Localization of Hippo signalling complexes and Warts activation in vivo

Hippo signalling controls organ growth and cell fate by regulating the activity of the kinase Warts. Multiple Hippo pathway components localize to apical junctions in epithelial cells, but the spatial and functional relationships among components have not been clarified, nor is it known where Warts activation occurs. We report here that Hippo pathway components in Drosophila wing imaginal discs are organized into distinct junctional complexes, including separate distributions for Salvador, Expanded, Warts and Hippo. These complexes are reorganized on Hippo pathway activation, when Warts shifts from associating with its inhibitor Jub to its activator Expanded, and Hippo concentrates at Salvador sites. We identify mechanisms promoting Warts relocalization, and using a phospho-specific antisera and genetic manipulations, identify where Warts activation occurs: at apical junctions where Expanded, Salvador, Hippo and Warts overlap. Our observations define spatial relationships among Hippo signalling components and establish the functional importance of their localization to Warts activation.

H ippo signalling is a conserved signal transduction pathway that controls organ growth and cell fate throughout the metazoa [1][2][3] . Hippo signalling is regulated by diverse inputs and subject to extensive crosstalk by other pathways, but major factors regulating the pathway include cellular organization and contacts with neighbouring cells. Thus, disruption of apicalbasal polarity in epithelial cells inactivates Hippo signalling 4 , whereas contact inhibition is associated with high levels of Hippo signalling 5 . Genetic and biochemical studies have identified many components of Hippo signalling and characterized their potential for interactions, but our understanding of the normal cellular organization and dynamics of the Hippo pathway in vivo has remained relatively poor.
Hippo signalling regulates growth by inhibiting a transcriptional co-activator protein, Yorkie (Yki) 6 . Yki and its mammalian homologues are crucial and potent promoters of growth, consequently disruption of Hippo signalling is associated with many cancers 7 , whereas excess Hippo pathway activity inhibits growth and induces apoptosis. The key direct regulator of Yki activity is the kinase Warts (Wts): phosphorylation of Yki by Wts inactivates Yki by restricting it to the cytoplasm 8,9 . A complex network of regulatory proteins converge to modulate Wts activity. Direct activators of Wts include Hippo (Hpo), a kinase that activates Wts, Salvador (Sav), a scaffolding protein that binds both Hpo and Wts, and Mob as tumour suppressor (Mats), a Wts co-factor ( Fig. 1a) 1,3 . Wts activation is promoted by cytoplasmic proteins associated with apical junctions, including Kibra, Merlin (Mer) and Expanded (Ex) [10][11][12][13][14] . Activated forms of Mer can bind to Wts, and it has been proposed that Mer activates Wts by recruiting it to membranes where it could be phosphorylated by Hpo 15 . However, inhibitors of Wts activity also localize to junctional membrane complexes, including Zyxin 16,17 , and Ajuba LIM protein (Jub) 18,19 . Relationships among the various junctional components of Hippo signalling have not been well defined, and where within the cell Wts activation occurs has not been determined.
We report here that Hippo pathway components in the Drosophila wing disc are organized into distinct complexes at apical junctions, including separate distributions for Sav, Ex, Wts and Hpo. These complexes are reorganized on Hippo pathway activation, when Wts shifts from associating with the Wts inhibitor Jub to the Wts activator Ex, and Hpo shifts from predominantly cytoplasmic to predominantly associating with Sav. We identify both increased expression of Wts and Ex, and increased Hpo activity, as contributing to this Warts relocalization. Using a phospho-specific antisera, we find that Wts is activated at sub-apical junctions where Ex, Sav, Hpo and Wts overlap. We also define the role of Ex within the Hippo pathway as functioning as a scaffold that recruits Wts to membrane locations where it interacts with and can be activated by Hpo. Our observations identify sites of Wts activation in vivo, define spatial relationships among Hippo signalling components and establish the functional importance of their localization to Wts activation.

