DCAF8, a novel MuRF1 interaction partner, promotes muscle atrophy

Muscle tissue represents the largest protein reservoir in the body that can be utilized as a source of amino acids for energy production during starvation. This so-called muscle atrophy involves the removal of contractile proteins like several different myosin heavy chain proteins (MyHC) by proteolytic systems, which in turn causes the reduction of the size of muscle fibers. The molecular details of muscle atrophy are of immanent medical importance because excessive loss of muscle mass counteracts the therapeutic treatment of cancer and other malignancies, aggravates the progression of most chronic diseases, prolongs the recovery phase after surgery, and increases overall morbidity and mortality. Proteolysis in muscles is mainly mediated by the ubiquitin-proteasome system (UPS). The UPS constitutes a cascade of E1 (ubiquitin-activating), E2 (ubiquitin-conjugating) and E3 (ubiquitin ligase) enzymes to attach ubiquitin onto substrate proteins, which initiates their degradation by 26S proteasomes. E3 ligases ensure the specificity of the UPS by selecting appropriate substrates and promoting the transfer of ubiquitin onto them. These components either act as individual enzymes or are organized in multi-subunit protein complexes (Hershko and Ciechanover, 1998; Buetow and Huang, 2016).

The ‘muscle-specific RING-finger’ proteins MuRF1, MuRF2 and MuRF3 (also known as TRIM63, TRIM55 and TRIM54, respectively) have been identified as important factors for the maintenance of protein homeostasis in muscle cells (Spencer et al., 2000; Foletta et al., 2011). However, the molecular function of MuRF1 is poorly understood and even less is known about MuRF2 and MuRF3. MuRF1 expression is highly upregulated during atrophy (Bodine et al., 2001), where it was reported to act as an autonomous ubiquitin ligase for the removal of proteins such as MyHC and actin (Clarke et al., 2007; Fielitz et al., 2007a; Cohen et al., 2009). Consistent with this, deletion of the MuRF1-encoding gene Trim63 in mice or the reduction of MuRF1 levels in muscle cells through siRNA treatment inhibits muscle atrophy (Cohen et al., 2009; Castillero et al., 2013). MuRF1 was also found to localize to the sarcomere, to associate with titin and to stabilize the sarcomeric M-line (McElhinny et al., 2002; Gotthardt et al., 2003). Moreover, MuRF1 appears to regulate muscular energy metabolism by targeting creatine kinase (Hirner et al., 2008; Koyama et al., 2008). The large diversity of the published MuRF-associated activities suggests that these proteins team up with different co-factors, which integrate them into distinct cellular processes.

Through genetic and biochemical techniques we identified DDB1 and cullin-associated factor 8 (DCAF8), also termed WD repeat (WDR) containing protein 42A (WDR42A), as a new MuRF1 interaction partner. WDR proteins commonly serve as substrate receptors for cullin RING ligases (CRLs). These protein complexes represent modular E3 enzymes that are built around one of the seven cullin scaffolding proteins. Through their C-terminus, the cullins recruit specific RING-finger proteins involved in ubiquitin conjugation. Additional adaptor and receptor proteins that determine the specificity of the ubiquitylating reaction bind to their N-terminus. CRL4 complexes, for instance, encompass an appointed combination of the cullins Cul4A or Cul4B, the RING-finger proteins RBX1 or RBX2, DNA damage-binding protein 1 (DDB1), and an accessory DDB1- and CUL4-associated factor (DCAF) (Jackson and Xiong, 2009). It is noteworthy that CRL complexes seem to primarily be involved in the control of gene expression and in the regulation of the cell cycle (Petroski and Deshaies, 2005; Sarikas et al., 2011) and to contribute to lysosomal protein degradation (Lee et al., 2017) rather than to directly promote the proteasomal degradation of structural proteins like MyHC.

We demonstrate here that MuRF1 associates with a CRL4A multi-protein complex by binding to DDB1 and DCAF8. Similar to MuRF1, DCAF8 is strongly upregulated during denervation-induced skeletal muscle atrophy in mice. Moreover, we show that C2C12 skeletal muscle cells lacking DCAF8 resist glucocorticoid-induced myotube atrophy and that DCAF8 and MuRF1 directly participate in the UPS-dependent MyHC degradation. Since chemical inhibition of CRL activity impedes atrophy in our model system, we propose that a CRL4A ubiquitin ligase complex containing DCAF8 contributes to the proteasomal degradation of sarcomeric proteins like MyHC and thereby promotes muscle atrophy.

The multitude of the reported MuRF functions suggests that these proteins act in combination with yet unknown co-factors, which mediate the role of these RING finger proteins in diverse cellular functions. We thus performed two proteome-wide high-throughput (HT) screens to identify such binding partners. To this end, we employed automated unbiased yeast two-hybrid (Y2H) screens using a ∼23,000 single-clone cDNA expression library as a prey and non-auto activating MuRF wild-type and truncation cDNA constructs as baits (Fig. 1A; Fig. S1, Table S1) (Fields and Song, 1989; Stelzl et al., 2005). In a second screen, we used stable isotope labeling by amino acids in cell culture (SILAC) and affinity purification (AP) of individual MuRF constructs, and analyzed the precipitates by mass spectrometry (MS). We transduced cDNAs encoding Myc(His)₆-tagged MuRF1, MuRF2, or MuRF3 by adenoviral vector (AdV) transfer into ‘heavy’ isotope-labeled H9c2 myocytes (lysine-8, arginine-10) and isolated the corresponding MuRF proteins by metal affinity purification. As a reference, we used ‘light’ isotope-labeled H9c2 myocytes (lysine-0, arginine-0) containing an empty vector [promoterless adenovirus (AdV-PL)] in the analysis (Fig. 1A, Fig. S2; Table S2).

The outputs of multiple screening rounds were scored (see Materials and Methods section) resulting in several putative MuRF-binding proteins (Tables S3–S12). Y2H employs plasmid-driven protein expression of bait (MuRF1–MuRF3) and prey (∼23,000 single-clone cDNA expression library, see above) constructs in the heterologous organism S. cerevisiae. Positive hits in this screening method do not necessarily represent partner proteins, since the candidates may not even be expressed in the same tissue in mammals. Similarly, the SILAC-AP-MS approach employs overexpressed and epitope-tagged MuRF proteins and this may result in the identification of factors that unspecifically and only weakly interact with the unnatural high amounts of the bait. Thus, we first validated our findings by comparing the hits of both experiments. In the end, 43 candidate proteins were considered to be enriched in both the SILAC-MS and the Y2H assays (Fig. 1A; Table S13). Importantly, the identification of already-known MuRF substrates and interacting proteins underscored the reliability of both screening approaches. Among others, we isolated the sarcomeric proteins nebulin and FHL-1, which have been shown to associate with the MuRF proteins, the MuRF substrates titin-cap/telethonin and troponin, and proteins linked to the UPS and autophagy such as UBE2I, USP13, UCHL3, ubiquitin and Sqstm-1 (also known as p62) (McElhinny et al., 2002; Kedar et al., 2004; Lange, 2005; Witt et al., 2005; Fielitz et al., 2007b; Hirner et al., 2008; Loch and Strickler, 2012; Woodsmith et al., 2012).

