Separating the adhesive and signaling functions of the Fat and Dachsous protocadherins

The Drosophila proteins Fat (Ft) and Dachsous (Ds) are required in several processes during the development of imaginal disc tissues. First, they help regulate the growth of discs. Removal of Ft and, to a lesser extent, Ds,induces a disc overgrowth phenotype, caused by both an increase in the rate of cell division and a failure to properly arrest disc growth at the end of larval development (Bryant et al.,1988; Clark et al.,1995; Garoia et al.,2000; Garoia et al.,2005; Rodriguez,2004). Second, both Ft and Ds are required for the normal establishment of planar cell polarity (PCP) in the eye, wing and abdomen(Adler et al., 1998; Casal et al., 2002; Ma et al., 2003; Matakatsu and Blair, 2004; Rawls et al., 2002; Simon, 2004; Strutt and Strutt, 2002; Yang et al., 2002). Here, Ft and Ds can reorient the subcellular polarization of the `core' planar polarity proteins that occurs during pupal stages (reviewed by Fanto and McNeil, 2004; Uemura and Shimada, 2003). Finally, Ft and Ds are required for proper proximodistal patterning of some appendages; reduced function causes shape changes and foreshortening in the blade and hinge of the wing, and the loss or fusion of tarsal segments from the leg (Bryant et al., 1988; Clark et al., 1995). Some ds alleles can affect PCP without affecting growth, and proximodistal defects can occur without affecting growth or PCP, as in homozygous flies lacking Four-jointed (Fj), a protein thought to alter Ft or Ds function(Brodsky and Steller, 1996; Ma et al., 2003; Strutt et al., 2004; Strutt and Strutt, 2002; Villano and Katz, 1995; Waddington, 1940; Zeidler et al., 2000).

As Ft and Ds are protocadherins (Clark et al., 1995; Mahoney et al.,1991), differences in cell adhesion have been invoked to explain the mutant phenotypes. When homozygous mutant clones are generated in wild-type discs, they round up and form unusually smooth boundaries with their neighbors, consistent with a change in adhesion(Adler et al., 1998; Garoia et al., 2000). The role of cell adhesion and junctional proteins in growth control in vertebrate cells is well known, and in Drosophila, failures in junctional proteins such as Discs large 1 can lead to a disc overgrowth phenotype (reviewed by Bilder, 2004; Hajra and Fearon, 2002; Johnston and Gallant, 2002). It is also easy to imagine that failures in adhesion could lead to changes in cell-cell communication, such as the signals thought to be mediated by the core PCP proteins to transmit polarity information from cell to cell (see Amonlirdviman et al., 2005), or the unknown cues that result in proper proximodistal patterning of the wing and leg.

However, several lines of evidence have suggested an alternative hypothesis, that Ft and Ds act partially or wholly as receptor and ligand,respectively, in a poorly understood signaling pathway (reviewed by Saburi and McNeill, 2005). Ft and Ds are unusual among members of the cadherin family, as they bind each other in a preferentially heterophilic fashion. Each is necessary to stabilize the other on the cell surface in imaginal discs(Ma et al., 2003; Strutt and Strutt, 2002). Overexpression of one leads to excess accumulation of the other in vivo; in vitro, Ds-expressing cells preferentially aggregate with Ft-expressing cells(Matakatsu and Blair, 2004)(this study).

Ft is expressed in a largely uniform fashion(Garoia et al., 2000; Ma et al., 2003). Ds, however,has spatially restricted domains and gradients of expression in the eye, wing,leg and abdomen, suggesting that it might provide positional cues for PCP;indeed, artificial boundaries and gradients of Ds can reorient PCP(Adler et al., 1998; Casal et al., 2002; Ma et al., 2003; Matakatsu and Blair, 2004; Rawls et al., 2002; Simon, 2004; Strutt and Strutt, 2002; Yang et al., 2002). Genetic epistasis experiments examining PCP in the eye suggest that Ft acts downstream of Ds (Yang et al., 2002). Thus, Ds may alter the activity of Ft, thereby regulating the polarity of cells along the axis of graded Ds expression(Fanto et al., 2003; Ma et al., 2003; Yang et al., 2002). By extension, Ds may similarly regulate Ft activity during growth control and proximodistal patterning (Clark et al.,1995).

