Spalt transcription factors are required for R3/R4 specification and establishment of planar cell polarity in the Drosophila eye

Planar cell polarity (PCP), also known as tissue polarity, occurs when epithelial cells are polarized along the plane of the epithelium(perpendicular to the apical-basal axis of the cell). PCP is evident in many different biological systems, such as the coordinated orientation of bristles in invertebrates, scales in fish, or feathers in birds. Recent work has suggested that the molecular mechanisms that establish PCP are conserved throughout evolution (reviewed by Fanto and McNeill, 2004; Keller,2002; Mlodzik,2002; Veeman et al.,2003).

In the Drosophila eye, PCP is manifested by the distinct specification of the photoreceptor (PR) fate of R3 and R4, and the rotational movement performed by the developing ommatidia. In the adult eye, each ommatidium contains six outer PRs (R1-R6), which are positioned in a trapezoidal arrangement, and two inner PRs (R7 and R8), which are located in the center of this trapezoid. The trapezoidal arrangement comes in two chiral shapes (generated through the asymmetric positioning of R3 and R4) that form a mirror-image symmetry on either side of the dorsoventral (DV) midline, also called the equator (Fig. 1)(reviewed by Tomlinson,1988).

PCP is generated during the third larval instar, when the R3 and R4 precursors are specified and the developing ommatidial clusters rotate 90°in opposite directions in the dorsal and ventral halves of the eye (clockwise in the dorsal and anticlockwise in the ventral; Fig. 1). The direction of rotation and the chirality adopted by the ommatidia are a direct consequence of the specification of the R3/R4 pair. The cell of the R3/R4 precursor pair closest to the equator adopts the R3 fate, while the other cell of the pair takes the R4 fate. The specification of R3 is mediated by the activation of the Frizzled/PCP (Fz/PCP) pathway in the R3 precursor, and leads to the activation of Notch signaling in the other cell of the pair, and its specification as R4. Gain- and loss-of-function experiments with several members of Fz/PCP or Notch signaling pathways lead to the random specification of R3 and R4, or the establishment of symmetric ommatidia, where both cells acquire either the R3 or the R4 fate(Cooper and Bray, 1999; Fanto and Mlodzik, 1999; Tomlinson and Struhl, 1999; Zheng et al., 1995). Several other genes are also involved in the establishment of PCP in the Drosophila eye and in other tissues, and are generally referred to as`core PCP' genes (Fanto and McNeill,2004). Among these are dishevelled (dsh)(Theisen et al., 1994), strabismus (stbm, also known as Van Gogh)(Taylor et al., 1998; Wolff and Rubin, 1998), flamingo (fmi, also known as starry night)(Chae et al., 1999; Das et al., 2002; Usui et al., 1999), diego (dgo) (Feiguin et al., 2001) and prickle/spiny legs (pk)(Gubb et al., 1999).

The spalt (sal) gene complex encodes two related transcription factors, spalt major (salm) and spalt-related (salr), which are required for the differentiation of the inner PRs (R7 and R8)(Mollereau et al., 2001). In sal null mutant (sal^–) retinas, the morphology of the rhabdomeres (the light-sensing structure of the PR), and the expression patterns of rhodopsins in R7 and R8, change to become identical to those of the outer PRs(Mollereau et al., 2001). More recently, we found that sal is required for R7 differentiation in the third instar larva, as the expression of several R7 markers is lost in sal null mutant clones (Domingos et al., 2004). In this last study, we found that the expression of Enhancer of split mδ0.5-lacZ(mδ0.5-lacZ – a direct target of Notch signaling in R4 and R7) (Cooper and Bray,1999; Cooper and Bray,2000) is lost in sal^– clones both in R4 and R7. This result was an indication that sal could also be required for R3/R4 specification and PCP establishment.

