Monogenetic, familial forms of Parkinson's disease (PD) only account for 5–10% of the total number of PD cases, but analysis of the genes involved therein is invaluable to understanding PD-associated neurodegenerative signaling. One such gene, parkin, encodes a 465 amino acid E3 ubiquitin ligase. Of late, there has been considerable interest in the role of parkin signaling in PD and in identifying its putative substrates, as well as the elucidation of the mechanisms through which parkin itself is activated. Its dysfunction underlies both inherited and idiopathic PD-associated neurodegeneration. Here, we review recent literature that provides a model of activation of parkin in the setting of mitochondrial damage that involves PINK1 (PTEN-induced kinase-1) and phosphoubiquitin. We note that neuronal parkin is primarily a cytosolic protein (with various non-mitochondrial functions), and discuss potential cytosolic parkin activation mechanisms.
Loss-of-function mutations in the E3 ubiquitin (Ub) ligase parkin are the most common causes of autosomal-recessive (AR) Parkinson's disease (PD), a disease characterized by progressively worsening motor deficits, wrought by the death of dopamine-producing neurons in the substantia nigra pars compacta (SNpc). Pathologically, PD is characterized by the neuronal presence of intracytoplasmic inclusions rich in aggregated α-synuclein (α-syn) called Lewy bodies (LBs). Parkin was found to be present in the LBs of both familial and sporadic PD brains, where it co-localized with α-syn . Heterozygous mutations in parkin have also been described in the more commonly occurring sporadic PD cases [2,3].
The domain structure of the parkin gene (which plays a crucial role in its activation state) comprises an N-terminal Ub-like (Ubl) domain, followed by a 60 amino acid (aa) linker, a zinc-binding domain unique to parkin called the RING0 domain , one canonical Really interesting new gene (RING) domain termed RING1, and two linear zinc-binding folds termed In-Between RING (IBR) , and RING2 [6–9]. Together, the RING1, IBR, and RING2 folds form the catalytic RING-in-between-RING (RBR) module (Figure 1). Unlike canonical RING E3 domains, the parkin RING2 domain shows sequential Zn2+ ion coordination and cannot recruit E2 enzymes, and hence is not a true RING domain, but it is referred to as one nonetheless due to historical reasons . Over a hundred somatic AR-PD-causing mutations in the parkin gene have been identified, which map to every single structural domain, suggesting that a dynamic coaction between these domains may be required to attain optimum parkin activity [10,11].
Domain structure of parkin.
Parkin can polyubiquitinate and monoubiquitinate its substrate proteins . Protein ubiquitination is a fundamental post-translational modification that governs cell fate and function. The attachment of single or multiple Ub moieties to proteins facilitates their subcellular localization or marks them for proteolytic turnover. Ubiquitination is a complex process, requiring the successive actions of E1 (Ub-activating), E2 (Ub-conjugating), and E3 (Ub ligase) enzymes . There are two major subgroups among E3 ligases: RING E3 ligases that transfer Ub moieties from the catalytic cysteine on an E2 Ub-conjugating enzyme directly onto the substrate protein and HECT E3 ligases, wherein an intermediate thioester is formed between the Ub and the E3 ligase itself before substrate ubiquitination . In addition, RBR ligases have recently been shown to utilize thioesters, where the RING1 domain mediates the transfer of Ub first to the RING2 domain, an event that precedes substrate ubiquitination (Figure 2). Thirteen such RBR ligases exist in humans, including parkin [15–17].
Mechanism of action of E3 ubiquitin ligases.
Over the last decade, extensive studies on the cell biology of parkin indicate that it is associated with another PD-associated protein, PTEN-induced kinase-1 (PINK1) . The parkin–PINK1 signaling pathway is activated upon mitochondrial damage/dysfunction, in which damaged mitochondria are selectively targeted for autophagic destruction in a parkin-dependent manner in a process designated as mitophagy. This review will highlight the mechanisms by which parkin is activated and review parkin's diverse and important biology.
