14-3-3 binding to LRRK2 is disrupted by multiple Parkinson's disease-associated mutations and regulates cytoplasmic localization

Autosomal dominant missense mutations within the gene encoding for the LRRK2 (leucine-rich repeat protein kinase 2) predispose humans to develop PD (Parkinson's disease) [1,2]. LRRK2 is a large enzyme (2527 residues), consisting of leucine-rich repeats (residues 1010–1287), a GTPase domain (residues 1335–1504), a COR [C-terminal of ROC (Ras of complex GTPase domain)] domain (residues 1517–1843), a serine/threonine protein kinase domain (residues 1875–2132) and a WD40 repeat (residues 2231–2276) [3]. Patients with LRRK2 mutations generally develop PD with clinical symptoms indistinguishable from idiopathic PD at approx. 60–70 years of age [4]. Mutations in LRRK2 account for 4% of familial PD, and are observed in 1% of sporadic PD patients [4]. Over 40 missense mutations have been reported that are scattered throughout LRRK2 [5,6].

The activity, as well as localization, of a subset of mutant forms of LRRK2 has been analysed in previous work using various forms of recombinant LRRK2 expressed and assayed using diverse approaches [3,7]. This has revealed that the most frequent mutation, comprising an amino acid substitution of the highly conserved Gly²⁰¹⁹ located within the subdomain VII-DFG motif of the kinase domain, enhances protein kinase activity approx. 2-fold [8–10], suggesting that LRRK2 inhibitors might be useful for the treatment of PD. It was also reported that various mutants such as LRRK2(R1441C) and LRRK2(Y1699C) accumulated within discrete cytosolic pools, resembling inclusion bodies that were suggested to consist of aggregates of misfolded protein [9].

In the present study we have demonstrated that endogenous LRRK2 interacts with endogenous 14-3-3 protein isoforms and that this binding is mediated via phosphorylation of LRRK2 at Ser⁹¹⁰ and Ser⁹³⁵. We provide evidence that disruption of 14-3-3-binding induces LRRK2 to accumulate within cytoplasmic pools, similar in appearance to those reported previously for the LRRK2(R1441C) and LRRK2(Y1699C) mutants [9]. Comparing the properties of 41 PD-associated mutant forms of LRRK2 strikingly revealed that ten of these displayed markedly reduced phosphorylation of Ser⁹¹⁰/Ser⁹³⁵ and 14-3-3 binding that included five common well-characterized pathogenic mutations (R1441C, R1441G, R1441H, Y1699C and I2020T). We also observed that most mutants displaying reduced Ser⁹¹⁰/Ser⁹³⁵ phosphorylation and association with 14-3-3 accumulated within cytoplasmic pools. These results provide the first evidence that a significant number of PD-associated mutations inhibit Ser⁹¹⁰/Ser⁹³⁵ phosphorylation, disrupting 14-3-3 binding and leading to accumulation of LRRK2 within cytoplasmic pools. Our present study will stimulate further work to explore whether disruption of 14-3-3 binding to LRRK2 is linked to the development of PD.

Tissue-culture reagents were from Life Technologies. P81 phosphocellulose paper was from Whatman and [γ-³²P]ATP was from PerkinElmer. All peptides were synthesized by Pepceuticals. The Flp-in T-REx system was from Invitrogen, and stable cell lines, generated according to the manufacturer's instructions by selection with hygromycin, have been described previously [11]. Restriction enzyme digests, DNA ligations and other recombinant DNA procedures were performed using standard protocols. All mutagenesis was carried out using the QuikChange® site-directed mutagenesis kit (Stratagene). DNA constructs used for transfection were purified from Escherichia coli DH5α cells using Qiagen or Invitrogen plasmid Maxi kits according to the manufacturer's instructions. All DNA constructs were verified by DNA sequencing, which was performed by the Sequencing Service, School of Life Sciences, University of Dundee, Dundee, Scotland, U.K., using DYEnamic ET terminator chemistry (Amersham Biosciences) with automated DNA sequencers (Applied Biosystems).

Lysis buffer contained 50 mM Tris/HCl, pH 7.5, 1 mM EGTA, 1 mM EDTA, 1% (w/v) (1 mM) sodium orthovanadate, 10 mM sodium β-glycerophosphate, 50 mM NaF, 5 mM sodium pyrophosphate, 0.27 M sucrose, 1 mM benzamidine and 2 mM PMSF and was supplemented with 1% Triton X-100. Buffer A contained 50 mM Tris/HCl, pH 7.5, 50 mM NaCl, 0.1 mM EGTA and 0.27 M sucrose. λ phosphatase reactions were carried out in buffer A supplemented with 1 mM MnCl₂, 2 mM DTT (dithiothreitol) and 0.5 μg of λ phosphatase.

