Stress-induced phospho-ubiquitin formation causes parkin degradation

Mutations in the E3 ubiquitin ligase parkin are the most common known cause of autosomal recessive Parkinson’s disease (PD), and parkin depletion may play a role in sporadic PD. Here, we sought to elucidate the mechanisms by which stress decreases parkin protein levels using cultured neuronal cells and the PD-relevant stressor, L-DOPA. We find that L-DOPA causes parkin loss through both oxidative stress-independent and oxidative stress-dependent pathways. Characterization of the latter reveals that it requires both the kinase PINK1 and parkin’s interaction with phosphorylated ubiquitin (phospho-Ub) and is mediated by proteasomal degradation. Surprisingly, autoubiquitination and mitophagy do not appear to be required for such loss. In response to stress induced by hydrogen peroxide or CCCP, parkin degradation also requires its association with phospho-Ub, indicating that this mechanism is broadly generalizable. As oxidative stress, metabolic dysfunction and phospho-Ub levels are all elevated in PD, we suggest that these changes may contribute to a loss of parkin expression.


Results
L-DopA causes parkin degradation. To assess the effect of L-DOPA on cellular levels of parkin, we treated neuronally differentiated PC12 cells with various concentrations of L-DOPA for 24 hours and determined relative parkin expression by Western immunoblotting (WB) (see Table 1 for antibody information). PC12 cells are catecholaminergic cells (producing principally DA) that were originally isolated from a rat pheochromocytoma and have been widely used to investigate catecholamine function and metabolism as well as for model studies of potential causes and treatments of PD 58,59 . Neuronally differentiated PC12 cells also possess levels of parkin that are easily detected by WB, making them a fitting model in which to evaluate the effect of stress on endogenous parkin. Of note, although human parkin contains an internal translation initiation site that gives rise to a shorter parkin isoform 60 , rat parkin lacks this alternative initiation site, so our analysis is of full-length rat parkin. Upon exposure to L-DOPA, we observed a dose-dependent loss of parkin protein that reached significance at concentrations of 100 µM and beyond (Fig. 1A). Given the robust parkin loss we observed with 200 μM L-DOPA (68.4 ± 5.2% parkin remaining with 200 μM L-DOPA compared to 0 μM L-DOPA, p = 0.01, N = 5), we chose this dose for further experiments. A time course study revealed that significant parkin loss after treatment with 200 μM L-DOPA is detectable around 6 hours post-treatment and is nearly complete by 26 hours (0 μM L-DOPA: 99.2 ± 5.2% relative parkin level at 26 hours, N = 4; 200 μM L-DOPA: 55.6 ± 3.2% relative parkin level, N = 4; p = 0.03) (Fig. 1B).
Because we observed some toxicity with 200 μM L-DOPA after 24 and 48 hours of treatment (20.6 ± 1.7% cell death after 24 hours, p = 0.005, N = 4; 41.8 ± 4.9% cell death after 48 hours, p = 0.046, N = 4) (Fig. S1A), we investigated whether parkin loss can occur independently of cell death. To do this, we assessed whether subtoxic doses of L-DOPA can still decrease parkin. To maximize the chances that we would see an effect on parkin with relatively low doses of L-DOPA as well as to ensure that sufficient time would elapse to detect any cell death from L-DOPA, we used a multiple-treatment design (Fig. S1B). Differentiated PC12 cells were treated with 25-100 μM L-DOPA a total of three times, with two days between treatments. Lysates were harvested for WB two days after the final L-DOPA treatment (six days after the first exposure to L-DOPA), while replicate cultures were harvested for survival assessment three days after the final L-DOPA treatment (seven days after the first exposure to L-DOPA). In these experiments, all doses of L-DOPA were non-toxic (100.2 ± 3.9% cells remaining with 100 μM L-DOPA, p = 1.0, N = 3) (Fig. S1C), but a dose-dependent loss of parkin was again observed, reaching significance with 100 μM L-DOPA (33.9 ± 2.8% parkin loss with 100 μM L-DOPA, p = 0.03, N = 3) (Fig. 1D). These results indicate that parkin loss from L-DOPA treatment can occur independently of cell death.
Next, we asked whether decreased production and/or increased degradation is responsible for parkin loss after L-DOPA treatment. We did not observe a significant decrease in parkin mRNA levels after L-DOPA treatment (relative mRNA level after 27 hours normalized to 3 hours of control treatment: 0 μM L-DOPA: 1.54 ± 0.63, N = 2; 200 μM L-DOPA: 1.31 ± 0.186, N = 2; p = 1.0) (Fig. S3A), indicating that L-DOPA's effect on parkin is not transcriptional. To assess whether L-DOPA might affect parkin turnover, we treated cells with the translation inhibitor cycloheximide (CHX) and measured parkin levels after various times with or without L-DOPA exposure (Fig. S3B). L-DOPA accelerated parkin loss in the presence of cycloheximide after an apparent lag time of several hours (CHX: 52.5 ± 5.8% parkin remaining after 48 hours, N = 3; CHX + L-DOPA: 36.2 ± 5.9% parkin remaining, N = 3; p = 0.004), such that the apparent half-life of the protein during the linear phase of decay fell from 41.9 hours to 27.2 hours without and with L-DOPA treatment, respectively. In contrast, there was no effect of L-DOPA on the half-life of ERK1 protein (Fig. S3B). These findings suggest that the loss of parkin that occurs with L-DOPA is due at least in part to its enhanced degradation.
If L-DOPA promotes parkin loss at least in part by increasing its turnover, then we would expect to see loss of overexpressed exogenous parkin. In an initial study, we did not observe consistent and specific L-DOPA-promoted loss of highly overexpressed (~30-fold) exogenous parkin driven from the EF-1α promoter (Fig. S4A), suggesting that very high levels of parkin are able to overwhelm the L-DOPA-induced parkin degradation mechanism. In contrast, when we more moderately overexpressed N-terminally-tagged parkin using the minimal human parkin promoter, leading to ~8-fold overexpression above endogenous parkin levels, we observed an L-DOPA-induced loss of exogenous parkin similar to that observed with the endogenous protein (40.7 ± 0.7% exogenous parkin loss with L-DOPA, p = 0.0003, N = 3) (Fig. S4B).
L-DopA-induced oxidative stress contributes to parkin loss. Given that L-DOPA has been shown to cause oxidative stress in cultured cells 56,61,62 , we examined whether L-DOPA-induced oxidative stress plays a role in parkin loss. To do this, we treated differentiated PC12 cells with L-DOPA in the presence of the antioxidant glutathione. Glutathione (GSH) significantly attenuated parkin loss from L-DOPA, but this rescue was  Supplementary Fig. 12. In C, the 500 μM immunoblot image shows non-adjacent bands from the same blot. not complete (43.3 ± 3.9% loss with L-DOPA alone vs. 18.9 ± 6.5% loss with L-DOPA + GSH, N = 3, p = 0.03; p = 0.03 for L-DOPA + GSH vs. GSH alone, N = 3) ( Fig. 2A). We did not observe enhanced protection of parkin from L-DOPA-induced loss by using higher doses of glutathione or by pre-treating cells with glutathione prior to L-DOPA exposure (Fig. S5). This result indicates that oxidative stress is at least partially responsible for L-DOPA-induced parkin loss but that a non-oxidative mechanism may account for about half of this loss.
