Notch signaling is a conserved signaling pathway implicated in embryogenesis and adult tissue maintenance. Notch signaling strength is strictly regulated, notably by maintaining a controlled pool of functional receptor at the cell surface. Mammalian non-activated Notch receptor is internalized, ubiquitylated by the Itch E3 ubiquitin ligase and degraded in the lysosomes. Here, we show that β-arrestins are necessary for Itch–Notch interaction and for Itch-driven ubiquitylation and degradation of Notch. Interestingly, β-arrestins do not directly bind Itch but heterodimerize with a member of another subfamily of arrestins called ARRDC1 or α-arrestin 1, which harbors PPxY motifs that allow direct interaction with Itch. Cells transfected with ARRDC1 mutated in PPxY motifs show reduced Itch-mediated Notch ubiquitylation and impaired lysosomal degradation of Notch, as observed in β-arrestin−/− or Itch−/− cells. Our data show for the first time that ARRDC1 and β-arrestins heterodimerize and cooperate in the same complex to promote non-activated Notch receptor degradation, thus acting as negative regulators of Notch signaling.
Notch signaling controls a variety of developmental processes as well as adult tissue and organ maintenance. It relies on the interaction between Notch ligand and receptor, which are transmembrane proteins present on the surface of adjacent cells. This binding triggers a succession of events leading to proteolytic cleavage of Notch receptor, which results in the production of a cytosolic fragment of the receptor. The latter migrates to the nucleus, where it acts as a transcriptional cofactor to regulate target genes transcription (Bray, 2006).
A number of components and post-translational modifications have been implicated in regulating the activity of Notch receptor, either positively or negatively (Brou, 2009). Among those, some affect Notch degradation in a ligand-independent manner, thus regulating the quantity of receptor present at the cell surface and therefore the strength of the Notch response. In mammals, non-activated Notch receptor is constitutively internalized and ubiquitylated by Itch/AIP4 E3 ubiquitin ligase in order to be processed for lysosomal degradation (Chastagner et al., 2008). Moreover, the deubiquitylating complex USP12/UAF1 is recruited by Itch to Notch and allows Notch deubiquitylation, probably before its entry into the intralumenal vesicles of the multivesicular bodies (MVBs) (Moretti et al., 2012).
These data suggest that Notch degradation in the lysosomes is a regulated and ordered process, where several proteins are necessary to recruit E3 ubiquitin ligases and deubiquitylating enzymes. The first evidence of arrestin involvement in Notch signaling was made in Drosophila (Mukherjee et al., 2005) by the identification in a two-hybrid screen of Kurtz (krz), the single Drosophila homolog of mammalian non-visual β-arrestins, as an interacting partner of Deltex, a known modulator of Notch activity. Loss of krz function results in an accumulation of Notch receptor, resulting in a partial gain-of-function effect on Notch signaling. On the basis of these genetic data in Drosophila and the absence of direct interaction between Itch and Notch in mammals (Chastagner et al., 2008), we asked whether arrestins could have a role in Notch regulation in mammals.
In mammals, β-arrestins (ARRBs) were originally shown to desensitize activated G-protein-coupled receptors (GPCR) and to promote receptor endocytosis. In addition, there is a growing list of examples showing that β-arrestins might also serve as multifunctional adaptors and scaffolds that mediate endocytosis of various types of receptors, have signaling functions or allow E3 ubiquitin ligase recruitment (Kovacs et al., 2009). Among these E3 ligases, Mdm2 and Deltex belong to the RING (a really interesting new gene) family, whereas Nedd4 and Itch are HECT (homologous to E6AP carboxyl terminus) family members. Because β-arrestins lack any canonical PPxY motif known to interact with WW domains of HECT family members (Ingham et al., 2004), the molecular basis of the interaction between β-arrestins and HECT E3 ubiquitin ligases remains unknown. By contrast, members of the related ARRDC/α-arrestin family (arrestin domain-containing proteins, ARRDC1–ARRDC5 and TXNIP in mammals) (Alvarez, 2008; Becuwe et al., 2012a; Masutani et al., 2012) exhibit PPxY motifs in their C-terminal tail (except ARRDC5). Until now, ARRDCs have been shown to serve, similar to β-arrestins, as adaptors for the recruitment of specific cargo to E3 ubiquitin ligases and endocytosis in yeast (Lin et al., 2008), and also, in a few cases, in mammals (Nabhan et al., 2012; Nabhan et al., 2010). Whereas heterodimerization of β-arrestins in cells has been demonstrated (Storez et al., 2005; Xiao et al., 2007), heterodimerization of α- and β-arrestins giving rise to functional complexes has been theoretically proposed (Polo and Di Fiore, 2008), but not yet proven. Here, we show that a complex is formed containing α-arrestins and β-arrestins to recruit the E3 ubiquitin ligase Itch on Notch, which leads to Notch ubiquitylation and lysosomal degradation.
