Autophagy-Dependent Regulation of the DNA Damage Response Protein Ribonucleotide Reductase 1

This article is featured in Highlights of This Issue, p. 375

Cellular exposure to genotoxic agents can result in DNA damage and promote mutations and cell death. In response to DNA damage, cells activate genome maintenance pathways associated with the DNA damage response. In Saccharomyces cerevisiae, the DNA damage response is activated by Mec1, Rad53, and Dun1 kinases to mediate cell-cycle checkpoint activation, to activate DNA repair and to promote the regulation of ribonucleotide reductase (RNR) activity (1, 2). RNR catalyzes the rate-limiting step in the production of dNTPs. RNR activity is high during S-phase and increases after DNA damage to elevate dNTP levels (3, 4). In S. cerevisiae, there are 4 RNR genes encoding 2 large subunits (RNR1 and RNR3) and 2 small subunits (RNR2 and RNR4), with the final RNR complex consisting of a tetramer of 2 large and 2 small subunits. To promote high RNR activity after DNA damage, the transcription of the RNR subunits RNR1, RNR3, and RNR4 is induced in a Mec1-dependent fashion (5). Similarly, after DNA damage, RNR activity is increased by Dun1-dependent phosphorylation of the Rnr1 inhibitor Sml1 (6). RNR activity has proved to be a vital regulator of DNA synthesis and cell-cycle progression after DNA damage; RNR regulation by transcription and protein kinase cascades can be considered prototypical for components of the DNA damage response.

After DNA damage, transcription and kinase-based signaling are global themes: the DNA damage response works in tandem with these and a broad range of other cellular processes to optimize DNA repair. Transcriptional profiling studies of S. cerevisiae have demonstrated that alkylation damage induces the transcription of hundreds of genes corresponding to proteins associated with DNA repair, cell-cycle checkpoints, amino acid metabolism and protein degradation (7–11). Protein degradation has been observed previously for components of the DNA damage response (12, 13) and, in theory, can work to optimize essential enzyme activity. In addition, protein degradation is used to remove damaged or unfolded proteins. Protein degradation can occur using either ubiquitin-proteasome or vacuolar-mediated pathways. Both proteasome and vacuolar-mediated protein degradation have been implicated in stress responses and components of both pathways are associated with the reaction to and recovery from DNA-damaging agents (14–16). Transcriptional profiling and targeted studies have highlighted the role of the proteasome after cellular exposure to alkylating agents and demonstrated that the proteasome-associated factor Rpn4 is involved in the regulation of the DNA damage response (9). Global phenotypic and subcellular localization studies have shown that the vacuolar H⁺–ATPase complex modulates the toxicity of the alkylating agent MMS (17, 18). Phenotypic studies also demonstrated that the vacuole plays an important role in promoting cellular viability after S. cerevisiae have been exposed to bleomycin (19).

The ability of the vacuole to modulate the toxicity of damaging agents may rest either in its role in the sequestration of toxic compounds or in its degradation activity. The degradation of cellular constituents in the vacuole requires a functional set of vacuolar ATPases, degradation enzymes and autophagy proteins. Yeast mutants that are compromised for vacuole function are defective in autophagy and display reduced degradation of proteins and organelles (20). Autophagy is a complex catabolic program that uses the vacuole and lysosome in yeast and humans, respectively. In the process of autophagy, parts of the cytoplasm, long-lived proteins, and intracellular organelles are sequestered into a membrane-bound structure termed the autophagosome. The autophagosome is then targeted to the vacuole for degradation (21). TOR, a phosphatidylinositol kinase regulated by nutritional stress, controls autophagy. TOR signaling is active under nutrient-rich conditions and is dormant under starvation conditions to induce autophagy, thereby promoting the breakdown and recycling of cellular macromolecules (22, 23). In addition to controlling autophagy, TOR regulates cell growth, cell cycle, protein synthesis, and nutrient import. In mammals, autophagy is regulated by mTOR and this catabolic process has been implicated in carcinogenesis and cancer progression (24, 25). The human Beclin 1 gene, an ortholog of yeast ATG6, is a critical component of the mammalian autophagy pathway and is classified as a haploinsufficient tumor suppressor (26). Beclin 1 is monoallelically deleted in breast, ovarian, and prostate cancers, and beclin-1 knockout mice show increased levels of spontaneous tumors (27, 28). Allelic loss of Beclin 1 leads to genome instability (29), but the mechanism connecting autophagy to genome maintenance is not well understood.

In this study, we postulated that many DNA damage response transcripts are regulated posttranscriptionally. To test our hypothesis, we have used transcriptional profiling and proteomic approaches to identify targets of posttranscriptional regulation following DNA damage. We determined that over 50% of the transcripts induced by MMS were regulated after transcription, with RNR1 and RNR4 as specific examples. We have used genetic and pharmacological methods to demonstrate increased RNR1 transcription, increased Rnr1 protein synthesis, increased Rnr1 targeting to a membrane-bound structure and increased Rnr1 degradation after DNA damage. Notably, we also demonstrate that Rnr1 targeting and degradation increased upon TOR inhibition, and in all cases the removal of Rnr1 from the soluble fraction required proteins participating in autophagy. We show that cells defective in autophagy have increased Rnr1 protein levels and they display a DNA damage phenotype under conditions of TOR inhibition. Under nutrient-limiting conditions we propose that the autophagy-dependent degradation of Rnr1 serves to optimize the DNA damage response. Ultimately, the following results highlight new regulatory roles for autophagy and TOR signaling during the DNA damage response and further links nutrient sensing to the control of RNR activity.

