The novel SH3 domain protein Dlish/CG10933 mediates fat signaling in Drosophila by binding and regulating Dachs

Much of the Hippo and planar cell polarity (PCP) signaling mediated by the Drosophila protocadherin Fat depends on its ability to change the subcellular localization, levels and activity of the unconventional myosin Dachs. To better understand this process, we have performed a structure-function analysis of Dachs, and used this to identify a novel and important mediator of Fat and Dachs activities, a Dachs-binding SH3 protein we have named Dlish. We found that Dlish is regulated by Fat and Dachs, that Dlish also binds Fat and the Dachs regulator Approximated, and that Dlish is required for Dachs localization, levels and activity in both wild type and fat mutant tissue. Our evidence supports dual roles for Dlish. Dlish tethers Dachs to the subapical cell cortex, an effect partly mediated by the palmitoyltransferase Approximated under the control of Fat. Conversely, Dlish promotes the Fat-mediated degradation of Dachs. DOI: http://dx.doi.org/10.7554/eLife.16624.001


Introduction
Heterophilic binding between the giant Drosophila protocadherins Fat and Dachsous (Ds) both limits organ growth, via regulation of the Hippo pathway, and orients planar cell polarity (PCP), through cell-by-cell polarization of Fat, Ds and their downstream effectors (Blair, 2012(Blair, , 2014Irvine and Harvey, 2015;Matis and Axelrod, 2013). Loss of Fat and, to a lesser extent, Ds, leads to the profound overgrowth of the Drosophila imaginal discs that give rise to adult appendages, and loss of either disorders the polarity of cell divisions, hairs and other morphological features in a variety of Drosophila tissues. But while players and pathways have been defined that are genetically downstream of Fat-Ds binding, only a little is known about the biochemical links between these and their most powerful regulator, the intracellular domain (ICD) of Fat.
A good deal of the recent work on Fat effectors has focused on the regulation of unconventional type XX myosin Dachs (Mao et al., 2006;Matakatsu and Blair, 2008;Rauskolb et al., 2011;Rogulja et al., 2008). Dachs is critical first because it provides the only known marker specifically sensitive to changes in the Fat/Ds branches of both the Hippo and PCP pathways. Dachs is normally concentrated in the subapical cell cortex, overlapping subapically-concentrated Fat and Ds. Loss of Fat greatly increases subapical Dachs levels, and polarization of Fat and Ds to opposite cell faces can in turn polarize Dachs to the face with less Fat (Ambegaonkar AA, 2012;Bosveld et al., 2012Bosveld et al., , 2016Brittle et al., 2012;Mao et al., 2006). Fat thus inhibits or destabilizes subapical Dachs, while Ds may do the opposite. Downstream changes in Hippo or PCP activities do not affect Dachs.
Dachs changes are also critical because they play a major role downstream of Fat. Dachs binds to and inhibits the activity of the kinase Warts (the Drosophila Lats1/2 ortholog), both reducing Warts levels and changing its conformation Rauskolb et al., 2011;Vrabioiu and Struhl, 2015). Warts is concentrated in the subapical cell cortex (Matakatsu and Blair, 2008;Sun et al., 2015), and thus the increased cortical Dachs of fat mutants should reduce the phosphorylation of Yorkie by Warts, allowing Yorkie to move into the nucleus to drive the transcription of growth-promoting target genes. Indeed, Dachs is necessary for the overgrowth and increased Yorkie target gene expression of fat mutants Mao et al., 2006). Dachs overexpression also causes overgrowth, although more weakly than the overgrowth caused by the loss of Fat, indicating that Dachs is partly sufficient Rauskolb et al., 2011).
Dachs can also bind to the core PCP pathway component Spiny legs (Sple) and alter its localization, thus influencing PCP in the subset of tissues that rely on Sple (Ayukawa et al., 2014). The increased levels of unpolarized Dachs in fat mutants may misdirect Sple, accounting for at least some of the PCP defects; fat mutant hair PCP defects are improved, although not eliminated, by loss of dachs (Ambegaonkar and Irvine, 2015;Blair, 2014;Mao et al., 2006;Matakatsu and Blair, 2008).
Dachs has not been shown to interact directly with Fat's ICD, and only three other proteins are known to affect Dachs accumulation in the subapical cell cortex, the casein kinase 1e Discs overgrown (Dco), Approximated (App) and F-box-like 7 (Fbxl7). Dco may act through Fat itself: Dco binds and phosphorylates the Fat ICD, and loss of Dco function causes strong overgrowth and increases subapical Dachs, similar to loss of Fat (Feng and Irvine, 2009;Matakatsu, Blair and Fehon, Unpublished;Pan et al., 2013;Rodrigues-Campos and Thompson, 2014;Sopko et al., 2009).