β-arrestins are necessary for Itch–Notch interaction
Previous studies in mammals have demonstrated that non-activated Notch-1 receptor (hereafter Notch-1 receptor is named as Notch) is ubiquitylated by the E3 ubiquitin ligase Itch, even if no direct interaction has been detected (Chastagner et al., 2008). To determine whether β-arrestins promote Itch recruitment to Notch, we performed a coimmunoprecipitation (Co-IP) assay after transfecting HEK 293T cells with expression vectors encoding β-arrestin 1 (β-arr1), Notch (Myc-tagged Notch full-length, FL) together with FLAG-tagged Itch dominant negative (DN) or two other E3 ubiquitin ligases of the HECT family (Nedd4 DN–FLAG and Nedd4.2 DN–FLAG). The E3 ligases were mutated in their active site (indicated as DN) to try to stabilize the transient interactions with their substrates. We pulled down the E3 ubiquitin ligases and analyzed the immunoprecipitates and the whole-cell extracts (WCEs) by western blot using an antibody against the intracellular part of Notch (Notch IC) allowing us to detect the membrane-anchored subunit of mature Notch heterodimer (designated as NFL). Notch weakly co-immunoprecipitated with Itch DN (Fig. 1A, lane 13) and overexpression of β-arrestin 1 increased this interaction more than twofold (Fig. 1A, lane 10 compared with lane 13). By contrast, the interaction between Nedd4/4.2 and Notch FL remained weak even in the presence of overexpressed β-arrestin 1 (Fig. 1A,B lanes 11, 14, and lanes 12, 15), as confirmed by the corresponding densitometry analysis (Fig. 1B). As shown in Fig. 1A (lane 13) and 1C (lane 2) the Itch-DN–Notch complex was established at a basal level even in the absence of overexpressed β-arrestin 1. The presence of endogenous β-arrestins (detected in the WCEs in Fig. 1C) was sufficient to allow some binding of Notch to Itch. To demonstrate this, we knocked down the endogenous β-arrestin 1 and β-arrestin 2 as described previously (Coureuil et al., 2010). The efficiency of siRNA targeting β-arrestins (siβ-arr) was confirmed by western blot (Fig. 1C). β-arrestin silencing almost completely abolished the Itch-DN–Notch-FL interaction compared with control siGFP (31% of signal left in lane 4 versus 2, as quantified in Fig. 1D). These results suggest that β-arrestins are necessary for the formation of the Itch–Notch-FL complex in mammalian cells.
β-arrestins are required for Itch-mediated Notch ubiquitylation and degradation, acting as negative regulators of Notch signaling
To determine whether the function of β-arrestins in the Notch pathway is related to the Itch-mediated Notch ubiquitylation, we monitored Notch ubiquitylation when endogenous β-arrestins were knocked down or not. We transfected HEK 293T cells with vectors encoding Itch, Notch and 6×His-tagged ubiquitin and we purified the ubiquitylated proteins on Nickel-charged beads in denaturing conditions. Ubiquitylated Notch products were specifically detected when the cells were transfected with His–Ub (a monoubiquitylated Notch indicated by an asterisk and a smear corresponding to the polyubiquitylated forms, indicated by NFLub, compare lane 2 with lane 1 in Fig. 2A). A low amount of ubiquitylated Notch was detected without overexpressed Itch, in contrast to what was observed when the cells were transfected with Itch-expressing plasmid and a control siRNA (NT) (Fig. 2A, lane 2 compared with lane 3). Silencing the endogenous β-arrestins significantly decreased Itch-mediated Notch ubiquitylation (Fig. 2A, lane 4), suggesting that β-arrestins are not only important for the formation of the complex, but are also required for the proper function of Itch in the Notch pathway.
To further assess the role of β-arrestins in Notch trafficking and degradation, we used mouse embryonic fibroblasts (MEFs) either WT or knocked out for both β-arrestins to avoid possible redundancy. We infected these cells with a retroviral vector encoding Notch, harboring a HA tag in its extracellular part and we selected stable clones expressing comparable amount of Notch, either WT or β-arrestin−/− cells (C2). The amount of cell surface Notch was quantified by flow cytometry and was found to be higher in WT cells compared with C2 (supplementary material Fig. S1A). Living cells were labeled with an anti-HA antibody to recognize Notch molecules present at the cell surface, then extensively washed and incubated at 37°C for various period of time to allow Notch endocytosis and trafficking. At time point 0 (no incubation at 37°C) HA–Notch was localized to the plasma membrane in both cell lines, as exemplified in Fig. 2B panels A and B, where the upper slice of each field was chosen to easily visualize the staining. After 30 and 60 minutes of incubation at 37°C, Notch was mostly detected in intracellular vesicles, showing that the receptor was internalized in WT as well as in β-arrestin−/− cells with comparable kinetics (Fig. 2B, panels C–F). At 90 and 180 minutes, almost no Notch signal was detectable in WT cells (G,I), whereas it was still clearly visible in C2 cells (H,J). Quantification of spot numbers revealed that they are seven times more frequent at time 180 minutes in C2 cells compared with WT cells (Fig. 2C). Thus, Notch degradation was delayed in the absence of β-arrestins with a kinetics similar to that described in Itch−/− cells (Chastagner et al., 2008). As shown in Fig. 2B, panels K,L, β-arr1 expression could complement the defect of β-arrestin−/− cells, because Notch degradation was restored after 180 minutes of incubation as quantified in Fig. 2D. In light of this result, we decided to assess whether β-arrestins, as previously demonstrated for Itch, were necessary to target non-activated Notch receptor towards lysosomes for degradation. We performed a receptor-uptake experiment in WT, C2 and Itch−/− cells (3F4) in the presence of leupeptin, an inhibitor of lysosomal proteases to impair Notch degradation and visualize it in late endosomes and lysosomes. In leupeptin-treated WT cells, Notch colocalized with LAMP1, a late endosome/lysosomal marker, after 3 hours of internalization (Fig. 2E, panels A–C and see quantification in supplementary material Fig. S1B). As expected, in 3F4 cells, Notch signal was much less associated with the lysosomal marker (Fig. 2E, panels G–I) (Chastagner et al., 2008). Similarly, in C2 cells, the percentage of colocalization between Notch and LAMP1 showed a significant decrease (Fig. 2E, panels D–F) compared with WT cells (supplementary material Fig. S1B), suggesting that in β-arrestin−/− cells as in Itch−/− cells, Notch did not reach the lysosomes for degradation and rather remained trapped in endocytic vesicles.
