Comparative analysis of Parkinson’s disease–associated genes in mice reveals altered survival and bioenergetics of Parkin-deficient dopamine neurons

why alterations in such diverse cellular processes lead to the selective loss of dopamine (DA) neurons in the substantia nigra (SNc) in the of individuals with PD. Recent work has shown that one of the reasons for the high vulnerability of SNc DA neurons is their high basal rate of mitochondrial oxidative phosphorylation (OXPHOS), resulting from their highly complex axonal arborization. Here, we examined whether axonal growth and basal mitochondrial function are altered in SNc DA neurons from Parkin-, Pink1-, or DJ-1–KO mice. We provide evidence for increased basal OXPHOS in Parkin-KO DA neurons and for reduced survival of DA neurons that have a complex axonal arbor. The surviving smaller neurons exhibited reduced vulnerability to the DA neurotoxin and mitochondrial complex I inhibitor MPP (cid:3) , and this reduction was associated with reduced expression of the DA transporter. Finally, we found that glial cells play a role in the reduced resilience of DA neurons in these mice and that WT Parkin overexpression rescues this phenotype. Our results provide critical insights into the complex relationship between mitochondrial function, axonal growth, and genetic risk factors for PD.

In PD, 5 chronic progressive loss of SNc DA neurons leads to some of the characteristic motor symptoms of the disease. Over the past years, many mutations in genes coding for proteins such as the E3 ubiquitin ligase Parkin, PTEN-induced putative kinase 1 (Pink1), protein deglycase DJ-1, leucine-rich repeat kinase 2 (LRRK2), and ␣-synuclein have been linked to familial forms of the disease.
Although the mechanisms underlying the effects of these mutations, such as alteration in mitochondrial function, mitophagy, lysosomal and proteasome function, and protein aggregation, are slowly starting to be better understood (1), little is known about the reason why alterations in such ubiquitous cellular mechanisms lead to selective loss of sub-populations of neurons, including SNc DA neurons. An emerging hypothesis suggests that the selective vulnerability of DA neurons and other cell groups in PD can be explained in large part by the fact that they establish an unusually complex axonal arborization (2)(3)(4). This characteristic has been associated with particularly high demands in ATP needed to sustain neurotransmission along profuse axons (5)(6)(7)(8).
Arguing in favor of this hypothesis, recent work showed that SNc DA neurons have a higher basal rate of mitochondrial OXPHOS and ATP production and a smaller reserve capacity compared with less vulnerable DA neurons of the ventral tegmental area (VTA) (9), characteristics that appear to be the result of the highly complex axonal arborization of these neurons and that are also associated with a high level of oxidative stress. Based on these results, a question that arises is whether bioenergetic parameters or axon growth are perturbed in genetic mouse models of PD. Here, we examined primary SNc and VTA DA neurons obtained from Parkin-, Pink1-, or DJ-1-KO mice. We found in vitro that basal survival, axonal growth, and mitochondrial function are altered in SNc DA neurons from Parkin-KO mice but not from Pink1-or DJ-1-KO mice, resulting in an increased oxygen consumption rate (OCR) but lower ATP content. We also found that astrocytes from Parkin-KO mice had a slower growth rate and lower OCR and that their replacement with WT astrocytes abolished the reduction in basal survival observed in Parkin-KO mouse SNc DA neurons. Finally, we show that overexpression of WT Parkin rescues the survival, axonal morphology, and bioenergetics of SNc Parkin-KO mouse cultures. Our data provide new insights in the complex relationship between cellular bioenergetics, axonal development, and survival of DA neurons in Parkin-, Pink1-, and DJ-1-KO mouse models of PD, highlighting a critical role for Parkin.