App suggests a mechanism in which Fat inhibits the tethering of Dachs to protein complexes in the subapical domain ( Figure 1). app mutants decrease subapical Dachs levels and reduce Dachs activity (Matakatsu and Blair, 2008). Thus, like dachs mutants, app mutants reverse the overgrowth and increased Yorkie target gene expression normally observed in fat mutants, and improve hair PCP. App is one of 20 Drosophila DHHC palmitoyltransferases, transmembrane proteins responsible Figure 1. Models of Fat-mediated regulation of subapical Dachs by Fat-inhibited subapical tethering and Fatstimulated destabilization. In wild type, Dlish binds Dachs, and when palmitoylated by Approximated helps tether Dachs to the subapical membrane where Dachs can inhibit Warts; Dachs contributes to tethering by binding the F-actin cytoskeleton. Fat binds Approximated and reduces its activity, reducing Dachs tethering. Fat also binds Fbxl7 and other ubiquitin ligases, destabilizing Dachs and trafficking it from the membrane. Dlish promotes the destabilization of Dachs, either indirectly by tethering it near Fat-bound ubiquitin ligases, or directly by helping Dachs complex with those ligases. In dlish mutants both the tethering and destabilization of Dachs are reduced; excess Dachs accumulates in the cytoplasm but no longer inhibits Warts. In fat mutants Approximated-mediated tethering increases and Fat-mediated destabilization of Dachs is lost, increasing subapical Dachs and the inhibition of Warts. DOI: 10.7554/eLife.16624.002 for adding palmitates to cytoplasmic proteins and thereby anchoring them to cell membranes (Fukata et al., 2016;Linder and Deschenes, 2007;Nadolski and Linder, 2007). App is also concentrated in the subapical cell membrane and can bind both Dachs and the Fat ICD (Matakatsu, Blair and Fehon, Unpublished). Thus, in the simplest model App palmitoylates or tethers Dachs, concentrating it in the cell cortex, and Fat works in part by sequestering or inhibiting App. However, Dachs is not detectably palmitoylated (Matakatsu and Blair, 2008;Matakatsu, Blair and Fehon, Unpublished).
Fbxl7 suggests a complementary mechanism based on ubiquitination. Fbxl7 binds Fat's ICD, and Fbxl7 localization and function are regulated by Fat; moreover, subapical Dachs increases in Fbxl7 mutants, causing moderate overgrowth Rodrigues-Campos and Thompson, 2014). Like other F-box proteins (Skaar et al., 2013), Fbxl7 could function in a Skp-Cullin1-F-box (SCF) ubiquitin E3 ligase complex, thereby targeting Dachs for proteasome-mediated degradation or trafficking away from the membrane (Figure 1). However, Fbxl7 has not been shown to bind Dachs or change its total levels in cells, and the two studies differ about its ability to stimulate Dachs ubiquitination. Nor does loss of Fbxl7 completely reproduce the strong overgrowth caused by loss of Fat or Dco. Nonetheless, the idea that Fat regulates a complex for the ubiquitination of Dachs is an attractive one (Figure 1), given Fat's ability to locally reduce the accumulation of Dachs. In fact, loss of Fat greatly increases the total levels of Dachs in imaginal disc cells (Matakatsu et al., Unpublished), a result we will confirm.
Below, we will describe the function of a novel Dachs binding protein, and show that its effects provide strong evidence for both the palmitoylation-dependent and degradation-dependent regulation of Dachs ( Figure 1). Our structure-function analysis of Dachs found regions required for its normal subapical localization, and we used this information as the basis for a screen for novel Dachs binding partners. We found a direct binding partner for the Dachs C-terminus, the previously uncharacterized SH3 domain protein CG10933, which we have renamed Dachs ligand with SH3s, or Dlish. We show that the activity and subapical concentration of Dlish are regulated by Fat, Dco and Dachs, and that Dlish in turn is required for the subapical concentration and full activity of Dachs in both wild type and fat mutant cells. Dlish localization also depends on App; furthermore Dlish binds to and is palmitoylated by App, and that palmitoylation can be suppressed by Fat. Loss of Dlish also increases the total levels of Dachs, likely by blocking Fat-mediated destabilization of Dachs. We propose that Dlish targets Dachs to subapical protein complexes in part via Fat-regulated, App-mediated palmitoylation. Dlish thereby concentrates Dachs where it can efficiently inhibit subapical Warts, and conversely links Dachs to the machinery for Fat-dependent destabilization.

Results
Hippo-specific subdomains of the Fat ICD affect the accumulation of subapical Dachs Overexpression of a version of Fat that lacks its extracellular domain (FatDECD) can rescue the overgrowth and heightened Yorkie target gene expression normally seen in fat mutant wings (Matakatsu and Blair, 2006). This rescue requires the presence of adjacent 'PH' and 'Hippo' domains in the ICD (Matakatsu and Blair, 2012) (see schematic in Figure 2J). We extended these studies by testing which domains reduce the increased subapical Dachs of fat mutants.
