Structural basis for Parkinson’s disease-linked LRRK2’s binding to microtubules

Snead, David M.; Matyszewski, Mariusz; Dickey, Andrea M.; Lin, Yu Xuan; Leschziner, Andres E.; Reck-Peterson, Samara L.

doi:10.1038/s41594-022-00863-y

Download PDF

Article
Open access
Published: 12 December 2022

Structural basis for Parkinson’s disease-linked LRRK2’s binding to microtubules

Nature Structural & Molecular Biology volume 29, pages 1196–1207 (2022)Cite this article

7720 Accesses
9 Citations
26 Altmetric
Metrics details

Subjects

Abstract

Leucine-rich repeat kinase 2 (LRRK2) is one of the most commonly mutated genes in familial Parkinson’s disease (PD). Under some circumstances, LRRK2 co-localizes with microtubules in cells, an association enhanced by PD mutations. We report a cryo-EM structure of the catalytic half of LRRK2, containing its kinase, in a closed conformation, and GTPase domains, bound to microtubules. We also report a structure of the catalytic half of LRRK1, which is closely related to LRRK2 but is not linked to PD. Although LRRK1’s structure is similar to that of LRRK2, we find that LRRK1 does not interact with microtubules. Guided by these structures, we identify amino acids in LRRK2’s GTPase that mediate microtubule binding; mutating them disrupts microtubule binding in vitro and in cells, without affecting LRRK2’s kinase activity. Our results have implications for the design of therapeutic LRRK2 kinase inhibitors.

Structure of LRRK2 in Parkinson’s disease and model for microtubule interaction

Article 19 August 2020

C. K. Deniston, J. Salogiannis, … A. E. Leschziner

Structure of LRRK1 and mechanisms of autoinhibition and activation

Article Open access 19 October 2023

Janice M. Reimer, Andrea M. Dickey, … Andres E. Leschziner

The LRRK2 N-terminal domain influences vesicle trafficking: impact of the E193K variant

Article Open access 02 March 2020

Algerta Marku, Maria Dolores Perez Carrion, … Giovanni Piccoli

Main

PD is the second most common neurodegenerative disease, affecting more than ten million people worldwide. Autosomal dominant missense mutations in LRRK2 are a major cause of familial PD^1,2,3,4, and mutations in LRRK2 are also linked to sporadic cases of PD^5,6. All PD-linked mutations in LRRK2 increase its kinase activity^7,8,9,10, and increased LRRK2 kinase activity in the context of a wild-type (WT) protein is also associated with sporadic PD cases¹¹. LRRK2-specific kinase inhibitors have been developed to treat PD and are in clinical trials (NCT04056689 and NCT03710707).

Although it remains unclear how it drives PD, LRRK2 has been functionally linked to membrane trafficking^12,13,14. Mutant LRRK2 causes defects in endo/lysosomal, autophagosomal, and mitochondrial trafficking^{15,16,17,18,19}, and LRRK2 regulates lysosomal morphology^20,21,22,23. Although the bulk of LRRK2 is found in the cytosol, it can associate with membranes under some conditions^{20,21,24,25,26,27}. A subset of Rab GTPases, which are master regulators of membrane trafficking²⁸, are phosphorylated by LRRK2, and PD-linked LRRK2 mutations increase Rab phosphorylation in cells^9,29. Phosphorylation of Rabs by LRRK2 is linked to alterations in ciliogenesis^25,26,28 and defects in endolysosomal trafficking^{16,20,22,23,30,31,32}. LRRK2 also co-localizes with microtubules in cells and in vitro^33,34,35,36. Cellular localization of LRRK2 with microtubules is seen with elevated expression levels and is enhanced by type-1 LRRK2-specific kinase inhibitors^33,35,37,38. In vitro, the catalytic half of LRRK2 alone can bind to microtubules³⁵. In addition, many PD-linked mutations (p.R1441C, p.R1441G, p.Y1699C, and p.I2020T) increase microtubule association in cells in conjunction with elevated expression of the mutant protein^33,37. It is not understood how LRRK2 perturbs cellular trafficking or how the cellular localization of LRRK2—cytosolic, membrane-associated, and/or microtubule-bound—contributes to its function or PD pathology. Developing tools that control the localization of LRRK2 in cells is crucial to determine LRRK2’s cellular function and to understand the molecular basis of PD.

LRRK2 is a large, multidomain protein (Fig. 1a). The amino-terminal half contains armadillo, ankyrin, and leucine-rich repeat domains. The carboxy-terminal half contains LRRK2’s enzymatic domains—both a Roco family GTPase (Ras-of-complex or ROC domain) and a kinase—as well as a scaffolding domain (C-terminal of ROC, or COR) and a WD40 domain. The COR domain is further subdivided into COR-A and COR-B. Here we refer to the catalytic half of LRRK2 as LRRK2^RCKW, named for its ROC, COR, kinase, and WD40 domains. Recent structures of LRRK2 have revealed the architecture of LRRK2 at near-atomic resolution^35,39. A 3.5-Å structure of LRRK2^RCKW showed that LRRK2’s catalytic half is J-shaped, placing the kinase and GTPase domains in close proximity³⁵. Later, a 3.5-Å structure of full-length LRRK2 revealed that the N-terminal half of LRRK2 wraps around its enzymatic half, with the leucine-rich repeats blocking the kinase’s active site in an autoinhibited state³⁹. A 14-Å structure of LRRK2 carrying the p.I2020T PD mutation bound to microtubules in cells was obtained using cryo-electron tomography (cryo-ET)³⁴. The cryo-ET map was used to guide integrative modeling, leading to a molecular model for the enzymatic half of LRRK2 bound to microtubules³⁴. This model was updated when the 3.5-Å cryo-EM structure of LRRK2^RCKW was docked into the cryo-ET structure³⁵. In these models, LRRK2^RCKW wraps around the microtubule using two dimerization interfaces, one between WD40 domains and the other between COR-B domains³⁵. In the models, the ROC GTPase domain faces the microtubule, although the cryo-ET structure did not reveal any direct interactions between LRRK2 and the microtubule³⁴. An isolated ROC domain has also been shown to interact with alpha- and beta-tubulin heterodimers⁴⁰.

**Fig. 1: Cryo-EM structure of microtubule-associated LRRK2^RCKW-I2020T.**

We previously investigated the possible functional consequences of LRRK2’s interaction with microtubules by looking at the impact of LRRK2 on the movement of the microtubule-based motor proteins dynein and kinesin in vitro³⁵. Both dynein and kinesin interact with their membranous cargos directly or indirectly via connections to Rab GTPases, including those Rabs phosphorylated by LRRK2 (refs. ^41,42,43,44). Using single-molecule assays, we showed that low nanomolar concentrations of LRRK2^RCKW blocked the movement of both dynein and kinesin on microtubules³⁵. Furthermore, we showed that the conformation of LRRK2’s kinase domain was essential for this effect³⁵. LRRK2 with its kinase domain ‘closed’ (the canonical active conformation) by LRRK2-specific type-1 kinase inhibitors blocked motility of dynein and kinesin³⁵, in agreement with studies showing that these inhibitors enhance the association of LRRK2 with microtubules in cells^33,35,37,38. In contrast, LRRK2 predicted to have its kinase in an ‘open’ or inactive conformation (in the presence of type-2 kinase inhibitors) no longer robustly blocked the movement of dynein or kinesin³⁵.

Despite these insights, how LRRK2 filaments form and interact with microtubules remains unknown. Here, we report cryo-EM structures of microtubule-bound filaments formed by LRRK2^RCKW. Our structures reveal direct interactions between LRRK2’s ROC domain and the microtubule. We show that microtubule binding is mediated by electrostatic interactions and involves the negatively charged C-terminal tubulin tails. We also present a cryo-EM map of the C-terminal half of LRRK1 (LRRK1^RCKW), LRRK2’s closest human homolog. Despite its structural similarity to LRRK2^RCKW, we show that LRRK1^RCKW does not bind to microtubules. We identify microtubule-facing basic amino acids that are conserved in LRRK2’s ROC domain, but not in LRRK1’s ROC domain, and are required for LRRK2’s interaction with microtubules in vitro and in cells. Mutation of these amino acids also renders LRRK2 unable to block the movement of kinesin in vitro. Together, our work reveals the structural basis for LRRK2’s filament formation and microtubule interaction and identifies mutations that perturb them in cells. These are essential advances for determining the cellular functions of LRRK2 and for the further development of therapeutic LRRK2 kinase inhibitors.

Results

Cryo-EM structure of microtubule-associated LRRK2^RCKW

To understand how LRRK2 oligomerizes on and interacts with microtubules, we set out to obtain a higher resolution structure of microtubule-associated LRRK2 filaments using an in vitro reconstituted system and single-particle cryo-EM approaches. We chose to work with LRRK2^RCKW because it can form filaments in vitro³⁵ and it accounts for the density observed in the cryo-ET reconstruction of full-length LRRK2 filaments in cells³⁴.

As has been previously observed³⁵, co-polymerization of tubulin with LRRK2^RCKW—either the WT protein, or that carrying the PD-linked mutations p.G2019S or p.I2020T—yielded microtubules decorated with LRRK2^RCKW (Extended Data Fig. 1a). Diffraction patterns calculated from images of these filaments showed layer lines, indicative of the presence of ordered filaments (Extended Data Fig. 1a). In the presence of MLi-2, a type-1 LRRK2-specific kinase inhibitor^45,46, we saw an additional layer line of lower frequency for all three constructs, suggesting that the filaments had longer-range order (Extended Data Fig. 1a). Unlike WT LRRK2^RCKW and LRRK2^RCKW-G2019S, LRRK2^RCKW-I2020T showed this additional layer line in the absence of MLi-2 as well (Extended Data Fig. 1a). Given these observations, we chose the LRRK2^RCKW-I2020T filaments that formed in the presence of MLi-2 for our cryo-EM work. The symmetry mismatch between microtubules, which are polar left-handed helices, and the LRRK2 filaments, which are right-handed and have pseudo-twofold axes of symmetry perpendicular to the microtubule, required that we largely uncouple their processing (Extended Data Fig. 1b,c and Methods). Our cryo-EM analysis resulted in several maps originating from an initial reconstruction of the filaments (Fig. 1b): a map of a LRRK2^RCKW tetramer that includes density for two microtubule protofilaments (6.6 Å) (Fig. 1c); a higher resolution map of the same LRRK2^RCKW tetramer excluding the microtubule (5.9 Å) (Extended Data Fig. 3g–i); maps of pseudo-twofold-symmetry-related LRRK2^RCKW monomers along a filament that face either the plus (‘+’) (5.0 Å) or minus (‘−’) (5.2 Å) end of the microtubule, revealing their different contacts with the microtubule (Fig. 1e,f and Extended Data Fig. 1c); and a consensus structure of LRRK2^RCKW that gives the highest resolution for the kinase domain (4.5 Å) (Extended Data Fig. 1c). The resolutions of our maps, even that of the consensus structure, are not sufficient to reveal how MLi-2 interacts with LRRK2.

The LRRK2^RCKW filaments are formed by two different homotypic dimer interfaces, involving either COR-B-COR-B or WD40-WD40 interactions (Fig. 1c), in agreement with what modeling has predicted^34,35. Each interface has a pseudo-twofold axis of symmetry perpendicular to the microtubule axis. The ROC domain points towards and contacts the microtubule (Fig. 1c–f). Our in vitro-reconstituted filaments of LRRK2^RCKW differ from those formed by full-length LRRK2 in cells³⁴, with LRRK2^RCKW forming a triple (rather than double) helix, with the strands packed closer together, likely owing to the absence of the N-terminal half of LRRK2. Despite these differences, the pitch of the helix is similar in both cases (Supplementary Table 1).

We have previously hypothesized that LRRK2’s kinase must adopt a closed conformation to form filaments around microtubules³⁵. Our current structure agrees with this prediction (Fig. 1g,h). To determine whether the closed conformation of the kinase was a consequence of the presence of MLi-2, which is expected to stabilize that state, we solved a structure of microtubule-associated LRRK2^RCKW filaments in its absence (Extended Data Fig. 2). Although these filaments are less well ordered (Extended Data Fig. 1a) and thus resulted in a lower resolution reconstruction (7.0 Å), the final map still fit a closed-kinase model of LRRK2^RCKW better than its open form (Extended Data Fig. 3a,b). Finally, the conformation of the kinase in the microtubule-associated LRRK2^RCKW-I2020T filaments appears to be more closed than that predicted by AlphaFold^47,48 for the active state of full-length LRRK2 (Extended Data Fig. 3c–f). We cannot determine at this point whether this difference is a consequence of the absence of the N-terminal half of LRRK2, the presence of the p.I2020T mutation in our filaments, a small difference in the AlphaFold modeling, or a consequence of the formation of the filaments themselves.

It has previously been proposed that the ROC domain would mediate binding of LRRK2 to microtubules, owing to its proximity to the microtubule surface in the cryo-ET map of the filaments in cells³⁴. However, the cryo-ET map showed no density connecting the ROC domain, or any other domain, to the microtubule³⁴. In contrast, our cryo-EM map showed clear density connecting LRRK2^RCKW and the microtubule (Fig. 1e,f). The fact that microtubules are directional polymers, with ‘+’ (fast-growing) and ‘–’ (slow-growing) ends, means that the ROC domains, which would otherwise be related by a twofold symmetry axis perpendicular to the microtubule, are in different local environments. In agreement with this, their connections to the microtubule became apparent only when LRRK2^RCKW monomers were refined individually (Fig. 1c,e,f and Extended Data Fig. 1).

LRRK2’s dimer interfaces are important for microtubule association

We next examined the role played by the WD40- and COR-B-mediated dimer interfaces in LRRK2’s ability to associate with microtubules. We built a model of the LRRK2^RCKW filament using rigid-body docking of individual domains from the LRRK2^RCKW structure (PDB: 6VNO)³⁵ (Fig. 2a,b). This revealed WD40–WD40 and COR-B–COR-B interfaces that are very similar to those seen previously with isolated WD40 domains⁴⁹, full-length LRRK2 COR-B–COR-B dimers³⁹, and LRRK2^RCKW dimers in the absence of microtubules³⁵. However, small differences exist between our model and the full-length LRRK2 dimer³⁹ (Extended Data Fig. 3g–j). It remains to be seen whether these differences are due to the absence of the N-terminal half of LRRK2 in the microtubule-associated filaments or to small conformational changes associated with filament formation.