The domain structure of parkin reveals multiple autoinhibitory mechanisms
Biochemical studies and crystallization structure analyses of parkin domains revealed that it exists in an autoinhibited state (Figure 3) . The parkin Ubl domain binds to a motif between the IBR and RING1 domains, thereby obscuring the E2 enzyme-binding site on RING1 [20–22]. Autosomal mutations in this region can repress this inhibition, leaving parkin able to autoubiquitinate at the IBR domain but also rendering it susceptible to proteasomal degradation [6,19]. Pathogenic PD mutations that mapped to the Ubl domain were found to prevent Ubl-mediated parkin inhibition . Additional studies showed two further degrees of inhibition; the linker region that separates the IBR and RING2 domain [also called the REP (repressor element of parkin) region] also binds to the RING1 domain, thereby preventing the interaction of parkin with E2 ubiquitin-conjugating enzymes and ultimately preventing it from transferring Ubs to substrates. Also, the RING0 domain physically occludes access to the catalytic C431 in the RING2 domain. Both these barriers must be overcome in order for parkin to attain substantial Ub ligase activity [7–9]. The Ubl domain binds to RING1 via hydrophobic interactions that are dependent on Ile44 [7,9,20,22,23], but can also bind to SH3 and Ub-interacting motifs on other proteins [24–29], suggesting that parkin autoinhibition could be prevented by competition for interaction with the Ubl domain. The W403 residue in the linker region between the IBR and RING2 domain, designated the REP, is thought to be essential for binding to the E2 Ub-binding site on the RING1 domain. Mutating this residue brought about an increase in parkin activity by facilitating an increased E2–RING1 interaction. Cells overexpressing W403A parkin showed faster parkin recruitment to mitochondria following treatment with carbonyl cyanide m-chlorophenyl hydrazine (CCCP), a mitochondrial depolarizing agent that leads to parkin activation and recruitment to the mitochondria . AR-PD parkin mutations map to regions that govern proper protein folding, such as zinc-co-ordinating residues, and other mutations which are required for E3 ligase activity. The C431F familial mutation directly causes a defect in the parkin catalytic mechanism. Other parkin mutation hotspots exist in gene regions that encode amino acids at domain interfaces, namely the RING0–RING2 interface and the REP–RING1 interface . Thus, it seems that the complex parkin quaternary structure has evolved to maintain it in a basally inhibited state, and mutations that prevent the formation of inter-domain interfaces or proper folding as well as mutations that directly affect parkin's catalytic activity often have pathological consequences.
Parkin exists in an autoinhibited state.
The PINK1 and phosphoubiquitin-mediated activation of parkin
Parkin activation is associated with the signaling biology of another PD-associated protein, PINK1. Early studies in Drosophila indicated that these proteins function in the same signaling pathway, mediating mitochondrial quality control in neurons as well as in other cell types [30,31]. These studies also indicated that PINK1 activation may lie upstream of parkin activation. Deficits in the PINK1–parkin pathway cause mitochondrial dysfunction, one of the major hallmarks of PD-associated neurodegeneration [32,33]. Under normal physiological conditions, PINK1 is constitutively imported into mitochondria, where it is cleaved by PARL protease and degraded by the proteasome . Upon the treatment of cells with the protonophore CCCP, loss of mitochondrial membrane potential occurs, causing accumulation of PINK1 on the outer mitochondrial membrane, followed by recruitment of parkin to the mitochondria [35–37]. Other mitochondria-specific agents such as paraquat, antimycin A, or oligomycin A also activate this pathway [38,39]. Parkin then mediates the polyubiquitination of outer membrane mitochondrial proteins, thereby facilitating the clearance of damaged mitochondria via autophagy (mitophagy) [40,41]. PINK1 was initially thought to recruit parkin by phosphorylating outer mitochondrial membrane proteins such as the GTPase mitofusin (Mfn2). Phosphorylated Mfn2 acted as a docking site for parkin . More recent studies suggest that phosphorylated Ub chains are important for the parkin recruitment to damaged mitochondria (see below) [40,43,44]. The autophagy receptors optineurin and NDP52 also bind to phosphorylated Ub chains to recruit the autophagy machinery to mitochondria [45,46]. Mitochondrial depolarization also activates TANK-binding kinase 1 (TBK1), which phosphorylates optineurin on S473 to enhance mitophagy . However, in this study, phosphoUb decreases the affinity of autophagy receptors for ubiquitin, in opposition to what the former study showed .
