Structure of Parkin Reveals Mechanisms for Ubiquitin Ligase Activation

See allHide authors and affiliations

Science  21 Jun 2013:
Vol. 340, Issue 6139, pp. 1451-1455
DOI: 10.1126/science.1237908

Parkin Enhanced?

Inactivation of parkin, an E3 ubiquitin ligase, is responsible for a familial form of Parkinson's disease and may be involved in sporadic forms as well. Trempe et al. (p. 1451, published online 9 May) present the crystal structure of full-length parkin in an autoinhibited configuration. Guided by the structure, mutations were designed that activated parkin both in vitro and in cells. Because parkin is neuroprotective, the structure provides a framework for enhancing parkin function as a therapeutic strategy in Parkinson's disease.


Mutations in the PARK2 (parkin) gene are responsible for an autosomal recessive form of Parkinson’s disease. The parkin protein is a RING-in-between-RING E3 ubiquitin ligase that exhibits low basal activity. We describe the crystal structure of full-length rat parkin. The structure shows parkin in an autoinhibited state and provides insight into how it is activated. RING0 occludes the ubiquitin acceptor site Cys431 in RING2, whereas a repressor element of parkin binds RING1 and blocks its E2-binding site. Mutations that disrupted these inhibitory interactions activated parkin both in vitro and in cells. Parkin is neuroprotective, and these findings may provide a structural and mechanistic framework for enhancing parkin activity.

Parkinson’s disease (PD) is a common neurodegenerative disease characterized by severe motor and nonmotor symptoms. More than 120 mutations in PARK2 (parkin) have been shown to cause autosomal recessive PD, with point mutations found in every domain of the protein (13). The parkin protein is a RING-in-between-RING (RBR) E3 ubiquitin ligase (4) that exhibits low basal activity in vitro (5). Parkin has been implicated in a range of biological processes, including autophagy of damaged mitochondria (mitophagy), cell survival pathways, and vesicle trafficking (6, 7). The activity of the PD-associated mitochondrial kinase PINK1 is required for parkin activation in mitophagy (812). Parkin consists of a ubiquitin-like (Ubl) domain and a 60–amino acid linker followed by RING0, a zinc finger unique to parkin (13), and three additional zinc finger domains characteristic of the RBR family (Fig. 1A). RBR ligases such as parkin, HOIL-1L interacting protein (HOIP), and Ariadne/HHARI use a RING-HECT hybrid mechanism (1416) whereby ubiquitin forms a thioester intermediate with a cysteine side chain in RING2 before being transferred to the primary amino group of a substrate to form an isopeptide bond. How RBR ligases accomplish these two reactions is unknown.

Fig. 1 Structure of parkin.

(A) Primary structure and domains of parkin. (B) Cartoon representation of the full-length structure. The repressing REP α helix is surrounded by unstructured regions (yellow dashed lines). Zinc atoms are shown as gray spheres. (C) Schematic representation of the full-length structure showing the occluded E2 binding and catalytic sites. (D) SAXS data were obtained for both full-length and R0-RBR (amino acids 141 to 465) parkin in solution. The data were fitted to single chains in each crystal structure. (E) Mapping of human PD mutations onto the structure. Position 240 is alanine in rat parkin. (F) Topology of zinc finger domains in parkin. Cysteine or histidine zinc ligands are shown in blue circles connected by lines from the N terminus to the C terminus. The RING1 domain displays the cross-brace motif characteristic of RING domains. RING0 displays a hairpin topology, whereas IBR and RING2 display sequential topologies. Single-letter abbreviations for amino acid residues: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; Y, Tyr.

We determined the crystal structure of rat parkin, obtaining a low-resolution structure of the full-length protein and a 2.8 Å resolution structure of a C-terminal fragment (amino acids 141 to 465) (Fig. 1B, figs. S1 and S2, and table S1). Collectively, the structures show that parkin forms a rigid core of the RING0, RING1, and RING2 domains stabilized by several key hydrophobic interactions (Fig. 1C). This core functions as a scaffold for interactions with the remaining domains of parkin. The in-between-RING (IBR) domain is attached through a flexible linker; its position varies by up to 13 Å between different chains in the structure of the C-terminal fragment (fig. S2B). The N-terminal Ubl domain is bound to RING1 through the hydrophobic surface centered around Ile44, consistent with the reported interaction between the Ubl and the C terminus (5). This surface of the Ubl also binds ubiquitin-interacting motifs (1719) (UIMs) and SH3 domains (20), implying that the Ubl must dissociate from RING1 in order to bind these partners. The long linker following the Ubl is not visible in the crystal structure, but a fragment of the second linker between the IBR and RING2 domains is visible. This fragment contains an α helix bound to RING1, called REP (repressor element of parkin). Small-angle x-ray scattering (SAXS) analysis revealed that both full-length parkin and R0-RBR parkin adopt the same conformation in solution as in the crystals (Fig. 1D and fig. S3).