**Fig. 2: Effect of mutations in LRRK2’s WD40 and COR-B domains on filament formation and microtubule binding.**

On the basis of our model, we made mutants designed to disrupt both interfaces and tested their ability to form filaments in cells and to bind microtubules or inhibit the motility of kinesin in vitro. At the WD40–WD40 interface, we mutated leucine 2343 or serine 2345 to aspartic acid (p.L2343D or p.S2345D; Fig. 2a), designed to introduce a charge clash. At the COR-B–COR-B dimer interface, we mutated arginine 1731 to leucine or aspartic acid (p.R1731L or p.R1731D; Fig. 2b), designed to disrupt the salt bridge with glutamic acid 1681. The expression levels of all mutant alleles were similar to that of full-length WT LRRK2 when transfected into 293T cells (Extended Data Fig. 4a–d). We then tested the ability of these mutations to disrupt filament formation by LRRK2 (WT except for the interface mutations) in cells, which is induced by MLi-2 (refs.^33,35,37,38) (Fig. 2c,e and Extended Data Fig. 4a,b). As has previously been shown, mutation of either the WD40–WD40 interface³⁴ or the COR-B–COR-B interface³⁹ reduced filament formation in cells. We found that p.S2345D, p.R1731L, and p.R1731D all significantly decreased MLi-2-induced LRRK2 filament formation, with p.S2345D and p.R1731D completely abolishing our ability to detect filaments in cells (Fig. 2c,e). Surprisingly, although the p.L2343D mutation has previously been shown to decrease dimerization of a purified WD40 domain⁴⁹, it did not reduce the formation of LRRK2 filaments in the presence of MLi-2 (Fig. 2c). We also tested each mutant’s ability to phosphorylate Rab10 in cells, and found that the WD40 dimerization interface mutants had no effect on LRRK2’s kinase activity, whereas the COR-B dimerization interface mutants roughly doubled it (Fig. 2d,f, and Extended Data Fig. 4c,d).

Next, we examined the effects of the mutations at the LRRK2 dimerization interfaces on LRRK2’s ability to bind microtubules or inhibit kinesin motility in vitro. To investigate LRRK2’s binding to microtubules, we incubated pure LRRK2^RCKW with in vitro-assembled, taxol-stabilized microtubules and quantified the fraction of LRRK2 that pelleted with microtubules after centrifugation. Although a point mutation at the WD40 dimerization interface (p.S2345D) did not affect LRRK2^RCKW’s pelleting with microtubules, a point mutation at the COR-B interface (p.R1731D) reduced it by about 50% (Fig. 2g and Extended Data Fig. 4e). Combining these mutations (p.R1731D and p.S2345D) largely abolished LRRK2^RCKW’s interaction with microtubules (Fig. 2g and Extended Data Fig. 4e). Cryo-EM imaging of microtubules incubated with the different mutants agreed with the binding data: we observed the layer lines that were indicative of filament formation with LRRK2^RCKW-S2345D, but not with LRRK2^RCKW-R1731D or LRRK2^RCKW-R1731D S2345D (Extended Data Fig. 4f). Previously, we showed that low nanomolar concentrations of LRRK2^RCKW blocked the movement of dynein and kinesin in vitro³⁵. To determine whether the dimerization interfaces are required for this inhibitory effect, we monitored the motility of single GFP-tagged human kinesin-1 (referred to as ‘kinesin’ here) molecules using total internal reflection fluorescence (TIRF) microscopy. As in the microtubule-binding experiments, we found that LRRK2^RCKW-S2345D blocked kinesin motility similarly to WT LRRK2^RCKW, although the COR-B interface mutant LRRK2^RCKW-R1731D or the combined interface mutant LRRK2^RCKW-R1731D S2345D no longer inhibited kinesin motility in vitro (Fig. 2h,i and Extended Data Fig. 4g). Importantly, 2D averages from cryo-EM images of LRRK2^RCKW-R1731D S2345D showed that the mutations do not alter the structure of the protein substantially (Extended Data Fig. 4h).

Electrostatic interactions drive binding of LRRK2^RCKW to microtubules

We next tested the hypothesis that LRRK2 binding to microtubules is mediated by electrostatic interactions between the negatively charged surface of the microtubule and basic residues in LRRK2’s ROC domain. In addition to the observed charge complementarity between our model of the LRRK2 filaments and the microtubule (Fig. 3a and Deniston et al.³⁵), other data support this hypothesis: (1) the symmetry mismatch between microtubules and the LRRK2 filaments suggests that there cannot be a single LRRK2-microtubule interface³⁴, (2) the cryo-ET reconstruction of filaments in cells showed no clear direct contact between LRRK2 and tubulin³⁴, and (3) the connections in our reconstruction only became apparent when LRRK2^RCKW monomers were refined individually (Fig. 1c,e,f and Extended Data Fig. 1). To directly test this hypothesis, we developed a fluorescence-based assay to monitor binding of LRRK2^RCKW to microtubules in vitro. We randomly chemically labeled primary amines of LRRK2^RCKW with BODIPY TMR-X (‘TMR’ here) and used widefield fluorescence microscopy to quantify the association of TMR-LRRK2^RCKW with Alexa Fluor 488-labeled microtubules tethered to a coverslip. Chemical labeling did not significantly impair LRRK2^RCKW kinase activity as assessed by Rab8a phosphorylation in vitro (Extended Data Fig. 4i,j). In our indirect assay of filament formation, TMR-LRRK2^RCKW also inhibited the microtubule-based motility of kinesin (Extended Data Fig. 4k), suggesting that its ability to bind microtubules was not compromised. Titration of increasing concentrations of TMR-LRRK2^RCKW to microtubules led to a dose-dependent increase in microtubule binding (Fig. 3b,c). Notably, LRRK2^RCKW bound to microtubules at low nanomolar concentrations, similar to the concentrations required to inhibit the motility of kinesin and dynein³⁵. Unlabeled LRRK2^RCKW also bound microtubules in a bulk microtubule co-sedimentation assay (Extended Data Fig. 4l,m).

**Fig. 3: LRRK2^RCKW interacts with the microtubule via electrostatic interactions.**

To determine whether electrostatic interactions contribute to the binding of LRRK2^RCKW to microtubules, we tested the effect of increasing concentrations of sodium chloride on this binding. We observed a dose-dependent decrease in microtubule binding (Fig. 3d and Extended Data Fig. 4n). We also observed a salt-dependent decrease in microtubule binding for unlabeled LRRK2^RCKW in the bulk co-sedimentation assay (Extended Data Fig. 4l,m).

Tubulin carries an overall negative charge, and the disordered, negatively charged, glutamate-rich C-terminal tails of tubulin are known to contribute to microtubule binding by many microtubule-associated proteins⁵⁰. We tested the contribution of the tubulin tails to the LRRK2^RCKW-microtubule interaction by removing them with the protease subtilisin, which cleaves tubulin near its C terminus⁵¹ (Extended Data Fig. 4o). Cleavage of tubulin tails decreased LRRK2^RCKW’s ability to bind microtubules by ~50% (Fig. 3e and Extended Data Fig. 4p). Together, these results show that LRRK2’s interaction with the microtubule is driven by electrostatic interactions and is mediated in part by the C-terminal tails of tubulin.

LRRK1^RCKW adopts a similar overall fold to LRRK2^RCKW

To identify specific residues in LRRK2 that might be important for mediating its interaction with microtubules, we used a comparative approach with its closest homolog, LRRK1. Although LRRK2 has been linked to both familial and sporadic PD^1,2,3,4,5,6, LRRK1 is not clinically associated with PD⁵², but instead is implicated in metabolic bone disease and osteopetrosis^53,54,55,56. Many of LRRK1’s domains are relatively well conserved with LRRK2, with 41%, 48%, 46%, and 50% similarity between the leucine-rich repeat (LRR), ROC, COR, and kinase domains, respectively. The N and C termini of LRRK1 and LRRK2 are more divergent; LRRK1 lacks the N-terminal armadillo repeats, and it was unclear at the time (this part of our work was done before the release of AlphaFold^47,48), on the basis of sequence analyses, whether LRRK1, like LRRK2, contained a WD40 domain, with only 27% sequence similarity in this region.

We began by solving a cryo-EM structure of the part of LRRK1 that corresponds to LRRK2^RCKW (residues 631 to 2015; referred to as LRRK1^RCKW) (Fig. 4a and Extended Data Fig. 5a). The resolution of the LRRK1^RCKW monomer (5.8 Å) was limited by the fact that the protein adopted the same strong preferred orientation that we had observed for LRRK2^RCKW (ref. ³⁵). Although LRRK2^RCKW forms trimers, which allowed us to solve its high-resolution structure³⁵, we saw no evidence of trimer formation by LRRK1^RCKW. Our structure, obtained in the presence of GDP but in the absence of ATP, shows that LRRK1^RCKW adopts the same overall J-shaped domain organization seen in LRRK2^RCKW and contains a WD40 domain (Fig. 4a,b). Our map revealed that the αC helix in the N-lobe of LRRK1’s kinase is about four turns longer than that in LRRK2 (Fig. 4c), a feature correctly predicted by the AlphaFold^47,48 model of LRRK1. Our structure also revealed a density corresponding to a C-terminal helix extending from the WD40 domain and lining the back of the kinase domain, as is the case for LRRK2, but the LRRK1 helix appears to be shorter (Fig. 4d). This disagrees with the LRRK1 structure predicted by AlphaFold, which has a longer C-terminal helix (Fig. 4d). The meaning of this difference is not clear at this time, as the AlphaFold structure was modeled in the active conformation (closed kinase), whereas our LRRK1^RCKW is in an inactive, open-kinase conformation and lacks the amino-terminal repeats. At the current resolution, LRRK2^RCKW and LRRK1^RCKW are otherwise very similar, confirming that the overall domain organization is conserved between these two proteins.

**Fig. 4: LRRK1^RCKW is structurally similar to LRRK2^RCKW but does not bind to microtubules.**

LRRK1^RCKW does not bind microtubules

Given the structural similarity between LRRK1 and LRRK2 (Fig. 4a,b), we wondered whether LRRK1 could also bind microtubules. To test this, we randomly chemically labeled LRRK1^RCKW with BODIPY TMR-X and used widefield fluorescence microscopy to quantify microtubule binding in vitro. We did not observe association of TMR-LRRK1^RCKW with microtubules (Fig. 4e,f). Unlabeled LRRK1^RCKW also failed to co-sediment with microtubules (Fig. 4g and Extended Data Fig. 5b) or block kinesin motility, even at a concentration of 100 nM (Fig. 4h–j). Together, these data show that, in contrast to LRRK2, LRRK1 does not interact with microtubules.

Basic residues in LRRK2’s ROC domain are important for microtubule binding

Next, we used our discovery that LRRK1^RCKW and LRRK2^RCKW share a similar structure, but only LRRK2^RCKW binds microtubules, to identify amino acids in LRRK2 that are important for microtubule binding. A sequence alignment of LRRK1 and LRRK2 revealed several basic patches in LRRK2’s ROC domain that are not well conserved in LRRK1 (Extended Data Fig. 5c). These basic patches create a positively charged surface facing the microtubule that is absent in LRRK1 (Fig. 5a,b). The patches correspond to residues 1356–1359 (KTKK in human LRRK2), 1383–1386 (KRKR in human LRRK2), and 1499–1502 (KLRK in human LRRK2). In our highest resolution maps, in which we refined individual LRRK2^RCKW monomers in the filament and their contacts with the microtubule (Fig. 1e,f), the strongest density connecting LRRK2^RCKW to tubulin involves the 1356–1359 and 1383–1386 basic patches (Fig. 5c and Extended Data Fig. 5d). To determine whether these basic patches are required for LRRK2’s interaction with microtubules, we mutated two basic residues to alanine in each patch in the context of LRRK2^RCKW (p.K1358A and p.K1359A or p.R1384A and p.K1385A) and tested the the ability of each mutant to bind to microtubules in vitro. Both mutants showed a significant decrease in microtubule binding in a microtubule co-sedimentation assay compared with WT LRRK2^RCKW (Fig. 5d). LRRK2^RCKW-K1358A K1359A also showed a significant reduction in its inhibition of kinesin motility in vitro compared with WT LRRK2^RCKW (Fig. 5e,f and Extended Data Fig. 5e). Finally, we introduced full-length GFP-LRRK2, carrying either of the two basic-patch mutations, into human 293T cells and quantified microtubule association in the absence or presence of MLi-2. In the absence of MLi-2, all three constructs (WT and the two basic-patch mutants) formed little or no filaments in cells (Fig. 5g and Extended Data Fig. 5f). Treatment with MLi-2 resulted in the appearance of filaments in a significant percentage of cells carrying WT LRRK2, but failed to induce filament formation in cells carrying thebasic-patch mutants (Fig. 5g and Extended Data Fig. 5f). We also tested whether GFP-LRRK2 carrying the PD-linked p.I2020T mutation, which forms filaments in cells in the absence of MLi-2 (refs. ^33,35,37,38) (Fig. 5h), is sensitive to a basic-patch mutation. Indeed, GFP-LRRK2-I2020T no longer formed microtubule-associated filaments in cells while carrying the p.K1358A and p.K1359A mutations (Fig. 5h and Extended Data Fig. 5g). Cryo-EM imaging of microtubules incubated with LRRK2^RCKW carrying either of the two basic-patch mutants did not show the layer lines that are indicative of filament formation (Extended Data Fig. 4f). Class averages from cryo-EM images of the soluble form of those mutants also showed that the mutations do not alter the structure of the protein substantially (Extended Data Fig. 4h).

**Fig. 5: Basic patches in the ROC domain are involved in LRRK2’s binding to microtubules.**

Although none of the most common PD-linked mutations in LRRK2 are found in these basic-patch regions, the recently reported p.R1501W variant⁵⁷ is found in the ROC domain facing the microtubule, near the basic patches we identified (Fig. 5c). To determine whether p.R1501W alters LRRK2’s interaction with microtubules, we expressed GFP-LRRK2-R1501W in 293T cells. In the presence of MLi-2, LRRK2-R1501W showed a ~50% reduction in the fraction of cells containing microtubule-bound filaments compared with WT LRRK2 (Fig. 5i and Extended Data Fig. 5h). Although the effect of the p.R1501W mutation was milder than that of the basic-patch mutations in the context of WT LRRK2, it was as extreme as the basic-patch mutants when combined with the p.I2020T mutation, where it also abolished filament formation (Fig. 5j and Extended Data Fig. 5g).

