Interactions between Fat and Dachsous and the regulation of planar cell polarity in the Drosophila wing

The fat (ft) and dachsous (ds) genes in Drosophila encode large, cadherin-related proteins(Clark et al., 1995; Mahoney et al., 1991). These proteins differ from classic cadherins in several respects (reviewed by Tepass et al., 2000). Ft and Ds have unusually large numbers of extracellular cadherin repeats (34 and 27,respectively). Ft also contains five EGF and two laminin-A globular-like domains. Their cytoplasmic domains also differ markedly from those of classic cadherins, although there are some limited similarities to the β-catenin binding sites in classic cadherins (Clark et al., 1995). Unlike classic cadherins, Ft and Ds are concentrated apcial to but not within the adherens junctions(Ma et al., 2003). Ft and Ds are required for several processes during Drosophila development,including the regulation of growth and the proximodistal patterning of appendages (Bryant et al.,1988; Buratovich and Bryant,1997; Clark et al.,1995; Garoia et al.,2000). They also are required for the correct orientation of planar cell polarity (PCP) in the eye, wing and abdomen(Adler et al., 1998; Casal et al., 2002; Ma et al., 2003; Rawls et al., 2002; Strutt and Strutt, 2002; Yang et al., 2002).

In PCP, cells of an epithelial sheet are polarized perpendicular to the apical-basal axis (Adler, 2002; Eaton, 2003; Fanto and McNeill, 2004; Strutt, 2003; Uemura and Shimada, 2003). In the Drosophila wing, the hairs produced by each epithelial cell normally point distally (see Fig. 3A). In one view, PCP in the wing relies in part upon signals passed from cell to cell via the Frizzled (Fz) PCP pathway, using the `core'planar polarity proteins (Fig. 1E) (Tree et al.,2002). From 18-30 hours after pupariation (AP), the core planar polarity proteins are redistributed to the proximal [Prickle (Pk), Van Gogh(Vang, also known as Strabismus)], distal [Fz, Dishevelled (Dsh)], or proximal and distal [Flamingo (Fmi), also known as Starry night] faces of individual cells (Axelrod, 2001; Bastock et al., 2003; Shimada et al., 2001; Strutt, 2001; Tree et al., 2002; Usui et al., 1999). This redistribution requires mutual interactions between these proteins, both within and between cells. The interaction between cells is thought to ensure that neighboring cells have a similar polarization.

However, these cell-cell interactions must be oriented in some fashion. It was recently proposed that this orienting cue or bias is provided by Ds and Ft(Ma et al., 2003; Yang et al., 2002). Ft is uniformly expressed in the wing blade(Garoia et al., 2000; Ma et al., 2003), and thus is unlikely to provide any polarity information on its own. Ds, by contrast, is expressed at higher levels in the proximal wing, and thus it has been proposed that a proximal to distal gradient of Ds provides the polarity cue for much or all of the wing (Ma et al.,2003) (reviewed in Fanto and McNeill, 2004). Although it is difficult to detect more than low levels of uniform of Ds in the distal wing at 5 hours AP(Strutt and Strutt, 2002)(Fig. 1A-C), by 17 hours AP strong Ds expression extends into the distal wing (see Fig. 6H), forming a proximal to distal gradient down the center of the wing by 26 hours AP(Fig. 1D) (see also Ma et al., 2003). Loss of ft or ds results in PCP defects throughout most of the wing,the redistribution of Fmi and Dsh to the wrong faces of cells, and the propagation of errors in Fz signaling for longer distances(Adler et al., 1998; Ma et al., 2003; Strutt and Strutt, 2002).

According to one view, Ft and Ds act, not as adhesion molecules, but as receptor (Ft) and ligand (Ds). Because clones lacking Ft or Ds function have opposite effects on the orientation of hairs in neighboring wild-type cells,it was hypothesized that Ds inhibits Ft, creating an opposing distal to proximal gradient of Ft activity (Fig. 1E) (Ma et al.,2003), similar to what is proposed in the eye(Yang et al., 2002). Examination of protein distribution in loss-of-function clones suggests that Ft and Ds interact either directly or indirectly in the wing, which is suggestive of a receptor-ligand relationship(Ma et al., 2003; Strutt and Strutt, 2002) (see below). Although Ft activity is as yet hypothetical, the intracellular domain of Ft can bind to the Drosophila homolog of mammalian Atrophin(encoded by grunge); the similarity between ft^-and grunge^- PCP phenotypes suggests that Grunge may have a role in mediating some or all of the activity of Ft(Fanto et al., 2003).