L-DOPA can generate reactive oxygen species (ROS) via non-enzymatic autoxidation and by conversion to dopamine, with subsequent autoxidation and oxidative metabolism of the latter 53,63,64 . To test whether conversion of L-DOPA to dopamine is required for its effect on parkin, we used carbidopa to block the activity of aromatic L-amino acid decarboxylase (AADC), the enzyme responsible for this conversion. As anticipated, 200 μM L-DOPA treatment alone for 16-24 hours significantly increased intracellular dopamine levels (to 3.0 ± 0.4 fold over control levels, p = 0.02, N = 5) (Fig. 2B). Co-treatment with carbidopa robustly diminished this effect (to 1.1 ± 0.3 fold over control levels, p = 0.02 vs. L-DOPA alone, N = 5) but did not impact the loss of parkin from L-DOPA treatment (30.9 ± 5.9% parkin loss with L-DOPA alone vs. 31.4 ± 4.7% loss with L-DOPA + carbidopa, p = 0.9, N = 4) (Fig. 2C). These results suggest that autoxidation of L-DOPA itself, rather than its conversion to dopamine, is the source of the oxidative stress that leads to parkin loss following L-DOPA treatment in our system. Consistent with this possibility, we observed "browning" of the cell culture medium after 24 hours of exposure to L-DOPA, even in the absence of cells (Fig. S6), which has been shown to be a result of quinone formation from L-DOPA autoxidation 62,[65][66][67] .
To confirm that oxidative stress is capable of inducing parkin loss, we treated cells with 200 μM hydrogen peroxide and observed a reduction of parkin protein comparable to that found with L-DOPA (39.4 ± 3.8%, p = 0.009 vs. untreated control, N = 3) (Fig. 2D), in line with a previous report 41 . Altogether, these results indicate that oxidative stress from L-DOPA autoxidation significantly contributes to L-DOPA-induced parkin loss. PINK1 plays a role in L-DOPA-induced parkin loss. The kinase PINK1 has been reported to play a role in parkin turnover, including that caused by hydrogen peroxide exposure 41,68 . We therefore examined whether PINK1 is involved in L-DOPA-dependent parkin loss. To do this, we knocked down PINK1 in differentiated PC12 cells using a previously-characterized shRNA 69 . PINK1 knockdown was very effective at the mRNA level, reducing PINK1 transcript levels to 14.1 ± 3.5% of control levels after 4 days (p = 0.0001, N = 4) (Fig. 3A).
Knockdown of PINK1 significantly attenuated L-DOPA-induced parkin loss, to about half that achieved with L-DOPA alone (36.7 ± 2.8% parkin loss with sh-ctrl vs. 17.7 ± 6.3% loss with sh-PINK1, p = 0.02 for sh-ctrl + L-DOPA vs. sh-PINK1 + L-DOPA, N = 6) (Fig. 3B). These findings indicate that PINK1 is indeed involved in L-DOPA-mediated parkin loss, but that a PINK1-independent mechanism also appears to contribute to this process.
Having implicated PINK1 in L-DOPA-induced parkin loss, we next assessed PINK1 activity using S65-phosphorylated ubiquitin (phospho-Ub) as a readout. At baseline, phospho-Ub is present at very low or undetectable levels in cells 70,71 . Following mitochondrial depolarization, PINK1 becomes stabilized on the outer mitochondrial membrane and activated, leading it to phosphorylate ubiquitin and resulting in an accumulation of phosphorylated poly-ubiquitin chains (phospho-poly-Ub) that is detectable by Western blotting 70,71 . As expected, we observed little phospho-Ub in untreated PC12 cells (Fig. 3C). As a positive control for PINK1 activation and phospho-Ub induction, we treated cells with the protonophore carbonyl cyanide 3-chlorophenylhydrazone (CCCP), which is widely used as a mitochondrial depolarizing agent 72 . Following CCCP, we observed the appearance of high molecular weight phospho-Ub ladders, consistent with formation of phospho-poly-Ub chains (Fig. 3C). Treatment of cells with L-DOPA also led to formation of phospho-poly-Ub (Fig. 3C), although the strength of the L-DOPA-induced phospho-poly-Ub signal was about 10 times weaker than that induced by CCCP (p = 0.04, N = 3 for L-DOPA and 4 for CCCP) (Fig. 3C,D). We did not observe the formation of unconjugated mono-phospho-Ub after L-DOPA treatment (Fig. 3C). Knockdown of PINK1 strongly attenuated the phospho-poly-Ub signal from CCCP exposure (to 27.5 ± 5.1% of sh-ctrl, p = 0.003, N = 4), and caused a similar downward trend in the L-DOPA-induced phospho-poly-Ub signal (to 40.5 ± 16.3% of sh-ctrl, p = 0.07, N = 3), indicating that PINK1 indeed appears to be involved in generating this signal (Fig. 3E,F). This suggests that L-DOPA, like CCCP, activates PINK1. Together, these results indicate that PINK1 activity contributes to L-DOPA-induced parkin loss.

Association with L-DOPA-induced phospho-Ub leads to parkin loss. Having implicated PINK1
in L-DOPA-induced parkin depletion, we next sought to uncover the mechanism by which PINK1 contributes to this loss. PINK1 activates parkin by phosphorylating Ser65 on both parkin and on ubiquitin [42][43][44]46,48,49,[73][74][75][76] , with parkin binding the latter non-covalently. Parkin binds to both unconjugated phospho-mono-Ub and to phospho-poly-Ub chains 43,73,77 . We reasoned that parkin phosphorylation, parkin binding to L-DOPA-induced phospho-Ub, or both of these events may underlie PINK1-mediated parkin loss. To determine the relative importance of these events, we introduced several point mutations in the "moderately" (~8-fold) overexpressed exogenous parkin construct described above (Fig. S4B). To test the importance of parkin phosphorylation, we generated a S65A parkin mutant. To test the importance of phospho-Ub binding to parkin, we generated parkin mutants H302A and K151E. Both of the latter mutations significantly abrogate the interaction between parkin and phospho-Ub, although K151E has been demonstrated to be more effective in this regard 46,47 . We delivered the parkin mutants to PC12 cells via lentiviral transduction and tested their responses to L-DOPA treatment.
