Research ArticleParkinson’s Disease

Ser1292 Autophosphorylation Is an Indicator of LRRK2 Kinase Activity and Contributes to the Cellular Effects of PD Mutations

See allHide authors and affiliations

Science Translational Medicine  12 Dec 2012:
Vol. 4, Issue 164, pp. 164ra161
DOI: 10.1126/scitranslmed.3004485

Abstract

Mutations in the leucine-rich repeat kinase 2 (LRRK2) gene are the most common cause of familial Parkinson’s disease (PD). Although biochemical studies have shown that certain PD mutations confer elevated kinase activity in vitro on LRRK2, there are no methods available to directly monitor LRRK2 kinase activity in vivo. We demonstrate that LRRK2 autophosphorylation on Ser1292 occurs in vivo and is enhanced by several familial PD mutations including N1437H, R1441G/C, G2019S, and I2020T. Combining two PD mutations together further increases Ser1292 autophosphorylation. Mutation of Ser1292 to alanine (S1292A) ameliorates the effects of LRRK2 PD mutations on neurite outgrowth in cultured rat embryonic primary neurons. Using cell-based and pharmacodynamic assays with phosphorylated Ser1292 as the readout, we developed a brain-penetrating LRRK2 kinase inhibitor that blocks Ser1292 autophosphorylation in vivo and attenuates the cellular consequences of LRRK2 PD mutations in vitro. These data suggest that Ser1292 autophosphorylation may be a useful indicator of LRRK2 kinase activity in vivo and may contribute to the cellular effects of certain PD mutations.

Introduction

Parkinson’s disease (PD) is the second most common neurodegenerative disease, affecting 1 in 100 people over the age of 60 years in the United States. The characteristic motor symptoms of PD derive from degeneration of dopaminergic neurons in the substantia nigra pars compacta and include tremor, rigidity, bradykinesia, and postural instability. Whereas these motor symptoms can be temporarily managed by dopamine replacement therapy, the nonmotor symptoms of PD, which include dementia, depression, sleep disorders, and sometimes psychosis, stem from degeneration in other parts of the brain and are not ameliorated by dopamine therapy (1). Currently, there are no disease-modifying therapies for PD.

Genetic studies have revealed a series of genes that are associated with familial and idiopathic forms of PD (2). Among these, mutations in the leucine-rich repeat kinase 2 (LRRK2) gene are the most common cause of familial PD, affecting 1 to 2% of patients in Western countries (3). Genome-wide association studies have further shown that polymorphisms in the LRRK2 locus increase the probability of developing sporadic PD (4, 5). The LRRK2 gene encodes a 2527–amino acid protein composed of a leucine-rich repeat (LRR), Ras of complex (ROC) domain, C terminus of Roc (COR) domain, kinase and WD40 domains (Fig. 1A). The most common LRRK2 mutation is located within the kinase activation loop (G2019S) (6). Various in vitro biochemical studies have shown that the LRRK2 G2019S mutation results in enhanced kinase activity. In cellular models, familial PD mutations including G2019S induce cell death and neurite outgrowth defects, whereas kinase inhibitors and LRRK2 kinase-dead mutations reduce these cellular phenotypes (711). It has thus been suggested that aberrant LRRK2 kinase activation might contribute to PD pathology (12). However, it remains controversial whether familial PD mutations besides G2019S increase LRRK2 kinase activity (13). The mechanistic link between LRRK2 kinase activity and neuronal toxicity also remains unclear, in part because in vivo substrates of LRRK2 remain unknown (13).

Fig. 1

Identification of an LRRK2 autophosphorylation site. (A) Diagram depicts the domain architecture of LRRK2 with sites for PD mutations and Ser1292 indicated. (B and C) Annotated MS/MS spectra that unambiguously identify pSer1292 and pThr1343. (D and E) Extracted ion chromatograms for pSer1292 SFPNEMGKLS#K and pThr1343 LMIVGNT#GSGK peptides from reactions of G2019S mutant LRRK2 at times from 0 to 60 min. The isotopically labeled internal standards of pSer1292 (13C615N1 Leu; +7.017 daltons) and pThr1343 (13C515N1 Val; +6.014 daltons) are shown at the back of each three-dimensional plot, with the signal in the front representing the analyte peptide derived from the digested sample. (F and G) Extracted ion chromatograms for regular and isotopically labeled pSer1292 peptides from FLAG-tagged LRRK2 immunoprecipitated from transfected HEK293 cells. Light pSer1292 can be detected in G2019S mutant LRRK2 but not in kinase-dead D1994A mutant LRRK2.

Like many kinases, LRRK2 is phosphorylated on multiple residues within its sequence through both autophosphorylation and the actions of other kinases. Early work showed phosphorylation of a series of LRRK2 residues immediately upstream of the LRR repeats (14). Whereas these constitutive modifications do not derive from autophosphorylation, recent studies suggest that they may be modulated indirectly by LRRK2 kinase activity (15, 16). Multiple groups have reported in vitro LRRK2 autophosphorylation sites including notable clusters within the guanosine triphosphatase (GTPase) and kinase domains (14, 1720). Cookson and colleagues used mass spectrometry (MS) to first demonstrate LRRK2 autophosphorylation of Thr1343 and Thr1491 within the GTPase domain. Ueffing and colleagues identified additional LRRK2 autophosphorylation sites, remarking that the LRRK2 protein displayed a low degree of autophosphorylation. Phospho-specific antibodies developed toward phosphorylated Thr1503 (pThr1503) (17) and pThr1967 (19) recognize specific phospho-epitopes on purified proteins after in vitro autophosphorylation reactions. Despite these efforts, few cellular or in vivo data on LRRK2 autophosphorylation are available, making it unclear whether autophosphorylation activity is involved in the pathology associated with LRRK2 PD mutations.

To investigate the cellular effects of LRRK2 kinase activity, we used quantitative MS to extend our understanding of LRRK2 autophosphorylation and identified Ser1292 as an in vivo LRRK2 autophosphorylation site. Furthermore, we investigated how familial PD mutations affect Ser1292 autophosphorylation in cultured cells and in animals, and whether Ser1292 autophosphorylation contributes to the cellular effect of PD mutations. Finally, we used phosphorylated Ser1292 (pSer1292) as a readout for LRRK2 kinase activity to develop assays for drug discovery and identified a potent, specific, and brain-penetrable LRRK2 kinase inhibitor.

