Hippo Stabilises Its Adaptor Salvador by Antagonising the HECT Ubiquitin Ligase Herc4

Signalling through the Hippo (Hpo) pathway involves a kinase cascade, which leads to the phosphorylation and inactivation of the pro-growth transcriptional co-activator Yorkie (Yki). Despite the identification of a large number of pathway members and modulators, our understanding of the molecular events that lead to activation of Hpo and the downstream kinase Warts (Wts) remain incomplete. Recently, targeted degradation of several Hpo pathway components has been demonstrated as a means of regulating pathway activity. In particular, the stability of scaffold protein Salvador (Sav), which is believed to promote Hpo/Wts association, is crucially dependent on its binding partner Hpo. In a cell-based RNAi screen for ubiquitin regulators involved in Sav stability, we identify the HECT domain protein Herc4 (HECT and RLD domain containing E3 ligase) as a Sav E3 ligase. Herc4 expression promotes Sav ubiquitylation and degradation, while Herc4 depletion stabilises Sav. Interestingly, Hpo reduces Sav/Herc4 interaction in a kinase-dependent manner. This suggests the existence of a positive feedback loop, where Hpo stabilises its own positive regulator by antagonising Herc4-mediated degradation of Sav.


dsRNA production and treatment
RNAi library generation work-plan can be found in the supplementary figures (S2-S4 Figs). dsRNAs were synthesized using the Megascript T7 kit (Ambion). DNA templates for dsRNA synthesis were PCR-amplified from genomic DNA using primers that contained the 5' T7 RNA polymerase-binding site sequence. dsRNA primers were designed using the DKFZ RNAi design tool (http://www.dkfz.de/signaling2/e-rnai/) RNAi screening was carried out as outlined in S3 Fig. Follow-up RNAi experiments were carried out in 6 well plates using S2 cells. 1-2x10 6 cells were plated per well and cells left to settle. After 3 h the medium was removed and replaced with 1 ml of serum free Schneider's medium containing dsRNAs (20 μg). Cells were soaked for 30 min and then 2 ml of full Schneider's medium was added. Transfections were carried out 24 h after dsRNA treatment.

Ubiquitylation assays
For ubiquitylation assays S2 cells were transfected with Myc-Sav and HA-Ubiquitin and, unless otherwise stated, treated with proteasome inhibitors for 4 h. Cells were harvested and lysed in Lysis buffer (see above). Lysates were immunoprecipitated with an anti-Myc antibody and Protein A Sepharose (Sigma), and subjected to western blot analysis. Ubiquitylation was detected using an antibody against HA-Ubiquitin. Lysis buffer containing 5mM N-Ethylmaleimide (NEM) (Sigma) was used to block deubiquitylating enzyme activity.

Proteasome inhibitor treatment
For proteasome inhibition experiments, cells were treated with 50μM MG132 (Calbiochem) and 50μM calpain inhibitor I (Ac-LLnL-CHO or LLnL) (Sigma) for 4 h before cell lysis. Ubiquitylation assays were carried out after proteasome inhibitor treatment, unless stated otherwise.

Western blot analysis of fly heads
Heads were cut off from frozen animals and lysed in RIPA buffer (10 mM Tris HCl pH 7.5, 150 mM NaCl, 1% (v/v) Triton X-100, 0.1% (w/v) SDS, 1% (w/v) Sodium deoxycholate) supplemented with 0.1M NaF, phosphatase inhibitor cocktails 2 and 3 (Sigma) and protease inhibitor cocktail (Roche). Tissue lysates were incubated on ice for 10 min and subsequently cleared by centrifugation at 16,000 g for 10 min at 4°C. 150ug of total lysate was loaded per lane.

