[1] 15w Mitochondrial SIRT3 confers neuroprotection in Huntington's disease by regulation of oxidative challenges and mitochondrial dynamics

[1] 154w SIRT3 is a major regulator of mitochondrial acetylome. Here we show that SIRT3 is neuroprotective in Huntington's disease (HD), a motor neurodegenerative disorder caused by an abnormal expansion of polyglutamines in the huntingtin protein (HTT). Protein and enzymatic analysis revealed that increased SIRT3 is a signature in several HD models, including human HD brain, which is regulated by oxidative species. While loss of SIRT3 further aggravated the oxidative phenotype, antioxidant treatment regularized SIRT3 levels. SIRT3 overexpression promoted the antioxidant effect in cells expressing mutant HTT, leading to enhanced mitochondrial function and balanced dynamics. Decreased Fis1 and Drp1 accumulation in mitochondria induced by SIRT3 expression favored mitochondrial elongation, while the SIRT3 activator ε-viniferin improved anterograde mitochondrial neurite transport, sustaining cell survival. Notably, SIRT3 fly-ortholog dSirt2 overexpression in HD flies ameliorated neurodegeneration and extended lifespan. These findings provide a link between oxidative stress and mitochondrial dysfunction hypotheses in HD and offer an opportunity for therapeutic development.

[1] 243w Sirtuin 3 (SIRT3) belongs to a larger family of NAD + -dependent lysine deacetylases (KDACs) with important roles in metabolism in a variety of organisms. SIRT3 is the major KDAC in mitochondria, controlling more than 1500 acetylation sites in mitochondrial enzymes involved in fatty acid oxidation, ATP production, pyruvate metabolism and antioxidant defenses [1,2]. SIRT3 is mainly located in the mitochondrial matrix and inner membrane (IMM), but has also been found in the nucleus [3,4] although its functional relevance in this compartment is not clear. In fact, observations demonstrate that unprocessed form of SIRT3 is enzymatically inactive and becomes fully activated upon import to mitochondria followed by cleavage by mitochondrial matrix processing peptidase (MPP) [5]. The mechanisms regulating SIRT3 are multi-faceted. During aging SIRT3 is downregulated, whereas its upregulation reverses aging-associated functional deficits by improving antioxidant activity of the mitochondrial enzyme superoxide dismutase 2 (SOD2) [6]. On the other hand, SIRT3 is upregulated in response to calorie restriction and fasting [7] or to pharmacological inhibition of the mitochondrial electron transport chain [8], highlighting its critical role in stress, adaptive responses and control of lifespan. In accordance, the Haigis lab has shown that under normal physiological conditions a pool of SIRT3 is sequestered at the IMM, whereas under mitochondrial stress such as matrix pH reduction, SIRT3 redistributes to other binding substrates potentiating their rapid deacetylation [9]. This elegant study provided clear evidence that the binding profile of SIRT3 dynamically changes according to mitochondrial environment.

[2] 48w Despite being the most abundant sirtuin in the brain [10], the neuronal role of SIRT3 has only recently been deeply explored [11]. Initial studies indicated that SIRT3 is neuroprotective by enhancing mitochondrial antioxidant capacity and biogenesis through the Peroxisome Proliferator-Activated Receptor Gamma (PGC-1α)/Estrogen-related receptor alpha (ERRα)-SIRT3 pathway [12].

[3] 336w In neurodegeneration models, as in the case of Alzheimer's disease App NL-G-F knock-in mice, SIRT3 was described to be critical for GABAergic synaptic homeostasis, improving behavior and neuronal adaptations to intermediate fasting [13]. Improvements in neuronal activity have been linked to the ability of SIRT3 to deacetylate mitochondrial p53, leading to increased mitochondrial DNA transcription [14]. The absence of SIRT3 has also been found to enhance striatal neuron toxicity in mice where mitochondrial complex II was irreversibly inhibited using 3-nitropropionic acid (3-NP) [11]. Indeed, systemic administration of 3-NP produces preferential degeneration in the caudate-putamen that resembles many behavioral and anatomical features of Huntington's disease (HD) [15]. HD is a fatal genetic neurodegenerative disorder that gradually impairs memory, cognitive skills and normal movements of the affected individuals and for which no disease modifying treatments exist. Our group has previously shown that pharmacological activation of SIRT1 by resveratrol ameliorates mitochondrial dysfunction and delays HD-related motor symptoms by modifying the expression of mitochondrial-encoded complex subunits [16]. Another study found that the semisynthetic resveratrol derivative trans-ε-viniferin mediates SIRT3-dependent AMP-activated protein kinase (AMPK) activation, replenishing cellular NAD + levels and cell survival in HD striatal cells [17]. Despite encouraging data that pharmacological activation of sirtuins prevents energy depletion that contributes to neurodegeneration, untangling the off-target effects of these polyphenols is a current challenge. Remarkably, no studies exist which explore the effects of directly targeting SIRT3 in HD. In the present study, we provide evidence that human caudate obtained from HD patients as well as neuronal and peripheral models expressing the mutant form of the HD gene display a stress-related increase in SIRT3 activity, which can be reverted upon antioxidant treatment. Under these conditions, we observed that SIRT3 is required to regulate mitochondrial radical production and mitochondrial membrane potential, eventually leading to decreased neurodegeneration in vitro and in a fruit fly model of HD. Additionally, we provide evidence that SIRT3 fine-tunes the mitochondrial network structure and distribution in HD models rather than increasing mitochondrial mass, thereby supporting overall bioenergetics in neuronal cells.

[1] 437w Mitochondrial SIRT3 activity is regulated by distinct processes; in aging an overall decrease of its activity is well accepted, while during neurodegeneration data are not conclusive. Here we used several HD models, schematically represented in Fig. 1, and human post-mortem brain tissue from HD patients to characterize SIRT3 expression levels and activity. Knock-in striatal STHdh Q111/Q111 cells expressing 111 CAG repeats in the chimeric human-mouse exon 1 of HD gene (hereafter Q111 cells) and cortical tissue isolated from YAC128 transgenic mice bearing 128 CAGs in the human HD gene presented a significant increase in SIRT3 mRNA levels (p < 0.01 in striatal cells, p < 0.05 in mice cortex (Fig. 1b,f). SIRT3 human and mouse transcripts codify proteins with a predicted molecular weight of 44 kDa and 36-37 kDa, respectively. They are cleaved by the MPP to become a 28 kDa enzymatically active mature form in the mitochondrial matrix [5,29]. Therefore, we evaluated both the full-length SIRT3 (FL-SIRT3) and processed enzymatically active mitochondrial SIRT3 (mSIRT3). The levels of processed mSIRT3 were also increased in Q111 cells (p < 0.05) (Fig. 1c), while the inactive ~37 kDa FL-SIRT3 was reduced in cytosolic-enriched extracts (p < 0.05) (Supplementary Fig. 1a), suggesting increased translocation of SIRT3 to mitochondria in Q111 cells. Of note, we have not detected the enzymatic active form of endogenous mSIRT3 in cytosol of striatal cells (Supplementary Fig. 1a). SIRT3 activity was also evaluated in mitochondrial-enriched extracts by the basic principle of a fluorophore and quencher coupled to an amino and carboxyl terminal of a SIRT3 acetylated substrate peptide in the presence of NAD + . A pronounced increase in fluorescence over time was observed in Q111 in relation to Q7 wild-type cells (p < 0.001) (Fig. 1d), correlating with increased substrate peptide cleavage. Recombinant SIRT3 and SIRT3 knockdown (KD) were used as positive and negative controls, respectively, to confirm the specificity of the reaction (Supplementary Fig. 1c). In 12 month-old YAC128 brain cortex (late symptomatic) no significant differences were detected in SIRT3 protein levels (Fig. 1g), suggesting that increased SIRT3 may be an early adaptation. In the impossibility of studying SIRT3 activity in frozen tissue, we evaluated acetylated levels of SOD2 at lysine 68 (k68), which has been described to be a primary target of SIRT3 [30]. Acetylated SOD2 was decreased in 12 month-old YAC128 compared to WT mouse cortex (p < 0.05; Fig. 1h). This result can be interpreted as an indication of increased SIRT3 activity, and although the contribution of other classes of mitochondrial deacetylases for k68 acetylation are unlikely [30], the regulation of SOD2 activity by other acetylation sites cannot be ruled out.