Results

Ser1292 as an in vivo LRRK2 autophosphorylation site

To better understand the biochemical activity of LRRK2, we carried out in vitro autophosphorylation assays using purified G2019S and kinase-dead (D1994A) LRRK2 constructs representing residues 970 to 2527, as previously described (18). Reactions were performed in the presence of adenosine triphosphatase (ATP) (1 mM) and guanosine 5′-O-(3′-thiotriphosphate) (GTPγS) (10 mM) and separated by SDS–polyacrylamide gel electrophoresis. Coomassie-stained LRRK2 bands were digested and analyzed by liquid chromatography (LC)–MS/MS. Database searches revealed a series of spectra from the G2019S sample confidently matching to phosphopeptides of LRRK2 (table S1), whereas no sites were observed in reactions with the D1994A kinase-dead mutant LRRK2. Most of our identified sites overlap with previously published work, including the cluster of phosphopeptides spanning the ROC domain (Fig. 1, B and C, and fig. S1, A to C). These results provide direct evidence for autophosphorylation at Ser1366 while supporting a previous report that LRRK2 more frequently modifies Thr residues than Ser in vitro (17, 21) (fig. S2, A and B, and table S1). Among the identified LRRK2 autophosphorylation sites, we noted an unusual frequency of Met residues in proximity to LRRK2 phosphorylation sites, with 9 of 14 peptides displaying a Met between one and five positions upstream of the modified residue. Among the identified spectra was one matching pSer1292. pSer1292 (Fig. 1B) was also identified in the G2019S negative control reaction, indicating that a fraction of LRRK2 was phosphorylated at Ser1292 in the expression system and maintained through purification.

To assess the relative abundance of phosphopeptides in vitro and to assist with site validation in vivo, we synthesized a series of internal standard peptides (table S2). These peptides comprised the same sequences as identified LRRK2 phosphopeptides, but included a single amino acid enriched in stable isotopes (denoted by underline) (22). Among these were isotopically labeled peptides directed toward pSer1292 SFPNEMGKLS#K (13C615N1 Leu; +7.017 daltons on underlined residue), pThr1343 LMIVGNT#GSGK (13C515N1 Val; +6.014 daltons), pThr1357 TTLLQQLMKT#K (13C615N1 Leu; +7.017 daltons), and pThr1368 SDLGMQSAT#VGIDVK (13C515N1 Val; +6.014 daltons), with S# or T# denoting the position of phosphorylation. Assessing the phosphorylation status of Ser1357 and Ser1368 required additional peptides to account for the ragged termini flanking these sequences (table S2).

In vitro kinase reactions were again performed, this time using purified wild-type, G2019S and R1441C mutant LRRK2 proteins (970 to 2527). Time points were collected at 0, 5, 10, 30, and 60 min for analysis (fig. S1D). GTPγS was omitted from these reactions given that published work (23, 24) and our pilot studies (fig. S3) both indicated that it had no effect on LRRK2 in vitro kinase activity. Extracted ion chromatograms for pSer1292 and pThr1343 peptides from the G2019S reactions are shown relative to their coeluting internal standards (Fig. 1, D and E). In line with the phosphomapping studies, a small amount of pSer1292 was observed at the onset of the kinase reaction (Fig. 1D), increasing steadily out to 60 min. In contrast, Thr1343 autophosphorylation occurred between 0 and 10 min before reaching a plateau. Individual sites displayed unique kinetics, with autophosphorylation of pThr1343 and pThr1357 rapidly reaching plateaus, whereas phosphorylation at Ser1292 and Thr1368 continued robustly even out to the 60-min time point (fig. S1, E to H).

We next used our quantitative LC-MS assay to assess LRRK2 autophosphorylation activity in living cells. FLAG epitope-tagged versions of G2019S or D1994A mutant LRRK2 were expressed and purified from the human embryonic kidney (HEK) 293 cell line. As suggested by previous work, most LRRK2 autophosphorylation events were undetectable, even when using isotopically labeled internal standards as references. The only autophosphorylation site observed in HEK293 cellular lysates was pSer1292. Extracted ion chromatograms revealed the presence of pSer1292 in immunopurified FLAG-labeled G2019S mutant LRRK2 (Fig. 1F) but not in immunopurified FLAG-tagged kinase-dead D1994A mutant LRRK2 (Fig. 1G). MS/MS spectra acquired on the coeluting internal standard and analyte peptides confirmed the presence of pSer1292 in immunopurified FLAG-tagged G2019S mutant LRRK2.

Given our finding that pSer1292 was an autophosphorylation site in vitro and in cultured cells, a rabbit polyclonal antibody was generated to specifically recognize this phospho-epitope (anti-pSer1292). Western blots using anti-pSer1292 antibody showed signal in lysates of HEK293 cells transfected with wild-type LRRK2 but not in cells transfected with either S1292A (phospho-site dead) or D1994A (kinase-dead) mutant constructs (Fig. 2A). The anti-pSer1292 antibody also detected a specific signal from whole-brain lysates of bacterial artificial chromosome (BAC) transgenic mice overexpressing wild-type LRRK2 (25) but not from BAC transgenic mice overexpressing kinase-dead LRRK2 (Fig. 2B). These results demonstrate that LRRK2 kinase activity is required for phosphorylation of Ser1292 in vivo.

