Biophysical analysis reveals autophosphorylation as an important negative regulator of LRRK2 dimerization

Leucine-rich repeat kinase 2 (LRRK2) is a large, multi-domain protein which is associated with Parkinson’s disease. Although high-resolution structures of LRRK2 are available, little is known about the complex dynamics behind the inter-domain regulation of LRRK2 and its perturbation by pathogenic variants. Previous studies have demonstrated that LRRK2 goes through an oligomerization cycle at the membrane, however it remains unclear in which form it exerts its kinase activity. Moreover, the LRRK2 monomer-dimer equilibrium and associated functional implications at a molecular level also need further investigation. In the present work, we used a multi-faceted approach to better understand LRRK2 oligomerization and suggest a functional model of how LRRK2 interacts with its substrates. To this end, we combined nano differential scanning calorimetry and mass photometry with molecular modelling. The thermal analysis resulted in a multistep denaturation profile, elucidating novel insights into the composite structural organization of the multi-domain protein LRRK2. Furthermore, LRRK2 shows a remarkable thermal stability, confirming its oligomeric nature. By using mass photometry, we could observe a monomer-dimer equilibrium which is altered by R1441G, a pathogenic variant within the Roc-COR interface. Most importantly, we could demonstrate that autophosphorylation induces LRRK2 monomerization, indicating a novel intramolecular feedback mechanism. Finally, we investigated the interaction of LRRK2 with its substrate, RAB10 by integrative computational modelling. The resulting models suggest that the monomeric form of LRRK2 is the favored protein conformation for the interaction with its substrate, leading to an increasing interest in the monomer-dimer equilibrium as a possible intervention point for the pathology.


Introduction
Parkinson´s disease (PD) is the second most common age-related neurodegenerative disease and is clinically characterized by movement impairments, bradykinesia, rigidity and resting tremors (1).
Although most forms of the disease are sporadic (idiopathic PD), familial forms of PD exist (2). Missense variants in the Leucine-Rich Repeat Kinase 2 (LRRK2) gene have been identified to cause familial forms of PD but risk variants within its genomic locus (PARK8) also play a role in the etiology of iPD (3,4).
Furthermore, a recent study suggests that, independent of mutations, increased LRRK2 activity plays a role in iPD (5). LRRK2 encodes a large ubiquitously expressed multi-domain protein of 286-kDa and 2527 amino acids that exerts both GTPase and kinase activities (6)(7)(8). LRRK2 is involved in multiple cellular processes including protein scaffolding, regulation of cytoskeletal dynamics and vesicle sorting but has also been identified as a regulator in innate immunity (9)(10)(11)(12).
Various PD-associated mutations in LRRK2 have been shown to augment kinase activity leading to an increased auto-phosphorylation as well as phosphorylation of Rab proteins, the recently identified physiological LRRK2 substrates (7,8,(13)(14)(15). Furthermore, pathogenic LRRK2 mutations within the Roc domain have been associated with decreased LRRK2 GTPase activity.
The protein belongs to the Roco-protein family of G-proteins. Members of this family have in common a Ras of complex proteins (Roc) G-domain and an adjacent conserved C-terminal of Roc (COR) acting as dimerization domain (16). Besides its enzymatic core, LRRK2 contains four predicted solenoid domains, commonly involved in protein-protein interactions. These domains include the N-terminal Ankyrin, Armadillo, and namesake Leucine-Rich Repeat (LRR) domains, along with a C-terminal WD40 domain (17).
Membrane localization plays an important role in the oligomerization and subsequent activation of LRRK2 (18) where it phosphorylates a specific subset of proteins (19). In addition, LRRK2 seems to be predominantly monomeric in the cytosol while forming oligomers at the membrane (20). Structural models as well as high-resolution structures, which became available recently, suggest that the LRRK2 N-terminus as well as the C-terminal WD40 domain may also be involved in auto-regulatory processes, likely keeping LRRK2 in an auto-inhibited state and, besides the Roc-COR module, may also contribute to the LRRK2 oligomerization (21)(22)(23).
Available high-resolution structures, together with biochemical studies already gave valuable insight into the local interaction of different domains and helped to identify first regulatory mechanisms of the LRRK2 kinase activity (22,24,25). Activity-dependent local transitions within the LRR-Roc linker region have also been recently identified by molecular dynamics (MD) simulations (26). However, the dynamic process of LRRK2 regulation by conformational and oligomerization changes remains largely unknown. For this reason, we have performed differential scanning calorimetry (nDSC) and mass photometry (MP) experiments on purified human full-length LRRK2, particularly looking at different conformational states defined by GDP-bound and GTPyS-bound LRRK2 protein, phosphorylation state and linking the information obtained with the pathological mutants.
In the present work, calorimetric investigation of the full-length protein gave valuable insights of the conformational stability of the apo-protein and the phospho-state dependent conformational modifications of LRRK2. Furthermore, MP measurements allowed the determination of the kinetics of the LRRK2 monomer-dimer and revealed functional roles of auto-phosphorylation in the oligomerization process. Finally, Crosslinking mass spectrometry (MS-XL)-driven integrative models of the LRRK2:Rab10 complex suggest that the monomer is the favorite state allowing kinase-substrate interactions. Unbiased AF-multimer predictions converge on predicting similar LRRK2-KIN-RAB10 interfaces, which are consistent with several of the XL as well as other biochemical evidences (e.g. RAB10 T73 phospho-site).

