Regulation of long-range planar cell polarity by Fat-Dachsous signaling

Development and function of many tissues requires coordination of cellular characteristics along a consistent direction. This coordinate organization is called planar cell polarity (PCP). Most forms of PCP are regulated by two well-conserved pathways: the Frizzled(Fz)/PCP and the Fat (Ft)-Dachsous (Ds) PCP pathway (reviewed by Axelrod, 2009; Goodrich and Strutt, 2011; Lawrence et al., 2007; Maung and Jenny, 2011; Thomas and Strutt, 2012). Ft and Ds are large cadherins (Bryant et al., 1988; Clark et al., 1995; Mahoney et al., 1991), the binding of which is regulated by the kinase Four-jointed (Fj) (Brittle et al., 2010; Ishikawa et al., 2008; Simon et al., 2010). Ds and Fj are expressed in gradients in the eye, regulating Ft activity (Simon, 2004; Yang et al., 2002) (Fig. 1B). Ft, Ds and Fj are also involved in the growth-regulatory Hippo pathway (reviewed by Grusche et al., 2010; Halder and Johnson, 2011; Lawrence et al., 2008; Staley and Irvine, 2012). A crucial question is how the orientation of neighboring cells is established and propagated across a tissue. Analysis of clones of cells lacking ft, ds or fj reveals that mutant tissue inside clones and wild-type tissue adjacent to clones has altered polarity (Casal et al., 2002; Casal et al., 2006; Fanto et al., 2003; Ma et al., 2003; Rawls et al., 2002; Strutt and Strutt, 2002; Yang et al., 2002). The reorganization of polarity in wild-type tissue, called domineering non-autonomy (Krasnow et al., 1995; Vinson and Adler, 1987), can propagate over many cell lengths, indicating that the Ft-Ds pathway has a role in communicating polarity information across a tissue.

Recent studies have revealed that Ft, Ds and the atypical myosin Dachs (D) are asymmetrically distributed (Ambegaonkar et al., 2012; Bosveld et al., 2012; Brittle et al., 2012; Mao et al., 2006; Rogulja et al., 2008). D localization redistributes several cells away from a border of Ds or Fj expression (Ambegaonkar et al., 2012; Bosveld et al., 2012; Brittle et al., 2012; Mao et al., 2006; Mao et al., 2011). These data, along with previous studies, have led to a model in which polarity information is propagated from cell to cell by polarized distribution of Ft and Ds, which in turn reorganizes Ft, Ds and D in adjacent cells (Casal et al., 2006; Ma et al., 2003; Matakatsu and Blair, 2006; Strutt and Strutt, 2002; Yang et al., 2002).

An alternative model to explain the propagation of polarity from Ft- and Ds-expressing cells is that Ft-Ds binding leads to production of a second signal, which is responsible for the propagation of polarity. The finding that the cytoplasmic domain of Ft binds to the transcriptional co-repressor Atrophin (Atro) led to the proposal that this second signal is transcriptionally regulated by Ft activity (Fanto et al., 2003). Loss of either ft or atro leads to increased expression of fj, providing a potential mechanism for regulation of PCP.

We reasoned that quantitative analysis of polarity disruptions caused by mutant clones of ft, ds and atro could provide insight into the mechanisms underlying the propagation of polarity, and allow us to distinguish between these models. Using these analyses, we show that asymmetric subcellular localization of Ft and Ds cannot account for the propagation of polarity in the Drosophila eye. We further show that the cytoplasmic domains of Ft and Ds can generate a signal that can propagate across a tissue independently of changes in Ft, Ds or D localization. We demonstrate that atro regulates polarity only in regions near the equator, indicating that other, as yet unknown polarity regulators mediate Ft signaling in polar regions. We show that loss of D does not block the propagation of polarity defects, and that increased expression of fj cannot account for polarity disruptions caused by loss of ft or atro. These data support a model in which long-range polarity propagation in the Ft-Ds pathway is mediated by a second signal regulated by Ft-Ds interactions.

Clones were generated using FLP/FRT with ey or hs drivers (Xu and Rubin, 1993) and MARCM (Lee and Luo, 2001). Adult clones were marked by loss of w⁺ or expression of UAS-W-RNAi. Adult eyes were sectioned according to Wolff (Wolff, 2000). Antibody staining of eye discs was carried out after 4% PFA fixation and blocking with 1% BSA. Antibodies used were rat α-Ft (1:1000), rabbit α-Ds (1:1000), rat α-Bar (1:100), mouse α-Elav (DSHB, 1:1000) and mouse α-β-gal (Promega, 1:1000).

