The TRC8 Ubiquitin Ligase Is Sterol Regulated and Interacts with Lipid and Protein Biosynthetic Pathways Free

Renal cell carcinoma (RCC) comprises 5% of epithelial cancers, with more than 38,000 new cases in the United States each year (1). Most are clear cell RCCs, and of these, a majority contain mutations or epigenetic silencing of von Hippel-Lindau with upregulation of hypoxia-inducible factor (HIF) α subunits and a constitutively activated hypoxic response (2). Although HIF deregulation is a major factor in RCC development, it does not explain all features of the disease (3). Similarly, loss of von Hippel-Lindau function alone is insufficient for tumor formation (4). Besides von Hippel-Lindau, additional genes have been implicated in hereditary renal cancer. These include fumarate hydratase and succinate dehydrogenase (5, 6), which affect HIFα prolyl hydroxylation and stability (7), the hepatocyte growth factor receptor MET (8), as well as tuberous sclerosis (TSC1/2) and Birt-Hogg-Dubé/folliculin (BHD/FLCN), which affect mammalian target of rapamycin (mTOR) activity and protein translation (9-11). However, the control of protein translation initiation involves not only mTOR, which regulates cap-mediated translation, but also other factors, including eIF3, which regulates ribosome and mRNA assembly as well as internal ribosomal entry site–mediated translation (12).

We identified the TRC8/RNF139 gene from its disruption by a constitutional translocation, t(3;8)(3p14.2;q24.1), in a family with hereditary RCC and thyroid cancer (13-15). Recently, a second family with hereditary RCC carrying an independent (3, 8) translocation affecting TRC8 was described (16). In addition to disruption of TRC8 in RCC by chromosomal translocation and tumor-specific loss of the wild-type allele, a rare sporadic mutation was also identified, consistent with a proposed tumor suppressor function (15, 16).

TRC8 encodes a multimembrane-spanning protein, located in the endoplasmic reticulum (ER), and contains a COOH-terminal RING-H2 domain with E3 ubiquitin ligase activity (15, 17-19). Recent work has identified MHC class I molecules as one target for TRC8-mediated ubiquitylation and degradation in cells expressing the cytomegalovirus protein US2 (20), but other targets remain unknown. In HEK293 cells, TRC8 overexpression inhibits growth and tumorigenesis in a RING-dependent manner (19). In Drosophila, expression of DTrc8 in the wing suppressed growth, whereas targeting expression to the eye caused disappearance of retinal cells (17, 21). A yeast two-hybrid screen with DTrc8 identified Csn5/Jab1 (17), the catalytic component of the COP9 signalosome, which acts to remove NEDD8 from E3 ubiquitin ligase complexes (22). Moreover, genetic interaction studies showed that Csn5 loss-of-function mutations partially reversed the effect of DTrc8 overexpression in the wing (21). The Mprlp, Padlp N-terminal (MPN) domain of Csn5 was specifically required for interaction with DTrc8. This domain is found in only a limited number of proteins, including subunits in the COP9 signalosome (Csn5 and Csn6), the 19S regulatory lid of the proteasome (Rpn8 and Rpn11), and eIF3 (eIF3f and eIF3h; ref. 23). Like Csn5, crosses with mutant Csn6 flies relieved TRC8-induced growth suppression, whereas those with Rpn8 and Rpn11 had no effect. Interactions with the eIF3 subunits were not tested.

In addition to the RING domain, TRC8 contains a predicted sterol-sensing domain (SSD) in its NH₂ terminus, suggesting that sterols may influence some aspects of TRC8 function (15). Interestingly, the clear cell phenotype in RCC is due to abundant inclusions of endogenously synthesized cholesterol esters and other lipids, suggesting that lipid homeostasis may be deregulated in this disease (24-26). SSDs are found in a limited number of proteins, some of which (e.g., HMGCR and SCAP) are directly involved in sterol biosynthesis (27). When sterol levels are high, the SSD of 3-hydroxy-3-methylglutaryl CoA reductase (HMG-CoA reductase; HMGCR) interacts with the ER anchor protein INSIG-1 (insulin-induced gene). The ER-associated degradation–associated E3 ubiquitin ligase gp78 binds INSIG-1 despite the lack of a SSD (28). As a result of this complex formation, HMGCR is polyubiquitylated by gp78, leading to its proteasome-mediated destruction (28, 29). Patched, Dispatched, and Neimann-Pick proteins also contain SSDs, although the function of this domain in these proteins is not well understood and no interactions with INSIG have been reported (30).