The structure explains a large number of mutations associated with PD (Fig. 1E). The mutations can be grouped into three classes: mutations such as Arg42 → Pro, Lys211 → Asn, Cys212 → Tyr, Cys289 → Gly, Thr351 → Pro, and Cys441 → Arg that disrupt zinc coordination or protein folding; mutations such as Gly430 → Asp and Cys431 → Phe that affect the catalytic site; and mutations that affect protein contacts. For example, in the Ubl domain, the mutation Arg33 → Gln could interact with the adjacent IBR domain. The position and flexibility of the IBR is probably important for parkin’s function: The mutation Gly328 → Glu in the hinge between RING1 and IBR causes PD, possibly via loss of conformational flexibility. In the crystal structure, the Gly328 backbone torsion angles (ϕ ~ 90°; ψ ~ 20°) are incompatible with glutamate. Some mutations such as Asp280 → Asn and Thr415 → Asn may affect aspects of parkin function other than its stability, catalysis, or interdomain contacts.

The parkin zinc finger domains fold with distinct topologies. RING1 is the only domain with a classical C3HC4 cross-brace zinc coordination topology typical of other RING fingers (Fig. 1F and fig. S4). Modeling based on the structures of five E2-RING complexes reveals the E2 binding site on RING1 (fig. S5). The other domains are atypical. RING0 binds two coordinated zinc atoms at each extremity of the domain with a hairpin. The IBR domain is similar to the reported nuclear magnetic resonance (NMR) structure of the isolated domain (21) (fig. S2D). The RING2 domain has the same topology as the IBR domain and coordinates two zinc atoms in a sequential fashion (Fig. 1F). Two of the zinc coordinating residues in RING2, Cys457 and His461, are not conserved in other RBR proteins (fig. S6) but are well conserved in parkin (fig. S7). Moreover, these two residues are part of a helix that mediates multiple interactions with a hydrophobic groove in RING0 (Fig. 2A).

Fig. 2 Autoinhibitory interactions in parkin.

(A) Interface between RING0 and RING2. The side chain of Phe463 inserts into a hydrophobic groove in RING0 formed by Phe146. (B) Surface view of the interface between RING0 and RING2 showing occlusion of the active site, Cys431 (orange). (C) Interface of RING1 (cyan), Ubl (red), and the REP motif (yellow). The predicted UbcH7 binding site on RING1 is colored violet. The binding site was determined by structural overlay with the cCbl-RING:UbcH7 crystal structure (PDB ID: 1fbv) (29). (D) Sequence alignment of the REP sequence in parkin from different species. Invariant residues are shown as white letters on a black background. (E) Autoubiquitination assays with mutated GST-parkin constructs (WT, wild type). Reactions were performed with 1 μM GST-parkin and 2 μM UbcH7 and incubated for 3 hours at 37°C. (F) Activation of parkin E3 ligase activity by mutagenesis. Reactions were performed as described above. Autoubiquitination was followed by the appearance of high–molecular weight smear in both the stacking and resolving sections of the blots, and by the loss of unmodified parkin in the Ponceau blot. (G) Autoubiquitination assays with untagged parkin constructs. Reactions were performed as described above for 2 hours. (H) Summary of autoubiquitination activity of parkin deletions (top) and missense mutations mapped onto the parkin structure and color-coded by activity (bottom). See fig. S8 for details.

A number of structural features suggest that the observed conformation is autoinhibited. First, the catalytic Cys431 is occluded by the RING0-RING2 interactions (Fig. 2B). Second, modeling of the E2 binding site on RING1 suggests that it is located beside the Ubl binding site and blocked by the REP linker (Fig. 2C and fig. S5). The REP is well conserved across parkin orthologs (Fig. 2D) and forms a two-turn helix that is held in place by Trp403, whose indole group inserts into a pocket in RING1 in part formed by the side chain of Arg234 (Fig. 2C), a residue mutated to glutamine in PD. Third, parkin must undergo a conformational change to mediate the transfer of ubiquitin, because the active-site cysteines on the E2 and parkin are ~50 Å apart in the E2-parkin models (fig. S5B).