Anti-LRRK2-(100–500) (S348C and S406C) and anti-LRRK2 (2498–2514) (S374C) antibodies have been described previously [8]. The antibody against LRRK2 phospho-Ser⁹¹⁰ (S357C) was generated by injection of the KLH (keyhole-limpet haemocyanin)-conjugated phospho-peptide VKKKSNpSISVGEFY (where pS is phospho-serine) into sheep and was affinity-purified by positive and negative selection against the phospho- and de-phospho-peptides respectively. Antibody against LRRK2 phospho-Ser⁹³⁵ (S814C) was generated by injection of the KLH-conjugated phospho-peptide NLQRHSNpSLGPIFDH into sheep and was affinity-purified by positive and negative selection against the phospho- and de-phospho-peptides respectively. Anti-GFP (green fluorescent protein) antibody (S268B) was raised against the recombinant GFP and affinity-purified against the antigen. Anti-FLAG M2 antibody and the affinity matrix were from Sigma (A2220). The Nanotrap GFP-binder affinity matrix was from ChromoTek. Rabbit polyclonal antibody recognizing 14-3-3 (K-19, SC-629) and control rabbit IgG (SC-2027) antibody were from Santa Cruz Biotechnology. Anti-Hsp90 (heat-shock protein 90) antibody was from Cell Signaling Technology (#4877). Anti-MARK3 (microtubule affinity-regulating kinase 3) was from Upstate Biotechnology (#05-680).

HEK-293 (human embryonic kidney) and Swiss 3T3 cells were cultured in DMEM (Dulbecco's modified Eagle's medium) supplemented with 10% FBS (fetal bovine serum), 2 mM glutamine and 1× antimycotic/antibiotic solution (1× penicillin/streptomycin/amphotericin B; Invitrogen). HEK-293 T-REx cell lines were cultured in DMEM supplemented with 10% FBS and 2 mM glutamine, 1× antimycotic/antibiotic solution, 15 μg/ml blastocidin and 100 μg/ml hygromycin. Cell transfections were performed by the polyethyleneimine method [12]. T-REx cultures were induced to express the indicated protein by inclusion of 1 μg/ml doxycycline in the culture medium for 24 h. Per 15 cm dish, cells washed once with PBS and lysed in situ with 1.0 ml of lysis buffer, on ice, then centrifuged at 16000 g at 4 °C for 10 min. Protein concentrations were determined using the Bradford method with BSA as the standard.

SILAC DMEM (high glucose without NaHCO₃, L-glutamine, arginine, lysine and methionine; Biosera #A0347) was prepared with 10% dialysed FBS (Hyclone) and supplemented with methionine, glutamine, NaHCO₃ and labelled or unlabelled arginine and lysine. Cells harbouring GFP-tagged proteins were cultured in SILAC DMEM for three passages at a 1:10 ratio with the following isotopic labelling. For GFP compared with wild-type LRRK2, L-arginine (84 μg/ml; Sigma–Aldrich) and L-lysine (146 μg/ml; Sigma–Aldrich) were added to the GFP ‘light’ medium, whereas ¹³C-labelled L-arginine and ¹³C-labelled L-lysine (Cambridge Isotope Laboratory) were added to the GFP–LRRK2 wild-type ‘heavy’ medium at the same concentrations. For the GFP compared with LRRK2(G2019S) experiments, L-arginine and L-lysine were added to the GFP ‘light’ medium and ¹³C/¹⁵N-labelled L-arginine and ¹³C/¹⁵N-labelled L-lysine (Cambridge Isotope Laboratory) to the GFP–LRRK2(G2019S) ‘heavy’ medium. The amino acid concentrations are based on the formula for normal DMEM (Invitrogen). Once prepared, the SILAC medium was mixed well and filtered through a 0.22-μm filter (Millipore). Metabolically labelled cells were induced to express GFP or the GFP–LRRK2 fusion protein for 24 h by inclusion of doxycycline in the culture medium.

Cells metabolically labelled and induced to express GFP, wild-type LRRK2 or LRRK2(G2019S) were lysed in lysis buffer supplemented with 1% Triton X-100 at 0.5 ml per 10 cm dish. For each condition individually, 9 mg of cell lysate was subjected to individual immunoprecipitation with a 20 μl of bed volume of GFP-binder agarose beads for 1 h at 4 °C. Beads were washed once with 5 ml and then with 10 ml of lysis buffer supplemented with 1% Triton-X 100 and 300 mM NaCl. Beads were then washed once with 5 ml and then once with 10 ml of storage buffer. Bead-associated proteins were eluted with 1× NuPAGE LDS sample buffer (Invitrogen) for 10 min at 70 °C then passed through a 0.22 μm Spin-X column (Corning). Control GFP eluates were combined with either eluates of wild-type LRRK2 or LRRK2(G2019S) in equal amounts and reduced and alkylated as above. Samples were resolved on a 12% Novex gel for only one half of the gel. Gels were stained with Colloidal Blue overnight and destained for 3 h. The entire lane was excised in nine bands in total and digested with trypsin as described previously [13]. The digests were separated on a Biosphere C₁₈ trap column [0.1 mm (internal diameter)×2 mm; Nanoseparations] connected to a PepMap C₁₈ nano column (75 μm×15 cm; Dionex Corporation) fitted to a Proxeon Easy-LC nanoflow LC-system (Proxeon Biosystems) with solvent A (2% acetonitrile/0.1% formic acid/98% water) and solvent B (90% acetonitrile/10% water/0.09% formic acid). Samples (10 μl; a total of 2 μg of protein) were loaded with a constant flow of 7 μl/min on to the trap column in solvent A and washed for 3 min at the same flow rate. After trap enrichment, peptides were eluted with a linear gradient of 5–50% solvent B over 90 min with a constant flow of 300 nl/min. The HPLC system was coupled to a linear ion-trap–orbitrap hybrid mass spectrometer (LTQ-Orbitrap XL, Thermo Fisher Scientific) via a nanoelectrospray ion source (Proxeon Biosystems) fitted with a 5 cm Picotip FS360-20-10 emitter. The spray voltage was set to 1.2 kV and the temperature of the heated capillary was set to 200 °C. Full-scan MS survey spectra (m/z 350–1800) in profile mode were acquired with the LTQ-Orbitrap with a resolution of 60000 after accumulation of 500000 ions. The five most intense peptide ions from the preview scan in the LTQ-Orbitrap were fragmented by collision-induced dissociation (normalized collision energy 35%, activation Q 0.250 and activation time 30 ms) in the LTQ-Orbitrap after the accumulation of 10000 ions. Maximal filling times were 1000 ms for the full scans and 150 ms for the MS/MS scans. Precursor ion-charge state screening was enabled and all unassigned charge states, as well as singly charged species, were rejected. The lock mass option was enabled for survey scans to improve mass accuracy. Data were acquired using the Xcalibur software.

