Promotion of mitochondrial biogenesis by necdin protects neurons against mitochondrial insults

Neurons rely heavily on mitochondria for their function and survival. Mitochondrial dysfunction contributes to the pathogenesis of neurodegenerative diseases such as Parkinson's disease. PGC-1α is a master regulator of mitochondrial biogenesis and function. Here we identify necdin as a potent PGC-1α stabilizer that promotes mitochondrial biogenesis via PGC-1α in mammalian neurons. Expression of genes encoding mitochondria-specific proteins decreases significantly in necdin-null cortical neurons, where mitochondrial function and expression of the PGC-1α protein are reduced. Necdin strongly stabilizes PGC-1α by inhibiting its ubiquitin-dependent degradation. Forced expression of necdin enhances mitochondrial function in primary cortical neurons and human SH-SY5Y neuroblastoma cells to prevent mitochondrial respiratory chain inhibitor-induced degeneration. Moreover, overexpression of necdin in the substantia nigra in vivo of adult mice protects dopaminergic neurons against degeneration in experimental Parkinson's disease. These data reveal that necdin promotes mitochondrial biogenesis through stabilization of endogenous PGC-1α to exert neuroprotection against mitochondrial insults.

M ammalian neurons require high mitochondrial activities to generate a large amount of ATP for their signalling events such as action potential generation and excitatory synaptic transmission 1 . Mitochondria are also involved in neuronal death and contribute to neuroprotection against various detrimental stresses 2 . Furthermore, mitochondrial abnormalities are suggested to contribute to the pathogenesis of neurodegenerative diseases such as Parkinson's disease (PD), amyotrophic lateral sclerosis (ALS), Alzheimer's disease and Huntington's disease 3,4 . However, little is known about the regulatory mechanisms of mitochondrial biogenesis in mammalian neurons under physiological and pathological conditions.
The peroxisome proliferator-activated receptor g coactivator-1 (PGC-1) family, which consists of PGC-1a, PGC-1b and PRC, plays a central role in governing a transcriptional regulatory network for mitochondrial biogenesis and respiratory function 5 . The PGC-1 family transcriptional coactivators enhance the activities of the nuclear respiratory factors NRF1 and NRF2, which induce transactivation of many genes encoding mitochondria-specific proteins involved in respiratory chain, mitochondrial DNA transcription/replication and protein import/assembly 6 . PGC-1a is the first identified PGC-1 family member 7 , and its expression and function have been most extensively studied 5 . In non-neuronal cells, expression of PGC-1a is dynamically regulated at the transcriptional and posttranslational levels in response to various environmental stimuli such as temperature, nutritional status and physical activity 5,8 . However, there is limited information on the regulation of neuronal PGC-1a and its involvement in mitochondrial biogenesis.
Necdin is a MAGE (melanoma antigen) family protein originally isolated from neurally differentiated embryonal carcinoma cells 9 . Necdin is expressed in virtually all neurons throughout the nervous system 10 . The necdin gene (gene symbols; Ndn for mouse, NDN for human) is expressed only from the paternal allele via genomic imprinting, a mammal-specific epigenetic regulation of gene expression 11,12 . Necdin interacts with the major transcription factors E2F1 and p53 to suppress cell proliferation and apoptosis [13][14][15][16] . Moreover, necdin binds to Sirt1, an NAD þ -dependent protein deacetylase involved in the regulation of energy homeostasis, and facilitates Sirt1-mediated deacetylation of the transcription factors p53 and FoxO1 in neurons 16,17 . These findings suggest that necdin interacts with major nuclear proteins to modulate the transcriptional regulation networks in mammalian neurons.
We here report that necdin facilitates neuronal mitochondrial biogenesis via PGC-1a stabilization by suppressing its proteolytic degradation in the ubiquitin-proteasomal system. Necdin forms a stable complex with PGC-1a in the nucleus of cortical neurons to maintain high mitochondrial activities. Furthermore, we demonstrate that necdin exerts potent neuroprotective effects on dopaminergic neurons against mitochondrial complex I inhibitors that are commonly used for modelling PD 18 . Our findings will provide a better understanding of the regulatory mechanism underlying neuronal mitochondrial biogenesis under physiological and pathological conditions.

