Structural and Functional Impact of Parkinson Disease-Associated Mutations in the E3 Ubiquitin Ligase Parkin

Abstract

Mutations in the PARKIN/PARK2 gene that result in loss-of-function of the encoded, neuroprotective E3 ubiquitin ligase Parkin cause recessive, familial early-onset Parkinson disease. As an increasing number of rare Parkin sequence variants with unclear pathogenicity are identified, structure-function analyses will be critical to determine their disease relevance. Depending on the specific amino acids affected, several distinct pathomechanisms can result in loss of Parkin function. These include disruption of overall Parkin folding, decreased solubility and protein aggregation. However pathogenic effects can also result from misregulation of Parkin auto-inhibition and of its enzymatic functions. In addition, interference of binding to co-enzymes, substrates and adaptor proteins can affect its catalytic activity too. Herein, we have performed a comprehensive structural and functional analysis of 21 PARK2 missense mutations distributed across the individual protein domains. Using this combined approach we were able to pinpoint some of the pathogenic mechanisms of individual sequence variants. Similar analyses will be critical in gaining a complete understanding of the complex regulations and enzymatic functions of Parkin. These studies will not only highlight the important residues, but will also help to develop novel therapeutics aimed at activating and preserving an active, neuroprotective form of Parkin.

INTRODUCTION

PARK2 (PARKIN; MIM# 602544) gene mutations are the most common cause of familial, recessive early-onset Parkinson disease (EOPD) (Kitada et al. 1998; Puschmann 2013). To date, over 170 mutations (including point mutations and exonic rearrangements) have been identified, however, the pathogenic relevance remains unclear for several of these sequence variants (Corti et al. 2011). The encoded Parkin protein is an E3 ligase that mediates the transfer of the small modifier Ubiquitin (Ub) to substrate proteins (Wenzel et al. 2011). Parkin can catalyze several different types of Ub modifications with distinct biological functions and numerous unrelated substrate proteins have been identified so far (Walden and Martinez-Torres 2012). Thus, the exact function of Parkin enzymatic activities and in particular its role in the pathogenesis of EOPD remains unclear. However, over the last few years, the Parkin/PINK1-dependent mitophagy pathway has been subject of intense research. Upon mitochondrial depolarization, the kinase PINK1 (mutations in the PINK1 gene also cause EOPD) activates Parkin and enables its translocation to damaged mitochondria (Geisler et al. 2010; Matsuda et al. 2010; Narendra et al. 2010b; Vives-Bauza et al. 2010). Subsequently Parkin labels damaged mitochondria with Ub to mark their degradation. Strikingly, EOPD mutations in both PINK1 and PARK2 result in failure of this protective mitochondrial quality control system. Of note, specific Parkin mutations appear to disrupt this sequential process at distinct steps, offering an opportunity to dissect the pathway through structure-function analyses.

First partial crystal structures of the Parkin protein show a ‘closed’, inactive conformation mediated through several intra-molecular interactions among the individual domains (Riley et al. 2013; Trempe et al. 2013; Wauer and Komander 2013). Auto-inhibition had been suggested before (Chaugule et al. 2011) and is consistent with the notoriously weak enzymatic activity of Parkin under steady-state conditions. PINK1 has been shown to phosphorylate a conserved serine residue (Ser65) in both, Parkin (Kondapalli et al. 2012; Shiba-Fukushima et al. 2012; Iguchi et al. 2013) and Ub (Kane et al. 2014; Kazlauskaite et al. 2014b; Koyano et al. 2014; Ordureau et al. 2014; Zhang et al. 2014) to fully activate Parkin enzymatic function during mitophagy. Using computational modeling and molecular dynamics simulations (MDS), we have recently established a complete structure for human Parkin at an all-atom resolution and developed a conformational pathway of activation (Caulfield et al. 2014). PINK1 phosphorylation initiates a cascade of structural changes that result in sequential release of auto-inhibitory self-interactions and eventually liberation of Parkin enzymatic activities.

Given the complex activation process of Parkin protein, mutations can affect its enzymatic function through several distinct pathomechanisms. First, PARK2 variants can result in reduced solubility and enhanced aggregation thereby affecting protein folding, stability and functions. Second, PARK2 mutations can affect the activation process through either enhanced auto-inhibition, failure in opening conformations or even premature release of its intra-molecular interactions. As Parkin is a preferred substrate for itself, hyperactivation of the E3 ligase might result in enhanced turnover and thus loss-of function. Third, mutations can also affect its ability to bind E2 co-enzymes, Ub moieties, substrates or adaptor proteins, which would negatively impact its translocation to mitochondria or the Ub transfer. In order to assess the pathogenicity of variants, a critical understanding of Parkin activation process, the role of its individual functional domains and of its enzymatic activity(ies) is required. We present a comprehensive structural and functional analysis of PARK2 missense mutations that provides a framework for the dissection of the underlying pathomechanisms. At the same time, these studies will be important to guide small molecule design that aims to activate Parkin or stabilize Parkin in its activated form.

MATERIALS AND METHODS

Nomenclature for the description of sequence variants

We have used the consensus GenBank RefSeq accession NM_004562.2 to number all variants within the PARK2 gene and protein. The DNA mutation numbering system we use is based on this cDNA sequence. Nucleotide numbering uses +1 as the A of the ATG translation initiation codon in the reference sequence, with the initiation codon as amino acid number 1.

Public databases for mutations

The Exome variant sever (http://evs.gs.washington.edu/EVS/) was used for the general minor allele frequencies of PARK2 mutations and cBioPortal (http://www.cbioportal.org/public-portal/) for the frequencies of PARK2 mutations in the cancer (Table 1).

Table 1.

