Regulation of Parkin E3 ubiquitin ligase activity

Abstract

Parkin is an E3 ubiquitin ligase mutated in autosomal recessive juvenile Parkinson’s disease. In addition, it is a putative tumour suppressor, and has roles outside its enzymatic activity. It is critical for mitochondrial clearance through mitophagy, and is an essential protein in most eukaryotes. As such, it is a tightly controlled protein, regulated through an array of external interactions with multiple proteins, posttranslational modifications including phosphorylation and S-nitrosylation, and self-regulation through internal associations. In this review, we highlight some of the recent studies into Parkin regulation and discuss future challenges for gaining a full molecular understanding of the regulation of Parkin E3 ligase activity.

Introduction

Parkinson’s disease (PD) is a neurodegenerative disorder characterised by tremors, bradykinesia and rigidity, symptoms caused by the selective and premature loss of dopaminergic neurons in the substantia nigra. PD is far from rare, affecting between 1 and 3 % of those over 65 years of age. Until recently, PD was thought to be sporadic, caused by environmental factors. However, in the last 15 years, evidence has mounted that in 5–10 % of PD cases, there are genetic components in both hereditary forms such as autosomal recessive juvenile Parkinson’s disease (ARJPD) and sporadic PD [1–6]. There are several genetic components, including the dominant genes α-synuclein and LRRK2, and the recessive genes DJ-1 and PINK1. However, the most frequently mutated gene, causing ~50 % of ARJPD, is PARK2, which encodes for the protein Parkin. A multitude of genetic mutations in PARK2 give rise to ARJPD including missense, exon duplications, rearrangements, deletions and truncations (Fig. 1a). In addition to the established role in ARJPD, Parkin is also a putative tumour suppressor [6]. The parkin gene is located on the long (q) arm of chromosome 6 between positions 25.2 and 27 close to a fragile chromosome site, FRA6E, known to be a hotspot for deletions and other alterations in cancer cells, which spans a large genomic region and is frequently lost in a number of human cancers [7–10]. With such a central role in human health and disease, it is no surprise that Parkin is highly regulated.

Fig. 1.

Schematic representation of Parkin exon and protein arrangement. a Exon arrangement and mutations that occur in the parkin gene, b the protein domain organisation of Parkin in relation to the exons, with amino acid numbering according to the human sequence to highlight the domains, c the spread of pathogenic point mutations in Parkin, with 22 indicating those mutations that fall outside of the defined domains

The parkin gene (PARK2) comprises 12 exons and codes for a 465 amino acid protein (Fig. 1b). It is widely expressed in a variety of tissues, predominantly in the muscle and brain [11, 12]. Parkin is a multidomain protein comprising an N-terminal ubiquitin-like domain (Ubl), a cysteine-rich RING0 domain [13], and 2 C-terminal really interesting new gene (RING) domains (RING1 and RING2) separated by a cysteine-rich, zinc-binding in between RING (IBR) domain (Fig. 1b). Pathogenic point mutations are dispersed throughout the amino acid sequence, with clusters in each of the domains (Fig. 1c); and for a summary of mutations see ([14] and http://www.molgen.ua.ac.be/PDmutDB). In common with the majority of RING-domain containing proteins, Parkin has E3 ubiquitin ligase activity [12, 15].

Ubiquitination is a posttranslational modification required for the regulation of many cellular pathways including cell cycle control, endocytosis, inflammation and DNA repair. Conjugation of ubiquitin to a target protein requires the coordinated action of a ubiquitin-activating enzyme (E1), ubiquitin-conjugating enzyme (E2) and a ubiquitin-ligase (E3). The E3 enzyme generally confers the substrate specificity by catalysing the conjugation of ubiquitin usually to the epsilon-amino group of a lysine in the substrate. Further ubiquitin molecules can then be attached to ubiquitin itself, by virtue of it having 7 lysine residues on the surface and the amino-terminus, all of which serve as acceptors for an activated ubiquitin (for reviews, see [16, 17]). A diversity of cellular fates awaits a ubiquitinated protein, depending on the type of modification, and all possible chains can be made [18]. A K48-linked ubiquitin chain can signal for destruction of the protein by the proteasome. K11-linked chains can also target for degradation. K63-linked chains play roles in DNA repair, protein trafficking and other non-proteasomal pathways [19–21]. Linear ubiquitination has been reported to function as a degradation signal, and in NFκB signalling [17, 22–24]. Other polyubiquitin modifications are less well functionally characterised, but there are also numerous degradation-independent functions of ubiquitin signalling through polyubiquitination, monoubiquitination and multiple monoubiquitination [25, 26].

