PTEN-induced kinase 1 (PINK1) and Parkin: Unlocking a mitochondrial quality control pathway linked to Parkinson's disease

Dissection of the function of two Parkinson's disease-linked genes encoding the protein kinase, PTEN-induced kinase 1 (PINK1) and ubiquitin E3 ligase, Parkin, has illuminated a highly conserved mitochondrial quality control pathway found in nearly every cell type including neurons. Mitochondrial damage-induced activation of PINK1 stimulates phosphorylation-dependent activation of Parkin and ubiquitin-dependent elimination of mitochondria by autophagy (mitophagy). Structural, cell biological and neuronal studies are unravelling the key steps of PINK1/Parkin-dependent mitophagy and uncovering new insights into how the pathway is regulated. The emerging role for aberrant immune activation as a driver of dopaminergic neuron degeneration after loss of PINK1 and Parkin poses new exciting questions on cell-autonomous and noncell-autonomous mechanisms of PINK1/Parkin signalling in vivo.


Introduction: PTEN-induced kinase 1/Parkin links to Parkinson's disease pathogenesis
Parkinson's disease (PD) has emerged as the fastestgrowing neurodegenerative disorder worldwide characterised by motor symptoms including tremor and nonmotor symptoms such as cognitive decline [1,2]. Pathological studies of postmortem brain tissue of patients with PD reveal characteristic loss of dopamine neurons from the substantia nigra pars compacta region in the brain and accumulation of a-synuclein positive inclusions known as Lewy bodies (LBs) [1]. The role of mitochondrial dysfunction in PD emerged in the 1980s with the discovery that mitochondrial toxins, such as 1methyl-4-phenyl-1,2,3,6-tetrahydropyridine, recapitulate selective nigral degeneration in humans and animals, and biochemical analysis of PD brain tissue revealed defects in mitochondrial respiratory chain enzyme complexes [3]. However, the molecular basis of how mitochondrial dysfunction leads to PD remained unknown till pioneering genetic studies that identified approximately 20 genes and loci (designated PARK) linked to monogenic forms of PD [4]. Autosomal recessive-inherited causal mutations in human PTEN-induced kinase 1 (PINK1) and the RING-IBR-RING E3-ubiquitin (Ub) ligase Parkin (encoded by PARK6 and PARK2 genes, respectively) were discovered in patients with earlyonset PD [5,6] (Figure 1). Molecular analysis demonstrated these proteins function together in a common mitochondrial signalling pathway in which PINK1 phosphorylates Ub (phospho-ubiquitin) and Parkin. This triggers Parkin activation and Ubiquitin-dependent removal of damaged mitochondria by autophagy (mitophagy). This model correlates very well with clinical data in which PINK1 and patients with Parkin PD have indistinguishable phenotypes with early-onset and sustained responsiveness to L-Dopa therapy and similar pathology with general paucity of LBs [7].
Parallel studies of postmortem brain tissue have demonstrated elevated levels of phospho-ubiquitin in aged and sporadic PD cases compared with controls [8,9]. It has also been reported that Parkin undergoes oxidative modifications that render it more insoluble and that it can sequestrate in LBs, leading to reduced availability of soluble Parkin for native functions including as a redox molecule [10,11]. A specific role for PINK1 and Parkin in a-synuclein-driven PD is also suggested from preclinical models in which a-synucleininduced mitochondrial pathologies and dopamine neurodegeneration are significantly worsened in either PINK1 or Parkin knockout mice [12,13].
Structural view of PTEN-induced kinase 1 and Parkin and impact of Parkinson's disease-causing mutations Human PINK1 encodes a 581 amino acid (AA) serine (Ser)/threonine (Thr) protein kinase containing an Nterminal mitochondrial targeting sequence, a catalytic kinase domain with three insertions and a C-terminal hydrophobic extension (Figure 1). Most PD-causing mutations lie within the kinase domain highlighting the importance of this region in its protective role against PD (Figure 1 within the ATP-binding pocket (Ala217Asp, Glu240Lys, Ala244Gly and Leu369Pro) and within the activation loop (Gly386Ala, Pro416Arg/Leu and Glu417Gly) both resulted in catalytically inactive PINK1, and mutation within Gly309Asp that lies in INS3 prevented Ub recognition with the remainder of PD mutations being likely to affect the structural integrity of PINK1 ( Figure 1) [14,15]. The current insect structures do not provide insights into the role of the N-terminus of PINK1 as well as the mechanism by which PINK1 undergoes dimerisation on activation in cells [18], and understanding these in the structural context of mammalian PINK1 will be essential.