Importantly, none of the effects described above are due to changes in protein expression levels (Extended Data Fig. 5i) or to changes in the kinase activity of LRRK2 (Extended Data Fig. 5i,j). LRRK2 with the basic-patch mutants and LRRK2-R1501W show similar levels of Rab10 phosphorylation in cells compared with WT LRRK2, and they do not alter the increased Rab10 phosphorylation seen in the context of p.I2020T (Extended Data Fig. 5i,j).

Discussion

Here we report a structure of in vitro reconstituted LRRK2^RCKW filaments bound to microtubules. Our structure confirmed our previous proposal that filament formation by LRRK2 requires its kinase to be in a closed (active) conformation³⁵. This provides a structural explanation for the observation that LRRK2-specific type-1 kinase inhibitors, which are expected to stabilize the closed conformation, induce filament formation in cells^34,35,37,38. We also report a structure of the catalytic half of LRRK1 (LRRK1^RCKW), LRRK2’s closest homolog, which shows that the overall fold is similar in both LRRK proteins. Despite this similarity, LRRK1 does not bind to microtubules in vitro. We identified microtubule-facing basic patches in LRRK2’s ROC domain that are not well conserved in LRRK1 and these are located in regions of the cryo-EM map showing density connecting LRRK2^RCKW to the microtubule. Mutating two basic amino acids in LRRK2’s ROC domain was sufficient to block microtubule binding both in cells and in vitro. Together, the results of this work provide important insights and tools for probing the cellular function and localization of LRRK2 and for designing LRRK2-specific kinase inhibitors.

The previous reconstruction of LRRK2 filaments in cells, at 14 Å, did not show any density connecting LRRK2 to the microtubule. The higher resolution of our map, and our ability to process LRRK2^RCKW monomers individually, allowed us to circumvent the symmetry mismatch between the LRRK2 filaments and the microtubule, and show that LRRK2^RCKW monomers related by (pseudo) twofold symmetry in the filament are indeed not truly symmetric and interact with the microtubule differently. The general features of the filaments—their curvature and basic patches facing a negatively charged surface (the microtubule)—raise the possibility that a similar geometry could be involved in the interaction between LRRK2 and membranes.

Although LRRK2 filaments were double-helical in cells³⁴, the in vitro reconstituted LRRK2^RCKW filaments were triple-helical and packed closer together. However, the helical parameters are very similar between the structures, suggesting that the underlying structure of the filaments is similar as well. The most likely explanation for the differences is the absence of the N-terminal repeats in our structure of the LRRK2^RCKW filaments. Although the N-terminal half of LRRK2 was present in the filaments reconstructed in cells, they were disordered and absent from the final map³⁴. Placing the AlphaFold model of LRRK2 into the cryo-ET map of filaments in cells showed major clashes between the filament itself (formed by the RCKW domains) and the N-terminal repeats (Extended Data Fig. 6). Thus, the filaments could not form unless the N-terminal repeats are undocked from the rest of the protein. Their presence, albeit in a flexible state, could explain the larger spacing, and thus lower number of helices, seen in LRRK2 versus LRRK2^RCKW filaments; the disordered N-terminal repeats could act as ‘spacers’ that prevent the filaments from packing closer together.

Our data suggest that LRRK2 can bind microtubules as very short oligomers. This binding mode is likely to be the preponderant one at the low concentrations used in our in vitro single-molecule assays. We base these observations on the mutants we designed to disrupt dimerization interfaces (COR-B and WD40). Any mutant that completely abolishes one dimerization interface would allow LRRK2 to form dimers (via the other interface) but would prevent the formation of longer oligomers. Although mutants predicted to break either the COR-B (p.R1731D) or WD40 (p.S2345D) interfaces abolished formation of LRRK2 filaments in cells, their effects on microtubule binding in vitro were far less extreme, with p.R1731D resulting in a ~50% decrease and p.S2345D having no significant effect. The p.S2345D mutant likely does not fully disrupt the WD40 interface, as it is able to form some filaments in vitro under the high concentrations used for cryo-EM, although these filaments are less well ordered. The mutants’ affinity for microtubules correlates with their ability to inhibit kinesin motility: p.R1731D is unable to inhibit the motor, whereas p.S2345D inhibits motility as much as WT LRRK2 does. Taken together, these data suggest that small LRRK2 oligomers, as small as a dimer, could act as roadblocks for microtubule-based transport. This possibility, along with the fact that type-1 inhibitors stabilize the conformation of LRRK2 that favors microtubule binding, should be considered when designing LRRK2 inhibitors.

An intriguing observation from our data is the increase in Rab10 phosphorylation in cells expressing LRRK2 with the p.R1731D mutation, designed to disrupt the COR-B interface. Since the available structural information shows that LRRK2 adopts the same autoinhibited conformation in its monomeric and dimeric forms, this was not a result we had predicted. One possible explanation is that conformational changes involved in the activation of LRRK2 are favored in its monomeric form, which the p.R1731D mutation would promote. Alternatively, this effect could reflect differences in cellular localization between the monomer and the dimer, which could in turn change their exposure to the Rab10 substrate.

Although LRRK2 readily binds microtubules at low concentrations in vitro, whether LRRK2 binds to and/or forms filaments around microtubules in cells expressing endogenous levels of LRRK2 remains an open question. Although the only reports of LRRK2 interacting with microtubules in cells so far have been under overexpression conditions^{18,33,34,35,38}, only a limited number of cell types have been imaged for LRRK2 localization, and to our knowledge there are no reports of live-cell imaging of endogenous LRRK2. Thus, an important future goal will be to determine the localization and dynamics of LRRK2 expressed at endogenous levels in PD-relevant cell types. A recent report suggests that a noncoding LRRK2 PD variant leads to increased LRRK2 expression in induced microglia⁵⁸. In addition, LRRK2 expression levels are elevated in a variety of immune cells in people with PD compared with age-matched healthy controls^59,60. These findings raise the possibility that increased expression of WT LRRK2 could be linked to PD. Our finding that the interaction of WT LRRK2^RCKW with microtubules acts as a potent roadblock for the microtubule-based motors dynein and kinesin³⁵ suggests a mechanism for how increased LRRK2 expression levels could be detrimental for membrane trafficking. All of the membrane cargos that LRRK2 has been implicated in trafficking—including lysosomes, endo-lysosomes, autophagosomes, and mitochondria¹⁴—are moved by dynein and kinesin^61,62,63. Elevated LRRK2 kinase activity leading to the phosphorylation of Rab GTPases is also linked to changes in membrane trafficking, and specifically in the recruitment of adapter proteins that can bind dynein and kinesin motors^64,65. Thus, examining the effects of increased LRRK2 expression in combination with increased LRRK2 kinase activity may be relevant for understanding the molecular basis of PD.

Methods

Cloning, plasmid construction, and mutagenesis

LRRK2^RCKW and Rab8a protein expression vectors were cloned as previously described³⁵. The LRRK1 sequence was codon optimized for expression in Spodoptera frugiperda (Sf9) cells and synthesized by Epoch Life Science. The DNA coding for WT LRRK1 residues 631–2015 (LRRK1^RCKW) was cloned through Gibson assembly into the pKL baculoviral expression vector (RRID: Addgene_110741), with an N-terminal His₆-Z-tag and TEV protease cleavage site. LRRK2 mutants were cloned using QuikChange site-directed mutagenesis (Agilent), or Q5 site-directed mutagenesis (New England Biolabs), following the manufacturer’s instructions. As previously described for LRRK2^RCKW (ref. ³⁵), the LRRK1^RCKW plasmid was used for the generation of recombinant baculoviruses according to bac-to-bac expression system protocols (Invitrogen).

For mammalian expression, GFP-LRRK2 was cloned into the pDEST53 vector (RRID: Addgene_25044) as previously described³⁵. LRRK2 mutants were cloned using QuikChange site-directed mutagenesis (Agilent) using standard protocols, except for liquid cultures of Escherichia coli, which were grown at 30 °C. EGFP-Rab10 (ref. ⁶⁶) was obtained from Addgene (RRID: Addgene_49472), and pET17b-Kif5b(1-560)-GFP-His⁶⁷ was obtained from Addgene (RRID: Addgene_15219).

LRRK2^RCKW and LRRK1^RCKW expression and purification

N-terminally His₆-Z-tagged LRRK2^RCKW was expressed in Sf9 insect cells (Thermo Fisher Scientific cat. no. 11496015) and purified as previously described³⁵. Protocols are also available at https://doi.org/10.17504/protocols.io.rm7vzyyrrlx1/v1 and https://doi.org/10.17504/protocols.io.81wgb6693lpk/v1. Briefly, ~1 L insect cells was infected with baculovirus and grown at 27 °C for 3 days. Pelleted Sf9 cells were resuspended in lysis buffer (50 mM HEPES pH 7.4, 500 mM NaCl, 20 mM imidazole, 0.5 mM TCEP, 5% glycerol, 5 mM MgCl₂, 20 μM GDP, 0.5 mM Pefabloc, and protease inhibitor cocktail tablets) and lysed by Dounce homogenization. Clarified lysate was incubated with Ni-NTA agarose beads (Qiagen), extensively washed with lysis buffer, and eluted in buffer containing 300 mM imidazole. Protein eluate was diluted twofold in buffer containing no NaCl, loaded onto an SP Sepharose column, and eluted with a 250 mM to 2.5 M NaCl gradient. Protein was cleaved by TEV protease overnight. Cleaved protein was isolated by running over a second Ni-NTA column. Protein was concentrated and run on an S200 gel filtration column equilibrated in storage buffer (20 mM HEPES pH 7.4, 700 mM NaCl, 0.5 mM TCEP, 5% glycerol, 2.5 mM MgCl₂, 20 μM GDP). Protein was concentration to a final concentration of ~20–30 µM, as estimated by absorbance at 280 nm using an extinction coefficient of 140,150 M⁻¹ cm⁻¹.

Purification of molecular motors

Human KIF5B^1-560(K560)-GFP was purified from E. coli using an adapted protocol that has been previously described⁶⁸. (Our current protocol can also be found at https://doi.org/10.17504/protocols.io.bp2l61xrdvqe/v1.) All protein purification steps were performed at 4 °C unless otherwise noted. pET17b-Kif5b(1-560)-GFP-His was transformed into BL-21[DE3]RIPL cells (Agilent cat. no. 230280) until an optical density at 600 nm of 0.6–0.8, and expression was induced with 0.5 mM isopropyl-β-d-thiogalactoside (IPTG) for 16 hours at 18 °C. Frozen pellets from 7.5 L of culture were resuspended in 120 ml lysis buffer (50 mM Tris, 300 mM NaCl, 5 mM MgCl₂, 0.2 M sucrose, 1 mM dithiothreitol (DTT), 0.1 mM Mg-ATP, and 0.5 mM Pefabloc, pH 7.5) supplemented with one cOmplete EDTA-free protease inhibitor cocktail tablet (Roche) per 50 ml and 1 mg/ml lysozyme. The resuspension was incubated on ice for 30 minutes and lysed by sonication. Sonicate was supplied with 0.5 mM PMSF and clarified by centrifugation at 40,000 rcf (118,272g) for 60 minutes in a Type 70 Ti rotor (Beckman). The clarified supernatant was incubated with 15 ml Ni-NTA agarose (Qiagen) and rotated in a nutator for 1 hour. The mixture was washed with 100 ml wash buffer (50 mM Tris, 300 mM NaCl, 5 mM MgCl₂, 0.2 M sucrose, 10 mM imidazole, 1 mM dithiothreitol (DTT), 0.1 mM Mg-ATP, and 0.5 mM Pefabloc, pH 7.5) by gravity flow. Beads were resuspended in elution buffer (50 mM Tris, 300 mM NaCl, 5 mM MgCl₂, 0.2 M sucrose, 250 mM imidazole, 0.1 mM Mg-ATP, and 5 mM βME, pH 8.0), incubated for 5 minutes, and eluted stepwise in 0.5-mL increments. Peak fractions were combined and buffer exchanged on a PD-10 desalting column (GE Healthcare) equilibrated in storage buffer (80 mM PIPES, 2 mM MgCl₂, 1 mM EGTA, 0.2 M sucrose, 1 mM DTT, and 0.1 mM Mg-ATP, pH 7.0). Peak fractions of motor solution were either flash-frozen at −80 °C until further use or immediately subjected to microtubule-bind-and-release purification. A total of 1 ml motor solution was incubated with 1 mM AMP-PNP and 20 μM taxol on ice in the dark for 5 minutes and subsequently warmed to room temperature. For microtubule bind and release, polymerized bovine brain tubulin was centrifuged through a glycerol cushion (80 mM PIPES, 2 mM MgCl₂, 1 mM EGTA, and 60% glycerol (vol/vol) with 20 μM taxol and 1 mM DTT) and resuspended as previously described⁶⁹, and was incubated with the motor solution in the dark for 15 minutes at room temperature. The motor–microtubule mixture was laid on top of a glycerol cushion and centrifuged in a TLA120.2 rotor at 80,000 r.p.m. (278,088g) for 12 minutes at room temperature. Final pellet (kinesin-bound microtubules) was washed with BRB80 (80 mM PIPES, 2 mM MgCl₂, and 1 mM EGTA, pH 7.0) and incubated in 100 μL of release buffer (80 mM PIPES, 2 mM MgCl₂, 1 mM EGTA, and 300 mM KCl, pH 7 with 7.5 mM Mg-ATP) for 5 minutes at room temperature. The kinesin release solution was spun at 72,000 r.p.m. (225,252g) in TLA100 for 7 minutes at room temperature. The supernatant containing released kinesin was supplemented with 660 mM sucrose and flash-frozen. A typical kinesin prep in the lab yielded 0.5 to 1.5 μM K560-GFP dimer.