SNc DA neurons in Parkin-but not PINK1-or DJ-1-KO PD mouse models show altered basal survival and growth
Experiments were performed using a mouse primary culture system, including DA neurons microdissected from the SNc and VTA of newborn Parkin-, Pink1-, or DJ-1-KO mice. The neurons were grown on a supporting monolayer of astrocytes of the corresponding genotype. As reported previously (9), in such cultures, WT SNc DA neurons show increased basal respiration, increased ATP production, and a larger axonal arborization compared with WT VTA DA neurons (Fig. S1).
We first evaluated the rate of spontaneous loss of DA neurons in vitro over a period of 11 days as an index of the neurons' basal resilience. At 11 DIV, a time point at which DA neurons are mature at the morphological and metabolic levels, we found that Parkin-KO SNc DA neurons showed a pronounced rate of spontaneous degeneration, with 56% more cell loss compared with WT SNc DA neurons (Fig. 1A). We did not observe any similar enhancement of SNc DA neuron loss in Pink1-KO or DJ-1-KO neurons (Fig. 1, B and C) or in VTA cultures from any of the three genotypes ( Fig. 1, D-F). Interestingly, there was also no change in survival at an earlier time point in SNc Parkin-KO culture (5DIV, Fig. S2A), arguing for the need of sustained stress for this increased vulnerability to be revealed. We next examined axonal arborization size in the remaining Parkin-KO mouse SNc DA neurons at 11 DIV and found that it was 40% smaller compared with WT ( Fig. 1, G and H), with no change in Parkin-KO mouse VTA neurons. Again, we did not observe any differences between Pink1-WT and -KO mouse DA neurons, between DJ-1-WT and -KO DA neurons (Fig. 1, I and J), or at an earlier time point in Parkin-WT mouse neurons versus KO mouse DA neurons (Fig. S2B). These results suggest that Parkin-KO mouse SNc DA neurons have a higher intrinsic vulnerability and that the most arborized DA neurons are degenerating preferentially or, alternatively, that axonal development is slowed in these neurons after prolonged development in vitro.

Parkin but not PINK1-or DJ-1-KO DA neurons show altered mitochondrial function
Because recent work has provided evidence for a close relationship between survival, axonal arborization size, and mitochondrial bioenergetics in DA neurons (9), we next evaluated mitochondrial OXPHOS by measuring cellular respiration. At 10 DIV, basal and maximal (uncoupled with CCCP) oxygen consumption rates (OCR) were measured in living neurons. The respiratory control ratio (RCR), corresponding to the ratio between basal and maximal OCR, was also determined to estimate the respiratory reserve capacity. We found that basal OCR was significantly increased by 50% in Parkin-KO mouse SNc DA neurons compared with WT ( Fig. 2A). However, maximal OCR or the RCR were unchanged in these neurons ( Fig. 2A). OCR was unchanged in SNc DA neurons from Pink1-KO mice or DJ-1-KO mice (Fig. 2, B and C) or at 5 DIV in SNc Parkin-KO mouse culture (Fig. S2C). Interestingly, basal OCR was significantly increased by 181% in Parkin-KO mouse VTA DA neurons, accompanied by an increase of maximal OCR of 120%, with no changes in RCR (Fig. 2D). OCR was unchanged in VTA DA neurons form Pink1 or DJ-1-KO mice.
Considering previous work suggesting a critical role for Parkin in the turnover of mitochondria (10 -12), a possible interpretation of the increase in basal OCR in Parkin-KO mouse DA neurons is that loss of Parkin leads to the presence of dysfunctional mitochondria that are partially uncoupled. If this were the case, a prediction is that the increase in OCR should not be accompanied by an increase in ATP production and could even be accompanied by a decrease in ATP. To test this hypothesis, we measured ATP content in each culture and discovered a significant reduction of 44% in ATP levels in Parkin-KO mouse SNc cultures, with no change in Pink1-KO or DJ-1-KO mouse cultures (Fig. 2, G-I). Such a possible uncoupling between OCR and ATP production appears to be particularly exacerbated in Parkin-KO mouse VTA DA neuron cultures, perhaps because more of these neurons survived to this impairment, compatible with their higher resilience in response to oxidative stress (9). Interestingly, contrary to SNc DA neurons (Fig. 3A), VTA DA neurons from Parkin-KO mice showed an increase in basal glycolysis (Fig. 3B), perhaps representing a compensatory mechanism for impaired mitochondrial function. This was not observed in SNc or VTA neurons from either PINK1-or DJ-1-KO mice (Fig. 3, C-F).
Considering our data arguing for defective mitochondrial function in SNc DA neurons from Parkin-KO mice, the distribution and density of mitochondria in DA neurons from this genetic background were also directly examined by infecting cells with a mitochondrially-tagged DsRed protein (DsRed2mito-lentivirus) (Fig. 4A). No significant difference was detected in the density of mitochondria in the soma, axon, or dendrites of SNc DA neurons from Parkin-KO mice compared with WT (Fig. 4B).

