Polyubiquitin Chains Linked by Lysine Residue 48 (K48) Selectively Target Oxidized Proteins In Vivo

Abstract

Aims: Ubiquitin is a highly conserved protein modifier that heavily accumulates during the oxidative stress response. Here, we investigated the role of the ubiquitination system, particularly at the linkage level, in the degradation of oxidized proteins. The function of ubiquitin in the removal of oxidized proteins remains elusive because of the wide range of potential targets and different roles that polyubiquitin chains play. Therefore, we describe in detail the dynamics of the K48 ubiquitin response as the canonical signal for protein degradation. We identified ubiquitin targets and defined the relationship between protein ubiquitination and oxidation during the stress response.

Results: Combining oxidized protein isolation, linkage-specific ubiquitination screens, and quantitative proteomics, we found that K48 ubiquitin accumulated at both the early and late phases of the stress response. We further showed that a fraction of oxidized proteins are conjugated with K48 ubiquitin. We identified ∼750 ubiquitinated proteins and ∼400 oxidized proteins that were modified during oxidative stress, and around half of which contain both modifications. These proteins were highly abundant and function in translation and energy metabolism.

Innovation and Conclusion: Our work showed for the first time that K48 ubiquitin modifies a large fraction of oxidized proteins, demonstrating that oxidized proteins can be targeted by the ubiquitin/proteasome system. We suggest that oxidized proteins that rapidly accumulate during stress are subsequently ubiquitinated and degraded during the late phase of the response. This delay between oxidation and ubiquitination may be necessary for reprogramming protein dynamics, restoring proteostasis, and resuming cell growth.

Introduction

The ubiquitin/proteasome system (UPS) is the most important intracellular proteolytic machinery in eukaryotes, responsible for the selective degradation of unneeded, misfolded, and damaged proteins (17, 31). In the UPS, the proteasome is the multicatalytic protease, while ubiquitin is the molecular marker that targets protein for degradation. The 26S proteasome holoenzyme is responsible for the recognition of ubiquitinated proteins as well as the unfolding and translocation of substrates for degradation. Ubiquitin, an evolutionarily conserved protein, is conjugated to select substrates by an enzymatic cascade composed of an E1 (ubiquitin activating enzyme), E2 (ubiquitin conjugating enzyme), and E3 (ubiquitin ligase) (17, 31). Cellular proteins can be mono-, multi-, or polyubiquitinated, with the latter triggering a multitude of functions depending on which ubiquitin lysine residue is used to link the ubiquitin molecules into a chain (36, 75). Polyubiquitin chains linked via lysine 48 (K48) are the most abundant and the canonical signal for protein degradation by the proteasome, while chains linked via lysine 63 (K63) play nondegradative roles (36, 73, 75). In response to environmental stressors, cells must control the balance of protein synthesis and degradation and we are beginning to explore the roles of distinctive ubiquitin linkages in these cellular contexts (5, 61).

Innovation.

The presence of oxidized proteins is a hallmark of cellular redox imbalance, and failure to remove these proteins can result in aggregation and toxicity. Here, we show that ubiquitin—a prominent posttranslational protein modifier—mediates the degradation of oxidized proteins and that ubiquitin linked by K48 is the relevant variant in this pathway. Our large-scale framework encompassing hundreds of oxidized and ubiquitinated proteins presents a time-resolved trajectory of the fate of proteins subject to oxidative modification, where many proteins are first oxidized during the stress response, followed by K48-ubiquitination and degradation once cell growth has resumed.

Oxidative stress is a prominent type of cellular stress generated by the imbalance of pro-oxidants against the cell's antioxidant capacity (25). When not deactivated properly, oxidizers can damage biomolecules, including DNA, lipids, and proteins. In the case of damaged proteins, very few mechanisms for protein repair exist (8–10), and therefore, protein degradation is the most efficient way to re-establish proteostasis. During oxidative stress, cells accumulate a large number of oxidatively damaged proteins that are primarily degraded by the proteasome (21, 23). However, when the UPS's capacity is overwhelmed or impaired by redox processes, the accumulation of damaged proteins can lead to toxic protein aggregates, which in turn can cause cell death and various human diseases (11, 24). Although the UPS's role in oxidative stress has been extensively studied [reviewed in Refs. (3, 57)], the molecular mechanism by which the cell recognizes and degrades oxidized proteins is not fully understood, especially regarding the requirement of ubiquitin. Comprehending how eukaryotic organisms cope with and regulate the degradation of oxidized proteins is of broad interdisciplinary interest and is fundamental to further understanding a variety of biological processes and stress-related diseases.

Contradictory results have sparked a debate on whether oxidized proteins are degraded in a ubiquitin-dependent or ubiquitin-independent manner. Authors have proposed that oxidized proteins are degraded by the 20S proteasome and the immunoproteasome in a ubiquitin-independent manner (22, 30, 32, 53, 60), while other groups have highlighted the importance of ubiquitin in the process (15, 37, 45, 56, 62). Still, it is a widely acknowledged that ubiquitin conjugates accumulate heavily during oxidative stress. Because of the diversity of ubiquitin functions, it is essential to dissect the roles of distinct ubiquitin linkages, particularly K48 ubiquitin, in the degradation of oxidized proteins. The studies available to date were conducted mainly using individual targets or reporter proteins without a systems-wide view. Therefore, there is a pressing need for comprehensive studies to investigate the role of protein ubiquitination during the stress response at the proteome level.

Here, we combined specific UPS antibodies and inhibitors to isolate oxidized and ubiquitinated proteins and analyzed them by high-resolution mass spectrometry. We computationally integrated the data to investigate the role of the ubiquitin in the yeast Saccharomyces cerevisiae subjected to oxidative stress. We showed that ubiquitin is required for the degradation of at least half of the pool of oxidized proteins and that oxidized proteins can be modified by K48 ubiquitin in vivo. Ubiquitinated and oxidized proteins share functional and sequence features, which might be predictive of oxidized proteins prone to ubiquitination and directed to the proteasome.