Recombinant insect PINK1 isoforms that are catalytically active  can directly phosphorylate parkin at S65 in its Ubl domain. S65 phosphorylation is required but not sufficient for parkin mitochondrial translocation . Since the Ubl domain interacts with the REP element and the RING1 domain of parkin, preventing its activation, it was shown that PINK1-mediated phosphorylation of parkin at S65 is required for parkin derepression. Microscopy was used to demonstrate that the parkin S65A mutation impairs its recruitment to mitochondria . In another study, however, mutating the S65 residue to an alanine residue resulted in no deficits in either localization or Mfn2 ubiquitination, as well as the dependence of parkin on PINK1 to be activated, suggesting that this mechanism might be incorrect , though the authors did not utilize microscopy to analyze S65A parkin recruitment kinetics. Tang et al.  recently demonstrated that S65A parkin is impaired in both recruitment and mitophagy.
A series of articles provided evidence that PINK1 was also a Ub kinase and that its phosphorylation of Ub was required for parkin activation [50–52]. Mass spectrometry analysis of CCCP-treated mitochondrial proteins from wild-type and PINK1-deficient cells showed that PINK1 phosphorylated Ub at S65, analogous to its phosphorylation of the Ubl domain of parkin at S65. Phosphoubiquitin (phosphoUb) activated parkin and brought about its mitochondrial translocation and was required for increased E3 ligase activity . Phosphomimetic Ub can directly activate parkin by accelerating the discharge of the E2–Ub conjugate . High-resolution structural analysis of parkin showed that phosphoUb binds to the RING1 domain of parkin, promoting Ubl release [21,22]. Binding of phosphoUb to parkin also promotes its phosphorylation by PINK1, leading to its maximal activation . Parkin lacking the inhibitory Ubl domain can also be activated by phosphorylated Ub in the absence of PINK1, as well as by the isolated, phosphorylated Ubl domain in trans. Although parkin lacking the inhibitory Ubl domain could translocate to mitochondria, it still displayed deficits in ubiquitinating mitochondrial components . This observation, coupled with the finding that parkin lacking the inhibitory Ubl domain prevents C431S parkin from trapping Ub , suggested that the Ubl domain may have both inhibitory and activating roles with regard to parkin activation. Evidence for this was provided by several studies. PhosphoUb forms an interface with the RING1 and IBR domains on parkin. It also interacts with the side chains of the RING0 domain. The phospho group of the phosphoUb and the hydrophobic Ile44 both interact with different subregions of the RING1 domain, the latter binding to a helical structure called a phosphoUb-binding helix (pUBH). PhosphoUb also binds to the β-hairpin loop in RING1. Finally, the C-terminus of phosphoUb interacts with the IBR region. Mutations in the putative phosphoUb-binding region include some PD-associated mutations, suggesting a putative pathogenic role wrought by the defective ability of parkin to be activated by phosphoUb. Not only do these mutations prevent phosphoUb binding to parkin, they also inhibit parkin mitochondrial translocation following CCCP treatment. PhosphoUb binding to the pUBH was shown to destabilize the interaction between the inhibitory Ubl domain and the RBR core . Other studies also provided evidence for the dual roles of Ubl in parkin activation [21,22,53]. It should be noted that the structure of phosphoUb-bound parkin resolved by Wauer and colleagues utilized parkin that was missing the first 16 kDa, or 140 residues. More recently, the structure of human parkin (with its Ubl domain intact) bound to phosphoUb has been described . Some studies suggest that parkin lacking the Ubl domain is still autoinhibited , while others showed that Ubl deletion permits parkin activation, albeit to a degree lower than when phosphoUb is added . The Ubl domain, which is released following phosphoUb binding, helps unravel the autoinhibited parkin conformation, thereby augmenting parkin activation. Further experiments revealed that the Ubl domain release is a prerequisite for its PINK1-dependent phosphorylation and activation [21,22,53,55]. Hence, a concerted model of parkin activation in which the closed autoinhibitory conformation mediated primarily by the Ubl and REP regions binding to the RING1 domain is released by phosphoUb binding to RING1, causing a reversible and transient increase in parkin activity. This releases the Ubl domain, leaving it free to be phosphorylated by PINK1, and giving rise to a fully active parkin molecule (Figure 4) .
PINK1 and phosphoubiquitin-mediated parkin activation model.
