A molecular triage process mediated by RING finger protein 126 and BCL2-associated athanogene 6 regulates degradation of G0/G1 switch gene 2

Oxidative phosphorylation generates most of the ATP in respiring cells. ATP is an essential energy source, especially in cardiomyocytes because of their continuous contraction and relaxation. Previously, we reported that G0/G1 switch gene 2 (G0S2) positively regulates mitochondrial ATP production by interacting with FOF1-ATP synthase. G0S2 overexpression mitigates ATP decline in cardiomyocytes and strongly increases their hypoxic tolerance during ischemia. Here, we show that G0S2 protein undergoes proteasomal degradation via a cytosolic molecular triage system and that inhibiting this process increases mitochondrial ATP production in hypoxia. First, we performed screening with a library of siRNAs targeting ubiquitin-related genes and identified RING finger protein 126 (RNF126) as an E3 ligase involved in G0S2 degradation. RNF126-deficient cells exhibited prolonged G0S2 protein turnover and reduced G0S2 ubiquitination. BCL2-associated athanogene 6 (BAG6), involved in the molecular triage of nascent membrane proteins, enhanced RNF126-mediated G0S2 ubiquitination both in vitro and in vivo. Next, we found that Glu-44 in the hydrophobic region of G0S2 acts as a degron necessary for G0S2 polyubiquitination and proteasomal degradation. Because this degron was required for an interaction of G0S2 with BAG6, an alanine-replaced G0S2 mutant (E44A) escaped degradation. In primary cultured cardiomyocytes, both overexpression of the G0S2 E44A mutant and RNF126 knockdown effectively attenuated ATP decline under hypoxic conditions. We conclude that the RNF126/BAG6 complex contributes to G0S2 degradation and that interventions to prevent G0S2 degradation may offer a therapeutic strategy for managing ischemic diseases.

Oxidative phosphorylation generates most of the ATP in respiring cells. ATP is an essential energy source, especially in cardiomyocytes because of their continuous contraction and relaxation. Previously, we reported that G 0 /G 1 switch gene 2 (G0S2) positively regulates mitochondrial ATP production by interacting with F O F 1 -ATP synthase. G0S2 overexpression mitigates ATP decline in cardiomyocytes and strongly increases their hypoxic tolerance during ischemia. Here, we show that G0S2 protein undergoes proteasomal degradation via a cytosolic molecular triage system and that inhibiting this process increases mitochondrial ATP production in hypoxia. First, we performed screening with a library of siRNAs targeting ubiquitin-related genes and identified RING finger protein 126 (RNF126) as an E3 ligase involved in G0S2 degradation. RNF126-deficient cells exhibited prolonged G0S2 protein turnover and reduced G0S2 ubiquitination. BCL2-associated athanogene 6 (BAG6), involved in the molecular triage of nascent membrane proteins, enhanced RNF126-mediated G0S2 ubiquitination both in vitro and in vivo. Next, we found that Glu-44 in the hydrophobic region of G0S2 acts as a degron necessary for G0S2 polyubiquitination and proteasomal degradation. Because this degron was required for an interaction of G0S2 with BAG6, an alanine-replaced G0S2 mutant (E44A) escaped degradation. In primary cultured cardiomyocytes, both overexpression of the G0S2 E44A mutant and RNF126 knockdown effectively attenuated ATP decline under hypoxic conditions. We conclude that the RNF126/BAG6 complex contributes to G0S2 degradation and that interventions to prevent G0S2 degradation may offer a therapeutic strategy for managing ischemic diseases.
Oxygen (O 2 ) is essential for cell survival because it plays a fundamental role in the production of ATP, mainly through oxidative phosphorylation (OXPHOS). 3 Hypoxia, which is defined as a decrease in O 2 supply, causes the depletion of intracellular ATP and triggers various adaptive cellular responses that maintain the ATP level. For example, hypoxia-exposed cells undergo metabolic switching from OXPHOS toward glycolysis (1-4), but we previously reported that in the early phase of hypoxia, OXPHOS activity rather increases by direct activators such as G 0 /G 1 switch gene 2 (G0S2) and HIGD1A (5,6). We also developed a method to measure mitochondrial ATP concentrations in living cells and showed that G0S2 enhanced mitochondrial ATP production via an interaction with F O F 1 -ATP synthase (5). G0S2 overexpression in cultured cardiomyocytes enhanced mitochondrial ATP production and protected cells from hypoxic damage. These data indicate the therapeutic potential of increasing G0S2 expression for ATP-depleted diseases.
A potential strategy for increasing protein expression is to target the ubiquitin-proteasome system (UPS) to inhibit protein degradation. UPS is the principal mechanism for maintaining cellular protein homeostasis. Ubiquitination, a type of protein posttranslational modification, entails the covalent conjugation of the small ubiquitin protein through the formation of an isopeptide bond between a glycine residue of ubiquitin and lysine residue of the substrate. This reaction occurs through a sequential enzymatic mechanism involving E1 (ubiquitin-activating enzyme), E2 (ubiquitin-conjugating enzyme), and E3 (ubiquitin ligase). E3 is a key enzyme that mediates substrate recognition and is thus responsible for the selectivity and specificity of this system (7). There has been progression recently in the development of therapeutic drugs to specifically inhibit E3 ligase itself or to regulate E3 ligase activity so as to increase the expression of its substrate (8). Because the amount of G0S2 protein is regulated by UPS (9), its degradation process is a potential therapeutic target to increase mitochondrial ATP levels. However, the mechanism regulating the levels of G0S2 protein and the corresponding E3 ligase remains unknown.
In this report, we performed functional screening using an siRNA library and identified RNF126 as a specific E3 ligase for G0S2 degradation. We further demonstrated that the inhibition of G0S2 degradation by RNF126 knockdown or by overexpression of a G0S2 degradation-resistant mutant that sequestered BAG6 preserved mitochondrial ATP concentrations in cardiomyocytes under hypoxic conditions. This result indicates that inhibiting G0S2 degradation may be a novel therapeutic strategy for ATP-depleting conditions such as ischemic heart diseases, mitochondrial diseases, and metabolic diseases.