The following stocks were used: ft^G-rv FRT40. ft^fd FRT40. ds^38k FRT40 (from David Strutt, University of Sheffield, UK); ds^38k ft^G-rv FRT40. ds^UAO71 ft^G-rv FRT40. atro³⁵ FRT80. atro^e46 FRT80 (Manolis Fanto, King’s College London, UK). GFP FRT40. GFP FRT80. GFP FRT40;UAS-FtΔECD (Seth Blair, University of Wisconsin, Madison, WI, USA); GFP FRT40;UAS-DsΔECD (Seth Blair); GFP FRT40;UAS-fz-RNAi (TRiP, Boston, MA, USA); GFP FRT40;UAS-ds-RNAi (TRiP); GFP FRT40;UAS-atro-RNAi (C.-C. Tsai, Rutgers University, Piscataway, NJ, USA); d^GC13 FRT40 [Ken Irvine (Rutgers University, Piscataway, NJ, USA) and David Strutt]; ft^G-rv d^GC13 FRT40 (Ken Irvine and David Strutt); fat2^58D FRT80 (Christian Dahmann, MPI-CBG, Dresden, Germany); ft^G-rv FRT40;UAS-FtΔECD; ft^G-rv FRT40;UAS-fz-RNAi; ds^38k FRT40;UAS-DsΔECD; ft^G-rv FRT40;UAS-DsΔECD; ds^38k ft^G-rv FRT40;UAS-DsΔECD; ds^38k FRT40;UAS-fz-RNAi; ds^38k FRT40;UAS-atro-RNAi; ft^G-rv FRT40 fj^D1; ds^38k FRT40 fj^D1; ds^38k ft^G-rv FRT40 fj^D1; fj^D1;atro³⁵ FRT80; GFP FRT40 fj^D1; fj^D1;GFP FRT80; ft^G-rv FRT40;fat2^58D FRT80; ds^38k FRT40;fat2^58D FRT80; GFP FRT40;fat2^58D FRT80; ser-lacZ.II-9.5 (Maria Dominguez, Universidad Miguel Hernández - CSIC, Alicante, Spain); and ft^G-rv FRT40;ser-lacZ. ds^38k FRT40;ser-lacZ. ds^38k ft^G-rv FRT40;ser-lacZ.

As ommatidial polarity determination occurs through photoreceptor R3/4 fate determination (Cooper and Bray, 1999; Fanto and Mlodzik, 1999; Fanto et al., 2003; Yang et al., 2002; Zheng et al., 1995), and it has been shown that R3/R4 mosaicism can affect ommatidial polarity, we excluded ommatidia that were mosaic for R3/R4 from our analysis of long-range polarity. We considered ommatidia in which R3 and R4 cells were both wild type as wild type for the purpose of long-range polarity establishment, and those in which both were mutant as being genetically mutant, unless otherwise specified. We tested this approach by measuring polarity effects in ommatidia in which all outer photoreceptors were wild type, surrounding ft, ds, ds ft and atro mutant clones. These showed similar trends as R3⁺/R4⁺ ommatidia (supplementary material Fig. S1), confirming the validity of our analyses. Inside clones, we quantified only entirely mutant ommatidia for polarity analysis. No R3⁺/R4^- or R3^-/R4⁺ mosaic ommatidia were included in any of our quantification, although we note that R3/R4 mosaic ommatidia tend to have reversed or rescued polarity according to their location in the clone, similar to fully mutant ommatidia (supplementary material Fig. S2).

Images were processed with Adobe Photoshop and statistical analysis was carried out using GraphPad Prism. Comparisons of phenotype strength were done by unpaired t-tests, and size- and position-dependence determined by linear regression.

To gain insight into the mechanisms underlying polarity propagation, we quantitatively analyzed alterations in polarity that occur in and around Ft-Ds pathway mutant clones (Fig. 1C). We found that ft^G-rv mutant clones cause reversal of polarity in 34% of genetically wild-type ommatidia in the first row of ommatidia outside the polar border of the clone (n=40 clones), whereas polarity outside the equatorial border was unaffected (Fig. 2A,C).