The transcriptional regulation of cholesterol and lipid biosynthetic enzymes is under the control of two ER-tethered transcription factors: SREBP-1 and SREBP-2. When sterol levels are high, the SREBP precursor proteins (preSREBP) are retained in the ER in a preSREBP-SCAP-INSIG complex. In low sterols, the SCAP-INSIG interaction is lost and SREBP precursors are transported to the Golgi, where they undergo release from the membrane by proteolytic cleavage and subsequently enter the nucleus (27). The SREBPs regulate their own transcription, and that of INSIG-1, as part of positive and negative feedback loops (31). In fission yeast, SREBP activation occurs in response to hypoxia (32), although a similar response has not been described in mammalian cells. However, INSIG-2 is hypoxia regulated (33). Given the apparent disruption of lipid homeostasis in clear cell RCC, this raises the question of whether other regulatory components of the SREBP pathway might be affected in the disease process.

With the presence of a putative SSD in TRC8, we asked whether TRC8 was affected by low sterol levels. We report here that TRC8 protein is stabilized when sterols are reduced and destabilized when sterols are replete. This effect is RING dependent and accompanied by sterol-stimulated ubiquitylation. TRC8 interacts with and stimulates ubiquitylation of INSIG. In addition, it destabilized preSREBPs and reduced expression of SREBP target genes. TRC8 knockdown had opposite effects in sterol-depleted cells. In addition, we found that TRC8 interacts with the MPN domain–containing subunits of the translation initiation factor eIF3, induced ubiquitylation of the complex, and reduced protein translation as assessed by polysome loading. Together, these observations suggest that TRC8 function may provide a regulatory link between the lipid and protein biosynthetic pathways. Of note, the hereditary kidney cancer gene BHD/FLCN may do a parallel function affecting AMP kinase (AMPK) and HMG-CoA reductase (see Discussion).

HEK293 FlpIn TRex (HEK293 FlpIn) cells were maintained in DMEM containing 50 μg/mL zeocin and 5 μg/mL blasticidin. CHO FlpIn TRex lines were generated as described by the supplier (Invitrogen). Stably transfected FI lines were selected in 50 μg/mL hygromycin B. Sterol-depleted medium contained 5% delipidized fetal bovine serum (LPDS; ref. 34), 10 μmol/L mevastatin, and 50 μmol/L mevalonate. Sterol-replete medium contained 5% FCS, cholesterol (12.5 μmol/L), and 25-hydroxycholesterol (25-HC; 10 μg/mL), or as noted in figure legends. Sodium mevalonate was prepared by saponification of mevalonolactone (Sigma-Aldrich). Antibody sources included TRC8/RNF139 (H00011236-A01; Abnova/Novus Biologicals), SREBP-1 (IgG-2A4) and SREBP-2 (IgG-1C6, COOH-terminal specific; Labvision/NeoMarkers), and SREBP-2 (IgG-1D2, NH₂-terminal specific; MBL International). Anti-FLAG M2 beads and antibody were from Sigma. HA (Y-11) and eIF3b (p116) antibodies, anti-myc 9E10 beads, and protein A/G PLUS beads were from Santa Cruz Biotechnology, Inc.; anti-eIF3 complex antibody was kindly provided by J.W. Hershey (University of California-Davis, Davis, CA). HIF1α antibody was from BD Biosciences. HMGCR monoclonal antibody was purified from A9 hybridoma (CRL-1811, American Type Culture Collection).

TRC8 expression constructs have been described (19). myc-INSIG-1 and myc-INSIG-2 were obtained from the American Type Culture Collection; the K156R, K158R double mutation was generated by site-specific mutagenesis using the QuikChange kit (Stratagene), as was the Y32E and Y32C mutations of TRC8-HA; pcDNA3-flag ubiquitin was a gift from R. Wojcikiewicz and D. Sliter (State University of New York, Syracuse, NY). Cells were lysed on ice using specific buffers (see Supplementary Materials and Methods for formulas). In general, 10 μg protein was resolved by SDS-PAGE using 4% to 15% gradient gels (Bio-Rad) and analyzed by Western blot using polyvinylidene difluoride membranes and standard protocols (Millipore). Antibody detection used horseradish peroxidase–conjugated secondary antibodies and chemiluminescence reagents (Pierce) or Alexa Fluor 488–conjugated secondary antibodies and a Typhoon 9400 Variable Mode Imager or CW800 and CW680 IRDye-conjugated secondary antibodies and a LI-COR Odyssey fluorescence scanner.