Although the viral titers between constructs varied slightly due to inherent variability in the viral production process, we found that all parkin mutants used in this study were expressed at similar levels (i.e. ~8-fold higher than endogenous parkin) in cells when controlling for titer (Fig. S7). www.nature.com/scientificreports www.nature.com/scientificreports/ L-DOPA induced a significant loss of wild-type overexpressed parkin after 24 hours of treatment, as expected (38.0 ± 2.9% loss, p < 0.0001, N = 8) (Fig. 4A,B). However, the mutants deficient in binding phospho-Ub were significantly resistant to loss compared to wild-type parkin (H302A: 22.9 ± 2.1% loss, N = 8, p = 0.006 vs. WT + L-DOPA; K151E: 20.7 ± 4.3% loss, N = 7, p = 0.003 vs. WT + L-DOPA) (Fig. 4A,B). Of note, the degree of protection from L-DOPA-induced loss afforded by the H302A and K151E mutations was ~50%, similar to the degree of protection afforded by PINK1 knockdown (Fig. 3B). This suggests that the contribution of PINK1 to L-DOPA-mediated parkin loss may be fully accounted for by parkin's interaction with PINK1-generated phospho-Ub. In agreement with this, the S65A parkin mutant was not at all protected from L-DOPA-induced loss compared to wild-type parkin (S65A: 36.8 ± 3.6% loss, N = 4, p = 0.82) (Fig. 4A,B), indicating that PINK1-mediated parkin phosphorylation is not required for parkin loss. We also tested whether a protective effect of S65A might be revealed in the context of a parkin mutant that is deficient in binding to phospho-Ub. However, we did not observe greater protection from L-DOPA with a H302A/S65A double mutant over the H302A single mutant (H302A/S65A: 16.1 ± 7.8% loss, N = 3, p = 0.48 vs. H302A alone) (Fig. 4A,B). The above findings indicate that parkin's interaction with phospho-Ub is required for a portion of its loss following L-DOPA treatment, while parkin phosphorylation appears to be dispensable. However, one potential reservation for this interpretation is that our mutant parkin constructs have an N-terminal tag to differentiate them from endogenous parkin, and such tags have been reported to disrupt parkin's autoinhibited conformation by opening the Ubl domain 78 . Because parkin phosphorylation has also been reported to open the Ubl domain 46,80 , we wished to ensure that the effect of parkin phosphorylation was not made functionally redundant by the presence of the tag. To do so, we took advantage of the finding that parkin phosphorylation is required for full parkin activation in response to CCCP 48,49 . We compared the levels of CCCP-induced phospho-poly-Ub in cells transduced with either wild-type or S65A parkin. We reasoned that since parkin phosphorylation contributes to its activation and since such activation leads to phospho-poly-Ub formation following mitochondrial depolarization 50 , we would expect to see decreased levels of phospho-poly-Ub in CCCP-treated cells expressing S65A parkin, but only if this mutant form is not already activated by the presence of the N-terminal tag. Indeed, cells transduced with S65A parkin had significantly decreased levels of phospho-poly-Ub after CCCP exposure compared with those expressing wild-type parkin (WT: 2.12 ± 0.18; S65A: 1.72 ± 0.23; p = 0.04, N = 5) (Fig. S8). Similarly, CCCP-treated cells expressing the H302A/S65A double parkin mutant had significantly decreased levels of phospho-poly-Ub compared to those expressing the H302A single mutant (H302A: 1.73 ± 0.13; H302A/ S65A: 1.19 ± 0.04; p = 0.007, N = 5) (Fig. S8). These results demonstrate that parkin phosphorylation plays a functional role in the CCCP model and is not made redundant by the presence of an N-terminal tag. In light of this, our observation that the S65A parkin mutant is not at all protected from L-DOPA-induced loss indicates that parkin phosphorylation status does not impact this loss. Instead, PINK1-mediated ubiquitin phosphorylation is required for PINK1-dependent parkin loss downstream of L-DOPA exposure.
L-DopA-induced oxidative stress-and phospho-Ub-dependent mechanisms of parkin loss are in the same pathway. The findings that glutathione treatment, PINK1 knockdown, and mutation of parkin's phospho-Ub binding site all rescued L-DOPA-induced parkin loss by about 50% (Figs 2A, 3B and 4A) suggest that parkin loss from L-DOPA treatment occurs via two distinct pathways, and that oxidative stress and PINK1 activity each play a role in only one of these pathways. To determine whether the oxidative stress-dependent (glutathione-sensitive) and the phospho-Ub-dependent mechanisms of parkin loss are in the same pathway, we investigated whether glutathione can abrogate the phospho-Ub signal induced by L-DOPA. Given that the majority of this signal is present in high molecular weight phospho-poly-Ub conjugates, we examined the effect of glutathione on phospho-poly-Ub. We found that glutathione prevented L-DOPA-induced phospho-poly-Ub formation almost completely (no treatment: 28.6 ± 7.8% phospho-poly-Ub relative to L-DOPA treatment, p = 0.002, N = 5; L-DOPA + GSH: 37.6 ± 5.6% relative to L-DOPA treatment, p = 0.002, N = 5) (Fig. 4C). Additionally, hydrogen peroxide exposure induced phospho-poly-Ub formation (Fig. 4D). These results indicate that oxidative stress from L-DOPA exposure leads to PINK1-mediated formation of phospho-Ub, its association with parkin, and parkin loss.

phospho-Ub generated by diverse stressors induces parkin degradation.
Having placed a portion of parkin loss from L-DOPA in the same pathway as L-DOPA-induced formation of phospho-Ub, we next asked whether, and to what extent, parkin interaction with phospho-Ub plays a role in its loss from other stressors. Given that hydrogen peroxide causes parkin loss (Fig. 2D) and induces phospho-poly-Ub formation (Fig. 4D), we first queried whether parkin loss from this oxidative stressor is also mediated by its interaction with phospho-Ub. To achieve this, we utilized neuronal PC12 cells moderately over-expressing either wild-type or various mutant forms of parkin. Indeed, while wild-type parkin levels decreased by half upon peroxide treatment (50.9 ± 4.3% parkin remaining; p = 0.001, N = 5), the H302A and K151E parkin mutants were nearly completely rescued from loss induced by this agent (K151E + peroxide: 96.2 ± 8.0% parkin remaining, N = 5, p = 0.97 vs. no treatment; H302A + peroxide: 86.4 ± 5.4% parkin remaining, N = 5, p = 0.2 vs. no treatment) (Fig. 4D,E). As with the L-DOPA model, we did not observe any protection from peroxide-induced loss with the S65A mutant (S65A: 54.1 ± 2.4% parkin remaining, N = 5; p = 0.84 vs. WT) (Fig. 4E). Based on these results, it appears that parkin's association with phospho-Ub is critical, and fully accounts for, its degradation following exposure to hydrogen peroxide.

phospho-poly-Ub is present in both cytosolic and mitochondrial fractions and leads to mitochondrial parkin translocation.
