Autophagy in neurodegeneration: New insights underpinning therapy for neurological diseases

In autophagy long-lived proteins, protein aggregates or damaged organelles are engulfed by vesicles called autophagosomes prior to lysosomal degradation. Autophagy dysfunction is a hallmark of several neurodegenerative diseases in which misfolded proteins or dysfunctional mitochondria accumulate. Excessive autophagy can also exacerbate brain injury under certain conditions. In this review, we provide specific examples to illustrate the critical role played by autophagy in pathological conditions affecting the brain and discuss potential therapeutic implications. We show how a singular type of autophagy-dependent cell death termed autosis has attracted attention as a promising target for improving outcomes in perinatal asphyxia and hypoxic-ischaemic injury to the immature brain. We provide evidence that autophagy inhibition may be protective against radiotherapy-induced damage to the young brain. We describe a specialized form of macroautophagy of therapeutic relevance for motoneuron and neuromuscular diseases, known as chaperone-assisted selective autophagy, in which heat shock protein B8 is used to deliver aberrant proteins to autophagosomes. We summarize studies pinpointing mitophagy mediated by the serine/threonine kinase PINK1 and the ubiquitin-protein ligase Parkin as a mechanism potentially relevant to Parkinson’s disease, despite debate over the physiological conditions in which it is activated in organisms. Finally, with the example of the autophagy-inducing agent rilmenidine and its discrepant effects in cell culture and mouse models of motor neuron disorders, we illustrate the importance of considering aspects such a disease stage and aggressiveness, type of insult and load of damaged or toxic cellular components, when choosing the appropriate drug, timepoint and duration of treatment. the complex interplay between autophagy and other intracellular processes that may also at some point be recruited by the pathological process, either as contributing factors or protective responses. As illustrated with PINK1 and Parkin, many proteins involved in autophagy and its regulation have multiple functions in the cell, and these have to be taken into consideration if we want to explore meaningful therapeutic avenues for neurological diseases. The case of CASA, on the other hand, illustrates the interplay between autophagy and the proteasome system, leaving room for envisaging strategies aimed at promoting alternative degradation routes when autophagy is impaired. Finally, while autophagy and mitophagy are emerging as central modulators of inflammatory and immune-related mechanisms, which are recognized as a component of the pathological process in various brain diseases, work discussed here also shows that the selective modulation of autophagy in neurons has broader impact on brain tissue homeostasis. Therefore, a better knowledge of the different cell types affected by the disease process and their complex interactions will provide invaluable insight into the most appropriate ways of using autophagy modulation for therapeutic purposes in the brain. signaling in cell culture and animal models of Parkinson's disease.


I. Insult severity, dysfunctional autophagy and hypoxic-ischaemic injury.
There is an existent literature that patterns of neuronal injury are highly dependent upon the type, intensity and duration of insult, especially oxidative stress and excitotoxicity, and presumably load of toxic protein aggregates (Galluzzi et al. 2016). Indeed, we found in cultured cortical neurons that caspase-independent PCD induced by the oxidative stressor hydrogen peroxide was greatly attenuated by 3-methyladenine and knockdown with short interfering RNA (siRNA) directed at Atg7 and Beclin1 (BECN1) as shown by use of specific autophagic and cell death morphological markers. This diverse evidence supports the involvement of autophagy-mediated neuronal death, and also caspase-independent PCD, in a model with different interactive cell death pathways (Higgins et al. 2011). Consistent with the concept for a determinant role in neuronal of insult intensity, in severe oxidative stress there was even more extensive crosstalk between different forms of cell death with siRNA directed at Atg7 being less effective versus injury progression, even though autophagy-associated cell death was found (Higgins et al. 2012). Additionally, we found in a comparative analysis of our extensive microarray data that there are many commonalities in autophagiclysosomal genes regulated downstream of oxidative stress and excitotoxicity, and that these events lead to downstream inhibition of autophagy and the autophagic-lysosomal pathway (Yap et al. 2016).
Given that both oxidative stress and excitotoxicity are documented to contribute to hypoxic-ischaemic injury (HI) (Khoshnam et al. 2017), studies of the possible recruitment of autophagy to stroke were a logical extension of our work. Indeed, in post-mortem human brain tissue from patients with a history of stroke we found abundant microtubule-associated protein light chain 3 (LC3)-immunolabelled autophagic vacuoles (Frugier et al. 2016). Dense Sequestosome 1 (SQSTM1) labelling was also found, although immunopositive neurons were further away from the penumbral area unlike those found after LC3 labelling -both SQSTM1 mRNA and protein levels were also increased. This increased level of SQTM1 may be

Accepted Article
This article is protected by copyright. All rights reserved indicative of inhibition of autophagy, however our current understanding of the recruitment of the autophagic-lysosomal system in HI is very much in its infancy with existent evidence suggestive that the "load" of cellular debris and damaged proteins may determine the mode of recruitment of autophagy (Wen et al. 2008, Li et al. 2010, Shi et al. 2012. Clearly, given the dearth of drugs for the clinical management of HI, autophagy and its management represents an attractive target (vide infra).