Parkin-KO mice glia show altered growth, extracellular acidification rate (ECAR), and OCR and influence basal survival of SNc DA neurons
The reduced resilience of SNc DA neurons cultured from Parkin-KO mice could be due to a cell-intrinsic perturbation, to perturbation of non-neuronal cells such as astrocytes or microglia, or a combination of the two. Previous work has indeed demonstrated reduced proliferation of astrocytes, increased density of microglia, and reduced neuroprotective effects of glial-conditioned medium from Parkin-KO mice (13). We therefore evaluated the implication of glial cells in the altered

Altered survival and bioenergetics in Parkin-KO mouse neurons
survival of the Parkin-KO mouse SNc DA neurons by growing Parkin-KO mouse DA neurons with WT glia. We found that this completely reverted this reduced survival phenotype (Fig.   5A). Compatible with previous results (13), there were 33% fewer cells in Parkin-KO mouse cortical glial cultures after 7 DIV, 36% less at 10 DIV, and 39% fewer cells at 10 DIV when a

Altered survival and bioenergetics in Parkin-KO mouse neurons
mitotic inhibitor was added after glial cells reached confluence, conditions that mimicked those used to prepare neuronal cultures (Fig. 5, B and C). The proliferation of mesencephalic glial cells was similarly reduced (Fig. S3). Glial cells in Parkin-KO mouse cultures were also perturbed at the metabolic level, as revealed by the fact that basal OCR was reduced by 24% and maximal OCR by 25%, with no change in RCR (Fig. 5D). In addition, basal glycolysis was increased by 52% (Fig. 5E), and no change was observed in ATP levels (Fig. 5F), suggesting the possibility of a compensation by glycolysis of the reduced OXPHOS. Similar measurements performed in Pink1-KO mice and DJ-1-KO cultures revealed a decreased basal and maximal OCR in glial cells from Pink1-KO mice but not DJ-1-KO mice, with no changes in any other measurements (Fig. S4).

Vulnerability to the neurotoxin MPP؉ is altered in Parkin, DJ-1, but not PINK1-KO mouse SNc DA neurons
The reduced basal survival of SNc DA neurons from Parkin-KO mice (Fig. 1A) together with the observation of reduced axonal arborization size (Fig. 1H) in these surviving neurons suggest the possibility that the surviving smaller neurons might be more resilient, as in a recent model (4,9). To test this possibility, we next examined the vulnerability of DA neurons at 10 DIV to cellular stress induced by the DA neuron-specific neurotoxin and mitochondrial complex I inhibitor MPPϩ. Interestingly, the surviving SNc DA neurons from the Parkin-KO mouse cultures were significantly less vulnerable to MPPϩ (Fig.  6A). No change in vulnerability was observed in SNc DA neurons from Pink1-KO mice (Fig. 6B). Finally, SNc DA neurons from DJ-1-KO mice were to the contrary more vulnerable to MPPϩ (Fig. 6C). Compatible with previous work showing a higher resilience of VTA DA neurons, MPPϩ caused much smaller cell loss in VTA DA neurons, and no difference was found between genotypes (Fig. 6, D-F).

Reduced DA transporter expression in surviving SNc DA neurons in Parkin-KO mice
Although reduced vulnerability of surviving SNc DA neurons from Parkin-KO mice could be due to a number of factors, an obvious possibility is that reduced toxicity could be due to

Altered survival and bioenergetics in Parkin-KO mouse neurons
lower levels of the DA transporter (DAT), which is the route of entry of MPPϩ in DA neurons (14). We measured DAT density using a radio ligand assay and found that the surviving DA neurons indeed showed 60% less DAT compared with WT SNc DA neurons (Fig. 7A). Also arguing in favor of this hypothesis, vulnerability to DAT-independent toxic treatments such as hydrogen peroxide (Fig. 7B) 7D).