Biophysical investigation of the LRRK2 apo state
To identify defined conformational states, the full-length LRRK2 thermal denaturation was assessed by nano-DSC (Fig. 1, blue profile). For this purpose, recombinant SF-tagged full-length LRRK2 was purified from HEK293T cells ( Supplementary Fig. 1). The observed profile shows a very broad shouldered and asymmetric endothermic signal corresponding at the thermal protein denaturation process. However, aggregation phenomena were observed at high temperatures (see Material and methods section), preventing us from performing a straight full equilibrium thermodynamic analysis in absolute enthalpic and entropic terms. Nonetheless, the intensive properties, i.e., both the denaturation temperature range and the calorimetric profile, are still highly informative and permit at least a qualitative assessment of the overall denaturation mechanism.
Indeed, the first part of the thermogram (region A, Fig. 1) indicates the presence of several thermodynamic domains that exhibit low stabilities, which are compatible with relatively less compact protein structural domains (27). In general, there is not straight correlation between thermodynamic and structural domains but, in many cases, thermal stability profile has been successfully used as a structure-correlated fingerprint to assess protein structural peculiarities (27)(28)(29). Being aware of the uncertainties and the need of further confirmation, we highlight that these preliminary indications are in line with the data obtained from the recently published Thermal Shift Assay data for a LRRK2 Kinase-WD40 domain construct (30). According to this analysis, the denaturation temperatures for Kinase and WD40 domains were measured to be 45°C and 56°C, respectively, collocating them in the lowtemperature portion of the overall protein thermal denaturation profile. By contrast, the thermogram portion over 70°C (region B, Fig. 1), reveals a calorimetric profile that is typical of a denaturation mechanism that complains a dissociation process (left tail asymmetry) (27,31) concomitant with the denaturation of a rather stable structures as indicated by the peak size if compared to the previous domains' thermal denaturation signal.