Fig. 2

Several familial PD mutations increase Ser1292 autophosphorylation. (A) Western blots detecting pSer1292 LRRK2 in lysates of HEK293 cells transiently expressing FLAG-tagged wild-type LRRK2 or S1292A or D1994A mutant LRRK2. (B) Western blots detecting pSer1292 autophosphorylation of LRRK2 in whole-brain lysates from BAC transgenic (Tg) mice overexpressing wild-type or kinase-dead mutant LRRK2. (C) Western blots detecting pSer1292 and total LRRK2 signal from lysates of HEK293 cells transiently transfected with wild-type LRRK2 or LRRK2 with the following PD mutations: N1437H, R1441G, R1441C, Y1699C, G2019S, I2020T, G2385R, or the R1441G/G2019S (RG/GS) double mutant. (D) Quantification of the ratio of pSer1292 to total LRRK2. Y axis is the logarithm of the ratio to ensure similar variance among different constructs. Means ± SEM for each construct are shown. *P < 0.0001, analysis of variance (ANOVA) followed by Tukey-Kramer test (mutant was significantly different from wild-type). The double mutation R1441G/G2019S (RG/GS) was significantly different from either of the single mutations. #P < 0.0001, one-way ANOVA followed by Tukey-Kramer test. (E) Western blot detecting pSer1292 and total LRRK2 in total brain lysates from BAC transgenic mice expressing wild-type and G2019S mutant LRRK2. (F) Quantification of the ratio of pSer1292 over total LRRK2 normalized to the ratio in LRRK2 wild-type BAC transgenic mice in (E) (n = 5 animals per genotype, mean ± SEM). ***P < 0.0001, Student’s t test. (G) Western blot detecting pSer1292 and total LRRK2 in lysates from different brain regions of 12-month-old BAC transgenic mice expressing FLAG-tagged wild-type and G2019S mutant LRRK2.

PD mutations increase LRRK2 autophosphorylation on Ser1292

In biochemical assays, the G2019S mutation of LRRK2 has been shown to increase kinase activity toward itself, myelin basic protein, and a generic peptide substrate (LRRKtide) (6, 14, 26). Published results suggest that LRRK2 PD mutations besides G2019S, such as R1441G/C, Y1699C, and I2020T, may not increase kinase activity in vitro (13). Using pSer1292 as a readout, we tested the effects of various LRRK2 familial PD mutations and a PD risk factor mutation found in Asian populations (G2385R) (27) on autophosphorylation activity in HEK293 cells. Lysates were prepared from HEK293 cells transfected with each of a series of FLAG-LRRK2 expression constructs and were Western-blotted for total LRRK2 and pSer1292. Five of the six familial PD mutations tested including N1437H, R1441G, R1441C, G2019S, and I2020T displayed statistically significant elevation (P < 0.0001) of pSer1292 compared to wild-type LRRK2 (Fig. 2, C and D). Consistent with previously published in vitro autophosphorylation results (17), combining R1441G and G2019S mutations (RG/GS) in the same construct further increased LRRK2 autophosphorylation (Fig. 2, C and D). Notably, the Y1699C and G2385R mutations did not increase autophosphorylation at Ser1292.

We next tested whether PD mutations increased pSer1292 autophosphorylation in BAC transgenic mice overexpressing wild-type or G2019S mutant LRRK2. In whole-brain lysates from the BAC transgenic mice, pSer1292 was >10 times higher for G2019S mutant LRRK2 compared to the wild-type protein (Fig. 2, E and F). Total LRRK2 protein concentrations were comparable between the two cohorts of BAC transgenic mice, but pSer1292 was significantly increased for transgenic mice overexpressing G2019S mutant LRRK2 compared to transgenic mice overexpressing wild-type LRRK2 in all brain regions examined including cerebellum, striatum, cortex, olfactory bulb, and midbrain (Fig. 2G). Together, these data demonstrate that some LRRK2 PD mutations enhance Ser1292 autophosphorylation in cultured cells and in brain lysates of BAC transgenic mice.

Biochemical mechanisms of LRRK2 Ser1292 autophosphorylation

The mechanism of LRRK2 autophosphorylation remains unclear. Several reports suggest that an intact GTP binding site in the ROC GTPase domain is required for LRRK2 kinase activation in vitro (14, 28, 29). To test whether GTP binding site mutations affect LRRK2 autophosphorylation in cells, we examined pSer1292 concentrations in HEK293 cells transfected with T1348N or G2019S/T1348N mutant LRRK2 constructs. In lysates from HEK293 cells expressing the T1348N mutant LRRK2, pSer1292 was not detectable (Fig. 3A). Autophosphorylation was greatly reduced in cells transfected with the G2019S/T1348N double mutant LRRK2 compared to those expressing G2019S mutant LRRK2 alone (Fig. 3A).

Fig. 3

LRRK2 autophosphorylation on Ser1292 requires an intact GTP binding site and may be intramolecular. (A) Representative Western blot detecting pSer1292 and total LRRK2 in lysates of HEK293 cells transfected with FLAG-tagged wild-type LRRK2 or LRRK2 with the following mutations: T1348N (GTP binding site mutation), G2019S, and T1348N/G2019S. (B) Schematic of two constructs, one with the R1441G/Y1699C/G2019S/S1292A (LRRK2 3M/S1292A) mutation and the other with the D1994A kinase-dead mutation, cotransfected in HEK293 cells (left). Representative Western blot detecting pSer1292 in lysates of HEK293 cells transiently expressing FLAG-tagged LRRK2 carrying the R1441G/Y1699C/G2019S triple mutation (3M), 3M/S1292A, or the D1994A mutation, or D1994A and 3M/S1292A mutations (right). (C) Schematic of two cotransfected constructs showing the full-length LRRK2 (amino acids 1 to 2527) and the truncated (amino acids 907 to 2527) kinase-dead mutant (D1994A) (left). Representative Western blot detecting pSer1292 and total LRRK2 in lysates from HEK293 cells coexpressing full-length G2019S mutant LRRK2 and truncated LRRK2 carrying the D1994A mutation (right).