Models in which polarity propagation by the Ft-Ds pathway is due to asymmetric distribution of Ft and Ds in one cell promoting asymmetric redistribution of Ds and Ft in adjacent cells predict that: (1) polarity propagation in clones lacking Ft or Ds could only re-orient polarity in one row of cells inside mutant clones; (2) the degree of propagation in wild-type tissue outside clones depends on the relative levels and affinities of Ft and Ds; and (3) inside mutant clones, ommatidia not contacting a wild-type cell should have randomized polarity. To test these predictions, we quantified the degree of polarity disruption inside ft clones. If polarity were randomized, quantification should show a polarity distribution of ∼50% normally oriented ommatidia and ∼50% reverse-oriented ommatidia. A significant departure from this distribution would indicate the presence of polarity signaling in the absence of Ft and Ds redistribution.

We found that inside ft clones, 89% of entirely mutant ommatidia in the first row inside the polar border of the clone had reversed polarity (n=36 clones) (Fig. 2G). Polarity was almost completely rescued on the equatorial side of ft clones, as 93% of entirely mutant ommatidia had normal polarity (n=32 clones) (Fig. 2H). The rescue and reversal of polarity were statistically different from the 50% normal/50% reversed polarity distribution expected of randomized ommatidial orientation (both P<0.0001). Significantly, and in further contrast to a redistribution model, in the interior of many ft clones polarity was normal on the equatorial side and reversed on the polar side over up to four ommatidial rows, in the complete absence of Ft protein (Fig. 2A; supplementary material Fig. S3A). Importantly, these rescue and reversal effects extended to only a limited distance, and polarity was randomized in the middle of large ft clones. Another null ft allele, ft^fd (Matakatsu and Blair, 2006), showed similar clonal polarity effects (Fig. 2B; supplementary material Fig. S4). These data strongly argue that polarity information can propagate ∼10-40 cells in the absence of Ft, in contrast to the predictions of an asymmetric redistribution-based model.

We examined whether the degree of polarity reversals caused by ft mutant clones varied with clone size along the DV axis and found that it did not - clones that were one to two ommatidial rows in size caused reversals outside clones as often as those that were over five rows in size (Fig. 2D). We also examined effects of DV position, and found that clones near the equator caused reversals as often as those near the poles (P=0.63) (Fig. 2E).

Analysis of the shape of clone borders indicated that boundary geometry is crucial in determining the extent of polarity disruption. ft clone borders that were parallel to the equator induced more polarity reversals (43%) than clone borders slanted relative to the equator (10%) (Fig. 2F; supplementary material Fig. S3C), a significant difference (P<0.0001). As different alignments of clone boundaries result in different degrees of polarity reversal, we conclude that boundary-specific signaling is likely to be important in the ft clonal polarity effect.

We quantitatively analyzed ds^38k clones and found that polarity can also propagate in the absence of Ds (Fig. 3A,B). Polarity reverses in 31% of wild-type ommatidia in the first row of ommatidia outside the equatorial border of ds clones (n=40 clones). Polarity outside the polar border was unaffected. Polarity information is also transmitted inside ds mutant clones. Sixty-six percent of entirely ds mutant ommatidia in the first row inside the equatorial border of the clone have reversed polarity (n=29 clones) (Fig. 3G), a statistically significant degree of reversal (P=0.0046). On the polar side of ds clones, 95% of entirely ds mutant ommatidia in the first row have normal polarity (n=36 clones) (Fig. 3F). As with ft mutant clones, polarity rescue extends for many ommatidial rows inside clones, though the reversal effect is weaker (supplementary material Fig. S3B). The degree of polarity reversal caused by ds clones does not vary with clone size (P=0.09) (Fig. 3C), but does vary with position: ds clones near the poles cause more polarity reversals, (P=0.007) (Fig. 3D). ds levels are higher near the poles, so the difference in levels inside and outside a mutant clone is greater there, which may be responsible for the stronger polarity effect. Unlike ft clones, ds clones do not show significantly different degrees of reversal along parallel or slanted borders (P=0.23) (Fig. 3E). We observed changes in ommatidial orientation inside and outside ft and ds clonal borders in developing larval eye discs as well (supplementary material Fig. S5). Together, these data indicate that although Ft and Ds are asymmetrically distributed in wild-type tissue (Ambegaonkar et al., 2012; Brittle et al., 2012), polarity changes in the interiors of ft and ds mutant clones cannot be accounted for by a chain-reaction propagation of asymmetric Ft-Ds distribution.