Overview of genetic, clinical and pathological findings on PARK2 mutations

	Mutation			Clinical frequency					Prediction tools
	protein: aa subst.	cDNA: bp subst.	Ref SNP #	# of patients1	homozygous	Freq. heterozygous2	LB Pathol.3	Freq. Cancer DB4	Poly phen−25	SIFT6	Mut. Taster7
	WT	c.1-1498	-		−	-	-	-	0.000	1.00
UBL	p.R33Q	c.98G>A	rs147757966	4	−	0.01	ND	NA	0.317	0.70
	p.P37L	c.110C>T	rs148990138	0	−	0.01	ND	0.10	0.948	0.03
	p.R42P	c.125G>C	NA	>10	+	0	ND	NA	0.967	0.01
L	p.T83A8	c.247A>G	rs141825163	0	−	0.01	ND	NA	0.000	0.52
RING0	p.C150G9	c.448T>G	NA	1	+	0	ND	NA	0.999	0.00
	p.K161N	c.483A>T	rs137853057	2	−	0	ND	NA	1.000	0.00
	p.W183A10	c.547T>G c.548G>C	NA	NA	NA	0	ND	NA	0.999	0.00
	p.K211N11	c.633A>T	rs137853060	>5	+	0	1+13	0.21	0.780	0.00
RING1	p.T240R	c.719C>G	rs137853054	1	+	0	ND	NA	0.286	0.02
	p.R256C	c.766C>T	rs150562946	1	−	0.05	ND	NA	0.964	0.02
	p.Y267H	c.799T>C	rs114696251	0	−	0	ND	NA	0.994	0.00
	p.R275W	c.823C>T	rs34424986	>40	+	0.29	3+14,15,2−15	0.4	1.000	0.00
IBR	p.G328E	c.983G>A	NA	0	−	0	ND	NA	0.998	0.09
	p.R334C	c.1000C>T	rs199657839	>5	+	0	ND	NA	0.564	0.12
REP	p.A398T12	c.1192G>A	NA	1	−	0	ND	0.7	0.870	0.12
	p.W403A10	c.1207T>Gc.1208G>C	NA	NA	NA	0	ND	NA	1.000	0.07
RING2	p.T415N	c.1244C>A	NA	>5	+	0	ND	NA	0.863	0.00
	p.G430D	c.1289G>A	rs191486604	>10	+	0.03	1+15, 1−15	NA	1.000	0.00
	p.C431F	c.1292G>T	rs397514694	3	+	0	ND	NA	1.000	0.00
	p.C431S10	c.1292G>C	NA	NA	NA	0	ND	NA	1.000	0.00
	p.F463A10	c.1387T>G c.1388T>C	NA	NA	NA	0	ND	NA	0.998	0.00

Table summarizes features of potentially pathogenic PARK2 mutations. NA = not available. ND = not determined. GenBank RefSeq accession NM_004562.2 was used to number all variants within PARK2 gene and protein.

¹Number of patients with homozygous or compound heterozygous mutations. Single heterozygous mutation carriers are not included.

²Minor allele frequency [%] in 4300 European American from Exome variant server (EVS; http://evs.gs.washington.edu/EVS/). No homozygous carriers listed for these mutations.

³Indicates presence (+) or absence (−) of Lewy body pathology in one or more autopsy cases (as indicated by number).

⁴Minor allele frequency [%] according to cBioPortal for Cancer Genomics (http://www.cbioportal.org/public-portal/)

⁵Prediction using PolyPhen-2 (http://genetics.bwh.harvard.edu/pph/): red = probably damaging; yellow = possibly damaging; green = benign. In addition probability scores are given. Note Parkin wild type is 0.000.

⁶Prediction using SIFT (http://sift.jcvi.org/): red = damaging; green = tolerated. In addition probability scores are given. Note Parkin wild type is 1.00.

⁷Prediction using Mutation Taster (http://www.mutationtaster.org/): red = disease causing; green = polymorphism. No probability scores available.

⁸This variant was found only in control subjects.

⁹Unpublished data.

¹⁰Artificial mutation.

¹¹One patient had 3 variants including homozygous p.K211N mutation and heterozygous Exon 3 deletion.

¹²One patient had two mutations p.L272I/p.A398T but p.L272I might not be pathogenic; 6 out of 200 (carrier frequency 3%, allele frequency 1.5%) Japanese control subjects had heterozygous p.L272I. EVS doesn’t have data of p.L272I.

¹³Pathological data from (van de Warrenburg et al. 2001)

¹⁴Pathological data from (Farrer et al. 2001)

¹⁵Pathological data from (Doherty et al. 2013)

Prediction of the functional consequences of variants on Parkin protein sequence

We used software programs available on the internet, namely PolyPhen-2 with HumVar-trained model (http://genetics.bwh.harvard.edu/pph2/), SIFT (http://sift.jcvi.org/) and Mutation Taster (http://www.mutationtaster.org/). The results are shown in Table 1.

Zone Equilibration of Mutants

To determine structural alterations induced by Parkin missense mutations, we used Zone Equilibration of Mutants (ZEMu) (Dourado and Flores 2014), implemented in MacroMoleculeBuilder (MMB). MMB is a multiscale, Internal Coordinate Mechanics (ICM) code (Flores et al. 2010; Flores et al. 2011), which models the 3D structure and dynamics of macromolecules. In brief, ZEMu establishes a flexibility zone (5 residues) in torsional coordinates, centered on the mutation of interest, leaving the remainder of the protein rigid and fixed. ZEMu then computes the dynamics in a larger, enclosing physics zone within which electrostatic and van der Waals forces are active. One can then calculate difference in free energies of unfolding (ΔG) between the thus-equilibrated wild type and mutant structures (ΔΔG = ΔGmut − ΔGwt) using the FoldX potential (Guerois et al. 2002). The root mean square error associated with ZEMu when the FoldX potential is used is 1.54 kcal/mol for all mutants in a full test set of 1245 mutants (1–15 simultaneous mutations) in 65 co-crystals, and 1.34 kcal/mol for a subset consisting only of single mutations.

Modeling of Parkin Structures, Modifications and Mutations

The modeling of full-length (465 amino acid) human Parkin protein encoded by the PARK2 gene, has been previously described (Caulfield et al. 2014). Parkin has several conserved domains: UBL (residues 1–76), linker region (residues 77–140), RING0 (or UPD) domain (residues 141–216), RING1 domain (residues 217–328), IBR domain (residues 329–378), REP region (residues 379–410), and RING2 domain (residues 411–465). Mutations p.R42P, p.K161N, p.K211N, p.T240R, p.R275W, p.R334C, p.A398T, and p.G430D were modeled on the phospho-Ser65 Parkin protein structure (Caulfield et al. 2014) using Maestro (Maestro-9.1 2010), or Visual Molecular Dynamics (VMD) (Humphrey et al. 1996).

