The Drosophila tumour suppressor Lgl and Vap33 activate the Hippo pathway through a dual mechanism

Deregulation of cell polarity and the epithelial-to-mesenchymal transition are hallmarks of human cancer (Hanahan and Weinberg, 2011; Muthuswamy and Xue, 2012; Stephens et al., 2018). Key regulators of cell polarity, Lethal (2) giant larvae [Lgl; also known as L(2)gl], Scribble (Scrib) and Discs large (Dlg; also known as Dlg1), were discovered in Drosophila as neoplastic tumour suppressors; these proteins regulate apical-basal cell polarity via their antagonistic interactions with the Par–aPKC and Crumbs (Crb) polarity complexes, and also limit cell proliferation (Stephens et al., 2018). Lgl functions to antagonise the activity of atypical protein kinase C (aPKC), and conversely aPKC phosphorylates and inhibits Lgl, in cell polarity regulation and tissue growth control (Stephens et al., 2018). In Drosophila, Lgl–aPKC plays a role in the control of tissue growth (cell proliferation and survival) that is distinct from their role in cell polarity regulation; lgl mutation or aPKC activation in eye epithelial tissue causes increased cell proliferation and survival without loss of apico-basal cell polarity (Grzeschik et al., 2007, 2010). Thus, the Lgl–aPKC axis regulates tissue growth independently of cell polarity effects.

The human Lgl orthologue, LLGL1, has a conserved function with Drosophila Lgl, as its expression in Drosophila rescues the tumorigenic defects of lgl mutants (Grifoni et al., 2004; Grzeschik et al., 2010). Reduced expression or mutations in LLGL1 (also known as HUGL1), are associated with hepatocellular carcinoma (Lu et al., 2009), malignant melanoma (Kuphal et al., 2006) and colorectal cancer (Schimanski et al., 2005). Similarly, aberrant localization or deletion of the second human Lgl homolog, LLGL2 (also known as HUGL2), is associated with gastric epithelial dysplasia and adenocarcinoma (Lisovsky et al., 2009), and with pancreatic intraepithelial neoplasia and ductal adenocarcinoma (Lisovsky et al., 2010). Mislocalization of LLGL1 and/or LLGL2 is also observed in other human cancers, including lung adenocarcinoma (Imamura et al., 2013) and ovarian cancer (Grifoni et al., 2007). As occurs in Drosophila, the mislocalization or dysfunction of LLGL1 and/or LLGL2 in cancer is also associated with altered aPKC localization or activity (Grifoni et al., 2007; Imamura et al., 2013).

We previously made the novel discovery that in Drosophila Lgl acts independently of its apico-basal cell polarity role to regulate the Salvador–Warts–Hippo (Hippo) negative tissue growth control pathway (Grzeschik et al., 2007, 2010). The core of the Hippo pathway involves the serine-threonine protein kinases Hippo (Hpo) and Warts (Wts), which respond to cell–cell contact and tissue architectural cues to control tissue growth via phosphorylating the co-transcriptional activator Yorkie (Yki), which regulates cell proliferation genes (e.g. Cyclin E) and cell survival genes (e.g. Diap1) (Grusche et al., 2010; Richardson and Portela, 2017). Lgl depletion or aPKC activation (which inhibits Lgl) impairs the Hippo pathway via mislocalization of the Hpo protein, away from the apical cortex, where it is normally activated by apical cues (Grzeschik et al., 2010; Parsons et al., 2014a). Thus, Lgl–aPKC controls the Hippo pathway by regulating Hpo localization and activity. In mammalian systems, deregulation of Lgl–aPKC impairs Hippo signalling and induces cell transformation, which mechanistically involves the association of aPKC with the Hpo orthologues, MST1 and MST2 (also known as STK4 and STK3, respectively), thereby uncoupling MST from the downstream kinases LATS1 and LATS2 (Wts in flies) and leading to increased nuclear YAP (Yki) activity (Archibald et al., 2015), consistent with what we observe in Drosophila (Grzeschik et al., 2010).

We have also shown that in Drosophila Lgl plays a novel regulatory role in ligand-dependent Notch signalling (Parsons et al., 2014b; Portela et al., 2015). Notch activation depends on the cleavage by Adam proteases to produce Notch^ext, and then processing by γ-secretase to produce the active form Notch^ICD (where ICD refers to the intracellular domain), which translocates to the nucleus to activate transcription of target genes, such as the E(spl) complex, as well as the cell proliferation and survival genes (Ntziachristos et al., 2014). We found that intracellular Notch accumulated in lgl mutant tissue, resulting in elevated Notch signalling that contributes to the overgrowth defects in lgl mutant tissue (Grzeschik et al., 2010; Parsons et al., 2014b; Portela et al., 2015). In mouse neural development, Lgl1 knockout causes elevated Notch signalling, which is associated with hyperproliferation and differentiation defects (Klezovitch et al., 2004), whereas in zebrafish, Lgl1 knockdown in the developing retina leads to elevated Notch signalling and neurogenesis defects, which are rescued by blocking Notch activity (Clark et al., 2012).

Intriguingly, we recently found that lgl mutant tissue exhibited increased LysoTracker incorporation (Parsons et al., 2014b; Portela et al., 2015), indicating that there is elevated vesicle acidification due to the activity of the Vacuolar-ATPase (V-ATPase). In the Notch signalling pathway, γ-secretase activity is dependent on vesicle acidification regulated by V-ATPase activity (Kobia et al., 2014; Vaccari et al., 2010). Therefore, in lgl mutant tissue the increased V-ATPase activity elevates γ-secretase activity and Notch cleavage, forming Notch^ICD, and leading to elevated Notch target gene expression. Consistent with this, genetically or chemically reducing vesicle acidification or V-ATPase function reduces the elevated Notch signalling in lgl mutant tissue (Parsons et al., 2014b; Portela et al., 2015). Thus, elevated Notch signalling in lgl mutant tissue is due to increased vesicle acidification and elevated γ-secretase activity.