Such signaling may be mediated by the intracellular domains of Ft or Ds. The extracellular domain of Ft contains 34 cadherin repeats, followed by five EGF-like and two laminin A-G domains; the extracellular domain of Ds contains 27 cadherin repeats (Fig. 1)(Clark et al., 1995; Mahoney et al., 1991)(reviewed by Tepass, 1999). Several Ft-like and Ds-like protocadherins have been identified in vertebrates and a second `Fat-like' protocadherin has been identified in Drosophila (also called Fat2), based on their similar arrangement of extracellular domains (reviewed by Tanoue and Takeichi, 2005). Intriguingly, a subset of these, the vertebrate proteins Fat4 (also called Fat-J), dachsous 1 (also called protocadherin 16) and dachsous 2, are also similar to Ft and Ds in their intracellular domains (Hong et al.,2004; Nakajima et al.,2001). Moreover, the intracellular domain of Ft can bind to the Drosophila Atrophin protein Grunge, and grunge mutants have PCP defects (Fanto et al.,2003).

But while suggestive, the evidence for signaling via Ds or Ft in PCP is not definitive. The role of Grunge is uncertain (see Discussion), and many of these results could be explained by spatially regulated changes in cell adhesion. Moreover, most of the evidence for signaling has focused on PCP, and different mechanisms might underlie growth control and proximodistal patterning.

Therefore, we have taken a structure-function approach, asking whether the various activities of Ft and Ds are mediated by their extracellular or intracellular domains. For this, we concentrated especially on two well-studied phenotypes: the strong overgrowth of ft mutant discs,and the effects of ft and ds on hair polarity in the wing and abdomen. Our results (summarized in Table 1) show that a form of Ft lacking almost the entire extracellular domain retains nearly wild-type activity in growth control and PCP, while a form of Ds that lacks the intracellular domain retained nearly wild-type activity in PCP. This is consistent with the model Ft has a receptor-like function mediated by its intracellular domain, while Ds has a ligand-like function in PCP. However, we will also present evidence suggesting that Ds can act independently of Ft in growth control, and that the intracellular domain of Ds has some activity in the proximodistal patterning of the wing. Because each assay raises separate issues, and may rely on divergent biological mechanisms, we have separated our presentation of the data below depending on the assay used.

UAS-ft (Matakatsu and Blair,2004) contains the entire coding sequence from the Ft-RA prediction (FlyBase). Nucleotide positions below are from ft-RA. For UAS-ftΔECD, a fragment from nucleotide 1 to 412 was amplified by PCR and fused with a fragment from nucleotide 13704 to the stop codon; the amino acid sequence at the fusion is QPT¹³⁷-C⁴⁵⁶⁹RGD. The deletion thus begins within the first cadherin domain and ends three amino acids N-terminal to the transmembrane domain predicted by TMHMM(Krogh et al., 2001). For UAS-ftΔICD, a fragment from nucleotide 13858 to the stop codon was replaced with HA tag sequence, yielding a C-terminal amino acid sequence of KQEKIG⁴⁶¹⁹-YPYDVPDYAstop (HA tag underlined). The deletion thus begins 11 amino acids C-terminal to the transmembrane domain, removing the regions with high homology to the intracellular domain of vertebrate Fat4.

UAS-ds (Matakatsu and Blair,2004) contains the entire region coding for GenBank Accession Numbers AAA79329 and Q24292, including 20 amino acids N-terminal of the SQA-AS signal cleavage site predicted by SignalP v.3.0(Bendtsen et al., 2004), but lacks the additional N-terminal 53 amino acids of the Ds-RA prediction(Flybase). Nucleotide positions below are from GenBank Accession Number L08811.2 and amino acid positions are from Ds-RA (FlyBase). For UAS-dsΔECD, a fragment from nucleotide 955 to 1225 was ligated with a fragment from nucleotide 10033 to the stop codon; the amino acid sequence at the fusion is RETR¹⁴³-G³⁰⁸⁰ TNL. The deletion thus begins N-terminal to the cadherin repeats and ends 17 amino acids N-terminal to the predicted transmembrane domain. For UAS-dsΔICD, a DNA fragment was amplified lacking nucleotides 10198 to the stop codon, yielding a C-terminal amino acid sequence of VKPHL³¹³⁴stop. The deletion, thus, begins 14 amino acids C-terminal to the transmembrane domain, removing the regions with high homology to the intracellular domains of vertebrate dachsous 1 and 2.

All constructs were confirmed by sequencing and cloned between the NotI (5′) and KpnI (3′) sites of pUAST. Detailed information for DNA constructs is available upon request.

To determine molecular lesions for ft^G-rv and ft^fd, genomic DNA from heterozygote adults and homozygote larvae were used as template for PCR and the resultant products were directly sequenced.

ft^G-rv and ft^fd are lethal alleles(Bryant et al., 1988). ft clones were generated in y w hs-Flp; ft^G-rv FRT^40A/y⁺FRT^40A or y w hs-Flp; ft^fdFRT^40A/ubi-GFP FRT^40A larvae. ft¹⁸ (ft^k07918) is a semi-lethal allele. ds⁰⁵¹⁴² (Bloomington Stock Center) and ds^UAO71 are strong alleles that lack detectable cell surface anti-Ds staining (Adler et al.,1998; Matakatsu and Blair,2004) (data not shown). ds^UAO71ft^G-rv has been described previously(Ma et al., 2003). ds^gal4 was isolated by mobilizing the P element in 30A-gal4, and is a strong ds allele.