ft is epistatic to ds in R3/4 fate determination (Yang et al., 2002). Quantitative analysis of ds ft clones revealed polar non-autonomous polarity effects similar to ft clones (Fig. 4A). However, although ft clones cause this effect in 34% of ommatidia, only 3% of ommatidia outside ds ft clones (n=40 clones) have reversed polarity (Fig. 4B). This effect was seen with two strong ds mutant alleles, ds^38k ft^G-rv and ds^UAO71 ft^G-rv (Fig. 4A; supplementary material Fig. S6A), and also when ds-RNAi is expressed inside ds ft clones (supplementary material Fig. S6C), suggesting that the polarity effects of ds ft clones are not due to residual ds function. Thus, loss of ds largely but not entirely suppresses the polar non-autonomous polarity effects caused by loss of ft in clones.

The polarity reversal observed inside the polar border of ft clones does not occur in ds ft clones (Fig. 4F). However, 73% of entirely ds ft mutant ommatidia in the first row inside the equatorial border have normal polarity, which is statistically different from randomized polarity (P<0.0001) (Fig. 4G). Thus, polarity information can be transmitted from clone borders in the absence of both Ft and Ds. The double mutant clonal phenotype is not size dependent (P=0.52) (Fig. 4C), and only slightly position dependent, perhaps reflecting the ds gradient (P=0.034) (Fig. 4D). ds ft clones do not show significantly different polarity effects along parallel or slanted borders (P=0.11) (Fig. 4E).

It has been suggested that polarity phenotypes associated with ft mutant clones are due to generation of an additional equator inside the clones (Rawls et al., 2002), leading to long-distance changes in ommatidial polarity. We did not find changes in the ser-lacZ reporter (Fig. 5), which marks regions of equator formation (Bachmann and Knust, 1998; Gutierrez-Aviño et al., 2009), in ft or ds clones. Furthermore, an ectopic equator would be expected to create a sharp polarity reversal, like the normal equator, in the middle of the clone. However, we find that consistent polarity, whether normal or reversed, is most strongly observed near clone borders, and is increasingly randomized farther inside clones (Fig. 2A, Fig. 3A; supplementary material Fig. S3A,B). We therefore conclude that apparent equators in clone interiors are a consequence of rescue and reversal of polarity at clone borders, and are not true equators, as observed (for example) in Wg pathway clones (Wehrli and Tomlinson, 1998).

Together, these data indicate that accumulation of Ft and Ds at cell borders establishes a polarity signal, and that polarity propagation can occur independently of Ft and Ds.

Expression of a version of Ft with most of the extracellular domain deleted (FtΔECD) can substantially rescue PCP when expressed ubiquitously in the wing (Matakatsu and Blair, 2006; Matakatsu and Blair, 2012). We found that expression of FtΔECD in clones resulted in rare disruptions in PCP within the clone, and no disruptions outside the clone (Fig. 7C). In contrast to the ability of FtΔECD to rescue polarity in the wing, FtΔECD is unable to rescue ft clonal polarity defects in the eye (Fig. 7D). This suggests that the Ft ECD is important for polarity in the eye, possibly via interaction with Ds. Recent work by Zhao et al. (Zhao et al., 2013) found that expression of FtΔECD in fully ft mutant eyes also does not rescue polarity, supporting this observation.

Expression of Ds lacking the entire extracellular domain (DsΔECD) in clones resulted in occasional non-autonomous polarity reversals outside the equatorial side of clones (Fig. 7J). This effect, which is qualitatively similar to, but weaker than, the ds clone effect, suggests that the Ds intracellular domain (ICD) acts in a dominant-negative manner in non-autonomous polarity signaling. Strikingly, expression of DsΔECD in ds clones resulted in polarity disruptions stronger than those of ds clones alone, with polarity reversal in 48% of wild-type ommatidia in the first row of ommatidia outside the equatorial border of the clone (n=18 clones) (Fig. 7K; supplementary material Fig. S7), significantly stronger than the 31% polarity reversal outside ds clones (P=0.023).

Importantly, ft clones that express DsΔECD do not exhibit the equatorial polarity reversals seen when DsΔECD is expressed in wild-type tissue. They instead behave like ft clones with polar reversals (Fig. 7L). Strikingly, ds ft clones expressing DsΔECD also have no polarity defects outside their equatorial borders (Fig. 7M), in contrast to 48% equatorial polarity reversals induced by ds clones expressing DsΔECD. Thus, as DsΔECD is unable to bind Ft in neighboring cells, DsΔECD-induced polarity changes must depend on the presence of Ft in cis.

