NEDD4 controls intestinal stem cell homeostasis by regulating the Hippo signalling pathway

The Hippo pathway plays crucial roles in regulating organ size and stem cell homeostasis. Although the signalling cascade of the core Hippo kinases is relatively well understood, little is known about the mechanisms that modulate the activity of the Hippo pathway. Here, we report identification of NEDD4, a HECT-type E3 ubiquitin ligase, as a regulatory component of the Hippo pathway. We demonstrate that NEDD4 ubiquitylates and destabilizes WW45 and LATS kinase, both of which are required for active Hippo signalling. Interestingly, MST1 protects WW45, but not LATS2, against NEDD4. We also provide evidence indicating that NEDD4 inactivation at high cell density is a prerequisite for the elevated Hippo activity linked to contact inhibition. Moreover, NEDD4 promotes intestinal stem cell renewal in Drosophila by suppressing Hippo signalling. Collectively, we present a regulatory mechanism by which NEDD4 controls the Hippo pathway leading to coordinated cell proliferation and apoptosis. The Hippo pathway plays a role in regulating organ size and stem cell renewal but the regulatory mechanisms that fine-tune this pathway are not well understood. Here the authors report on the role of NEDD4 as a negative regulator of the Hippo signalling components, WW45 and LATS kinase, and in controlling cell proliferation and intestinal stem cell homeostasis.

D etermination and maintenance of the proper organ size are of paramount importance in all living organisms. The Hippo pathway represents one of the crucial mechanisms in control of organ size via modulating cell proliferation and apoptosis. Most of the Hippo signalling components were initially identified from genetic studies in Drosophila melanogaster [1][2][3] . Specifically, WW domain containing scaffold protein Salvador (Sav), Ste-20 family kinase Hippo (Hpo), NDR kinase Warts (Wts) and transcription co-activator Yorkie (Yki) are core components of the Hippo pathway in Drosophila 4-6 . Sav forms a complex with Hpo and Wts through its C-terminal Sav-Rassf-Hpo (SARAH) domain and WW domains, respectively, and the trimeric interaction facilitates phosphorylation and activation of Wts by Hpo 6,7 . Activated Wts phosphorylates Yki and phosphorylated Yki is subsequently sequestered in the cytoplasm through interaction with the 14-3-3 protein. Inactivation of Yki leads to decreased expression of the cell cycle regulator Cyclin E and the Drosophila inhibitor of apoptosis protein DIAP1 (refs 8,9). The Hippo signalling cascade and core components are highly conserved from Drosophila to mammals (Sav for WW45 or SAV1, Hpo for MST1/2, Wts for LATS1/2, and Yki for YAP and TAZ in mammals, respectively), and mammalian homologues are able to rescue the overgrowth phenotype of Drosophila mutants 6,10,11 . It has recently been reported that the mammalian Hippo pathway regulates cell-to-cell contact inhibition as well as organ size [12][13][14][15] . Thus, dysregulation of the mammalian Hippo pathway is linked to tumorigenesis. Core Hippo components, MST1/2, LATS1/2 and WW45, are either hypermethylated or mutated in various tumours and cancer cells 1,16,17 . YAP is elevated in some human cancers, including breast, liver and colorectal cancer, and this overexpression leads to inhibition of apoptosis and enhancement of the epithelialmesenchymal transition as well as anchorage-independent growth [18][19][20] . Several lines of evidence also indicate that the Hippo signalling modulates homeostatic self-renewal and regeneration of intestinal stem cells (ISCs) 21,22 .
Although the signalling cascade of the core kinase pathway is relatively well understood, much less is known about the regulatory mechanisms that fine-tune the Hippo pathway. Given that Sav proteins become stable in Drosophila when the 26S proteasome is inhibited and Wts protein levels are post-transcriptionally regulated 7,23 , it is plausible that the ubiquitin-proteasome system plays a role in regulating the Hippo pathway. Ubiquitin-activating (E1), conjugating (E2) and ligating (E3) enzymes work in concert to accomplish protein ubiquitylation. Among these, members of the E3 ubiquitin (Ub) ligase family confer substrate specificity by binding substrate proteins 24 . NEDD4 (neural precursor cell expressed, developmentally down-regulated protein 4; also known as NEDD4-1) is an E3 Ub-ligase composed of a N-terminal C2 domain, four WW domains that bind to PY motifs (L/PPXY) and a C-terminal HECT (Homologous to E6-AP carboxyl terminus) domain 25 . The C2 domain is crucial for Ca 2 þ -dependent plasma membrane localization of NEDD4 (refs 26,27). NEDD4 is known to positively regulate cell proliferation and growth. This is supported by the observation that the NEDD4 knockout (KO) mouse shows apparent growth retardation compared with wild-type littermates and that NEDD4 KO mouse embryonic fibroblast cells grow slowly. Such phenotypic outcomes originate from the NEDD4-mediated positive regulation of IGF-1 and insulin signalling 28 . It has recently been reported that NEDD4 family member proteins, including ITCH 29,30 and NEDD4 (ref. 31), negatively regulate LATS1, suggesting that NEDD4 plays a role in Hippo signalling.