Molecular Dynamics Methods

Molecular dynamics (MD) was carried out using OPLS2005, Amber or Charmm force fields with well documented parameters (Krieger et al. 2009; Maestro-9.1 2010; Krieger et al. 2012; Schrödinger 2012; Schrödinger 2013), which were further tested with the current release of NAnoscale Molecular Dynamics 2 engine (Brooks et al. 1983; Cornell et al. 1994; D.A. Pearlman 1995; MacKerell et al. 1998; Krawetz 1999). The protein with hydrogens consists of ~7,100 atoms. In all cases, we neutralized with counter-ions, and then created a solvent with 150 mM Na+ Cl- to recreate physiological strength. TIP3P water molecules were added around the protein at a depth of 15–18 Å from the edge of the molecule depending upon the side (Jorgensen et al. 1983). Our protocol has been previously described in the literature (Caulfield et al. 2011). Solvated protein simulations consist of a box with between >1.15 x 10⁵ atoms including proteins, counter-ions, solvent ions, and solvent waters. Simulations were carried out using the particle mesh Ewald technique with repeating boundary conditions with a 9 Å nonbonded cut-off, using SHAKE with a 2-fs timestep. Pre-equilibration was started with 100,000 steps of minimization followed by 10000 ps of heating under MD, with the atomic positions of protein fixed. Then, two cycles of minimization (100000 steps each) and heating (2000 ps) were carried out with restraints of 10 and 5 kcal/(mol·Å²) applied to all protein atoms. Next, 50000-steps of minimization were performed with solute restraints reduced by 1 kcal/(mol·Å²). Then, 1000 ps of unrestrained MD were carried out, and the system was slowly heated from 1 to 310 K. The production MD runs were carried out with constant pressure boundary conditions (relaxation time of 1.0 ps). A constant temperature of 300 K was maintained using the Berendsen weak-coupling algorithm with a time constant of 1.0 ps. SHAKE constraints were applied to all hydrogens to eliminate X-H vibrations, which yielded a longer simulation time step (2 fs). Methods for equilibration and production run protocols have been published (Reblova et al. 2006; Reblova et al. 2007; Caulfield and Medina-Franco 2011; Caulfield and Devkota 2012). Equilibration was determined from a flattening of RMSD over time after an interval of >20ns. Production run simulations for each mutant were carried out for >120 ns. Translational and rotational center-of-mass motions were initially removed. Periodically, simulations were interrupted to have the center-of-mass removed again by a subtraction of velocities to account for the “flying ice-cube” effect (Cheatham III and Young 2001). Following the simulation, the individual frames were superposed back to the origin, to remove rotation and translation effects. Analyses scripts were generated or derived from VMD repositories (Humphrey et al. 1996).

Cloning and mutagenesis

FLAG-Parkin and pEGFP-Myc-Parkin wild type have been described before (Geisler et al. 2010). Mutant Parkin was cloned using site-directed mutagenesis and sequence verified using BigDye Terminator v.3.1 and an ABI 3100 Genetic Analyzer (Applied Biosystems).

Cell culture

Human HeLa were obtained from the ATCC (American Type Culture Collection and were maintained in DMEM plus 10% FBS at 37°C under humidified conditions and 5% CO₂.

High Content Imaging

To objectively quantify a defined Parkin relocalization, we employed automated High Content Imaging (HCI) that has recently been described (Caulfield et al. 2014; Fiesel et al. 2014). In brief, cells were seeded with 4000 cells/well in 96-well imaging plates (BD Biosciences) and allowed to attach overnight. Cells were transfected with empty vector, EGFP-Myc-Parkin wild type or mutants using Lipofectamine 2000 (Invitrogen). 48h after transfection cells were treated with 10μM CCCP for 0, 1, 2, or 4h. Cells were washed 1x in PBS and fixed for 20min in 4% paraformaldehyde. Nuclei were stained with Hoechst 33342 (1: 5000, Invitrogen) for 10min and cells washed twice in PBS. Plates were imaged on a BD Pathway 855 with a 20x objective using a 3x3 montage (no gaps) with laser autofocus every second frame. Raw images were processed using the build-in AttoVision V1.6 software. Regions of interest (ROIs) were defined as nucleus and cytoplasm using the build-in ‘RING - 2 outputs’ segmentation for the Hoechst channel after applying a shading algorithm. As a measure of Parkin relocalization, the ratio of GFP signal intensity in the cytosol/nucleus was calculated. To exclude non-transfected cells and to ensure comparable transfection levels among analyzed cells, only ROIs with a GFP signal at least 30% higher than background were taken into consideration.

Immunostaining

HeLa cells were plated onto glass coverslips coated with poly-D-lysine (Sigma), fixed with 4% paraformaldehyde and permeabilized with 1% Triton-X-100 in PBS. Cells were incubated with primary anti-TOM20 (1:2000, Proteintech group 11802-1-AP) and anti-p62 antibodies (1:500, BD Biosciences 610832) followed by incubation with secondary antibodies anti-mouse IgG Alexa Fluor-647 and anti-rabbit Alexa Fluor-568 (Molecular Probes) diluted 1:1000. Nuclei were stained with Hoechst 33342 (1:5000). Coverslips were mounted onto slides using fluorescent mounting medium (Dako). Confocal fluorescent images were taken with an AxioObserver microscope equipped with an ApoTome Imaging System (Zeiss).

Oxyester analysis

HeLa cells were transfected with FLAG-Parkin p.C431S variants using Lipofectamine 2000 according to manufacturer’s protocol and medium was replaced 4h later. The next day, cells were treated with 10μM CCCP for 0, 1, 2, 4, or 16h. Cells were harvested in preheated (95°C) SDS lysis buffer (50mM Tris pH7.6, 150mM NaCl, 1%SDS). Lysates were homogenized by 10 strokes through a 23G needle. Protein concentration was determined by use of bicinchoninic acid (Pierce Biotechnology). To verify the band shift by oxyester formation, aliquots of lysates were treated with or without NaOH (100mM final) for 1h at 37°C. NaOH was neutralized by addition of equal amounts of HCl before samples were run on 8–16% Tris-Glycine gels and transferred onto polyvinylidene fluoride membranes (Millipore). Membranes were incubated with anti-FLAG antibody (1:100,000, Sigma F3165) overnight at 4°C followed by HRP-conjugated anti-mouse secondary antibodies (1:15,000; Jackson ImmunoResearch 115-035-003). Bands were visualized with ImmobilonWestern Chemiluminescent HRP Substrate (Millipore) using a LAS-3000 Imager (Fuji).

Statistical analysis

Statistical analysis was performed with one-way ANOVA followed by Tukey’s post-hoc test. Error bars indicate S.E.M..

RESULTS

Mutations in Parkin

For our structure-function analyses, we selected 21 PARK2 missense mutations distributed across the individual domains of the Parkin protein (Figure 1A/B). Clinical, genetic and pathological findings that have been reported for these sequence variants can be found in Table 1. Although the majority of these substitutions (13) have been identified in patients with EOPD, their pathogenicity remains unclear. Out of these, nine were found in homozygous carriers, and four from patients with compound heterozygous mutations. With some overlap, six mutations were listed on the Exome variant sever with a significant heterozygous frequency. Interestingly, four mutations were also identified from the Cancer database (cBioPortal). Several reports already suggested a role for loss of Parkin in tumorigenesis in addition to PD (Xu et al. 2014).