Misexpression experiments used actin5C (act)-gal4,AyGal4, daughterless (da)-gal4, engrailed(en)-gal4; tubulin (tub)-gal4 and UAS-GFP (Bloomington Stock Center); spalt(sal)-gal4, hedgehog (hh)-gal4, apterous(ap)-gal4, UAS-ds and UAS-ft(Matakatsu and Blair, 2004);and the UAS stocks generated above.

Transfection of S2 cells and cell aggregation assays were as described previously (Matakatsu and Blair,2004).

Fixation, anti-Ds and anti-Ft staining and visualization in vivo were as described previously (Matakatsu and Blair,2004), with the exception that EGTA was added to the Brower fix buffer when used for nuclear antigens. We used the following additional primary antibodies: mouse anti-Engrailed 4F11 (1:10)(Patel et al., 1989), rabbit anti-Vestigial (1:500) (Williams et al.,1991), rabbit anti-Distal-less (1:200)(Panganiban et al., 1995), rat anti-DE-cadherin (1:20) (Oda et al.,1994) (Development Studies Hybridoma Bank) or goat anti-DE-cadherin (1:500) (Santa Cruz), and rat anti-HA (1:100) (Roche) or rabbit anti-HA (1:200) (Santa Cruz).

For western blots, protein was extracted from S2 cells or larval discs and CNS using sample buffer, run on 3-8% Tris-Acetate gels (Invitrogen), and transferred to PVDF membranes (Millipore) in transfer buffer (48 mM Tris, 390 mM glycine, 0.1% SDS, 10% methanol). Blots were stained with anti-Ft, anti-Ds(see above) or anti-HA (Santa Cruz), followed by appropriate HRP-linked secondary antisera (Jackson) and the Super Signal detection kit (Pierce).

We made two deletion constructs for both Ft and Ds, removing most of the extracellular (FtΔECD, DsΔECD) or intracellular (FtΔICD,DsΔICD) domains (Fig. 1;details in Materials and methods). Because the available anti-Ft antiserum was generated to the intracellular domain(Yang et al., 2002), we used a version of FtΔICD with a C-terminal HA tag.

To check the stability and subcellular localization of these deleted proteins, we expressed these in imaginal discs and S2 cells using the Gal4-UAS system (Brand and Perrimon,1993). For those that could be directly compared by antibody staining (all but FtΔICD), levels of expression were equivalent, and well above endogenous anti-Ft or anti-Ds levels in wing discs (see Fig. S1 in the supplementary material). However, the ΔICD and ΔECD proteins had different subcellular localizations. Both FtΔICD and DsΔICD were largely localized at the cell cortex(Fig. 2B,E), similar to what we observed previously with full-length Ft and Ds(Fig. 2A,D)(Matakatsu and Blair, 2004). By contrast, the FtΔECD and DsΔECD proteins were diffusely localized, except for a few vesicle-like structures, and only very low levels of protein were located at the cell cortex(Fig. 2C,F; data not shown). This is reminiscent of the mislocalization of wild-type Ft and Ds that occurs in vivo after the loss of their Ds- or Ft-binding partners(Ma et al., 2003; Strutt and Strutt, 2002), and thus may represent a failure in the stabilization of the ΔECD proteins at the cell surface by binding to endogenous Ds or Ft.

We therefore next checked whether the deleted proteins retained the ability shown by full-length Ft or Ds to stabilize endogenous cell surface Ds or Ft,respectively, after misexpression in wing discs(Fig. 2A,D)(Matakatsu and Blair, 2004). FtΔICD and DsΔICD constructs did retain this ability(Fig. 2B,E). However,misexpression of the FtΔECD or DsΔECD constructs in wing discs did not result in the stabilization of endogenous Ft or Ds, respectively(Fig. 2C,G).

To confirm that removal of the extracellular domains prevented the binding of FtΔECD to Ds and DsΔECD to Ft, we checked the ability of these constructs to mediate heterophilic cell aggregation in vitro. S2 cells co-transfected or separately transfected with ft and dsaggregate (Fig. 3B,F)(Matakatsu and Blair, 2004). Cells co-transfected with dsΔICD and ft also aggregated (Fig. 3C), as did cells separately transfected with ftΔICD and ds (Fig. 3G). Thus,the deletions of the intracellular domains of Ft or Ds did not obviously impair their ability to bind each other. By contrast, S2 cells co-transfected with dsΔECD and ft or ftΔECD and ds did not aggregate(Fig. 3D,E).