Results

Complete degradation of oxidized proteins requires ubiquitin and the proteasome

To investigate the role of ubiquitin in the removal of oxidized proteins, we developed a system to analyze individual components of the UPS at different stages of the oxidative stress response. First, we cultivated yeast cells in minimum proline dextrose (MPD)-sodium dodecyl sulfate (SDS) medium to permeabilize cells to the UPS and translation inhibitors (40, 51). Cultivating cells in this condition also overcame the need for genetic manipulation, avoiding further phenotypic alterations. Cells were then preincubated in the presence or absence of inhibitors of the proteasome (MG-132), protein ubiquitination (PYR-41), and protein synthesis (cycloheximide [CHX]), followed by treatment with 0.6 mM hydrogen peroxide (H₂O₂) for 45 min at 30°C (Fig. 1A). This treatment was sufficient to induce accumulation of oxidized proteins without compromising cellular viability (Supplementary Fig. S1A, B). After stress induction, cells were allowed to recover for up to 8 h in fresh media (Fig. 1A). This workflow enabled us to perform a comprehensive characterization of the dynamics of protein ubiquitination and oxidation and showed that the ubiquitination system and the proteasome are important for the removal of oxidized proteins (Fig. 1B).

FIG. 1.

Ubiquitin is required for the degradation of oxidized proteins during oxidative stress. (A) Schematic overview of the workflow used in this analysis. Yeast was permeabilized by incubation with 0.003% of SDS in MPD medium (40). When specified, cells were preincubated for 15 min with inhibitors before stress induction. Oxidative stress was induced for 45 min with 0.6 mM H₂O₂, and cells were transferred to fresh media and allowed to recover for 8 h. The equivalent concentrations of inhibitors were added to the recovery medium where applicable. (B) Whole-cell extract was immunoblotted (IB) and probed using distinct antibodies to monitor the dynamics of protein oxidation through carbonyl formation (anti-DNP), pan-ubiquitin, K48 ubiquitin, and the 20S proteasome (20SPT). MG-132 (75 μM) was used to inhibit proteasome activity, PYR-41 (75 μM) to inhibit E1 ubiquitin activation enzyme, and CHX (50 μg/mL) to inhibit translation. Anti-actin was used as a loading control. (C) Whole-cell extract from yeast cells was immunoblotted and probed using a K63 ubiquitin antibody. Anti-GAPDH was used as a loading control. (D) Proteasomal activity in cell lysates from yeast treated with the indicated inhibitors. Activity was measured following incubation of cellular lysate with 100 μM suc-LLVY-AMC fluorogenic substrate. Error bars, standard deviation. CHX, cycloheximide; DNP, dinitrophenyl; H₂O₂, hydrogen peroxide; MPD, minimum proline dextrose; SDS, sodium dodecyl sulfate; Suc-LLVY-AMC, succinyl-Leu-Leu-Val-Tyr-7-amino-4-methylcoumarin.

Global ubiquitination and oxidation presented different dynamics in response to oxidative stress induced by H₂O₂. To study protein oxidation, we derivatized proteins in the cellular lysate with 2,4-dinitrophenylhydrazine (DNPH), a carbonyl probe used as proxy for global protein oxidation (39). We observed that oxidized proteins quickly accumulated in response to stress, and their levels started to decrease after 2 h of recovery (Fig. 1B, top left panel). The proteasome is key for the removal of oxidized proteins [best reviewed in Ref. (31)], and to confirm its role, we preincubated cells with the proteasome inhibitor MG-132. Proteasome inhibition was acute and long-lasting (Supplementary Fig. S1C), and completely blocked the removal of oxidized proteins even after 8 h of recovery (Fig. 1B, MG-132). Because carbonylation is an irreversible protein modification (39, 42, 47), MG-132 stabilized oxidized proteins by inhibiting protein degradation and not the reduction of oxidation. This modification is also not reversed by thiol reducing agents such as dithiothreitol (DTT) (Supplementary Fig. S1D). Longer exposure to MG-132 also reduced the abundance of the proteasome itself (Fig. 1B). Notably, incubation with a low concentration of SDS (0.003%), required for induction of cellular permeability, did not significantly affect the abundance of oxidized proteins (Supplementary Fig. S1E, F), the levels of ubiquitination (Supplementary Fig. S1F), or the activity of deubiquitinating enzymes (Supplementary Fig. S1G). Moreover, while it is known that ionic detergents can stimulate the proteasome activity (54), the presence SDS only minimally enhanced the proteasome activity after H₂O₂ treatment (Fig. 1B and Supplementary Fig. S1H), collectively demonstrating that H₂O₂ and not SDS is responsible for the modulation of the UPS. Thus, our results confirm that the proteasome activity is necessary for the removal of oxidized protein independent of our culture conditions.

Next, we showed that accumulation of ubiquitinated proteins peaked immediately after stress induction, rapidly decreased to an intermediate level, and dropped back to steady-state after 8 h of recovery (Fig. 1B, no inhibitor). Since global ubiquitination can now be deconvoluted into individual linkages using selective antibodies, we analyzed the dynamics of K11, K48, and K63 ubiquitins, the three most abundant linkages in yeast, during the stress response (73). Although others have demonstrated that K11 ubiquitin linkages are abundant in yeast (73), no significant K11 accumulation was detected in response to oxidative stress (Supplementary Fig. S1I). Instead, the initial burst of protein ubiquitination was a mixture of K63 and K48 ubiquitin linkages. K63 ubiquitin linkages sharply accumulated in response to stress and quickly decreased during the recovery phase (Fig. 1C). We previously showed that K63 ubiquitin regulates translation in response to stress and is not involved in proteasomal degradation (5, 61). Therefore, we focused our analysis on K48 ubiquitin, which is the canonical signal for protein degradation and has not been extensively explored in the context of the oxidative stress response.

Proteins conjugated with K48 ubiquitin accumulated in different phases of the stress response. We observed an early increase in K48 ubiquitinated targets after stress induction and a more pronounced increase after 2 h in the recovery phase (Fig. 1B). When the proteasome was inhibited by MG-132, we observed an extensive accumulation of K48 ubiquitinated targets without a subsequent return to steady-state levels. This result confirmed that K48 ubiquitin plays a major role in proteasomal degradation under these conditions (Fig. 1B, MG-132). Importantly, the accumulation of K48 after 2 h in the recovery medium coincided with a decline in abundance of oxidized proteins, suggesting that ubiquitin is involved in their degradation. To investigate the role of ubiquitin in the dynamics of oxidized proteins, we incubated cells with PYR-41, an inhibitor of the E1 activating enzyme, which blocks the entire ubiquitination cascade at its initial step (74). Inhibition of the E1 decreased and slowed the accumulation of global and K48 ubiquitin targets, partially preventing the removal of oxidized proteins (Fig. 1B, PYR-41). Even after 8 h in the recovery phase, we still detected approximately twice as many oxidized proteins as found in the control without inhibitors (Fig. 1B, PYR-41). Importantly, PYR-41 did not impair the proteasome activity or abundance (Fig. 1B, D), suggesting that lack of ubiquitination itself was responsible for the stabilization of oxidized proteins. If PYR-41 fully inhibited protein ubiquitination, it stands to reason that we would observe an even higher stabilization of oxidized proteins.