Another mechanism for regulating Ft activity may be provided by the distally expressed Four-jointed (Fj) protein(Fig. 1B,C,E)(Brodsky and Steller, 1996; Villano and Katz, 1995; Zeidler et al., 2000). Like ds and ft mutants, null alleles of fj cause proximodistal defects in wing and leg. Although fj null homozygotes have little effect on PCP, creating artificial boundaries or gradients of fj expression can alter PCP, suggesting that it acts as a redundant cue (Zeidler et al., 1999; Zeidler et al., 2000). Clones lacking fj affect anti-Ft and anti-Ds staining in a manner consistent with a reduced interaction between Ds and Ft(Ma et al., 2003; Strutt and Strutt, 2002). Recent evidence indicates that Fj acts in the Golgi, suggesting that Fj affects the interaction between Ft and Ds by modifying the forms generated or the cellular localization of these proteins(Strutt et al., 2004).

The Ft-Ds signaling model is based on loss-of-function phenotypes, as the very large size of these molecules has made the construction of misexpression constructs difficult. We have, however, developed fully functional constructs for the misexpression of full-length Ft and Ds, and we use these to test various predictions of the model. First, Ft and Ds must bind preferentially. Although the changes in protein distribution caused by loss-of-function clones are consistent with a heterophilic interaction(Ma et al., 2003; Strutt and Strutt, 2002), this has never been directly demonstrated. We show that Ft and Ds mediate preferentially heterophilic cell adhesion in vitro and in vivo. Second,artificial gradients of ft and ds expression in the wing should be sufficient to reorient PCP throughout the wing; again, we show that this prediction is largely met.

The extensive PCP defects induced by loss of ds suggest that the gradient of ds expression acts as a global cue required for orienting PCP throughout the wing. However, it remains possible that ds acts,not as an instructive cue, but as a permissive factor, and that the crucial polarity information is being provided other cues. We have therefore tested the role of the ds gradient by driving uniform dsmisexpression. Surprisingly, we find that a gradient of ds is not necessary for correct PCP in all but the most proximal region of the wing. In most of the wing, uniform ds expression can rescue the dsmutant PCP phenotype. This is true even in the absence of any putative redundant information from Fj. Thus, other unknown cues are sufficient to orient PCP in absence of information from the pattern of dstranscription.

We have also used misexpression to test the timing of Ds activity in PCP,by analyzing whether Ds acts during the stages when the core polarity proteins are being redistributed, or whether it acts during an earlier, recently identified period of Fz activity (Strutt and Strutt, 2002). If the loss of fz is limited to this early period, distinct PCP defects are produced that are similar to those observed in ds mutants, but differ from the typical PCP defects observed in fz mutants. Because of this similarity in phenotypes, it was suggested that Ds acts during this early period(Eaton, 2003; Strutt and Strutt, 2002). Our data supports this hypothesis.

Finally, we show that Ft misexpression is sufficient to drive PCP in an orientation opposite that that caused by Ds, and that the activity of ectopic Ft depends partly, but not wholly, on the presence of Ds and Fj.

ds⁰⁵¹⁴² (Bloomington Stock Center); fj^9-II (Villano et al., 1995); fj^d1(Brodsky et al., 1996); ft^G-rv and ft^fd (Bryant et al.,1988); UAS-fj(Zeidler et al., 1999); UAS-fz-GFP and actin-fz-GFP(Strutt, 2001); AyGal4 UAS-GFP (Ito et al.,1997); tub-gal80^ts(McGuire et al., 2003); and dsh¹ tub-gal4, en-gal4, sal-gal4,ptc-gal4, dll-gal4, dagal4 and ap-gal4.