To identify novel proteins that link Lgl to the V-ATPase, we undertook affinity purification-mass spectrometry (AP-MS) analysis of Lgl in Drosophila S2 tissue culture cells (Portela et al., 2018). Lgl interacted with aPKC and Par6 with high significance as expected. Among the novel interactors of Lgl that bound at high significance, the standout protein involved in endocytosis was the VAMP-(v-SNARE)-associated protein, Vap33, which is an orthologue of human VAPA and VAPB (Portela et al., 2018). Vap33 (VAPA/B) physically and genetically interacts with endocytic regulators and is involved in endo-lysosomal trafficking, with mutations in Drosophila Vap33 and human VAPB resulting in endocytic defects, including the accumulation of the early endosome Rab5 marker (Sanhueza et al., 2015), a phenotype we also observed in Drosophila lgl mutant tissue (Parsons et al., 2014b; Portela et al., 2015). We confirmed the binding of Vap33 to Lgl by co-immunoprecipitations (co-IPs) from S2 cells and in vivo in Drosophila epithelial cells (Portela et al., 2018) by using proximity ligation assays (PLAs) (Soderberg et al., 2006). Comparison of our data with the global Drosophila proteomics network has revealed that Vap33 and Lgl form a network with V-ATPase subunit proteins (Guruharsha et al., 2011; Portela et al., 2018). Interaction with human VAPA, VAPB and V-ATPase proteins is also evident in the human proteome (Huttlin et al., 2015; Portela et al., 2018). We also find that Vap33 overexpression rescues the elevated Notch signalling in lgl mutant eye epithelial clones and the lgl mutant adult eye phenotype, whereas knockdown of Vap33 enhances these eye defects (Portela et al., 2018). Vap33 overexpression also reduces V-ATPase activity, as assayed by LysoTracker levels (Petzoldt et al., 2013; Portela et al., 2018). Moreover, in lgl-knockdown S2 cells and in lgl mutant tissue, the interaction between Vap33 and V-ATPase components is decreased (Portela et al., 2018). Thus, Lgl binds to and facilitates the binding of Vap33 to V-ATPase components, which inhibits V-ATPase activity, thereby controlling vesicle acidity, γ-secretase activity and Notch signalling.

Although our previous studies have dissected how Lgl regulates the Notch pathway, it is currently not known precisely how Lgl regulates the Hippo pathway. Thus, in this study, we focused on the mechanism of this regulation. We show that the V-ATPase inhibits the Hippo pathway and conversely that Vap33 activates the Hippo pathway. Mechanistically, Vap33 is connected to the Hippo pathway by interacting with the cytoskeletal regulators, RtGEF (Pix), Git and Arf79F (Arf1), which bind to the Hpo protein kinase. Our findings are consistent with a model whereby Lgl–Vap33 promotes Hippo signalling via a dual mechanism through interaction with RtGEF, Git and Arf79F, and by inhibiting the V-ATPase.

Our previous studies have revealed the involvement of Lgl in the negative regulation of the V-ATPase (Portela et al., 2018), which is an important regulator of the Notch pathway (Kobia et al., 2014; Vaccari et al., 2010). Given that the lgl mutant adult eye phenotype (which is due to both Notch and Hippo pathway deregulation) is suppressed by reducing V-ATPase levels [more so than individually inhibiting Notch signalling or reducing Yki and Scalloped (Sd, a TEAD family transcription factor) activity (Grzeschik et al., 2010; Parsons et al., 2014b; Portela et al., 2015)], we suspected that the V-ATPase activity might be involved in the regulation of the Hippo pathway. The V-ATPase comprises several subunits, and knockdown of any of these proteins reduces V-ATPase function (Collins and Forgac, 2020; Dow, 1999; Dow et al., 1997; Eaton et al., 2021). Indeed, reducing V-ATPase levels and activity by using a Vha68-2 RNAi line resulted in higher Hippo pathway activity as assessed by reduced Yki nuclear staining, and lower expression of the Yki target, expanded (ex)-lacZ (expanded gene promoter driving expression of lacZ) (Fig. 1B) compared to the wild-type control (Fig. 1A, quantified in E). Conversely, knockdown of the core Hippo pathway gene, wts, resulted in impaired Hippo pathway signalling, as assessed by increased Yki nuclear staining and Yki target gene expression (ex-lacZ) (Fig. 1C), and a similar level of pathway impairment was observed upon activating the V-ATPase by overexpressing another V-ATPase subunit, namely Vha44 (Petzoldt et al., 2013) (Fig. 1D, quantified in E), which has been previously shown to elevate V-ATPase activity (Petzoldt et al., 2013). Overexpression of Vha44 has been previously shown to activate the JNK signalling pathway; however, in the eye epithelium JNK activity does not block Hippo signalling but instead is involved in activating Hippo (Doggett et al., 2011; La Marca and Richardson, 2020). Thus, the effect observed upon Vha44 overexpression on the elevation of the Yki target, ex-lacZ, occurs in the background of the elevated JNK-mediated inhibition of Yki activity. From these results we conclude that the V-ATPase negatively regulates the Hippo signalling pathway, leading to the upregulation of Yki activity.