In the present study, we report the role of NEDD4 as a key regulator of the Hippo signalling pathway. NEDD4 negatively regulates the stability of WW45 and LATS kinase, which leads to alterations in cell proliferation and ISC homeostasis.

Results
NEDD4 directly interacts with WW45. To understand how the Hippo signalling pathway is controlled, we set out to identify a specific regulator using affinity purification combined with mass spectrometry. WW45 was chosen as the bait since WW45 is the sole adaptor protein that makes a core complex with MST1/2 and LATS1/2. We thus expressed FLAG-WW45 in HEK293T cells and performed affinity purification. Subsequent liquid chromatography-mass spectrometry analysis revealed several WW45interacting proteins. Specifically, our mass spectrometry data showed that WW45-immunoprecipitates contained MST1 and LATS2, both of which are known to interact with WW45, as well as a novel interacting protein, NEDD4 E3 Ub-ligase (Fig. 1a). We confirmed the in vivo interaction between WW45 and NEDD4 using both transiently expressed and endogenous proteins in HEK293T cells (Supplementary Fig. 1a and Fig. 1b). We performed an in vitro pulldown assay using GST-WW45 purified from Escherichia coli and immunopurified FLAG-NEDD4 proteins. NEDD4 was readily detected in the fractions eluted from the GST-WW45 affinity column but not in eluates from the glutathione S-transferase (GST) column, indicating that the interaction between these proteins was direct (Fig. 1c). Next, we constructed N-and C-terminal-truncated WW45 mutants (199C N198, and N321) to map the binding region for NEDD4. Whereas WW45 interacts with its binding partners mostly through either the WW domains or the SARAH domain 6 , it bound to NEDD4 via its N-terminal region (Fig. 1d). We also constructed several NEDD4 deletion mutants and found that the very N-terminus (amino acids 1-100) was crucial for binding to WW45 ( Fig. 1e and Supplementary Fig. 1b). Taken together, NEDD4 is a newly identified WW45-interacting protein and they interact with each other directly through their N-terminal regions.
NEDD4 ubiquitylates and destabilizes WW45. To determine the functional significance of the NEDD4-WW45 interaction, we tested whether NEDD4 affects the cellular level of WW45. We found that WW45 levels varied inversely with the levels of cotransfected NEDD4 (Fig. 2a). Moreover, NEDD4-mediated destabilization of WW45 was blocked by the proteasome inhibitor MG132. Co-expression of NEDD4 C967S , which lacks ubiquitin ligase activity owing to the replacement of Cys 967 with Ser 32 , and N-terminal deletion mutant NEDD4 DN280 did not affect WW45 levels, indicating that both the substrate-binding and E3 catalytic activity of NEDD4 were required for WW45 destabilization (Fig. 2b). Furthermore, ectopic expression of NEDD4 led to downregulation of endogenous WW45 ( Supplementary Fig. 2a), which was also confirmed by a cycloheximide (CHX) chase assay ( Supplementary Fig. 2b). To investigate whether endogenous NEDD4 contributes to the regulation of WW45, we depleted endogenous NEDD4 in HEK293T cells using RNA interference (RNAi). All three short interfering RNAs (siRNAs) against NEDD4 (siNEDD4#1, siNEDD4#2 and siNEDD4#3) reduced NEDD4 levels and increased WW45 levels accordingly (Fig. 2c). This is also confirmed by the CHX chase assay with siRNAs that the halflife of WW45 was not declined in the presence of siNEDD4, but not either siScr or siITCH as a control ( Supplementary  Fig. 2c). Next, we investigated whether NEDD4 mediates the proteasomal degradation of WW45 by its ubiquitylation. GST-WW45 purified from E. coli was ubiquitylated in vitro  (a) WW45-interacting proteins were identified. Either mock-or FLAG-WW45-transfected HEK293T cells were immunoprecipitated with anti-FLAG M2 resin. Cells were pre-treated with 20 mM MG132 for 4 h. Immunoprecipitates were separated by SDS-polyacrylamide gel electrophoresis and subjected to coomassie blue staining or immunoblotting with indicated antibodies. Peptide sequences of NEDD4 detected from the analysis of mass spectrometry are shown. (b) WW45 and NEDD4 are endogenously associated with each other. Immunoprecipitation (IP) with IgG or anti-WW45 antibody was performed with HEK293T lysates