siRNA screening of ubiquitin ligases for the degradation of G0S2
We first examined G0S2 protein stability in rat neonatal cardiomyocytes. Cells were treated with cycloheximide to inhibit protein synthesis, and the amount of G0S2 protein was analyzed by immunoblotting (Fig. 1A). G0S2 was rapidly degraded with a half-life of ϳ20 min. Proteasome inhibitors, but not lysosomal inhibitors, increased G0S2 protein amounts, indicating degradation by UPS in cardiomyocytes (Fig. 1B). Although G0S2 was reported to be regulated by UPS (9), the E3 ligases involved had yet to be identified. Therefore, to identify the E3 ubiquitin ligase specific for G0S2, we performed functional screening using an siRNA library targeting ubiquitin-related genes encoding E1, E2, and E3. We first established C2C12 myoblast cell lines stably expressing EGFP-fused G0S2 to monitor G0S2 protein degradation (C2C12/EGFP-G0S2 cells). Treatment of these cell lines with the proteasome inhibitor MG132 induced G0S2 protein accumulation, which manifested as increased EGFP intensity (Fig. S1A). As a positive control for the screening, we selected siRNA for UBA1, an E1 ubiquitinactivating enzyme, and confirmed a 2-fold increase in the fluorescence of EGFP-G0S2 after UBA1 knockdown (Fig. S1B). To quantitatively measure the fluorescence intensity of each C2C12/EGFP-G0S2 cell in a high-throughput manner, we used an automated high-content imaging system. In the primary screening, siRNAs from the library were transfected into C2C12/EGFP-G0S2 cells in 96-well plates. After 72-h incubation, cells were imaged and their fluorescence intensities were analyzed ( Fig. 1C; Table S1). As positive hits, we selected the genes that demonstrated a log 2 -fold change in fluorescence that was greater than 0.6 ( Fig. 1D). As a counter assay, we established cell lines expressing the EGFP-fused CL1 degron (ACKNWFSSLSHFVIHL) (C2C12/EGFP-CL1 cells), which has been shown to be rapidly degraded by proteasome (10) (Fig. 1C;  Table S1). The counter assay, together with an assessment of cell number in primary screening (Table S1), identified one gene, RNF126, as a candidate for the G0S2-specific E3 ligase (Fig. 1E). Because the screening used pools that included four individual siRNAs at once, individual siRNA for RNF126 was separately transfected to C2C12 stably expressing HA-tagged G0S2 cells, and G0S2 protein levels were examined. We confirmed ϳ60% knockdown of RNF126 by each siRNA (Fig. 1F), and G0S2 protein levels were markedly increased by every siRNA (Fig. 1G).