Given the importance of phospho-Ub for parkin loss from oxidative stress and mitochondrial depolarization, we next queried where in the cell phospho-Ub is found. As before, we examined high molecular weight phospho-poly-Ub because it is the dominant species of phospho-Ub we observe. Previous work has shown that phospho-poly-Ub is present in both the cytosol and on mitochondria following PINK1 activation by mitochondrial depolarization 70 . In line with these findings, sub-cellular fractionation Given the substantial presence of phospho-poly-Ub in the mitochondrial fraction after L-DOPA treatment and the affinity of parkin for phospho-Ub 46,47,73,79 , we predicted that we would observe mitochondrial parkin translocation following L-DOPA treatment, analogously to that observed following CCCP exposure 50,81 . In agreement with this, we saw a modest but significant increase of parkin in the mitochondrial fraction after 14-17 hours of L-DOPA treatment (1.31 ± 0.08 with L-DOPA vs. without, p = 0.01, N = 7) (Fig. 6A,E), a time before L-DOPA-induced parkin loss is complete (Fig. 1B). As a positive control, we examined mitochondrial translocation of parkin following 6 hours of CCCP exposure -a time at which parkin loss is underway but not complete. Similarly to the effect of L-DOPA treatment, there was an increase of parkin in the mitochondrial fraction after CCCP treatment (1.55 ± 0.08 with CCCP vs. without, p = 0.007, N = 4) (Fig. 6B,F). Altogether, these www.nature.com/scientificreports www.nature.com/scientificreports/ results suggest that parkin associates with stress-induced phospho-poly-Ub both in the cytosol and on mitochondria before its degradation. phospho-Ub-induced parkin degradation is proteasomal. We next sought to address the mechanism of parkin degradation following its association with phospho-Ub. Two previous studies reported that parkin degradation is proteasomal after mitochondrial depolarization 40,41 . In agreement with these reports, the proteasomal inhibitor epoxomicin fully abrogated CCCP-induced parkin loss (CCCP: 63.5 ± 7.6% of untreated; CCCP + epoxomicin: 94.5 ± 7.6% of untreated, p = 0.02, N = 5) (Fig. 7A). Significantly, it also abrogated L-DOPA-induced parkin loss, but by about half (L-DOPA: 61.1 ± 3.0% of untreated; L-DOPA + epoxomicin: 80.5 ± 3.3% of untreated, p = 0.001, N = 8), consistent with half of L-DOPA-induced parkin loss being phospho-Ub-dependent (Fig. 7B). Parkin has been reported to be cleaved by caspases 82,83 , but caspase inhibition using the pan-caspase inhibitor Z-VAD-FMK did not abrogate parkin loss from L-DOPA treatment (L-DOPA: 41.0 ± 5.7% parkin loss, L-DOPA + Z-VAD-FMK: 40.8 ± 4.0% parkin loss, p = 0.98, N = 4) (Fig. S9). Taken together, these data suggest that phospho-Ub-dependent parkin degradation is mediated by the proteasome in the various models tested here. L-DOPA does not appear to affect parkin solubility. In addition to degradative mechanisms, various stressors have also been reported to diminish available parkin levels by disrupting its structure and decreasing its solubility [84][85][86][87][88][89][90] . To assess whether this might contribute to apparent parkin loss after L-DOPA exposure, we harvested L-DOPA-treated and non-treated cells in either our standard lysis buffer or a strong lysis buffer containing 8 M urea in addition to 2% lithium dodecyl sulfate (LDS). The strong lysis buffer did not produce significantly higher yields of parkin recovery (standard lysis buffer: 30.7 ± 4.7% parkin loss, 8 M urea buffer: 20.9 ± 4.8% parkin loss, p = 0.32, N = 5) (Fig. S10). These findings suggest that L-DOPA does not reduce apparent cellular levels of parkin by reducing its solubility.
phospho-Ub-dependent parkin loss does not require its activity in cis. During mitophagy initiation, PINK1 and parkin participate in a positive feedback loop wherein parkin-mediated poly-ubiquitination of mitochondrial proteins provides PINK1 with ubiquitin substrates for phosphorylation, generating www.nature.com/scientificreports www.nature.com/scientificreports/ phospho-poly-Ub 50 . In accordance with this process, parkin activity has been found to increase the total amount of cellular phospho-Ub following mitochondrial depolarization 70 . If this process occurred in our models, we would expect stress-induced phospho-poly-Ub formation and phospho-Ub-induced parkin loss to depend on parkin's ligase activity.
To examine this possibility, we compared the levels of stress-induced phospho-poly-Ub in PC12 cells transduced with our "moderately overexpressed" wild-type parkin, a catalytically inactive parkin mutant (C431S), or an empty control vector. In CCCP-treated cells, expression of wild-type parkin increased the phospho-poly-Ub signal by about 3-fold over that of cells expressing empty vector (WT: 2.85 ± 0.30 relative to empty vector, p = 0.01, N = 5) (Fig. 8A,B). This increased phospho-poly-Ub signal can be attributed to parkin activity because www.nature.com/scientificreports www.nature.com/scientificreports/ catalytically inactive parkin almost completely abrogated the increased phospho-poly-Ub signal (C431S: 1.31 ± 0.07, p = 0.01 for WT vs. C431S, N = 5) (Fig. 8A,B). This indicates that, in the CCCP model, parkin is a limiting component for the progression of the positive feedback loop, because addition of more parkin allows the loop to generate more phospho-poly-Ub. By contrast, the L-DOPA-induced phospho-poly-Ub signal was not detectably increased in cells expressing wild-type parkin compared to empty vector or C431S parkin (WT: 0.83 ± 0.02, p = 0.04 vs. empty vector; C431S: 1.25 ± 0.12, p = 0.19 for WT vs. C431S, N = 3) (Fig. 8D,E). A similar result was observed in the hydrogen peroxide model (WT: 1.19 ± 0.17, p = 0.66 vs. empty vector; C431S: 0.92 ± 0.13, p = 0.51 for WT vs. C431S, N = 5) (Fig. 8G,H). These results indicate that, in the L-DOPA and hydrogen peroxide models, either parkin isn't involved in building the poly-ubiquitin chains that become phosphorylated by PINK1, or, alternatively, parkin does build the poly-ubiquitin chains, but this activity isn't the limiting factor for the progression of the positive feedback loop with PINK1.
phospho-Ub-dependent parkin loss does not require parkin autoubiquitination. Of note, given that parkin has been shown to only autoubiquitinate in cis 91 , the finding that C431S parkin was not protected from L-DOPA-induced depletion indicates that this loss is not due to autoubiquitination. Parkin autoubiquitination has been reported following mitochondrial depolarization, reflected by the formation of higher-molecular weight parkin species as detected by WB 36,40,[92][93][94] . It was previously suggested that such autoubiquitination is what targets parkin for proteasomal degradation following mitochondrial depolarization 40 . In the studies where parkin ubiquitination was observed, the ubiquitinated parkin signal was generally strongest for the band corresponding to mono-ubiquitinated parkin, but multiply-ubiquitinated parkin species were also detected, though with diminished intensity. To extend our observations with C431S parkin, we treated cells with either L-DOPA or CCCP in the presence or absence of epoxomicin and carried out Western blotting to detect higher-molecular weight parkin www.nature.com/scientificreports www.nature.com/scientificreports/ forms (Fig. 7A,B). In no case did we detect higher-molecular weight parkin bands, consistent with the finding that autoubiquitination is not required for parkin loss.