Partial rescue of Parkin-KO mouse SNc DA neurons by WT Parkin overexpression
Finally, to further strengthen the idea that the reduced basal survival, axonal length, and increased basal OCR in SNc DA neurons from Parkin-KO mice is a direct consequence of the

Altered survival and bioenergetics in Parkin-KO mouse neurons
loss of Parkin expression, we overexpressed WT Parkin using an AAV vector (Fig. 8A). We found that the basal survival of SNc DA Parkin-KO mouse neurons was partially rescued by Parkin overexpression (Fig. 8B) and that their axonal length and basal OCR was also back to WT levels (

Discussion
Since the discovery of the first human mutations causing familial forms of Parkinson's disease, extensive work has been performed to identify the mechanisms by which loss-or gainof-function of gene products, including Parkin, Pink1, DJ-1, LRRK2, and ␣-synuclein, trigger the disease process. One of the approaches used has been to create knockout animals in species such as flies and mice in the hope of obtaining valid models of the disease. Unfortunately, in mice, constitutive-KO of these genes typically does not lead to age-dependent loss of DA neurons in vivo (15)(16)(17). A possible explanation could be that it is difficult to model age-dependent neurodegeneration in species with a short life span or that mice need to be exposed to a second hit to trigger the disease. Another possibility is that the cumulative burden of oxidative stress in mouse DA neurons is not sufficiently high to lead to cell loss. Whatever the case may be, using these models is still of interest to examine some of the early stages of the disease, prior to cell loss. Indeed, various cellular and subcellular dysfunctions have been reported in such models during the past decade (18 -22). In this work, we aimed to determine whether basal resilience and parameters linked to cellular bioenergetics such as mitochondrial OXPHOS and axonal arborization size are altered in primary cultures prepared from Parkin-, Pink1-, or DJ-1-KO mice.

Morphological and bioenergetic alterations in Parkin-KO mouse SNc cultures
We discovered that basal survival in vitro, axonal growth, and mitochondrial functions are altered in SNc DA neurons from Parkin-KO mice, with little if any changes in Pink1-KO mice or DJ-1-KO mice. Our finding of reduced basal survival of SNc DA neurons in Parkin-KO mice argues that SNc DA neurons in these mice have a higher basal vulnerability to cellular stress, a conclusion that is in keeping with previous results (9,23,24). Interestingly, we found that the subset of SNc DA neurons that survived up to 11 DIV in Parkin-KO mouse cultures possessed a smaller axonal arborization and produced less ATP, findings that are compatible with previous work proposing that cellular bioenergetics are directly related to axon length and to vulnerability (9). Further experiments directly evaluating the links between ATP levels and axon growth in DA neurons would help establish more direct links between these parameters. Our observation of reduced axonal arborization size is also in line with recent work showing reduced complexity of neuronal processes in induced pluripotent stem cell-derived human neurons carrying Parkin mutations (25). However, somewhat surprisingly, surviving Parkin-KO mouse SNc DA neurons at 10 days in vitro showed higher basal OCR, suggesting an uncoupling between oxygen consumption and ATP production. This observation could be the result of the accumulation of damaged mitochondria and as such would be compatible with previous work showing mitochondrial dysfunction in Parkin-KO mouse mutant Drosophila (26). However, no increase in OCR was reported in mouse cortical or striatal primary cultures or in mesencephalic post-nuclear supernatants prepared from adult Parkin-KO mice (27). Because such cultures do not contain any DA neurons and because OCR measurements from mesencephalic post-nuclear supernatants are necessarily contaminated by non-DA neurons and other cells, it is possible that a change in OCR occurring principally in DA neurons may have been missed in these previous studies.

Impaired mitochondrial function in SNc Parkin-KO cultures
Considering that Parkin is normally recruited by Pink1 to damaged mitochondria to initiate mitophagy, it is possible that the increase in OCR with a reduction in ATP levels that we observed in residual SNc DA neurons after 10 days in vitro is the result of the accumulation of malfunctioning, uncoupled mitochondria. Although we cannot exclude this possibility, our observation of a lack of change in mitochondrial density, evaluated with Mito-DsRed, does not provide support for such an interpretation. One possibility is that other mechanisms compensate for the lack of Parkin-mediated mitophagy. Such mechanisms could either get rid of some of the damaged mitochondria in a Parkin-independent manner and/or slow down the production of new mitochondria, explaining the lack of change of total mitochondrial density. Although recent work suggests that the rate of mitophagy in axons may be very limited (28)