Immunofluorescence
Wing imaginal discs from third-instar larvae were dissected, fixed for 20 min in 4% formaldehyde, washed three times in PBS supplemented with 0.1% Triton X-100 (PBS-T) and preblocked in PBS supplemented with 0.3% Triton X-100 and 10% normal goat serum (NGS) for 1 h. Discs were incubated overnight in the primary antibody diluted in PBS-T containing 10% NGS, followed by three washes with PBS-T, and incubation with a secondary antibody in PBS-T/10% NGS for 2 h at room temperature. After three further washes, discs were mounted in Vectashield mounting medium with DAPI (Vectorlabs). Rat anti-Ci 155 antibody (Developmental Studies Hybridoma Bank, monoclonal 2A1) was used at 1:20. Secondary anti-rat Alexa 647 antibody (Molecular Probes) was used at 1:200. Fluorescence images were acquired on a Zeiss LSM710 confocal laser scanning microscope (40x objective lens).

Generation of herc4 CRISPR mutant
Introduction of a frame-shift mutation in the herc4 gene was performed using the CRISPR/ Cas9 system [49]. Suitable target sites for guide RNAs (sgRNA) were determined using http:// www.flyrnai.org/crispr/. Sense and anti-sense oligos of the selected sgRNA were annealed, phosphorylated and cloned into the expression vector pCFD3 according to the protocol described on http://www.crisprflydesign.org/grna-expression-vectors/. DNA was injected into embryos of y 1 , M{nos-Cas9.P}ZH-2A w Ã transgenic flies (Bloomington stock number 54591) by the Fly Facility at the University of Cambridge Department of Genetics. G0 flies were then crossed with the w;;TM3/TM6 balancer strain; subsequently, G1 male flies were crossed with the same balancer strain, followed by genomic DNA screening for frame-shift mutations. We also balanced a wild type progeny of the injected founders to use as a control strain (herc4 ctrl ).

Analysis of Drosophila adult wings
Adult wings were mounted using Euparal mounting medium (Agar Scientific) and imaged using a Zeiss Axio-plan 2 imaging system, using Optovar optics (1.6x) and a 2.5x objective (Plan-NEOFLUAR, Zeiss) connected to a Leica DFC420C camera. Wings were subsequently measured and quantified using ImageJ.

Results
Sav is ubiquitylated and degraded by the proteasome Sav is an unstable protein and its stability is regulated by the proteasome system ( Fig 1A, [7,9,45,46]). As proteasomal degradation is triggered by poly-ubiquitylation of target proteins, we performed a cell-based ubiquitylation assay by co-expressing Myc-tagged Sav with HAtagged ubiquitin, followed by Myc immunoprecipitation (Fig 1A). This revealed the presence of Sav poly-ubiquitylated species, which were stabilised by treatment with proteasome inhibitors. Ubiquitin conjugation primarily occurs through the formation of an isopeptide bond between the ubiquitin C-terminus and the epsilon amino group of lysine residues on the substrate [50]. Sav contains eighteen lysine residues, clustered at both termini of the protein, ten are contained within amino acids 12 to 57 and eight are located in the area containing the conserved WW and SARAH domains ( Fig 1B). To establish the domain where Sav ubiquitylation occurs, we analysed Sav truncations spanning either the N or C terminal part of the protein (SavΔC and SavΔN). As shown in Fig 1B, Sav ubiquitylation occurs within the conserved C-terminal domain, comprising the WW and SARAH domain.