treated with 20 mM MG132 for 4 h and analysed by immunoblotting with indicated antibodies. Asterisk and arrowhead designate IgG and immunoprecipitated WW45, respectively. (c) NEDD4 directly interacts with WW45 in vitro. Purified FLAG-NEDD4 and either GST or GST-WW45 were incubated with glutathione sepharose for 1 h. Precipitates were analysed by immunoblotting. (d) The N-terminus of WW45 is important for interaction with NEDD4. HisMax-NEDD4 was co-transfected with various FLAG-WW45 deletion mutants (WT, wild-type; 199C, N-terminal deletion; N198 and N321, C-terminal deletion, amino acids (a.a.) 1-198 and a.a. 1-321 region, respectively) in HEK293T cells. At 24 h post transfection, cells were treated with 20 mM MG132 for 4 h. Cell lysates were immunoprecipitated with anti-FLAG M2 resin and immunoblotted with indicated antibodies. (e) N-terminus of NEDD4 is required for association with WW45. MYC 9 -WW45 was co-transfected with various FLAG-NEDD4 constructs (WT, wild-type; N280, a.a. 1-280; WW, a.a. 281-620; DN280, a.a. 281-1000) in HEK293T cells. At 24 h post transfection, cells were treated with 20 mM MG132 for 4 h. Cell lysates were immunoprecipitated with anti-FLAG M2 resin and immunoblotted with indicated antibodies. upon incubation with bacteria-expressed His-NEDD4 (Fig. 2d). To examine whether NEDD4 promotes ubiquitylation of WW45 in vivo, MYC 9 -WW45 was co-expressed in HEK293T cells with FLAG-NEDD4 vectors (NEDD4 WT , NEDD4 C967S or NEDD4 DN280 ). WW45 was extensively ubiquitylated by NEDD4 WT , but barely ubiquitylated in the presence of either NEDD4 C967S or NEDD4 DN280 (Fig. 2e). Endogenous WW45 ubiquitylation by NEDD4 was also validated with siNEDD4 ( Supplementary Fig. 2d). Despite the major effect of NEDD4 on the ubiquitylation status of WW45, total cellular ubiquitin conjugates were not affected by the overexpression or knockdown (KD) of NEDD4. Taken together, these results demonstrate that NEDD4 directly binds to and ubiquitylates WW45, resulting in its proteasomal degradation.
NEDD4 associates with and destabilizes LATS kinase. Because WW45 is a key component of the Hippo pathway and thereby regulates apoptotic processes, we investigated whether NEDD4 is able to control apoptosis through WW45. When NEDD4 was depleted by siNEDD4 in HEK293T cells, apoptosis was induced and further augmented by the treatment of an apoptosis-inducing drug, staurosporine (STS; Fig. 3a). However, simultaneous KD of both NEDD4 and WW45 reversed STS-induced cell death albeit partially compared with what was seen with NEDD4 KD alone (the efficacy of shWW45 is shown in Supplementary Fig. 3a). It was also validated by the amount of cleaved poly [ADP-ribose] polymerase (PARP) and fluorescence-activated cell sorting analysis ( Fig. 3b and Supplementary Fig. 3b,c). These results possibly suggested that NEDD4 might act on additional targets besides   WW45 protein. Intriguingly, we noticed that NEDD4 depletion stabilized not only WW45 but also LATS2 (Fig. 3b). This is in agreement with the observation that protein levels for both LATS2 and WW45 were enhanced by the treatment of MG132 ( Fig. 3c and Supplementary Fig. 3d). Thus, it is plausible that NEDD4 may also regulate LATS2 as its specific E3 Ub-ligase. Coimmunoprecipitation assays showed that NEDD4 bound to LATS2 mostly through amino acids 281-660 region that encompass the WW domains ( Fig. 3d and Supplementary  Fig. 3e). Furthermore, LATS2 protein levels were decreased by NEDD4 WT in a dose-dependent manner and recovered by MG132 treatment, whereas unaffected by the NEDD4 C967S mutant ( Fig. 3e and Supplementary Fig. 3f). CHX chase assay also validated that LATS2 stability was rapidly decreased in the presence of NEDD4 ( Supplementary Fig. 3g). In contrast to WW45, both NEDD4 WT and NEDD4 DN280 mutant were able to destabilize LATS2 (Fig. 3 f) as NEDD4 DN280 can bind to LATS2 but not to WW45. Moreover, LATS2 was polyubiquitylated in vitro ( Supplementary Fig. 3h) and in vivo (Fig. 3g) by NEDD4 WT but not by NEDD4 C967S mutant. We also showed that NEDD4 was able to destabilize LATS1, another mammalian homologue of Drosophila Warts ( Supplementary Fig. 3i). Taken together, these data indicate that NEDD4 negatively regulates LATS1 and LATS2 as well as WW45.