II.
Selective neuronal deletion of Atg7 is protective in neonatal brain injury.

The neonatal brain differs from the adult brain.
Autophagy is crucial for maintaining homeostasis in response to stress in most eukaryotic cells, including neurons, mediating neuroprotection following several forms of brain damage. Involvement of autophagy in neuronal cell death has therefore been controversial, and efforts have been made to distinguish cell deathpromoting autophagy from autophagic responses that occur parallel to cell death, or even counteracting degenerative mechanisms. Emerging evidence indicates that inappropriate activation of autophagy can directly mediate neuronal cell death (Piras et al. 2017, Ginet et al. 2014b, Ginet et al. 2014a, Clarke & Puyal 2012). In the immature, neonatal brain, multiple mechanisms of cell death are often activated simultaneously, even in the same neuron (Puka-Sundvall et al. 2000, Blomgren et al. 2001, and the relative contributions of the various mechanisms are strikingly different from the adult brain (Zhu et al. 2005, Hu et al. 2000. In mice and rats, mixed morphological phenotypes indicative of necrotic, apoptotic, and autophagic features are present following HI-induced neuronal cell death in the immature brain, each confirmed by specific biochemical and morphological criteria (Li et al. 2010, Northington et al. 2011, Blomgren et al. 2007). Levels of caspase-dependent, caspase-independent cell death and lipidation of LC3 provided evidence for greater activity of the apoptotic and autophagic machinery in the immature brain than in the adult brain (Zhu et al. 2005). This is reasonable, given the massive growth and continuous remodelling, including removal of redundant cells and synapses, in the immature brain, where autophagy plays an integral role in maintaining cellular and tissue homeostasis (Oppenheim 1991, Semple et al. 2013. Neuronal autophagy has been reported to be enhanced in rodent models of perinatal cerebral ischemia or HI (Zhu et al. 2005, Liu et al. 2008, Koike et al. 2008, Carloni et al. 2010.
Neuroprotective measures to treat injury in the immature brain should most likely be quite different from those applied in the adult brain (Wang et al. 2009, Zhu et al. 2005, and autophagy has evolved as a particularly promising strategy in the immature brain

Accepted Article
This article is protected by copyright. All rights reserved Methodological aspects.
The role of autophagy in neonatal brain injury is controversial, with independent studies showing protective effects in brain injury in rodents following inhibition or induction of autophagy by pharmacological agents (Vaslin et al. 2009, Carloni et al. 2010. Confounding aspects are linked to the fact that pharmacological autophagy-modulating agents have poor specificity and that they have been used intraperitoneally or intravenously, although they may not readily cross the blood-brain barrier (Galluzzi et al. 2016). In addition, the methodologies used to monitor autophagic responses in vitro and in vivo, have not always been discriminative. Cellular markers used include expression of LC3, ATG5, BECN1; phosphorylation of AMP-activated protein kinase, Unc-51 Like Autophagy Activating Kinase 1 or substrates of mammalian target of rapamycin complex 1; interaction of BECN1 with B-cell lymphoma 2 (BCL-2) family members; LC3 lipidation; lysosomal acidification; and the degradation of p62 or autophagic substrates. Cytoplasmic accumulation of autophagosomes and autolysosomes, LC3 lipidation, and other morphological or biochemical features of autophagy have been exploited to imply increased autophagy in neurons and other cells in brain tissue (Klionsky et al. 2016). However, cytoplasmic vacuolization and LC3 lipidation can also increase when lysosomal function is impaired, or when autophagosomes cannot fuse with lysosomes. To distinguish between active autophagic responses and autophagy blockade, it is necessary to take into account autophagic flux (Klionsky et al. 2016), an aspect that has not been considered in many of the reports in the literature. Genetic approaches are more specific, but also have disadvantages. Genetically targeting a component of the autophagic machinery, for example Atg5 or Atg7, is more specific than pharmacological agents, but the targeted protein may also play additional roles unrelated to autophagy.
Furthermore, the genetic deletion is often constitutive and life-long, leading to adaptive changes of the tissue and its metabolism. Making the deletion inducible, not only in terms of which cells that will be targeted, but also when the deletion will occur, will make the interpretations of the results more accurate.
Autophagy is also involved in the release of danger signals and inflammation, both systemically and locally in the central nervous system, and these contribute to the overall short and long term outcomes of any brain injury. Future studies taking these aspects into consideration are needed to more accurately dissect the effects of autophagy in brain injuries.

Transient and specific inhibition of autosis reduces injury to the immature brain.
In 2013, a study was published demonstrating that selective over-activation of autophagy could cause a pattern of cell death with morphological characteristics different from apoptosis or necrosis, termed "autosis" (Liu et al. 2013) (Table 1). Autosis occurs in vitro and in vivo in cerebral HI of the immature brain, and is inhibited by Na + /K + -ATPase antagonists approved for clinical use. Lithium has been demonstrated to induce autophagy (Sarkar et al. 2005). However, we found it to be neuroprotective in a rat

