Relative sensitivity of LRRK2 autophosphorylation and Rab GTPase phosphorylation to MLi-2 inhibition in vivo
Our aim in this series of studies was to mimic a likely clinical scenario where mutant LRRK2 is inhibited to levels similar to that seen with wild type protein, thus nullifying the potential toxic effects of enhanced pathogenic kinase activity. In order to be able to model this situation, we first needed to identify reliable markers of LRRK2 activity in vivo and understand the relationship between peripheral organ and brain target engagement. To this end, we compared the response of phosphorylation of LRRK2 itself and downstream Rab substrates to MLi-2 in vivo using two acute treatment paradigms.
First, we investigated the dose responsiveness of LRRK2 autophosphorylation and phosphorylation of Rab GTPases to acute kinase inhibition in vivo. G2019S KI mice were given an acute dose of MLi-2 at 1, 3, 10, 30, 60, or 90mg/kg or vehicle via oral gavage and sacrificed 1hr post dose. We observed significant pS1292 dephosphorylation starting at the lowest dose of 1mg/kg in kidney and lung (Fig. 1A, B). Maximal dephosphorylation was achieved in the brain, kidney and lung at 10mg/kg for both pS1292 and pS935 (Fig. 1A - C). Maximal S1292 dephosphorylation in the brain did not exceed 60% signal decrease with increasing MLi-2 concentrations, suggesting that maximal inhibition was achieved. The S935 phospho-site had lower IC50 compared to pS1292 in the brain, suggesting that this marker is more sensitive to LRRK2 inhibition in vivo
Phosphorylated Rab GTPases showed more variable responses to MLi-2 across tissues. The T73 Rab10 phospho-site responded significantly to MLi-2 treatment in peripheral tissues but not in the brain (Fig. 1A, D). Phospho- S106 Rab12 showed a robust response to MLi-2 in both brain and peripheral tissues with maximal dephosphorylation at 30-60mg/kg (Fig. 1A, E; antibody validation experiments are shown in Supplementary figs. 1 and 2). The T71 Rab29 phospho-site responded with treatment in all tissues but with higher variability compared to Rab12, especially at lower doses (Fig. 1A, F). At 10mg/kg, we saw ~50% decrease in Rab10, Rab12 and Rab29 phosphorylation in the periphery while higher doses retained ~30% residual phosphorylation signal. These results demonstrate that 10mg/kg of MLi-2 is adequate to acutely inhibit LRRK2 autophosphorylation in brain and peripheral tissues. Based on these results, we selected a dose of 10mg/kg MLi-2 for subsequent time course analyses and washout experiments.
Time course of acute MLi-2 administration reveals that phosphorylation of Rab GTPases recovers faster than LRRK2 phosphorylation
To further discriminate how LRRK2 and Rab phosphorylation events vary between brain and peripheral tissues, we compared time courses and recovery after washout of inhibitor application. An acute dose of MLi-2 at 10mg/kg was administered via oral gavage to G2019S KI mice and sacrificed at 0.5, 1, 3, 12, 24, or 72 hours post dose. A time point of 0 hours was included where animals were given an equivalent dose of vehicle and euthanized immediately thereafter to measure baseline phosphorylation.
Phospho- S1292 and S935 LRRK2 showed rapid dephosphorylation at 0.5hrs, with maximal dephosphorylation achieved at 1hr post dose across all tissues (Fig. 2A - C). The dephosphorylation patterns of both phospho-sites tightly correlated between brain, kidney, and lung tissues. Interestingly, maximal dephosphorylation for all pRabs was seen at 1hr in all tissues. However, these sites recovered more rapidly than pLRRK2 (~12hrs versus 24hrs; Fig. 2A, D - F). Measures of T73 Rab10 signal in brain tissue did not show MLi-2-dependent dephosphorylation compared to peripheral tissues (Fig. 2A, D). Dephosphorylation of T71 Rab29 was achieved in all three tissues to varying degrees, with kidneys showing the strongest response to MLi-2, reaching ~75% dephosphorylation compared to 20-30% seen in brain and lung tissue (Fig. 2A, F). pS106 Rab12 mimicked pLRRK2 most closely in that all tissues showed similar dephosphorylation patterns over time and reached a maximum of 60% dephosphorylation (Fig. 2A, E). It was also noted that Rab12 phosphorylation showed the least variability across mice compared to Rab10 and Rab29. These data suggest that Rab GTPases are dephosphorylated within the first hour after LRRK2 inhibition in a similar fashion to LRRK2 dephosphorylation, with Rab12 performing similarly to pLRRK2 in the brain and periphery. In contrast, the kinetics of Rab GTPase re-phosphorylation show a quicker recovery (~12hrs) compared to LRRK2 (~24hrs).
In-diet MLi-2 administration can diminish G2019S-dependent hyperphosphorylation to wild type levels
To evaluate the molecular effects of chronic LRRK2 inhibition, we first conducted a dose response experiment of MLi-2 in diet. With clinical relevance in mind, our aim was not to fully inactivate LRRK2, as inferred from a complete dephosphorylation of S1292, but to ameliorate the hyperphosphorylation to a range observed in wildtype animals.