We then validated the results from the HT screens further in individual immunoprecipitation (IP) reactions (selected validated interactors are summarized in Fig. 1B). FLAG-tagged candidate proteins were co-expressed with Myc(His)₆-tagged MuRF proteins in COS-7 cells under the control of the same promoter to yield near-stoichiometric cellular protein levels, immunopurified from cell extracts, and the precipitates were analyzed by immunoblotting with anti-Myc and anti-FLAG antibodies. Importantly, we were able to reproduce published data for the binding of established MuRF partner proteins in these assays (Fig. 1C,D; Fig. S3). For instance, the selective autophagy receptor protein Sqstm1 was shown to interact with MuRF1 (Khan et al., 2014; Troncoso et al., 2014) and indeed the FLAG-tagged isoform 1 of Sqstm1 efficiently precipitated MuRF1–Myc(His)₆ in our experiments (Fig. 1C). Interestingly, MuRF1 was not purified with Sqstm1 isoform 2, which lacks 84 amino acids at its N-terminus, implying that the N-terminal region of isoform 1 confers binding to MuRF1 (Fig. 1C). Similarly, USP13 was described to specifically bind MuRF1 and MuRF2, but not MuRF3 (McElhinny et al., 2002), which matches the results of our IP experiments (Fig. S3A,D,E).

We then investigated binding of new putative MuRF interaction partners in the IP assay. For some of the high scoring hits of the HT screens, like the AAA-ATPase p97 (also known as VCP), we were not able to confirm a physical interaction with the corresponding MuRF protein in these experiments (Fig. S3A,D; Table S14). However, FLAG-tagged variants of 14 of the tested HT candidates precipitated at least one MuRF construct (Fig. 1C,D; Fig. S3, Table S14). Surprisingly, only one of these, the myosin light chain kinase 2 (MYLK2) has been reported to be preferentially expressed in muscle tissue (UniProt database, entry Q9H1R3). MYLK2–FLAG precipitated all three MuRF–Myc variants when co-expressed in COS-7 cells, but the biological significance of this observation was not pursued further (Fig. 1D).

Remarkably, we observed binding of several components of the UPS, such as the ubiquitin conjugating enzyme UBE2E3, the ubiquitin specific proteases USP13 and USP35, the F-Box and WD repeat containing protein FBXW2, the regulator of cullin-type ubiquitin ligases NEDD8, and the WD repeat containing protein DCAF8 to at least one MuRF protein (Fig. 1C; Fig. S3, Table S14). Of these, DCAF8 caught our particular attention. In the IP experiments, DCAF8–FLAG specifically precipitated MuRF1 and MuRF3 (Fig. 1D). Importantly, DCAF8 was previously linked to the regulation of muscle mass since mutations in the coding gene that impair the interaction of the protein with DDB1 are associated with the development of axonal hereditary motor and sensory neuropathy (HMSN2), which leads to a progressive muscular atrophy of the extremities (Klein et al., 2014). DCAF8 also contains so-called WD repeat domains and has been reported to constitute a bona fide substrate recruiting factor of cullin RING ligase 4- (CRL4-)type ubiquitin ligase complexes (Jin et al., 2006). Aside from the cullin regulator NEDD8, DCAF8 was therefore another CRL4 component that was enriched in our screens (Fig. 1B).

We subsequently investigated whether DCAF8 constitutes a functional interaction partner of the MuRF proteins or represents a substrate of these putative ubiquitin ligases. Thus, we determined the half-life of DCAF8 and the MuRF proteins in transiently transfected COS-7 cells. In short, cells were treated with cycloheximide (CHX) to block the synthesis of new polypeptides and time-dependent changes of the amount of selected proteins were recorded by immunoblotting. In these assays all MuRF–Myc(His)₆ proteins and FLAG-tagged DCAF8 were stable for 4 h (Fig. 2A). FLAG-tagged DCAF8 that was expressed with any of the MuRF–Myc(His)₆ proteins displayed no significant change in stability. Likewise, all MuRF proteins were stable when DCAF8 was co-expressed indicating that DCAF8 does not initiate their turnover and vice versa (Fig. 2B). In line with these findings the half-life of endogenous DCAF8 in COS-7 cells was also unaffected by increased MuRF1 levels (Fig. 2C). Hence, DCAF8 seems not to be degraded in a MuRF1-dependent manner but rather represents a stable binding partner of this RING-finger protein.

As MuRF1 solely resides in striated muscle tissue, such as skeletal muscle and the heart (Bodine et al., 2001), we were interested in the expression pattern of DCAF8. A polyclonal DCAF8 antibody reacted with proteins displaying an apparent molecular mass of ∼70 kDa in immunoblots of lysates from of COS-7 cells (Fig. 2C) and C2C12 myoblasts (Fig. 3A), which roughly matches the calculated molecular mass of the DCAF8 protein of ∼67 kDa. Interestingly, in C2C12 cells, another DCAF8 species was observed at 80 kDa, which appeared to be more prevalent and may represent phosphorylated DCAF8 (five putative phosphorylation sites are annotated for DCAF8 in the UniProt database, entry Q8N7N5). Both signals were not detected in cells treated with DCAF8-specific siRNAs. Using this antibody, we observed, in immunoblots of mouse tissue samples, comparable levels of DCAF8 in the Gastrocnemius plantaris (GP) and Tibialis anterior (TA) skeletal muscle, in the heart and in other organs (Fig. 3B). Analogous to what was found for the C2C12 cell lysate, DCAF8 was detected as a doublet with molecular masses of 70 and 80 kDa in the majority of these tissues, but their relative abundance varied. Whether these signals and other species, of 35 kDa and 50 kDa, respectively, recorded in pancreas, liver and kidney lysates derive from posttranslational modifications, or limited proteolysis remains to be determined. Notably, the UniProt database lists several entries for truncated human, mouse and even claw frog DCAF8. Thus, the polyclonal antibody may react with different and tissue-specific isoforms of DCAF8.