RNF126 regulates G0S2 protein degradation by ubiquitination
To examine whether RNF126 acts as an E3 ligase for G0S2 in cells, we used CRISPR-Cas9 technology to generate two RNF126 KO cell lines, specifically RNF126-deficient (RNF126 KO) and RNF126-deficient EGFP-G0S2-expressing (RNF126 KO/EGFP-G0S2) cells (Fig. S2). Endogenous G0S2 protein expression was elevated in RNF126 KO cells ( Fig. 2A), and RNF126 KO/EGFP-G0S2 cells exhibited an increased EGFP-G0S2 protein half-life as shown by cycloheximide treatment (Fig. 2, B and C). Next, to examine whether the G0S2 protein accumulation was because of RNF126 deficiency, human RNF126 (hRNF126) was dose-dependently expressed in RNF126 KO cells (Fig. 2D). In hRNF126-expressing cells, accumulation of G0S2 protein was decreased. We further tested the in vivo ubiquitination of G0S2 in both RNF126 KO cells and control cells. After the transfection of HA-G0S2, cell lysates were subjected to a TUBE2-Agarose pulldown assay, followed by immunoblotting (Fig. 2E). A ladder of polyubiquitinated G0S2 was clearly observed in control cells. On the other hand, in RNF126 KO cells, the polyubiquitinated form of G0S2 was decreased (Fig. 2E). Re-expression of hRNF126 in RNF126 KO cells restored G0S2 ubiquitination. These data suggest that RNF126 acts as an E3 ligase for G0S2 and specifically regulates its degradation in vivo.

Knockdown of RNF126 in neonatal cardiomyocytes preserves mitochondrial ATP production under hypoxia
To investigate the physiological role of RNF126 as a G0S2 E3 ligase, we used primary cultured cardiomyocytes. The knockdown of RNF126 using siRNA in cardiomyocytes confirmed G0S2 protein accumulation without affecting its mRNA expression (Fig. 3, A and B). In a previous report, we used a mitochondria-targeted FRET-based ATP biosensor, Mit-ATeam, to show that overexpression of G0S2 enhanced mitochondrial ATP production in neonatal cardiomyocytes under hypoxic conditions (5,11). In the present study, we used this probe to examine whether RNF126 affected mitochondrial ATP production in cardiomyocytes under hypoxic conditions. Because G0S2 expression is induced by hypoxic stimuli, we examined whether hypoxic stimuli change G0S2 turnover. Even under hypoxic condition, G0S2 was also degraded at the same rate of that in normoxic condition (Fig. S3A) and its degradation was inhibited by proteasome inhibitor (Fig. S3B). Moreover, hypoxic stimuli did not change the protein expression of RNF126 (Fig. S3C). We transfected RNF126 siRNA or control siRNA into cardiomyocytes and then infected cells with Mit-ATeam-encoding adenovirus. Forty-eight h after adenovirus Molecular triage-mediated G0S2 degradation infection, mitochondrial ATP concentrations were measured by a Mit-ATeam assay during hypoxia. The mitochondrial ATP concentration ([ATP] mito ) gradually declined in cardiomyocytes transfected with either control or RNF126 siRNA. Notably, the knockdown of RNF126 before the onset of hypoxia reduced this decline in [ATP] mito (Fig. 3, C and D). These data indicated that increased G0S2 protein levels caused by RNF126 knockdown enhanced mitochondrial ATP production.