Immunoprecipitations were done using published protocols (35, 36). Briefly, washed cells were suspended in 0.5 mL of CHAPSO buffer (Supplementary Materials and Methods) and ruptured by passage 15 times through a 22-gauge needle. Cleared supernatants (500 μg protein) were incubated with beads and primary antibodies overnight, washed thrice, and analyzed by Western blot. Triton X-100 and radioimmunoprecipitation assay lysates were sonicated and clarified by centrifugation before immunoprecipitation. GST pull-downs were done as previously described using radiolabeled TNT reaction products and a GST-DTrc8 fusion protein (17, 21). Subcellular fractionation used a protocol adopted from Sever et al. (29). Washed cell pellets were suspended in 0.5 mL of Cell Disruption Buffer and passed 15 times through a 22-gauge needle. Lysates were cleared at 1,000 × g, and the supernatant was centrifuged at 100,000 × g for 20 min to recover membranes. The pellet was resuspended in 50 μL of Membrane Solubilization Buffer. The low-speed pellet was resuspended in 100 μL Nuclear Extraction Buffer and rotated for 4 h in the cold. Following centrifugation at 100,000 × g for 20 min in a TLA 120.1 rotor, the supernatant was recovered as the nuclear fraction.

Preparation of RNA and cDNA and performance of quantitative RT-PCR using hot-start conditions have been previously described (37, 38), except that the assays were carried out in a 10 μL volume using SYBR Green in ABI 7500 instrument with 9600 emulation. Primer sequences used for real-time PCR (5′ to 3′) are listed in Supplementary Materials and Methods. Primers specific for TRC8 have been described (19). C_t values were normalized by subtracting the geometric mean of four control genes: GAPDH, HMBS, UBC1, and HPRT1 (39). Affymetrix microarray data analysis methods are in Supplementary Materials and Methods. Polysome profiling was done on cultures of HEK293 FlpIn cells induced or not to express wild-type or mutant TRC8-HA for 24 h with 100 ng/mL doxycycline. Sucrose gradient separations of polysomes were done according to protocols from Johannes and Sarnow (40).

Multiple anti-TRC8 siRNAs (Ambion) were tested for efficacy, and Ambion ID 136327 (TRC8-Amb-27, sense strand sequence, GCUUGACGAUUAUGUCUACtt) was most effective. Cells were transfected with Lipofectamine 2000 according to the manufacturer's directions (Invitrogen, Inc.), and assays were done 72 h after treatment. The tet-inducible anti-TRC8 short hairpin RNA construct was prepared by annealing complementary oligonucleotides using the following sequences: forward, GATCCGCTTGACGATTATGTCTACTTCAAGAGAGTAGACATAATCGTCAAGCTTTTTTGGCCC; reverse, TCGAGGGCCAAAAAAGCTTGACGATTATGTCTACTCTCTTGAAGTAGACATAATCGTCAAGCG. Annealed oligos were ligated into BamHI/XhoI cleaved pSuperior.puro vector, propagated in Escherichia coli, sequence verified, and stably transfected into HEK293 FlpIn TRex cells using puromycin (3 μg/mL) and blasticidin selection (5 μg/mL).

Fly culture, genetic manipulations, and evaluation for genetic suppression were done as previously described (21) using the following strains for testing MPN domains: y[1]w[67c23];P{w[+mC] = lacW}eIF-3p40[k09003]/CyO for eIF3h (eIF-3p40); FCK-20 dp bw/SM5q-TM6b for MPND (CG4751); Df(2R)nap1/In(2LR)Gla, Dp(2;2)BG, wg[Gla-1] for eIF3f (CG8335); Df(2R)CB21/CyO;ry506 for U5 small nuclear ribonucleoprotein–specific factor (CG8877); Df(3R)110, ru[1]th[1]st[1]kni[ri-1]rn[roe-1]p[p]e[s]ca[1]/TM3, Sb[1] for eIF3f (CG9769); and Df(3R)tll-g, ca[1]/TM6B, Tb[1] for AMSH (CG2224). Mitotic recombinant clones were prepared by generating embryos containing the actin promoter-FRT-yellow-FRT-Gal4 cassette (AYG) along with heat shock–inducible FLP recombinase (FLP122), UasGFP, and UasDTrc8^7.1 (17). Resulting larvae were heat shocked for 20 min (24 h after egg laying) and then development was continued until adults eclosed. Patches of adult cuticle containing y⁻, DTrc8⁺ clones were identified by color, recorded, and analyzed using a Zeiss DSM940A digital scanning electron microscope. Flies were mounted on standard stubs with double-sided carbon SEM tape and a silver colloidal suspension. Samples were vacuum dried at room temperature and coated with 50 μm of gold using a sputter coater before examination.