As the kinase activity of MST1 is dispensable for WW45 stability 7,33 , we hypothesized that MST1 physically interrupts the interaction between NEDD4 and WW45. Consistently, along with the protective role of MST1 from the NEDD4-mediated degradation of WW45, there was a clear reduction in the binding proportion of WW45 with NEDD4 (Fig. 4c). Of note, no physical interaction was observed between NEDD4 and MST1. It was further validated by co-immunoprecipitation assay of WW45 that the binding degree to MST1 and that of NEDD4 were inversely correlated (Supplementary Fig. 4b).
NEDD4 retains YAP in the nucleus and increases the YAP transcriptional activity. As the activation of the Hippo signalling pathway is reflected by YAP localization, which is critical for YAP activity 13,34 , we investigated the effect of NEDD4 on the subcellular localization of YAP using nuclear/cytoplasmic fractionation and immunofluorescence staining. In the absence of Hippo activation signals, YAP shuttles between the cytoplasm and the nucleus 13,34 . When the Hippo pathway was activated by ectopic expression of WW45 and LATS2, YAP was found preferentially in the cytoplasm. Such cytoplasmic localization of YAP was further augmented by MST1 co-expression. In contrast, YAP was retained in the nucleus to a significant degree when NEDD4 was co-expressed ( Fig. 5a and Supplementary Fig. 5a).
These results indicate that NEDD4 keeps YAP in the nucleus by inhibiting the Hippo signalling. Because nuclear YAP functions as a transcriptional co-activator, we investigated how NEDD4 affects the expression of YAP downstream target genes using quantitative real-time reverse transcription PCR (RT-PCR) assay. As shown in Fig. 5b, NEDD4 WT but not NEDD4 C967S activated both CTGF (connective tissue growth factor; IGFBP-8) and BIRC5 (baculoviral IAP repeat-containing protein-5; apoptosis inhibitor 4; SURVIVIN) gene expression, both of which are known target genes of YAP 9,13 . Next, we tested the effect of NEDD4 on the transcriptional co-activator activity of YAP using the TOPflash luciferase reporter assay 35 . YAP was able to activate the TOPflash but not the FOPflash reporter. LATS2 co-transfection suppressed the TOPflash reporter almost to the basal level, and conversely, additional NEDD4 transfection led to the re-activation of the TOPflash reporter in accordance with the reduction of LATS2 protein (Fig. 5c). On the other hand, WW45-binding defective NEDD4 DN280 was not able to re-activate the WW45-driven TOPflash reporter, unlike NEDD4 WT . Intriguingly, there was no difference of LATS2-driven TOPflash reporter activity between NEDD4 DN280 and NEDD4 WT (Fig. 5d), suggesting that NEDD4 may differentially regulate WW45 and LATS2 protein levels. Taken together, NEDD4 modulates the subcellular localization and the transcriptional activity of YAP through the destabilization of WW45 and LATS2.