Parkin E3 ligase activity is absolutely dependent on the RING2 domain, as any truncation of this domain or disruption of the fold through mutation of the zinc-coordinating residues required for correct folding result in inactive Parkin. However, an N-terminal truncation encompassing only the IBR-RING2 domains is active [27–29]. Parkin was originally shown to have ubiquitin ligase activity through autoubiquitination assays [12, 15]. Since then, there has been an ever-growing list of putative Parkin substrates, but confirmation of many of these substrates and their functional consequences remain controversial. Part of the complexity is no doubt due to the different techniques used to determine Parkin substrates, ranging from yeast 2-hybrid assays to co-immunoprecipitations from different cell types and proteomic analyses of Parkin-null mouse striata [15, 30–52]. In addition, there is often an assumption that a ‘genuine’ substrate of Parkin should accumulate in the brains of Parkin-linked PD patients and/or Parkin-null mice. Certainly, this is a good indication of a potential substrate; however, such an assumption ignores the data that suggest Parkin can also catalyse monoubiquitination [28, 34, 43, 48, 53] and that the outcome of polyubiquitination could be non-degradative [32, 48, 53]. In addition, Parkin-null mice do not exhibit loss of nigrostriatal dopaminergic neurons, so may not faithfully recapitulate the human scenario [54–56]. However, the list of diverse substrates does implicate Parkin in a variety of cellular pathways, something which would seem to align with the myriad disease-associated mutations. Consequently, the identification of genuine Parkin substrates is of great importance to gain insights into Parkin function and disease. Several groups have identified and reported putative Parkin substrates that in principle interact with Parkin and are either mono- or polyubiquitinated in a Parkin-dependent manner. The substrates are listed in Table 1, along with a summary of the pathways involved, techniques used to identify the substrate, and accumulation in patient brains or Parkin-null mice if identified. Although some of the putative substrates have been reported to accumulate in brain samples, research in experiments performed in parkin null mice and/or ARJPD patients have been conflicting [30, 42, 55], and their roles in the pathogenesis of Parkinson’s disease remain to be confirmed [57].

Table 1.

Putative Parkin substrates, the pathways involved, techniques used to identify the substrate, and accumulation in patient brains or Parkin-null mice