BAG6 regulates G0S2 ubiquitination and degradation
RNF126 is a RING finger-type E3 ubiquitin ligase and several substrates of RNF126 have been reported (12)(13)(14)(15)). In addition, RNF126 was recently reported to be an E3 ligase that degrades mislocalized or tail-anchored transmembrane proteins together with the BAG6-UBL4A-TRC35 complex (16). Thus, we examined the involvement of the BAG6 complex in RNF126-mediated G0S2 degradation. First, we knocked down BAG6 by specific siRNA in neonatal rat cardiomyocytes and C2C12 cells expressing HA-G0S2, then analyzed G0S2 protein amounts. BAG6 knockdown increased G0S2 protein expression in both cardiomyocytes and C2C12 cells (Fig. 4, A and B) and also inhibited G0S2 ubiquitination (Fig. 4C). Cycloheximide chase analysis showed that the halflife of G0S2 was significantly longer than that of control cells in

Molecular triage-mediated G0S2 degradation
BAG6 knocked-down cells (Fig. 4, D and E). These data suggest that BAG6 is also essential for G0S2 ubiquitination and degradation as a partner protein of RNF126. Furthermore, we examined whether BAG6 knockdown influences mitochondrial ATP decline under hypoxic condition in cardiomyocytes using Mit-ATeam assay. Hypoxia did not affect BAG6 protein expression (Fig. S3C). Intriguingly, despite the accumulation of G0S2 protein ( Fig. 4A), BAG6 knockdown did not affect the mitochondrial ATP production under hypoxia (Fig. 4, F and G), suggesting that BAG6 acts as not only a part of E3 ligase complex but also a molecular switch of newly synthesized protein (see "Discussion").

Identification of the degron in G0S2
To further reveal the mechanism of G0S2 degradation, it is important to identify the degron in G0S2 that is involved in substrate recognition by the E3 ligase (17,18). We created several deletion mutants in the N-terminal and C-terminal domains of G0S2, as well as alanine-scanning mutants in the hydrophobic region (Fig. 5A). These G0S2 mutants were transfected into C2C12 cells. After treatment with MG132, cell lysates were analyzed by immunoblotting (Fig. 5B). Among these mutants, only the Glu-44 to Ala (E44A) mutant was unaffected by MG132, showing a vast increase in basal protein expression. This suggests that amino acid Glu-44 of G0S2 plays an important role in G0S2 degradation. Then, we assessed the in vivo ubiquitination and protein half-life of G0S2 E44A. G0S2 E44A was less ubiquitinated (Fig. 5C) and had a longer half-life (Fig. 5, D and E) than the WT. These data suggested that the domain around Glu-44 is a degron of G0S2 that is recognized by the RNF126/BAG6 complex. Thus, we tested the interaction of G0S2 with BAG6 in cells expressing WT or E44A mutant of G0S2. A co-immunoprecipitation experiment revealed that BAG6 preferentially interacted with G0S2 WT rather than with E44A (Fig. 5F). These data suggested that BAG6 interacted with G0S2 and recruited RNF126 for ubiquitination.
Because BAG6 not only acts as a scaffold protein of RNF126 but also has a shield-like function to protect the hydrophobic domain of the substrate (19), we considered that BAG6 knockdown or the expression of G0S2 E44A might have caused G0S2 protein aggregation. Thus, to examine the effect of BAG6 knockdown on protein aggregation, C2C12 cells expressing HA-G0S2 WT or E44A were transfected with siRNA for BAG6 and then solubilized with buffer containing 1% Triton X-100. In BAG6-depleted cells, protein aggregates were detected in the detergent-insoluble fraction (20,21), whereas G0S2 E44A mutant as well as WT existed only in the detergent-soluble fraction (Fig. 5G). These data suggest that the G0S2 E44A escaped BAG6-mediated degradation without causing its protein aggregation.