G-nucleotide dependency of LRRK2 conformational states
The oligomeric state of the bacterial LRRK2 orthologue CtRoco correlated to defined G-nucleotide states, with a monomeric GTP and dimeric GDP-bound conformation (32). In order to assess the effects of defined G-nucleotide states on the full-length LRRK2 thermal stability, nDSC measurements were performed by alternatively loading the protein with GDP or GTPγS, a non-hydrolysable GTP analogue. Fig. 1 shows the thermal denaturation profile of the GDP-loaded full-length LRRK2 (red profile). We observed again a very broad endothermic and asymmetric thermogram corresponding to the thermal denaturation of the protein, in first approximation similar to the profile observed for LRRK2 in its apostate. Unfortunately, post-denaturation aggregation phenomena were observed also in this case, though to a minor extent. Accordingly, we limit at comparing the position and shape of the calorimetric profiles being the differences between the two conditions clearly visible. Indeed, the low-temperature portion of the thermogram results to be reduced with respect to the second part, which, by contrast, seems to be enhanced and shifted towards higher temperatures (T max at about 86°C). Accordingly, we may argue that the presence of this G-nucleotide affects the protein overall stability and, it might do this by enhancing the dimer association interaction leading to a much more stable dimer structure compared to the apo-state.
Conversely, as far as the effect of GTPγS on LRRK2 is concerned, the nDSC thermogram shown in Fig. 1 (green profile) still discloses some endothermic thermodynamic steps in the range between 40 and 80°C, which unfortunately cannot be clearly discriminated due to the predominant aggregation and precipitation exothermic phenomena preventing any further analysis at these experimental conditions.
To circumvent the limitations of nDSC and further characterize the monomer-dimer equilibrium, we subjected LRRK2 to Blue-Native PAGE analysis and mass photometry. Blue-Native Page analysis with purified WT LRRK2 revealed that, regardless of the nucleotide binding state, a single band corresponding to a dimeric protein (Fig. 2, Supplementary Figs. 1C, D). However, Blue-Native PAGE requires a considerable high sample concentration (approx. 2µM) to allow the detection of LRRK2 by colloidal Coomassie, which might force its dimerization. In contrast, MP, an interferometric scattering microscopy, allows to determine the molecular mass distribution within a protein sample at low concentrations and minimal volumes (33). Confirming our previous observations, the analysis showed two distinct particle contrasts, which transformed to molecular weights corresponding to the expected weights of the LRRK2 monomer and dimer, respectively. At the measured concentration, the protein was mainly in the monomer state (Supplementary Materials). Subsequently, we measured LRRK2 conformations under different nucleotide states (non-hydrolysable GTP analogues and GDP state) as well as concentrations (25,50,75, and 100nM). However, no significant differences were observed between GDP and GTP-analogue bound states of LRRK2 at the concentrations tested with this technique (Fig. 3A, Supplementary Fig. 2). Interestingly, the concentration-dependent changes in the monomer/dimer ratio from MP allowed an estimation of the dissociation constant (K d ) about 200nM for LRRK2 dimers.
Together, our data shows there is monomer and dimer state of LRRK2, in vitro. While the nucleotide might make some contributions to LRRK2 stability, we did not observe a nucleotide-dependent conformation as seen in the homogenous bacterial protein (32).

Autophosphorylation induces LRRK2 monomerization
Recently we have revealed a feed-back regulatory mechanism of the kinase domain on GTPase activity.
In that study, we could show that the kinetical properties of LRRK2-mediated GTP hydrolysis are negatively regulated by auto-phosphorylation (34). To assess if this negative feedback is mediated via LRRK2 monomerization-dimerization we auto-phosphorylated the purified protein under conditions allowing sufficient phospho-transfer (35) and analyzed its oligomeric state. In the presence of ATP, Blue-Native PAGE analysis revealed in addition to the band corresponding to the LRRK2 dimer, a second band at the expected molecular weight of the monomer (approx. 300kDa) (Fig. 2B).
Consistently, mass photometry demonstrated that LRRK2 pre-incubated with ATP shows significantly less dimerization ( Fig. 3B and Supplementary Fig. 3). To confirm that the decrease in dimeric protein was indeed due to the ATP-induced auto-phosphorylation, we repeated the experiments in presence of the LRRK2-specific kinase inhibitor MLi2. No difference was observed between ATP-treated or untreated LRRK2 in presence of MLi-2, confirming that the observed shift in the monomer/dimer equilibrium was indeed the result of auto-phosphorylation. Similarly, no monomerization could be induced by ATP-treatment for the kinase-dead mutant LRRK2 K1906M (Fig. 3B).
In conclusion, these data implicate that the LRRK2 autophosphorylation regulates the monomer/dimer equilibrium.

