Ubiquitin Chain Elongation Enzyme Ufd2 Regulates a Subset of Doa10 Substrates

Abstract

Ufd2 is the founding member of E4 enzymes that are specifically involved in ubiquitin chain elongation but whose roles in proteolysis remain scarce. Here, using a genome-wide screen, we identified one cellular target of yeast Ufd2 as the membrane protein Pex29. The ubiquitin chains assembled on Pex29 in vivo by Ufd2 mainly contain Lys-48 linkages. We found that the ubiquitin-protein E3 ligase for overexpressed Pex29 is Doa10, which is known to be involved in protein quality control. Interestingly, not all Doa10 substrates are regulated by Ufd2, suggesting that E4 involvement is not specific to a particular E3, but may depend on the spatial arrangement of the E3-substrate interaction. Cells lacking UFD2 elicit an unfolded protein response, expanding the physiological function of Ufd2. Our results lead to novel insights into the biological role of Ufd2 and further underscore the significance of Ufd2 in proteolysis.

Introduction

Proteasomal substrates are often first covalently modified by ubiquitin (Ub),³ a highly conserved 76-residue protein, through the concerted actions of several enzymes, including a Ub-activating enzyme (E1), a Ub-conjugating enzyme (E2), and a Ub-protein ligase (E3) (1, 2). In a few cases, an additional enzyme E4 is required specifically for Ub chain extension (3).

The first E4 enzyme identified is yeast Ufd2, which belongs to a family of proteins containing a conserved U-box motif (∼70 amino acids) at their C termini (3, 4). Although it lacks the metal-chelating residue of a RING (Really Interesting New Gene) motif found in many E3s, the U-box structurally resembles the RING domain. Current knowledge of Ufd2 function mainly derives from the studies of artificial model substrates. Ufd2 was initially isolated in a screen for yeast mutants that stabilized Ub^V76-V-βgal, a chimeric protein bearing a non-cleavable Ub moiety that targets the β-galactosidase portion for ubiquitylation and proteolysis (5). The dissection of the proteolytic pathway that degrades the Ub^V76-V-βgal fusion led to the discovery of the Ub fusion degradation (UFD) pathway (5). The E3 enzyme, which acts on UFD substrates, is the HECT domain-containing protein Ufd4 (5). Deletion of UFD2 caused significant reduction of Ub^V76-V-βgal bearing more than two Ub molecules, suggesting a role for Ufd2 in Ub chain synthesis in vivo (5). It was later found that Ufd2 only elongates an existing, short oligo-Ub chain on model UFD substrates assembled by the Ufd4 E3 (4). Interestingly, although its function in the UFD pathway is U-box-dependent (6), mammalian Ufd2 also promotes the degradation of p73, a homolog of the tumor suppressor p53, in a U-box-independent manner with no apparent role in p73 ubiquitylation (7), suggesting that Ufd2 can work in proteolysis without its E4 activity.

The physiological function of Ufd2 is poorly understood. The loss of yeast Ufd2 was found to cause partial stabilization of several proteins, including hydroxymethylglutaryl-CoA reductase Hmg2, Deg1-Sec62, and the transcription factor Spt23 (6). However, it is unclear whether Ufd2 acts as an E4 enzyme for their degradation, because the requirement of the Ufd2 U-box and the ubiquitylation pattern of these proteins in wild-type versus ufd2Δ cells were not established. Ufd2 is highly conserved in eukaryotes. Mouse and Caenorhabditis elegans Ufd2 each act as the E4 enzyme in the degradation of the polyQ protein Ataxin-3 (8) and the myosin chaperone UNC-45 that is crucial for muscle formation (9). Although it is mostly concentrated in skeletal muscle and neuronal tissues, mouse Ufd2 is also expressed in heart, lung, liver, spleen, and ovary. Interestingly, mutations in mouse UFD2 lead to cardiac defects and neurodegeneration (10). Ufd2a, one of the two human Ufd2 homologues, is a candidate neuroblastoma tumor suppressor, because it is localized in a region closely associated with neuroblastoma and subject to mutations in tumors (11). Known E4s also include CHIP and p300, which are important for the ubiquitylation of the Pael receptor and the tumor suppressor p53, respectively (3). The multiubiquitylation activities of three E3s (i.e. Ufd4, CHN, and Parkin) that work with Ufd2 or CHIP appear to be poor in in vitro ubiquitylation reactions (4, 9, 12). Thus far, the involvement of E4 in proteolysis is largely deemed limited, because many E3s are seemingly capable of catalyzing multiubiquitylation and only a handful of substrates require E4s. Hence, the biological roles of E4s in proteolysis remain sparse.