[2] 137w As human models we used lymphoblasts derived from HD affected patients containing heterozygous expansion mutation (Fig. 1i-l) versus unaffected voluntary control siblings and HD patient-derived iPSC expressing either normal or 72 CAG-expanded form of HD gene (HD-iPSC) (Fig. 1m-o) versus AMS4-iPSC cells (Ctr-iPSC). HD lymphoblasts showed increased mRNA levels of SIRT3 (p < 0.01), as well as increased mitochondrial SIRT3 protein levels (p < 0.001) and activity (p < 0.05) (Fig. 1j-l). FL-SIRT3 was detected in the nucleus, where it was described to interact with histones and transcriptional factors [3,4], but no changes in SIRT3 nuclei levels were detected between HD and Ctr lymphoblasts (Supplementary Fig. 1b). HD-iPSC followed the same tendency, showing a significant increase in mitochondrial SIRT3 protein levels (p < 0.05) and a trend for increased mRNA levels (p = 0.06) (Fig. 1n, o).

[3] 221w To understand if the overall increase in SIRT3 activity observed in HD models also occurs in HD patients brains, samples from the two main regions affected in HD brain, caudate and parietal cortex, were tested from HD patients with grade II-III and non-affected controls. As observed previously in late-symptomatic HD mice, no significant changes were observed in SIRT3 levels in human postmortem HD tissues at late stage of the disease (Fig. 2a, e). Interestingly, when evaluating ac (K68)SOD2 as a substrate for SIRT3 (acSOD2/SIRT3 ratio), we observed a significant downregulation in HD caudate (p < 0.05) but not in parietal cortex (Fig. 2b, e). Moreover, we detected a modest inverse correlation between SIRT3 and acSOD2 levels in HD caudate that was not observed in control samples (Fig. 2c). Remarkably, this inverse correlation was not detected in HD parietal cortex (Fig. 2d). Due to the delicate nature of the material and limited sample size it is rather precipitate to draw a definite conclusion about SIRT3 levels in human HD caudate; however, these data indicate that decreased acSOD2 levels are associated with increased SIRT3 activity and that this increase is region specific. Different degrees of SIRT3 expression and activity were also found across the different HD models tested, which likely correlate with disease degree, the severity of the pathogenic mechanisms and oxidative stress.

[4] 167w SIRT3 has been extensively described as a shield against oxidative stress and aging. Therefore, we hypothesized that increased activity of SIRT3 in HD cells may occur as a defense mechanism rather than a toxic response. To test how increased levels of SIRT3 affect mitochondrial oxidation in HD cells, we overexpressed SIRT3 in striatal cells and evaluated mitochondrial-derived ROS. Overexpression was confirmed by immunofluorescence (Fig. 3a and b) and immunoblotting (Supplementary Fig. 1d). Interestingly, SIRT3-GFP colocalizes more in mitochondria (labeled with mitochondrial targeting sequence with an RFP tag, MitoDsRed) from Q111 than in Q7 cells as observed by the matching histogram (★) (Fig. 3b). This accumulation of SIRT3 in HD mitochondria (p < 0.05), occurs together with a decrease in nuclear location (p < 0.0001) (Fig. 3c), supporting the data obtained for endogenous SIRT3 (Fig. 1c). Increased activity of SIRT3 was further confirmed in cellular extracts of SIRT3 overexpressing cells through decreased lysine acetylation, with no effects of GFP expression on acetyl-lysine levels (Fig. 3d, Supplementary Fig. 1e).

[5] 8w Single cell levels of anion superoxide (O 2

[6] 18w ) were measured using the fluorescent probe MitoSOX. Q111 cells (transfected or GFP control) displayed exacerbated O 2

[7] 93w •-levels in comparison with Q7 cells (p < 0.001) (Fig. 3e and f, Supplementary Fig. 2a and b), as previously described by us [31]. As a proof-of-concept mitochondrial complex III inhibitor antimycin A (AA) and the mitochondrial or cytosolic ROS scavengers mitoTEMPO (MitoT) and TEMPOL, respectively, were used. MitoT significantly reduced mitochondrial O 2 •-levels in both non-treated and AA-treated Q111 cells (p < 0.0001) (Supplementary Fig. 2a and b), while TEMPOL was ineffective/non-significant, attesting the specific effect of MitoT on mitochondrial-derived ROS. In SIRT3 overexpressing Q111 cells, the levels of O 2

[8] 78w •-significantly dropped, by about 76%, reaching the levels of wild-type cells (p<0.0001) (Fig. 3e and f). Remarkably, MitoT showed no effect under SIRT3 overexpressing conditions, sustaining that SIRT3 acts as a stress sensitive enzyme to promote mitochondrial antioxidant activity in HD mutant cells. Instead, the knock down (KD) of SIRT3 in HD cells subjected to mitochondrial complex III inhibition increased the susceptibility of these cells to show increased mitochondrial ROS levels, without affecting cell viability (Supplementary Fig. 2c-e).

[9] 75w To ultimately show that SIRT3 constitutes a stress response protein in HD, we further treated striatal cells and human lymphoblasts with the antioxidant MitoT. While wild-type/control cells displayed stable levels of mSIRT3 after MitoT incubation, supporting a balanced oxidative environment, in MitoT-treated HD cells the levels of mitochondrial SIRT3 returned to the levels of control cells (Fig. 3g). These data indicate that mSIRT3 regulates and is regulated by the oxidant status of cells expressing mHTT.

[10] 191w Previously we have shown depolarization of mitochondria following excessive ROS levels in Q111 cells [33]. To understand if enhanced ROS scavenging achieved by SIRT3 overexpression in HD cells correlated with a healthy mitochondrial function, we evaluated mitochondrial membrane potential (Δψ m ) using the membrane permeant cationic fluorescent probe TMRM + before and after mitochondria depolarization with FCCP in the presence of oligomycin (Fig. 4a). As described previously by us and others, Q111 striatal cells showed a decreased mitochondrial accumulation of TMRM + compared to wild-type cells, as observed by decreased probe release after FCCP (p < 0.01) (Fig. 4a-c), reflecting decreased Δψ m . This decrease in Δψ m may account for by impaired mitochondrial maximal respiration and ATP production observed in these cells (p < 0.05) (Fig. 4e and f). SIRT3 overexpression in both Q7 and Q111 striatal cells led to more TMRM + accumulation, with a complete recovery of Δψ m in mutant cells (Fig. 4a-c). Interestingly, SIRT3 KD decreased mitochondrial maximal respiration (p < 0.05) but did not affect ATP production in Q111 cells (Fig. 4e and f), indicating that HD mitochondria still retained some metabolic reserve.