Parkin encodes a 465 AA Ub E3 ligase containing an Nterminal Ub-like (Ubl) domain, RING0, RING1, inbetween-RING (IBR), repressor element and RING2 domains. The three RING and IBR domains each coordinate two zinc ions that are essential for protein folding. A previously uncharacterised conserved linker region between the Ubl domain and RING0 termed the activating element (ACT) (AA residues 101e109) has also been identified [19], and it has recently been reported that phosphorylation of a conserved Ser 108 residue within the ACT element by Unc-51 autophagyactivating kinase 1(ULK1) is required for optimal activation of Parkin in cells, and this presages PINK1mediated phosphorylation of Ser65 within the Ubl domain ( Figure 1) [20]. Other key functional residues of Parkin include the catalytic cysteine residue (Cys431) that lies within the RING2 domain [21,22] and His302 on the RING1 domain that has been identified as an essential phospho-ubiquitin-binding residue [23e25]. Numerous PD-causing mutations have been identified including rearrangements that drastically destabilise the protein and missense mutations across the whole protein that disrupt critical regulatory AA residues, for example [Cys431Phe] within RING2 that disrupts the Ub thioester formation required for Ub transfer to substrates [21,22], Ser65Asn that prevents phosphorylation and activation by PINK1 [26], and Lys161Asn Lys211Asn mutants that disrupt the phosphoacceptor pocket on RING0 for pSer65-Ubl thereby preventing release of RING2 and Parkin activation and associated with defective recruitment to damaged mitochondria in cells [19,27].
Upstream mechanism of stress-evoked mitophagy: regulation of Parkin activation by PTEN-induced kinase 1 Under basal conditions, PINK1 is present at low levels owing to proteolytic turnover, and Parkin is localised in the cytosol where it exists in an autoinhibited conformation mediated by three autoinhibitory interfaces [28]. On mitochondrial depolarisation, induced by uncoupling agents (e.g. carbonyl cyanide m- chrolophenyl hydrazone/oligomycin-antimycin), PINK1 is stabilised at the outer mitochondrial membrane (OMM) in association with components of the translocase of the outer membrane complex where it undergoes dimerisation, autophosphorylation and activation ( Figure 2)  A recent study has found that mitochondrial depolarisation-dependent PINK1-Parkin activation also converges with the adenosine 5-monophosphate (AMP)activated protein kinase (AMPK) signalling pathway [20]. AMPK is rapidly activated after any stress that triggers mitochondrial damage depleting ATP levels and concomitantly elevating AMP. In response to mitochondrial depolarisation, AMPK is fully activated in the cytosol within 2 min and directly phosphorylates its substrates Raptor, acetyl-CoA carboxylase (ACC) and ULK1 [20]. Strikingly, Parkin is phosphorylated by ULK1 at Ser108 within the ACT element at this early time point of minutes followed by rapid recruitment of ULK1 and AMPK to the mitochondria [20]. This presages PINK1-dependent phosphorylation of Parkin and Ub whose temporal dynamics may be slowed by the requirement of PINK1 stabilisation on the OMM (Figure 2) [20].  [42]. In contrast, in physiologically more relevant cell types that express endogenous levels of Parkin including human embryonic stem cell (ESC)-derived dopamine neurons, human iNeurons and mouse primary cortical neurons, K63 chains were exclusively or predominantly upregulated after mitochondrial depolarisation [39,43]. On PINK1/Parkin activation, OPTN1 recruitment to mitochondria is associated with activation of TANK-binding kinase 1 (TBK1) that phosphorylates OPTN1 to increase its affinity for Ub chain binding that promotes further OPTN1 recruitment to damaged mitochondria and TBK1 activation and also enhances binding to microtubule-associated protein 1A/1B-light chain 3A (LC3 A) to drive mitophagy [37,44]. TBK1 also phosphorylates ras-associated binding (Rab)7 at Ser72 that has been reported to be critical for mitophagy via recruitment of autophagy-related protein 9A-positive vesicles to damaged mitochondria [37]. ULK1, double FYVEcontaining protein 1 and WD repeat phosphoinositideinteracting protein 1 stimulate LC3 recruitment and subsequent mitoautophagosome formation. Furthermore, it has been reported that NDP52 and TBK1 jointly recruit a second wave of the ULK1 complex to ubiquitylated cargo at the mitochondria, leading to ULK1 kinase activation, which occurs independently of energy-sensing pathways (AMPK and mechanistic target of rapamycin kinase) and Parkin Ser108 phosphorylation [45]. Consistent with this, reconstitution studies have recently demonstrated that cargo-loaded NDP52, OPTN and TAX1BP1 stimulate LC3 lipidation in the presence of purified autophagy initiation complexes expressed at appropriate physiological (nM) concentrations and furthermore that NDP52 and TAX1BP1 but not OPTN promote ULK1 complex recruitment to the membrane [46]. Recently, it has been proposed that Parkin generates short chains including branched Ub species in both HeLa cells and human iNeurons [42,47], and in future studies, it will be interesting to directly assess the role of specific Ub chain linkages (e.g. K6 vs K63) as well as chain length in a reconstitution system.

