Nedd4-induced monoubiquitination of IRS-2 enhances IGF signalling and mitogenic activity

Insulin-like growth factors (IGFs) induce proliferation of various cell types and play important roles in somatic growth and cancer development. Phosphorylation of insulin receptor substrate (IRS)-1/2 by IGF-I receptor tyrosine kinase is essential for IGF action. Here we identify Nedd4 as an IRS-2 ubiquitin ligase. Nedd4 monoubiquitinates IRS-2, which promotes its association with Epsin1, a ubiquitin-binding protein. Nedd4 recruits IRS-2 to the membrane, probably through promoting Epsin1 binding, and enhances IGF-I receptor-induced IRS-2 tyrosine phosphorylation. In thyroid FRTL-5 cells, activation of the cyclic AMP pathway increases the association of Nedd4 with IRS-2, thereby enhancing IRS-2-mediated signalling and cell proliferation induced by IGF-I. The Nedd4 and IRS-2 association is also required for maximal activation of IGF-I signalling and cell proliferation in prostate cancer PC-3 cells. Nedd4 overexpression accelerates zebrafish embryonic growth through IRS-2 in vivo. We conclude that Nedd4-induced monoubiquitination of IRS-2 enhances IGF signalling and mitogenic activity. Phosphorylation of insulin receptor substrate (IRS)-1/2 by insulin-like growth factor (IGF)-I receptor tyrosine kinase is essential for IGF signalling. Here, the authors show that monoubiquitination of IRS-2 by the ubiquitin ligase Nedd4 recruits IRS-2 to the cell membrane and increases IRS-2 phosphorylation and IGF signalling.

I nsulin-like growth factor (IGF)-I and IGF-II induce cell proliferation, differentiation, survival and migration in many cell types 1 . In vivo, IGFs play important roles in prenatal and postnatal development 2 . Reduced IGF activities cause growth retardation, whereas elevated IGF activities cause overgrowth and cancer development 3 . The growth-promoting activities of IGFs or their homologues are evolutionarily conserved in metazoans 4 . IGFs bind to the IGF-I receptor (IGF-IR) in the plasma membrane and induce the activation of its intrinsic tyrosine kinase and auto-phosphorylation. The activated IGF-IR phosphorylates intracellular substrates including insulin receptor substrate (IRS)-1 and IRS-2. Phosphotyrosyl IRSs are in turn recognized by SH2 domain-containing proteins such as the p85 regulatory subunit of phosphatidylinositol 3-kinase (PI3K). These events trigger activation of downstream signalling pathways, leading to various biological actions 5 .
IGF activities are often modulated by other hormonal factors 6,7 . IRSs serve as 'signal nodes' on which IGF signals and other signals converge. Indeed, IGF-induced IRS tyrosine phosphorylation is potentiated by other hormonal stimuli, leading to amplification of IGF activities [8][9][10] . In addition, oncogenic pathways frequently cause excess IRS-mediated signalling, which contribute to tumour growth and metastasis 11 . We have shown that IRSs form high-molecular-mass complexes (IRSomes) and different complexes are formed depending on IRS isoform, cell types or hormone stimulation 12 . Subsequent study showed that IRS-associated proteins potentially modulate the availability of IRSs to receptor tyrosine kinases 12 , prompting us to identify IRS-associated proteins and to understand their functions.
In this study, we identified HECT-type ubiquitin ligase Nedd4 as a novel IRS-2-associated protein. Nedd4 is expressed in a variety of tissues and regulates the trafficking and/or degradation of its substrate proteins 13,14 . IGF-IR is reported to be a putative Nedd4 substrate, but the effects of Nedd4 on IGF-IR are controversial 15,16 . Nedd4 knockout mice show fetal growth retardation and perinatal lethality 16 , indicating a crucial role for Nedd4 in embryonic growth. Nedd4 is overexpressed in several types of cancers and functions as an oncogenic protein [17][18][19][20][21] . We provide multiple lines of evidence, suggesting that Nedd4 conjugates monoubiquitin to IRS-2 and recruits IRS-2 to the plasma membrane, thereby enhancing IRS-2 tyrosine phosphorylation by IGF-IR and IGF-I mitogenic activity.

Results
Nedd4 is an IRS-2-associated protein. We previously reported that pretreatment of thyroid FRTL-5 cells with thyroid stimulating hormone (TSH) or other cAMP-generating reagents enhances IGF-I-induced IRS-2 tyrosine phosphorylation and increased cell proliferation 10,22,23 . Our study also suggested that some IRS-2associated protein(s) might increase the availability of IRS-2 to the IGF-IR tyrosine kinase 12 . Using co-immunoprecipitation followed by matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (MS) analysis, we identified Nedd4 as an IRS-2associated protein in FRTL-5 cells treated with dibutyryl cAMP ( Supplementary Fig. 1a,b). This association was confirmed by coimmunoprecipitation and immunoblotting ( Supplementary  Fig. 1c). Co-immunoprecipitation analysis using HEK293 cells expressing exogenous IRS-2 together with Nedd4 or Nedd4-2, a related protein in the Nedd4 family 13 , revealed that IRS-2 is associated only with Nedd4 ( Supplementary Fig. 1d).