We found a strong correlation between constructs that rescued overgrowth and Hippo pathway activity and those that suppressed Dachs. hh-gal4-driven expression of a construct containing only the transmembrane, the PH and both the Hippo N and C domains (FtDECD-PH+Hippo) reduced overgrowth, Yki targets and Dachs levels ( Figure 2A-C), as did larger constructs containing these domains ( Figure 2D) or a smaller construct containing only the Hippo domains (FtDECD-Hippo) (Figure 2-figure supplement 1). Deletions lacking only one of the two Hippo domains could also weakly reduce Dachs levels ( Figure 2E,F), in line with their weak effects on Hippo pathway markers (Matakatsu and Blair, 2012). Constructs lacking both the PH and Hippo domains did not detectably suppress subapical Dachs, even those that retained the N-terminal PCP-regulating region (FtDECDD1-C; Figure 2G), or the region C-terminal to the Hippo domains (FtDECDDN-5 or DN-6; Figure 2H,I).
This does not mean regions outside the Hippo domains lack function, especially at the lower expression levels of the endogenous gene. While point mutations in the Hippo domains cause overgrowth Bossuyt et al., 2014), fat mutants lacking the phosphorylation sites or the 'D' domain just C-terminal to Fat's Hippo domains ( Figure 2J) also cause moderate overgrowth and increased Dachs levels, and deleting the C-terminal 'F' domain misorients Dachs polarization Pan et al., 2013;Rodrigues-Campos and Thompson, 2014). However, the D and F domains are not sufficient to visibly affect Dachs levels in the absence of the Hippo domains, even at the high levels of overexpression used here, suggesting a modulatory or cooperative role.

Regions C-terminal to the Dachs myosin domain are critical for Dachs function and subcellular localization
To investigate the structures responsible for Dachs localization and activity, we tested whether expression of Dachs deletion constructs in dachs mutant flies mimicked the expression of full-length Dachs, which can rescue viability, increase the expression of Yorkie activity reporters such as ban3-GFP and concentrate at the subapical cell cortex ( Figure 3B,H).
First, we found the portion of Dachs N-terminal to the myosin head was not needed for either function or subapical localization (D-DN; Figure 3C,I). This region contains a putative coiled-coil motif that is well-conserved in predicted Dachs homologs from other insects and some arthropods, but is missing from the shorter predicted homologs from most other taxa (   The Dachs myosin head domain binds F-Actin (Cao et al., 2014) and Warts (Rauskolb et al., 2011). Dachs lacking the myosin head (D-DMH) did not rescue dachs mutants ( Figure 3D) and had no effect on wild type wings. D-DMH also localized to the cytoplasm, and did not concentrate in the sub-apical cell cortex ( Figure 3J), consistent with the hypothesis that the head tethers Dachs to the cortical cytoskeleton.
C-terminal to the myosin domain is a single regulatory IQ domain that binds calmodulin (Cao et al., 2014) and several regions that lack Pfam and SMART motifs but are well-conserved in predicted homologs from all taxa (Figure 3-figure supplement 1). Removing the entire C terminus (D-D1C) impaired Dachs activity and localization, as did a smaller deletion that left the IQ domain intact (D-D2C) ( Figure 3E,F,L,M). Removing only the C-terminal 98 amino acids (D-D3C) also impaired Dachs activity ( Figure 3G); its concentration in the sub-apical cell cortex was slightly improved compared with the larger C-terminal deletions, but it was still disorganized and cytoplasmic levels were abnormally high ( Figure 3N). Overexpression of the C-terminal deletions did not affect the growth of wild type wings, and D-D1C did not affect the endogenous Dachs recognized by a C-terminal-specific anti-Dachs (Figure 3-figure supplement 1B).

The SH3 protein Dlish directly binds to the Dachs C terminus
Given the importance of the C-terminal portions or Dachs, we used this region as bait in a yeast two-hybrid screen for potential binding partners. Of the 26 confirmed positives, 17 were to CG10933 ( Table 1). CG10933 was also found to complex with Dachs in a mass-spec screen of the Hippo pathway interactome (Kwon et al., 2013), so we selected it for further study. We confirmed binding between Dachs and CG10933 by reciprocal co-immunoprecipitation (co-IP) in S2 cells ( Figure 4B). The co-IP was specific for the C-terminus of Dachs ( Figure 4C). Binding was direct, as we observed it in both yeast two-hybrid and by co-IP between GST-purified CG10933 and in vitrotranslated Dachs constructs containing the C-terminus ( Figure 4E). CG10933 binding to larger C-terminal Dachs constructs containing more of the myosin head was reduced, but still visible, possibly due to interference between the truncated head and the CG10933-binding region in the C-terminus.
CG10933 encodes a novel cytoplasmic protein containing three SH3 domains ( Figure 4A). Predicted CG10933 homologs are found in many animal taxa, from Placozoa to Cephalochordata, that also contain a predicted Dachs homolog (Figure 4-figure supplement 1). Like Dachs, however, CG10933 lacks an obvious vertebrate ortholog. We have renamed CG10933 Dachs ligand with SH3s, or Dlish.