Accepted Article
This article is protected by copyright. All rights reserved model of neonatal HI, where it reduced the levels of LC3 lipidation, consistent with autophagy inhibition (Li et al. 2011, Li et al. 2010, Xie et al. 2014. To circumvent the limited specificity of pharmacological agents, we investigated the effect of HI in mice with neuron-specific deletion of the essential autophagy gene, Atg7. Tissue loss was decreased by more than 40%, and caspase-dependent and -independent cell death was attenuated in several brain regions compared to wild-type mice . These findings suggest that autophagy inhibition may be a promising therapeutic avenue for the treatment of human neonates developing severe hypoxic-ischemic encephalopathy (HIE). These results extend similar observations in the hippocampus of a milder HI model (Koike et al. 2008), supporting a lethal role of autophagy in brain regions highly vulnerable to HI in neonates, such as cortex, thalamus and striatum.
Moreover, investigation of the abundance of LC3 II-and cathepsin D-positive cells in susceptible areas of the brain of human term newborns who died from severe asphyxia with HIE revealed increased neuronal autophagy compared to brains from neonates who died from other causes, without brain injury .
A different injury model, genotoxic stress induced through ionizing radiation, was used to model the effects of cranial radiotherapy in children treated for brain malignancies. Irradiation typically targets actively proliferating cells, making it effective in the treatment of high grade brain tumours. Very few cells in the brain proliferate under normal conditions, making the brain relatively radioresistant, but the neurogenic niches in the hippocampus and the subventricular zone, but also in the cerebellum, harbouring neural stem and progenitor cells, display overt cell death after irradiation (Naylor et al. 2008, Roughton et al. 2012, Fukuda et al. 2004. The loss of neurogenic capacity observed after a single moderate dose of irradiation increased over time, such that the difference compared to controls was bigger the longer we waited, up to more than one year , Bostrom et al. 2013. In paediatric patients, damage to normal brain tissue adjacent to the tumour is a major concern associated with adverse side effects (Kahalley et al. 2016). The young developing brain is more sensitive to irradiation than the adult brain (Fukuda et al. 2005, Duffner 2010). Radiotherapy can cause long-term cognitive impairment, secondary malignancies, and perturb growth and puberty. Irradiation-induced depletion of neural stem and progenitor cells may underlie some of the cognitive deficits. Using the same Atg7 deletion mouse strain as above, we subjected 10-day-old mice to a single dose of 6 Gy whole brain irradiation and found that the number of dying stem and progenitor cells in the dentate gyrus of the hippocampus was reduced by 60% (Wang et al. 2017). The ensuing inflammatory reaction, as judged by microglial activation and cytokine levels, was also reduced to a similar extent, but since Atg7 was deleted only in neurons it is likely that the reduced inflammation observed was due to the lower number of dying cells rather than autophagy inhibition. The large reduction of neural stem and progenitor cell death was somewhat surprising, given that Atg7 was only deleted in neurons, and the progenitors commit to a neuronal fate approximately 3 days after the stem cells

Accepted Article
This article is protected by copyright. All rights reserved start proliferating (Steiner et al. 2006, Brandt et al. 2003, and require 4 weeks to be mature neurons (Esposito et al. 2005, van Praag et al. 2002. The deletion of Atg7 was driven by a neuron-specific nestin promoter (Komatsu et al. 2006), and it is unclear when this promoter was activated and hence when Atg7 was deleted in the neural stem and progenitor cells. Dying stem and progenitor cells display primarily signs of apoptosis (Fukuda et al. 2004) and it remains to determine to what extent autophagy is involved in this type of genotoxic cell death, which is very different from the type of neuronal cell death observed after HI.
Furthermore, a recent study suggests that Atg7 deletion merely delayed neural progenitor/neuronal death, since tissue "loss" (lack of growth) and levels of neurogenesis-related markers were similar to WT brains 5 days after irradiation, on postnatal day 15 (Wang et al. 2019). However, prior to irradiation, Atg7-deficient brains displayed a regenerative state, as judged by the higher levels of transcripts related to neurogenesis and oligodendrocyte precursors, indicating that loss of Atg7 in neurons is stressful to the brain tissue, a foreboding of the neurological symptoms that inevitably will develop several weeks later. Importantly, 5 days after irradiation, myelin development was less impaired in the Atg7-deficient animals than in WT mice (Wang et al. 2019). In summary, this relatively recent publication suggests that autophagy inhibition in neurons has indirect protective effects also in other cell types. Autophagy may thus be considered a target for preventing brain injury triggered by irradiation.
Together, these findings point to the importance of elucidating in further detail the mechanisms through which Atg7 deletion renders neurons resistant to ischaemic injury. It also remains to investigate if autosis is an important mechanism of cell death in all types of neurons in the brain, and if it is less important in the adult brain. The role of autophagy in the demise of neural stem and progenitor cells in the neurogenic niches and oligodendrocyte progenitor cells after genotoxic stress through ionizing radiation also requires further investigation, since these cells typically display signs of classic apoptosis rather than the mixed morphologies observed in neurons dying after HI. Prolonged overall inhibition of autophagy, though, would not be a fruitful strategy, since autophagy is of fundamental importance in the homeostasis of the developing brain. This is evidenced by the reduced lifespan of neuronally Atg7-deficient mice; their brains degenerate and they usually die at 6-10 weeks of age (Komatsu et al. 2006). In addition to direct inhibition of cell death-related mechanisms, the lack of autophagic capacity, even if restricted to one cell type, may cause tissue stress, which in turn elicits protective, regenerative responses. In summary, transient and specific autophagy inhibition strategies should be considered to improve clinical outcome after perinatal asphyxia and HIE, and possibly also after radiotherapy-induced brain damage.