We next assessed the role of β-arrestins in Notch activation. We transfected WT and C2 cells with a Notch reporter vector, an internal Notch-insensitive control vector and an expression vector encoding β-arrestin 1. WT and C2 cells were then co-cultured with cells expressing Delta-like 1 (a Notch ligand) for 18 hours. As a reference, we performed the co-culture assay in the presence of DAPT, a potent inhibitor of gamma-secretase activity and therefore of Notch transcriptional effects. In both WT and C2 cells, Notch activation was significantly induced compared with the respective DAPT-treated co-cultures (Fig. 2F). 100% of Notch activation corresponded to a 6.7-fold stimulation for WT cells and 4.7-fold stimulation for C2 cells. In the presence of overexpressed β-arrestin 1, no significant change in Notch activation was observed in WT cells (93% of Notch activation), whereas C2 cells showed a significant decrease of Notch activation (67% of Notch activation). We confirmed that the exogenous β-arrestin 1 was indeed expressed in both cell lines by western blot analysis (supplementary material Fig. S2). As β-arrestins are probably not limiting in WT cells, their negative effect was detectable only when complementing β-arrestin-knockout cells with β-arrestin 1. This decreased Notch activation observed in C2 cells complemented with β-arrestin 1 was in accordance with our previous results showing that β-arrestins push Notch towards the degradative pathway.
Together, these results show that, in the absence of β-arrestins, Itch-mediated Notch ubiquitylation and lysosomal localization are strongly perturbed. Moreover, β-arrestin complementation restores Notch degradation and significantly affects Notch activation, confirming that β-arrestins are negative factors involved in Notch trafficking.
ARRDC1/α-arrestin 1 interacts with β-arrestins and Itch
β-arrestins lack canonical PPxY motifs, which are known to interact with WW domains of HECT proteins (Ingham et al., 2004). We thus wondered whether β-arrestins could directly interact with Itch. We performed a GST pull-down assay, where in vitro translated [35S]methionine-labeled β-arrestin 1 was incubated with either purified GST–Itch or GST alone adsorbed on glutathione-agarose beads. As shown in Fig. 3A, no interaction was detected between β-arrestin 1 and Itch (lane 6), which is inconsistent with a direct binding between the two native proteins. We thus asked whether another member of the arrestin family could interact with Itch. α-arrestins/ARRDCs (arrestin-domain containing proteins) have been classified after phylogenetic analysis as members of the large family of proteins that includes the well-characterized β-arrestins (Alvarez, 2008). Among the six mammalian α-arrestins, ARRDC1–ARRDC4 harbor conserved PPxY motifs (Alvarez, 2008; Rauch and Martin-Serrano, 2011). We first tested by coimmunoprecipitation from transiently transfected cells, the interaction of ARRDC1–ARRDC4 with Itch. We found that only ARRDC1 and ARRDC3 were able to coimmunoprecipitate with Itch-DN (supplementary material Fig. S3A). For technical reasons we used ARRDC1 in all our subsequent experiments.
We performed in vitro translations and a GST pull-down assay with HA–ARRDC1 and HA–ARRDC1-ΔPY (which lacks the PPxY-containing C-terminal part, as schematized in Fig. 3B) in parallel with β-arrestin 1. In contrast to β-arrestin 1, ARRDC1 directly bound Itch, whereas GST–Itch interaction with the mutant ARRDC1-ΔPY was almost undetectable (Fig. 3A, lanes 2 and 4). Therefore the PPxY-containing domain of ARRDC1 is necessary to directly bind Itch. A PY* mutant (which contains a double point mutation in the PPxY motifs; Fig. 3B) was not able to coimmunoprecipitate with Itch DN from transfected cells, in contrast to WT ARRDC1, confirming that the PPxY motifs of ARRDC1 indeed account for Itch interaction (supplementary material Fig. S3B).
We next asked whether β-arrestins could heterodimerize with ARRDC1 and so cooperate as adaptors for Itch recruitment. In transfected HEK 293T cells, β-arrestin 1 co-immunoprecipitated with WT ARRDC1 constructs (Fig. 3C, compare lanes 5 and 7 with lanes 2 and 4). This interaction does not depend on PPxY motifs because the binding was also detectable with PY* and ΔPY mutants (lanes 3, 6 and 8), suggesting that ARRDC1 and β-arrestin 1 interact through the arrestin domain. Therefore, β-arrestin 1 is not able to directly interact with Itch, but it can heterodimerize with ARRDC1, a direct Itch interactor. In addition, in vitro pull-down experiments and coimmunoprecipitations from transfected cells showed that ARRDC1 is able to directly interact with β-arrestin 1 and β-arrestin 2 (supplementary material Fig. S3C,D). In light of these results, we hypothesized that ARRDC1 could form a complex with β-arrestin 1 and β-arrestin 2 that is able to recruit Itch to Notch.