The presence of a putative SSD (Supplementary Fig. S1; ref. 15) prompted us to determine if TRC8 was affected by sterols. We initially examined Chinese hamster ovary (CHO) cells stably transfected with a hemagglutinin (HA)–tagged, tet-inducible TRC8 (TRC8-HA-FlpIn; ref. 19), which is detectable at low levels in the absence of doxycycline because of leaky expression. When cells were grown in medium with 5% FCS containing abundant lipoproteins, TRC8 showed little change over time (Fig. 1A, lanes 1-4). However, under sterol-depleted conditions, there was substantial accumulation of both TRC8 (lanes 6-8) and the positive control HMGCR (29). Following cholesterol depletion in HEK293 cells containing TRC8-HA, the addition of sterols caused disappearance of both TRC8 and HMGCR with similar kinetics (Fig. 1B). In the presence of cycloheximide, the decay of wild-type TRC8 was substantially delayed in the absence of sterols (Fig. 1C), yielding a t_1/2 of 29 hours compared with 10 hours in the presence of sterols (Fig. 1D). Similar results were observed in CHO cells (Supplementary Fig. S2). Analysis of endogenous TRC8 protein, detected in membrane fractions prepared from sterol-starved HEK293 cells, yielded similar results. Compared with sterol-replete cells (Fig. 2A, lanes 1 and 2), endogenous TRC8 accumulated when sterols were removed, as did the positive control HMGCR (lanes 5, 6, 9, and 10). Quantification of these results indicated ∼3-fold more endogenous TRC8 in sterol-deprived cells after 24 hours (Fig. 2B). Real-time reverse transcription-PCR (RT-PCR) indicated that sterol starvation did not change TRC8 mRNA levels, whereas increases were observed for the SREBP target gene INSIG-1 as expected. Stable transfection of a tet-inducible knockdown construct (Supplementary Fig. S3A-C) resulted in specific loss of TRC8 following doxycycline treatment (Fig. 2A, lanes 3 and 4), and TRC8 in knockdown cells remained low following sterol depletion (lanes 7, 8, 11, and 12). Thus, endogenous TRC8 was increased by sterol deprivation without concurrent effects on mRNA levels. We have not determined whether mutating the putative SSD affects these responses.

Many E3 ubiquitin ligases are subject to autoubiquitylation, prompting us to examine TRC8-HA in FlpIn cells transfected with flag-tagged ubiquitin (flag-Ub). Anti-HA immunoprecipitations revealed high–molecular weight ubiquitin-conjugated products in cells expressing wild-type TRC8 (Fig. 2C). To determine if sterols influence the degree of TRC8 ubiquitylation, flag-Ub–transfected cells were lipid starved overnight and then sterols were added. An increase in ubiquitylated TRC8 was observed in cells 3 hours after sterol addition (Fig. 2D). Thus, TRC8 is ubiquitylated in vivo. We then asked if a mutation in the RING domain, C547S, C550S, affected the response to sterols. These results are shown in Fig. 1C and D. Of note, although the sterol responsiveness was absent in the RING-mutated protein, its half-life was reduced to 8 to 9 hours regardless of sterols. This may be the result of protein instability due to mutation. Thus, we cannot determine at the present time whether the effects of sterols on the half-life of the wild-type protein are RING dependent.