UAS-ds contains the full-length ds coding sequence,reconstructed from fragments amplified by RT-PCR (Takara). UAS-ftcontains full-length ft genomic sequence, reconstructed from DNA fragments obtained from P1 clones DS06482 and DS06843, or amplified from genomic DNA by PCR. All constructs were confirmed by sequencing and cloned into pUAST (Brand and Perrimon,1993). Detailed information is available by request. To induce expression in S2 cells, we co-transfected UAS-ft or UAS-dswith pAWGal4 (gift from Y. Hiromi), which drives Gal4 under the control of an actin promoter in S2 cells. To examine the effects of ft or ds misexpression in vivo, UAS constructs were injected into w¹¹¹⁸ embryos and transfectants were isolated using standard methods. In western blots of transfected S2 cells, we detected bands corresponding to full-length and processed Ft and Ds proteins (data not shown); each construct can substantially rescue the PCP defects of the corresponding mutant.

S2 cells were co-transfected with UAS constructs and pAWGal4. To identify transfectants UAS-GFP was co-transfected into S2 cells. Cell aggregation assays were performed as described (Usui et al., 1999; Oda et al.,1994). Cells were normally agitated on a rotator at 50 or 100 rotations per minute (rpm); 150 rpm was used for high rates of agitation.

S2 cells were fixed in 4% formaldehyde in PBS for 30 minutes at room temperature, with rhodamine-labeled phalloidin (Molecular Probes, 1:2000)added as a counterstain. Wing discs and pupal wings were fixed in 4%formaldehyde in Brower fix buffer lacking EGTA at room temperature for 40 minutes. The following primary antibodies were used: rat anti-Ds (1:20,000)(Yang et al., 2002), rabbit anti-Ds (1:100) (Strutt and Strutt,2002), rat anti-Ft (1:2000)(Yang et al., 2002), mouse anti-DSRF (Geneka, 1:1000), rabbit anti-pMad (1:2000)(Tanimoto et al., 2000) and mouse anti-βgal (Developmental Studies Hybridoma Bank, 1:1000). Fluorescent secondary antibodies were visualized using a Biorad MRC 1024 confocal microscope.

sal-gal4/tub-gal80^ts; UAS-ds/+embryos or white prepupae were collected and reared at either 20°C or 30°C. Larvae or pupae were upshifted to 30°C or downshifted to 20°C to induce or repress ds expression, respectively. For each time span we scored more than fifty adults for temperature shifts before pupariation and more than sixteen adults for temperature shifts at or after pupariation.

Almost all of the classic cadherins mediate homophilic cell adhesion(Takeichi, 1995). To test whether the same was true of Ft or Ds, we transfected ft and ds constructs into Drosophila S2 cells, which normally show little or no cell aggregation (Fig. 2A). Although transfection of Drosophila E-cadherin(DE-cadherin; also known as shotgun)(Oda et al., 1994) or the protocadherin encoded by flamingo(Usui et al., 1999) elicited cell aggregation (cell clusters of >100 or 50 cells, respectively), we were unable to elicit any detectable homophilic aggregation amongst ft-transfected cells or ds-transfected cells(Fig. 2B, and data not shown). By contrast, cells co-transfected with ft and ds aggregated,forming clusters of approximately 50 cells(Fig. 2C). The aggregation was weaker that the homophilic adhesion between cells expressing DE-cadherin, but similar to the aggregation between cells expressing flamingo, as measured by their sensitivity to the rate of agitation. At higher rates of agitation (see Materials and methods) cells expressing DE-cadherin still aggregated, whereas aggregation between cells expressing flamingo, or cells co-transfected with ft and ds, was lost. The aggregation between cells co-transfected with ft and ds was inhibited by the addition of EGTA, suggesting that, as with classic Cadherins, aggregation is dependent on Ca²⁺(Fig. 2D). Similarly, cells separately transfected with ft and ds aggregated when placed together (Fig. 2G-J).