Figure 1. Organization and structure of human PARK2 gene and Parkin protein.

A) PARK2 gene/mRNA structure. PARK2 exons are drawn to scale and size in base pairs (bp) is indicated. Exons are color-coded according to the functional domains of the Parkin protein as described in (B). A bi-directional promoter of 65bp drives expression of PARK2 and the Parkin-co-regulated gene (PACRG). B) Parkin protein 2D structure. The individual functional domains are drawn to scale and are color-coded [red = Ubiquitin-like (UBL) domain; gray = linker region; green = RING0 (R0)/Unique Parkin domain (UPD); blue = RING1 (R1); purple = in-between-RING (IBR); yellow = Repressor element of Parkin (REP); pink = RING2 (R2)]. The amino acids (aa) are indicated below, the analyzed missense mutations at the top. At the bottom, Zn²⁺ coordinating residues of the RING finger domains are given. Residues Ser65 and Cys431, which are important for Parkin activation and enzymatic function are highlighted in green. C) A surface presentation is given for the closed, inactive Parkin structure with domains labeled as in (B). D) Ribbon presentation of the closed Parkin conformation. Dark gray spheres indicate coordinated Zn²⁺ ions. Mutated residues are highlighted as sticks with carbon in gray, nitrogen in blue, oxygen in red, and sulfur in yellow.

Crystal structures of Parkin (Riley et al. 2013; Trempe et al. 2013; Wauer and Komander 2013) showed a ‘closed’, inactive conformation in which the intertwined domains literally fold Parkin back onto itself (Figure 1C/D). We recently suggested an ‘unfolding’ pathway in which the intra-molecular interactions of Parkin are sequentially released upon PINK1-dependent phosphorylation (Caulfield et al. 2014). Missense mutations in Parkin appear to differently interfere with certain aspects of auto-inhibition release and/or enzymatic activity. In addition to disease-associated Parkin mutations, we also included artificial variants (p.W183A, p.W403A, p.C431S and p.F463A) to shed more light onto the complex regulation of Parkin.

Local conformational and energetic changes in Parkin caused by mutations

We performed a structural and energetic analysis of mutations across three crystal structures of human [PDB IDs: 4I1H (Riley et al. 2013) and 4BM9 (Wauer and Komander 2013)] and rat [PDB ID: 4K95 (Trempe et al. 2013)] Parkin. Qualitative inspection of the intra-molecular interactions surrounding the mutation sites (referred to here as structure gazing, Supp. Figure S1), was complemented by quantitative predictions of free binding energies (ΔΔG) with Zone Equilibration of Mutants (ZEMu) (Dourado and Flores 2014) (Table 2). Overall, structure gazing and energetic analyses allowed a detailed profiling and revealed significant conformational changes for the majority of Parkin mutations.

Table 2.

Overview of local structural changes induced by Parkin mutations

	PDB ID	4BM9, human aa 142–465		4I1H, human aa 141–465		4K95, rat aa 1–465			comment
	protein: aa subst.	struct. gazing	ZEMu	struct. gazing	ZEMu	struct. gazing	ZEMu	change1
	WT	0	0	0	0	0	0	−
UBL	p.R33Q	-	~	-	~	>0	1.7	+	decrease in UBL stability but no probable effect on UBL-IBR interaction
	p.P37L	-	-	-	-	>0	1.0	+	decrease in UBL stability
	p.R42P	-	-	-	-	≫0	6.4	+	introduction of a β-sheet breaker and disruption of the UBL fold
L	p.T83A	-	-	-	-	-	-		-
RING0	p.C150G	≫0	ND	≫0	ND	≫0	ND	++	loss of Zn2+ coordination and complete disruption of the RING0 fold
	p.K161N	0	−0.2	0	−0.6	0	−4.2	−	no effect on RING0 stability, but loss of charge in putative PBS
	p.W183A	≫0	2.5	≫0	6.6	≫0	3.3	++	loss of van der Waals interactions within RING0-RING1-RING2 pocket
	p.K211N	0	−1.4	0	1.2	0	0.5	−	no effect on RING0 stability, but loss of charge in putative PBS
RING1	p.T240R2	-	-	-	-	>0	2.0	+	RING1-UBL/REP-RING1 interfaces destabilized; E2 binding affected
	p.R256C	>=0	1.8	>=0	−0.6	>=0	−0.4	−	no significant effect
	p.Y267H	>0	3.7	>0	4.7	>0	11.3	++	loss of van der Waals interactions and destabilization of the RING1 fold
	p.R275W	≫ 0	7.8	≫0	1.7	≫0	2.8	++	disrupts charge distribution and local rearrangements in the RING1-IBR interface
IBR	p.G328E	≫0	2.9	≫0	3.3	≫0	2.5	+	loss of flexibility in the loop region and disturbance of backbone arrangement
	p.R334C	0	1.4	0	0.5	0	0.0	−	no significant effect
REP	p.A398T	>0	1.4	>0	0.9	>0	4.5	+	decrease of inter-domain stability, with possible effect on the release of REP
	p.W403A	≫0	3.6	≫0	1.2	>0	−1.0	+	decrease of inter-domain stability, with possible effect on the release of REP
RING2	p.T415N	0	0.5	0	1.0	0	0	−	no significant effect
	p.G430D	≫0	4.9	≫0	8.5	≫0	3.8	++	disturbance of the backbone arrangement and desolvation of active site
	p.C431F	>0	16.0	0	0.4	>0	14.6	++	high local destabilization and loss of the catalytic center residue
	p.C431S	0	−1.8	0	0.7	0	0.9	−	forms a more stable oxyester with Ub, but otherwise very similar to WT
	p.F463A	≫0	4.1	≫0	4.4	≫0	1.1	++	loss of van der Waals interactions within RING0-RING1-RING2 pocket

Table summarizes features of potentially pathogenic PARK2 mutations. NA = not available. ND = not determined. Genebank RefSeq accession NM_004562.2 was used to number all variants within PARKIN gene and protein. Mutations were introduced into the respective structures and energetically local rearrangements around the mutation site were determined. Shown are qualitative evaluations of free binding energies for Parkin structures (ΔΔG = ΔGmut − ΔGwt) from structure gazing and quantitative predictions from ZEMu in kcal/mol. For residues Lys211, and Thr240 the backbone was maintained rigid so as not to perturb the adjacent Zn finger motifs. Calculations were performed across all three crystals whenever possible. A minus (−) indicates residues not present/resolved in the crystals as e.g. both human structures lack the N-terminus. The residue Thr83 was not resolved in any of the crystals. ND = not determined as ZEMu was not yet validated for metal-complexes as in the case of the Zn2+ coordinating residue C150 within RING0.