ARRDC1 cooperates with β-arrestins on Notch signaling
To determine whether ARRDC1 plays a role in Notch regulation we first decided to assess whether it could influence β-arrestin–Notch interaction. We pulled down β-arrestin 1 (β-arr1–GFP) in transfected HEK 293T and we observed that Notch coimmunoprecipitated with β-arrestin 1 (Fig. 4A, compare lane 7 with lane 2). When ARRDC1 WT was overexpressed, β-arrestin–Notch complexes were less abundant (Fig. 4A, lane 5). Overexpression of ARRDC1 could increase the quantity of the degradative complexes formed with endogenous Itch and therefore further promote Notch degradation. This hypothesis was corroborated by the fact that ARRDC1 PY*, which is not able to bind Itch, stabilized β-arrestin-Notch complexes (lane 6), probably avoiding the whole Itch-containing complex assembly and therefore Notch degradation. In all cases, the levels of Notch were comparable in the WCEs, reflecting the fact that a small percentage of whole cellular Notch is undergoing degradative trafficking at a given time. Together, these results suggest that complexes containing Notch and α–β heterodimeric arrestins are formed and that they are able to target Notch for degradation. To further prove and localize the interaction of endogenous ARRDC1 with Notch, we detected Notch–ARRDC1 complexes by using in situ PLA technology (Fig. 4B, panels G,M). Notch- and ARRDC1-containing spots were twice as numerous when cells were transfected with ARRDC1 PY* (see quantification on Fig. 4C), confirming the stabilization of the complexes when Itch cannot be recruited.
To further address the functionality of such complexes, we first asked whether ARRDC1 like β-arrestins, could affect Itch-dependent Notch ubiquitylation. Low amounts of ubiquitylated Notch (NFLub) were detected without overexpressed Itch (Fig. 4D, lane 2) and these levels greatly increased when cells overexpressed Itch and ARRDC1 WT (lane 3). By contrast, the presence of the mutated ARRDC1 (PY*) drastically decreased Itch-mediated Notch ubiquitylation (lane 4), in accordance with the hypothesis that ARRDC1 could have an adaptor function for Itch recruitment, leading to Notch ubiquitylation. This finding agrees with the result obtained using siRNA targeting β arrestins in Fig. 2A and strongly supports the idea that both subfamilies of arrestins are involved together in assembly and function of the Itch–Notch-containing complexes.
To further assess the role of ARRDC1 in Notch trafficking and degradation, we used WT and C2 MEFs transiently transfected with either ARRDC1 (WT or HA-WT) or the mutated forms (PY* or ΔPY) to perform a receptor-uptake experiment on living cells. Expression of each ARRDC1 construct was confirmed by western blotting (supplementary material Fig. S6A). Notch was mostly detected in intracellular vesicles at early time points, with a comparable kinetics in all cases (supplementary material Fig. S4), meaning that overexpression of any ARRDC1 construct had no effect on the early steps of Notch endocytosis. As shown in Fig. 5A and quantified in Fig. 5B, the presence of ARRDC1 WT affected neither Notch degradation observed in WT cells (panels E,I compared with A,C) nor the delayed degradation observed in C2 (panels F,J compared with B,D). This shows that ARRDC1 did not complement the trafficking defect of β-arrestin−/− cells, in contrast to what was observed when β-arrestin 1 was transfected into C2 cells (Fig. 2B, panels K,L). Therefore, these factors are not interchangeable. We observed Notch-positive spots remaining in WT cells transfected with ARRDC1 PY* or ΔPY (Fig. 5A, panels G and K), as was the case in all C2 conditions (Fig. 5A, panels B,D,F,H,J,L and quantification Fig. 5B). Thus ARRDC1 PY* or ΔPY overexpression converted WT cells into cells that behaved like β arrestin−/− cells. To further assess whether this effect was due to an impairment of Notch lysosomal degradation, we looked whether the remaining Notch-positive vesicles (visible after 3 hours of incubation in Fig. 5A) colocalized with LAMP1. As expected, Notch signal colocalized with LAMP1 in the presence of leupeptin either in WT cells transfected with control vector (Fig. 5C, panels A–C), or in C2 cells transfected with β-arrestin 1 expression vector (Fig. 5C, panels M–O). In WT cells expressing ARRDC1 PY* or ΔPY (Fig. 5C, panels D–F and G–I, respectively), Notch co-labeling with LAMP1 was barely detected, as in C2 cells (panels J–L). These findings strongly suggest that the overexpression of mutated ARRDC1 (PY* or ΔPY) in WT cells impairs Notch targeting to lysosomal degradation, as observed in β-arrestin−/− cells. The vesicles where Notch was trapped in C2 cells or in ARRDC1-PY*-transfected WT cells, after 180 minutes of internalization were partially colocalized with the early endosome marker EEA1 in both conditions (supplementary material Fig. S5), suggesting that ARRDC1 and β-arrestins are required after early endocytosis for Notch degradation.