Molecular triage-mediated G0S2 degradation RNF126 directly ubiquitinates G0S2 in vitro in a BAG6-dependent manner
Next we tested whether the RNF126/BAG6 complex was sufficient to ubiquitinate G0S2 by constructing an in vitro ubiquitination assay. We prepared recombinant GST-RNF126 using Escherichia coli and recombinant G0S2 and BAG6 from HEK293T cells (Fig. S4, A-C). UBCH5B was chosen as the E2 ubiquitin-conjugating enzyme for RNF126 because only the knockdown of UBCH5B increased G0S2 protein expression during our siRNA library screening. Moreover, UBCH5B was reported to act as an E2 enzyme with RNF126 (13, 15,16). In an in vitro ubiquitination assay, recombinant G0S2 was slightly ubiquitinated by GST-RNF126 only in the presence of E1, E2, ATP, and ubiquitin (Fig. 6A, lane 4). In contrast, the addition of recombinant BAG6 strongly enhanced G0S2 ubiquitination by RNF126 (Fig. 6A, lane 5; Fig. S4D). A catalytically inactive mutant of RNF126 (C231A/C234A, CA) (22), which was created by substituting two critical cysteine residues in a RING domain with alanine residues, did not ubiquitinate G0S2, even in the presence of BAG6 (Fig. 6B, lane 3). The G0S2 E44A mutant that inhibited BAG6 interaction was less ubiquitinated than the WT, as well as the G0S2 lysine-less mutant (Fig. 6B,  lanes 4 and 5; Fig. S4E). We also examined whether G0S2 was ubiquitinated at the type of Lys-48 or Lys-63 linkage by combining an in vitro ubiquitination assay and subsequent immunoprecipitation. Immunoblotting with Lys-48 linkage-specific antibody revealed that RNF126 ubiquitinated G0S2 at Lys-48linked ubiquitination, in consistent with previous reports (Fig. S4F) (12,22). These data indicate that the BAG6 complex is sufficient for G0S2 recognition and ubiquitination.

Inhibition of G0S2 interaction with BAG6 leads to the preservation of mitochondrial ATP concentration in hypoxia
To confirm the importance of BAG6-G0S2 interaction in G0S2 degradation, G0S2 E44A-or WT-encoding lentivirus was expressed in cardiomyocytes. Treatment with proteasome inhibitors results in accumulation of the G0S2 WT protein but not the E44A mutant (Fig. 7A). Cycloheximide chase analysis also revealed that compared with WT, the G0S2 E44A mutant had a greatly increased half-life and higher basal protein expression (Fig. 7, B and C). We next examined the changes in [ATP] mito during hypoxia with the Mit-ATeam assay. Whereas overexpression of G0S2 WT enhanced [ATP] mito , as reported previously, G0S2 E44A also enhanced [ATP] mito to a greater extent than WT (Fig. 7, D and E). We further tested whether this augment of mitochondrial ATP production led to restore the cellular viability in hypoxic stress. The prehypoxia overexpression of G0S2 E44A in cardiomyocytes ameliorated hypoxia-induced cell death (Fig. 7F). These results suggest that inhibiting G0S2 degradation via BAG6 sequestration enhances mitochondrial ATP concentration under hypoxic condition and maintains cellular viability.

Discussion
Our recent study showed that G0S2 played an important role in mitochondrial ATP production under hypoxic conditions  (5). Overexpression of G0S2 in cardiomyocytes attenuated the decline in mitochondrial ATP levels during hypoxia and protected cells from hypoxic damage. These results indicate that increasing in G0S2 expression may be a novel therapeutic target for ischemic disease. Thus, for the purpose of developing innovative drugs to increase G0S2 expression, in this study we focused on G0S2 protein turnover.

Molecular triage-mediated G0S2 degradation
We demonstrated rapid G0S2 protein degradation in cardiomyocytes and identified RNF126 as an E3 ligase for G0S2 by functional screening using an siRNA library. RNF126 has been implicated in a number of physiological processes, such as mitochondrial metabolic flux (23) and membrane receptor trafficking (22). RNF126 was also shown to directly ubiquitinate and promote the degradation of the cyclin-dependent kinase inhibitor p21 (14) and the mitochondrial iron-sulfur protein frataxin (13), but the mechanism of substrate recognition by RNF126 was not fully understood.
On the other hand, recent studies using an in vitro translation system revealed that an RNF126/BAG6 complex targeted mislocalized membrane proteins for proteasomal degradation (16,19,24). The mislocalized proteins that failed to be degraded because of BAG6 or RNF126 depletion were rapidly aggregated and accumulated in the cytosol, and these defective proteins are theorized to cause the onset of various diseases (16,19,24). Therefore, the RNF126/BAG6 complex plays a major role in the cytosolic pre-emptive quality control system. In this paper, we demonstrated that BAG6 was required for G0S2 ubiquitination and degradation by RNF126. BAG6 knockdown increased G0S2 protein levels and prolonged protein turnover by inhibiting degradation

Molecular triage-mediated G0S2 degradation
without causing protein aggregation. Given these data, newly synthesized native G0S2 protein is considered to be constitutively degraded by RNF126 and BAG6 rather than eliminated as a mislocalized protein. Thus, in certain conditions G0S2 may change its conformation or undergo modifications to become a functional protein.