¹(−) indicates no/minor change, whereas (+) or (++) indicate moderate to major changes, respectively. Empty box indicates that the respective mutation has not been analyzed in the assay.

²As Thr240 is not resolved in any of the human crystals, calculations were performed only with the rat Parkin structure. However, human Thr240 corresponds to Ala240 in rat (PDB ID 4K95).

All three mutations (p.R33Q, p.P37L and p.R42P) in the UBL domain of Parkin decreased stability to some extent, consistent with a less stable Ub fold (Safadi et al. 2011). p.T83A could not be analyzed, as the linker region is not resolved in any of the available Parkin crystals. Structure gazing suggested that p.C150G would result in gross distortion of the overall RING0 fold due to its key role in coordination of Zn²⁺ ions, which prevented prediction of ΔΔG. While for the RING0 mutations p.K161N and p.K211N no significant changes were observed besides a loss of charge in the putative phospho-binding site (Wauer and Komander 2013), p.W183A introduced considerable alterations of the local structure. Within the E2 binding site in RING1, p.T240R significantly affected RING1-UBL binding, while p.R256C showed only minor structural variation. Due to their contribution to the domain-specific fold, both p.Y267H and p.R275W resulted in significant predicted conformational alterations. Within the IBR, p.G328E was predicted to reduce flexibility of the region, whereas p.R334C showed no effect on domain stability. p.A398T and p.W403A, mutations within the REP region, which blocks the E2 binding site in RING1, resulted in moderate decrease in local stability, consistent with a partial release of Parkin auto-inhibition (Riley et al. 2013; Trempe et al. 2013; Wauer and Komander 2013). Within RING2, p.T415N showed only minor structural alterations, while p.G430D introduced major ΔΔG changes. In contrast to the catalytic center mutation p.C431F, p.C431S, which traps Ub as a more stable oxyester, showed no alterations thus confirming its suitability to monitor Ub charging of Parkin. Last, p.F463A showed considerable changes in the local structure of RING2, consistent with the proposed disruption of auto-inhibition that could result in a more active form of Parkin (Riley et al. 2013; Trempe et al. 2013; Wauer and Komander 2013).

Global conformational changes in Parkin caused by mutations

To investigate conformational defects of Parkin mutations upon activation by PINK1 phosphorylation, we performed MDS of >120 ns and calculations over time (Figure 2). We selected variants from defined protein domains of Parkin with a focus on mutations with strong genetic evidence (see Table 1). Included were p.R42P, p.K161N, p.K211N, p.T240R, p.R275W, p.R334C, p.A398T, and p.G430D. Structures illustrating local changes of the mutations before and after MDS can be found in Supp. Figure S2A–H, respectively. Supp. Movies S1–S8 show zoom-ins from the entire MDS over time (compare Supp. Movies S1–S8A for wild type Parkin with S1–S8B for the respective mutations). Structures illustrating global changes of the entire Parkin molecule after MDS can be found in Supp. Figure S3. We analyzed the root mean square deviation (RMSD) to assess direct structural changes of the mutations and local rearrangements within 6Å of the mutant site. In addition, we investigated global conformational changes through calculations of the root mean square fluctuation (RMSF) across the entire structure.

Figure 2. Molecular Dynamics Simulations of Parkin mutations.

Individual Parkin mutations were chosen from each of the defined domains for modeling and analyses by free MDS over time. A) p.R42P, B) p.K161N, C) p.K211N, D) p.T240R, E) p.R275W, F) p.R334C, G) p.A398T, and H) p.G430D. Two graphs are shown for each of the mutations. Left graph: RMSD measurements over time are given for the mutated residue (black) and for the adjacent amino acids within 6Å of the mutant (gray), which demonstrate local structural changes. For zooms into the mutation sites before and after MDS see Supp. Figure S2A–H. For the entire MDS of the mutation region see Supp. Movies S1A–S8A for wild type Parkin and Supp. Movies S1B–S8B for the respective variants. Right graph: RMSF measurements indicating global structural changes are given for the mutant residue (black) and wild type (gray) as a reference. Fluctuations of individual residues over the entire length of Parkin were analyzed over 120ns MDS. RMSF peaks indicate residues with most mobility over the course of the simulation as an average. Dashed lines separate the individual domains of Parkin, which are labeled on the top of the figure. An arrow points at the respective mutant residue. For wild type Parkin, peaks of maximal fluctuation showed RMSF of around or above 2.3Å. Mutant Parkin variants introduced peak shifts that are indicative of structural rearrangements. For a view of the entire Parkin molecules after 120 ns MDS see Supp. Figure S3.

Although ZEMu predicted a larger ΔΔG for p.R42P, we did not observe a significant increase in local RMSD (Figure 2A). Rather, mutational effects appeared to propagate from the UBL into particularly the linker domain and the C-terminal part. Mutations in the UBL have been shown to reduce the stability of the Ub fold upon e.g. increasing temperature compared to wild type (Safadi et al. 2011) and have been suggested to release the auto-inhibition of Parkin as judged by auto-ubiquitination (Chaugule et al. 2011). Additional calculation of the RMSD of the entire UBL domain (Supp. Figure S4A) and of the distance between the UBL and IBR domains (Supp. Figure S4B) as a measure during Parkin activation (Caulfield et al. 2014) corroborated these findings. Consistent with ZEMu predictions, MDS did not reveal any significant structural alterations for p.K161N or p.K211N, besides a charge change in a pocket in RING0 (Figure 2B/C).

As suggested by the energetic profile of p.T240R, MDS over time revealed major conformational rearrangements in RING1 following an abrupt switch in orientation of the mutant residue (Figure 2D). As the mutant residue in the E2 binding site flipped towards the REP region, it engaged in a stabilized interaction, which is indicated by lower RMSF peaks compared to wild type. Although the primary defect of p.R275W is first observed in RING1, as it expands to accommodate the more bulky amino acid, MDS revealed the propagation of significant structural changes throughout the entire Parkin molecule (Figure 2E). p.R275W showed a rapid RMS increase while the surrounding region expanded over 120 ns to converge with changes induced by the mutation. Several RMSF peaks indicate further global structural effects including rearrangements between RING1 and REP, exposure of RING0, and re-orientation of RING2.