Here, we demonstrate that sal is required for the establishment of proper ommatidial chirality. We show that the PCP defects observed in sal^– clones are due to incorrect specification of the R3/R4 cells, as several R3/R4 markers are not correctly expressed. We find that sal is required for R3/R4 specification upstream of seven-up (svp), a gene that is also required for R3/R4 specification and PCP establishment (Fanto et al., 1998; Mlodzik et al.,1990). Finally, we show that, posterior to row seven, svprepresses sal in R3/R4 in order to maintain R3/R4 identity and to inhibit the transformation of these cells to an R7 cell fate.

The following transgenic and mutant fly stocks were used: Df(2L)32FP5,FRT40A/CyO (deficiency spanning both salm and salr)(Barrio et al., 1999); FRT82B, svp^e22 (Fanto et al., 1998); svp^rA028(Hoshizaki et al., 1994); E(spl)mδ0.5 (Cooper and Bray, 1999); Dl-lacZ 1282(Fanto and Mlodzik, 1999); sev-svp (Hiromi et al.,1993); and sev-N^act(Fortini et al., 1993).

Clones of mutant eye tissue were generated by the Flp/FRT technique(Golic, 1991). Flipase expression was induced under the control of the eyeless(Newsome et al., 2000) or heat-shock promoters (larvae were heat-shocked at 37°C for 1 hour, 48 hours after egg laying).

Third instar larval eye discs were dissected in 1×PBS, fixed in 1×PBS + 4% formaldehyde for 20 minutes at room temperature, and washed 3 times with PBX (1×PBS + 0.3% Triton X-100). Primary antibodies were incubated in BNT (1×PBS, 1% BSA, 0.1% Tween 20, 250 mM NaCl) overnight at 4°C. Primary antibodies were as follows: rabbit anti-Salm(Kuhnlein et al., 1994), antiβ-gal (Cappel), rat anti-ELAV (DSHB), mouse anti-Ro (DSHB), rabbit anti-BarH1 (Higashijima et al.,1992) and mouse anti-Fmi (Usui et al., 1999). Samples were washed 3 times with PBX and incubated with appropriate secondary antibodies (Cy3, Cy5, FITC from Jackson Immuno-Research Laboratories) for 2 hours at room temperature. Samples were mounted in Vectashield (Vector Laboratories) and analyzed on a Zeiss LSM 510 confocal microscope. Tangential sections of adult eyes were performed as described (Tomlinson and Ready,1987).

In the third instar eye imaginal disc, salm is expressed in the R3/R4 precursor pair, starting in row three posterior to the morphogenetic furrow and progressively fading in these cells from row seven onwards (as determined by co-staining with mAb22C10)(Domingos et al., 2004). A direct consequence of the R3/R4 specification is the induction of mδ0.5-lacZ expression in R4(Cooper and Bray, 1999), as well as the initial 45° ommatidial rotation that occurs by row six(Strutt et al., 2002). The onset of Salm expression in R3/R4 precedes the expression of svp in R3/R4 by one row (Fig. 1E), and precedes the expression of mδ0.5-lacZ in R4 by one to two rows (Fig. 1F). Although the svp-lacZ and mδ0.5-lacZ lines may not faithfully report the expression of these genes because of the perdurance of the β-galactosidase protein, the expression of sal in R3/R4 is concomitant with the specification of R3/R4 and establishment of PCP, and is consistent with the possibility that sal is part of the molecular mechanisms that regulate these cellular processes.

In sal^– ommatidia, the rhabdomeres of both R7 and R8 acquire the typical `large' morphology of outer PRs(Mollereau et al., 2001). This leads to the disruption of the normal position of rhabdomeres in each ommatidium, thus making the evaluation of PCP defects impossible. However, on the borders of sal^– clones, it is possible to detect mosaic ommatidia with wild-type R7 and R8, allowing the analysis of PCP defects.

