Diego interacts with Prickle and Strabismus/Van Gogh to localize planar cell polarity complexes

Epithelial cells are polarized apicobasolaterally and within the plane of the epithelium. The latter is known as planar cell polarization (PCP). PCP is evident in many tissues in vertebrates and invertebrates, and is often central to the development and function of an organ(Adler, 2002; Djiane et al., 2000; Keller, 2002; Mlodzik, 2002; Wallingford et al., 2002; Wallingford et al., 2000).

In Drosophila, PCP manifests itself in many adult tissues, most notably the eye, wing, notum (dorsal thorax) and abdomen(Adler, 2002; Casal et al., 2002; Mlodzik, 2002; Uemura and Shimada, 2003). Frizzled (Fz), which is the founding member of the seven-pass transmembrane Fz-receptor family of Wnt receptors (Bhanot et al., 1996; Vinson and Adler, 1987) requires Dishevelled (Dsh), a multi-domain cytoplasmic protein, to transduce signals(Boutros and Mlodzik, 1999). Downstream of Dsh, the Wnt/Fz signaling pathway bifurcates into two distinct pathways: the canonical Wnt/β-catenin pathway and the Fz/PCP pathway(reviewed by Mlodzik, 2002; Veeman et al., 2003). In addition, a conserved group of genes is involved in PCP generation by regulating Fz/PCP signaling. These genes include the atypical cadherin flamingo [fmi; also known as stan – FlyBase(Chae et al., 1999; Usui et al., 1999)], the cytoplasmic LIM-domain protein prickle [pk(Gubb et al., 1999)], the four-pass transmembrane protein Strabismus [stbm; also known as Van Gogh (Vang – FlyBase)(Chae et al., 1999; Taylor et al., 1998; Usui et al., 1999; Wolff and Rubin, 1998)] and Diego [see below (Feiguin et al.,2001)].

In the Drosophila wing, PCP signaling leads to the emergence of an actin hair at the distal vertex of each cell (with respect to the body). Although the PCP proteins are initially distributed uniformly around the apical cortex of each wing cell, as development and PCP establishment progress, these proteins are sorted towards either the distal (e.g. Fz, Dsh),the proximal (e.g. Stbm, Pk) or both (e.g. Fmi) cell margins(Axelrod, 2001; Bastock et al., 2003; Strutt, 2001; Tree et al., 2002; Usui et al., 1999). The actin hair emerges distally as a final outcome of the PCP protein distribution(Adler, 2002).

In the fly eye, PCP is reflected in the precise arrangement of ommatidia with respect to the anteroposterior (AP) and dorsoventral (DV) axes(Fig. 1A). The AP orientation follows the directed progression of a morphogenetic furrow (MF), leaving in its wake a series of ommatidial preclusters(Wolff and Ready, 1993). The precluster that emerges from the MF includes the precursors for the photoreceptors R2, R3, R4, R5 and R8. Initially, the R3/R4 cells are symmetrically positioned within the cluster(Fig. 1A). At this stage,Fz/PCP signaling breaks the initial symmetry within the R3/R4 pair and specifies their individual fates as two distinct photoreceptors(Adler, 2002; Mlodzik, 1999; Mlodzik, 2002). Following R3 and R4 cell fate determination, preclusters begin a 90° rotation in opposite directions in each half of the eye field(Fig. 1A,B). At the end of ommatidial rotation, the symmetric photoreceptor arrangement is broken and ommatidial chirality is established by the specific arrangement of the R3 and R4 photoreceptors (Fig. 1A).

During PCP establishment and R3/R4 cell fate specification, it is thought that Fz/PCP signaling activity is graded: highest at the DV midline or the equator, and decreasing towards the poles. Although it remains unclear how a Fz/PCP activity gradient is generated, the transmembrane protein Four Jointed and the cadherins Fat and Dachsous have been implicated in this process(Ma et al., 2003; Yang et al., 2002; Zeidler et al., 1999; Rawls and Wolff, 2002). Within each precluster, Fz/PCP signaling is activated to higher levels in the R3 precursor, which is initially closer to the equator, when compared with the R4 precursor. This results in increased Delta (Dl) expression in R3, leading to activation of its ligand, Notch, in the neighboring R4 cell(Cooper and Bray, 1999; Fanto and Mlodzik, 1999; Tomlinson and Struhl, 1999; Zheng et al., 1995). Thus, the correct chirality decision and the direction of rotation are dependent on the correct R3/R4 cell fate decision.

