Reciprocal cross-regulation between RNF41 and USP8 controls cytokine receptor sorting and processing

The cellular sensitivity to a cytokine signal is tightly controlled to avoid aberrant, potentially pathological signaling. Many studies have focused on feedback inhibition mechanisms like phosphatases (Xu and Qu, 2008), protein inhibitors of activated STAT (PIAS) proteins (Shuai and Liu, 2005) and suppressors of cytokine signaling (SOCS) proteins (Croker et al., 2008), which suppress the signal transduction and gene transcription initiated after binding of the cytokine to its receptor at the cell surface. The cellular response to a cytokine additionally depends on the number of signaling-competent receptors exposed at the cell surface and their subsequent intracellular routing. Following de novo synthesis and delivery to the plasma membrane receptors are internalized into early endosomes, irrespective of ligand binding. These endosomes operate as a signaling platform by coupling activated receptors to specific signaling pathways (Pálfy et al., 2012) and serve as a central sorting station (Clague et al., 2012). From here, receptors can either be recycled back to the plasma membrane or sorted out for lysosomal degradation via late endosomes or multivesicular bodies (MVB). Incorporation of activated receptors from the outer membrane into intraluminal vesicles of MVBs sequesters the receptor away from cytoplasmic signaling molecules, thus attenuating cytokine signaling. The MVB content gets cleared by lysosomal degradation after fusion with lysosomes.

Ubiquitin serves as a trafficking signal and mediates the sorting and fate of receptors. K48-linked polyubiquitylation tags proteins for proteasomal degradation, while K63-linked polyubiquitylation and monoubiquitylation are implicated in the endocytosis and lysosomal delivery of various receptors (Komander, 2009; Piper and Lehner, 2011). Adaptor proteins recognize ubiquitylated proteins via ubiquitin-binding domains (UBD) and mediate their progression along the endocytic pathway. The endosomal complex required for transport (ESCRT) proteins are essential for the selection of cargo towards lysosomal degradation (Raiborg and Stenmark, 2009). Phosphorylation and ubiquitylation of these and other endocytic adaptor proteins allows regulation of the recruitment and intracellular traffic of cargo. Several E3 ubiquitin ligases are involved in the endocytosis and lysosomal targeting of cytokine receptors. They coordinate ubiquitin transfer from a ubiquitin-conjugating enzyme (E2) to a specific substrate by serving as a scaffold. C-cbl, NEDD4, SOCS3, β-TRCP and CHIP play a role in ligand-dependent internalization and lysosomal degradation of the activated erythropoietin, growth hormone, prolactin, gamma common, type I interferon, gp130, granulocyte colony-stimulating factor (G-CSF) and interleukin-10 (IL-10) receptor complexes. (da Silva Almeida et al., 2012; Gesbert et al., 2005; Jiang et al., 2011; Kumar et al., 2004; Kumar et al., 2007; Malardé et al., 2009; Meyer et al., 2007; Slotman et al., 2012; Tanaka et al., 2008; Wölfler et al., 2009). However, little is known about the mechanisms that govern cytokine receptor trafficking in the absence of cytokine stimulation.

We recently reported that the E3 ubiquitin ligase RNF41 (also known as Nrdp1 or FLRF), affects the basal routing of JAK2-dependent cytokine receptors. We observed that cathepsin L cleaves the leptin receptor (LR), LIF receptor (LIFRα), IL-6 receptor (IL-6Rα) and generates a C-terminal stub (CTS), which is degraded in the lysosomes. RNF41 blocks CTS formation, which reflects reduced lysosomal degradation of these receptors, and simultaneously enhances their ectodomain shedding by transmembrane ADAM (a disintegrin and metalloproteinase) proteases (Wauman et al., 2011). Ectodomain shedding generates soluble receptors, which can capture circulating ligand. This can either antagonize signaling as it prevents normal ligand binding to full length, membrane-bound receptors or benefit signaling by prolonging cytokine bio-availability (Levine, 2008). Moreover, in case of IL-6, soluble receptors allow cytokine stimulation of cells that lack the respective membrane-bound receptor, this way enabling trans-signaling. RNF41 thus acts at the crossroads between cytokine receptor routing towards lysosomes or cellular compartments where ectodomain shedding occurs. RNF41 belongs to the family of really interesting new gene (RING) E3 ubiquitin ligases and differentially ubiquitylates various substrates. RNF41 mediates K48-polyubiquitylation of MyD88, Parkin and BRUCE, while it induces K63-polyubiquitylation of TBK1 and unspecified polyubiquitylation of ErbB3, retinoic acid receptor alpha (RARα), interleukin-3 and erythropoietin receptors (Jing et al., 2008; Qiu and Goldberg, 2002; Qiu et al., 2004; Wang et al., 2009; Yu and Zhou, 2008). RNF41 however does not influence the ubiquitylation of LR, LIFRα, IL-6Rα or JAK2, which is common to all these type I cytokine receptor complexes. As RNF41 was shown to interact with TSG101 and USP8 (Markson et al., 2009; Wu et al., 2004), both coupled to the ESCRT complex, we speculated that one of these proteins might be the direct substrate of RNF41 and assist in cytokine receptor trafficking.

DUBs reverse ubiquitylation, enabling spatiotemporal control of protein ubiquitylation. The balance between ubiquitylation and deubiquitylation regulates protein stability and determines the efficiency of lysosomal sorting (Wright et al., 2011). USP8 (also known as Ubpy) regulates the degradation of various transmembrane proteins at the sorting endosome by modulating the ubiquitin dynamics of both cargo and sorting proteins. USP8 interacts with signal transducing adaptor molecule (STAM) and stabilizes STAM and hepatocyte growth-factor-regulated substrate (Hrs), which together constitute the ESCRT-0 complex and govern the early steps of receptor trafficking en route to the lysosomes (Niendorf et al., 2007; Row et al., 2006). By stabilizing the ESCRT-0 complex, USP8 directs the chemokine receptor 4 (CXCR4) towards lysosomal degradation, while USP8-mediated deubiquitylation of the epidermal growth factor receptor (EGFR) favors receptor recycling (Berlin et al., 2010b; Niendorf et al., 2007). Moreover, USP8 stabilizes RNF41 levels and reverts RNF41 ubiquitylation (Cao et al., 2007). We here reveal a reciprocal regulation between RNF41 and USP8. RNF41 ubiquitylates, destabilizes and relocalizes USP8. In addition, USP8 knockdown reduces lysosomal degradation of the LR and LIFRα and enhances their recycling and subsequent shedding, phenotypically resembling the effect of RNF41 ectopic expression. Loss of USP8 thus accounts for the functional role of RNF41 in cytokine receptor degradation and shedding.

