The Parkinson's disease (PD) protein leucine-rich repeat kinase 2 (LRRK2) exists as a mixture of monomeric and dimeric species, with its kinase activity highly concentrated in the dimeric conformation of the enzyme. We have adapted the proximity biotinylation approach to study the formation and activity of LRRK2 dimers isolated from cultured cells. We find that the R1441C and I2020T mutations both enhance the rate of dimer formation, whereas, the G2019S kinase domain mutant is similar to WT, and the G2385R risk factor variant de-stabilizes dimers. Interestingly, we find a marked departure in the kinase activity between G2019S–LRRK2 homo-dimers and wild-type-G2019S hetero-dimers. While the homo-dimeric G2019S–LRRK2 exhibits the typical robust enhancement of kinase activity, hetero-dimers comprised of wild-type (WT) and G2019S–LRRK2 exhibit kinase activity similar to WT. Dimeric complexes of specific mutant forms of LRRK2 show reduced stability following an in vitro kinase reaction, in LRRK2 mutants for which the kinase activity is similar to WT. Phosphorylation of the small GTPase Rab10 follows a similar pattern in which hetero-dimers of WT and mutant LRRK2 show similar levels of phosphorylation of Rab10 to WT homo-dimers; while the levels of pRab10 are significantly increased in cells expressing mutant homo-dimers. Interestingly, while the risk variant G2385R leads to a de-stabilization of LRRK2 dimers, those dimers possess significantly elevated kinase activity. The vast majority of familial LRRK2-dependent PD cases are heterozygous; thus, these findings raise the possibility that a crucial factor in disease pathogenesis may be the accumulation of homo-dimeric mutant LRRK2.
Mutations in the gene encoding leucine-rich repeat kinase 2 (LRRK2) are the most frequent cause of familial forms of Parkinson's disease (PD), and are also commonly found in the more common sporadic form . The gene encodes a large 2527 amino acid protein belonging to the ROCO/RIPK families, with multiple enzymatic as well as protein–protein interaction domains. Biochemical, structural, and fluorescence and electron microscopic evidence supports the existence of LRRK2 dimers. The first indication that LRRK2 might exist in a dimeric conformation came from the crystal structure determination of the ROC GTPase domain [2,3]. In fact, to date, only two ROC domain structures have been reported, one for the LRRK2 domain, and the second for the ROC-COR tandem of Chlorobium tepidum [4,5]. While both studies provided structural support for the existence of a dimer, they diverged in their interpretation of the final structure of the G-domain, based on several factors. First, the overlay of the structural co-ordinates of human LRRK2-ROC on the corresponding C. tepidum structure, despite having high sequence homology, did not support the conclusion of the domain-swapped dimer, as originally suggested by Deng et al. . Secondly, the C. tepidum ROC-COR structure more closely resembled the known structures of other small GTPases, and forms stable dimers in solution ; whereas the dimer proportion of the isolated human LRRK2 ROC domain was less prevalent . In contrast, in a subsequent study, the isolated LRRK2 ROC domain in solution, in this case with extended boundaries, elutes in FPLC fractions consistent with both a dimer as well as a monomer . Dimers of cell-derived LRRK2 have been visualized using single-particle transmission electron microscopy , revealing the presence of LRRK2 complexes that were double immunogold-labeled with distances between the gold particles consistent with an ordered dimer. This technique is constrained by an incomplete labeling efficiency; however, it is the first direct observation of full-length LRRK2 dimers. Recently, a structural model of full-length LRRK2 was assembled, based on chemical cross-linking and negative-stain electron microscopic images of purified protein, which confirms from a structural point of view the existence of LRRK2 dimers ; and the in silico homology modeling and docking findings that are presented lend computational support to this evidence.
Complementary studies employing biochemical approaches to identify dimeric- or higher-order LRRK2 complexes, native gel electrophoresis , co-immunoprecipitation of differently tagged LRRK2 molecules, gradient centrifugation or size-exclusion chromatography (SEC; [10,11]) have each revealed the presence of LRRK2 with an apparent molecular mass consistent with dimeric or larger LRRK2 complexes. However, the major limitation of those approaches relying on the identification of LRRK2-positive species based on apparent molecular mass is that they cannot discriminate between truly dimeric LRRK2 and a LRRK2 monomer interacting with other proteins in a complex; and are further hampered by potentially misleading molecular mass comparisons that are influenced by globular conformational effects. Secondly, while co-immunoprecipitation of two differently tagged LRRK2 molecules overexpressed in cells can reveal the existence of both LRRK2 forms in a potential protein complex; however, whether this interaction is direct cannot be determined by this approach.
Dimerization can have a great impact on protein function, particularly for GTPases and kinases (e.g. ). In separate studies, the groups of West  and LaVoie  used complimentary approaches to compare kinase activity (total auto-phosphorylation) between high molecular weight (HMW) dimeric species or low- molecular-weight (LMW) LRRK2 monomer. Glycerol gradient centrifugation or SEC techniques were used to define monomeric vs. dimeric or other species of LRRK2, and in both studies LRRK2 auto-phosphorylation was highly enriched in the HMW or dimeric fractions in comparison with the LRRK2 monomer [10,11]. Berger et al.  expanded on this finding to demonstrate that the dimeric species of LRRK2 became enriched at the plasma membrane, with proportionally little dimeric LRRK2 detected in the cytosol both in overexpression studies and at endogenous levels in human lymphoblast cells. Thus, while LRRK2 auto-phosphorylation activity was clearly enriched in the dimeric pool of LRRK2 identified by both groups, whether this activity is fully attributable to truly dimeric LRRK2, or monomeric LRRK2 acting within a multi-protein complex cannot be definitively concluded in these studies.
We have developed a novel approach to address this question by labeling LRRK2 dimers in situ, and selectively purifying these species in order to directly compare their relative kinase activities. By combining this approach with an ELISA-based assay that couples the quantification of phosphorylation with the precise amount of LRRK2 present in each reaction, we confirm that true dimers of LRRK2 purified from total cytoplasmic extracts indeed exhibit greater kinase activity on a per molecule basis than monomeric LRRK2. Importantly, we show that while dimeric LRRK2 exhibits greater kinase activity, the specific composition of the dimeric complex is critical. In hetero-dimeric complexes of WT and mutant LRRK2, a single copy of the G2019S kinase domain mutant is insufficient to increase in vitro activity, whereas homo-dimers of G2019S–LRRK2 exhibit a robust increase. This approach has the capability to dramatically expand our understanding of the regulation of LRRK2 dimer formation, how this affects its cellular functions in neurons and other cell types, and how this is potentially linked to its pathological function in the presence of PD-causing mutations.
Wild-type (WT) LRRK2 cDNA with an N-terminal Flag epitope tag (in pcDNA3.1; [13,14]) was used as a template for all subsequent expression constructs. For introduction of pathogenic PD-linked mutations, disease-associated variants, or functional mutations, the Quikchange II site-directed mutagenesis kit (Agilent Technologies; CA, U.S.A.) was used according to the manufacturer's instructions. Plasmids subjected to mutagenesis underwent full sequencing to verify the absence of errors.