What is downstream of Ft-Ds signaling? The transcriptional co-repressor Atro binds to the cytoplasmic domain of Ft, and, like ft, loss of atro results in changes in polarity (Erkner et al., 2002; Fanto et al., 2003; Zhang et al., 2002), but no quantitative analysis was previously performed. Quantification of polarity disruptions around atro³⁵ clones showed that 17% of wild-type ommatidia in the first row outside the polar border of atro clones have reversed polarity (n=38 clones) (Fig. 6A,C). Similarly to ft clones, the degree of polarity reversals induced by atro clones did not change with clone size (P=0.56) (Fig. 6D). However, it did change with clone position: clones near the equator caused more reversals and those near the poles caused significantly fewer reversals (Fig. 6E).

Strong atro alleles cause neurodegeneration in mutant tissue, making polarity inside clones indecipherable. To examine polarity inside atro clones, we analyzed a weaker allele: atro^e46. Clones of atro^e46 rarely cause non-autonomous effects, but do cause polar reversal and equatorial rescue inside clones, qualitatively similar to the effect of ft clones (Fig. 6B).

Loss of atro leads to increased fj expression (Fanto et al., 2003). To test whether increased expression of fj contributes to polarity disruptions caused by atro, we analyzed atro clones in fj^D1-null background. Strikingly, atro clones in fj^- animals caused polarity reversals in 55% of wild-type ommatidia in the first row of ommatidia outside the polar border of the clone (n=13 clones) (Fig. 7S), significantly stronger than atro clones in a wild-type background (P<0.0001). Thus fj is not required for atro clones to affect polarity, and in fact loss of fj enhances atro non-autonomous polarity disruptions.

The degree of polarity reversal does not significantly differ along parallel or slanted atro clone borders (P=0.08) (Fig. 6F). As atro suppresses ft expression in the adult eye (Napoletano et al., 2011), one possible mechanism of atro activity is through regulating Ft accumulation, promoting polarity disruptions. However, in developing discs, loss of atro has no effect on Ft or Ds levels or localization (supplementary material Fig. S8). These data further argue against changes in Ft and Ds localization being responsible for propagation of polarity information, and suggest that a downstream signal establishes polarity.

ds and atro mutant clones have opposite polarity phenotypes, allowing examination of their epistatic relationship. ds and atro are on different chromosomes, so we could not generate double mutant clones. We instead made ds clones expressing atro-RNAi. These clones caused both polar non-autonomous polarity reversals (more frequently than atro-RNAi-only clones), and equatorial non-autonomous polarity reversals (less frequently than ds clones (Fig. 7T,U; supplementary material Fig. S9).

We examined the roles of other polarity regulators in the Ft-Ds pathway. D is an unconventional myosin that acts downstream of Ft (Feng and Irvine, 2007; Mao et al., 2006). D subcellular localization is regulated by Ft and Ds in the wing and eye (Ambegaonkar et al., 2012; Brittle et al., 2012), and loss of d suppresses polarity randomization in ft and ds mutant eyes, probably through its effect on the Hippo pathway (Brittle et al., 2012). However, d^GC13 mutant flies have extremely mild polarity defects in the eye (Mao et al., 2006) (supplementary material Fig. S10A). Of 45 wild-type ommatidia examined on the polar borders of 11 d mutant clones, only one ommatidium had reversed polarity. To examine the contribution of d to ft-dependent polarity propagation, we examined ft d clones and found that they were similar to ft clones in their effects on polarity both inside and outside clones (Fig. 7G; supplementary material Fig. S10B). ft d clones caused polarity reversals in 31% of wild-type ommatidia in the first row of ommatidia outside the polar border of the clone (n=19 clones), similar to the effect of ft clones (34% polarity reversal, P=0.72), and also display reversal and rescue inside clones. Thus, D does not mediate production and propagation of the Ft-Ds long-range PCP signal in the eye.

fat2, the only ft paralog in Drosophila (Viktorinová et al., 2009), regulates polarization of ovarian follicle cells. We found that neither eyes wholly mutant for fat2^58D nor fat2 mutant clones had polarity defects (supplementary material Fig. S11A,B). Furthermore, the non-autonomous polarity caused by ft and ds mutant clones was not altered by loss of fat2 in the background (supplementary material Fig. S11C,D). fat2, thus, appears not to be involved in polarity in the Drosophila eye.