Impact of PD mutations on the monomer-dimer cycle
To investigate whether disease-associated variants affect LRRK2 function by disturbing the monomerdimer cycle, we compared the conformational states of different PD variants located in the Roc domain (R1441G) as well as in the kinase domain (G2019S) with the LRRK2 wild-type in dependence of the nucleotide state. In a Blue-Native PAGE analysis no clear defect of particular PD variants on monomerdimer equilibrium were observed ( Supplementary Fig. 1C). The mutant proteins, at the high concentration required for the experiment, clearly show a predominant dimeric form, though a faded signal correspondent to the monomer can maybe be observed ( Fig. 2D) Next, MP profiles were determined using LRRK2 protein purified in the presence of different nucleotides at a fixed concentration (75nM). Among the variants tested, only the R1441G mutant showed a nucleotidedependent shift in the monomer/dimer equilibrium. In fact, a higher degree of dimerization was found for GppNHp-bound LRRK2 compared to the GDP-bound state. In contrast, there was no nucleotidedependent effect observed for G2019S, suggesting that the underlying PD mutations located in different domains might feature distinct molecular pathomechanisms (Fig. 3C, Supplementary Fig. 4).

Integrative modelling suggests preferential binding of RAB10 to the LRRK2 monomer
In addition to its relevance in GTPase activity, the monomer-dimer equilibrium may also play a role in the activation as well as the interaction pattern of LRRK2.
To better understand how LRRK2 physically interacts with its effectors, we followed an integrative computational modelling approach. For the initial assessment of conformational changes upon Rab binding, we employed a crosslinking approach combined with mass spectrometric mapping of the linked lysine residues which we have previously used for building a model of dimeric LRRK2 as well as to map binding sites of LRRK2-specific nanobodies (21,36). Recombinant LRRK2 as well as the Rab proteins were purified from HEK293T cells ( Supplementary Fig. 1). When comparing the resulting crosslinking datasets of different experimental conditions generated by the CID-cleavable crosslinker DSSO, we could already draw two major conclusions: Crosslinks between RAB10 and LRRK2 were only obtained for GTP-loaded LRRK2 and in the presence of RAB29. While crosslinks were seen between RAB10 and Rab29, no crosslinks were seen between LRRK2 and Rab29. Multiple binding sites for Rab proteins have been described in the N-terminal Armadillo domain, which we previously found not to be well accessible to crosslinking, likely due to its compact folding (21,(37)(38)(39). As Rabs are not known to form direct protein:protein complexes while interaction with LRRK2 at micromolar affinity (38), the crosslinks between RAB10 and RAB29 might be the result of constraints provided by LRRK2. In the modelling approach, we limited the analysis to the LRRK2:RAB10 complexes as we only obtained crosslinks between LRRK2 and RAB10 passing the significance filters of the analysis software. When comparing conditions +/-RAB10 we observed a re-arrangement of the intramolecular crosslinks of LRRK2, demonstrating that a major conformational change is necessary to accommodate the substrate in proximity to the enzymatic core. In details, RAB10 binding is concomitant with a loss of XL-mediated contacts mediated by the ankyrin (ANK) domain with both the N-term domains such as armadillo formed between RAB10 and LRRK2 on top-scoring, most representative models, revealed that best results are achieved when only treating the LRR-Roc linker region as flexible (Fig. 4C). This finding is in well agreement with a molecular dynamics (MD) analysis performed recently, proposing an ordered to disordered transition of the linker region (26). Integrative modeling simulations also revealed that monomeric input structures, particularly the AF2 model, achieved maximum MS-XL satisfaction, while lowest agreement between the XL data set and the models was obtained when using dimeric fulllength LRRK2 structures (40) as input (Fig. 4C).
As an independent approach, we also predicted RAB10 complexes by using AlphaFold-multimer, by considering either monomeric or dimeric LRRK2 states (see Methods). We validated the predicted models by mapping MS-XL and assessing their spatial satisfaction in the structural models. Also in this case, we found that the predictions with the monomeric LRRK2 led to a higher number of satisfied crosslinks than the dimeric state (Figs. 4C-E), suggesting that the former is responsible of mediating the interaction with the substrate. Intriguingly, the best complexes between monomeric LRRK2 and RAB10 predicted by AF-multimer involve the same RAB10 interface, mediating MS-XL through Lys 136 on α4 helix, in addition to contacts between RAB10 switch regions with the ARM domain (Fig. 4E).
Among the alternative models predicted by AF-multimer, we also found one where the RAB10's T73 is very close to the KIN catalytic site (Supplementary Fig. 6A). A broader modelling approach for different LRRK2:RAB complexes based on interactions curated in the InAct database is shown in Supplementary   Fig. 5C. In agreement with previous studies, multiple Rab binding sites are detected in the Armadillo domain (38,41). However, district contact patterns with the enzymatic core (Roc-COR-Kinase) of LRRK2 have also been detected.
In particular, we found that monomeric LRRK2 is subjected to a large conformational change of both the ROC and COR domains, which are roto-translated with respect to the FL, monomeric LRRK2 APO state (Figs. 4D, 5A). The Roc-COR conformational change is propagated via the LRR-Roc linker region to the N-terminus (LRR, ANK and ARM), which is maximally displaced to accommodate the positioning of the RAB10 substrate, which on one side is docked on the ARM domain, and in the other it forms an interface with the KIN domain. More in details, we analyzed the hinge and screw axes of the conformational change required to bind RAB10 via established protocols (e.g. DynDom (42)), which highlighted a multi-step conformational change happening through multiple screw axes (Figs. 5A).
One is located at the KIN-COR interface, allowing a rigid-body-like roto-translation of the Roc-COR toward the WD40-KIN (Figs. 5A-C). Another one is centered on the α0 helix linking the Roc and LRR, describing a rotation of the LRR, to allow the large displacement of the N-term (Figs. 5A, D-E). A third axis is found at the border between the ANK and ARM domain, and it is instrumental in presenting RAB10 in close proximity to the KIN domain ( Supplementary Fig. 6). The AF-multimer predicted complexes between RAB10 and dimeric-LRRK2 never display such a large conformational displacement of the Roc-COR as well as the N-term domains, while RAB10 frequently docked on just the ARM domain. This is likely due to the constraint exerted by the dimeric interface, which prevents Roc-COR to rearrange and unleash the large conformational displacement of the N-term, as observed in the monomeric LRRK2 complex (Figs. 5A).
In conclusion, the integrative computational analysis suggests that LRRK2 binds its substrate RAB10 preferably in its monomeric form.