Cell lines and transfection
HEK293T cells were cultured in DMEM (Sigma; MO, U.S.A.) medium supplemented with 10% FBS and antibiotics, and transiently transfected in 10 cm or 6-well tissue culture plates using calcium phosphate:DNA precipitates. In co-expression experiments, vectors were transfected at a ratio of 1 : 1.
Proximity biotinylation of LRRK2 dimers/oligomers
For the selective labeling of LRRK2 dimers and oligomers, we adapted the proximity biotinylation approach described by Fernandez-Suarez et al. . We cloned LRRK2 in-frame, at the 3′ end of BirA, separated by a six base linker. The BirA-LRRK2 fusion also contains an N-terminal Flag epitope tag. Additionally, we created two LRRK2 constructs containing a truncated acceptor peptide, (-3) AP ; first, with an N-terminal c-myc epitope tag and the AP placed at the C-terminal (myc-LRRK2-AP), and second, with the c-myc tag and AP positions reversed (AP-LRRK2-myc). A schematic of the constructs used is shown in Figure 1a. BirA and AP-LRRK2 fusion constructs were co-transfected at a 1 : 1 ratio in HEK293T cells. When the two LRRK2 molecules are present as dimers in the cell, BirA-LRRK2 will biotinylate AP-LRRK2, in the presence of biotin, ATP, and MgCl2. After 48–72 h following expression, the cultures were given a pulse with biotin (50 µM, diluted in PBS supplemented with 5 mM MgCl2) for 5 min at 37°C followed by extensive washing in sterile PBS. For the biochemical detection and purification of biotinylated LRRK2 dimers, cytoplasmic extracts were prepared in lysis buffer A (50 mM Tris–HCl, pH 7.5; 1 mM EGTA; 0.27 M sucrose; 1% Triton X-100; supplemented with protease and phosphatase inhibitor cocktails). Five µg or less (as indicated), of total clarified cell extract was bound to streptavidin-coated ELISA plates for 1 h at 37°C under constant agitation. The supernatant was removed and retained, and the wells were washed and prepared for kinase activity assay and ELISA quantification of captured dimeric LRRK2 (see below). The supernatant following the initial binding to streptavidin (SA)-coated ELISA plates was then transferred to a second ELISA plate pre-coated with mouse monoclonal Flag antibodies (Pierce-Thermo Scientific), and incubated for 2 h at 37°C under constant agitation to capture non-biotinylated, monomeric, Flag-LRRK2. As for the SA-coated plate, these plates were prepared for kinase activity assays, followed by LRRK2 detection ELISA. As a control for the biotinylation reaction, AP-LRRK2 was co-expressed with Flag-LRRK2 (without BirA biotin ligase) and processed in parallel as the BirA/AP-LRRK2 pair.
Establishment of proximity biotinylation approach to label LRRK2 dimers in cells.
To examine the intracellular localization of LRRK2 dimers, HEK293T cells or primary neurons were co-transfected as before with BirA-LRRK2 and AP-LRRK2 (as indicated) at an equivalent molar ratio. Seventy-two h following transfection, the cells were given a biotin pulse as above, washed extensively with PBS and fixed in 4% paraformaldehyde for 20 min at 4°C. The fixed cells were then stained with anti-Flag antibodies to demonstrate the localization of BirA-LRRK2, followed by co-staining with streptavidin conjugated with AlexaFluor-streptavidin (SA-488) to label biotinylated LRRK2. Control cells expressing AP-LRRK2 together with Flag-LRRK2 (without BirA) revealed only very slight background signal under the conditions used here, which represents labeling of endogenously biotinylated proteins. Some cells were pretreated for 2 h with 100 nM of MLi-2 , to induce the redistribution of LRRK2 into cytoplasmic filaments. Images were acquired on a Leica TSP SP5II confocal microscope; with image stacks converted and processed using FIJI and Adobe Photoshop.
Transmission electron microscopy analysis of LRRK2 dimers/monomers
Immunogold transmission electron microscopic (TEM) visualization of LRRK2 in dimer-depleted samples was performed as previously described . Biotinylated, dimeric, LRRK2 was captured on SA-coated ELISA plates, and following binding, the supernatant containing cellular proteins and non-biotinylated (presumably monomeric) LRRK2 was retained. The supernatant was adsorbed to carbon-coated copper grids, and immunolabeled (anti-Flag) with 5 nm anti-mouse gold particles as described . Transmission EM images were acquired as described , and the relative proportion of remaining LRRK2 dimers, indicated by doublet nanogold particles separated by ∼100–200 Å, compared with monomeric LRRK2, labeled with singlet gold particles, was determined.
Size-exclusion chromatography and western immunoblotting
For the separation of globular protein complexes by SEC, HEK293T cells transiently expressing BirA- or AP-LRRK2, or Flag-LRRK2 without the biotin ligase (as indicated), were washed and given the biotin pulse as described above, washed extensively again and collected in ice-cold PBS, centrifuged and re-suspended in PBS containing protein and phosphatase inhibitors and incubated on ice for 20 min. The cells were then disrupted by homogenization in a glass Dounce homogenizer and centrifuged again at low speed (2000 rpm for 10 min) to remove nuclei and un-broken cells. The clarified lysate (1 mg total protein) was then injected into a 500 µl loop and separated using lysis buffer as the running buffer through a Superose 6 10/300 column. Fractions (250 µl) were collected in a 96-well plate and stored at −80°C until use. In parallel, native protein molecular mass markers (thyroglobulin, 669 kDa, apoferritin, 440 kDa, BSA, 66 kDa, and cytochrome c, 12.5 kDa are shown in relation to the fractions probed for LRRK2) were run under identical conditions and their elution used as a reference for the approximate size of LRRK2-containing protein complexes. To detect the elution of biotinylated dimeric LRRK2 present in each fraction, an aliquot (5 µl) was bound to SA-coated wells for 1 h at 37°C. The total amount of biotinylated LRRK2 present in each fraction was determined with anti-LRRK2 (clone N241A/34) conjugated in-house to HRP; or with anti-Flag (clone M2; Sigma). The chemiluminescence signal was produced by incubation with Pico ELISA substrate (ThermoScientific) for 5 min, and read on a Tecan Spark 10M multi-modal ELISA reader. The total amount of LRRK2 present within each fraction was determined by ELISA for total LRRK2 using wells pre-coated with anti-LRRK2 antibody (c41-2) and anti-LRRK2 detection antibody as above (HRP-N241A/34). The proportion of LRRK2 dimers within each fraction is expressed as a ratio of biotinylated LRRK2 to total LRRK2. In some cases, before separation on the Superose 6 10/300 column, lysates were incubated with SA-agarose resin to deplete biotinylated LRRK2 dimers. Following depletion, clarified extracts or extracts incubated with agarose beads alone, were injected in the Superose 6 10/300 column and fractions analyzed as described above.