We further investigated the IRS-2-binding site in Nedd4 in greater detail. The Nedd4 C2 domain can bind intramolecularly to the HECT domain 24 . Overexpression of HECT domain in HEK293 cells decreased the association of Nedd4 1-551 with IRS-2 (Fig. 1c), indicating that the Nedd4 HECT domain and IRS-2 may competitively bind to the Nedd4 C2 domain. This raised the possibility that the HECT domain and IRS-2 may bind to an overlapping site in the C2 domain. To test this idea, we modelled the Nedd4 intramolecular binding sites based on the reported structure of Smurf2, a Nedd4-type ubiquitin ligase 24 . Based on the modelling results ( Supplementary Fig. 2b), several mutants were generated. Mutation of Ile 95 and Leu 96 to Ala, or Ser 126 and Leu 127 to Ala inhibited the association of Nedd4 1-502 with Nedd4 503-887 ( Supplementary Fig. 2c). These data suggest that these residues are part of the intramolecular binding sites. These mutations also inhibited the association of Nedd4 with IRS-2 ( Fig. 1d), indicating their roles in IRS-2 binding.
Nedd4 induces IRS-2 monoubiquitination. We investigated whether Nedd4 ubiquitinates IRS-2 using a cell-free ubiquitination assay. The results showed that Nedd4 caused a prominent upward mobility shift of IRS-2, a typical pattern of protein ubiquitination (Fig. 2a). Furthermore, overexpression of Nedd4 in HEK293/HEK293T cells increased the ubiquitination signal in IRS-2 immunoprecipitates (Fig. 2b). The signal was observed when proteins that bound to IRS-2 (for example, Nedd4) were removed by denaturing the samples before immunoprecipitation (Fig. 2b), indicating that the observed signal arose from ubiquitinated IRS-2 itself. The detected signal was 20-30 kDa larger than the intact IRS-2, indicating the conjugation of a small number of ubiquitin molecules. The exact number was unclear, as the IRS-2 mobility may be changed by other posttranslational modifications in addition to ubiquitination. Nedd4-induced IRS-2 ubiquitination was not affected by IGF-I stimulation ( Supplementary Fig. 3a). Nedd4 overexpression also induced IGF-IR ubiquitination as previously reported 15 , but it did not induce IRS-1 ubiquitination ( Supplementary Fig. 3a).
Two antibodies were used to examine whether Nedd4-induced IRS-2 ubiquitination is monoubiquitination or polyubiquitination. Anti-ubiquitin antibody P4D1 can recognize both monoubiquitin and polyubiquitin chains, whereas FK1 only recognizes polyubiquitin chains 25 . IRS-2 ubiquitination was detected only with P4D1 (Fig. 2c). We also compared Nedd4-dependent IRS-2 ubiquitination in cells overexpressing haemagglutinin (HA)tagged ubiquitin (HA-Ub) or modified HA-Ub in which all Lys residues were replaced with Arg (HA-UbK0). The molecular weight of IRS-2 conjugated with HA-UbK0 was similar to that with HA-Ub (Fig. 2d). IRS-2 conjugated with HA-UbK0 was less than that with HA-Ub, possibly due to low expression level of HA-UbK0 in our experimental system. IRS-2 ubiquitination was also detected when cells were overexpressing Nedd4 Y592A or Nedd4 F694A that can conjugate single ubiquitin to substrates but lack ubiquitin chain elongation activity 26 (Fig. 2e). Taken together, these results indicated that IRS-2 ubiquitination by Nedd4 occurs as monoubiquitination at single or multiple lysine residues.
To map the IRS-2 ubiquitination sites, we generated a series of IRS-2 mutants in which all Lys residues in a defined region are replaced with Arg (Fig. 2f). Mutation of all Lys residues in the C1-C3 region (C1-C3 KR) abolished IRS-2 ubiquitination (Fig. 2g,h), indicating that IRS-2 ubiquitination sites are located ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7780 in this region. Mutation in the C3 region largely decreased IRS-2 ubiquitination (Fig. 2g,h), indicating that Lys residues in the C3 region are more important. Further analysis using site-directed mutagenesis targeting Lys residues in the C3 region revealed that Lys 1331 is the most frequently ubiquitinated site (Fig. 2g,h). In parallel experiments, we used liquid chromatography-tandem mass spectrometry (LC-MS/MS) to identify ubiquitination sites in HEK293T cells overexpressing Nedd4. We detected several IRS-2 peptides containing the ubiquitin remnant motif (K-e-GG) marking ubiquitinated proteins after trypsinolysis ( Supplementary  Fig. 3b). Importantly, the peptide containing Lys 1331 with the ubiquitin remnant was detected ( Supplementary Fig. 3c). Quantitative MS analysis revealed that the peptide containing Lys 1331 with ubiquitin remnant was intensely increased by Nedd4 overexpression ( Fig. 2I and Supplementary Fig. 3d,e).