Altered survival and bioenergetics in Parkin-KO mouse neurons
in OCR combined with decreased ATP production argues for malfunctioning mitochondria (9,22,29,30). In this context, the resulting metabolic stress could favor the survival of neurons with smaller bioenergetic needs such as those that have a smaller axonal arborization. Arguing in favor of this interpretation, VTA DA neurons, which are less vulnerable in PD and normally have a smaller axonal arborization than SNc DA neurons (9), did not show altered survival in the absence of Parkin under our experimental conditions. We found that even if they showed a much higher basal OCR, their ATP production and axonal arborization were unchanged. This highlights the higher basal resilience of these neurons. Further experiments quantifying basal reactive oxygen species production and antioxidant defense mechanisms in VTA DA neurons would be helpful to further characterize and explain such resilience. Our finding of a lack of change in bioenergetics and axonal growth in DA neurons from Pink1-and DJ-1-KO mice is puzzling, especially concerning the implication of Pink1 in mito-

Altered survival and bioenergetics in Parkin-KO mouse neurons
chondrial quality control, in a common pathway with Parkin. A partial explanation could come from recent work showing that mitochondrial quality control in tissues with elevated metabolic demands such as the brain can be Pink1-independent (31). If this is also the case in our model, this could explain why we only see alterations in Parkin-KO mouse cultures. Except for a recent study reporting age-dependent unilateral loss of SNc DA neurons in a new sub-strain of DJ-1-KO mice (15), there have been no previous reports of altered basal growth and survival of DA neurons in Pink1-or DJ-1-KO mouse lines. However, previous work has reported impaired respiration and increased glycolytic activity in mixed neuronal cultures from the midbrain of Pink1-KO mice (32). The relative abundance of DA neurons in such cultures was not reported, making it unclear whether results reflected the bioenergetic properties of SNc DA neurons. In addition, mitochondrial membrane potential was used to indirectly evaluate OXPHOS, making a comparison with our present results difficult. Concerning DJ-1, previous work using a neuron-like cell line showed reduced OCR and ATP production (33). It is very difficult to compare our findings to such results because of the major morphological and bioenergetic differences between such cell lines and primary DA neurons, in addition to the major difference in cell maturity between the two studies (18 h in the Heo et al. study (33) compared with 10 days for the present study). In this study, although increased cell loss was apparent after 11 days in Parkin-KO mouse cultures, we found no change after 5 days. Similarly, loss of Pink1 or DJ-1 might therefore need more time to reveal an increased vulnerability, as suggested by work on aged Pink1-KO mice (34). Further studies evaluating neuronal survival in DA neurons from Pink1-or DJ-1-KO mice after longer time points in culture might thus be useful to determine whether changes in vulnerability do exist but simply need a higher level of cellular stress or time to be revealed.

WT Parkin overexpression rescues morphological and bioenergetic defects in Parkin-KO mouse SNc cultures
Because the KO strain used in this work was constitutive, the increased vulnerability of SNc DA neurons we detected in Parkin-KO mouse cultures could result from changes occurring during embryonic development of the neurons and not be a direct consequence of the lack of Parkin expression in cells. To explore this possibility, we performed a rescue experiment and reintroduced WT Parkin using an AAV vector. We found that such overexpression was sufficient to rescue the changes in basal survival, axonal growth, and mitochondrial OXPHOS in Parkin-KO mouse SNc DA neurons, observations that are compatible with previous work showing that overexpression of WT Parkin is neuroprotective (35,36). An important limitation of the present rescue experiment is that Parkin expression was driven through a strong CBA promoter, likely leading to expression levels much higher than WT levels (Fig. S5).

Parkin-KO glial cell alterations and implications for neuronal defects
The increased vulnerability of SNc DA neurons we detected in Parkin-KO mouse cultures could result from cell-autonomous alterations, from non-cell-autonomous mechanisms, such as a change in glial cell function, as reported previously (13), or a combination of the two. To explore the implication of astrocytes, cells that play a key role in the trophic support of DA neurons (37,38), we prepared mixed cultures containing WT