Results
Organization of Hippo signalling components in wild type. Physical interactions among Hippo pathway components have been characterized mostly based on overexpression in cultured cells. However, in many cases, the extent to which such interactions occur in vivo remains unclear, as regulated localization to separate sites within the cell could prohibit interactions, whereas localization to shared sites could promote them. To assess the potential for regulated localization to modulate Hippo signalling in vivo, we compared the subcellular distributions of Hippo pathway components expressed at endogenous levels using antibodies or tagged genomic constructs for Crumbs (Crb), Ex, Wts, Echinoid (Ed), Sav, Hpo, Mats and Jub in wing disc epithelial cells, and then examined how their localization was affected by pathway activation.
Kibra, Mer and Ex interact genetically and exhibit overlapping apical distributions [10][11][12][13][14] ; we focused on Ex, as it is a major Hippo pathway activator in the wing disc, whereas Kibra and Mer have weaker effects. Ex localizes to apical junctions largely through association with the transmembrane protein Crb, which participates in homophilic interactions between cells; Crb and Ex exhibit extensive co-localization (Fig. 1c) [20][21][22] . Along the apical-basal axis, these proteins accumulate at a site known as the sub-apical region or marginal zone, which is just apical to the E-cadherin-containing adherens junctions ( Supplementary  Fig. 1a) 23 . Within the plane of the sub-apical region, Crb and Ex localization are discontinuous, with small gaps in their membrane accumulation around each cell ( Fig. 1c; Supplementary Fig. 1a). Sav localizes to apical junctions largely through association with the transmembrane protein Ed, which also participates in homophilic interactions between cells, and Ed and Sav co-localize in wing discs (Fig. 1d) 24 . Ed-Sav complexes overlap Crb-Ex complexes ( Supplementary Fig. 1a), but have a broader distribution within the horizontal plane, as there are no obvious gaps in their distribution (Fig. 1d), and they also partially overlap the more basal adherens junction complexes ( Supplementary  Fig. 1a). Wts co-localizes with the Wts inhibitor Jub (Fig. 1f), which is responsible for recruiting Wts to adherens junctions 18 . In late third instar wing discs, Jub-Wts complexes were B0.8 mm basal to Crb-Ex complexes (Fig. 1e). Wts and Jub exhibit a punctate distribution within the plane of the adherens junctions, which is distinct from the Crb-Ex distribution (Fig. 1e). Finally, Hpo localization is predominantly cytoplasmic, but appears slightly elevated at apical junctions 25 , in a punctate distribution, partially overlapping Sav ( Fig. 1g; Supplementary Fig. 1b,c). The co-localization of Ex with Crb, Sav with Ed, and Wts with Jub was confirmed by calculation of Pearson's correlation coefficient for each of these pairs of interacting proteins at apical junctions (Fig. 1h). Altogether, this characterization identified four distinct localization profiles for Hippo pathway components: Crb-Ex, Ed-Sav and Jub-Wts complexes at distinct apical junctions, and a distribution of Hpo that includes both broad cytoplasmic accumulation and a modest junctional accumulation overlapping Sav (Fig. 1b).
Detection of Wts activation in vivo. Although the cellular location of Wts activation has not been identified, activation of Wts is associated with Hpo-mediated phosphorylation at Thr1077, and an antisera that recognizes phospho-Thr1077 (pWts) on western blots has been reported 10 . Staining wild-type wing disc cells with this antisera revealed only nonspecific background staining, comprising both dispersed puncta and a Wts-independent mitotic stain ( Supplementary Fig. 2a). However, the observation that Wts co-localizes with an inhibitor, Jub, rather than with the activators Ex, Sav or Hpo, implies that most Wts protein in vivo is inactive. Thus, we wondered whether it might be possible to visualize pWts in vivo under conditions of elevated Hippo pathway activity. Hippo pathway activation inhibits Yki, thereby decreasing growth and increasing apoptosis, whereas Yki activation promotes growth and blocks apoptosis [1][2][3] . Yki also enhances expression of several upstream Hippo pathway components (Fig. 1a) 11,13 , including Ex, Mer and Kibra, and as we show here, Wts (Fig. 2b,     to activate Hippo signalling upstream of Yki (Fig. 1a), while simultaneously blocking downstream apoptosis. Thus, we expressed Yki containing a Ser-to-Ala mutation at the main Wts phosphorylation site (Yki S168A ), which renders it constitutively activated and insensitive to upstream Hippo pathway activation 8,9 . Because of lethality associated with highlevel activation of Yki, Yki S168A was expressed under UAS-Gal4 control in a conditional manner, by combining en-Gal4, which drives expression of UAS transgenes in posterior cells, with tub-Gal80 ts , which expresses a temperature-sensitive form of the Gal4 repressor Gal80. Examination of wing discs in which Yki S168A was expressed in posterior cells of wing discs revealed a clear, Wts-dependent, pWts signal above the background staining ( Fig. 2a; Supplementary Fig. 2c) Thus, under these conditions of Hippo pathway activation, activated Wts can be detected in vivo.