(tag+)Substrate	Method of identification	Parkin	Biological pathway	Accumulation	References
cMyc-SEPT5 (CdCrel-1)	Yeast 2-hybrid, full-length Parkin as bait	Endogenous and cMyc-Parkin	Synaptic function	Nd	[15]
cMyc-Synphilin-1	Found in Lewy body-like inclusion	FLAG-Parkin	Unclear, but may play a role in neurodegeneration	No	[32, 137]
αSP22	Co-IP from frontal cortex of normal brain tissue	Endogenous	Unclear	Yesa	[33]
cMyc-SEPT4 (CDCrel-2)b	Yeast 2-hybrid, full-length Parkin as bait	cMyc-Parkin	Synaptic function	Yes	[34]
FLAG-Pael-R	Yeast 2-hybrid, full-length Parkin as bait	Endogenous and GST-Parkin	Unknown	Yes	[35, 138]
Cyclin E	Pull downs using Parkin to IP	FLAG-Parkin	Cell cycle control	Yesc	[36]
cMyc-p38/JTV-1	Yeast 2-hybrid, Parkin 135–290 as bait	HA-Parkin	Protein biosynthesis	Yesd	[31, 42]
GFP-synaptotagmin XI	Yeast 2-hybrid, full-length Parkin as bait	HA-Parkin	Synaptic function	No	[139]
α/β-tubulin	Yeast 2-hybrid, full-length Parkin as bait	FLAG-Parkin	Cytoskeleton	No	[38]
GST-polyQ proteins	Co-localisation with model proteins	FLAG-Parkin	Neurotoxicity	No	[39]
FLAG-dopamine transporter	N/A	Untagged-Parkin	Neurotransmission	No	[40]
HA-SIM2 (single-minded 2)	N/A	FLAG-Parkin	Transcription	No	[41]
cMyc-FBP1	Interaction with AIMP2	cMyc-Parkin	Transcription	Yes	[140]
His-Eps15#	Ubiquitination performed in vitro and in cells	GST/FLAG-Parkin	Endocytosis	No	[43]
GFP-RanBP2	Yeast 2-hybrid, full-length Parkin as bait	cMyc-Parkin	Nuclear import/SUMOylation	No	[44]
FLAG-IKKγ and FLAG-TRAF2	Co-transfection and co-immunoprecipitation	Untagged-Parkin	NF-κB signalling	No	[45]
PICK1 and cMyc-PICK1b	GST pulldowns from mouse synaptosomes	GST/FLAG-Parkin	Synaptic function	No	[53]
DJ-1(L166P mutant)	N/A	Endogenous	Neurotoxicity	No	[46]
3xFLAG-PDCD2-1	Yeast 2-hybrid, Parkin 1-238 used as bait	Endogenous and cMyc-Parkin	Apoptosis/cell proliferation	Yes	[47]
EGFP-Bcl-2b	GST pulldowns from HEK293 cells	GST/FLAG-Parkin	Anti-apoptosis	No	[48]
VDAC1/p62-SQTM1	N/A	FLAG-Parkin	Autophagy/Mitophagy	No	[49]
FLAG-Mitofusins 1 and 2	N/A	Endogenous	Mitochondrial fusion	No	[50, 60–63]
cMyc-Drp1	N/A	GFP-Parkin	Mitochondrial fission	No	[141]
FLAG-PARIS	Yeast 2-hybrid, full-length Parkin as bait	cMyc-or His-Parkin	Transcription	Yes	[52]
MIRO	N/A	YFP-Parkin	Mitochondrial function	No	[51]

Nd not determined

^aGoldberg et al. [55] found slight increase in αSP22 in Parkin-null mice

^bMono-ubiquitination identified

^cNot found to accumulate in Parkin-null mice [42]

^dKo et al. [42] report significant accumulation in ARJPD patient brains and Parkin-null mice, but these findings were not confirmed by Periquet et al. [30, 142]

A growing body of evidence suggests that one functional consequence of Parkin dysfunction is loss of the quality control clearance of depolarised and fragmented mitochondria through mitophagy [58, 59]. Consistent with this hypothesis, one of the few putative substrates that has been identified and corroborated in independent laboratories are the mitochondrial proteins Mitofusin (1 and 2) [50, 60–63], which are required for mitochondrial fusion. Selective degradation of the mitofusins may prevent fusion of damaged mitochondria and therefore promote mitophagy [64]. In addition, two recent potential substrates have been identified that are also implicated in mitochondrial homeostasis [51, 52], adding to the hypothesis that Parkin dysfunction may result in aberrant mitochondrial regulation.

What follows is a review and discussion of the various means by which Parkin activity may be regulated, outside of interactions with putative substrates.

Regulation of Parkin activity through internal interactions

Historically, Parkin has been considered as a constitutively active protein due to its apparent autoubiquitination activity in vitro and in cell culture [12, 15]. However, a recent study suggests that wild-type native Parkin adopts an autoinhibited conformation mediated by an intramolecular interaction between the N-terminal Ubl-domain and putative peptides in the C-terminal RING regions [65]. In addition, the same C-terminal regions of Parkin interact with ubiquitin and this interaction is required for efficient ligase activity even in the active state. An independent study revealed that MBP-Parkin could catalyse E2-independent autoubiquitination in the presence of high concentrations of E1 [27]. Furthermore, removal of the N-terminal regions of Parkin, including the Ubl and RING0 domains, allowed Parkin to catalyse polyubiquitin chains as well as monoubiquitination, while the presence of the Ubl domain restricted the modification to monoubiquitination. Interestingly, a study of the mechanistic reactivity of the E2 UbcH7 revealed that HHARI (human homolog of Ariadne-1), a member of the RING-IBR-RING family of E3 ligases, forms a catalytic thioester-bound intermediate with ubiquitin in a manner analogous to the HECT-type E3 ligases [66]. Taken together, these studies suggest a mode of Parkin self-regulation whereby the N-terminal Ubl domain inhibits Parkin autoubiquitination, but once released, whether by pathogenic mutations or interactions with other proteins (see later), Parkin interacts covalently or non-covalently with ubiquitin in a semi-catalytic intermediate to catalyse an efficient transfer of ubiquitin [65, 66]. The key to the catalytic activity is the recognition of charged ubiquitin, rather than E2, explaining the observation that, in the presence of high concentrations of E1, an E2 can be dispensed with [27] (Fig. 2). Many early reports on Parkin mutations hypothesised a loss-of-catalytic-activity as the mechanism of Parkin-associated pathogenesis, but several biochemical studies reveal that loss of activity is not a unifying mechanism [28, 29, 65, 67]. Coupled with earlier studies that show a C-terminal fragment comprising IBR-RING2 is sufficient for Parkin activity [28, 29], it is clear that the N-terminal Ubl and RING0 domains are not required for Parkin activity per se, but are required for regulating its activity.