Given that we have previously shown that Lgl functions with Vap33 in negatively regulating the V-ATPase (Portela et al., 2018), we then tested whether Vap33 is also involved in Hippo pathway regulation. Here, we used the canonical Yki target, Diap1 (Harvey and Tapon, 2007), to assess Hippo pathway activity. Although Jak-STAT signalling has been shown to induce Diap1 expression in the wing disc during development (Recasens-Alvarez et al., 2017), expression profiling after activation of the Jak-STAT signalling in the eye epithelium did not identify Diap1 as a target (Flaherty et al., 2009). Additionally, there are no reports that Lgl depletion in eye disc clones elevates Jak-STAT signalling (Stephens et al., 2018), but instead loss of cell polarity in mutant clones of the apico-basal polarity gene, scrib, in the eye disc results in expression of the Jak-STAT pathway ligand Upd and non-cell autonomous induction of Jak-STAT signalling in the surrounding wild-type cells (Bunker et al., 2015; Fahey-Lozano et al., 2019; Schroeder et al., 2013). Thus, Diap1 expression is a reliable reporter of Hippo pathway activity in the eye epithelium. As we have previously reported (Grzeschik et al., 2010), in lgl mutant clones the levels of the Hippo pathway target protein Diap1 were increased (Fig. 2A, quantified in G) and a distorted adult eye phenotype was observed (Fig. 2B). We found that when Vap33 was overexpressed in clones, the expression of Diap1 was reduced (Fig. 2C, quantified in G) and adult eyes appeared only slightly disorganized (Fig. 2D), consistent with Yki activity being reduced. Importantly, overexpression of Vap33 in lgl mutant clones resulted in a reduction of the elevated Diap1 protein levels (Fig. 2E) that were observed in lgl mutant clones (Fig. 2A) to below wild-type levels, similar to what is seen upon Vap33 overexpression alone (Fig. 2C, quantified in Fig. 2G), and rescued the lgl mutant mosaic distorted adult eye phenotype towards that of the Vap33 overexpressing mosaic adult eye (Fig. 2F compared with B and D), indicating that Vap33 is epistatic to Lgl. Taken together, these results show that Vap33 activates the Hippo pathway and that Vap33 is epistatic to lgl impairment in the activation of the Hippo pathway.

Given that our results implicated Vap33 in the regulation of the Hippo pathway, we sought to determine whether this occurs via protein–protein interactions. Interestingly, the global human proteomics analysis has elucidated the protein interaction network of the human Vap33 orthologues VAPA and VAPB, and revealed amongst the interacting proteins the Hpo orthologues MST1 and MST2 (Huttlin et al., 2015). To determine whether Drosophila Vap33 also interacts with Hpo we conducted affinity purification-mass spectrometry (AP-MS) on endogenously expressed Vap33–YFP in vivo using immunoprecipitation with a GFP nanobody (Neumuller et al., 2012). We decided to use an in vivo approach because the common Drosophila cell line (S2 cells, thought to be derived from haemocytes) used for protein interaction analysis are not polarized cells, and by using Drosophila tissues, we hoped to reveal physiologically relevant interactors important in polarized epithelial tissues. We prepared proteins from Vap33–YFP-expressing embryos and conducted the experiment in triplicate together with controls. As expected, among the AP-MS Vap33 interactors, we observed the V-ATPase component Vha68-2 as a medium confidence interactor [significance analysis of interactome (SAINT) score ∼0.77] and three other V-ATPase components as lower confidence interactors (Table S3). Surprisingly, Lgl and Hpo were not detected, suggesting that under the conditions used in embryonic cells these proteins do not form strong interactions with Vap33. However, pertinently, with respect to the Hippo pathway, we detected as high confident interactors (SAINT scores ∼1), the Hpo-interacting proteins, RtGEF (also known as Pix; a Rho-type guanine nucleotide exchange factor) and Git (G protein-coupled receptor kinase interacting ArfGAP), are actin cytoskeletal regulators that form a protein complex (Frank and Hansen, 2008; Zhou et al., 2016), and also bind to and activate Hpo (Dent et al., 2015).

To confirm these interactions, we conducted co-immunoprecipitation (co-IP) analyses in S2 cells. We co-transfected Vap33 tagged with V5 and RtGEF tagged with HA and conducted IP-western blot analysis (Fig. 3). Vap33–V5 co-immunoprecipitated with RtGEF–HA in both directions (Fig. 3A,B). We also investigated whether Vap33 and Hpo could form a complex by co-transfecting Vap33–V5 and Hpo–Flag in S2 cells and conducting IP followed by western blotting. Indeed, Vap33 co-immunoprecipitated with Hpo in both directions (Fig. 3C,D). Thus, Vap33 binds to both RtGEF and Hpo, supporting our AP-MS data, and consistent with previous studies showing that Hpo interacts with RtGEF and Git (Dent et al., 2015), and that the human Vap33 orthologues VAPA and VAPB interact with the Hpo orthologues MST1 and MST2 (Huttlin et al., 2015).

