|Home | About | Journals | Submit | Contact Us | Français|
Ubiquitination is one of the most prevalent protein posttranslational modifications in eukaryotes, and its malfunction is associated with a variety of human diseases. Despite the significance of this process, the molecular mechanisms that govern the regulation of ubiquitination remain largely unknown. Here, we have used a combination of yeast proteome chip assays, genetic screening, and in vitro/in vivo biochemical analyses to identify and characterize eight novel in vivo substrates of the ubiquitinating enzyme Rsp5, a homolog of the human ubiquitin-ligating enzyme Nedd4 in yeast. Our analysis of the effects of a deubiquitinating enzyme, Ubp2, has demonstrated that an accumulation of K63-linked poly-ubiquitin chains results in processed forms of two substrates, Sla1 and Ygr068c. Finally, we have shown that the localization of another newly identified substrate, Rnr2, is Rsp5-dependent. We believe that our approach constitutes a paradigm for the functional dissection of an enzyme with pleiotropic effects.
Post-translational modification (PTM), the covalent crosslinking of a modifying group to one or more amino acids of a protein, is of great interest because of its capacity to modulate the function, location, and stability of proteins as well as their interactions with other proteins . Ubiquitination, one of the most prevalent PTMs in eukaryotes, has emerged as an important mechanism for intracellular signaling. Ubiquitin (Ub) is a highly conserved protein of ~8 kDa that is covalently attached to lysine (K) residues of target proteins, thereby drastically changing the fate of its substrates . Ubiquitination occurs through a three-step process involving Ub-activating (E1), Ub-conjugating (E2) and Ub-ligating (E3) enzymes . E3s determine substrate specificity, and mutations of these enzymes and/or their substrates can lead to a variety of human disorders, including neurodegenerative diseases and cancer .
The two major classes of E3 enzymes are the RING and HECT domain-containing E3s. HECT E3s differ from RING E3s in that they participate directly in the ubiquitination reaction by forming a ubiquitin-thioester intermediate and subsequently catalyzing the ubiquitination of the substrate . Many of the E3s are indispensable for life because they serve as “hubs” to convey upstream signals and/or direct the fate of their downstream targets. Since the substrates of most E3 enzymes are unknown, we chose to take a proteome-wide approach to identifying these molecules, using yeast protein arrays as a platform . We chose Rsp5 as our candidate E3 because it is essential for yeast viability, and it has pleiotropic effects on various intracellular pathways, including endocytosis , mitochondrial inheritance , maintenance of the actin cytoskeleton , drug resistance , biosynthesis of fatty acids , and protein sorting at the trans-Golgi network . Rsp5 is also the closest yeast ortholog to Nedd4, a human HECT E3 that is involved in a congenital human hypertensive disorder known as Liddle's syndrome. Identification of the downstream targets of the yeast E3 enzyme should help identify the mechanisms by which Rsp5 signaling operates to regulate various crucial biologic functions in yeast.
The yeast protein chips were fabricated in-house as described previously .
The genotypes of the strains used in this study are listed in Table 1. Yeast strain FW1808 contains a temperature-sensitive (Ts) allele of Rsp5 (rsp5–1), derived from the wild-type (WT) strain FY56 . To overproduce fusion proteins, constructs were transformed in strains FW1808 and FY56 using standard protocols . UBP2 was deleted in strains FW1808 and FY56 using a standard yeast homologous recombination protocol [15, 16]. Deletion of UBP2 was confirmed by PCR analysis. Empty pEGH and pEGH-Rsp5 were used to transform the rsp5–1 and WT strains for the drug sensitivity experiments. Genes of interest were also chromosomally tagged with the 13×Myc epitope or C-terminal GFP (S65T) , on both the rsp5–1 and WT backgrounds.
GST proteins were purified from 50 ml of culture at the desired temperature, as described previously . The concentration of each purified protein was either estimated on gels stained with Coomassie blue (using bovine serum albumin [BSA] as a standard) or was determined using the BCA™ protein assay kit (Pierce). To remove the GST tag, GST fusion proteins were digested with thrombin (Sigma) at 22° C for 2.5 h according to the manufacturer's instruction.