Results
Necdin promotes neuronal mitochondria-related gene expression. To investigate whether necdin modulates specific gene transcription networks in brain neurons, we performed microarraybased gene expression profiling in necdin-null cortical neurons (GEO accession; GSE63498). In gene ontology analysis for reduced gene expression in necdin-null neurons, the term mitochondrion in the cellular component category was the most significantly enriched (Fig. 1a). Of 61 downregulated genes (Fig. 1b and Supplementary Table 1), 10 genes encoding mitochondria-specific proteins were selected, and their expression levels were determined by quantitative reverse transcription-PCR (qRT-PCR) (Fig. 1c). In necdin-null neurons, the mRNA levels of Tomm20, Tomm22, Tomm40, Timm9 and Timm50, which encode mitochondrial import receptors, decreased by 41-53%, and those of Ndufs3 and Atp5c1, which encode electron transport chainassociated enzyme components, decreased by 24% and 32%, respectively. However, cytochrome c (Cyc1), Atp5d and Atp5f1 mRNA levels were unchanged. We also quantified the expression of mitochondrial biogenesis-regulatory genes by qRT-PCR. Nrf1, Gabpa (also known as Nrf2) and Tfam mRNA levels decreased significantly in necdin-null neurons, whereas no significant change in PGC-1a (Ppargc1) mRNA expression was observed. The gene expression signature of necdin-null neurons indicates that necdin promotes mitochondrial biogenesis in cortical neurons.
We determined the effect of necdin on mitochondrial metabolic activity in primary cortical neurons using 3-[4,5dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide (MTT), a tetrazolium dye metabolized into water-insoluble formazan mainly in mitochondria. Necdin-null primary neurons exhibited normal morphology, but contained reduced MTT-formazan deposits (Fig. 2a). In quantitative MTT assay, MTT-formazan levels in necdin-null neurons decreased by 41% in the conditions where cellular damage was absent as assessed by lactate dehydrogenase (LDH) release (Fig. 2b). We further analysed oxidative phosphorylation complex I activity in primary cortical neurons by immunocapture assay. Complex I activity was reduced in necdin-null neurons by 60% (Fig. 2c,d). Cellular ATP levels were also significantly reduced in necdin-null neurons, but not in necdin-null neural progenitor cells (NPCs), indicating that necdin promotes the mitochondrial activity predominantly in differentiated neurons (Fig. 2e). We then examined whether mitochondrial membrane potential is reduced in necdin-null neurons. The levels of chloromethyl-X-rosamine, a mitochondrial membrane potential-dependent fluorescent probe, were reduced by 28% in necdin-null neurons (Fig. 2f).
PGC-1a serves as a master regulator of mitochondrial biogenesis in mammalian cells 5,7 . Because we found no significant change in PGC-1a expression at the mRNA level in necdin-null neurons, we analysed the expression of PGC-1a at the protein level using a novel antibody raised against mouse PGC-1a (PGCAN) (Supplementary Fig. 1). In NPCs and cortical neurons, PGC-1a was detected as a major 128-kDa band by western blotting, and wild-type cortical neurons expressed higher PGC-1a levels than wild-type NPCs and necdin-null cortical neurons (Fig. 2g). Furthermore, the expression level of PGC-1a in necdin-null neurons was 55% of the wild-type control level (Fig. 2h,i). We then analysed the expression of the PGC-1a protein in various tissues at embryonic day 14.5 (E14.5) by western blotting (Fig. 2j). PGC-1a was clearly detected in the brain, slightly in the skeletal muscle (gastrocnemius) and hardly detectable in the heart and liver. ATP levels in the brain and muscle, where necdin was highly expressed, were significantly lower in necdin-null mice than in wild-type mice at E14.5   (a) Gene expression profiling of primary cortical neurons prepared from wild-type and necdin-null mice at E14.5 was performed using DNA microarrays. Enriched gene ontology terms of genes whose expression decreased (420% at Po0.05) in necdin-null cortical neurons. P value cutoff (yellow line), 1.3 ¼ À log 10  ( Fig. 2k). The brain levels of PGC-1a were the highest at E14.5 and postnatal day 0 (P0), and reduced markedly at adult stages (Fig. 2l). Expression levels of MTCO1 and necdin were the highest at E14.5 and decreased during postnatal development. Necdin-null mice had lower levels of PGC-1a and MTCO1 than wild-type mice throughout brain development. Interestingly, ATP levels in the brain also decreased sharply during development (o10% and o1% of the E14.5 level at P0 and 17 months, respectively) and were significantly reduced in the brain of necdin-deficient mice (Fig. 2m). We further examined the effect  (h,i) PGC-1a expression in cortical neurons was analysed by immunoblotting (h) and quantified by densitometry (i) (n ¼ 3). (j) PGC-1a and necdin expression levels in the brain, gastrocnemius muscle, heart and liver prepared from WT and KO mice at E14.5 were analysed by western blotting.