We then examined whether interactions between Lgl or Vap33 and RtGEF and/or Git occurred in Drosophila larval epithelial tissues, by using the in situ PLAs (Soderberg et al., 2006). We have previously used the PLA method to detect robust interactions between Lgl and Vap33 in Drosophila epithelial tissue, whereas controls lacking expression of one protein or without one of the antibodies showed minimal positive foci (Portela et al., 2018). First, using larval tissues (eye-antennal epithelial, salivary gland and brain) from endogenously GFP-tagged Lgl flies (Portela et al., 2018; Venken et al., 2011) and antibodies against GFP and Vap33, we confirmed that PLA foci were observed for Lgl and Vap33 (Fig. 4A; Fig. S1G,H) (Portela et al., 2018), suggesting that these proteins physically interact in vivo. We also observed foci for Lgl–GFP and atypical protein kinase C (aPKC), by using GFP and aPKC antibodies in eye-antennal discs and salivary gland cells (Fig. S1A,C). Importantly, no PLA signals were observed in the single antibody negative controls (Fig. S1B,D–F). We then conducted PLAs on eye-antennal epithelial and salivary gland tissues using flies expressing the endogenously tagged Lgl–GFP and Git–RFP transgene (Dent et al., 2019), using anti-GFP and -RFP antibodies, which revealed multiple PLA foci (Fig. 4B; Fig. S1J). Likewise, PLAs performed on Git–RFP transgenic fly eye-antennal epithelial and salivary gland tissues using anti-RFP and -Vap33 antibodies, revealed multiple PLA foci (Fig. 4C; Fig. S1K). Altogether, the PLA data confirm that Git interacts in vivo with Vap33 and Lgl in the Drosophila larval eye-antennal epithelium (Fig. 4), as well as in other polarized cells in the salivary gland and brain (Fig. S1).

Interestingly, in mammalian cells, Git interacts with the ADP ribosylation factor Arf1 and inhibits its activity (Zhou et al., 2016), and examination of the Drosophila Hippo pathway proteomics dataset (Kwon et al., 2013) revealed that the Drosophila Arf1 ortholog Arf79F is a high confidence interactor with Hpo (SAINT score >0.8), which was validated by co-IP analysis (Kwon et al., 2013). Based on these physical interactions, we also tested by PLA whether Arf79F interacted with Git, RtGEF and Hpo. Indeed, using flies containing a Hpo–GFP–Venus transgene (Pojer et al., 2021) and antibodies against Arf79F, we confirmed via PLA that Hpo and Arf79F interact in eye-antennal epithelial and brain tissue (Fig. 4D; Fig. S1I). Similarly, using PLAs on eye-antennal discs and salivary glands from Git–RFP or RtGEF–RFP transgenic flies (Dent et al., 2019) and anti-RFP and -Arf79F antibodies, Arf79F was confirmed to interact with Git and RtGEF (Fig. 4E,F; Fig. S1L,M). Thus, we have revealed that Arf79F binds to Git, RtGEF and Hpo in eye-antennal tissue (Fig. 4), as well as salivary gland or brain tissues (Fig. S1). The lower levels of PLA signals in some samples might be due to low levels of expression of the proteins or lower levels of protein–protein interactions.

To determine whether Vap33 genetically interacts with RtGEF, we examined the expression of the Hippo pathway target Diap1 in RtGEF mutant (RtGEF^P1036) clones that overexpress Vap33. Individually, RtGEF mutant clones showed increased Diap1 expression (Fig. 5A), as expected (Dent et al., 2015), and resulted in a slightly enlarged adult eye phenotype (Fig. 5B). The elevated Diap1 expression observed in RtGEF mutant clones was rescued to normal levels upon Vap33 overexpression (Fig. 5C, quantified in E), and the adult eye size was normalized (Fig. 5D). Thus, consistent with the protein interactions, RtGEF and Vap33 genetically interact in the regulation of the Hippo pathway in a manner consistent with both Vap33 and RtGEF acting to induce Hippo pathway activity.

Next, we examined genetic interactions between Git and a component of the V-ATPase, Vha68-2, in Hippo pathway regulation. Previous analysis of Git and RtGEF knockdown tissue had revealed that Yki targets are upregulated; however, the adult mosaic eyes were similar to the wild-type control (Dent et al., 2015; data not shown). As we have previously observed, Vha68-2 mutant clones are small although they do not show indications of cell death and are retained into pupal development (Portela et al., 2018). Consistent with our analysis of ex-lacZ (Fig. 1C), Vha68-2 mutant clones also showed reduced expression of the Hippo pathway target Diap1 (Fig. 6C) relative to wild-type clones (Fig. 6A, quantified in G), and resulted in reduced and disorganized adult eyes relative to the wild-type control (Fig. 6D compared with B). When Git was knocked down using a transgenic RNAi line in Vha68-2 mutant clones, partial rescue of the reduced Diap1 expression was observed (Fig. 6E, quantified in G) and the disorganized Vha68-2 mutant mosaic adult eye phenotype was strongly rescued resulting in a phenotype that was similar to wild-type adult eyes (Fig. 6F compared with D and B). These results show that the reduced Yki activity (low Diap1 expression) that occurs in Vha68-2 mutant clones is rescued by Git1 knockdown and suggest that the activation of Hippo signalling (reduced Diap1 expression) that occurs upon reducing V-ATPase activity depends upon Git. A possible interpretation for this genetic interaction is that V-ATPase activity might oppose the action of Git1 and/or RtGEF to activate Hpo.