Another unresolved question is whether LRRK2 autophosphorylation occurs in cis or trans. One possibility is that pSer1292 becomes modified via an intramolecular reaction (cis), whereas an alternate mechanism postulates that an LRRK2 molecule transfers a phosphate to Ser1292 on another LRRK2 molecule (trans). We reasoned that if LRRK2 autophosphorylation occurred in trans, the active kinase domain of LRRK2 should be able to phosphorylate Ser1292 on a kinase-dead LRRK2 molecule (Fig. 3B). To test this, we cotransfected HEK293 cells with R1441G/Y1699C/G2019S/S1292A (3M/S1292A, phospho-site dead) and D1994A (kinase-dead) LRRK2 expression constructs. Neither of these mutant proteins alone can autophosphorylate via either intra- or intermolecular mechanisms (Fig. 3B). Arguing against intermolecular LRRK2 autophosphorylation, no pSer1292 was observed in lysates from HEK293 cells cotransfected with LRRK2 3M/S1292A and D1994A constructs (Fig. 3B). To substantiate this finding, we tested whether full-length LRRK2 was able to phosphorylate an N-terminally truncated version of LRRK2. Either N-terminally truncated (907 to 2527) G2019S or D1994A mutant LRRK2 was cotransfected with full-length mutant G2019S LRRK2 in HEK293 cells. Using anti-pSer1292 antibody, Western blots showed that whereas the full-length and the truncated G2019S LRRK2 protein was phosphorylated, little or no pSer1292 was observed on the truncated D1994A LRRK2 (Fig. 3C). These results indicate that full-length G2019S LRRK2 cannot efficiently phosphorylate truncated LRRK2 D1994A via an intermolecular mechanism. Given these findings, we suggest that LRRK2 autophosphorylation at Ser1292 occurs in cis via an intramolecular mechanism.

Ser1292 autophosphorylation is required for neurite outgrowth defects due to LRRK2 PD mutations

Ser1292 is located at the junction of the LRR and the ROC GTPase domains of LRRK2. Although the sequence conservation of LRRK2 across species is limited, Ser1292 and the sequence surrounding it are highly conserved from worm to human (fig. S4). This hints at the possibility that LRRK2 autophosphorylation at Ser1292 may be important for its biological function. Several reports have shown that in cultured primary neurons, overexpression of LRRK2 protein carrying PD mutations causes neurite outgrowth defects that are dependent on altered kinase activity (79). To investigate whether pSer1292 contributes to neurite outgrowth defects elicited by LRRK2 PD mutations, we cultured and transfected primary hippocampal neurons from rat embryos at embryonic day 18 with various LRRK2 mutant constructs (Fig. 4A). Rat embryonic primary hippocampal neurons expressing R1441G/G2019S double mutant LRRK2 showed a robust neurite outgrowth defect compared with those expressing wild-type LRRK2 or a green fluorescent protein (GFP) alone (Fig. 4, B and C). Under our experimental conditions, the single mutation R1441G or G2019S did not show robust neurite outgrowth defects (Fig. 4D). Consistent with published reports, introduction of the D1994A kinase-dead mutation into the R1441G/G2019S double mutant LRRK2 construct abolished this neurite outgrowth defect (Fig. 4, B and C). Likewise, mutation of the LRRK2 autophosphorylation site (S1292A) also rescued the neurite outgrowth defect associated with the R1441G/G2019S double mutant (Fig. 4, B and C). Notably, a presumably “phosphomimetic” mutant S1292D also reversed the neurite outgrowth defect induced by the R1441G/G2019S LRRK2 double mutant (Fig. 4C).

Fig. 4

Ser1292 is required for LRRK2 PD mutations to induce neurite outgrowth defects. (A) Neurolucida tracing of representative cultured rat embryonic hippocampal neuron labeled with GFP alone as control. (B) Neurolucida tracing of rat embryonic primary hippocampal neurons expressing GFP and wild-type LRRK2, or LRRK2 containing different PD mutations. Scale bars, 100 μm. (C) Quantification of total dendrite length of hippocampal neurons expressing indicated LRRK2 constructs normalized to the average dendrite length of neurons transfected with GFP (n = 40 to 85 cells). Mean diamonds are shown for each group. Statistically significant difference between mean of R1441G/G2019S and mean of any of the other groups (*P < 0.0001). (D) Quantification of total dendrite length of hippocampal neurons expressing indicated LRRK2 constructs normalized to the average dendrite length of neurons transfected with GFP (n = 18 to 23 cells per group). Mean diamonds are shown for each group. No statistically significant difference was observed among the groups. (E) An overlay of concentric circles, spaced at 20-μm intervals, centered at the cell body was used for Sholl analysis. Sholl analysis shows the number of branching intersections for neurites crossing each circle (mean ± SEM). *P < 0.05, one-way ANOVA followed by Tukey-Kramer test (n = 40 to 85 cells per construct).

In addition to shortening total neurite length, Sholl analysis of branch points showed that the R1441G/G2019S LRRK2 double mutant also reduced the complexity of the dendrites of cultured rat embryonic primary hippocampal neurons (Fig. 4E). As with neurite length phenotypes, introduction of either the D1994A or the S1292A mutation to the G2019S/R1441G LRRK2 double mutant expression construct ameliorated the reduction in neurite branching and restored neurite complexity (Fig. 4E).

Ser1292 autophosphorylation may affect LRRK2 PD mutant protein localization

To determine whether autophosphorylation of Ser1292 alters LRRK2 kinase activity, we transiently transfected FLAG-tagged versions of wild-type and S1292A LRRK2 into HEK293 cells. The kinase activities of immunoprecipitated LRRK2 were assessed in vitro using a radioactive autophosphorylation assay and the peptide substrate LRRKtide (26). No significant differences in kinase activity were observed between wild-type and S1292A LRRK2 proteins (fig. S6), suggesting that pSer1292 does not directly regulate LRRK2 kinase activity in vitro. However, it remains possible that the S1292A mutation may affect LRRK2 kinase activity in vivo. A second direct readout of LRRK2 kinase activity in vivo would be needed to test this hypothesis.

Next, we examined the subcellular localization of LRRK2 PD mutant proteins within cultured rat embryonic primary cortical neurons using immunofluorescence microscopy. FLAG-tagged LRRK2 constructs were transfected into cultured rat embryonic primary cortical neurons and subsequently detected using anti-FLAG antibody. Wild-type LRRK2 protein was diffusely distributed within the neurons; however, neurons expressing the R1441G/G2019S double mutant protein showed a filamentous distribution of LRRK2 (Fig. 5, A and B). The LRRK2 double mutant colocalized with the microtubule marker tubulin and the autophagy adaptor protein p62 (Fig. 5D). In some neurons, the R1441G/G2019S double mutant LRRK2 was observed in large puncta (Fig. 5, A and C). These puncta were positive for autophagy markers p62, LC3, and K63 polyubiquitin (Fig. 5D), indicating that they are autophagosomes. The kinase-dead mutation (D1994A) in the R1441G/G2019S LRRK2 double mutant abolished both filamentous and puncta distribution (Fig. 5, A to C). Whereas the S1292A mutation substantially reduced filamentous distribution of the R1441G/G2019S LRRK2 double mutant (Fig. 5, A and B), it had little effect on puncta localization (Fig. 5, A and C).