We next observed that inhibition of protein synthesis delayed the accumulation of oxidized proteins in response to stress. We reasoned that the delayed accumulation of oxidized proteins in cells treated with PYR-41 could be related to protein synthesis regulation. Medicherla and colleagues showed that newly synthesized proteins comprised a large fraction of the targets affected by oxygen radicals (45). Therefore, inhibition of translation could affect the dynamics of protein oxidation and ubiquitination during the stress response and could lead to depletion of ubiquitin itself (26). In the presence of CHX, oxidized proteins as well as ubiquitin conjugates accumulated gradually in response to stress, indicating that protein synthesis is an important source of oxidation targets (Fig. 1B, CHX). Still, de novo protein synthesis was not required for the removal of oxidized proteins and the levels of oxidized proteins still fell during the recovery phase (Fig. 1B, CHX). We tested whether the levels of oxidized proteins could have decreased as a result of dilution through cell division. Our results showed that CHX treatment severely impacted cell growth (Supplementary Fig. S1J), indicating that reduction in oxidized proteins was likely mediated by protein degradation rather than cell division. Collectively, our results identified important correlations between the accumulation and removal of oxidized and ubiquitinated targets during the oxidative stress response. Our data strongly suggest that protein ubiquitination plays an important role in the proper removal of oxidized proteins, most likely via the formation of K48 linkages.

Oxidized proteins are modified by K48 polyubiquitin chains

To investigate our hypothesis on the role of K48 ubiquitin in the degradation of oxidized proteins, we sought to show that a fraction of oxidized proteins was modified by K48 polyubiquitin chains in vivo. The literature has provided examples of oxidized (46, 47) and ubiquitinated proteins (52, 65, 66), but no work has examined the ubiquitination of oxidized proteins at the ubiquitin linkage level. The resolution of our immunoblots highlighted that K48 ubiquitin and protein oxidation were distributed throughout a wide range of molecular weights, suggesting that a suite of proteins could bear both modifications. For this line of experimentation, we cultivated cells in synthetic dextrose (SD) medium and selected three representative time points from our previous analysis to investigate the presence of both modifications on intracellular proteins. The conditions examined were as follows: (i) cells before the H₂O₂ treatment (Control); (ii) cells after 45 min of H₂O₂ stress induction (Stress); and (iii) cells that were allowed to recover from stress for 2 h in fresh SD medium (Recovery). We chose 2 h of recovery because it is the stress phase that occurred immediately before the disappearance of oxidized proteins, thereby making it the most likely time at which protein species that are both oxidized and K48 ubiquitinated would be detectable. After cell growth and stress induction, proteins were extracted and derivatized with DNPH for oxidation analysis under denaturing conditions. Next, ubiquitinated proteins were isolated using the ubiquitin trap tandem ubiquitin binding entities (TUBE) system (LifeSensors) and blotted for oxidation or ubiquitination (Scheme in Fig. 2A). The eluted fraction displayed similar variation in abundance as that found in samples without enrichment (input) and only a small number of ubiquitin conjugates were detected in the unbound fraction, highlighting the efficiency of the isolation method (Fig. 2B). Furthermore, we identified oxidized proteins in both the eluted and the unbound fractions following TUBE enrichment (Fig. 2B). This result is in agreement with our initial findings (Fig. 1B) that a fraction of ubiquitinated proteins are also oxidized during oxidative stress. To confirm our results, we performed the complementary assay, in which, following DNPH derivatization, oxidized proteins were isolated by immunoprecipitation and probed for ubiquitin via Western blot (Fig. 2C). For this experiment, we saturated beads with cellular lysate to isolate similar amounts of oxidized proteins from each one of the three conditions (Control, Stress, and Recovery). Loading similar amounts of oxidized proteins would allow us to detect changes in the ubiquitin levels according to the stress phase. We then blotted the isolated proteins with antibodies against total-ubiquitin, K63-, K11-, and K48-specific ubiquitin linkages. Oxidized proteins are heavily labeled with ubiquitin, particularly during the stress and recovery phases in the high-molecular-weight range (Fig. 2C). Even after prolonged exposure of the blots, we did not detect K63 or K11 ubiquitin linkages conjugated to oxidized proteins (Fig. 2C). Still, oxidized proteins were detected containing K48 ubiquitin chains (Fig. 2C). The K48 ubiquitin modification was abundant in the stress and the recovery phases. The highest abundance of K48 ubiquitin was detected after 2 h in the recovery phase (Fig. 2C), which correlated with the second wave of K48 ubiquitination (Fig. 1B), likely directing oxidized proteins for selective degradation by the proteasome.

FIG. 2.

Oxidized proteins are K48 polyubiquitinated. (A) Schematic overview of the workflow selected for oxidized and ubiquitinated protein enrichment and mass spectrometry analysis. Untreated (Control), H₂O₂ stress (Stress), and cells that recovered in fresh media for 2 h (Recovery) were used as representative samples of different stages of oxidative stress. The TUBE system was used to isolate ubiquitinated proteins, anti-DNP was used to isolate oxidized (carbonylated) proteins for blots, and biotin hydrazide was used to isolate oxidized proteins for mass spectrometry. (B) Ubiquitinated and (C) oxidized proteins were isolated by TUBE affinity purification and by anti-DNP immunoprecipitation, respectively. The input, the TUBE/DNP unbound and eluted fractions were immunoblotted against pan-ubiquitin (Ub) and oxidized proteins (DNP). In (C), beads were saturated with cell lysate to normalize the amount of oxidized proteins retrieved from each condition. Samples were immunoblotted against pan-ubiquitin, K63, K11, and K48. Anti-DNP blotting was used as loading control. *Immunoblotting for high-molecular-weight range (>50 kDa) was subject to longer exposure times. DNPH, 2,4-dinitrophenylhydrazine; LC-MS/MS, liquid chromatography with tandem mass spectrometry; TUBE, tandem ubiquitin binding entities.