IRS-2 monoubiquitination enhances IGF-I signalling. We investigated the effects of Nedd4 overexpression on IGF-I signalling in HEK293 cells. Although Nedd4 overexpression did not affect the levels of IGF-IR or IGF-I-induced IGF-IR tyrosine phosphorylation (Fig. 3a,b and Supplementary Fig. 4a), it enhanced IGF-I-induced IRS-2 tyrosine phosphorylation and increased the amounts of PI3K bound to IRS-2 ( Fig. 3a,b). Timecourse experiments showed that Nedd4 increased the magnitude of IGF-I-induced IRS-2 tyrosine phosphorylation at 1 and 3 min (Fig. 3c). In contrast, Nedd4 overexpression resulted in modest decreases in the total IRS-1 levels and IGF-I-induced IRS-1 tyrosine phosphorylation ( Supplementary Fig. 4b,c). Nedd4 overexpression did not affect the amounts of PI3K bound to IRS-1 ( Supplementary Fig. 4b).
We next investigated the underlying mechanism. In mouse embryonic fibroblasts, Nedd4 was reported to maintain cell-surface IGF-IR levels by suppressing the abundance of the adaptor protein Grb10 (ref. 16). However, Nedd4 overexpression neither affected cell-surface IGF-IR levels nor Grb10 levels in HEK293 cells ( Supplementary Fig. 4c,d)  was reduced by the deletion of the Nedd4 amino-terminal region (IRS-2-binding region; Supplementary Fig. 4e) and was abolished by the substitution of Nedd4 Cys 854 with Ser, which inactivates Nedd4 ubiquitin ligase activity 14 (Fig. 3a,b). These results suggest that the Nedd4-induced IRS-2 ubiquitination leads to the enhancement of IRS-2 tyrosine phosphorylation.
To elucidate whether monoubiquitination of IRS-2 is required for the increased IGF signalling, we used several IRS-2 ubiquitination site mutants. Both IRS-2 C1-C3 KR mutation  and K1331R mutation suppressed Nedd4-dependent increases in IRS-2 tyrosine phosphorylation (Fig. 3d), indicating that monoubiquitination at IRS-2 Lys 1331 is required for Nedd4-induced enhancement of IRS-2 tyrosine phosphorylation. As Lys 1331 is near the IRS-2 carboxy terminus (Fig. 2f), we also generated IRS-2 fused with monoubiquitin at the C terminus. When cells overexpressing this chimeric protein were stimulated with IGF-I, it was tyrosine phosphorylated and bound to PI3K more intensely than intact IRS-2 (Fig. 3e). Therefore, fusion of a single ubiquitin molecule to IRS-2 is sufficient to enhance its tyrosine phosphorylation in response to IGF-I.
Ubiquitin is recognized by a number of ubiquitin-binding proteins. These proteins target the ubiquitinated proteins to proteasome, sorting machinery or de-ubiquitinating enzymes 27 . Ile 44 of ubiquitin has been shown to be crucial for its interaction with ubiquitin-binding proteins 27 . When the Ile 44 in the IRS-2-ubiquitin chimeric protein was substituted to Ala, Nedd4-dependent increases in IRS-2 tyrosine phosphorylation was suppressed (Fig. 3e), suggesting that recognition of IRS-2 by ubiquitin-binding protein(s) is required for the Nedd4-dependent increases in IRS-2 tyrosine phosphorylation.
Epsin1 mediates the Nedd4 effects on IGF-I signalling. To identify proteins that recognize ubiquitinated IRS-2, we studied several ubiquitin-binding proteins in the endocytosis machinery, including Eps15, Eps15R, Epsin1, HRS and TSG101 (ref. 28). Among them, the association of Epsin1 with IRS-2 was increased by overexpression of Nedd4 (Fig. 4a) and the IRS-2-ubiquitin chimeric protein ( Supplementary Fig. 5a). Nedd4 overexpression did not affect the association of the other ubiquitin-binding proteins with IRS-2 ( Supplementary Fig. 5b). The association of IRS-2 with Epsin1 was suppressed by the mutation of IRS-2 ubiquitination sites ( Fig. 4b) and by the deletion of Epsin1 ubiquitin-interacting motifs (UIMs) 29 (Fig. 4b and Supplementary Fig. 5c). These results indicate that Epsin1 UIMs bind to the ubiquitin conjugated on IRS-2. Further analysis revealed that Epsin1 knockdown suppressed the effect of Nedd4 overexpression on IRS-2 tyrosine phosphorylation, while it had no effect on IGF-IR tyrosine phosphorylation (Fig. 4c). These results indicate that ubiquitinated IRS-2 is bound to Epsin1 and this molecular interaction is required for the enhanced IRS-2 tyrosine phosphorylation in response to IGF-I.
Epsin1 is localized at the plasma membrane, especially at sites where clathrin-coated pits will be formed 30 . Subcellular fractionation analysis showed that Nedd4 overexpression increased IRS-2 levels in the membrane fraction when cells were stimulated with IGF-I (Fig. 4d). These data suggested that Nedd4 recruits IRS-2 to the plasma membrane, where IRS-2 is effectively phosphorylated by the IGF-IR tyrosine kinase.