III.
Modulation of the selective, chaperone assisted type of autophagy to protect against neuronal death.

Accepted Article
This article is protected by copyright. All rights reserved As mentioned above, the term autophagy covers several types of degradative processes, but usually refers to macroautophagy, a pathway generally characterized by high capacity, but lower specificity in the removal of misfolded proteins compared to proteasomal degradation, and also capable of clearing aggresomes or aggregates (aggrephagy) (Cuervo 2011).

The CASA complex.
Macroautophagy involves the formation of autophagosomes engulfing organelles/aggregate, and their subsequent fusion to lysosomes for content degradation (Klionsky et al. 2016). However, autophagosomes can also receive cargoes of aberrant damaged, oxidize or misfolded proteins utilizing more selective pathways. Aberrant proteins that must be cleared from cells can be selectively delivered by specific chaperones (Arndt et al. 2010). These aberrant proteins, particularly in their misfolded forms are potentially neurotoxic and chaperones facilitate their degradation in their monomeric/oligomeric form, generally (and possibly) prior to their aggregation (Arndt et al. 2010). This type of macroautophagy assisted by specific chaperones has been named chaperone-assisted selective autophagy (CASA) (Arndt et al. 2010, Cuervo 2011. CASA differs from other forms of autophagy, like chaperone-mediated autophagy (CMA). In CASA misfolded proteins are delivered to autophagosomes, which then fuse with lysosomes, while in CMA a specific set of chaperones deliver misfolded proteins directly to lysosomes, thus escaping the autophagosome pathway. CASA takes place in muscle and brain, as well as in cancer cells (Crippa et al. 2010b, Crippa et al. 2010a, Piccolella et al. 2017, Arndt et al. 2010, Cuervo 2011) and CASA involves a peculiar chaperone of the family of the small HSPs, known as HSPB8 . In this pathway, HSPB8 dimerizes and form a stable complex with its cochaperone BAG3, a nucleotide exchange factor (NEF) for HSP70s. Of note, mutation found in HSPB8 or in BAG3 have been associated with motor neuronal or neuromuscular diseases (Fontaine et al. 2006, Irobi et al. 2010, Ghaoui et al. 2016, Adriaenssens et al. 2017, Bouhy et al. 2018, Guilbert et al. 2018, Fang et al. 2017, Konersman et al. 2015, Selcen et al. 2009), suggesting they play an essential role in the protein quality control system (PQC) in these two tissues. Together, HSPB8 and BAG3 specifically interact with the ATP-dependent chaperone HSP70 conjugated to the E3 ubiquitin ligase carboxyl terminus of Hsc70interacting protein (CHIP). HSPB8 is the limiting factor of the complex and recognizes misfolded protein, while BAG3, via its PxxP, allows the interaction with dynein (and the protein 14-3-3) for delivery of the entire complex to the microtubule organizing center (MTOC) where autophagosomes are assembled.
Because of its function, the complex HSPB8-BAG3-HSP70-CHIP has been named CASA complex (Arndt et al. 2010). The CASA complex component CHIP ubiquittes the misfolded protein cargo for the autophagosomal receptor (SQSTM1/p62) recognition and insertion into autophagosomes ).

Accepted Article
This article is protected by copyright. All rights reserved The role of HSPB8 in the CASA complex and in protective activity in NDs.
Indeed, HSPB8 restores a normal autophagy flux which is found to be blocked in several NDs. Conversely, in most cases, HSPB8 down-regulation resulted in increased accumulation of these mutant proteins (SOD1, TDP-43, dipeptide repeats coded by C9ORF72) supporting its role in the clearance of misfolded proteins (Crippa et al. 2010b, Crippa et al. 2016a, Rusmini et al. 2013). HspB8 also efficiently removes misfolded proteins containing PolyQ tracts, such as polyQ-AR (Carra et al. 2005, Carra et al. 2008a, Carra et al. 2009), or those involved in Alzheimer's and PD (Wilhelmus et al. 2006, Bruinsma et al. 2011, Seidel et al. 2011. In most neurodegenerative diseases with alterations of the autophagic pathway the pro-autophagic role of HSPB8 mediated by the enhanced activity of the CASA complex appears to be crucial to prevent misfolded protein accumulation in affected neurons (or surrounding cells).

HSPB8, with BAG3 in the CASA complex mediate the interplay between autophagy and proteasome systems.
It is of note that the CASA complex also serves as routing system of misfolded proteins to autophagy to prevent possible proteasome overwhelming and /or impairment ( Figure 2). When dynein mediated retrograde transport is blocked, a condition present in a number of NDs (Sau et al. 2011) (and that we reproduced genetically using siRNAs against dynein or pharmacologically using the selective dynein inhibitor erythro-9-[3-(2-hydroxynonyl)]adenine (EHNA)), the complex HSP70-CHIP cannot bind the HSPB8-BAG3 complex, and the CASA activity is perturbed and cannot dispose of misfolded proteins via autophagy . Under these conditions, specific transcription factors (still unknown)