Relocalization of Wts and Hpo induced by pathway activation.
Since Wts normally co-localizes with an inhibitor rather than its activators, we wondered whether activation of Wts is associated with changes in the localization of pathway components.
Remarkably, under conditions of Wts activation induced by activated Yki, Wts shifts from overlapping with E-cadherin (Ecad) at adherens junctions to overlapping with Ex at Crb-Ex junctions (Fig. 2d,e). Expression of wild-type Yki, which does not increase Yki activity 26 , did not affect Wts localization ( Supplementary Fig. 2d). The shift in Wts localization induced by activated Yki was evident both in the more apical localization of Wts and in a more even distribution around the circumference of the cell, matching Ex, and was quantified by calculating Pearson's correlation coefficient, which confirmed both increased Ex-Wts co-localization and decreased E-cad-Wts co-localization ( Fig. 2f,g). Total levels of Wts protein, as assayed by western blotting, were also increased by Yki S168A (Fig. 2b), and this affect appears to be at least partially transcriptional because expression of -galactosidase from a wts-lacZ enhancer trap was also increased by Yki S168A (Supplementary Fig. 2b). A similar shift in Wts localization, from overlapping E-cad to a more apical location, could also be induced by Yki S168A expression in eye imaginal discs ( Supplementary Fig. 3a).
Examination of additional Hippo pathway components revealed that Hpo localization is also altered in Yki S168Aexpressing cells. Apical junction accumulation of Hpo was greatly increased (Fig. 2h), whereas total Hpo levels were not affected (Fig. 2b), indicating that Hpo shifts from a cytoplasmic towards an apical junction localization. The increased accumulation of Hpo at apical junctions overlapped Sav (Fig. 2h), but was less extensive, as some gaps in the Hpo distribution relative to Sav were observed (Fig. 2h). Thus, Wts and Hpo, the two key kinases of the Hippo signalling pathway, are both relocalized concomitant with Hippo pathway activation induced by Yki S168A (Fig. 1b).
To clarify the relationship between Hpo and Wts relocalization and Wts phosphorylation, we compared their distribution in cells expressing Yki S168A . pWts staining overlaps with Wts and Ex, although occasional sites of Wts accumulation without detectable pWts could be identified (Fig. 3a,b). pWts staining is also encompassed within apical Hpo accumulation (Fig. 3b), and simultaneous examination of pWts, Wts and Hpo revealed that pWts is detected at apical junctions where Hpo and Wts overlap (Fig. 3b). Thus, Wts is activated in vivo when and where the subcellular localizations of Hpo and Wts intersect. These results support a model in which pathway activation is controlled by regulating the subcellular localization of Wts and Hpo. We note that key accessory proteins, including Sav and Ex, also co-localize at these sites where pWts is detected, and could play an essential role in assembling a Wts activation complex.