[11] 148w To further address the mechanisms contributing for the improvement of mitochondrial bioenergetics prompt by SIRT3, we evaluated mitochondrial mass and the levels of the two major transcription factors influencing mitochondrial biogenesis and transcription, mitochondrial transcription factor A (TFAM) and its upstream regulator PGC-1α. Mutant cells showed decreased protein levels of PGC-1α (p < 0.001) and TFAM (p < 0.05) (Supplementary Fig. 3a, b, d), accompanied by a substantial reduction in mitochondrial mass (p < 0.0001) (Fig. 4d). SIRT3 overexpression reestablished the levels of TFAM in mitochondria in HD mutant cells (p < 0.01) to the levels of wild-type cells (Supplementary Fig. 3b and d). Despite that, no significant changes in PGC-1α levels, mtDNA copy number or mitochondrial content were observed in SIRT3-overexpressing Q111 cells (Fig. 4d, Supplementary Fig. 4a and c), indicating that the increase in TFAM levels is not sufficient to restore mitochondrial biogenesis in these cells.

[12] 457w ROS-mediated loss in Δψ m is described to mediate mitochondrial fission [34]. Considering the preceding data, we speculated that SIRT3 may regulate mitochondrial dynamics. HD models are characterized by a large number of malfunctional small round-shaped mitochondria [35], which is partially related to the binding and hyperactivation of the mitochondrial fission GTPase dynamin-related protein 1 (Drp1) by mHTT [36]. Using mitochondrial and cytosolic-enriched extracts we confirmed that Q111 cells display Drp1 accumulation in mitochondria, compared to Q7 cells that have a more uniform distribution of Drp1 (Fig. 5a). Co-localization analysis of Drp1 with MitoDsRed corroborated the previous result (Supplementary Fig. 4a and b), ultimately suggesting that mutant cells display increased fission, since Drp1 is a cytosolic protein that is only translocated to the outer mitochondrial membrane (OMM) to mediate mitochondrial constriction sites for fission, together with its adaptor Fis1 and endoplasmic reticulum [37]. Remarkably, analysis of the mitochondrial network revealed that both wild-type and mutant cells exhibit the three categories of mitochondrial network: fragmented, tubular or mixed (Fig. 5b). Both roundness, that indicates the degree of circularity, and aspect ratio, that positively correlates with length, were analyzed for each independent cell to allow categorization of the mitochondrial network. Cells with roundness lower than 0.5 (or aspect ratio > 2.43; percentile 25%) were categorized as tubular, while cells with roundness higher than 0.55 (or aspect ratio < 2; percentile 75%) were categorized as fragmented, as demonstrated on Fig. 5c. Cells in between these cutoffs were considered as displaying mixed mitochondrial network. The uncoupler CCCP was used as a positive control for the fragmented network. Both values for aspect ratio and roundness fitted in the values obtained previously for the 'fragmented' category, validating the model (Supplementary Fig. 4c and d). Cells displaying swollen mitochondria were discarded from this analysis, since they represented a minority of the whole population for either genotype (data not shown). Following the described categorization, we observed that Q111 cells show a substantial increase of fragmented mitochondria in detriment of tubular mitochondria (Fig. 5d and e; Supplementary Fig. 4e-g). Remarkably, overexpression of SIRT3 significantly reduced the number of mutant cells with fragmented mitochondria (p < 0.01 for roundness; p < 0.05 for aspect ratio) to similar levels of GFP-expressing wild-type cells (Fig. 5d and e, Supplementary Fig. 4g). This effect may be partially related to the decrease of Drp1 levels in mitochondria of mutant cells in the presence of SIRT3 (p < 0.05) (Fig. 5f). Additionally, SIRT3 overexpression also reduced total levels of the Drp1 adaptor on OMM, Fis1 (p < 0.05) (Fig. 4g). Contrarily, SIRT3 did not affect mitochondrial fusion markers mitofusin 2 and optic atrophy 1 (Opa1) levels (Supplementary Fig. 4h and i), suggesting that SIRT3 primarily influences mitochondrial network through regulation of fission.

[13] 15w Overall, these data suggest that SIRT3 affects Drp1 relocation, circumventing mitochondrial fragmentation induced by mHTT.

[14] 314w Previous work by the Bossy-Wetzel group have shown that restoration of mitochondrial fusion, either by impairing Drp1 enzymatic activity or inducing Mfn2 function, directly correlate with rescue of HD neurons from neuritic trafficking defects and cell death [36]. ε-Viniferin (Vini) is a resveratrol dimer that has been shown to improve mitochondrial function and enhance the expression of mitochondrial homeostasis-related proteins through SIRT3 activation [17,38,39]. Here, we used primary neurons isolated from YAC128 mice embryos to evaluate the effect of ε-viniferin, and consequent SIRT3 activation, on mitochondrial morphology and movement in this HD model. The increase in mSIRT3 protein and mRNA levels were confirmed (p < 0.05), whereas the levels of SIRT1 and SIRT5 remained unchanged with Vini incubation (1 μM, 24 h) (Fig. 5h, Supplementary Fig. 5a and b), suggesting an effect of ε-viniferin towards SIRT3, among other members of the sirtuin family known to directly influence mitochondrial function. As in striatal cells, YAC128 neurons also showed more fragmented mitochondria (2.3 aspect ratio in relation to 2.8 aspect ratio in WT neurons), which was completely prevented after ε-viniferin treatment (p < 0.05) (Fig. 5i). Striatal MitoDsRed-transfected neurons were imaged in a spinning disk confocal with 5 s intervals; YAC128 striatal neurons had approximately 90% stationary mitochondria and only ~3.5% were detected to move towards synaptic terminals (Fig. 5j and k). Additionally, the low percentage of moving mitochondria moved slower in YAC128 striatal neurites compared to the wild-type controls (mean velocity of 35.4 μm/s in YAC128 neurons, compared to 60.5 μm/s in WT conditions) (p < 0.05) (Fig. 5m, video 1, 2). After ε-viniferin treatment, the percentage of stationary mitochondria in YAC128 neurons reduced significantly to 60% (p < 0.0001), mainly favoring anterograde transport that displayed about 7.6-fold increase in comparison with YAC128 vehicle-treated striatal neurons (p < 0.001) (Fig. 5j and k). Additionally, ε-viniferin reestablished mitochondrial velocity (p = 0.0533) (Fig. 5m, video 1-3).

[15] 32w Together, these data indicate that ε-viniferin-induced SIRT3 expression completely reestablished mitochondrial neuritic elongation and trafficking in YAC128 neurons, an effect that has been previously described to be dependent on SIRT3 activation [38].

[16] 12w Supplementary video related to this article can be found at htt ps://doi.org/10.1016/j.freeradbiomed.2020.11.031

[17] 71w Finally, we explored if SIRT3 was able to mitigate cell death and neuronal degeneration that characterize most HD models. Q111 cells showed a 3.4-fold increase in apoptotic cell death, compared to Q7 cells, measured by nuclear condensation using Hoechst/PI co-staining (Fig. 6a and b). On the contrary, Q111 cells overexpressing human SIRT3 displayed a significant reduction in DNA condensation (p < 0.0001), entailing a complete regularization of apoptotic mechanisms (Fig. 6b).