While ZEMu predicted no structural change for p.R334C in the IBR region, the substitution of a basic with a polar amino acid side chain appeared to considerably perturb the inter-domain interactions over time (Figure 2F). RMSD for the mutant residue and the local region showed an increase from the initial conformation. The RMSF peaks revealed most fluctuation to occur in the N-terminal region with additional changes at the C-terminus. Consistent with ZEMu calculations, p.A398T showed a moderate change in RMS followed by a rapid return to conformational space similar to the original conformation (Figure 2G). Nevertheless, major fluctuations were found in the UBL, linker region, and RING0 as well as in the C-terminal part.

p.G430D is adjacently positioned to the active site in RING2 and introduces a negative charge that likely changes the environment important for thiol-ester formation between Cys431 of Parkin and Gly76 of Ub. MDS corroborated the effect predicted by ZEMu (RMS > 3Å immediately and throughout simulation), however, did not reveal any major structural rearrangements supporting a simple electrostatic disruption of the active site.

High Content Imaging of the relocalization of Parkin missense mutations

As a first functional read-out, we investigated the ability of EGFP-Parkin variants to relocalize from the cytosol to perinuclear regions using high content imaging (HCI). This objective assay relies on the ratio of GFP signal intensities between cytosol and nucleus (Figure 3A) and correlates with activation and translocation of Parkin to damaged mitochondria (Caulfield et al. 2014; Fiesel et al. 2014). In total, we quantified >5.000 cells per Parkin mutant at 0, 2, and 4h after treatment with the mitochondrial uncoupler carbonyl cyanide m-chlorophenyl hydrazone (CCCP). A heat-map provides the average GFP ratios and gives an initial overview of the functionality of Parkin mutants (Figure 3B). To quantify defects in Parkin activation, percentages of cells were calculated that showed a defined GFP threshold of > 1.8, which corresponds to the average ratio of wild type Parkin after 2h CCCP treatment. Significances were calculated to wild type Parkin and to the inactive p.C431S mutation at 0, 2, and 4h of CCCP treatment (Figure 3C–E, respectively).

Figure 3. Cell-based High Content Imaging of Parkin mutations.

HeLa cells transiently expressing wild type or mutant EGFP-Parkin were left untreated (0h) or treated with the uncoupler CCCP for 2h or 4h. Cells expressing EGFP only or the catalytically inactive EGFP-Parkin p.C431S mutation served as negative controls. A) Principle of HCI Parkin translocation assay. Images have been acquired using automated microscopy and show mitochondria (TOM20) in red, EGFP-Parkin in green and nuclei (Hoechst) in blue. Quantification of Parkin re-localization is assessed by measuring maximal intensity in a cytoplasmic ring around the nucleus divided by the mean intensity of the nuclear GFP signal. Cytoplasmic ring and inner nuclear regions are schematically shown in the merge image. The GFP ratio is given for each GFP-positive cell in white as an example and reflects Parkin translocation to mitochondria. Untransfected cells or cells below threshold expression of Parkin are marked with a white asterisk and have been excluded from the analysis. B) The average GFP ratio of wild type and mutant Parkin is shown as a heat map. C-E) Bar graphs give the percentages of cells with a defined mitochondrial translocation of Parkin at 0h (C), 2h (D), and 4h (E) of CCCP treatment. A GFP threshold of >1.8 was chosen that corresponds to the average ratio of Parkin wild type translocation after 2h CCCP treatment. Significance levels to wild type (black asterisks, above bars) and p.C431S (gray asterisks, within bars) are provided (n>4 plates with at least 4 wells per condition, one-way ANOVA, Tukey’s post-hoc, p < 0.0001, F=147.3, ns = not significant, *** p < 0.0005). F) Given are representative merge images at higher resolution that show co-localization of GFP-Parkin (green) and mitochondria (anti-TOM20, red). Nuclei (Hoechst) are shown in blue. Scale bars correspond to 10μm. For images of the individual channels at all time points (0, 2, and 4h CCCP), see Supp. Figure S5A, B, and C, respectively.

Upon CCCP treatment, p.R33Q, p.P37L, p.T83A, p.A398T, and p.W403A were indistinguishable from wild type Parkin, but highly significant to p.C431S. While Parkin wild type showed no major differences between 2h or 4h CCCP, a clear delay, but not a complete disruption of Parkin translocation was observed for p.R42P, p.K161N, p.W183A, p.Y267H, p.R275W, and p.F463A. However, Parkin p.K211N, p.T240R, and the RING2 mutations p.T415N, p.G430D, and p.C431F showed strongly reduced if not abrogated relocalization. The accuracy of automated measurements was corroborated by co-staining with mitochondrial markers at higher resolution (Figure 3F). Additional co-staining for the autophagy adaptor protein p62 that is co-recruited to mitochondria in a Parkin-dependent manner further validated the findings (Supp. Figure S5).

Of note, none of the Parkin mutants showed mitochondrial localization without CCCP treatment. However, p.C150G and to a lesser extent also p.W183A and p.R275W showed significant differences compared to wild type Parkin at 0h CCCP at least in the highly sensitive HCI assay. This likely reflects their aggregation, which leads to a less even distribution under basal conditions, and not their co-localization with mitochondria as confirmed with higher resolution microscopy (Figure 3F and Supp. Figure S5 for single channel images).

Ubiquitin charging of Parkin missense mutations

As a second functional read-out for activity, we analyzed Ub charging of Parkin using double mutant constructs that harbored the p.C431S variant, which stabilizes the Ub moiety on the catalytic center. Ub-charged Parkin p.C431S appears as an 8kD band shift that is only sensitive to chemical cleavage by NaOH. None of the Parkin double mutants had noticeable Ub charging in the absence of CCCP (Figure 4A/B). Upon treatment, Parkin p.R256C, p.A398T, p.W403A and p.F463A showed levels comparable to ‘wild type’ Parkin (i.e. p.C431S only). In contrast, several mutants displayed significantly diminished or abrogated Ub charging even after long times of CCCP incubation (16h). These included the UBL domain mutants, p.R33Q, p.P37L, and p.R42P, the linker mutant p.T83A as well as the RING0 mutations p.K161N, p.W183A, and p.K211N. The RING1 mutants p.T240R, p.Y267H, and p.R275W had particularly low Ub charging, if any at all.

Figure 4. Activation of Parkin E3 Ub ligase functions.