PTEN-induced kinase 1/parkin mitophagic signalling in neurons
Initial mass spectrometry studies in HeLa cells overexpressing Parkin not also identified myriad substrates, mainly localised to the mitochondria, but also included substrates in the cytosol and other cell compartments [41]. Recent comparative analysis of Lys-e-Gly-Gly (di-GLY) proteomic data sets of human iNeurons and primary mouse cortical neurons in which PINK1/Parkin is expressed under endogenous levels has revealed a common set of 49 diGLY Ub sites spanning 22 OMM proteins [43,47]. This represents a defined Parkindependent ubiquitylation signature of damaged mitochondria, and in future studies, it will be interesting to assess this in Dopaminergic (DAergic) neurons as well as non-central nervous system (CNS) cell types. In human iNeurons, it has been demonstrated that increased Parkin-mediated ubiquitylation of OMM substrates is associated with mitochondrial stressevoked mitophagy that is absolutely PINK1 dependent. This is distinct from basal levels of mitophagy detected in iNeurons that are independent of PINK1 [47]. Basal mitophagy is also independent of PINK1 and Parkin in knockout models in vivo [26,48], although these mice lines are reported to exhibit no robust phenotype or only mild phenotypes without striatonigral degeneration, for example, Parkin knockout mice [49,50]. Analysis of Parkin-dependent mitophagy in mouse hippocampal neurons suggests a high degree of spatiotemporal regulation with Parkin/OPTN and TBK1 recruitment occurring in the cell body and rarely in the axon, suggesting alternate pathways may regulate mitophagy in axons [36]. Complementary in vivo analysis of Parkin knockout mice crossed with the 'mutator' mouse model, that induces mitochondrial stress owing to mutation in Polg, led to age-dependent DA neuron loss and levodopa-responsive motor dysfunction [51]. This confirms that endogenous Parkin has a critical role in protecting nigral neurons against mitochondrial dysfunction/accumulation of mitochondrial DNA (mtDNA) mutations in vivo consistent with its functions on neurons in vitro.

PTEN-induced kinase 1 and parkin mechanisms of immunology linked to neurodegeneration
Innate immune signalling Akin to bacterial components, escape of mtDNA to the cytosol can trigger innate immune signalling pathways that trigger toll-like receptor 9-mediated type I interferon (IFN) or nuclear factor kappa-light-chainenhancer of activated B cell inflammatory responses [52] or activation of the NLR family pyrin domaincontaining 3 inflammasomes [53]. In a similar manner, the combination of mitochondrial stress and failure to remove damaged mitochondria by loss of PINK1/Parkin has also been shown to stimulate innate immune responses via activation of cyclic GMP-AMP synthase/ stimulator of interferon gene (STING)-mediated type 1 IFN response (Figure 3) [54]. PINK1 or Parkin knockout mice crossed with mutator mice or exposed to exhaustive exercise exhibited high levels of cytokines including interleukin (IL)-6, IL-12, and IFN-beta in the serum [54]. Strikingly genetic inactivation of STING blocked exercise-induced cytokine production and prevented the development of neurodegeneration in the double-mutant Parkin/mutator mice [54]. This contrasts with findings in Drosophila in which loss of STING does not affect behavioural and mitochondrial phenotypes of PINK1 and Parkin mutant flies [55]. Clinically, patients with Parkin are also found to have higher serum levels of IL-6 than those of controls [54], and this has been recently confirmed in independent  cohorts of patients with PD harbouring biallelic PINK1 and Parkin mutations [56]. Furthermore, a recent study has also found that cytosolic mtDNA is increased in idiopathic PD brain tissues and in zebrafish models of mutant PINK1, glucocerebrosidase (GBA1) and ATP13A2 that are sensed by IFI16 to induce type I interferon responses [57]. These exciting findings raise many questions regarding whether age-dependent mutated or oxidised mtDNA (that might accrue in PD) are differentially sensed to selective immune machinery and which cell types do these mechanisms occur to trigger nigral DAergic neuronal loss in PD and whether this is cell-autonomous or not.