Fig. 5

Ser1292 is required for association of LRRK2 PD mutant proteins with microtubules. (A) Representative images of primary rat embryonic cortical neurons transfected with FLAG-tagged LRRK2 constructs as indicated. Staining is with antibodies against FLAG and the microtubule marker MAP2. Puncta are indicated by white arrows. Scale bar, 10 μm. (B and C) Quantification of percentage of neurons displaying filamentous (B) or punctate (C) distribution of LRRK2 in (A). Each bar is an average ± SEM of percentages from three independent experiments. P < 0.005; P < 0.001, one-way ANOVA test. (D) Representative images of FLAG-tagged R1441G/G2019S mutant LRRK2 transfected into cultured rat embryonic primary cortical neurons. Neurons were costained for FLAG, tubulin, or p62, or LC3, or K63 polyubiquitin. Scale bars, 5 μm.

LRRK2 kinase inhibitors block Ser1292 autophosphorylation in HEK293 cells

A high-throughput in vitro assay was performed to identify small-molecule inhibitors of LRRK2 kinase activity (30, 31). To evaluate the cellular activity of these agents, we engineered an inducible HEK293 cell line to express the R1441G/Y1699C/G2019S triple mutant LRRK2 tagged with FLAG (LRRK2 3M cells) (fig. S7A). For quantifying pSer1292 in cell lysates, we developed an assay using electrochemiluminescent detection technology [Meso Scale Discovery (MSD)]. Doxycycline induction of LRRK2 kinase activity in the LRRK2 3M cell line increased the pSer1292 signal in corresponding lysates by ~20-fold (fig. S7B). When doxycycline-induced LRRK2 3M cells were incubated with the broadly acting kinase inhibitor sunitinib (21), dose-dependent reduction of the pSer1292 signal was observed with maximal inhibition equivalent to the baseline signal from noninduced cell lysates (Fig. 6A). Using this assay, we tested more than 600 compounds from a variety of chemical classes for their ability to block phosphorylation of Ser1292 in LRRK2 3M cells. The median inhibitory concentration (IC50) for reduction of pSer1292 signal correlated with the LRRK2 biochemical IC50 (Fig. 6B).

Fig. 6

LRRK2 kinase inhibitors block pSer1292. (A) Dose-dependent reduction of pSer1292 signal in lysates from doxycycline-induced LRRK2 3M (R1441G/Y1699C/G2019S)–inducible HEK293 cells incubated with sunitinib and G1023 for 2 hours. Signals were normalized to doxycycline-induced cell lysates without inhibitor as 100% and noninduced cell lysates as 0%. (B) Six hundred compound data set showing correlation of cellular IC50 values for pSer1292 in the LRRK2 3M HEK293 cell line [as in (A)] and biochemical IC50 values for LRRK2. Slope = 0.95, R2 = 0.77. Open symbols are out-of-range IC50 values and were excluded from the regression fit. (C) Structure of the LRRK2 inhibitor G1023. (D to F) Western blot (D) and quantification of pSer1292 (E) and pSer935 (F) relative to total LRRK2 in tissue lysates from the spleens of BAC transgenic mice expressing FLAG-tagged G2019S mutant LRRK2. (G to I) Western blot (G) and quantification of pSer1292 (H) and pSer935 (I) relative to total LRRK2 in tissue lysates from hippocampus of BAC transgenic mice expressing FLAG-tagged G2019S mutant LRRK2. For quantification of pSer1292 and pSer935, ratios of pSer1292 or pSer935 to total LRRK2 were calculated for each animal and normalized to those for mice treated with vehicle. *P < 0.05; **P < 0.01; ***P < 0.001, one-way ANOVA test followed by the Tukey-Kramer test. (J) Mouse brain concentrations of pSer1292 versus brain concentrations of unbound G1023 drug. Closed circles represent observed data, and lines represent the prediction from a direct inhibition model. pSer1292 LRRK2 is normalized to pSer1292 concentrations observed in mice dosed with vehicle (100%). Mean is shown for mice treated with vehicle (n = 2) or with doses (10, 30, or 100 mg/kg) of G1023 (n = 3 per dose).

An LRRK2 kinase inhibitor blocks pSer1292 in mouse brain

In the course of screening, G1023 was identified as a potent inhibitor of LRRK2 kinase activity. G1023 is a diaminopyrimidine inhibitor (Fig. 6C) with a biochemical IC50 of 4 nM that displays dose-dependent LRRK2 inhibition in the LRRK2 3M cell-based assay with an IC50 of 9 nM (Fig. 6A). The cellular IC50 against wild-type LRRK2 has not been determined for this compound. When G1023 was tested against a panel of 190 kinases at 0.1 μM, this drug showed greater than 50% inhibition against only one other kinase.

After a single oral dose (30 mg/kg) in BAC transgenic mice, G1023 showed a brain half-life of 6.6 hours. To test the in vivo activity of this drug, we gave BAC transgenic mice expressing G2019S mutant LRRK2 a single oral dose of the G1023 kinase inhibitor at 10, 30, or 100 mg/kg. Peripheral (spleen) and brain tissues were harvested 6 hours after injection to measure concentrations of the drug as well as concentrations of pSer1292, pSer935, and total LRRK2. In the spleen, G1023 displayed dose-dependent inhibition of pSer1292 (Fig. 6, D and E). In the brain (hippocampus), G1023 also effectively reduced pSer1292, although a higher dose (100 mg/kg) was required to obtain the same level of inhibition as observed for spleen (Fig. 6, G and H). A relationship was observed between unbound drug concentrations in the brain and pSer1292 inhibition (Fig. 6J). Using a pharmacodynamic model, we determined the unbound brain IC50 for G1023 to be 12 nM, consistent with an IC50 of 9 nM for G1023 in cultured cells.