We finally monitored the role of endogenous arrestins in Notch signaling by co-culture assay. We transfected human-Notch-expressing cells (U2OS-FL) with Notch reporter, a control vector (as described in Fig. 2D) and different siRNAs: NT (non-targeting) or targeting ARRDC1, β-arrestin, CSL, USP12, respectively, as indicated in Fig. 5D. The efficiency of both siRNAs targeting β-arrestin and ARRDC1 was quantified by qRT-PCR (see supplementary material Fig. S6B). After co-culturing Delta-like-1-expressing cells with U2OS-FL cells transfected with siNT, Notch activation was significantly induced compared with that in the DAPT-treated co-cultures (33.3 fold, see Fig. 5D). As expected, knockdown of CSL, the key DNA-binding subunit of the Notch-containing transcriptional complex (Bray, 2006), led to a significant reduction of Notch activation. By contrast, silencing of USP12, a negative regulator of Notch signaling (Moretti et al., 2012), significantly increased Notch activity (233% over siNT). Strikingly, knockdown of either β-arrestins or ARRDC1 also induced Notch reporter activity (161% and 127%, respectively) showing that ARRDC1 and β-arrestins are negative regulators of Notch signaling. Moreover, the simultaneous ablation of both ARRDC1 and β-arrestins further increased the effect of single knockdown (184% for the combination vs 127% and 161% for siARRDC1 and siβ-arrestin respectively) suggesting that indeed ARRDC1 and β-arrestins cooperate as negative regulators of Notch signaling.
Taken together, these data show that ARRDC1 is involved in Itch-mediated Notch ubiquitylation and lysosomal degradation at the same step, but not redundantly, with β-arrestins. Moreover, ARRDC1 in combination with β-arrestins acts as a negative regulator of Notch signaling in accordance with ARRDC1 and β-arrestins being members of the same complex.
Arrestins are new factors involved in mammalian Notch degradation
It has been shown in various systems that full-length Notch receptor has a limited half-life at the cell surface and is constantly internalized and degraded through a lysosomal pathway (Chastagner et al., 2008; Sakata et al., 2004; Vaccari et al., 2008; Wilkin et al., 2008). This mechanism is a way of maintaining a functional receptor, and of eventually regulating Notch signal strength by acting on the receptor level at the cell surface. Notch trafficking towards the lysosomes involves ubiquitylation by the E3 ubiquitin ligase Itch/AIP4 in mammals (Chastagner et al., 2008) and Su(dx) or Nedd4 in Drosophila (Sakata et al., 2004; Wilkin et al., 2004). However, whereas Drosophila Notch contains a canonical PPxY motif that could account for direct binding of the receptor to the E3 ligase, this is not the case for mammalian Notch molecules (except Notch 3). Therefore, scaffolding proteins are probably necessary, at least in mammals, to form and stabilize complexes that allow Notch ubiquitylation and degradation. Our work shows that Notch degradation in the absence of ligand is regulated by β-arrestins, together with the α-arrestin ARRDC1. Biochemical data demonstrate that β-arrestins are necessary for Notch–Itch interaction, although they cannot bind Itch directly. Moreover, ARRDC1, which contains PPxY motifs, is able to interact with Notch and directly with Itch, and with β-arrestin 1 and β-arrestin 2. In addition, complexes containing Notch and β-arrestin are stabilized in the presence of mutant ARRDC1 constructs that are unable to recruit Itch. These data strongly suggest that Itch is recruited to Notch by the ARRDC1 subunit of β-arrestin–ARRDC1 heterodimers. This hypothesis is corroborated by functional data, because we show that Notch ubiquitylation and lysosomal degradation both depend on the presence of both β-arrestins and α-arrestins. Interestingly, in β-arrestin-knockout cells, ARRDC1 overexpression cannot restore Notch degradation, in contrast to overexpression of β-arrestin 1. Conversely, a mutant form of ARRDC1, affected in its ability to recruit Itch, can convert WT cells to cells that behave like β-arrestin-deficient cells, because Notch degradation is blocked at the same step in both conditions. In addition, ARRDC1 and β-arrestins act as negative regulators of Notch signaling, as shown in a co-culture assay, allowing us to monitor the involvement of these endogenous proteins on Notch activation. Taken together, our observations substantiate the idea that α-arrestins and β-arrestins are not redundant, but instead work together in complexes recruiting Itch to Notch.
How and when are arrestins heterodimers recruited to Notch?
α-arrestins and β-arrestins are cytosolic proteins that were described as being recruited to the plasma membrane upon cargo activation or upon environmental signals in yeast (Polo and Di Fiore, 2008). However, recent data localized these proteins to the plasma membrane (Lin et al., 2008), along the endocytic pathway (Becuwe et al., 2012b; O'Donnell et al., 2010; Patwari et al., 2009), or even in the nucleus (Boularan et al., 2007; Hara et al., 2011) during cargo recognition. Similar to β-arrestins, ARRDCs have been described to interact with clathrin and clathrin adaptors, suggesting that they could act on early endocytosis (Becuwe et al., 2012a; O'Donnell et al., 2010). However, our results suggest that this is not the case for Notch trafficking, because the formation of early endocytic vesicles occurs similarly in WT, β-arrestin−/− or PY*-ARRDC1-overexpressing cells, as well as in Itch−/− cells. In these conditions we observed an increased number of Notch-containing vesicles that colocalized with EEA1, but not a complete overlap of staining. Therefore recruitment of arrestins and Itch to Notch probably takes place after early endocytosis. This result reflects the fact that arrestins are multifunctional molecules that fulfil complex and diverse functions, depending on the substrates. For example, β-arrestin1 can interact with STAM-1, a component of ESCRT-0 to control CXCR4 downregulation (Malik and Marchese, 2010). Recent data on retroviral and vesicle budding mechanisms (Nabhan et al., 2012; Rauch and Martin-Serrano, 2011) show that ARRDC1 can directly interact with HECT ubiquitin ligases, including Nedd4 and Itch, and also with ALIX and Tsg101, components of the ESCRT pathway. In this case, the proposed model is that ARRDCs could provide a physical link between HECT ubiquitin ligases and ESCRT machinery, via interactions with ALIX and Tsg101 through their C-terminus, and with uncharacterized factors through their arrestin domains. Similarly, during Notch trafficking, arrestin heterodimers (bound through their arrestin domain) could be recruited to Notch when the receptor becomes associated with specific factors of the endocytic machinery in sorting endosomes. This scaffold could recruit Itch (through the C-terminus of ARRDC1) and target Notch for subsequent degradation (see model in Fig. 6).