Some proteins with SSDs, such as SCAP and HMGCR, interact with the ER anchor proteins INSIG-1/INSIG-2 (27, 41-43). To explore this, we transfected TRC8-HA-FlpIn cells with myc-tagged INSIG-1 or INSIG-2 and did anti-HA (Fig. 3A) or reciprocal anti-myc immunoprecipitations (Supplementary Fig. S4A). In the absence of TRC8 induction, neither INSIG-1 nor INSIG-2 was detectable in the immunoprecipitation pellets (Fig. 3A, lanes 1 and 4). However, with the addition of doxycycline, both INSIG-1 and INSIG-2 were coprecipitated with TRC8 (lanes 2 and 5). This interaction was observed regardless of the sterol status in HEK293 cells (Supplementary Fig. S4B) and was not affected by mutation of INSIG residues K156R and K158R (Supplementary Fig. S4A, kk), which abrogate INSIG-1 ubiquitylation by gp78 (44, 45). Interestingly, the TRC8 RING mutant increased levels of INSIG-1/INSIG-2 in both lysates and immunoprecipitation pellets (Fig. 3A, lanes 8 and 11). Proteasome inhibition with MG132 increased the level of INSIG-1 in both lysates and immunoprecipitation pellets, whereas INSIG-2 levels were only marginally affected. INSIG binds SCAP and HMGCR through their respective SSDs, and this interaction is dependent on a tyrosine residue within the conserved YIYF motif present in the first transmembrane segment (27, 46). The presence of a similar motif and conserved tyrosine (Y32) in the first transmembrane of TRC8 (Fig. 3B; Supplementary Fig. S1) prompted us to test appropriate mutants for INSIG interaction. Compared with the wild-type control, Y32 mutations reduced the ability of TRC8 to coprecipitate INSIG-1 (Fig. 3C). These results are consistent with INSIG-1 binding being mediated, at least in part, by the SSD of TRC8.

Next, we cotransfected myc-INSIG-1 and flag ubiquitin into TRC8-FlpIn cells, then induced TRC8 for 24 hours, and did immunoprecipitations using anti-flag beads. We observed increased INSIG-1 in the immunoprecipitation pellet in wild-type TRC8-expressing cells accompanied by high–molecular weight, ubiquitin-conjugated INSIG-1 (Fig. 3D). In contrast, there was a marked decrease of ubiquitylated INSIG-1 in cells expressing the TRC8 RING mutant (lanes 11 and 12), although some high–molecular weight forms were detected. The INSIG-1–K156, K158R mutant was immunoprecipitated from wild-type TRC8-expressing cells at levels comparable with normal INSIG-1 (lane 8), suggesting that lysines other than K156 and K158 may be targeted for ubiquitylation by TRC8. Thus, we conclude that INSIG-1 and INSIG-2 physically interact with TRC8 and that TRC8 enhances ubiquitylation of INSIG-1 in a RING-dependent manner. We have not determined whether these effects also involve gp78.

In the presence of sterols, INSIG forms a complex in the ER with SCAP and the preSREBPs (27). To determine if TRC8 affected preSREBPs, we analyzed total lysates before and after TRC8-HA induction (Fig. 4A). Increasing amounts of doxycycline-induced TRC8 led to a dose-dependent reduction of both preSREBP-1 and preSREBP-2. In control cells, doxycycline treatment had no effect. A time course experiment showed that loss of preSREBP-1 was detectable after 6 hours of TRC8 induction and reduced levels of preSREBP-2 were evident after 18 hours (Supplementary Fig. S5).

We previously reported that growth and cell cycle inhibition by TRC8 were dependent on a ubiquitylation-competent RING domain (19). As shown in Fig. 4B, a RING deletion mutant (ΔRING) failed to downregulate SREBP-1/SREBP-2 (lanes 5-8) and similar results were obtained with the C547S;C550S RING mutant. In contrast, expression of the ubiquitylation-competent mutant S557A;R559A caused loss of the preSREBPs nearly as well as wild-type TRC8. An analysis of additional mutations (data not shown) consistently showed that both preSREBPs were sensitive only to ubiquitylation-competent TRC8. Furthermore, the loss of preSREBPs was completely blocked by MG132 (Fig. 4C), verifying that the effect of TRC8 is proteasome dependent. Interestingly, MG132-treated cells consistently contained less TRC8, suggesting that destabilization of ectopic TRC8-HA can still occur in the presence of proteasome inhibition.

To confirm that the preSREBP loss was not due to proteolytic activation and nuclear accumulation, and after verifying that the precursors were appropriately processed in HEK293 cells (Supplementary Fig. S6), nuclear and membrane fractions were analyzed after 24 hours of TRC8 induction (Fig. 4D). As expected, preSREBP-1/preSREBP-2 levels were reduced in the membrane fraction of TRC8-induced cells (lanes 7 and 8) compared with controls. Nuclear SREBPs did not increase on TRC8 expression; rather, the levels seemed reduced (lanes 7 and 8). These experiments were done with excess cholesterol to accumulate SREBP precursors; similar results were obtained in standard culture medium (Supplementary Fig. S7). In addition, levels of the cleaved COOH terminus of preSREBP-2, which are known to increase during processing, were reduced by TRC8 (Fig. 4A). Together, these results indicate that TRC8-induced loss of preSREBPs is proteasome dependent and not the consequence of enhanced processing.