NEDD4 controls cell-density-dependent activation of the Hippo pathway. It has been established that cell-densitydependent activation of the Hippo pathway leads to YAP phosphorylation followed by inhibition of YAP transcriptional activity 13 . Consistently, YAP was seen in both nucleus and cytoplasm at low cell density but was localized in the cytoplasm when the cells were confluent (Fig. 6a). When NEDD4 was suppressed as in NEDD4 KD NIH3T3 stable cells, YAP was localized in the cytoplasm regardless of cell density. It is confirmed by transiently expressed NEDD4 in NIH3T3 cells that co-expression of NEDD4 WT , but not NEDD4 C967S , led to nuclear retention of YAP regardless of cell density ( Supplementary Fig. 6a). These results indicate that the E3 Ub-ligase activity of NEDD4 is necessary for turning off the celldensity-dependent activation of the Hippo signalling ( Supplementary Fig. 6a). We also showed that WW45, LATS2 and phosphorylated YAP protein levels were elevated as cell density increased (Fig. 6b), confirming that the Hippo pathway was activated when cells reached confluence. Meanwhile, unlike the protein counterparts, mRNA levels of NEDD4, LATS2 and WW45 remained unchanged regardless of the cell density, indicating that WW45 and LATS2 protein levels are posttranslationally regulated ( Supplementary Fig. 6b). It is of note that although NEDD4 acts as a negative regulator of WW45 and LATS2, NEDD4 protein levels were also increased at high cell density (Fig. 6b). It is well established that E3 Ub-ligases auto-ubiquitylate to control their own stability [36][37][38] . This appears to apply to NEDD4 as well. Specifically, we noticed that the protein level of the E3 Ub-ligase-defective NEDD4 C967S mutant was always higher than that of NEDD4 WT (Figs 2b and 3d), indicating that the stability of NEDD4 itself might be regulated by auto-ubiquitylation. Thus, we speculated that the elevation of NEDD4 protein levels at high cell density (shown in Fig. 6b) was due to a reduction in the E3 Ub-ligase activity of NEDD4. To address this hypothesis, we analysed the E3 Ub-ligase activity and stability of NEDD4 at low and high cell densities. Autoubiquitylation of NEDD4 at high cell density was declined drastically compared with that at low cell density ( Supplementary  Fig. 6c). Consistently, NEDD4 at high cell density was more stable than NEDD4 at low cell density, providing a basis for the increase in protein levels and activation of the Hippo pathway components (Supplementary Fig. 6d). Next, we examined the dependency of WW45 and LATS2 protein levels on NEDD4 using NEDD4 KD NIH3T3 stable cells. When NEDD4 was scarce, protein levels of WW45 and LATS2 were constantly high regardless of cell density (Fig. 6c) and their ubiquitylation was barely detected (Supplementary Fig. 6e). However, their specific interaction was not altered as cell density changed ( Supplementary Fig. 6f). These data suggest that the stabilization of WW45 and LATS2 at high cell density is mainly dependent on NEDD4 activity. YAP was phosphorylated even at low cell density in NEDD4 KD cells (Fig. 6c), indicating that elevated protein levels of WW45 and LATS2 by NEDD4 depletion lead to YAP phosphorylation. Moreover, quantitative real-time RT-PCR analyses revealed that YAP transcriptional activity was also dependent on NEDD4. Expression levels of YAP target genes, Ctgf, Birc5, Inhba (Inhibin beta A chain) and Mcl1 (induced myeloid leukemia cell differentiation protein Mcl-1 homologue; Bcl-2-related protein EAT/mcl1) 9,13 , were decreased at high cell density in NIH3T3 control cells but remained low in NEDD4 KD cells regardless of cell density (Fig. 6d and Supplementary Fig. 6g). Furthermore,  (Figs 6e,f). Taken together, our results demonstrate that cell-density-dependent activation of the Hippo pathway is due to the changes of NEDD4 E3 Ub-ligase activity. In addition, NEDD4 depletion leads to the activation of the Hippo pathway accompanied by reduced proliferation. NEDD4 acts as an essential regulator of the Hippo/Yorkie pathway in Drosophila. The Hippo signalling pathway was originally discovered in Drosophila and is highly conserved from flies to humans 1,2,6,10,21 . However, it is still unknown whether Drosophila Nedd4, the homologue of human NEDD4, functions in the Hippo pathway. Therefore, we sought to confirm and characterize the novel functions of Nedd4 as an essential modulator for the Hippo pathway using the genetically