Quantitative proteomics reveals shared targets of ubiquitination and oxidation

Our results showed that oxidized proteins can be modified by K48 ubiquitin in response to stress. Still, we have very little information on the identity of these targets and their physiological role in the cell. Since we induced oxidative stress exogenously with H₂O₂ treatment, and protein oxidation is an unspecific chemical modification, we hypothesized that cell wall, cell membrane, redox-related, and highly abundant proteins would be the preferential targets of oxidation. To identify the simultaneous targets of ubiquitination and oxidation, we applied a quantitative proteomic workflow to measure proteins that were modified during the stress response. Because both K48 and K63 ubiquitins accumulate in response to oxidative stress, we used the yeast strain GMS413 to focus on the targets of K48 ubiquitin. In this strain, ubiquitin's lysine residue 63 is mutated to arginine (K63R), thereby preventing K63 chain extension. As a result, this strain accumulates mostly K48 ubiquitin chains in response to stress. As before, cells were treated and collected (i) after H₂O₂ treatment and (ii) after 2 h of recovery in fresh media, the period during which oxidation and ubiquitin are detectable prior degradation (Figs. 1B and 2C). Ubiquitin conjugates were isolated using the TUBE system (Supplementary Fig. S2A) and oxidized proteins were isolated after carbonyl derivatization with biotin-hydrazide (42) (Supplementary Fig. S2B). This method was used as an alternative for isolating oxidized proteins because unlike dinitrophenyl (DNP)-derivatization, it can be used to produce mass spectrometry-compatible samples, since it eliminates the requirement for a high concentration of detergent (Supplementary Fig. S2B). Using mass spectrometry-based proteomics, we identified 1141 proteins from the ubiquitin enrichment experiment and 690 from the oxidized protein enrichment experiment. To quantify proteins that were enriched because of their modification status over experimental contaminants, we used a stable isotopic-labeled amino acid in cell culture (SILAC) approach (stable isotope labeling by amino acids in cell culture) (50). Here, we isotopically labeled proteins from the stress and the recovery phase and compared them with the untreated condition, which was unlabeled. In an SILAC experiment, contaminant proteins can be statistically removed from our data set because they will present with similar abundance ratios when compared across samples. Using this method, we defined a data set of 739 ubiquitinated and 411 oxidized proteins that provided a quantitative framework for follow-up analysis (Fig. 3 and Supplementary Table S1).

FIG. 3.

Mass spectrometry identified proteins that are oxidized and ubiquitinated. (A) Matrix depicting the logarithmic ratio of expression changes for ubiquitin and oxidation during stress and recovery phases relative to the untreated cells. Yeast strain GMS413 containing the K63R ubiquitin mutation was cultivated in SILAC medium labeled with heavy isotopes for lysine and arginine (Control) and light isotopes for Stress and Recovery. Each row denotes one quantified protein. Hierarchical clustering was performed in Perseus, and the four largest clusters were analyzed for function enrichment using the DAVID annotation tool (FDR <0.02) (28). (B) Venn diagram displaying the overlap between populations of ubiquitinated and oxidized proteins. (C) Immunoblotting of selected proteins after ubiquitin (TUBE) and oxidation enrichment (biotin-hydrazide). Representative proteins Htb2 (14 kDa) from cluster B and Rpt1 (52 kDa) from cluster D were analyzed by immunoblot during the stress response. FDR, false discovery rate; Htb2, histone H2B; SILAC, stable isotopic-labeled amino acid in cell culture.

To infer biological relevance, we clustered our data into a protein expression matrix (Fig. 3A) and searched for functional enrichment using the DAVID annotation tool (28). Among the four largest clusters, clusters A and B were significantly enriched for membrane proteins, and proteins containing transmembrane domains, which supported our initial hypothesis. Ribosomal proteins were highly represented in cluster C, while cluster D was composed of proteasome components and nucleotide binding proteins (Fig. 3A). This analysis suggested that large, abundant macromolecular complexes were modified by ubiquitination and oxidation. Our data also revealed that both modifications share a large core of common targets (Fig. 3A, B), where ∼80% of the oxidized proteins were identified in the ubiquitin data set. To validate the presence of both modifications on a protein target, we selected one protein from cluster B (histone H2B [Htb2]) and one protein from cluster D (Rpt1). Htb2 is known to be K48 ubiquitinated (18, 65) and was previously identified as a target of oxidation (47). Ubiquitination of oxidized Htb2 increased under stress, and conjugated DNP was detected during all phases of the stress response (Fig. 3C). Rpt1 is a proteasome 19S regulatory subunit known to be ubiquitinated (44). We detected more ubiquitinated Rpt1 during the stress phase and intense oxidation during the recovery phase (Fig. 3C). Both proteins exhibited a laddering pattern characteristic of protein ubiquitination (Fig. 3C).

To refine our analysis even further, we applied the statistical tool EBprot, which uses protein- and peptide-level information to evaluate differential protein expression (35). Analysis of differential expression works under the assumption that the majority of the proteome remains unaltered in response to a perturbation. Still, both oxidation and ubiquitination increase under stress, and therefore, enrichment will provide groups of proteins that are mostly upregulated. Using this probabilistic framework that rewards significant differential expression when multiple peptides are quantified per protein, we identified significant changes in ubiquitination and oxidation following stress at a 5% false discovery rate (FDR) (see Materials and Methods for details). Our analyses returned a total of 413 proteins significantly modified by ubiquitin, 270 by oxidation, and an overlap of approximately half of the protein data set (Fig. 4, Supplementary Fig. S3A, Supplementary Tables S2, and S3). These results corroborated our previous data suggesting that a large fraction of proteins that are oxidation targets can also be targets for ubiquitination. To further understand this group of modified proteins, we evaluated their protein association networks and their functional enrichment. A high-confidence protein interaction map was generated (interaction score >0.7) and the four main overlapping protein complexes were clustered using the ClusterONE tool (Figs. 5 and 6). The largest interaction cluster among ubiquitinated (Fig. 5) and oxidized proteins (Fig. 6) is composed of proteins involved in translation. Proteasome proteins are also shared between both modifications. Individual functions such as ATP binding and vacuole/Golgi are represented in the interaction network for ubiquitinated proteins (Fig. 5), while metabolic proteins and tRNA synthetases are represented among the oxidized proteins (Fig. 6). Gene ontology analysis on the EBprot data set revealed that ubiquitinated proteins were significantly (FDR <2%) involved in ion transport, proteasome, and translation during the stress phase (Supplementary Fig. S3B). During the recovery phase, we identified significant enrichment in categories related to the Golgi apparatus, transmembrane proteins, and vacuoles (Supplementary Fig. S3B). Interestingly, the functional categories significantly associated with the oxidized proteins did not change between the stress and the recovery phase (Supplementary Fig. S3B). This is in agreement with our initial characterization that oxidized proteins accumulated mostly during stress induction, but K48 ubiquitinated targets accumulated in two different time points (Fig. 1B). Proteins that were differentially oxidized were annotated as involved in nutrient metabolism, mitochondria, translation, and protein folding (Fig. 6).