Mats-Wts association is independent of Wts activation. The Wts co-factor Mats has been reported to localize to both the cytoplasm and adherens junctions 27 . It was also reported that forced membrane localization of Mats, achieved by expressing a myristoylated Mats construct, Myr:Mats:green fluorescent protein (GFP), could recruit Wts to membranes and increase Wts activation 27 . Thus, we investigated the relationship between Wts relocalization and activation, and Mats localization. As available Mats antisera did not yield specific staining in our hands, we created a genomic GFP:Mats transgene (Fig. 4a). The specificity of GFP:Mats expression was confirmed by its depletion using mats RNA interference (RNAi; Fig. 4b). In addition to a broad cytoplasmic distribution, GFP:Mats localizes to discrete puncta at adherens junctions, which overlap Wts puncta (Fig. 4c). Thus, Mats co-localizes with Wts even in its inactive (Jub-associated) state. Mats is recruited to adherens junctions by Wts, because GFP:Mats localization to junctional puncta was eliminated by wts RNAi (Fig. 4d). Conversely, Wts localization to adherens junctions does not require Mats (Fig. 4e). Relocalization of Wts induced by Yki S168A induced a concomitant relocalization of GFP:Mats to more apical junctions (Fig. 4f,g). Thus, we infer that in vivo Mats can associate with Wts at both inactive (Jub) and active (Ex) sites.
Assembling a Wts activation complex. The implication that Wts activation could be ascribed to a relocalization of Wts and Hpo that facilitates their interaction prompted us to examine the requirements for Hpo pathway components in this process. Since the Hpo that accumulates at the apical membrane overlaps with Sav, we used RNAi to knockdown sav in cells expressing Yki S168A . This severely diminished apical accumulation of Hpo in Yki S168A -expressing cells, indicating that this recruitment of Hpo is Sav dependent (Fig. 3c). This observation is consistent with cell fractionation experiments in cultured cells reporting an ability of Sav to promote membrane association of Hpo 15 . The modest apical accumulation of Hpo in wild-type cells was also Sav dependent ( Supplementary Fig. 3d). Thus, while most Hpo is normally cytoplasmic, Sav can recruit Hpo to apical junctions, and this recruitment can be enhanced by conditions that promote Hippo pathway activity.
Wts overlaps with Jub in wild-type but not in Yki S168Aexpressing cells, and Jub is not required for the relocalization of Wts ( Supplementary Fig. 3b,c). Conversely, relocalized Wts in Yki S168A -expressing cells overlaps extensively with Ex (Fig. 2e,g), and Ex was required for Wts relocalization: in cells co-expressing Yki S168A and ex-RNAi, Wts remained at adherens junctions rather than shifting to Crb-Ex junctions ( Fig. 3d; Fig. 2f). Knockdown of Ex also eliminated detectable pWts staining in Yki S168A -expressing cells (Supplementary Fig. 3e). Thus, Ex is required for both the relocalization and the activation of Wts induced by Yki S168A in wing disc cells. Conversely, knockdown of Mer did not detectably influence Wts relocalization ( Fig. 2f; Supplementary Figs 1d and 4a), consistent with Mer playing only a minor role in Wts activation in wing discs.
We also investigated requirements for Hpo, Mats and Sav in Wts relocalization. When Hpo, Mats or Sav levels were reduced by RNAi in Yki S168A -expressing cells, Wts relocalization was reduced, as revealed by its partial co-localization with both Ex and E-cad (Fig. 2f,g; Supplementary Fig. 4b-f). We also confirmed that Wts phosphorylation was severely diminished by sav RNAi, consistent with its role in localizing Hpo to apical junctions where Wts phosphorylation occurs ( Supplementary Fig. 3f). These results indicate that Hpo, Sav and Mats each contribute to the relocalization of Wts from Jub to Ex.   Wts-Ex binding and Wts activation. To assess whether the co-localization of Wts and Ex could be reflective of direct association, we assayed for physical interaction between Wts and Ex. Indeed, when tagged forms of Ex and Wts were co-expressed in cultured Drosophila S2 cells, we discovered that Ex could be co-precipitated with Wts, but not with a control protein (GFP), indicating that Wts and Ex can interact (Fig. 5a). This coprecipitation was increased approximately threefold when Hpo was co-transfected together with Ex and Wts, and was increased 410-fold when Hpo and Mats were both co-transfected with Ex and Wts (Fig. 5a). Thus, physical association between Wts and Ex is promoted by Hpo and Mats. To confirm that endogenous Wts and Ex can associate in vivo, we prepared lysates of wing discs expressing GFP:Wts (which is also Flag tagged). Significant coprecipitation was observed from lysates prepared from discs expressing Yki S168A , but not from wild-type wing disc lysates (Fig. 5c). Together with the genetic requirements for Ex, Mats, Sav and Hpo in Wts relocalization, these observations indicate that relocalization of Wts is associated with Wts-Ex binding, and further imply that increased binding between Wts and Ex induced by Hpo and Mats contributes to the relocalization of Wts observed in vivo in Yki S168A -expressing cells. Intriguingly, we observed a slight shift in Ex mobility when Hpo was cotransfected (Fig. 5a,b), suggesting that Hpo can phosphorylate Ex.