Altered survival and bioenergetics in Parkin-KO mouse neurons
glia and Parkin-KO mouse DA neurons. In such cultures, we observed no significant differential loss of SNc DA neurons over 10 days in vitro, arguing for a major implication of glial cell dysfunction in the reduced survival of SNc DA neurons in cultures prepared from Parkin-KO mice. Compatible with glial cell dysfunction in such mice, we found slower proliferation of Parkin-KO mouse glial cells, an observation that is in line with previous work (13,39). Such slower proliferation of glial cells (mostly astrocytes) led to a reduced density of such cells at 10 DIV, accompanied by a reduction in OCR, but an increase in glycolysis, with no changes in ATP levels. It is possible that the reduction in glial cell number itself influenced neuronal survival in Parkin-KO mouse cultures. However, we feel that this is unlikely because we found that very large reductions in glial cell density are required to negatively impact DA neuron survival (data not shown). Here, the reduction in glial cell density was modest (ϳ33%) and was not accompanied by a decrease in ATP production. Future work will be required to determine whether the secretion of glial cell-derived neurotrophic factors or the regulation of lactate or pyruvate shuttling is altered in Parkin-KO mouse astrocytes, providing a potential explanation for the reduced survival of SNc DA neurons.

Lower DAT level in surviving Parkin-KO SNc neurons
SNc DA neurons are well-known to be particularly vulnerable to mitochondrial toxins such as MPPϩ, which enter DA neurons through the DAT (14). DA neurons in Pink1-and DJ-1-KO mice have been reported to be more vulnerable to different toxin models, suggesting increased cell vulnerability (21, 40 -45). Surprisingly, this is not the case for Parkin-KO mice (30, 46 -49). We therefore also investigated this issue in this study using MPPϩ, the active metabolite of MPTP. Interestingly, we found increased vulnerability to MPPϩ only in SNc DA neurons cultured from DJ-1-KO mice, but not in cultures prepared from Pink1-or Parkin-KO mice, arguing that some of the properties of DJ-1, such as its antioxidant role, is important to prevent toxicity to MPPϩ, even if it is not necessary for survival up to 11 DIV under our basal culture conditions. Because the MPTP experiments cited previously were performed in adult mice, it is possible that the developing state of our neurons prevented us from seeing such increased vulnerability in Pink1-KO mouse cultures. Surprisingly, in Parkin-KO mouse SNc cultures, the neurons surviving after 10 DIV were actually less vulnerable to MPPϩ. This could perhaps be due to

Altered survival and bioenergetics in Parkin-KO mouse neurons
lower expression of DAT in surviving DA neurons, a downregulation reported previously when this mouse was first characterized (48) and more recently in Parkin mutant human midbrain DA neurons derived from induced pluripotent stem cells (50). Compatible with this, we did observe a lower expression of DAT in Parkin-KO mouse cultures. In addition, we found no change in survival in response to DAT-independent oxidative stress such as that induced by rotenone or H 2 O 2 . This last finding is in apparent contradiction with previously published data showing that embryonic midbrain DA neurons from Parkin-KO mice show an increased vulnerability to rotenone (51). The different type of cultures used (embryonic versus postnatal), density of DA neurons (low versus high), age of cultures (7 versus 10 DIV), and the different duration of rotenone application are all factors that should be considered to explain such different results. Interestingly, in this study, we found no differential effect of MPPϩ on the survival of Parkin-KO mouse DA neurons when Parkin-KO mouse SNc DA neurons were cultured together with WT glia; this finding could be taken to suggest that the differential expression of DAT alone is not sufficient to explain the differential survival of SNc DA neurons to MPPϩ in Parkin-KO mouse cultures. However, a more comprehensive explanation is that in Parkin-KO mouse SNc cultures, the impaired functioning of astrocytes leads to the preferential death of highly arborized DA neurons, a population of neurons that are otherwise more vulnerable to cellular stressors in general (9). Because of this differential survival, the remaining SNc DA neurons in Parkin-KO mouse cultures would have lower axonal arborization and reduced DAT expression with consequent reduced vulnerability to MPPϩ. DAT levels were not quantified in cultures prepared from Pink1-KO or DJ-1-KO mice. Our data provide new insights into the complex relationship between mitochondrial function, axonal growth, and genetic risk factors linked to PD. Previous work has shown that overexpression of Parkin, Pink1, or DJ-1 in DA neurons is neuroprotective against mitochondrial toxins (35,36,52). Such results argue for a direct role of these gene products in regulating the vulnerability of DA neurons. The present results also uncover a role of Parkin in regulating mitochondrial function and OXPHOS in DA neurons. However, our findings also argue for the existence of more indirect roles through an impact on glial cells. Further work will be required to examine whether such changes might be key to explain the heightened vulnerability of SNc DA neurons over multiple decades, as occurs in PD.