Nedd4 enhances IGF-I signalling in thyrocytes. Prolonged pretreatment of nontransformed thyroid FRTL-5 cells with TSH or other cAMP-generating reagents is known to enhance IGF-I-dependent IRS-2 tyrosine phosphorylation and DNA synthesis 10,22,23 . As shown in Fig. 5a and Supplementary Fig. 6a, dibutyryl cAMP treatment of FRTL-5 cells increased the association of endogenous Nedd4 with IRS-2. Pull-down analysis using GST-IRS-2 (aa 1-329) or GST-Nedd4 (aa 1-226) showed that cAMP treatment increases the affinity of Nedd4 for IRS-2 (Fig. 5b). In comparison, no association of IRS-2 with Nedd4-2 was detected (Fig. 5a). It should be noted that cAMP treatment changed the IRS-2 electrophoretic mobility ( Fig. 5a and subsequent figures). Our previous study showed that cAMP treatment does not change IRS-2 levels and the cAMP treatmentinduced IRS-2 mobility shift is due to its Ser/Thr phosphorylation 10 . Although Nedd4 was reported to interact with IGF-IR previously in other cell types 15 , our analysis indicated that IGF-IR was not co-immunoprecipitated with IRS-2 ( Supplementary Fig. 6b), indicating that Nedd4 associates with IGF-IR and IRS-2, independently. Next, we tested the effects of cAMP treatment on IRS-2 ubiquitination. Immunoblotting analysis with P4D1 and FK1 revealed that cAMP treatment increased monoubiquitination of IRS-2 ( Fig. 5c). Subcellular fractionation showed that cAMP treatment markedly increased the amounts of IRS-2 in the membrane fraction (Fig. 5d).
We then studied the role of Nedd4 in IGF-I signalling and mitogenic activity in FRTL-5 cells. Nedd4 knockdown neither affected IGF-I-induced IGF-IR tyrosine phosphorylation (Fig. 5e,f) nor Grb10 levels ( Supplementary Fig. 6c). cAMP pretreatment enhanced IGF-I-induced IRS-2 tyrosine phosphorylation and this was suppressed by Nedd4 knockdown (Fig. 5e,f) but not by Nedd4-2 knockdown ( Supplementary Fig. 6d). cAMP pretreatment enhanced IGF-I-induced DNA synthesis and this was abolished by Nedd4 knockdown (Fig. 5g). These data indicated that Nedd4 is required for cAMP-dependent enhancement of IGF-I signalling and mitogenic activity.
Nedd4 enhances IGF-I signalling in prostate cancer cells. We also investigated the possible role of Nedd4 in IGF-I action in prostate cancer cells. We confirmed that Nedd4 associated with IRS-2 in prostate cancer DU145 cells and PC-3 cells (Fig. 6a). Nedd4 knockdown did not affect IGF-I-induced IGF-IR tyrosine phosphorylation in PC-3 cells, whereas it decreased IRS-2 tyrosine phosphorylation and the amounts of PI3K bound to IRS-2 ( Fig. 6b). Nedd4-2 was not detected in these cells ( Supplementary  Fig. 7). Epsin1 knockdown also decreased IGF-I-induced IRS-2 tyrosine phosphorylation (Fig. 6c), supporting our conclusion that Epsin1 mediates Nedd4-induced enhancement of IGF-I signalling. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7780 ARTICLE Cultured PC-3 cells secrete IGF-I and -II into media, which stimulates their proliferation 31,32 . We observed that IGF-I stimulation increased DNA synthesis in PC-3 cells and this was decreased below the basal levels by Nedd4 knockdown as well as by IRS-2 knockdown (Fig. 6d), indicating that both Nedd4 and IRS-2 are necessary for their proliferation in response to endogenously secreted IGFs and exogenously added IGF-I. Together, these data indicate the requirement of the Nedd4-IRS-2 complex for maximum IGF-I signalling as well as mitogenic activity in PC-3 cells.
To elucidate Nedd4 functions in an in vivo setting, we used the zebrafish embryo model. The IGF signalling pathway is highly conserved and previous studies suggest that IGFs regulate (a-c) Effects of Nedd4 overexpression on IGF-I signalling. HEK293 cells overexpressing Nedd4 or Nedd4 C854S (a ubiquitin ligase inactive mutant) together with IRS-2 were serum starved, followed by IGF-I stimulation (100 ng ml À 1 ) for 1 min (a) or the indicated durations (c). Lysates were subjected to immunoprecipitation and immunoblotting using the indicated antibodies. The results of a were quantified by densitometric analysis. IGF-I-induced tyrosine phosphorylation levels of IGF-IRb and IRS-2 were normalized to their protein levels in immunoprecipitates and p85 PI3K bound to IRS-2 was normalized to IRS-2 levels in immunoprecipitates. The means ± s.d. of three independent experiments are shown in b. *Significant difference from control (Po0.05, one-way analysis of variance (ANOVA) followed by Tukey-Kramer test). (d,e) Effects of the expression of IRS-2 ubiquitination site mutants (d) and an IRS-2 -ubiquitin chimeric protein (e) on IRS-2 tyrosine phosphorylation. HEK293 cells overexpressing indicated proteins were subjected to experiments similar to a. Ub I44A , ubiquitin unable to interact with ubiquitin-binding domains. IGF-I-induced tyrosine phosphorylation levels of IRS-2 and PI3K bound to IRS-2 were normalized to IRS-2 protein levels in immunoprecipitates. Means ± s.d. of six (d) or three (e) independent experiments are shown. *Significant difference from control (Po0.05, one-way ANOVA followed by Tukey-Kramer test); NS, not significant. zebrafish embryonic growth by promoting cell proliferation and survival, without changing developmental patterning 33,34 . As shown in Fig. 7a-c, Nedd4 overexpression increased zebrafish embryo growth (body length) and developmental timing (somite number) without major organ loss or patterning abnormalities. These biological actions of Nedd4 were abolished by knocking down Irs2, using morpholino antisense oligos (MO) (Fig. 7a-c). Specific translation block with Irs2 MO and Nedd4 overexpression were confirmed ( Supplementary Fig. 8a-c). Taken together, these data indicate that Nedd4 accelerates zebrafish embryonic growth through Irs2.