Supplementary Table 5S lists the strains and oligonucleotides used in this study. All mutants were made using their corresponding G418 knockout cassette from the Saccharomyces Gene Deletion Project and were selected on YPD plates supplemented with G418 (200 μg/mL). Mutants were confirmed by PCR. Media preparation and other yeast manipulations were performed using standard methods. The nutritional stress medium consisted of 0.17% yeast nitrogen base without amino acids, ammonium sulfate and 2% glucose.

Total RNA was isolated using the RNeasy Mini Kit (Qiagen) with 250 U of Zymolyase (Associates of Cape Cod). RNA was isolated from the spheroplasts and examined spectroscopically, by agarose gel electrophoresis and Bioanalyzer analysis (Agilent). Northern blot analysis was performed using 10 to 12 μg of RNA with detection facilitated using the Chemiluminescent Nucleic Acid Detection Module (Pierce). Affymetrix Gene Chip (Yeast 2.0) analysis was performed as previously described (30). RNA from untreated and MMS-treated cells (n = 2) was labeled, applied to chips, and the resulting data analyzed using CyberT software (31). Polysome profiles were performed as previously described (32).

Pelleted cells were lysed with 50 μL of a boiling SDS solution (50 mmol/L Tris-HCl, pH 7.5, 5% SDS, 5% glycerol, 50 mmol/L DTT, 5 mmol/L EDTA, 2 μg/mL Leupeptin, 2 μg/mL Pepstatin A, 1 μg/mL Chymostatin, 0.15 mg/mL Benzamidine, 0.1 mg/mL Pefabloc, 8.8 μg/mL Aprotanin, and 3 μg/mL Anitpain). Lysed cells were centrifuged for 5 minutes at 13,000 rpm and the resulting supernatant was used as protein extract. Membrane protein was extracted using Mem-PER Eukaryotic Membrane Protein Extraction Kit (Pierce). Protein concentrations were measured by NanoDrop 1000 (Thermo Scientific) using BSA as a standard. Western blots were performed as described (33) with an anti-TAP antibody (Open Biosystems), anti-Rnr1 antibody (Santa Cruz Biotechnology) and anti-β-tubulin antibody (Abcam). Individual bands were quantitated using AlphaEaseFC software (Imgen Technologies). iTRAQ analysis was performed using a methodology previously reported (34, 35).

Rnr1 protein degradation was measured in the presence of either the proteasome inhibitor MG132 (Sigma) or the vacuolar inhibitor PMSF (Pierce) in erg6Δ cells. Cultures were grown at 30°C to 5 × 10⁶ cells/mL in YPD media and were preincubated with either 100 μmol/L MG132 or 1 mmol/L PMSF as described (14).

Lysates from MMS-treated and untreated cells were added to the High Bind Plate from Meso Scale Discovery (MSD). Plates were incubated at room temperature for 1 hour, blocked, and washed according to manufacturer instructions. A 1:500 dilution of Anti-tap antibody was added followed by an MSD Sulfo-tag labeled secondary antibody. The plate was washed with TBST and read immediately after adding MSD Read Buffer using the SECTOR 2400 from MSD.

Fluorescence microscopy was performed using an endogenously expressed Rnr1-GFP fusion protein (Invitrogen). Cells were grown to mid-log phase in YPD medium, MMS-treated as indicated, and examined on a Nikon Eclipse TS100 fluorescence microscope. Images were taken with a RT3 camera (Diagnostic Instruments) and SPOT Basic and ImageJ software were used to capture images and quantitate the focus area, respectively. Statistical significance was determined using a t test. The vacuolar membrane was stained using FM4-64 (Invitrogen) following the manufacturers protocol. For confocal microscopy, the vacuolar lumen was stained using CellTracker Blue (Invitrogen) and observed under a Leica Confocal microscope using GFP and DAPI channels. The DNA content of untreated, alpha factor synchronized, and MMS-treated cells were measured by flow cytometry after PI staining, as described (36).