Animals
All procedures involving animals were conducted in strict accordance with the Guide to Care and Use of Experimental Animals (2nd Ed.) of the Canadian Council on Animal Care. The experimental protocols were approved by the animal ethics committee (CDEA) of the Université de Montréal. Housing was at a constant temperature (21°C) and humidity (60%), under a fixed 12-h light/dark cycle, with free access to food and water.

Primary neuronal cultures and drug treatments
Cultures were prepared according to a previously described protocol (53), with minor variations (9). Dissociated neurons microdissected from the SNc or VTA of postnatal day 0 -2 (P0 -P2) Parkin-KO mice (48), Pink1-KO mice (54), or DJ-1-KO mice (40) were seeded on a monolayer of corresponding cortical astrocytes grown on collagen/poly-L-lysinecoated glass coverslips. The total seeded neuron density was 100,000 cells/ml for all experiments, except for the metabolic flux experiments for which the density was 250,000 cells/ml. All cultures were incubated at 37°C in 5% CO 2 and maintained in 2/3 of Neurobasal TM -A medium enriched with 1% penicillin/ streptomycin, 1% Glutamax TM -1, 2% B-27 supplement, and 5% fetal bovine serum (Invitrogen) plus 1/3 of minimum essential medium enriched with 1% penicillin/streptomycin, 1% Glutamax TM -1, 20 mM glucose, 1 mM sodium pyruvate, and 100 l of MITOϩ serum extender. In some experiments, MPPϩ (Sigma) (5 or 10 M) or H 2 O 2 (Sigma) (100 -150 M) were added at 10 DIV for 24 h. In other experiments, the complex I blocker rotenone (Sigma) (50 nM) was added at 8 DIV for 72 h. For the rescue experiments, AAV encoding GFP (AAV-CBA-emGFP) or WT Parkin fused with GFP (AAV-CBA-emGFP-FLAG-Parkin WT) was applied to the cells the day of the culture, leading to infection of more than 95% of the cells.

Metabolic flux experiments
The rate of oxygen consumption derived from mitochondrial OXPHOS and the ECAR derived from glycolysis were assessed using an extracellular flux analyzer (Seahorse Biosciences), as described previously (9). Cells were plated on XF24 tissue culture plates and maintained in culture for 5 or 10 days. Before experiments, cells were incubated for 1 h at 37°C in a CO 2 -free incubator in bicarbonate-free Dulbecco's modified Eagle's medium (Sigma) supplemented with 2 mM GlutaMax TM -1 (Invitrogen), 1 mM sodium pyruvate (Sigma), 25 mM D-glucose (Sigma), 63.3 mM NaCl (Sigma), and phenol red (Sigma). The pH was adjusted to 7.4 with NaOH. These concentrations of substrate represent saturating concentrations to ensure that they did not limit the rate of respiration and glycolysis. Oxygen consumption was sequentially measured under basal conditions, in the presence of the mitochondrial uncoupler CCCP (0.5 M), to assess the maximal oxidative capacity and in the presence of the mitochondrial inhibitors rotenone (1 M) and antimycin A (1 M) to assess non-mitochondrial oxygen consumption. After the assays, the cells were immediately fixed for immunofluorescence. For 10 DIV cultures, the OCR was normalized to cell number, as assessed by DAPI staining. Because neurons were cultured together with astrocytes, parallel measurements were also performed from pure astrocyte cultures. After subtraction of the OCR measured from astrocytes, the resulting OCR values measured from mixed cultures were normalized in a second step on the total number of neurons identified with MAP-2 immunostaining. For 5 DIV cultures, OCR and all ECAR measurements, the contribution of glial cells was not subtracted because values in co-cultures were close to those of astrocytes. Values were normalized on total cell number and then on the total number of neurons.