Discussion
In this study, we identified Nedd4 as a novel IRS-2-associated protein and elucidated the molecular mechanisms of the complex formation, IRS-2 ubiquitination and role of this interaction in IGF action in normal and cancer cells in vitro, as well as in zebrafish embryos in vivo.
Wiesner et al. 24 suggested that the Nedd4 C2 domain binds to its HECT domain and this intramolecular binding inhibits its ubiquitin ligase activity. In this study, we discovered that IRS-2 and the Nedd4 HECT domain competitively bind to the C2 domain. IRS-2 binds to the Nedd4 C2 domain in a region overlapping with the HECT-binding site. Thus, we speculate that the dissociation of the C2 and HECT domains not only restores Nedd4 ligase activity but also opens up a substrate-binding site, which enables its recognition of some substrates, including IRS-2. It cannot be ruled out that other factor(s) mediate Nedd4-IRS-2 interaction. Crystal structure analysis will aid in understanding the molecular basis of the regulation of this complex formation.
This study demonstrates that Nedd4 induces monoubiquitination at single or multiple Lys residues in the IRS-2 C-terminal region. Ubiquitinated IRS-2 in FRTL-5 cells showed a little higher molecular weight than that in HEK293 cells (Figs 2b and 5c), indicating a different number of monoubiquitinated lysine residues. Nedd4 conjugated more ubiquitin molecules to IRS-2 in a cell-free system than in cells (Fig. 2a). Some ubiquitin molecules may be removed from IRS-2 by de-ubiquitination enzymes in cells. In fact, we observed significant levels of deubiquitination enzyme USP7 associated with IRS-2 in HEK293 cells 35 . Further analysis is needed to elucidate how Nedd4 induces monoubiquitination of IRS-2.
Several ubiquitin ligases other than Nedd4 have been shown to ubiquitinate IRS-1/2, but they all induce proteasomal degradation of IRS-1/2 and suppress IGF/insulin signalling [36][37][38][39] . In contrast, we demonstrated that Nedd4 enhances IGF-I signalling in this study, indicating a unique role for Nedd4 in IGF signal regulation. Conjugation of polyubiquitin chains consisting of four or more ubiquitin molecules to substrate proteins is required for their proteasomal degradation 40 , which can explain why Nedd4 does not induce IRS-2 degradation. Our data suggest that Nedd4 increases the association of IRS-2 with Epsin1 and this in turn promotes the translocation of IRS-2 from the cytosol to the plasma membrane. Epsin1 UIMs mediate the association of ubiquitinated IRS-2 with Epsin1. Epsin1 is localized to the plasma membrane and facilitates the formation of clathrincoated invaginations through its other domains and motifs (schematically represented in Fig. 4b) 30,41 . We speculate that Epsin1 may recruit the ubiquitinated IRS-2 to the position adjacent to the plasma membrane through physical interaction. As IGF-IR internalizes through the clathrin-dependent pathway 42 , it is attractive to postulate that Nedd4 may induce the co-localization of IRS-2 and IGF-IR at a specific plasma membrane compartment, where high levels of Epsin1 are present and clathrin-coated pits are forming. Their co-localization facilitates phosphorylation of IRS-2 by the IGF-IR tyrosine kinase. This model is strongly supported by our result that Epsin1 knockdown suppressed the effect of Nedd4 on IRS-2 tyrosine phosphorylation (Fig. 4c). It is unclear why different IGF-Idependency of IRS-2 translocation was shown in subcellular fractionation of HEK293 cells and FRTL-5 cells (Figs 4d and 5d). Thyroid epithelial cells synergistically proliferate in response to IGF-I and TSH in vitro [43][44][45] , which is thought to be important for thyroid morphogenesis and thyroid hormone homeostasis 46 . We previously reported that TSH or other cAMP-generating reagents enhance IGF-I-induced cell proliferation in thyroid FRTL-5 cells 44 . In this study, we showed that cAMP stimulus induces the association of Nedd4 with IRS-2, IRS-2 monoubiquitination and recruitment of IRS-2 to the membrane fraction. Nedd4 is required for the cAMP-dependent enhancement of IRS-2 tyrosine phosphorylation and DNA synthesis induced by IGF-I. We conclude that cAMP stimulus triggers the mechanisms by which Nedd4 enhances IRS-2-mediated signalling. In various endocrine tissues, IGFs and tropic hormones synergistically regulate cell proliferation, differentiation and hormone synthesis 6 . As tropic hormones generally use the cAMP pathway, we speculate that Nedd4 may play roles in the cross-talk between IGFs and other hormones.