Fig. 2.

Parkin regulation through internal interactions. Parkin exists in an autoinhibited state which can be relieved by removal of the Ubl domain, or release of the Ubl domain from the C-terminal regions of Parkin i. Non-covalent ii or covalent interaction with ubiquitin allows for productive transfer iii. Polyubiquitination occurs in the absence of the Ubl and RING0 domains, while substrate (synphilin-1 [27]) ubiquitination is only supported by the IBR-RING2 fragment iv

Regulation of Parkin through external interactions

In addition to the putative substrates identified for Parkin E3 ligase activity, there are multiple proteins reported to interact with Parkin. The picture is complicated by the fact that Parkin itself is prone to misfolding [29, 68–70], and therefore under certain conditions may appear to interact with other poorly folded proteins, but these may not be specific interactions. However, in this section, we will summarise the current evidence for regulation of Parkin activity through interactions with multiple external proteins.

PINK1 (PTEN-induced kinase 1)

PINK1 is a highly conserved mitochondrial kinase, mutations in which are also linked to ARJPD [71]. Mutations cluster in the kinase domain, suggesting a loss-of-function of the kinase activity [72, 73]. Interestingly, in common with Parkin-null mice, PINK1-null mice do not show nigrostriatal degeneration [74–76]. However, a genetic link between Parkin and PINK1 has been identified in which Parkin and PINK1-mutant cells and flies exhibit similar phenotypes. Parkin overexpression can rescue loss of PINK1, but not vice versa, placing Parkin downstream of PINK1 [77–82]. It is reported that PINK1 can phosphorylate Parkin [83, 84], but there is no evidence that Parkin ubiquitinates PINK1. In 2008, the Youle group reported the translocation of Parkin to dysfunctional mitochondria [85]. Subsequent studies by several groups found that PINK1 is required to translocate Parkin to dysfunctional mitochondria [49, 67, 86] (reviewed in [59]). In HeLa cells exposed to the uncoupling agent carbonyl cyanide 3-chlorophenylhydrazone (CCCP), overexpressed Parkin co-localises to the mitochondria and induces mitophagy. Such localisation appears to be abolished in fibroblasts from PINK1-null mice [86]. Interestingly, the PINK1-dependent mitochondrial localisation of Parkin is required to activate ‘latent’ Parkin activity, supporting the notion that Parkin activity is inhibited and requires activation [67]. A note of caution, however, comes from a recent study conducted in vivo which finds that, while mCherry-Parkin localises to the mitochondria in transiently transfected HeLa cells, the same construct administered in mice fails to co-localise with mitochondria in dopaminergic neurons. Equally, removal of Parkin does not modify the pathway in their readout, leading to the suggestion that Parkin-mediated mitophagy is a rare event in vivo [87]. Further studies in animal models will be required to resolve these discrepancies; however, it appears that some interaction between PINK1 and Parkin, direct or indirect, is involved in mitochondrial homeostasis. Whether the regulation of Parkin by PINK1 is limited to localisation or extends to modifications and influence in choice of substrate remains to be determined.

