LRRK2 phosphorylates membrane-bound Rabs and is activated by GTP-bound Rab7L1 to promote recruitment to the trans-Golgi network

LRRK2 phosphorylates Rab7L1

Recent phosphoproteomic analyses have revealed that LRRK2 phosphorylates a subset of Rab proteins. Phosphorylation occurs at threonine residues in the switch II effector binding domain of the closely related Group I Rab8 and Rab10 proteins (16). The LRRK2-binding partner Rab7L1 [also known as Rab29 (9,13)] is a Group III Rab, phylogenetically distant from Rab8 and Rab10 (see Supplementary Material, Fig. S1A). To determine whether LRRK2 phosphorylates Rab7L1, we co-transfected LRRK2-encoding plasmids with eGFP-Rab plasmids into HEK293 cells (Fig. 1). Phos-tag analysis of the effects of LRRK2-mutations R1441C, Y1699C, and G2019S, together with LRRK2-kinase-dead (D1994A), demonstrated robust LRRK2 phosphorylation of Rab10 (Fig. 1A and B) and Rab8 (Supplementary Material, Fig. S2). The R1441C and Y1699C mutations enhanced phosphorylation over WT-LRRK2, but the G2019S mutation was similar to WT-LRRK2 for these substrates. Co-transfection of Rab10 or Rab8 with Y1699C-LRRK2 led to the phosphorylation of ∼30% and ∼40%, respectively, of the Rab proteins, whereas no phospho-Rab protein could be detected with kinase-dead LRRK2 co-transfections. These results indicate there are no other kinases resident in HEK293 cells that might significantly phosphorylate these Rab proteins. Co-expression of LRRK2 with Rab7L1 led to a similar level of Rab phosphorylation (Fig. 1C). In contrast to Rab8 and Rab10 substrates, G2019S-LRRK2 protein phosphorylation of Rab7L1 could not be distinguished from the effects of the Y1699C and R1441C mutations and phosphorylated ∼30% of Rab7L1 protein in cells.

LRRK2 phosphorylation of Rab switch II domains requires Rab nucleotide binding

LRRK2 autophosphorylates its own switch-effector domains in the Rab-like ROC domain as well as in Rab8 and Rab10 (16,20–22). The switch II region is known to be relatively disordered in the inactive GDP-bound state compared with the active GTP-bound state (23), and could alter LRRK2-mediated phosphorylation. We created a panel of constructs to manipulate Rab nucleotide binding through altering intrinsic hydrolysis activity (QL mutations, see Supplementary Material, Fig. S1B) and nucleotide binding properties (TN mutations). Substitution of the P-loop threonine to asparagine (TN mutant) that shifts equilibrium towards nucleotide-free and GDP-bound states ablated LRRK2 phosphorylation (Fig. 1 and Supplementary Material, Figs S2 and S3). The TN mutations did not affect protein stability or expression compared with WT-Rab8, 10, and Rab7L1 (Fig. 1D and G and Supplementary Material, Fig. S2). The QL mutation in Rab10, designed to disrupt Rab10 GTPase activity and promote GTP-bound Rab10, resulted in significantly more LRRK2 phosphorylation activity compared with WT-Rab10 (Fig. 1E). In contrast, QL-Rab7L1 protein, which was shown to decrease GTP-binding affinity (13), cannot be phosphorylated by LRRK2 (Fig. 1F). On a subcellular level, the QL-Rab7L1 protein had a diffuse distribution different than WT-Rab7L1 and more similar to TN-Rab7L1 (Fig. 1G). These results are consistent with our past observations that the QL mutation in Rab7L1 results in a dominant-negative protein, in contrast to the equivalent mutation in Rab8/10 (13). In summary, LRRK2 phosphorylation of Rab protein substrates requires a nucleotide-bound, possibly GTP-bound, state.