FIG. 4.

Protein interaction network for proteins ubiquitinated during oxidative stress. EBprot analysis of differentially expressed ubiquitinated (left) and oxidized proteins (right). (A) Identification of the null component to be fitted (green) by EBprot algorithm. Dotted lines specify the range of ratios for the nondifferentially expressed proteins. (B) Histogram of fitted mixture model of peptide ratio distribution for differentially (solid red lines) and nondifferentially modified peptides. (C) Probability scores from peptide-level EBprot scores versus the median of logarithmic peptide ratios.

FIG. 5.

Protein interaction network for proteins ubiquitinated during oxidative stress. Interaction maps for ubiquitinated protein data sets determined by EBprot association analysis were determined by STRING and visualized with Cytoscape (58). The top four clusters of protein association (translation, UPS, ATP binding/mitochondria, vacuole/Golgi apparatus) were determined by the ClusterONE tool (49).

FIG. 6.

Protein interaction network for proteins oxidized during oxidative stress. Interaction maps for oxidized protein data sets determined by EBprot association analysis were determined by STRING and visualized with Cytoscape (58). The top four clusters of protein association (translation, metabolism, tRNA synthetase, proteasome) were determined by the ClusterONE tool (49).

We next identified 100 proteins whose peptides contained a diglycyl (GG) lysine modification (Supplementary Table S4). A GG modification is the remnant of a ubiquitin chain that was cleaved by trypsin during the digestion step of sample preparation for mass spectrometry and is used to precisely identify ubiquitination sites on peptides (52). Although the detection of GG sites is usually preceded by a GG-peptide enrichment due to the low comparative abundance to unmodified peptides (33), abundant peptides can be detected even without enrichment. First, we analyzed the GG peptides generated by the degradation of the ubiquitin chains themselves. Each particular ubiquitin chain generates signature GG peptides surrounding the ubiquitin link and identifies the types of chains present in a sample (34). This analysis revealed that K48, K11, and K6 linkages were detected but not K63, confirming that the K63R mutation prevented the generation of K63 ubiquitin chains (Supplementary Table S4). Analyzing our mass spectrometry data, we identified a handful of GG-containing peptides in the samples that were enriched for oxidized proteins (Supplementary Table S4). Among them, we identified proteins involved in translation and metabolic proteins, in agreement with our gene ontology (Fig. 3), our network mapping (Fig. 5), and our hypothesis that abundant proteins are oxidized and serve as a target for ubiquitination (Supplementary Table S4). Due to the absence of K63 chains, these proteins are likely en route to the proteasome for degradation.

Co-occurrence of oxidization and ubiquitination affects highly abundant proteins

To test whether abundant proteins are targets of protein ubiquitination and oxidation, we developed the yeast protein annotation database (yPAD), an R script that statistically tests for the enrichment of multiple protein sequence features among different protein groups (Supplementary Fig. S4). Using EBprot high-confidence list of proteins that are ubiquitinated or oxidized (Supplementary Tables S2 and S3), we tested for features related to protein abundance, disorder, degradation signals, and protein stability (Supplementary Tables S5 and S6). Since mass spectrometry analysis favors the identification of abundant proteins, protein groups (oxidized, ubiquitinated) were compared only with proteins identified by mass spectrometry (universe) and not with the entire yeast proteome. This analysis revealed that ubiquitinated proteins differ from the mass spectrometry universe by having a higher proportion of aromatic residues, a higher proportion of cysteine, serine, and tryptophan residues, a higher copy number of molecules per cell, and higher indices of translation efficiency (ribosome occupancy, ribosome density, and translation) (Table 1 and Supplementary Table S5). Oxidized proteins also present a significantly higher incidence of codon adaptation (Table 1 and Supplementary Table S5). Proteins that are both oxidized and ubiquitinated have a significantly high codon adaptation index and a higher isoelectric point compared with the mass spectrometry universe (Table 1, Supplementary Tables S5 and S6). A bias in codon usage is evident in highly expressed proteins (19, 59) and—taken together with our previous experiments—this bias corroborated the hypothesis that abundant proteins are susceptible to oxidation and can be targeted by K48 ubiquitin for degradation.

Table 1.

Characteristics of Ubiquitinated and Oxidized Proteins

	Sequence features
Ubiquitinated	High codon adaptation index (CAI and FOP scores)a
	Cysteine, serine, tryptophan rich
	High aromaticity
	High protein copy number (4)
	High translation efficiency (4)
Oxidized	High codon adaptation index (CAI and FOP scores)a
Ubiquitinated and oxidized	High isoelectric point
	High codon adaptation index (CAI and FOP scores)a

Ubiquitinated and oxidized proteins identified by EBprot as differentially expressed were compared with all the proteins identified by mass spectrometry. Specific sequence and structural features were tested using yPAD, an in-house set of R-scripts to test a variety of characteristics (Supplementary Tables S5 and S6). Features with t-test <0.005 are highlighted in this table.

^aCAI, FOP, from yeast SGD.

CAI, codon adaptation index; FOP, frequency of optimal codons; SGD, Saccharomyces Genome Database; yPAD, yeast protein annotation database.

Discussion

The function of ubiquitination during oxidative stress and its role in the degradation of oxidized proteins have been investigated for over three decades (14, 20, 22, 56, 57). Still, the complexity of the process and technical limitations have rendered an existing body of knowledge incomplete and often contradictory. Here, we developed a quantitative, time-resolved, proteome-wide system for mapping both oxidation and ubiquitination simultaneously during the cellular response to oxidative stress in yeast (Figs. 1 and 2). First, our results showed that ubiquitinated and oxidized proteins accumulate at different stages of the stress response. Still, their disappearance occurs at the same time and depends on proteasome and E1 activity (Fig. 1B). We also showed that ubiquitinated proteins occur in two phases: first they accumulated rapidly during stress induction and then at a later stage of the recovery phase (Fig. 1B). This rapid increase in ubiquitin conjugates is one of the earliest markers of mild oxidative stress across several tissues (1, 16, 55, 67) and could become a potential tool for diagnosis of stress-related diseases.