A reaction mixture consisting of 5 μM E1 (Uba1), 25 μM E2 (UbcH5), and 0.04 μg/μL Ub, with or without the addition of 0.075 μg/μL E3 (GST-Rsp5 or Ubr1), was prepared in reaction buffer (25 mM Tris, pH 7.6, with 50 mM NaCl, 10 mM MgCl2, 4 mM ATP, and 0.5 mM DTT). A yeast proteome chip was incubated with 100 μL of the reaction mixture for 90 min at 37° C, and then subjected to three15-min washes with 0.5 M NaCl, followed by three 15-min washes with 0.5% SDS at room temperature. The chip was then probed with anti-Ub (3,000-fold dilution) (Covance) and anti-GST (5,000-fold dilution) (Chemicon) antibodies, and detected with Cy3- (1:200) and Cy5- (1:200) labeled secondary antibodies, respectively (Jackson ImmunoResearch). The signals were acquired and analyzed by using GenePix software to determine the relative ubiquitination levels of each of the proteins on the chip.
The plasmid constructs of 86 candidate substrates and 64 random proteins were transformed into strains FW1808 and FY56. The growth of each transformant was monitored by plating five-fold serial dilutions of the cells in quadruplicate on SC-Ura agar containing either 2% glucose or 2% galactose at 30° C (permissive temperature) or 34° C (semi-permissive temperature) for 3–4 days.
In-liquid ubiquitination reactions were carried out in test tubes using the buffer system mentioned above at 37° C for 90 min. The outcome of the ubiquitination reactions was determined by immunoblot analysis using an anti-GST antibody.
After the yeast culture had been shifted from 30° C to 37° C for 2 h, the GST-tagged proteins of interest were purified from rsp5–1, ubp2Δ, rsp5–1, ubp2Δ, and WT cells as described previously, followed by immunoblot analysis using the anti-Ub antibody. The same blot was later stripped and re-probed with the anti-GST antibody as a quantity control.
Cells expressing Myc-tagged Rnr2 were grown to log phase and treated with 100 mg/mL cycloheximide. The relative Rnr2-Myc amounts at the indicated time points (0', 5', 10', 30', 60', 90', 120', and 150') were determined by immunoblot analysis using an anti-c-Myc (9E10) antibody (Santa Cruz).
The drug sensitivity of the rsp5–1 and WT strains was assessed by plating five-fold serial dilutions of the cells in quadruplicate on agar with or without the drugs (Table 4) [18–20] for 3–4 days at 30° C and 34° C. Strains containing the empty vector were used as controls. Meanwhile, the rsp5–1 cells were transformed with a low-copy plasmid carrying RSP5 to determine whether the hypersensitivity of rsp5–1 to HU could be reversed at the semi-permissive temperature.
Approximately 1× 107 yeast cells were harvested at mid-log phase and fixed in 70% (v/v) ethanol overnight at 4° C. Fixed cells were sequentially incubated with 2 mg/ml RNase A solution for 2 h at 37° C, then 5 mg/mL pepsin solution (in 4.5 μl/ml HCl) for 1 h at 37° C and 50 μg/mL (1×) propidium iodide (PI; in 0.1 M Tris, pH 7.5, with180 mM NaCl and 70 mM MgCl2) overnight at 4° C. The samples were then resuspended in 0.1× PI, sonicated twice on low power for 5 sec, and analyzed using a Becton Dickinson FACSCalibur. Data were collected on 20,000 cells per sample.
We first took advantage of a combination of protein chip technology, genetic screening, and biochemical assays to identify and characterize in vivo substrates of Rsp5 (Fig. 1A). After optimizing surface chemistries and detection methods, we chose a FullMoon surface for the reactions and anti-Ub antibodies for detection. Each ubiquitination reaction was set up by incubating a proteome chip with a mixture of Ub monomer, ATP, and the E1 (Uba1), E2 (UbcH5), and E3 (Rsp5) enzymes (Fig. S1A) . To ensure that only covalently bound ubiquitins were detected, the chips were washed under highly stringent and denaturing conditions after the reactions took place. To measure the Ub signals and the relative amounts of the spotted proteins, the chips were incubated with anti-Ub and -GST antibodies, followed by incubation with Cy3- and Cy5-labeled secondary antibodies to detect the anti-Ub and -GST antibodies, respectively (Fig. 1B, Fig. S1A). As a negative control, a separate proteome chip was incubated with the same reaction mixture lacking Rsp5. We also performed the ubiquitination reaction using Ubr1, a RING domain-containing E3 ligase, as an additional control. Each assay was performed in duplicate to ensure reproducibility.
After normalizing the Cy3 (ubiquitin) signals against the Cy5 (GST) signals and removing regional artifacts in the data using Lowess normalization, we determined the degree of Rsp5-dependent ubiquitination by comparing the normalized signals between the Rsp5 and the negative control experiments (without Rsp5) (Table S1). The example in Fig. 1B indicates that Ygr068c was clearly ubiquitinated by Rsp5, whereas it remained unmodified when Rsp5 was not included or was replaced by Ubr1. We decided to focus on the top 100 in vitro substrates of Rsp5 for further analysis and characterization. By comparing these hits to the top 40 substrates of Ubr1, as determined in the same fashion, we found that only Vma6 and Nkp2 were shared by both enzymes; this result indicates that specific substrates could be identified using in vitro ubiquitination reactions on a proteome chip.