The raw mass spectrometric data files obtained for each experiment were collated into a single quantified dataset using MaxQuant (version 1.0.13.13; http://www.maxquant.org) and the Mascot search engine (Matrix Science, version 2.2.2) software. Enzyme specificity was set to that of trypsin. Other parameters used within the software: variable modifications, methionine oxidation; database, target-decoy human MaxQuant (ipi.HUMAN.v3.52.decoy) (containing 148380 database entries); labels, R6K4 (for GFP compared with wild-type LRRK2) or R10K8 [for GFP compared with LRRK2(G2019S)]; MS/MS tolerance, 0.5 Da; top MS/MS peaks per 100 Da, 5; maximum missed cleavages, 2; maximum of labelled amino acids, 3; FDR (false discovery rate), 1%.

Endogenous and recombinant LRRK2 was immunoprecipitated from 50 mg of Swiss 3T3 lysate or T-REx cells induced to express FLAG–LRRK2 cell lysate using anti-LRRK2-(100–500)– or anti-FLAG–agarose respectively. Immunoprecipitates were eluted from the affinity matrices using 2× LDS sample buffer or 200 μg/ml FLAG peptide then filtered through a 0.2 μm Spin-X column before reduction with 10 mM DTT and alkylation with 50 mM iodoacetamide. Samples were heated for 10 min at 70 °C and resolved on 4–12% Novex gels before staining with Colloidal Blue (Invitrogen). Bands corresponding to LRRK2 were excised and digested with trypsin as described previously [13]. Samples were analysed on an LTQ-Orbitrap XL mass spectrometer as described above, except that the top five ions were fragmented in the linear ion-trap using multistage activation of the neutral loss of phosphoric acid from the parent ion (neutral loss masses=49, 32.33 and 24.5 for z=2, 3 and 4 respectively). Mascot generic files were created from the raw files using raw2msm (a gift from Professor Matthias Mann, Max Planck Institute of Biochemistry, Martinsried, Germany) and were searched on a local Mascot server (http://www.matrixscience.com) using the IPI (International Protein Index) mouse database for endogenous LRRK2 or the IPI human database for recombinant LRRK2.

Cell lysates (10–30 μg) were resolved by electrophoresis on SDS/polyacrylamide gels or Novex 4–12% gradient gels, and electroblotted on to nitrocellulose membranes. Membranes were blocked with 5% (w/v) skimmed milk in TBST [Tris-buffered saline with Tween 20: 50 mM Tris/HCl, pH 7.5, 0.15 M NaCl and 0.1% (v/v) Tween 20]. For phospho-specific antibodies, primary antibody was used at a concentration of 1 μg/ml, diluted in 5% skimmed milk in TBST with the inclusion of 10 μg/ml dephosphorylated peptide. All other antibodies were used at 1 μg/ml in 5% (w/v) milk in TBST. Detection of immuno-complexes was performed using either fluorophore-conjugated secondary antibodies (Molecular Probes) followed by visualization using an Odyssey infrared imaging system (LI-COR Biosciences) or by horseradish-peroxidase-conjugated secondary antibodies (Pierce) and an enhanced chemiluminescence reagent. For immunoprecipitations, antibody was non-covalently coupled to Protein G–Sepharose at a ratio of 1 μg of antibody/μl of beads, or anti-FLAG M2–agarose was utilized. Cell lysates were incubated with coupled antibody for 1 h. To assess Ser⁹³⁵ phosphorylation, total LRRK2 levels and 14-3-3 binding in mouse tissues, LRRK2 was immunoprecipitated from 6 mg of whole-tissue lysate using 15 μg of antibody coupled to 15 μl of Protein G–Sepharose. Ser⁹¹⁰ phosphorylation was assessed following immunoprecipitation from 10 mg of tissue lysate. Immuno-complexes were washed twice with lysis buffer supplemented with 0.3 M NaCl and twice with buffer A. Precipitates were re-suspended in LDS sample buffer and subjected to immunoblot analysis. DIG (digoxigenin)-labelled 14-3-3 for use in overlay far-Western blot analysis was prepared as described in [14]. To directly assess 14-3-3 interaction with LRRK2, immunoprecipitates were electroblotted on to nitrocellulose membranes and blocked with 5% skimmed milk for 30 min. After washing with TBST, membranes were incubated with DIG-labelled 14-3-3 diluted to 1 μg/ml in 5% BSA in TBST overnight at 4 °C. DIG-labelled 14-3-3 was detected with horseradish-peroxidase-labelled anti-DIG Fab fragments (Roche).