An interesting observation was that phosphorylation of Ser935 in LRRK2, a site known not to be an autophosphorylation site, showed LRRK2 kinase inhibition in response to G1023, although higher concentrations of the drug were needed than with inhibition of pSer1292 LRRK2 (Fig. 6, D, F, G, and I). To test whether phosphorylation of Ser910 and Ser935 is regulated by Ser1292 autophosphorylation, we examined whether the S1292A mutation affected pSer910 and pSer935. In transfected HEK293 cells, the S1292A mutation did not affect pSer910 and pSer935 concentrations (fig. S8A). Likewise, the S935A mutation did not affect the concentration of pSer1292 (fig. S8B).

The drug G1023 attenuates a neurite outgrowth defect induced by LRRK2 PD mutations

Given that LRRK2 kinase activity and Ser1292 autophosphorylation may be required for the toxicity of LRRK2 PD mutations in cell-based assays, we examined the effects of LRRK2 kinase inhibition on the neurite outgrowth defects elicited by LRRK2 carrying PD mutations. Hippocampal neurons were cultured from BAC transgenic mice expressing G2019S mutant LRRK2 and their wild-type nontransgenic littermates. Mouse hippocampal neurons from the BAC transgenic mice expressing the G2019S mutant displayed shorter total neurite length when compared to hippocampal neurons derived from nontransgenic littermates (Fig. 7, A and C). Hippocampal neurons from the G2019S mutant LRRK2 BAC transgenic mice when treated with the kinase inhibitor G1023 showed reduced Ser1292 phosphorylation (Fig. 7B) and reversal of the neurite outgrowth defects (Fig. 7, A and C).

Fig. 7

LRRK2 kinase inhibitor ameliorates neurite outgrowth defects. (A) Representative images of cultured primary embryonic hippocampal neurons from BAC transgenic mice expressing G2019S mutant LRRK2 and their nontransgenic littermates treated with dimethyl sulfoxide (DMSO) (vehicle) or 100 nM G1023. Scale bar, 20 μm. (B) Western blot of pSer1292 in lysates of cultured primary hippocampal neurons from G2019S mutant LRRK2 expressing BAC transgenic mice or nontransgenic control animals. Cultured neurons were treated with vehicle or 100 nM G1023. (C) Quantification of total neurite length of mouse primary hippocampal neurons from (A) (mean ± SEM). ***P < 0.0001, one-way ANOVA test followed by the Tukey-Kramer test.

Discussion

Here, we report the identification and functional characterization of an in vivo LRRK2 autophosphorylation site at Ser1292 using cultured HEK293 cells, embryonic rat neurons, and neurons from BAC transgenic mice. Using pSer1292 as a readout of LRRK2 kinase activity, we developed cell-based and pharmacodynamic assays of LRRK2 kinase activity for drug discovery. The LRRK2 kinase inhibitor G1023 effectively reduced pSer1292 in cultured HEK293 cells and the brains of BAC transgenic mice overexpressing G2019S mutant LRRK2.

Previously, Ser910 and Ser935 on LRRK2 have been reported to be constitutive phosphorylation sites indirectly regulated by LRRK2 kinase activity. Several structurally diverse LRRK2 kinase inhibitors have been shown to inhibit Ser910 and Ser935 phosphorylation. Consistent with these findings, the kinase inhibitor G1023 also effectively reduced phosphorylation of these sites in the brains of drug-treated BAC transgenic mice. However, because Ser910 and Ser935 are not phosphorylated by LRRK2 kinase (16), the concentration of the G1023 compound required for reduction of phosphorylation of these sites was much higher than that for reduction of pSer1292.

We observed that several confirmed familial PD mutations of LRRK2, both mutations in the ROC GTPase domain (N1437H and R1441G/C) and mutations in the kinase domain (G2019S and I2020T), induced elevated autophosphorylation at Ser1292. Of the known PD mutations, only G2019S had been consistently shown in previous work to increase LRRK2 kinase activity in vitro (13). Because the G2019 residue of LRRK2 is located in the kinase activation loop, it was proposed that the G2019S mutation changes the conformation of that loop to activate the kinase (6). However, it remained perplexing that an adjacent mutation at I2020T would not increase LRRK2 kinase activity in vitro (13). Here, we report that both G2019S and I2020T mutations can increase Ser1292 phosphorylation in HEK293 cells, suggesting that both PD mutations in the LRRK2 kinase domain increase kinase activity. A possible explanation for the past findings may be that substrates used in the biochemical assays were not representative of endogenous substrates and may not accurately reflect LRRK2 kinase activity.

One outstanding question is how LRRK2 PD mutations in the ROC domain increase Ser1292 autophosphorylation. Mutation of the GTP binding site in the ROC domain has been shown to abolish LRRK2 kinase activity in vitro (28). Consistent with this finding, we demonstrate that an intact GTP binding site is required for Ser1292 autophosphorylation in HEK293 cells. Moreover, PD mutations in the LRRK2 ROC domain (R1441G/C and N1437H) have previously been shown to reduce GTPase activity and increase GTP binding (14, 32, 33). We show here that the same mutations (R1441G/C and N1437H) increase Ser1292 autophosphorylation. On the basis of these two pieces of complementary data, we postulate that GTP binding may be important for Ser1292 autophosphorylation in cells. Intriguingly, PD mutations in the ROC domain have not consistently shown elevated in vitro kinase activity (13). Two recent papers suggest that addition of GTPγS or GDP (guanosine diphosphate) does not alter in vitro kinase activity (23, 24). Indeed, our data from in vitro Ser1292 autophosphorylation assays suggest that GTP binding is not essential for in vitro kinase activity (fig. S3). It is possible that auxiliary proteins may modulate the observed differences between the in vitro biochemical activity of LRRK2 and the activity of LRRK2 in cultured cells. Alternatively, ROC domain mutations could modulate Ser1292 autophosphorylation by altering the conformation of LRRK2 in cells. Identification of auxiliary factors and interrogation of the underlying mechanisms regulating LRRK2 kinase activity will be an important focus of future research.