To assess the role of the Ubl domain and the RING0-RING2 and REP-RING1 interactions on parkin activity, we tested a series of deletions and single-site mutants in autoubiquitination assays. Wild-type parkin displayed weak but detectable activity, monitored by the appearance of high–molecular weight polyubiquitinated proteins and the loss of unmodified parkin (Fig. 2, E and F). This activity required E1, E2, and ubiquitin (Fig. 2F). Autoubiquitination assays with the K0 ubiquitin variant, where all lysines were mutated to arginines, confirmed that the high–molecular weight species were polyubiquitin chains (fig. S8A). Mutation of different unliganded cysteines in parkin confirmed the unique role of Cys431 in the catalysis of ubiquitination (fig. S8B) (14). The majority of PD mutations tested (Lys161 → Asn, Met192 → Val, Asp280 → Asn, Gly328 → Glu, Glu444 → Gln) reduced activity (Fig. 2E). Mutation of His433, located next to the active site, Cys431 (Fig. 2A), strongly reduced activity (Fig. 2E); this finding suggests that His433 could assist Cys431 in catalyzing ubiquitin transfer from a charged E2 or in the downstream acyl transfer reaction onto a substrate lysine. Mutation of residues in the predicted E2 binding site (Ala240 → Arg, Thr242 → Ala, Asp243 → Ala) all abolished parkin autoubiquitination activity (Fig. 2E). The Ubl has been reported to mediate parkin autoinhibition (5), and its position next to the E2 binding site and the REP linker also suggested this (Fig. 2C). Surprisingly, deletion of the Ubl, either alone or together with the following linker, had little effect on parkin activity, regardless of whether the proteins were tagged with glutathione-S-transferase (GST) at their N terminus (Fig. 2, F and G, and fig. S9). In contrast, the additional deletion of RING0 massively derepressed parkin activity (Fig. 2, F and G, and fig. S9). Further supporting the role of RING0 in autoinhibition, point mutations in RING0 (Phe146 → Ala) or RING2 (Phe463 → Ala) that had been predicted to disrupt the RING0-RING2 interface both increased parkin activity (Fig. 2, F and G, and fig. S9). Moreover, mutation of the zinc ligand Cys457 also activated parkin, confirming that the integrity of the C-terminal helix in RING2 is required to maintain the autoinhibition of parkin (fig. S8). To test the role of the REP-RING1 interface, we mutated Trp403 to alanine. The resulting W403A mutant also showed increased autoubiquitination activity, thereby confirming that the REP effectively represses the ligase activity of parkin. Thus, the REP and RING0 domains play a preeminent role in repressing parkin ligase activity through their interactions with RING1 and RING2, respectively (Fig. 2H).

We then asked whether the REP regulated parkin activity by affecting E2 binding to RING1. We used a ubiquitin-loaded E2 (E2∼Ub) discharging assay to measure the rate of ubiquitin transfer from the E2 enzyme UbcH7. Wild-type parkin was moderately efficient at removing ubiquitin from UbcH7, concurrent with the slow formation of high–molecular weight polyubiquitinated protein (Fig. 3, A and B, and fig. S10). As expected, the W403A mutation increased the UbcH7~Ub discharging activity and polyubiquitin chain formation. The direct effect of the W403A mutant on E2 binding was measured using NMR spectroscopy. Full-length and R0-RBR parkin bound UbcH7 weakly and led to line broadening and signal loss from UbcH7 residues at the E2-RING1 interface (Fig. 3, C to F, and fig. S11). Mutation of Trp403 in the REP strongly increased binding and led to complete loss of signals from the interfacial residues. The double W403A-C431S (Cys431 → Ser) mutant also strongly bound UbcH7 but was completely impaired in UbcH7~Ub discharging (Fig. 3, A and B), which suggests that UbcH7 binding is upstream of Cys431-mediated catalysis of ubiquitination.

Fig. 3 Activation of parkin through increased UbcH7 binding.

