zeidler et al., 1999). Factor X may act through the Frizzled (Fz) receptor and other Wnt pathway signaling components (Wehrli and Tomlinson, 1998; Zheng et al., 1995).

pnr has the effects on polarity that we would predict from its function as a positive regulator of wg expression. Small dorsal clones of cells mutant for pnr have no effect on polarity, probably because Wg produced by adjacent wild-type cells can diffuse in to rescue the clone. This also applies to the dorsalmost regions of large pnr clones. The non-autonomous polarity inversions seen in wild-type tissue adjacent to pnr mutant clones may be caused by changes in the level of either Wg, factor X or Fng. Since pnr is only expressed at the dorsal margin of the eye disc, the border of repolarization is not related to the boundary of the pnr clone, which may extend beyond the region of Pnr function. Instead it should be related to the reduction in Wg activity, which would depend on the extent to which the pnr clone overlaps the wg domain. Although loss of pnr induces an ectopic morphogenetic furrow which is often perpendicular to the normal furrow, the midline of this ectopic eye field does not appear to act as an equator to organize the rotation of ommatidia (Chanut and Heberlein, 1995; Ma and Moses, 1995; Strutt and Mlodzik, 1995).

pnr has been shown to encode a transcription factor of the GATA family (Ramain et al., 1993), and it is therefore possible that its positive regulation of wg expression may reflect direct transcriptional activation. Pnr has been best characterized as a transcriptional activator of the ac and sc genes in the notum, where it acts through defined GATA sites in the enhancer mediating ac-sc expression in the dorsocentral proneural cluster (Garcia-Garcia et al., 1999). In the dorsalmost region of the notum, Pnr activity is inhibited by Ush, a zinc finger protein that dimerizes with Pnr to convert it from a transcriptional activator to a repressor (Cubadda et al., 1997; Garcia-Garcia et al., 1999; Haenlin et al., 1997). This mechanism is unlikely to operate in the eye because removal of ush function from very large regions of the eye disc has no effect. As misexpression of ush in the eye disc can induce pnr loss-of-function phenotypes, it is likely that ush is normally absent there rather than rendered non-functional.

wg also appears to be a target of pnr regulation in the notum, where it is transcribed in a stripe at the edge of the pnr domain (Calleja et al., 1996; Garcia-Garcia et al., 1999). This pattern seems to be a result of its transcriptional activation even by low levels of Pnr, but its repression by Pnr/Ush heterodimers within the Ush domain (Garcia-Garcia et al., 1999). The absence of ush function in the eye would explain the overlapping expression patterns we observe for pnr and wg. However, the action of Pnr on wg has not been proved to be direct in any system, and it is not known whether Pnr regulates the establishment or the maintenance of wg transcription. In the notum, the iro-C genes are expressed in a domain adjacent to the pnr domain, rather than the overlapping domain we see in the eye (Diez del Corral et al., 1999; Gomez-Skarmeta et al., 1996; Kehl et al., 1998; Leyns et al., 1996); iro-C function is also required for expression of the stripe of wg. However, in the wing pouch itself wg acts upstream of ara and caup to negatively regulate their expression at the wing margin (Gomez-Skarmeta and Modolell, 1996). Thus the regulatory relationship between these genes varies in different developmental contexts, suggesting that other factors contributing to their expression patterns remain to be identified.

We thank Nancy Bonini, Fernando Casares, Barry Dickson, Manfred Frasch, Jose-Luis Gomez-Skarmeta, Marc Haenlin, Ulrike Heberlein, Ken Irvine, Todd Laverty, Philippe Ramain, Gary Struhl, Uwe Walldorf and the Bloomington Drosophila stock center for fly stocks and reagents. We are grateful to Franck Pichaud and Fernando Casares for communicating results prior to publication. We thank the members of the Treisman laboratory for helpful discussions, and Francoise Chanut, Russ Collins, Jeff Lee and Mark Van Doren for comments on the manuscript. This work was supported by grants from the National Institutes of Health (GM56131), the National Science Foundation (IBN-9728140) and the Edward Mallinckrodt, Jr. Foundation.