Gene ontology and statistical analyses revealed no significant protein motifs (e.g., PXY motifs) shared by the substrate candidates. Forty-two proteins shared the same subcellular localization with Rsp5; however, Rsp5 has been localized to multiple subcellular compartments, including the Golgi, cytoplasm, endosomal membrane, plasma membrane, and mitochondria. Furthermore, none of them shared the same biological process with Rsp5 (Fig. 1C). Among the 145 proteins that had previously been shown to bind to either the full-length or the WW domains of Rsp5 [21, 22], only five (Sla2, Met12, Bna5, Ygr068c, and Yjl084c) were found on the hit list, and four (excluding Sla2) of them contain a PXY motif . Therefore, these data were unlikely to help us generate a robust hit list for further validation. These results therefore prompted us to conduct genetic and alternative in vitro assays before carrying out the more rigorous in vivo investigations.
We picked 86 top candidates from the hit list and 64 other proteins at random to evaluate in terms of their potential synthetic dosage lethality or suppression interaction with RSP5 (Fig. 1A). The 150 genes we chose for this analysis were then overexpressed on both RSP5 temperature-sensitive (rsp5–1) and wild-type (WT) strain backgrounds (Table 1, Table 2) , in order to monitor potential differences in colony growth at both a semi-permissive temperature (34° C) and permissive temperature (30° C). Of the 86 candidates, 28 (32.6%) showed an obvious synthetic growth defect or suppression (Figs. 2 and S1C). Among these, Sla2 and Ygr068c have known physical interactions with Rsp5 [9, 24], while Sla1 and Taf3 could be co-purified with Rsp5 [25, 26]. In contrast, only two (Rim11 and Slt2) of the 64 (3.1%) random genes showed dosage lethality/suppression interaction with Rsp5 (data not shown). This dramatic difference in the likelihood of observing dosage lethality interactions suggests that combining the results for protein chip assays and genetic screening may significantly improve the probability of identifying in vivo substrates.
The authenticity of the 28 identified proteins with positive dosage lethality/suppression interactions, as well as 28 other proteins from the 86 top candidates, was examined by in vitro ubiquitination assays. The extent of ubiquitination of each protein was determined by immunoblot analysis with anti-GST antibodies (Fig. 3, Fig. S1C). Eight proteins (Bro1, Nsl1, Rnr2, Rpn10, Sla1, Sla2, Taf3, and Ygr068c) of the positive group and three (Nkp2, Ygr206c, and Bna5) of the negative group were readily ubiquitinated by Rsp5 in solution.
To better evaluate each of the validation steps performed thus far and to identify bona fide Rsp5 substrates, we decided to include all the 28 proteins that were positive in terms of dosage lethality/suppression interactions, the three proteins that were positive in liquid assays but negative for dosage lethality, and Rim11 and Slt2, which were positive in the genetic screens but negative in protein chip assays, in our further experiments to determine whether their ubiquitination is Rsp5-dependent in vivo. These 33 proteins were purified from both WT and rsp5–1 strains grown at the non-permissive temperature, and equal amounts of the purified proteins were then subjected to immunoblot analysis using anti-ubiquitin antibodies to detect their ubiquitinated forms (Fig. 4 and Fig. S1D). The same blot was then stripped and re-probed with anti-GST to visualize all their forms (Fig. 4 and Fig. S1D). The ubiquitinated substrates migrate slower than the unmodified forms, and this shift always correlates with the molecular weight of the unmodified form, excluding the possibility of contamination by other ubiquitinated proteins during the purification. Eight proteins (Rpn10, Rnr2, Nsl1, Nkp2, Sla1, Sla2, Taf3, and Ygr068c) showed Rsp5-dependent ubiquitination (Figs. 4 and S1D, Table 3). Therefore, these proteins are confirmed to be novel, in vivo substrates of Rsp5. In contrast, neither Rim11 nor Slt2 showed Rsp5-dependent ubiquitination, suggesting that positive results in dosage lethality/suppression screening can reflect indirect effects.