To determine if the modulation of preSREBPs was physiologic, we used RNA interference to knock down endogenous TRC8. Transiently transfected small interfering RNAs (siRNA) targeting TRC8 variably increased preSREBP-2 levels compared with mock-transfected or scrambled siRNA controls following several days of culture (Fig. 5A). We reasoned that depletion of serum lipoproteins with effects on TRC8 might account for this variability. In addition, sterol starvation would induce SREBP processing. To separate these effects, we analyzed membrane and nuclear fractions from stably transfected TRC8 knockdown cells cultured with or without sterols (Fig. 5B). In medium containing abundant sterols, knockdown of TRC8 had little effect on either the precursor or nuclear forms of SREBP-2 (lanes 1-4). However, after 24 hours of sterol deprivation (lanes 5-8), TRC8 knockdown resulted in higher levels of preSREBP-2 (lanes 7 and 8) compared with controls (lanes 5 and 6). Similarly, nuclear fractions from TRC8 knockdown cells contained more processed SREBP-2. Furthermore, in sterol-starved cells, when SREBP processing was subsequently blocked acutely by the addition of excess sterols, TRC8 knockdown cells accumulated higher levels of preSREBP-2 (Fig. 5C, lanes 7 and 8). Thus, inhibition of endogenous TRC8 affected SREBPs in a manner opposite to its overexpression, particularly in cultures subjected to prior sterol deprivation.

We next examined the consequence of TRC8 expression on SREBP target genes. Our initial microarray analysis compared cells expressing wild-type TRC8 with the C547S;C550S RING mutation. Significant differential expression was observed for several hundred genes (GC-Robust Multi-array Average; see Supplementary Materials and Methods), including 21 associated with lipid metabolism and 16 known SREBP targets (Supplementary Table S1; ref. 31). All SREBP targets were downregulated by wild-type TRC8. Eight genes (SCD1, INSIG-1, LDLR, HMGCS, HMGCR, MVLK, DHCR7, and SQLE) were further analyzed by quantitative RT-PCR. After 24 hours of TRC8 induction, six were significantly repressed from controls (Fig. 5D), whereas the remaining two trended lower without achieving statistical significance (data not shown). FASN, SREBF1, and MAC30, although not detected by microarray analysis, were also analyzed and found to be suppressed after 48 or 72 hours of TRC8 induction. There was no apparent distinction between SREBP-1 and SREBP-2 target genes, as examples of each [e.g., SCD1 (SREBP-1a) and MVLK (SREBP-2)] were modulated by TRC8.

In contrast to overexpression, knockdown of endogenous TRC8 was associated with higher levels of SREBP target genes (Supplementary Fig. S8). However, in time course studies, this effect was only observed after cells were sterol starved for 12 to 24 hours (Fig. 5E; Supplementary Fig. S9). In control cells, acute sterol deprivation consistently led to SREBP target gene induction by 3 hours, typically returning to baseline by 12 hours. However, 12- and 24-hour sterol deprivation led to significantly higher expression of FASN and INSIG-1 in TRC8 knockdown cells (Fig. 5E). Similar results were obtained for LDLR (Supplementary Fig. S9), SCD1, and SREBP-1 (data not shown). Together, these observations indicate that accumulation of TRC8 following longer-term sterol deprivation negatively regulates preSREBP levels with corresponding effects on nucSREBPs and their target genes.

We previously reported that Drosophila Trc8 genetically and physically interacted with the MPN domain of the COP9 signalosome subunit Csn5/Jab1 (17). To pursue this observation, we generated mitotic recombinant clones that ectopically expressed myc-tagged DTrc8 within a background of normal cells (Supplementary Fig. S10). In the cuticle of adult flies, such clones were associated with aberrant bristle formation (Supplementary Fig. S11). Microchaetes were smaller and deformed, whereas macrochaetes were shortened. In addition, the bases of affected macrochaetes were thinner and had lost their characteristic fluting pattern. These effects are highly suggestive of Minute mutations, which primarily involve ribosomal protein genes and are associated with reduced translation efficiency (47).

We previously reported that loss-of-function mutations or hemizygous deletions of two MPN domain genes, Csn5 and Csn6, partially restored growth to DTrc8-inhibited tissues in flies (21). The Drosophila genome contains 10 proteins with MPN domains, leading us to test each in a genetic interaction screen for suppression of the DTrc8 phenotype. Besides Csn5 and Csn6, mutations in three other MPN domain genes were suppressive: eIF3f (two genes) and eIF3h. All three encode subunits of the eukaryotic translation initiation factor eIF3 (48), and hemizygous loss of each restored some growth to DTrc8-inhibited wings (Fig. 6A; Table 1).