Previous studies in S. cerevisiae have reported that MMS induces hundreds of different transcripts, depending on the exposure conditions (7–11). We hypothesized that many of the MMS-induced transcripts would be posttranscriptionally regulated, and we have matched transcriptional profiling and protein level data to identify potential targets. Mild exposure to the DNA damaging agent MMS promotes a delay in cell-cycle progression, induces minimal cell death, and promotes global reprogramming of the yeast transcriptome (37–39). We have performed transcriptional profiling studies to identify transcripts regulated after MMS exposure. Similar to results reported in the literature, Affymetrix mRNA analysis indicated that MMS exposure induced the levels of 145 transcripts (2.0-fold, P < 0.05) and decreased the levels of 44 transcripts (2.0-fold, P < 0.05) after DNA damage (Fig. 1A, Supplementary Table S1). Specifically, MMS exposure increased the levels of the DNA damage response transcripts RNR1, RNR3, RNR4, and MAG1 (Fig. 1B). Similar to previous studies (7–11), we determined that the 145 MMS-induced transcripts corresponded to protein activities overrepresented (P < 0.05) in the following functional categories: DNA damage response, deoxyribonucleotide metabolism, DNA repair, cell-cycle arrest, and protein/peptide degradation (Supplementary Table S2).

To determine whether an increase in a specific transcript was matched at the protein level, we used global proteomic and targeted protein analyses. We matched protein expression changes to our 145 MMS-induced transcripts. A global proteomic analysis of soluble proteins was performed with proteins isolated from both MMS-treated and untreated cells using tandem mass spectrometry of trypsinized total proteins. The corresponding peptides were labeled with isobaric tags for relative and absolute quantitation (iTRAQ) reagents. iTRAQ identified peptides corresponded to approximately 600 proteins (∼10% coverage), with 33 proteins upregulated (2.0-fold) and 13 proteins downregulated (2.0-fold) in the MMS-treated samples (Supplementary Table S3). Upregulated proteins were overrepresented in the functional categories of stress response, translation initiation, and oxidative stress (P < 0.05, Supplementary Table S4). We paired the 145 damage-induced transcripts with total proteins identified by iTRAQ and identified 16 transcripts with corresponding protein data (Supplementary Table S3). In this limited overlap, only 25% of the MMS-induced transcripts displayed concordant protein level increases, with the other 75% displaying no change or a decrease at the protein level. The identification of protein level changes corresponding to the 129 other MMS-induced transcripts was not feasible at this point as matching peptides were not identified by mass spectrometry. To help fill in the missing information, we used a high-throughput ECL-based technology from MesoScale Discoveries and analyzed 29 specific TAP-tagged strains that corresponded to the transcripts with the highest MMS induction (Fig. 1A). These strains are specific to individual TAP-tagged proteins and allow for the measurement of endogenous protein levels, using a single antibody. We also analyzed Rnr4-TAP as it was transcriptionally induced by MMS and two other components of the RNR complex were picked for analysis using the ECL-based method. Our targeted analysis of the 30 TAP-tagged proteins found only 50% of the MMS-induced transcripts to be concordant with protein level increases. This finding further demonstrates that a large number of MMS-induced transcripts or corresponding proteins were regulated posttranscriptionally, translationally or by protein degradation. We note that a schematic detailing these transcript and protein based methodologies and results is provided in Supplementary Figure S1A.

In evaluating the DNA damage response, we have determined that 2 transcriptionally-induced components fail to show a corresponding protein level increase after DNA damage: the large and small RNR subunits Rnr1 and Rnr4 (Fig. 1B). We also analyzed the MMS-induced transcripts ARG3 and PAC11, and corresponding proteins (Supplementary Fig. S1B), and determined that both failed to have concurrent protein level increases. This suggests that these transcripts were regulated posttranscriptionally. As both Rnr1 and Rnr4 are recognized components of the DNA damage response, we were intrigued by the discrepancy at the protein level. Thus we keyed in on components of the RNR complex for the rest of our study and to probe the posttranscriptional regulation of DDR components. We note that our assay only detected levels of soluble Rnr1 and Rnr4 proteins. In addition, we only detected a slight increase in Rad53 protein levels after MMS damage, somewhat matching the transcriptional increase. However, our protein level analysis may have been distorted by extensive phosphorylation of Rad53 after MMS damage, which we observed as a high-molecular-weight species on Western blots (Supplementary Fig. S1C). Thus, we have omitted Rad53 from our list of proteins decoupled from a transcriptional increase. Northern and Western blot analysis (Fig. 1B) of RNR1 and RNR4 transcripts and corresponding proteins further demonstrated that the MMS-induced transcriptional increases for each do not lead to an increase in soluble protein. In contrast, we detected matched transcript and protein level increases for the DNA damage response genes RNR3 and MAG1 (Fig. 1B). These results suggest that posttranscriptional regulation of the RNR1 and RNR4 transcripts or corresponding proteins is occurring.