Measurement of LRRK2 kinase activity
To determine the phosphorylation of the recently identified substrate of LRRK2, Rab10, we co-transfected plasmids encoding human Rab10 (Flag-tagged; Origene, MD, U.S.A.) together with the homo- and hetero-dimeric WT and mutant LRRK2 construct pairs used in the proximity biotinylation studies. In some experiments, the kinase inhibitor MLi-2 was added (100 nM; 2–4 h). At 48 h following expression, cells were washed in ice-cold PBS and lysed in buffer A, and the clarified cellular extract kept at −80°C until analyzed. For determination of phosphorylation of Rab10 at Thr73, proteins were separated by SDS–PAGE and the membranes were probed with phospho-Rab10 (T73; clone MJF-R21, Abcam), and Flag (Rabbit polyclonal; Sigma) for total Rab10 overexpression. Alternate 6% SDS–PAGE samples were run to facilitate separation of BirA-LRRK2 and AP-LRRK2, in order to demonstrate expression of both proximity biotinylation constructs. Expression of both LRRK2 molecules was visualized by probing the membranes with anti-LRRK2 (clone UDD3; or c41-2; Abcam). The ratio of band intensities between phospho-Rab10 and total Rab10 (anti-Flag) was plotted relative to total protein loaded (anti-GAPDH).
For the in vitro assessment of LRRK2 kinase activity, we performed a modification of the assay previously described in Melachroinou et al. . Briefly, HEK293T cells were transfected as indicated. Seventy-two hours following expression, cells were subjected to the biotin pulse as above, collected in ice-cold PBS, re-suspended in lysis buffer A, incubated on ice for 20 min, and then centrifuged at 13 000 rpm for 10 min. The supernatant was collected and 5 µg of clarified lysate added to SA-coated 96-well plates (Thermo Scientific; NJ, U.S.A.) and incubated for 1 h at 37°C while shaking. The supernatant was removed and subsequently immediately bound to Flag-coated ELISA plates for 2 h at 37°C under constant agitation. The wells in both the SA- and Flag-coated ELISA plates were washed 4X with ELISA wash buffer (50 mM Tris, pH 7.4; 150 mM NaCl; 0.1% Tween-20), followed by two washes with kinase reaction buffer (20 mM Tris, pH 7.5; 20 mM NaCl; 10 mM MgCl2; 2 mM DTT). For the kinase reaction, reaction buffer containing 100 µM ATP, 5 µM 6His-tagged NICtide (Innovagen; Lund, SE), and protease/phosphatase inhibitors (Roche) was added to each well and incubated for 30 min at 30°C. The reaction was stopped by the addition of 450 µl of ice-cold binding buffer (50 mM Tris, pH 7.6; 150 mM NaCl; 0.5% NP-40). Fifty µl of the diluted reaction was added to nickel-coated 96-well plates (Thermo Scientific; PA, U.S.A.) and incubated at 37°C for 1 h, followed by four washes with ELISA wash buffer. The wells were then incubated with anti-phospho-ThrXArg (Cell Signaling) for 1 h at room temperature, followed by four washes, and incubation with HRP-conjugated anti-rabbit secondary antibodies for 1 h at room temperature. The wells were washed and incubated for 5 min with chemiluminescent substrate (Thermo Scientific). The levels of LRRK2 were quantified directly in the 96-well plates by ELISA. Following the kinase reaction, the buffer containing phosphorylated peptide was removed, and the wells washed 4× with ELISA wash buffer. The same wells were then incubated with HRP-conjugated rabbit anti-LRRK2 (clone N241A/34), or anti-Flag (clone M2) for 1 h at room temperature. The wells were washed again, and incubated for 5 min with chemiluminescent substrate (Thermo Scientific). Thus, LRRK2 kinase activity, as determined by the relative amount of phosphorylated NICtide, is expressed as a ratio to the amount of LRRK2 present within the same well as determined by ELISA. The raw chemiluminescence data were converted to ng LRRK2/ml using the standard curves generated with full-length human recombinant LRRK2.
Alternatively, the auto-phosphorylation activity of LRRK2 was assessed by the incorporation of phosphate at the Ser1292 residue using a phospho-specific antibody (generously provided by Genentech; see ). Lysate from transfected HEK293T cells were sequentially bound to SA and Flag-coated ELISA plates in duplicate as described above, and washed extensively. The ELISA plates were processed in parallel for pS1292 or total LRRK2 by ELISA. The kinase reaction was performed in the absence of any added substrate, and parallel reactions performed in the absence or presence of ATP. Reactions in the absence of ATP revealed the steady-state level of S1292 auto-phosphorylation; whereas reactions performed in the presence of ATP indicated the in-well activity of LRRK2 with respect to auto-phosphorylation at this residue. As a control for the specificity of the antibody, lysate from cells expressing S1292A-LRRK2 were processed in parallel. The chemiluminescence signal from the pS1292 ELISA was normalized to the total LRRK2 in each sample to correct for variances in LRRK2 expression.
For the statistical comparisons used, we employed a one-way ANOVA, followed by Tukey's HSD post-hoc test for multiple comparisons. The P-value is indicated in the figure legend; and the data are presented as the mean ± standard error of the mean. The number of replications from independent experiments is provided in the figure legend.
Expression of BirA-LRRK2 and AP-LRRK2, and the in situ labeling of LRRK2 dimers
To label LRRK2 dimers in cells, with the purpose of not only detecting them but also selectively purifying active dimers from LRRK2 monomers, we adapted the proximity biotinylation approach of labeling protein:protein interactions (PPIs; ). Figure 1a shows a schematic of the biotin ligase (BirA) and acceptor peptide fusion construct pairs. Both LRRK2 fusions are expressed in HEK293T at similar levels when singly transfected, or when co-transfected at a 1 : 1 ratio (Figure 1b,c). The BirA fusion adds ∼25 kDa to the LRRK2 sequence, which causes its migration to be slower.
In a series of validation steps for the purification and detection of biotinylated AP-LRRK2 in our ELISA platform, we first assessed the concentration and time-dependency of the biotin pulse in cells co-expressing BirA- and AP-LRRK2. Transfected HEK293T cells were incubated with increasing concentrations of biotin for 5 min (Figure 2a), or a pulse consisting of 50 µM biotin for increasing incubation times (Figure 2b) and washed extensively in PBS. An increasing amount of LRRK2 dimers are labeled as the concentration of biotin is raised, and the time of incubation is increased. Both parameters quickly saturated the labeling of dimers within the cells, with no further increase in LRRK2 dimer labeling detected beyond 50 µM of biotin, or 5–10 min of the biotin pulse. Separately, transfected cells were given a biotin pulse of 50 µM biotin for 5 min, and processed as above. Increasing amounts of cell extract were then added to SA-coated ELISA plates. A concentration-dependent increase in chemiluminescent signal using two different LRRK2-specific antibodies independently as detection antibodies (UDD3 & N241A/34; Figure 2c and not shown) was observed in cells co-expressing BirA- and AP-LRRK2. As a control for the biotinylation labeling, and to rule out that LRRK2 is interacting with an endogenously biotinylated protein, we co-transfected AP-LRRK2 together with Flag-LRRK2 lacking the BirA biotin ligase. Additionally, AP-LRRK2 was co-transfected with BirA ligase alone to rule out a nonspecific biotinylation of LRRK2 independent of dimer formation. In these cells, we failed to detect any LRRK2 captured in SA-coated ELISA plates (Figure 2d), indicating that a specific BirA-LRRK2/AP-LRRK2 dimer is necessary for biotinylation.