The Golgi-localized kinase Fj acts both upstream and downstream of Ft and Ds (Ishikawa et al., 2008; Strutt et al., 2004; Yang et al., 2002). fj expression is negatively regulated by ft and atro, and positively regulated by ds and d (Cho et al., 2006; Fanto et al., 2003; Mao et al., 2006; Yang et al., 2002). We therefore examined the role of fj in ft and ds clonal polarity disruptions in the eye.

We found that in a fj-null background, polarity reversals caused by ft and ds mutant clones increased. ft clones in a fj^- background caused polar polarity reversals in 63% of wild-type ommatidia (n=9 clones) (Fig. 7B), compared with 34% in a wild-type background. ds clones in a fj^- background caused equatorial polarity reversals in 68% of wild-type ommatidia (n=23 clones) (Fig. 7I), compared with 31% in a wild-type background. Similarly, ds ft clones in a fj^- background caused polar polarity reversals in 16% of wild-type (n=6 clones) (Fig. 7P), compared with 3% in a wild-type background. Thus, fj is not required for non-autonomous polarity of ft and ds clones, and loss of fj strengthens these effects, as with atro clones. Interestingly, even though both ds and fj gradients are absent inside ds mutant clones in a fj^- background, there is extensive rescue and reversal of polarity around clone borders (supplementary material Fig. S12).

The relationship between The Ft-Ds and Fz/PCP pathways is complex and context dependent (Adler et al., 1998; Aigouy et al., 2010; Casal et al., 2006; Hogan et al., 2011; Ma et al., 2003; Matakatsu and Blair, 2004; Sagner et al., 2012; Strutt and Strutt, 2002; Yang et al., 2002). ft clones and ds clones expressing fz-RNAi had no consistent rescue or reversal of polarity inside clones. Instead, polarity was randomized, with reversed, misrotated and symmetrical ommatidia (supplementary material Fig. S13), similar to fz clones (Zheng et al., 1995). Wild-type ommatidia surrounding ft and ds mutant clones expressing fz-RNAi were often misrotated, complicating analysis. However, we detected polarity reversals in wild-type ommatidia on polar borders of both types of clones, characteristic of the fz phenotype (Fig. 7E,F,N; supplementary material Fig. S13A-C). Furthermore, we occasionally observed ommatidia with reversed polarity outside equatorial borders of ds clones expressing fz-RNAi. These ommatidia were also usually misrotated (supplementary material Fig. S13D). Thus, ds clones expressing fz-RNAi have stronger polarity effects than fz-RNAi-only clones, and weaker equatorial polarity effects than ds mutant-only clones.

Ft and Ds can regulate each others subcellular localization, and changes in their levels in clones can induce redistribution outside clone boundaries (Ambegaonkar et al., 2012; Bosveld et al., 2012; Brittle et al., 2012; Ma et al., 2003; Matakatsu and Blair, 2004; Strutt and Strutt, 2002). It has recently been proposed that a redistribution cascade may be responsible for long-range polarity propagation (Ambegaonkar et al., 2012; Brittle et al., 2012). A redistribution cascade cannot, of course, occur inside ft or ds clones, and yet we observe very strong polarity reorganization well inside these clones. This can only occur if the long-range signaling is occurring through something other than Ft or Ds redistribution. Surprisingly, we find that polarity propagation extends much further in the absence of Ft or Ds, than in wild-type tissue. In ft and ds clones, rescue or reversal of polarity can in some cases extend to a distance of approximately four ommatidial rows or ∼40 cells along the DV axis. The increased range of polarity reversals in the absence of Ft or Ds suggest that the normal distribution of Ft and Ds outside the clone provides a signal that counteracts and dampens the polarity propagation from the clone border. Thus, although the normal asymmetric distribution of Ft and Ds could play a role in creating or orienting the polarity signal, it is not necessary for transmission of the signal.

One surprising result was that ommatidia along borders of ft mutant clones that are parallel to the equator are more likely to be reversed than those along slanted borders. What is the basis of this difference? We speculate that this is due to different degrees of accumulation of Ft-Ds interactions at parallel and slanted borders. In the interior of a ft clone, Ds is diffusely distributed. At the border of the ft clone, all available Ds redistributes to bind Ft in wild-type cells touching the clone (Ma et al., 2003; Simon et al., 2010; Strutt and Strutt, 2002). In ft clones that have a border that spreads over several cell diameters along the AP axis, parallel to the equator, this causes redistribution of Ds to only one cell border along the DV axis. Along slanted borders, by contrast, Ds redistribution dissipates over both DV and AP sides of the cell, reducing the degree of accumulation (Fig. 8).