The deubiquitinating enzyme Ubp2 has been reported to form a complex with Rsp5 and to antagonize Rsp5-dependent poly-ubiquitination by removing the K63-linked poly-Ub chains . We predicted that the deletion of UBP2 would cause an accumulation of ubiquitinated substrates of Rsp5. To test this hypothesis, we determined the proportion of the total protein that was ubiquitinated for each of the eight validated substrates expressed on the rsp5–1, rsp5–1ubp2Δ, WT, and ubp2Δ backgrounds. Deletion of UBP2 in cells with intact Rsp5 activity resulted in a significant and specific increase in the ubiquitination signals in all eight of the substrates (Table 1, Figs. 5A and S2A).
Intriguingly, two of the substrates (Sla1 and Ygr068c), when overexpressed, were processed in ubp2Δ strains to specific smaller products; however, they were not processed when Rsp5 function was dampened, indicating that this processing is Rsp5-dependent (Fig. 5B). These results suggest that Ubp2 specifically protects these two nonessential proteins from a previously unidentified type of processing, by removing poly-ubiquitin chains added by Rsp5.
When we assessed the endogenous protein levels of six of the substrates (Rpn10, Rnr2, Nsl1, Sla1, Taf3, and Ygr068c), we found that only Nsl1 increased slightly when Rsp5 function was impaired (Fig. S2C). Protein turnover analysis revealed that the protein levels of Rnr2 remained stable, even after 150 min of treatment with cycloheximide, in the case of both the rsp5–1 mutant and WT, indicating a role for Rsp5 beyond targeting proteins for degradation (Fig. S2D).
Two of the in vivo substrates identified in this study, Rnr2 and Nsl1, reside in pathways not currently known to be regulated by Rsp5. We also found that the rsp5–1 mutant was hypersensitive to hydroxyurea (HU), a specific inhibitor of ribonucleotide reductase (RNR), but not to methyl methanesulfonate (MMS) and camptothecin (CPT) (Fig. 6A), all of which lead to DNA damage responses by different mechanisms (Table 4) [18–20]. Moreover, introduction of low-copy-number plasmids containing the wild-type Rsp5 completely suppressed the hypersensitivity to HU in rsp5–1 strains at 34° C (Fig. S3A). Hypersensitivity to HU on an rsp5–1 mutant background could not be explained by cell cycle arrest (Fig. S3B), nor could it be explained by the ubiquitination status of Rnr2 (Fig. S3C).
To further elucidate how Rsp5 regulates Rnr2, we asked whether the Rnr2 localization depends on Rsp5 activity. Using chromosomally GFP-tagged RNR2 in WT and rsp5–1 strains, we assessed the subcellular localization of Rnr2 at permissive and non-permissive temperatures and examined the effect of treatment with HU. We found that the majority of the Rnr2 molecules were redistributed from the nuclei to the cytoplasm in rsp5–1 in the presence of HU at the non-permissive temperature (Fig. 6B, Fig. S4). However, when Rsp5 was active, the same dose of HU had no effect on the subcellular localization of Rnr2 (data not shown). In WT cells, the localization of Rnr2 was not affected by either the temperature shift or HU treatment (Fig. 6B, Fig. S4). In the presence of HU, the pattern of Rnr2 localization in rsp5–1 obviously differed from that for the WT at the non-permissive temperature. Furthermore, when Rsp5 activity was restored by introducing a low-copy-number plasmid carrying RSP5 in the rsp5–1 strain, the localization of Rnr2 showed the same pattern as in the WT strain (Fig. 6B). Taken together, these results demonstrate that the Rsp5-depnedent ubiquitination of Rnr2 contributes to the substrate's resistance to HU, perhaps by regulating the subcellular localization of Rnr2.
Using traditional techniques to elucidate the molecular function of an enzyme with multiple roles in many pathways has always been challenging; identifying all the downstream substrates of such enzymes usually requires a systematic approach. The emerging protein chip technology offers a new tool for globally identifying in vitro substrates of various enzymes. Like other types of large-scale, high-throughput screening (e.g., yeast two-hybrid screening and gene expression profiling), investigators using this approach now face two challenges: how to identify bona fide, direct in vivo targets and how to establish a biological connection between a new target and its upstream modulator. Data integration has been proposed as a means of improving the accuracy of the “hits” derived from large-scale screening , but this strategy does not always work when obvious enrichment is lacking. Therefore, careful examination and evaluation of the robustness, reliability, and inherited bias of the proteomic approach is important for identifying the true substrates of an enzyme.
In this study, the use of chip assay allowed us to quickly narrow down the potential substrates from 5,800 to about 100 proteins. By using genetic screening and a less sensitive, solution-based ubiquitination reaction, we were able to rapidly reduce the number of candidates to eight; seven (87.5%) of these were further validated as true substrates of Rsp5 by more rigorous in vivo analyses. This combination of the three methods dramatically improved the probability of identifying bona fide substrates of Rsp5.