The PCP gene diego (dgo) encodes a cytoplasmic protein with six Ankyrin repeats at its N-terminal region(Feiguin et al., 2001). The Ankyrin repeat motif has been implicated in membrane targeting of proteins(Bennett and Chen, 2001). Dgo has been shown to colocalize with and depend on Fmi for membrane recruitment. In addition, previous studies in the eye have shown that dgo mutants dominantly enhance a fmi GOF phenotype(Das et al., 2002; Feiguin et al., 2001). However, the role of Dgo in Fz/PCP signaling or its regulation has not yet been established.

We provide evidence that dgo is required to maintain the apical localization of other PCP factors. We demonstrate that Dgo is redundant to Pk and Stbm in this context, maintaining the apical localization of Fmi,following its Fz-dependent membrane recruitment. This role of Dgo, Pk and Stbm is supported by physical interactions between Pk and Stbm with Dgo. These data suggest a positive feedback loop initiated by Fz that results in the apical maintenance of all the other PCP factors (Fmi, Stbm, Pk, Dgo and Dsh).

Overexpression studies were performed using the Gal4/UAS system(Brand and Perrimon, 1993), or by direct expression of the dgo-coding sequence in a modified sev-enhancer vector. Transgenic flies were generated by standard P-element mediated transformation(Spradling and Rubin, 1982). Interaction crosses were grown at 29°C and w¹¹¹⁸ was used as control. For imaginal disc staining, the respective mutant chromosomes were established over the TM6B or SM5a:TM6B balancers. The flip-out clones of UAS-GFP-Dgo were generated using the hsFLP,actin>CD2>Gal4 strains as described(Pignoni and Zipursky, 1997)and marked with UAS-lacZ.

Mutant alleles used were dgo³⁰⁸, dgo ³⁸⁰(Feiguin et al., 2001), fmi^E59, fmi^E45(Lu et al., 1999), pk-sple¹³, stbm ^6cn, dsh¹,dsh^v26, fz^R52 and fz^K21(http://flybase.bio.indiana.edu).

Primary antibodies were: mouse anti-Fmi, rat anti-Dsh (generous gifts of T. Uemura; (Usui et al., 1999),rabbit anti-Dgo (Feiguin et al.,2001), rabbit anti-Stbm (Rawls and Wolff, 2003), rabbit anti-Pk(Tree et al., 2002), rabbit anti-Bar (gift from K. Saigo), rat anti-Sal (gift from R. Barrio) and anti-β-gal (Cappel, Promega). The β-Gal lines used were Dl-lacZ¹²⁸² (from Marc Haenlin) and svp-lacZ(Baker et al., 1990). Rhodamine phalloidin (Molecular Probes) was used to visualize actin. Secondary antibodies coupled to fluorochromes were from Jackson Laboratories. Imaginal disc staining were performed as described(Fanto et al., 2000; Feiguin et al., 2001). Discs were mounted in Mowiol and viewed with a Leica confocal microscope; images were assembled in Adobe Photoshop. Confocal images shown are single optical sections.

Tangential eye sections were prepared as described(Tomlinson et al., 1987). For genetic interaction analysis, eyes were sectioned at the equatorial region and ommatidia scored for polarity. Three to 12 sections from independent eyes were scored for each genotype.