Molecular triage-mediated G0S2 degradation
It was also recently revealed that BAG6 plays a major role as a switching hub in proteasomal degradation or importing proteins into membrane (25). Newly synthesized membrane proteins in ribosomes are first trapped by SGTA and then transferred to the BAG6 complex (BAG6-UBL4A-TRC35). In the case of the tail-anchored (TA) protein, the targeting factor TRC40 delivers the TA protein trapped by SGTA to the endoplasmic reticulum for insertion, whereas BAG6 recruits the E3 ligase RNF126 for protein ubiquitination and degradation (25,26). Although this precise triage system has been demonstrated using an elegant in vitro translation method in previous reports, it remains unclear how it is determined whether newly synthesized TA proteins are degraded or imported into membranes.
Generally, degradation signals within a protein that make it short-lived are called degrons (17,27). In experiments using deletion mutants and alanine scanning, we determined that Glu-44 is a specific degron within the G0S2 protein. G0S2 has unique hydrophobic region with only one negatively charged glutamic acid. Loss of a negative charge in Glu-44 by alanine

Molecular triage-mediated G0S2 degradation
substitution not only increased G0S2 protein stability but also enhanced its role in mitochondrial ATP production. We further revealed that the G0S2 E44A mutant reduced interaction with the BAG6 complex, indicating that this mutant might be more efficiently delivered from the BAG6 complex to subsequent membrane targeting factors. Although some model substrates such as SEC61␤ or VAMP2 were used to analyze the cytosolic quality control of nascent membrane proteins (19,25), none of these substrates contains negatively charged amino acids within its transmembrane domain. Therefore, in contrast to general TA proteins, the specific regulatory mechanism of Glu-44 -mediated G0S2 degradation, such as masking of its negative charge by binding of an unidentified protein factor, or the posttranslational modification of surrounding residues may be crucial for the escape of G0S2 from the degradation pathway and subsequent membrane targeting that enables appropriate functioning (Fig. S5).
Taken together, we showed that G0S2 protein turnover was regulated by molecular triage system involving RNF126 and BAG6 in the cytosol, although G0S2 does not seem to be a TA protein. Moreover, we identified the degron of G0S2, which is not seen in other substrates of BAG6, and the alanine-replaced mutant inhibited the degradation of G0S2 without aggregation and enhanced mitochondrial ATP production in hypoxia. These findings suggest that further functional analysis may facilitate the development of innovative drugs that inhibit G0S2 degradation by targeting its degron.

Cell culture and transfection
Primary neonatal cardiomyocytes from 1-or 2-day-old Wister rats were isolated and cultured as described previously (29). C2C12 and HEK293T cells were prepared and cultured in DMEM (Sigma-Aldrich) supplemented with 10% FBS and 1% penicillin-streptomycin. Hypoxic condition (1% O 2 ) was provided by MCO-5M multi-gas incubator (Sanyo) unless described otherwise. Lipofectamine 2000 and Lipofectamine 3000 were used for plasmid transfection into HEK293T and C2C12 cells, respectively. To knock down endogenous genes, cells were transfected with siRNAs for targeting genes using Lipofectamine RNAiMAX.