In this study, we also detected the Nedd4-IRS-2 complex in prostate cancer cells. Nedd4 knockdown suppressed IGF-Iinduced IRS-2 tyrosine phosphorylation and DNA synthesis, suggesting that Nedd4 enhances IRS-2-mediated signalling and IGF-I mitogenic activity. It has been reported that Nedd4 promotes prostate cancer development through the ubiquitination and degradation of phosphatase and tensin homologue (PTEN) 17 . However, PC-3 cells are deficient in PTEN 47 , excluding the possibility that the effects of Nedd4 are mediated by the regulation of PTEN in our experiments. The Nedd4-IRS-2 binding may represent a novel drug target at least in prostate cancer. Accumulating evidence suggests that Nedd4 is overexpressed in various cancers and functions as a tumour-promoting factor [17][18][19][20][21] . It is important to investigate the contribution of the Nedd4-IRS-2 complex to the development of various cancers.
The results of this study show that Nedd4 accelerates zebrafish embryonic growth through Irs2. In mice, Nedd4 deficiency causes delayed embryonic growth 16 . In both cases, Nedd4 modulates embryonic growth without major developmental abnormality of any specific organs. Given that IGF signalling deficiency in various species causes growth retardation without patterning abnormalities [2][3][4]33 , these results imply that Nedd4 is an evolutionarily conserved factor that plays a positive role in embryonic growth, possibly through modulating IGF-IR-IRS-2 signalling activities.
It has been reported that Nedd4 acts on multiple IGF signal transducers. As described above, in mouse embryonic fibroblasts, Nedd4 is required to maintain the cell-surface IGF-IR levels 16 . In contrast, others reported that Nedd4 induces degradation of IGF-IR 15,48,49 . A recent report suggested that Nedd4 may enhance IGF-I-induced IRS-1 tyrosine phosphorylation by antagonizing PTEN, which functions as a tyrosine phosphatase for IRS-1 (ref. 50). Nedd4 has also been reported to regulate the stability and subcellular localization of Akt and PTEN 17,51,52 . In our study, Nedd4 affected neither total nor cell-surface IGF-IR levels, and had little effect on the signalling downstream of IRS-1. These discrepancies may be due to different experimental conditions or cellular models used. It is also possible that these multiple sites of Nedd4 action allow the fine-tuning of IGF signalling in a context-dependent manner.
In conclusion, our study demonstrated that Nedd4 enhances IGF-I signalling and mitogenic activity through a novel mechanism (Fig. 7d). Nedd4 associates with IRS-2 and conjugates monoubiquitin to IRS-2. Ubiquitinated IRS-2 is in turn  recognized by Epsin1, which possibly recruits IRS-2 to the plasma membrane. Consequently, Nedd4 enhances IGF-I-induced tyrosine phosphorylation of IRS-2 by the IGF-IR kinase, which leads to the augmentation of IGF-I signalling and mitogenic activity.
Rat IRS-1 complementary DNA and human IRS-2 cDNA were prepared as described elsewhere 53 . Mouse Nedd4 cDNA was kindly donated by Dr Sharad Kumar (Queensland University of Technology, Brisbane, Australia.). Mouse Nedd4-2 cDNA was amplified by PCR from mouse brain cDNA library. Ubiquitin cDNA was kindly provided by Dr Takeshi Imamura (Japanese Foundation for Cancer Research, Tokyo, Japan). Rat Epsin1 cDNA and human USP15 cDNA were amplified by PCR from H4IIE hepatocyte cDNA library and Huh7 hepatocyte cDNA library, respectively. Partial zebrafish Irs2 cDNA (NM_200315) was amplified by PCR from zebrafish embryo cDNA library. It contains 5 0 -untranslated region (UTR) and translation start site of Irs2. cDNAs with deletion or site-directed mutation were generated by standard restriction enzyme-or PCR-based methods.
Other chemicals were of the reagent grade available commercially.