amenable Drosophila model organism. For this, we examined expression levels of expanded (ex), a well-known target of Hippo signalling, by using ex-lacZ reporter in wing imaginal discs with Nedd4 mutant clones. The result showed that ex-lacZ expression was diminished in Nedd4 mutant cells when compared with adjacent wild-type cells (Fig. 7a). These data demonstrate that Nedd4 acts as a regulator of Hippo pathway in wing imaginal disc. ARTICLE In Drosophila, it has recently been reported that the Hippo pathway is rapidly inactivated in intestinal enterocytes (ECs) in response to tissue injury, leading to the activation of transcriptional co-activator Yki 21,22,39,40 . Activated Yki is known to induce the expression of cytokine Unpaired-3 (Upd3) in ECs that in turn activates JAK-signal transducers and activators of transcription (STAT) signalling pathway in progenitor cells such as enteroblasts (EBs) and ISCs 21,22,39,41 . It has been shown that Hippo pathway inactivation in EC and subsequent JAK-STAT pathway activation in progenitor cells are of central importance for the epithelial renewal process (that is, ISC proliferation and differentiation into EB and EC) to replenish the damaged EC 21,22 . To investigate whether Nedd4 in EC is involved in epithelial cell renewal process via regulating the Hippo pathway, we examined infection-induced Upd3 expression in EC in a Nedd4-KD condition. When we introduced the expression of Nedd4-RNAi, we found that infection-induced Upd3 expression was greatly reduced (Fig. 7b), indicating that Nedd4 regulates the Hippo pathway in ECs. Two independent Nedd4-RNAi lines showed similar results. Furthermore, the infection-induced JAK-STAT activation (by monitoring STAT-dependent green fluorescent protein (GFP) reporter) in progenitor cells and augmented ISC proliferation/differentiation rate (by monitoring number of dividing cells using an anti-phosphorylated histone 3 (PH3) antibody and Su(H)Gbe-lacZ-positive EB accumulation) found in wild-type flies were almost completely abolished in animals carrying EC-specific Nedd4-KD (Fig. 7c-e). These results indicate that Nedd4-mediated Hippo pathway inactivation and subsequent Upd3 induction in ECs are required for epithelial cell renewal process following pathogen infection. Consistent with this notion, inactivation of Hippo pathway in EC, by using Wts-KD or Sav-KD, is sufficient to induce all events necessary for epithelial cell renewals (that is, Upd3 induction in ECs, STAT-GFP reporter activation in progenitor cells, ISC proliferation and EB accumulation) even in the absence of pathogen infection (Fig. 7f-i). Importantly, the high epithelial cell renewal rate induced by downregulating the Hippo pathway activity in EC was greatly attenuated by boosting the Hippo pathway activity through Nedd4-KD in EC (Fig. 7f-i). Consistent with Hippo-modulated Upd3 expression, high expression levels of different Hippo target genes such as DIAP1 and ex were also found in flies with reduced Hippo pathway activity (that is, Wts-KD flies), and this high expression level could be significantly diminished by enhancing the Hippo pathway activity (that is, Wts/Nedd4 double KD; Fig. 7j). As reduced Wts protein level in the intestine of Wts-KD animal can be augmented in Wts/Nedd4 double KD animals ( Supplementary  Fig. 7a), the suppression of the high epithelial cell renewal rates of flies with reduced Hippo pathway activity by Nedd4-KD is most likely due to a reciprocal compensatory effect of the two opposing KD phenotypes (for example, Hippo pathway downregulation by Wts-KD versus Hippo pathway upregulation by Nedd4-KD). Taken together, these results demonstrate that Nedd4 genetically interacts with the Hippo pathway, and that Nedd4 acts as an essential signalling component in the Hippo pathway in gut cell homeostasis.

Discussion
In the present study, we have identified NEDD4 as a new regulatory component of the Hippo signalling pathway. Our results show that NEDD4 plays a crucial role in regulating cell proliferation and apoptosis in both mammalian cells and Drosophila. NEDD4 depletion leads to escalated apoptosis and reduced proliferation, which are typical outcomes of the activated Hippo pathway. Accordingly, NEDD4 negatively regulates core components of the Hippo pathway, WW45 and LATS1/2. Both proteins are polyubiquitylated by NEDD4 and subsequently degraded by the 26S proteasome, leading to inactivation of the Hippo pathway.