LRRK2 phosphorylates Rab proteins in switch II domains. However, since Rab7L1 sequence diverges from Rab8 and Rab10, a serine residue (Ser72) in Rab7L1 is substituted for the threonine residue LRRK2 phosphorylates in the Rab8 and Rab10 switch II domains [Supplementary Material, Fig. S1A, (16)]. In Rab7L1, a threonine residue lies adjacent to the conserved phenylalanine. To clarify the site of LRRK2-mediated Rab7L1 phosphorylation, we generated an S72D and T71E mutation in Rab7L1. In contrast to S72D-Rab7L1 which showed similar LRRK2-phosphorylation as WT-Rab7L1, the T71E substitution nearly ablated LRRK2-mediated Rab7L1 phosphorylation (Fig. 1H and I). WT-LRRK2, R1441C, Y1699C, and G2019S-LRRK2 were all unable to phosphorylate T71E-Rab7L1 (Fig. 1J). These results suggest that the major LRRK2 phosphorylation site on Rab7L1 is the Thr71 residue.

LRRK2 phosphorylation of Rab substrates requires Rab prenylation

Subcellular distribution profiles of the Rab proteins suggested both cytosolic as well as membrane-associated pools of proteins (Fig. 1G). In splitting cell lysates into soluble cytosolic fractions and membrane-enriched triton-x100 extracted fractions, we could see near equal Rab protein expression between the two fractions, in contrast to the canonical proteins GAPDH that lacks membrane associations and aquaporin-1 (VDAC) that lacks a cytoplasmic component (Fig. 2A and B). Both QL and TN-Rab7L1 showed low levels of protein in the membrane-enriched fractions compared with WT-Rab7L1. Co-transfections of the QL and TN-Rab variants together with R1441C-LRRK2 demonstrated that most of the LRRK2-phosphorylated Rab proteins localized to membrane-associated fractions {[∼6: 1 Rab10, ∼10: 1 Rab8, and ∼3: 1 Rab7L1, (w/v)], Fig. 2A–D and Supplementary Material, Fig. S4}.

Pathogenic LRRK2 mutations may enhance the proportion of membrane-bound and kinase activity LRRK2 protein (13,24). We split cell lysates from cells co-expressing WT-Rab10 or Rab7L1 together with WT, R1441C, Y1699C and G2019S-LRRK2 into soluble cytosolic fractions and membrane-enriched fractions. All the LRRK2 mutations preferentially phosphorylated membrane-localized WT-Rab10 and Rab7L1 (Fig. 2E–H). Prenylation of C-terminal cysteine residues in Rabs is known to increase the affinity of the Rabs to membranes and is required for Rab function (25). We mutated the residues required for prenylation (Supplementary Material, Fig. S1B) and verified in vivo that GTP and GDP-binding of the resultant CC-Rab proteins was undisturbed (Supplementary Material, Fig. S5). Subcellular localization of the CC-Rab proteins revealed a diffuse distribution compared with the WT-Rab proteins, consistent with the intended effects of the CC mutations in blocking prenylation (Fig. 2I–L). In co-transfection experiments, LRRK2 was largely unable to phosphorylate CC-Rab mutants (Fig. 2M–P), as only slight LRRK2-phosphorylation in CC-Rab7L1 or CC-Rab10 could be detected above background in the cytoplasmic fraction and none in the membrane-enriched fraction. As expected, the overall CC-Rab concentration in the membrane-enriched fraction was also diminished.

Rab7L1 activates LRRK2 autophosphorylation

From our results, LRRK2 phosphorylation of other Rab proteins may require the Rab substrate to be membrane and GTP-bound. As LRRK2 autophosphorylates its own Rab-like ROC GTPase domain, and membrane-associated LRRK2 shows higher kinase activity than cytosolic LRRK2 (24,26), we next investigated whether Rab proteins might affect LRRK2 autophosphorylation. In comparing the intensity of FLAG-tagged WT-LRRK2 and FLAG-tagged WT-Rab7L1, we could determine that increasing levels of Rab7L1 led to increasing the proportion of LRRK2 autophosphorylated at Ser1292. At an equimolar stoichiometry, Rab7L1 enhanced pSer1292 levels in WT-LRRK2 to 3fold compared with cells that lacked FLAG-Rab7L1 expression (Fig. 3A and B). In contrast, WT-LRRK2 could only phosphorylate Rab7L1 to 4% when LRRK2 outnumbers Rab7L1 molecules 8: 1 (Supplementary Material, Fig. S6).