We further showed that oxidized proteins can be K48 ubiquitinated in vivo (Fig. 2). Indeed, ubiquitin appears to modify at least half of the oxidized proteins in the data set that are then degraded by the proteasome (Figs. 1B, 2B, 2C, 3, and Supplementary Fig. S3A). Notably, oxidized proteins devoid of ubiquitin could potentially become ubiquitinated during a different phase of the stress response, immediately before their degradation. Our results reinforce that a time-resolved analysis of the ubiquitinated and oxidized proteome is essential in understanding the interplay between these modifications. Temporal analysis of the stress response further helps elucidate the function of the UPS, given that it is composed of various redox-sensitive components that may be activated and inactivated at different stages of the stress response. Moreover, posttranslational modification and gene expression regulation also impact target recognition, ubiquitination, shuttling, and degradation of proteins. Combined with our earlier findings (5, 61), the immediate response to oxidative stress is marked by K63 polyubiquitination (Fig. 1C), while we show here that the recovery phase mostly involves K48 ubiquitin in a proteasome-dependent pathway (Fig. 1). Thus, our work sheds light on the chronological order of events (Fig. 7), where protein oxidation occurs immediately after stress induction, but ubiquitination and degradation occur during a later phase of the stress response (Figs. 1B and 3).

FIG. 7.

Model of the dynamics of protein oxidation and ubiquitin linkages during the stress response. In response to stress, cells accumulate oxidized proteins, K48 and K63 ubiquitins. During the early phase of the response, a second wave of K48 ubiquitin accumulates modifying oxidized proteins, while K63 ubiquitin is removed by deubiquitinating enzymes. Later, K48 ubiquitinated proteins are degraded by the proteasome. Gray arrows indicate the directions of the molecules.

The next challenge is to understand how the oxidation of individual proteins allows them to be recognized by the ubiquitination and degradation system and to define new regulatory processes governing protein dynamics in response to stress and in stress-related diseases. Following protein dynamics at the single-molecule level will be crucial in comprehending the intricacies of this process and the role of the UPS in recognizing and removing damaged and unneeded proteins. Although we show for the first time that both oxidation and K48 ubiquitination affect the same population of molecules, it is still unclear whether some oxidized proteins are degraded independently of ubiquitination (Figs. 1B and 2C). Our results and other works (16, 37, 62) restimulate the discussion on the roles of ubiquitination in the removal of oxidized proteins by showing that a fully functional ubiquitination system is essential for the efficient degradation of oxidized proteins and for the restoration of cellular proteostasis following oxidative stress.

Using high-resolution proteomics, we quantified the oxidation and ubiquitination states of hundreds of yeast proteins in response to oxidative stress. These proteins were involved in a variety of molecular pathways (Figs. 3, 5, 6, and Supplementary Fig. S3) and showed high codon adaptation bias (Table 1 and Supplementary Table S5), indicating a conspicuous relationship to protein abundance (29, 59). Since the UPS regulates the turnover of around 90% of intracellular proteins (38) and more than 1000 ubiquitinated and oxidized substrates have been described in yeast (46, 47, 52, 65, 66), it is not surprising that proteins containing both modifications are involved in a variety of functional pathways and molecular processes. Our data suggest that distinct sequence and structural features may govern the susceptibility of a given protein to damage and degradation via K48 ubiquitin. In this study, we analyzed protein oxidation in the context of oxidative stress, however, our data showed that a basal level of carbonylation is present even in the absence of H₂O₂ (Fig. 1). Carbonyl formation occurs under a variety of physiological processes, including cellular arrest, nutrient starvation, energy production, and even as part of oscillatory metabolic activities (2, 41). Further studies will be necessary to fully comprehend how cells control the levels of oxidized proteins produced under unstressed conditions and whether the same principles of degradation proposed here apply to these proteins. Moreover, the development of new tools to isolate and enrich for additional types of oxidative modification will lead to a more comprehensive understanding of the complexity of this chemical perturbation. As the chemical process is solely dependent on protein abundance and reactivity to reactive oxygen species (64), highly abundant proteins are likely an important source for oxidation and ubiquitination. In agreement with this hypothesis, we showed that oxidized proteins can carry K48 ubiquitin (Fig. 2C) and we identified ubiquitin signatures (GG-modified peptides) in the oxidized protein pool (Supplementary Table S4).

The advancement of mass spectrometry-based proteomics has allowed us to elevate the investigation of this relationship to a systems level (27, 69, 72). In humans, concomitant impairment of UPS function, increased levels of reactive oxygen species, and accumulation of oxidized, ubiquitinated, and aggregated proteins are hallmarks of several neurodegenerative diseases and aging (31, 68). As proposed by others (43), the fate of a protein likely depends on the sophisticated kinetically favorable competition between oxidation and repair; ubiquitination and deubiquitination/degradation; unfolding and chaperone-mediated folding; and finally, degradation and aggregation. The framework presented here contributes to our understanding of these equilibria, and future work should focus on identifying regulators of this pathway and the mechanism by which these proteins are recognized and degraded by different components of the UPS.

Identifying targets of oxidative stress and ubiquitination has promising implications for the diagnosis and treatment of human disease states. A higher preponderance of oxidized (e.g., carbonylated) proteins is associated with a number of human diseases, including amyotrophic lateral sclerosis, rheumatoid arthritis, and muscular dystrophy (7). An intriguing future application of the strategies used here might be to conduct comparative analyses of oxidized and ubiquitinated proteomes in normal and disease models. Identifying frequently oxidized species under disease conditions could inform the development of both diagnostic and prognostic markers. Furthermore, by identifying species that are specifically ubiquitinated under disease states, we can draw hypotheses about whether the clearance of these proteins increases cell viability. Given the recent development and clinical use of bifunctional drugs that direct specific protein targets for ubiquitin-mediated proteolysis (13, 48), the ability to identify proteins that, when cleared, increase cell viability and specifically direct them to the proteasome using small molecules presents an attractive precision therapeutic strategy.