Transfected cell lysates (500 μg) were subjected to immunoprecipitation with a 5 μl bed volume of anti-FLAG–agarose for 1 h. Beads were washed twice with lysis buffer supplemented with 300 mM NaCl, and twice with buffer A. Peptide kinase assays were set up in a total volume of 50 μl with immunoprecipitated LRRK2 in 50 mM Tris/HCl, pH 7.5, 0.1 mM EGTA, 10 mM MgCl₂ and 0.1 mM [γ-³²P]ATP (~300–500 c.p.m./pmol) in the presence of 200 μM long variant of the LRRKtide peptide substrate (RLGRDKYKTLRQIRQGNTKQR) [10,11] or the Nictide peptide substrate (RLGWWRFYTLRRARQGNTKQR) [11]. Reactions were terminated by applying 30 μl of the reaction mixture on to P81 phosphocellulose paper and immersing in 50 mM phosphoric acid. After extensive washing, the radioactivity in the reaction products were quantified by Cerenkov counting. One half of the remaining reaction was subjected to immunoblot analysis using the Odyssey infrared imaging system and specific activity is represented as c.p.m./independent density values.

An N-terminally biotinylated di-phosphorylated peptide encompassing Ser⁹¹⁰ and Ser⁹³⁵ (biotin–KKKSNpSISVGE-FYRDAVLQRCSPNLQRHSNpSLGPIF) was conjugated to streptavidin–agarose (1 μg peptide/μg of agarose). Aliquots of agarose beads (10 μl) were treated with or without λ phosphatase for 30 min at 30 °C, with λ phosphatase being in the presence or absence of 50 mM EDTA. Conjugated beads were then incubated with 3 mg of HEK-293 cell lysate at 4 °C for 1 h. Following two washes with lysis buffer supplemented with 0.5 M NaCl, beads were boiled in LDS sample buffer and samples subjected to immunoblot analysis for 14-3-3.

HEK-293 Flp-in T-REx cells harbouring GFP-tagged LRRK2 and the PD-associated mutations were plated in four-well glass bottomed CC2-coated chamber slides (Nunc). After plating (1 day), cells were induced with 1 μg/ml doxycycline and 24 h later, cells were fixed in 4% (w/v) paraformaldehyde buffered in PBS (pH 7.0–7.6) (purchased from USB, #19943). Cells were mounted in ProLong Gold (Invitrogen) and imaged under the same settings for each mutant with a Zeiss LSM 700 confocal microscope using an α-Plan-Apochromat ×100 objective.

We employed quantitative SILAC-based MS to identify proteins associated with immunoprecipitates of stably expressed full-length GFP–LRRK2 (Figure 1A), as well as the GFP–LRRK2(G2019S) mutant (Figure 1B), derived from HEK-293 cells. The top hits, which were enriched at 10–30-fold higher levels with GFP–LRRK2 or GFP–LRRK2(G2019S) compared with GFP alone, comprised various isoforms of 14-3-3 (Figure 1). The two other major interactors observed were two isoforms of the Hsp90 chaperone associated with its kinase-specific targeting CDC37 (cell division cycle 37) subunit (enriched 5–15-fold). Hsp90 and CDC37 associated with both wild-type LRRK2 and the LRRK2(G2019S) mutant, and have been reported previously to interact with LRRK2 [15]. No other significant interactors of LRRK2 were detected in this screen.

We found that endogenous 14-3-3, as well as Hsp90, was co-immunoprecipitated with endogenous LRRK2 from Swiss 3T3 cells (Figure 2A). We also observed that endogenous LRRK2 was co-immunoprecipitated with an antibody that recognizes endogenous 14-3-3 isoforms from Swiss 3T3 cells (Figure 2B). Plasmids encoding expression of seven isoforms of human 14-3-3 were transfected into HEK-293 cells stably expressing full-length FLAG–LRRK2. Following affinity purification, apart from the atypical σ isoform, all other isoforms of 14-3-3 interacted with FLAG–LRRK2 (Figure 2C). As a control we also demonstrated that all 14-3-3 isoforms except 14-3-3σ interacted with endogenous MARK3, a previously characterized 14-3-3 interactor [16].

14-3-3 isoforms mostly interact with specific phosphorylated residues on their binding partners [17,18]. To verify whether association of 14-3-3 with LRRK2 was dependent on phosphorylation, we incubated endogenous LRRK2 (Figure 2D) or overexpressed FLAG–LRRK2 (Figure 2E) in the presence or absence of λ phosphatase. λ phosphatase markedly reduced interaction of 14-3-3 with LRRK2 assessed using the overlay assay. This effect was suppressed by the inclusion of the λ phosphatase inhibitor EDTA in the assay (Figures 2D and 2E). Residual binding of 14-3-3 to LRRK2 following λ phosphatase treatment is presumably due to incomplete dephosphorylation of LRRK2.