To further analyze the potential posttranscriptional regulation of RNR components after DNA damage, we have performed detailed analysis of RNR1 transcript and protein levels over 3 doses of MMS (0.006%, 0.0125%, and 0.025% MMS). Cell viability results demonstrated that there was little killing using these MMS doses (Fig. 1C). Notably, we observed transcriptional induction of the RNR1 transcript using all 3 concentrations of MMS, yet there was no observable increase in Rnr1 protein levels in any of the exposed cells (Fig. 1D). Thus our observation that the transcriptional increase in RNR1 is decoupled from protein levels increases holds true for both sublethal and mildly lethal MMS exposures. MMS is a classic DNA damaging agent that promotes strand breaks and replication blocks, as well as readily methylating RNA and proteins. To rule out MMS-induced protein damage as a cause for our observed discrepancy, we have also analyzed the levels of the RNR1 transcript and Rnr1 protein following bleomycin-induced double-strand breaks. Similar to our MMS studies, we found that RNR1 transcript levels increased in response to bleomycin treatment; yet, Rnr1 protein levels remained unchanged (Fig. 1E). The transcriptional increase in RNR1 after bleomycin treatment is more pronounced compared with MMS treatment, suggesting that the decoupling of protein level increases from transcriptional increases is more dramatic after treatment with agents that directly cause double strand breaks. The MMS and bleomycin results support our hypothesis that protein-level changes can be decoupled from transcriptional changes during the DNA damage response and that posttranscriptional regulation affects soluble Rnr1 protein levels.

As a consequence of observing MMS-induced increases in only 1 of the 2 homologous large subunits of the RNR complex (i.e., Rnr3 and not Rnr1), we have focused in-depth studies on the potential posttranscriptional regulation of the RNR1 transcript and the regulation of the Rnr1 protein. To further validate our observation that increased RNR1 transcription does not lead to increased levels of soluble protein, we monitored RNR1 and RNR1-TAP transcript and protein levels in both MMS-treated and untreated controls (10 to 60 minutes postdamage) using Northern and Western blots (Fig. 2A). Rnr1 protein levels were also analyzed using 2 different antibodies: an anti-TAP antibody to examine the levels of endogenously expressed Rnr1-TAP and an anti-Rnr1 antibody to analyze the native protein. We determined that at 10 and 60 minutes after MMS treatment, RNR1 transcription increased 2.0- to 3.5-fold. Similar MMS-induction results were obtained for the RNR1-TAP transcript. Rnr1 and Rnr1-TAP protein levels in the soluble fraction showed little increase at any time point when compared with untreated controls (Fig. 2B). Analogous Northern and Western blot results were observed at 20-, 30-, and 45-minute time points (Supplementary Fig. S1D). Our time course results using both endogenous and TAP-tagged Rnr1 further demonstrate that levels of soluble Rnr1 protein do not dramatically increase after MMS damage. In addition, the parallel behavior of the endogenous and TAP-tagged Rnr1 proteins supports that the C-terminal TAP tag on Rnr1 does not promote protein degradation. We have performed similar time-course studies using our quantitative ECL approach to track levels of endogenous Rnr1, and in all cases we observed minimal change in soluble Rnr1 protein levels after MMS treatment (data not shown). The observed decoupling of the MMS-induced RNR1 transcriptional increase from a soluble protein level increase suggests that a fraction of the RNR1 transcript or Rnr1 protein pools were affected by posttranscriptional, translational, or posttranslational regulation.

Protein levels can be regulated during translation or by protein degradation pathways. Translation occurs in the ribosomes and polysome profiles are typically used to identify transcripts engaged with the translation machinery. We performed a polysome profile to determine if the RNR1 transcript is actively translated after MMS damage (Fig. 2C). Analysis of the profile for RNR1 from the untreated sample indicated that a portion of the transcript was found in the polysomes, supporting active translation of the transcript under basal conditions. After MMS treatment, we observed that a majority of the RNR1 transcripts were contained in the polysome portion of our profile, indicating that RNR1 is actively translated after damage. Taken together, our RNR1 polysome profile results for untreated and MMS-treated cells suggest that Rnr1 is regulated posttranslationally.

We reasoned that our failure to observe an increase in soluble Rnr1 protein levels after MMS damage was most likely caused by the degradation of newly synthesized or existing Rnr1 protein. Protein degradation occurs via proteasome- or vacuole-dependent pathways, with the later pathway using membrane-bound structures to remove protein from the soluble fraction found in the cytoplasm and deliver it to the vacuole. To determine if either proteasome- or vacuole-dependent pathways promoted the degradation of Rnr1 after MMS damage, we analyzed MMS-induced Rnr1 protein levels in the presence of the proteasome inhibitor MG132 or the vacuolar inhibitor PMSF (14, 40). Preincubation of cells with PMSF followed by MMS treatment promoted a notable increase in the soluble levels of Rnr1 protein at 60 minutes (Fig. 3A), whereas the proteasome inhibitor MG132 did not alter the levels of observed protein. We have also analyzed Rnr4 protein levels after treatment with PMSF and found that inhibition of vacuole-dependent degradation did not affect soluble Rnr4 protein levels (Supplementary Figure S2A). This negative result suggests that regulation of Rnr4 occurs outside of vacuole-associated protein degradation pathways. In contrast, our positive results showing an increase in Rnr1 after PMSF treatment lead us to propose that Rnr1 is ultimately degraded in a vacuolar-dependent fashion.