To create pCRIITopo_Dgo and pCRIITopo_DgoAnk, Dgo cDNA was amplified with primers Dgo_upper_Not (TATGCGGCC-GCGATGCAGCATGGATCCTCC) and Dgo_AnkStop_Sal(ATAGTCGACTCATTTCTCCTTGCGATTCCG) or Dgo_lower_Sal(ATAGTCGACTCAAACTAGACTCGAGACATT), respectively, and cloned into pCRIITopo according to instructions of the manufacturer (Invitrogen). Inserts were then cloned as NotI(blunt)/SalI fragments into the NsiI(blunt)/XhoI sites of pβTH(Jenny et al., 2003) to give pβTH_Dgo and pβTH_DgoAnk. pGex4TI_DgoAnk was made by cloning the NotI(blunt)/SalI fragment of pCRIITopo_DgoAnk into the BamHI(blunt)/XhoI sites of pGex4TI. All other constructs are as described previously (Jenny et al.,2003). In vitro translations and GST pull-downs were carried out as described previously (Jenny et al.,2003). pEGFP-Dgo was made by cloning the NotI(blunt)/SalI fragment of pCRIITopo_Dgo into the BglII(blunt)/SalI sites of pEGFP_C1 (Clontech). GFP-Dgo was then transferred into SpeI/XbaI(blunt) sites of pUASP(Rorth, 1998) as NheI/SalI(blunt) fragment. Sixtyto 80-hour-old larvae grown at 25°C, were heat-shocked for 1 hour at 38°C to induce Gal4 expressing flip-out clones. After the heat shock, larvae were kept at 18°C until dissection of the larval eye and pupal wing discs. Wing discs were stained as described (Feiguin et al.,2001).

To create pCRIITopo_Dgo and pCRIITopo_DgoAnk, Dgo cDNA was amplified with primers Dgo_upper_Not (TATGCGGCCGCGAT-GCAGCATGGATCCTCC) and Dgo_AnkStop_Sal(ATAGTCGA-CTCATTTCTCCTTGCGATTCCG) or Dgo_lower_Sal(ATAGTC-GACTCAAACTAGACTCGAGACATT), respectively, and cloned into pCRIITopo according to instructions of the manufacturer (Invitrogen). Inserts were then cloned as NotI(blunt)/SalI fragments into the NsiI(blunt)/XhoI sites of pβTH(Jenny et al., 2003) to give pβTH_Dgo and pβTH_DgoAnk. pGex4TI_DgoAnk was made by cloning the NotI(blunt)/SalI fragment of pCRIITopo_DgoAnk into the BamHI(blunt)/XhoI sites of pGex4TI. All other constructs are described elsewhere (Jenny et al.,2003). In vitro translations and GST pull-downs were carried out as described previously (Jenny et al.,2003).

Loss-of-function mutant dgo eyes show typical PCP defects that include randomized chirality, and misrotated and symmetrical ommatidia. These defects are reminiscent of the phenotype of other primary PCP genes, except that in the dgo^– mutants, a higher percentage of ommatidia remain as symmetrical clusters(Fig. 1D, ∼30% symmetrical clusters in dgo^– compared with 10% in fz^–). Similarly, overexpressed Dgo in the developing 3rd instar eye disc causes a typical gain-of-function PCP phenotype (not shown). Furthermore, the analysis of the orientation of preclusters in dgo^– clones in larval eye imaginal disc (with antibodies to Spalt, marking R3/R4 cells, and Bar, marking R1/R6 cells)revealed an abnormal orientation of the photoreceptor preclusters from early developmental stages (Fig. 1E;not shown). Taken together, these data and the wing analysis (Feguin et al.,2001) indicate that dgo behaves like a primary PCP gene, revealing a characteristic role in PCP establishment.

To better define the role of Dgo in PCP establishment, we next analyzed Dgo localization in 3rd instar eye discs. Dgo is detected apically at the membrane in all cells ahead of and behind the MF(Fig. 2A; not shown). The initial uniform apical localization changes posterior to the MF in rows 2 to 3 in the 5 cell precluster (Fig. 2A). At this stage, Dgo is detected at higher levels in R3/R4 cells within the dorsoventral axis and, in particular, at the membranes between R3 and R4. Dgo is absent from membranes where the R3/R4 cells abut the R2/R5 pair. By row 5, Dgo is maintained at the border between the R3/R4 cells,whereas it is now detected at lower levels in other membrane regions of R3 and becomes enriched to higher levels in R4(Fig. 2A). The asymmetric enrichment in the R4 cell becomes resolved by row 6 and persists through row 10. In addition, Dgo is detected at high levels at the posterior side of the R2/R5 and R8 cells from row 2 onwards (Fig. 2A). In summary, these data indicate that the localization of Dgo has a pattern characteristic of other PCP proteins in the eye disc.