To determine which phosphorylated residue(s) mediate binding to 14-3-3, we performed detailed phospho-peptide Orbitrap MS analysis of endogenous LRRK2 immunoprecipitated from mouse Swiss 3T3 cells (Figure 3A). This revealed three clear phosphorylation sites, namely Ser⁸⁶⁰, Ser⁹¹⁰ and Ser⁹³⁵ (Figure 3B). These residues lie in the N-terminal non-catalytic region of LRRK2 just prior to the leucine-rich repeats (Figure 3C). We also analysed phosphorylation of overexpressed full-length human FLAG–LRRK2 expressed in HEK-293 cells, which confirmed that Ser⁸⁶⁰, Ser⁹¹⁰ and Ser⁹³⁵ were major sites of phosphorylation (Figures 3A and 3B). In addition, we found three other phosphorylation sites in the overexpressed human FLAG–LRRK2 preparation, namely Ser⁹⁵⁵, Ser⁹⁷³ and Ser⁹⁷⁶ (Figures 3B and 3C). The phospho-peptides encompassing Ser⁹⁵⁵, Ser⁹⁷³ and Ser⁹⁷⁶ were also detected in our analysis of endogenous LRRK2, but owing to the lower abundance of these peptides, we were unable to assign the exact phosphorylation sites (results not shown).

We observed that mutation of Ser⁸⁶⁰, Ser⁹⁵⁵, Ser⁹⁷³, Ser⁹⁷⁶ or both Ser⁹⁷³/Ser⁹⁷⁶ phosphorylation sites to an alanine residue did not affect binding of 14-3-3 to full-length FLAG–LRRK2 (Figure 3D). Strikingly, however, mutation of Ser⁹¹⁰ and/or Ser⁹³⁵ to an alanine residue abolished 14-3-3 interaction, indicating that phosphorylation of both of these residues is necessary for binding of LRRK2 to 14-3-3 isoforms (Figure 3D). Mutation of Ser⁹¹⁰ and/or Ser⁹³⁵ or any of the other identified phosphorylation sites did not affect LRRK2 protein kinase activity (Figure 3D). To study whether phosphorylation of Ser⁹¹⁰ and Ser⁹³⁵ was sufficient to mediate interaction with 14-3-3 isoforms, we generated a di-phosphorylated biotinylated peptide encompassing Ser⁹¹⁰ and Ser⁹³⁵. We observed that when this was conjugated to streptavidin–agarose, it efficiently affinity-purified 14-3-3 isoforms from HEK-293 cell extracts (Figure 3E). Incubation of this peptide with λ phosphatase to dephosphorylate Ser⁹¹⁰ and Ser⁹³⁵ prevented interaction with 14-3-3 isoforms, an effect that was not observed when the λ phosphatase inhibitor EDTA was included (Figure 3E).

We next generated phospho-specific antibodies recognizing LRRK2 phosphorylated at Ser⁹¹⁰ or Ser⁹³⁵. These antibodies were specific, as mutation of Ser⁹¹⁰ to an alanine residue ablated recognition of LRRK2 with an anti-phospho-Ser⁹¹⁰ antibody and, similarly, mutation of Ser⁹³⁵ abolished recognition with the anti-phospho-Ser⁹³⁵ antibody (Figure 3F). We consistently observed that mutation of Ser⁹¹⁰ to an alanine residue reduced phosphorylation of Ser⁹³⁵ approx. 2-fold and the similar mutation of Ser⁹³⁵ reduced phosphorylation of Ser⁹¹⁰ approx. 2-fold as quantified by the Odyssey system (Figure 3F). Utilizing these antibodies, we demonstrate that endogenous LRRK2 immunoprecipitated from mouse brain, kidney and spleen was phosphorylated at Ser⁹¹⁰, as well as Ser⁹³⁵, and also bound 14-3-3 (Figure 3G).

Sequence alignments indicate that the Ser⁹¹⁰ and Ser⁹³⁵ sites, as well as residues surrounding them, are highly conserved in mammalian species (Figure 3H). This region encompassing Ser⁹¹⁰ and Ser⁹³⁵ is not present in Caenorhabditis elegans or Drosophila melanogaster LRK-1, or, indeed, mammalian LRRK1. Comparison of the residues surrounding Ser⁹¹⁰ and Ser⁹³⁵ indicates some striking similarities (Figure 3I, i.e. basic residues at the −3 and −4 positions, a serine residue at the −2 position, an asparagine residue at the −1 position and a large hydrophobic residue at the +1 position).

A common role of 14-3-3 proteins is to influence the subcellular localization of the protein to which it binds. We therefore studied whether 14-3-3 binding might affect LRRK2 cellular localization. To ensure low-level expression and as uniform as possible, we generated Flp-in T-REx 293 cells that stably express the wild-type and non-14-3-3-binding Ser⁹¹⁰/Ser⁹³⁵ mutant forms of full-length GFP–LRRK2. Immunoblot analysis revealed that the wild-type and mutant GFP–LRRK2 forms were expressed at similar levels (Figure 4A). We next studied the cellular localization using confocal microscopy and found that, consistent with a previous report [19], wild-type LRRK2 was uniformly distributed throughout the cytosol and excluded from the nucleus (Figure 4B). In contrast, the non-14-3-3-binding LRRK2(S910A), LRRK2(S935A) and LRRK2(S910A/S935A) mutants accumulated within cytosolic pools resembling inclusion bodies (Figure 4B).