To investigate the role of sal in R3/R4 specification and PCP generation, we analyzed a large number of such mosaic sal^– ommatidia and scored them for PCP defects. Within 1391 mosaic ommatidia, we could identify 29 different mosaic configurations with a normal number of PRs, where it was possible to score polarity. Fig. 2A shows examples of sal^– clones, containing mosaic ommatidia, which display typical PCP defects with chirality inversions and mis-rotations. Interestingly, this analysis of mosaic clusters reveals a requirement of sal in R3 for PCP establishment. In the 16 configurations that always adopt the correct chiral form, R3 is always sal⁺ (Fig. 2B). In mosaic ommatidia adopting the wrong chiral form, the cell in the R4 position is invariably sal^–(Fig. 2C). Presumably, in such ommatidia, the sal^– precursor for R3 developed incorrectly as an R4. In ommatidia where only the R3 precursor was sal^–, we found seven cases that adopted the wrong chirality (an example is shown in Fig. 2A, top panel), and eight cases with the correct chirality (data not shown). These results demonstrate randomization of the R3/R4 chirality choice when the R3 precursor is sal^–. We also found 15 ommatidia with symmetric R4/R4 (14) or R3/R3 (1) configurations (data not shown). In each of these 15 symmetric ommatidia, at least one cell of the R3/R4 pair was sal^–. Thus, the PCP requirement of sal is similar to that of fz(Tomlinson and Struhl, 1999; Zheng et al., 1995), and demonstrates that sal is required in R3 for correct ommatidial chirality and PCP establishment.

To determine at which stage of development the PCP defect in salmutants occurred, we analyzed orientation of ommatidial rotation in sal^– clones in larval third instar eye discs. The BarH1 antibody (Higashijima et al.,1992), which labels R1 and R6, allows the visualization of the progressive rotational movement of developing ommatidia in the eye disc. In sal^– clones, 17.9% (n=212) of the ommatidia display rotational errors (Fig. 3A). However, this may be an underestimation of the number of affected ommatidia, as from the BarH1 staining it is not possible to distinguish between ommatidia with a correct or flipped chirality. Thus, we conclude that PCP defects in sal mutants are present from the early time of R3/R4 specification. The R3 and R4-specific expression of Rough(Kimmel et al., 1990) and the neuronal marker Elav are unaffected in sal^– clones(Fig. 3), indicating that neuronal specification and certain aspects of R3/R4 subtype identity do not require the sal genes.

Next, we investigated at which level of the PCP pathway sal is required. We have shown that sal is required in the R3 precursor for correct establishment of chirality (Fig. 2). According to current models, PCP signaling in the eye is a two tiered process: (1) activation of Fz/PCP signaling and upregulation of Delta (Dl) in the R3 precursor; and (2) activation of Notch signaling in the neighboring R4 cell. The atypical cadherin Flamingo (Fmi) is thought to play a dual role during the establishment of PCP(Das et al., 2002; Strutt et al., 2002). Initially, from row three to five, Fmi is asymmetrically localized in the equatorial side of R3 and R4, and promotes Fz/PCP signaling in R3. Subsequently, Fmi is enriched predominantly in R4, where it functions to downregulate Dl expression and antagonize Fz/PCP signaling(Das et al., 2002). In sal^– clones, as with fz^–or dsh^– clusters(Das et al., 2002), neither the asymmetric localization of Fmi in the equatorial side of R3/R4 nor the subsequent enrichment in R4 are observed. Instead, Fmi is present on all sides of the apical membrane cortex of R3/R4(Fig. 3B). We have also observed a defect in the asymmetric localization of Fz-GFP and Dgo in R3/R4 in sal^– clones (data not shown).

Activation of Fz/PCP signaling leads to the transient transcriptional upregulation of Dl in R3 within approximately two to three ommatidial rows, which can be observed by in situ hybridization(Parks et al., 1995), or with a Dl enhancer detector line (Fanto and Mlodzik, 1999). During this period, most sal^– ommatidia fail to upregulate Dl in R3,and both cells of the R3/R4 pair show a low level of expression(Fig. 3C). We can also find clusters where the cell in the R4 position has higher levels than R3, or where both cells in the R3/R4 pair have increased levels of Dl expression(Fig. 3C). These results suggest that sal is required for the correct interpretation of the Fz/PCP-mediated polarity signal and for the upregulation of Dlexpression in R3.