USP8 suppresses RNF41 ubiquitylation and stabilizes RNF41 protein levels (Cao et al., 2007). Conversely, ectopic expression of RNF41 lowers the protein levels of endogenous USP8 in HEK293T cells (Fig. 1A). Based on the crystal structure of the Rhodanese domain of USP8 in complex with the C-terminal substrate binding domain of RNF41 (Avvakumov et al., 2006), we created an RNF41 Q266A R269E mutant (further referred to as RNF41 AE) with strongly reduced binding to USP8 in MAPPIT, AlphaScreen and GST-pulldown assays (Fig. 1B). The RNF41 AE mutant is properly folded and functional as it preserves the interaction with other RNF41-interacting proteins (data not shown) and is readily recognized by an antibody detecting endogenous RNF41. RNF41 AE mutant expression no longer lowers endogenous USP8 protein levels (Fig. 1C), indicating that suppression of USP8 by RNF41 depends on direct interaction between RNF41 and USP8. The RNF41 AE mutant even enhances endogenous USP8 levels, although it hardly interacts with USP8. Most likely, this mutant counters endogenous RNF41 activity by sequestering additional endogenous factors necessary for USP8 suppression. Expression of an Etagged RNF41 C34S H36Q mutant (further referred to as RNF41 SQ), with impaired E3 ubiquitin ligase activity due to loss of recruitment of the E2 conjugating enzyme (Qiu and Goldberg, 2002), also strongly stabilizes USP8 expression due to direct competition between this mutant and endogenous RNF41 for USP8 binding (Fig. 1C). This reveals that the E3 ubiquitin ligase activity of RNF41 is indispensable to reduce USP8 levels and suggests that USP8 is a substrate of RNF41. This was confirmed by ubiquitylation analysis of exogenous USP8 performed in the presence of both MG132 and chloroquine to stabilize ubiquitin linkage preceding proteasomal or lysosomal degradation. RNF41 but not RNF41 SQ expression enhances USP8 ubiquitylation (Fig. 1D). Residual interaction with USP8 (Fig. 1B) might explain why the RNF41 AE mutant still moderately provokes USP8 ubiquitylation. Finally, RNF41 dramatically changes the intracellular localization of USP8. Exogenous USP8 is homogeneously distributed throughout HeLa cells. However, co-expression of RNF41 redistributes USP8 to large intracellular vesicles (Fig. 1E).

We previously demonstrated that RNF41 expression blocks cathepsin L-mediated LR, LIFRα and IL-6Rα C-terminal stub (CTS) formation. These degradation products are generated in a ligand-independent manner and are stabilized by overnight incubation with chloroquine, an inhibitor of lysosomal degradation (Wauman et al., 2011) (Fig. 2A,C, lane 2 versus 1). The LR constitutively internalizes in a leptin-independent manner (Belouzard et al., 2004). After internalization, it was previously reported that the LR does not appear to recycle, but is transported via Rab5-positive early endosomes and Rab7-positive late endosomes to the lysosomes, where LR degradation occurs (Belouzard and Rouillé, 2006). Rab GTPases coordinate intracellular vesicle traffic and several Rab mutants are well established to interfere with specific vesicle budding, uncoating, motility and fusion events along the endocytic pathway (Stenmark, 2009). Co-expression of the dominant-negative Rab5 S34N mutant, which prevents endocytic transport from the plasma membrane to early endosomes or the dominant-negative Rab7 T22N mutant, which precludes exit from the early to late endosomes, almost completely blocks LR CTS formation in LR-expressing HEK293T cells (Fig. 2A) (Feng et al., 1995; Stenmark et al., 1994). This indicates that cathepsin L-dependent CTS formation occurs downstream of Rab5 and Rab7 activity, most likely in the MVBs, prior to fusion with lysosomes and further receptor degradation (which is inhibited by chloroquine). Inhibition of fast LR recycling upon expression of dominant-negative Rab4 S22N mutant (de Wit et al., 2001) has no effect on LR CTS formation (Fig. 2A). Minor trapping of LR in the slow recycling pathway via dominant-negative Rab11 S25N mutant expression might explain slightly reduced LR CTS generation (Fig. 2A) (Ren et al., 1998). The N-terminal RING domain of RNF41 is crucial for the inhibitory effect of RNF41 on CTS formation (Wauman et al., 2011). This implies that RNF41 suppresses CTS generation and thus cytokine receptor degradation via ubiquitylation of a so far unknown substrate. Since we now identified USP8 as a new substrate of RNF41-dependent degradation (Fig. 1D), we examined whether RNF41 inhibits LR CTS generation as a result from its interaction with endogenous USP8. In contrast to wild-type RNF41, expression of the RNF41 AE mutant leaves LR CTS generation unaffected (Fig. 2A, lane 8 versus 7). This reveals that direct interaction between RNF41 and USP8 is essential for the functional effect of RNF41 on LR CTS formation. Given the central role of USP8 in lysosomal routing of various receptors, we analyzed whether USP8 primarily influences CTS generation. Silencing of endogenous USP8 protein using two different siRNAs (Fig. 2B) mimics the effect of RNF41 ectopic expression and severely impairs (in case of USP8 siRNA1, which is slightly less efficient in suppressing USP8) or completely abolishes (in case of USP8 siRNA2) LR CTS generation. This shows that the presence of USP8 is essential for ligand-independent LR incorporation into MVBs, where the CTS is generated, and further degradation in the lysosomes. Similar results on stub formation were obtained for the LIFRα chain. Expression of Rab5 S34N, Rab7 T22N, wild-type RNF41 (Fig. 2C) or silencing of endogenous USP8 protein (Fig. 2D) blocks LIFRα CTS formation, while expression of Rab4 S22N or RNF41 AE has no effect on the LIFRα CTS (Fig. 2C). As for the LR CTS, Rab11 S25N expression partially precludes LIFRα CTS formation. Together, this illustrates that USP8 promotes cytokine receptor CTS generation.