Accepted Article
This article is protected by copyright. All rights reserved activate the de novo transcription of another nucleotide exchange factor (NEF/BAG) called BAG1. Also BAG1 is capable of selectively binding the complex HSP70/CHIP; but in this case, misfolded proteins are routed to the proteasome instead of autophagy (Behl 2016, Gamerdinger et al. 2011. Indeed, BAG1 overexpression facilitates proteasomal removal of ARpolyQ, mutant SOD1 and mutant TDP-43 , and this BAG1 activity is fully counteracted by the inhibition of the proteasome activity, but not by inhibition of autophagy . Since, the inhibition of the proteasome also results in the upreguation of both HSPB8 and of BAG3 (Crippa et al. 2010b), while blockage of the delivery of misfolded protein to autophagy results in upregulation of BAG1 , this gives rise to a very nice equilibrium between the proteasome and autophagy selective degradation. The equilibrium is maintained by the ratio BAG3:BAG1, which determines the relative amount of the proteins BAG3 (associated with HSPB8) and BAG1 capable of associating with the dimer HSP70/CHIP and of selecting the proper degradative pathway for a misfolded protein in a given neuron.

HSPB8 protects against misfolded protein toxicity in animal models of ALS.
Few animal model studies have been performed to study the effect of HSPB8 overexpression in NDs.
However On these bases, a large screen aimed to identify commercially available drugs that are able to enhance HSPB8 expression in motor neuronal cells for therapeutic purposes (Crippa et al. 2016b) led to the identification of the drug colchicine as a potent inducer of HSPB8 together with a series of other genes involved in autophagy activation (transcription factor EB, p62, LC3) suggesting that colchicine could represent a useful candidate to be tested in misfolded protein associated NDs (Crippa et al. 2016b). Of note, estrogens are also potent inducers of HSPB8 expression (Piccolella et al. 2017), confirming previous

Accepted Article
This article is protected by copyright. All rights reserved published data (Sun et al. 2007), and this might in part explain some of the gender differences described in misfolded protein associated NDs neurodegenerative diseases, including ALS.

Parkinson's disease-linked proteins promote mitophagy at the endoplasmic reticulum-mitochondria interface.
Mitophagy is a relatively recent term describing the selective autophagy of mitochondria under specific stress conditions, including during apoptosis induction in the presence of caspase inhibitors, following nutrient deprivation or toxin-induced mitochondrial damage (Tolkovsky et al. 2002, Elmore et al. 2001, Kissova et al. 2004, Lemasters 2005, or during specific developmental programs, such as metabolic reshuffling in the developing heart (Gong et al. 2015) or red blood cell maturation (Sandoval et al. 2008, Schweers et al. 2007. Mitophagy became relevant to neurodegeneration when the mitochondrial serine/threonine kinase PINK1 and the ubiquitin-protein ligase Parkin, the products of two genes (PINK1 and PARK2) mutated in autosomal recessive forms of PD, were found to jointly promote the clearance of depolarized mitochondria in cell lines treated with the protonophore carbonyl cyanide m-chlorophenyl hydrazone (CCCP) (Narendra et al. 2008, Geisler et al. 2010, Matsuda et al. 2010.
We now know that a number of stress-related stimuli activate this program, including specific inhibitors of the respiratory chain, unfolded protein stress in the mitochondrial matrix, and impairment of the These events are probably concomitant to the separation of the damaged mitochondrion from the rest of the

Accepted Article
This article is protected by copyright. All rights reserved network, operated by the dynamin-related GTPase Drp1 and regulated by the ER-mitochondria interface (Twig et al. 2008, Tanaka et al. 2010, Buhlman et al. 2014, Friedman et al. 2011. Drp1 is cooperatively recruited by PINK1 and Parkin on depolarized mitochondria, suggesting that mitochondrial fission and mitophagy are orchestrated at the ER-mitochondria interface (Buhlman et al. 2014, Erpapazoglou & Corti 2015. patients (Tieu et al. 2004). HSD17B10 was co-recruited with Parkin near the TOM complex and PINK1, suggesting regulation of its mitochondrial import by a PINK1/Parkin-dependent mechanism (Bertolin et al. 2015). Using an original biosensor for exploring the presequence-mediated protein import pathway in living cells, we recently showed that this process is more generally facilitated by the PINK1/Parkin system (Jacoupy et al. 2019).

PINK1 and
Dissecting the various mechanisms by which the PINK1/Parkin sensor/effector system maintains mitochondrial quality has highlighted the central role played by mitochondrial protein import in monitoring the mitochondrial functional state. There is a remarkable parallel between the characteristics of the PINK1/Parkin system and a key regulatory component of the mitochondrial unfolded protein response (UPR mt ) in the small invertebrate organism C. elegans, Activated Transcription Factor Associated with Stress 1 (ATFS-1). ATFS-1 is constitutively imported into functional mitochondria by an N-terminal mitochondrial targeting signal, but under mitochondrial stress it translocates to the nucleus to activate a protective transcriptional response mediated by hundreds of genes with various functions, including coping with the accumulation of unfolded proteins in mitochondria, fighting against an excess of reactive oxygen species, and remodelling of cellular metabolism (Nargund et al. 2012). By analogy, it is likely that PINK1