Here, we employed a genome-wide screen for substrate discovery and identified the peroxisomal membrane protein Pex29 as a ubiquitylation target of yeast Ufd2. The U-box is essential for the involvement of Ufd2 in Pex29 degradation. Pex29 is regulated by the well known Ub-protein E3 ligase Doa10. Using quantitative mass spectrometry, we demonstrated that Ufd2 decorates Pex29 with a predominantly Lys-48-linked Ub chain in vivo, an effective targeting signal for the proteasome. Furthermore, consistent with the link to the Doa10 pathway, we found that deletion of UFD2 activates an unfolded protein response (UPR) that is likely caused by its inability to efficiently destroy misfolded proteins. Interestingly, not all Doa10 substrates are regulated by Ufd2, indicating that the Ufd2 requirement is not solely determined by E3s. We suspect that the involvement of Ufd2-like ligases in vivo may be much more prevalent than previously anticipated. We propose that the length of the ubiquitin chain assembled by E3s in vivo is influenced by the spatial positioning of the E3 and the specific substrate. Consequently, the growing Ub chain may be out of reach of the E3 or impeded for further extension by the same E3 due to another part of the substrate or to other substrate-binding proteins and requires assistance from a Ufd2-like Ub ligase.

EXPERIMENTAL PROCEDURES

Yeast Strains and Plasmids

The synthetic genetic array-compatible S288c strain Y8835 (MATa ura3::natR can1:: STE2pr-Sp_his5 lyp1 his3 1 leu2 0 met15 0) and isogenic ufd2 (UFD2::natR) cells were used for the genome-wide screen (13). Haploid strains bearing doa10Δ, hrd1Δ, pex10Δ, ufd4Δ, ubc4Δ, ubc7Δ, ire1Δ, rpn10Δ, or UFD2-TAP in the BY4741 background were obtained from Open Biosystems (Huntsville, AL). Yeast strain expressing 13Myc-tagged Doa10 was a gift from Dr. Jeffery Brodsky (University of Pittsburgh). Strains EJY130 (UFD2::LEU2) and ufd1-1 were obtained from Dr. Alex Varshavsky (California Institute of Technology). The yeast strain YHR207 (UFD2^U-boxΔ-TAP in the BY4741 background) was generated by replacing the Ufd2 sequences after amino acid 856 with a tandem affinity purification tag (TAP)-HIS3 cassette in BY4741 by homologous recombination. The replacement was confirmed by PCR and immunoblotting. GST-His₆-tagged PEX29 was subcloned to a vector plasmid pRS314 bearing the CUP1-promoter.

Cultures were grown in rich (YPD) or synthetic media containing standard ingredients and 2% glucose (SD medium), or 2% raffinose (SR medium), or 2% galactose (SG medium), or 2% raffinose plus 2% galactose (SRG medium), or 1% oleic acid. For the synthetic dosage lethality (SDL) screens, haploids were selected on synthetic dextrose medium (2% glucose; 1.7 g/liter yeast nitrogen base without ammonium sulfate and amino acids; 1 g/liter monosodium glutamic acid; 2 g/liter amino acid dropout mix lacking uracil, arginine, lysine, and histidine), supplemented with the following antibiotics: 100 mg/liter clonNAT (Werner BioAgents), 200 mg/liter Geneticin (Invitrogen), 50 mg/liter l-canavanine (Sigma), and 50 mg/liter S-(2-aminoethyl)-l-cysteine hydrochloride (Sigma) (13).

Genome-wide Synthetic Dosage Lethal Screen

The URA3-marked overexpression library and the screening procedures used in this study were previously described (13). Briefly, two isogenic synthetic genetic array-compatible strains, Y8835 and ufd2Δ, were mated to an ordered yeast array expressing 5200 unique galactose-inducible genes. The arrayed strains were subjected to diploid selection, sporulation, and two rounds of haploid selection, to give rise to an output array of duplicated colonies carrying the desirable natR-marked deletion and one unique galactose-inducible gene. The haploid arrays were finally replica-pinned onto media containing 2% glucose (uninduced condition) or 2% galactose (induced gene expression condition). The colonies on the glucose and galactose plates were photographed after 2 and 3 days of incubation, respectively. Overexpressed genes that uniquely caused a reduction in colony size of >20% in the ufd2Δ background were considered for downstream analysis.