Hpo stabilises Sav by inhibiting its ubiquitylation
We and others have observed that exogenous Sav is stabilised by Hpo co-expression, while Hpo depletion causes endogenous Sav downregulation (Fig 2A and 2B, [7,9,45,46,48]). In order to further investigate the role of Hpo in regulating Sav stability, we carried out  ΔN) were tested for ubiquitylation as described above. Right panel: schematic representation of Sav protein structure and distribution of Lysine residues. Sav contains eighteen Lysine residues (K), which could serve as potential ubiquitylation sites: ten are located near the N-terminal end of the protein, whereas the remaining eight are located in the C-terminus.
doi:10.1371/journal.pone.0131113.g001 ubiquitylation assays using full-length Sav or Sav truncations (SavΔC and SavΔN) in the presence or absence of Hpo. Co-expression of Hpo resulted in a reduction in Sav ubiquitylated species ( Fig 2C), indicating that Hpo regulates Sav stability by preventing its ubiquitylation.
Since Sav ubiquitylation occurs in the C-terminal domain responsible for Hpo binding, Hpo could sterically interfere with ubiquitin ligase binding or ubiquitin transfer. Alternatively, since Hpo is known to phosphorylate Sav ( [7,9,45,46,51] and this report), phosphorylation might prevent ligase binding or ubiquitylation. To investigate these possible mechanisms, we analysed the effect of several Hpo truncations and mutants on Sav ubiquitylation ( Fig 3A). Hpo constructs that bind but fail to phosphorylate Sav, either by deletion of the kinase domain (HpoΔN) or point mutation in the T-loop domain (at residue 195 -HpoT195A), were unable to prevent Sav ubiquitylation ( Fig 3B). Similarly, we tested two C-terminal truncations of Hpo lacking the SARAH domain but still containing the Hpo PPxY motif, the longer of which still binds Hpo robustly (HpoΔC2), while the shorter truncation is reduced in its ability to bind Hpo (HpoΔC3), as well as a version of HpoΔC3 carrying a mutation in the PPxY motif (HpoΔC3/Y591G) motif, which cannot bind Sav at all (Fig 3C). These experiments showed that the strength of Hpo/Sav binding correlates with the effect of Hpo on Sav ubiquitylation, with HpoΔC2 strongly blocking Sav ubiquitylation, and HpoΔC3 and HpoΔC3/Y591G having increasingly weaker effects (Fig 3C). These results indicate that Hpo requires both Sav binding and kinase activity to inhibit Sav ubiquitylation, thereby excluding steric interference as the sole mechanism for Hpo action. A ubiquitin ligase RNAi screen identifies Herc4 as a Sav E3 ligase To identify proteins responsible for Sav ubiquitylation, we carried out a cell-based RNAi screen using a Drosophila dsRNA sub-library containing known and candidate genes involved in ubiquitylation (378 genes involved in ubiquitin conjugation and ubiquitin binding). The effect of dsRNAs on Sav-Myc expression was analyzed by standard Western-blot analysis. Using this approach we identified the HECT domain protein Herc4 (CG9153) as a Sav E3 ligase ( Fig 4A). Drosophila Herc4 contains a RLD (RCC1-like domains) domain comprising 7 RCC1-like repeats and a HECT domain [52] (Fig 4B). RLD domains are structural scaffolds with a β-propeller architecture believed to be involved in protein-protein interactions, whereas the HECT domain is a conserved E3 ligase domain [53,54]. HECT domains bind specific ubiquitin-conjugating enzymes (E2s), accept ubiquitin from the E2 to form a ubiquitin-thioester intermediate with the HECT active cysteine (Fig 4B), then transfer ubiquitin to the substrate. Herc4 belongs to the subfamily of small Herc E3 ligases, which in humans comprises 4 members, Herc3-6 [52]. In Drosophila there are only 2 small Herc E3 ligases, Herc2 and Herc4. Drosophila Herc4 is marginally more related to human Herc4 than human Herc3 (45% and 44% identity between CG9153 and hsHerc4 and hsHerc3, respectively; Fig 4C).