Plasmid and viral vector constructions
Mouse G0S2 coding sequence (NM_008059.3) was subcloned into pCDH-EF1 cloning vector with puromycin or G418 resistant gene via 2A peptide (System Biosciences) for plasmid transfection and lentiviral production. The deletion mutants of N-terminal and C-terminal portions of G0S2 or alanine substitution mutants were generated by PCR-based mutagenesis using N-terminal HA-tagged G0S2 WT (HA-G0S2) as a template. N-terminal EGFP ORF was inserted into N terminus of G0S2 by normal restriction enzyme method (EGFP-G0S2). For lentiviral particle production, plasmids encoding various G0S2 WT or mutants were transfected into HEK293T cells together with pRSV-Rev (Addgene no. 12253), pMD 2.G (Addgene no. 12259), and pMDL g/p RRE plasmids (Addgene no. 12251) using Lipofectamine 2000. Seventy-two h after transfection, cell supernatants were collected, and then filtered viral supernatants to remove any nonadherent packaging cells were stocked at Ϫ80°C. Mouse RNF126 (NM_144528.3) and human RNF126 (NM_194460.3) were amplified by PCR from mouse and human heart cDNA, respectively, and subcloned into pCDH-EF1 vector or pENTR1A-Dual selection vector. Adenovirus production was performed as described previously (5). The plasmid encoding human BAG6 (pRK5-FLAG-BAG6) was obtained from Addgene (no. 61836).

siRNA library screening
C2C12 cell lines stably expressing EGFP-fused G0S2 or CL1 degron for siRNA library screening were generated as follows: Lentiviral particles encoding EGFP-G0S2 or EGFP-CL1 (ACK-NWFSSLSHFVIHL) were generated as described above and then infected into C2C12 cells, followed by puromycin selection (5 g/ml, Sigma-Aldrich). Single clones obtained by limiting dilution were used for siRNA library screening. siRNA library screening was performed as follows: For siRNA library screening, we targeted 523 genes encoding factors related to UPS from siGENOME SMART pool siRNA Library-Mouse Ubiquitin Conjugating Subset1ϳ3 (Dharmacon). This library was dissolved in siRNA buffer (60 mM KCl, 6 mM HEPES, 0.9 mM MgCl 2 ) to adjust final concentration in 1 M and dispensed with each well containing 3 pmol of a pool in 96-well plate. siUBA1 and siCTL were used as the positive and negative controls, respectively. Transfection reagent Lipofectamine RNAiMAX was first added to siRNA library-containing plates according to the manufacturer's protocol. After incubation in brief, C2C12/EGFP-G0S2 or C2C12/EGFP-CL1 cells were seeded onto 96-well plates with the appropriate cell numbers. Twenty-four h after transfection, media were changed to Fluo-roBrite DMEM (Gibco) supplemented with 10% FBS, 1% penicillin-streptomycin, and 1ϫ GlutaMax to avoid the toxicity of Molecular triage-mediated G0S2 degradation transfection reagent, and cells were incubated for additional 48 h. After staining nuclei with Hoechst 33342, cells were imaged using IN Cell Analyzer 6000 (GE Healthcare), and EGFP fluorescence intensities or cell numbers of each well were quantitatively analyzed by Developer ToolBox (GE Healthcare). Each assay was performed and normalized to the siCTL in three independent experiments. The following siRNAs were used in this study: siUBA1, CACUUACUUUUGAUGUUAA; siRNF126#1, UCACGCAGCUCCUCAAUCA; siRNF126#2, GCAACCACCUGUUCCACGA; siRNF126#3, CAUCUUCG-ACGAUAGCUUU; siRNF126#4, UCACGCUGCCACAGGG-AUA; siBAG6#1, GGGUACCUAUUAUCCAGCA; siBAG6#2, CCUUCAAUCUUCCUAGUGA; siBAG6#3, GCACAUGAU-UAGGGAUAUA.

RNA extraction and quantitative RT-PCR
Total RNA was extracted from C2C12/HA-G0S2 using RNA-Bee Reagent (Tel-Test) and converted to cDNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) according to the manufacturer's instructions. Quantitative RT-PCR was performed with SYBR technology and Step One Plus Real-Time PCR Systems (Applied Biosystems). All of the samples were processed in duplicate. The level of each transcript was quantified by the threshold cycle method using GAPDH as an endogenous control.
After extensive washing, the proteins were eluted three times with HA or FLAG peptide. Recombinant GST-fused RNF126 proteins (WT and C231A/C234A) were purified as follows: The mouse RNF126 cDNA was subcloned into pGEX-6P-1 Vector. The RING domain mutant (C231A/C234A) was generated by PCR-based mutagenesis using pGEX-6P-1/RNF126 WT as a template. pGEX-6P1/RNF126 WT or RNF126 C231A/C234A mutant was transformed into BL21-Rosetta chemically competent E. coli (Invitrogen), and then protein expression was induced with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside (Sigma-Aldrich) at 15°C for 6 h. The cells were collected by centrifugation and lysed by sonication in buffer A. After addition of 2% Triton X-100, the cell lysates were agitated at 4°C for 30 min and pulled down with GSH-Sepharose 4 Fast Flow (GE Healthcare) at 4°C for 1 h. After being washed three times, the proteins were eluted with 10 mM reduced GSH and ultrafiltered in buffer A using a Nanosep 50K Device (Pall Corp.).