Comparison with the Drosophila model
The first description of the involvement of arrestins in Notch signaling was made in Drosophila (Mukherjee et al., 2005). The Drosophila genome encodes a single β-arrestin, Kurtz (krz), that is involved in development and survival; however, many of the classical functions of β-arrestins are conserved in the fly. In addition to Deltex (dx) and krz, shrub has been recently identified as affecting subcellular localization of Notch (Hori et al., 2011). shrub encodes the Drosophila homologue of vps32, a core component of the ESCRT III complex. Whereas Drosophila shrub antagonizes dx, it enhances the downregulating activity of krz on Notch (Hori et al., 2011). The proposed model in Drosophila is that Krz and the E3 ubiquitin ligase Deltex could form a complex with Notch and regulate Notch degradation via an endosomal-lysosomal pathway. In addition to the identification of Krz by two-hybrid screen using Deltex as bait, this model is mainly based on genetic interaction experiments using the adult wing phenotype as a readout of Notch signaling activation. Our data complete and extend these results in mammals, even though some discrepancies might exist between both systems. It is of note that although Drosophila Notch does contain a PPxY motif, Krz is still necessary for Notch degradation, suggesting that Drosophila β-arrestin could actually stabilize degradation complexes in the same way as we show here for mammalian β-arrestins. Regarding α-arrestins, there are at least 13 members in the fly ARRDC family (Alvarez, 2008). Further genetic studies are needed to see whether some of these genes can have a role in Notch signalling; however, compensatory effects could complicate the analysis.
Because we observed a direct interaction of human Deltex1 with β-arrestins (supplementary material Fig. S7), one putative model that would integrate all the data would be that Notch internalization could be directed by Deltex (Yamada et al., 2011), which harbors conserved internalization motifs in its C-terminal part. Once non-activated Notch reaches an adequate compartment, arrestins could bind the receptor and recruit Itch to eventually trigger Notch degradation through ESCRT-dependent mechanisms. Alternatively if Notch is activated by its ligand and cleaved by ADAM-10, the γ-secretase complex could recognize it after Deltex-mediated internalization, and the activation process could follow. Further experiments are needed to prove this hypothesis and build a full model.
α-arrestins and β-arrestins can heterodimerize and are required together in Notch signaling: is this a general way of working?
Our data show that β-arrestin 1 is able to heterodimerize in vitro and in vivo with ARRDC1 through the structurally conserved arrestin domain. Do α–β complexes pre-exist in the cell cytoplasm? A strong argument against this possibility comes from a recent census of human soluble complexes (Havugimana et al., 2012), which did not identify arrestin-containing preformed complexes. In addition, a previous proteomic analysis using cells stably expressing β-arrestin 1 or β-arrestin 2 identified more than 300 proteins able to interact with one or both β-arrestins (Xiao et al., 2007). Among them, β-, S- and X-arrestins were detected, suggesting that these proteins are able to heterodimerize in solution, however neither α-arrestins nor E3 ubiquitin ligases are listed in the paper. Therefore α–β complexes could form when recruited to a specific substrate. Our work demonstrates for the first time that α-arrestins and β-arrestins heterodimerize and function coordinately to fine-tune Notch recognition and recruitment of the E3 ubiquitin ligase Itch, eventually targeting Notch for lysosomal degradation. The same type of model could apply in other contexts. In the case of β2AR (β2-adrenergic receptor), challenging results suggest that β-arrestin 2, as well as ARRDC3, are necessary for Nedd4-dependent ubiquitylation and degradation of the receptor after its activation by isoproterenol (Han et al., 2012; Nabhan et al., 2010; Shea et al., 2012; Shenoy et al., 2008; Shenoy et al., 2001). Therefore, and in accordance with our hypothesis, it is tempting to speculate that β-arrestin 2 and ARRDC3 (like β-arrestin 1 and ARRDC1 in our case) could heterodimerize to fulfil this scaffolding function. Of note, Shea and colleagues (Shea et al., 2012) visualized a vesicular colocalization of ARRDC3/4 with β-arrestin 1 after stimulation of β2AR, as well as a co-immunoprecipitation between α-arrestins and β-arrestins. More generally, various types of arrestin heterodimers could be formed, depending on the cargo and its subcellular localization, and could recruit the specific E3 ubiquitin ligase involved in the process. Given that there are two β-arrestins and a larger number of α-arrestins (six in mammals), the various combinations would thus increase the repertoire of intermediary factors able to specifically recognize defined cargo during their trafficking.
Materials and Methods
Antibodies and peptides
Antibodies for western blots were supplied by Abcam [anti-ARRDC1, anti-ARRDC3 and ARRDC4, EEA1, anti-Notch1 (ab49990)], Bethyl (polyclonal anti-eIF3a), Covance (monoclonal anti-HA), Invitrogen (polyclonal anti-GFP), Sigma (monoclonal anti-FLAG M2; polyclonal anti-FLAG; monoclonal anti-α-tubulin, anti-Myc 9E10), BD Transduction Laboratories (monoclonal anti-Itch and anti-β-arrestin) and Novagen (monoclonal anti-S-Tag). Anti-LAMP1 was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA 52242. Secondary antibodies for immunofluorescence were supplied by Molecular Probes (Alexa Fluor conjugates). Anti Notch IC antibody has been already described (Logeat et al., 1998). FLAG peptide was provided by GeneCUST and used at 2 mg/ml.