We then tested whether Arf79F is required for the regulation of the Hippo pathway by Lgl. As we previously observed, lgl mutant tissue shows impaired Hippo pathway signalling, as revealed by the elevated levels of Diap1 (Fig. 7C compared to the control, A, quantified in M) and results in a distorted, disorganized adult eye phenotype relative to the wild-type control (Fig. 7D compared with B). We therefore sought to determine whether Arf79F genetically interacts with lgl in its regulation of the Hippo pathway. When Arf79F was knocked down, using a RNAi line (which we showed effectively reduced Arf79F protein levels in eye disc clones, Fig. S2A,B), a decrease in Diap1 expression was observed and clones were smaller than wild-type clones (Fig. 7E), suggesting that the Arf79F knockdown clones have reduced tissue growth, consistent with Hippo pathway upregulation. However, the Arf79F adult mosaic eyes showed only slight disorganization (Fig. 7F compared with the wild-type control in B). Knockdown of Arf79F in lgl mutant clones also resulted in low Diap1 levels, similar to Arf79F knockdown alone (Fig. 7G compared with E, quantified in M), suggesting that Arf79F is epistatic to lgl in the regulation of the Hippo pathway. Consistent with this, the lgl mutant mosaic distorted adult eye phenotype was normalized upon Arf79F knockdown (Fig. 7H compared with D and B). Additionally, we used a dominant-negative version of the activator of Arf79F, Sec71 (Sec71^DN) to reduce Arf79F function. When expressed individually in clones, Sec71^DN had no significant effect upon Diap1 levels (Fig. 7I, quantified in M), and did not obviously affect the adult eye phenotype (Fig. 7J). However, when Sec71^DN was expressed in lgl mutant clones, lower Diap1 levels were observed (Fig. 7K, quantified in M), and the lgl mutant mosaic adult eye phenotype was normalized towards wild-type (Fig. 7L compared with D and B). Interestingly, although Sec71^DN expression did not have a significant effect on Diap1 expression on its own, when expressed in lgl mutant clones it reduced Diap1 levels to below that of the control (Fig. 7M), suggesting that Lgl depletion depends on Sec71 (and thus Arf79F) to inhibit Hippo signalling. These results show that, consistent with the binding of Arf79F with Hpo (Fig. 4D), knockdown of Arf79F reduces Diap1 expression, suggesting that Arf79F acts to inhibit the Hippo pathway. Furthermore, knockdown of Arf79F (or expression of Sec71^DN) rescued the elevated expression of Diap1 in lgl mutant clones (Fig. 7M), indicating that Arf79F levels/activity are required for the inhibition of Hippo signalling in lgl mutant tissue. Thus, Arf79F is a negative regulator of Hippo signalling that acts downstream of Lgl. Since mammalian Git proteins act to inhibit Arf1 activity (Zhou et al., 2016), our results are consistent with a mechanism where Lgl and Vap33 promote RtGEF/Git activity to inhibit Arf79F (Arf1), and thereby activate the Hippo pathway.

In this study, we have revealed a new mechanism for the control of the Hippo pathway by the cell polarity regulator Lgl. We show herein that V-ATPase activity leads to inhibition of Hippo signalling, and conversely that Vap33 activates the Hippo pathway downstream of Lgl. We discovered that Vap33 physically interacts with RtGEF and Git using in vivo AP-MS. RtGEF and Git are Hpo interactors (Dent et al., 2019), and we confirmed that Vap33 interacts with RtGEF and Hpo in S2 cells, and that Vap33 and Lgl are in close proximity with Git in Drosophila cells. We also show that the ADP ribosylation factor Arf79F, the mammalian orthologue of which (ARF1) binds to the mammalian Git orthologues GIT1 and GIT2 (Zhou et al., 2016), and is in close proximity with Git, RtGEF and Hpo in Drosophila cells. Functionally, we show that overexpression of Vap33 rescues the elevated Diap1 levels (impaired Hippo signalling) in RtGEF mutant clones. Furthermore, the reduced Diap1 expression (elevated Hippo signalling) upon V-ATPase impairment (Vha68-2 mutant clones) is rescued by knockdown of Git. We also show that Arf79F knockdown leads to reduced Diap1 expression (increased Hippo signalling) downstream of Lgl. Altogether, our data are consistent with a model where Lgl and Vap33 activate the Hippo pathway by a dual mechanism: (1) Lgl and Vap33, through interaction with RtGEF, Git and Arf79F, positively regulate Hippo pathway activity, and (2) Lgl and Vap33, through interaction with components of the V-ATPase, block V-ATPase activation and prevent its negative regulation of the Hippo pathway (Fig. 8). Precisely how the V-ATPase functions to regulate Hippo signalling is unknown. However, the V-ATPase acts to increasing vesicle acidification, and this has been shown to impact the efficacy of many signalling pathways (Collins and Forgac, 2020; Eaton et al., 2021; Pamarthy et al., 2018). The V-ATPase might therefore act to inhibit Hippo pathway signalling by blocking the interaction of Lgl, Vap33, RtGEF, Git and Arf79F with Hpo on endosomes, thereby altering Hpo localization and inhibiting its activity. Consistent with this notion, we previously observed that Lgl colocalizes with endocytic vesicle markers (Parsons et al., 2014b), and that Hpo localization is altered in lgl mutant tissue (Grzeschik et al., 2010). Furthermore, Hpo signalling is associated with endosomal regulators (Verghese and Moberg, 2019), and the subcellular localization of Hpo has been shown to be important for its activation, and for Hpo-mediated phosphorylation and activation of its downstream protein kinase Wts (Sun et al., 2015).