We next analyzed the intracellular localization of MuRF1 and DCAF8. As determined by fluorescence microscopy, RFP and GFP fusion proteins of DCAF8 and MuRF1, respectively, resided as granules within the cytosol of COS-7 cells (Fig. 3C). Importantly, the majority of both constructs stained overlapping structures in these cells. Likewise, immunostaining of fixed undifferentiated and differentiated C2C12 cells with specific anti-MuRF1 or anti-DCAF8 antibodies (see below) revealed that endogenous DCAF8 and MuRF1 were excluded from the nucleus and localized in the cytosol (Fig. 3D). Still, we noticed that in COS-7 and in C2C12 cells a fraction of the MuRF1 staining did not overlap with the DCAF8 signal and vice versa, which suggests that both proteins also associate with other partners.

DCAF8 has been proposed to serve as substrate receptor in CRL4 complexes in non-muscle cells (Jin et al., 2006; Li et al., 2017). Such modular E3 ligases usually contain a Cul4A or Cul4B cullin scaffolding protein, which recruits the RING-finger protein RBX1 or RBX2, the adaptor protein DDB1, and different DCAF proteins involved in substrate binding (Angers et al., 2006). To test whether DCAF8 bound to MuRF1 was still part of such a CRL4 ubiquitin ligase or whether MuRF1 binding sequesters DCAF8 from such cullin complexes, we performed FLAG-IP experiments from COS-7 cells that transiently overexpressed combinations of epitope-tagged variants of MuRF1, DCAF8 and DDB1. As expected from our previous experiments, we purified MuRF1–Myc(His)₆ with DCAF8–FLAG, indicative of a physical interaction (Fig. 4A). Of note, cells transfected with constructs for the expression of epitope-tagged MuRF1 contained higher amounts of this protein than ones harboring additional plasmids encoding DCAF8–FLAG or other proteins (Fig. 4D,E). We made a similar observation when we analyzed cells transfected for the overexpression of epitope-tagged DCAF8 (Fig. 4C) or DDB1 (Fig. 4E). For the following reasons, this effect most likely originates from the transfection procedure and does not implicate enhanced degradation of MuRF1, DCAF8 and/or DDB1. First, we noticed a decrease in the content of any epitope-tagged protein in cells that were co-transfected with another construct irrespective of the encoded gene (Fig. 4D,E). Second, we observed that the overexpression of DCAF8 did not change the stability of the MuRF1 protein in COS-7 cells (Fig. 2B) nor did overexpression of MuRF1 affect the amount of endogenous DCAF8 in COS-7 cells (Fig. 2C). Finally, a stable knockout of DCAF8 in C2C12 myocytes did not contain increased protein levels of DDB1 or DCAF8 (see below, Fig. 5C). We also detected DCAF8–FLAG in precipitates of DDB1–HA and retrieved DDB1–FLAG upon purification of DCAF8–Myc(His)₆ (Fig. 4B,C), which is in agreement with published data (Li et al., 2017). Interestingly, we found MuRF1–Myc(His)₆ in precipitates of DDB1–FLAG (Fig. 4D). Since we also co-purified epitope-tagged MuRF1 and DDB1 with DCAF8–FLAG, our findings suggest that these three proteins form a complex (Fig. 4E). To analyze this idea in more detail, we downregulated DCAF8 with siRNA in COS-7 cells and isolated DDB1–FLAG. Even in absence of DCAF8 we co-precipitated MuRF1–Myc(His)₆ with DDB1–FLAG (Fig. 4F).

Next, we employed antibodies specific for DCAF8 or DDB1 in IPs from C2C12 myotube lysates to confirm complex formation of the endogenous proteins. Intriguingly, in precipitates using a polyclonal anti-DCAF8 antibody we found DDB1, Cul4A, RBX1 and, most importantly, also MuRF1 (Fig. 5A). We conducted a similar experiment using a polyclonal anti-DDB1 antibody. Again, we co-precipitated the components of a CRL4A complex, Cul4A, DCAF8, RBX1 and MuRF1, along with DDB1 from C2C12 myotube lysates (Fig. 5B). These results indicate that DCAF8 is part of a cullin-type ligase in myocytes, even when bound to MuRF1. Given that DCAF proteins associate with CRL4 complexes via binding to DDB1 we wondered, whether MuRF1 also interacts with this adaptor protein. Therefore, we established a stable C2C12 cell line lacking DCAF8 using CRISPR/Cas9 (Fig. S4C). Strikingly, even in absence of endogenous DCAF8, we still detected MuRF1 in precipitates of DDB1 (Fig. 5C), which agrees with the knockdown co-IP results in COS-7 cells (Fig. 4F). Taken together, these results imply that MuRF1 interacts with CRL4A complexes in myocytes via direct binding to DDB1 and DCAF8.

So far, we have been able to show a physical interaction of MuRF1 with DCAF8 and a CRL4A-type ubiquitin ligase, but up to now a specific function of such a complex in myocytes has not been described. MuRF protein levels change substantially during development and myogenic differentiation of C212 cells (Centner et al., 2001; Perera et al., 2012) and we wanted to determine whether the cellular content of its interaction partner(s) are regulated in a similar manner. Thus, we differentiated C2C12 myoblasts into myotubes by reducing the amount of serum in the medium for 4 days and monitored the levels of Cul4A, RBX1, DDB1, DCAF8 and MuRF1 by immunoblotting (Fig. 6A,B). Throughout differentiation we observed a moderate increase in the cellular content of DDB1 and a more pronounced rise in the amounts of Cul4A and DCAF8. Intriguingly, we also detected a substantial augmentation of MuRF1 and RBX1 protein levels. This finding correlates with published results on MuRF1 expression (Centner et al., 2001; Perera et al., 2012) and with the reported involvement of CRL-type ligase complexes, which contain RBX1, in myocyte differentiation (Blondelle et al., 2017).

Because the amount of MuRF1 strongly increases in atrophic muscle (Bodine et al., 2001; Foletta et al., 2011), we investigated whether DCAF8 expression is also upregulated during this process. We induced muscle atrophy in mice by denervation of the sciatic nerve (Schmidt et al., 2014) and determined the amount of DCAF by immunoblotting. Compared to sham operated mice, denervation led to a significant decrease in the mass of TA and GP muscles after 7, 14 and 21 days of surgery (Schmidt et al., 2014). Strikingly, we observed a significant increase of DCAF8 protein content in de-nervated TA and GP muscle tissue compared to innervated muscles in three individual animals (Fig. 6C–F; Fig. S5). These results demonstrate that DCAF8 levels in muscle cells significantly increase during differentiation and atrophy, which correlates with increased expression of MuRF1 during atrophy.