Rab8a expression and purification

Rab8a was expressed and purified as previously described³⁵. The protocol is also available at https://doi.org/10.17504/protocols.io.6qpvr63mzvmk/v1. Briefly, N-terminally His₆-ZZ tagged Rab8a was expressed in BL21(DE3) E. coli cells (Agilent cat. no. 200131) by addition of 0.5 mM IPTG for 18 hours at 18 °C. Cells were pelleted, resuspended in lysis buffer (50 mM HEPES pH 7.4, 200 mM NaCl, 2 mM DTT, 10% glycerol, 5 mM MgCl₂, 0.5 mM Pefabloc, and protease inhibitor cocktail tablets), and lysed by sonication on ice. Clarified lysate was incubated with Ni-NTA agarose (Qiagen). Protein was washed with wash buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 2 mM DTT, 10% glycerol, 5 mM MgCl₂) and eluted in buffer containing 300 mM imidazole. Eluate was incubated with IgG Sepharose 6 fast flow beads. Following further washing, Rab8a was cleaved off IgG sepharose beads by incubation with TEV protease at 4 °C overnight. Cleaved Rab8a was isolated by incubation with Ni-NTA agarose beads, followed by washing with buffer containing 25 mM imidazole. Purified Rab8a was run on an S200 gel filtration column equilibrated in S200 buffer (50 mM HEPES pH 7.4, 200 mM NaCl, 2 mM DTT, 1% glycerol, 5 mM MgCl₂). Protein was then concentrated and exchanged into 10% glycerol for storage.

Cryo-electron microscopy: sample preparation and imaging of filaments

LRRK2^RCKW filaments were prepared as previously described³⁵, with the exception that 10% glycerol was used instead of 10% DMSO in all the samples, except for the one that led to the initial data set (‘19dec14f’), as glycerol promotes the formation of 11- and 12-protofilament microtubules. For ‘+MLi-2’ samples, we added MLi-2 to LRRK2^RCKW to a final concentration of 5 μM after incubation with tubulin. The updated protocol is also available at protocols.io (https://www.protocols.io/view/reconstituting-lrrk2rckw-on-microtubules-for-cryo-3byl4kjb8vo5/v1).

Cryo-EM data were collected on a Talos Arctica (FEI) operated at 200 kV, equipped with a K2 Summit direct electron detector (Gatan). Automated data collection was performed using Leginon⁷⁰ (version 3.4, https://emg.nysbc.org/redmine/projects/leginon, RRID: SCR_016731) with a custom-made plug-in to automate the targeting to areas of the sample that contained LRRK2^RCKW filaments. The only exception was the first data set (‘19dec14f’), which was collected using Leginon’s regular raster target finder. The ‘19dec14f’ data set was subsequently used for training the machine-learning component of the custom-made plug-in used for all other datasets. The code for the plug-in is available at https://github.com/matyszm/filfinder (https://doi.org/10.5281/zenodo.5854954).

The ‘Apo’ reconstruction was obtained using two datasets: 836 micrographs from ‘19dec14f’ and 1,010 micrographs from ‘20aug12b.’ The ‘MLi-2’ reconstruction was also obtained from two datasets: 926 micrographs from ‘20sep10b’ and 1,430 micrographs from ‘20sep30c’. Final micrograph counts include only micrographs with at least one usable LRRK2^RCKW filament. The dose per data set varied between 5 and 5.5 electrons Å^{− 2} s^{− 1}. To accommodate for that range, we varied the exposure time between 10 and 11 seconds, with 200-ms frames, for a total number of frames between 50 and 55, and a total dose of 55 electrons Å − 2. The images were collected at the nominal magnification of ×36,000, resulting in an object pixel size of 1.16 Å. The defocus was set to −1.5 μm, with a final range of defoci from −0.5 to −2.5 μm owing to the nature of the lacey carbon grids and the collection strategy. All datasets are available on EMPIAR (see Table 1 for accession codes).

Table 1 Cryo-EM data collection, refinement, and validation statistics

Full size table

Cryo-electron microscopy: reconstruction of LRRK2^RCKW bound to a microtubule

Movie frames were aligned using UCSF MotionCor2 (ref. ⁷¹) (version 1.4.5, https://emcore.ucsf.edu/ucsf-software, RRID: SCR_016499) with the dose-weighting option on. CTF estimation was done with CTFFIND4 (ref. ⁷²) (version 4.1.14, http://grigoriefflab.umassmed.edu/ctffind4, RRID: SCR_016732) using the non-dose-weighted aligned micrographs. All micrographs containing filaments were kept regardless of the CTF estimated resolution. Data processing up to the symmetry expansion step is detailed in the protocol available at protocols.io (https://doi.org/10.17504/protocols.io.bwwnpfde). In brief, manual selected filaments from a subset of micrographs were 2D classified (Relion 3.1, https://github.com/3dem/relion, RRID: SCR_016274)⁷³, with the best classes then acting as a reference for automated filament picking (Relion 3.1). The separation distance of the particles was set to 30 Å, which ensures each particle contains one new LRRK2^RCKW dimer per strand. These particles were then filtered first by classifIcation on the basis of the presence of a microtubule, then followed up by another 2D classification focusing on the presence of ordered LRRK2^RCKW filaments if MLi-2 was present, or a blurred, disordered layer if working with apo filaments. The selected particles were then 3D classified into six classes (Relion 3.1), each corresponding to a specifically sized microtubule (from 11 to 16 protofilaments). This step is inspired by MiRP⁷⁴ and used their provided reference scaled to the appropriate pixel size (https://github.com/moores-lab/MiRP, commit at time of download: 3e3b699). Filaments with MLi-2 tend to favor 11-protofilament microtubules, whereas the apo filaments favored larger sizes. We kept all the 11-protofilament microtubules for the MLi-2 dataset and all the 12-protofilament microtubules for the apo dataset.

In order to more accurately reconstruct the LRRK2^RCKW filaments, the microtubule had to be digitally subtracted from the particles. To accomplish this, we refined the structure of the microtubule for each dataset (Relion 3.1) and subtracted it from the particles (Relion 3.1, using legacy subtraction mode). This allowed us to 2D classify (Relion 3.1) focusing on LRRK2^RCKW filaments. Particles falling into ordered 2D classes were further 3D classified (Relion 3.1). The initial reference for each subgroup (with or without MLi-2) was always a featureless cylinder and was initialized with the helical symmetry reported for microtubule-associated LRRK2 filaments in cells³⁴. Subsequent rounds used the output as the reference and were allowed to refine the symmetry, often showing multiple classes. Once the symmetry was found, a local refinement was done with the original un-subtracted particles to give a LRRK2^RCKW filament containing some of the original microtubule density. Since our LRRK2^RCKW filaments each have three strands, we used symmetry expansion to extract an individual dimer from each strand. We centered the new particles on the subtraction mask and decreased the box size to 300 pixels while keeping the Å/px scale the same. This step was performed with the new subtraction function in Relion 3.1. This resulted in 206,649 particles for the MLi-2 dataset and 49,629 particles for the apo dataset. See Extended Data Figures 1b,c and 2 for the data-processing workflow. After symmetry expansion, the newly generated particles were exclusively processed in CryoSPARC⁷⁵ (version 3.2.0, https://cryosparc.com/, RRID: SCR_016501). The first step was always to align the particles to the centered subtraction mask in order to align the particles to each other. For the particles from the MLi-2 dataset, we were able to compare the Psi Euler angle to the original angle assigned during the microtubule-only alignment. Because the particles could be flipped during the LRRK2^RCKW refinement, only particles showing 0° ± 20° and 180° ± 20° were kept. Particles with a ~180° flip were flipped back to align them to the microtubule. This left us with 133,246 particles for the MLi-2 dataset. This step was skipped for the apo particles owing to the lower particle count.

The MLi-2 dataset was processed in two ways, resulting in different levels of detail in either the kinase or ROC regions. The first processing strategy was designed to achieve a better kinase reconstruction. Here, we allowed the filtered particles to be freely aligned again, ignoring the microtubule orientation. This was followed by two local refinements: the first focused on a LRRK2^RCKW tetramer, and the second on a single LRRK2^RCKW monomer. The second strategy was designed to better resolve the contacts between the ROC domain and the microtubule. Here, we performed only local refinements on the particles with the fixed microtubule orientation. To make sure the microtubule was properly aligned, we performed a local refinement focusing only on the microtubule, which resulted in a map with no ambiguity in the tubulin orientation. We then did a local refinement focused on a LRRK2^RCKW dimer, followed by a 3D Variability Analysis (3DVA, in cryoSPARC)⁷⁶ focused on a single LRRK2^RCKW monomer with the goal of being able to separate particles that have intact LRRK2^RCKW in them. We analyzed the components generated and determined that components 1 and 2 ranged from a well formed LRRK2^RCKW to having discontinuities or weak densities. We kept only particles with a negative value for at least one of the two components. Following that, we did another refinement for a LRRK2^RCKW dimer using the filtered particles, and then used a smaller mask to cover either the ‘+’ or ‘−’ LRRK2^RCKW (see Extended Data Fig. 1c) along with a small part of the microtubule. This resulted in maps showing the ROC domains interacting with the microtubule.

For the apo sample, only the freely aligned approach was used because the particle count and resolution were too low to filter the particles by microtubule orientation. After freely aligning the particles to the recentered subtraction mask, we performed a local refinement focused on the LRRK2^RCKW tetramer. To help the alignment, we used a 20-Å low-passed LRRK2^RCKW tetramer reference built by rigid-body fitting four copies of LRRK2^RCKW into the early 9-Å reconstruction. This new reconstruction was still noisy, most likely owing to multiple conformations being present. Although Relion Class3D did not work on this dataset, we were able to use 3DVA again to help us find a component to separate apo LRRK2^RCKW into classes. Component 1 resulted in a more detailed reconstruction at both positive and negative ends of the spectra than the starting structure. We reconstructed both sets, and both were able to reach ~7-Å resolution, the data with the positive component 1 resulted in a more continuous map and was chosen as the final map.

Model creation and refinement: LRRK2 ROC domain interacting with the microtubule

For modeling, we used the maps in which the refinement had been focused on the interactions of the ROC domain with the microtubule, facing either towards the plus or minus end of the microtubule. For the initial model, we used LRRK2’s model from the AlphaFold Protein Structure Database^47,48 (Q5S007) as it had the most complete loops available for the ROC domain (using residues 1332–1525). Since AlphaFold models lack ligands, we added GDP on the basis of the placement in previous structures^35,39. Because the ROC domain occupies only a small portion of the map and some microtubule density is present, we added tubulin dimers (PDB code: 1TUB) to provide a restraint during refinement. Initial refinement was done using Rosetta (ver 3.13, https://www.rosettacommons.org/, RRID: SCR_015701) and additional refinement scripts⁷⁷. Two hundred models were generated from each map. Tubulin dimers were removed from the model before further quantification. Models with the best energy score and fit to the density were manually inspected. Small modeling errors were corrected in Isolde⁷⁸ by hand and refined one more time in Rosetta using Relax with the map density loaded in as a restraint. Five models were selected for each map and converted to poly-alanine models except for residues of interest (K1358, K1359, R1384, K1385, R1501).

Cryo-electron microscopy: sample preparation and imaging of LRRK1^RCKW

The protocol for preparing LRRK1^RCKW grids is available at protocols.io (https://doi.org/10.17504/protocols.io.b3rqqm5w). Briefly, the protein was spun down after thawing, and kept on ice until grid making. We used UltrAuFoil Holey Gold 1.2/1.3 300 mesh grids and plasma cleaned them in a Solarus II (Gatan) using the QuantiFoil Au preset. Immediately before freezing, LRRK1^RCKW was added to ‘LRRK2 buffer’ (20 mM HEPES pH 7.4, 80 mM NaCl, 0.5 mM TCEP, 2.5 mM MgCl₂, 20 μM GDP) to the desired concentration (2–6 μM protein). We used a Vitrobot Mark IV (FEI) to freeze our samples.

Cryo-EM data were collected on a Talos Arctica (FEI) operated at 200 kV, equipped with a K2 Summit direct electron detector (Gatan). Automated data collection was performed using Leginon⁶² (version 3.4, https://emg.nysbc.org/redmine/projects/leginon, RRID: SCR_016731). Reconstruction was done with four datasets (‘19dec11a’: 847 micrographs, ‘19dec21c’:926 micrographs, ‘20sep11a’: 904 micrographs, and ‘21jan18d’: 952 micrographs). One of the datasets (‘20sep11a’) was collected at a 20^o tilt. The exposure of the micrographs varied to achieve a total dose of 55 electrons Å⁻². The images were collected at a nominal magnification of ×36,000x resulting in an object pixel size of 1.16 Å. The defocus was set to −1.5 μm, which gave a range of defoci of −0.8 to −1.8 μm over all datasets. All datasets are available on EMPIAR (Table 1).

Cryo-electron microscopy: reconstruction of LRRK1^RCKW

Movie frames were aligned in cryoSPARC⁷⁵ (version 3.2.0, https://cryosparc.com/, RRID: SCR_016501) using the ‘patch motion correction’ program. CTF estimation was also done in cryoSPARC using the ‘patch CTF estimation’ program. Images were manually screened for any obvious defects and removed from further processing if defects were found. Particle picking was done with a mixture of a crYOLO⁷⁹ (version 1.6, https://cryolo.readthedocs.io/, RRID:SCR_018392) set previously trained for LRRK2^RCKW (ref. ³⁵) and simple blob picking followed by a round of 2D classification to remove obvious contaminants. Both methods gave similar results, and both were used depending on whether the picking was done on the fly (blob picker) or later (crYOLO). The final particle count was 645,743.

At this point, 2D classification was used on the combined particles. Only classes showing an intact RCKW-like shape were kept. Ab initio reconstruction gave us two classes, with two-thirds of the particles ending in the intact class. We recovered additional intact particles from the broken class after another round of 2D classification. Combining class 1 and the good 2D classes gave us 131,821 particles, from which we were able to obtain a 5.8-Å map with some stretched features, likely owing to preferred orientation. To lower the impact of preferred orientation, we used ‘Rebalance 2D’ with the rebalance factor set to 0.7, making sure the smallest supergroup is at least 70% of the size of the largest. While the resolution dropped to 6.5 Å, the severity of the stretching was reduced.

Despite this improvement, the map contained discontinuous density on the edges of the mask, suggesting problems with the automatically generated mask. We remade the mask by basing it on homology models of LRRK1^RCKW domains (ROC, COR, and Kinase; SWISS model⁸⁰), and LRRK2’s WD40 domain that we rigid-body fitted into the current best LRRK1^RCKW density and used molmap in ChimeraX (ref. ⁸¹) (version 1.2.5, https://www.cgl.ucsf.edu/chimerax/, RRID: SCR_015872) to create a map to serve as the mask. This map was low-passed to 15 Å, dilated by 8 px, and soft padded by another 8 px. This was then used to refine the structure one more time. This new map still contained artifacts in the ROC and COR-A region. We used 3DVA⁷⁶ (cryoSPARC version 3.2.0, https://cryosparc.com/, RRID: SCR_016501) to analyze the structure and found a component showing slight movement of these domains. We selected to focus on particles in the more ‘closed’ state. Refining these new particles gave us a better-defined map without artifacts at 5.8 Å resolution after using cryoSPARC’s Non-Uniform Refinement.