G2019S KI mice were fed a customized rodent chow supplemented with MLi-2 to achieve 10, 30, or 60 mg/kg/day dosing. For reference, we included wildtype and G2019S KI mice that were fed control chow for 10 days. In the treated animals, we observed that 60mg/kg/day diminished S1292 phosphorylation to wildtype levels in brain and kidney, whereas a 10mg/kg/day dose was sufficient to decrease phosphorylation to wildtype levels in lung tissue (Fig. 3A, B). This suggests some peripheral tissues with enrichment of LRRK2 may be more sensitive to drug-induced inhibition. Increasing doses of MLi-2 show a dose response in pS935 LRRK2 (Fig. 3A, C). These results confirm a dose of 60mg/kg/day is sufficient to inhibit G2019S LRRK2 to wild type levels in vivo across tissues.
Chronic MLi-2 treatment in G2019S KI mice results in sustained LRRK2 and Rab12 dephosphorylation
To extend these results into a chronic timescale, G2019S KI mice were given customized chow supplemented with MLi-2 to reach a 60 mg/kg/day dose for 10 days or 10 weeks. Control groups of wildtype, G2019S KI, and LRRK2 KO mice receiving untreated chow were included for reference of baseline phosphorylation patterns. For the purpose of this experiment, we refer to the 10-day cohort as ‘short-term’ and the 10-week cohort as ‘long-term’ treatment groups. The schematic in Fig. 4A depicts the design of this experiment, in which brain, kidney, and lung tissues were collected and processed for Western blot analyses, and the 10-week cohort tissues were additionally prepped for total and phospho-proteomics. Body weight and estimated food intake for each mouse was recorded daily to determine the daily dose of MLi-2 each mouse received (Fig. 4B - G). Both short- and long-term cohorts received the appropriate dose of MLi-2 and had comparable chow intake of ~4g, while weight increase was observed only in the 10-week cohort, particularly in the LRRK2 KO animals.
In both short-term and long-term cohorts, pS1292 levels of G2019S KI mice treated with MLi-2 were significantly decreased to levels comparable to wildtype mice (Fig. 5A - C). In the long-term cohort, mice treated with MLi-2 showed further dephosphorylation compared to wildtype levels in lung (Fig. 5C). S106 Rab12 showed significant dephosphorylation consistent throughout tissues, similar to the response of LRRK2 dephosphorylation (Fig. 5A, H - I), proving pRab12 is a reliable readout of LRRK2 activity and inhibition in this model. Additionally, phosphorylation of S935 LRRK2 was significantly reduced in all tissues compared to both G2019S and wildtype untreated animals in both cohorts as expected (Fig. 5A, D - E). In contrast, and similar to the tissue and dose-specific responses to acute inhibition, Rab10 and Rab29 did not respond consistently to LRRK2 inhibition in this chronic paradigm (Supplementary fig. 3). Furthermore, total LRRK2 levels in the short-term cohort were significantly decreased in MLi-2-treated mice compared to their untreated counterparts in kidney and lung tissues, which was exacerbated in the long-term treatment across all tissues (Fig. 5A, F - G). This suggests chronic inactivation of LRRK2 leads to protein degradation, consistent with previous in vivo and in vitro studies using MLi-2 and other LRRK2 inhibitors (13,24).
Unbiased proteomics reveal both therapeutic and dysregulatory effects in endolysosomal, trafficking, and mitochondrial pathways with chronic LRRK2 inhibition in mice
The above studies identified a chronic dosing regimen in which amelioration of the hyperphosphorylation of the S1292 autophosphorylation LRRK2 site and pS106 Rab12 in G2019S KI mice can be achieved to levels seen with wildtype LRRK2 at the endogenous level in vivo. We next used a series of proteomics approaches to determine, in an unbiased manner, what the consequences of this treatment might be to tissue proteome.
Proteomic analysis revealed 115 total proteins and 34 phospho-proteins that were differentially expressed in the kidney between chronic MLi-2-treated and untreated G2019S LRRK2 mice (false-discovery rate (FDR) adjusted p < 0.05; fold change (FC) > 1.4); (Fig. 6A, B). Among the top differentially abundant proteins, there was a strong enrichment for endolysosomal, trafficking and mitochondrial proteins (Fig. 6A, C). Multiple lysosomal proteins showed differential abundance, including cathepsin B (Ctsb), legumain (Lgmn), galactosidase beta 1 (Glb1), Lysosomal-associated membrane protein 1 (Lamp1), and N-acetylglucosamine-6-sulfatase (Gns). In addition, a number of proteins involved in vesicular trafficking, lipid metabolism, iron uptake and mitochondrial function were also significantly altered in kidneys of chronically treated animals. Hierarchical clustering of differential proteins in the G2019S MLi-2 treated group showed most similarity to the LRRK2 KO animals (Fig. 6C), suggesting that chronic inhibition of LRRK2 may mimic features of an absence of LRRK2 in the periphery. Among the significant phosphoprotein hits, additional trafficking and mitochondrial proteins were identified, including sorting nexin 1 (pS188 Snx1) and vacuolar sorting protein 4b (pS102 Vps4b) (Fig. 6D). Analysis using Gene Ontology databases of significant hits from total and phospho-proteins showed enrichment of the endolysosomal system as well as mitochondrial membrane (Fig. 6E).