Mitochondrial network quantification
Mitochondria were labeled by infecting neurons with a lentivirus encoding DsRed2-mito (mitochondrially-targeted red fluorescent reporter protein) at 30 multiplicities of infection at the time of plating and fixed at 10 DIV. Images were obtained by capturing confocal 0.5-m z-stacks (10 -15 images) using a ϫ60 oil-immersion objective (NA 1.42) on a Olympus Fluoview FV1000 confocal using Fluoview version 3.1b software. Image analysis was performed using ImageJ software. Only TH-positive infected DA neurons were selected for analysis. Regions of interest were determined to separately measure DsRed2-mito signal in the cell body, isolated dendrite segments, and isolated axon segments. The percentage of TH-positive area covered by DsRed signal was measured in each cellular compartment.

Global neuronal morphology and survival assessment
At 10 DIV, TH-positive axonal arborization surface was estimated by capturing five confocal image stacks per coverslip randomly throughout the coverslip with a ϫ20 water immersion objective (NA 0.5). Using a homemade macro in ImageJ, MAP2 and TH signals were thresholded to remove background and then binarized. MAP2 signal area was then removed from TH signal area, and the remaining axonal signal surface was measured. For survival assessment, the number of TH neurons with clear round nuclei on the coverslip were estimated by scanning the coverslip vertically and horizontally (cross-counting) with a ϫ20 dry objective (NA 0.40) on a Nikon Eclipse TE200 inverted microscope.

Parkin-KO mouse glial cell proliferation measurements
Glial cells were grown on coverslips for 1, 3, 7, and 10 DIV. On a separate series of coverslips, FUDR was added once the glial cell reached confluence as is typically performed for neuronal cultures. The number of glial cells was measured by counting cells in five random pictures per coverslip after DAPI staining.

Measurement of total cellular ATP
Cellular ATP content was determined using the "ATPlite" kit (PerkinElmer Life Sciences) according to the manufacturer's instructions. The cells were collected by trypsinization and centrifugation at 500 ϫ g and then resuspended in PBS, pH 7.4. Measurements were performed on a Wallac Victor plate reader (PerkinElmer Life Sciences). One coverslip from each sample was labeled with DAPI to quantify total cell number. ATP values were then normalized on total cell number. Oligomycininsensitive ATP production (from glycolysis) was subtracted. Values obtained from measurements from glial cells alone were subtracted. The remaining values, representing the ATP derived from OXPHOS in neurons only, were then normalized on the total number of neurons identified with MAP-2 immunostaining.

Radioligand assay
10 DIV cultures were incubated with trypsin at 37°C for 10 min and collected in 20% FBS in PBS after subsequent centrifugation at 900 and 1500 ϫ g for 5 min. One ml of PBS was added between the two centrifugations. The cell pellets were resuspended in 200 l of a cold binding buffer solution (Na 2 HPO 4 10.1 mM, KH 2 PO 4 1.8 mM, KCl 2 mM, NaCl 137 mM) and then stored at Ϫ80°C until use. Tissue homogenates were washed by centrifugation at 13,000 ϫ g (4°C) for 15 min. Supernatants were discarded, and pellets were resuspended in the same buffer. DAT labeling was performed by adding 20 pM [ 125 I]RTI-55 ligand (specific activity, 2200 Ci/mmol, Perkin-Elmer Life Sciences, Woodbride, Ontario, Canada) to the tissue homogenates in a final volume of 200 l. Each sample was evaluated in triplicate. Nonspecific binding was estimated in the presence of 1 M of the selective DA reuptake inhibitor GBR12909 (Sigma). Incubation at room temperature for 1 h was terminated by rapid filtration through Whatman GF/C glass filters under vacuum, followed by three rinses (4 ml each) with the same ice-cold buffer. Radioactivity was counted by ␥ scintigraphy.

Statistics
Parametric statistical tests were used because samples contained data with normal distributions. Data were always obtained from a minimum of three separate sets of experiments and presented as mean Ϯ S.E. The level of statistical significance was established at p Ͻ 0.05 in one-or two-way analyses of variance or two-tailed t tests. A ROUT outlier analysis was performed when needed (Q ϭ 1%). Statistical analyses were performed with the Prism 7 software (GraphPad Software, * ϭ p Ͻ 0.05; ** ϭ p Ͻ 0.01; *** ϭ p Ͻ 0.001; and **** ϭ p Ͻ 0.0001). The Tukey post hoc test was used when all the means were com-pared with each other, and the Sidak post hoc test was used when only subsets of means were compared.