We next decided to investigate the Ser⁹¹⁰/Ser⁹³⁵ phosphorylation and 14-3-3-binding properties of 41 PD forms of LRRK2. These mutations include the six most common pathogenic mutations that have been described to date (R1441C, R1441G, R1441H, Y1699C, G2019S and I2020T) [5,6]. Most of the other mutations studied have only been observed in low numbers of PD patients and further work is required to assess the contributions these mutations make to the development of disease [6]. The location of different mutations in LRRK2 analysed is indicated in the Figure 5 (inset). We expressed full-length wild-type and mutant forms of LRRK2 with an N-terminal FLAG epitope tag in HEK-293 cells. LRRK2 was immunoprecipitated and levels of protein were determined by quantitative immunoblotting analysis (Odyssey system). Similar levels of LRRK2 forms were subjected to immunoblot analysis, which revealed that most of the mutants were phosphorylated at Ser⁹¹⁰ and Ser⁹³⁵ to a similar extent as the wild-type enzyme and interacted with 14-3-3 (Figure 5, lower panel). However, Ser⁹¹⁰/Ser⁹³⁵ phosphorylation and hence 14-3-3 binding were abolished in four mutants (R1441G, Y1699C, E1874stop and I2020T) (Figure 5, lower panel). Phosphorylation of Ser⁹¹⁰/Ser⁹³⁵ and 14-3-3 binding were significantly reduced in six other mutants (M712V, R1441H, R1441C, A1442P, L1795F and G2385R) (Figure 5, lower panel).

We also compared the relative protein kinase specific activity of the 41 mutant forms of LRRK2 employing the LRRKtide peptide substrate [10] (Figure 5). Consistent with previous work [8–10], the LRRK2(G2019S) mutant possessed approx. 3-fold higher specific activity than wild-type LRRK2. Two other mutants, LRRK2(R1728H) and LRRK2(T2031S), also exhibited 2- and 4-fold increased activity respectively compared with wild-type LRRK2. Apart from the R1874stop mutation that lacks the kinase domain and is therefore inactive, all other mutants tested possessed similar activity to wild-type LRRK2.

To obtain further evidence that the LRRK2(R1441C) PD mutation disrupts 14-3-3 binding, we compared levels of 14-3-3 associated with endogenous LRRK2 derived from previously reported littermate wild-type and homozygous LRRK2(R1441C) knock-in mice [20]. LRRK2 was immunoprecipitated from spleen, kidney and brain from three separate mice of each genotype. Immunoblotting and 14-3-3 overlay analysis demonstrated that level of LRRK2 expression was similar in the wild-type and knock-in mice, however, the level of Ser⁹¹⁰/Ser⁹³⁵ phosphorylation and associated 14-3-3 was markedly reduced in tissues derived from LRRK2(R1441C) knock-in mice compared with wild-type (Figure 6). The largest effect of the knock-in mutation was observed in kidney.

To ensure low-level expression and as uniform as possible of the wild-type and mutant forms of full-length GFP–LRRK2, we generated the Flp-in T-REx 293 cells that stably express the wild-type and mutant forms of GFP–LRRK2. Immunoblot analysis revealed that GFP–LRRK2 forms were expressed at relatively similar levels (Figure 7). The localization of wild-type, kinase-dead and many other LRRK2 mutants studied were uniformly distributed throughout the cytosol and excluded from the nucleus with no accumulation within cytoplasmic pools observed (Figure 7 and Supplementary Figure S1 to view larger images at http://www.BiochemJ.org/bj/430/bj4300393add.htm). Strikingly, most mutants that displayed reduced Ser⁹¹⁰/Ser⁹³⁵ phosphorylation and binding to 14-3-3 accumulated within cytosolic pools resembling inclusion bodies (R1441C, R1441G, R1441H, A1442P, Y1699C, L1795F and I2020T). Not counting the truncated E1874stop mutant, which displays diffuse cytoplasmic localization, only two other mutants (M712V and G2385R) displaying reduced Ser⁹¹⁰/Ser⁹³⁵ phosphorylation and 14-3-3 binding (Figure 5), but did not accumulate in cytoplasmic pools (Figure 7 and Supplementary Figure S1). It is possible that the ability of these mutants to still interact weakly with 14-3-3 isoforms may prevent their accumulation within cytoplasmic pools. Only a single mutant (R1067Q) was found to interact with 14-3-3 and accumulate within cytoplasmic pools (Figure 7 and Supplementary Figure S1). The reasons for this require further work.