Ft and Ds also stabilized each other in vitro and in vivo. In unaggregated S2 cells transfected with ft or ds, most of the anti-Ft and anti-Ds staining was localized in internal vesicle-like structures, and we could detect only low levels at the cell surface(Fig. 2E,F). Similarly, in clusters containing both ft-and ds-transfected cells, we detected only low levels of Ds or Ft at the interface between the ds-expressing or the ft-expressing cells(Fig. 2G,H, and data not shown). However, anti-Ft and anti-Ds staining was substantially stronger at the interface between ft-transfected and ds-transfected cells (Fig. 2G-J). In vivo,anti-Ft staining was heightened in ds-overexpressing regions in late third instar wing discs (Fig. 2K,L), and anti-Ds staining was heightened in ft-overexpressing regions (Fig. 2M,N). These effects were post-transcriptional, as we could not detect any equivalent changes in mRNA levels with in situ hybridization (data not shown). Clones overexpressing one protein also appeared to alter the distribution of the other in adjacent wild-type cells, `capping' the protein to the surface facing the clone while reducing levels on the other surfaces of the cell (Fig. 2L,N).

These changes in protein stability and distribution are consistent with our own and previously reported results from ft^-, ds^- and ds^- ft^- double mutant clones (Ma et al.,2003; Strutt and Strutt,2002) (data not shown). In the wing pouch Ds is reduced in ft^- clones, whereas Ft is more diffuse (although the overall staining is elevated) in ds^- clones. Moreover, Ft and Ds are redistributed at mutant clone boundaries in a manner consistent with the capping of proteins on the cell surface. For example, wild-type cells at the boundaries of ds^- ft^- double mutant clones lose both anti-Ft and anti-Ds staining on the surface facing the clone(Ma et al., 2003). Thus, both the in vitro and in vivo results indicate a preferentially heterophilic interaction between Ft and Ds. This type of binding is unusual for cadherin family members, and is consistent with the ligand-receptor relationship proposed by the signaling model.

To test whether Ds is sufficient to reorient PCP in the wing, we used the UAS-GAL4 system to drive patterned UAS-ds expression. We observed PCP defects if we used drivers that drove expression in the spalt(sal) pattern, which forms a gradient orthogonal to the proximodistal axis of the wing (Fig. 3B,E). Similarly, misexpressing ds in the Distal-less(Dll) pattern, which forms a distal to proximal gradient that is opposite to the endogenous ds expression pattern, partially reversed hair polarity (Fig. 3C). The defects were similar to those caused by misexpressing fz using dll-gal4 (Adler et al.,1997) or sal-gal4(Fig. 3F). PCP defects could also be induced by creating sharp boundaries of ds misexpression,using either the posterior-specific driver engrailed (en)(Fig. 4B) or patched(ptc), which is expressed at high levels in cells just anterior to the fourth longitudinal vein (data not shown). en-gal4 and ptcgal4 induced PCP defects in wild-type tissue adjacent to the region of misexpression. Similar non-autonomy was found surrounding ds mutant clones, indicating that the PCP defects are propagated to adjacent cells (Adler et al.,1998; Ma et al.,2003; Strutt and Strutt,2002). Thus, an artificial gradient or boundary of Ds misexpression can alter wing hair polarity.

From 18 to 30 hours AP, the Fz protein is redistributed to the distal face of each cell in the wing (Fig. 1E) (Strutt,2001). Some polarity mutations prevent this polarization, but ds^- and ft^- simply reorient the polarization without disrupting it (Ma et al., 2003; Strutt and Strutt,2002). Similarly, patterned ds misexpression reoriented the polarization of Fz-GFP within cells without disrupting it(Fig. 3G,H). The adult PCP defects induced by localized ds misexpression depended on the presence of intact Fz/Dsh PCP signaling, as misexpression of ds in the dsh¹ mutant background generated PCP defects resembling those of dsh¹ rather than those of dsmisexpression (data not shown). Misexpression of ds also did not generate the multiple wing hair phenotypes that normally result from misexpression of fz (Fig. 3E; compare with Fig. 3F). Thus, Ds affects the direction, rather than the presence, of polarized Fz redistribution.

The data above indicate that Ds is sufficient to reorient PCP. However,that does not test whether the endogenous gradient of Ds is required for PCP. We therefore used strong misexpression of ds to override the endogenous ds pattern and create wings with uniform dsexpression. We used either Gal4 driven from a tubulin promoter(tub-gal4), looked at the posterior of wings with posteriorly-expressed engrailed (en)-gal4, or looked at dorsal wings with dorsally-expressed apterous(ap)-gal4.