In vitro glutathione S-transferase (GST) pull-down experiments confirmed that DTrc8 specifically interacted with eIF3f and eIF3h in addition to Csn5 (Fig. 6B). To validate the interaction in mammalian cells, TRC8-HA immunoprecipitations were done from HEK293 cells expressing wild-type or the C547S;C550S mutant. Both wild-type and RING mutant TRC8 coimmunoprecipitated eIF3 (Fig. 6C). When cells were transfected with flag-Ub and immunoprecipitated for eIF3, we observed high–molecular weight ubiquitylated products that were increased by expression of TRC8 (Fig. 6D). These results suggest that one or more components of the eIF3 complex may be ubiquitylated by TRC8.

eIF3 contains 13 subunits and represents part of a larger translation initiation complex. In addition, the availability of suitable antibodies against multiple components is limiting. Therefore, we used sucrose gradients to analyze polysome profiles from control and TRC8-inducible cells to determine if TRC8 affected translation (Fig. 6E). In three independent experiments, TRC8 induction suppressed the polysome profile, consistent with an inhibition of protein translation. In contrast, the ΔRING mutant had no demonstrable effect (data not shown). Taken together, these results are consistent with TRC8-inhibiting protein translation in a ubiquitylation-dependent manner through its interaction with eIF3 subunits. Further investigations will be required to determine which eIF3 subunits, or components of the larger complex, are affected and whether this is modulated by sterol levels.

The results reported here suggest that TRC8, an E3 ubiquitin ligase, may provide a regulatory link between pathways of lipid homeostasis and protein translation initiation (Fig. 7). The involvement of TRC8 in cholesterol homeostasis is based on its predicted SSD and sterol-dependent protein stability, its interaction with and ubiquitylation of INSIGs, and the regulation of preSREBPs with corresponding effects on SREBP target genes. Our results suggest that upregulation of TRC8 during prolonged periods of sterol deprivation may serve to dampen, or fine tune, the SREBP response.

TRC8 protein stability was found to vary in a sterol-dependent manner (Figs. 1 and 2), with kinetics similar to those of HMG-CoA reductase, which contains a SSD and is polyubiquitylated by gp78 on binding INSIG-1/INSIG-2 (27, 28). Both wild-type and RING mutant TRC8 proteins bound INSIG-1/INSIG-2 (Fig. 3) without apparent regard to sterol levels, at least in overexpression experiments, and TRC8 stimulated the polyubiquitylation of INSIG-1 in a RING-dependent manner. Mutation of conserved Tyr³² in the TRC8 SSD inhibited this interaction (Fig. 3C). Tyr³² seems homologous to Tyr²⁹⁸ of SCAP, mutations of which impair INSIG binding (27). Thus, the SSD of TRC8 is the likely interaction site for INSIG. These results are consistent with TRC8 mediating the ubiquitylation of INSIG-1 and possibly INSIG-2, in contrast to gp78, which we note, acts only on INSIG-1 (45). Polyubiquitylation of TRC8 itself was observed (Fig. 2C) and this increased with added sterols (Fig. 2D), consistent with its destabilization by excess cholesterol. Because TRC8 mRNA levels did not change on sterol starvation, these results suggest that a primary regulator of TRC8 activity is the cholesterol content of the ER membrane.

Ectopic TRC8 caused a substantial reduction in preSREBP levels, an effect that was RING dependent (Fig. 4). Moreover, preSREBP loss was completely blocked by proteasome inhibition, consistent with the ubiquitylation function of TRC8. However, it is currently unknown if the SREBP precursors are direct targets of this activity. Reduction of SREBP target gene expression showed that induced TRC8 inhibits these transcription factors (Fig. 5D; Supplementary Table S1). In contrast, TRC8 knockdown resulted in higher levels of precursor and nuclear SREBPs and increased target genes following chronic sterol deprivation (Fig. 5). Therefore, our results indicate that this ubiquitin ligase acts, at least in part, to down-modulate the SREBP pathway, especially after TRC8 accumulates during prolonged sterol deprivation.