Single-molecule microscopy and motility assays

Single-molecule kinesin motility assays were performed as previously described³⁵. Protocol is also available at https://doi.org/10.17504/protocols.io.ewov14qykvr2/v1. Imaging was performed with an inverted microscope (Ti-E Eclipse; Nikon) equipped with a ×100 1.49-NA oil immersion objective (Plano Apo; Nikon). The microscope was equipped with a LU-NV laser launch (Nikon), with 405 nm, 488 nm, 532 nm, 561 nm, and 640 nm laser lines. The excitation and emission paths were filtered using appropriate single bandpass filter cubes (Chroma). The emitted signals were detected using an electron multiplying CCD camera (Andor Technology, iXon Ultra 888). The xy position of the stage was controlled by ProScan linear motor stage controller (Prior). Illumination and image acquisition were controlled by NIS Elements Advanced Research software (Nikon).

Single-molecule motility assays were performed in flow chambers assembled as previously described⁸². Biotin-PEG-functionalized coverslips (Microsurfaces) were adhered to glass slides using double-sided scotch tape. Each slide contained four flow chambers. Taxol-stabilized microtubules (approximately 15 mg ml⁻¹) with 10% biotin-tubulin and 10% Alexa 405-tubulin were prepared as previously described⁸². For each motility experiment, 1 mg ml⁻¹ streptavidin (in 30 mM HEPES, 2 mM magnesium acetate, 1 mM EGTA, 10% glycerol) was incubated in the flow chamber for 3 minutes. A 1:150 dilution of taxol-stabilized microtubules in motility assay buffer (30 mM HEPES, 50 mM potassium acetate, 2 mM magnesium acetate, 1 mM EGTA, 10% glycerol, 1 mM DTT and 20 μM Taxol, pH 7.4) was added to the flow chamber for 3 minutes to adhere polymerized microtubules to the coverslip. Flow chambers containing adhered microtubules were washed twice with LRRK2 buffer (20 mM HEPES pH 7.4, 80 mM NaCl, 0.5 mM TCEP, 5% glycerol, 2.5 mM MgCl₂, and 20 μM GDP). Flow chambers were then incubated for 5 minutes with either LRRK2 buffer alone or LRRK2 buffer containing the indicated concentration of WT or mutant LRRK2^RCKW. Before the addition of kinesin motors, the flow chambers were washed three times with motility assay buffer containing 1 mg ml⁻¹ casein. The final imaging buffer for motors contained motility assay buffer supplemented with 71.5 mM βME, 1 mM Mg-ATP, and an oxygen scavenger system, 0.4% glucose, 45 μg/ml glucose catalase (Sigma-Aldrich), and 1.15 mg/ml glucose oxidase (Sigma-Aldrich). The final concentration of kinesin in the motility chamber was 1 nM. K560-GFP was imaged every 500 ms for 2 minutes with 25% laser (488) power at 150-ms exposure time. Each sample was imaged no longer than 15 minutes. Each technical replicate consisted of movies from at least two fields of view containing between five and ten microtubules each.

Single-molecule motility assay analysis

Kymographs were generated from motility movies using ImageJ macros as described previously⁸². See https://doi.org/10.17504/protocols.io.ewov14qykvr2/v1 for a summary. Specifically, maximum-intensity projections were generated from time-lapse sequences to define the trajectory of particles on a single microtubule. The segmented line tool was used to trace the trajectories and map them onto the original video sequence, which was subsequently re-sliced to generate a kymograph. Brightness and contrast were adjusted in ImageJ for all videos and kymographs. Motile and immotile events (>1 second) were manually traced using ImageJ and quantified for run lengths and percentage motility. Run-length measurements were calculated from motile events only. For percent motility per microtubule measurements, motile events (>1 second and >785 nm) were divided by total events per kymograph. Bright aggregates, which were less than 5% of the population, were excluded from the analysis. Data visualization and statistical analyses were performed in GraphPad Prism (version 9.2, https://www.graphpad.com/, RRID: SCR_002798) and ImageJ⁸³ (version 1.53, https://imagej.nih.gov/ij/, RRID: SCR_003070). All tabular data are available at https://doi.org/10.5281/zenodo.6463635.

Microtubule sedimentation binding assay

Protocol is also available at https://doi.org/10.17504/protocols.io.36wgq73b5vk5/v1. Porcine brain tubulin was purchased from Cytoskeleton Taxol-stabilized microtubules were polymerized at a final concentration of ~2.5 mg/mL, and free tubulin was removed by ultracentrifugation at 108,628g for 15 minutes at 37 °C through a 64% glycerol cushion. The resulting microtubule pellet was resuspended in LRRK2 binding buffer (20 mM HEPES pH 7.4, 110 mM NaCl, 0.5 mM TCEP, 5% glycerol, 2.5 mM MgCl₂, 20 μM GDP, and 20 μM taxol). Tubulin concentration was determined by comparison of the polymerized microtubule stock to actin standards on SDS–PAGE.

For a typical LRRK^RCKW microtubule co-sedimentation assay, 200 nM LRRK^RCKW was incubated at room temperature for 10 minutes with various concentrations of microtubules in buffer containing 20 mM Hepes pH 7.4, 110 mM NaCl, 0.5 mM MgCl₂, 0.5 mM TCEP, 5% glycerol, 20 μM GDP, and 20 μM taxol. Microtubules were then pelleted by ultracentrifugation (15 minutes, 108,628 g, 25 degrees). To quantify the depletion of LRRK2^RCKW, samples of the supernatant were taken and boiled for 10 minutes in SDS buffer. Samples were run on 4–12% polyacrylamide gels (NuPage, Invitrogen) and stained with SYPRO-Red Protein Gel Stain (Thermo Fisher) for protein detection. Binding curves were fit in GraphPad Prism (version 9.2, https://www.graphpad.com/, RRID: SCR_002798) with a nonlinear regression hyperbolic curve. All tabular data are available at https://doi.org/10.5281/zenodo.6463635.

Microtubule cleavage by subtilisin

Taxol-stabilized microtubules with 10% biotin-tubulin and 10% Alexa 488-tubulin were prepared as described in https://doi.org/10.17504/protocols.io.bp2l6bdedgqe/v1. Subtilisin (Sigma-Aldrich P580), or an equivalent volume buffer, was added to 0.7 mg/mL and incubated for 10 minutes at 37 °C. The cleavage reaction was quenched by addition of 2.8 mM PMSF. To remove free tubulin and protease, the reaction mixture was centrifuged at 108,628g for 15 minutes at 37 °C through a 60% glycerol cushion. The microtubule pellet was resuspended in ×1 BRB80 (80 mM PIPES, 2 mM MgCl₂, and 1 mM EGTA, pH 7.0) with 1 mM DTT and 20 μM taxol. Cleavage was confirmed by SDS–PAGE electrophoresis (Extended Data Fig. 4o).

TMR labeling

BODIPY TMR-X NHS Ester (Thermo Fisher) was used to fluorescently label LRRK2^RCKW and LRRK1^RCKW. For a typical 40 μL labeling reaction, dye was added at a ratio of 1:1 to ~20 µM LRRK2^RCKW, followed by incubation at room temperature for 1 hour. Excess dye was removed by two consecutive buffer exchanges through Micro Bio-Spin P-6 desalting columns (Bio-Rad). Protein concentration and labeling efficiency were estimated using a NanoDrop Microvolume Spectrophotometer. The protocol is also available at https://doi.org/10.17504/protocols.io.ewov1nq5ogr2/v1.

Widefield fluorescence microtubule binding assay

Imaging was performed with an inverted microscope (Nikon, Ti-E Eclipse) equipped as described above (single-molecule microscopy and motility assays). See https://doi.org/10.17504/protocols.io.kxygxz7bdv8j/v1 for an online version of this protocol.

LRRK2^RCKW microtubule-binding experiments were performed in flow chambers made as described above (single-molecule microscopy and motility assays). LRRK^RCKW was labeled with TMR (TMR labeling, above); taxol-stabilized microtubules were polymerized from a mixture of unmodified, biotinylated, and Alexa Fluor 488-labeled bovine tubulin, as previously described (REF). To attach microtubules to the coverslip, flow chambers were incubated with 0.5 mg/mL streptavidin for 3 minutes, washed twice in buffer (30 mM Hepes pH7.4, 50 mM potasisum acetate, 2 mM magnesium acetate, 1 mM EGTA, 10% glycerol, 1 mM DTT, and 0.2 mM taxol), and then incubated with microtubules for 3 minutes. Microtubules were washed twice in buffer (20 mM Hepes pH 7.4, 80 mM NaCl, 0.5 mM MgCl₂, 0.5 mM TCEP, 5% glycerol, 20 μM GDP), and then incubated with varied concentrations of LRRK2^RCKW (6.25 nM–50 nM) for 5 minutes. Multiple fields of view were imaged along the flow chamber with the objective in widefield illumination, with successive excitation at 488 nm (15% laser power, 100 ms exposure) and 561 nm (25% laser power, 100 ms exposure).

Image analysis was performed with ImageJ⁸³ (version 1.53, https://imagej.nih.gov/ij/, RRID: SCR_003070). Average TMR-LRRK2^RCKW fluorescence intensity per microtubule was calculated from a 1-pixel-wide line drawn along the long axis of the microtubule; overall average background fluorescence intensity was subtracted. These background-subtracted intensities were averaged over all microtubules per field of view, normalized by microtubule length, to yield a single data point. Eight fields of view at each concentration of LRRK2^RKCW were then averaged. All tabular data are available at https://doi.org/10.5281/zenodo.6463635.

In vitro Rab8a phosphorylation

LRRK kinase assays were performed as previously described³⁵ with LRRK^RCKW and Rab8a purified as described above. For a typical kinase reaction, 38 nM LRRK^RCKW was incubated with 3.8 μM Rab8a for 30 minutes at 30° in buffer containing 50 mM Hepes pH 7.4, 80 mM NaCl, 10 mM MgCl₂, 1 mM ATP, 200 uM GDP, 0.5 mM TCEP. Phosphorylation of Rab8a at residue T72 by LRRK^RCKW was monitored by western blot using a commercially available antibody (Abcam cat. no. ab230260, RRID: AB_2814988) as previously described^35,84.

Immunofluorescence, confocal microscopy, and image analysis

LRRK2 filament assays were performed as previously described³⁵. Briefly, cells were plated on fibronectin-coated glass coverslips and grown for 24 hours before transfection with PEI. Cells were transfected with 500 ng of indicated GFP-LRRK2 plasmids. After 24–48 hours, cells were incubated at 37 °C with DMSO or MLi-2 (500 nM) for 2 hours. Stocks of the kinase inhibitor MLi-2 (10 mM; Tocris) were stored in DMSO at −20 °C.

Cells were rinsed briefly with ice-cold 1× PBS on ice, then fixed with ice-cold 4% PFA, 90% methanol, and 5 MM sodium bicarbonate for 10 minutes at −20 °C. Coverslips were subsequently washed three times with ice-cold PBS and then incubated with blocking buffer (1% BSA, 5% normal goat serum, 0.3% Triton X-100 in 1× PBS) for 1 hour at room temperature. Primary antibodies were diluted in antibody dilution buffer (1% BSA, 0.1% Triton X-100 in 1× PBS) and incubated at 4 °C overnight. The following day, coverslips were washed three times with 1× PBS and incubated with secondary antibodies diluted in antibody dilution buffer for 1 hour at room temperature. After secondary incubation, coverslips were washed three times with 1× PBS. Cells were briefly rinsed in ddH₂O and mounted on glass slides using CitiFluor AF-1 mounting medium (TedPella). Coverslips were sealed with nail polish and stored at 4 °C. Antibodies used for immunofluorescence were used at a 1:500 dilution and included: chicken anti-GFP (Aves Labs cat. no. GFP-1020, RRID: AB_1000024) and goat anti-chicken-Alexa Fluor 488 (Thermo Fisher Scientific cat. no. A-11039, RRID:AB_2534096).). DAPI was used at 1:5,000, according to the manufacturer’s recommendation (Thermo Fisher Scientific cat. no. D1306, RRID: AB_2629482).

For the LRRK2 filament analysis, experimenters were blinded to conditions for both the imaging acquisition and analysis. Cells were imaged using a Yokogawa W1 confocal scanhead mounted to a Nikon Ti2 microscope with an Apo ×60 1.49-NA objective. The microscope was run with NIS Elements using the 488 nm and 405 nm lines of a six-line (405 nm, 445 nm, 488 nm, 515 nm, 561 nm, and 640 nm) LUN-F-XL laser engine and a Prime95B camera (Photometrics).

ImageJ⁸³ (version 1.53, https://imagej.nih.gov/ij/, RRID: SCR_003070) was used to quantify the percentage of cells with LRRK2 filaments, as previously described. Maximum-intensity projections were generated from z-stack confocal images. Using the GFP immunofluorescence signal, transfected cells were identified. Cells were scored for the presence or absence of filaments using both the z-projection and z-stack micrographs as a guide. To calculate the percentage cells with filaments, the number of cells with filaments was divided by the total number of transfected cells per technical replicate (defined as one 24-well coverslip). Per coverslip, eight fields of view were imaged containing a total of 50 and 150 cells per replicate. The quantification of all cellular experiments come from compiled data collected on at least three separate days. All statistical analyses were performed in GraphPad Prism (version 9.2, https://www.graphpad.com/, RRID:SCR_002798).

Western blot analysis and antibodies

For western blot quantification of LRRK2 protein expression and Rab10 phosphorylation, cells were plated on 6-well dishes (200,000 cells per well) 24 hours before transfection. Cells were transfected with 500 ng of GFP-LRRK2 construct and 500 ng of GFP-Rab10 using polyethylenimine (PEI, Polysciences). After 36 hours, cells were rinsed with ice-cold 1× PBS, pH 7.4 and lysed on ice in RIPA buffer (50 nM Tris pH7.5, 150 mM NaCl, 0.2% Triton X-100, 0.1% SDS, with cOmplete protease inhibitor cocktail and PhoStop phosphatase inhibitor). Lysates were rotated for 15 minutes at 4 °C and clarified by centrifugation at maximum speed in a 4 °C microcentrifuge for 15 minutes. Supernatants were then boiled for 10 minutes in SDS buffer. Experiments were performed in duplicate or triplicate and repeated on at least 3 separate days.