(k) ATP levels in the tissues indicated were measured by chemiluminescence assay (n ¼ 4). (l) Expression of PGC-1a, MTCO1 and necdin in the brain of WT and KO mice at different ages was analysed by western blotting. (m) ATP levels in the brain at indicated ages were measured by chemiluminescence assay (n ¼ 5 of necdin on mitochondrial degradation in cortical neurons using carbonyl cyanide m-chlorophenyl hydrazone, a mitochondrial uncoupler, as described previously 19 . There was no difference in the expression levels of the autophagy marker LC3-II between wild-type and necdin-null neurons treated with carbonyl cyanide m-chlorophenyl hydrazone ( Supplementary Fig. 2), suggesting that necdin fails to affect mitochondrial degradation.
Necdin and PGC-1a are colocalized in neuronal nucleus. We next investigated the distribution patterns of necdin and PGC-1a in the neocortex by immunohistochemistry. In E14.5 mouse forebrain, virtually all cortical neurons expressed both PGC-1a and necdin (Fig. 3a). Most PGC-1a-expressing cells overlapped with necdin-expressing cells in the cortical plate where differentiated neurons are present. PGC-1a and necdin were colocalized in the nucleus of primary cortical neurons as analysed by confocal laser-scanning microscopy (Fig. 3b). Furthermore, fluorescence microphotometry revealed that PGC-1a and MTCO1 immunoreactivities were significantly reduced in the neocortex of necdin-null mice ( Fig. 3c-e). These reductions were observed in the cortical plate but not in the ventricular/ subventricular zone, where undifferentiated neural precursors are present. Western blot analysis revealed that expression of PGC-1a and MTCO1 decreased by 46% and 77%, respectively, in necdin-null forebrain extracts ( Fig. 3f,g). Tissue ATP levels were markedly reduced in the dorsal cortical area, which contains the cortical plate, of necdin-null mice but not in the ventral area containing undifferentiated cell populations (Fig. 3h). These data suggest that necdin upregulates the protein levels of PGC-1a and MTCO1 in the brain in vivo.
To clarify whether necdin inhibits proteolytic degradation of PGC-1a by the ubiquitin-proteasome pathway, we examined the effects of the protein synthesis inhibitor cycloheximide (CHX) and the proteasome inhibitor MG132 on PGC-1a expression levels in transfected HEK293A cells. Necdin markedly increased the PGC-1a level irrespective of the presence or absence of CHX ( Fig. 4g) or MG132 (Fig. 4h), indicating that necdin strongly inhibits the degradation of PGC-1a in the proteasome. We then examined the effect of necdin on ubiquitin-dependent degradation of PGC-1a using Rnf34, a PGC-1a E3 ubiquitin ligase 22 . Rnf34 reduced the PGC-1a level, and necdin completely inhibited the reduction (Fig. 4i). In addition, necdin strongly suppressed Rnf34-mediated ubiquitination of PGC-1a (Fig. 4j). Necdin also protected PGC-1a against ubiquitination mediated by Fbxw7, another PGC-1a E3 ubiquitin ligase 23 (Fig. 4k). These data indicate that necdin stabilizes PGC-1a by inhibiting its degradation in the ubiquitin-proteasomal system.
Necdin prevents oligomycin-induced neurodegeneration. To examine whether necdin promotes mitochondrial biogenesis and function, we transferred the mouse necdin gene into primary cortical neurons using lentivirus vectors (LVs). LV-mediated overexpression of necdin increased the PGC-1a levels by 49% and 91% in wild-type and necdin-null neurons, respectively, as compared with the Emerald Green Fluorescent Protein (EmGFP) control levels ( Fig. 5a,b). Similarly, necdin overexpression increased the MTCO1 levels 2.1-and 2.7-fold in wild-type and necdin-null neurons, respectively (Fig. 5c). We also examined the effects of necdin overexpression on mitochondrial activities in primary cortical neurons. LV-mediated overexpression of necdin increased MTT levels by 43 and 73% in wild-type and necdin-null neurons, respectively (Fig. 5d). Similarly, necdin overexpression increased the ATP levels by 55% and 63% in wild-type and necdin-null neurons, respectively (Fig. 5e). In necdin-null neurons, necdin increased the MTT and ATP levels to those similar to wild-type control levels.