A) HeLa cells were transfected with FLAG-Parkin p.C431S (herein referred to ‘wild type, WT’) or the indicated double mutants to monitor the Ub charging and thus activation of Parkin. Cells were left untreated or incubated with CCCP for 16h to assess maximal levels of C431S-Ub oxyester formation. Representative western blots show unmodifed Parkin (closed arrowhead) as well as Ub-charged Parkin that appears as a band shift (open arrowhead) and is sensitive to NaOH treatment. Graph represents quantification of Parkin-Ub/total Parkin levels normalized to Parkin ‘WT’ after 16h CCCP from several independent experiments (n≥3, one-way ANOVA, Tukey’s post-hoc, *** p < 0.0005, ** p < 0.005). B) To assess released auto-inhibition and possibly accelerated Ub charging of presumed activating Parkin mutants, HeLa cells were transfected and treated for shorter time point with CCCP (i.e. 1, 2, or 4h). Representative Western blots are shown as in (A). Graphs represent quantification of Parkin-Ub/total Parkin levels after 1, 2, or 4h CCCP treatment normalized to Parkin ‘WT’ after 16h CCCP from several independent experiments (n≥3, one-way ANOVA, Tukey’s post-hoc, *** p < 0.0005, ** p < 0.005, * p < 0.05). C-D) HCI analysis was performed upon 1h CCCP treatment. Shown are (C) the average GFP ratio and (D) the percentage of defined Parkin relocalization in GFP-positive cells (n>2 plates with at least 3 wells per condition, one-way ANOVA, Tukey’s post-hoc, p < 0.0001, F=6.635, ns = not significant, *** p < 0.0005).

Although we did not find appreciable amounts of Ub charging for the supposedly activating Parkin mutations in the absence of CCCP, we performed a time course experiment using shorter incubations with the depolarizer to assess a potentially accelerated Ub-charging (Figure 4B). Indeed for the mutations p.A398T, p.W403A, and p.F463A we found relatively more Ub-charged Parkin compared to p.C431S alone. This is in agreement with their predicted ‘looser’ structure that is caused by a partial release of some of the auto-inhibitory self-interactions. In analogy to the Ub charging analysis, we performed HCI after only 1h CCCP treatment to monitor a potentially accelerated Parkin relocalisation. However, consistent with the initial imaging, we did not find accelerated redistribution of potentially activating Parkin mutants after 1h of CCCP treatment (Figure 4C and D).

DISCUSSION

As we enter the era of high-throughput genetic analyses, we will identify more and more rare sequence variants of unknown significance in disease-related genes. thus structure-function analyses will be critical to inform geneticists, biologists, and clinicians. A comparison between different available programs that are used to predict the pathogenicity of Parkin variants (Table 1) revealed only modest consistency among these tools. However, given the complexity of regulations and enzymatic activities, multiple mechanisms may result in loss of Parkin functions, at steady state as well as upon its activation. It is thus crucial to study the structural and functional alterations of mutants by several parameters, which will not only help to pinpoint the individual defects, but at the same time, will contribute to a better understanding of Parkin functions. Our findings together with previously published data are summarized in Table 3.

Table 3.

Overview of structural and functional changes induced by Parkin mutations

	Mutation	auto-Ub assay w/o PINK11/2		substr.-Ub3	in vitro E3 ligase activity with PINK14			cellular assays with CCCP5		structural analyses of ‘closed’ & ‘open’ Parkin5
	protein: aa subst.	in vitro1	in vitro2	in cellulo	E2 dis-charge	Miro1-Ub	free chains	Ub charge	relocalization	Xray ZEMu	MDS RMSD	MDS RMSF
	WT							−6		−	−	−
UBL	p.R33Q									+
	p.P37L									+
	p.R42P									+	−7	+
L	p.T83A
RING0	p.C150G								8	++
	p.K161N									−	−	−
	p.W183A									++
	p.K211N									−	−	−
RING1	p.T240R									+	++	−−
	p.R256C									−
	p.Y267H									++
	p.R275W									++	+	+
IBR	p.G328E									+
	p.R334C									−	+	+
REP	p.A398T									+	+	++
	p.W403A									+
RING2	p.T415N									−
	p.G430D			9						++	−	−
	p.C431F									++
	p.C431S									−
	p.F463A									++

Table summarizes functional defects of individual Parkin mutations from previous publications that overlapped by more than five missense variants in comparison with results obtained in the present study. Genebank RefSeq accession NM_004562.2 was used to number all variants within PARK2 gene and protein. Empty boxes indicate that the respective mutation has not been analyzed in the particular assay. Results of functional alterations are color-coded summary compared to wild type Parkin: blue = enhanced activity/function; green = similar/equal to wild type; yellow = reduced activity/function; red = loss of activity/function. Results of structural alterations are indicated by (−) no/minor change, (+) moderate change, (++) major change. (−−) indicates less fluctuation than wild type suggestive of an even more ‘closed’ conformation (only p.T240R).

¹In vitro auto-ubiquitination assay using recombinant MBP-Parkin purified from E. coli (Matsuda et al. 2006). Assay was not quantified in this study.

²In vitro auto-ubiquitination assay using HA-Parkin immunoprecipitated from COS7 cells (Hampe et al. 2006). Assay was not quantified in this study.

³In cellulo substrate-ubiquitination observed in stable MYC-Parkin expressing SH-SY5Y cells that overexpressed the aminoacyl-tRNA synthetase subunit p38 or Synphilin-1 together with HA-Ubiquitin (Sriram et al. 2005). Assay was not quantified in this study.

⁴E3 ligase activity of untagged Parkin purified from E. coli as measured by Ub discharging of the E2 co-enzyme, ubiquitination of the model substrate Miro1 and formation of free poly-Ub chains (Kazlauskaite et al. 2014a).

⁵Functional and structural alterations determined in this study.

⁶For Ub charging assays, p.C431S without additional mutation represents ‘wild type’ Parkin.

⁷p.R42P showed no increase in RMS for the mutation site or the surrounding 6Å region, however showed increased RMSD when the entire UBL domain was analyzed (see Supp. Figure S4).

⁸p.C150G forms cytosolic aggregates and therefore does not translocate to damaged mitochondria.

⁹p.G430D showed no E3 ligase activity towards the aminoacyl-tRNA synthetase subunit p38, but not towards Synphilin-1 (Sriram et al. 2005).

Herein, we performed a detailed structure-function analysis of 21 Parkin missense mutants in order to uncover the pathogenic mechanism of individual mutants. Structure gazing and energetic profiling with ZEMu across crystal structures, showed either no, only moderate or significant changes in free binding energies in the auto-inhibited conformation for about a third of the Parkin mutants each and provided initial insights into potential functional defects. Using MDS of a PINK1-activated Parkin molecule, we could further corroborate structural effects of mutations selected from each domain, but could also show that moderate local changes can translate into significant rearrangements upon activation and opening of Parkin. Analyses of the mutation sites and the surrounding regions (RMSD) as well as fluctuation of all Parkin residues over time (RMSF) revealed that local changes indeed propagate throughout the entire molecule.