As RNF41 suppresses USP8 levels (Fig. 1A) and as USP8 knockdown blocks CTS generation (Fig. 2B,D), reduction of endogenous USP8 by RNF41 might explain the inhibitory effect of RNF41 on CTS formation described in our previous work (Wauman et al., 2011). Under this scenario, ectopic expression of USP8 should compensate for the loss of endogenous USP8 levels upon RNF41 expression and should restore CTS generation. Indeed, co-expression of USP8 reverses the inhibitory effect of RNF41 on LR CTS formation in a dose-dependent manner (Fig. 3, lane 5 versus 3). Moreover, the catalytic activity of USP8 is essential as the inactive USP8 C786S mutant is unable to rescue LR CTS generation (Fig. 3, lane 7 versus 3). In contrast to wild-type USP8, this catalytic inactive mutant likely fails to counteract RNF41-mediated suppression by self-deubiquitylation, which might explain the lack of expression of the USP8 C786S mutant upon RNF41 co-expression (Fig. 3, lane 7 versus 5). Importantly, USP8 and RNF41 influence each other's expression. RNF41 induces USP8 ubiquitylation and suppression (Fig. 1A,D), while on the other hand, wild-type USP8 enhances RNF41 stability (Fig. 3, lane 5 versus lane 3), as reported before (Wu et al., 2004). This points to a reciprocal cross-regulation between the DUB USP8 and the E3 ubiquitin ligase RNF41, controlling cytokine receptor trafficking.

Soluble LR and LIFRα can be generated by alternative splicing or ectodomain shedding (Maamra et al., 2001; Tomida, 2000). We previously demonstrated that RNF41 expression blocks CTS formation and concomitantly enhances LR and LIFRα ectodomain shedding by metalloproteases of the ADAM family, resulting in increased soluble cytokine receptor levels (Wauman et al., 2011). Therefore, cell culture supernatants from the cells transfected in Fig. 2 were collected to simultaneously analyze the effect of dominant-negative Rab proteins, RNF41 mutants and USP8 knockdown on cytokine receptor shedding. Rab7 T22N, Rab4 S22N or Rab11 S25N expression has no significant effect on LR or LIFRα shedding (Fig. 4A,C), indicating that neither recycling nor blocking of receptor transport from the early endosomes to the lysosomes influences basal LR and LIFRα ectodomain shedding. Expression of Rab5 S34N, RNF41 or USP8 knockdown on the other hand clearly increases soluble LR (Fig. 4A,B) and LIFRα (Fig. 4C,D) levels, while RNF41 AE induced no significant increase compared to wild-type RNF41. This again parallels the potentiating effect of RNF41 expression on LR and LIFRα ectodomain shedding, further underscoring that suppression of endogenous USP8 accounts for the enhanced shedding by RNF41. To investigate whether RNF41 expression or loss of USP8 levels enhance LR ectodomain shedding via enhanced fast or slow recycling of the LR to the plasma membrane, we explored if Rab4 S22N or Rab11 S25N co-expression impairs the enhanced shedding. Rab11 S25N or Rab4 S22N co-transfection both impair the enhanced shedding upon RNF41 transfection (Fig. 4E) or USP8 knockdown (Fig. 4F), although Rab11 S25N has a more potent effect. This shows that RNF41-dependent suppression of endogenous USP8 mainly stimulates slow recycling of LR to the plasma membrane, where ectodomain shedding occurs.

The decision to recycle receptors back to the plasma membrane, where receptors are susceptible to ectodomain shedding, or to direct them towards the lysosomes for degradation is largely a function of ESCRT-0. The ESCRT-0 complex consists of Hrs, constitutively associated with either STAM1 or redundant STAM2 (also known as Hrs-binding protein or Hbp). ESCRT-0 binds ubiquitin via its ubiquitin-binding domains (UBD). This way it concentrates ubiquitylated cargo prior to sorting for lysosomal degradation. Direct binding of USP8 to the SH3 domain of STAM proteins deubiquitylates and stabilizes the ESCRT-0 complex (Niendorf et al., 2007; Row et al., 2006). Since we demonstrated that RNF41 clearly destabilizes USP8, we next determined whether this indirectly affects the ESCRT-0 integrity. RNF41 expression markedly reduces endogenous Hrs, STAM1 and STAM2 protein levels (Fig. 5), depending on its direct interaction with USP8 and intact E3 ubiquitin ligase activity, as the RNF41 AE and RNF41 SQ mutants have no effect. The effect on ESCRT-0 instability is more pronounced upon RNF41 expression compared to the effect resulting from expression of the catalytic inactive USP8 C786S mutant (Fig. 5, lane 2 versus 5). RNF41 has also been reported to interact with TSG101 (Markson et al., 2009), part of the ESCRT-I complex. However, endogenous TSG101 levels are unaffected upon RNF41 expression, showing that the destabilizing effect of RNF41 is restricted to ESCRT-0.