We also examined the effects of expression of FtΔECD and DsΔECD on non-autonomous polarity outside clones. We saw only subtle changes in polarity upon expression of FtΔECD. However, we saw clear non-autonomous polarity effects upon clonal expression of DsΔECD. Surprisingly, this caused a polarity change qualitatively similar to loss of ds in clones, suggesting that DsΔECD had a dominant-negative effect on polarity signaling. Importantly, DsΔECD signals in a Ft-dependent manner, as equatorial polarity disruptions are lost in ft clones. As there is no extracellular domain of Ds in DsΔECD, this suggests that Ds ICD interacts with Ft ICD in cis to regulate non-autonomous polarity propagation.

Why do ds ft clones cause polarity reversals that are a weak version of those caused by ft clones? ft and ds single mutant clones cause strong accumulation of Ds or Ft, respectively, in mutant cells just inside the clone border. In ds ft double mutant clones, however, there would be no such strong accumulation at the border (Fig. 8). There is only loss of signal within clones, which causes milder polarity disruptions.

Why are the border effects of ft and ds clones opposite in orientation? We speculate that Ft-Ds heterodimers form both in cis (in the same cell) and in trans (with neighboring cells). In wild-type cells, both cis and trans forms exist. However, inside ft clones, although the Ds protein in a ft^-/- cell can bind to Ft in adjacent cells, it cannot form a cis heterodimer. Similarly, in ds clones, although Ft can bind to Ds in adjacent cells, that Ft molecule has no Ds cytoplasmic domain to bind it in cis. We propose that the signals produced upon Ft-Ds binding are affected by both cis and trans interactions. This difference in signaling by cis and trans dimers could lead to differential regulation of the same long-range signaling molecule, or to regulation of two different signaling molecules. Although Ft and Ds are both cadherins, they have very different cytoplasmic domains, suggesting they may transmit information about binding in different ways.

What is downstream of Ft-Ds signaling? The Ds intracellular domain is known to affect Hippo pathway targets in the wing (Matakatsu and Blair, 2012), and is required to receive boundary signals that regulate Hippo targets in the eye (Willecke et al., 2008). However, loss of Hippo pathway genes does not lead to non-autonomous polarity disruptions in clones (Blaumueller and Mlodzik, 2000; Harvey et al., 2003; Tapon et al., 2002; Udan et al., 2003; Wu et al., 2003). Although D localization can change in response to Ds levels (Ambegaonkar et al., 2012; Brittle et al., 2012), we find that d clones do not cause non-autonomous polarity disruptions, and ft d double mutant clones have similar polarity disruptions to ft clones. Thus, D is not a component of polarity signal propagation in the eye.

Atro binds to Ft ICD and atro clones qualitatively phenocopy ft clones (Fanto et al., 2003) (Fig. 6), suggesting Atro could mediate the Ft-Ds polarity signal. Loss of atro does not lead to changes in Ft or Ds levels or localization (supplementary material Fig. S8), indicating that atro does not affect polarity by changing Ft-Ds accumulation. Interestingly, we found that atro mainly functions near the equator, as loss of atro near the poles has little effect. This suggests atro functions downstream of Ft and Ds near the equator, with another unidentified polarity regulator acting as a Ft-Ds polarity effector near the poles.

Ds, Ft and Fj interactions have been extensively studied in the Drosophila abdomen wing, eye and larval denticles (Ambegaonkar et al., 2012; Bosveld et al., 2012; Brittle et al., 2012; Casal et al., 2002; Casal et al., 2006; Donoughe and DiNardo, 2011; Harumoto et al., 2010; Ma et al., 2003; Repiso et al., 2010; Simon, 2004; Strutt et al., 2004; Yang et al., 2002). However, the effects of loss of fj on non-autonomous polarity caused by ft and ds clones have not been examined. fj levels are altered in ft, ds and atro mutant clones (Fanto et al., 2003; Yang et al., 2002). If changes in fj levels were essential for change in polarity signaling at ft, ds or atro clone borders, the phenotypes of these clones would be weaker in a fj^- animal. Instead, we found that non-autonomous polarity disruptions of ft, ds or atro clones are enhanced in fj^- animals. Thus, changes in fj expression cannot be responsible for their non-autonomous polarity effects. The strengthened polarity effects could be explained in terms of diminution of a standing polarity gradient in the wild-type cells surrounding the clone, as fj and ds establish complementary and redundant polarity gradients in the eye (Simon, 2004; Zeidler et al., 1999). Loss of fj strongly enhances polarity phenotypes of Ft, Ds and Ds-ECD overexpression, consistent with this conclusion (Casal et al., 2006; Simon, 2004; Strutt et al., 2004).