phospho-Ub-dependent parkin loss does not require mitophagy. Given that phospho-poly-Ub
formation is an early step in the process of mitophagy, we asked whether mitophagy is necessary for phospho-Ub-induced parkin loss. First, we determined whether there is evidence of robust mitophagy in our models. Induction of robust mitophagy by CCCP has been previously shown to cause rapid ubiquitination of the mitochondrial GTPases mitofusin-1 and -2 (Mfn1 and 2), followed by a gradual decrease in mitochondrial proteins in all compartments, all of which is readily apparent by Western blotting 36,[95][96][97] . To assess whether such robust mitophagy takes place in our models, we assessed the levels of several mitochondrial proteins from different mitochondrial compartments after CCCP and L-DOPA treatment: Mfn2 (OMM), VDAC (OMM), Tom20 (OMM), Tim23 (IMM), and UQCRC1 (IMM/matrix), all of which have been shown to decrease by Western blot during mitophagy 36,95-97 . Following CCCP treatment, we observed robust poly-ubiquitination and loss of The leftmost half of (G) also appears in Fig. 4D. Images of blots have been cropped; uncropped images are shown in Supplementary Fig. 12.
In the L-DOPA model, we did not observe obvious ubiquitination or a significant loss of any of the mitochondrial proteins we evaluated through 48 hours of L-DOPA treatment (Fig. 9B). To confirm the efficacy of L-DOPA treatment in these experiments, we assessed parkin levels and observed the anticipated decrease (54.1 ± 2.2% loss after 48 hours relative to time zero, p = 0.001 vs. no treatment, N = 3) (Fig. 9B). L-DOPA treatment also induced loss of the long isoform of OPA1 (45.0 ± 0.6% decrease after 48 hours relative to time zero, p = 0.01 vs. no treatment, N = 3) (Fig. 9B), consistent with a promotion of mitochondrial stress. These data suggest a presence of mitochondrial stress but a lack of robust mitophagy in the L-DOPA model.
Given that Western blotting for mitochondrial proteins could miss low levels of mitophagy, we sought to more directly assess whether phospho-Ub-dependent parkin loss requires mitophagy. Mitophagy results in the destruction of mitochondrial remains by lysosomes 50 , and blocking lysosomal degradation using the inhibitor bafilomycin A1 has been shown to effectively prevent this process 99 . To determine whether mitophagy is necessary for phospho-Ub-induced parkin loss, we examined the effect of bafilomycin A1 on L-DOPA-induced parkin loss. Bafilomycin has been previously shown to be effective at inhibiting lysosomal activity in PC12 cells 100,101 . Bafilomycin treatment failed to attenuate L-DOPA-induced parkin loss despite effectively interfering with autophagic flux (as assessed by LC3BII to LC3BI ratio) (Fig. 10) (L-DOPA: 41.1 ± 3.5% parkin loss, L-DOPA + bafilomycin A1: 36.6 ± 1.7% parkin loss, p = 0.41, N = 3). This finding indicates that phospho-Ub-dependent parkin loss does not require mitophagy or another autophagic pathway.

Discussion
In this study, we investigated the mechanism by which stressors, L-DOPA in particular, decrease cellular levels of parkin protein. We found that L-DOPA causes parkin loss via two distinct pathways: an oxidative stress-dependent pathway driven by L-DOPA autoxidation and an oxidative stress-independent pathway, each of which is responsible for about half of parkin loss (Fig. 11). We conclude that oxidative stress plays a role in only one of these pathways because parkin mutants deficient in binding phospho-Ub were only partly protected from L-DOPA-induced loss, whereas they were fully protected from loss induced by the oxidative stressor hydrogen peroxide. Furthermore, exposure to the antioxidant glutathione, which is critical for neuronal protection against ROS 102 , only attenuated L-DOPA-induced parkin loss by half. Although glutathione lacks membrane permeability 103,104 and does not react efficiently with all kinds of ROS 105 , glutathione treatment has been previously shown to be effective at completely preventing the autoxidation of both L-DOPA and the dopamine analog 6-hydroxydopamine (6-OHDA) to their respective quinones 106,107 . Additionally, glutathione treatment has been found to completely prevent cell death induced by L-DOPA 106 , 6-OHDA 107 , dopamine [108][109][110] , hydrogen peroxide 107 , and the oxidant nitric oxide 111 . In support of glutathione's capacity to prevent L-DOPA-induced oxidative stress in our hands, we observed an almost complete abrogation of L-DOPA-induced phospho-poly-Ub formation upon glutathione treatment. Nevertheless, it's possible that not all aspects of L-DOPA-induced oxidative stress were fully neutralized by glutathione in our model, which would have led us to underestimate the contribution of oxidative stress to parkin loss.
Our dose-response experiment with L-DOPA indicated that relatively high concentrations of L-DOPA (≥100 μM) are necessary to see a significant effect on parkin levels. These concentrations are significantly higher than L-DOPA levels reached in the cerebro-spinal fluid (CSF) of PD patients receiving L-DOPA, around 0.5-1 μM 112,113 . Although it's clear that patients are not acutely exposed to such high doses of L-DOPA as we use in vitro, they are exposed chronically to low doses of L-DOPA over the course of years. It's feasible that, over time, a low L-DOPA concentration could have the same effect on parkin that a high dose of L-DOPA has acutely.
Because oxidative stress has been well-documented in PD 32-34 , we focused on the oxidative stress-dependent pathway of parkin loss. We found that, in this pathway, oxidative stress leads to PINK1-mediated ubiquitin phosphorylation, mostly of high molecular weight poly-ubiquitin conjugates. Furthermore, we found that parkin's association with phospho-Ub induced by various stressors leads to its proteasomal degradation, but not through autoubiquitination, and that parkin activity in cis is not required for this process. Finally, we found that phospho-Ub-induced parkin degradation does not require mitophagy.
Our observation that phospho-Ub is nearly undetectable in untreated cells but induced upon L-DOPA exposure in a PINK1-dependent manner, coupled with the finding that L-DOPA appears to elicit mitochondrial stress (as assessed by loss of OPA1-long), suggests that L-DOPA activates PINK1, which in turn promotes poly-ubiquitin phosphorylation. Nevertheless, we can't rule out the possibility that L-DOPA-mediated inactivation of the recently-discovered phospho-Ub phosphatase, PTEN-L 114 contributes to the mechanism behind phospho-poly-Ub accumulation following L-DOPA treatment.
Our experiments with parkin point mutants reveal that parkin binding to stress-induced phospho-Ub is critical for its PINK1-dependent loss, while parkin phosphorylation, per se, is dispensable. The finding that S65A parkin was not protected from loss induced by any of the stressors we used was unexpected because phosphorylation of S65 increases the binding affinity of parkin for phospho-Ub 46,47,73,79 . Given the importance of phospho-Ub binding for parkin loss, we anticipated that S65A parkin would exhibit greater protection from stress-induced degradation due to a lower affinity for phospho-Ub. On the contrary, both wild-type and S65A parkin were (2019) 9:11682 | https://doi.org/10.1038/s41598-019-47952-5 www.nature.com/scientificreports www.nature.com/scientificreports/ equally depleted under the conditions of our studies, though we cannot rule out the possibility that wild-type parkin is lost more rapidly in response to stress than the S65A mutant.