Purification and quantification of LRRK2 dimers in situ.
To determine if both interacting LRRK2 pairs are present in the ELISA well capturing biotinylated AP-LRRK2, we incubated lysates of cells co-expressing BirA/AP-LRRK2 or WT-Flag (not fused with BirA ligase)/AP-LRRK2 in SA-coated ELISA plates, but instead of anti-LRRK2 as the detection antibody, we used anti-Flag, which selectively labels the BirA-LRRK2 fusion. We found that in cells co-expressing BirA/AP LRRK2, but not un-modified Flag-LRRK2 (without the BirA ligase), both molecules of LRRK2 are present, the biotinylated AP-LRRK2 captured in the SA-coated ELISA well, and it's interacting Flag-tagged BirA-fusion partner (Figure 2e). This is a critical determination for later studies described below assessing the kinase function of specific dimeric species.
A slight but specific biotinylation of AP-LRRK2 is detected even in the absence of added biotin, with the BirA ligase able to utilize the minimal endogenous biotin present in the growth medium and FBS (Figure 2f). Subsequent experiments were therefore conducted in OptiMEM growth media, which is biotin-free, in order to avoid over-estimating the proportion of LRRK2 dimers by capturing biotinylated AP-LRRK2 that has dissociated from its interacting partner during the 2–3 days prior to cell collection. Thus, by keeping the cells in biotin-free growth media for the duration of the experiment ensures that the only LRRK2 dimers that are labeled are those present during the 5 min biotin pulse. Interestingly, the position of the acceptor peptide resulted in different labeling efficiencies, with the relative amount of LRRK2 dimers markedly higher in cells expressing LRRK2 fused with a N-terminal acceptor peptide in comparison with the acceptor peptide placed at the C-terminal of LRRK2 (Figure 2g). This suggests that when folded, and in a dimeric conformation, the two N-terminal regions of LRRK2 are in closer proximity. Finally, we determined whether the presence of the BirA ligase fused to LRRK2 altered the kinase activity. In our ELISA-coupled in vitro kinase assay, LRRK2 purified from cells expressing either Flag-tagged WT LRRK2 or BirA-WT LRRK2 showed equal levels of phosphorylation of the NICtide peptide substrate (Figure 2h). However, purified protein from cells expressing BirA-G2019S–LRRK2 exhibited a significantly higher level of phosphorylation of NICtide compared with either WT protein (Figure 2h), to an extent similarly reported by other groups using other approaches [18,19]. Thus, we have shown that the proximity biotinylation approach can be adapted to specifically label, in cells, dimeric LRRK2 complexes, which can be purified and quantified by ELISA.
We next performed a series of experiments to examine the elution of labeled LRRK2 dimers using SEC. First, to rule out the possibility that the presence of the BirA and AP sequences were artificially inducing LRRK2 dimer formation, we co-expressed AP-LRRK2 either with Flag-BirA-LRRK2 as before, or with Flag-LRRK2 lacking the biotin ligase, and examined the elution of total LRRK2 by SEC. Cell extracts from both conditions were separated, and the amount of LRRK2 present in each fraction determined by dot blot and ELISA. In both cases, the results shown are for total LRRK2, not just purified dimers; thus, the elution profile spans a broader range of possible LRRK2 species, including monomeric LRRK2 (either alone or bound to other protein interactors), and dimeric LRRK2. The elution pattern was similar in both cases (Supplementary Figure S1a,b), suggesting that the globular structure of LRRK2 dimers and other LRRK2 complexes was not affected by the addition of the biotin ligase; nor was the elution pattern shifted in a way that would reflect an artificial induction of dimerization. Next, we prepared parallel cell extracts containing biotinylated WT-LRRK2 dimers, incubating one with SA-coated resin, to deplete labeled dimers from the extract, and the other with control resin. The un-bound flow-through of each sample was further separated by SEC, and the presence of dimeric LRRK2 present in selected fractions determined by ELISA, with anti-Flag as the detector antibody. A parallel fractionation was run with native molecular mass standards to indicate the approximate size of proteins within specific fractions (in Supplementary Figure S1d). Following incubation with SA-coated resin, the relative proportion of labeled dimeric LRRK2 present in the indicated range of fractions was significantly reduced (Supplementary Figure S1c), relative to total LRRK2 (not shown). This indicates that dimeric LRRK2 can be purified from cell extracts following in situ labeling with biotin.
De-phosphorylation alters the localization, but not the formation of LRRK2 dimers
Treatment of cells with virtually every class of LRRK2 kinase inhibitor leads to a rapid de-phosphorylation of the cluster of N-terminal Ser residues, such as Ser910/Ser935 and others . This de-phosphorylation is accompanied by a loss of binding at these sites by 14-3-3, and a re-distribution of LRRK2 into highly ordered cytoplasmic filamentous structures . Interestingly, these structures are also observed in cells overexpressing certain pathogenic mutant forms of LRRK2 [22,23]. We asked whether de-phosphorylation of LRRK2 leads to an alteration in dimer formation. Cells co-expressing BirA/AP-LRRK2 were treated with a pharmacological LRRK2 kinase inhibitor (MLi-2, 100nM; ) for 2 h before labeling LRRK2 dimers with a biotin pulse and cell lysis. We confirmed that such treatment induced the de-phosphorylation of LRRK2, as evidenced by a decrease in pS935 levels (Figure 3a). We find that de-phosphorylation of LRRK2 by kinase inhibitor treatment does not significantly alter LRRK2 dimer formation (Figure 3b), at least during the period of exposure to the inhibitor.
De-phosphorylated and mutant LRRK2 cytoplasmic filaments are comprised of dimeric species.