The PCP genes so far analyzed in detail show an asymmetric localization across the R3/R4 membranes as the result of PCP signaling. Thus, for example,although initially on both sides of the R3/R4 membrane, Fz becomes subsequently enriched to the R3 side. By contrast, Stbm becomes enriched on the R4 side (Strutt et al.,2002). This asymmetric localization is also evident in the relevant stages of PCP establishment in the wing, where Fz and Stbm are localized to the opposing sides of the proximal/distal cellular boundary(Strutt et al., 2002). To determine the precise localization of Dgo in this context, we have generated transgenic flies with a functionally active GFP-Dgo (it rescues the dgo^– phenotype, not shown), and analyzed its localization in mosaic clusters (see Materials and methods for details). This analysis revealed that GFP-Dgo is first detected on both sides of the R3/R4 membrane (Fig. 2B,C), but subsequently becomes restricted to the R3 side of the R3/R4 cell membrane(Fig. 2D,E; in addition, it is detected in R4 on the polar side, like other PCP genes). Consistent with this observation, dgo shows a genetic requirement in R3 for correct chirality establishment (J.R.K. and M.M., unpublished). Similarly, in pupal wings when the localization of the PCP factors is resolved within the proximodistal axis, GFP-Dgo is enriched on the distal side of the proximal cell, and is largely absent from the opposing membrane(Fig. 2F,F′). A schematic presentation of these localization studies is shown in Fig. 2G-I.

Taken together, these data indicate that Dgo always localizes in a manner similar to Fz and Dsh both in the eye, on the R3 side of the R3/R4 cell boundary, and in the wing on the distal side of the pupal wing cell.

A mutation in a PCP gene can affect the localization of other PCP proteins in three distinct ways in the eye: (1) asymmetric R4-like enrichment occurs but is random with respect to the R3/R4 precursor cell (reflecting chirality flips); (2) no asymmetric pattern is observed, but apical localization is maintained; and (3) apical localization is compromised.

As Dgo is predicted to be a cytoplasmic protein, a question of particular interest is how it becomes recruited to the membrane. Previous studies in the wing and eye have shown that Dgo localization is affected by fmi, but Fmi itself was not sufficient to recruit Dgo to the membrane(Das et al., 2002; Feiguin et al., 2001).

In fz-null clones, Dgo completely loses its apical membrane localization (Fig. 3A,B) and,strikingly, this is evident not only posterior to the MF(Fig. 3B), but also anterior to the MF prior to a PCP requirement (Fig. 3A). This fz requirement is unique to Dgo, as fzhas no effect or a different effect on the localization of Pk, Fmi and Stbm:the apical localization of Pk is not perturbed significantly, except for the loss of the characteristic PCP protein localization pattern, as seen in the case of Fmi in fz mutants (Fig. 3E; not shown) (Das et al.,2002). In addition, Dgo seems to be found at higher levels at the clone borders (vertical arrowheads in Fig. 3B), suggesting that a difference in Fz signaling levels promotes PCP complex formation, and this is also observed for Fmi localization(Das et al., 2002). Dgo is much less affected in mutant clones for any other PCP gene: in fmi^– clones, residual Dgo is still present at the membrane, although it largely loses its membrane association(Das et al., 2002; Feiguin et al., 2001); in pk^– and stbm^– tissue, Dgo localization is only slightly affected(Fig. 3G,H) with a short delay(approximately one row) in the asymmetric R4-like enrichment. In tissue that is dgo^–, none of the other PCP factors is affected and they localize normally to the apical cortex, resulting in the typical asymmetric, albeit randomized, R4-like enrichment(Fig. 3C,D,F)(Das et al., 2002).

In summary, these data suggest that while Fz is the key PCP protein that recruits Dgo to the apical membrane cortex, Fz localization is unaffected in dgo^– tissue, indicating that dgo acts downstream of fz for its localization. In addition, Dgo alone does not affect the apical or asymmetric localization of other PCP proteins (see below).