Given that NEDD4 negatively regulates the stability of WW45 and that MST1 interferes with this destabilization, a new underlying mechanism of WW45 stabilization by MST1/2 can be speculated. We have shown that MST1/2 inhibits WW45 ubiquitylation by hindering its interaction with NEDD4 and the binding degree of MST1 to WW45 was greatly increased not only in the overexpressed conditions but also in the confluent culture conditions, which indicates that WW45 protein levels are regulated by a fine balance between NEDD4 and MST1. As the binding regions of WW45 to MST1 and to NEDD4 are different, a direct binding competition between them is not likely, as seen in the case of RASSF6 and MST2 (ref. 42). Although overall MST1 protein levels were not changed by cell density, phosphorylated MST1, which is associated with its activity, was increased at high cell density, therefore, it is plausible that MST1 becomes more active and predominantly interacts with WW45 at high cell density. Meanwhile, NEDD4 becomes inactive, which was monitored by autoubiquitylation and substrate ubiquitylation, but its interaction to WW45 remains unchanged at high cell density. Thus, it is most likely that the shifted binding preference of WW45 from NEDD4 to MST1 at high cell density may contribute to determine the WW45 protein levels.
On the other hand, high levels of either WW45 or MST1 were not able to protect LATS1/2 against NEDD4, which is consistent with previous findings that Warts protein levels can be regulated by alternative mechanism independent of Salvador and Hippo in Drosophila 34,43 . KIBRA was recently identified as a component of the mammalian Hippo signalling 44 in addition to genetic studies in Drosophila [45][46][47] . Similar to NEDD4, the WW domain of KIBRA is necessary for binding to LATS2 and KIBRA thus stabilizes and activates LATS2 by inhibiting ubiquitylation. Although how KIBRA is involved in the NEDD4-mediated LATS2 regulation remains to be elucidated, it is possible that KIBRA and NEDD4 compete with each other for LATS2. It has also been reported that LATS1 is regulated by ITCH, an E3 Ubligase that belongs to the NEDD4 family 29,30 . Thus, we sought to further delineate the specificity of NEDD4 on regulating the Hippo pathway. N-terminal amino-acid sequence alignment of (also known as NEDD4-2), ITCH, WWP1 and WWP2, showed that only NEDD4 has specific N-terminal 100-amino-acid region, which is critical for binding to WW45 compared with other family members ( Supplementary Fig. 1b). On the other hand, ITCH can interact with LATS1, but not with either WW45 or LATS2 ( Supplementary Fig. 1c-e), followed by LATS1 destabilization. It has recently been reported that LATS1 stability can be controlled by NEDD4 (ref. 31). Therefore, it would be possible that NEDD4 and ITCH regulate LATS1 together or independently of the other depending on physiological or cellular context and NEDD4 could function as the E3 Ub-ligase for both LATS1 and LATS2. Activation of the Hippo pathway is manifested by the activation of LATS whose stabilization and phosphorylation are both required 34,43 . As elevated LATS protein levels are able to act on YAP even without its phosphorylation 9,13 , it is very important to maintain LATS protein levels precisely. In this regard, it is worthy of note that NEDD4 is able to control LATS activation directly through the negative regulation of LATS stability and indirectly through the negative regulation of WW45, which is essential for the core signalling cascade of WW45-MST-mediated LATS phosphorylation. Thus, NEDD4 is able to block the activation of LATS efficiently and thereby tightly regulates the Hippo signalling pathway ( Fig. 8; see the corresponding legend for further details).
It has been reported that environmental cues (for example, cell density) determine the activation of the Hippo pathway 13,49 . When cells in culture reach confluence, the Hippo pathway is activated, thus ensuing phosphorylation and inactivation of YAP ultimately induce contact inhibition 13 . We also observed that phosphorylation and subcellular localization of YAP are indeed modulated by changes in cell density. In addition, it has been reported that WW45-and LATS2-KO mouse embryonic fibroblast cells lose contact inhibition, similar to cancer cells 50,51 . However, it is still largely unknown how the Hippo pathway becomes active as cell density increases. As shown in the present study, NEDD4 Ub-ligase activity seems essential for Hippo signalling in this regard as well. At low cell density, NEDD4 inhibits the activation of the Hippo pathway by maintaining WW45 and LATS proteins at low levels. NEDD4 becomes inactive at high cell density, resulting in the stabilization of WW45 and LATS followed by subsequent activation of the Hippo pathway. One of the intriguing observations presented here is that NEDD4 activity is altered by changes in cell density. NEDD4 becomes stabilized at high cell density. NEDD4 is known to be auto-inhibited by intramolecular interaction between C2 and HECT domains 37 , and calcium influx into cells releases C2 domain from HECT domain thereby activating the E3 Ub-ligase activity of NEDD4 (ref. 27). As epidermal growth factor receptor (EGFR) and/or acetylcholine receptor signalling pathways are able to elevate intracellular calcium levels and then activate NEDD4 (ref. 27), it would be of interest to test whether the Hippo pathway is in turn regulated by EGFR and/or acetylcholine receptor signalling via the activation of NEDD4. Moreover, NEDD4 WT is able to block the accumulation of cytoplasmic YAP even at high cell density, whereas the catalytically inactive NEDD4 C967S mutant fails to do so. NEDD4 is known to promote cell growth as an oncoprotein and NEDD4 KO mouse shows severe growth retardation 28 . Consistently, NEDD4 KD cells show slower cell proliferation and higher apoptosis, indicating that the Hippo pathway is abnormally activated by the dysregulation of WW45 and LATS. Importantly, simultaneous double KD of NEDD4 with WW45 or LATS2 partially restores cell growth or reduces apoptosis in part. Taken together, these data indicate that NEDD4 positively controls cell proliferation through the negative regulation of WW45 and LATS. This would in turn imply that NEDD4 affects cell growth and tumorigenesis via Hippo signalling.