The expression of mδ0.5-lacZ, a marker of Notch signaling activation in R4 (second tier), is lost in sal^– ommatidia(Fig. 3D)(Domingos et al., 2004). Interestingly, in mosaic ommatidia, where only one cell of the R3/R4 pair is sal^–, either R3 or R4, we observe reduced expression of mδ0.5-lacZ. This indicates that sal is required for normal levels of mδ0.5-lacZ expression,both non-cell-autonomously in R3 (arrows in Fig. 3D), and cell-autonomously in R4 (arrowhead in Fig. 3D),which is surprising given the specific genetic requirement in R3 only for ommatidial polarity (Fig. 2). The non-cell autonomous requirement of sal in R3, for normal mδ0.5-lacZ expression in R4, is consistent with the deficient upregulation of Dl in R3(Fig. 3C). It is also possible that the lack of asymmetric localization of PCP proteins, as seen in the case of Fmi in sal^–(Fig. 3B), is responsible for the Notch activity modulation, as previously proposed(Strutt et al., 2002). The autonomous requirement of sal in R4 for mδ0.5-lacZ expression either could be due to a defective activation of Notch signaling, or sal may be required for transcriptional activation of E(spl)mδ in parallel to Notch signaling (see also Discussion).

To further investigate the role of sal in R3/R4 specification, we asked whether sal is required for svp expression in R3/R4. It was previously shown that svp is also required in R3 for proper R3/R4 specification and PCP establishment in the eye(Fanto et al., 1998). Our data indicate that sal genes are cell-autonomously required for the expression of svp (svp^rA28; svp-lacZ)(Hoshizaki et al., 1994) in R3 and R4 (Fig. 3F). Conversely,in svp^– (svp^e22) clones, the initiation of salm expression in R3/R4 is normal, although salm is not repressed in more posterior rows(Fig. 4A). As in sal^– clones, in svp^–clones Fmi is not properly localized in R3/R4(Fig. 4B) and the expression of mδ0.5-lacZ is lost(Fig. 4C). These results suggest that sal acts upstream of svp during R3/R4 specification.

To test this hypothesis, we attempted to rescue the expression of mδ0.5-lacZ in sal^– clones by the exogenous expression of svp in the R3/R4 pair [under the control of the sevenless (sev) promoter – sev-svp(Hiromi et al., 1993)]. Strikingly, sev-svp can induce mδ0.5-lacZexpression in at least one cell of the pair, in many cases the one in the R4 position (Fig. 4D). This result indicates that exogenous svp expression can rescue mδ0.5-lacZ expression in sal^– clones, and thus svp acts downstream of sal in this context. In addition, exogenous expression in R3/R4 of a constitutively active form of Notch (sev-N^act)(Fortini et al., 1993)upregulates mδ0.5-lacZ expression in both R3 and R4,independently of their sal genotype(Fig. 4E). Thus, although sal is normally required for mδ0.5-lacZexpression, constitutively activated Notch can overcome the salrequirement, demonstrating that Notch activation acts downstream of or in parallel to sal to regulate E(spl)mδ expression.

In svp mutants, R3, R4, R1 and R6 fail to adopt their normal fate and are transformed into the R7 fate(Mlodzik et al., 1990). The PCP defects observed in svp mutant ommatidia were attributed to a failure of the R3/R4 cells to interpret the Fz/PCP signal, because of their transformation to R7 (Fanto et al.,1998). Consistent with this, we have observed that in svp^– clones in the larval eye disc, both R3 and R4 express the R7 marker prospero (data not shown). In addition, in svp^– clones, salm is not repressed in R3/R4 by row seven, but continues to be expressed in more posterior rows(Fig. 4A) [salnormally starts to be expressed in R7 by rows seven to nine, and is both required and sufficient for R7 differentiation during larval stages(Domingos et al., 2004)]. Thus,it is likely that, in svp^– clones, the ectopic expression of salm in R3/R4 posterior to row seven is responsible for their transformation into R7.