To explore the potential role of the vacuole in Rnr1 protein degradation, we used a panel of yeast mutants specific to a component of the Golgi-associated retrograde protein complex involved in vacuolar protein sorting (vps54Δ) and 5 VMA gene mutants (vma2Δ, vma4Δ, vma6Δ, vma7Δ, and vma21Δ). VMA genes code for vacuolar H⁺–ATPases that acidify vacuoles and promote proteolysis by vacuolar peptidases (41). Rnr1 protein levels were measured in untreated and MMS-treated cells for the panel of mutants, as well as for wild-type controls. We observed an approximately 1.5-fold increase in basal Rnr1 protein levels in the vma4Δ, vma6Δ, vma7Δ, and vma21Δ cells relative to wild-type cells (Supplementary Fig. S2B), suggesting that some vacuolar-mediated degradation of Rnr1 occurs under normal growth conditions. We observed MMS-induced increases in Rnr1 protein levels in the vps54Δ, vma2Δ, vma4Δ, vma6Δ, vma7Δ, and vma21Δ cells, relative to each mutant-specific untreated control (Figs. 3 B and C). In all cases, the MMS-induced increase of Rnr1 in the vacuole-compromised mutants, relative to untreated mutants, was contrasted by little MMS-induced change in soluble Rnr1 protein levels in wild-type cells after damage (Fig. 3C). Our results with the pharmacological inhibitors PMSF and MG132 and data using vacuole-compromised mutants support the hypothesis that vacuole-dependent degradation of Rnr1 is occurring.

We reasoned that for vacuole-dependent degradation of Rnr1 to occur, a portion of the Rnr1 protein pool must be relocalized to the vacuole. It has previously been demonstrated that subcellular relocalization of the RNR small subunits from the nucleus to the cytoplasm occurs, and this supports the argument that RNR activity can be regulated by subunit location (42). We have demonstrated that a fraction of Rnr1 is degraded in a vacuole-dependent fashion (Fig. 3). We used endogenously expressed Rnr1-GFP and Rnr3-GFP to study the location of the large subunits pre- and post-MMS damage. Fluorescence analysis under basal conditions demonstrated that Rnr1 was diffuse and cytoplasmic with some discrete foci (38% ± 7% of total cells, n = 3) close to the vacuole in the cell (Fig. 4A and B). We note that these foci can also be observed in the Rnr1-specific data found in the Yeast GFP Fusion Localization Database (yeastgfp.yeastgenome.org; ref. 43). In addition, we note that even when cells have discrete foci, we also detected a diffuse cytoplasmic signal for Rnr1-GFP. We determined that MMS treatment induced the transcription of RNR1 and significantly increased the formation of discrete foci (73% ± 8%, n = 3; Fig. 4B), thus indicating a relocalization of a portion of Rnr1-GFP after MMS damage. In contrast, Rnr3-GFP remained diffuse and predominantly cytoplasmic under basal and MMS-treated conditions (Supplementary Fig. S3A). We have also analyzed the growth rate and viability of the Rnr1-GFP strain, before and after MMS treatment and demonstrated that it behaves like wild-type cells (Supplementary Fig. S3B–C). If the GFP tag destabilized the essential protein Rnr1, growth rate and MMS phenotypes would be expected, and this was not observed. Further, data from half-life experiments indicates that both native Rnr1 and Rnr1-GFP are long-lived proteins with half-lives greater than 8 hours (Supplementary Fig. S3D). Together, these results support that the C-terminal GFP tag on Rnr1 is not hindering RNR activity or destabilizing the Rnr1 protein. Further they demonstrate that the Rnr1 protein is long lived under basal conditions. We note that for Rnr1-GFP, DNA damage appeared to induce an increase in focus size, a finding that we confirmed using quantitative image analysis. When compared with the untreated cells (478 ± 290 arbitrary units), the focus area was increased approximately 2.2-fold in MMS-treated cells (1,030 ± 478 arbitrary units). The increase in the number and area of Rnr1-GFP foci in the MMS-treated cells, relative to untreated cells, reveals that a large fraction of the Rnr1 protein pool is localized to foci following DNA damage.

To further analyze the location of Rnr1-GFP foci, we used fluorescence and confocal microscopy with 2 different vacuole-specific dyes: FM4-64 stains the vacuolar membrane and CellTracker Blue selectively stains the vacuolar lumen. Rnr1-GFP–expressing cells were treated with MMS and counterstained with each of the vacuole markers. Merged images specific to Rnr1-GFP foci and the vacuole membrane stained with FM4-64 indicated that the Rnr1-GFP foci were docked at the vacuole membrane and poised to enter the vacuole (Fig. 4C). We obtained a similar result using CellTracker Blue: the Rnr1-GFP foci were observed overlapping the periphery of the vacuole (Fig. 4D). The fluorescence and confocal microscopy results support the hypothesis that Rnr1-GFP foci are targeted to the vacuole and that Rnr1 degradation occurs in the vacuole.