Expression Shut-off Assay and Proteasome Inhibition Treatment

Yeast cells carrying plasmids expressing both hexahistidine (His₆)- and GST-tagged proteins from the GAL1 promoter were grown at 30 °C to an A₆₀₀ of ∼1 in SR-ura medium with auxotrophic supplements and 2% raffinose as the carbon source. Protein expression was induced with 2% galactose for 3 h and then repressed by the addition of 2% glucose. Cycloheximide (100 μg/ml) was also added to stop translation. Samples were withdrawn at the indicated time points and harvested by centrifugation. Proteins were extracted by glass bead lysis of cells, processed for immunoprecipitation with glutathione-Sepharose beads (Amersham Biosciences), and resolved by 7% SDS-PAGE. Immunoblots were probed with anti-His₆ antibody (Abcam) followed by detection with goat anti-mouse horseradish peroxidase conjugate using ECL reagents (Amersham Biosciences). The stable protein Rpt5 was used as a loading control in the expression shutoff experiments.

Proteasome inhibition was performed as described previously (14). Briefly, yeast cells were grown in synthetic media using proline as the only nitrogen source. SDS (0.003%) was added to the media 3 h before galactose induction. MG132 (75 μm, Biomol) was added 30 min before the addition of glucose. Samples were collected at indicated time points and processed as described above.

Detection of Ubiquitylated Substrates

Yeast cells expressing GAL1-regulated substrate GST-Pex29 and Myc or Ha-tagged Ub were grown to log phase in SR medium. Then 2% galactose was added to induce protein expression for 3 h. Cells were lysed with glass beads and immunoprecipitated with glutathione-Sepharose beads or Myc beads for 2 h at 4 °C. The immunoprecipitates were resolved by SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and immunoblotted with anti-Myc or anti-Ha antibody (Covance) as indicated, followed by treatment with anti-mouse horseradish peroxidase conjugates and ECL reagents. Non-ubiquitylated Pex29 was detected by anti-His₆ antibody.

Ub Chain Linkage Analysis

Ubiquitylated Pex29 species was isolated via two sequential purification steps. Specifically, 150 ml of yeast cells expressing GST-His₆-tagged Pex29 and Myc-tagged Ub were grown in galactose-containing medium. Cells were lysed by glass beads disruption in buffer L (50 mm sodium phosphate, pH 7.5, 300 mm NaCl, 10 mm imidazole, 1% Triton X-100). Cell lysates were first incubated with rProteinA beads at 4 °C for 30 min to eliminate nonspecific proteins sticking to the beads. Supernatants were mixed with nickel-nitrilotriacetic acid beads for 2 h. GST-His₆-Pex29 was eluted with 250 mm imidazole in lysis buffer and subsequently incubated with glutathione-Sepharose beads for 2 h. The immunoprecipitates were resolved by SDS-PAGE. Gel regions were cut out as indicated and prepared for Ub-AQUA analysis as published (15, 16). A fraction of immunoprecipitates was subjected to protein sequencing by mass spectrometry (17).

Cross-linking Assay

The protein-protein interactions were assessed by cross-linking assay described previously (18, 19). Briefly, cells were harvested by centrifugation and suspended in XL buffer containing 0.4 mg/ml dithiobis(succinimidyl-propionate) as previously described (19). Cells were incubated for 30 min at 30 °C. Cells were quenched with 1 m Tris (pH 6.8) and then lysed. Extracts were incubated with Dyna protein G beads (Invitrogen) coated with the indicated antibody for 3 h, and washed extensively with lysis buffer. The immunoprecipitates were washed, treated with SDS to release bound proteins, which were resolved on SDS-PAGE, transferred, and then probed with the indicated antibody.

RESULTS

A Functional Genomics Screen for Proteolytic Substrate Identification

To identify the physiological target of Ufd2, we employed an SDL screen. The premise of SDL is that increased levels of a protein may have little toxic effect on the growth of a wild-type strain but cause growth retardation or even lethality in a mutant strain (20). Previous studies suggest that this strategy can be adapted to isolate substrates of the Ub/proteasome system. For example, overexpression of several cell cycle proteins (e.g. Scc1 and Sic1) in proteolysis-deficient cells (e.g. ubr1 mutant and cdc34 mutant) led to extremely slow cell growth (20, 21).

The genome-wide SDL screen was carried out as previously described (13). More than 5000 yeast genes fused to both GST and His₆ tags were separately introduced into wild-type or ufd2Δ mutant cells. We compared the growth of wild-type versus ufd2Δ cells that expressed each of these yeast genes from a regulatable GAL1 promoter. The expression of these genes was repressed in dextrose-containing medium, but induced in galactose-containing medium. Haploids carrying these plasmids were pinned onto both SD (dextrose medium) and SG (galactose medium) plates (Fig. 1A). Following incubation for 2–3 days at 30 °C, the colony sizes on these plates were scored (13).