The NEDD4-regulated Hippo signalling cascade is further validated in Drosophila. Our study suggests that infectioninduced Nedd4 activation leads to the inactivation of the Hippo pathway and subsequent Upd3 expression in ECs that in turn activates the JAK-STAT pathway for the transition from ISC to EB and EC. Nedd4-KD animals having defects in epithelial cell renewal showed severe mortality to gut infection ( Supplementary  Fig. 7b), and this high mortality could be significantly ameliorated by enhancing the Hippo pathway activity (that is, double KD flies for Sav-RNAi plus Nedd4-RNAi) or by re-introducing Nedd4 (Supplementary Fig. 7b). In addition to the involvement of Nedd4 in epithelial cell renewal in the gut, we found that Nedd4 is also involved in gut size determination via regulating the Hippo pathway ( Supplementary Fig. 7c). Furthermore, Nedd4-KD is sufficient to decrease adult wing sizes ( Supplementary Fig. 7d). All of these observations indicate that Nedd4 acts as an essential signalling component in gut cell homeostasis and host survival during gut-pathogen interactions by modulating the Hippo pathway activity.
Overall, our study defines a previously unknown level of control in the Hippo signalling and sheds a new light on the regulatory mechanism of the Hippo pathway establishing NEDD4 as an essential signalling component of the pathway.
Assays for ubiquitylation and pulse-chase analysis. For the in vitro ubiquitylation assay, purified GST-WW45 or His-LATS2 (0.8 mg) from E. coli was incubated at 37°C for 2 h with E1 (1 mg), Ubc5 (2 mg) and Ub (10 mg), ATP-regenerating system (50 mM Tris-HCl (pH 7.5), 5 mM MgCl 2 , 10 mM creatine phosphate, 5 U ml À 1 of phosphocreatine kinase and 5 mM ATP) and 1 Â complete protease inhibitor cocktail in the presence or absence of NEDD4 (1 mg) purified from HEK293T cells. After incubation, each sample was subjected to SDS-polyacrylamide gel electrophoresis and immunoblotting with anti-WW45 or anti-LATS2 antibody. For the pulse-chase analysis, transfected HEK293T cells were treated with CHX (200 mg ml À 1 ) for indicated times. Cell lysates were subjected to immunoblotting with appropriate antibodies.
Flow cytometry. HEK293T cells were transfected with appropriate siRNAs including scrambled control and treated with STS. Cells were collected, washed with PBS and fixed with 70% ethanol. Fixed cells were permeabilized by 0.3% Triton X-100 and stained in propidium iodide solution containing RNase A (250 mg ml À 1 ). Propidium iodide-stained cells were then analysed for their DNA content using a fluorescence-activated cell sorting instrument (BD Biosciences).
Immunofluorescence staining. Cells on gelatin-coated coverslips were fixed with 2% formaldehyde in PBS for 30 min at room temperature and permeabilized with 0.5% Triton X-100 in PBS. All subsequent dilutions and washes were performed with PBS containing 0.1% Triton X-100 (PBS-T). Nonspecific binding sites were masked by incubation for 30 min with blocking solution consisting of 10% goat serum, 1% BSA and 1% gelatin in PBS. Cells were incubated with primary antibody for 1 h,