Cytosolic parkin and mechanisms of cytosolic parkin activation
The current model for parkin activation based on the studies described thus far envisions a cytosolic, inactive form of parkin, which is activated following mitochondrial damage upon phosphorylation and mitochondrial recruitment by PINK1. However, localization studies indicate that parkin is primarily cytosolic and it only becomes localized to the mitochondria after mitochondrial damage [35–37]. Thus, parkin is likely to have non-mitochondrial functions. Indeed, over the course of the last 20 years, several non-mitochondrial parkin substrates have been identified and characterized, indicating that there are probably other parkin activation mechanisms that come into play to activate parkin at non-mitochondrial localizations. For instance, many of these substrates are presynaptic proteins. Parkin polyubiquitinates the α-syn-interacting protein synphilin-1, a protein which is enriched specifically in LBs in PD brains [57–61]. Synaptic vesicle-associated proteins, such as cell division cycle related-1 (CDCrel-1) and CDCrel-2, were also found to be parkin substrates [62–67], as was synaptogamin XI, a protein involved with synaptic vesicle formation and docking [68,69]. Recently, parkin was also found to polyubiquitinate synaptotagmin IV, another member of the synaptogamin family of membrane-trafficking proteins . The parkin Ubl binds the SH3 domain of Bin/Amphiphysin/Rvs homology (BAR) proteins such as endophilin-A to ubiquitinate and regulate synaptic proteins and transmission . It is up-regulated in endophilin-mutant mice . Parkin interacts with Eps15 to regulate epidermal growth factor (EGF) receptor trafficking and endocytosis . The C-terminus of parkin also functions as a PDZ-binding motif that binds to calcium, calmodulin-associated serine/threonine kinase (CASK) to regulate synaptic transmission and plasticity . Parkin negatively regulates the number and strength of excitatory synapses through its E3 ligase activity . Parkin controls DA utilization in human DA neurons probably through the regulation of the levels of monoamine oxidase [74,75]. Consistent with its role as a ubiquitin ligase for multiple synaptic element components, induced pluripotent stem cell (iPSC)-differentiated DA neurons from parkin-mutant patients exhibited spontaneous dopamine D1 class receptor-dependent excitatory postsynaptic currents (EPSCs), which were abolished following parkin overexpression, providing credence that accumulation of non-mitochondrial parkin substrates may contribute to PD-associated pathophysiology . Mutations in parkin also reduce the complexity of neuronal processes in iPSC-derived neurons . Proteomic analysis of parkin knockouts indicates that there are defects in synaptic function due to the absence of parkin .
Parkin also polyubiquitinates cytoskeletal components such as α- and β-tubulin where it stabilizes microtubules [79,80]. It is associated with actin filaments in neuronal and non-neuronal cells . Parkin activation reduces LIM kinase 1-induced actin filament accumulation . Cytoskeletal alterations are found in parkin-mutant human primary fibroblasts . Parkin associates with the cytosolic NF-κB complex proteins NF-κB essential modifier (NEMO) and tumor necrosis factor (TNF) receptor-associated factor 2 (TRAF2) . It also binds to the Rpn10 and S5a subunits of the 26s proteasome via its Ubl domain [27,28]. Aminoacyl tRNA synthetase complex-interacting multifunctional protein 2 (AIMP2), a protein required for the assembly and stability of the aminoacyl tRNA synthetase complex in translation, is a parkin substrate [85,86], along with the far upstream element-binding protein-1 (FBP-1) [87–89]. AIMP2 accumulates in adult conditional parkin knockout mouse brains, as well as in the brains of familial and sporadic PD patients [85–87,89]. It was subsequently shown to elicit cell death via activation of poly(ADP-ribose) polymerase-1-dependent, parthanatos activation . Parkin directly acts as a transcriptional repressor, preventing the expression of p53 via its RING1 domain. Familial PD parkin variants did not show this effect . More recent studies demonstrate that parkin contributes to mitochondrial biogenesis in dopaminergic neurons by polyubiquitinating and mediating the proteasomal clearance of the transcriptional repressor parkin-interacting substrate (PARIS) [92,93]. Upon parkin dysfunction, PARIS accumulated and repressed the expression of the master mitochondrial biogenesis gene, peroxisome proliferator-activated receptor gamma co-activator 1-alpha (PGC1α). Other groups have found a similar role for PARIS in models of parkin inactivation [94–96]. Like AIMP2, increased PARIS expression occurs in both sporadic and familial PD brains, as well as in adult-onset models of ventral midbrain-specific parkin ablation [92,93].