Ubiquitylation regulates cell surface expression of both unstimulated and activated receptors by affecting their proteasomal degradation, internalization and subsequent intracellular sorting. E3 ubiquitin ligases confer substrate specificity by mediating ubiquitin transfer from the E2 ubiquitin-conjugating enzyme to a specific substrate. Moreover, ubiquitylation is reversible and thus highly dynamic. The activity of E3 ligases is counterbalanced by the activity of deubiquitylating enzymes (DUBs), which allow spatial and temporal control of receptor trafficking and turn-over (Clague et al., 2012). We previously demonstrated that RNF41 affects the basal turnover of several JAK2-dependent cytokine receptors, including the LR, LIFRα and IL-6Rα in a dual way: RNF41 inhibits their lysosomal degradation and concomitantly enhances their ectodomain shedding (Wauman et al., 2011). However, the substrate of RNF41 responsible for these effects remained unidentified. In this study, we further reveal the underlying molecular mechanism by demonstrating for the first time that RNF41, besides inducing its own auto-ubiquitylation and subsequent proteasomal degradation (Qiu and Goldberg, 2002), also enhances USP8 poly-ubiquitylation (Fig. 1D) and reduces endogenous USP8 protein expression (Fig. 1A). Based on the crystal structure of the interaction between the C-terminal substrate binding domain of RNF41 and the Rhodanese domain of USP8 (Avvakumov et al., 2006), we identified Q266 and R269 as two critical RNF41 residues required for efficient USP8 binding (Fig. 1B). Mutation of these residues impairs USP8 ubiquitylation and protein suppression (Fig. 1C,D), showing that direct interaction between RNF41 and USP8 is required. Moreover, proper RNF41 E3 ubiquitin ligase activity is essential as disruption of the E3 ubiquitin ligase activity by two point mutations in the RING domain of RNF41 (C34S and H36Q) inhibits RNF41 auto-ubiquitylation (Qiu and Goldberg, 2002) as well as USP8 ubiquitylation (Fig. 1D). As a result, ectopic expression of this RNF41SQ mutant is more stable than wild-type RNF41 (Fig. 1C and Fig. 5) (Wu et al., 2004) and stabilizes endogenous USP8 levels (Fig. 1C). Interestingly, E-tagged RNF41 SQ expression additionally enhances endogenous RNF41 levels, migrating at a slightly lower molecular mass (Fig. 1C; Fig. 5, lane 4). This can be readily explained as USP8 DUB activity conversely suppresses RNF41 auto-ubiquitylation (Cao et al., 2007) and enhances RNF41 stability. Disruption of the USP8 DUB activity destabilizes endogenous RNF41 protein expression (Wu et al., 2004). Moreover, USP8 might deubiquitylate itself as catalytic inactive USP8 is poly-ubiquitylated (Mizuno et al., 2005) and less stable than wild-type USP8 upon coexpression with wild-type RNF41 (Fig. 3, lane 7 versus 5; Fig. 5, lane 5 versus 6), as reported before (Wu et al., 2004). Together, this reveals an important reciprocal regulation between the E3 ubiquitin ligase RNF41 and the DUB USP8. The expression and activity of RNF41 and USP8 appears to be determined by a cross-regulatory balance wherein the relative expression levels of both RNF41 and USP8 control each other's and their own stability via their ubiquitin ligase and deubiquitylase activity, respectively. Eventually, this RNF41–USP8 balance will determine the outcome of cytokine receptor trafficking.

ESCRT-0, consisting of a complex between Hrs and either STAM1 or STAM2, partitions cargo arriving at the sorting endosomes between recycling and progression to the lysosomes and thereby acts as a critical regulatory checkpoint in the determination of a receptor's fate. Increasing evidence suggests that ubiquitin ligase and DUB interaction with ESCRT members modulates trafficking outcomes (Hurley, 2010; Wright et al., 2011). USP8 is one of the DUBs that associates with the STAM2 SH3 domain through three conserved central RXXK domains or to charged multivesicular body proteins (CHMP) of the ESCRT-III via the USP8 N-terminal microtubule interacting and trafficking (MIT) domain (Berlin et al., 2010a; Row et al., 2007). USP8 influences the post-internalization trafficking of several plasma membrane proteins in a substrate-specific way. USP8 directly deubiquitylates protease-activated receptor 2 (PAR₂) and δ-opioid receptor (DOR), two G-protein coupled receptors (GPCRs), and the ion channel KCa3.1, which is a required intermediate step for their lysosomal targeting and degradation (Balut et al., 2011; Hasdemir et al., 2009; Hislop et al., 2009). In case of the EGFR, contradictory results have been published: although some reports demonstrated that USP8 stimulates EGFR degradation (Alwan and van Leeuwen, 2007; Bowers et al., 2006; Row et al., 2006), USP8 more likely protects the receptor from lysosomal degradation by direct deubiquitylation of the receptor and promoting EGFR recycling (Berlin et al., 2010a; Mizuno et al., 2005; Niendorf et al., 2007). The chemokine receptor 4 (CXCR4) is stabilized after depletion or catalytic inactivation of USP8, although it is no direct target of USP8. In this case, USP8 directs CXCR4 lysosomal degradation by supporting ESCRT-0 stability (Berlin et al., 2010b). Here, we primarily assign a role for USP8 in promoting constitutive lysosomal degradation of cytokine receptors, similar to CXCR4. We previously reported that cathepsin L cleaves cytokine receptors and generates a C-terminal stub (CTS), which is degraded in the lysosomes (Wauman et al., 2011). Using dominant-negative Rab5 and Rab7 mutants, we further specify that this CTS is generated in the MVBs (Fig. 2A,C). Blocking lysosomal degradation with chloroquine stabilizes the CTS and can be used as a read-out to monitor the sorting of LR and LIFRα towards lysosomes. USP8 knockdown strikingly phenocopies ectopic RNF41 expression as both treatments block LR (Fig. 2A,B) and LIFRα CTS formation (Fig. 2C,D), and concomitantly enhance their ectodomain shedding (Fig. 4A–D). In line with the effect on CTS formation, it was shown before that an intermediate EGFR degradation product was no longer formed in USP8 knockdown cells (Row et al., 2006). USP8 co-expression successfully counteracts the inhibitory effect of RNF41 on LR CTS formation (Fig. 3), while disrupting RNF41 interaction with USP8 preserves LR and LIFRα CTS generation (Fig. 2A,C) and has no significant effect on ectodomain shedding (Fig. 4A,C), indicating that these functional effects of RNF41 are secondary to RNF41-dependent USP8 repression. Dominant negative Rab4 and Rab11 mutants that prevent slow or fast recycling, respectively, both interfere with the enhanced LR ectodomain shedding upon RNF41 expression or USP8 knockdown (Fig. 4E,F), indicating that enhanced recycling underlies the enhanced ectodomain shedding. Moreover, loss of USP8 seems to actively stimulate cytokine receptor sorting along the recycling pathway, as merely blocking receptor entry into the lysosomal degradation pathway by expression of dominant-negative Rab7 is not sufficient to enhance ectodomain shedding. Importantly, in addition to suppressing endogenous USP8 levels, ectopic expression of RNF41 specifically depletes endogenous Hrs, STAM1 and STAM2 protein levels, without affecting ESCRT-I TSG101 protein expression (Fig. 5). A direct effect of RNF41 on either of the ESCRT-0 components can be excluded as RNF41 solely interacts with USP8 in MAPPIT assays, which do recapitulate the interaction between USP8 and either STAM1 or STAM2 (data not shown). Moreover, RNF41 is clearly more potent in suppressing ESCRT-0 proteins than catalytic inactive USP8 C786S (Fig. 5, lane 2 versus 5). RNF41 expression thus phenotypically resembles the described effect of USP8 knockdown on ESCRT-0 stability (Berlin et al., 2010a; Niendorf et al., 2007; Row et al., 2006). USP8 binding to the SH3 domain of STAM proteins results in STAM and Hrs deubiquitylation and protection against proteasomal degradation by functionally opposing their ubiquitylation by the E3 ubiquitin ligase AIP4 (or Itch). This way, USP8 sustains ESCRT-0 integrity and allows CXCR4 sorting towards lysosomes, irrespective of ligand binding (Berlin et al., 2010b). We propose a model in which USP8 similarly indirectly promotes constitutive lysosomal degradation of cytokine receptors by stabilizing the ESCRT-0 complex. Loss of USP8, either via knockdown or ectopic expression of RNF41, abrogates ESCRT-0 functionality and reroutes receptors from the lysosomal degradation pathway (where cathepsin L cleavage occurs), back to the plasma membrane (where ectodomain shedding occurs) by stimulating slow and fast recycling (Fig. 6). The balance between USP8 DUB and RNF41 E3 ligase expression and activity this way eventually defines the half-life and cell surface expression of cytokine receptors at steady-state.