The following constructs have been already described: murine Myc-tagged Notch1 full-length (FL), deleted of a C-terminal fragment (Schroeter et al., 1998), human Notch1 retroviral construct (Rand et al., 2000), GST–β-arrestin-1/2 (Hara et al., 2011), GFP–β-arrestin-1 and FLAG–β-arrestin-1/2 (Scott et al., 2002; Storez et al., 2005), β-arrestin 1 and HA–β-arrestin-1 (Bhandari et al., 2007), HA–ARRDC1 (Rauch and Martin-Serrano, 2011), GST–Itch (Angers et al., 2004).
ARRDC1, ARRDC1 PY* (PPxYs were mutated in PAxYs), S-tagged ARRDC2, ARRDC3, ARRDC4 were gifts from Olivier Staub (Department of Pharmacology and Toxicology, University of Lausanne). VSV-tagged DTX was described previously (Chastagner et al., 2006), 6×Ubi-His was from M. Treier (EMBL, Heidelberg, Germany). CSL-Luc was a gift from T. Honjo (Kyoto University, Japan) and is referred to as pGa981-6 (Minoguchi et al., 1997). HA-tagged ARRDC1 and HA-tagged ARRDC1 ΔPY were generated by inserting PCR products in pcDNA3-HA backbone vector (ECOR1/XhoI), using as forward primer: 5′-CTAGCGGAATTCGGGCGAGTGCAGCTCTTC-3′ and as reverse primers 5′- TGCATGCTCGAGTCAGCTCTCAGGGGTCAG-3′ for HA-ARRDC1 and 5′- TGCATGCTCGAGTCAAAGAATCAAGGTGCTGGTAGT-3′ for HA-ARRDC1 ΔPY. All constructs were verified by sequencing.
Cell lines and transfections
Cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS). OP9-Dll1 and U2OS-FL cell lines were described previously (Six et al., 2004; Moretti et al., 2012). Mouse embryonic fibroblast (MEF) cells isolated from β-arrestin 1 and β-arrestin 2 double-knockout (DKO) embryos and matched wild-type (WT) embryos were provided by Robert J. Lefkowitz (Duke University, Durham, NC). These cells were retroviral transduced: high titers of recombinant HA-tagged Notch FL viruses were obtained 48 hours after transfection of the Plat-E ecotropic packaging cell line with retroviral expression plasmids. Clonal populations were obtained by limiting dilution. Transfections were performed in HEK 293T with calcium phosphate or jetPRIME™ (Polyplus) according to the manufacturer's instructions. MEF-FL, OP9 DLL1 and U2OS-FL were transfected with jetPRIME™ (Polyplus) or Fugene HD (Promega).
GST pull-down analysis
GST fusion constructs were transformed into E. coli DG1 cells and expression was induced by IPTG. Approximately equal amounts of glutathione-S transferase (GST) alone or in fusion with Itch, as estimated from a Coomassie-stained gel, were bound to glutathione-Agarose beads (Sigma). In vitro transcription and translation were made according to the manufacturer's instructions in reticulocyte lysates (Promega), in the presence of [35S]methionine. The products were incubated for 1 hour at 4°C with GST charged beads in a buffer containing 50 mM Tris-HCl (pH 8.0), 1% Triton X-100, 300 mM NaCl. The bound proteins were washed extensively in the same buffer, eluted in Laemmli buffer, resolved on SDS-PAGE and detected by autoradiography.
HEK293T cells were harvested 36 hours after transfection and lysed in 8 M urea, 0.1 M NaH2PO4, 10 mM Tris-HCl (pH 8), 1% Triton X-100 and 20 mM Imidazole at room temperature. His–Ub-conjugated proteins were purified on chelating Sepharose beads (Pharmacia), previously charged with Nickel. Ubiquitylated proteins were washed extensively with the same buffer and then with a Tris-HCl, pH 6.3, buffer. The Ubiquitin-conjugated products were then eluted in Laemmli buffer for western blot analysis.
Cell extracts and immunoprecipitation
HEK293T cells were collected 36 hours after transfection and lysed in 50 mM Tris-HCl (pH 7.9), 5 mM MgCl2, 300 mM NaCl, 1% Triton X-100, supplemented with protease inhibitor cocktail (Roche) and N-ethylmaleimide (NEM), an inhibitor of deubiquitylases (Sigma). The lysates were cleared by centrifugation at 14,000 rpm for 20 minutes at 4°C. Immunoprecipitations were performed with the appropriate antibodies in the same buffer. When indicated, the immunoprecipitates were eluted by FLAG peptide competition (2 mg/ml) for 1 hour at 4°C. Samples were denatured in Laemmli buffer for SDS-PAGE resolution and immunoblotting.