It is important to mention that some of the above studies on non-mitochondrial parkin substrates used immunoprecipitation, pull-downs, and expression systems, and some included zinc-chelating agents to validate their findings without proper controls, which may give rise to false positives. Many, but not all, of these substrates need to be independently verified and future direct biophysical studies need to be performed to augment the cell biologic and animal model data that exist supporting their role as parkin substrates. Although the above discussion on non-mitochondrial parkin substrates is not exhaustive, it is clear that parkin has non-mitochondrial roles. This is not surprising because parkin is primarily a cytosolic protein that only localizes to the mitochondria after mitochondrial damage [35–37]. What then might account for cytosolic parkin activation in the absence of mitochondrial depolarization? There is evidence to indicate that cytosolic parkin can be activated and inactivated by alternative routes other than phosphoUb- and PINK1-mediated activation. Indeed, S-nitrosylation of parkin was one of the first endogenous activators described for parkin [97–99]. Subsequent studies showed that S-nitrosylation of C323 activates parkin to induce mitochondrial protein degradation . Since weakening interdomain interactions that are required for parkin autoinhibition can activate parkin, it is conceivable that nitrosylation at the RING1 domain weakens the Ubl/RING1 or REP/RING1 interactions, opening up parkin quaternary structure and facilitating its activation . Interestingly, S-nitrosylation of parkin first activates and then reduces its E3 ligase activity probably through sequential S-nitrosylation on different cysteines [97–99,101]. In postmortem sporadic PD brain lysates, there is marked evidence of increased S-nitrosylative stress and parkin S-nitrosylation compared with non-disease controls, suggesting that parkin has reduced activity in sporadic PD [98,101]. Parkin is also activated by sulfhydration where it enhances the degradation of parkin substrates .
Carboxy-terminus of Hsp70-interacting protein (CHIP), also known as STUB1, functions as an E4-like molecule that positively regulates parkin's E3 activity [103,104], where it controls the levels of the Pael receptor . CHIP directly enhances parkin-mediated in vitro ubiquitination of the Pael receptor . Although it may be difficult if not impossible to crystallize parkin modified by S-nitrosylation or sulfhydration due to the transient nature of these modifications, knowing the sites of S-nitrosylation and sulfhydration will facilitate molecular modeling studies. If CHIP can be co-crystallized with parkin, it would provide important insights into how CHIP positively regulates parkin's E3 activity. Cytosolic phosphoUb is another mechanism accounting for cytosolic parkin activation. PhosphoUb is present in the cytosol and is PINK1-dependent, suggesting that PINK1 phosphorylation of Ub could activate cytosolic parkin [40,52,105,106].
Inactivation of cytosolic parkin contributes to neurodegeneration in sporadic PD
Since parkin plays many roles in non-mitochondrial cellular compartments, mechanisms that would prevent it from being activated might contribute to the degenerative process in PD. Several post-translational modifications inhibit parkin's E3 ligase activity or maintain it in a catalytically inactive state (Figure 5). As noted above, sequential S-nitrosylation inhibits parkin and there is evidence of excessive S-nitrosylation in sporadic PD [97–99,101]. In macrophages, parkin can be cleaved by caspase-1, leading to its inactivation [107,108]. Dopamine, which can auto-oxidize upon exposure to oxidizing agents to form dopamine-quinones, covalently attacks parkin cysteine residues, inhibiting its activity . The dopamine-mediated inhibition of parkin was mapped to the quinone modification of C268 and C231 in the RING1 domain . Other forms of oxidative stress can also prevent parkin activation [110–112]. Casein kinase I and cyclin-dependent kinase 5 (cdk5) caused reduction in parkin solubility, leading to its inactivation . Cdk5 phosphorylates parkin S131, leading to its inhibition . c-Abl-mediated tyrosine phosphorylation of parkin at Y143 inhibits its E3 Ub ligase activity [87,89]. In sporadic PD postmortem SN, c-Abl is overactive, leading to tyrosine phosphorylation of parkin and accumulation of the cytosolic parkin substrates, PARIS, AIMP2 and FBP-1, suggesting that this mechanism of inactivation or prevention of activation is plausible [87,89,92]. Other cytosolic parkin substrates such as STEP61 (striatal-enriched protein tyrosine phosphatase 61) also accumulate in sporadic PD . Knockout or pharmacologic inhibition of c-Abl prevents the inactivation of parkin and the accumulation of these parkin substrates, leading to a reduction in the loss of dopamine neurons and behavioral restoration in animal models of PD [87,89,115–117]. The use of c-Abl inhibitors has advanced to encouraging preliminary clinical studies demonstrating restoration of function in PD patients . However, it remains to be seen whether c-Abl inhibition exerts neuroprotection via preventing parkin inhibition.