In vivo/in vitro ubiquitination of G0S2
In vivo ubiquitination assay was performed as follows: C2C12/HA-G0S2 or RNF126 KO cells were lysed with buffer A containing 0.5% Nonidet P-40 and immunoprecipitated with TUBE2-Agarose at 4°C, for 4 h. After extensive washing, proteins were eluted with SDS sample buffer. In vitro ubiquitination assay was performed as follows: in vitro ubiquitination assay was performed in a mixture containing 50 nM human ubiquitin-activating enzyme (UBE1), 250 nM human UBCH5B (UBE2D2), 2.5 nM purified GST-RNF126, 50 nM purified FLAG-BAG6 from 293T cells, and 73.5 nM purified HA-G0S2 from 293T cells in reaction buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 2 mM ATP, 1 mM DTT, 10 M ubiquitin, 100 M ZnCl 2 ) at 37°C for 60 min. Reaction was stopped by adding SDS sample buffer. Samples were electrophoresed and analyzed with immunoblotting.

Preparation of detergent-soluble and -insoluble fraction
C2C12 cells were transfected with 30 nM either siCTL or siBAG6. Forty-eight h after transfection, HA-G0S2 WT or E44A was transfected into cells expressing siRNA. Forty-eight h after plasmid transfection, cells were treated with 10 M MG132 for 4 h and harvested with buffer A containing 1% Triton X-100. Cell lysates were centrifuged at 20,000 ϫ g for 20 min at 4°C, and the resulting precipitates were dissolved in SDS sample buffer in equivalent to a supernatant.

FRET-based measurement of mitochondrial ATP concentration (Mit-ATeam assay)
FRET-based measurement of mitochondrial ATP concentration in cardiomyocytes was measured as described previously (5). Briefly, FRET signal was measured in cardiomyocytes infected with adenoviral encoding Mit-ATeam 1.03 with an Olympus IX-81 inverted fluorescence microscopy (Olympus) using a PL APO 60ϫ, 1.35 N.A., oil immersion objective lens (Olympus). Mit-ATeam assay was performed under 1% O 2 condition at 48 h after siRNA transfection or lentivirus infection.

Molecular triage-mediated G0S2 degradation Cell viability assay
Cardiomyocytes seeded on 96-well plates were infected with lentivirus encoding G0S2 WT or E44A mutant for 24 h, and then they were exposed to hypoxia for 24 h. Hypoxic condition (less than 0.1%) was achieved by the AneroPack System (Mitsubishi Gas Chemical, Inc.). After hypoxia, cells were stained with 1 g/ml propidium iodide (Sigma-Aldrich) and 0.5 g/ml Hoechst 33342 (Dojin Chemical, Inc.) at 37°C for 30 min. The stained nuclei were then imaged using IN Cell Analyzer 6000. The data were analyzed as percentage of propidium iodidepositive nuclei/total nuclei.

Quantification and statistical analysis
Our data are expressed as mean Ϯ S.D. of at least three independent experiments except for Mit-ATeam assay as mean Ϯ S.E. The two-tailed Student's t test was used to analyze differences between two groups unless otherwise noted. For the data shown in Fig. 1F, one-way ANOVA with Dunnett's post hoc test was used to compare multiple groups with siCTL. For the data of Mit-ATeam, as shown in Figs. 3D, 4G, and 7E, two-way ANOVA with Tukey-Kramer's test was used for comparisons of multiple groups at individual time points. For the data shown in Fig. 7F, one-way ANOVA with Tukey-Kramer's test was used.