Cell culture and transfection. MS analysis of IRS-2-associated proteins. Immunoprecipitates with antibodies against IRS-2 C terminus were subjected to SDS-PAGE, followed by gel staining using SilverQuest (Life Technologies). Gel slices were destained according to the protocols attached to the gel staining kits and then equilibrated with 50 mM NH 4 HCO 3 for 15 min. After the reagent was removed, gel slices were mixed with dehydration buffer (50 mM NH 4 HCO 3 , 50% CH 3 CN) and vortexed for 15 min. After the dehydration procedure was repeated again, gels were dried by vacuum evaporator centrifuge VEC-260 (ASAHI GLASS, Chiba, Japan). Gels were then reswollen in reducing buffer (10 mM dithiothreitol, 25 mM NH 4 HCO 3 ) and incubated for 1 h at 56°C. After the equal volume of alkylation buffer (55 mM iodoacetamide, 25 mM NH 4 HCO 3 ) was added, the mixtures were shaded and vortexed for 30 min at room temperature. Gels were then equilibrated with 50 mM NH 4 HCO 3 , dehydrated with 50% CH 3 CN and dried by VEC-260 as described above. After reswelling on ice in digestion buffer (10 mg ml À 1 Trypsin Gold, Mass Spectrometry Grade (Promega, Madison, WI, USA), 50 mM NH 4 HCO 3 , 5 mM CaCl 2 ), gels were incubated for 16 h at 37°C. To extract the tryptic peptides, 5% trifluoroacetic acid (TFA) was added and the supernatant was recovered. Gels were then vortexed in 5% TFA/30% CH 3 CN for 5 min and sonicated for 20 min, and the supernatant was recovered. The same procedure was repeated using 5% TFA/50% CH 3 CN and 5% TFA/70% CH 3 CN sequentially. All supernatants were collected and concentrated to 10 ml by VEC-260 and TFA (final concentration, 0.1%) was added to the samples. To desalt the sample, ZipTip mC18 (Millipore) was swollen in 0.1% TFA/50% CH 3 CN and equilibrated with 0.1% TFA/2% CH 3 CN, and the samples were then adsorbed and washed with 0.1% TFA/2% CH 3 CN. The tryptic peptides were eluted with 1 ml of 0.1% TFA/66% CH 3 CN onto a 100-well gold sample plate (Applied Biosystems, Framingham, MA, USA), mixed with 1 ml of matrix solution (10 mg ml À 1 a-cyano-4-hydroxy cinnamic acid in 0.2% TFA/60% CH 3 CN) and dried at room temperature. The samples were analysed by matrixassisted laser desorption/ionization-time-of-flight MS Voyager-DE STR (Applied Biosystems) in peptide-sensitivity reflector mode. The spectra were obtained by the accumulation of 200-1,000 consecutive laser shots and calibrated by mass values of auto-digested trypsin peptides. The mass data were subjected to MASCOT PMF search tools (Matrix Science, London, UK). Na 3 VO 4 , 10 mg ml À 1 p-nitrophenyl phosphate, protease inhibitor cocktail (Sigma-Aldrich, P8340)). In pull-down assay, lysates were incubated with 100 pmol of GST-fused proteins for 2 h and then 10 ml of GST-Sepharose beads (GE Healthcare). After 1 h incubation, beads were washed four times with ice-cold lysis buffer and proteins were subjected to SDS-PAGE followed by immunoblotting.
Immunoprecipitation and immunoblotting. Cells were lysed with ice-cold lysis buffer. The lysates were centrifuged at 15,000g for 10 min at 4°C. The protein assay of the supernatant was carried out using a protein assay kit (Bio-Rad, Hercules, CA, USA). Equal amounts of proteins of each sample were mixed with 1/4 volume of 4 Â Laemmli's buffer (0.4 M Tris-HCl pH 6.8, 8% SDS, 20% glycerol, 10% 2-mercaptoethanol, 0.2% bromophenol blue). The mixtures were boiled for 5 min and subjected to SDS-PAGE. In immunoprecipitation, equal amounts of proteins of each sample were mixed with indicated antibodies (their concentrations were as recommended by manufacturers). Samples were incubated at 4°C for 12 h and then mixed with 10 ml of Protein A-Sepharose or Protein G-Sepharose (GE Healthcare). After the additional incubation for 1 h, precipitates were washed with ice-cold lysis buffer three times and diluted with 1 Â Laemmli's buffer. After boiling for 5 min, the supernatant was subjected to SDS-PAGE. The immunoprecipitation using anti-FLAG antibody-conjugated agarose beads followed by elution with FLAG peptide (Sigma-Aldrich) was performed according to the manufacturer's recommended protocols. The eluates were mixed with 1/4 volume of 4 Â Laemmli's buffer, boiled for 5 min and subjected to SDS-PAGE. After SDS-PAGE, proteins were transferred onto a polyvinylidene difluoride membrane. The indicated first antibodies and HRP-conjugated second antibodies were hybridized according to standard immunoblotting protocols (their concentrations were as recommended by manufacturers). Chemiluminescence reactions were carried out using SuperSignal West Pico/Femto (Thermo Fisher Scientific, Rockford, IL, USA) and the luminescence was exposed onto X-ray film (X-Omat, Kodak, Tokyo, Japan). Densitometric analysis was carried out using ImageJ 1.43u programme (National Institutes of Health, Bethesda, MD, USA). Uncropped scans of blots are shown in Supplementary Figs 9-13.
In sample preparation for the detection of ubiquitination signals as well as the interaction of IRS-2-Epsin1 (or other ubiquitin-binding proteins), cells were lysed with lysis buffer supplemented with 2 mM N-ethylmaleimide, to inhibit de-ubiquitinating enzymes. In addition, protein samples for the detection of ubiquitination signals were denatured by mixing with 2% SDS and boiling for 5 min, to disassemble protein complex, and then diluted to 20 times by lysis buffer followed by immunoprecipitation. After SDS-PAGE and transfer to a polyvinylidene difluoride membrane, the membrane was denatured by incubating in denaturing buffer (6 M guanidine-HCl, 20 mM Tris-HCl pH7.5, 5 mM b-mercaptoethanol, 1 mM phenylmethyl sulphonyl fluoride) for 30 min and then washed several times with 0.1% Tween20/PBS before immunoblotting using anti-ubiquitin antibody.