In their interactions with the Hippo pathway, Git and RtGEF function to activate Hpo (Dent et al., 2015), and our analyses indicate that Arf79F acts to inhibit Hippo signalling given that the elevated Hippo pathway activity (increased Diap1 expression) in lgl mutant clones was normalized upon knockdown or inhibition of Arf79F. Previous studies examining the interactions between mammalian orthologues of Git and Arf79F, have shown that GIT1 and GIT2 act to inactivate ARF1 (Zhou et al., 2016), and therefore it is likely that Git (and RtGEF) act to block Arf79F activity to mediate Hpo activation (Fig. 8). Lgl and Vap33 might then function to promote Git- and RtGEF-mediated inhibition of Arf79F, facilitating activation of Hpo. Arf79F (ARF1) is a regulator of vesicular trafficking (Adarska et al., 2021), and might inhibit Hpo by relocalizing it away from its activators, such as Expanded and Fat, at specific apical subcellular domains (Sun et al., 2015). Interestingly, from our previous Lgl interactome analysis, Arf79F can be linked to Lgl through the Arf79F-binding protein Rab5 (Guruharsha et al., 2011; Kwon et al., 2013), which also binds to Lgl and aPKC at high confidence (SAINT score ∼1) (Portela et al., 2018 and our unpublished data). Thus, it is possible that the early endosomal regulator Rab5 might also be involved in linking Lgl, Vap33 and aPKC to Arf79F in early endosomes, where they could also interact with Vap33, Git, RtGEF and Hpo.

It is unclear whether the RtGEF- and Git-regulated protein kinase Pak (Zhou et al., 2016) is involved in the regulation of Hippo signalling, given that although individual knockdown of two Pak paralogs in Drosophila, Pak1 and Pak3, did not affect Hippo pathway signalling (Dent et al., 2015), it is possible that there might be redundancy between Pak1, Pak3 and the third Drosophila Pak paralogue Mushroom bodies tiny (Mbt) (Melzer et al., 2013; Menzel et al., 2008, 2007; Schneeberger and Raabe, 2003) in the regulation of Hippo pathway signalling. Pak proteins have been shown to be involved in cell polarity, F-actin regulation and morphogenesis in Drosophila (Asano et al., 2009; Baek et al., 2012; Bahri et al., 2010; Conder et al., 2007; Felix et al., 2015; Harden et al., 1996; Menzel et al., 2008, 2007; Schneeberger and Raabe, 2003; Vlachos and Harden, 2011); however, Mbt also plays a role in tissue growth during Drosophila developmental (Lim et al., 2019; Melzer et al., 2013), suggesting that Mbt is a candidate for investigation of a potential role downstream of RtGEF and Git in Hippo pathway regulation.

Our analyses here and in our previous study (Portela et al., 2018), have revealed that Lgl interacts with Vap33 to positively regulate the Hippo pathway and to negatively regulate the Notch pathway. Mechanistically, Lgl and Vap33 inhibit the Notch pathway by inhibiting V-ATPase activity and reducing the vesicle acidification that is required for the cleavage of the Notch receptor and release of Notch^ICD from vesicles (Portela et al., 2018), where it can then translocate into the nucleus and promote transcription of its target genes (Fig. 8). Interestingly, Arf79F (Arf1) is required for Notch signalling in Drosophila haemocyte differentiation, where it promotes Notch trafficking (Khadilkar et al., 2014). Thus, it is possible that Lgl, Vap33, RtGEF and Git might also inhibit Arf79F to reduce Notch signalling, and that in lgl mutant tissue elevated Arf79F activity might also contribute to the elevated Notch activation. Indeed, we found that Arf79F is required for the elevated Notch pathway signalling in lgl mutant clones, given that Arf79F knockdown in lgl mutant clones normalized the expression of the Notch pathway reporter, E(spl)m8-lacZ (Fig. S3). Whether RtGEF and Git are also regulators of Notch signalling remains to be determined.

In summary, our study has revealed a new mechanism for Hippo pathway regulation by the cell polarity and tumour suppressor protein Lgl. Mechanistically, Lgl and Vap33 function in a dual manner to promote Hippo signalling by both reducing V-ATPase activity, and thereby preventing V-ATPase from inhibiting Hippo signalling and by acting through RtGEF and/or Git to prevent Arf79F from inhibiting Hpo. By promoting Hippo signalling as well as inhibiting Notch signalling, Lgl and Vap33 function to limit tissue growth. Mutation of Lgl therefore results in tissue overgrowth due to inhibition of the Hippo pathway and activation of the Notch pathway. However, despite the dual regulation of Hippo and Notch signalling by Lgl and Vap33, the lgl mutant phenotype does not completely mirror the Vap33 mutant phenotype (Portela et al., 2018), which is likely due to Vap33 having additional roles in vesicle trafficking, proteolysis and signalling (Borgese et al., 2021; Choi et al., 2012; Deivasigamani et al., 2014; Gao et al., 2023; Mao et al., 2019; Sanhueza et al., 2015; Tsuda et al., 2008), and Lgl having additional roles in cell polarity regulation (Grifoni et al., 2013; Stephens et al., 2018). The dual mechanism of Notch and Hippo pathway regulation by Lgl and Vap33 might be important in limiting tissue growth during development and might play a role in tissue homeostasis, enabling cell proliferation to occur after tissue damage, thereby enabling wound repair. Indeed, Hippo pathway signalling is inhibited and Notch pathway signalling is activated after tissue wounding, and both pathways are involved in tissue regeneration (Blanco et al., 2010; Chen et al., 2012; Grusche et al., 2011; Kux and Pitsouli, 2014). However, the involvement of Lgl, Vap33, Git, RtGEF and Arf79F and the V-ATPase in the regulation of Hippo and Notch signalling during the response to tissue wounding remains to be determined. Furthermore, whether the mammalian orthologues of Lgl and Vap33 also act via these mechanisms to control tissue growth in mammals remains to be determined.