Herc4 controls Sav stability and ubiquitylation in vitro and in vivo
HECT E3 ligases physically interact with their substrates [54]. To determine whether Herc4 associates with Sav, we performed co-immunoprecipitation (co-IP) experiments in S2 cells using tagged Herc4 and Sav. Herc4 robustly associated with Sav, but not with a control protein, Rep3 (Fig 5A). We identified Herc4 based on the ability of herc4 RNAi to increase Sav protein levels in S2 cells (Fig 5B). To further investigate the functional significance of this finding, the role of Herc4 in Sav ubiquitylation was analysed. We performed a cell-based ubiquitylation assay in the presence of the deubiquitylating enzyme inhibitor NEM to preserve ubiquitylated species. As shown in Fig 5C, herc4 RNAi resulted in a marked reduction of Sav ubiquitylation compared to control RNAi treatment, indicating that Herc4 is required for Sav ubiquitylation.
To assess the specificity of Herc4 as a Sav E3 ligase, we analyzed the effect of Herc4 overexpression on Sav ubiquitylation. Expression of Herc4 enhanced Sav ubiquitylation and reduced Sav protein levels in cell lysates, while a catalytically impaired Cysteine to Alanine mutant (Herc4C/A) had a reduced effect (Fig 5D). In agreement with our previous observations ( Fig  1B), the C-terminal portion of Sav (SavΔN), but not the N-terminal (SavΔC), binds to and is ubiquitylated by Herc4 (Fig 5D). In contrast, two other HECT domain E3 ligases, Su(dx) and Nedd4, did not promote Sav ubiquitylation (S1 Fig). Next, we explored the effect of Herc4 on Sav stability in vivo. We generated a GFP-tagged sav transgene, expressed under the control of the ubiquitin promoter (ubi-GFP::sav). Using the GAL4/UAS system, UAS-Herc4 was expressed in the posterior compartment of wing imaginal discs, the larval precursor to the adult wing, using the hh-Gal4 (referred to as hh>herc4) driver. Cubitus interruptus (Ci) staining was used to differentiate the anterior (Ci-positive) and posterior (Ci-negative) compartments of the wing imaginal disc. Herc4 overexpression caused depletion of Sav levels (Fig 6, right panels), indicated by the reduction in the GFP signal. These results are consistent with our biochemical data showing that Herc4 promotes Sav ubiquitylation and proteasomal degradation.

Mapping the Herc4/Sav interaction domains
To map the domains required for the interaction between Herc4 and Sav, we carried out co-IP experiments using fragments of either protein (Fig 7). Using this strategy, we were able to map the Herc4 interaction domain to the RLD domain as deleting the whole RLD abolishes binding between the two proteins. Deletion of RCC1-like repeats 1-3 had no effect on binding, neither did deletion of the entire HECT domain (Fig 7A). Thus, Herc4 binds Sav via RCC1-like domains 4-7.
Herc4 binds to the conserved C-terminus of the Sav protein (Fig 5D). To precisely define the region responsible for Sav/Herc4 binding, we analysed C-terminal fragments of Sav in co-IP experiments with Herc4 ( Fig 7B). The domain responsible for the association of Herc4 to Sav can be narrowed down to the region stretching from the end of the WW domains to the beginning of the SARAH domain (aa 492-546), as deleting this domain abolishes binding between Herc4 and Sav (Fig 7B). Adding the SARAH domain (SavΔN4) renders the interaction more robust, and elicits the appearance of a ladder of higher molecular species, which could indicate poly-ubiquitylation (Fig 7B).