Materials and Methods

Cell strain and culture

S. cerevisiae strain RJD1171, MATa his3Δ200 leu2–3,112 lys2–801 trp1Δ63 ura3–52RPT1^FH::Ylplac211 (URA3) (70), was cultivated into the MPD medium (0.17% yeast nitrogen base without ammonium sulfate, 0.1% proline, 2% glucose, 1.92 g/L of amino acid dropout supplement without uracil) (51). Cells were allowed to divide at least six times before induction of permeability in log phase optical density (OD)₆₀₀ ∼0.2. Permeability was induced by a 90-min incubation in the presence of 0.003% SDS (40). Oxidative challenge was promoted by a 45-min pulse treatment with 0.6 mM H₂O₂. When specified, cells were incubated for 15 min with inhibitors before the oxidative stress induction. Proteasome inhibitor MG-132 (Santa Cruz Biotechnology) was incubated at 75 μM, the E1 ubiquitin activation enzyme was inhibited by 75 μM PYR-41 (EMD Millipore), and 50 μg/mL CHX (Sigma) was used to block protein synthesis. After oxidative stress induction, cells were transferred to fresh media and were allowed to recover up to 8 h in the presence or absence of the specific inhibitors. Yeast SILAC strain GMS413, MATa lys2–801 leu2–3,112 ura3–52 his3-Δ200 trp1–1[am] ubi1-Δ1::TRP1 ubi2-Δ2::ura3 ubi3-Δub-2 ubi4-Δ2::LEU2 arg4::URA3 lys9::KanMX6 [pUB39 ubiquitin (Ub) K63R, LYS2][pUB100, HIS3], was derived from SUB413 (63). Quantitative SILAC experiments were performed in cells grown in SD medium containing the amino acid dropout mixture depleted in arginine and lysine (Sunrise Science Products). SILAC media were supplemented with light or heavy isotopes of arginine and lysine (L-Arg6 ¹³C, L-Lys8 ¹³C, ¹⁵N—Cambridge Isotopes) where untreated control was grown into heavy medium and stress and recovery phases into unlabeled light medium.

Cell permeability assay

The determination of yeast cell permeability to inhibitors was performed based on Ref. (6). Yeast cells were cultivated in MPD medium to log-phase and permeability was induced by 0.003% SDS. For each time point, cellular population was normalized by OD measurement (OD₆₀₀ 0.3) and incubated with 2 μg/mL of crystal violet (Fisher Scientific) for 10 min at 30°C. After incubation, cells were spun down at 7000 g for 3 min. The supernatant was collected, and absorbance was determined at 590 nm.

Serial dilution

Yeast cells were cultivated in MPD medium and after specific treatment, cellular concentration was normalized to OD₆₀₀ of 0.2. From the initial dilution, cells were sequentially diluted at a 1:5 ratio, and each dilution point was spotted into an yeast peptone dextrose-rich medium plate. A qualitative analysis of the colony forming units was assessed after 36–48 h.

Protein preparation

Cell disruption was performed by glass bead agitation at 4°C in standard buffer 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 25 mM iodoacetamide (IAM), 1 × protease inhibitor cocktail set I (EMD Millipore). The extract was centrifuged at 13,000 g for 30 min at 4°C, and protein concentration was determined by the Bradford assay (BioRad). Mass spectrometry: Protein denaturation was achieved by 2,2,2-trifluoroethanol (TFE) addition to 100 μg of protein at a final concentration of 50%. Cysteine residues were reduced by 15 mM DTT for 45 min at 55°C and alkylated by 55 mM IAM in the dark at room temperature for 30 min. Samples were diluted with 50 mM Tris-HCl pH 8.0 to reach 5% of TFE. Trypsin was added to a final concentration of 1:50 w/w enzyme:protein, and the mixture was incubated for 16 h at 37°C. Reaction was stopped with 1% (v/v) formic acid (FA), desalted using Hypersep Spin Tip C18 (Thermo), lyophilized, and resuspended in 5% acetonitrile (ACN), 0.1% FA. Proteasome activity: The cell lysate was prepared as described above in the absence of protease inhibitor and IAM. To preserve proteasome integrity, 4 mM ATP and 4 mM DTT were added to buffer. DNPH derivatization: Proteins were derivatized in 10 mM DNPH solution (solubilized in trifluoroacetic acid) in the presence of 6% SDS for 30 min. Reaction was neutralized in a 1:0.75 ratio of protein sample and buffer containing 2 M Tris and 30% glycerol before electrophoresis (39).

Fluorescence assays for proteasome and deubiquitinase activity determination

The chymotrypsin-like activity of the proteasome and the ubiquitin hydrolase activity of deubiquitinating enzymes were determined in vitro by the incubation of cell lysate with fluorogenic substrate. Protein concentration was determined by the Bradford assay. For the proteasome activity, 100 to 200 μg of cell extract was incubated for 30 min in the presence of 100 μM succinyl-Leu-Leu-Val-Tyr-AMC substrate (EMD Millipore). The reaction was stopped by addition of four volumes of 50 mM Tris-HCl pH 7.5 containing 1.25% SDS. The emission of fluorescence was captured at 440 nm with respective excitation at 365 nm. For the deubiquitinase activity, 25 μg of cell extract was incubated for 40 min in the presence of 1.5 μM Ub-AMC substrate (LifeSensors). The emission of fluorescence was captured at 460 nm with respective excitation at 380 nm.

Antibodies

Anti-ubiquitin (pan), anti-ubiquitin (K48-specific linkage), and anti-biotin were purchased from Cell Signaling. Anti-ubiquitin K63-specific linkage, anti-ubiquitin K11-specific linkage, and anti-20S proteasome core subunits were obtained from EMD Millipore. Anti-actin was from Abcam, and anti-DNP and anti-FLAG were from Sigma-Aldrich. Anti-mouse and anti-rabbit secondary antibodies conjugated with horseradish peroxidase were acquired from GE LifeSciences, and anti-rabbit TrueBlot was obtained from eBiosciences.

Isolation of ubiquitinated and oxidized proteins

Ubiquitinated proteins were isolated from cell lysate by TUBE2 (LifeSensors) affinity purification according to the manufacturer's protocol. Briefly, 1 mg of protein was incubated with TUBE-agarose beads for 2 h at 4°C under agitation. The beads were washed five times before elution. Oxidized (carbonylated) proteins (400 μg) were derivatized with four volumes of 10 mM DNPH solution for 30 min in the dark. The proteins were precipitated with 10% ice-cold trichloroacetic acid and centrifuged at 10,000 g for 10 min. Pellet was washed twice with 1:1 ethyl acetate:ethanol and resuspended in 0.5 M Tris-HCl pH 7.5, 200 mM NaCl, 0.02% Tween-20 with additional sonication to disrupt the protein pellet. DNPH-derivatized proteins were immunoprecipitated using magnetic Dynabeads Protein G (Invitrogen) loaded with 7 μg antibody per mg beads of anti-DNP (Sigma). Immunoprecipitation was performed for 90 min at 4°C under agitation and washed 5 times in Tris-buffered saline-Tween 20 (TBS-T) buffer. TUBE affinity purification was eluted with TFE for mass spectrometry analysis. TUBE pulldown and DNP immunoprecipitated samples were boiled in SDS-PAGE sample buffer for Western blotting analysis.