FIGURE 1.

Identification of Pex29 as a substrate of the Ufd2-Rad23/Dsk2 pathway. A, representative images from the plasmid overexpression screen. Each of ∼5280 yeast open reading frames regulated by the GAL1 promoter was transformed separately into wild-type Y8835 and ufd2Δ cells. Transformants were grown on media containing either glucose (SD, expression off) or galactose (SG, expression on). Each open reading frame is represented by two spots on the plate to reduce false positives. After 2 days for SD plates and 3 days for SG plates, colony sizes were scored for possible hits. Boxed spots are PEX29 on each plate. Strains and growth conditions are labeled at the top of each image. B, overexpression of Pex29 leads to slower growth in ufd2Δ or rad23Δ dsk2Δ double mutant cells. GST-His₆-tagged Pex29 isolated from the genome-wide screen was transformed into wild-type, or indicated mutants. These cells were grown to similar densities, and 5-fold serial dilutions were spotted onto SD or SG media. C, efficient degradation of Pex29 requires Ufd2, Rad23, and Dsk2. Wild-type (Tyr-8835) and mutant cells containing a GAL1 promoter-driven GST-His₆-Pex29 were first grown in raffinose-containing medium. Expression of Pex29 was induced by the addition of galactose. Samples were taken after promoter shutoff at the time points indicated and analyzed by anti-His₆ Western blots. Equal amounts of protein extracts were used and confirmed by blotting with anti-Rpt5 antibody in the expression shutoff experiments (lower panels). The identities of proteins are indicated on the left. The pull-down (IP) and Western blot (blot) are indicated to the right of the panels. D, quantification of the data in C for Pex29.

The Degradation of Pex29 Requires Ufd2 and Rad23/Dsk2

Following the initial screen, we performed a serial spotting assay to ascertain their toxicity in ufd2Δ cells (Fig. 1B and data not shown) and then used an expression shut-off assay to determine the stability of the encoded proteins in wild-type or ufd2Δ cells. We found that Pex29 was degraded in wild-type cells, but significantly stabilized in ufd2Δ cells (Fig. 1C), suggesting that Pex29 is degraded by the Ufd2 pathway.

Previously, we demonstrated that Ufd2 works with the Ub-binding proteins Rad23 and Dsk2, directly coupling substrate ubiquitylation to its delivery to the proteasome (22). Consistent with this notion, the overexpression of Pex29 led to growth retardation in cells lacking RAD23 and DSK2 (Fig. 1B), and Pex29 was degraded in a Rad23/Dsk2-dependent manner (Fig. 1, C and D).

Pex29 is a peroxisomal membrane protein that regulates the number and size of peroxisomes (23). Peroxisomes are membrane-enveloped organelles that carry out multiple metabolic functions linked to lipid metabolism, including the oxidation of fatty acids and the metabolism of nitric oxide and oxygen free radicals (24).

Ufd2 Is Important for the Assembly of a Lys-48-linked Ub Chain on Pex29

Because Pex29 degradation was not previously demonstrated, we first evaluated whether overexpressed Pex29 is degraded by the proteasome. We found that Pex29 was stabilized in the presence of the proteasome inhibitor MG132 (Fig. 2A, the quantification is included in supplemental Fig. S1), suggesting that Pex29 is a proteasomal substrate. Ufd2 can promote substrate degradation without its enzymatic activity in a U-box-independent manner (7). We assessed the requirement of the U-box for Pex29 degradation. The deletion of the U-box, which abolished its E4 activity (6), did not affect the protein level of endogenous Ufd2 appended with a TAP tag (Fig. 2B). Pex29 was stabilized in ufd2^U-boxΔ cells to the same extent as in the ufd2 null mutant, suggesting that its enzymatic activity is essential for Pex29 degradation (Fig. 2C).

FIGURE 2.