Parkin as part of a multiprotein complex

PINK1 and Parkin have been reported to collaborate with yet another protein implicated in recessive Parkinsonism, DJ-1 [88]. As reported by Xiong et al. [88], differentially-tagged Parkin, DJ-1 and PINK1 co-elute at a size of ~200 kDa by size-exclusion chromatography (Fig. 3a). This complex then promotes the ubiquitination and degradation of unfolded proteins. Mitochondrial fragmentation caused by α-synuclein overexpression could be rescued by co-expression of each protein, but not certain mutants [89]. However, loss of DJ-1 gives rise to a different mitochondrial phenotype than seen with loss of PINK1 or Parkin, suggesting that this complex may not be required for mitochondrial homeostasis. Furthermore, a subsequent independent study also analysed complex formation using endogenous DJ-1 with overexpressed Parkin and PINK1, separated by size-exclusion chromatography [90], and found that DJ-1 did not co-elute with PINK1 and Parkin. A very recent study also fails to find a stable complex between PINK1 and Parkin [91] in either an endogenous or overexpression setting. Therefore, at this point, it remains unclear whether Parkin is part of a complex including DJ-1 and PINK1, but since all 3 proteins offer protection against mitochondrial stress, it seems likely that there is at least a convergence of pathways to maintain mitochondrial integrity.

Fig. 3.

Regulation of Parkin activity through complex formation. a Parkin, PINK1 and DJ-1 form a complex required for mitochondrial integrity. b Parkin may be part of an SCF-like complex. In a typical complex (left), Skp1 is the adaptor for the F-box protein, which then recruits the substrate. Cullin-1 associates with the RING protein, Rbx1, which in turn recruits the E2. In the SCF-like complex, Parkin adopts the roles of both Skp1 and Rbx1 (right). c Parkin forms a complex with Hsp70 and CHIP

Parkin has also been proposed to be a component of a Skp1-Cullin-Fbox (SCF)-like complex [36] (Fig. 3b). Along with Cullin-1 and the F-box protein, hSel10, Parkin is reported to ubiquitinate and mediate the proteasomal degradation of Cyclin E. The SCF or Cullin-RING ligases typically provide a Cullin-1 scaffold with a RING protein for efficient ubiquitin transfer, Skp1, to act as an adaptor for the F-box, [92, 93], while the F-box protein recruits the substrate. In the scenario proposed by Staropoli et al. [36], Parkin would provide the RING component and act as the adaptor to recruit the F-box, as Parkin does not co-immunoprecipitate either Skp1 or the RING protein, Rbx1 (Fig. 3b).

A third potential Parkin complex is the interplay between Parkin, the chaperone Hsp70, and the U-box protein CHIP (carboxy terminus of Hsc70 interacting protein) [94]. Both Hsp70 and Parkin bind the putative Parkin substrate, unfolded PaelR. Hsp70 apparently inhibits Parkin E3 ligase activity towards PaelR, allowing it to refold. However, CHIP inhibits Hsp70 activity [95, 96] and displaces Hsp70 from PaelR, thus allowing the ubiquitination of PaelR [94] (Fig. 3c). This intriguing complex provides an elegant model for competing refolding and degradation. It is worth noting that a stable endogenous Parkin complex, obtained from brains, elutes at 110 kDa, which is too small for the complexes described above [97]. It remains to be determined which of these complexes, if any, predominate in regulating Parkin activity towards different substrates.

Negative regulators

14-3-3η is a member of the 14-3-3 chaperone like protein family. Myc-14-3-3η was found to bind FLAG-Parkin in a manner that inhibited Parkin ubiquitination activity both towards itself and towards synphilin-1 [98]. This inhibitory effect is relieved by overexpression of α-synuclein, and is specific to the η isoform of 14-3-3. Sato et al. [98] speculate that the inhibition of Parkin autoubiquitination allows for levels of Parkin to remain constant, and this is relieved in the presence of increasing concentrations of α-synuclein, allowing activation of Parkin.