MuRF1 was shown to facilitate the ubiquitylation of MyHC, a highly abundant muscle protein, which is degraded during atrophy (Fielitz et al., 2007a). The upregulation of DCAF8 during atrophy prompted us to investigate whether this protein is also involved in MyHC proteolysis. To this end, we employed COS-7 cells, which endogenously express DCAF8 (see Fig. 2C) and the core components of CRL4A-type ligase complexes, and analyzed the turnover of transiently expressed FLAG-tagged MyHC (FLAG-MyH7) in CHX decay assays (see above). As a control we transfected a construct for the expression of FLAG-tagged Cas9, which is unrelated to the UPS. Importantly, the FLAG–MyHC protein level decreased by 30% after 30 min of CHX treatment in wild-type (WT) cells while it remained unchanged in COS-7 cells lacking DCAF8 (COS-7 ΔDCAF8; Fig. 7A,B; Table S15). Importantly, expression of DCAF8–FLAG in COS-7 ΔDCAF8 cells restored the turnover of co-transfected FLAG–MyHC (39.0±6.1% decrease, mean±s.e.m.; Fig. 7A,B; Table S15). Likewise, the expression of MuRF1-Myc(His)₆ in COS-7 ΔDCAF8 cells also induced FLAG–MyHC degradation to some extent (22.5±5.3% decrease after 30 min). These data imply that DCAF8 as well as MuRF1 contribute to the degradation of MyHC in COS-7 cells.

Subsequently, we recorded the steady state levels of FLAG–MyHC by immunoblotting in COS-7 cells that transiently co-expressed either FLAG-tagged Cas9 or MuRF1–Myc(His)₆. Compared to Cas9-transfected cells, we observed an 80% reduction of the FLAG–MyHC protein amount in COS-7 cells harboring the MuRF1 construct. Treatment of these cells with MLN4924, which prevents the neddylation of CRL-type ubiquitin ligases and thereby impedes their activity (Blondelle et al., 2017), or the proteasomal inhibitor lactacystin restored the FLAG–MyHC content to WT levels. This observation confirms an involvement of CRL-type ligases and the proteasome in the degradation of FLAG–MyHC (Fig. 7C,D).

In COS-7 ΔDCAF8 cells we found an ∼2-fold increase in the FLAG–MyHC protein content compared to COS-7 WT cells (Fig. 7C,D; Table S16). Expression of either MuRF1–Myc(His)₆ or DCAF8–FLAG in these cells restored the MyHC levels to that detected in COS-7 WT cells (Fig. 7C,D; Table S16). As expected, transfection of COS-7 ΔDCAF8 cells with both the DCAF8–FLAG and MuRF1–Myc(His)₆ constructs yielded a similar low MyHC protein content as in COS-7 WT transfected with MuRF1 (Fig. 7C,D). The addition of MLN4924 or lactacystin to these cells substantially increased the amount of FLAG–MyHC (Fig. 7C,D). Taken together, the experiments in the heterologous COS-7 cells show that MuRF1 and DCAF8 on their own are capable of facilitating the turnover of MyHC and that both proteins display additive effects in the degradation of MyHC. However, the functional relation between MuRF1 and DCAF8 in COS-7 cells remained unclear.

We therefore sought to validate our observations in C2C12 cells, which represent a more physiologically relevant system to study muscle-specific activities. To this end, we differentiated C2C12 cells that were downregulated for expression of individual genes into myotubes and induced atrophy using the glucocorticoid dexamethasone (Dexa). Dexa treatment serves as an established method to investigate muscle atrophy in cell culture but widely different concentrations of this substance have been used in individual studies (Hasselgren, 1999; Clarke et al., 2007; Menconi et al., 2008; Massaccesi et al., 2016). In initial experiments we exposed the C2C12 myotubes to 10 µM Dexa (defined as a ‘mild Dexa treatment’) and, as recently published, this treatment caused a decrease in MyHC protein levels in scrambled siRNA-treated control cells (Fig. 8A) (Massaccesi et al., 2016). In cells impeded for the expression of either MuRF1 or DCAF8 the loss of MyHC was less pronounced. The combined downregulation of MuRF1 and DCAF8 had a similar impact on the MyHC protein as the single knockdowns, implying that both proteins act in the same pathway for MyHC reduction. Because the atrophic response in the control cells was rather weak and the results in the single DCAF8 knockdown samples showed large deviations, we repeated this experiment with the C2C12 ΔDCAF8 cell line (Fig. S4) and treated the cells after differentiation with 100 µM Dexa (defined as a ‘strong Dexa treatment’) (Clarke et al., 2007). Again, we observed a substantial decrease in the amount of MyHC after Dexa treatment in WT myotubes but not in DCAF8-knockout cells (Fig. 8B). MyHC levels remained largely unchanged, when cells were exposed to Dexa and the neddylation inhibitor MLN4924 or the proteasomal inhibitor MG132 in this assay. Other than in the COS-7 experiments (see above), where MyHC turnover was already strongly induced in presence of either DCAF8 or MuRF1 alone, the results from C2C12 myocytes imply that the degradation of MyHC during atrophy relies on MuRF1 as well as on DCAF8, and involves the activity of CRL-type E3 ubiquitin ligase complexes and the proteasome.

Given the upregulation of DCAF8 expression in atrophic muscle and its involvement in MyHC degradation we were interested whether DCAF8 directly contributes to muscle atrophy. An established method to evaluate muscle atrophy is the quantification of changes in the diameter of differentiated C2C12 myotubes after Dexa treatment (Menconi et al., 2008; Son et al., 2015; Han et al., 2017). We downregulated DCAF8 and/or MuRF1 by siRNA in this assay to evaluate their impact in this process. Mild Dexa treatment caused a significant decrease in myotube diameter in C2C12 cells incubated with scrambled siRNA (Fig. 8C), which matches published data (Son et al., 2015; Han et al., 2017). As expected, knockdown of MuRF1 markedly impaired the Dexa-induced decrease of myotube size. Strikingly, the reduction of myotube diameter was also inhibited upon knockdown of DCAF8. In this assay, the combined downregulation of DCAF8 and MuRF1 prevented shrinkage of the myotubes to approximately the same extent as the single knockdowns (Fig. 8E). These findings support the notion that DCAF8 and MuRF1 form a functional unit to mediate muscle atrophy in cultured myotubes.