Accepted Article
This article is protected by copyright. All rights reserved and Parkin are more broadly involved in the modulation of the mitochondrial stress response in mammalian cells, considering the well-known cell-protective properties of Parkin (Imai et al. 2000, Darios et al. 2003, Johnson et al. 2012, Bouman et al. 2011, and the fact the PINK1/Parkin system is induced during the UPR mt (Jin & Youle 2013). Establishing the consequence of PINK1 and Parkin dysfunction on the mitochondrial stress response, and investigating the possibility that these proteins coactivate transcription factors near dysfunctional mitochondria, are major challenges for future research. Activating Transcription Factor 5 (ATF5) may mediate the UPR mt response in mammals, but whether this transcription factor interacts with the PINK1/Parkin system is unknown (Fiorese et al. 2016). Notably, PINK1 and Parkin contribute to the activation of transcription factors of the microphtalmia/transcription factor E (MiT/TFE) family during mitophagy, including the master regulator of lysosomal biogenesis, transcription factor EB (TFEB) (Nezich et al. 2015). Morover Parkin acts as transcription factor itself, although the precise mechanisms regulating its nuclear translocation and transcriptional activity remain to be determined (da

Fluorescent reporters with pH-sensitive components illuminate PINK1/Parkin-dependent mitophagy in cultured neurons and in vivo.
We still lack sufficient evidence for a role of PINK1/Parkin-dependent mitophagy in mitochondrial maintenance and neuronal survival in vivo and in the context of PD. The question of whether mitophagy is relevant to neurons has been intensively debated (Grenier et al. 2013, Van Laar et al. 2011. Studies based on the simultaneous expression in primary neurons of fluorescent reporters targeted to the mitochondrial compartment, the autophagosome and the lysosome have, however, provided evidence in neurons for events of engulfment of mitochondria into autophagosomes, or of fusion of autophagosomes containing mitochondria with lysosomes (Cai et al. 2012, Ashrafi et al. 2014, Hsieh et al. 2016. These events were observed to a lesser degree in neurons from Parkin-and PINK1-deficient mice, supporting their relation to PINK1/Parkin-dependent mitophagy (Ashrafi et al. 2014). More convincingly, mitophagy has been investigated in cultured primary cells, including neurons and in vivo, in Drosophila and mice, using ratiometric or dual-fluorescence reporters with pH-sensitive components for direct tracking of mitochondria in lysosomes (Bingol et al. 2014, Bonello et al. 2019, McWilliams et al. 2016, McWilliams et al. 2018, Lee et al. 2018, Cornelissen et al. 2018, Shin et al., 2019. These studies have demonstrated the occurrence of mitophagy under physiological conditions in neurons, including the highly vulnerable dopaminergic (DA) neuron in the mouse (McWilliams et al. 2018) and fly brains (Lee et al. 2018, Cornelissen et al. 2018. However, it remains controversial whether PINK1 and Parkin coregulate this type of physiological mitophagy detected in the absence of exogenous insults, particularly during aging (Cornelissen et al. 2018, Lee et al. 2018, McWilliams et al. 2018, which is recognized today as the major risk factor for PD (Collier et al. 2017). Notably, a previous study reported strong immunostaining for

Accepted Article
This article is protected by copyright. All rights reserved phosphorylated ubiquitin partially colocalizing with mitochondria and lysosomes in the aged and PDaffected human brain, but it remains to be clarified whether this reflects impaired or increased mitophagy (Fiesel et al., 2015).

Beyond neurons: a role in innate immunity.
Finally, there is mounting evidence that PINK1 and Parkin jointly regulate immune-related mechanisms, which is of particular interest, considering the neuroinflammatory component in PD (Greene et al. 2005, Manzanillo et al. 2013, Matheoud et al. 2016, Torres-Odio et al. 2017, Sun et al. 2018, Kim et al. 2014, Kang et al. 2016, Zhong et al. 2016, Sumpter et al. 2016, Sliter et al. 2018. In some cases, these regulatory roles of PINK1 and Parkin have been clearly linked to modulation of mitophagy (Kim et al. 2014, Zhong et al. 2016, Sumpter et al. 2016. Specifically, PINK1/Parkindependent mitophagy has been recognized to counterbalance the action of the nucleotide-binding domainlike receptor protein 3 (NLRP3) inflammasome, a proinflammatory multiprotein complex activated in response to pathogens as well as endogenous signals associated with cell and tissue damage (Sumpter et al. 2016, Zhong et al. 2016). This complex plays a central role in a number of inflammatory/autoimmune disorders, and is emerging as key player in neurodegenerative diseases, including PD (Labzin et al. 2018).
Specifically, the neurotransmitter DA has been shown to attenuate NLRP3 inflammasome signalling in immune cells, whereas mitochondrial toxins, such as MPTP or rotenone, exacerbate NLRP3-dependent inflammation, leading to DA neurodegeneration in cell and animal models (Yan et al. 2015, Sarkar et al. 2017. We have recently reported exacerbation of the NLRP3 inflammasome pathway in bone-marrow-

V. Focus on autophagy as therapeutic target in brain pathologies.
With our improved understanding of the regulation of autophagy, its diverse regulatory mechanisms and a greater awareness that it represents a unique target with multiple therapeutic options for the clearance of toxic aggregates and damaged organelles, there has been an extraordinary rise of interest in its clinical applicability and options for translation (e.g. (Feng et al. 2017, Tramutola et al. 2017) (Lipinski et al. 2015, Maiese 2016, Boland et al. 2018. Indeed these reviews are pertinent to topics as different in pathology as Alzheimer's disease and traumatic brain injury, reflecting the wide applicability of such therapeutic options. A seminal publication that initiated cognisance of this area, indeed a veritable tour de force, overviewed multiple classes of drugs and potential molecular targets regulating autophagic mechanisms (Rubinsztein et al. 2007).