BAG5 (bcl-2-associated athanogene 5) is a member of the BAG-domain-containing family that are thought to confer resistance to cell death, and has been shown to interact with the chaperone Hsp70, and Parkin [99]. This interaction has been found to enhance the degeneration of dopaminergic neurons through inhibition of both the ubiquitin ligase activity of Parkin and the chaperone activity of Hsp70, leading to the sequestration of Parkin into protein aggregates and blocking the neuroprotective function of Parkin when the proteasome system is overwhelmed [99]. Recently, BAG5 has also been described as a negative regulator of CHIP [100], which is thought to function together with Parkin (see above) [94], raising the intriguing possibility of BAG5 being a major negative regulator of the refolding-or-degradation pathway administered by the CHIP/Hsp70/Parkin complex. It is possible that 14-3-3η and BAG5 both secure wild-type Parkin in an autoinhibited conformation, or that they block additional protein binding sites for E2s or putative substrates (Fig. 4a). Further studies will be required to determine the molecular details of such interactions and the mechanism(s) of the apparent negative regulation of Parkin by external binding partners.

Fig. 4.

Regulation of Parkin through external interactions. Negative regulation of Parkin through interaction with 14-3-3η and BAG5 (left). The inactivation could be achieved by stabilising an autoinhibited conformation of Parkin, or by obscuring substrate and/or E2-binding sites. Positive regulation of Parkin through interaction with SUMO-1, Eps15 and Endophilin-1A (right). The activation could be achieved through release of the autoinhibition, or through stabilising an E2- or substrate-bound form

Positive regulators

The small ubiquitin-like protein SUMO-1 is reported to associate non-covalently with Parkin [101]. Parkin is found predominantly in the cytosol [102], but is also present in the nucleus [31, 69, 103, 104]. The association with SUMO-1 enhances the nuclear translocation of Parkin and increases its autoubiquitination [101]. Interestingly, although an increase in ubiquitinated Parkin was observed, no appreciable difference in Parkin protein levels was detected upon overexpression of SUMO-1 [101], suggesting that the increased autoubiquitination activity does not necessarily result in proteasome-dependent degradation of Parkin. Parkin is implicated in gene transcription as a repressor of p53 [104], but this activity is independent of its ligase function, suggesting a functional requirement for subcellular localisation of Parkin outside substrate ubiquitination.

In our recent study describing the autoinhibition of Parkin [65], we observed that the addition of Ubl-domain-interacting proteins, Eps15 [43] and Endophilin-1A [105], released the autoinhibition allowing Parkin self-ubiquitination to occur. This raises the possibility of effectors of Parkin function and also suggests a potential for substrate-mediated activation [106]. However, it is as yet unclear whether such activation of Parkin includes a greater activity towards substrates rather than autoubiquitination, or if effector-led activation would result in proteasomal degradation of autoubiquitinated Parkin. Chew et al. [27] found that only the C-terminal IBR-RING2 of Parkin is capable of ubiquitinating synphilin-1, even though other Parkin constructs, including full-length MBP-tagged Parkin, are competent for self-ubiquitination. This suggests that there are differences between self and substrate ubiquitination. It is possible that positive regulators of Parkin may simply relieve an autoinhibited conformation, or enhance E2 or substrate binding (Fig. 4b). The possibility of allosteric regulation of Parkin by effectors is an exciting area for further exploration.

Components of the ubiquitination/deubiquitination machinery

Parkin is reported to act with multiple ubiquitin-conjugating E2 enzymes, including UbcH7, UbcH8 and the K63-linkage specific dimer, UbcH13/Uev1a. It is not yet clear what parameters influence the pairing up of Parkin with a particular E2, and deconvoluting this remains an important question as E2s often influence the choice of ubiquitin modification [107, 108].

Another important mechanism for regulating reversible ubiquitination is deubiquitination, mediated by the deubiquitinatinases (DUBs) [109]. Very recently, ataxin-3 has been identified as a binding partner of Parkin [110]. Ataxin-3 is a polyglutamine repeat protein involved in the neurodegenerative disorder, Machado-Joseph disease [111]. It contains 3 ubiquitin-interacting motifs (UIMs), a polyglutamine region which is expanded in the disease state, and an N-terminal Josephin deubiquitinase domain. Durcan et al. [110] report bimodal binding between Parkin and ataxin-3, with one interaction between the Ubl domain of Parkin and the UIMs of ataxin-3, and a second interaction between the IBR-RING2 of Parkin and the Josephin domain of ataxin-3. Polyubiquitination (but not mono-) of Parkin enhances this interaction, and ataxin-3 deubiquitinates Parkin. Interestingly, the polyglutamine expansion mutant of ataxin-3 promotes the removal of Parkin via the autophagy pathway and reduces Parkin levels in vivo [110, 112]. More recently, the same group have described an intriguing regulation of Parkin autoubiquitination that is coupled to the E2 Ubc7 [113]. Within a complex comprising ataxin-3, Parkin and Ubc7, the charged ubiquitin thioester-bound to the E2 is conjugated onto ataxin-3 rather than Parkin (Fig. 5). Furthermore, the authors find that, in contrast to their earlier report, ataxin-3 is unable to deubiquitinate chains that have already been assembled on Parkin, and is only able to deubiquitinate Parkin when Parkin is actively autoubiquitinating. These studies suggest an intricate interplay between E2, E3 and DUB enzymes that will require further studies to resolve.