In addition to the different structural calculations, we investigated the translocation of Parkin to damaged mitochondria by automated HCI and confocal microscopy as a functional readout for its activation. During the preparation of this manuscript another study reported a similar translocation assay (Ordureau et al. 2014), though in live cells over a time course of 120 min; our data are in excellent agreement. As an additional read-out for activation of Parkin in cells, we investigated Ub charging using the p.C431S variant that showed no significant structural changes compared to wild type Parkin. These analyses complement data from previous publications that have used alterative measures for Parkin activity such as E2 discharge and in vitro ubiquitination assays or cell-based monitoring of substrate modification and degradation. While earlier systematic reports from the pre-mitophagy era potentially analyzed Parkin in an auto-inhibited conformation (Sriram et al. 2005; Hampe et al. 2006; Matsuda et al. 2006), more recent in vitro studies were designed to incorporate functional PINK1 as an important activator of Parkin (Kazlauskaite et al. 2014a). While in vitro assays are always limited by the selection of the individual components, recent advances in quantitative mass spectrometry from cells will certainly help to better determinate the activity(ies) of Parkin.

Of note, mutations in the N-terminal UBL domain can affect the activation and enzymatic functions of Parkin. While a ΔUBL version of Parkin translocated to damaged mitochondria (Geisler et al. 2010; Sarraf et al. 2013; Ordureau et al. 2014), albeit slower than full-length, Ub-charging of a C431S double variant was not found (Zheng and Hunter 2013). Strikingly, this truncation exhibited significant E2 discharge and auto-ubiquitination activity, but could not catalyze Miro1 ubiquitination or form low molecular weight poly-ubiquitin chains (Kazlauskaite et al. 2014a). This suggests in addition to an auto-inhibitory role of the UBL domain (Chaugule et al. 2011) a regulatory function for the enzymatic activity of Parkin.

The RING0 mutations p.K161N and p.K211N most likely exhibit their dysfunctions as a result of a simple charge change with a consequent local hydrophobic restructuring and desolvation of the surrounding pocket. This site has been suggested to mediate phospho-binding (Wauer and Komander 2013) and thus could interfere with the binding of phosphorylated substrates on damaged mitochondria which might also explain the observed translocation delay.. While Ub charging was strongly reduced for both mutations, E2 discharge in vitro appeared comparable to wild type or only modestly reduced (Kazlauskaite et al. 2014a). Both p.K161N and p.K211N exhibited a strong reduction of free poly-Ub chains formation, but some preservation of substrate ubiquitination. Interestingly, neither of the mutations affected binding to phospho-Ub (Ordureau et al. 2014).

RING1 is the only canonical RING domain of Parkin and harbors an E2 enzyme binding site that is blocked in the inactive conformation by the REP, which folds back from a region between IBR and RING2. It has been shown before that p.T240R does not interact with the E2 enzyme UbcH7 (Shimura et al. 2000). In accordance with this, MDS indicated that it might prevent the release of the REP, a prerequisite for E2 binding (Caulfield et al. 2014), whereas the interaction surface for E2 binding stays largely intact. p.Y267H and p.R275W might exhibit their dysfunction through complete restructuring of RING1 and thus loss of E2 binding, consistent with decreased or eliminated Ub charging as well as strongly reduced E2 discharge and lack of E3 ligase activity in the presence of PINK1 (Kazlauskaite et al. 2014a). Interestingly, both mutants were able to translocate to mitochondria comparable to wild type Parkin, albeit somewhat slower, while p.R275W could not cluster damaged organelles or mediate their final degradation (Geisler et al. 2010; Narendra et al. 2010a; Okatsu et al. 2010).

In contrast to RING-type E3 ligases, but similar to HECT-type E3 ligases, Parkin and related enzymes have been shown to receive the Ub moiety on their active cysteine sites in RING2 before further transfer onto a lysine residue of a substrate protein in a HECT/RING hybrid mechanism (Wenzel et al. 2011). In general mutations of or around the active site C431 showed no translocation to damaged mitochondria. In addition these RING2 mutants strongly reduced, if not completely abrogated, E2 discharge and E3 ligase activity (Kazlauskaite et al. 2014a), with the exception of p.G430D which retained some functions. Our structural analyses supported the hypothesis that p.G430D most likely affects Parkin enzymatic functions by charge changes and consequent desolvation around the active site.

Several mutations have been predicted to release at least partially the auto-inhibitory self-interactions thereby potentially rendering Parkin more active a more active (Riley et al. 2013; Trempe et al. 2013; Wauer and Komander 2013). Parkin p.R33Q was recently found to be phosphorylated at significantly higher levels than wild type in vitro, which translated into increased E3 Ub ligase activity (Kazlauskaite et al. 2014a). However, we could not find enhanced Ub charging or accelerated translocation to damaged mitochondria similar to another study (Ordureau et al. 2014). Nevertheless, mutations in the REP region (p.A398T and p.W403A) and p.F463A in RING2 showed enhanced Ub charging at earlier time points, yet no accelerated mitochondrial translocation although both appear as inter-dependent events.

In summary, we provide a comprehensive structure-function analysis of several Parkin missense mutations. Going forward, combinations of various different assays will be key to exactly pinpoint the particular dysfunctions of individual Parkin variants. Moreover, these studies will guide structure-based drug design for small molecule Parkin activators. While some mutations simply cause misfolding, insolubility and aggregation of Parkin though to varying degrees (Cookson et al. 2003; Gu et al. 2003; Henn et al. 2005; Sriram et al. 2005; Schlehe et al. 2008) as in the case of p.C150G, others appear to disrupt individual aspects of its activation and enzymatic functions. Accumulating data might suggest that Parkin exhibits distinct enzymatic activities that can be uncoupled from each other. Along these lines, diverse ubiquitinations of Parkin itself and of its substrates have been observed including mono-Ub as well as free and attached poly-Ub chains of different topologies. In addition, E2-independent ubiquitinations have been proposed for Parkin (Chew et al. 2011) as well as PINK1-dependent, but C431-independent, modifications of Parkin have been identified (Ordureau et al. 2014). In addition, Parkin might self-interact and potentially multimerize (Lazarou et al. 2013) and/or activate itself in trans (Zhang et al. 2014). Studying mutants also in the presence of wild type and in different combinations as seen in compound heterozygotes, where individual mutations with partial loss of function might be able to compensate distinct defects for each other, could help to fully understand regulation and functions of the neuroprotective E3 Ub ligase Parkin.