Of the yeast strains harboring knockout mutations of the eight in vivo substrates identified in this study, three (rnr2Δ, nsl1Δ, and taf3Δ) are lethal, two (slaΔ1 and sla2Δ) are temperature sensitive, and two (rpn10Δ and nkp2Δ) show reduced fitness (Table 3) . It is intriguing that many downstream targets of Rsp5 are also essential for viability. Although previous studies have suggested that the essential requirement for Rsp5 is related to the oleic acid pathway , our data seem to indicate that the vital importance of Rsp5 is correlated with its effects on several additional essential pathways. On the basis of the known functions of the substrates we have identified, it is likely that Rsp5 plays a pivotal role in a complicated network that regulates various crucial downstream events, including proteasome function, DNA synthesis, chromosome segregation, cytoskeleton assembly/ endocytosis, and transcription (Fig. 6C). These results should encourage in-depth studies related to the function of ubiquitin E3 ligase.
Among the 145 reported Rsp5-interacting proteins containing the PXY motif , only five were ubiquitinated in vitro by Rsp5 on the protein chip, and two of the five were confirmed as in vivo substrates of Rsp5. This situation may be explained by the notion that a significant portion of these proteins acts as adaptors for Rsp5. Emerging evidence suggests that many Rsp5-interacting proteins recruit Rsp5 to particular subcellular compartments to facilitate the ubiquitination of their substrates . Moreover, the WW domains of Rsp5 may interact only with phosphorylated PXY motifs, and some adaptor proteins may mediate substrate interactions from which proline-rich PXY motifs are absent . Therefore, it would be useful to carry out the Rsp5 ubiquitination on protein chips in the presence of an adaptor protein or after pre-treatment with specific kinases.
Previous studies have identified 11 proteins as bona fide substrates of Rsp5, as determined by Rsp5-dependent ubiquitination in vivo [12, 14, 24, 26, 29, 33]. Most of these substrates either have a low protein abundance on chips because they are membrane proteins and are therefore difficult to express and purify (Fur4, Gap1, Lsb1, Sna4, and Ydl203c), or because they have proved to be unstable in separate attempts at purification (Rpb1 and Ste2). For the rest (Rvs167, Mga2, Sna3, and Ydl203c), we observed only moderate ubiquitin signals for Rvs167 and Sna3, suggesting that the protein chip approach has its own bias against certain proteins. Moreover, Gupta et al. used a similar proteome-wide approach but found a different spectrum of substrates (Fig. S5) . The discrepancy can conceivably be explained by the different strategies used to validate the candidates. In our study, we found that combining the result of on-chip biochemical experiments with genetic interaction profiling significantly increased the probability of identifying biologically relevant substrates (Fig. 1A). Our results further suggest that protein-protein interaction is not required for substrate identification, since only four of the eight validated substrates have been previously shown to interact with Rsp5.
Among the validated in vivo substrates of Rsp5, Rnr2 is of particular interest. Rnr2 is a highly conserved ribonucleotide reductase (RNR) that coverts nucleotides to deoxynucleotides in a reaction dependent on a diferric-tyrosyl cofactor . A heterozygous null mutant of RNR2 is associated with hypersensitivity to DNA damage and to treatment with HU, a chemical inhibitor of the RNRs . After DNA damage, Rnr4 is redistributed within cells, perhaps reflecting an as yet-unidentified posttranslational mechanism . Our results suggest that Rnr2 localization is determined by Rsp5 activity as well as HU treatment. Rnr2 was found in both the nucleus and the cytoplasm in the WT strain, but the majority of the Rnr2 was localized to the nucleus in the rsp5–1 mutant. The fact that the RNR complex needs to be present in the cytoplasm in order to be functional  may help explain why the rsp5–1 mutant is hypersensitive to HU at the semi-permissive temperature.
We conclude that a combination of proteome microarray-based biochemical assays and genetic interaction screens offers a powerful platform for identifying bona fide substrates of enzymes involved in various cellular pathways and our approach constitutes a paradigm for the functional dissection of an enzyme with pleiotropic effects.
This manuscript is dedicated to the memory of Dr. Cecile Pickart. We thank Drs. Jef Boeke and Eric Cooper for critical comments and helpful discussion, Dr. Andrew Emili for providing strains FW1808 and Fy56, and Dr. Deborah McClellan for editorial assistance. This work is supported in part by funding from the NIH (U54RR020839, EY015684, GM28470). JYL is supported in part by NTU and NTUH.