Discussion
Although LRRK2 is considered as one of the most promising targets for PD treatment and thus object of intense research, in particular its activation and regulatory dynamics at a molecular level as well as the impact of PD variants on these mechanisms is yet to be determined at a more detail. Roco proteins, including LRRK2, have been shown to undergo an oligomerization cycle which is considered to be a central mechanism of switch between active and inactive states (43). For this reason, a better understanding of the oligomerization dynamics might give additional insight into the molecular pathomechanisms associated with LRRK2-PD. We used a multifaceted approach, combining biophysical and biochemical analysis with computational modelling to investigate LRRK2 oligomerization, which lead to the identification of a novel feedback mechanism driven by autophosphorylation. In addition, integrative computational models allowed new conclusions about how LRRK2 interacts with its substrates and the role of oligomerization in this process.
In the present work, we determined the thermal denaturation profiles of human LRRK2 protein. The analysis of large multidomain proteins by nDSC is generally challenging and, to our knowledge, has not done before for full-length LRRK2. By providing a detailed analysis of such profiles we could confirm the dimeric nature of purified LRRK2 and observe that its dimer has a remarkable thermal stability indicated by a melting temperature of approximately 85°C. Furthermore, by mass photometry, an optical approach to determine molecular masses in solution, we could estimate a dissociation constant of around 200nM which is one order of magnitude lower compared to GDP-bound Chlorobium tepidum (Ct) Roco, a bacterial Roco protein (32). The observed oligomerization of CtRoco is nucleotide dependent. However, based on the thermodynamic data, a G-nucleotide-dependent oligomerization process, similar to what has been seen for the bacterial homologues, has not been observed.
Nevertheless, a clear effect of GDP binding on purified full-length LRRK2 can be observed, indicated by differences in thermal profile between 40 to 60 °C. Interestingly, the profile in this temperature range correlates with thermal data previously determined for the LRRK2 Kinase-WD40 module (22).
Furthermore, a slight increase in dimer stability can also be observed, indicated by a slight shift of a characteristic large endothermic peak towards a higher melting temperature. However, in presence of GTPγS, a non-hydrolysable GTP analogue, the observed exothermic profile reflects an extensive structural change that leads to unstable and aggregated material, which cannot be attributed solely to a GTP-dependent monomerization as observed for CtRoco. None of the used investigation methods neither nDSC nor mass photometry could give a clear indication of a GTP-driven monomerization.
However, we cannot exclude that GTP is inducing release of the LRRK2 RocCOR dimer interface, but that other domains keep LRRK2 in a dimeric conformation.
In contrast, by analyzing the most relevant PD-associated variants segregating with the disease, we could demonstrate that a variant within the Roc-COR interface, R1441G, shows a higher extent of monomerization when bound to a non-hydrolysable GTP analogue, indicating that indeed a perturbed monomer-dimer cycle could be a relevant pathomechanism.
Furthermore, and rather unexpected, we could identify a mechanism based on LRRK2 autophosphorylation. By mass photometry and Blue-Native PAGE, we could demonstrate that forced in vitro autophosphorylation leads to LRRK2 monomerization. We and others have previously shown that the Roc domain of LRRK2 is subject to autophosphorylation at multiple sites, which are in proximity to the dimerization interface within the Roc-COR module (35,44,45). Furthermore, we recently identified a novel intramolecular feedback regulation of the LRRK2 Roc G domain that depends on auto-phosphorylation of the G1+2 residue (T1343) in the Roc P-loop motif (34). In contrast to wild-type, ATP dependent monomerization is abolished for T1343A, suggesting it plays a key role in regulating auto-phosphorylation dependent monomerization. Interestingly, in cells the T1343A mutant shows a similar increased Rab10-phosphorylation compared to the LRRK2 G2019S PD mutant, suggesting auto-phosphorylation-meditated regulation of the oligomerization state is an important step in the LRRK2 activation cycle. In order to investigate the possible physiological role of an (auto)phosphorylation-induced monomerization, we focused on the interaction of LRRK2 with its substrates. In particular, changes in the oligomerization state have previously been