Fly stocks were generated in house or obtained from other laboratories or stocks centres as detailed in Table S1. All Drosophila genotypes for all the samples analysed in the figures and detailed in Table S2. All fly stocks and crosses were raised and undertaken on a standard cornmeal, molasses and yeast medium within temperature-controlled incubators kept at 25°C.

Mosaic analysis with a repressible cell marker [MARCM (GFP+); Lee and Luo, 2001) clones were generated as previously described (Grzeschik et al., 2010), using the following stock: ey-FLP, UAS-GFP; Tub-GAL80, FRT40A; Tub-GAL4/TM6B MARCM 2L. Crosses were set up and left overnight at room temperature. Adults were then turned daily into new vials and allowed to lay for ∼24 h at 25°C. Third-instar larval (L3) animals were then collected after ∼144 h (6 days after the egg laying period). Samples were collected and prepared as described across multiple days to be similarly aged, and then stored at 4°C in 80% glycerol as necessary prior to mounting.

Heat shock induced Flip-out clones were generated using the stock: hs-FLP; ex-lacZ; Act>CD2>GAL4, UAS-GFP. Specifically, clones were induced by heat shocking larvae at 37°C for 15 min at ∼72 h after egg laying. L3 imaginal discs were then dissected and prepared as described 72 h after clone induction.

L3 eye-antennal discs, were dissected in phosphate-buffered saline (PBS), fixed in 4% paraformaldehyde for 30 min, washed in PBS plus 0.1 or 0.3% Triton X-100 (PBT), and blocked in PBT plus 1% BSA (PBT/BSA). The tissues were incubated with the primary antibody in PBT/BSA overnight at 4°C. After washing off the primary antibodies, the tissues were incubated with the secondary antibodies in PBT for 1 h at room temperature. After washing off the secondary antibodies the samples were mounted in 80% glycerol or Vectashield (Vector Laboratories).

Antibodies used were: mouse β-galactosidase (Sigma, cat. no. G4644, 1:500), rabbit GFP (Invitrogen, cat. no. A11122, 1:500), mouse GFP (Invitrogen, cat. no. A11120, 1:500); mouse RFP (Invitrogen, cat. no. RF5R, 1:100) rabbit Yki (gift from Ken Irvine, Waksman Institute, Rutgers University, New Jersey, USA, 1:400), mouse Diap1 (gift from Bruce Hay, Division of Biology, California Institute of Technology, California, USA, 1:100), rabbit Vap33 (gift from Hugo Bellen, Department of Molecular and Human Genetics, Baylor College of Medicine, Houston, Texas, USA, 1:1000), rabbit aPKC (Santa Cruz Biotechnology, cat. no. sc216,  1:1000), rabbit Arf79F (gift from Maneesha S. Inamdar, Molecular Biology and Genetics Unit, Jawaharlal Nehru Centre for Advanced Scientific Research, Jakkur, Bangalore, India, 1:500; Khadilkar et al., 2014) and guinea pig Arf79F (gift from Fengwei Yu, Temasek Lifesciences Laboratory, Singapore, 1:200; Wang et al., 2017).

Secondary antibodies used were anti-mouse-IgG conjugated to Alexa Fluor 568, 633, 647, anti-rabbit-IgG conjugated to Alexa Fluor 568 and 633, and anti-guinea pig-IgG conjugated to Alexa 568 and 633 (all at 1:500). DNA was stained with 1 µM DAPI (stock prepared at 1 μg/ml, used at 1:1000; Sigma-Aldrich, cat. no. D9542).

Flies expressing endogenously YFP-tagged Vap33 (line #115288, Kyoto Stock Center) were grown in population cages, with embryos laid overnight collected on apple juice and agar plates. Extraction and protein purification were carried out as described previously (Neumuller et al., 2012). Briefly, extracts were prepared using Default Lysis Buffer (DLB) [50 mM Tris-HCl pH 7.5, 125 mM NaCl, 5% glycerol, 0.2% IGEPAL, 1.5 mM MgCl₂, 1 mM DTT, 25 mM NaF, 1 mM Na₃VO₄, 1 mM EDTA, and 2× Complete protease inhibitor (Roche)], spun down to clarify, incubated first with empty agarose beads and then with GFP-Trap resin (ChromoTek), followed by washes and elution in SDS sample buffer. The eluates were loaded on SDS-polyacrylamide gels and electrophoresed so that the dye front migrated ∼1 cm in the separating gel. The gel was then stained with Coomassie Blue, and the lane was cut into two 5 mm×5 mm pieces and sent for mass spectrometry analysis (Taplin Mass Spectrometry Facility, Harvard Medical School, USA). Samples were digested with trypsin in-gel and peptides were analysed using a Thermo Scientific Orbitrap mass spectrometer. The mass spectrometry analysis was conducted on three biological replicates for both experimental samples (Vap33-YFP) and controls (yw fly line). Data were analysed using the Significance Analysis of Interactome (SAINT) program (Choi et al., 2011), and a complete table of Vap33–YFP-interacting proteins is included as Table S3. Mass spectrometry data were deposited to the ProteomeXchange Consortium via the PRIDE partner repository (Perez-Riverol et al., 2022) with the data set identifier PXD035110.