The portion of Dachs that is deleted in D-D3C contains a proline-rich region that conforms to the consensus for both type I and type II SH3-binding domains (Kaneko et al., 2008), and that is conserved in most Dachs homologs (Figure 3-figure supplement 1). Co-IP depended on this binding domain (SH3Bd-3), but not two other candidate proline-rich sites, SH3Bd-1 and -2 ( Figure 4D

Dlish complexes with the Fat ICD and its binding partners
In S2 cells, we have not detected binding between Dlish and the Dachs binding proteins Warts and Zyxin, the Warts-binding protein Ajuba, or the Hippo pathway components Hippo, Sav, Mats, or Yorkie. However, Dlish can co-IP with FatDECD ( Figure 5A). Known binding partners for the Fat ICD include Dco (Feng and Irvine, 2009;Sopko et al., 2009) App (Matakatsu, Blair and Fehon, Unpublished) and Fbxl7 Rodrigues-Campos and Thompson, 2014). Dco and App can co-IP Dlish ( Figure 5A), as can Fbxl7 (see below). DsICD, a Ds construct which lacks both its extracellular and transmembrane domains, has been reported to co-IP Dachs (Bosveld et al., 2012), and can co-IP Dlish as well ( Figure 5A).
While in vitro co-IPs should be treated with caution, these results raise the possibility that, in vivo, Dlish and its direct binding partner Dachs participate and function in the subapical protein complexes regulated by the ICDs of Fat and Ds. As summarized in Figure 2J, different domains of Fat's ICD have different binding and signaling activities, so we next examined Dlish binding to these domains. Fbxl7 and Dco bind to domains C-terminal to the Hippo domain Feng and Irvine, 2009;Rodrigues-Campos and Thompson, 2014;Sopko et al., 2009). We have found that Dlish can co-IP with the FatDECDDN-6 construct containing this region ( Figure 5B). Dlish can also co-IP with FatDECD-PH+Hippo, which includes the PH and Hippo domains most active in the regulation of growth and Dachs levels, but which had no previously known binding partners. Dlish does not co-IP with FatDECDD4-C, which includes the PH but not the Hippo domains and which cannot regulate growth. We did not observe co-IP with the smaller FatDECD-Hippo construct that lacks the PH domain. Although in vivo the PH region is not required for activity of the adjacent Hippo domains (e.g.  it does strengthen that activity (Matakatsu and Blair 2012), and in the different environment found in S2 cells may be more critical for proper folding or the formation of protein complexes.

Fat and Dachs affect Dlish accumulation in the subapical cell cortex
We next examined how Dlish might be regulated by Fat and Dachs in vivo. In situ hybridization showed that Dlish is expressed ubiquitously in imaginal discs ( Figure 6A). To analyze the subcellular distribution of Dlish we generated anti-Dlish. The reduced staining within regions of Gal4-driven UAS-dlish-RNAi allowed us to unambiguously localize endogenous Dlish: low levels of Dlish were found diffusely in the cytoplasm, but high levels were concentrated at the subapical cell cortex, overlapping the region of highest anti-Dachs staining ( Figure 6B . The Dlish increase was also observed in fat mutant discs, and was reduced by expression of FatDECD-PH+Hippo ( Figure 6G). The Dlish increase is not caused by the overgrowth and heightened Yorkie activity of fat or dco mutants, as hh-gal4-driven expression of an activated form of Yorkie (UAS-Yki S168A ) caused strong, fat mutantlike overgrowth but had no effect on subapical Dlish ( Figure 6-figure supplement 1D). The subapical Dlish accumulation is also regulated by Dachs, as subapical Dlish was increased by overexpression of Dachs ( Figure 6H) and decreased in dachs or fat dachs mutant clones ( Figure 6I-J).
dlish is required for normal Dachs localization and fat mutant overgrowth Not only does subapical Dlish depend on Dachs, but both Dachs localization and function depend on, and can be driven by, Dlish. Moderate reductions of Dlish led to adult phenotypes characteristic of reduced Dachs or increased Fat function. dlish 04 , caused by insertion of a P element containing UAS-YFP.Rab26.Q250L into the first coding exon of the dlish gene (Zhang et al., 2007), is homozygous and hemizygous viable. Western blotting shows that dlish 04 still produces protein of approximately the normal size, but the placement of the P element insertion makes it likely that it is missing the six most N-terminal amino acids, which are highly conserved in homologs from other taxa There was no obvious change in Fat or Ds levels or localization in dlish 04 homozygous clones, or in regions of UAS-dlish-RNAi-mediated knockdown (Figure 7-figure supplement 1F,G). There was, however, a profound effect on anti-Dachs staining that was reminiscent of the Dachs mislocalization caused by the removal of its Dlish-binding C terminus. Dachs levels increased throughout the cytoplasm, and the concentration of Dachs at the subapical cell cortex was slightly to severely disrupted, depending on the strength of the Dlish knockdown ( Figure 7L Dlish is also sufficient to change Dachs localization and activity. Overexpression of Dlish caused the overgrowth of adult wings (Figure 7,J), and in discs moderate increases in subapical Dachs and increased Yki activity ( Figure 7P,Q). Dlish tagged at the C-terminus with FLAG was more effective than untagged Dlish despite both being expressed at similar levels, suggesting that the FLAG tag alters Dlish conformation or binding in a manner that increases its activity.