HA antibody-uptake assay
Cells were grown on glass coverslips for 24 hours, and when indicated, transfected (Fugene HD, Promega) with plasmids for the following 24 hours. After 1 hour of incubation in serum-free medium, cells were incubated for 15 minutes with anti-HA, washed again and incubated in serum-free medium at 37°C for various periods of time (0, 30, 60, 90 and 180 minutes). Cells were then quickly washed with cold PBS, fixed with 4% paraformaldehyde for 20 minutes and permeabilized with 0.2% Triton X-100 for 5 minutes before incubation with appropriate primary and secondary antibodies. Cells were mounted in Mowiol (Calbiochem, Merck Biosciences, Darmstadt, Germany) and images acquired with 0.24 µm sections using an Axioplan 2 imaging system with ApoTome (Carl Zeiss MicroImaging, Le Pecq, France). When indicated, cells were treated with Leupeptin (Sigma) at 20 µM.
WT, C2 MEFs or U2OS-FL cells were grown in 24-well plates and transiently transfected with reporter genes (Notch reporter: CSL-firefly luciferase; internal control: TK-Renilla luciferase) in triplicate with FuGeneHD transfection reagent (Roche, Mannheim, Germany), then co-cultured with OP9-Dll1 cells for 18 hours. The co-cultured cells were lysed using Passive Lysis buffer (Promega). A fraction of cell lysates was used to measure Firefly and Renilla luciferase activities in a luminometer Centro XS (Berthold). For western blot analysis, co-culture wells were lysed with 8 M urea, 0.1 M NaH2PO4, 10 mM Tris-HCl (pH 8), 1% Triton X-100 and 20 mM Imidazole.
β-arrestin expression was silenced using a validated siRNA (Coureuil et al., 2010). siARRDC1 target sequence was: CAGCCUCGUGUUCUAUAUCUU. siUSP12 target sequence was UAGCAGAUCUCUUCCAUAG. ON-TARGET plus smart pool targeting human CSL (L-007772-00-0005), non-targeting pool (D-001810-10−05) or GFP Duplex I (P-002048-01-20,) were from Thermo Scientific.
Cells were harvested and washed twice with PBS, fixed in 4% paraformaldehyde and labeled with anti-HA followed by incubation with appropriate secondary antibody (Alexa Fluor 488). Finally, the cells were washed and analyzed by using Cyan ADP Flow Cytometer (Beckman Coulter). The data were analyzed using FlowJo (Tree Star) software.
Colocalization images quantification using Icy
The colocalization events were quantified using Icy software (de Chaumont et al., 2012), http://icy.bioimageanalysis.org, composed of (1) efficient fluorescence-labeled spot detection based on wavelet transform and (2) distance object-based colocalization. The colocalization distance was defined as a positive hit when the pixel distance between Notch and LAMP1 centroids was less or equal to 5 pixels (0.5 µm). Percentage of colocalization was defined as a ratio of colocalization value over spot detection number.
Quantification and statistical analysis
Protein bands and images were quantified by densitometry and analysed with Quantity One Software (BioRAD Laboratories) and ImageJ. Data were analyzed using Student's t-test. Error bars correspond to s.e.m.
In situ PLA technology
The PLA technique was used to detect in situ interaction between Notch and ARRDC1 using primary antibodies specific for these proteins (rabbit polyclonal anti-ARRDC1 and mouse anti-Notch, Abcam) coupled to a Duolink II kit (http://www.olink.com/). This technique allows detection of a single fluorescent spot when Notch and ARRDC1 are closely located (<40 nm) (Söderberg et al., 2006).
U2OS-FL cells were lysed and RNAs were purified using RNeasy mini kit (Qiagen). A reverse transcription followed by quantitative PCR was then performed. CFX96TM real-time PCR detection system and the CFX ManagerTM Software (Bio-Rad) were used for analysis. The following oligonucleotides were used: human β-arrestin 1, 5-GCCCAAGGTCACACAGAAAG-3′ (forward) and 5-AGGGCAAGCGATGAAGACT-3′ (reverse); human ARRDC1, 5-CAGTGCCCACTACCAGCAC-3′ (forward) and 5-ATAGGGGTAGCCCCAAGAAC-3′ (reverse); human UBCH5B, 5-TGAAGAGAATCCACAAGGAATTGA-3′ (forward) and 5′-CAACAGGACCTGCTGAACACTG-3′ (reverse). Relative expression of β-arrestin 1 and ARRDC1 was normalized using UBCH5B.
We thank Sébastien Léon and Franck Coumailleau for critical reading of the manuscript. We thank J. Aster (Harvard medical school, Boston, MA), R. Kopan (Washington University, St Louis, MO), Robert Lefkowitz, Stefano Marullo and Mark Scott (Institut Cochin, Université Paris Descartes, Paris, France), Adriano Marchese (Stritch School of Medicine, Loyola University Chicago, Maywood, IL) Juan Martin-Serrano (Department of Infectious Diseases, King's College London School of Medicine, London, UK), Olivier Staub, M. Treier A. Angers (Université de Montreal, Canada) and T. Honjo for generous gifts of materials; Marcello Albanesi and P. H. Commere for the Flow Cytometry analysis (Institut Pasteur, France).
L.P. and P.C. performed experiments; L.P. and C.B. designed experiments; L.P., V.M.Y. and C.B. analyzed data; L.P., A.I. and C.B. wrote the paper.
This project was supported in part by Institut Pasteur; Centre national de la recherche scientifique (CNRS); Association pour la Recherche sur le Cancer [grant number SFI20101201664]; and Ligue Nationale Contre le Cancer [grant number LNCC RS11/75-21] to C.B. L.P. was supported by the Pasteur-Paris University (PPU) International PhD program, SFD (Société Française de Dermatologie) and FRM (Fondation pour la Recherche Médicale) fellowships.