An important question that our work raises is how parkin binding to phospho-Ub following cellular stress leads to its loss. One possibility is that the conformational change in parkin that occurs upon binding phospho-Ub [45][46][47]79,115 makes it vulnerable to degradation. If this is the case, the conformational change must be specific to phospho-Ub binding and not to that induced by parkin phosphorylation, because we show that only phospho-Ub binding is important for parkin loss. Both events have been proposed to release parkin's Ubl domain from its interaction with the RING1 domain [45][46][47]79,80 , and both have also been shown to facilitate the access of E2 ubiquitin-conjugating enzymes to parkin's E2 binding site 46,74 . Accordingly, neither of these effects is likely to be the crucial conformational change involved in parkin degradation. Instead, a conformational change that www.nature.com/scientificreports www.nature.com/scientificreports/ may promote parkin loss is movement of the IBR domain away from the Ubl domain. This movement has, thus far, only been attributed to phospho-Ub binding 47,[115][116][117] , raising it as a plausible candidate trigger for parkin loss. Alternatively, instead of inducing a critical conformational change, parkin's association with phospho-Ub may cause parkin loss by bringing it into proximity with proteins that promote its degradation.
Our finding that the proteasome mediates parkin degradation downstream of its association with phospho-Ub is in line with prior work that reported proteasomal parkin degradation following mitochondrial depolarization 40,41 or PINK1 overexpression 68 . Additionally, our results support the findings of others that parkin can be degraded proteasomally 8,86,[118][119][120] . The latter and additional studies 91,121-125 suggested that proteasomal parkin degradation is mediated by autoubiquitination, but this possibility was not directly tested. Our observation that catalytically inactive parkin was not protected from L-DOPA-induced loss indicates that parkin's proteasomal degradation downstream of stress-induced phospho-Ub binding is not due to autoubiquitination, as this only occurs in cis 91 . This finding is contrary to a model proposed to explain depolarization-induced parkin loss 40 , but is consistent with the finding that parkin autoubiquitination appears to protect it from degradation 36 .
An additional finding of interest is that parkin is a limiting factor for the generation of phospho-poly-Ub in the CCCP model but not in the L-DOPA model, because introduction of exogenous parkin led to an increase in phospho-poly-Ub formation in the former but not the latter. This indicates that, in the L-DOPA model, either a different component of parkin's positive feedback loop with PINK1, such as ubiquitination substrate or PINK1 activity, is the limiting factor for phospho-poly-Ub production, or that a different E3 ubiquitin ligase builds the phospho-poly-Ub chains following L-DOPA treatment. One piece of evidence in favour of the second possibility is that the pattern of phospho-poly-Ub staining after L-DOPA and CCCP treatment, though quite similar, is not identical (Supplementary Fig. 11). Parkin has been found to be partly functionally redundant with two other E3 ubiquitin ligases 126,127 , and the activity of these or other redundant ligases could underlie the similar banding patterns observed after L-DOPA and CCCP treatment.
We observed that CCCP treatment did not cause a loss of mitochondrial proteins, which is consistent with an absence of robust mitophagy induction. However, it is possible that low levels of mitophagy may have been present that were undetectable by our methodology. The lack of evident mitochondrial loss in the CCCP model indicates that even in the presence of potent mitochondrial damage and under conditions in which a large proportion of cellular Mfn2 has been poly-ubiquitinated, robust mitophagy may not necessarily take place, consistent with a previous report 40 . Our data also suggest that L-DOPA similarly does not induce robust mitophagy, although we cannot rule out the possibility that L-DOPA promotes mitochondrial biogenesis.
Our observation that bafilomycin treatment doesn't affect L-DOPA-induced parkin loss indicates that mitophagy is not necessary for phospho-Ub-induced parkin loss. This finding is inconsistent with suggestions that parkin degradation downstream of mitochondrial depolarization is linked to mitochondrial elimination [36][37][38][39] . Instead, our results indicate that parkin's association with phospho-Ub is the key determinant of its depolarization-and oxidative-stress-induced loss and that this loss can occur independently of mitophagy.
An important implication of our finding that phospho-Ub is involved in parkin loss is that a parkin activator (phospho-Ub) appears to be intimately linked with parkin degradation. We observed this phenomenon across different models of cellular stress, suggesting that it is broadly generalizable. Therefore, it is possible that degradation is a cellular mechanism by which activated parkin is "turned off ".
A recent study found that phospho-Ub levels are elevated in the substantia nigra of patients with Lewy body disease, a category that includes PD 128 . Given our findings, this observation suggests that parkin levels may be lowered in these patients due to enhanced phospho-Ub-induced turnover. This decrease could in turn contribute to PD pathology. As such, finding ways to uncouple parkin's association with phospho-Ub from its degradation may be an attractive therapeutic avenue for PD treatment. Collagen was dissolved in 5 mL 0.2% (v/v) acetic acid and diluted 1:20 in sterile, deionized water prior to coating plates. 139 μL/cm 2 or 17.6 μL/cm 2 of diluted collagen was added to multi-well plates or 10-cm dishes, respectively, and plates were left uncovered to dry overnight in a laminar flow hood. Cells were plated the following day.
To differentiate PC12 cells, undifferentiated cells from one 10-cm dish were first resuspended in 4 mL RPMI with 1% horse serum and 1% penicillin-streptomycin. Then, cells were passed through a 5-mL repeat pipettor tip to disperse them, counted using a hemacytometer, and plated at a density of ~3 × 10 5 cells/mL (1 mL/well in a 12-well plate) in RPMI with 1% horse serum, 1% penicillin-streptomycin, and 50 ng/mL human recombinant NGF (kind gift of Genentech, Gemini Bio-Products #300-174P). The remaining undifferentiated PC12 cells were  Figure 11. Model of L-DOPA-induced parkin loss. L-DOPA decreases parkin via two pathways: oxidative stress-dependent and -independent. The mechanism of the latter pathway remains unclear. The former pathway results from L-DOPA autoxidation, which induces PINK1 stabilization on the mitochondrial outer membrane. PINK1 phosphorylates poly-ubiquitin chains in both the cytosol and on mitochondria, though it's unclear to which substrate protein(s) these poly-ubiquitin chains are conjugated (indicated by single question marks in the model). Parkin associates with these phospho-poly-Ub chains. Following its association with phospho-poly-Ub, parkin is degraded via the proteasome independently of autoubiquitination and mitophagy. www.nature.com/scientificreports www.nature.com/scientificreports/ plated on a freshly-coated 10 cm dish. Cells were used in experiments after 5-10 days of differentiation. In experiments with lentiviral transduction, cells were transduced after 3-5 days of differentiation; for shRNA constructs, cells were treated with drugs 4-5 days after transduction; for overexpression constructs, cells were treated with drugs 3-5 days after transduction.
For both differentiated and undifferentiated cells, half of the medium was replaced every 2-3 days with fresh medium and cells were kept in an incubator under 7.5% CO 2 .
Primary embryonic rat cortical neurons. Cortical neurons were grown on plates coated with poly-D-lysine (P1149, Millipore Sigma). 139 μL/cm 2 of 0.1 mg/mL poly-D-lysine was added to multiwell plates, and plates were left covered in the cell culture incubator overnight. The following day, the poly-D-lysine solution was aspirated and the wells were washed at least three times with sterile deionized water. The wells were allowed to dry and cells were plated the same day.