associated with the activation of LRRK2 (18). For this purpose, we used our previously established integrative modelling approach relying on chemical crosslinking-derived constraints (21). Interestingly, we could only see crosslinks between RAB10 and LRRK2 in the presence of GTP and RAB29. In addition, changes in the XL pattern between the LRRK2 apo and the Rab-bound state imply major conformational changes upon activation/switch to a substrate-binding competent state. These findings are in agreement with the recent tetrameric structure of RAB29-bound LRRK2 showing the kinase domain in an active conformation (25). In contrast, a recently high-resolution structure of the LRRK2 dimer is considered to represents an inactive conformation of the kinase domain (40). By integrative modelling, we could indeed show that less intermolecular crosslinks are satisfied when RAB10 is docked on the dimeric LRRK2. In contrast, the LRRK2 monomeric state gave models in better agreement with the experimental data, irrespective of the different simulation and prediction conditions employed. Interestingly, we got more intermolecular crosslinks satisfied when predicting the complex of monomeric state using AlphaFoldmultimer, which was ran in an unbiased, crosslinking-agnostic way. In the latter case, we achieved more reliable complexes without using any available LRRK2 structural templates, again confirming that the FL APO conformation is not competent for RAB10 binding. The higher-scoring AF-multimer predicted interface of monomeric LRRK2 with RAB10 entails the α4 helix, which is also engaged in inter-molecular crosslinks with LRRK2 via Lys 136. Notably, this interface is compatible a ternary complex between monomeric LRRK2-RAB10-RAB29 is predicted by AF-multimer (Supplementary Figure 3B). Intriguingly, the interface between Ras domain's α4 helix, within the Rad domain Cterminal lobe, and a kinase domain has been already observed on homologues proteins (e.g. RHOE-ROCK1 complex, PDB: 2V55). Also, the predicted interfaces between RAB10 and LRRK2's LRR and ARM domains are reminiscent of structures involving homologous Ras GTPases (e.g., RAN-RANBP1, PDB:1K5D and RHOA-SmgGDS, PDB:5ZHX, respectively). This suggests that AF-multimer is harnessing evolutionary and structural information of homologous proteins to predict the conformational changes required to dock RAB10 on LRRK2. In addition to the docking site on the ARM domain, another critical docking hotspot for RAB10 on LRRK2 is the LRR-Roc linker, which is predicted in contact with RAB10 as well as other Rab substrates (Supplementary Figure 3C). This linker therefore appears to have a dual role: on one hand, it enables the conformational dynamics of LRRK2 necessary to unleash its kinase activity, as shown by MD simulation (26) as well as by our hinge analysis on AF-multimer predicted conformers. On the other, it might contribute an additional docking site for Rab substrates binding.
Our unsupervised, comparative analysis of monomeric LRRK2-RAB interfaces also suggests that activity modifier Rabs, such as RAB29, have a different ARM docking mode than Rab substrates (Supplementary Figure 3C), and that the differential ARM docking mode might dictate the functional role played by a given Rab interactor. Modeling analysis indicates that LRRK2 preferably binds and phosphorylates RAB10 in its monomeric state, suggesting distinct roles for LRRK2 oligomer species, where the dimeric form may be responsible for the GTPase activity whilst the LRRK2 monomer exerts kinase activities distinct from the oligomeric state. In fact, our data are implicating that the dimeric state favors auto-phosphorylation events while, in contrast, the monomeric form is the substratecompetent state acting as upstream-effector in a canonical sense. This is also in agreement with the observation that LRRK2 forms more monomers in presence of RAB29 in a Blue-Native gel analysis while no increase of higher-order oligomers is observed (Supplementary Fig. 1D). However, to understand the defined role of the LRRK2 monomeric state in substrate phosphorylation and the dynamics of kinase-active LRRK2 potentially shuttling between different oligomeric states at the membrane, certainly more work is needed.
In conclusion, we show the LRRK2-PD mutations located in different domains might be mediated by different pathophysiological mechanisms. More importantly, our study describes a novel feedback mechanism relaying on autophosphorylation induced monomerization by multifaceted approach. The integrative modelling indicates the LRRK2 monomer is the preferred conformation for binding and phosphorylating its physiological substrates, which might be the down-stream process of monomerization.