We then examined whether necdin-null cortical neurons are susceptible to oligomycin, a mitochondrial ATP synthase inhibitor. Oligomycin significantly increased apoptosis of necdinnull neurons during the 4-24 h period ( Fig. 5f), suggesting that endogenous necdin suppresses oligomycin-induced apoptosis of cortical neurons. In the LDH-based cytotoxicity assay, oligomycin increased the LDH levels by B50% in wild-type neurons and 2.3-fold in necdin-null neurons under LV-EmGFP-infected conditions (Fig. 5g), indicating that necdin-null neurons are highly susceptible to oligomycin. Remarkably, LV-mediated overexpression of necdin completely protected necdin-null neurons against oligomycin-induced damage. These results indicate that necdin promotes mitochondrial function to prevent oligomycin-induced degeneration of cortical neurons.
Necdin prevents neurodegeneration in experimental PD models. The mitochondrial complex I inhibitor MPP þ (1-methyl-4phenylpyridine) and its precursor MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine) selectively damage dopaminergic neurons and are commonly used for experimental PD models 18 . We first examined whether necdin prevents MPP þ -induced degeneration of SH-SY5Y cells, a human neuroblastoma cell line susceptible to MPP þ . LV-mediated overexpression of necdin increased the expression of PGC-1a and MTCO1 in SH-SY5Y cells as analysed by immunocytochemistry and western blotting ( Fig. 6a,b). Necdin markedly increased the expression levels of PGC-1a 2.3-and 18.4-fold (compared with EmGFP controls) and those of MTCO1 9.3-and 8.6-fold (compared with EmGFP controls) in SH-SY5Y cells in the absence and presence of MPP þ , respectively. Necdin increased the MTT levels 1.3-and 2.4-fold (compared with EmGFP controls) in the absence and presence of MPP þ , respectively (Fig. 6c). Similarly, necdin increased the ATP levels in SH-SY5Y cells by 38% and 79% in the absence and presence of MPP þ , respectively (Fig. 6d). In the cytotoxicity assay, MPP þ significantly increased the LDH levels in uninfected and LV-EmGFP-infected SH-SY5Y cells, whereas forced expression of necdin reduced the LDH levels near MPP þuntreated levels (Fig. 6e). These results suggest that necdin improves mitochondrial function to reduce MPP þ -induced damage of SH-SY5Y cells. We next investigated whether necdin prevents MPTP-induced degeneration of dopaminergic neurons in the substantia nigra (SN) in vivo using necdin-null mice in the C57BL/6J background, which is susceptible to MPTP. Tyrosine hydroxylase-expressing (TH þ ) dopaminergic cells in the SN pars compacta (SNpc) were quantified after MPTP treatment ( Fig. 7a and Supplementary  Fig. 5a). There was no difference in the number of SNpc TH þ cells between wild-type and necdin-null mice without MPTP treatment, whereas SNpc TH þ cells in necdin-null mice were significantly reduced in number when treated with MPTP (Fig. 7b,c). The numbers of TH þ dopaminergic neurons were reduced throughout the SNpc area of necdin-null mice. This suggests that endogenous necdin in SNpc dopaminergic neurons is neuroprotective against MPTP toxicity.
Because necdin potentially prevents neuronal apoptosis through suppressing the E2F1-Cdc2 and p53-Bax axes 15 Supplementary Fig. 5b,c). However, there were no significant differences in the expression levels of these mRNAs between wild-type and necdin-null mice, suggesting that these pathways are not involved in MPTP-induced degeneration of nigral dopaminergic neurons seen in necdin-null mice.
Necdin deficiency promotes neurodegeneration. We investigated the neuroprotective effect of endogenous necdin on SN dopaminergic neurons using necdin-null mice. We first analysed the number of TH þ neurons in the SNpc of wild-type and necdin-null mice at their adult (17 weeks old) and late adult (60 weeks old) stages (Fig. 9a). Although there was no significant difference in the number of SNpc TH þ cells between wild-type and necdin-null mice at 17 weeks of age, the number of SNpc TH þ cells was significantly reduced by 15% in necdin-null mice at 60 weeks of age. We then analysed the expression levels of TH, PGC-1a and MTCO1 in the SN by western blotting (Fig. 9b,c). Similar to the results of TH þ neuron counting, the expression level of TH was decreased by 21% in the SNpc of necdin-null mice 60 weeks old (Fig. 9d). Furthermore, the expression levels of PGC-1a and MTCO1 in the SN decreased by 58% and 63%, respectively, in necdin-null mice 60 weeks old, whereas there were no differences in the SN levels of PGC-1a and MTCO1 between wild-type and necdin-null mice 17 weeks old (Fig. 9e,f). On the other hand, PGC-1a and MTCO1 levels in the cortex of necdin-null mice decreased at 17 weeks of age by 66% and 78%, respectively, and at 60 weeks of age by 37% and 67%, respectively. These results suggest that endogenous necdin prevents degeneration of nigral dopaminergic neurons by maintaining the level of mitochondrial biosynthesis.