In the present study we have established that 14-3-3 isoforms interact with endogenous LRRK2 and this is mediated by phosphorylation of Ser⁹¹⁰ and Ser⁹³⁵. The finding that a di-phosphorylated peptide encompassing Ser⁹¹⁰ and Ser⁹³⁵ interacted with 14-3-3 isoforms in HEK-293 cell extracts and that LRRK2 S910A/S935A do not interact with 14-3-3 suggests that phosphorylation of these residues is necessary and sufficient to enable LRRK2 to interact with 14-3-3 isoforms. 14-3-3 proteins interact dynamically with many intracellular proteins, exerting widespread influence on diverse cellular processes. They operate by binding to specific phosphorylated residues on target proteins. The finding that LRRK2 interacts with 14-3-3 isoforms could not be predicted by analysis of the primary sequence, because the residues surrounding the phosphorylation sites at positions 910 and 935 do not adhere to the optimal mode 1 and 2 consensus binding motifs for a common mode of 14-3-3 interaction [17,18]. However, many proteins that interact with 14-3-3 do so via diverse non-predictable atypical binding motifs, presumably because other structural features contribute to the interactions [17].

Our data suggest that phosphorylation of both Ser⁹¹⁰ and Ser⁹³⁵ is required for stable interaction of 14-3-3 with LRRK2, as mutation of either Ser⁹¹⁰ or Ser⁹³⁵ abolishes 14-3-3 interaction. 14-3-3 molecules form dimers with each monomer having the ability to interact with a phosphorylated residue [18]. Thus a 14-3-3 dimer has the capacity to interact with two phosphorylated residues. It is possible that one dimer of 14-3-3 interacts with both phosphorylated Ser⁹¹⁰ and phosphorylated Ser⁹³⁵. We also observed that mutation of either Ser⁹¹⁰ or Ser⁹³⁵ to an alanine residue induced a significant dephosphorylation of the other residue (Figure 3F). This could be explained if 14-3-3 binding protected Ser⁹¹⁰ and Ser⁹³⁵ from becoming dephosphorylated by a protein phosphatase. Thus abolishing 14-3-3 binding by mutation of either Ser⁹¹⁰ or Ser⁹³⁵ would promote dephosphorylation of the other site. 14-3-3 binding to other targets such as phosphoinositide 4-kinase IIIβ [21] or CDC25C [22] protects these enzymes from dephosphorylation, presumably by sterically shielding phosphorylated residues from protein phosphatases.

14-3-3 proteins were originally identified 42 years ago as acidic proteins that were highly expressed in the brain [23]. Since then 14-3-3 proteins have been implicated in the regulation of numerous neurological disorders including PD [24,25]. For example, 14-3-3η binds to parkin, a protein mutated in autosomal recessive juvenile parkinsonism, and negatively regulates its E3 ligase activity [26]. 14-3-3 proteins interact with α-synuclein [27] and have been found in Lewy bodies in brains of patients with PD [28]. Additionally, 14-3-3θ, 14-3-3ε and 14-3-3γ were recently shown to suppress the toxic effects of α-synuclein overexpression in a cell-based model of neurotoxicity [29]. Our data suggest that 14-3-3 binding to LRRK2 may be relevant to PD, as, strikingly, ten out of 41 mutations studied displayed reduced phosphorylation of Ser⁹¹⁰/Ser⁹³⁵ and binding to 14-3-3 isoforms.

How 14-3-3 interaction influences LRRK2 function requires further investigation. Our data suggest that this interaction does not control LRRK2 protein kinase activity, as mutation of Ser⁹¹⁰ and/or Ser⁹³⁵ does not influence LRRK2 catalytic activity (Figure 3D). However, our experiments indicate that 14-3-3 binding influences the cytoplasmic localization of LRRK2, as disruption of 14-3-3 binding by mutation of Ser⁹¹⁰/Ser⁹³⁵ caused LRRK2 to accumulate within cytoplasmic pools. Previous work has revealed that LRRK2 mutants, including LRRK2(R1441C) and LRRK2(Y1699C), when expressed in cells accumulate within cytoplasmic pools resembling inclusion bodies [9,19,30], a finding we were able to confirm in the present study (Figure 7 and Supplementary Figure S1). These cytoplasmic pools were suggested to comprise aggregates of misfolded unstable LRRK2 protein [9]. If this were the case, it may suggest that 14-3-3 plays a role in stabilizing LRRK2. Strikingly, five out of the six most common pathogenic mutations (R1441C, R1441G, R1441H, Y1699C and I2020T) accumulated within cytoplasmic pools and displayed reduced Ser⁹¹⁰/Ser⁹³⁵ phosphorylation and binding to 14-3-3 under conditions in which wild-type LRRK2 and most other LRRK2 mutants analysed bound 14-3-3 and did not accumulate within cytosolic pools (Figure 7). We also validate these findings by demonstrating that Ser⁹¹⁰/Ser⁹³⁵ phosphorylation and 14-3-3 binding is markedly reduced in three mouse tissues derived from homozygous R1441C knock-in mice that display impaired dopaminergic neurotransmission [20]. Further work is required to establish why the R1441C, R1441G, R1441H, Y1699C and I2020T mutations disrupt phosphorylation of Ser⁹¹⁰ and Ser⁹³⁵, hence inhibiting 14-3-3 binding. It is possible that these mutations are partially misfolded or unstable and are hence not properly recognized by the upstream protein kinase that phosphorylates the Ser⁹¹⁰ and Ser⁹³⁵ residues.