Generation of LRRK2 expression constructs.
The generation of N-terminal, SF-tagged (NSF), full-length LRRK2 expression constructs has been previously described (35). N-terminal HA-FLAG RAB10 and RAB29 were generated as described in (46). (Polysciences) solution as described previously (21). After transfection, cells were cultured for 48 h.

Protein purification
Purification of (N)Strep-Flag-tagged LRRK2 as well as (N)HA-Flag RAB10 and RAB29 as well as RAB32 from 14 cm dishes (600 cm 2 ) of confluent HEK293T cells was performed as previously described (21).
Briefly, after removal of the medium cells were lysed in 1 mL of lysis buffer (

SDS and BN gel electrophoresis
SDS and Blue Native gel electrophoresis were performed as described in (21). Briefly, SDS/PAGE was performed using 4-12% NuPage (Life Technologies) gradient gels. BN gel electrophoresis was performed on 4-12% NativePage gels (Life Technologies) according to the vendor's protocols. The gels were stained with colloidal Coomassie.

G-Nucleotide exchange
Nucleotide exchange of the G-proteins was performed a previously described (47). Briefly, freshly purified LRRK2 protein was incubated in Elution buffer supplemented with 10 mM EDTA and a 10-fold molar excess of G-Nucleotide (either GDP or GTPS) for at least 15 up to 30 minutes at 37 °C followed by an additional incubation on ice for 10 min. Finally, MgCl 2 was added to the protein solution at a final concentration of 10 mM. The same protocol was used for nucleotide exchange for RAB10 and RAB29.
For the XL-MS and the Blue Native gel analysis, the Rabs were loaded with either GTP or the nonhydrolysable analogue GTPS.