[18] 296w To ultimately verify if SIRT3 also had neuroprotective functions in vivo we took advantage of a widely used fruit fly model of HTT toxicity [27] and the GAL4/UAS system to direct expression of mHtt exon 1 fragment (Htt93Q) either pan-neuronally with an elav-GAL4 driver or to muscles (which have a high mitochondrial load) with a mef2-GAL4 driver (Fig. 6c). Pan-neuronal expression of mHtt results in an obvious loss of photoreceptor neurons (rhabdomeres) leading to a disorganization of ommatidia [40] a robust readout of neurodegeneration. While mammalian sirtuin family is composed by seven different SIRTs, there are just five members of the Sir2 family in Drosophila melanogaster. dSirt2 presents the highest sequence homology with hSIRT3 (54% similarity), followed by hSIRT2 (49% similarity) (Supplementary Fig. 6a). Additionally, it was shown that uniquely amongst the Drosophila sirtuins dSirt2 is an important regulator of mitochondrial function that impacts on acetylation of OXPHOS subunits (hSIRT3 recognized targets) [41]. While dSirt2 is not localized to mitochondria, of the Drosophila sirtuins it appears to be most functionally similar to mammalian SIRT3. Taking this information into consideration, we overexpressed dSirt2 in the HD fly model and analyzed lifespan and the number of rhabdomeres. While no changes in lifespan were observed in elavHtt93Q dSirt2 flies, compared to elav Htt93Q (Supplementary Fig. 6b), HD flies with muscle dSirt2 overexpression displayed a significant lifespan increase (p < 0.0001), with an extended median survival of 1.5 days and a maximum lifespan of 18-days (Fig. 6d), highlighting the importance of SIRT3 for mitochondria health. Moreover, pan-neuronal expression of dSirt2 in elav Htt93Q flies led to a mean recovery of ~50% of rhabdomeres lost by day 7 (p < 0.0001) (Fig. 6e). All together these results strongly suggest that hSIRT3/dSirt2 can act as an important modifier of neurodegeneration in HD.

[1] 256w Mitochondrial deregulation is an early event that characterizes all HD models and HD patients [16,[42][43][44]. Our group and others have previously shown that sirtuin-activation compounds enhance mitochondrial biogenesis and function in vivo and in vitro, ameliorating HD phenotype [16,45]. These effects are mainly driven by SIRT1 nuclear targets PGC-1α, forkhead box O3A (Foxo3a) and cAMP response element binding protein (CREB) transcriptional pathways [16,45,46]. Interestingly, human HD cybrids, a cell model in which the contribution of mitochondria from patients is separate from the nucleus, also exhibit mitochondrial-associated metabolic deficits, suggesting that targeting mitochondria itself could be a better and more direct strategy to mitigate these deficits. Within the sirtuin family, SIRT3 has been shown to regulate the majority of the mitochondrial acetylome [1]. Here we provide evidence that SIRT3 mediates neuroprotective effects against mHTT-induced cell toxicity in vitro and in vivo through regulation of mitochondrial-driven oxidative stress and mitochondrial morphofunction. We found that HD mouse and human models and caudate from HD patients show increased SIRT3 levels and/or activity in response to high levels of oxidative damage [31,33,47]. Remarkably, SIRT3 levels regularize to normal levels after mitochondrial directed-antioxidant treatment, corroborating the previous assumption. In fact, the observed increase is not detrimental. We showed that SIRT3 promotes the translocation of Drp1 from mitochondria to cytosol, reducing mitochondrial fission that is classically increased in HD, as confirmed in our models. This event may account for the recovery of mitochondrial function and distribution in HD neural cells, rather than the increase in mitochondrial biogenesis, preventing cell death and increasing lifespan.

[2] 578w Several circumstances seem to regulate SIRT3 expression in the central nervous system. A previous study from the Duan laboratory using the same HD striatal cells reported decreased levels of the processed active mSIRT3 isoform [17] after serum deprivation, while SIRT3 baseline levels have not been quantified. This study diverges from describing the increase in SIRT3 levels as a stress response, as we and others [48] have observed. Concordantly with our data, pharmacological interference with the mitochondrial electron transport chain using the complex III inhibitor antimycin A, which induces increased mitochondrial ROS levels, upregulated SIRT3 transcript expression in hippocampal neurons [8]. SIRT3 mRNA levels were also increased in the presence of beta-amyloid peptide (Aβ), the aggregation-prone molecule that characterizes Alzheimer's disease pathology [49]. Moreover, mitochondrial SIRT3 was reported to be increased in neurons where NAD + was depleted following stimulation with NMDA [50]. Although most of these studies describe SIRT3 activation under acute stress, we have proven that SIRT3 is also upregulated in HD models undergoing chronic stress. In fact, this lasting chronic activation may explain why ROS levels remain increased in our HD cell model and why further SIRT3 activation is protective. This postulation is supported by experiments showing that the powerful mitochondrial-directed antioxidant MitoT significantly decreased ROS and SIRT3 levels, and its protective effect was disguised by SIRT3 overexpression in HD cells. Therefore, we consider that despite increased SIRT3 activity, HD cells need an additional stimulus to disrupt the oxidative steady state where they stand to recover mitochondrial homeostasis. Oxidative damage is one of the main players that drives mitochondrial dysfunction and cell death. We have shown that an oxidative stimulus like H 2 O 2 in striatal Q111 cells causes a significant impairment in Δψ m and induces caspase-3 activation [33,51]. Recently, a well-designed study aimed to understand the role of SIRT3 in adaptive responses of neurons to metabolic and excitatory challenges. Similarly to what we observed here, the authors found that neurons lacking SIRT3 exhibited increased vulnerability to excitotoxic, oxidative and metabolic stress induced by glutamate, H 2 O 2 and 3-nitropropionic acid (3-NP, an irreversible inhibitor of complex II), respectively [52]. SIRT3 knockout mice showed decreased survival after 7 days of 3-NP administration and rotarod performance was significantly impaired compared to wild-type mice. Here as well, we showed that SIRT3 is linked to the activation of lifespan mechanisms, as already described several years ago for nuclear SIRT1 and SIRT6 [53,54], since dSirt2/SIRT3 expression extended survival in flies with high metabolic demand expressing mHtt. Authors also showed that mitochondria of these SIRT3 knockout mice exhibited decreased levels of ATP and swelling associated with calcium dyshomeostasis [52]. The decrease in ATP production is likely linked to alterations in the electrochemical gradient across the mitochondrial membrane as SIRT3 is tightly associated with ATP5O under physiological conditions, but upon loss of membrane potential this attachment with ATP synthase is reduced [9]. Loss of Δψ m in our HD cells was also recovered after SIRT3 overexpression. Besides the above described effects, the sustained increase in SIRT3 was previously described to enhance the expression of subunits of mitochondrial complexes such as cyt c oxidase subunit 2 and 4, enhancing cellular respiration [55]. In addition, two mass spectrometry assays identified several mitochondrial complex subunits that were hyperacetylated on two or more lysine residues in SIRT3 knockout mice, including complex I/NADH dehydrogenase 1α, 1β subcomplexes and flavoprotein 1, complex II/SDH flavoprotein and iron-sulfur subunits, and complex IV/cyt c oxidase subunits 4-7 [2,56].