Wing PCP was largely unaffected by the presence of uniformly misexpressed ds (tub-gal4, Fig. 4C; posterior of en-gal4, Fig. 4B; dorsal of ap-gal4, data not shown). This was true even though the levels of anti-Ds staining being driven were at (en-gal4, Fig. 2K) or well above(tub-gal4, Fig. 4F;compare with Fig. 4E) the high levels observed in the proximal regions of the wild-type wing. To rule out the possibility of signaling from the endogenous Ds, we repeated these experiments in a ds⁰⁵¹⁴² mutant background, which has a strong PCP phenotype (Fig. 4A) and lacks detectable cell surface anti-Ds staining. Uniform misexpression of Ds(tub-gal4 or posterior of en-gal4) almost completely rescued the ds⁰⁵¹⁴² PCP phenotype distal to the anterior cross vein, without inducing any additional PCP defects(Fig. 4D,H, and data not shown). Thus, distal to the anterior cross vein, PCP must depend on cues other than the pattern of ds transcription.

We were unable, however, to rescue ds⁰⁵¹⁴² PCP defects in the most proximal regions of the wing(Fig. 4D,J). This is the region of the wing that normally has high Ds expression at 5 hours AP(Fig. 1A-C), and thus may require higher levels of Ds activity than could be supplied by misexpression. However, as noted above, the levels of anti-Ds driven using tub-gal4were well above normal proximal levels(Fig. 4E,F). Moreover, raising the temperature at which the larvae were reared to 30°C, which results in higher activity of the cold-sensitive Gal4, did not reduce the proximal PCP phenotype (data not shown). As there is also a sharp gradient of Ds expression at the distal boundary of this expression domain(Fig. 1A-C), we favor an alternative explanation, that the Ds gradient observed here is locally required for normal PCP (see Discussion).

What orients PCP in the more distal portions of the wing? One redundant cue suggested by previous studies is the distally expressed Fj(Fig. 1B,C). As discussed above, Fj may be required in vivo for strong interactions between Ft and Ds. Cotransfection of fj into cells co-transfected with ft and ds did not detectably enhance their aggregation or reduce their sensitivity to high rates of agitation. This did not appear to be due to endogenous Fj, as we could not detect Fj in untransfected S2 cells by western blot. However, subtle quantitative effects would likely be difficult to detect in this assay. It is possible that S2 cells do not respond to fj in a normal manner; for instance, if Fj regulates the membrane localization of Ft or Ds in epithelial cells, this may differ in the non-epithelial S2 cells, as they lack any obvious apical-basal polarity.

If Fj does modulate Ft-Ds interactions in vivo, a gradient of Fj could spatially regulate the activity of even uniformly misexpressed Ds, accounting for the failure of uniform Ds to disrupt PCP. However, our tests using a null allele of fj indicate that the contribution of fj to wing PCP is minor. ds⁰⁵¹⁴² fj^d1 double mutant wings did not show an appreciably stronger PCP defect than that observed in ds⁰⁵¹⁴² wings (Fig. 5A; compare with Fig. 4A). Moreover, uniform misexpression of ds in a fj null background caused only minor PCP defects in the proximal wing(tub-gal4, Fig. 5B;posterior of en-gal4, not shown), and PCP in most of the wing was normal. Overriding the endogenous gradients with simultaneous uniform misexpression of fj and ds had no effect on PCP(tub-gal4, Fig. 5C;posterior of en-gal4, data not shown). Similarly, the PCP defects caused by patterned ds misexpression using sal-gal4 were only slightly more extensive in a fj null background(Fig. 5D,F; compare with Fig. 3B and Fig. 5E). Thus, the wing must have some sources of polarizing information that can act in the absence of a gradient of ds or fj transcription.

Does Ds act during the polarized redistribution of the core polarity proteins, or at an earlier stage? To answer this question, we made use of the PCP defects induced by patterned ds misexpression using sal-gal4, and regulated the timing of misexpression with a temperature-sensitive version of the Gal4 inhibitor Gal80(McGuire et al., 2003).