The effects of Dlish on subapical Dachs were even more striking in fat mutant cells. Instead of the strong increase normally observed in fat mutant tissue, subapical Dachs was reduced in the dlish- Because the GFP is in nuclei, the clone marker image is from the more basal, nuclear focal plane and slightly out of register with subapical Dlish. (L-N 2 ) Posterior, hh-gal4 driven expression of UAS-dlish-RNAi decreases cortical, subapical Dachs (L) and increases basolateral, cytoplasmic Dachs (M); N 1 and N 2 show the Dachs changes in the same cross-section Z series projection that shows Dlish in Figure 6B 2 , with Engrailed (En) marking the posterior (p) versus anterior (a). (O 1 ,O 2 ) dlish 04 homozygous clone, marked by absence of GFP, increases basolateral, cytoplasmic Dachs. (P 1 -Q 2 ) Posterior, hh-gal4driven expression of UAS-dlish-FLAG increases expression of the Yorkie activity reporter ex-lacZ (P 1 ,P 2 ) and increases subapical Dachs (Q 1 ,Q 2 ); anterior marked with Ci (P 1 ) or posterior with GFP (Q 1 ). DOI: 10.7554/eLife.16624.015 The following source data and figure supplement are available for figure 7: Source data 1. Data and analysis for comparison of adult wing sizes between nub-gal4 and nub-gal4 UAS-dlish-RNAi for Figure 7C. RNAi-expressing region of fat mutant discs ( Figure 8A). Moreover, dlish-RNAi also substantially reversed the overgrowth and increased Yki activity normally caused by loss of fat ( Figure 8A,B), similar to the effects of dachs reduction in fat mutants. And in marked contrast to the severe disc overgrowth and 100% pupal lethality of fat fd homozygotes, the majority of dlish 04 fat fd double homozygotes produced viable adults with mispatterned but nearly normally sized wings ( Figure 8C).
In one view, the primary function of Dlish is to direct or tether Dachs to the subapical cell cortex; the cortical Dachs is then more active because it is better placed to bind and inhibit cortical Warts. Alternatively, Dlish could be directly mediating the regulation of Warts by Dachs, independent of its effects on Dachs localization. We favor the first alternative, as we have not found binding between Dlish and Warts or other Warts-binding proteins (see above). In addition, we found that tethering Dachs to the cell cortex by adding a CAAX prenylation site at its C-terminus can override the requirement for Dlish. Both UAS-dachs and UAS-dachs-CAAX induced similar overgrowth when expressed in the posterior of the wing with hh-gal4 ( Figure 9A,B). In a dlish 4506 homozygous background, however, the overgrowth induced by Dachs overexpression was greatly reduced, while the overgrowth induced by Dachs-CAAX was undiminished; there was a similar difference in their effects on the Hippo pathway marker ban3-GFP (Figure 9C,D,G,H). These different activities correlated with different protein localizations. In dlish 4506 wing discs, overexpressed Dachs-CAAX was largely cortical ( Figure 9F), while overexpressed Dachs was largely cytoplasmic ( Figure 9E), contrasting with its cortical localization in wild type discs ( Figure 3H).

Dlish regulates total Dachs levels
Reduced Dlish activity not only affects the subcellular localization and activity of Dachs, but also the amount of Dachs in cells. dlish 04 or act-Gal4 UAS-dlish-RNAi increased total Dachs levels, as assessed by western blotting protein from imaginal discs and normalized for cell number and size using a-Tubulin ( Figure 10A; dlish-RNAi data in Figure 10-figure supplement 1A).
Loss of fat causes a similar increase in total Dachs levels in imaginal discs (Matakatsu, Blair and Fehon, Unpublished) ( Figure 10A). Yet dlish knockdown increases cytoplasmic Dachs, whereas in fat mutants the increase is subapical. To explain this, we hypothesize that as Dlish targets or stabilizes Dachs in the subapical cell cortex, it places Dachs in or near complexes where it can also be destabilized by Fat (Figure 1). When Fat is lost destabilization no longer occurs, so the total and subapical levels of Dachs increase. When dlish is reduced, the Fat-dependent destabilization of Dachs cannot occur but subapical targeting is also lost, and so Dachs accumulates in the cytoplasm.