[3] 284w The relationship between mitochondrial energy defects and mitochondrial dynamics was first explored in detail by the Bossy-Wetzel lab. The authors showed that while the drop in ATP is a quick response to 3-NP-mediated complex II inhibition, mitochondrial fission only occurred after a late dramatic rise in ROS levels [57]. Moreover, depletion of Δψ m (h) YAC128 and WT primary striatal neurons were treated with DMSO (Veh) or Viniferin (Vini, 1 μM) for 24 h. SIRT3, SIRT1 and SIRT5 proteins were labeled and quantified (Supplementary Fig. 5a) by western blotting and normalized for actin using specific antibodies (n = 3-4). seems to be another trigger event for mitochondrial fission, recruiting Drp1 to constriction sites at the mitochondrial outer membrane [58]. Our data indicate that the increase in filamentous mitochondria observed after SIRT3 overexpression occurs together with the drop in ROS levels. In accordance, nitric oxide was shown to induce S-nitrosylation of Drp1, increasing its GTPase activity and mitochondrial fragmentation [59]. In primary cultured neurons expressing mHTT a significant increase in nitric oxide production was observed and correlated with increased S-nitrosylation of Drp1, whereas transfection with non-nitrosylatable Drp1 mutant abrogated excessive mitochondrial fragmentation and dendritic spine loss [60]. Similarly, ROS-mediated modification of reactive cysteine thiol to form cysteine sulfenic acid to increase the sulfenylation of Drp1 at Cys644 also induced mitochondrial fragmentation [61]. Concordantly, the antioxidant agent acetyl-L-carnitine was able to decrease levels of Drp1 [62]. Interestingly, SIRT3 may also directly modulate Drp1 activity since one acetylation site was recently found for Drp1 at K642. Hyperacetylation of Drp1 at this site was observed to induce its oligomerization and phosphorylation at the activating serine 616 site in cardiomyocytes, recruiting Drp1 to OMM and promoting cell death [63].

[4] 243w Evidences that inhibition of mitochondrial fission is sufficient to hinder neurodegeneration have been accumulating. An impeccably study performed by Mochly-Rosen lab showed that inhibition of Drp1 binding to its adaptor Fis1 using the heptapeptide P110, is enough to decrease ROS levels and inhibit the propagation of the inflammatory response and neuronal cell death in neurons expressing 73Q and in R6/2 HD neurons [64]. These studies further support that strategies aiming a more balanced mitochondrial dynamics like SIRT3 activation, likely contribute to cope with neurodegeneration and increase cell survival. Nevertheless, it remains a challenge to pharmacologically modulate this enzyme to further consider clinical studies. Several studies, including the present work, have resorted to the use of pharmacological modulators that are not selective for a certain deacetylase. Viniferin is a resveratrol dimer recently shown to activate SIRT1 [65], although we observed no changes in SIRT1 levels in our model. This pleotropic effect is not necessarily disadvantageous, provided that it only operates at a pathological level, however it is difficult to reconcile whether the observed effects are exclusively due to SIRT3 activation. In the same way, unselective inhibition of deacetylases can also elicit unexpected side effects. Therefore, future research on crystal structure of the catalytic core of sirtuins and on specific binding modes of molecules will be able to offer valuable information for the design of novel targeting strategies. This might be particularly beneficial when considering the antioxidant and neuroprotective role of SIRT3 in HD pathogenesis.

[1] 244w Primary striatal neuronal cultures were generated from the offspring of crosses between WT mice (used as controls), or between hemizygous YAC128 male mice and WT females. Embryos from timed pregnant females were collected on day E15.5-16.5 of gestation. Striatum were dissected at 25× magnification in ice-cold Hank's balanced salt solution (HBSS, 5.4 mM KCl, 0.4 mM KH 2 PO 4 , 137 mM NaCl, 4.2 mM NaHCO 3 , 0.3 mM NaH 2 PO 4 .H 2 O, 5 mM glucose, 5.4 mM sodium pyruvate, 5.4 HEPES) supplemented with 0.3% fatty acid free BSA [18] and then transferred to Hibernate E medium for approximately 4 h or until the genotyping was concluded. Dissected tissue was pooled for each genotype and digested with HBSS supplemented with 0.003% trypsin. Soybean was used to block the trypsin. Neurons were cultured in Neurobasal medium supplemented with 2% B-27, 1 mM glutamine, 20 μg/ml gentamicin and plated at a density of 130 × 10 3 cells/cm 2 in poly--D-lysine (0.1 mg/ml)-coated plates. For low density cultures cells were plated at 85 × 10 3 cells/cm 2 on coverslips. Cultures were maintained at 37 • C in a 5% CO 2 incubator and half of the medium was replaced after 3 days in vitro (DIV) with fresh supplemented Neurobasal medium containing 5-fluoro-2 ′ -deoxyuridine (5 μM C f ) to inhibit glial division. Half of medium was replaced again with fresh medium at DIV7 and cells were used at DIV12/13.

[2] 167w In primary neurons, transfection with pDsRed2-Mito vector was performed at 8 DIV using the calcium phosphate precipitation method. Briefly, plasmid was diluted in TE (1 mM Tris-HCl pH 7.3, 1 mM EDTA), followed by the addition of CaCl 2 (2.5 M CaCl 2 in 10 mM HEPES, pH 7.2). The DNA solution was carefully added to 2 × HEBS (12 mM dextrose, 50 mM HEPES, 10 mM KCl, 280 mM NaCl and 1.5 mM Na 2 HPO 4 .2H 2 O, pH 7.2) while bubbling air through the solution. The mixture was then incubated for 25 min at room temperature. The precipitates were added dropwise to the coverslips in Neurobasal medium and incubated for 80 min at 37 • C. The DNA-Ca 2+ -phosphate precipitates were dissolved in freshly-made dissolution medium (Neurobasal medium with 20 mM HEPES, pH 6.8) and incubated for 7 min at room temperature. The transfected neurons were then washed with Neurobasal medium and transferred back to their original dishes containing conditioned culture medium.

[3] 31w A working stock concentration of 10 mM viniferin was prepared in DMSO. Subsequent dilutions were done in culture media. 1 μM viniferin or DMSO (0.1%) was added 24 h before experiments.

[4] 107w Immortalized striatal neurons derived from knock-in mice expressing full-length normal (referred as Q7 or wild-type cells) or full-length mutant Htt (mHtt) with 111 glutamines (referred as Q111 or mutant cells) were obtained from Coriell Institute for Medical Research (USA) and maintained as previously described [19]. Cells were used until passage 16 by recommendation of cell line producer Dr M. MacDonald (Dept. of Neurology, Massachusetts General Hospital, Boston, USA). Transfection of striatal cells with selected plasmids or siRNAs (Table 1) were performed with Lipofectamine TM 3000 transfection reagent accordingly to the manufacturer's instructions. MitoTEMPO (MitoT; 2 μM) and TEMPOL (2 μM) were added and incubated for 24 h.

[5] 37w Human lymphoblastoid cell lines were obtained from Coriell Institute for Medical Research, derived from HD-affected patients containing heterozygous expansion mutation, four males (CAG repeats: 43/15, 45/ 15, 42/18, 49/17) and one female (CAG repeats: 47/18), or from

[6] 40w Reagents and resources. unaffected voluntary control siblings, three males and one female, defined in this work as control (CTR) lymphoblasts. Cells were cultured in RPMI-1640 medium as previously described [16]. MitoT (2 μM) was added and incubated for 24 h.