ds mutant clones expressing atro-RNAi cause both polar and equatorial non-autonomous polarity reversals. Analysis of knockdown phenotypes is complicated, but observation of reversals on both sides suggests some parallel function between ds and atro. Analysis of ds clones expressing fz-RNAi suggests a similar relationship between ds and fz.

To summarize, we observe that: (1) polarity information is transmitted to a large (but not unlimited) distance inside ft and ds clones; and (2) clone size does not affect phenotype but (3) clone shape affects PCP signal generation. These data, along with previously observed changes in localization of Ft and Ds at clone borders (Ma et al., 2003; Matakatsu and Blair, 2004; Strutt and Strutt, 2002), suggest a model in which polarity signals from clone borders act in concert with changes in signaling inside mutant clones. We propose that normal Ft-Ds signaling, acting in a gradient because of graded ds and fj expression, results in a graded production of long-distance secondary signal. This creates a standing gradient of polarity information. Inside ft and ds clones, the gradient of polarity information is reduced or lost. These changes are reinforced by accumulation of Ds or Ft at clone borders. The resultant changes in secondary signal oppose the standing gradient in surrounding wild-type tissue, leading to non-autonomous polarity reversals outside the clone. Inside the clone, with no standing gradient to oppose it, the aberrant signal from clone borders leads to extensive reversal or rescue of polarity.

Though our data strongly supports the existence of a long-range signal, its identity is open to speculation. The Fz/PCP pathway itself is an obvious candidate, but it could be another signaling pathway or an as yet unknown factor. With the current absence of a clear mechanistic link between the Ft-Ds and Fz/PCP pathways, or between these pathways and their downstream polarity effectors, the determination of the secondary signal is an unresolved issue.

Extensive structure-function analysis has identified regions of the intracellular domain of Ft that function in PCP regulation (Matakatsu and Blair, 2012; Pan et al., 2013). However, these works examined wholly mutant tissue. Further studies are needed to address the specific regions of the Ft and Ds intracellular domains essential for regulation of long-range polarity signaling.

It is interesting to consider how the same factors, Ft, Ds and Fj, working towards the same goal, i.e. establishing tissue polarity, appear to work in different ways in different contexts (Casal et al., 2006; Matakatsu and Blair, 2012; Willecke et al., 2008). For example, we found that FtΔECD does not rescue polarity effects of ft clones in the eye, unlike in the wing and abdomen (Casal et al., 2006; Matakatsu and Blair, 2006; Matakatsu and Blair, 2012). This suggests that the Ft extracellular domain is important for polarity regulation in the eye, perhaps through binding Ds. In the abdomen, ds ft clones have no polarity effects outside clones, whereas there are clear, though limited, polarity disruptions outside ds ft clones in the eye. Casal and co-workers argue that Ds and Ft are both needed in the receiving cell to propagate Ft and Ds polarity information in the abdomen (Casal et al., 2006). This may be due to tissue-specific differences in signaling.

Our data support a Ft-Ds signaling model in the Drosophila eye wherein different levels of Ft and Ds lead to modulation of a long-range polarity effector that directs further polarity establishment in the eye. Propagation of this signal is independent of changes in the asymmetric distribution of Ft and Ds. In at least part of the eye, this process may be mediated by transduction of Ft-binding information to the nucleus by Atro. Further work is needed to understand the identity of the long-range signal(s), and how Ft and Ds levels and interactions alter signal strength.

We are grateful to Seth Blair, Christian Dahmann, Maria Dominguez, Manolis Fanto, Ken Irvine, David Strutt, C.-C. Tsai, the Bloomington Drosophila Stock Center, the Transgenic RNAi Project at Harvard Medical School, the Vienna Drosophila RNAi Center and the Developmental Studies Hybridoma Bank for fly stocks and antibodies. We thank Rodrigo Fernandez-Gonzalez, Peter Lawrence, Alison McGuigan, Gary Struhl and Rudi Winklbauer for insightful discussion and comments on the manuscript.