Ufd2 regulates Pex29 ubiquitylation and degradation. A, Pex29 is degraded by the proteasome. Wild-type yeast cells expressing Pex29 were treated with or without the proteasome inhibitor MG132 (14). Pex29 degradation was monitored as described in Fig. 1C. B, deletion of the U-box does not affect Ufd2 levels. The C-terminal U-box was replaced by a TAP tag. Wild-type and mutant Ufd2-TAP were detected by immunoblotting. C, Pex29 degradation is impaired in ufd2^U-boxΔ mutant cells. GST-tagged Pex29 was transformed into wild-type and ufd2 mutants. Pex29 degradation was assayed as described above. D, reduced Pex29 multiubiquitylation in ufd2Δ. GST-His₆-tagged Pex29 was co-transformed with the plasmid YEp105 expressing the CUP1-promoter regulated Myc-tagged Ub alleles into wild-type or ufd2Δ cells. Pex29 was precipitated with GST beads and analyzed by immunoblotting first with anti-Myc antibody and later with anti-His₆ antibody. Ubiquitylated and non-ubiquitylated Pex29 proteins are indicated on the left. Loading was ascertained by blotting with anti-Rpt5 antibody. E, Ufd2 promotes Lys-48-linked Ub chains in vivo. Pex29 was isolated from wild-type cells through a two-step purification (i.e. nickel-nitrilotriacetic acid, glutathione beads) and resolved by SDS-PAGE. Gel regions below and above ∼140 kDa were excised and prepared for Ub-AQUA analysis (15). The amounts (picomoles) of Lys-48- and Lys-63-linked chains in both regions are shown.

Ubiquitylated Pex29 species can be easily detected in wild-type cells. To further determine the role of Ufd2 in Pex29 ubiquitylation, we co-transformed the plasmid expressing the GST-His₆-tagged Pex29 with the plasmid bearing Myc-tagged Ub into wild-type or ufd2Δ mutant cells. Pex29 was efficiently multiubiquitylated in wild-type cells (Fig. 2D, lane 3). Interestingly, deletion of UFD2 reduced highly ubiquitylated Pex29 species, but increased lower forms of ubiquitylated Pex29 (Fig. 2D, lane 6). Ub-Ub linkage mediated through Lys-48, but not Lys-63 is deemed to be favorable for proteasome recognition and degradation (1, 2). To understand the nature of Ub chain conjugated onto Pex29, we employed the Ub mutants defective for Ub chain elongation, which are often employed in many similar studies. Mutation in Lys-48 markedly reduced Pex29 ubiquitylation. The results suggest the involvement of Lys-48-linked Ub chain (Fig. 2D).

To further ascertain the Ub-chain linkages involved, we purified Pex29 from wild-type cells. Samples were resolved by SDS-PAGE. The regions corresponding to less or more ubiquitylated Pex29 species were cut out and analyzed by quantitative mass spectrometry (15, 16, 25, 26). Interestingly, longer Pex29 Ub chains are composed of mainly Lys-48 linkages, and shorter Pex29 Ub chains are made of a mixture of Lys-63 and Lys-48 chains (Fig. 2E). Consistent with the analysis using Ub mutants, these data suggest that Ufd2 likely promotes Lys-48-linked Ub chain synthesis in vivo.

A caveat for these kind of studies is that other ubiquitylated proteins may be purified along with Pex29 and contributed to Ub counts. We then further analyzed Pex29 immunoprecipitates by mass spectrometry (17). Among the proteins identified, Ub (∼63%) and Pex29 (∼14%) are the most abundant species (data not shown). An amount of ∼21% proteins identified comprises several proteins (i.e. Ssa1, Ura2, Tdh3, Tef1, and Yef2) that are known to be often present in immunoprecipitation as non-ubiquitylated contaminants even under denaturing conditions (27). It remains formally possible that small amounts of other proteins present in the fraction may contribute to Ub counts, but less likely because they account for <2% total proteins in the immunoprecipitates.

Pex29 Degradation Is Doa10-dependent

Because Ufd2 seems to act as an E4 for Pex29 ubiquitylation, we looked for the E3 component for Pex29. As Pex29 overexpression caused growth retardation in ufd2Δ cells (Fig. 1B), we reasoned that increased levels of Pex29 would be toxic to cells lacking a relevant E3 enzyme. We considered four Ub ligases that may have a role in Pex29 ubiquitylation, a peroxisomal E3 protein Pex10 (28, 29), Ufd4, which is the E3 component for UFD substrates (4), and two endoplasmic reticulum (ER) membrane proteins Hrd1 and Doa10 (30), because newly synthesized peroxisomes are derived from the ER (24). Interestingly, Pex29 overexpression led to slow growth only in cells lacking DOA10 (Fig. 3A). Furthermore, Pex29 was stabilized in the doa10Δ mutant, but efficiently degraded in pex10Δ, ufd4Δ, or hrd1Δ cells (Fig. 3B and data not shown). We also detected the co-localization of Doa10 and Pex29 outside of the nucleus (supplemental Fig. S2) by immunostaining (31). Consistent with the involvement of Doa10 in Pex29 turnover, Pex29 was stabilized in cells lacking Ubc7, a Doa10-associated E2 enzyme (Fig. 3C) (31). Importantly, Pex29 ubiquitylation was diminished in doa10Δ cells, suggesting that the Ub ligase Doa10 is involved in Pex29 ubiquitylation (Fig. 3E).