Adaptive immune signalling
Parallel studies in immortalised cell lines indicate that activation of PINK1/Parkin can induce autophagyindependent mitochondrial turnover including mitochondria-derived vesicles (MDVs) that shuttle cargoes to lysosomes for degradation [58,59]. Mechanistically, PINK1/Parkin-dependent MDVs appear to be distinct to that of mitophagy, occurring independent of Dynamin-related protein 1 (Drp1) and requiring the Soluble NSF attachment proteins (SNAP) REceptor (SNARE) syntaxin-17 to mediate MDV endolysosome fusion [59,60]. Whilst MDV formation is dependent on Parkin catalytic activity, how Ub chains or specific substrates are involved in this process remains unknown. Physiologically, MDVs have been linked to mitochondrial antigen presentation (MitAP), and loss of PINK1 and/or Parkin drives MDV formation in immune cells which facilitates MitAP (Figure 3) [61]. Furthermore, exposure of PINK1-knockout mice to intestinal infection by Gram-negative bacteria led to a MitAP response and the induction of cytotoxic mitochondrial CD8þ T cells in the periphery, triggering dopaminergic pathological deficits in the brain and L-Dopa-responsive motor deficits [62]. This raises exciting follow-up studies on the cell specificity of these mechanisms in vivo and which specific substrates of PINK1 and Parkin control MDVs and MitAP. Generation of MDVs and MitAP is dependent on SNX9 and Rab9, and SNX9 levels have been observed to be regulated by Parkin, although it is unknown whether SNX9 is ubiquitylated by Parkin [61]. PINK1 has also been reported to induce phosphorylation of a highly conserved Ser111 residue on a subset of Rab GTPases, including Rab8A, Rab8B and Rab13. The regulation of Rabs by PINK1 appears to be indirect, and it remains unknown whether phosphorylation of these proteins influences MDV formation and trafficking [63].

Concluding remarks
PINK1 and Parkin research progress to date has been exemplar in demonstrating how understanding the normal function of PD genes is a critical first step to elucidate the fundamental mechanisms by which mutations lead to DAergic cell loss. This has revealed the central role mitochondrial homeostasis plays in neuronal integrity and survival in the ageing brain. The discovery that loss of PINK1 and Parkin can induce DAergic dysfunction via immune hyperactivation opens up new avenues for investigation on cell-autonomous and noncell-autonomous mechanisms of PINK1/Parkin signalling in vivo. Activated microglia and astrocytes play an important role in neuroinflammation in the CNS, and in future work, it will be critical to determine the molecular role of PINK1 and Parkin in these cell types. Therapeutically, activation of PINK1 and Parkin is a potential strategy to prevent neurodegeneration; however, screening and identification of small molecule activators remain challenging in drug discovery research. Defining the negative regulators of PINK1/Parkin signalling has revealed mitochondrial localised Ub-specific protease 30 as a candidate deubiquitinase [64]. Ubspecific protease 30 promotes efficient import of mitochondrial proteins, and its inhibition can enhance PINK1-catalysed phospho-ubiquitin accumulation after mitochondrial damage [47,65]. Another major question is defining parallel mitophagic pathways that may compensate for loss of PINK1 and Parkin. In neurons, multiple proteins are ubiquitylated in response to mitochondrial damage independent of PINK1 and Parkin; however, the ubiquitin E3 ligases that mediate this are unknown [43,47]. Mitochondrial ubiquitin ligase 1 (MUL1; also known as mitochondrial ubiquitin ligase activator of NF-kB (MULAN) and mitochondrial associated protein ligase (MAPL)) functions with Parkin to regulate degradation of paternal mitochondria from sperm, but the mechanism by which MUL1 induces mitophagy is unknown [66]. Furthermore, mitochondrial fusion factors mitofusin1/2 have been reported to be ubiquitylated by MUL1, Membrane-associated RING finger protein 5 (MARCH5), Mahogunin RING finger protein 1 (MGRN1) and HECT, UBA and WWE domain-containing protein 1 (HUWE1) as well as Parkin under distinct cellular stress, but no studies have systematically addressed their interdependence under mitophagic stress conditions [67]. Looking at the future, these and many other questions (Box 1) remain to be addressed to define the key mitochondrial mechanisms relevant to PD pathogenesis.