Thermal analysis measurements
Calorimetric measurements were carried out in HEPES-based purification buffer with a TA Instruments Nano-DSC (6300) equipped with capillary cells (sensitive volume of 0.3 mL) at 0.5°C·min -1 scan rate. A heating-cooling cycle followed by a second heating scan was scheduled in the temperature range from 5°C to 110°C for all experiments. A DSC with a capillary cell design was selected in order to prevent or at least mitigate protein aggregation phenomena (48).
Data were analyzed by means of the software THESEUS (49) following procedures reported in previous studies (50,51). Briefly, the apparent molar heat capacity C P (T) of the sample was scaled with respect to the pre-denaturation region (namely, the region corresponding to T < T i , where T i is the highest temperature at which only the native protein is present, also indicated as molar heat capacity of the "native state", C P,N (T)). The baseline subtraction provided the excess molar heat capacity C P exc (T) across the scanned temperature range. The area underlying the recorded peaks, so treated, directly corresponds to the relevant denaturation enthalpy, Δ d H°. The heat capacity drop, Δ d C P , across the signal was affected by a rather large error and therefore it was not considered in the present work.
Furthermore, despite the use of a capillary nano-DSC, all the samples showed relevant aggregation phenomena, and the measurements were affected by high uncertainties as regards the observed enthalpy values. Nonetheless, a good reproducibility of the overall calorimetric profiles was obtained.
Accordingly, here we only considered these intensive parameters (peak temperatures etc.), and the corresponding thermograms were reported by expressing the C P exc in terms of arbitrary units.

Mass photometry
To evaluate the monomer/dimer ratio of LRRK2, mass photometry on a Refeyn Two MP (Refeyn) instrument was performed (52). Prior to the experiment, the instrument was calibrated using a molecular weight standard (Native Marker, Invitrogen; diluted 1:200 in HEPES-based elution buffer).

Integrative modeling.
We predicted the 3D structure of full length LRRK2, either monomeric or dimeric, in complex with For each cluster, we calculated the number of satisfied MS/cross-links by measuring the number of Cα pairs, corresponding to cross-linked Lys, whose distance was shorter than 35 Å. We recorded the fraction of satisfied cross-links, given by the number of satisfied cross-links over the total, for all the cluster members. We finally reported the fraction of satisfied cross-links for the best scoring solution, as well as the maximum fraction obtained for a single cluster conformer and the aggregate fraction obtained by considering all the cluster members together.
For visualization purposes, we generated all atoms models by starting from the Cα traces generated from IMP and using the automodel() function from Modeler (58).
To encourage the exploration of a broader conformational space and avoid being biased by experimental templates, we generated the models without the utilization of structural templates. This was done by setting the flag --max_template date=01-01-1900 to ensure that AlphaFold-multimer did not rely on any experimental structural data.
In the following analysis we focused on the best model out of the 25 generated. The evaluation of the models was based on the default scoring system used by AlphaFold-multimer (0.2*pTM + 0.8*ipTM).
We generated AF-multimer complexes of monomeric and dimeric LRRK2 in complex with RAB10, as well as of monomeric LRRK2 in complex with both RAB10 and RAB29.
We also predicted the 3D complexes of monomeric LRRK2 with Rab GTPases reported to interact with LRRK2 in IntAct (60). We performed analysis of 3D interfaces using our previously established protocol (61), through which we considered residue-residue contacts as those having the Cβ spatially closer than 8 Å (Cɑ for glycine). We generated a contact fingerprint where each Rab interactor is described by a vector, whose index corresponds to LRRK2 positions, and numerical values to the number of predicted complexes involving a given Rab showing that contact. We employed the contact fingerprint to unsupervised cluster LRRK2-RAB complexes on the basis of their interface structural characteristics.
Structural analysis--The mapping of chemical crosslinks to structural models was performed through a custom script. We computed the distance between the Cɑ coordinates of the crosslinked residues and we considered the crosslink satisfied if the distance was smaller than 35 Å.
To identify the structural rearrangements of the models with respect to the monomeric experimental structure (PDB:7li4) we used DynDom6D (42) (v1.0, setting the grid size to 8 Å, the block factor to 3 and the minimum domain size to 100).

Figures and Tables
in the presence of either GDP (red trace) or GTPγS (green trace).