To determine if filamentous LRRK2, induced by kinase inhibitor dependent de-phosphorylation or the presence of certain pathogenic mutations, consists of dimeric LRRK2, we treated HEK293T cells expressing WT/WT-LRRK2 constructs with the specific LRRK2 inhibitor MLi-2. Cells were given a biotin pulse, fixed and labeled with anti-Flag antibodies and fluorescent-SA. In parallel, cells were transfected with I2020T BirA/AP constructs, and fixed and processed in the same way. Representative confocal images are shown in Figure 3c. Un-treated cells displayed a diffuse pattern of SA and Flag labeling throughout the cell, excluding the nucleus. In contrast, as has been shown by multiple groups, pharmacological kinase inhibition led to a re-distribution of LRRK2 into discrete cytoplasmic filaments (Figure 3c). Similar filamentous structures are observed in cells expressing the I2020T kinase domain mutant LRRK2 (Figure 3c), as has been previously reported by us , and others . We have previously reported that mutant LRRK2 filaments partially localize with microtubules , and are resistant to mild detergent extraction in situ . We found that filamentous LRRK2, induced by pharmacological inhibitor induced de-phosphorylation, or caused by the presence of a pathogenic mutation such as I2020T, is populated with dimeric LRRK2 species; with a near complete localization of fluorescent-SA signal to Flag-LRRK2 positive filaments in both conditions.
Dimer formation is altered by pathogenic PD-causing mutations in LRRK2
We, and others, have previously reported that certain pathogenic mutations in LRRK2 that are causative for PD alter the cellular distribution of overexpressed protein in a way that is similar to the re-distributions of LRRK2 following de-phosphorylation ([20,22,23]; and Figure 3c above). This is accompanied biochemically by a shift in elution of LRRK2 by SEC towards higher molecular mass fractions . To determine if PD-causing mutations alter LRRK2 dimer formation, we introduced several pathogenic mutations into the BirA and AP-LRRK2 constructs, and co-expressed each pair in HEK293T cells. Initially, we co-expressed both proximity biotinylation pairs harboring LRRK2 mutations; i.e. in a homozygous-like state. Subsequently, to mimic more closely the disease state, which is predominantly heterozygous (albeit, within the confounds of an overexpression system), we co-expressed WT BirA-LRRK2 together with mutant AP-LRRK2.
Extracts were bound to SA-coated wells and the amount of dimeric LRRK2 present determined using N241A/34-LRRK2-HRP or M2-Flag-HRP conjugates. Parallel samples were bound to c41-2 anti-LRRK2 coated ELISA plates and quantified using the N241A/34-LRRK2-HRP detection antibody, in order to normalize LRRK2 dimers to total LRRK2 expression levels. The common kinase domain mutant, G2019S–LRRK2, in a homozygous-like condition, exhibited a slight but non-significant increase in relative LRRK2 dimers in comparison with WT/WT LRRK2 dimers. In contrast, the other kinase domain mutant, I2020T-LRRK2, induced a significant increase in dimer formation in comparison with WT-LRRK2 (Figure 4), in a homo-dimeric conformation. For the ROC G-domain mutation, R1441C-LRRK2, we detected a significant increase in dimer formation in the homo-dimeric condition only (Figure 4) in comparison with WT/WT LRRK2 dimers.
LRRK2 dimer formation is not uniformly altered in different pathogenic mutant forms of LRRK2.
Kinase activity in dimeric vs. monomeric LRRK2
Overexpression of mutant LRRK2, particularly the G2019S mutant form, in cultured cells leads to a robust induction of LRRK2 auto-phosphorylation, and phosphorylation of model or physiological substrates. However, these reports, apart from the studies noted [10,11], do not distinguish between discrete pools of monomeric, hetero-dimeric, or homo-dimeric LRRK2. With the method of labeling LRRK2 dimers in situ described here, we are able to selectively measure the activity of each pool of LRRK2 derived from cells. Furthermore, the key feature of this system is the ability to selectively purify specific dimeric complexes; for example, in to model hetero-dimeric conditions, the overexpressed AP mutant LRRK2 fusion can only be biotinylated by the WT-BirA LRRK2 fusion. Thus, by assessing activity exclusively in the biotinylated pool of LRRK2, representing dimeric LRRK2, we are also able to directly compare the kinase function of homo-dimeric and hetero-dimeric species. We compared LRRK2 kinase activity, assessed by quantification of NICtide phosphorylation or auto-phosphorylation at Ser1292, in each pool of LRRK2 quantified in ng/ml by ELISA. The levels of biotinylated dimeric LRRK2 were estimated by a subtractive ELISA method whereby LRRK2 present in samples prior to, and following binding to SA-ELISA plates, thus following depletion of biotinylated dimeric LRRK2 (Figure 5a,b), were measured relative to recombinant full-length LRRK2. By this approach, we estimate the proportion of WT LRRK2 dimers within the extracts of HEK293T cells prepared here, to be ∼35% of the total detectable LRRK2 (Figure 5b). Confirming previous reports demonstrating increased kinase activity in species of LRRK2 isolated in fractions consistent with a dimeric complex, we find purified biotinylated dimeric LRRK2 displays greater phosphorylation of peptide substrates (NICtide; Figure 5c), and auto-phosphorylation at Ser1292 (pS1292; Figure 5d) in comparison with non-biotinylated, monomeric LRRK2. The pS1292 auto-phosphorylation score represents the ratio of phosphate incorporation into the S1292 residue following parallel kinase reactions carried out in the absence or presence of ATP in the reaction mixture, and then normalized to dimeric LRRK2 levels. The relative enrichment of kinase activity in the dimeric pool of LRRK2 is similar, or even greater, to what was reported by the groups of West and LaVoie [10,11]. In order to visualize the depletion of dimeric LRRK2 following incubation of cell extracts on SA-coated ELISA plates, we utilized immuno-transmission EM as previously described . Following a biotin pulse, extracts of cells expressing WT BirA- and AP-fused LRRK2 were incubated in SA-coated ELISA plates. Following capture of biotinylated dimeric LRRK2 by SA, the supernatant containing non-biotinylated (presumably monomeric) LRRK2 was retained and processed for immuno-gold labeling of Flag-BirA-LRRK2. Representative TEM micrographs are shown in Supplementary Figure S2a,b, where single-labeled monomeric LRRK2 as well as double-labeled LRRK2 dimers are visible. Quantification of single vs. double-labeled particles in lysates following SA incubation in comparison with input lysates shows a depletion of biotinylated LRRK2 dimers (Supplementary Figure S2c). While the immuno-labeling efficiency in TEM is <100%, these findings suggest that the majority of LRRK2 present within the supernatant following incubation with SA, to deplete biotinylated LRRK2 dimers, is in a monomeric conformation. These findings are consistent with the reduction in dimeric LRRK2 determined by ELISA following depletion of labeled dimers using SA-resin combined with SEC (see Supplementary Figure S1c).
In vitro kinase activity of purified LRRK2 is concentrated in the dimeric pool.