To test this hypothesis, we have analyzed the number of large (R1-R6) and small (R7 and R8) rhabdomeres in svp^–/sal^– double mutants(Fig. 5). In svp^– clones, most ommatidia have three to five cells with small rhabdomeres because of the transformation of R3, R4, R1 and R6 into R7 (Fig. 5A,D)(Mlodzik et al., 1990). In sal^– clones, most ommatidia have eight large and no small rhabdomeres, due to the transformation of R7 and R8 to the outer PRs subtype (Fig. 5B,D)(Mollereau et al., 2001). Strikingly, svp^–/sal^–double mutant ommatidia have the same appearance as single sal^– mutant clusters(Fig. 5C,D). This result demonstrates that sal is required, downstream of svpmutation, for the transformation of R3/R4 into R7. In conclusion, salis required upstream of svp during R3/R4 specification (rows three to seven), but repression of sal by svp posterior to row seven is required to avoid the transformation of these cells into R7(Fig. 6 and Discussion).

PCP establishment in the Drosophila eye requires the correct specification of the R3/R4 pair of cells. It is thought that R3/R4 specification occurs as consequence of a higher level of Fz signaling in the equatorial cell of the R3/R4 precursor pair, which specifies the R3 fate. This leads to the upregulation of Dl in R3 and activation of Notch signaling in the polar cell of the pair, specifying it as R4(Cooper and Bray, 1999; Fanto and Mlodzik, 1999; Tomlinson and Struhl, 1999; Zheng et al., 1995).

We show that sal is also required for PCP establishment in the Drosophila eye. The analysis of sal^–mosaics reveals that sal is specifically required in R3 for establishment of ommatidial chirality (Fig. 2). The analysis of sal^– clones in the larval eye reveals that sal is required for the asymmetric localization of Fmi in the R3/R4 precursor pair(Fig. 3B) and upregulation of Dl in R3 (Fig. 3C). In a similar manner to sal, fz is required in R3 for the establishment of ommatidial chirality (Tomlinson and Struhl, 1999), and in fz mutants, unlike in stbmor dgo mutants, Fmi is not localized asymmetrically in R3/R4(Das et al., 2002). However, fz mutants also have a non-autonomous effect, disrupting PCP in ommatidia located outside the mutant clone(Zheng et al., 1995), which is not observed in sal mutants. Thus, sal is required for the correct interpretation of the fz-mediated polarity signal in R3, but not for the propagation of the polarity signal across the equatorial-polar axis. This also indicates that the expression of Fz should not be affected in sal^– clones. Finally, in sal^– clones, all PCP proteins tested (Fmi, Fz, Dgo)exhibit a defect in their asymmetric localization(Fig. 3B and data not shown),but their overall expression remains unaffected. A possible interpretation of these results is that sal transcription factors induce the expression of a yet unidentified factor, which is required for the asymmetric localization of PCP genes.

Therefore, our results suggest that sal is required upstream or in parallel to the Fz/PCP pathway for R3/R4 specification. Also, in support of this model, sal expression is not affected in R3/R4, either in gain-or loss-of-function experiments with members of the Fz/PCP and Notch signaling pathways [fz^R52 clones – data not shown; dsh¹, sev-Gal4/uas-Dsh and sev-Gal4/uas-N^icd(Cooper and Bray, 1999); sev-Fz and sev-N^*(Fanto and Mlodzik, 1999); fmi^– clones and sev-Fmi (Das et al.,2000)].