Our finding that several PD mutations spanning different domains of LRRK2 increase Ser1292 autophosphorylation suggests that phosphorylation of Ser1292 may be implicated in PD pathogenesis driven by mutant LRRK2. Notably, one confirmed familial PD mutation (Y1699C) and a PD risk factor mutation (G2385R) did not increase Ser1292 autophosphorylation in HEK293 cells, suggesting that these mutations may affect PD pathogenesis through mechanisms independent of Ser1292 autophosphorylation. In future experiments, it will be important to test how different LRRK2 PD mutations affect autophosphorylation of Ser1292 in postmortem brain tissue from PD patients.

Given that several LRRK2 PD mutations increased autophosphorylation of Ser1292 and that the Ser1292 site is conserved across species, we postulated that Ser1292 autophosphorylation may be important for LRRK2 function. PD mutations in LRRK2 have been reported to cause cellular abnormalities in both cultured cell lines and primary neurons, with toxicity being dependent on kinase activity (79). We observed that expression of the R1441G/G2019S double mutant LRRK2 in rat embryonic hippocampal neurons reduced neurite length and branching and that the kinase-dead LRRK2 mutation ameliorated the neurite outgrowth defect. This is consistent with the previous finding that expressing LRRK2 single PD mutations affects neurite outgrowth (9). The fact that we did not observe a statistically significant effect when expressing single PD mutations may be due to different experimental conditions. The S1292A LRRK2 mutation, which abolishes Ser1292 autophosphorylation, also abrogated the neurite outgrowth defect. These data suggest that Ser1292 autophosphorylation may be implicated in disease-related mechanisms downstream of LRRK2 kinase activity. A presumably phosphomimetic mutation S1292D also reversed the neurite outgrowth defect. The simplest explanation is that S1292D prevents phosphorylation but does not mimic pSer1292. Future proteomics studies will need to elucidate changes in the protein-protein interactions mediated by Ser1292 autophosphorylation.

The cellular mechanisms through which LRRK2 PD mutations induce neuronal abnormalities remain elusive. Recently, it was suggested that LRRK2 PD mutations cause defects in nuclear envelope organization (34). It was previously observed that LRRK2 PD mutant proteins accumulate on microtubules in primary neurons (35). In addition, several studies have implicated LRRK2 in the autophagy pathway (3639). Our findings that LRRK2 PD mutant proteins become localized to microtubules and colocalize with the autophagy adaptor protein p62 in primary neurons are consistent with the published data. Protein cargo destined for degradation by autophagy in neurons is thought to move along the microtubule network en route to lysosomes (40). It is possible that there is aberrant trafficking of protein cargo destined for autophagy in neurons expressing LRRK2 PD mutations, but this possibility requires further exploration.

There has been great interest in developing LRRK2 kinase inhibitors for treatment of PD, and several LRRK2 kinase inhibitors have been previously described (10, 4143). Although these inhibitors were shown to be efficacious in ameliorating the toxic effects of LRRK2 PD mutations in vitro, the lack of selectivity and brain penetrance of these molecules may limit their use in vivo. Furthermore, without an in vivo measure of LRRK2 activity, it is difficult to make accurate dose predictions.

Here, we report a brain-penetrating LRRK2 kinase inhibitor, G1023. In cultured mouse BAC transgenic primary hippocampal neurons, this compound effectively ameliorated neurite outgrowth defects caused by the G2019S mutation in LRRK2. This compound has useful pharmacokinetic properties in our preclinical mouse model. It can cross the blood-brain barrier with unbound brain concentrations above the cellular IC50 after a single dose. G1023 blocked LRRK2 kinase activity in the brains of BAC transgenic mice. The IC50 of unbound G1023 in mouse brain was similar to the IC50 of G1023 obtained from an LRRK2 cellular assay. At the in vivo concentration required for LRRK2 inhibition, this compound was relatively selective.

Although our study seeks to address several important questions about LRRK2 biology, many challenges remain in applying our findings translationally. First, we still need a sensitive assay for detection of the endogenous pSer1292 signal in animal and human brain tissues. Also, the efficacy of G1023 remains to be tested in mutant LRRK2 and other PD animal models to establish preclinical proof of concept that LRRK2 inhibition may be useful for treating PD. Key to these efforts will be establishing robust behavioral and histopathological endpoints in these models (44). Nevertheless, the tools described in our study should serve as a useful foundation for further development of LRRK2 inhibitors.

Materials and Methods

LRRK2 BAC transgenic mice

Mice expressing the G2019S and wild-type LRRK2 BAC transgene under the control of the endogenous LRRK2 promoter were described previously (45). Mice were maintained on a C57BL/6N background and housed in a specific pathogen–free facility under conditions of controlled temperature and humidity. Mice had ad libitum food and water, and lived on a 14:10 light-dark cycle.

MS analysis

Excised gel pieces were reduced, alkylated, and subjected to overnight trypsin digestion. For phosphomapping analyses, digested peptides were extracted, dried, and resuspended in 10% acetonitrile/5% formic acid. For quantitative analysis, a mixture of isotopically labeled internal standard peptides (Cell Signaling Technology) was introduced to the digested gel pieces before extracting the peptides. After extraction, dried peptides were resuspended in 3% acetonitrile/5% formic acid/10% H2O2 at least 30 min before analysis to allow for complete oxidation of methionine-containing peptides. Peptides were separated with a standard gradient composed of H2O/acetonitrile/formic acid delivered by a nanoACQUITY UPLC (Waters) and analyzed on an LTQ Orbitrap XL (Thermo) mass spectrometer as previously described (46). The mass spectrometer was operated in data-dependent mode with a 60,000-resolution full-MS scan collected in the Orbitrap and MS/MS for the eight most intense ions collected in the ion trap. Spectral data were searched with Mascot using the target-decoy strategy against the human UniProt database and rough-filtered to 5% false discovery rate using a linear discriminant analysis (47). Putative phosphorylation sites were localized and confidence scores were assigned with the Ascore algorithm (48), with scores of 13 and 19 representing the 95% and 99% confidence levels, respectively. Confidently matching phosphopeptides with uncertain localizations (Ascore <13) have not been reported. For the reported peptide-spectral matches, manual validation of at least one representative spectrum for each unique phosphopeptide sequence was required. For quantitative analysis, full-MS scans were integrated with Qual Browser software (Thermo) to determine the peak areas for heavy and light mass/charge ratio (m/z) ions of the most abundant charge states of each LRRK2 phosphopeptide.