When we initiated misexpression during larval stages we observed PCP defects. Defects were rare if induction occurred after 0 hours AP(Fig. 6A). Induction at 0 hours AP resulted in scattered ectopic anti-Ds staining at 5 hours AP at 30°C(the equivalent of 6 hours AP at 25°C; Fig. 6D), and strong misexpression at 22 hours AP at 30°C (the equivalent of 24 hours AP at 25°C; Fig. 6E). Thus, PCP was usually normal despite the misexpression of ds during the stage when the polarized redistribution of core polarity proteins occurs (18-30 hours AP at 25°C).

Conversely, if we induced misexpression early but suppressed misexpression beginning in mid-third instar, few PCP defects were induced; however, if we did not begin suppression until later larval or early pupal stages, strong PCP defects were observed (Fig. 6B). If the suppression of misexpression began at 0 hours AP,ectopic anti-Ds staining was still visible at 5 hours AP at 20°C (the equivalent of 4 hours AP at 25°C; Fig. 6F), but was undetectable at 24 hours AP at 20°C (the equivalent of 17-19 hours AP at 25°C; Fig. 6G). Thus, PCP defects were induced despite the apparently normal ds expression during the stage of Fz redistribution.

As Ft is uniformly expressed in the wing, it is not thought to play an instructive role in wing PCP. Nonetheless, we used misexpression to answer two questions raised by the Ft-Ds signaling model. First, is Ft misexpression sufficient to drive PCP in an orientation opposite that caused by Ds? Second,how does the activity of ectopic Ft depend on the presence of Ds and Fj?

UAS-ft showed substantial activity in vivo: when driven with a low-level driver (da-gal4) it rescued the pupal lethality of ft^G-rv/ft^fd heterozygotes and showed substantial rescue of the mutant wing PCP and tarsal segment defects (data not shown). Unfortunately, driving stronger expression with tubgal4 in wild-type wings resulted in larval lethality prior to the stage when PCP defects could be assessed. However, driving ft misexpression with en-gal4 (posterior) or ap-gal4 (dorsal) was not lethal, and resulted in strong PCP defects. These defects were observed not only near the boundaries of misexpression, but also within regions of apparently uniform misexpression in the posterior or dorsal wing blade(Fig. 7A,I,J for en-gal4, Fig. 7B for ap-gal4). Driving ft expression centrally using ptc-gal4 also resulted in strong PCP defects(Fig. 7C,K), as did driving ft expression in a gradient orthogonal to the proximodistal axis using sal-gal4 (Fig. 7D,L). Interestingly, all of these PCP defects were observed in a central region near the two cross-veins. This is also the region that shows the most common PCP defects in ft mutant clones(Strutt and Strutt, 2002). As expected from the Ft-Ds signaling model, which posits high Ft activity in the distal wing, driving high distal expression with dll-gal4 did not induce any PCP defects (data not shown).

The Ft-Ds signaling model hypothesizes that Ds inhibits Ft activity, and that Fj modulates that inhibition. Thus, hairs should point towards regions of high Ft or Fj activity, but away from regions of high Ds activity. In most regions of the wing this prediction was met. For example, hairs near the boundary of en-gal4-driven or ptc-gal4-driven ftmisexpression tended to point posteriorly(Fig. 7A,C,I,K), whereas anterior cells near the boundary of en-gal4-driven dsmisexpression tended to point anteriorly(Fig. 4B). However, some exceptions were also observed, where hairs appeared to point towards regions of low ft expression. Some of this may be due to propagation of PCP errors from within the domains of misexpression to boundary regions. However,it should also be remembered that Ft and Ds stabilize each other(Fig. 2), and thus regions misexpressing high levels of one protein may also have high levels of the other. The effects of fj misexpression on PCP are weak, but hairs do reorient towards region of high fj(Zeidler et al., 2000)(Fig. 7G). Because endogenous Fj may mask the effects of fj misexpression, we tried the same experiments in a fj mutant background; this increased the strength of the PCP phenotype (compare Fig. 7H,P with Fig. 7G,O).

If Ds spatially regulates Ft activity, then the PCP defects induced by uniform ft mixexpression might be reduced by a reduction in Ds levels. Indeed, in a ds⁰⁵¹⁴² background, the posterior PCP defects induced by the posterior expression of ft were weakened(Fig. 7E,M). However, they were not eliminated, despite the absence of detectable cell surface Ds in the ds⁰⁵¹⁴² background, indicating that the misexpressed Ft has substantial activity even in the absence of its putative ligand. Similarly, if Fj normally plays a role strengthening the Ds-Ft interaction,the effects of misexpressing ft should be reduced in a fj^- background. As expected, posterior PCP defects were weakened in fj^- wings(Fig. 7F,N).