Previous studies have suggested that Fz and Fmi promote each other's asymmetric distribution, but that Fmi remains apically localized in fz^– tissue (Das et al., 2002; Strutt,2001). Fmi has also been shown to promote the apical and asymmetric localization of the cytoplasmic PCP proteins Dgo, Dsh and Pk(Das et al., 2002; Strutt et al., 2002). In addition, in the eye fz again affects the asymmetric localization of Stbm, Fmi and Pk, but not their apical localization (not shown)(Das et al., 2002; Strutt et al., 2002). Nevertheless, very little is known about how the apical localization of Fmi and Fz themselves is established and, importantly, maintained prior to PCP signaling in the eye or the wing.

In order to address this issue, we have analyzed the localization of Fz and Fmi in all single and in several double mutant PCP gene combinations. Single mutant pk and stbm clones show no significant defects in apical or asymmetric Fz or Fmi localization(Fig. 4A; not shown)(Bastock et al., 2003; Das et al., 2002; Strutt et al., 2002),indicating that stbm and pk alone are not crucial for apical Fmi or Fz localization. However, in dgo^–,pk^– double mutant clones (in 3rd instar eye discs anterior and posterior to the MF) the apical localization of Fmi and Fz is strongly reduced (Fig. 4B,C;not shown). A similar effect is observed in dgo^–,stbm^– double mutant clones(Fig. 4D,E; not shown),suggesting that Dgo acts in a redundant manner with Pk and Stbm for this function. In contrast to dgo^–, pk^–and dgo^–, stbm^– double mutants, pk^–, stbm^– double mutant tissue shows a different effect. Whereas the apical localization of Fmi and Fz anterior to the MF is reduced (Fig. 4F; not shown), similar to, for example, dgo^–, pk^– double mutant clones,apical Fmi and Fz localization is basically unaffected posterior to the MF(Fig. 4F,G). Although, the apical localization of Fmi was lost, Fmi protein was present in all single and double mutant backgrounds at comparable levels to wild type (determined by western blot analysis; not shown).

These data suggest that there is a degree of redundancy among the PCP genes and that certain requirements for their localization can be mediated by more than one core PCP gene. In contrast to the single mutants of either stbm,pk or dgo, or double mutants of stbm^–,pk^– (posterior to the MF), where localization of Fmi or Fz is not significantly affected, the dgo^–, stbm^– and dgo^–,pk^– double mutant backgrounds always compromise apical Fmi and Fz localization throughout the eye disc. It is likely that these effects are mediated through a feedback loop via Fmi (see below and Discussion).

To explore this observation of potential redundancy further, we analyzed Stbm and Dsh localization in dgo^–,pk^– double mutant tissue. The localization of Dsh and Stbm is not significantly affected in dgo^– mutant clones (Fig. 5A,C). In pk^– mutants, Dsh and Stbm are localized apically,although this localization appears to be slightly diffused(Fig. 5D; not shown). Similarly, apical Dsh localization is not affected in stbm^– tissue in the eye [not shown; this is different from the wing (Bastock et al.,2003; Strutt et al.,2002)]. By contrast, neither Dsh nor Stbm are detected at apical membranes in dgo^–, pk^– double mutant clones (Fig. 5B,E). As localization of Fmi (which also affects Stbm and Dsh localization; Fig. 5F) is similarly affected in the dgo^–, pk^– double mutant background, this effect could be indirectly mediated via Fmi (see above)(Das et al., 2002; Strutt et al., 2002). The requirement of Fmi for the maintenance is further supported by the fact that in fmi^–, the apical localization of Fz is significantly compromised both posterior and anterior to the furrow(Fig. 5G).

In summary, these observations raise an important question about the basis of the differences observed between the double and single mutant clones. A possible explanation for the differences in localization patterns could be that an initial complex is comprised of a number of factors, and the absence of one factor can be tolerated for some aspects of function, e.g. the maintenance of apical localization of Fmi and thus by inference also of the other PCP genes (see Discussion).

To provide biochemical evidence for a PCP multiprotein complex, we tested for physical interactions between these factors in a yeast two-hybrid matrix with available PCP proteins (A.J. and M.M., unpublished). This initial test suggested that Dgo interacts physically with Pk and Stbm (data not shown). GST pull-down assays were used to confirm these interactions independently and map the interacting domains (Fig. 6). Using in vitro translated constructs of full-length Dgo and the Ankyrin repeat region of Dgo (Fig. 6A), we showed that Dgo-Ank is sufficient to interact with a stretch of 131 amino acids close to, but not including, the very C terminus of Pk (Fig. 6A). Similarly,Dgo-Ank interacts with a Stbm domain of ∼80 amino acids in its cytoplasmic tail (Fig. 6B). Interestingly,the regions of Pk and Stbm required for the interaction with Dgo-Ank are the same regions that mediate an interaction between Pk and Stbm themselves(Jenny et al., 2003).