Our observations from experiments using both native Rnr1 and a GFP-tagged version predicted that a portion of the Rnr1 protein pool was being packaged into a membrane-bound vesicle and targeted for degradation. To test this prediction, we isolated the supernatant and membrane fraction of cells from mock- and MMS-treated cells. As expected the levels of native Rnr1 in the supernatant from untreated and MMS-treated cells was similar. In contrast, we have determined that there is an approximately 2-fold increase in native Rnr1 in the membrane fraction after MMS damage (P < 0.0135), relative to the amount of Rnr1 found in the membrane fraction of untreated cells (Figs. 4E and F). This finding supports that Rnr1 is transported to the vacuole in membrane-bound autophagosomes for degradation. RNR activity is regulated during the cell cycle. We reasoned that the removal Rnr1 from the soluble fraction could help optimize dNTP levels or promote cell-cycle transitions and in either case Rnr1 targeting to the autophagosome should be cell-cycle–dependent. In addition, the autophagosome results, when combined with previous data, predict that Rnr1 foci formation and transport to the vacuole should also be dependent on autophagy and TOR.

MMS promotes an S-phase arrest and induces the DNA damage response. During this response, cells will optimize enzyme activities to promote efficient DNA replication. We reasoned that if Rnr1 targeting to the vacuole is used to optimize or control DNA synthesis, then Rnr1-GFP foci should increase during S-phase. To test this prediction, cells were synchronized, released from G₁, and analyzed for Rnr1-GFP foci formation as a function of time (0, 15, 30, 45, 60, 90, and 120 minutes). Cell-cycle characteristics were also quantitated by FACS analysis at each of the 7 time points. We observed (Fig. 5A) foci at increasing levels after cells were released from G₁. Peak foci formation occurred at 60 minutes (Fig. 5B), a time that coincides with a majority of the cells being in late S or early G₂ phases, with approximately half in each phase. As this is a population of cells, this point represents the S to G₂ transition point. Peak mRNA expression for RNR1 occurs during S, whereas the foci were present during both S and early G₂; this most likely occurs because of the high translation level of RNR1 throughout S-phase and its increased targeting to the vacuole during late S and into early G₂. Our results using synchronized and released cells confirm our prediction that foci formation is cell-cycle dependent.

Targeting macromolecules or organelles to the vacuole requires Atg proteins that participate in autophagy. Yeast atg mutants are defective in autophagy and have been reported to be corrupted in vacuole-dependent protein degradation (44). Based on our Western blot and microscopy results, we postulated that Rnr1 degradation would be dependent on Atg proteins. To test this hypothesis, we focused our study on native Rnr1 protein levels in atg2Δ, atg5Δ, atg8Δ, atg9Δ, and atg14Δ cells. Cultures of atg mutants were either left untreated or treated with 0.0125% MMS for 1 hour, after which Rnr1 protein levels were measured by Western blot. Relative to each untreated mutant, we observed MMS-induced increases in Rnr1 protein levels in our panel of atg mutants (Fig. 6A), with 2.0- to 3.5-fold increases for Rnr1 identified in atg2Δ, atg5Δ, atg8Δ, atg9Δ, and atg14Δ (Fig. 6B). In contrast, wild-type cells demonstrated little change in soluble Rnr1 protein levels after MMS damage. Atg2, Atg5, Atg8, Atg9, and Atg14 are proteins involved in autophagosome assembly or vacuolar docking, and deficiencies in any of these proteins increased soluble Rnr1 protein levels. Western blot results reveal that proteins participating in autophagy are involved in the removal of Rnr1 from the soluble fraction. In addition, they demonstrate that deficiencies in either autophagosome assembly or vacuolar docking result in increased levels of soluble Rnr1 protein after MMS damage.

Four lines of evidence support that Rnr1 is sequestered into autophagosomes and subsequently released into the vacuole for degradation. Included among these were the observations that Rnr1-GFP foci localized at the vacuole and that Rnr1 protein levels were increased in membrane fractions after MMS treatment, relative to untreated cells (Fig. 4E and F). In addition, we also observed MMS induced increases in soluble Rnr1 protein levels in vacuole-compromised and atg mutants (Fig. 3 B and C and 6A and B). Consequently, we postulated that defects in the early stages of the autophagy pathway would prevent the formation of autophagosomes, which are represented as discrete Rnr1-GFP foci. We deleted specific atg genes (atg2Δ, atg5Δ, atg8Δ, atg9Δ, and atg14Δ) in the Rnr1-GFP strain and analyzed Rnr1-GFP levels and their corresponding foci, both before (data not shown) and after MMS-exposure (Fig. 6C). Rnr1-GFP was found throughout the cytoplasm before and after MMS exposure in atg2Δ, atg5Δ, atg8Δ, atg9Δ, and atg14Δ; in contrast, we observed MMS-induced foci in wild-type cells. We note that in Figure 6, experiments with atg mutants were performed using both native Rnr1 (Fig. 6A and B) and a C-terminal GFP-tagged version (Fig. 6C), and results associated with each form of Rnr1 implicate a role for autophagy proteins in regulating soluble protein levels.