Cortices were harvested from E17-E18 rat embryonic brains in cold HBSS without calcium chloride, magnesium chloride, or magnesium sulfate (Thermo Fisher Scientific) using a dissection microscope. The cortices were dissociated by a 10-minute incubation in 0.05% trypsin at 37 °C and subsequent trituration in HBSS with two fire-polished pasteur pipettes of decreasing aperture size. Neurons were then passed through a cell strainer, counted, and plated at 5 × 10 5 cells/mL on poly-D-lysine-coated plates in Neurobasal medium supplemented with 2% B27 supplement (Thermo Fisher Scientific), 500 mM Glutamax (Thermo Fisher Scientific), and 1% penicillin-streptomycin.
Half of the medium was replaced with fresh medium twice a week, and cells were kept in an incubator under 5% CO 2 . Cells were used for experiments after 6-9 days in vitro.
All experiments involving preparation of rat cortical cultures were carried out in accordance with the guidelines set forth in the NIH Guide for the Care and Use of Laboratory Animals (2011) and with approval of the Columbia University Institutional Animal Care and Use Committee (Animal Care Protocol AC-AAAR7401).
Immortalized mouse embryonic fibroblasts. Immortalized MEFs were a kind gift from Dr. Thong Ma from the Un Kang lab (Columbia University). The isolation and immortalization of these cells were described in 68 . Cells were grown in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin under 5% CO 2 .
Tagged parkin under minimal parkin promoter. Two steps were involved in generating a "moderately overexpressed" parkin construct. First, rat parkin cDNA was amplified from a previously generated parkin construct www.nature.com/scientificreports www.nature.com/scientificreports/ (described in 7 ) using primers to attach HA-and FLAG-tags on the N-terminus. The primers used were: F: ctagcctcgaggtttaaacATGTACCCATACGATGTTCCAGATTACGCTGCAGCAAATGATATCCTGGATTAC AAGGATGACGACGATAAGctcccgcggacgcgtacgATAGTGTTTGTCAGG; R: tgcagcccgtagtttaaacctaCACGTC AAACCAGTGATC. This amplified, N-tagged cDNA was then inserted into a pWPI vector (Addgene) that had been linearized at the PmeI restriction site. In-Fusion cloning (Clontech) was used for this purpose, according to the manufacturer's instructions.
In the second step, the EF-1α promoter of the pWPI-HA-FLAG-parkin construct was cut out at the PpuMI and PacI sites, and the minimal human parkin promoter amplified from a commercially available luciferase construct (SwitchGear Genomics) was inserted in its place using the NEBuilder ® HiFi DNA Assembly kit from New England Biolabs. The primers used to amplify the minimal human parkin promoter were: F: cagacccacctcccaaccccgagggga cccCACCTTCTATAAGGCCCTTTG; R: gataccgtcgagattaattaaatttaaatCGTACCTATCATGGTCAC.
Site-directed mutagenesis. Site-directed mutagenesis of "moderately overexpressed" parkin was carried out using the Q5 Site-Directed Mutagenesis kit from New England Biolabs, according to the manufacturer's instructions.
The next day, cells were transfected with 21 μg expression vector, 16.5 μg psPAX2, and 7.5 μg VSVg plasmids per plate using the calcium phosphate transfection method. Virus-containing medium was harvested 2 and 3 days after transfection, pooled, passed through a 0.45 μM filter, and concentrated using the Lenti-X concentrator (Clontech, #631231), according to the manufacturer's instructions.
To evaluate viral titer, a range of viral solution volumes was added to differentiating PC12 cells, cells were fixed 3 days later, and GFP expression was evaluated by immunofluorescence. Viral infection efficiency was calculated by dividing the number of GFP-positive cells by the number of nuclei in the same field (assessed by Hoechst stain) using Fiji. For knockdown experiments, viral solution volumes yielding ≥70% infection efficiency were used, with an effort to match infection efficiencies between viruses. For exogenous parkin experiments, viral titer was assessed by Western blot, with viral volumes for subsequent experiments adjusted to achieve equal expression levels between constructs.
L-DopA, hydrogen peroxide, and cccp treatments. Prior to drug treatment, half of the medium was removed from cells. An equal volume of fresh medium containing drugs at 2X the working concentration was added to cells to initiate the experiment.
For L-DOPA experiments, a 10 mM L-DOPA stock solution was freshly prepared before every experiment in 50 mM HCl and filter-sterilized. L-DOPA stock solution or an equal volume of 50 mM HCl was diluted in fresh medium before addition to cells. In experiments involving pretreatment of cells with an inhibitor, the inhibitor was added with fresh medium and L-DOPA was added directly to each well. 3% (w/v) hydrogen peroxide was diluted with sterile distilled water 1:1000 before subsequent dilution in fresh medium.
CCCP was dissolved in DMSO to yield a 10 mM stock solution, which was filter-sterilized and added to fresh medium before addition to cells. The stock solution was stored at −80 °C and used multiple times.
inhibitor & antioxidant treatments. Inhibitors and antioxidants were dissolved in water when possible and in DMSO when not water-soluble. In most cases, stock solutions of the drugs were diluted in fresh medium before addition to cells (with the exception of cycloheximide, in which case fresh medium was added to cells the day before treatment and cycloheximide stock solution was added directly to cells). In such cases, drugs and cellular stressors (L-DOPA, H 2 O 2 , CCCP) were added to cells at the same time. In the cases of carbidopa and cycloheximide, the latter were added to cells before treatment with L-DOPA. Cells were pretreated with carbidopa for 1.5 hours and cycloheximide for ~15 minutes before L-DOPA addition.
Western blotting. Cell lysates were harvested in 1X Cell Lysis Buffer (Cell Signaling Technology) supplemented with Complete Mini protease inhibitor tablet -EDTA (Millipore Sigma) or directly in Western blot loading buffer (29% Cell Lysis Buffer, cOmplete Mini protease inhibitor tablet, 1X NuPage Reducing Agent (Thermo Fisher Scientific), 1X NuPage LDS Sample Buffer, 36% water). Lysis buffers were added directly to culture plate wells.
When harvested in Cell Lysis Buffer, lysates were kept on ice and subsequently sonicated with a probe sonicator at the lowest amplitude with a pattern of 1 second ON/1 second OFF, for a total of 10 seconds ON. The protein concentration in lysates was then determined using the Bradford assay using Protein Assay Dye Reagent Concentrate (Bio-Rad). Upon determining protein concentrations, samples were diluted with water to match the concentration of the most dilute sample and supplemented with NuPage Reducing Agent and NuPage LDS Sample Buffer.
Following addition of Reducing Agent and LDS Buffer to cell lysates, these were boiled for 20 minutes in a dry heat bath at 95-100 °C. After boiling, sample proteins were separated by SDS-PAGE and detected by Western blotting. The NuPAGE electrophoresis system was used, with precast Bis-Tris gels, MOPS SDS running buffer (or MES SDS running buffer for low molecular weight proteins), and NuPAGE transfer buffer. Generally, 4-12% gels were used and ~20 μg of protein was added per well. For one sample per gel, half the sample was added in a well adjacent to the full volume of sample to serve as a standard (see below). Proteins were transferred onto a nitrocellulose membrane for ~2 hours at 35 V at 4 °C. Following transfer, membranes were blocked for 1 hour at room temperature with gentle shaking in 5% milk in TBST (TBS + 1% Tween 20).