Biotin-hydrazide derivatization and isolation for mass spectrometry

Cell lysate was prepared and protein concentrations were measured and normalized as described above. To covalently label oxidized proteins, 280 μg of total protein was diluted in 100 mM MES-buffered saline pH 4.7, and EZ-link hydrazide-biotin (Thermo) was added to a final concentration of 5 mM. Samples were incubated at room temperature for 1 h. Unreacted biotin-hydrazide was removed and buffer was exchanged by dialysis (Slide-A-Lyzer MINI dialysis devices, 10K MWCO; Thermo) into dialysis buffer (50 mM Tris HCl pH 7.5, 50 mM NaCl) overnight. To isolate biotinylated proteins, dialyzed lysate was incubated with T1 streptavidin Dynabeads (Invitrogen) for 1 h at 4°C with agitation followed by three washes in TBS-T buffer. For Western blot analysis, samples were eluted by boiling in 2 × Laemmli sample buffer (BioRad). For mass spectrometry analysis, samples were eluted in elution buffer (8 M urea, 50 mM Tris HCl, pH 8.0) at room temperature twice for 20 min with agitation. Samples were digested for 3 h with 2 μg of trypsin/LysC (Promega) at 37°C. Samples were diluted with 50 mM Tris HCl, pH 8.0, to urea concentration of 0.72 M and were allowed to continue digesting at 37°C overnight. Tryptic digestion was halted by adding 1% (v/v) FA. Samples were lyophilized to ∼20 μL, resuspended in 150 μL buffer C (95% H₂O, 5% ACN, 0.1% FA), and washed using a Hypersep C18 Spin Tip (Thermo). The eluted sample was again lyophilized and resuspended in buffer C. Samples were stored at −80°C until liquid chromatography with tandem mass spectrometry analysis.

Mass spectrometry analysis

Tryptic digest was separated on an Agilent Zorbax 300 Stablebond-C18 (3.5 μm, 0.075 × 150 mm) by a reverse-phase chromatography with a gradient of 5%–90% ACN from 3 to 5 h. The eluted peptides were in-line injected into an LTQ-Orbitrap Elite mass spectrometry (Thermo). A data-dependent analysis was performed over the top 20 most intense peaks from each mass spectrometry (MS) full scan using dynamic exclusion settings. Each biological sample was injected three to four times as technical replicates to increase coverage and improve quantification. Data analysis: The output RAW data were processed using the MaxQuant suite (12) to detect, assign, sequence, and quantify the protein abundance assisted by the Andromeda search engine. The data were searched against the yeast S. cerevisiae database (UniProt). Protein identification was performed at 20 ppm tolerance at the Fourier transform-mass analyzer and 6 ppm at the ion trap-MS, with a posterior global false discovery rate at 1% based on the reverse sequence of the FASTA file (UniProt). Two trypsin missed cleavage sites were allowed, oxidation of methionine was searched as variable posttranslational modifications and cysteine carbamidomethylation as fixed. Only proteins with more than one peptide were considered for further analysis, and for the core data set (oxidation/ubiquitinated/expression), a given protein had to be present in all biological replicates.

EBprot method

EBprot computes statistical significance of differential expression for each protein based on the SILAC ratios of its constituent peptides (35). The software assigns a high probability score of true differential expression to a protein if it is represented by consistently large ratios from multiple peptides or if the magnitude of ratio is sufficiently large for proteins with few identified peptides. The score calculation for proteins with a varying number of peptides is automatically calibrated from the entire data set in a data-dependent manner. In EBprot, the user can specify the range of log-transformed ratios for the nondifferentially expressed proteins, which assists the algorithm to accurately identify null distribution of peptide ratios. For both differential protein abundance comparisons, we instructed EBprot to identify the null distribution from the 10 percentile point to the 50 percentile point of the overall distribution (green solid lines in Fig. 4). The algorithm next identifies the distribution of the remaining proteins, that is, differentially expressed ubiquitinated or oxidized proteins (red solid lines in Fig. 4). EBprot reports the final probability score for each protein along with the median fold change (log-transformed, base 2) and the associated false discovery rate.

yPAD method

yPAD refers to a collection of R scripts that were developed to allow user-generated protein lists to be queried against a custom database of relevant protein features from various sources (Supplementary Table S5) in a generalized form. yPAD takes as input, two protein lists in commonly used protein identifier formats such as gene symbol, SGID, UniProt, or open reading frame. In the present study, one list comprised oxidized proteins and one for ubiquitinated proteins. These two lists are queried via identifiers across multiple nonrelational table schemas correcting for key conversion when necessary to produce p-value statistics using both parametric and nonparametric analyses of found feature distributions. Furthermore, the nonoverlapping union of the inputted protein lists is stored as a separate group to elucidate novel feature enrichment of each gene list.

Abbreviations Used

ACN

acetonitrile

CHX

cycloheximide

DNP

dinitrophenyl

DNPH

2,4-dinitrophenylhydrazine

DTT

dithiothreitol

FA

formic acid

FDR

false discovery rate

GG

diglycyl

H₂O₂

hydrogen peroxide

Htb2

histone H2B

IAM

iodoacetamide

MPD

minimum proline dextrose

OD

optical density

SD

synthetic dextrose

SDS

sodium dodecyl sulfate

SILAC

stable isotopic-labeled amino acid in cell culture

TBS-T

Tris-buffered saline-Tween 20

TFE

2,2,2-trifluoroethanol

TUBE

tandem ubiquitin binding entities

Ub

ubiquitin

UPS

ubiquitin/proteasome system

yPAD

yeast protein annotation database

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work was supported, in part, by the U.S. National Institutes of Health K99/R00 award ES025835 (G.M.S.). C.V. acknowledges funding by the NIH/NIGMS grant 1R35GM127089–01 and the Zegar Family Foundation Fund for Genomics Research at New York University. H.C. acknowledges the funding and support of the Singapore Ministry of Education 2013 #T2-2-084. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the article.

Supplementary Material

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4

Supplementary Table S1

Supplementary Table S2

Supplementary Table S3

Supplementary Table S4

Supplementary Table S5

Supplementary Table S6