Both the dynamic adaptation of muscle mass to cellular stimuli and the quality control in muscle cells relies on the balance of the synthesis of sarcomeric proteins and their turnover by the ubiquitin proteasome system (UPS). Key components of this process are the MuRF1, MuRF2 and MuRF3 proteins, which have been implicated in the degradation of sarcomeric proteins like MyHC. In this work, we employed Y2H screens to identify a first set of novel putative MuRF interaction partners. This screening method is based on overexpressed bait and prey proteins in non-muscle cells. Thus, we sought to re-examine our results by SILAC-AP-MS of epitope-tagged overexpressed MuRF proteins, and identified 43 myocyte proteins that were also found in the Y2H approach. As the SILAC-AP-MS approach can also yield candidates that unspecifically bind to the overexpressed MuRF proteins, we confirmed our findings in co-IP experiments with transfected constructs in heterologous COS-7 cells and, in the case of DCAF8 and DDB1, also in co-IPs of endogenous proteins in C2C12 myocytes.

In summary, we report on the identification of cellular factors that specifically associate with MuRF1, MuRF2 and MuRF3. These newly identified binding partners may either represent substrates of the MuRF proteins or constitute co-factors that modulate their activity. MYLK2, a protein kinase that is required for the induction of muscle contraction, was among the most prominent hits in our analysis. Remarkably, MuRF1 was previously shown to target muscle-specific creatine kinase and thereby fine-tune muscle energy metabolism (Koyama et al., 2008). Our data now suggest that MYLK2 represents another factor for the control of muscle function that is regulated by MuRF-dependent proteolysis.

In our proteome-wide screens, we identified DCAF8, a member of the DDB1 and Cul4-associated Factor (DCAF) family, as a stable interaction partner of MuRF1 and MuRF3. Our data now show a direct involvement of DCAF8 in MyHC degradation. At a first glance, the functional implications of this finding remain unclear. In COS-7 cells, DCAF8 alone is able to target ectopically expressed MyHC for proteasomal degradation. Notably, these cells do not contain endogenous muscle myosin nor do they accomplish atrophy-like processes. The co-expression of MuRF1 in such cells induces the turnover of MyHC even in absence of DCAF8. Most likely, some of the MyHC is cleared from COS-7 cells via a DCAF8-dependent quality control system that probably targets the highly abundant ectopically expressed MyHC to prevent proteotoxic stress. The addition of MuRF1 to these cells may then mobilize an additional and more specific proteolytic pathway. By contrast, MuRF1 and DCAF8 seem to co-operate in MyHC degradation in C2C12 myotubes. These cells contain endogenous DCAF8 and MuRF1 and generally constitute a far more suitable system to study muscle biogenesis. Downregulation of DCAF8 or MuRF1 or the combined loss of both factors impairs the turnover of endogenous MyHC to a similar extent. This implies that these proteins act in the same pathway for MyHC degradation in myotubes. Importantly, DCAF8 and MuRF1 levels increase in response to atrophic stimuli in C2C12 cells, and the individual or combined loss of these proteins impedes Dexa-induced morphological transformations to a comparable extent. Our data therefore support a model for a collaborative activity of DCAF8 and MuRF1 in MyHC degradation during muscle atrophy.

DCAF proteins typically constitute substrate receptors for CRL4 ubiquitin ligase complexes. Indeed, a DCAF8-containing CRL4A complex was recently shown to ubiquitylate histone H3 in non-muscle cells, which is important for postnatal liver maturation (Li et al., 2017). Notably, CRL4A–DCAF8 complexes seem to be important factors for the development and/or homeostasis of neuronal cells. Mutations in the gene encoding DCAF8, which reduce its association with DDB1, were recently associated with the development of a HMSN2 variant. This heredity sensory neuropathy of the extremities results in progressive muscular atrophy (Klein et al., 2014). Although, in this disease, muscular atrophy is a secondary affect, this observation already indicated an important role of DCAF8 and therefore of CRL4-type ligase complexes in muscle biology. We now confirm that DCAF8 associates, via the adaptor DDB1, with a cullin-type ligase encompassing CUL4A and RBX1. Surprisingly, this complex also interacts with the RING-finger protein MuRF1.

In most cases, CRL-type ubiquitin ligases contain a single RING-finger protein, RBX1 or RBX2, which stimulates the transfer of ubiquitin from a ubiquitin conjugating enzyme to an acceptor site within a substrate (Petroski and Deshaies, 2005; Jackson and Xiong, 2009; Sarikas et al., 2011). However, a Cul4A complex harboring two of such proteins has already been reported: COP1, a RING-finger protein that was initially thought to constitute an autonomous ubiquitin ligase, turned out to be an integral component of a CRL4A ubiquitin ligase for the degradation of c-Jun (Wertz et al., 2004). Within this complex, RBX1 is required for substrate ubiquitylation, while COP1 apparently serves different functions like defining the target range of the ubiquitin ligase (Wertz et al., 2004). Our results now suggest that topologically similar CRL4A complexes encompassing two RING finger proteins are formed in myocytes.

What would be the role of the interaction of MuRF1 and a CRL4A ligase in myocytes? In analogy to COP1, the recruitment of CRL4A to MuRF1 could specify the selection of MuRF1 substrates. We noticed that MuRF1 binds to Sqstm1 isoform 1 but not to the isoform 2, which lacks 84 amino acids at its N-terminus (Fig. S6). In contrast, DCAF8 binds both Sqstm1 species. This implies that DCAF8 and MuRF1 interact with different regions in a client protein. Such divergent binding properties may expand the substrate range of MuRF1 and/or the CRL4A ligase. Indeed, MuRF1 and DCAF8 establish separate routes for the proteolysis of MyHC in the heterogeneous COS-7 cell system indicating that each of these proteins alone is capable of targeting proteins for degradation. Since MuRF1 directly binds to DDB1 and associates with the CRL4A complex even in the absence of DCAF8, this RING-finger protein may interact with several CRL4A ubiquitin ligases exposing distinct preferences for target proteins. Intriguingly, we observed that in some cellular areas the DCAF8 and MuRF1 staining did not overlap. Hence, a MuRF1–CRL4A complex in myocytes may dynamically recruit different DCAF substrate receptors to target a specific subset of sarcomeric proteins for degradation. Future mass spectrometric analysis of gene-edited myocytes will elucidate the formation of such MuRF–CRL4 complexes. Owing to the lack of appropriate antibodies we could not investigate whether MuRF1 also associates with other CRL4 components like additional DCAF proteins, Cul4B or RBX2.