FIGURE 3.

Pex29 degradation is mediated by the Doa10 pathway. A, Pex29 overexpression leads to growth retardation in doa10Δ cells. 5-fold serial dilutions of wild-type and various isogenic mutants bearing a plasmid for inducible PEX29 expression were plated as in Fig. 1B. B and C, Doa10 and Ubc7 are involved in Pex29 degradation. Pex29 stability was determined in wild-type, doa10Δ, ubc4Δ, and ubc7Δ cells. D, Ufd1 is required for Pex29 degradation. E, Doa10 is required for Pex29 ubiquitylation. Pex29 ubiquitylation patterns in wild-type or mutant cells were determined as in Fig. 2D. The p81 plasmid bearing the GAL1-promoter-regulated Ha-tagged Ub was used. F, elevated UPR activity in cells lacking UFD2. Levels of β-galactosidase activity in yeast cells harboring a plasmid that contains the LACZ gene under the control of the KAR2 promoter. The strain genotypes are indicated at the bottom. The ire1Δ strain defective in UPR signaling is included as a negative control. Values shown are the means derived from three measurements. Bars represent ±S.D.

The degradation of many Doa10 substrates requires a multisubunit Ufd1-Npl4-Cdc48 ATPase complex, which facilitates the presentation of ubiquitylated substrates to the proteasome (30, 32). Pex29 was stabilized in ufd1–1 mutant cells (Fig. 3D). Doa10 belongs to a branch of the protein quality control system termed ER-associated protein degradation (ERAD), which targets misfolded secretory proteins for ubiquitylation and degradation (30). One way that protein misfolding arises is when a protein is expressed in the absence or excess of its binding partners. The synthesis of yeast peroxisomal gene products is often induced under the stress conditions requiring peroxisomes (e.g. oleic acid-containing medium) (24). Here we overexpressed Pex29 under the non-inducible condition for peroxisomes. De novo synthesis of peroxisomes starts from the ER. Lacking other peroxisomal components, overexpressed Pex29 is likely misfolded and localized on the ER membrane, which is in line with the involvement of Doa10, a well established ERAD E3, in its degradation. Consistent with this notion, Pex29 ubiquitylation is significantly impaired in the presence of oleic acid (supplemental Fig. S3).

While this study was in progress, in-line with the Doa10-Ufd2 connection in Pex29 degradation, Ufd2 was recently shown to promote the assembly of Ub chains onto misfolded Ste6*, another Doa10-regulated ERAD substrate (33), that further validates the use of SDL readout as a means to identify enzyme substrates. In addition, our study reveals that Ufd2 is involved in the in vivo synthesis of Lys-48 Ub chain linkages (Fig. 2E) and that a subset of Doa10 substrates requires Ufd2 (see Fig. 5).

FIGURE 5.

Ufd2 is not required for all Doa10 or Cdc48 substrates. A and B, degradation of two commonly employed Doa10 substrates Deg1-βgal and Ubc6* in wild-type, ufd2Δ, or doa10Δ cells. The expression shut-off assays for these two substrates were preformed as previously described (32). C and D, degradation of two Cdc48 substrates Ha-tagged KHN and Sec61-2 in wild-type, or ufd2Δ. The protein stability assays were done as described above.

Because the primary function of Doa10 is to destroy misfolded proteins, we wondered whether ufd2Δ cells would exhibit the phenotypes associated with defects in protein quality control. Misfolded ER proteins trigger a signaling cascade, the UPR pathway, which includes transcriptional induction of UPR genes (e.g. ER chaperone Kar2) and ERAD (30). Cells lacking functional ERAD show constitutive activation of UPR. Interestingly, higher UPR activity (3-fold) was observed in ufd2Δ cells (Fig. 3F), supporting a role for Ufd2 in the destruction of misfolded proteins (33).

Pex29 Interacts with Doa10 and Ufd2

Next, we wanted to examine whether Pex29 binds Doa10 and Ufd2. To this end, we employed a cross-linking strategy, which was used successfully to detect the bindings of several ERAD substrates to E3s involved in ERAD (e.g. Hrd1 and Doa10) (18, 19). Specifically, we introduced the plasmid-bearing GST-His₆-tagged Pex29 into the Doa10-Myc strain, in which the chromosomal copy of DOA10 is linked to 13× Myc tag. We found that Pex29 binds Doa10 specifically (Fig. 4A). Similarly, Pex29 associates with Ufd2 (Fig. 4B), further supporting that overexpressed Pex29 is regulated by Doa10 and Ufd2. We failed to detect the binding between Doa10 and Ufd2 (Fig. 4C), suggesting that Doa10 and Ufd2 do not form a stable complex. This is consistent with the previous study of the UFD pathway in that the interaction between Ufd4 (E3) and Ufd2 (E4) was not detected (4).