[7] 159w Protein extracts were denatured with denaturing buffer (50 mM Tris-HCl pH 6.8, 2% SDS, 5% glycerol, 600 mM DTT, 0.01% bromophenol blue) at 95 • C, for 5 min. Equivalent amounts of protein (15-30 μg) were separated on an 8-12% SDS-PAGE and electroblotted onto polyvinylidene fluoride membranes (PVDF). For dot blotting, 5 μg of the total extracts were directly applied onto a previously methanolactivated and equilibrated PVDF membrane. Membranes were blocked for 1 h in Tris-buffered saline (TBS) solution containing 0.1% Tween (TBS-T) and 5% BSA, followed by an overnight incubation with primary antibodies listed in Table I (at 4 • C) with gentle agitation. Membranes were then washed 3 times with TBS-T and incubated with secondary antibodies conjugated with alkaline phosphatase (1:20,000; 1 h) at room temperature, with gentle agitation. Immunoreactive bands were visualized by alkaline phosphatase activity after incubation with ECF substrate in a ChemiDoc Touch Imaging System (Bio-Rad). Bands were quantified using the Image Lab software.

[8] 172w RNA from different samples was extracted with PureZOL RNA isolation reagent according to the manufacturer's protocol. Briefly, mice cortex was lysed after potter homogenization, whereas striatal cells, iPSCs and lymphoblasts were scraped or homogenized with a pipette. RNA was precipitated with isopropyl alcohol and the final pellet was resuspended with diethylpyrocarbonate 0.01% (v/v) water. The concentration of RNA was measured using a Nanodrop 2000 (Thermo-Scientific, Walthem, MA, USA) and the RNA integrity was confirmed by A260/280 > 1.9. RNA was transcribed into cDNA in a template-primer mix using the NZY First-Strand cDNA Synthesis kit (Nzytech, Portugal). The reaction was performed using a CFX96 qPCR system (Bio-Rad, Hemel Hempstead, UK) with the following steps: 5 min at 25 • C, 30 min at 42 • C and 5 min at 85 • C. qPCR was performed using the SsoFast Eva Green Supermix or the iQTM SYBR® Green supermix (in the case of iPSC) using the following primers at 300 nM: human SIRT3 (forward: CGGCTCTACACGCAGAACATC; reverse: CAGCGGCTCCCCAAAGAA-CAC), human GAPDH (forward: 5 ′ -ATTCCACCCATGGCAAATTC-3'; reverse:

[9] 169w 5 ′ -GGGATTTCCATTGATGACAAGC-3 ′ ), human tubulin (forward: 5 ′ -CCAGGGCTGTGTTTGTAGACC-3'; reverse: 5 ′ -CAA-TAGTGTAGTGTCCACGGGC-3 ′ ), human S18 (forward: 5 ′ -GAGGAT-GAGGTGGAACGTGT-3'; reverse: 5' -TCTTCAGTCGCTCCACGTCT-3 ′ ), mouse SIRT3 (forward: 5 ′ -GCCTGCAAGGTTCCTACTCC-3'; reverse: 5 ′ -TCGAGGACTCAGAACGAACG-3 ′ ) and mouse GAPDH (forward: 5 ′ -CGGCCTTCCGTGTTCCATC-3'; reverse: 5 ′ GAGTTGCTGTTGAAGT CGCA-3 ′ ). Amplification of 10 ng (50 ng for SIRT3 mouse gene) template was performed with an initial cycle of 30 s at 95 • C, followed by 40 cycles of 5 s at 95 • C plus 5 s at 59 • C (for cell extracts) or 61 • C (for tissue extracts). At the end of each cycle, Eva/SYBR Green fluorescence was recorded to enable determination of Cq. After amplification, the melting temperature of the PCR products was determined by performing melting curves, with 0.5 • C increments every 5 s, from 65 • C to 95 • C, with fluorescence recorded after each temperature increment. For each set of primers amplification efficiency was assessed.

[10] 227w SIRT3 activity was measured using the commercially available SIRT3 activity assay kit following the manufacture's' instructions. The assay is based in a fluorophore and quencher coupled to amino terminal and carboxyl terminal of a SIRT3 substrate peptide, respectively. When SIRT3 deacetylates the substrate peptide, this will be cut by the peptidase added simultaneously, separating the quencher and the fluorophore, and thus fluorescence is emitted. Briefly, two 100 mm 2 petri dish plates were used per condition to extract mitochondria using unsupplemented ice-cold sucrose buffer. After homogenization with a Potter-Elvejhem 377 homogenizer at 300 rpm, cell extract was centrifuged at 1088 xg for 12 min (4 • C). The supernatant was further centrifuged at 12,000 xg for 20 min (4 • C) and the resulting mitochondrial-enriched pellet was resuspended in a provided SIRT3 assay buffer without supplements. The amount of protein was immediately quantified and 6 μg of mitochondria were distributed per well (in triplicates) in a black 96-well plate, containing fluoro-substrate peptide, peptidase and NAD + . The fluorescence was monitored for 40 min at 1 min intervals using a microtiter plate reader fluorometer with excitation at 350 nm and emission at 450 nm. The rate of reaction was expressed as slope per min. As positive and negative controls, we used a recombinant SIRT3 provided by the kit and mitochondrial extracts of SIRT3 knockdown Q7 cells, respectively.

[11] 27w Striatal cells cultured on 18-mm coverslips were incubated in experimental media (in mM: 132 NaCl, 4 KCl, 1 CaCl 2 , 1.2 NaH 2 PO 4 .

[12] 112w H 2 O, 1.4 MgCl 2 , 6 Glucose, 10 HEPES, pH 7.4) supplemented with 1 μM MitoSOX red, 300 nM Mitotracker deep red and 1 μg/mL Hoechst 33342 for 25 min at 33 • C. Coverslips were washed and mounted in experimental media in a pre-warmed insert and images were collected using an Axio Observer Z1 system, a fully motorized inverted widefield microscope (Zeiss, Jena, Germany) equipped with a large stage incubator for temperature and humidity control and EC plan-neofluar/1.3NA 40x lens. MitoSOX fluorescence was imaged at 510 nm excitation and 580 nm emission. 20 images were collected per coverslip. Fluorescence intensities were calculated using Fiji software as previously described [22].

[13] 120w Striatal cells were plated at a density of 25,000 (Q7)/35,000 (Q111) cells per well in 96-well assay plates and transfected with siRNA the following day. 48 h after transfection, cells were washed in experimental media and incubated for 40 min with 7 μM MitoPY1 at 37 • C and 5% CO 2 . MitoPY1 fluorescence was imaged at 503 nm excitation and emission at 528 nm [23] using a Microplate Spectrofluorometer Gemini EM (Molecular Devices, USA). Basal levels were measured for 10 min followed by exposure to 2 μM antimycin A (AA), an inhibitor of mitochondrial complex III. Fluorescence was measured for an additional 30 min. The presented results are the slope of fluorescence after stimuli, normalized for protein content.