Alternatively, the CRL4A complex may assist MuRF1 in the ubiquitylation of its targets. In some instances, the generation of poly-ubiquitin on substrate proteins occurs in discrete steps that each rely on the activity of distinct enzymes (Weber et al., 2016). The processing of appointed proteins during atrophy may require an initial ubiquitylating event, which then initiates the generation of the proteasomal targeting signal by a different ubiquitin ligase. MuRF1 has been reported to act as an autonomous E3-ligase (Fielitz et al., 2007a) and, to our knowledge, neither CRL4-type ligases nor DCAF8 have been previously suggested to associate with MuRF1. Thus, previous in vitro studies on MuRF1 substrate ubiquitylation only contained a minimal set of components, in particular the E1 and an E2 enzyme, MuRF1, ATP, ubiquitin and a substrate protein. The addition of purified CRL4A components to such reactions may reveal new insights into how the interplay of these components facilitate the degradation of sarcomeric proteins, for example, by accelerating their ubiquitylation and/or by forming distinct poly-ubiquitin signals.

However, resolving this interesting question requires the establishment of a powerful in vitro system with purified components to reconstitute the precise molecular mode of action of MuRF1 and CRL4A(–DCAF8) substrate ubiquitylation. Establishing this system surpasses the focus of the presented work due to the complexity of the involved machinery but will be pursued in future studies.

Aside from the now described CRL4A complex, other cullin-type ubiquitin ligases have been implicated in the regulation of muscle mass. The F-box proteins atrogin-1 and muscle ubiquitin ligase of SCF complex in atrophy-1 (MUSA1, also known as FBXO30) serve as substrate receptors of Skp1–cullin1–F-box (SCF) E3 ligase complexes (Bodine and Baehr, 2014). Similar to MuRF1 and DCAF8, the expression of these proteins is upregulated during muscle atrophy and downregulation of either factor in a model system affects this process. Notably, CRL1 ligases encompassing atrogin-1 or MUSA1 seem to control muscle waste by targeting transcription factors for degradation. In contrast, we describe here that CRL4A complexes directly promote the turnover of the sarcomeric protein MyHC and thereby facilitate atrophy.

Our work also revealed a stable association of MuRF3 with DCAF8. While we did not investigate the nature of this interaction in more detail, this observation suggests a functional relationship between a DCAF8-containing CRL4 complex and MuRF3, similar to our observations on DCAF8, CRL4A-type ligases and MuRF1. Again, such a complex may target a different protein pool from the MuRF1–DCAF8–CRL4A axis and represent an adaptation to fine-tune the complex proteolytic events during muscle atrophy. We also observed a physical interaction of MuRF3 and the F-box protein FBXW2, indicating that the individual MuRF proteins can direct also other CRL complexes to muscle-specific functions.

Further work is required to characterize the MuRF–CRL relationship in more detail, to determine the molecular function and the regulation of their components, and to identify their target proteins. Given that the MuRF proteins directly facilitate the turnover of sarcomeric proteins, research on these ubiquitin ligases is of medical importance, because it may reveal very specific entry points for the treatment of pathogenic muscle atrophy.

cDNA expression plasmids were generated using standard cloning techniques and their identity was verified by sequencing. Plasmids used in this study were: pGEM-T (Promega); pZsGreen1-C1, pZsGreen1-N1, pMCherry-C1, pmCherry-N1 (Clontech); pcDNA3.1(–), pcDNA3.1(-)FLAG (Invitrogen), pGEX-6P-1 (GE Healthcare); and pMAL–c2x, pMAL-c4e (NEB); pQE30, pQE60 (QIAGEN). Y2H expression plasmids were generated with the Gateway technology (Invitrogen) as recommended by the manufacturer. Y2H vectors were: pDONR221 (Invitrogen), used as entry vector, and pBTM116-D9 and pACT4-DM (Goehler et al., 2004; Stelzl et al., 2005; Suter et al., 2013), as bait and prey destination vectors, respectively. Additionally, pFLAG-CMV-D11 (FLAG-tagged destination vector) was used for co-IP experiments (Invitrogen). pcDNA3-FLAG-DDB1 (Addgene #19918) and pcDNA3-HA2-DDB1 (Addgene #19909), deposited by Yue Xiong (Hu et al., 2008), were also used.

Anti-Myc (Sigma-Aldrich M5546; 1:1000), anti-FLAG (Sigma-Aldrich F3165; 1:400 or Cell Signaling #2368; 1500), anti-HA (Sigma-Aldrich H9658; 1:1000) and anti-Myc (Millipore 06-549; 1:500) were used for immunodetection of epitope-tagged proteins. Anti-HA (Thermo Fisher Scientific PA1-985, IP: 5 μg), anti-DCAF8 [Bethyl Laboratories, Inc. A301-556A and A301-557A; immunoblotting (IB): 1:500–1:1000, immunohistochemistry (IHC): 1:50–1:100, IP: 5 μg), anti-DDB1 (Bethyl Laboratories, Inc. A300-462A; IB: 1:500–1:1000, IP: 5 μg), anti-Cul4A (Bethyl Laboratories, Inc. A300-739A; 1:500), anti-RBX1 (Cell Signaling #11922; IB: 1:1000), anti-β-actin (Cell Signaling #4967; 1:1000), anti-MuRF1 (abcam ab57865; IB: 1:500, IHC: 1:50–1:100), anti-GAPDH (Millipore MAB374; 1:30,000), anti-USP13 (Thermo Fisher Scientific, PA5-3122; 1:1000), anti-USP35 (Novusbio, NBP1-28733; 1:1000), anti-vinculin (Sigma-Aldrich SAB4200080; 1:1000) and anti-MyHC chain (MF20) (R&D Systems, MAB4470; 1:1000) were used for immunodetection of endogenous proteins in IB or IHC experiments or for IP. A polyclonal anti-MuRF1 antibody was generated by immunizing rabbits with a C-terminal fragment (aa 185 - 355) of Mus musculus MuRF1(-GST) purified from E. coli (IB: 1:500). Secondary antibodies were: horseradish peroxidase-coupled antibodies (Sigma-Aldrich A9044 and A0545; 1:10,000) and IRDye 680RD anti-mouse-IgG for immunodetection by enhanced chemiluminescence or fluorescence emission; for immunohistochemistry detection: Alexa Fluor 488 and 555 (Invitrogen A11001 and A21422; 1:500).