(A) UbcH7~Ub discharge assays with wild-type GST-parkin and mutants. Reactions were stopped with sample buffer containing tris(2-carboxyethyl)phosphine (TCEP) to reduce disulfide bonds but keep thioester bonds intact. Products were visualized using immunoblotting for polyubiquitin chains and silver staining for UbcH7. See fig. S10A for details. (B) Densitometric quantification of UbcH7~Ub discharging. Similar results were obtained under varying conditions (fig. S10, B and C). (C) NMR analysis of parkin binding to 15N-labeled UbcH7. Changes in the HSQC spectra of UbcH7 (150 μM) were caused by the addition of wild-type R0-RBR (70 μM) or wild-type R0-RBR–W403A (70 μM). Peaks that undergo the most broadening (signal loss) correspond to the RING-binding site in UbcH7. (D) Quantification of amide NMR signal loss for Phe63 as a function of different parkin construct concentrations. Spectra were normalized for dilution; plots show the number of scans and the peak intensity lost relative to the free UbcH7 spectrum. (E) Quantification of amide NMR signal loss for UbcH7 resonances after addition of 216 μM R0-RBR parkin. (F) Model of parkin bound to UbcH7. The REP occludes the UbcH7-binding site on RING1. Note the long distance between the active-site cysteines in UbcH7 (Cys86) and parkin (Cys431).

We next addressed the role of the REP in the regulation of parkin activity in vivo. We expressed green fluorescent protein (GFP)–tagged parkin in HeLa cells and used time-lapse microscopy to examine the kinetics of wild-type and mutant GFP-parkin recruitment to mitochondria after their depolarization with carbonyl cyanide m-chlorophenyl hydrazone (CCCP), a proton ionophore (22). Recruitment of wild-type parkin to mitochondria began 30 min after CCCP treatment, with half of the cells showing recruitment at 45 min, whereas no recruitment of the catalytically inactive mutant was observed during the first hour (Fig. 4, A to C, fig. S12, and movies S1 and S2) as previously reported (22, 23). In contrast, the W403A mutant was recruited earlier than wild-type parkin with a lead time of ~10 min, implying that releasing autoinhibition by the REP predisposes parkin for mitochondrial recruitment. Nonetheless, recruitment of the W403A mutant to mitochondria was still PINK1-dependent (Fig. 4, D and E), similar to wild-type parkin (1012). Thus, parkin activation appears to be a multistep process involving relocalization and release of ubiquitination activity, which are likely regulated through phosphorylation (10, 2426), cysteine modification (27, 28), and ligand binding to the Ubl domain (5, 17, 20).

Fig. 4 Effects of parkin activation in cells.

(A) Parkin recruitment to mitochondria upon membrane depolarization. HeLa cells were transduced with a baculovirus expressing a mitochondrial marker fused to red fluorescent protein (CellLight mitochondria-RFP, Invitrogen) and transfected with GFP-parkin wild-type (WT), W403A, or C431S plasmids. Cells were treated with CCCP and visualized by time-lapse microscopy. Recruitment can be visualized by the appearance of punctate GFP fluorescence (arrows) superposed on the mitochondrial RFP (omitted for clarity; see fig. S12). Scale bar, 20 μm. (B) Quantification of GFP-parkin recruitment to mitochondria. The percentage of cells showing recruitment of GFP-parkin to mitochondria was determined every 5 min. Error bars denote SEM (N = 4). *P < 0.05 (Student t test). (C) HeLa cells were transfected with GFP-parkin WT, W403A, or C431S plasmids for 24 hours. Whole-cell lysate proteins (20 μg) were immunoblotted for GFP and voltage-dependent anion channel (VDAC). (D) HeLa cells were transfected with nontargeting (NT) or PINK1 small interfering RNAs (siRNAs) for 24 hours and treated with CCCP for 3 hours. Whole-cell lysate proteins were immunoblotted for PINK1 and actin. (E) Parkin W403A recruitment to mitochondria is PINK1-dependent. HeLa cells were transfected with NT or PINK1 siRNAs before transfection of GFP-parkin WT or W403A (green) plasmids. Cells were treated with 20 μM CCCP for 3 hours and immunostained for Tom20 (red). Scale bar, 20 μm.

Whereas many PD-linked mutations reduce or abolish parkin activity, our findings reveal that it is possible to derepress parkin activity both in vitro and in vivo. Because parkin is neuroprotective in a number of PD models (7), the structure-based mechanism of activation presented here provides a potential framework to enhance parkin activity for therapeutics in PD.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S12

Table S1

Movies S1 and S2

References (3040)

References and Notes

  1. Acknowledgments: We thank C. Arrowsmith for the His6-UbcH7 construct, the Canadian Light Source and CHESS for data acquisition, and group members for insights and helpful discussion. We acknowledge support from Parkinson Society Canada and from Canadian Institutes of Health Research grants MOP-62714 (E.A.F.) and MOP-14219 (K.G.). Coordinates and structure factors of the R0-RBR and full-length parkin crystal structures were deposited in the Protein Data Bank under accession codes 4K7D and 4K95.
View Abstract

Navigate This Article