Western blotting

Cultured cells and mouse brains were homogenized in radioimmunoprecipitation assay buffer [25 mM tris-HCl (pH 7.6), 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS] containing Complete protease and phosphatase inhibitor cocktail (Roche). Cell/tissue lysate was cleared via centrifugation at 20,000g for 30 min at 4°C. Protein concentration of the supernatant was measured with BCA assay (Pierce). Lysate (20 μg) was loaded onto 3 to 8% tris-acetate gels or 4 to 12% bis-tris gels (Invitrogen) and immunoblotted with the following antibodies: pSer1292 (custom-generated at Yenzyme), pSer935, pSer910, LRRK2 (Epitomics), FLAG, and tubulin (Sigma). The LI-COR Odyssey system was used for Western blot detection.

Epstein-Barr virus (EBV) cells were lysed directly in 1% SDS, and 75 μg of total protein was loaded per lane on 3 to 8% tris-acetate gels.

Generation of an inducible LRRK2 stable cell line

FLAG-tagged full-length human LRRK2 was inserted into pTre2Hygro vector (Clontech). QuikChange Site-Directed Mutagenesis kit was used to introduce three Parkinson’s mutations: R1441G, Y1699C, and G2019S (Stratagene). The tetracycline-regulated Tet-On system was used to generate an LRRK2-expressing stable cell line (49). HEK293 Tet-On cells (Clontech) were transfected with the LRRK2 plasmid mentioned above together with GFP. Hygromycin was used for stable cell selection, and 48 single clones were selected with clone disk. LRRK2 expression levels were compared with Western blots to identify the highest expresser line. The concentration of hygromycin used for selection was determined by a killing curve assay.

Inhibitor treatment of LRRK2 cells

HEK293 Tet-On LRRK2 G2019S/R1441G/Y1699C (LRRK2 3M)–inducible stable cell line was cultured in complete media [high-glucose Dulbecco’s modified Eagle’s medium, 10% Tet System fetal bovine serum, 1% Glutamax, 1% nonessential amino acids, G418 (100 μg/ml), and hygromycin B (30 μg/ml)]. Cells were induced with doxycycline (1 μg/ml) for 16 hours at 37°C. Cells were treated with inhibitors for 2 hours and then lysed with cold MSD tris lysis buffer.

Primary embryonic neuron culture, transfection, immunofluorescence, image acquisition, and analysis

Dissociated primary embryonic neuron cultures were prepared on the basis of the methods previously described (50). Detailed primary neuron culture, transfection, immunofluorescence, image acquisition, and analysis methods are included in the Supplementary Materials.

Animal studies

G1023 was dissolved in 1% RC591 in water. Nine- to 12-week-old G2019S BAC transgenic mice were assigned to treatment groups with balanced age and sex. Animals received oral dosing of either compound (10, 30, or 100 mg/kg, n = 3 per group) or 1% RC591 vehicle (n = 2). The dosing volume was adjusted with 1% RC591 to ensure constant volume of 10 ml/kg. At 6 hours after dose, animals were euthanized and tissues were rapidly harvested. All animal experiments were reviewed by Genentech Institutional Animal Care and Use Committee and in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Supplementary Materials

www.sciencetranslationalmedicine.org/cgi/content/full/4/164/164ra161/DC1

Materials and Methods

Fig. S1. Quantitative analysis of LRRK2 autophosphorylation in vitro.

Fig. S2. Preferential LRRK2 autophosphorylation on threonine.

Fig. S3. GDP or GTPγS preloading does not affect pSer1292 in an in vitro autophosphorylation assay.

Fig. S4. Ser1292 autophosphorylation site of LRRK2 is conserved from worm to human.

Fig. S5. S1292A mutation does not affect LRRK2 kinase activity in vitro.

Fig. S6. Ser1292 is highly phosphorylated in the LRRK2 triple mutant cells.

Fig. S7. Mutation of Ser1292 does not affect phosphorylation of Ser935 and vice versa.

Table S1. LC-MS/MS identification of in vitro LRRK2 autophosphorylation sites.

Table S2. Internal standard peptides used in characterization of LRRK2 autophosphorylation.

References and Notes

  1. Acknowledgments: We thank M. Sheng and R. Watts for critical reading of the manuscript. Funding: This work is supported by Genentech and funding from the following sources to Z.Y.: NIH/National Institute of Neurological Disorders and Stroke grants NS060809-01 and NS072359-01 and Michael J. Fox Foundation for Parkinson’s Research. Author contributions: Z.S. and S.Z. carried out the experiments using HEK293 cell transient transfection. Z.S. carried out the experiments using transfected embryonic rat primary neurons. S.Z. carried out the experiments using cultured embryonic primary neurons from BAC transgenic mice. Z.S. and S.Z. participated in interpreting the data and preparing the figures. D.B. and D.S.K. designed and carried out MS experiments. Cell-based assay and compound screening were established and carried out by T.K., S.Z., J.D., and J.G.M. Chemistry efforts for G1023 were carried out by D.J.B., J.G.-T., H.C., A.A.E., and Z.K.S. In vivo studies and pharmacokinetic/pharmacodynamic studies were designed and carried out by C.E.L.P., S.L.D., H.O.S., S.Z., X.Z., X.D., X.L., A.A.E., Z.K.S., and K.S.-L. F.C. and M.P.v.d.B. contributed to characterization of EBV-immortalized B cell lines. X.L. and Z.Y. contributed transgenic mouse models. Q.S. performed statistical analysis of the data. H.Z. oversaw all studies, designed the experiments, and interpreted the data. D.S.K. and H.Z. wrote the paper in collaboration with all co-authors. Competing interests: All authors, except Z.Y. and X.L., are current or former employees of Genentech, now owned by Roche. Patents have been filed on an LRRK2 cell–based assay using pSer1292 as the readout (patent WO2012 075046) and on compound G1023 (patent WO2011 151360).
View Abstract

Navigate This Article