Discussion
The present study has shown that necdin upregulates the expression of neuronal PGC-1a at the post-translational level. In primary cortical neurons and developing brain tissues, the levels of necdin parallel with those of PGC-1a and the mitochondrial DNA-encoded protein MTCO1. In addition, PGC-1a and MTCO1 levels in the cerebral cortex in vivo are much lower in necdin-null mice than in wild-type mice at adult stages. These results suggest that necdin is an intrinsic positive regulator of PGC-1a-mediated mitochondrial biogenesis. Virus vector-mediated overexpression of necdin in primary cortical neurons and nigral dopaminergic neurons in vivo upregulates the expression of PGC-1a and MTCO1, indicating that endogenous expression levels of those proteins are controllable by exogenous necdin. Thus, necdin is likely to act as a molecular rheostat to control neuronal mitochondrial biogenesis via PGC-1a. We speculate that necdin and PGC-1a, both of which are expressed in virtually all neurons, cooperate to maintain high mitochondrial activities required for neuronal function and survival.
We have employed the mitochondrial oxidative phosphorylation inhibitors in combination with viral vector-mediated necdin gene delivery to demonstrate the neuroprotective effects of necdin. The combination of primary cortical neurons and the ATP synthase inhibitor oligomycin showed that necdin-null neurons are highly susceptible to oligomycin-induced apoptosis, which is fully prevented by overexpression of necdin. To investigate the effects of necdin on neurodegeneration in vivo, we have selected nigral dopaminergic neurons, which are highly susceptible to the mitochondrial complex I inhibitor MPTP, a lipophilic protoxin metabolized in glial cells to MPP þ , which is concentrated in dopaminergic neurons via the dopamine transporter 18 . Before the analysis of in vivo PD model, we confirmed the toxic effect of the active MPTP metabolite MPP þ on SH-SY5Y neuroblastoma cells, a human catecholaminergic cell line that exhibits apoptotic changes in response to MPP þ (refs 26,27). Necdin-mediated protection of SH-SY5Y cells against MPP þ toxicity led to further demonstrate the neuroprotective effect of necdin on nigral dopaminergic neurons in vivo against MPTP-induced degeneration. These findings implicate that mitochondrial dysfunction-induced neurodegeneration is preventable by enhancing neuronal mitochondrial biogenesis through overexpression of necdin.
The present study clarified that necdin strongly stabilizes PGC-1a by inhibiting its degradation in the ubiquitinproteasome system. Necdin interacted with PGC-1a via the N-terminal region that contains the PEST (proline, glutamic acid, serine, threonine) sequence, a domain critical for ubiquitinproteasomal degradation of PGC-1a in the nucleus 28 Fbxw7-mediated ubiquitination of PGC-1a. We speculate that necdin forms a stable nuclear complex with PGC-1a in neurons and protects PGC-1a from ubiquitination mediated by these E3 ubiquitin ligases. The N-terminal aminoacid sequences of PGC-1a and PGC-1b are well conserved 5 , and these PGC-1 family proteins increase mitochondrial density in primary cortical neurons 21 . Thus, we infer that necdin protects PGC-1a and PGC-1b against proteasomal degradation to promote their effects on mitochondrial biogenesis in mammalian neurons.
A genome-wide expression study revealed that PGC-1aregulated genes controlling cellular bioenergetics are underexpressed in SN neurons affected by PD 31 . PGC-1a-null mutant mice exhibit striatal lesions 32 , and their SN neurons are highly susceptible to MPTP 33 . Moreover, PGC-1a overexpression protects SN neurons against neurodegenerative insults 34,35 . These findings imply that PGC-1a contributes to the resistance of SN dopaminergic neurons against PD-associated neurodegeneration. In contrast to the beneficial effects of PGC-1a, sustained overexpression of PGC-1a in the SN induces degeneration of dopaminergic neurons 36,37 . These inconsistent results suggest that control of exogenous PGC-1a expression levels is crucial for the neuroprotective effects. In contrast, AAV-mediated overexpression of necdin increases endogenous PGC-1a protein levels by inhibiting its ubiquitinproteasomal degradation at the post-translational level. Thus, we speculate that the necdin gene delivery increases the PGC-1a levels in dopaminergic neurons within a physiologically tolerable extent that is sufficient for effective neuroprotection.