Lysates were run on 4–12% polyacrylamide gels (NuPage, Invitrogen) for 50 minutes at 180 V and transferred to polyvinylidene difluoride (Immobilon-FL, EMD Millipore) for 4 hours at 200 mA constant current. Blots were rinsed briefly in MilliQ water and dried at room temperature for at least 30 minutes. Membranes were briefly reactivated with methanol and blocked for 1 hour at room temperature in 5% milk (wt/vol) in TBS. Antibodies were diluted in 1% milk in TBS with 0.1% Tween-20 (TBST). Primary antibodies used for immunoblots were as follows: mouse anti-GFP (1:2,500 dilution; Santa Cruz Biotechnology cat. no. sc-9996, RRID:AB_627695), rabbit anti-LRRK2 (1:5,000 dilution; Abcam cat. no. ab181386), rabbit anti-GAPDH (1:3,000 dilution; Cell Signaling Technology cat. no. 2118, RRID: AB_561053), recombinant rabbit anti-phospho-T72-RAB8 (1:1,000 dilution; Abcam cat. no. ab230260, Lot: GR3216587-1), and rabbit anti-phospho-T73-RAB10 (1:2,500 dilution; Abcam cat. no. ab230261, RRID: AB_2811274). Secondary antibodies (1:15000) used for western blots were IRDye goat anti-mouse 680RD (LI-COR Biosciences cat. no. 926-68070, RRID:AB_10956588) and IRDye goat anti-rabbit 800CW (LI-COR Biosciences cat. no. 926-32211, Lot: C90229-05, RRID: AB_10956166). Primary antibodies were incubated overnight at 4 °C, and secondary antibodies were incubated at room temperature for 1 hour. For quantification, blots were imaged on an Odyssey CLx controlled by Imaging Studio software (version 5.2, https://www.licor.com/bio/image-studio/, RRID: SCR_015795), and intensity of bands quantified using Image Studio Lite software (version 5.2, RRID: SCR_013715). All tabular data are available at https://doi.org/10.5281/zenodo.6463635.

Cell line

Human 293T cells were obtained from ATCC (ATCC cat. no. CRL-3216, RRID: CVCL_0063) and maintained at 37 °C with 5% CO₂ in Dulbecco’s Modified Eagle Medium (DMEM, Corning) supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin–streptomycin (Corning). Cells were routinely tested for mycoplasma contamination and were not authenticated after purchase.

Sequence alignment

Protein sequences of LRRK2 (Q5S007) and LRRK1 (Q38SD2) were obtained from UniProt. Sequence alignments were performed with Clustal Omega web services⁸⁵ (https://www.ebi.ac.uk/Tools/msa/clustalo/, RRID: SCR_001591) and annotated using Jalview⁸⁶ (version 2.11, http://www.jalview.org/, RRID: SCR_006459).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Cryo-EM maps and molecular models have been deposited in the EM Data Bank and Protein Data Bank, respectively. EMDB accession numbers are as follows: EMDB-25649: LRRK2^RCKW + MT + MLi-2 (helical); EMDB-25658: LRRK2^RCKW + MT + MLi-2 (tetramer + MT); EMDB-25664: LRRK2^RCKW + MT + MLi-2 (tetramer only); EMDB-25672: LRRK2^RCKW + MT + MLi-2 (plus end); EMDB-25674: LRRK2^RCKW + MT + MLi-2 (minus end); EMDB-25897: LRRK2^RCKW + MT + MLi-2 (focused on kinase); EMDB-25908: LRRK2^RCKW + MT + MLi-2 (microtubule only). PDB accession numbers are as follows: PDB-7THY: LRRK2^RCKW + MT + MLi-2 (minus end); PDB-7THZ: LRRK2^RCKW + MT + MLi-2 (plus end). Movie frames for the micrographs used in this study were deposited in EMPIAR; accession numbers are also listed in Table 1. All reagents and data will be made available upon request by the corresponding authors. Source data are provided with this paper.

Code availability

The code for the plug-in developed to automate the targeting to areas of the sample that contained LRRK2^RCKW filaments during data collection in Leginon is available at https://github.com/matyszm/filfinder (https://doi.org/10.5281/zenodo.5854954).

References

Funayama, M. et al. A new locus for Parkinson’s disease (PARK8) maps to chromosome 12p11.2-q13.1. Ann. Neurol. 51, 296–301 (2002).
Article CAS Google Scholar
Zimprich, A. et al. The PARK8 locus in autosomal dominant parkinsonism: confirmation of linkage and further delineation of the disease-containing interval. Am. J. Hum. Genet. 74, 11–19 (2004).
Article CAS Google Scholar
Zimprich, A. et al. Mutations in LRRK2 cause autosomal-dominant parkinsonism with pleomorphic pathology. Neuron 44, 601–607 (2004).
Article CAS Google Scholar
Paisán-Ruíz, C. et al. Cloning of the gene containing mutations that cause PARK8-linked Parkinson’s disease. Neuron 44, 595–600 (2004).
Article Google Scholar
Simón-Sánchez, J. et al. Genome-wide association study reveals genetic risk underlying Parkinson’s disease. Nat. Genet. 41, 1308–1312 (2009).
Article Google Scholar
Satake, W. et al. Genome-wide association study identifies common variants at four loci as genetic risk factors for Parkinson’s disease. Nat. Genet. 41, 1303–1307 (2009).
Article CAS Google Scholar
West, A. B. et al. Parkinson’s disease-associated mutations in leucine-rich repeat kinase 2 augment kinase activity. Proc. Natl Acad. Sci. USA 102, 16842–16847 (2005).
Article CAS Google Scholar
Gloeckner, C. J. et al. The Parkinson disease causing LRRK2 mutation I2020T is associated with increased kinase activity. 15, 223–232 (2005).
Steger, M. et al. Phosphoproteomics reveals that Parkinson’s disease kinase LRRK2 regulates a subset of Rab GTPases. eLife 5, 809 (2016).
Article Google Scholar
Sheng, Z. et al. Ser1292 autophosphorylation is an indicator of LRRK2 kinase activity and contributes to the cellular effects of PD mutations. Sci. Transl. Med. 4, 164ra161 (2012).
Article Google Scholar
Di Maio, R. et al. LRRK2 activation in idiopathic Parkinson’s disease. 10, eaar5429 (2018).
Bae, E.-J. & Lee, S.-J. The LRRK2–RAB axis in regulation of vesicle trafficking and α-synuclein propagation. Biochim. Biophys. Acta, Mol. Basis Dis. 1866, 165632 (2020).
Article CAS Google Scholar
Hur, E.-M., Jang, E.-H., Jeong, G. R. & Lee, B. D. LRRK2 and membrane trafficking: nexus of Parkinson’s disease. BMB Rep. 52, 533–539 (2019).
Article Google Scholar
Bonet-Ponce, L. & Cookson, M. R. LRRK2 recruitment, activity, and function in organelles. FEBS J. https://doi.org/10.1111/febs.16099 (2021).
Dodson, M. W., Zhang, T., Jiang, C., Chen, S. & Guo, M. Roles of the Drosophila LRRK2 homolog in Rab7-dependent lysosomal positioning. Hum. Mol. Genet 21, 1350–1363 (2012).
CAS Google Scholar
Manzoni, C. et al. Pathogenic Parkinson’s disease mutations across the functional domains of LRRK2 alter the autophagic/lysosomal response to starvation. Biochem. Biophys. Res. Commun. 441, 862–866 (2013).
Article CAS Google Scholar
Orenstein, S. J. et al. Interplay of LRRK2 with chaperone-mediated autophagy. Nat. Neurosci. 16, 394–406 (2013).
Article CAS Google Scholar
Godena, V. K. et al. Increasing microtubule acetylation rescues axonal transport and locomotor deficits caused by LRRK2 Roc-COR domain mutations. Nat. Commun. 5, 5245 (2014).
Hsieh, C.-H. et al. Functional impairment in miro degradation and mitophagy is a shared feature in familial and sporadic Parkinson’s disease. Cell Stem Cell 19, 709–724 (2016).
Article CAS Google Scholar
Eguchi, T. et al. LRRK2 and its substrate Rab GTPases are sequentially targeted onto stressed lysosomes and maintain their homeostasis. Proc. Natl Acad. Sci. USA 115, E9115–E9124 (2018).
Article CAS Google Scholar
Bonet-Ponce, L. et al. LRRK2 mediates tubulation and vesicle sorting from lysosomes. Sci. Adv. 6, eabb2454 (2020).
Article CAS Google Scholar
Henry, A. G. et al. Pathogenic LRRK2 mutations, through increased kinase activity, produce enlarged lysosomes with reduced degradative capacity and increase ATP13A2 expression. Hum. Mol. Genet. 24, 6013–6028 (2015).
Article CAS Google Scholar
Hockey, L. N. et al. Dysregulation of lysosomal morphology by pathogenic LRRK2 is corrected by TPC2 inhibition. J. Cell Sci. 128, 232–238 (2015).
CAS Google Scholar
Berger, Z., Smith, K. A. & LaVoie, M. J. Membrane localization of LRRK2 is associated with increased formation of the highly active LRRK2 dimer and changes in its phosphorylation. Biochemistry 49, 5511–5523 (2010).
Article CAS Google Scholar
Purlyte, E. et al. Rab29 activation of the Parkinson’s disease-associated LRRK2 kinase. EMBO J. 37, 1–18 (2018).
Article CAS Google Scholar
Gomez, R. C., Wawro, P., Lis, P., Alessi, D. R. & Pfeffer, S. R. Membrane association but not identity is required for LRRK2 activation and phosphorylation of Rab GTPases. J. Cell Biol. 218, 4157–4170 (2019).
Article CAS Google Scholar
Schapansky, J., Nardozzi, J. D., Felizia, F. & LaVoie, M. J. Membrane recruitment of endogenous LRRK2 precedes its potent regulation of autophagy. Hum. Mol. Genet 23, 4201–4214 (2014).
Article CAS Google Scholar
Pfeffer, S. R. Rab GTPases: master regulators that establish the secretory and endocytic pathways. Mol. Biol. Cell 28, 712–715 (2017).
Article CAS Google Scholar
Steger, M. et al. Systematic proteomic analysis of LRRK2-mediated Rab GTPase phosphorylation establishes a connection to ciliogenesis. eLife 6, e80705 (2017).
Article Google Scholar
Rivero-Ríos, P. et al. The G2019S variant of leucine-rich repeat kinase 2 (LRRK2) alters endolysosomal trafficking by impairing the function of the GTPase RAB8A. J. Biol. Chem. 94, 4738–4758 (2019).
Article Google Scholar
Rivero-Ríos, P. et al. Alterations in late endocytic trafficking related to the pathobiology of LRRK2-linked Parkinson’s disease. Biochem. Soc. Trans. 43, 390–395 (2015).
Article Google Scholar
Rivero-Ríos, P., Romo-Lozano, M., Fernández, B., Fdez, E. & Hilfiker, S. Distinct roles for RAB10 and RAB29 in pathogenic LRRK2-mediated endolysosomal trafficking lterations. Cells 9, E1719 (2020).
Article Google Scholar
Kett, L. R. et al. LRRK2 Parkinson disease mutations enhance its microtubule association. Hum. Mol. Genet. 21, 890–899 (2012).
Article CAS Google Scholar
Watanabe, R. et al. The in situ structure of Parkinson’s disease-linked LRRK2. Cell 182, 1508–1518 (2020).
Article CAS Google Scholar
Deniston, C. K. et al. Structure of LRRK2 in Parkinson’s disease and model for microtubule interaction. Nature 588, 344–349 (2020).
Article CAS Google Scholar
Leschziner, A. E. & Reck-Peterson, S. L. Structural Biology of LRRK2 and its interaction with microtubules. Mov. Disord. 36, 2494–2504 (2021).
Article CAS Google Scholar
Blanca Ramírez, M. et al. GTP binding regulates cellular localization of Parkinson’s disease-associated LRRK2. Hum. Mol. Genet. 26, 2747–2767 (2017).
Article Google Scholar
Schmidt, S. H. et al. The dynamic switch mechanism that leads to activation of LRRK2 is embedded in the DFGψ motif in the kinase domain. Proc. Natl Acad. Sci. USA 116, 14979–14988 (2019).
Myasnikov, A. et al. Structural analysis of the full-length human LRRK2. Cell 184, 3519–3527 (2021).
Article CAS Google Scholar
Gandhi, P. N., Wang, X., Zhu, X., Chen, S. G. & Wilson-Delfosse, A. L. The Roc domain of leucine-rich repeat kinase 2 is sufficient for interaction with microtubules. J. Neurosci. Res. 86, 1711–1720 (2008).
Article CAS Google Scholar
Wang, Y. et al. CRACR2a is a calcium-activated dynein adaptor protein that regulates endocytic traffic. J. Cell Biol. 218, 1619–1633 (2019).
Etoh, K. & Fukuda, M. Rab10 regulates tubular endosome formation through KIF13A and KIF13B motors. J. Cell Sci. 132, jcs226977 (2019).
Article CAS Google Scholar
Niwa, S., Tanaka, Y. & Hirokawa, N. KIF1Bβ- and KIF1A-mediated axonal transport of presynaptic regulator Rab3 occurs in a GTP-dependent manner through DENN/MADD. Nat. Cell Biol. 10, 1269–1279 (2008).
Article CAS Google Scholar
Christensen, J. R. et al. Cytoplasmic dynein-1 cargo diversity is mediated by the combinatorial assembly of FTS-Hook-FHIP complexes. Elife 10, e74538 (2021).
Article CAS Google Scholar
Scott, J. D. et al. Discovery of a 3-(4-pyrimidinyl) indazole (MLi-2), an orally available and selective leucine-rich repeat kinase 2 (LRRK2) inhibitor that reduces brain kinase activity. J. Med. Chem. 60, 2983–2992 (2017).
Article CAS Google Scholar
Fell, M. J. et al. MLi-2, a potent, selective, and centrally active compound for exploring the therapeutic potential and safety of LRRK2 kinase inhibition. J. Pharmacol. Exp. Ther. 355, 397–409 (2015).
Article CAS Google Scholar
Tunyasuvunakool, K. et al. Highly accurate protein structure prediction for the human proteome. Nature 596, 590–596 (2021).
Article CAS Google Scholar
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
Article CAS Google Scholar
Zhang, P. et al. Crystal structure of the WD40 domain dimer of LRRK2. Proc. Natl Acad. Sci. USA 116, 1579–1584 (2019).
Article CAS Google Scholar
Janke, C. & Magiera, M. M. The tubulin code and its role in controlling microtubule properties and functions. Nat. Rev. Mol. Cell Biol. 21, 307–326 (2020).
Article CAS Google Scholar
Serrano, L., Avila, J. & Maccioni, R. B. Controlled proteolysis of tubulin by subtilisin: localization of the site for MAP2 interaction. Biochemistry 23, 4675–4681 (1984).
Article CAS Google Scholar
Civiero, L. & Bubacco, L. Human leucine-rich repeat kinase 1 and 2: intersecting or unrelated functions? Biochem. Soc. Trans. 40, 1095–1101 (2012).
Article CAS Google Scholar
Howaldt, A. et al. Adult osteosclerotic metaphyseal dysplasia with progressive osteonecrosis of the jaws and abnormal bone resorption pattern due to a LRRK1 splice site mutation. J. Bone Miner. Res. 35, 1322–1332 (2020).
Article CAS Google Scholar
Miryounesi, M. et al. A novel homozygous LRRK1 stop gain mutation in a patient suspected with osteosclerotic metaphyseal dysplasia. Ann. Hum. Genet. 84, 102–106 (2020).
Article CAS Google Scholar
Guo, L. et al. Identification of a novel LRRK1 mutation in a family with osteosclerotic metaphyseal dysplasia. J. Hum. Genet 62, 437–441 (2017).
Article CAS Google Scholar
Iida, A. et al. Identification of biallelic LRRK1 mutations in osteosclerotic metaphyseal dysplasia and evidence for locus heterogeneity. J. Med. Genet. 53, 568–574 (2016).
Article CAS Google Scholar
Masuzugawa, S. et al. A novel rare variant of LRRK2 associated with familial Parkinson’s disease: p.R1501W. Parkinsonism Relat. Disord. 76, 46–48 (2020).
Article Google Scholar
Ryan, K. J. et al. A human microglia-like cellular model for assessing the effects of neurodegenerative disease gene variants. Sci. Transl. Med. 9, eaai7635 (2017).
Article Google Scholar
Atashrazm, F. et al. LRRK2-mediated Rab10 phosphorylation in immune cells from Parkinson’s disease patients. Mov. Disord. 34, 406–415 (2019).
Article CAS Google Scholar
Cook, D. A. et al. LRRK2 levels in immune cells are increased in Parkinson’s disease. NPJ Parkinsons Dis. 3, 1–12 (2017).
Google Scholar
Reck-Peterson, S. L., Redwine, W. B., Vale, R. D. & Carter, A. P. The cytoplasmic dynein transport machinery and its many cargoes. Nat. Rev. Mol. Cell Biol. 19, 1–17 (2018).
Google Scholar
Bonifacino, J. S. & Neefjes, J. Moving and positioning the endolysosomal system. Curr. Opin. Cell Biol. 47, 1–8 (2017).
Article CAS Google Scholar
Kruppa, A. J. & Buss, F. Motor proteins at the mitochondria-cytoskeleton interface. J. Cell Sci. 134, jcs226084 (2021).
Article CAS Google Scholar
Boecker, C. A., Goldsmith, J., Dou, D., Cajka, G. G. & Holzbaur, E. L. F. Increased LRRK2 kinase activity alters neuronal autophagy by disrupting the axonal transport of autophagosomes. Curr. Biol. 31, 2140–2154 (2021).
Article CAS Google Scholar
Boecker, C. A. & Holzbaur, E. L. F. Hyperactive LRRK2 kinase impairs the trafficking of axonal autophagosomes. Autophagy 17, 2043–2045 (2021).
Article CAS Google Scholar
Rzomp, K. A., Scholtes, L. D., Briggs, B. J., Whittaker, G. R. & Scidmore, M. A. Rab GTPases are recruited to chlamydial inclusions in both a species-dependent and species-independent manner. Infect. Immun. 71, 5855–5870 (2003).
Article CAS Google Scholar
Woehlke, G. et al. Microtubule interaction site of the kinesin motor. Cell 90, 207–216 (1997).
Article CAS Google Scholar
Nicholas, M. P., Rao, L. & Gennerich, A. An improved optical tweezers assay for measuring the force generation of single kinesin molecules. Methods Mol. Biol. 1136, 171–246 (2014).
Article CAS Google Scholar
Redwine, W. B. et al. The human cytoplasmic dynein interactome reveals novel activators of motility. eLife 6, 379 (2017).
Article Google Scholar
Suloway, C. et al. Automated molecular microscopy: the new Leginon system. J. Struct. Biol. 151, 41–60 (2005).
Article CAS Google Scholar
Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy. Nat. Methods 14, 331–332 (2017).
Article CAS Google Scholar
Rohou, A. & Grigorieff, N. CTFFIND4: fast and accurate defocus estimation from electron micrographs. J. Struct. Biol. 192, 216–221 (2015).
Article Google Scholar
Zivanov, J. et al. New tools for automated high-resolution cryo-EM structure determination in RELION-3. eLife 7, 163 (2018).
Article Google Scholar
Cook, A. D., Manka, S. W., Wang, S., Moores, C. A. & Atherton, J. A microtubule RELION-based pipeline for cryo-EM image processing. J. Struct. Biol. 209, 107402 (2020).
Article CAS Google Scholar
Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017).
Article CAS Google Scholar
Punjani, A. & Fleet, D. J. 3D variability analysis: resolving continuous flexibility and discrete heterogeneity from single particle cryo-EM. J. Struct. Biol. 213, 107702 (2021).
Article CAS Google Scholar
Wang, R. Y.-R. et al. Automated structure refinement of macromolecular assemblies from cryo-EM maps using Rosetta. eLife 5, 352 (2016).
Article Google Scholar
Croll, T. I. ISOLDE: a physically realistic environment for model building into low-resolution electron-density maps. Acta Crystallogr D. Struct. Biol. 74, 519–530 (2018).
Article CAS Google Scholar
Wagner, T. et al. SPHIRE-crYOLO is a fast and accurate fully automated particle picker for cryo-EM. Commun. Biol. 2, 218 (2019).
Article Google Scholar
Waterhouse, A. et al. SWISS-MODEL: homology modelling of protein structures and complexes. Nucleic Acids Res. 46, W296–W303 (2018).
Article CAS Google Scholar
Pettersen, E. F. et al. UCSF ChimeraX: Structure visualization for researchers, educators, and developers. Protein Sci. 30, 70–82 (2021).
Article CAS Google Scholar
Roberts, A. J., Goodman, B. S. & Reck-Peterson, S. L. Reconstitution of dynein transport to the microtubule plus end by kinesin. eLife 3, e02641 (2014).
Article Google Scholar
Schneider, C. A., Rasband, W. S. & Eliceiri, K. W. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods 9, 671–675 (2012).
Article CAS Google Scholar
Lis, P. et al. Development of phospho-specific Rab protein antibodies to monitor in vivo activity of the LRRK2 Parkinson’s disease kinase. Biochem. J. 475, 1–22 (2018).
Article CAS Google Scholar
McWilliam, H. et al. Analysis Tool Web Services from the EMBL-EBI. Nucleic Acids Res. 41, W597–W600 (2013).
Article Google Scholar
Waterhouse, A. M., Procter, J. B., Martin, D. M. A., Clamp, M. & Barton, G. J. Jalview Version 2—a multiple sequence alignment editor and analysis workbench. Bioinformatics 25, 1189–1191 (2009).
Article CAS Google Scholar