If the loss of Dlish increases Dachs levels by blocking Fat-mediated destabilization, then it will reduce the sensitivity of total Dachs levels to the presence or absence of fat; the effects of removing both Fat and Dlish will not be additive. Our data agree with this prediction. In  Figure 2A), as is subapical Dachs (A 2 ,A 3 ) and the Yorkie activity reporter ban3-GFP (B 1 ,B 3 ). (C) fat fd dlish 04 double mutant wing is abnormally patterned but nearly-normally sized when compared with the green outline of a control wing from Figure 7A. Unlike the 100% lethality of fat fd , 50-60% of fat fd dlish 04 survive to adulthood (44/230 from selfcrossed dlish 04 fat fd / CyO-TM6,Tb vs. 77/230 expected from self-crossed non-lethal / CyO-TM6,Tb). DOI: 10.7554/eLife.16624.018 trials using the hypomorphic dlish 04 background, simultaneous loss of fat caused no or only a weak additional increase in total Dachs levels compared those caused by dlish 04 alone; the Dachs increase in dlish 04 fat fd double homozygotes was similar to that seen in fat fd mutants, and significantly weaker than the increase expected if the fat and dlish effects were additive ( Figure 10A). Similar results were obtained in a smaller number of trials combining fat loss with UAS-dlish-RNAi driven by tubulin promoter (tub)-gal4 or actin promoter (act)-gal4 (Figure 10-figure supplement 1A).
While screening for candidates that might also regulate Dachs levels, we found that Dlish can co-IP in vitro with the Fat-regulated E3 ubiquitin ligase Fbxl7, the related F-box protein Slimb and their binding partner Cullin1 ( Figure 10B). Loss of Fbxl7 increases the levels of subapical Dachs in imaginal discs Rodrigues-Campos and Thompson, 2014), and we found a similar effect on subapical Dlish ( Figure 10C). Homozygous slimb 1 clones also slightly increased subapical Dachs and Dlish ( Figure 10D,E), as well as the known Slimb target Expanded (Ribeiro et al., 2014;Zhang et al., 2015). Dlish and Fat-binding ubiquitin ligases therefore help regulate the levels and localization of the Dlish-Dachs complex, providing one route for Fat's regulation of pathway protein levels (see Discussion).

Approximated binds and regulates Dlish
We next addressed how changes in Dlish might help direct or tether Dachs to the subapical cell cortex. As noted above we found that Dlish can co-IP the transmembrane protein App (Figures 5A,  11A). Dlish-App binding was confirmed in a GST pulldown of App produced by S2 cells, and requires the Dlish N-terminus containing the first SH3 domain-( Figure 11B). Since the second SH3 domain of Dlish binds Dachs ( Figure 4F), Dlish could bridge App and Dachs using different domains.
This was particularly intriguing given the subapical localization of App, and the similarity between the effects of dlish and app knockdown on Dachs localization, levels and activity in wild type and fat mutant tissue (see Introduction) (Matakatsu and Blair, 2008;Matakatsu et al., Unpublished). This raised the possibility that App acted through Dlish and thus the Dlish-Dachs complex. In support of this hypothesis, we found that clonal loss of app in fat mutant wing discs decreased subapical and increased cytoplasmic Dlish; loss of app in wild type discs also increased cytoplasmic Dlish ( Figure 11C-E). Dlish and App also showed a suggestive genetic interaction: while overexpression of App slightly reduced wing growth in wild type wings, it significantly strengthened the overgrowth caused by overexpression of a weakly active untagged Dlish ( Figure 11F-J). Overexpression of app has a similar effect on the overgrowth of dachs-overexpressing wing discs (Matakatsu and Blair, 2008). Thus App increases the effectiveness of the Dlish-Dachs complex.
As a subapical transmembrane protein, App could be acting as a physical tether for Dlish and Dachs; however, App is also a DHHC palmitoyltransferase, and so might palmitoylate and help anchor the Dachs-Dlish complex near the membrane. Dachs is not detectably palmitoylated in vitro or in vivo (Matakatsu and Blair, 2008;Matakatsu, Blair and Fehon, Unpublished). We therefore directly examined the palmitoylation of Dlish using the Acyl Biotinyl Exchange (ABE) method, which uses N-ethylmaleimide to block free cysteines, hydroxylamine (HAM) to break palmitoyl-thioester bonds, and sulfhydryl-reactive biotin to label those bonds (Drisdel et al., 2006;Drisdel and Green, 2004). We detected HAM-dependent biotinylation of Dlish in S2 cells and, as a positive control, App itself (Matakatsu, Blair and Fehon, Unpublished), but not of the negative control GFP ( Figure 12A). Moreover, HAM-dependent biotinylation of Dlish was increased by co-expression with App (13 of 14 repeats; examples are shown in Figure 12A,C, and averages and significance are shown in Figure 12D). This strongly argues that a substantial fraction of the biotinylation represents App-sensitive palmitoylation.
App can also bind to the ICD of Fat (Matakatsu, Blair and Fehon, Unpublished). We have confirmed binding in S2 cells, and found that App can co-IP not only with the domains C-terminal to the Hippo domains (FatDECDDN-6), but also the domains most strongly active in Dachs and Hippo pathway regulation (FatDECD-PH+Hippo) ( Figure 12B). Like Dlish, App failed to bind to FatDECDD4-C and FatDECD-Hippo, which have, respectively, no or reduced Hippo activity in vivo (Matakatsu and Blair, 2012). These results raise the possibility that Fat binding might regulate the activity of App on substrates like Dlish.