RNAi-sav
The correlation between Wts relocalization to Ex and Wts activation prompted us to further examine the role of Ex in promoting Wts activation. Detection of Wts activation in S2 cells by western blotting with pWts antisera required co-transfection of Hpo and Wts (Fig. 5b). Ex then further stimulates Wts activation, because addition of Ex increased pWts, both in the absence or presence of exogenous Mats (Fig. 5b). This stimulation of Wts activation could be explained by the ability of Ex to bind to both Hpo 10 and Wts (Fig. 5a), thereby acting as a scaffold to link them together.

Wts or Ex overexpression induces pWts in vivo.
Expression of Yki S168A could potentially upregulate multiple Hippo pathway components that contribute to relocalization and phosphorylation of Wts in vivo. Since activation of Yki increased Wts levels, and overexpression of Wts is sufficient to increase Hippo pathway activity, we investigated the influence of Wts overexpression on its localization and phosphorylation. When Wts was transiently overexpressed using a UAS-Myc:Wts transgene under the control of en-Gal4 and tub-Gal80 ts , it could be detected in the cytoplasm, but a distinct sub-apical accumulation was visible, where it overlaps both E-cad and Ex (Fig. 6a). Quantitation revealed similar increases in Wts induced by UAS-Yki S168A and UAS-Myc:Wts ( Supplementary Fig. 2e). Weak pWts staining was detected overlapping this accumulation of Myc:Wts at apical junctions (Fig. 6c); hence, simply increasing Wts levels is sufficient to enable detection of pWts in vivo.
Recently, it was reported that myristoylation of Wts enhanced its activation, presumably through membrane targeting 15   Interestingly, when expressed in wing disc cells, Myr:V5:Wts was not evenly distributed on membranes as expected for myristoylated proteins, but accumulated at higher levels at apical junctions, overlapping E-cad and Ex ( Supplementary  Fig. 5a). In contrast, a different myristoylated protein, Myr:Mats:GFP 27 , was evenly distributed on cellular membranes ( Supplementary Fig. 5e). These observations suggest that interaction with Jub and Ex exerts a greater influence on Wts localization than the myristoylation tag. Expression of Myr:V5:Wts was also sufficient to induce pWts staining, and pWts accumulation above background was only observed at apical junctions, rather than along all membranes ( Supplementary Fig. 5c). As an additional control for pWts staining specificity, we took advantage of a construct expressing a myristoylated Wts protein with a mutation at the Hpo phosphorylation site (Myr:V5:Wts T1077A ) 15 ; this protein accumulated in a similar pattern as wild-type Myr:V5:Wts, but no specific pWts staining was detected ( Supplementary  Fig. 5b,d). These overexpression experiments with wild-type or myristoylated Wts imply that visible Wts activation only happens at a discrete junctional site where Wts overlaps with Wts activators. Indeed, co-expression of Ex with Wts resulted in a more robust activation of Wts than overexpression of Wts alone (Fig. 6c-e). Quantitation revealed that the increase in Ex by UAS-Yki S168A was two-to threefold greater than that induced by UAS-ex ( Supplementary Fig. 2e).