Recombinant AAV displays efficient transduction of SN dopaminergic neurons in vivo and has been often used for analysing the pathogenic mechanism of PD and developing therapeutic strategies. For example, AAV-mediated overexpression of a-synuclein, a major structural component of Lewy bodies seen in neurons undergoing degeneration in PD, induces PD-like neurodegeneration 38 . AAV-mediated overexpression of Parkin, a ubiquitin E3 ligase whose mutations cause recessively inherited early-onset PD 39 , ameliorates a-synuclein-mediated or MPTPinduced degeneration of dopaminergic neurons 25,40 . Moreover, AAV-mediated delivery of PARIS (also known as ZNF746), a substrate of Parkin, induces selective dopaminergic neuron loss, which is prevented by AAV-mediated overexpression of Parkin 34 . Gene therapy for PD using AAV-mediated gene delivery has reached human clinical trials 41 . Thus, we propose that AAV-mediated necdin gene delivery provides a novel strategy for mitochondrial biogenesis-based neuroprotection in PD.
Human necdin gene (NDN) is located on chromosome 15q11-q12, and its expression is absent in neurons affected by Prader-Willi syndrome (PWS), a classic genomic imprintingassociated neurodevelopmental disorder 11,12 . The present study has shown that the brain levels of mitochondrial proteins and mitochondrial activities are significantly reduced in necdin-null neurons. These findings raise the possibility that necdin-null neurons affected in PWS have low mitochondrial activities and reduced ATP levels during neuronal development. Motor proteins such as myosin, kinesin and dynein families are powered by ATP hydrolysis and play key roles in neuronal morphogenesis and network formation during brain development 42 . Mice carrying mutated paternal Ndn exhibit a variety of neurodevelopmental abnormalities such as reduced neuron number, impaired neuronal migration and abnormal axon extension in the embryonic brain [43][44][45][46][47][48][49] . Although there is limited information about the neuropathological abnormalities in the PWS-affected brain, earlier studies using magnetic resonance imaging have revealed multiple morphological abnormalities such as ventriculomegaly, reduced sizes of specific areas and differences in grey and white matter volumes in the brain of patients with PWS [50][51][52] . Thus, we speculate that reduced mitochondrial activities in necdin-null neurons contribute, at least in part, to these neurodevelopmental abnormalities seen in PWS. Our findings suggest that necdin and PGC-1a cooperate to promote mitochondrial biogenesis in various types of neurons and prevent mitochondria-associated neurodegeneration. Necdin-null mice exhibit low expression of the PGC-1a protein in the brain, where the number of nigral dopaminergic neurons is significantly reduced in late adulthood. Necdin is abundantly expressed in spinal cord motor neurons 10 , and their mitochondrial dysfunction causes ALS 53 . PGC-1a is also suggested to contribute to the pathogeneses of ALS 54 . Intriguingly, necdin expression in spinal cord motor neurons of SOD1 G93A mutant mice, an SOD1 gene mutant ALS model, increases significantly at the early presymptomatic stage and decreases at the late symptomatic stage 55 . We speculate that necdin expression is upregulated for neuroprotection at the early stage of neurodegeneration but declines at the advanced stage. A multicohort transcriptional meta-analysis has revealed that necdin expression is specifically diminished in major human neurodegenerative diseases 56 . These findings raise the possibility that necdin is involved in neuronal resistance or resilience against neurodegenerative insults. Thus, gene therapy using virus vector-mediated necdin gene delivery into specific neurons at risk will be a promising avenue for prevention or therapeutic intervention of neurodegenerative diseases. The present findings also warrant further studies on the neuron-specific mechanism of mitochondrial biogenesis and its association with neurodegenerative diseases.