FIGURE 4.

The interactions among Pex29, Doa10, and Ufd2. A and B, Pex29 binds Doa10 and Ufd2. Co-immunoprecipitation analysis of the interaction between GST-His₆-tagged Pex29 and Doa10–13Myc or Myc-tagged Ufd2 was done as previously described (18). Briefly, proteins were extracted from cells expressing galactose-inducible GST-His₆-tagged Pex29 and Myc-tagged Ufd2 or endogenous Doa10–13Myc and immunoprecipitated with various antibodies as indicated. Immunoprecipitates were resolved by SDS-PAGE and probed with indicated antibodies. The antibodies for immunoprecipitation (IP) and immunoblot (blot) are shown to the left of the panels. C, Doa10 and Ufd2 do not form a stable complex. The binding was done as in A.

Not All Doa10 Substrates Are Regulated by Ufd2

Ufd2 was also shown to promote Ub chain elongation onto the Doa10-regulated misfolded Ste6* mutant (33). We then determined whether other Doa10 substrates require Ufd2. We evaluated Deg1-βgal degradation in cells lacking Ufd2. The degradation of Deg1-βgal was unaltered in ufd2Δ cells (Fig. 5A), indicating that not all Doa10 substrates require Ufd2. An obvious difference among these substrates is their localization, whereas Ste6* and Pex29 are membrane localized, Deg1-βgal is soluble (32). We then assessed the requirement of Ufd2 in the degradation of a membrane localized Doa10 substrate, misfolded Ubc6*. Interestingly, the degradation of Ubc6* was not affected in ufd2Δ cells (Fig. 5B), showing that membrane localization is not sufficient to require Ufd2. Our data indicate that not all Doa10 substrates require Ufd2 for their degradation, suggesting that the relationship between an E3 and an E4 is not strictly linear.

Ufd2 was shown to bind Cdc48, which may help Ufd2 recognize its substrates (4). Thus far, all Ufd2 substrates seem to require Cdc48 for their degradation. We then evaluated whether other Cdc48 substrates are Ufd2 targets. Besides Ubc6*, we found that two Cdc48 substrates, Sec61-2 and KHN, are not regulated by Ufd2 (Fig. 5, C and D). Our results indicate that the requirement of E4 for degradation is substrate, but not E3 or Cdc48 specific.

DISCUSSION

A Ub-protein ligase E3, with the help of E1 and E2 enzymes, initiates Ub conjugation and often catalyzes further Ub chain elongation for degradation. In only a few cases, another enzyme E4 is required specifically for Ub chain extension. Because Ufd2 can also promote substrate degradation in the U-box- and ubiquitylation-independent manners (7), it remains to be seen for most of putative E4 substrates (e.g. Spt23 and Hmg2) (3, 6) whether the E4-ubiquitylation activity is required in vivo. Nevertheless, the involvement of E4 in proteolysis is so far deemed rather limited, because many E3s are quite proficient in multiubiquitylation at least in vitro. When, where, and why Ufd2-like E4s are needed remain to be elucidated. We demonstrated that Ufd2 promotes the ubiquitylation of Pex29 and facilitates its degradation in a U-box-dependent manner. Our study leads to novel insights into the biological roles of Ufd2-like Ub ligases.

First, Ufd2-like enzymes are likely involved in many degradation pathways despite the demonstrated multiubiquitylation activity of the responsible E3s. The E3s previously found to collaborate with Ufd2 (e.g. CHN and Ufd4) have only a few physiological substrates, and their multiubiquitylation activities are poor (4, 9). Without the addition of E4s, Doa10 seems to be quite competent in assembling multi-Ub chains in vitro, suggesting that Doa10 confers potent multiubiquitylation activity (31). Hence the involvement of Ufd2 in the Doa10 pathway is surprising, and further suggests that the efficient multiubiquitylation ability of E3s does not preclude the employment of Ufd2-like enzymes for their substrate ubiquitylation in vivo. Our data and recent work by Brodsky et al. (33) also establish Ufd2 as the E4 for some substrates of Doa10, which is one of the major E3s involved in protein quality control and has a wide range of substrates localized in the ER membrane, cytosol, and nucleus (30, 32). We therefore propose that Ufd2-like Ub ligases play far broader roles in proteolysis than previously anticipated.