We therefore tested whether the addition of FatDECD could alter App-stimulated Dlish palmitoylation in S2 cells, and found a marked decrease compared with the addition of App alone (6 of 6 repeats; example in Figure 12C; averages and significance in Figure 12D and Figure 12source data 1). This result is consistent with the model that Fat reduces subapical Dachs by interfering with or modifying the interaction between App and Dlish, reducing App's ability to palmitoylate Dish and tether the Dlish-Dachs complex.

Discussion
The unconventional myosin Dachs is an important effector Fat/Ds-regulated Hippo signaling, as its heightened subapical levels in fat mutants inhibit and destabilize Warts, freeing Yorkie to increase the expression of growth-promoting genes Rauskolb et al., 2011;Vrabioiu and Struhl, 2015). We used a structurefunction analysis of Dachs as a springboard to search for new binding partners that are critical for Dachs localization and function, and have found Dlish (CG10933), a novel SH3 domain protein. Dlish binds directly to the Dachs C-terminus, and loss of Dlish disrupts Dachs localization, levels and function: subapical accumulation of Dachs is reduced and cytoplasmic and total levels increase, both in wild type and fat mutant tissue, while activity is lost. Importantly, Dlish is regulated by Fat, as loss of Fat greatly increases Dlish levels in the subapical cell cortex and, like Dachs, Dlish is needed for much of the fat mutant overgrowth.
Dlish also binds the ICD of Fat and other Fatbinding proteins, including two that likely mediate part of its function: the palmitoyltransferase App and the F-box protein Fbxl7. Thus Dlish provides a new biochemical link from the Fat ICD to Dachs regulation. Our evidence indicates that Dlish plays two different and opposing roles (Figure 1). First, it helps tether Dachs in the subapical cell membrane, in part via Fat-regulated, Appdependent palmitoylation, so that Dachs can more efficiently inhibit Warts. Second, it links Dachs to Fat-organized machinery for Dachs destabilization, including Fbxl7, and thus helps reduce Dachs levels.

Dlish, App and subapical tethering
Dlish and Dachs cooperate to target or tether the Dlish-Dachs complex, as each is necessary, and to a weaker extent sufficient, for the subapical concentration of the other. The Dachs contribution is likely through tethering the complex to the cortical cytoskeleton, as we found that loss of the F-Actin-binding myosin head blocks the subapical localization of Dachs. This would agree with recent biochemical analyses that suggest that Dachs has no motor function, acting rather as an F-Actinbinding scaffolding protein (Cao et al., 2014).
The Dlish contribution, on the other hand, depends at least in part on its ability to bind the transmembrane DHHC palmitoyltransferase App. Loss of App (Matakatsu and Blair, 2008;Matakatsu, Blair and Fehon, Unpublished) and Dlish have very similar effects on Dachs localization and activity. We found that loss of App disrupts the subapical accumulation of Dlish in vivo, and that App can stimulate palmitoylation of Dlish in vitro. Thus, palmitoylation of Dlish likely stimulates membrane association of both Dlish and its binding partner Dachs.
App also has additional effects on Fat pathway activity. First, App has palmitoyltransferase-independent activity and can co-IP Dachs in vitro (Matakatsu, Blair and Fehon, Unpublished). Thus, while palmitoylation of Dlish may mediate some of App's activity, subapical App may simultaneously help localize the Dlish-Dachs complex by physical tethering. And while both palmitoylation and tethering of the Dlish-Dachs complex is likely critical for the fat mutant phenotype, App also has a function that depends on the presence of Fat, as App can bind, palmitoylate and inhibit the activity of Fat's ICD (Matakatsu, Blair and Fehon, Unpublished).
An important question is whether the absence of Fat regulates the App-dependent tethering of the Dlish-Dachs complex. The Fat ICD can complex with both Dlish (this study) and App (Matakatsu, Blair and Fehon, Unpublished). We found that Dlish and App can bind not only the C-terminal region of the Fat ICD where Fat is palmitoylated, but also the PH and Hippo domains which we showed played the strongest role in Dachs regulation. An attractive mechanism is that Fat inhibits the interaction between App and Dlish, reducing App's ability to palmitoylate and tether the Dlish-Dachs complex. In the absence of Fat, App and Dlish are freed to tether Dachs, and Dachs now inhibits and destabilizes Warts, causing overgrowth ( Figure 1). In support of this model, we found that overexpression of Fat's ICD in vitro can reduce App-stimulated palmitoylation of Dlish.