Methods
Ndn knockout mice. Ndn knockout mice (Ndn tm/Ky ) were generated and maintained as described 46 . Heterozygous male mice (Ndn þ m/ À p ) (425 generations on the ICR background and 20 generations on the C57BL/6J background) were crossed with wild-type (Ndn þ m/ þ p ) female mice to obtain Ndn þ m/ þ p and Ndn knockout (Ndn þ m/-p ) littermates. Genotypes of all mice were analysed by PCR for mutated Ndn locus. C57BL/6J mice were used for demonstrating MPTP-induced neurodegeneration and phenotypes of Ndn knockout mice (13, 17 and 60 weeks of age). The study was approved by the Animal Experiment Committee (Approval No. 24-04-0) and Recombinant DNA Committee (Approval No. 3642) of Institute for Protein Research, Osaka University, and were performed in accordance with national, institutional and the ARRIVE guidelines.
Cell cultures. Primary cortical neurons were prepared from the forebrain of ICR mice (Japan SLC) at E14.5 as described 16 . The cortex was dissected and incubated in 0.5 ml of Ca 2 þ /Mg 2 þ -free Hanks' balanced salt solution with 0.05% trypsin for 5 min at 37°C. Tissues were dissociated with 10% fetal bovine serum (FBS) in Dulbecco's modified eagle medium (DMEM) and centrifuged at 200g for 3 min. Pellets were resuspended in Neurobasal medium (Life Technologies) supplemented with 2 mM L-glutamine, kanamycin/penicillin and B-27 supplement (1:50 dilution, Life Technologies), plated at a density of B1 Â 10 5 cells per cm 2 in culture dishes, and incubated for 4 days in vitro (4 DIV) before analyses. For preparation of primary NPCs, dissociated cortical cells were cultured as floating neurospheres in DMEM/F12 (Life Technologies) supplemented with B-27 (1:50 dilution), 20 ng ml À 1 epidermal growth factor (PeproTech) and 20 ng ml À 1 basic fibroblast growth factor (PeproTech) for 48 h at 37°C under humidified 5% CO 2 conditions. HEK293A cells (Life Technologies) and SH-SY5Y cells (gift from Dr June Biedler, Memorial Sloan-Kettering Cancer Center) 26 were cultured in DMEM containing 10% FBS at 37°C under humidified 5% CO2 conditions.   Complex I activity assay. Complex I activity was measured using an assay kit based on the immunocapture with immobilized anti-Complex I antibodies combined with in-gel activity measurement (dipstick assay, ab109720, Abcam). Cell lysates (30 mg protein) were used for the assay according to the manufacturer's protocol. Complex I activity was measured by incubating with NADH and nitrotetrazolium blue. Signals of resulting blue-purple precipitates on the dipstick were analysed by densitometry and quantified using ImageJ 1.44 software.
ATP assay. ATP levels in cell lysates were analysed by luciferase chemiluminescence-based assays (CellTiter-Glo Luminescent Cell Viability Assay kit, Promega) according to the manufacturer's protocols. Briefly, cell lysates were mixed with CellTiter-Glo Reagent for 2 min by shaking vigorously and settled for 10 min. Chemiluminescence in reaction mixtures was measured with a luminometer (Lumat LB9501, Berthold).
Mitochondrial membrane potential assay. For determination of mitochondrial membrane potential in neurons, cortical neurons at 4 DIV were incubated with 100 nM chloromethyl-X-rosamine (MitoTracker Red, Life Technologies) at 37°C for 30 min, and analysed by flow cytometry with FACSCalibur (BD Biosciences).
Fluorescence microphotometry. Cortical sections or transfected cells were immunostained for PGC-1a and MTCO1. Signal intensities were quantified by fluorescence microphotometry as described 17  Co-transfection assays. HEK293A cells were transfected with combinations of expression vectors by the calcium phosphate method and collected after 24 h. For CHX treatment, transfected HEK293A cells were incubated with DMEM/10% FBS containing 500 mM CHX (Sigma-Aldrich) for 60 min. HEK293A cells were transfected with combinations of cDNAs and treated with 10 mM MG132 (Peptide Institute) for 3 h before harvest. For PGC-1a ubiquitination assay, full-length cDNAs encoding the PGC-1a E3 ubiquitin ligases Rnf34 (NM_030564) and Fbxw7 (NM_080428) were cloned from E14.5 mouse forebrain cDNAs used as a PCR template, sequenced, attached with the V5-encoding sequence at the 3 0 -end, and subcloned into pcDNA3.1 þ . Proteins were immunoprecipitated with anti-PGC-1a antibody (PGCAN,1:50) and analysed by western blotting.