Antagonism between Herc4 and Hpo determines Sav stability
As both Herc4 and Hpo have a role in regulating Sav ubiquitylation and stability, we tested a possible functional interaction between the two proteins. Combined expression of Hpo and Herc4 resulted in reduced Herc4-induced Sav ubiquitylation (Fig 8A). We also noted that upon Hpo co-expression there was reduced association between Herc4 and Sav (Fig 8A). Similarly, Hpo reduced Sav levels in Herc4 immunoprecipitates (Fig 8B). Interestingly, the kinase dead version of Hpo (HpoT195A) did not affect Sav/Herc4 binding (Fig 8B), indicating that the effect of Hpo is phosphorylation-dependent. These results suggest that phosphorylation of Sav by Hpo prevents Herc4/Sav binding, leading to Sav stabilization. Depletion of hpo causes Sav to be destabilized (Fig 2B), suggesting that under those conditions Sav is no longer protected from Herc4-mediated degradation. Indeed, co-depletion of herc4 and hpo restored Sav levels (Fig 8C). Thus, the interplay between Hpo and Herc4 determines Sav stability.
herc4 depletion prevents Sav destabilization caused by loss of apical Hpo pathway regulators Recent reports have highlighted the importance of the cell cortex as an organising centre for Hpo pathway activity [44,47,[55][56][57][58]. Several apically localized proteins, such as the Kibra/ Expanded/Merlin (KEM) complex and the immunoglobulin super-family trans-membrane protein Echinoid (Ed), have been proposed to recruit Hpo/Sav to the cell membrane. In particular, Ed has been shown to bind Sav and prevent its degradation [55].
Expression of Ed and Kibra in S2 cells stabilises co-transfected Sav (Fig 9A and [55]). Conversely, loss of either Kibra or Ed causes a marked reduction of Sav levels (Fig 9B and 9C), suggesting that loss of apical tethering of Sav makes it a target for ubiquitylation and degradation. To investigate this assumption further, we tested the effect of depleting herc4 in this context. S2 cells were transfected with Sav-Myc and treated with dsRNA for kibra and ed in combination with herc4 RNAi or lacZ RNAi as a control. herc4 RNAi rescued the destabilising effect of ed and kibra RNAi (Fig 9B and 9C). Thus, when apical members of the Hpo pathway are depleted, Sav is released into the cytosol, where Herc4 promotes its ubiquitylation and subsequent degradation by the proteasome.

Growth regulatory function of Herc4
One of the main effects of Hpo pathway disruption in vivo is an increase in tissue size. To investigate the impact of Herc4 on tissue growth control via Sav, we combined GAL4/UASdriven Herc4 expression with transgenic lines expressing either wild type Sav or a version which is refractory to inhibition by the Salt-Inducible Kinases (Sik2/3) and therefore more active (Sav S413A ) [59]. The transgenes were expressed under the control of the wing-specific driver rotund-Gal4 (rn-Gal4). Overexpression of wild type Sav modestly reduced the size of the wing, while the Sav S413A mutant had a stronger effect (Fig 10A). Expression of UAS-herc4 was sufficient to partially suppress the undergrowth phenotype caused by Sav overexpression. This effect was dependent on the E3 ligase activity of Herc4, since overexpression of a HECT mutant version of Herc4 (UAS-Herc4C/A) had no effect on wing size (Fig 10A and 10B). Thus, Herc4 expression is able to suppress Sav growth-inhibitory function in vivo.
To study the effect of loss of Herc4 activity we generated a herc4 mutant using CRISPR technology [49]. Using an sgRNA that targets the second exon of herc4, we generated a mutation, herc4 C6.3 , carrying a one base-pair deletion at codon position C92 (cysteine 92) (Fig 11A). This mutation induces a frame-shift resulting in the production of a 30aa peptide (followed by 24aa of out-of-frame sequence), which lacks most of the Herc4 protein (Fig 11B). Homozygous herc4 C6.3 mutant animals were viable and fertile and had no apparent morphological abnormalities. This is not surprising, as modest overexpression of Sav, for instance in the eye under the GMR promoter, does not lead to a visible phenotype [5]. We therefore tested whether herc4 C6.3 could modify the undergrowth phenotype elicited by strong overexpression of Sav in the wing under the rn-GAL4 driver (Fig 11C and 11D). Interestingly, loss of herc4 C6.3 does not modify the undergrowth induced by savS413A, which is already an activated form. By contrast, the herc4 C6.3 mutant significantly enhanced the undergrowth induced by wild type Sav overexpression. To test whether loss of herc4 affects Sav protein levels in vivo, we analysed Sav protein levels in fly heads by Western blot. Fig 11E shows that Sav protein levels were increased in herc4 C6.3 heads compared to control heads. Together, these results support a role for Herc4 as a Sav E3 ligase.