Membranes were probed with primary antibodies diluted in TBST with 5% BSA overnight at 4 °C. The next day, membranes were subjected to 3 five-minute washes in TBST, incubated for 1 hour at room temperature with (2019) 9:11682 | https://doi.org/10.1038/s41598-019-47952-5 www.nature.com/scientificreports www.nature.com/scientificreports/ LI-COR IRDye secondary antibodies in 5% BSA/TBST in the dark, and washed three times again. Following incubation with secondary antibodies, membranes were kept in TBS in the dark.
Membranes were imaged using an Li-Cor Odyssey CLX scanner and band intensities were quantified using Image Studio Lite software (ver. 4.0.21). Briefly, after background subtraction, the signal from the lane in which a half volume of sample was loaded was assigned a value of "0.5" using the "concentration standard" feature. Similarly, the signal from the lane corresponding to full volume of sample was assigned a value of "1". The program's linear interpolation feature was then used to assign values to the remaining bands on the blot. This process was carried out both for proteins of interest and for loading controls, after which the former were normalized to the latter. For most experiments, bands of interest were normalized to the average of the corresponding ERK1 and ERK2 signals (which was confirmed not to be affected by L-DOPA treatment), though sometimes Ponceau stain was also used as a loading control. Subcellular fractionation. To prepare crude mitochondrial and cytosolic fractions from cells, the following protocol was followed. Cells and lysates were kept on ice for the duration of the procedure, and all centrifugation steps were performed at 4 °C. First, cells were sprayed from the dish in PBS and pelleted by centrifugation. Cells from four wells of a 24-well plate were combined in one tube. 70 μL fractionation buffer was added to the cell pellet. The fractionation buffer consists of 220 mM mannitol, 70 mM sucrose, 20 mM Hepes-KOH (pH 7.5), 2 mg/ ml BSA, 1 mM EDTA. Halt protease/phosphatase inhibitor (Thermo Fisher Scientific #78440) was added to the buffer right before use.
Cells were lysed by passage through a 27G needle thirty times, until there were about 8 free nuclei for every unbroken cell. To remove cell debris & nuclei, samples were centrifuged at 1,000 × g for 10 minutes. The supernatant was set aside and the nuclear fraction was washed with 60 μL fractionation buffer and centrifuged a second time. The supernatant from this second spin was combined with the supernatant from the first spin and centrifuged at 10,000 × g for 10 minutes to separate the mitochondria-rich pellet from the cytosolic fraction. The mitochondria-rich pellet was washed three times by resuspending the pellet in 60 μL fractionation buffer and re-pelleting at 10,000 × g for 10 minutes. Then, the mitochondrial pellet was resuspended in 13 μL Cell Lysis Buffer (Cell Signaling Technology) with cOmplete Mini Protease inhibitor (Millipore Sigma). The mitochondrial and cytosolic fractions were supplemented with NuPAGE Reducing Agent and LDS sample buffer (Thermo Fisher Scientific), boiled for 20 minutes at 95 °C, and used for Western blotting following the standard protocol.
qpcR. Analysis of mRNA by qPCR was performed as in 7 . Briefly, total cellular RNA was extracted using TRI reagent (Molecular Research Center). cDNA was synthesized using the first-strand cDNA synthesis kit (Origene). qPCR was performed using FastStart SYBR Green Master Mix (Roche) and an Eppendorf Realplex Mastercyler. High performance liquid chromatography. Cells were washed twice with cold PBS, detached from wells via forceful pipetting, and pelleted at maximum speed in a table-top centrifuge for 5 minutes at 4 °C. The supernatant was then removed and the cell pellet was resuspended in 75 μL PBS. 25 μL of this cell mixture was removed, pelleted, and resuspended in 40 μL Cell Lysis Buffer with protease inhibitors. These samples were then sonicated, and the Bradford assay was performed to determine protein concentrations. The remaining 50 μL of cell mixture was mixed with an equal volume of 0.5 M trichloroacetic acid and vortexed for 10 seconds. These samples were then centrifuged at 10,000 × g for 2 minutes at 4 °C, and the supernatant was transferred to a new tube. This supernatant was used for HPLC.
Samples were separated on Brownlee VeloSep RP-18, 3 μM, 100 ×3.2 mm column (Perkin Elmer, Waltham, MA) using a Gilson (Middleton, WI) isocratic 307 pump. The mobile phase consisted of 45 mM NaH 2 PO 4 , pH 3.2, 0.2 mM EDTA, 1.4 mM heptanesulfonic acid and 5% methanol. L-DOPA peaks were detected using an ESA (Chelmsford, MA) Coulochem II electrochemical detector at 300 mV oxidation potential. Relative L-DOPA levels were calculated by normalizing the area under the L-DOPA HPLC peak to the protein concentration from the same sample.
Immunofluorescence (for quantification of viral titer). Cells were fixed for 15 min in 4% paraformaldehyde and then washed 3 times with 1X PBS. Cells were then blocked with Superblock blocking buffer supplemented with 0.3% Triton X-100 for 1-2 h at room temperature and incubated overnight at 4 °C with primary antibody in Superblock/Triton X (chicken anti-GFP, 1:1000, #A10262, Thermo Fisher Scientific) with gentle shaking. The next day, primary antibody was washed off in 3 × 8-minute washes with gentle shaking using 1X PBS. Cells were then incubated with fluorescent secondary antibody in Superblock/Triton X for 1 hour with gentle shaking (AlexaFluor-488 anti-chicken, 1:1000, #A11039, Thermo Fisher Scientific). Cells were then washed 3 × 8 minutes in 1X PBS with gentle shaking. Hoechst 33328 was added in the first wash (1:2500). Cells were kept in 1X PBS before imaging using an inverted fluorescence microscope.
Quantification of cell death. Cell death was assessed as previously described 129 . Briefly, cells were treated with a lysis buffer (150 μL lysis buffer per cm 2 ) that disintegrates the plasma membrane while keeping nuclei intact, and viable nuclei were counted manually using a hemacytometer. At least 200 nuclei were counted per condition. 10X lysis buffer recipe: 5 g of cetyldimethyl-ethanolammonium bromide, 0.165 g of NaCl, 2. www.nature.com/scientificreports www.nature.com/scientificreports/ Statistical analysis. Statistical analysis was performed using GraphPad Prism 6 and using an online multiple comparison correction resource (http://alexandercoppock.com/statistical_comparisons.html) developed by Alexander Coppock, as described in 130 .
For most experiments, values were normalized to the control condition in that experiment before replicate experiments were combined. Paired t-tests were performed in most cases, followed by Holm correction for multiple comparisons. In time-course analyses, unpaired t-tests at each time point followed by Holm-Sidak's multiple comparisons test were used. For analysis of differences between parkin mutants, one-way ANOVA of stressor-treated mutants was performed, followed by Holm-Sidak correction for multiple comparisons. For analysis of the cycloheximide experiment, repeated measures 2-way ANOVA of data from the 12-48 hour time points was used with Sidak's multiple comparisons test.

Data Availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.