Furthermore, E4s are not specifically tied to particular E3s, because Ufd2 is only required for a subset of Doa10 substrates (Figs. 1 and 5). Early pioneering work demonstrated that Ufd2 regulates the degradation of artificially designed Ub^V76-V-βgal but not Ub^V76-V-DHFR, both Ufd4 (E3) substrates (5). Our results extend this finding of differential involvement of Ufd2 for the targets of the same E3 to physiological ubiquitylation substrates, which is likely applicable to other E4s but, to our knowledge, has not been demonstrated. Because the relationship between an E3 and an E4 is not exclusive, whether a substrate requires E4 is hard to predict and needs to be examined on an individual basis.

Why do some substrates require a Ufd2-like Ub ligase? E3 is certainly not the sole determinant, because some Doa10 substrates do not require Ufd2 (Fig. 5). Then, why does Doa10 have difficulties in extending Ub chains on some of its substrates (e.g. Ste6* and Pex29)? Is there a common feature shared by Ufd2 substrates? The ubiquitylation efficiency may be influenced by several elements of a given substrate, including the degradation signal sequence that serves as the E3 binding site, lysine residues that are the site of Ub attachment, and the remainder of the protein. The degradation signal itself does not confer this Ufd2 specificity, because the same signal sequence (i.e. Ub) is used in Ub^V76-V-βgal and Ub^V76-V-DHFR (5). The sequences outside of the degradation signal and lysines alone do not determine the requirement for Ufd2 either, because the same β-galactosidase moiety (∼110 kDa) is used in Ub^V76-V-βgal and Deg1-βgal and makes up the bulk of the fusion proteins. Membrane localization is not a major determining factor, because Ufd2 is not restricted to either soluble or membrane proteins (e.g. Ub^V76-V-βgal, Ste6*). The oligomerization status of substrates is not a key determinant either, because differential requirements of Ufd2 were detected for monomeric Ubc6* (Fig. 5B) and Ste6* (33), oligomeric Ub^V76-V-βgal (4), and Deg1-βgal (Fig. 5A). An influential element for Ufd2 involvement could be the one or more ubiquitylation sites, the selection of which in vivo is poorly understood and subject to the regulation by multiple factors (26, 34, 35). Upon binding to the specific region of a substrate, an E3 likely surveys its surroundings for accessible lysine residues; in other words, the site to which E3 binds and the sequences/conformation around that site would affect the choice(s) of the ubiquitylation site. For a substrate like Sic1, its E3 (i.e. SCF) has numerous choices within its vicinity for the efficient assembly of multi-Ub chains and would not need an E4 for assistance (35). On the other hand, if an E3 encounters a limited selection of lysines, continuous Ub chain growth may become beyond its reach for further extension or later be hindered by the other part of the substrate or other substrate-binding proteins, in which case a Ufd2-like ligase may be required to extend the chain to sufficient length for rapid degradation.

The Ub chain assembled onto Pex29 by Ufd2 in vivo uses mainly Lys-48 linkage. Currently, due to both lower amounts of specific ubiquitylated protein and technical challenges in Ub chain analysis, only a few in vivo Ub chain linkages decorated on substrates are known. In contrast to earlier assumptions that Ub-protein ligases synthesized Ub chains with the same lysine linkage, these recent results, including ours, indicate that mixed chain linkages are used in vivo, in agreement with in vitro studies as well (15, 16, 25, 26). Interestingly, only Lys-48 and Lys-63 linkages were detected on Pex29 (Fig. 2E). Although both types of Ub chains can be recognized by the proteasome, Lys-48-linked chains are considered to be more commonly used for proteasome-mediated degradation and Lys-63-linked chains often lead to non-proteolytic signal transduction. Whether the mixing of Lys-63- and Lys-48-chains on Pex29 is accidental or biologically meaningful remains to be determined.

Ufd2 is involved in protein quality control (e.g. misfolded Ste6*, also Fig. 3F). Ufd2-mediated proteolysis allows efficient handling of misfolded proteins, the accumulation of which could cause protein aggregation and cellular toxicity. The involvement of Ufd2 may also provide more time and an opportunity for the protein folding machinery to rescue misfolded proteins from destruction, instead of an energy consuming and futile cycle of protein synthesis, folding and degradation.

The functioning of Ufd2-like enzymes in ubiquitylation may not always be essential for substrate degradation. However, without the help of Ufd2, inefficient ubiquitylation leads to hyperactive UPR, which have been implicated in neurodegenerative diseases and cancers. Understanding the functions and mechanisms of Ufd2-like Ub ligases will elucidate an important and yet poorly defined step in the Ub-proteasome system and may lead to effective therapeutic strategies.