To directly compare hetero- vs. homo-dimeric LRRK2 activity in cells expressing pathogenic mutant forms of the protein, we co-expressed plasmids encoding mutant (R1441C, G2019S, or I2020T) LRRK2 fused to the BirA biotin ligase, together with either WT AP-LRRK2 or each of the mutant forms of AP-LRRK2. Thus, the final combinations were (as BirA/AP fusion pairs): WT/WT; WT/R1441C, R1441C/R1441C; WT/G2019S G2019S/G2019S; and WT/I2020T, I2020T/I2020T-LRRK2. As before, the cells were given a brief biotin pulse, lysed and bound to SA-coated ELISA plates to capture biotinylated dimeric LRRK2. In each condition, the presence of the BirA-LRRK2 partner in the SA-ELISA plate was confirmed by processing parallel wells for the Flag ELISA (e.g. see Figure 2d). Western immunoblots probed with total LRRK2 antibodies showed expression of both LRRK2 fusion proteins, seen as lower molecular mass band representing AP-LRRK2 and a higher molecular mass band representing the larger BirA-LRRK2 fusion (Figure 6a). Parallel aliquots of diluted cell extract were bound to c41-2 anti-LRRK2-coated and SA-coated ELISA plates. The dimeric biotinylated LRRK2 present in the SA-ELISA plate was subjected to an in-well NICtide phosphorylation assay, where the phosphorylation of the peptide substrate is subsequently quantified by pThr-X-Arg labeling. Following the kinase reaction, the amount of LRRK2 present in the SA- and total LRRK2-coated ELISA plates was determined by ELISA; and normalized to the total amount of LRRK2 present. The phosphorylation of NICtide is then normalized to the relative amount of dimers present in each well. In cells expressing either hetero- or homo-dimeric complexes of R1441C or I2020T-LRRK2, the in vitro kinase activity is not significantly different than WT dimers (Figure 6b). Interestingly, while in the case of homo-dimeric G2019S–LRRK2, the in vitro phosphorylation of NICtide is significantly elevated in comparison with WT-LRRK2, as is generally reported for overexpressed G2019S; hetero-dimeric G2019S/WT LRRK2 complexes exhibit kinase activity at similar levels to WT LRRK2 (Figure 6b). This indicates that in a dimeric complex, a single copy of mutant G2019S–LRRK2 is insufficient to significantly increase the total composite kinase activity.
In vitro phosphorylation of peptide substrates by purified mutant LRRK2 dimers.
To assess the stability of purified dimers during the kinase reaction, we quantified dimer formation in samples either prior to, or following, the in vitro kinase reaction. Dimers were measured by ELISA with anti-Flag as the detector, and normalized to total levels of LRRK2 determined by ELISA on parallel anti-LRRK2 coated plates. Surprisingly, we found that although WT/WT and G2019S dimers remain stable following the kinase reaction, dimers of I2020T or R1441C were significantly reduced following the kinase reaction (Figure 6c). Interestingly, both of these mutant forms of LRRK2 exhibit elevated dimer formation at baseline, in comparison with WT-LRRK2 (see Figure 4), and both segregate together in terms of 14-3-3 binding  and filament formation .
We also estimated LRRK2 kinase activity via quantification of induced auto-phosphorylation at Ser1292. As we found for the phosphorylation of NICtide, homo-dimeric complexes of G2019S–LRRK2 exhibited significantly elevated auto-phosphorylation at the Ser1292 site compared with WT homodimers, or WT/G2019S hetero-dimeric LRRK2 (Figure 7). Additionally, this activity was completely abolished by inclusion of the LRRK2 inhibitor czc-25146 (1 µM) in the reaction.
In vitro auto-phosphorylation at Ser1292 is increased only in homo-dimeric G2019S–LRRK2.
In addition to the conformation and activity of the pathogenic mutations in LRRK2 (R1441C, G2019S, and I2020T), we also examined the behavior of the strong risk factor G2385R in this system. The G2385R mutation was introduced into the BirA- and AP-fusion expression constructs, as for the other disease-linked mutations, and we assessed the amount of LRRK2 dimer formation, as well as the in vitro kinase activity in the purified pool of hetero- and homo-dimers. In contrast with G2019S, which had no effect on dimer formation, or the R1441C and I2020T mutations, which stabilize LRRK2 dimers, the G2385R mutation had a remarkable de-stabilizing effect on LRRK2 dimers (Figure 8a), both in the hetero-dimeric as well as homo-dimeric conformation. When we measured the phosphorylation of NICtide in purified G2385R hetero- or homo-dimers, the level of kinase activity was surprisingly high (Figure 8b), showing a degree of enhancement similar to that seen in purified G2019S–LRRK2. This indicates that, as with other forms of LRRK2, the kinase activity is concentrated within the dimeric pool of G2385R mutant LRRK2, which in this case is proportionally much smaller compared with WT or the other pathogenic mutations.
The Gly2385Arg risk factor variant of LRRK2 de-stabilizes dimer formation, yet retains robust activity within purified dimers in vitro.
To assess the activity of different species of dimeric LRRK2 in a cellular setting, in contrast with the in vitro activity of purified dimers as assessed above, we examined the phosphorylation of LRRK2 itself as well as the recently identified endogenous substrate, Rab10 . To examine the effect of dimer complex composition on the phosphorylation of LRRK2 in cells, we assessed by Western immunoblot the phosphorylation of LRRK2 at two distinct sites, Ser935 and Ser1292. Phosphorylation at Ser1292 is an auto-phosphorylation event, whereas Ser935 is phosphorylated by other kinases. In cells overexpressing BirA- and AP-fused WT LRRK2, or hetero- or homo-dimeric R1441C BirA and AP-fused LRRK2, we detected a specific decrease in phosphorylation at Ser935 only in cells overexpressing the homo-dimeric LRRK2 construct pair (Figure 9a). This is consistent with previous reports showing a decrease in phosphorylation of Ser935 in cells expressing LRRK2 mutated at this site . In contrast, phosphorylation at Ser1292, an auto-phosphorylation event, was increased in cells expressing homo-dimeric R1441C-LRRK2 (Figure 9a). While this is an indirect measure of the behavior of hetero- vs. homo-dimeric LRRK2 in cells, these findings are consistent with our in vitro data with purified mutant LRRK2 dimers, and indicates that at the cellular level, functional changes associated with mutant LRRK2 are more pronounced in the homo-dimeric conformation.
Dimer-specific changes in phosphorylation of LRRK2 and its substrate Rab10 in cells expressing homo-dimeric or hetero-dimeric mutant LRRK2.