Strong ft alleles, such as the combination ft^G-rv/ft^fd, show extensive overgrowth of imaginal disc tissue at late third instar(Fig. 4B, Fig. 5B; see Fig. S2B,I in the supplementary material) and are lethal during pupal stages(Bryant et al., 1988). Both ft^G-rv and ft^fd remove detectable anti-Ft staining in discs (Ma et al.,2003) (data not shown), but the molecular defects in ft^G-rv and ft^fd had not previously been determined. We therefore sequenced both mutants. Both ft^fd and ft^G-rv contain stop codons predicted to truncate the proteins within the cadherin domains(Fig. 1). As this is prior to the transmembrane and intracellular domains, ft^G-rv and ft^fd should be null for any adhesive or receptor function;it remains possible that the truncated proteins retain some activity as ligands for other proteins.

UAS-ft can rescue both the disc overgrowth and pupal lethality of ft^G-rv/ft^fd when driven using act-gal4 (Fig. 4C) or da-gal4 (see Fig. S2C,J in the supplementary material)(Matakatsu and Blair, 2004). Similarly, expression of UAS-ft in the posterior of ft^G-rv/ft^fd discs using en-gal4 rescued overgrowth in a region-autonomous fashion; this assay has the advantage that the degree of rescue can be assessed in a single disc(Fig. 5C).

We therefore compared the abilities of the deleted Ft constructs to rescue lethality and overgrowth. Surprisingly, UAS-ftΔECD was nearly as effective as UAS-ft in rescuing ft^G-rv/ft^fd disc overgrowth when misexpressed using either act-gal4(Fig. 4D), da-gal4(see Fig. S2D,K in the supplementary material) or en-gal4(Fig. 5D), and rescued lethality with either act-gal4 or da-gal4. As shown above,FtΔECD cannot bind Ds and fails to accumulate at high levels at the cell surface, and thus should lack any adhesive function. Nonetheless, FtΔECD is sufficient for growth control at the levels of expression being driven in these experiments. This strongly suggests that the ability of Ft to control growth is mediated, not via any putative adhesive function, but through its intracellular domain.

We showed above that FtΔICD can bind to and stabilize Ds in vivo and in vitro. However, in contrast to UAS-ftΔECD,UAS-ftΔICD was unable to rescue either the disc overgrowth or pupal lethality of ft^G-rv/ft^fd when driven using act-gal4 (Fig. 4E, Fig. 5E). Thus,binding to Ds is not sufficient to confer wild-type Ft activity in the absence of the intracellular domain. This failure cannot simply be attributed to low levels of the FtΔICD protein, as FtΔICD had a dominant-negative effect on growth control, opposite to the effects of UAS-ft or UAS-ftΔECD. The overgrowth normally observed in ft^G-rv/ft^fd discs was markedly enhanced by expressing UAS-ft ΔICD with either act-gal4 or da-gal4 [compare Fig. 4B with Fig. 4E and Fig. S2B,I with Fig. S2E,L (supplementary material)]. Misexpression of UAS-ftΔICD in otherwise wild-type wing discs also caused a range of overgrowth phenotypes; the effect was moderate with hh-gal4, but very strong with the stronger ap-gal4 driver(Fig. 6B-D). Overgrowth was largely or wholly limited to the region of misexpression, as the region lacking misexpression was a nearly normal size (e.g. ventral with the dorsally expressed ap-gal4; compare Fig. 6A with 6B). This dominant-negative effect apparently depends on the existence of an intact transmembrane domain, as misexpression of a Ft construct lacking both the transmembrane and intracellular domains did not cause significant overgrowth or enhance the overgrowth observed in ft^G-rv/ft^fd discs; it also did not stabilize Ds in vivo and in vitro (H.M., unpublished).

It is unlikely that FtΔICD is acting only by interfering with the activity of endogenous Ft, as FtΔICD can enhance overgrowth in ft^G-rv/ft^fd discs. Nonetheless, such interference could contribute to the phenotype; for example, FtΔICD might titrate out some factor required for the normal stability or localization of endogenous Ft. We therefore examined the effects of FtΔICD on endogenous Ft, using an antiserum that was generated against the intracellular domain of Ft (Yang et al., 2002) and that does not cross-react with FtΔICD(Fig. 6E). We found that the levels of endogenous Ft on the cell surface were actually increased in the region of FtΔICD misexpression (Fig. 6F). As the levels of endogenous ft mRNA were unchanged by FtΔICD misexpression (data not shown), endogenous Ft protein is probably being stabilized by FtΔICD, although without providing enough Ft activity to suppress overgrowth. This stabilization is surprising given the lack of any evidence for homophilic binding in `trans' between cells expressing Ft (Stutt and Strutt, 2002; Ma et al., 2003; Matakatsu and Blair, 2004), and we cannot detect aggregation in vitro between cells expressing full-length Ft and FtΔICD (data not shown). However, it is possible that stabilization results from `cis' dimerization between the Ft and FtΔICD expressed on the surface of a single cell, as has been proposed to occur between other cadherins (e.g. Takeda et al., 1999).