Accepted Article
This article is protected by copyright. All rights reserved We became aware of the FDA approved drug rilmenidine, considered to act via a mTOR-independent mechanism, after it was reported to have beneficial actions in a mouse model of Huntington's disease (Rose et al. 2010). At the time we were working on 3,4-methylene-dioxymethamphetamine (ecstasy) induced injury of serotonin (5-HT) neurons, and (i) since autophagy was considered to be involved in the brain's response to amphetamines (Galluzzi et al. 2016), and (ii) rilmenidine binding sites exist in 5-HT raphe nuclei (King et al. 1995), the inclusion of rilmenidine in our study was an obvious step forward. Indeed ecstasy-induced neurotoxicity of 5-HT neurons involved an autophagic component and rilmenidine completely blocked the toxicity of ecstasy (Mercer et al. 2017). Thus this FDA-approved molecule, subsequently documented to be well-tolerated and suitable for human usage (Underwood et al. 2017), likely also possesses a profile suitable for use in drug abuse. Success here led to our further studies exploring the potential of rilmenidine in cellular and animal models of MND. We found beneficial effects on rilmenidine in a MN cell line and in MNs derived from human stem cells expressing mutant superoxide dismutase 1 (SOD1): there was an increase in autophagic flux following treatment, as assessed by monitoring LC3 lipidation levels and abundance of autolysosomes, using a tandem mCherry-GFP-LC3B reporter construct, accompanied by a 20% reduction in the load of mutant SOD1 (Perera et al. 2018).
However, we were disappointed at its lack of effectiveness in the SOD1 G93A mouse, where it actually worsened the pathological and functional outcome (Perera et al. 2018). What our in vivo study does highlight is that rilmenidine was certainly efficacious enhancing autophagic flux, but that there was a concurrent accumulation of misfolded SOD1, and mitochondrial depletion with autophagosome accumulation in MNs. In ongoing work in a transgenic mouse model overexpressing the normal (WT) and ALS-linked Q331K variants of the human TDP-43 protein in neurons (TDP-43 WTxQ331K mice) (Mitchell et al., 2015), we have found similar findings with rilmenidine with reduced TDP-43 immunoreactivity in MNs ( Supplementary Fig. 1), so moving forward it will be interesting to determine whether the enhanced auto-/mitophagic flux found with rilmenidine (Perera et al. 2017) will prove useful in other TDP-43 models. Given the positive outcomes here rilmenidine may prove effective in less aggressive forms of MND and certainly a key issue is that the timepoint of drug intervention may well be critical relative to toxic aggregate load and stage of disease progression. Indeed, our evidence for activation of autophagy in TARDBP in cell culture models, when taken the finding that rapamycin produces benefits in a mouse TARDBP model (Wang et al. 2012), strongly suggests that autophagy is likely to be a valid target in MND.
Although toxic aggregates and dysfunctional proteostasis (Yerbury et al. 2016) have been linked to diverse proteins in MND, emergent reports increasingly suggest that successful management may be possible by targeting autophagic signalling (Deng et al. 2017). For example, recent elegant work screening clinically approved drugs revealed a number of small molecules with beneficial actions using FUS-expressing stem cells and a Drosophila model (Marrone et al. 2018).

Accepted Article
In addition to the current great interest in potentially clinically effective small molecules, numerous strategies, including gene therapy, RNA interference or antisense oligonucleotides, continue to attract attention focusing on for example mTOR-dependent and-independent pathways, and broadly on signalling linked to Beclin1, Atg5, AMP-dependent protein kinase, PICALM and histone deacetylases , Hu et al. 2017, Menzies et al. 2017, Rahman & Rhim 2017. Although at the present time subsets of such published data are relatively small as they relate to neurodegeneration and human brain pathologies, there continue to be rapid developments in this area with new advances in our understanding of the regulation and dynamics of autophagic flux, and its interface with the ubiquitin proteasome, as highlighted by the preceding sections.