Dlish and the regulation of Dachs levels
Our evidence further indicates that Dlish targets Dachs for Fat-dependent destabilization. Loss of Fat increases not only subapical Dachs, but also total Dachs levels (Matakatsu, Blair and Fehon, Unpublished), a result we have confirmed. In the presence of Dlish the increased Dachs remains subapical. Loss of Dlish also increases the total levels of Dachs, but now that increase is cytoplasmic, and much less effective at inhibiting Warts. These Dachs increases are unlikely to have independent causes, as they are not additive; total Dachs levels are similar after loss of Fat, Dlish or both.
We propose that in wild type cells there is a flux of the Dachs-Dlish complex from the cytoplasm to the subapical cell cortex, where a Fat-dependent complex destabilizes Dachs (Figure 1). Normally the tethering effects of Dlish predominate over the Fat-dependent destabilization, and moderate levels of subapical Dachs are maintained. Destabilization is lost without Fat; this combines with Dlish-mediated tethering to increase subapical Dachs. Without Dlish the subapical tethering of Dachs is disrupted, access to Fat-dependent destabilization is lost and the now cytoplasmic Dachs increases. The model thus explains why the excess, largely cytoplasmic Dachs caused by reduced Dlish function is not greatly influenced by the presence or absence of Fat. (E 1 -E 6 ) app clones marked by absence of GFP in a fat mutant wing disc. Subapical Dlish levels decrease (E 2 ) and basolateral, cytoplasmic levels increase (E 5 ), while control DE-cadherin is unchanged (E 3 ,E 6 ). (F-J) Posterior, hh-gal4-driven expression of UAS-app decreases the wing area in a wild type wing, but increases the overgrowth induced by UAS-dlish (untagged). Brackets show p<0.01 by a single-tailed Student's T and Whitney-Mann tests; data and analysis shown in Figure 11-source data 1. DOI: 10.7554/eLife.16624.023 The following source data is available for figure 11: Source data 1. Data and analysis of wing size change after hh-gal4-driven expression of UAS-app, UAS-dlish (untagged), or both, for Figure 11J. DOI: 10.7554/eLife.16624.024 In addition to any effects caused by changing the subcellular localization of Dachs, Dlish may also provide a direct link to the machinery for protein ubiquitination, as Dlish can co-IP with the E3 ubiquitin ligase Fbxl7, as well as the related F-box ubiquitin ligase Slimb. Fbxl7 is particularly intriguing, as it binds to and is regulated by Fat's ICD, and reduces subapical Dachs, perhaps via ubiquitination Rodrigues-Campos and Thompson, 2014). Slimb can bind and ubiquitinate Expanded (Ribeiro et al., 2014;Zhang et al., 2015), a subapical regulator of Hippo signaling with links to Fat and Dachs function (Irvine and Harvey, 2015). But while we found that loss of Fbxl7 or Slimb increases subapical Dachs and Dlish, these effects are weak (see also Bosch et al., 2014;Matakatsu, Blair and Fehon, Unpublished), and the large increase in total Dachs levels caused by loss of Dlish or Fat must involve additional partners. The following source data is available for figure 12: Source data 1. Data and analysis of ABE assays for Figure 12D. DOI: 10.7554/eLife.16624.026

Fat, Dachs and Dlish in other taxa
Mutations in Fat's closest mammalian homolog Fat4 (FatJ) and its Ds-like ligands strongly disrupt PCP-like processes, and have in humans been associated with the multisystem defects of Hennekam and Van Maldergem syndromes (Alders et al., 2014;Mao et al., 2011;Saburi et al., 2008;Zakaria et al., 2014). There has been some debate, however, about whether the mammalian proteins retain direct regulation of Hippo activity (Bagherie-Lachidan et al., 2015;Bossuyt et al., 2014;Pan et al., 2013;Sadeqzadeh et al., 2014;Kuta et al., 2016). Nonetheless, Fat4 has been linked to Hippo changes in both normal development and tumors (Qi et al., 2009;Van Hateren et al., 2011), mutations in Fat4 or Dachsous1 change the balance of precursors and mature neurons in the developing neuroepithelium of both humans and mice, and the mouse defect can be reversed by knockdown of the Yki homolog Yap (Cappello et al., 2013). But the mechanisms underlying these effects are unknown, and Fat4 cannot regulate Hippo signaling in Drosophila (Bossuyt et al., 2014).
It is therefore important to note that while homologs of Dachs and Dlish are found throughout the animal kingdom, they are apparently absent from vertebrates. This suggests that the Dachs-Dlish branch of the Fat-Ds pathway, with its powerful effect on Warts activity, is also lacking. Nonetheless, it has been suggested that Drosophila Fat and Ds can affect Hippo pathway activity in a Dachs-independent manner (Degoutin et al., 2013;Gaspar et al., 2015). It is also clear that Drosophila Fat has Dachs-independent effects on PCP; indeed the N-terminal 'PCP' domain of the Fat ICD that did not affect Dachs in this study is sufficient to improve the PCP defects of fat mutants (Matakatsu and Blair, 2012). These or alternative pathways may still be present in mammals.

Quantification of wing sizes
Wings from adult females for each genotype were mounted on glass slides in mineral oil, and images were measured using NIH ImageJ.