Download references

Acknowledgements

This research was funded in part by Aligning Science Across Parkinson’s (grant number ASAP-000519) through the Michael J. Fox Foundation for Parkinson’s Research (MJFF) (A. E. L. and S. L. R.-P.). The work was also funded by the Howard Hughes Medical Institute (where S. L. R.-P. is an investigator); the Michael J. Fox Foundation (grant number 18321 to A. E. L. and S. L. R.-P.); the A.P. Giannini Foundation (postdoctoral fellowship to D. M. S.); the Molecular Biophysics Training Grant at UC San Diego (NIH Grant T32 GM008326 supported A. M. D.); and the National Institutes of Health (R35GM141825 to S. L. R.-P. and R01GM107214 to A. E. L). We also thank the UC San Diego Cryo-EM Facility, the Nikon Imaging Center at UC San Diego and Eric Griffis, and the UC San Diego Physics Computing Facility for IT support. For the purpose of open access, the authors have applied a CC BY 4.0 public copyright license to all author accepted manuscripts arising from this submission.

Author information

These authors contributed equally: David M. Snead, Mariusz Matyszewski, Andrea M. Dickey.
These authors jointly supervised this work: Andres E. Leschziner, Samara L. Reck-Peterson.

Authors and Affiliations

Department of Cellular and Molecular Medicine, University of California, San Diego, La Jolla, CA, USA
David M. Snead, Mariusz Matyszewski, Andrea M. Dickey, Yu Xuan Lin, Andres E. Leschziner & Samara L. Reck-Peterson
Aligning Science Across Parkinson’s (ASAP) Collaborative Research Network, Chevy Chase, Maryland, MD, USA
David M. Snead, Mariusz Matyszewski, Andrea M. Dickey, Yu Xuan Lin, Andres E. Leschziner & Samara L. Reck-Peterson
Department of Biochemistry and Molecular Biology, Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD, USA
David M. Snead
School of Biological Sciences, University of California, San Diego, La Jolla, CA, USA
Mariusz Matyszewski
Department of Molecular Biology, School of Biological Sciences, University of California, San Diego, La Jolla, CA, USA
Andres E. Leschziner
Department of Cell and Developmental Biology, School of Biological Sciences, University of California, San Diego, La Jolla, CA, USA
Samara L. Reck-Peterson
Howard Hughes Medical Institute, Chevy Chase, Maryland, MD, USA
Samara L. Reck-Peterson

Authors

David M. Snead
View author publications
You can also search for this author in PubMed Google Scholar
Mariusz Matyszewski
View author publications
You can also search for this author in PubMed Google Scholar
Andrea M. Dickey
View author publications
You can also search for this author in PubMed Google Scholar
Yu Xuan Lin
View author publications
You can also search for this author in PubMed Google Scholar
Andres E. Leschziner
View author publications
You can also search for this author in PubMed Google Scholar
Samara L. Reck-Peterson
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

M. M. collected and processed the LRRK2 cryo-EM data. D. M. S. and M. M. collected and processed the LRRK1 cryo-EM data. Y. X. L., D. M. S., A. M. D. and M. M. purified proteins for cryo-EM and biochemistry experiments. D. M. S. and A. M. D. performed the biochemical and single-molecule assays with the help of Y. X. L. A. M. D. performed the cellular assays. S. L. R.-P. and A. E. L. directed and supervised the research. D. M. S., M. M., A. M. D., A. E. L., and S. L. R.-P. wrote and edited the manuscript.

Corresponding authors

Correspondence to Andres E. Leschziner or Samara L. Reck-Peterson.

Ethics declarations

Competing interests

The authors have no competing interests to declare.

Peer review

Peer review information

Nature Structural & Molecular Biology thanks Dario Alessi and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Editor recognition statement (if applicable to your journal): Primary Handling editor: Florian Ullrich, in collaboration with the Nature Structural & Molecular Biology team. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Cryo-EM structure determination of microtubule-associated filaments of LRRK2^RCKW[I2020T] in the presence of MLi-2.

a, Optimization of in vitro reconstituted microtubule-associated LRRK2^RCKW filaments. Top row, cryo-EM images of an individual microtubule (left) or individual microtubule-associated LRRK2^RCKW filaments. Middle, Diffraction patterns calculated from the images above. Arrowheads point to layer lines arising from the microtubule (white) or from the LRRK2^RCKW filaments (open), and their frequencies are indicated below the images. Bottom, 2D class averages from multiple images equivalent to those shown at the top. The type of LRRK2^RCKW (WT, G2019S, or I2020T) and the presence or absence of MLi-2 during filament reconstitution are indicated at the top. b, Schematic of data processing pipeline used to obtain the different reconstructions of the microtubule-associated LRRK2^RCKW[I2020T] filaments in the presence of MLi-2 (see Methods for details). Local resolution maps, Fourier Shell Correlation, directional FSC plots, and the distribution of voxel resolutions are shown for all reconstructions discussed in the text.

Extended Data Fig. 2 Cryo-EM structure determination of microtubule-associated filaments of LRRK2^RCKW[I2020T] in the absence of MLi-2.

Schematic of data processing pipeline used to obtain the reconstruction of microtubule-associated LRRK2^RCKW[I2020T] filaments in the absence of MLi-2 (see Methods for details). Local resolution maps, Fourier Shell Correlation, directional FSC plots, and the distribution of voxel resolutions are shown for all reconstructions discussed in the text.