Conclusions and perspectives.
The examples discussed in this review are representative of a growing body of literature showing how autophagy has not only physiological roles in neurodevelopment and the maintenance of neuronal homeostasis, but also different levels of involvement in pathological processes affecting the brain, including neurodegenerative disorders and stroke. As we have seen here, a set of observations involve dysfunction of selective types of autophagy in specific neurodegenerative disease. For example, defective CASA has been linked to motor neuron disorders such as ALS and SBMA and mitophagy is suspected to be impaired in specific forms of Parkinson's disease, as well as in other neurodegenerative diseases, according to studies not discussed in this review (Wong & Holzbaur 2014, Cai & Jeong 2020. In other pathological conditions of the brain, autophagy appears to be uncontrollably activated and to cause neuronal death, as in the case of perinatal asphyxia. From an assessment of the current state of knowledge, it is clear that we are still some distance from a holistic understanding of how autophagy is recruited in different brain diseases and in which cases it is merely associated with cell death rather than mediating or even causing cell death (Klionsky et al. 2016). Thus, we still need appropriate in vivo documentation of the various forms of autophagy in different disease contexts and across different disease stages, as well as their relationship to different forms of cell death to fully appreciate whether and when they play beneficial or detrimental roles. Findings presented herein argue that steps to take advantage of the beneficial modulation of autophagy by pharmacological strategies, and the specific modalities of such strategies, are likely to be quite different depending on the age of the affected individual, the type of injury/disease, the specific stage of progression, as well as the overall duration and severity of the disease. For instance, it has already been reported in the SOD1 G93A mouse model of MND that short-term treatment with clemastine is beneficial, but long-term treatment fails to alleviate disease (Apolloni et al. 2016). As an additional take-home message, we need to better understand

Accepted Article
This article is protected by copyright. All rights reserved the complex interplay between autophagy and other intracellular processes that may also at some point be recruited by the pathological process, either as contributing factors or protective responses. As illustrated with PINK1 and Parkin, many proteins involved in autophagy and its regulation have multiple functions in the cell, and these have to be taken into consideration if we want to explore meaningful therapeutic avenues for neurological diseases. The case of CASA, on the other hand, illustrates the interplay between autophagy and the proteasome system, leaving room for envisaging strategies aimed at promoting alternative degradation routes when autophagy is impaired. Finally, while autophagy and mitophagy are emerging as central modulators of inflammatory and immune-related mechanisms, which are recognized as a component of the pathological process in various brain diseases, work discussed here also shows that the selective modulation of autophagy in neurons has broader impact on brain tissue homeostasis. Therefore, a better knowledge of the different cell types affected by the disease process and their complex interactions will provide invaluable insight into the most appropriate ways of using autophagy modulation for therapeutic purposes in the brain. a statement in the "Conflict of interest disclosure" section: "ARRIVE guidelines were not followed for the following reason: does not apply " (edit phrasing to form a complete sentence as necessary).
=> if Yes, insert in the "Conflict of interest disclosure" section: "All experiments were conducted in compliance with the ARRIVE guidelines." unless it is a Review or Editorial

Accepted Article
This article is protected by copyright. All rights reserved

Accepted Article
This article is protected by copyright. All rights reserved Medical Art by Servier is licensed under a Creative Commons Attribution 3.0 Unported License. https://creativecommons.org/licenses/by/3.0/.

Figure 2
The HSC70/CHIP-mediated routing system between proteasome and autophagy. BAG1 is a nucleotide exchange factor which is capable to complex the chaperone HSC70 and E3 ubiquitine ligase CHIP. This complex recognize monomeric misfolded proteins to be driven to proteasome degradation.
Alternatively, misfolded monomeric, oligomeric and/or aggregated misfolded proteins can be recognized by another nucleotide exchange factor BAG3 associated to the small HSP named HSPB8. Once the complex HSPB8/BAG3 recognizes target misfolded proteins, the entire complex can associate to the HSC70/CHIP complex, thus sequestering it from the BAG1 binding. The HSPB8/BAG3/HSC70/CHIP complex has been named the CASA complex, since it mediates a peculiar form of autophagy known as "Chaperone-Assisted Selective Autophagy (CASA)". In this case misfolded aggregating proteins, bound to the CASA complex, are routed by a dynein-dependent transport to the microtubule organization center (MTOC) where aggresomes are formed and inserted into nascent autophagosomes. The HSC70/CHIPmediated routing system thus generates a delicate equilibrium which is responsible for the choice of two alternative pathways for damaged/misfolded protein degradation. Of note, the blockage of the dyneinmediated transport of the CASA complex correlates which enhanced BAG1 expression. Conversely, proteasome inhibition results in HSPB8 and BAG3 overexpression. This transcriptionally regulated expression of the routing system factors maintains the fine tuned equilibrium between proteasome and autophagy degradation in cells.

Figure 3
Mechanisms by which the PINK1/Parkin system keeps mitochondria functional. In addition to regulating the degradation of damaged mitochondrial components and whole organelles through the MDV and mitophagy pathways, the PINK1/Parkin system promotes mitochondrial biogenesis by at least two distinct mechanisms: the proteasomal degradation of PARIS (Parkin Interacting Substrate/ZNF746), a trascriptional repressor of the master regulator of mitochondrial biogenesis PGC-1α (peroxisomeproliferator-activated receptor γ coactivator-1α); and the displacement of translational repressors, including the Parkin substrate hnRNP-F, from transcripts encoding certain nuclear-encoded respiratory chain components, anchored on the outer mitochondrial membrane in a PINK1-dependent manner. During mitophagy, PINK1 and Parkin also promote the nuclear translocation and activation of transcription factors of the MiT/TFE family, leading to lysosomal biogenesis and facilitation of mitophagy. The mechanism underlying the PINK1/Parkin-dependent activation of these transcription factors remain to be determined.

Accepted Article
This

Disposal of cell corpses
• Phagocytic uptake and lyososmal degradation

Caspase inhibitors •
Genetic ablation of pro-apoptotic factors