The small GTPase Rab10, a recently identified phospho-substrate of LRRK2  is a key GTPase participating in a variety of cellular functions including vesicular and endosomal trafficking, phagocytosis in macrophages, and neuronal process growth (see  for review). LRRK2-mediated phosphorylation at the Thr73 residue (in human Rab10) is significantly enhanced by most mutations associated with PD, particularly mutations at the R1441 residue, and also in cells expressing the risk factor G2385R [24,26]. We assessed phosphorylation of Rab10 at Thr73 using a phospho-specific antibody in HEK293T cells co-expressing WT human Flag-Rab10 together with the BirA- and AP-fused LRRK2 constructs. A representative Western immunoblot probed with anti-LRRK2 (clone c41-2) shows expression of both LRRK2 constructs, as well as overexpressed Flag-Rab10 (Figure 9b). In cells expressing homo-dimeric R1441C, G2019S, and G2385R mutant LRRK2, phospho-T73 Rab10 levels are significantly elevated in comparison with WT LRRK2 dimers (Figure 9b; quantification in Figure 9c normalized to total Rab10). Additionally, we also found elevated, but not to the level of statistical significance, phospho-T73 Rab10 in cells expressing hetero-dimeric WT/R1441C LRRK2 (Figure 9c), which is consistent with previous studies showing the most robust increase in Rab10 phosphorylation occurring with R1441C/G mutant LRRK2 [24,26]. As stated, while these experiments only indirectly assess the effects on LRRK2 function and substrate phosphorylation in cells, when taken in the context of our above findings with in vitro purified hetero- and homo-dimeric mutant LRRK2, they support the notion that in most cases a single copy of mutant LRRK2 is insufficient to significantly alter the function of dimeric LRRK2 complexes.
Here we describe a method for the direct in situ labeling of LRRK2 dimers, and their subsequent purification, using an adaptation of the proximity biotinylation approach. In this technique, LRRK2 is biotinylated by its binding partner, a second LRRK2 molecule bearing the bacterial biotin ligase (BirA). Other approaches for assessing protein:protein interactions, such as bi-molecular fluorescence complementation (BiFC), are methods designed predominantly for the detection of protein dimers or oligomers. The formation of LRRK2 dimers, as well as higher order oligomers, is fairly well established, having been demonstrated by multiple groups and complimentary techniques (reviewed in ). Our approach offers a highly sensitive way to not only detect, but to purify LRRK2 dimers in their native conformation through the exploitation of the very small biotin tag and the high affinity interaction between biotin and streptavidin. While this technique cannot discriminate between dimeric LRRK2 complexes, and larger multi-meric LRRK2 complexes, the current evidence favors the existence of dimeric LRRK2 as the dominant multi-meric conformation, as opposed to larger scale complexes composed of three or more LRRK2 molecules. Specifically, several recent reports, using immuno-coupled negative stain EM imaging , cryo-EM imaging , and immunotransmission EM imaging , demonstrated the existence of LRRK2 dimers in highly purified preparations of cell-produced LRRK2. Particularly in the immunolabeled preparations of purified LRRK2, apart from singly labeled LRRK2, only dimeric species, consisting of two closely aligned immuno-gold particles have been reported. Thus, while higher-order complexes, with an estimated mass exceeding the predicted size of a LRRK2 dimer, have been detected, it is likely that these comprises dimeric LRRK2 interacting with other proteins.
Little is known about the regulation of LRRK2 dimerization; however, it has been demonstrated ([10,11]; present work) and consistent with other related kinases, its activity is tightly linked to this conformational transition. For example, the dimerization of the related kinase RIPK3 has been shown to be both sufficient and necessary for the induction of necroptosis . This is consistent with the finding that hetero-dimers of RIPK1 and RIPK3 were able to induce necroptotic death, however only when additional RIPK3 monomers are recruited and bound to the RIPK1/RIPK3 complex , likely via a direct interaction with the resident RIPK3. Similarly, the kinase activity of RIPK1, an interacting partner of LRRK2 [31,32], integral to its induction of necroptosis, is dependent upon its dimerization. It was recently shown that mutation of a single residue within the DD of RIPK1, while having no direct effect on basal intrinsic kinase activity, blocks dimerization of RIPK1 and the induced auto-phosphorylation at Ser166 . Kinase-dependent necroptotic cell death was restored in mutant RIPK1 upon forced dimerization . Whether mutation of a specific residue within LRRK2 can similarly prevent dimerization and block pathogenic mutant LRRK2-induced neuronal death is unclear; however, given that the kinase activity of LRRK2 is known to be predominantly associated in dimeric complexes, and that inhibition of kinase activity in models of mutant LRRK2 neurodegeneration is protective, this notion merits investigation. A separate study from this same group recently demonstrated that a specific pool of insoluble and ubiquitinated RIPK1 (iuRIPK1) interacted with LRRK2 . Loss of LRRK2 in this specific complex was partially protective against the induction of RIPK1-dependent apoptosis (RDA); likely due to a role as a protein scaffold, stabilizing the presence of iuRIPK1 in TNF-α/5Z7 (a TAK1 inhibitor)-induced complex I, as LRRK2 kinase inhibitors did not produce similar protection against RDA . Perhaps in this unqiue complex with RIPK1, as LRRK2 kinase inhibition had no effect, LRRK2 is primarily monomeric; although this is speculative, as it was not established in the study of Amin et al. . While LRRK2 clearly has myriad cellular roles, its involvement in death signaling, whether in the context of mutant LRRK2 interacting with FADD and caspase-8 [13,32] or LRRK2 stabilizing iuRIPK1 in RDA, is characterized by the unifying feature of multi-protein complex formation. Many distinct death-inducing protein complexes, involving different combinations of the RIP protein kinase family, have been described. These complexes mediate the induction of multiple forms of cell death including apoptosis and necroptosis, as well as certain sub-types within each class (e.g. RDA, or extrinsic/intrinsic apoptosis in the case of mutant LRRK2 [13,32]. While the focus of this study was to characterize the kinase activity within specific mutant LRRK2 dimeric species, this has clear implications for the study of the kinase-dependent induction of neuronal death in models of LRRK2-PD. We have previously shown that in cells expressing a double mutant form of LRRK2 comprised of pathogenic (e.g. I2020T-LRRK2) and a kinase inactivating mutation (K1906R), increased association of mutant LRRK2 with the death adaptor protein FADD, was normalized to WT-LRRK2 levels , as is the ability of mutant LRRK2 to form filaments . This suggests that the death-inducing complex formed by mutant LRRK2 and death-signaling proteins such as FADD is dimeric.
The phosphorylation of LRRK2 itself can control many aspects of its activity (e.g. ), as well as its localization (e.g. S935; [20–22]), raising the possibility that phosphorylation or auto-phosphorylation at specific sites within LRRK2 can signal the transformation to a dimeric state. GTP binding inhibitors decrease filament formation , which we show here, are populated with LRRK2 dimers. In our assay, we find that treatment of cells with LRRK2 kinase inhibitors, such as MLi-2, do not significantly alter the formation of LRRK2 dimers labeled in situ. Such treatment would induce broad de-phosphorylation of LRRK2, presumably at most if not all LRRK2-dependent phospho-sites. If phosphorylation at a specific residue triggers the transition from monomeric to dimeric LRRK2, we would predict that MLi-2 treatment would suppress dimer formation. The fact that it does not raises the possibility that multiple sites of regulation, with opposing actions, exist within LRRK2. A comprehensive study of the intrinsic regulation of LRRK2 dimerization is underway, and will be reported separately.