African green monkey fibroblast (COS-7, CRL-1651), mouse skeletal muscle myoblast (C2C12, CRL-1772), and rat cardiac muscle myoblast (H9c2, CRL-1446) cells were received from the American Type Culture Collection (ATCC), cultured and maintained as recommended by the ATCC. C2C12 myoblast cells were differentiated into myotubes by reducing the FBS (PAA, A15-151) concentration of the culture medium from 10% to 2% for at least 5 days. COS-7 and C2C12 cells were transfected with plasmid DNA using Lipofectamine 300 and P3000 reagent or Lipofectamine and PLUS reagent (Invitrogen), respectively, as recommended by the manufacturer. H9c2 cells were transduced with adenovirus (AdV) to express Myc(His)₆ fusions of MuRF1, 2 or 3 for SILAC-AP-MS analysis. A promoterless adenovirus (AdV-PL) was used as control. All adenovirus were generated by SIRION BIOTECH (Martinsried, Germany). H9c2 cells were transduced with a multiplicity of infection (MOI) of 200 and incubated for at least 24 h. Subsequent protein synthesis was allowed in all experiments for a at least 48 h. To reduce the protein amount of DCAF8 (catalog number L-054718-01-0005) and MuRF1 (catalog number L-067480-01-0005), C2C12 cells were transfected with the respective commercially available ON-TARGETplus siRNA SMARTpool (Dharmacon Inc) or with non-targeting scrambled (catalogue number: D-001810-10-05) siRNA. 25 nM siRNAs were transfected using DharmaFECT3 according to the manufacturer's instructions for 72–96 h (medium renewal after 24 h). DCAF8 (catalog number: L-013062-01-0005) was knocked down in COS-7 cells with DharmaFECT2.

H9c2 cells were grown in medium supplemented with either light (lysin-0, arginin-0) or heavy (lysin-8, arginin-10) SILAC amino acids (aa) for 24 h. The SILAC-labeled H9c2 cells were transduced with Adv containing either a control plasmid (labeled with light aa) or the MuRF [MuRF1–Myc(His)₆, MuRF2–Myc(His)₆ or MuRF3–Myc(His)₆] expression plasmid (heavy aa). After transduction, surplus AdV was removed by washing and protein synthesis was allowed for 48 h. APs with TALON Metal Affinity Resin was then performed as described above. In each experiment, the SILAC labeling efficiency was ≥94%.

Proteins isolated by the APs were dissolved in 6 M urea, 2 M thiourea, 20 mM Hepes pH 8.0. Disulfide bonds were first reduced with dithiothreitol (DTT) and then cysteine groups were alkylated with 2-chloroacetamide (CAA). The proteins were proteolyzed with Lys-C and trypsin in a two-step procedure (Kanashova et al., 2015). Peptides were extracted, desalted and stored on reversed-phase (C18) StageTips (Ishihama et al., 2006). After elution, the peptides were lyophilized and resolved in 3% trifluoroacetic acid (TFA), 5% acetonitrile in preparation for LC-MS analysis. Peptides were separated on a 15 cm in-house made reverse-phase column, 75 μM ID, 3 μM C18 (Dr Maisch) by a gradient from 4% to 42% B in 135 min, with a flow-rate of 300 nl/min, and detected on a LTQ-Orbitrap XL. MS acquisition was performed at a resolution of 60,000 in the scan range from 300 to 170 m/z. The MS/MS spectra were collected in the LTQ part of the instrument with the dynamic exclusion set to 30 s. The data were processed using the MaxQuant software package (Cox and Mann, 2008) with the rat IPI-database v3.87, Cysteine carbamylation set as a fixed and methionine oxidation as a variable modification. Peptides with heavy:light (H:L) ratios higher than 1.5 were considered to be enriched with the corresponding MuRF protein.

The protein interactions from this publication have been submitted to the IMEx (www.imexconsortium.org) consortium through IntAct (Orchard et al., 2014) and assigned the identifier IM-25805.

The protein–protein interaction (PPI) network shown in Fig. 1B was generated with Cytoscape 2.8.2 (Shannon et al., 2003) and shows all the PPI that were validated in co-IP experiments.

For immunofluorescence microscopy, cells were plated on coverslips coated with gelatin (Sigma, G1393), transfected with the indicated constructs and, 24 h later, fixed for 20 min with 4% (v/v) paraformaldehyde in PBS. Coverslips were then washed with PBS and treated with PBS containing 0.2% Triton X-100 for 20 min at room temperature. After another wash with PBS, the samples were incubated with either 5% (v/v) goat serum or bovine serum albumin. The fixed cells were then incubated overnight at 4°C with the primary antibody, washed with PBS, and incubated for 1 h with the secondary antibody at room temperature. Afterwards, cells were washed three times with PBS and mounted with ProLong^® Gold antifade reagent containing DAPI (Thermo Fisher Scientific, 36931). The actin cytoskeleton was visualized with phalloidin conjugated to FITC (green) or –TRITC (red); fixed and permeabilized cells were incubated with 1 μl FITC or TRITC (0.5 mg/ml) in 100 μl PBS for 40 min at room temperature, subsequently washed with PBS and mounted. All images were acquired with an inverted fluorescence microscope (Axio Observer Z1, Zeiss) using a cooled CCD camera (AxioCam MRm) and AxioVision (rel. 4.7.2.0) software.

Stable C2C12 DCAF8-knockout cells were generated using the CRISPR/Cas9 technique (Ran et al., 2013). In brief, guide oligonucleotides targeting exon 2 of murine DCAF8 were designed using the CRISPOR web tool (www.crispor.tefor.net/crispor.py). The oligonucleotides were annealed and cloned according to the Zhang lab protocol (accessed via www.addgene.org/crispr/zhang/) into a modified pX330 vector coding also for a mCherry-tagged Cas9 (a kind gift of Dr Ralf Kühn, MDC Berlin). C2C12 cells were then transfected as described above. At 36 h post transfection single mCherry-positive cells were sorted by the MDC-FACS Core Facility. Individual clones were expanded and the DCAF8 knockout was verified by immunoblotting and sequencing of the genomic locus. To this end, genomic DNA was extracted (QuickExtract™ DNA Extraction Solution 1.0, Epicentre), the locus of DCAF8 exon 2 was PCR-amplified and the amplicon sequenced. Guide sequence and oligonucleotides used to generate stable DCAF8 knockout C2C12 cells were: guide sequence (including PAM), 5′-GCACCGTGGACAGCGCAAACGGG-3′; oligonucleotide 1 (including overhang for BbsI cloning into pX330), 5′-CACCGCACCGTGGACAGCGCAAAC-3′; oligonucleotide 2 (including overhang for BbsI cloning into pX330), 5′-AAACGTTTGCGCTGTCCACGGTGC-3′; validation primer forward direction, 5′-GCAAACCTGAAACCTGAGGC-3′; validation primer reverse direction, 5′-GCTGTAGGCTCCTGGATGTG-3′. Validation primers were used for amplification and sequencing of DCAF8 exon 2.