Co-transfection experiments with WT-Rab7L1 expressed together with WT, R1441C, Y1699C, or G2019S-LRRK2 showed that pSer1292 can be simulated by Rab7L1 irrespective of LRRK2 mutation status (Fig. 3C and D). To further study the mechanism of Rab7L1-mediated LRRK2 kinase activation, we co-transfected R1441C-LRRK2, which shows the most robust Rab7L1 activation (Fig. 3C and D), together with WT, TN, and CC-Rab7L1. The TN-Rab7L1 protein that LRRK2 cannot phosphorylate could not enhance pS1292-LRRK2 levels. Likewise, expression of the CC-Rab7L1 mutant that holds Rab7L1 cytoplasmic had no effect on pS1292-LRRK2 levels (Fig. 3E and F). These results suggest that Rab7L1 must be GTP and membrane-bound to stimulate LRRK2 autophosphorylation. In contrast to Rab7L1, co-expression of LRRK2 together with WT, TN, or CC-Rab10 did not affect pS1292-LRRK2 levels (Fig. 3G and H).

In parsing out pS1292-LRRK2 protein into cytosolic soluble fractions and triton-x100 extracted membrane-enriched fractions, we found that nearly all of the increased levels of pS1292-LRRK2 caused by Rab7L1 co-expression were occurring at the membrane (Fig. 3I and J). While the proportion of pS1292-LRRK2 in the membrane-enriched fraction is increased in the presence of WT-Rab7L1 (Fig. 3K), expression of TN or CC-Rab7L1 diminished total LRRK2 protein in the membrane-enriched fraction relative to the cytosolic fraction (Fig. 3L). As expected, the pS1292-LRRK2 signal was not present in a S1292A-LRRK2 mutant control construct, whether or not Rab7L1 was expressed (Supplementary Material, Fig. S7). These results suggest that WT-Rab7L1, but not mutant forms of Rab7L1 that lack membrane and/or nucleotide binding, activate LRRK2 autophosphorylation in membrane-enriched fractions.

Rab7L1 promotes LRRK2 recruitment to the trans-Golgi network

Previously, we demonstrated that LRRK2 and Rab7L1 lead to the clearance of Golgi-derived vesicles through endolysosomal pathways (9,13). We tested whether the increase in LRRK2 protein recruited to membrane-enriched fractions caused by WT-Rab7L1 expression is due to enhanced LRRK2 relocalization to the trans-Golgi network. We imaged the co-localization of LRRK2 and Rab7L1 (Fig. 4A) or TN-Rab7L1 (Fig. 4B) together with the trans-Golgi marker TGN46 using an automated high content imaging assay previously described (37). All three pathogenic LRRK2 mutations R1441C, Y1699C, and G2019S increased LRRK2 recruitment to the TGN and co-localization with Rab7L1 (Fig. 4A). Conversely, TN-Rab7L1 expression did not allow for LRRK2 localization to the TGN (Fig. 4B), consistent with measurements from membrane-enriched fractions (Fig. 3). Therefore, blocking the ability of either LRRK2 or Rab7L1 to bind GDP/GTP (T1348N-LRRK2, or TN-Rab7L1, respectively) abolishes LRRK2 co-localization with TGN46, and all pathogenic LRRK2 mutations increase localization to the TGN (Fig. 3C).