We show that sal is required cell-autonomously in R4 for normal levels of mδ0.5-lacZ expression(Fig. 3D). This requirement of sal in R4 could be due to a defect in the activation of Notch signaling (e.g. sal may be required for the expression of Notch or Su(H)). Alternatively, sal may be required for transcriptional activation of E(spl)mδ, in parallel to Notch signaling. We favor the latter possibility, as the expression of a transgenic line, where lacZ is under the regulation of 12 Suppressor of Hairless multimerized-binding sites[12Su(H)-lacZ (Go et al.,1998)], is not affected when R4 is sal^–(data not shown). The 12Su(H)-lacZ transgenic line is a reporter for Su(H)-dependent Notch signaling, and thus, sal is not required for the expression or activation of Notch, Su(H) or other components required for signaling. In addition, exogenous expression of a constitutively activated Notch (sev-N^act) can rescue mδ0.5-lacZ expression in sal^– clones (Fig. 4E). Altogether, these results suggest that sal acts in parallel to Notch signaling for the transcriptional activation of E(spl)mδ. Finally, although there is a reduction of E(spl)mδ expression when R4 is sal^–,this does not correspond to chirality defects in mature ommatidia(Fig. 2). This suggests that other genes may be redundant to sal in R4 for PCP establishment.

Several pieces of evidence demonstrate that sal is required upstream of svp for R3/R4 specification: (1) sal is required for svp expression in R3/R4 (Fig. 3F); (2) both sal and svp are required in R3 for the establishment of proper ommatidial chirality(Fig. 2)(Fanto et al., 1998); (3) in both sal and svp mutants Fmi is not asymmetrically localized in R3/R4 (Fig. 3B, Fig. 4B) and mδ0.5-lacZ expression is lost in R4(Fig. 3D, Fig. 4C); and (4) exogenous expression of svp in R3/R4 (sev-svp) can rescue the expression of mδ0.5-lacZ in sal^– clones (Fig. 4D).

In addition, we show that, posterior to row seven, svp is required to repress sal expression in R3/R4(Fig. 4A), and sal is responsible for the transformation of R3/R4 into R7 in svp mutants(Fig. 5). Based on our current and previous results, which demonstrate that sal is both necessary and sufficient for R7 differentiation posterior to row seven(Domingos et al., 2004), we propose a model for the action of sal and svp during R3/R4 specification (Fig. 6): from rows three to seven, sal is required for svp expression in R3/R4 and for R3/R4 specification; posterior to rows seven to nine, repression of sal by svp in R3/R4 is necessary for the maintenance of R3/R4 identity and the inhibition of R7 fate. This dual regulation between sal and svp helps to understand the complex sal^– phenotype in R3/R4. Strikingly, although svp expression is lost in sal^– R3/R4, these cells do not get transformed into R7, but keep an outer PR identity. Thus, in the absence of sal, the presumptive R3/R4 remain as outer PRs with an unspecified subtype identity.

In conclusion, our results demonstrate that sal is required in R3 to allow normal Fz/PCP signaling to specify the R3 and R4 cell fates. Ommatidia mutant for sal show defects that are very similar to those observed in fz and dsh mutants, as judged by the loss of asymmetric Fmi localization at the equatorial side of the R3/R4 precursors,and by the lack of Dl and E(spl)mδ upregulation within the R3/R4 pair. In addition, sal is required upstream of svpfor normal R3/R4 specification. Finally, our results show that, posterior to row seven, svp represses sal in R3/R4 in order to maintain R3/R4 identity and to inhibit transformation of these cells to the R7 cell fate.

We thank R. Schuh, R. Kühnlein, S. Artavanis-Tsakonas, U. Banerjee, K. Saigo, T. Uemura, S. Bray, the Bloomington Stock Center and the Developmental Studies Hybridoma Bank for fly strains and antibodies. We thank K. Gaengel, A. Jenny and G. Das for sharing reagents and helpful discussions. We thank members of the Steller Laboratory for helpful discussions. P.M.D. was supported by the Foundation for Science and Technology, Portugal. C.S.M. was supported by the Gulbenkian Foundation and the Foundation for Science and Technology, Portugal. This work was supported by the Strang Foundation, and by NIH grants from the National Eye Institute to B.M. (R01 EY14025) and to M.M.(RO1 EY13256). H.S. is an investigator of the Howard Hughes Medical Institute.