As FtΔICD can bind and stabilize Ds(Fig. 2B, Fig. 3G, Fig. 6E), another explanation for the overgrowth is that FtΔICD binds to Ds in a way that prevents full Ds activity. Again, this is unlikely to be the sole mechanism, as the overgrowth phenotype induced by FtΔICD(Fig. 6B) can be much stronger than that of strong ds mutants (see below). Nonetheless, we will show below that removal of ds also enhances the overgrowth observed in ft^G-rv/ft^fd discs(Fig. 4G, see Fig. S2G,N in the supplementary material) in a manner similar to FtΔICD misexpression. Thus, FtΔICD may bind to and block the activities of both Ds and Ft (see Discussion).

Recent studies have suggested that reducing the function of Ft and Ds can lead to modulation of Wg signaling, especially in the prospective hinge region of the wing disc, and expand the inner (distal) ring of wg expression in the hinge (Cho and Irvine,2004; Rodriguez,2004). Moreover, it has recently been reported that ftoverexpression inhibits the expression Distal-less (Dll) and the `quadrant'enhancer of vestigial (vg-QE), and that expression of the vg-QE is heightened in ft^- clones(Jaiswal et al., 2006). As expression of vg and Dll is stimulated by Wg signaling(Blair, 1994; Neumann and Cohen, 1997),these results raise the possibility that gains in wg or Wg signal transduction underlie some portion of the disc overgrowth phenotype.

However, although we find that ft^fd clones occasionally lead to distortions and apparent expansions in anti-Vg and anti-Dll staining,most clones, even those with obvious overgrowth, showed no obvious change in staining (see Fig. S3A-D in the supplementary material). To test this in another way, we examined the expression of Vg and Dll in two assays. In the first, we examined ft^G-rv/ft^fd discs in which overgrowth had been rescued in the posterior using en-gal4and UAS-ft or UAS-ftΔECD(Fig. 5). In the second we compared the expression of Dll in wild-type regions with regions in which overgrowth was induced by posterior misexpression of UAS-ftΔICD, using hh-gal4(Fig. 6C). We did not observe convincing increases in the expression of Vg or Dll in regions of overgrowth(Fig. 5C,D; Fig. 6C). Posterior expression of UAS-ftΔICD also did not obviously increase the width of the distal ring of wg-lacZ expression in the hinge region when compared with adjacent anterior cells(Fig. 6D), in contrast to the expansions observed in ft mutant clones(Cho and Irvine, 2004). Thus,gains in the expression of wg or Wg targets were not reliable correlates of overgrowth.

It was also reported that Wg signaling stimulates and Ft represses the transcription of the Drosophila E-cadherin (DE-cadherin, also called Shotgun) (Jaiswal et al.,2006). We misexpressed UAS-ftΔICD with the strong, dorsal-specific ap-gal4 driver, and observed a slight decrease in anti-DE-cadherin staining in the dorsal region(Fig. 6G). This is similar to what was reported with UAS-ft(Jaiswal et al., 2006),despite the opposite effects of full length Ft and FtΔICD on growth control. Thus, overgrowth does not reliably correlate with gains in DE-cadherin.

Although weaker ds alleles do not induce obvious overgrowth phenotypes, strong ds mutations can induce mild overgrowth phenotypes(e.g. ds⁰⁵¹⁴² or ds^UAO71; Fig. 4F, see Fig. S2F,M in the supplementary material) (see also Rodriguez, 2004). This raises the possibility that Ds regulates growth partly or wholly by binding to Ft and regulating the activity of its intracellular domain. Alternatively, it may be that Ds regulates growth independently of Ft.

Unfortunately, the mildness and variability of the ds mutant overgrowth made it difficult to test these alternatives using deleted Ds constructs. Instead, we tested the epistatic relationship between dsand ft in growth control, reasoning that if Ds regulated disc growth solely by regulating the activity of the intracellular domain of Ft, then removing ds should have no additional effect in the ft^G-rv/ft^fd mutant background. As noted above, we observed a marked enhancement of the disc overgrowth phenotype in ds^UAO71 ft^G-rv/ds^UAO71ft^fd wing and eye-antennal discs(Fig. 4G, see Fig. S2G,N in the supplementary material). Thus, Ds must have some growth control activity that is not mediated by the intracellular domain of Ft.