Discussion
The scaffold protein Sav is a core component of the Hpo signalling pathway and its genetic inactivation is sufficient to increase tissue growth in flies and mice [5,6,41,43]. Sav dimerises with Hpo through their respective SARAH domains [7][8][9][10][11], and recruits Hpo to the plasma membrane, thereby promoting Wts phosphorylation [44]. In addition, Sav associates with the downstream kinase Wts via a WW domain/PPxY interaction [5]. Sav may therefore scaffold the Hpo/Wts core kinase cassette, thereby further strengthening Wts phosphorylation and activation. Indeed, mammalian Sav1/WW45 immunoprecipitates with endogenous Lats1 and Mst2 [21], and Sav strengthens the Hpo/Wts association in Drosophila tissue culture cells [59].
Sav levels appear to be tightly regulated by proteasomal degradation, and its association with Hpo/MST promotes both its phosphorylation and stabilisation ( [7,9,45,46], this report). Using a point mutation in the ATP-binding pocket (Lysine to Arginine, K/R), we and others had observed that Hpo/MST kinase activity appeared dispensable for Sav stabilisation, opening the possibility for the involvement of another kinase [9,45]. However, since a version of Hpo carrying a mutation in the T-loop (T195A) and a truncation lacking the kinase domain both fail to phosphorylate and stabilise Sav (Fig 3), it seems most likely that Hpo kinase activity is in fact responsible for stabilising Sav, suggesting that the HpoK/R mutant has residual activity.
In order to shed light on the regulation of Sav stability, we performed a screen for Sav E3 ligases, leading us to identify Herc4 as a negative regulator of Sav (Fig 4). We show that herc4 depletion reduces Sav ubiquitylation in cell culture ( Fig 5) and that herc4 mutants have elevated Sav levels in vivo (Fig 11). However, since the increase in Sav levels in herc4 mutants are relatively modest, it is possible that other Sav ubiquitin ligases operate in parallel to Herc4. The Herc family is a poorly characterised group of six proteins containing a C-terminal HECT ubiquitin ligase domain and at least one N-terminal RLD [52]. Mammalian Herc1 and 2 are large (~500 KDa) proteins containing multiple RLDs as well as other domains (WD40, zinc fingers), while Herc3-6 have shorter N-termini containing one RLD domain. As a β-propeller structure, the RLD domain is likely to constitute a substrate-recruitment motif [53]. Indeed, we showed that Sav binding requires RCC1 repeats 4-7 (Fig 7). Drosophila has a single Herc1/2 ortholog (called Herc2) and a single Herc3-6 ortholog (CG9153, which we name Herc4 based on sequence homology). Herc4 mutant mice are viable but males display a spermatozoon maturation defect [60]. In addition, Herc4 was found to be upregulated in over half of a cohort of invasive ductal breast carcinoma [61]. It would therefore be interesting to examine Sav1/WW45 levels in these tumours, as well as assess their differentiation status, since recent work shows that Sav is required for terminal differentiation in the mouse breast epithelium [42].
In this report, we have also shown that the ability of Herc4 to degrade Sav is antagonised by Hpo (Fig 8). Furthermore, Hpo co-expression is able to displace Herc4 from Sav in a kinase domain-dependent manner. This could be due to phosphorylation by Hpo of the interaction  surface on either Sav or Herc4. Our data are consistent with a model in which apical recruitment of Sav by proteins such as Ed and the KEM complex can stabilise Sav, perhaps by bringing it in close proximity to active Hpo at the plasma membrane. This would lead to stabilisation of Sav, promote Wts activation by Hpo and, ultimately, lead to Yki inhibition. We therefore propose that Hpo can strengthen the activation of the Hpo pathway core kinase cassette by phosphorylating its own adaptor Sav, thereby preserving it from Herc4-dependent degradation.
Supporting Information