To investigate whether other means of activating the Hippo pathway also could be sufficient for detection of Wts relocalization or phosphorylation, we transiently expressed Ex, Hpo, Sav or Myr:Mats for 24 h, using en-Gal4 and Gal80 ts . These reduced total levels of Wts ( Fig. 6b; Supplementary Fig. 5f-h), consistent with the discovery that Hippo signalling regulates Wts levels.
Some detectable Wts remained at adherens junctions, but this Wts overlaps E-cad rather than Ex, and pWts staining was not detectable. Possibly, the reduced levels of Wts prevented us from detecting pWts. Alternatively, as all signal transduction pathways require mechanisms to reverse pathway activation, the lack of detectable pWts staining without Wts overexpression might reflect a rapid turnover of pWts (for example, via protein degradation or dephosphorylation), such that pWts normally cannot accumulate to high enough levels for detection using available antisera. Conversely when Wts is overexpressed (by direct overexpression or by Yki activation), the machinery dedicated to turnover of pWts might become saturated, allowing pWts to accumulate to detectable levels. Thus, we reasoned that it might be possible to transiently identify pWts production within a short time window after Hippo pathway activation. Indeed, we found that a faint pWts signal could be detected 12 h after induction of Ex expression (Fig. 6f). Thus, pWts can be detected in vivo at Ex sites either after increasing Wts or increasing Ex, and a robust pWts signal is visible when both Wts and Ex are increased, either by co-expression or as a consequence of transcriptional upregulation by Yki.

Discussion
Wts is a key control point within the Hippo pathway, where multiple upstream regulatory processes converge 1-3 . A fundamental gap in our understanding of Hippo signal transduction has been the cellular location of Wts activation. Our observations here establish that Wts activation in wing disc epithelial cells occurs at sub-apical junctions where Hpo, Sav, Ex and Wts overlap. They further implicate co-recruitment of Hpo and Wts kinases to a common scaffold as a central feature of Wts from lysates of wing discs from wild-type, GFP:Wts and GFP:Wts en-Gal4 UAS-RFP UAS-Yki S168A tub-Gal80 ts shifted to 29°C for 24 h before dissection. This blot is representative of three independent biological replicates. Numbers to the right of blots indicate molecular weights.  Our studies indicate that a key step in Wts activation in disc epithelia is its relocalization from Jub to Ex (Fig. 7). No special mechanism is needed to transport Wts from Jub to Ex, as Wts localization could simply be governed by equilibrium binding with a limited cytoplasmic pool. That is, if Wts normally binds relatively strongly to Jub, and relatively weakly to Ex, it could, depending on its concentration, accumulate at Jub sites but not at Ex sites. Expression of activated Yki induced a robust relocalization of Wts from Jub to Ex, and our studies identify three factors that contribute to the visible accumulation of Wts at Ex sites under these conditions. First (as discussed below), Yki activation appears to increase Hpo activity. We also found that hpo RNAi suppressed the relocalization of Wts from Jub to Ex, and that increased Hpo activity promotes Ex-Wts binding, as assayed by co-immunoprecipitation experiments. These observations are consistent with the hypothesis that Wts shifts from Jub sites towards Ex sites due to an increased Ex-Wts binding affinity induced by Hpo. Second, Yki activation increases levels of Ex, which under equilibrium binding would also increase the recruitment of Wts to Ex sites. The relocalization of Wts back to adherens junctions in the absence of Ex indicates that the shift in Wts localization is Ex dependent, and implies that Jub and Ex can compete for association with Wts. A third factor that contributes to detection of Wts-Ex co-localization is the increase in Wts protein levels induced by activated Yki, which could lead to Wts concentrations high enough to bind even lower-affinity Ex sites, and indeed we observed that simply overexpressing Wts was sufficient to induce Wts-Ex overlap, without removing Wts from adherens junctions where it co-localizes with Jub. We suggest that an additional consequence of increased Wts levels that enables detection of Wts and pWts overlapping Ex could be a saturation of pWts removal. While at present this remains speculative, all signal transduction pathways require mechanisms to turn off after they have been activated, so there should exist mechanisms that either degrade or dephosphorylate pWts. Relatively low levels of pWts due to rapid turnover could also help explain why pWts was undetectable in wild-type wing discs.