Autophagy is increased when TOR activity is inhibited by nutrient starvation or rapamycin. We reasoned that if Rnr1 is degraded via autophagy, then Rnr1 protein levels should be sensitive to TOR activity. Rnr1 protein and RNR1 transcript levels were analyzed in wild-type cells, both before and after treatment with rapamycin for 60 minutes (Fig. 7A). Although the RNR1 transcript was detected under all conditions and induced after MMS treatment, there was a large decrease in Rnr1 protein levels in cells treated with rapamycin when compared with those left untreated. Notably, RNR1 transcript levels in the rapamycin + MMS samples are higher than in the untreated sample, yet the Rnr1 protein levels are much lower in this treated sample. Further, we found that this rapamycin-induced decrease in Rnr1 protein levels are diminished in cells corrupted in autophagy (atg9Δ; Fig. 7B). The rapamycin-induced decrease in Rnr1 protein levels suggest that foci form quickly in response to treatment. To test our prediction we analyzed Rnr1-GFP foci levels 15 minutes after cells had been subjected to rapamycin treatment (Fig. 7 C and D). We found a significant increase (∼35%, P < 0.007) in Rnr1-GFP foci levels in rapamycin-treated cells, when compared with those left untreated. In addition, we analyzed native Rnr1 protein and transcript levels in cells where TOR was inhibited by nutritional stress (Fig. 7E). We found that similar to our results with rapamycin, nutritional stress promoted a substantial decrease in soluble Rnr1 protein levels in wild-type cells. This decrease in Rnr1 protein levels was abrogated in the autophagy mutant's atg9Δ and atg14Δ. Our observation of high Rnr1 protein levels in atg9Δ and atg14Δ cells, relative to wild-type cells, was a striking result because the RNR1 transcripts were at higher levels in wild-type cells. Ultimately, these results provide evidence that TOR inhibition and the induction of autophagy will promote the removal of Rnr1 from the soluble protein fraction.

In wild-type cells, the degradation of Rnr1 after TOR inhibition most likely prevents DNA synthesis after DNA damage, and we reasoned this should promote an efficient response to DNA damage. Conversely, we postulated that a failure to degrade Rnr1 levels after TOR inhibition should lead to a DNA damage phenotype in atg mutants. To test our prediction, we exposed cells to rapamycin in the absence or presence of a DNA damaging agent (Fig. 7F). We used rapamycin in plate-based experiments to induce autophagy, as atg mutants can grow in the presence of this compound but fail to grow on nutritional stress media. Thus, the use of rapamycin provided us with a methodology to test for a DDR-associated growth phenotype in cells that fail to degrade Rnr1 via autophagy. Conditions were chosen so neither wild-type or atg cells exhibited a growth defect in the presence of rapamycin or MMS alone. Plate-based killing assays using rapamycin demonstrate that atg (atg2Δ, atg5Δ, atg8Δ, atg9Δ, atg12Δ, and atg14Δ) mutants are more sensitive to MMS, relative to wild-type cells. Taken together, our results support the idea that TOR inhibition and the activation of autophagy plays an important role in promoting the degradation of Rnr1, and our phenotypic studies demonstrate that defects in this process lead to defects in the DNA damage response.

Targeted gene- and protein-specific studies report protein regulation at the levels of protein synthesis or protein degradation (15, 45); yet, few global techniques have been developed to efficiently identify multiple targets of posttranscriptional regulation. Our developed methodology uses 2 standard high-throughput approaches tethered to computational analysis to systematically identify targets of posttranscriptional regulation. Our approach has the potential to identify a plethora of novel regulatory strategies because it can be applied to other perturbations and model systems. What we have demonstrated is that matched transcript-protein level studies can be used to identify discrepancies in standard gene regulatory models (i.e., transcript is induced to make more protein). The identification of any discrepancy or unusual regulatory pattern can be challenging to comprehend but detailed study ultimately leads to a better understanding of biology. Here we focused our studies on the large subunit of the RNR complex to elucidate a novel pathway for the posttranslational regulation of Rnr1 protein levels.

Because dNTP levels play an important role in checkpoint arrest, mutagenesis, and cell viability after DNA damage, RNR activity is extensively regulated in eukaryotic organisms (3). In mammals, RNR activity is regulated by S-phase transcription, the p53-inducible subunit p53R2, proteasome-dependent degradation of the small subunit R2, and allosteric mechanisms (46). In yeast, transcriptional activation of the subunits comprising the RNR complex requires derepression of Rfx1 by Dun1-dependent phosphorylation (47), with phosphorylation also being used to inactivate the RNR inhibitor Sml1. Recently, translation of the RNR1 and RNR3 transcripts has been reported to be dependent on Trm9-catalyzed tRNA modifications (48) and our current study demonstrates that the soluble levels of the large Rnr1 subunit can be regulated via autophagy. Protein degradation of RNR subunits has now been observed in both yeast and human systems, thus supporting the hypothesis that subunit degradation is a conserved regulatory mechanism controlling dNTP levels in vivo.

In general, proteasome-dependent degradation of specific proteins uses ubiquitination and is associated with the removal of short-lived proteins, whereas vacuolar-dependent degradation is considered a nonspecific mechanism to remove long-lived proteins via autophagy. Our results run contrary to this dogma: we observed a very specific removal of Rnr1 from the soluble protein pool by autophagosome transport to the vacuole. Although autophagy is generally considered to be a bulk degradation process, examples of specific targeting have been reported for the cytosolic cysteine protease Lap3 (49), the gluconeogenic enzyme fructose-1,6-bisphosphatase (FBPase; ref. 50), and the metabolic enzyme acetaldehyde dehydrogenase (Ald6; ref. 51). Overexpression of Ald6 has been shown to be cytotoxic, implying that the targeted degradation of this protein is advantageous. Removal of proteins that produce dominant-negative phenotypes may be a theme associated with the specific targeting of proteins via autophagy.