Fig. 5.

Regulation of Parkin by the deubiquitinase ataxin-3. Parkin autoubiquitination is inhibited when Ataxin-3 interacts with Parkin via the E2. Subsequent charged ubiquitin molecules are transferred to Ataxin-3 but not to Parkin

There are multiple other interactors of Parkin that as yet are not identified as having regulatory effects on Parkin activity, including Klokin1 and Ambra1, both implicated in mitochondrial homeostasis [114, 115]. Intriguingly, the proliferating cell nuclear antigen (PCNA) is also reported to interact with Parkin and therefore suggests a role for Parkin in promoting DNA repair [116]. These proteins may or may not turn out to have regulatory activity regarding choice of substrate, subcellular localisation, or mediating additional Parkin–protein interactions.

Regulation of Parkin through posttranslational modifications

The regulation of Parkin function via posttranslational modifications has been recently and extensively reviewed [117, 118]. Below, we summarise the current understanding of the different posttranslational modifications and their effects specifically on Parkin E3 ligase activity.

Phosphorylation

Phosphorylation of Parkin has been observed by casein kinase-1, protein kinase A, protein kinase C, cyclin-dependent kinase 5 (cdk5), c-Abl, and, as discussed above, potentially by PINK1 [83, 84, 119–123]. Interestingly, with the exception of the proposed phosphorylation by PINK1, modification of Parkin by phosphorylation appears to inactivate Parkin in each of the systems studied. The interaction between Parkin and Endophilin-1A (described above in “Positive regulators”) appears to be mediated by phosphorylation, in that full-length Parkin only efficiently interacts with Endophilin-1A in conditions that favour phosphorylation [105]. Indeed, it is tempting to speculate that phosphorylation of Parkin may enhance its interactions with substrates and/or E2s, although formal evidence for such activities is still required.

S-nitrosylation

S-nitrosylation is a reversible posttranslational modification involving the covalent attachment of a nitric oxide (NO) group to the thiol of a cysteine residue to form an S-nitrosothiol species. NO is a free radical signalling molecule generated by various nitric oxide synthases, and there is increasing evidence that nitrosative stress is a key player in protein misfolding and neurodegeneration [124–128]. Two studies report the S-nitrosylation of Parkin, although they differ in the location of the modified cysteine(s) [129, 130]. However, Parkin is highly cysteine-rich, and co-ordinates 8 Zinc atoms, all of which are required for correct Parkin folding [13]. Therefore, it seems plausible that S-nitrosylation of any of the zinc-coordinating cysteines would impact on Parkin function. Chung et al. [129] report a decrease in Parkin activity, while Yao et al. [130] report an increase. However, the modulation of function seen by Yao et al. is biphasic, with an initial increase in self-ubiquitination, followed 6 h later by a decrease. Thus, the two reports seem to converge on an eventual loss of function of Parkin.

Oxidation

As with S-nitrosylation, oxidative stress is also implicated in neurodegeneration. A recent study identifies the oxidation of cysteines within Parkin (spread throughout the cysteine-rich RING0, RING1, IBR and RING2 domains) via comprehensive mass spectrometry [131]. Also in common with S-nitrosylation, the authors report a biphasic initial increase in Parkin activity, followed by a loss of Parkin itself. Coupled with an earlier report that Parkin is inactivated following oxidative stress [132], it seems likely that oxidation of Parkin leads to eventual loss of Parkin function, most likely through changes in solubility, aggregation and degradation.