The physiological relevance of the physical interactions between Dgo-Pk and Dgo-Stbm is further supported by genetic interaction data (for example, a GOF stbm phenotype is enhanced by dgo gene dose)(Jenny et al., 2003), and in vivo localization studies. Whereas in single mutant pk^– or stbm^– tissue Dgo localization was not affected (Fig. 3G,H), in pk^–, stbm^– double mutant tissue Dgo was reduced at the apical cortex (Fig. 3I). Taken together, these data suggest it is likely that an initial multiprotein complex is required to maintain all PCP proteins apically and that redundancy exists between Dgo, Pk, and Stbm for this particular role (see Discussion).

We demonstrate here that Dgo maintains a potential PCP complex through the apical localization of Fmi, which in turn is required for the apical localization of the other PCP factors, including Dgo. Interestingly, this function of Dgo is redundant with Pk and Stbm, and is only apparent in double mutant clones of Dgo in combination with either gene. This observation is supported by the molecular interaction between Dgo and Pk or Stbm.

A crucial region for PCP signaling in the eye is in rows 2-5 in the 3rd instar larval disc behind the morphogenetic furrow (MF; see also Introduction). Four lines of evidence support this assumption: (1) cells that take part in PCP signaling (R3/R4) are specified as photoreceptor subtypes in this region (Wolff and Ready,1991); (2) Frizzled-Notch signaling-dependent transcription in the R4 cell is initiated in this region, as detected by the m∂0.5 reporter for the E(spl)m∂ gene(Cooper and Bray, 1999); (3)the sev-enhancer, which is active in R3/R4 cells in this region, can drive a PCP gene in order to fully rescue the respective mutant phenotype(Boutros et al., 1998; Tomlinson and Struhl, 1999);and (4) in the region ahead of the MF to the first row behind it, the PCP proteins are uniformly apically localized in all cells, before they begin at row 2 to display the characteristic PCP protein localization pattern (e.g. Fig. 2A)(Das et al., 2002; Rawls and Wolff, 2003; Strutt et al., 2002).

Following their initial symmetric apical localization, the PCP factors become asymmetrically enriched across the respective cell boundaries in the proximodistal axis in the wing or the dorsoventral axis in the eye. Although several models have been proposed as to how these complexes might be formed and maintained, the mechanism behind the early aspect of PCP establishment remains largely unclear (Strutt,2003; Tree et al.,2002). Our data suggest a complex mechanism that involves redundancy among several PCP genes (see model, Fig. 7).

Based on the analysis of single mutant clones in the eye, only Fz and Fmi affect PCP gene localization in a general non-redundant manner (and Stbm affects Pk localization). The single and double mutant clone data indicate the following.

1. Fz is required for membrane localization of Dgo(Fig. 2A,B) and this step precedes any apparent PCP signaling requirement. Fz also affects the apical localization of Dsh but not of Fmi, Pk, and Stbm significantly.

2. Dgo alone does not affect the apical localization of other PCP genes, but instead it shares this function redundantly with Stbm and Pk.

3. Pk alone does not affect the apical localization of other PCP proteins significantly, but does so in conjunction with Dgo and Stbm.

4. Fmi is responsible for the apical localization of Fz(Strutt et al., 2002)(Fig. 5G).

In addition to these initial requirements for apical localization and maintenance, the subsequent asymmetric resolution of the respective PCP proteins to the R4 cell is affected and often delayed in mutant backgrounds(this work) (Strutt et al.,2002; Tree et al.,2002).