Several of our gain-of-function findings are consistent with previous loss-of-function findings, and support the model that Ft-Ds signaling is sufficient to influence wing PCP. Ft and Ds preferentially bind in vitro. Patterned misexpression of ds is sufficient to alter wing PCP,consistent with its proposed role as a ligand. The effects of ft or ds misexpression on the direction of hair polarization are usually the opposite of those previously reported from ft or ds loss of function. The direction of hair polarity induced by ectopic ft or ds are usually opposite from each other, consistent with the proposal that Ds binding inhibits Ft activity. Finally, the effects of Ft misexpression are reduced in a ds mutant background, consistent with the proposed role of Ft as a receptor.

Nonetheless, our data also show that the proximal to distal gradient of ds expression is not necessary for PCP throughout the wing, despite the distal defects observed in loss-of-function ds mutants. Instead,our experiments show that uniform ds misexpression can rescue the PCP defects caused by a ds mutation in all but the most proximal portions of the wing. Thus, ds is permissive for PCP in most of the wing, and there must be another polarity cue in the distal wing that is sufficient to orient PCP in the presence of uniformly transcribed ds. Our experiments indicate that this distal cue is not provided by the distally expressed Fj protein: distal PCP is not disrupted either by uniform misexpression of both ds and fj, or by uniform misexpression of ds in a fj null mutant.

It remains possible that the distal cue functions by regulating Ft-Ds signaling. Our studies tested the PCP inputs from the patterns of dsand fj transcription, but unknown factors might post-transcriptionally regulate the forms of Ds or Ft protein produced, or their availability at the cell surface. It also is possible that Ft activity is spatially regulated by binding partners other than Ds. ft mutants have stronger PCP and disc overgrowth defects than do ds mutants, and misexpression of ft still causes PCP defects in a ds mutant lacking detectable cell surface protein.

Alternatively, the cue may be provided by a mechanism that is completely independent of Ft or Ds. One often-proposed candidate is signaling via the Drosophila Wnts, especially given their patterned (distal or marginal) expression (Baker,1988; Gieseler et al.,2001; Janson et al.,2001; Kozopas and Nusse,2002) and the involvement of the Wnt receptor Fz in PCP. Vertebrate Wnt7a also helps regulate PCP in the inner ear(Dabdoub et al., 2003). However, although the misexpression of Drosophila wnt4 can disrupt wing PCP (Lawrence et al.,2002), PCP defects have not been reported in DrosophilaWnt mutants.

Although a ds gradient is not required for PCP in most of the wing, it is possible that such a gradient is required locally in the portion of the wing near and proximal to the anterior cross vein. We were unable to rescue proximal ds mutant PCP defects with uniform Ds expression, and our data suggests that this was not simply a failure caused by insufficient Ds levels. This is the location of a strong proximal to distal boundary or gradient of endogenous Ds expression at 5 hours AP (see Fig. 1A-C). Thus, we favor the view that this sharp Ds gradient acts as a PCP cue in the proximal wing. If so, this indicates that the cues that orient PCP in the wing are not generally distributed; rather, the wing may be a patchwork of different regions that rely on different cues. This would provide a mechanism for locally altering PCP during evolution without globally affecting polarity in the wing.

The hypothesis that Ft acts as a receptor and Ds acts as a ligand for PCP is based, not only on the uniform expression pattern Ft, but also on epistasis experiments in the eye, where the PCP activity of ds clones appears to depend on the presence of ft(Yang et al., 2002). Wing PCP can also be disrupted by the expression of a truncated form of Ds lacking its intracellular domain, which is consistent with Ds acting as a ligand (H.M. and S.S.B., unpublished).