The question remains how USP8 and RNF41 are precisely recruited to cytokine receptor complexes at the sorting endosome. ESCRT-0 assembles as a heterotetrameric complex on the endosomal membrane, providing ten ubiquitin-binding domains (UBD) (Lange et al., 2012; Mayers et al., 2011), which recognize and sort ubiquitylated cargo. However, the ESCRT-0 complex can also mediate ubiquitin-independent trafficking. Hrs recognizes a hydrophobic amino acid cluster in cytokine receptors and drives ubiquitin-independent lysosomal degradation of the interleukin-2 (IL-2) receptor β chain (Amano et al., 2011; Yamashita et al., 2008), while STAM proteins associate with JAK kinases (Endo et al., 2000; Takeshita et al., 1997). Moreover, the recruitment, activity and expression levels of USP8 and RNF41 can be further regulated by extracellular growth factors, LPS stimulation and intracellular phosphorylation events (Cao et al., 2007; Meijer et al., 2013; Naviglio et al., 1998; Wang et al., 2009). In this respect, we tried to avoid possible stimulation-dependent regulation of any of the two proteins, by employing a serum-free experimental setup. Finally, RNF41 and USP8 activity are also affected by their subcellular localization. Both endoplasmic reticulum (ER)- and endosome-localized RNF41 are implicated in ErbB3 degradation (Cao et al., 2007; Fry et al., 2011), while growth factor stimulation and DUB inactivation enhance recruitment of USP8 to early endosomes (Mizuno et al., 2005; Row et al., 2006). In addition to this, we show that RNF41 expression causes a drastic intracellular redistribution of USP8 in HeLa cells. In agreement with other publications, USP8 is normally dispersed throughout the cytosol with some additional staining at the plasma membrane and in punctuate, perinuclear vesicles (Cao et al., 2007; Row et al., 2006). However, RNF41 co-transfection results in USP8 accumulation in yet uncharacterized intracellular vesicles (Fig. 1E).

In conclusion, we identified the DUB USP8 as a substrate of the E3 ubiquitin ligase RNF41 and state that USP8 loss accounts for the reduced lysosomal degradation and enhanced ectodomain shedding of JAK2-dependent cytokine receptors upon ectopic expression of RNF41 described in our previous work (Wauman et al., 2011). Since RNF41 affects USP8 and ESCRT-0 stability, a more generic role for RNF41 on cargo sorting seems plausible, which will be the subject for further investigation. Moreover, imbalance between RNF41 and USP8 levels in the context of cytokine receptors and other proteins is medically relevant as aberrant RNF41 expression is linked to the onset of breast cancer, prostate cancer and chronic lymphocytic leukemia (CLL), while USP8 is a multiple myeloma survival factor (Chen et al., 2010; Hanlon et al., 2009; Tiedemann et al., 2012; Yen et al., 2006).