In vitro binding analysis. Deletion mutants of PGC-1a were subcloned into pMAL-C2 vector to make MBP fusion proteins. MBP-fused mutant proteins were affinity purified with amylose resin, and incubated with His-tagged necdin (200 ng) at 4°C for 30 min in 0.5 ml of the binding buffer containing 20 mM Tris-HCl, pH7.5, 200 mM NaCl and 1 mM EDTA as described 16 . After washing, bound His-tagged necdin was eluted with 20 mM maltose and detected by western blotting with anti-necdin antibody. MBP fusion proteins were detected by Coomassie Brilliant Blue staining.
Viral vectors. Recombinant LVs were produced in HEK293FT cells by transfecting SIN vector plasmids and two or three helper plasmids using calcium phosphate method as described 59 . Necdin and EmGFP cDNAs were subcloned into pENTR1A entry vector (Life Technologies) to construct the destination vectors CSII-EF1a-necdin-IRES-EmGFP and CSII-EF1a-IRES-EmGFP to make LV-Ndn and LV-EmGFP, respectively. EmGFP (Life Technologies) was used for an expression indicator and negative control for necdin overexpression. The viral titre was measured by serial dilution on HEK293FT cells and determined as GFP-positive cells by fluorescence-activated cell sorting analysis. For AAV serotype-1 vector preparation, pAAV-MCS carrying cytomegalovirus promoter (Stratagene) carrying human necdin cDNA (NM_002487) and humanized GFP were used to make AAV-NDN and AAV-GFP, respectively, as described 25 . For high-titre viral stocks, AAV vectors were purified by ultracentrifugation in a density gradient with OptiPrep (Axis-Shield PoC AS), which was removed by ultrafiltration using Centricon Plus-20 (10,000 molecular weight cut-off, Millipore). Averaged titres of AAV-NDN and AAV-GFP were 1 Â 10 12 genomes per ml.
Oligomycin-induced neurotoxicity assay. Primary cortical neurons were cultured for 4 days and treated with 20 mg ml À 1 oligomycin (Sigma-Aldrich). For quantifying apoptotic neurons, cultures were stained with 3.3 mM Hoechst 33342 (Sigma-Aldrich) for 5 min before fixation, and cortical neurons carrying condensed or fragmented nuclei were counted. For LV infection, primary cortical cells were infected with LVs at multiplicity of infection (m.o.i.) of 2, incubated in DMEM/F12-based medium containing 20 ng ml À 1 epidermal growth factor and 20 ng ml À 1 basic fibroblast growth factor for 30 min, and cultured for 4 days in Neurobasal medium for neuronal differentiation. Mean viral infection efficiency was more than 92%. For LDH release assay, neurons were treated with oligomycin for 6 h. MPTP-induced neurodegeneration analysis. In necdin-null mice of C57BL/6J background, 13-week-old mice were treated with MPTP-HCl (30 mg per kg body weight per day, Sigma-Aldrich) dissolved in saline for 5 consecutive days. Control mice without MPTP treatment were injected with saline. AAV vectors were stereotaxically injected into the SN of 13-week-old male C57BL/6J mice (Japan SLC) as described 25 . AAV-infected mice were injected intraperitoneally 42 days after AAV infection, with MPTP for 5 consecutive days. Control mice without MPTP treatment were injected with saline. MPTP-treated mice were killed 21 days after the last injection of MPTP. MPTP was handled in accordance with the guidelines 60 . For quantifying TH-expressing (TH þ ) cells, coronal 20-mm-thick brain sections were cut serially using a cryostat (CM1900, Leica Microsystems). Sections were stained with anti-TH antibody and counterstained with Nissl. TH-and Nissl-double-positive neurons in the SNpc were analysed by unbiased stereological counting method 25 . Cells having optimally visualized nuclei and nucleoli were counted to avoid double counting. TH þ cells in every fourth 20-mm section were counted so that 15 sections (total 60 sections for B1.2 mm) cover the entire SNpc extent. For western blotting, brain blocks including the entire SN were cut coronally at 2-mm thickness. A ventral part of the midbrain including the SN (B1.2 mm from the ventral end) was dissected horizontally, and immediately frozen in liquid nitrogen for tissue extraction. Pole test was performed 3 days before western blot analysis according to the method described 61 at 20:00 . Mice were placed on the top of a 48-cm-long 1-cm-diameter wooden rod. Mice performed three trials with 30-s intervals, and success rates to reach the floor within 80 s were measured.
Statistics. Statistical significance was tested using an unpaired Student's t-test, one-way analysis of variance followed by Tukey-Kramer post hoc test, or w 2 -test. A significance of Po0.05 was required for rejection of the null hypothesis.