A major question arising from our findings is: Why do cells transcribe and translate Rnr1 when it is unnecessary? In answering this question, we note that our results detailing Rnr1 protein levels involved 2 distinct yet overlapping responses: DNA damage and nutrient stress (via TOR inhibition). In theory, each response has a different objective and thus a different endpoint for Rnr1 protein levels. The DNA damage response requires a precise amount of Rnr1 to promote DNA synthesis under damaging conditions and to control cell-cycle transitions. The degradation of excess Rnr1 protein during DNA damage may be required to promote the formation of an Rnr1–Rnr3 assembly in the final RNR complex, as opposed to an Rnr1–Rnr1 assembly. The Rnr1–Rnr3 assembly is reported to be the most catalytically active form (52). Ghaemmaghami and colleagues have reported that Rnr1 protein levels are approximately 200-fold higher than Rnr3 levels (53). Thus, a low level of Rnr1 protein localization to autophagosomes, as seen in our cell-cycle data for early to mid S-phase, and its subsequent degradation may be part of the mechanism driving subunit formation and optimizing catalytic activity of the RNR tetramer. Because high dNTP levels are needed to drive translesion and replicative polymerases during the DNA damage response, optimal RNR activity is essential to cellular proliferation (3, 4, 54). In addition, our cell-cycle data demonstrating peak Rnr1-GFP foci at the S–G₂ border suggests that increased Rnr1 degradation could also be used to facilitate transition into G₂. There is a decrease in dNTP levels during G₂, relative to S, and Rnr1 degradation could be one of the mechanisms used to promote this reduction. The autophagy-dependent degradation of Rnr1 could play multiple roles, ultimately we believe at low levels it acts to optimize RNR activity during S-phase and at high levels it promotes a transition to G₂.

Nutrient stress, which is also mimicked by rapamycin treatment, is the other distinct condition used in our study and promotes the removal of all Rnr1 protein. The Rnr1 protein was not detected 60 minutes after TOR inhibition, and its absence should lead to a decrease in dNTP levels and a halt to DNA synthesis. Increased autophagy drives this degradation of Rnr1, and that may be a mechanism to quickly align DNA synthesis with current nutrient conditions. TOR inhibition has also been demonstrated to lead to a reduction in RNR1 transcription (55), and we have observed this transcriptional control in our studies. Our report supports and demonstrates that the cell uses 2 mechanisms to decrease Rnr1 protein levels after TOR inhibition, an RNR1 transcriptional decrease and a removal of existing Rnr1 from the soluble pool, followed by degradation. We note that we have observed a long half life for Rnr1 protein. The 2 TOR-based strategies to control Rnr1 protein levels limit the synthesis of new protein and ensure that any existing protein is removed from the soluble fraction. In situations of nutrient stress, cell division decreases and autophagy ensures that the necessary precursors are available to maintain the cell. In this context, Rnr1 degradation might represent one of the control mechanisms used to arrest cells and prevent DNA synthesis during times of nutritional stress. Our phenotypic studies demonstrate the importance of coordinated response to DNA damage under conditions of TOR inhibition, as atg mutants fail to grow under these conditions. Ultimately, our report links defects in autophagy to a DNA damage phenotype and highlights the importance of protein degradation for optimizing the DNA damage response.

In conclusion, we have used systems biology and targeted studies to identify posttranscriptional regulatory targets and to demonstrate that autophagy plays an important role in regulating soluble Rnr1 protein levels. A remaining question deals with how Rnr1 is targeted by the autophagy machinery. It is possible that MMS-induced protein damage, aggregation, or a posttranslational modification provides the signal, but this has yet to be determined and is the focus of future work. Previous studies in human cell systems report that defects in autophagy promote genome instability, arising from increased metabolic stress that generates increased levels of endogenous damaging agents (56). In addition, mTOR inhibition and the induction of autophagy have been linked to p53 and Ataxia-telangiectasia mutated (ATM) signaling after DNA damage, with ATM activation also occurring by reactive oxygen species in the cytoplasm (57, 58). Nonetheless, these aforementioned examples all demonstrate links between autophagy and known components of the DNA damage response. Our findings in yeast further demonstrate that autophagy and DNA damage signaling are intricately coordinated, and lead us to speculate that human cells defective in autophagy may have altered levels of DNA damage response proteins. Coupled with the increase in DNA damage, a misregulation of DNA damage response proteins could dramatically increase genome instability and carcinogenesis in autophagy deficient cells.

No potential conflicts of interest were disclosed.

This work was supported by grants awarded to T.J. Begley from the NIH (ES01225101 and ES015037) and a James D. Watson Award through NYSTAR.

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