Uniform expression of UAS-ft with act-gal4 or da-gal4 (data not shown) not only rescued the viability of ft^G-rv/ft^fd flies, but produced wings with largely normal PCP in distal regions; defects were largely limited to the region proximal and anterior to the ACV(Fig. 7D, see Fig. S4I,J in the supplementary material). Nearly identical results were obtained using act-gal4 and UAS-ftΔECD(Fig. 7E, see Fig. S4K,L in the supplementary material). The normal PCP in the distal wing probably constitutes rescue of the mutant state. Although ft^G-rv/ft^fd flies do not survive to produce adult wings, and the morphology of the mutant wings is too disrupted to assess PCP at pupal or pharate stages, homozygous ft^G-rv or ft^fd clones disrupt PCP in a central region (Ma et al.,2003; Strutt and Strutt,2002) that extends into the distal wing (distal L2-L3 intervein; Fig. 7C). The weaker viable ft¹⁸ mutation also produces PCP defects that extend distal to the PCV (Fig. 7B).

For an additional test of the rescue of ft mutant PCP defects, we turned to a different system, the abdomen of the adult fly. ft^G-rv/ft^fd abdomens have fairly normal morphology and can be examined at pharate stages; these had characteristic swirls of hairs that had lost their normally posterior polarity(Fig. 7J)(Casal et al., 2002). This phenotype was partially rescued by misexpression of UAS-ft(Fig. 7K) or UAS-ftΔECD (Fig. 7L) using act-gal4. Thus, a form of Ft that cannot bind Ds can substantially rescue PCP.

The severe overgrowth and early larval-pupal lethality induced by UAS-ftΔICD in ft^G-rv/ft^fd flies prevented us from assessing the rescuing ability of FtΔICD in either the wing or abdomen PCP assays. Therefore, as a final test of the PCP activities of deleted Ft constructs, we made use of the fact that expression of UAS-ft can disrupt PCP in wild-type wings (Matakatsu and Blair,2004). Expression of UAS-ft, UAS-ftΔECD or UAS-ftΔICD with act-gal4 all caused mild perturbation of PCP in the wing (Fig. 7F-H, see Fig. S4C-H in the supplementary material). Similar effects were observed in abdomens (data not shown). As this assay does not allow us to distinguish whether the defects were being caused by gains or losses in Ft activity, we do not know whether FtΔICD is having the same dominant-negative effects on PCP that it has on growth control.

Strong ds mutants, such as ds⁰⁵¹⁴², survive and produce adult wings with widespread, characteristic PCP defects(Fig. 8B)(Adler et al., 1998). Uniform misexpression of UAS-ds with tub-gal4 can rescue the PCP defect in all but the most proximal regions of the wing blade(Fig. 8C)(Matakatsu and Blair, 2004),and expression using ds-gal4 partially rescued PCP in the proximal wing (Fig. S5D). Nearly identical rescue of PCP was obtained using UAS-dsΔICD and tub-gal4(Fig. 8D) or ds-gal4(see Fig. S5E in the supplementary material). Thus, the extracellular domain of Ds is sufficient to drive normal PCP. By contrast, driving UAS-dsΔECD using ds-gal4 did not obviously rescue the PCP phenotype (see Fig. S5F in the supplementary material) (driving UAS-dsΔECD in ds⁰⁵¹⁴² mutants using tub-gal4 caused lethality before PCP could be assessed). These results are consistent with the model that Ds acts chiefly as a ligand for Ft in PCP.

To further compare the effects of DsΔICD and DsΔECD on PCP, we drove expression in wild type flies. We showed previously that wing hairs reorient away from regions of high ds misexpression [e.g. after misexpression in a gradient orthogonal to the proximodistal axis of the wing using sal-gal4 (Fig. 8J)] distally (using dll-gal4) or posteriorly (using en-gal4) (Matakatsu and Blair,2004). Driving misexpression of UAS-dsΔICDwith sal-gal4, dll-gal4 or en-gal4 also caused hairs to point away from the region of misexpression(Fig. 8K and data not shown). By contrast, patterned misexpression of UAS-dsΔECD did not cause any change in wing PCP (en-gal4, Fig. 8H; sal-gal4, Fig. 8L; dll-gal4, ap-gal4,tub-gal4; data not shown). Thus, the intracellular domain of Ds does not have any detectable activity in PCP in the absence of the extracellular domain, despite being driven at levels in excess of the endogenous protein(see Fig. S1E in the supplementary material). Again, this is consistent with the model that Ds acts chiefly as a ligand in this process.