However, we have shown here that misexpressed Ft retains PCP activity in a ds mutant that eliminates detectable cell surface Ds. Thus, Ft activity is apparently not strictly dependent on patterned Ds expression. Again, this is consistent with the greater severity of ft mutant phenotypes compared with ds, and with our finding that uniform misexpression of ft but not ds can cause PCP defects. As we do not have any evidence for homophilic Ft binding, the unbound Ft molecule may have basal PCP activity. Alternatively, low-level homophilic binding or heterophilic binding to some unknown ligand may activate Ft in the absence of Ds.

It is not yet known how Ft-Ds interactions regulate the polarized redistribution of the core polarity proteins in the older pupal wing. The cytoplasmic domains of Ft and Ds contain potential regions for β-catenin binding (Clark et al., 1995),and ft and ds mutants can enhance the effects ofβ-catenin (Armadillo) misexpression(Greaves et al., 1999). However, although expression of DE-cadherin in vitro results in a detectable concentration of Armadillo at the cell membrane(Oda et al., 1994), we have not detected similar effects after expression of ft or ds(data not shown). Moreover, clones homozygous for a strong armadillomutation do not affect PCP (Axelrod et al.,1998). It has also been suggested that the cytoplasmic domain of Ft binds to and changes the activity of Grunge, the Drosophilahomolog of the Atrophin transcriptional co-repressor(Fanto et al., 2003), but it is not known whether this interaction is altered by Ft-Ds binding.

Our studies examining the timing of Ds activity suggest that its effects on the polarization of the core polarity proteins are likely to be indirect, as Ds acts before the polarized redistribution of the core polarity proteins within cells can be detected. Patterned misexpression of Ds at later stages,during the time of core protein polarization, had no effect on PCP. The period sensitive to ds misexpression is roughly congruent with the period of early Fz activity identified by Strutt and Strutt(Strutt and Strutt, 2002); if loss of Fz is limited to a period from 6 to 24 hours AP it leads to distinct, ds-like PCP defects. Thus, early Fz and Ds activity may be linked, or they may share a common target.

The only known sign of cell polarization during the stages sensitive to Ds and early Fz activity is the redistribution of the Widerborst PP2A regulatory subunit from the anterior-proximal side to the distal side of wing cells at some time between 8 and 18 hours AP(Hannus et al., 2002). Reductions in Widerborst activity can disrupt the polarized redistribution of Fmi and Dsh, suggesting an instructive role. However, Widerborst polarization is not affected by ectopic Fz expression, making it less likely that Widerborst polarization mediates early Fz activity.

A final interesting feature of our results is the preferentially heterophilic binding we observed between Ft and Ds in vitro. This result is consistent with our own and previous in vivo analyses of protein distribution within and adjacent to ft and ds mutant and overexpression clones (Ma et al., 2003; Strutt and Strutt, 2002) (this study). With the exception of the desmosomal cadherins (Chitaev et al., 1997; Syed et al., 2002), this kind of binding is unusual for cadherin-like proteins.

A number of mammalian Fat-like (Fat1, Fat2, Fat3, XP_227060) and Ds-like(Protocadherin 16, Cdh23) proteins have been identified(Bolz et al., 2001; Bork et al., 2001; Cox et al., 2000; Dunne et al., 1995; Mitsui et al., 2002; Nakajima et al., 2001; Nakayama et al., 1998; Ponassi et al., 1999). Mutations and knockouts have been examined for a few of these; however,conjectures about the bases of the mutant phenotypes have largely assumed that these proteins mediate homophilic cell adhesion(Bolz et al., 2001; Bork et al., 2001; Ciani et al., 2003; Wilson et al., 2001). It will be interesting to see whether the preferentially heterophilic interactions observed in Drosophila are preserved in similar mammalian proteins.

We thank P. Adler, P. Bryant, S. Hayashi, Y. Hiromi, B. Sanson, M. Simon,D. Strutt, T. Uemura, the Developmental Studies Hybridoma Bank, the Berkley Drosophila Genome Project and the Bloomington DrosophilaStock Center for fly stocks, DNA clones and antibodies. We also thank Dr T. Usui for advice on the cell aggregation assay, Dr J. Pawley for use of his confocal microscope, W. Feeny for help with illustrations, and Dr M. Tanaka-Matakatsu for discussions. This work was supported by NSF, the NIH and a Guyer Postdoctoral Fellowship to H.M.