[14] 154w Striatal cells cultured on 18-mm coverslips were incubated in experimental media with 2 μg/ml Hoechst 33342 and 10 nM tetramethylrhodamine, methyl ester (TMRM + ), a concentration sufficient low to avoid quenching of the fluorescent signal in the matrix [24], for 30 min at 33 • C. Coverslips were washed twice and mounted in a pre-warmed insert containing experimental media with 10 nM TMRM + , and images were collected using LCI PlanNeofluar/1.3NA 63x lens on a Carl Zeiss Axio Observed Z1 inverted confocal microscope using the CSU-X1M spinning disc technology with Zen Black 2012 software (Zeiss, Jena, Germany). TMRM + fluorescence was monitored in a controlled temperature by excitation at 543 nm and emission at 458 nm, as described previously [25]. Four single positions were captured for each coverslip using the multi-position tiles tool. Images were collected every 10 s with definite focus between each 6 cycles for a total of 54 cycles.

[15] 51w Cells were exposed to 2.5 μg/mL oligomycin (cycle 18) and 2.5 μM FCCP (cycle 36) to induce TMRM + release from mitochondria. Fluorescence intensity at each time point was analyzed in Fiji software using the time series analyzer plugin (v 3.0) developed by Balaji J. (2007) at Dept. of Neurobiology, UCLA.

[16] 95w Genomic DNA was using Purelink genomic DNA kit according to the manufacturer's protocol. The concentration of gDNA was measured and 10 ng was used for each reaction. The mtDNA of each sample was amplified using sequence-specific primers to tRNATry/CO1 (forward: CAGTCTAATGCTTACTCAGC; reverse: GGGCAGTTACGATAACATTG), a gene region that is rarely deleted and contain only a few rare single nucleotide polymorphisms; whereas nuclear DNA (nDNA) was amplified using sequence-specific primers to beta-2-microglobulin (β2M). The reaction was performed using SsoFast Eva Green Supermix and thermal cycling conditions described above. mtDNA levels were expressed in relation to nDNA levels.

[17] 157w MitoDsRed-transfected striatal neurons were washed and incubated in Na + medium (140 mM NaCl, 5 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM Glucose, 10 mM HEPES, pH 7.4), and mitochondrial movement studies were carried out at 37 • C. Neuron projections were imaged every 5 s for a total of 145 frames, using a 63x objective with NA = 1.4, on a Carl Zeiss Axio Observed Z1 inverted confocal microscope using CSU-X1M spinning disc. Mitochondrial movement analysis was done using the Kymograph Macro in Fiji (Rietdorf and Seitz, 2004). Briefly, histograms were matched to the first frame to correct fluorescence variations using Bleach Correction plugin developed by Miura and Rietdorf, and time lapse-dependent x-y drift was corrected applying TurboReg plugin. ROIs were designated using segmented line following mitochondria trajectory across projections. Kymographs generated in a x-y dimension (distance vs time) were used to obtain the slope to calculate mitochondrial velocity.

[18] 107w The nuclear morphology of the transfected striatal wild-type and mutant cells was analyzed by live fluorescence microscopy, using a double staining procedure with Hoechst 33342 and propidium iodide (PI). Following a washing step with Krebs medium, the cells mounted on 18-mm coverslips were incubated with 2 μg/ml Hoechst 33342 and 2 μg/ml PI for 10 min, at 33 • C. Cells were then washed 5 times in saline medium, in order to remove extracellular dyes, and further examined in an Axioskop 2 plus upright epi-fluorescence microscope (Zeiss, Jena, Germany) with PlanNeofluar/0.75NA 40x lens. Viable and apoptotic cells were counted in Fiji software using the Cell Counter plugin.

[19] 52w The number of visible rhabdomeres per ommatidium was scored for 20 to 100 ommatidia per fly, with 16-26 flies examined per genotype at day 1 and 7 post-eclosion. Heads from aged flies were fixed to glass slides using fingernail polish, and rhabdomeres were examined at 40× magnification using a Nikon Optiphot-2 microscope.

[20] 119w Results are expressed as the mean ± SEM (standard error of the mean) of the number of independent experiments or animals indicated in figure legends. Comparisons between multiple groups were performed by non-parametric one-way analysis of variance (ANOVA) using Kruskal-Wallis test. Correction for multiple comparisons was done by two-way ANOVA followed by Tukey post-hoc test. Comparison between two groups was performed by non-parametric Mann-Whitney test. For the analysis of life spam log-rank Mantel-Cox test was performed. The Ftest was performed to analyze the interaction term. ROUT (Q = 1%) method based on the false discovery rate was used to identify outliers. Significance was accepted at p < 0.05. All statistical analyses were performed using Prism software (GraphPad Version 8.0).

[1] 63w Caudate and parietal cortex samples from 6 unaffected controls and 8 HD patients (Vonsattel grade II-III) (Table 2) were obtained from German Brain Bank "Neurobiobank München" (http://www.neurobi obank.org). The postmortem intervals did not exceed 22 h. The ethical permit to use human samples was approved by Institutional Review Board from Faculty of Medicine, University of Coimbra, by December 21, 2015 (reference no. CE-136/2015).

[2] 84w Colonies of hemizygous YAC128 [line HD53] and wild-type (WT) mice, with FVB/N background, were housed in the Animal Facility of the CNC, Coimbra, Portugal, under conditions of controlled temperature (22-23 • C) and under a 12-h light/12-h dark cycle. Food and water were available ad libitum. All mouse experiments were carried out in accordance with the guidelines of the Institutional Animal Care and Use of Committee and the European Community directive (2010/63/EU) and protocols approved by the Faculty of Medicine, University of Coimbra (ORBEA_189_2018/11042018).

[3] 145w Heterozygous human iPSC-HD4-iPS (XY) with 19/72 CAG repeats were generated by Park and colleagues [20] and kindly provided by Prof. George Daley (Harvard Medical School, Boston, Massachusetts, USA), whereas wild-type AMS4-iPS (XY) were generated by de Almeida and collaborators [21]. Cells were maintained on a layer of mitotically inactivated murine embryonic fibroblasts (MEFs) for a variable number of passages or allowed to grow under feeder-free conditions on Geltrex® prior to experiments. MEFs were acquired from AMSBIO® inactivated or left to expand and further inactivated with mitomycin C after three passages. Manual dissection was routinely used to passage the cells. iPSCs were cultured in KSR media containing DMEM/F12 supplemented with 20% knockout serum replacement, 2 mM glutamine, 1 mM non-essential amino acids, 1% penicillin/streptomycin, 100 μM 2-mercaptoethanol and 10 ng/ml recombinant human FGF 2 . Cultures were fed daily and passaged at least once a week.

[4] 136w Total fractions: Extracts were prepared in ice-cold lysis buffer (20 mM Tris, 100 mM NaCl, 2 mM EDTA, 2 mM EGTA, 50 mM NaF, 1 mM Na 3 VO 4 , 1% Triton; pH 7.4) supplemented with 1 mM 1,4-dithiothreitol (DTT), 100 μM phenylmethylsulfonyl fluoride (PMSF), 100 nM okadaic acid, 1 μg/mL protease inhibitor cocktail (chymostatin, pepstatin A, leupeptin, and antipain), 1 μM TSA (HDAC inhibitor) and 10 mM nicotinamide (SIRT inhibitor). In particular, for dot blotting assays, 0.05% Triton was used since higher concentrations can inhibit protein binding to the membrane. The homogenates (total fractions) were then frozen/thawed three times in liquid nitrogen and centrifuged at 20,800 xg for 10 min, at 4 • C, in order to remove cell debris; the resulting supernatant was collected and stored at -80 • C for later use.