Although DsΔECD showed no activity in PCP, it did show biological activity in a third phenotype common to ft and ds mutants:the alteration of proximodistal wing blade patterning. Adult wings from ds and weaker ft alleles are foreshortened along the proximodistal axis of the wing blade, as indicated by the abnormally close proximity between the anterior crossvein (ACV) and the posterior crossvein(PCV); with stronger alleles, regions of one or both crossveins are lost(Fig. 7B, Fig. 8B). We have as yet been unable to rescue the crossvein spacing defects of ft or dsmutants using full-length Ft or Ds (Fig. 7D, Fig. 8C; data not shown). In fact, overexpression of full-length Ft or Ds in wild-type flies produced a crossvein spacing defect that resembled the mutant phenotype(Fig. 7F, Fig. 8E). This suggests that the normal spacing between the crossveins requires a precise level or pattern of Ft and Ds that we were unable to reproduce. A similar situation occurs with Fj, a distally expressed protein that probably modulates the activity of Ds or Ft (Ma et al., 2003; Strutt and Strutt, 2002; Strutt et al., 2004). Both losses and gains in fj reduce crossvein spacing(Zeidler et al., 2000), and the fj crossvein spacing defect is rescued only when fjtransgene transcription is driven directly by cloned fj enhancers(Strutt et al., 2004).

Although we cannot use the crossvein spacing defect to distinguish between gains and losses in Ft or Ds function, it is an extremely sensitive, and thus useful, indicator of the perturbation of Ft or Ds function. All four of the Ft and Ds deletion constructs induced crossvein spacing defects(Fig. 7G,H, Fig. 8F-H). This included UAS-dsΔECD. Although misexpression with tub-gal4 caused a milder crossvein defect than observed with the other Ds constructs (Fig. 8G),strong defects were observed with drivers such as en-gal4(Fig. 8H).

Chief amongst our findings (summarized in Table 1) is that Ft activity is not simply a byproduct of changes in cell-cell adhesion. The FtΔECD construct lacks almost the entire extracellular domain and cannot bind or stabilize Ds in vitro or in vivo. Nonetheless, it can rescue the lethality,overgrowth and PCP defects of ft alleles that should be null for any adhesive or receptor function, and in a wild-type background can disrupt proximodistal patterning. This suggests that the intracellular domain of Ft can act in the absence of binding between endogenous Ft and Ds, or indeed between Ft and any other extracellular ligand, as long as sufficient levels are expressed.

Conversely, we found that a form of Ft lacking the intracellular domain(FtΔICD) failed to rescue overgrowth in ft mutants. In fact,this form acted as a strong dominant negative, inducing overgrowth of wild-type and ft mutant imaginal discs. This occurred despite the ability of FtΔICD to stabilize endogenous cell surface Ds and Ft,raising the possibility that FtΔICD binds to Ds and Ft is a way that blocks their activities. We cannot, however, rule out the possibility that FtΔICD alters the activity of some additional, unknown player. Although lethality prevented us from determining whether FtΔICD could rescue ft mutant PCP defects, expression of FtΔICD in wild-type wings also disrupted PCP. These PCP defects were weaker than those observed in ft mutants, suggesting that FtΔICD might have stronger effects on growth control than PCP.

In contrast to Ft, the extracellular domain of Ds was sufficient for its effects on PCP. The DsΔICD construct lacks almost the entire intracellular domain, but nonetheless can rescue the PCP defects of strong ds mutants and disrupt PCP in wild-type wings. The DsΔECD construct, however, cannot bind or stabilize Ft and cannot rescue dsmutant PCP defects or influence PCP in wild-type wings. Our results thus support the hypothesis that in PCP Ds acts chiefly as a ligand for Ft,modulating its activity.

Nonetheless, we cannot rule out the possibility that the intracellular domain of Ds has some PCP activity within the context of the whole protein,and the conservation of large regions of the Ds intracellular domain in its vertebrate homologs dachsous 1 and dachsous 2 suggests that Ds may have activity beyond that of a ligand. Thus, it is intriguing that expression of DsΔECD can disrupt another ds-sensitive phenotype, crossvein spacing in wild-type wings. As crossvein spacing defects can result from either gains or losses in Ds or Ft function, it is possible that this defect is caused by disrupting the function of endogenous Ds, and thus the ability of that Ds to signal via Ft. However, DsΔECD did not cause any obvious change in the levels of endogenous Ds (see Fig. S1 in the supplementary material). Moreover, loss of Ds normally causes visible destabilization of cell surface Ft (Ma et al.,2003; Strutt and Strutt,2002), and we did not see any changes in Ft levels in cells misexpressing DsΔECD (Fig. 2G).

ds mutations can also enhance the overgrowth observed in mutants that lack the intracellular domain of Ft, indicating that in overgrowth, Ds activity is not completely dependent on regulating the activity of the intracellular domain of Ft. In this respect, overgrowth differs from PCP, as ft mutants and ds ft double mutants produce identical PCP phenotypes (Ma et al., 2003). Our result could be explained if Ds regulates growth via its intracellular domain. Alternatively, Ds may be acting as an extracellular ligand for a binding partner other than Ft.