The approach we have developed provides a vital tool for the study of the functional regulation and outcomes of LRRK2 dimer formation. Because of the fact that in this system, the AP acceptor peptide can only become biotinylated by the bacterial BirA biotin ligase; and conversely, the BirA ligase can only label this AP motif, we can be certain that the biotinylated LRRK2 dimers that are purified are of the exact composition that we intend (e.g. hetero-dimeric WT and mutant LRRK2, vs. homo-dimeric mutant LRRK2). This permits the direct comparison of their relative activities, and can also be exploited for the identification of specific interacting partners, such as other death-signaling proteins or specific phosphatase complexes, that may interact selectively with homo-dimeric or hetero-dimeric LRRK2. Moreover, we can also gain structural insight into the formation of LRRK2 dimers, to complement the in silico and biochemical findings recently reported . In our system, when the AP motif was fused to the C-terminus of LRRK2, the biotinylation efficiency was greatly reduced in comparison with N-terminally fused AP. This suggests that, at least under the conditions employed here, within the dimeric complexes labeled during the brief biotin pulse, the two N-termini are in close apposition. In the structural model of full-length LRRK2 recently described by Guaitoli et al. , the final model proposes a complex structural organization with several notable features. First, it suggests that the N-terminal ankyrin domain is in close proximity to the kinase domain, within the folded dimeric complex, possibly exerting a regulatory effect on kinase activity. Secondly, the final model, constructed in part based on EM maps, suggests that the two N-terminal armadillo domains are situated on the top surface of the dimer complex in apposition to one another. Depending upon how the fused BirA ligase fits within this proposed concave opening between the two N-terminal ends of LRRK2 in relation to the AP motif located on the other LRRK2 binding partner, their proximity could conceivably be within the range necessary for BirA-mediated biotinylation of the AP motif. This is in contrast with the report of Sejwal et al. , employing cryo-EM techniques to examine the structure of isolated LRRK2 complexes, which suggests that the LRRK2 dimer is arranged in an antiparallel orientation. While the resolution attained in this study, employing a cryo-EM approach, is higher than previously achieved, in the absence of additional constraints provided by chemical cross-linking and SAXS, as employed in the study of Guatoli et al., the domain organization of the LRRK2 dimer proposed by Sejwal et al. can only be inferred indirectly. It is possible as well that the different buffer conditions used in the two studies may alter the binding partners co-purifying with dimeric LRRK2, which could alter the alignment of the interacting LRRK2 pairs.
Clinical studies thus far have shown that mutant LRRK2 (in these reports, G2019S–LRRK2) has greater kinase activity in comparison with control subjects, at least in peripheral biofluids [35,36]. Urinary exosomes contain LRRK2 with significantly increased phosphorylation at the Ser1292 residue, an auto-phosphorylation site, indicating that LRRK2 present in the cells from which these exosomes originate is more active. Moreover, a recent study from Di Maio et al.  used a novel approach to label phosphorylated LRRK2 (at Ser1292), together with a loss of LRRK2 interaction with 14-3-3, in iPD brain dopamine neurons and the rotenone model of PD. These changes in LRRK2 were accompanied by increased phosphorylation of Rab10 in dopamine neurons of iPD brain compared with healthy controls . This indicates that even in the absence of mutations in LRRK2, PD pathogenesis is broadly linked to increased activation of LRRK2. Similarly, in vivo models of mutant LRRK2-induced neurodegeneration have also established kinase-dependency in the loss of neuronal survival [38–40]. In the present study, we have established that purified homo-dimers of mutant LRRK2 possess greater kinase activity in comparison with hetero-dimers. In the case of G2019S and R1441C mutant LRRK2 this was evident in multiple independent indices of kinase activity: in vitro phosphorylation of a peptide substrate, in vitro auto-phosphorylation of Ser1292, and in cells co-expressing the endogenous substrate Rab10. For the R1441C mutation, localized to the ROC GTPase domain, it has previously been reported that mutation of LRRK2 at this site causes an extremely robust increase in phosphorylation of Rab10, greater than G2019S–LRRK2 [24,26]; and we have similarly found that R1441C-LRRK2 dimers lead to greater phosphorylation of Rab10.
Another key finding in this study concerned the apparent stability of dimers comprised specific mutant forms of LRRK2. Homo-dimers containing two copies of G2019S–LRRK2 remain at constant levels, similar to WT homo-dimers, following activation of its kinase. In contrast, mutant homo-dimers of I2020T- or R1441C-LRRK2 were reduced following a kinase reaction. Both of these mutant forms of LRRK2 exhibit increased levels of dimer formation, relative to WT, at baseline; thus following the kinase reaction, dimer formation was reduced to WT levels. This cannot be a factor of increased overall kinase activity, as dimers of G2019S–LRRK2 are similar to WT, yet its kinase activity is markedly increased. It is possible that I2020T and R1441C LRRK2 share an overlapping, and unique, de novo panel of substrates; a notion supported by their shared propensity to re-localize to cytoplasmic filamentous structures [22,23]. Since auto-phosphorylation of specific ROC domain residues can alter its GTPase activity, and perhaps induce localized structural changes , one possibility is that I2020T and R1441C-induced auto-phosphorylation at certain residues prime the dimeric complex for more rapid re-cycling back to a monomeric state. Moreover, it was similarly shown that dimers of LRRK2 show reduced stability following an in vitro GTPase assay , further indicating the link between LRRK2 enzymatic function and conformation.
The vast majority of familial LRRK2-dependent PD cases are heterozygous. These findings raise the possibility that a critical factor in disease pathogenesis may be the ratio of mutant homo-dimers to WT/mutant hetero-dimers; suggesting that a gradual buildup of mutant homo-dimers, associated with significantly elevated kinase activity, may precede the onset of PD symptoms. While the clinical picture from the limited number of homozygous G2019S carriers suggests that there is not a gene-dosage effect, that the phenotype is similar to heterozygous carriers ; our current findings indicate that the G2019S mutation does not increase dimer formation, as happens for the I2020T kinase domain mutation. Thus, the build-up of dimeric mutant LRRK2 may still be a critical factor in PD pathogenesis. Similarly, the recent findings of Di Maio et al.  that demonstrate increased LRRK2 activity, as evidenced by auto-phosphorylation at Ser1292 and phosphorylation of Rab10, when taken in the context of our findings, likely reflect increased LRRK2 dimerization. This highlights the need for more sensitive tools to assess not only the activation state of LRRK2 in clinical samples of those affected by, or at risk of developing PD, but to also determine its conformational state.
H.J.R. conceived the study and wrote the manuscript; M.L., E.M., A.M., K.M. performed experiments and analyzed the results; E.G. performed the T.E.M. image analysis and analyzed the results.
This work was supported by a grant from the Michael J. Fox Foundation for Parkinson's Research (H.J.R.; grant ID 8418.01).
The authors would like to acknowledge the technical assistance of Matina Maniati (BRFAA), and the Electron Microscopy Core Facility (Padova).
The Authors declare that there are no competing interests associated with the manuscript.