The pXP2d2-rPAP1-luciferase reporter, pMET7-hLR-HA, pMET7-hLIFRα-HA, empty pMG1 prey, pMET7-FLAG-SVT, pMET7-hRNF41-Etag and pMET7-Etag-hRNF41 vectors were described previously (Eyckerman et al., 2001; Wauman et al., 2011). The latter construct was used as a template to PCR untagged hRNF41 using 5′-GCGGAATTCGCCATGGGGTATGATGTAACCCG-3′ and 5′-CGCTCTAGATTATATCTCTTCCACGCCATGCG-3′ and clone it into an empty, EcoRI-XbaI opened pMET7 vector, generating pMET7-hRNF41. MAPPIT bait receptor vectors are generated based on the standard pSEL receptor construct, containing the extracellular part of the human EpoR and the transmembrane and intracellular parts of the mouse LR reported elsewhere (Eyckerman et al., 2001). RNF41 bait vector was generated by amplifying hRNF41 using 5′-GCGGAGCTCGATGGGGTATGATGTAACCCGTTTCC-3′ and 5′-CGCGCGGCCGCTTATATCTCTTCCACGCCATGCG-3′ with pMET7-hRNF41-Etag as a template and subsequent substitution of FKBP12 by RNF41 via SacI-NotI digestion of the pSEL-FKBP12 vector (Eyckerman et al., 2001). The Etag in pMET7-Etag-hRNF41 was cut out using EcoRI-NotI and replaced by a FLAG tag using annealed 5′-AATTCCATGGATTACAAGGATGACGACGATAAGGC-3′ and 5′-GGCCGCCTTATCGTCGTCATCCTTGTAATCCATGG-3′ to generate the pMET7-FLAG-RNF41 vector. pMET7-FLAG-Rab4A, pMET7-FLAG-USP8, pMET7-Etag-USP8 and pMG1-USP8 prey constructs were generated via an LR reaction (Invitrogen) to transfer the Rab4A or USP8 gene from a Gateway entry clone of the human ORFeome v5.1 collection to a pMET7-FLAG, pMET7-Etag or previously described pMG1 prey destination vector (Lievens et al., 2009). The pMET7-FLAG and pMET7-Etag destination vectors were generated by amplifying the ccdB gateway cassette by PCR using 5′-CCCCAATTGACAAGTTTGTACAAAAAAGC-3′ or 5′-CCCCCGCGGGCACAAGTTTGTACAAAAAAGC-3′, respectively and 5′-GGGTCTAGATCAAACCACTTTGTACAAG-3′ and the pMG1 prey destination vector as a template. MfeI-XbaI digest of the first PCR fragment was ligated into an EcoRI-XbaI opened pMG1-SVT vector (Eyckerman et al., 2001) with EcoRI site cloned C-terminal of the FLAG tag using site-directed mutagenesis with primers 5′-GATGACGACGATAAGGAATTCTCGACCGTGGTAC-3′ and 5′-GTACCACGGTCGAGAATTCCTTATCGTCGTCATC-3′. SacII-XbaI digest of the second PCR fragment was ligated into a SacII-XbaI opened pMET7-Etag-Cis vector (Lavens et al., 2006). Human RNF41 was cloned from the pMET7-Etag-hRNF41 vector into a pGEX-4T-2 vector using NotI-XbaI, downstream of glutathione S-transferase, generating the GST-RNF41 vector. An XbaI site was generated in pGEX-4T-2 by ligation of annealed 5′-CAGTCTCTAGACAC-3′ and 5′-GTGTCTAGAGACT-3′. RNF41 AE, derived from the RNF41 AE bait, was ligated into the pGEX-4T-2 vector using SalI-NotI, generating the GST-RNF41 AE vector. To generate pcDNA3.1-FLAG-USP8, USP8 was cloned from the pMG1-USP8 prey vector with a NotI site inserted by site-directed mutagenesis using 5′-ACAAAAAAGGCGGCCGCATGCCTGC-3′ and 5′-GCAGGCATGCGGCCGCCTTTTTTGT-3′, into the pCDNA3.1 vector using NotI-XmaI. PGEX-4T-PGC1a vector encoding GST-PGC1a and pcDNA3-HA-ubiquitin were a kind gift from Dr M. Stallcup (University of Southern California, USA) and Rudy Beyaert (BCCM plasmid collection), respectively. HEK293T cDNA was used to amplify Rab5A using 5′-GCGGAATTCATGGCTAGTCGAGGCGCAA CAAGACCC-3′ (with EcoRI site) and 5′-CGCTCTAGATTAGTTACTACAACACTGATTCCTGG-3′ (with XbaI site); Rab7A using 5′-GCGAGATCTACCTCTAGGAAGAAAGTGTTGC-3′ (with BglII site) and 5′-CGCTCTAGATCAGCAACTGCAGCTTTCTGCCGAGGCCTTGG-3′ (with XbaI site); Rab11A using 5′-GATTATGCTGCGGCCGCCATGGGCACCCGCGACG-3′ (with NotI site) and 5′-GTTGTCTCTAGATTAGATGTTCTGACAGCACTGCAC-3′ (with XbaI site). The PCR products were ligated after a FLAG (for Rab5A and Rab7A) or HA (for Rab11A) sequence in a pMET7 vector, opened with the indicated restriction enzymes. Site-directed mutagenesis was used to generate the following mutants: pMET7-FLAG-Rab5 S34N with 5′-TGGGAGAGTCCGCGGTTGGCAAAAACAGCCTAGTGCTTCG-3′ and 5′-CGAAGCACTAGGCTGTTTTTGCCAACCGCGGACTCTCCCA-3′; pMET7-FLAG-Rab7 T22N with 5′-TCTGGAGTCGGGAAGAATTCACTCATGAAC-3′ and 5′-GAGTGAATTCTTCCCGACTCCAGAATCTCC-3′; pMET7-FLAG-Rab4 S22N with 5′-TGCAGGAACTGGGAAAAATTGCTTACTTCATCAGTTTATT-3′ and 5′-AATAAACTGATGAAGTAAGCAATTTTTCCCAGTTCCTGCA-3′; pMET7-HA-Rab11 S25N with 5′-GATTCTGGTGTTGGAAAGAATAACCTGCTGTCTGATTTACTG-3′ and 5′-CGAGTAAATCGAGACAGCAGGTTATTCTTTCCAACACCAGAATC-3′; pMET7-RNF41 AE(-Etag), pMET7-FLAG-RNF41 AE and RNF41 AE bait with RNF41 Q266A R269E mutation using 5′-GCCACACTAGAGACTCGAGCGATGAACGAACGCTACTATGAGAAC-3′ and 5′-GTTCTCATAGTAGCGTTCGTTCATCGCTCGAGTCTCTAGTGTGGC-3′; pMET7-RNF41 SQ-Etag with RNF41 C34S H36Q mutation using 5′-GTACAGGCACCTCATAGTGAACAAGCTTTCTGCAACGCC-3′ and 5′-GGCGTTGCAGAAAGCTTGTTCACTATGAGGTGCCTGTAC-3′; pMET7-FLAG-USP8 C786S with 5′-TAACTTAGGAAATACTAGTTATATG-3′ and 5′-ATTGAGTTCATATAACTAGTATTTC-3′. pMET7-solIL-5Rα was generated via ligation of EcoRI-XhoI cut soluble IL-5Rα from the λgt11-hIL5Rα12 vector provided by Roche, into an empty pMET7 vector. All constructs were verified by DNA sequence analysis.