Extended Data Fig. 3 Structural analysis of microtubule-associated filaments of LRRK2^RCKW[I2020T].

a, A model for LRRK2^RCKW[I2020T] with a closed kinase, obtained by docking the individual domains into the cryo-EM map of LRRK2^RCKW[I2020T] filaments obtained in the presence of MLi-2, was docked into a cryo-EM map (7 Å) of filaments obtained in the absence of the inhibitor (Extended Data Fig. 2). b, A model for LRRK2^RCKW with an open kinase (PDB:6VNO) was docked into the same map. The colored arrows in (a) and (b) highlight structural elements in the model that protrude from the density when the kinase is in an open conformation. c, The LRRK2^RCKW domains (ROC, COR-A, COR-B, Kinase N-lobe, Kinase C-lobe, WD40) (PDB:6VNO) were fitted individually into one of the monomers in the cryo-EM map of microtubule-bound LRRK2^RCKW[I2020T] formed in the presence of MLi-2. d, The LRRK2^RCKW portion of the AlphaFold model of LRRK2 was aligned to the COR-B domain in (c) and is shown here inside the same cryo-EM map. The colored arrows highlight regions where part of the model protrudes from the density. (Note: there is no arrow pointing to the loop in the ROC domain as this loop was not seen or modeled in the microtubule-bound structure). e, The kinase from the AlphaFold model of LRRK2 was fitted into the cryo-EM map (same as in (d)) and is shown here superimposed on the N- and C-lobes of the kinase as fitted in (c). Note that while the C-lobes superimpose well, the N-lobe fitted individually in (c) is more closed than that modeled in the AlphaFold LRRK2. f, The N- and C-lobes of the kinase from the AlphaFold LRRK2 model were now fitted individually into the cryo-EM map (as in (c)) and are shown superimposed on the N- and C-lobes of LRRK2^RCKW from (a). The blue arrow between panels (e) and (f) highlights the downward movement of the N-lobe of AlphaFold’s LRRK2 when the two lobes are fitted individually into the cryo-EM map. g, The LRRK2^RCKW domains (ROC, COR-A, COR-B, Kinase N-lobe, Kinase C-lobe, WD40) (PDB:6VNO) were fitted individually into the central dimer of the cryo-EM map of a tetramer of microtubule-bound LRRK2^RCKW[I2020T] obtained in the presence of MLi-2. h, i, Different closeup views of the map in (g), showing either (h) the ROC, COR-A, COR-B and kinase N-lobe from the LRRK2^RCKW model (PDB:6VNO), or (i) the corresponding portion from the structure of full-length LRRK2 (PDB:7LHT) docked as a single body into the cryo-EM map. The colored arrows highlight parts of the model that fit the cryo-EM density better when the domains are fitted in individually (h) rather than as a rigid body (i). j, Superposition of the model used in (i) and the COR-B domain from (h) to show that the differences among the ROC, COR-A and N-lobe of the kinase between the two models ((h) and (i)) is not due to major differences at the COR-B:COR-B interface, which is similar.

Extended Data Fig. 4 Mechanism of LRRK2^RCKW binding to microtubules.

a, Representative images of 293 T cells expressing GFP-LRRK2 (wild-type or the WD40 mutants L23443D and S2345D) and treated with DMSO or 500 nM MLi-2 for 2 hours. Scale bar is 10 µm. b, Representative images of 293 T cells expressing GFP-LRRK2 (wild-type or the COR-B mutants R1731L and R1731D) and treated with DMSO or 500 nM MLi-2 for 2 hours. Scale bar is 10 µm. c,d, Rab10 phosphorylation in 293 T cells overexpressing WT LRRK2 or LRRK2 carrying mutations in the WD40 (c) or COR-B (d) domains. LRRK2[I2020T], which is known to increase Rab10 phosphorylation in cells, was tested as well. 293 T cells were transiently co-transfected with the indicated plasmids encoding for GFP-LRRK2 (wild type or mutant) and GFP-Rab10. Thirty-six hours post-transfection the cells were lysed, immunoblotted for phospho-Rab10 (pT73), total GFP-Rab10, and total GFP-LRRK2. Quantification of data in (c) and (d) is shown in Fig. 2d, f, respectively. e, A representative gel of the supernatants from the microtubule pelleting assays used to generate the data shown in Fig. 2g. f, Cryo-EM analysis of filament formation in vitro by LRRK2^RCKW mutants. Top row, cryo-EM images of an individual microtubule (left) or combinations of microtubules and LRRK2^RCKW mutants. Middle, Diffraction patterns calculated from the images above. Arrowheads point to layer lines arising from the microtubule (white) or from the LRRK2^RCKW filaments (open), and their frequencies are indicated below the images. Bottom, 2D class averages from multiple images equivalent to those shown at the top. g, Example kymographs of single-molecule kinesin motility assays in the presence or absence of 50 nM LRRK2^RCKW wild-type or indicated mutant. h, Comparison of 2D class averages from cryo-EM images of different LRRK2^RCKW mutants with the corresponding 2D projection from a LRRK2^RCKW molecular model (PDB: 6VNO). Two different views are shown for each mutant. i, Representative in vitro kinase reaction. Rab8a phosphorylation was measured via western blotting with a phospho-T72-specific Rab8a antibody, and total LRRK2^RCKW concentration was measured by Sypro Red staining. Phosphorylation reactions were terminated after 30 minutes. j, Quantification of data shown in (i). For each reaction, phospho-Rab8a band intensity (chemiluminescence) was divided by LRRK2^RCKW band intensity (Sypro red); for each western blot, an average normalized value was calculated for all replicates of unlabeled LRRK2^RCKW, and all data was then normalized to this value. Data are mean +/− s.d., n = 5 replicates, p=ns, two-tailed unpaired t-test. k, Cumulative distribution of run lengths for kinesin in the absence or presence of 25 nM TMR-LRRK2^RCKW. The run lengths were significantly different between 0 nM and 25 nM TMR-LRRK2^RCKW (Mann-Whitney test). Mean decay constants (tau) are shown. The effect on kinesin motility is similar to that previously shown using unlabeled LRRK2^RCKW. l, m, Representative microtubule pelleting assay gel for LRRK2^RCKW in the presence of 100 mM and 150 mM sodium chloride. Co-sedimentation was measured as depletion from the supernatant. For each reaction, 200 nM LRRK2^RCKW was mixed with a given concentration of microtubules, microtubules were pelleted by a high-speed spin, and a gel sample was taken of the supernatant. Quantification of data represented in (l) is shown in (m). Data are mean ± s.d., n = 4. The solid line represents a hyperbolic curve fit to the data. n, Representative images of coverslip-tethered Alexa Fluor 488-labeled MTs (cyan) bound to 100 nM TMR-LRRK2^RCKW (magenta) in the presence of increasing concentrations of sodium chloride, used to generate the data in Fig. 3d. o, Subtilisin treatment of taxol-stabilized microtubules. Serial dilutions of taxol-stabilized microtubules were treated with subtilisin (left) or left untreated as a control (right). The experiment was performed twice with similar results. p, Representative images of untreated (top) and subtilisin-treated (bottom) Alexa Fluor 488-labeled MTs (cyan) bound to 50 nM TMR-LRRK2^RCKW (magenta), used to generate the data shown in Fig. 3e.

Source data

Extended Data Fig. 5 Basic residues within the LRRK2 ROC domain are not conserved in LRRK1 and are involved in LRRK2’s binding to microtubules.

a, Cryo-EM structure determination of LRRK1^RCKW. Local resolution map, Fourier Shell Correlation, directional FSC plot, and the distribution of voxel resolutions are shown. b, Representative gel of supernatant from microtubule pelleting assay for 200 nM LRRK2^RCKW or LRRK1^RCKW with increasing tubulin concentrations. We performed 4 technical replicates. c, Sequence alignment of the ROC domains of LRRK2 and LRRK1 across several species made using Clustal Omega. Putative microtubule-contacting residues conserved in LRRK2 but not in LRRK1 are boxed. d, Close ups of the basic patches tested in this study, shown in the context of the cryo-EM maps for the ‘+’ and ‘−’ LRRK2^RCKW monomers in our reconstruction of the microtubule-associated filaments (Fig. 1e, f). The models shown here correspond to those in Fig. 5c. e, Example kymographs of kinesin motility in the presence of 100 nM LRRK2^RCKW (wild-type or K1358A/K1359A mutant). f, Representative images of 293 T cells expressing GFP-LRRK2 (wild type or ROC mutants) and treated with DMSO or 500 nM MLi-2 for 2 hours, corresponding to data plotted in Fig. 5g. Scale bar is 10 µm. g, Representative images of 293 T cells expressing GFP-LRRK2 (I2020T, I2020T/ROC mutant, or I2020T/R1501W), corresponding to data plotted in Fig. 5h, j. Scale bar is 10 µm. h, Representative images of 293 T cells expressing GFP-LRRK2 (wild type or R1501W mutant) and treated with DMSO or 500 nM MLi-2 for 2 hours, corresponding to data plotted in Fig. 5i. Scale bar is 10 µm. I, j, Rab10 phosphorylation in 293 T cells overexpressing WT LRRK2 or LRRK2 carrying the indicated mutations in the ROC domain. LRRK2[I2020T], which is known to increase Rab10 phosphorylation in cells, was tested as well. 293 T cells were transiently transfected with the indicated plasmids encoding for GFP-LRRK2 (wild-type or mutant) and GFP-Rab10. Thirty-six hours post-transfection the cells were lysed, immunoblotted for phospho-Rab10 (pT73), total GFP-Rab10, and total GFP-LRRK2. Quantification of immunoblotting data (i) is shown in (j) as ratios of pRab10/total GFP-Rab10 normalized to the average of all wildtype values. Individual data points represent separate populations of cells obtained across at least three independent experiments. Data are mean +/− s.e.m. ****p < 0.0001, ***p = 0.0004, **p = 0.0052, ***p = 0.0344, one-way ANOVA followed by a Fisher’s LSD test.

Source data

Extended Data Fig. 6 Modeling of full-length LRRK2 into the cryo-ET reconstruction of microtubule-associated LRRK2[I2020T] filaments in cells.

a, Cryo-ET reconstruction of microtubule-associated LRRK2[I2020T] filaments in cells¹. The LRRK2 strands that form the double-helical filaments are shown in light and dark orange. For this figure, the density corresponding to the microtubule was replaced with a 10 Å representation of a molecular model of a microtubule. b, We docked copies of our 5.9 Å reconstruction of a LRRK2^RCKW[I2020T] tetramer from the microtubule-associated filaments (Fig. 1b, c) into the regions indicated by the parallelograms in (a). c, Next, we docked two copies of the AlphaFold model of full-length LRRK2 (AF-Q5S007), which is in the active state, as is the case with LRRK2^RCKW[I2020T] in our filaments, into each of the 5.9 Å maps. The pairs of AlphaFold models in each map correspond to the COR-B:COR-B dimer. This panel shows a region corresponding to the rectangle in (b). d, Three different views of the models docked in (c). Below each model, close-ups show regions where adjacent filaments clash. These clashes involve a domain in the N-terminal repeats of one LRRK2, and either the same domain on another LRRK2, or the WD40 domain. For clarity, one of the LRRK2’s is shown in grey instead of the standard rainbow coloring.

Supplementary information

Supplementary Information

Supplementary Table 1 and Supplementary Material Reference

Reporting Summary

Peer Review File

Supplementary Video

Supplementary Video 1 Cryo-EM reconstruction of LRRK2^RCKW-I2020T bound to a microtubule in the presence of MLi-2. The video introduces cryo-EM structures obtained from in vitro reconstituted LRRK2^RCKW-I2020T filaments bound to microtubules in the presence of MLi-2: the initial helical reconstruction of the filaments; the structure of a LRRK2^RCKW tetramer bound to two microtubule protofilaments (6.6 Å); and the separate refinements of LRRK2^RCKW monomers facing either the minus (–) (5.2 Å) or plus (+) (5.0 Å) ends of the microtubule. The movie ends by showing a molecular model for a LRRK2^RCKW dimer bound to two microtubule protofilaments.

Source data

Source Data Fig. 2

Statistical Source Data

Source Data Fig. 3

Statistical Source Data

Source Data Fig. 4

Statistical Source Data

Source Data Fig. 5

Statistical Source Data

Source Data Extended Data Fig. 4

Unprocessed westernblots and gels

Source Data Extended Data Fig. 4

Statistical Source Data

Source Data Extended Data Fig. 5

Unprocessed Western Blots and gels

Source Data Extended Data Fig. 5

Statistical Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Snead, D.M., Matyszewski, M., Dickey, A.M. et al. Structural basis for Parkinson’s disease-linked LRRK2’s binding to microtubules. Nat Struct Mol Biol 29, 1196–1207 (2022). https://doi.org/10.1038/s41594-022-00863-y

Download citation

Received: 13 January 2022
Accepted: 10 October 2022
Published: 12 December 2022
Issue Date: December 2022
DOI: https://doi.org/10.1038/s41594-022-00863-y

This article is cited by

Structure and regulation of full-length human leucine-rich repeat kinase 1
- Riley D. Metcalfe
- Juliana A. Martinez Fiesco
- Ping Zhang
Nature Communications (2023)
Structure of LRRK1 and mechanisms of autoinhibition and activation
- Janice M. Reimer
- Andrea M. Dickey
- Andres E. Leschziner
Nature Structural & Molecular Biology (2023)

Subjects

Abstract

Similar content being viewed by others

Main

Results

Cryo-EM structure of microtubule-associated LRRK2RCKW

LRRK2’s dimer interfaces are important for microtubule association

Electrostatic interactions drive binding of LRRK2RCKW to microtubules

LRRK1RCKW adopts a similar overall fold to LRRK2RCKW

LRRK1RCKW does not bind microtubules

Basic residues in LRRK2’s ROC domain are important for microtubule binding

Discussion

Methods

Cloning, plasmid construction, and mutagenesis

LRRK2RCKW and LRRK1RCKW expression and purification

Purification of molecular motors

Rab8a expression and purification

Cryo-electron microscopy: sample preparation and imaging of filaments

Cryo-electron microscopy: reconstruction of LRRK2RCKW bound to a microtubule

Model creation and refinement: LRRK2 ROC domain interacting with the microtubule

Cryo-electron microscopy: sample preparation and imaging of LRRK1RCKW

Cryo-electron microscopy: reconstruction of LRRK1RCKW

Single-molecule microscopy and motility assays

Single-molecule motility assay analysis

Microtubule sedimentation binding assay

Microtubule cleavage by subtilisin

TMR labeling

Widefield fluorescence microtubule binding assay

In vitro Rab8a phosphorylation

Immunofluorescence, confocal microscopy, and image analysis

Western blot analysis and antibodies

Cell line

Sequence alignment

Reporting summary

Data availability

Code availability

References

Acknowledgements

Author information

Authors and Affiliations

Contributions

Corresponding authors

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Extended data

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

This article is cited by

Search

Quick links

Cryo-EM structure of microtubule-associated LRRK2^RCKW

Electrostatic interactions drive binding of LRRK2^RCKW to microtubules

LRRK1^RCKW adopts a similar overall fold to LRRK2^RCKW

LRRK1^RCKW does not bind microtubules

LRRK2^RCKW and LRRK1^RCKW expression and purification

Cryo-electron microscopy: reconstruction of LRRK2^RCKW bound to a microtubule

Cryo-electron microscopy: sample preparation and imaging of LRRK1^RCKW

Cryo-electron microscopy: reconstruction of LRRK1^RCKW