In Supplementary Table S1 (at http://www.BiochemJ.org/bj/430/bj4300393add.htm) we subdivide the 41 LRRK2 mutations we have analysed into six groups on the basis of the impact that the mutations analysed in the present study have on protein kinase activity, Ser⁹¹⁰/Ser⁹³⁵ phosphorylation and 14-3-3-binding, as well as cellular localization. Only three out of the 41 mutations analysed markedly enhanced LRRK2 protein kinase 2–4-fold. Strikingly, the LRRK2(T2031S) mutation was more active than the LRRK2(G2019S) mutant, displaying nearly 4-fold higher activity than wild-type LRRK2. The LRRK2(T2031S) mutation has only been reported in a single Spanish patient with a family history of PD [31]. Interestingly, Thr²⁰³¹ lies within the T-loop of the LRRK2 kinase domain, a region where many kinases are activated by phosphorylation. Recent studies have suggested that LRRK2 is capable of autophosphorylating itself at Thr²⁰³¹ [32,33], although so far we have never been able to observe detectable phosphorylation of this site in endogenous or overexpressed LRRK2 by MS, perhaps indicating stoichiometry of phosphorylation of this residue is very low. We have found that changing Thr²⁰³¹ to an alanine residue to prevent phosphorylation led to a similar increase in activity as the T2031S mutation (R.J. Nichols, unpublished work) and mutation to a glutamate residue to mimic phosphorylation inactivates LRRK2 (N. Dzamko, unpublished work). Further work is warranted to understand the role of Thr²⁰³¹ and how its mutation to a serine residue and/or phosphorylation or other covalent modification affects LRRK2 catalytic activity. The LRRK2(R1728H) mutation displayed 2-fold increased kinase activity and was identified in a single patient with a family history of PD [6]. Arg¹⁷²⁸ is located outside the kinase domain in the COR domain (Figure 5).

In future work it will be important to investigate further how 14-3-3 binding affects on LRRK2 stability, interaction with substrates or other regulators. It would also be interesting to undertake a more detailed characterization of the cytoplasmic pools resembling inclusion bodies that non-14-3-3-binding LRRK2 mutants accumulate in and to determine whether these do indeed comprise aggregates of misfolded protein. Our results analysing some of the properties of the 41 PD-associated forms of LRRK2 emphasize that most mutations do not exert their effects in the same way as the G2019S mutation by simply increasing catalytic activity. Finding out how the diverse mutations affect LRRK2 regulation, the ability to interact with binding partners and stability, as well as downstream functions, will provide valuable insights into how LRRK2 exerts its influence in PD. It would also be important to further explore links between LRRK2 Ser⁹¹⁰/Ser⁹³⁵ phosphorylation, 14-3-3 binding and its relevance to development of PD. However, as 28 LRRK2 mutations possessed similar activity, 14-3-3 binding and cellular localization as wild-type LRRK2, assuming that these comprise pathogenic mutations, there is clearly still much to learn about how the majority of the mutations described affect LRRK2 to cause PD. Finally, it would be interesting to undertake LRRK2 interactor screens from neuronal cell lines or brain tissues that might reveal cofactors other than 14-3-3 that would be relevant to understanding how LRRK2 is regulated and functions.

CDC

cell division cycle

DIG

digoxigenin

DMEM

Dulbecco's modified Eagle's medium

DTT

dithiothreitol

FBS

fetal bovine serum

GFP

green fluorescent protein

HEK-293

human embryonic kidney

Hsp90

heat-shock protein 90

IPI

International Protein Index

KLH

keyhole-limpet haemocyanin

LRRK2

leucine-rich repeat protein kinase 2

MARK3

microtubule affinity-regulating kinase 3

PD

Parkinson's disease

ROC

Ras of complex GTPase domain

COR

C-terminal of ROC

SILAC

stable isotope labelling of amino acids

TBST

Tris-buffered saline with Tween 20

Jeremy Nichols and Nicolas Dzamko performed most of the experiments. Nicholas Morrice and David Campbell carried out the MS analysis. Maria Deak and Thomas Macartney carried out the cDNA cloning. Alban Ordureau performed the studies that contributed to the identification of the Ser⁹¹⁰ phosphorylation site. Youren Tong and Jie Shen provided tissues from LRRK2(R1441C) knock-in mice. Alan Prescott supervised and performed the localization studies. Nicolas Dzamko, Jeremy Nichols and Dario Alessi planned the experiments, analysed the results and wrote the manuscript. Dario Alessi supervised the project.

We thank the Sequencing Service (School of Life Sciences, University of Dundee, Dundee, Scotland, U.K.) for DNA sequencing, the Post Genomics and Molecular Interactions Centre for Mass Spectrometry facilities (School of Life Sciences, University of Dundee, Dundee, Scotland, U.K.) and the protein production and antibody purification teams [Division of Signal Transduction Therapy (DSTT), University of Dundee, Dundee, Scotland, U.K.] co-ordinated by Hilary McLauchlan and James Hastie for purification of antibodies. We also thank Miratul Muqit for helpful discussions.

This work is supported by the Medical Research Council [Medical Research Council Technology (MRCT) Industry collaborative award to N.D.]; and The Michael J. Fox Foundation. The Division of Signal Transduction Therapy Unit is supported by the following pharmaceutical companies: AstraZeneca, Boehringer-Ingelheim, GlaxoSmithKline, Merck-Serono and Pfizer.