Myc:Wts
Our discovery of Ex-Wts binding, together with earlier studies that identified Ex-Hpo binding 10 , implicate Ex as a scaffold that could promote Wts activation by co-localizing it with Hpo, and thus define a role for Ex distinct from previous suggestions that it functions as an activator of Hpo. Similarly, recent studies in cultured cell models showed that activated forms of Mer could bind Wts, and suggested a model in which Mer promotes Wts activation by recruiting it to membranes where it could be activated by Hpo 15 . This suggests that in tissues where Mer, rather than Ex, plays key roles in Wts activation, such as glia 28 , Mer, which can also associate with Hpo, through Sav 10,28 , could play an analogous role in assembling a Wts activation complex 15 . It is thus noteworthy that the best characterized upstream branches of 'Hippo' signalling characterized in Drosophila: Fat, Ex and Mer can all now be said to act principally at the levels of Wts regulation rather than Hpo regulation. Moreover, we note that Kibra, which has been suggested to act at a similar point in the Hippo pathway as Mer and Ex [10][11][12] , has also been reported to be able to physically interact with both Hpo and Wts, and thus might also act principally as a scaffold that links them together rather than as a promoter of Hpo activation.
Indeed, external signals that impinge directly on Hpo activity have not yet been identified. Our discovery that Hpo localization to Sav is greatly increased by Yki activation reveals that regulators of Hpo localization exist, and implies that they are subject to negative feedback regulation downstream of Yki. As Hpo kinase activity can be promoted by Hpo dimerization 29,30 , we propose that the increased recruitment of Hpo to Sav could elevate Hpo activity by increasing its local concentration, and thereby its dimerization. Relocalization of Hpo might also affect its interactions with kinases that can modulate Hpo activity [31][32][33] . Recruitment of Hpo to Sav also concentrates Hpo near Ex, where it would more efficiently phosphorylate Ex-bound Wts. However, since most junctional Wts in disc epithelia is normally complexed with Jub rather than Ex, a mechanism-based solely on Hpo recruitment to apical junctions would not be expected to induce robust Wts activation. Importantly, then, our studies revealed that Hpo can increase Ex-Wts binding, possibly by phosphorylating Ex. Increased Ex-Wts binding would help recruit Wts to Ex, where it could then be phosphorylated by Hpo. Thus, it is now possible to suggest a sequential model for Hippo pathway activation in which Hpo is first recruited to membranes and activated, activated Hpo then phosphorylates Ex to recruit Wts and finally Hpo phosphorylates and activates Wts complexed with Ex. While further studies will be required to validate this model, it provides a framework that could guide future investigations, and our current studies clearly emphasize the importance of determining the in vivo localization of endogenous pathway components.

Methods
Drosophila genetics. Previously described tagged genomic transgenes were used to analyse localization of Crb, Wts or Jub expressed at endogenous levels 18 before dissection. The effectiveness of hpo, mats, Mer and sav RNAi is shown in Fig. 4b, Supplementary Figs 1c,d and 3d, other RNAi lines have been validated in earlier publications.
Immunofluorescence and image quantitation. Dissected wing discs were fixed in 4% paraformaldehyde for 8 min and stained using as primary antibodies rat anti-Ed  41 , and the distance between centroids of Ex and Wts puncta, was calculated using Volocity software (PerkinElmer). Horizontal images shown are projections through 10-20 sections, encompassing a slice of at least 2 mm, and include all of the apical junctions studied. The optical resolution of the Â 63 1.4 numerical aperture objective used is estimated to be B0.4-0.5 mm in the z axis with a 1-Airy unit pinhole, and hence sufficient to detect the estimated separation of 0.8 mm. Statistical analysis (t-tests for pairwise comparisons) was performed using Graphpad Prism software.