Affinity purifications (AP)  and immunopurifications of Myc(His)₆- or FLAG-tagged proteins from eukaryotic cells was performed in a 50 mM phosphate buffer pH 7.4 containing 150 mM NaCl. This lysis buffer was supplemented with cOmplete protease inhibitor cocktail (Roche) and 0.5–1% (v/v) Triton X-100. The cells were lysed on a rotating wheel for 30 min at 4°C and lysate was cleared by centrifugation (16,000 g, 4°C, 10 min). His-APs were performed using TALON Metal Affinity Resin (Clontech) in presence of 10 mM imidazole in the lysis buffer over night at 4°C and after washing of the beads with lysis buffer bound proteins were eluted with 200 mM imidazole in lysis buffer. For FLAG immunopurifications, the lysate was incubated with anti-FLAG M2 Affinity Gel (Sigma-Aldrich, A2220) at 4°C for 1 to 4 h and after washing of the beads the bound material was eluted by boiling (95°C, 10 min) in protein sample buffer. Endogenous DCAF8 and DDB1 were purified from C2C12 cell lysates with 5 μg of anti-DCAF8 or anti-DDB1 antibodies and protein G–Sepharose (GE healthcare, 50 µl slurry) overnight at 4°C and after washing eluted from the beads in protein sample buffer. Samples were analyzed by SDS-PAGE and immunoblotting. The signals were visualized with specific antibodies and horseradish peroxidase-coupled antibodies using Western Lightning Plus-ECL (PerkinElmer) in a LI-COR Odyssey FC according to the manufacturer's instructions.

A region encoding aa 185–355 of mouse MuRF1 (Fig. S4) was cloned into pGEX-6P-1 and the resulting construct transformed into E.coli BL21 cells. The expression of a MuRF1–GST fusion construct was induced by the addition of 0.5 mM IPTG for 4 h. The E. coli cells were lysed in a French press and the lysate cleared by centrifugation at 17,200 g at 4°C for 30 min. GST purifications were performed using glutathione–Sepharose 4 fast flow (GE Healthcare) and eluted by cleaving the GST tag with GST-tag PreScission protease according to the manufacturer's instructions. Rabbits were then immunized three times by injecting 200 µg of purified MuRF1 protein. The specificity of the antiserum was tested in immunoblots using lysates of muscle cells derived from WT, MuRF1 knockout or MuRF1/MuRF3 double knockout mice (Fig. S4B).

COS-7 cells were transfected with the given constructs and after 24–48 h the translation of polypeptides was inhibited by adding 100 μg/ml cycloheximide (CHX). Cells were harvested at the given time points after CHX addition, lysed as mentioned above, boiled with protein sample buffer, and analyzed by SDS-PAGE and immunoblotting. Quantitative analysis of the immunoblots was performed with fluorescently labeled secondary antibodies and LI-COR Odyssey FC using the Image Studio™ Lite (LI-COR Biosciences) or FIJI software (Schindelin et al., 2012). For steady state analysis, cells were transfected and harvested after 48 h of transient expression of the respective constructs. Where indicated, 1 µM of MLN 4924 (Boston Biochem) or 5 µM lactacystin were added 4 h prior to cell harvest in order to inhibit neddylation, and thereby activation of CRL-ligases, or in order to inhibit the proteasome, respectively. The FLAG-tagged motor region of myosin heavy chain 7 (FLAG–MyHC) was used as substrate (Fielitz et al., 2007a). Transiently expressed FLAG-tagged Cas9, which is not related to the UPS, served as control.

C2C12 myotubes were transfected with siRNA to inhibit expression of the given genes and atrophy was induced by adding 10 μM Dexa for 48 h (mild atrophy induction). The progress of atrophy was determined by measuring the diameter of myotubes. Cell images were captured at 20× total magnification using an inverted fluorescence microscope (Axio Observer Z1, Zeiss) and a cooled CCD camera (AxioCam MRm). At least ten random pictures per experiment were taken and, in each sample, the diameter of 10–20 myotubes was quantified using the Zeiss AxioVision software. Each myotube was measured at its ends and in the middle in three double-blind experiments. Mean±s.e.m. values were calculated and Dexa-treated cells were compared with untreated myotubes. After imaging cells were harvested, boiled in sample buffer, and analyzed by SDS-PAGE and immunoblotting. For strong atrophy induction, 100 µM Dexa was added to C2C12 myotubes for 24 h. Where indicated, 1 µM MLN4924 (BostonBiochem) or 12.5 µM MG132 (Sigma) in 0.05% DMSO (v/v, final concentration) was added to the medium to inhibit neddylation of CRL complexes and the proteasome, respectively.

Denervation experiments were performed as published (Schmidt et al., 2014). Briefly, skeletal muscle atrophy was induced in adult male C57BL/6N mice (6–8 weeks of age) by dissection of the left sciatic nerve. Mice were anesthetized with isoflurane, placed on a heating pad to assure a constant body temperature of 37°C, as measured by a rectal probe. The sciatic nerve of the left leg was cut and a 3 mm piece was excised (defined as the de-nervated sample). The right leg remained innervated and was used as control (defined as the innervated control). Mice were killed at baseline after 7, 14 and 21 days of surgery (n=6). Gastrocnemius plantaris and tibialis anterior muscles were obtained, snap frozen in liquid nitrogen and stored at −80°C until further analysis. The skeletal muscle mass as well as the length of the tibia served as a reference.

All animal procedures, denervation and food deprivation, were performed in accordance with the guidelines of the Max-Delbrück Center for Molecular Medicine and the Charité-Universitätsmedizin Berlin and were approved by the Landesamt für Gesundheit und Soziales (LaGeSo, Berlin, Germany) for the use of laboratory animals (permit number G 0129/12). These rules follow the ‘Principles of Laboratory Animal Care’ (NIH publication No. 86-23, revised 1985) and the current version of German Law on the Protection of Animals.

Data are presented as mean±s.e.m. Differences between indicated myotube groups were analyzed by two-sided unpaired Student's t-tests. Statistical analysis was performed using Microsoft Excel^®. P>0.05 was taken as not significant (n.s.); *P≤0.05, **P≤0.01, ***P≤0.001.

We thank all authors for comments on the manuscript. We thank Ernst Jarosch and Friedrich C. Luft for editing the manuscript. Ernst Jarosch is acknowledged for his help in the anti-MuRF1 antibody production. We thank Raymond J. Deshaies for discussion of CRL4 complex-related results. We thank Ralf Kühn for the modified pX330, Ines Lahmann (Carmen Birchmeier lab) and Christin Suenkel (Nikolaus Rajewsky lab) for technical advice on CRISPR/Cas9 experiments, and the MDC FACS core facility for cell sorting. We thank Corinna Volkwein, Mandy Mustroph and Angelika Wittstruck for excellent technical assistance.