Constructs used were: pMT-Vap33-V5 (Portela et al., 2018), pAc5.1-Flag-Hpo (Dent et al., 2015) and pAC5-1-HA-RtGEF (Dent et al., 2015). Drosophila S2 cells were maintained in standard Schneider's S2 medium with fetal bovine serum (Gibco) at 25°C, and transfections were performed using Effectene transfection reagent (Qiagen). CuSO₄ was added to culture medium at a final concentration of 0.35 mM for inducing expression of Vap33–HA. Cells were lysed using DLB (see above) and spun down (21,000 g for 15 min) to remove debris. Clear cell lysates were incubated with anti-V5, anti-HA or anti-Flag beads (Sigma) for 2 h at 4°C. Beads were washed three times with lysis buffer, and protein complexes were eluted with SDS buffer. Proteins were separated via SDS-PAGE, blotted, incubated with primary and secondary antibodies, and signal was detected using the Odyssey imaging system (LI-COR). Primary antibodies used were: mouse anti-V5 (Sigma, cat. no. V8012, 1:1000), rabbit anti-Flag (Sigma, cat. no. F7425, 1:1000), rabbit anti-HA (Sigma, cat. no. H6908, 1:1000). Secondary antibodies used were: IRDye 800CW donkey anti-rabbit IgG (LI-COR), IRDye 680CW donkey anti-mouse IgG (LI-COR).

The interactions between Lgl and aPKC, Vap33 or Git; Arf79F and Git, RtGEF or Hpo; and Vap33 and Git in Drosophila larval tissues were detected in situ using the Duolink^® In Situ Red Starter Kit Mouse/Rabbit (Sigma, DUO92101) according to the manufacturer's instructions. Briefly, primary antibody incubation was applied using the same conditions as immuno-histofluorescence staining. Duolink secondary antibodies against the primary antibodies were then added. These secondary antibodies were provided as conjugates to oligonucleotides that were able to form a closed circle through base pairing and ligation using Duolink ligation solution when the antibodies were in close proximity (Soderberg et al., 2006; a distance estimated to be <40 nm; Koos et al., 2014). The detection of the signals was conducted by rolling circle amplification using DNA polymerase incorporating fluorescently labelled nucleotides into the amplification products. The resulting positive signals were visualized as bright fluorescent dots, with each dot representing one interaction event. For a technical negative control, one of the primary antibodies was not added, with no positive signals were obtained from that assay. An additional negative control was performed in a tissue without one of the antigens (GFP) and the full protocol was performed in those tissues. As a positive control, antibodies against two well-known interactors in the tissues, Lgl–aPKC and Lgl–Vap33, were used. The tissues were visualized using confocal microscopy (Zeiss Confocal LSM 780 PicoQuant FLIM or Zeiss LSM 800 Airyscan laser scanning confocal).

Primary antibody pairs used were mouse GFP with rabbit Vap33 (Fig. 4A; Fig. S1G,H), mouse RFP with rabbit GFP (Fig. 4B; Fig. S1J), mouse RFP with rabbit Vap33 (Fig. 4C; Fig. S1K), mouse GFP with rabbit Arf79F (Fig. 4D; Fig. S1I), mouse RFP with rabbit Arf79F (Fig. 4E,F; Fig. S1L,M) and mouse GFP with rabbit aPKC (Fig. S1A,C). A GFP-tagged version of Lgl was used to detect interactions with Vap33, since the Lgl and Vap33 antibodies were both raised in rabbit. GFP-tagged Hpo was also used since the available Hpo antibody was raised in rats, and PLA antibodies designed for use with rat-raised antibodies were unavailable. RFP-tagged versions of RtGEF and Git were used for similar reasons.

Fluorescent-labelled samples were mounted in 80% glycerol or Vectashield (Vector Laboratories) and analysed by confocal microscopy (LEICA TCS SP5, Zeiss Confocal LSM 780 PicoQuant FLIM or Zeiss LSM 800 Airyscan laser scanning confocal). Images were processed using Leica LAS AF Lite and Fiji (Image J 1.50e). Images were assembled using Photoshop 21.2.3 (Adobe). Adult eyes were imaged on a dissecting microscope using a Scitec Infinity1 camera. Images were processed, analysed, and assembled using some combination of LAS AF Lite (Leica), Zen 2012 (Zeiss), Fiji and Photoshop 21.2.3 (Adobe).

Relative Ex-lacZ (Fig. 1), or E(spl)m8-lacZ (Fig. S3) βGal stainings, and Diap1 stainings (Figs 2, 5, 6 and 7) within eye discs was determined using images taken at the same confocal settings. Average pixel intensity was measured using the measurement log tool from Fiji or Photoshop 5.1 (Adobe). In the case of E(spl)m8-lacZ analyses, clones were chosen just posterior to the morphogenetic furrow of each eye disc. Average pixel intensity was measured in mutant clones and the adjacent wild-type tissue of the same areas and expressed as a ratio of pixel intensity of the mutant clone relative to the wild-type tissue. To analyse and plot data, we used Microsoft Excel 2013 and GraphPad Prism 9. We performed a D'Agostino and Pearson normality test, and the data found to have a normal distribution were analysed by a two-tailed unpaired t-test with Welch correction. In the case of multiple comparisons, we used a one-way ANOVA with a Bonferroni post test. Error bars represent s.e.m.

We are grateful to all those who contributed fly stocks or antibodies to this study, to Bloomington, VDRC and NIG Stock Centers, OzDros and Flybase. H.E.R. was supported by a Senior Research Fellowship from the National Health and Medical Research Council (NHMRC) Australia, and funds from La Trobe University and La Trobe School of Molecular Science. We are grateful for the assistance of the La Trobe University Bioimaging Platform.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.261917.reviewer-comments.pdf