HEK293T and HeLa cells were cultured in an 8% CO₂ humidified atmosphere at 37°C and grown in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen) with 10% fetal calf serum (Perbio). For protein expression analysis, 1.5×10⁵ HEK293T cells were seeded in a 12-well and transfected using calcium phosphate on the next day with 500 ng of the indicated RNF41/USP8 wild-type or mutant constructs on a total of 2 µg DNA transfected. On the next day, cells were washed and 24 hours post-transfection serum-deprived overnight prior to lysis. For MAPPIT analysis, 10,000 HEK293T cells were seeded in a 96-well and transfected on the next day with 50 ng of pXP2d2-rPAP1-luciferase reporter, 250 ng bait and 250 ng prey constructs using calcium phosphate. For AlphaScreen analysis, 3×10⁵ HEK293T cells were seeded in a 6-well and transfected on the next day with 1 µg of both E- and FLAG-tagged proteins using calcium phosphate. For ubiquitylation assays, 2.4×10⁶ HEK293T cells were seeded in a 60 mm Petri dish and transfected with 3 µg pMET7-FLAG-USP8 with or without 3 µg pcDNA3-HA-ubiquitin and RNF41-encoding plasmid DNA on the next day using calcium phosphate. For CTS cleavage and enzyme-linked immunosorbent assay (ELISA) assays, 2×10⁵ HEK293T cells were seeded in 12-wells and transfected on the next day with 1 µg of receptor-encoding plasmid with either 1 µg of mock/RNF41/USP8/Rab-encoding vectors or 500 ng mock/RNF41-expressing plasmid together with 500 ng mock/Rab-expressing plasmid using the calcium phosphate method. One day post-transfection, cells were left untreated (DMSO) or treated overnight with 25 µM chloroquine in serum-free OPTIMEM medium (Invitrogen). For confocal imaging, 2×10⁵ HeLa cells were seeded on no. 1.5 glass coverslips (Zeiss) in a 6-well coated with poly-D-lysine (Sigma-Aldrich). The next day, cells were transfected with JetPrime (Polyplus) according to the manufacturer's guidelines with 500 ng of pMET7-Etag-USP8 plasmid with or without 500 ng of pMET7-hRNF41 vector. In all experiments, the pMET7-solIL-5Rα construct was used to normalize for the amount of transfected DNA and load of the transcriptional and translational machinery.

Cells were washed with PBS, lysed in 2×SDS gel loading buffer (62.5 mM Tris-HCl pH 6.8, 3% SDS, 10% glycerol, 5% β-mercaptoethanol and 0.01% Bromophenol Blue sodium salt sonicated) and sonicated using the Bioruptor Plus (Diagenode). After boiling, cell lysates were resolved by SDS-PAGE and transferred to nitrocellulose membranes (Amersham Biosciences). Blotting efficiency was checked by Ponceau S staining (Sigma). Blots were blocked in StartingBlock blocking buffer (Pierce), when using Odyssey infrared imaging (LI-COR) or in 5% milk upon enhanced chemiluminescence (ECL) detection. β-actin, HA-tagged, FLAG-tagged and E-tagged proteins were revealed using a rabbit or mouse anti-β-actin (1:5000, Sigma), monoclonal rat anti-HA (3F10) (1:5000, Roche), rabbit anti-FLAG (1:5000, Sigma) and mouse anti-E-tag (1:10,000, Phadia) antibody, respectively, followed by an anti-rabbit or anti-mouse DyLight 800- or DyLight 680-conjugated antibody (1:15,000, Pierce) or an anti-rat Alexa-Fluor-680-conjugated antibody (1:5000, Molecular Probes), diluted in Odyssey blocking buffer (Li-Cor) + 0.1% Tween 20. Rabbit anti-USP8 A302-929A (1:5000, Bethyl), rabbit anti-RNF41 A300-048A (1:20,000, Bethyl), rabbit anti-STAM1 H-175 (1:1000, Santa Cruz), rabbit anti-STAM2 HPA035529 (1:2000, Sigma), mouse anti-TSG101 4A10 (1:1000, Abcam) or rabbit anti-Hrs kindly provided by Harald Stenmark (Raiborg et al., 2001) were revealed by SuperSignal West Pico Chemiluminescent Substrate (Pierce) using peroxidase-conjugated anti-mouse or anti-rabbit antibodies (1:5000, Jackson ImmunoResearch).

After overnight transfection, cells were left untreated or stimulated for 24 hours with human Epo (5 ng/ml). Luciferase activity from triplicate samples was measured by chemiluminescence in a TopCount luminometer (PerkinElmer) and expressed as fold induction (stimulated/non-stimulated relative light units) relative to the signal generated by a JAK2 binding prey, which corrects for possibly varying expression levels of the different baits used.