How is the initial apical localization of all these factors maintained? As outlined above, none of the single mutant PCP genes, except fz and fmi, has a significant effect on the whole complex. However, in double mutant clones for either dgo and pk, or dgoand stbm, localization of the PCP proteins is severely affected. Most strikingly, the apical localization of Fmi and Fz is affected in these double mutant combinations (Fig. 4B-E). In addition, the localization of Stbm and Dsh are also affected (Fig. 5B,E). This could be either a direct effect of Dgo and Pk or could be mediated through their effect on Fmi [as in fmi^– tissue, Stbm and Dsh as well as Fz are reduced apically; Fig. 5F,G (Das et al.,2002)]. These data suggest that the cytoplasmic PCP proteins,which are initially recruited to the membrane by Fz (i.e. Dgo and Dsh) and Stbm (i.e. Pk), form a protein complex that is required to maintain Fmi apically (Fig. 7A). This interpretation is supported by our observation that Dgo physically interacts with Stbm and Pk, and thus possibly stabilizes the initial complex(Fig. 6). Thus, our studies reveal that Dgo, Stbm and Pk are required to maintain apical Fmi localization,possibly through the physical interactions among themselves and possibly other PCP factors, during the early stages preceding PCP signaling (i.e. anterior to MF in eye). In turn, apical Fmi promotes the maintenance of an initial PCP complex at adjacent cell membranes to facilitate their signaling specific interactions.

We can speculate on further implications of these data. During later stages of PCP signaling, the localization of the PCP factors is resolved into two types of complexes on adjacent cell membranes. The differential localization of either Fz/Dgo or the Stbm/Pk complex in the neighboring cells (R3 versus R4) suggests that asymmetric localization of PCP factors is maintained across the border of the R3 and R4 cells in the eye and across proximodistal cell borders in the wing (Tree et al.,2002). In the eye, the PCP proteins analyzed in this manner indeed localize to specific sides of the R3/R4 cell border(Fig. 2)(Strutt et al., 2002). Similarly, proximodistal localization in the wing correlates with the respective R3/R4-specific localization. For example, the localization of Fz and Diego in the distal side of a wing cell correlates with the localization on the R3 side of the R3/R4 border; conversely, Stbm localization to the proximal side of a wing cell correlates with its localization on the R4 side of the R3/R4 border (Jenny et al.,2003; Tree et al.,2002). The localization to either the R3 or R4 side also corresponds to the genetic requirements in either cell, as established in mosaic analyses (Fanto and Mlodzik,1999; Tomlinson and Struhl,1999; Wolff and Rubin,1998; Zheng et al.,1995). Thus, as Dgo, which is initially recruited by Fz, localizes to R3 and the pk/stbm complex localizes to R4, it is likely that at later stages during PCP signaling (posterior to MF) Fmi localization is maintained and stabilized through feedback loops on both sides of the R3/R4 boundary (see model in Fig. 7B).

A prediction from such a scenario is that Fz/Dgo are performing this function in R3 and the Stbm/Pk complex(Jenny et al., 2003) in R4. As Fmi is known to function as a homophilic cell-adhesion molecule(Usui et al., 1999), the removal of the feedback loop on one side could be overcome through the homophilic recruitment of Fmi from the other side. Only when both feedback loops are weakened on either side, can Fmi localization become affected(Fig. 7B). This is supported by the different effects of the respective double mutants posterior to the MF;those that affect both sides of the R3/R4 boundary, e.g. dgo and stbm (R3side/R4side) or dgo and pk (R3side/R4side)can cause Fmi delocalization, whereas double mutants affecting only one cell,e.g. pk and stbm (both R4side), have no significant effect.

We are grateful to T. Uemura, T. Wolff and J. Axelrod for generously providing us with Fmi, Stbm and Pk reagents. We thank J. Treisman, P. Adler,D. Strutt, R. Barrio, K, Basler, S. Bray, B. Dickson, E. Knust, T. Laverty, G. Rubin, K. Saigo, the Developmental Studies Hybridoma Bank and the Bloomington Stock Center for fly strains and antibodies. Microscopy was performed at the MSSM-Microscopy Shared Resource Facility, supported by a NIH-NCI shared resources grant (1 R24 CA095823-01); we thank P. Carmen and S. Henderson for their assistance at the MSSM the microscopy facility. We thank all members of the Mlodzik laboratory, and F. Cole for helpful discussions and comments on the manuscript. This research was supported by the NIH/NIGMS (grant RO1 GM62917 to M.M.).