Rab proteins function as poor LRRK2 kinase substrates in vitro

Recent work has demonstrated that Rab8/10 proteins, purified from e. coli., serve as exceptional LRRK2 substrates in vitro (16). While the crystal structures of Rab7L1 and GDP-bound Rab10 are not known, both GDP-bound (PDB ID: 4LHV) and GTP-bound Rab8 (PDB ID: 4LHW) crystal structures are solved (27). Structural alignment showed the switch II motif, where the LRRK2 dependent phosphorylation site Thr72 resides (Supplementary Material, Fig. S1), is highly unstructured in the GDP-bound form (Fig. 5A). In reasoning that an unstructured domain may poorly satisfy conformational requirements for efficient LRRK2 phosphorylation, we purified Rab10 and Rab7L1 from mammalian cells and bound the protein to a solid surface (FLAG-resin). In combining this form of Rab protein with recombinant full-length G2019S-LRRK2 protein, we could measure limited LRRK2-mediated phosphorylation of GTP-bound Rab10 and not GDP-Rab10 (Fig. 5B and C and Supplementary Material, Fig. S8). Only a small fraction of GTP-Rab10 could be phosphorylated by G2019S-LRRK2 (∼3%) in these conditions, and we were unable to visualize any phosphorylation of GTP or GDP-bound FLAG-Rab7L1 in vitro. These results highlight the possible requirements we demonstrated in cells of Rab prenylation and membrane interaction, as well as GTP-binding, for physiologically meaningful interaction with LRRK2.

LRRK2 phosphorylation of Rab10 blocks interaction with TBC1D4/AS160

As LRRK2 requires Rab8, Rab10, and Rab7L1 substrates to be membrane and GTP-bound for phosphorylation in cells, we next tested whether LRRK2-mediated Rab phosphorylation may affect interaction with GAP proteins that revert the GTP-membrane-bound state back to an inactive GDP-bound state. Although the identities of GAP proteins for Rab7L1 have not yet been described, TBC1D4/AS160 is a known critical GAP for Rab8 and Rab10 function (28–30). We first confirmed that all the LRRK2-mediated phosphorylation we could visualize in Rab10 was localized to the threonine 73 position as described (16). Accordingly, substitution of the LRRK2 phosphorylation site residue (Thr73) to either alanine (TA) or glutamic acid (TE) largely blocked LRRK2 phosphorylation, similar to the effects of the LRRK2 kinase inhibitor MLi2 (Fig. 5D). The phospho-mimic mutation T72E but not T72A increases the perinuclear localization of Rab10, suggesting phosphorylation could potentially change the localization of Rab10 (Fig. 5E). However, immunoprecipitation assays demonstrated that substitution of the switch II LRRK2 phosphorylation site blocked the interaction between Rab10 and TBC1D4/AS160 (Fig. 5F and G). These results suggest that LRRK2 modification of the critical Thr-switch II residue may block interaction with TBC1D4/AS160 and thus prolong a Rab membrane and GTP-bound state, with altered interaction with effector proteins.

DNA plasmids

N-terminal Myc-tagged LRRK2 constructs have been previously described (36). The LRRK2 mutation construct G2019S-S1292A was generated using Phusion Site-Directed Mutagenesis Kit (Life Technologies). To generate Rab10 expression plasmids, WT, Rab10 DNA sequence were synthesized and sub-cloned into pCDNA3.1GFP-N1 plasmid or pCDNA3.1-Flag plasmids from Genscript. Rab8 constructs are obtained from Addgene (Plasmids 86075, 86076, 86077) courtesy of the Lei Lu laboratory. pEGFP-Rab7L1 plasmids are generated as previously described (9). Novel mutations in the Rab expression constructs were all made with the Phusion Site-Directed Mutagenesis Kit (Life Technologies). The N-terminal Flag-tagged AS160 expression plasmid was a kind gift from Gustav Lienhard. All plasmids were purified with Qiagen Maxi-prep kits following manufacturer’s instructions.

Cell culture

HEK 293 (clone FT, Invitrogen) and SH-SY5Y (clone CRL-2266, ATCC) cells were maintained in 10% fetal bovine serum and DMEM. All cells were seeded at ∼60% confluency 24-h prior to transfections of plasmids using polyethylenimine at a 3: 2 ratio (w/v) with plasmids. Cells were harvested or imaged 16-h post-transfection. Cell imaging was accomplished with Zen Blue software (Carl Zeiss) on a Cell Observer with incubation and cool LED imaging. To separate lysates into soluble cytosolic fractions and membrane-enriched fractions, cells were scraped into lysisbuffer containing 150 mM NaCl, 50 mM Tris-HCl, pH 7.4, 10mM MgCl₂ 1x PhosSTOP and Protease inhibitor cocktails (Roche) followed by ultracentrifugation at 150, 000 x g for 20 min. Pellets were next resuspended into the same lysis buffer supplemented with 1% triton-x100.