Post-translational modifications that increase or decrease parkin activity.
Concluding remarks and future perspectives
PD is a progressive neurodegenerative disorder characterized by the selective death of dopamine neurons in the SNpc of the ventral midbrain. The discovery of rare, inheritable variants of the disease opened up the cell biology of PD. Understanding the function of PD-related gene-encoded proteins has helped in the understanding of the relatively specific neurodegeneration that characterizes familial and sporadic PD. Since its identification as a PD-associated gene in 1998, there has been a surge of articles studying parkin biology . Parkin was shown to be an E3 Ub ligase [67,120,121]. X-ray crystallography-derived structural resolution of parkin revealed that it exists in an autoinhibited conformation through the Ubl and REP regions binding to the RING1 domain, preventing E2 enzyme binding, and the RING0 domain, which occludes the catalytic C431. Following mitochondrial depolarization, parkin translocates to the mitochondria where it is activated by the stabilization of PINK1, leading to phosphorylation of both Ub and parkin at S65, removing the autoinhibition. This mechanism accounts for the dual role of PINK1 and parkin in mitophagy.
Parkin also has non-mitochondrial roles, and the list of non-mitochondrial parkin substrates is constantly growing and now encompasses a large number of proteins with varying functions. As this list of proteins grows, it will help to elucidate the signaling pathways that underlie PD-associated neurodegeneration. Cytosolic parkin may be activated by S-nitrosylation and cysteine sulfhydration as well as cytosolic phosphoUb. There are also many post-translational modifications (PTMs) that prevent parkin's activation and probably contribute to the mechanism(s) of how defects in parkin function contribute to the pathogenesis of PD. Future studies will be required to understand how these alternative activation mechanisms work from a structural standpoint and how the PTMs prevent parkin activation. Since protein crystallization freezes a protein in a conformational state that is amenable to crystallization, it will be important in future studies to utilize cryo-electron microscopy (cryo-EM) to gain a more comprehensive understanding of parkin's activation and inhibitory mechanisms . However, current cryo-EM technology is limited to resolving protein structures above 150 kDa. Hence, cryo-EM-mediated resolution of parkin or even a parkin-E2-phosphoUb complex is not possible yet. Advances in structural biology techniques may help to elucidate the various activation states of parkin.
aminoacyl tRNA synthetase complex-interacting multifunctional protein 2
carbonyl cyanide m-chlorophenyl hydrazine
cell division cycle related
cyclin-dependent kinase 5
C-terminus of Hsp70-interacting protein
far upstream element-binding protein-1
Presenilins-associated rhomboid-like protein
repressor element of parkin
Really interesting new gene
substantia nigra pars compacta
striatal-enriched protein tyrosine phosphatase 61
STIP1 homology and U-Box containing protein 1
TANK-binding kinase 1
This work was supported by National Institutes of Health/National Institute of Neurological Disorders and Stroke [grant P50 NS38377] and the JPB Foundation. T.M.D. is the Leonard and Madlyn Abramson Professor in Neurodegenerative Diseases. N.P. is supported by a post-doctoral fellowship from the Maryland Stem Cell Research Foundation [2017-MSCRFF-3838].
The authors acknowledge the joint participation by the Diana Helis Henry Medical Research Foundation and Adrienne Helis Malvin Medical Research Foundation through its direct engagement in the continuous active conduct of medical research in conjunction with the Johns Hopkins Hospital and the Johns Hopkins University School of Medicine and the Foundation's Parkinson's Disease Program M-2, H-2014, M-2014. The authors thank I-Hsun Wu for assistance with figure illustrations.
The Authors declare that there are no competing interests associated with the manuscript.