Ubiquitination

Perhaps the most perplexing of all the modifications of Parkin is that it is still not clear whether Parkin autoubiquitination is something that occurs in the normal situation in cells. Certainly, it is clear that, when Parkin is heavily ubiquitinated in cells, it is degraded in a proteasome-dependent manner (e.g. [65, 131]), thus effectively inactivating Parkin. However, it is also plausible that ubiquitination of Parkin is a regulatory mechanism required for efficient ubiquitination of substrates [133].

Neddylation

A recent report suggests that Parkin can be conjugated with the ubiquitin-like protein NEDD8 [134]. Through a series of truncations and lysine-to-arginine point mutations, the authors attempt to locate a specific lysine residue that is neddylated. As their results reveal that all the mutants are neddylated, they conclude that the sites for neddylation occur throughout the sequence. However, it is interesting to note that the authors do not state whether they mutated the lysine residue in the Myc tag N-terminal to each Parkin species. Certainly, in the case of ubiquitination, Parkin has been shown to ubiquitinate N-terminal tags that are fused, in vitro and in cells [65, 135], and therefore it is possible that a similar mechanism is occurring in the case of the observed neddylation. The authors also report that neddylation enhances the interaction of Parkin with UbcH8 (but not UbcH7) and with the putative substrate, the p38 subunit of aminoacyl transferase, and that the outcome of these interactions is an increased ubiquitin ligase activity [134]. Indeed, a subsequent paper also reports increased Parkin E3 ligase activity upon neddylation, coupled with a stabilisation of PINK1 [136]. It will be interesting to see if Parkin activity, in common with the cullin-RING ligases [92, 93], is regulated by neddylation.

Perspective

Studies reporting effects on Parkin function can often be conflicting. This is likely due to the wide array of different techniques, cell types, organisms, expression systems, levels of purity of recombinant proteins, N- or C-terminal tags, temporal aspects of experiments and so on. However, one apparently unifying thread is that Parkin-related Parkinson’s disease is due to a loss of Parkin function. One common mechanism for a loss of Parkin function appears to be degradation of Parkin itself by the proteasome. By this means, pathogenic Parkin mutations that are ‘activating’ ultimately result in loss of Parkin; the cases of S-nitrosylation and oxidation which appear to be both activating and inactivating are likely due to an initial increase in Parkin autoubiquitination, followed by a subsequent destruction of Parkin by the proteasome, which is therefore inactivated. However, in the case of regulatory binding partners, the opposite appears to be the case, whereby binding to SUMO-1, Eps15 or Endophilin-1A enhances Parkin activity in vitro, but not degradation in cells. A summary of Parkin functions and regulations is shown in Fig. 6.

Fig. 6.

Summary of regulation of Parkin function. Parkin is activated by effectors to carry out its normal functions in ubiquitination of proteasomal substrates, and to promote mitophagy. When Parkin is compromised by pathogenic mutations, or oxidation/nitrosylation, Parkin is inactivated through proteasomal degradation

One mechanistic question that has yet to be addressed is whether, after a certain threshold of self-ubiquitination has been achieved, Parkin retains activity towards itself, substrates or both. Indeed, separating the effects of Parkin autoubiquitination and substrate ubiquitination remains a challenge, exacerbated by a lack of consensus on which criteria denote a genuine substrate. Certainly, one breakthrough would be the identification of a ‘model’ Parkin substrate that could be ubiquitinated by wild-type Parkin in vitro. Such a tool would allow for a detailed mechanistic understanding of how Parkin transfers ubiquitin to a substrate, and how that transfer is regulated. In the cellular environment, it will be important to determine what the native state of Parkin is, what conformational changes it undergoes and whether substrates are direct or indirect. Is it possible that Parkin alone directly recognises such a diverse array of substrates? Or is it likely to have indirect substrates mediated through complexes such as the SCF/CRLs? Do effectors act as substrate-binding adaptors? Or does Parkin perform a generic protein quality control role on one level, with a more specific subset of substrates on another? In addition, how Parkin interacts with different E2s and under what conditions each E2 is selected is a challenge for understanding the multiple modifications apparently catalysed by Parkin. Finally, a molecular understanding of how Parkin is recruited into various complexes and locations will be critical to our knowledge of Parkin dysfunction. Understanding how Parkin is regulated in terms of activity, protein levels, substrate recognition and subcellular localisation will be key to restoring compromised Parkin activity.