Additionally, we analyzed the human homologs of total and phospho-protein hits in silico to identify a LRRK2 protein-protein interactome using PINOT (19). We converted the 115 total proteins and 34 phospho-proteins that were differentially expressed in the kidney between chronic MLi-2-treated and untreated G2019S LRRK2 mice to their human orthologues and identified 76 matches within the LRRK2 interactome. Five of the matching proteins (AHCYL1, EEF2, HSP90AA1, HSP90AB1 and RANBP2) were present in the first layer while 71 in the second layer of LRRK2 interactions. The difference between the average random result from 100,000 simulated experiments (40 matches) and the real result (76 matches) was highly significant (p=2.78*10-11). The matching proteins were extracted from the LRRK2 interactome and their connectivity with LRRK2 visualized (Supplementary fig. 4). This high degree of connectivity suggests that these proteins are in fact related to LRRK2 biology rather than an effect of MLi-2 treatment itself.
Proteomic analysis of brain tissue of 10-week MLi-2-treated mice revealed a number of mitochondrial proteins that showed statistically significant p-value (adjusted p<0.05) compared to untreated G2019S KI controls, albeit with modest fold differences between treatment groups (Fig. 7A, B). These include Cytochrome C, NADH:Ubiquinone Oxidoreductase Subunit V3 (Nduvf3), the mitochondrial ATP synthase Atp5g1 and Voltage Dependent Anion Channel 2 (Vdac2) (Fig. 7A, B). Phospho-proteomics analysis revealed a decrease in S58 phosphorylation of the ion transporter Fxyd7 and increase in S109 phosphorylation of the PP2A inhibitor Ensa (Fig. 7B). Cumulative Gene Ontology analysis revealed enrichment for proteins residing in different mitochondrial compartments as well synaptic proteins (Figure 7C). These data suggest that even though the brain is more resilient in terms of potential endolysosomal defects that may be associated with LRRK2 inhibition, there are small changes in mitochondrial function resulting from chronic LRRK2 inhibition. Volcano plots comparing untreated groups of G2019S KI and wildtype mice showed that a number of proteins had the opposite trend to that seen in the chronic MLi-2 cohort (Fig. S5). For example, the transmembrane ion transporters Sfxn2 and Sfxn3, showed a modest upregulation in kidneys of G2019S LRRK2 compared to wildtype mice and this was rescued in the chronic MLi-2 G2019S LRRK2 cohort (Figs. 6A and Supplementary fig. 5A). For some mitochondrial proteins identified in our brain proteomics screen, Cytochrome C, Vdac1, Ndufs3, and Ndufv2 were downregulated in G2019S LRRK2 brain compared to wildtype and this was reversed in the chronic MLi-2 G2019S LRRK2 cohort (Figs. 7A and Supplementary fig. 5B). This suggests that treatment can significantly alter the expression of mitochondrial proteins in G2019S KI mice in a direction consistent with therapeutic potential.
Validation of endolysosomal, trafficking, and mitochondrial proteins in chronically LRRK2 inhibited mice reveal both a rescue of mutant-driven effects and dysregulatory patterns in vivo
We next wanted to validate protein hits observed from our proteomics screens in kidney tissue. Through Western blot analyses, we found two distinct patterns of effects, one we characterized as beneficial, based on a reverse in protein expression from G2019S KI animal levels back to wildtype levels and the other dysregulatory, based on a mimicking effect comparable to LRRK2 KO animals (Fig. 8A). In the former category, the non-glycosylated form of Lamp1, the ESCRT-0 protein Hgs, and the iron and serine mitochondrial transporter Sfxn3 were all shown to be significantly increased in protein levels in untreated G2019S KI mice compared to wildtype animals. These levels were ameliorated back to wildtype levels after MLi-2 treatment for 10 weeks (Fig. 8B). Conversely, the glycosylated form of Lamp1 and the lysosomal hydrolase Legumain were both significantly increased in the same animals, patterns of which have been previously characterized in LRRK2 KO animals, and is recapitulated here in our LRRK2 KO mice in this cohort (25). In addition, we also discovered that the motor adaptor protein Jip4 was significantly reduced in LRRK2 KO mice, and this was also seen in G2019S KI mice treated with MLi-2 (Fig. 8C). Interestingly, this phenotype is kidney-specific, as no significant difference of these proteins was observed in brain or lung tissues (not shown). Additionally, these differences were not seen in mice that were treated with MLi-2 for only 10 days (Supplementary fig. 6), suggesting that chronic inhibition of LRRK2 is necessary to see these effects.