[5] 193w Different fractions were obtained to evaluate protein levels of full length (FL-SIRT3), cleaved/activated mitochondrial SIRT3 (mSIRT3), and Drp1 in different compartments of the cells Mitochondrial-and cytosolic-enriched fractions: Cells were scraped or resuspended in icecold sucrose buffer (250 mM sucrose, 20 mM HEPES/KOH (pH 7.5), 100 mM KCl, 1.5 mM MgCl 2 , 1 mM EGTA, and 1 mM EDTA), supplemented with 1 mM DTT, 100 μM PMSF, 100 nM okadaic acid, 1 μg/mL protease inhibitor cocktail, 1 μM TSA and 10 mM nicotinamide. Then, cells were homogenized with a Potter-Elvejhem 377 homogenizer with a Teflonpestle, at 300 rpm and centrifuged at 1088 xg for 12 min (4 • C) to pellet the nuclei and cell debris. The supernatant was further centrifuged at 12,000 xg for 20 min (4 • C) and the resulting pellet (mitochondrialenriched fraction) resuspended in supplemented sucrose buffer. Trichloroacetic acid (15%) was added to the supernatant, and precipitated proteins were centrifuged at 16,300 xg for 10 min (4 • C). The resulting pellet (cytosolic-enriched fraction) was re-suspended in supplemented sucrose buffer and brought to pH 7 with KOH. Both protein extracts were stored at -80 • C for later use.

[6] 124w Nuclear-enriched fractions: Nuclear fractions were obtained using the Nuclear/Cytosol Extraction Kit from BioVision, following the manufacturer instructions. Briefly, cells were resuspended in cytosol extraction buffer A containing DTT and proteases inhibitors and vigorously vortex for 15 s. Extracts were then incubated on ice for 10 min. To remove the cytoplasmic fraction, ice-cold cytosol extraction buffer B was added to the cell suspension, and after a 5 s vortex, centrifuged at 16,000 xg for 5 min (4 • C). The remaining pellet containing nuclei was resuspended in nuclear extraction buffer and vortex for 15 s for every 10 min, for 40 min. The final extracts were centrifuged at 16,000 xg for 10 min (4 • C) and stored at -80 • C for later use.

[7] 140w Striatal cells seeded in 12 well plates were transfected in Opti-MEM medium as previously described, and then medium was replaced by 300 μL regular high-glucose DMEM without phenol-red for 24 h. Culture medium was collected, and cells were scraped in 300 μL 10 mM HEPES (pH 7.4), 0.01% Triton X-100. Cell extracts were obtained by freezethaw cycles in liquid nitrogen. Both culture medium and extracts were centrifuged at 20,800 xg for 5 min (4 • C) to remove cell debris. In a 96well plate, the samples (in duplicates) were incubated with 1.58 mM pyruvate and 0.19 mM NADH. Extracellular and intracellular LDH was determined spectrophotometrically by following the rate of conversion for reduced nicotinamide adenine dinucleotide to oxidized NAD + , at 340 nm. The amount of released LDH in the medium was calculated using the ratio between extracellular/total LDH.

[8] 133w OCR in striatal cells was measured using a Seahorse XFe-24 flux analyzer (Seahorse Bioscience, Billerica, MA, USA), following the manufacturer's instructions. Striatal cells were seeded in XF24 cell culture microplates at a density of 25,000 (Q7)/35,000 (Q111) cells per well. Prior to the experiments cells were washed, and culture medium was replaced by assay medium DMEM-5030 supplemented with 25 mM glucose and 2 mM glutamine, pH adjusted to 7.4, at 37 • C. Four baseline measurements of OCR were sampled prior to sequential injection of mitochondrial complex V inhibitor oligomycin (1 μM), protonophore FCCP (1 μM) and antimycin A plus rotenone (1 μM each) to completely inhibit mitochondrial respiration. Accordingly, mitochondrial basal respiration, maximal respiration and ATP production was automatically calculated and recorded by the Seahorse software. Data was normalized for protein levels.

[9] 343w MitoDsRed-and GFP/SIRT3-GFP-transfected striatal cells were fixed with 4% paraformaldehyde (pre-warmed at 37 • C) for 20 min, permeabilized in 0.1% Triton X-100 in PBS for 2 min and blocked for 1 h, at room temperature in 3% (w/v) BSA in PBS. Anti-TFAM and anti-Drp1 primary antibodies were prepared in 3% (w/v) BSA in PBS and incubated overnight at 4 • C. Nuclei were labeled with 4 μg/mL Hoechst 33342 for 20 min and mounted using Mowiol 40-88. Confocal images were obtained using a Plan-Apochromat/1.4NA 63x lens on an Axio Observer.Z1 confocal microscope (Zeiss Microscopy, Germany) with Zeiss LSM 710 software. Mitochondrial morphology and protein colocalization analysis was achieved using Macros designed by Doctor Jorge Valero (Center for Neuroscience and Cell Biology -University of Coimbra, presently at Achucarro -Basque Center for Neuroscience, Spain) in Fiji. Briefly, image background was normalized using the function Subtract Background, included in Fiji (rolling ball radius: 10 μm for mitochondria and proteins involved in mitochondrial dynamics; 30 μm for SIRT3-GFP/GFP, for 16-bit images). Mitochondria-targeting MitoDsRed images were extracted to grayscale. FindFoci function was then used to allow the identification of peak intensity regions [26] in order to show mitochondria-specific fluorescence. A threshold was applied to optimally resolve individual mitochondria. Mitochondrial outlines were traced through the Analyze Particles function. Aspect ratio (the ratio between the major and minor axis of mitochondria) was used as an index of mitochondrial length alongside Roundness (a relation between the area of mitochondria and its major axis). For proteins involved in mitochondrial dynamics a threshold was set similarly to the one described above, and Set Measurements function was employed to obtain Integrated Density (product of Area and Mean Gray Value) and later normalized to cellular area using the designed ROI area as an approximation. To obtain information about protein co-localization in mitochondria, a selection of mitochondrial ROIs was done, and the respective Integrated Density inside the ROIs was considered. SIRT3-GFP and GFP were analyzed in the same way, considering also the value of Integrated Density inside the nuclei. Each value derived represents a single cell.

[10] 108w Flies were raised on maize media, in LD12:12 at 25 • C.The elav-GAL4 [c155] and the mef2-GAL4 driver lines were obtained from the Bloomington Stock Centre, Indiana. The w;UAS Htt93Q/CyO; + flies were a gift from Larry Marsh and Leslie Thompson (University of California, Irvine) [27]. The w; + ;UAS dSirt2 line was a gift from Karen Chang (University of Southern California) [28]. Fly crosses were performed to generate a Htt93Q dSirt2 line (genotype: w; UAS Htt93Q/CyO; UAS dSirt2). Experimental flies were generated by crossing w;UAS Htt93Q/CyO; + and w; UAS Htt93Q/CyO; UAS dSirt2 flies to either elav-GAL4 (for pan-neuronal expression) or mef2-GAL4 (for expression in the muscles).

[11] 56w 48 to 245 virgin females were collected for each experimental genotype and kept at 25 • C in groups of 10 per vial. The number of surviving flies was scored in each vial on every day, and the flies moved to fresh food twice a week. Survival curve analysis were performed using the Log-rank (Mantel-Cox) test.