Our results support the hypothesis the Ft signals via its intracellular domain in growth control, PCP and proximodistal patterning. Similarly, it is likely that the intracellular domain of Ds contributes to proximodistal patterning and perhaps growth control. The conservation of long stretches of the intracellular domain of Ft and Ds in the vertebrate homologs Fat4,dachsous 1 and dachsous 2 also suggests that there is conserved binding to intracellular factors.

There are no known binding partners for the intracellular domain of Ds or dachsous-like proteins. The intracellular domain of Drosophila Ft also lacks the ENA-VASP binding sites that mediate at least some of the function of vertebrate Fat1 in vitro(Moeller et al., 2004; Tanoue and Takeichi, 2004; Tanoue and Takeichi, 2005). The intracellular domain of Drosophila Ft can bind the atrophin Grunge, and genetic evidence suggests a link between Grunge and PCP(Fanto et al., 2003). However,it is not yet clear if Grunge acts downstream of Ft, nor is it clear how atrophins, which act as transcriptional co-repressors(Erkner et al., 2002; Zhang et al., 2002; Zoltewicz et al., 2004), could polarize cells. grunge mutants also do not apparently reproduce the effects of ft mutants on disc growth(Fanto et al., 2003; Zhang et al., 2002) or on wg expression in the prospective wing hinge(Cho and Irvine, 2004).

Some evidence suggests that Ds and Ft regulate growth and patterning by altering either the expression of wg in the prospective wing hinge or the response to Wg signaling (Cho and Irvine, 2004; Rodriguez,2004; Jaiswal et al.,2006). However, our results make it unlikely that this can explain all but a small part of the overgrowth phenotype. The overgrowth induced by ft mutations or FtΔICD occurred without any consistent change in the expression of Wg target genes Dll or Vg, or in the expression of wg. Moreover, FtΔICD induced overgrowth in the entire wing disc, but whereas increased Wg signaling can induce overgrowth in the hinge(Neumann and Cohen, 1996), in the prospective wing blade Wg signaling reduces growth(Johnston and Sanders, 2003). Our results are consistent with the failure of mutants in the Wg signaling pathway to modify the ft overgrowth phenotype(Garoia et al., 2000; Garoia et al., 2005; Resino and Garcia-Bellido,2004).

A recent study has suggested a possible link between overgrowth and Ras signaling; mild reductions in Ras function that have little effect on the growth of wild-type cells can block the overgrowth observed in ftmutant clones (Garoia et al.,2005). It remains to be seen whether Ft can actually affect Ras signaling, or whether this represents the convergence of the two pathways on a shared target.

Because Ds is expressed in an apparently graded fashion along the axes of polarity, it was suggested that Ds provides a global cue that orients PCP in the eye, wing and abdomen (Casal et al.,2002; Ma et al.,2003; Yang et al.,2002). But whereas patterned Ds misexpression is sufficient to reorient PCP, and patterned Ds expression does appear to be necessary for normal PCP in the eye, in the wing uniform Ds expression is able to rescue most of the ds mutant PCP defects(Matakatsu and Blair, 2004; Simon, 2004). This suggests that most of the PCP defects in ds mutant wings are caused, not by a change in the spatial regulation of Ds-Ft signaling, but rather by the loss of a basal level of signaling required for the proper activity of some other polarizing cue. These results left open the possibility that Ft activity is being spatially regulated by an extracellular ligand other than Ds. However,we show here that ft mutant PCP defects can be substantially rescued by uniform expression of FtΔECD, a form of Ft that cannot bind Ds, or probably any other ligand.

There is, however, a region in the proximal wing where we were unable to rescue PCP defects with uniform expression of either Ds(Matakatsu and Blair, 2004),Ft, or FtΔECD. This is also the region of the wing where there is a boundary or sharp gradient between proximal regions with high and distal regions with low ds expression(Matakatsu and Blair, 2004)(see Fig. S5A,B in the supplementary material). Thus, it remains possible that Ds and Ft activities are permissive in much of the wing but, in the proximal wing, spatially instructive. The different sensitivities of different regions to changes in Ds and Ft may reflect localized differences in the strength of other partially redundant polarizing cues.

We thank Drs P. Adler, J. Axelrod, P. Bryant, S. Carroll, S. Hayashi, Y. Hiromi, H. McNeill, M. Simon, D. Strutt and T. Uemura; the Developmental Studies Hybridoma Bank and the Bloomington Drosophila Stock center for fly stocks, DNA and antibodies; Dr J. Pawley for use of his confocal microscope; and Dr M. Tanaka-Matakatsu for discussions. This work was supported by grants from NIH (R01-NS028202) and NSF (IBN-0416586).