Immunoblotting

Proteins and lysates were analysed via SDS-PAGE followed by transfer to PVDF membranes and analyses with Classico ECL reagent (Millipore) and signal recorded digitally on a Chemidoc Touch (BioRad). Intensities were calculated with the latest version of ImageLab software. The following antibodies were used: N241A/34 anti-LRRK2 (Antibodies Inc), MJFR-19–7-8 anti-pS1292-LRRK2 (Abcam), anti-eGFP antibody (Abcam), anti-FLAG M2 (Sigma), VDAC, β-actin, and GAPDH (Santa Cruz).

Phos-tag gel

Phosphoproteins were separated using a ‘phos-tag’ approach, as described (39) with acrylamide-gel matrices supplemented with 100 µM MnCl₂ and 50 µM of Phos-Tag reagent (Wako Chemicals). For separation of GFP-Rabs and phospho-GFP-Rabs, 5% phos-tag gels were used. For separation of FLAG-Rab10 and phospho-FLAG-Rab10, 10% gel was used. Gels were run at 100 V for 1.5 h in 25 mM Tris, 192 mM glycine, 0.1% SDS running buffer, followed by 2x washing in running buffer supplemented with 5 mM EDTA before transfer to PVDF (Immobilon-FL) membrane at 35 V for 10 h.

Protein purification

N-terminal Flag-Rab10 and Flag-Rab7L1 plasmids were transfected to HEK293 cells as described and harvested into lysis buffer (described above) supplemented with 100 μM GDP (Sigma). Lysates were cleared by centrifugation at 50,000 × g and incubated with M1-Flag-resin (Sigma). Resins were washed five times with lysis buffer and eluted with the same lysis buffer supplemented with 100 μg ml⁻¹ 3x Flag-peptide (Sigma). Recombinant full-length N-terminal FLAG-LRRK2 protein was purchased from Invitrogen. All protein purities were assessed by coomassie blue staining PAGE analysis and were all above >95% purity as assessed by ImageLab software (BioRad).

Immunoprecipitations

For Rab10-GAP interaction experiments, anti-GFP antibody resin loaded with eGFP-tagged WT, T73A, and T73D-Rab10 were incubated with lysates containing AS160 protein. After incubation for 4 h at 4 °C, resins were washed five times in lysis buffer and eluate analysed via immunoblot as above. In all immunoprecipitation experiments, anti-eGFP resin with immobilized Rab10 or Rab7L1 were used as negative controls to assess non-specific binding.

In vitro kinase assay

25 nM full-length G2019S-LRRK2 protein (Life Technologies) was incubated with 100 nM recombinant Flag-Rab10 or Flag-Rab7L1 in a kinase reaction buffer containing 150 mM NaCl, 50 mM Tris-HCl, pH 7.4, 10 mM MgCl_2, and 1 mM ATP, in the presence of 500 μM of GTP or GDP as indicated, with or without MLi2 compound at 100 nM (synthesized in-house). Reactions were allowed to proceed for 1 h at 30 °C prior to termination.

Trans-Golgi localization assay

Rab7L1-induced LRRK2 recruitment to the trans-Golgi network was measured using an automated assay as previously described (40). Briefly, transiently 3x-flag LRRK2/2x-myc Rab7L1 transfected HEK293 cells were fixed and stained 24-h post-transfection for FLAG (LRRK2), Myc (Rab7L1) and TGN46. Transfected cells were imaged and quantified on a Cellomics VTI high content imager using Spot Detector bioapplication for % of cells with LRRK2 and TGN46 positive spots from total number of LRRK2 transfected cells. At least 1000 cells in each of 6-independent wells were measured for each condition, and this was repeated at least twice. Cells were also imaged on a Zeiss LSM880 confocal microscope.

Statistics

Analyses were performed in GraphPad Prism 5.0. Graphs were generated in GraphPad Prism 5.0 and arranged in Adobe Illustrator 9.0. All statistical tests are reported in figure legends.