Alpha screen experiments were conducted according to the manufacturer's protocol (PerkinElmer). Typically, cells were co-transfected overnight with the appropriate E- and FLAG-tagged expression vectors, washed on the next day and lysed another 24 hours later in TAP lysis buffer (50 mM Tris–HCl pH 7.5, 125 mM NaCl, 5% glycerol, 0.2% NP40, 1.5 mM MgCl₂, 25 mM NaF, 1 mM Na₃VO₄, Complete™ Protease Inhibitor without EDTA Cocktail from Roche). Lysates were cleared by centrifugation and incubated for 1 hour at 4°C with 0.7 µg/ml anti-Etag antibody, biotinylated with sulfo-NHS-Biotin (Pierce) according to the manufacturer's guidelines. Anti-FLAG M2 acceptor beads and Streptavidin donor beads were subsequently added (20 ng/ml each) for another incubation time of 1 hour (at 4°C) and 30 minutes (at room temperature), respectively. Samples were measured in triplicate using the EnVision platereader (PerkinElmer). Part of the cellysate was simultaneously analyzed for protein expression using western blotting.

One day post-transfection, cells were washed and treated overnight with 25 µM chloroquine and 5 µM MG132 in serum-free OPTIMEM medium (Invitrogen). Cells were washed with PBS and lysed in 250 µl of SDS lysis buffer (2% SDS, 150 mM NaCl, 10 mM Tris-HCl, pH 8.0, 2 mM sodium orthovanadate, 50 mM sodium fluoride, 10 mM N-ethylmaleimide and Complete™ Protease Inhibitor without EDTA Cocktail from Roche). Lysates were sonicated using the Bioruptor Plus (Diagenode), boiled for 10 minutes and diluted by adding 2250 µl dilution buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100) and 30–60 minutes of incubation at 4°C under rotation. Lysates were cleared by 30 minutes centrifugation at 20,000 g and the supernatants were precleared by 1 hour incubation with Sepharose 4B beads, rotating at 4°C. After overnight, rotating incubation with anti-FLAG M2 Sepharose beads (Sigma) at 4°C, beads were washed 3× with 1 ml of wash buffer (10 mM Tris-HCl, pH 8.0, 1 M NaCl, 1 mM EDTA, 1% NP-40) and eluted in 2× SDS gel loading buffer. Samples were subjected to 7.5% SDS-PAGE and transferred onto nitrocellulose membranes. Ubiquitylation was assayed by western blotting against HA–ubiquitin.

E. coli strain BL21(DE3) was transformed with vectors containing GST-RNF41, GST-RNF41 AE or GST-PGC1a. Protein production was induced by 0.2 mM isopropyl-D-thiogalactoside at A600 of 0.6. The bacteria were further cultured overnight at 25°C. After centrifugation at 5000 r.p.m. for 10 minutes, the bacterial pellet was resuspended and sonicated in NETN buffer (20 mM Tris-HCl pH 8; 100 mM NaCl; 6 mM MgCl₂; 1 mM EDTA; 0.5% NP40; 1% DTT and Complete™ Protease Inhibitor Cocktail from Roche). The E. coli lysate was centrifuged at 12,000 r.p.m. for 10 minutes to obtain the soluble GST proteins in the supernatant. GST proteins were immobilized on Glutathione Sepharose 4B beads (GE Healthcare) for 1 hour at 4°C and washed three times with NETN buffer. FLAG-tagged USP8, produced using the TNT T7 Quick Coupled Transcription/Translation System (Promega), was added overnight and the beads are washed three times with NETN buffer. FLAG-tagged proteins bound to the GST-proteins were eluted by 10 minutes boiling in 2× SDS gel loading buffer and analyzed by SDS-PAGE and western blotting.

24 hours post-transfection, cells were cultured overnight in OPTIMEM medium (Invitrogen). Next day, cellular supernatant was collected and soluble LR or LIFR levels were determined using an in-house developed leptin-SEAP based ELISA method or the human sLIF-R/gp190 ELISA kit (BioVendor). Cells were simultaneously lysed for C-terminal stub (CTS) cleavage analysis using western blotting.

Transient knockdown was accomplished using USP8 siRNA1 siGENOME D-005203-02 (Dharmacon) or USP8 siRNA2 siGENOME D-005203-03 (Dharmacon) or Renilla luciferase siRNA P-002070-01-20 as a negative control (Dharmacon). 1.25×10⁵ HEK293T cells in a 12-well were revere transfected with 50 µM siRNA using Dharmafect1 (Dharmacon). After 24 hours, cells were transfected with a control plasmid or plasmids encoding the LR or LIFR using calcium phosphate. 72 hours following siRNA transfection cells were incubated overnight in serum-free OPTIMEM medium with or without 25 µM chloroquine. Next morning, cell supernatant was collected to detect soluble receptors by ELISA and cells were simultaneously lysed for C-terminal stub (CTS) cleavage analysis and evaluation of silencing efficiency using western blotting.

24 hours post-transfection, cells were rinsed with PBS and fixed for 15 minutes at room temperature in 4% paraformaldehyde. After three washes with PBS, cells were permeabilized with 0.4% Triton X-100 in PBS for 5 minutes and blocked in PBS with 1% BSA and whole donkey serum for 1 hour. Samples were further incubated for 1 hour at room temperature with goat anti-Etag A190-132A (1:4000, Bethyl) and mouse anti-RNF41 A-6 (1:500, Santa Cruz) antibodies. After four washes in PBS, cells were incubated for 1 hour at room temperature with donkey anti-goat Alexa Fluor 488 and donkey anti-mouse Alexa Fluor 647 secondary antibodies. Images were acquired using a 60× 1.35 NA objective on an Olympus IX-81 laser scanning confocal microscope and analyzed using Fluoview 1000 software.

We thank Professor Harald Stenmark (Center for Cancer Biomedicine Oslo, Norway), Professor Michael R. Stallcup (University of Southern California, USA) and Professor Rudy Beyaert (BCCM/LMBP Plasmid Collection, Belgium) for kindly providing the anti-Hrs antibody, the GST-PGC1a plasmid and the pcDNA3-HA-ubiquitin plasmid, respectively.