As noted, the Drosophila genome encodes only a limited number of proteins with MPN domains. Among these are the COP9 signalosome subunits Csn5 and Csn6, the proteasome lid subunits Rpn8 and Rpn11, and the eIF3 subunits eIF3f and eIF3h. We previously reported that Drosophila Trc8 genetically and physically interacts with Csn5 and Csn6 but not Rpn8 or Rpn11 (21). The interaction with Csn5 was specifically shown to involve the MPN domain. Here, we show that Drosophila and human TRC8 proteins also interact with eIF3f and eIF3h (Fig. 6). Mitotic recombinant clones overexpressing DTrc8 resulted in a striking phenocopy of Minute mutations, which characteristically affect ribosomal protein genes and cause reduced protein translation (47). Reducing the copy number of eIF3f or eIF3h reduced the growth-inhibitory phenotype of DTrc8 overexpression, much like reduction of the COP9 signalosome subunits Csn5 and Csn6 (21). Genetic interaction between eIF3 subunits and DTrc8 was substantiated by showing direct physical binding using GST pull-downs. In fact, interactions between DTrc8 and eIF3f or eIF3h were stronger than observed with Csn5. In HEK293 cells, TRC8 and eIF3 coprecipitate. Overexpression of TRC8 and immunoprecipitation of the eIF3 complex was associated with an increase in high–molecular weight ubiquitylated products. In addition, TRC8 overexpression caused a reproducible reduction of the polysome profile in a RING-dependent manner, consistent with inhibition of protein translation. Further studies will be required to identify which eIF3 subunits or interacting factors are affected. Because downregulation of eIF3f or eIF3h suppresses DTrc8-induced growth inhibition in flies, we suspect that these components form a bridge by which TRC8 affects its targets. Similarly, we suspect that INSIG forms a bridge by which TRC8 can access the precursor SREBPs, although this has not been formally examined.

Taken together, these findings suggest that TRC8 is positioned to coordinate lipid homeostasis and protein translation during extended periods of cholesterol limitation (Fig. 7). We propose that when TRC8 levels are low, under normal growth conditions (Fig. 7A) and during early times of sterol deprivation (Fig. 7B), SREBP activation occurs with minimal hindrance. However, during periods of chronic sterol loss (Fig. 7C), TRC8 will accumulate, exerting a suppressive effect on growth (21), on protein translation, and on synthesis of excessive ER transmembrane proteins, including the SREBPs. Activation of TRC8 will then be reversed by the accumulation of sterols, which would occur through the continued action of nuclear SREBPs and by stabilized HMGCR protein. We note that the effects of TRC8 on SREBPs are never absolute; some precursor always remains following TRC8 expression (Fig. 4A-D), whereas nuclear forms and transcription targets are only modestly reduced (Figs. 4D and 5D). Thus, lipid biosynthesis will continue in the presence of TRC8, although it may be slowed, whereas the reduction of growth following TRC8 accumulation would provide ample opportunity for cells to restore necessary lipids and achieve homeostasis.

The need for cells to coordinately inhibit growth during times of lipid depletion is implicit because membrane biogenesis depends on available phospholipids and cholesterol. We suspect that TRC8 accumulation also serves to protect the ER from the stress of accelerating transmembrane protein synthesis during a time of cholesterol insufficiency. The fact that SREBP activation results in a positive feedback loop (31, 49) may necessitate such a countermeasure. Intriguingly, cross-talk between lipid and protein translation pathways may also be a function of the kidney cancer–associated FLCN (11). FLCN forms a complex with FNIP1 and AMPK, which regulates mTOR activity through TSC1/2. Loss of FLCN leads to higher mTOR activity under certain stress conditions, including serum starvation (11). Nearly 20 years ago, AMPK was identified as the kinase that phosphorylates and inhibits HMGCR and acetyl CoA carboxylase (50, 51), rate-limiting steps for cholesterol and fatty acid biosynthesis, respectively (52, 53). FLCN, via interaction with AMPK, is positioned to influence lipid and protein biosynthetic pathways at the level of HMGCR, acetyl CoA carboxylase, and mTOR. Our results suggest that TRC8, which causes another form of hereditary kidney cancer, links these same pathways at the level of the preSREBPs and eIF3. How interrupting these points of coordination affects kidney cancer development represents a new dimension in this disease and a new investigational direction.

No potential conflicts of interest were disclosed.

We thank K. Lorick and A.M. Weissman for numerous discussions and suggestions, E. Robbins, S. McNamara, J. Nair-Menon, M. Waldman, M. Berlinsky, and B. Marsh for technical assistance, data analysis and support. The facilities of the DNA Sequencing and cDNA Microarray Cores of the University of Colorado Cancer Center were used during this study.

Grant Support: NIH grant RO1 CA076035.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.