LC-MS/MS analysis to determine IRS-2 ubiquitination sites. FLAG-tagged IRS-2 was immunoprecipitated with anti-FLAG antibody-conjugated beads. The immunoprecipitates were mixed with 100 ml of trypsin solution (10 ng ml À 1 Trypsin Gold Mass Spectrometry Grade, 50 mM ammonium bicarbonate, 0.01% RapiGest SF (Waters, Milford, MA, USA)), trypsin and Asp-N solution (10 ng ml À 1 Trypsin Gold, 6.6 ng ml À 1 Asp-N (Promega), 50 mM ammonium bicarbonate, 0.01% RapiGest SF), or trypsin and chymotrypsin solution (10 ng ml À 1 Trypsin Gold, 10 ng ml À 1 chymotrypsin (Promega), 100 mM Tris-HCl pH8.0, 10 mM CaCl 2 , 0.01% RapiGest SF). The samples were incubated at 37°C for 12 h. After a brief centrifugation, supernatants were recovered, mixed with 100 ml of TFA and then desalted using GL-Tip SDB and GL-Tip GC (GL Sciences, Tokyo, Japan). The samples were diluted with IAP buffer (Cell Signaling), mixed with anti-ubiquitin remnant motif antibody-conjugated beads and then incubated at 4°C for 2 h. The immunoprecipitates were washed with IAP buffer twice and with water three times. Peptides were eluted by 0.15% TFA, desalted with GL-Tip SDB and GL-Tip GC, and then subjected to LC-MS/MS analysis (nano-liquid chromatography column (EASY-nLC 1000, Thermo Fisher Scientific) coupled to Q Exactive mass spectrometer (Thermo Fisher Scientific)) 54 . Data were analysed by Proteome Discoverer software (Thermo Fisher Scientific) and ProteinPilot software (AB SCIEX, Framingham, MA). Peptide quantification was carried out by parallel reaction monitoring, a MS/MS-based quantification method. The extracted peptides were analysed by Q Exactive in targeted MS/MS mode. Raw files were processed by PinPoint software (Thermo Fisher Scientific). Ion chromatograms were extracted with a mass tolerance of 5 p.p.m. The area under the curve of selected fragment ion was calculated. The AUCs of each individual fragment ions were then summed to obtain AUCs at the peptide level 54 .
Biotinylation of cell surface proteins. Cells were incubated with 1 mg ml À 1 NHS-LC biotin (Thermo Fisher Scientific)/PBS for 20 min at 4°C and then washed twice with 100 mM glycine/PBS. The cells were then lysed with lysis buffer, followed by immunoprecipitation and blotting with HRP-conjugated streptavidin (Sigma-Aldrich) and indicated antibodies. Subcellular fractionation. Cells were homogenized in ice-cold detergent-free buffer (40 mM Tris-HCl pH 7.4, 120 mM NaCl, 1 mM EDTA, 1 mM EGTA, 10 mM pyrophosphate, 50 mM NaF, 500 mM Na 3 VO 4 , 10 mg ml À 1 p-nitrophenyl phosphate, protease inhibitor cocktail). Samples were then subjected to ultracentrifugation (100,000g, 1 h). The supernatants were recovered as cytosol fractions. The precipitates were solubilized with lysis buffer as described above (containing 1% Triton X-100), followed by centrifugation (15,000g, 10 min) and the supernatants were recovered as membrane fractions.
Thymidine incorporation into DNA. Cells were cultured in 48-well plates or 24-well plates and [Methyl-3 H]thymidine (0.3 mCi per well; 1 mCi ml À 1 , GE Healthcare) was added to each well for indicated durations. The labelling was stopped by adding 500 ml of 1 M ascorbic acid. The cells were washed twice with ice-cold PBS and twice with ice-cold 10% trichloroacetic acid. Trichloroacetic acidprecipitated materials were solubilized with 250 ml of 0.2 N NaOH/0.1% SDS and mixed into 5 ml clear-sol II (Nakalai tesque). The radioactivity was measured by liquid scintillation counter (Aloka, Tokyo, Japan).
Zebrafish experiments. Wild-type zebrafish (Danio rerio) were kept at 28°C under a light/dark cycle of 14 and 10 h, and fed twice daily. The wild-type embryos obtained by natural breeding were raised at 28.5°C and staged 55 . Capped RNAs were synthesized using mMESSAGE mMACHINE Kit (Life Technologies). Venus mRNA (200 pg per embryo), Nedd4-Venus mRNA (1,200 pg per embryo) or 5 0 -UTR Irs2 -Venus RNA (250 pg per embryo) was co-injected with either control MO or Irs2 MO (4 ng per embryo) into embryos at one-to two-cell stage. Embryos were then placed in embryo-rearing medium 34 and kept at 28.5°C. Body length and somite number were measured at 22 h post fertilization. Body length was defined as the curvilinear distance from the middle of the otic vesicle to the end of tail through the trunk midline. The expression of Nedd4-Venus was confirmed by fluorescence and immunoblotting. Specific translation block with Irs2 MO were confirmed by the decreases of 5 0 -UTR Irs2 -Venus fluorescence. All experiments were conducted in accordance with the guidelines approved by the Ethical Committee of Experimental Animal Care at Ocean University of China.