In this study we investigated for the first time the quaternary structure of full-length LRRK2 and of its closest homolog LRRK1 by transmission electron microscopy. A major hurdle to this type of analysis has been the obtainment of pure full-length LRRK2 in sufficient quantities suitable for structural characterization. Several attempts have been made to obtain LRRK2 fragments from prokaryotic sources 
, however the only proteins reliably demonstrating kinase activity have been purified from eukaryotic cells as truncated proteins 
. Here, we developed a protocol to purify full-length recombinant soluble LRRK1 and LRRK2 proteins from mammalian cells by immunoaffinity chromatography. We used a combination of spectroscopic measurements and enzymatic assays to demonstrate that LRRK1 and LRRK2 retained native folding and enzymatic activities and were therefore suitable for single particle analysis by transmission electron microscopy using negative staining and immunogold labeling.
The regions of highest homology between LRRK1 and LRRK2 are in the LRR domain and the catalytic ROC-COR-kinase tridomain. Our results suggest that sequence similarity is reflected in functional similarity for some domains. GTP binding affinities of LRRK1 and LRRK2 were comparable using an isotopic displacement assay and are in the same range as previously reported for LRRK2 using a similar assay 
. The only other ROCO protein for which GTP binding affinities have been measured is the C. tepidum
ROCO for which a similar GTP dissociation constant was reported (0.5 µM for GppNHp, 13.4 µM for GDP) 
While both LRRK proteins can bind to ATP as predicted for kinases, their kinase activity revealed striking differences. In particular we could not observe LRRK1 kinase activity toward itself or against generic or LRRK2 specific substrates, an observation that may be, in part, explained by the presence of an extra loop between β-sheet 5 and α-helix C1 of the N-terminal lobe. This loop possibly hinders substrate binding. Methods
to assess LRRK2 phosphorylation have been well characterized and include autophosphorylation 
or the phosphorylation of LRRK2 specific peptides such as LRRKtide or Nictide 
. In all cases here, the phosphorylation activity of LRRK1 was below detection limits for substrate phosphorylation assays and very low for autophosphorylation activity. These results are in agreement with another recent study where the authors did not observe significant autophosphorylation activity for LRRK1 
. It is interesting to note that in a previous study where we assessed LRRK1 and LRRK2 autophosphorylation activity using proteins bound to affinity resins 
, we observe only small differences in autophosphorylation activity, indicating the importance of working with soluble and pure protein preparations to improve the signal to noise ratio. We conclude that LRRK1 phosphorylation properties are fundamentally different from those of LRRK2, substrate specificity differs and while autophosphorylation is robust in LRRK2, it is not likely to play a significant role for LRRK1. Therefore, although the kinase domains are similar for LRRK1 and LRRK2, at a functional level the two proteins are likely to be very different.
To explore whether LRRK1 and LRRK2 are dimeric, we used these biochemically validated proteins to investigate their quaternary structure. Although crystallography represents the ideal approach to gain insight into the details of a protein structure, transmission electron microscopy (TEM) has the unique advantage of requiring diluted samples for image analysis and is ideal for molecules larger than 300 kDa such as LRRK1 and LRRK2. Several lines of evidence support the notion that active LRRK2 is dimeric 
. However, these studies are based on protein fractionation of cell lysates where predicting protein molecular weight, and by inference monomer vs dimer state, is difficult because of the presence of heterologous binding partners.
Here we show for the first time that highly purified LRRK1 and LRRK2 are capable of forming dimers as revealed by the presence of double-gold labeled LRRK1 and LRRK2 particles imaged by TEM. The distance between two gold particles is smaller than the mean particle size of LRRK2 suggesting that the N-terminal region, which is labeled by anti-Flag antibodies, possesses a well-defined structure. In particular, the width of the half height of the curve that interpolates the distribution of distances between gold particles is around 60 Å. Considering that the complex of primary and secondary gold-labeled antibodies is oriented randomly, the N-terminal region is likely to possess a structure that places the epitopes in a defined position. For LRRK1, the width of the half height of the curve is approximately 100 Å suggesting that the N-terminal region is possibly more flexible. Importantly, addition of a denaturing agent such as 6M GdHCl, that induces unfolding of the protein as witnessed by the red-shift of tryptophan fluorescence emission, caused a sharp decrease in the number of double-gold labeled particles compatible with the dimer size, supporting the notion that folded proteins are compact dimers. Collectively, these results support the hypothesis that LRRK proteins are capable of forming dimers in the absence of other binding partners. We also analyzed the effect of point mutations disrupting kinase activity, GTP binding or the hyperactive pathogenic G2019S mutation on dimer formation. Although we observed a reduction in the number of immunogold doublets in the G2019S and kinase dead mutants, it is difficult to conclude that these proteins are less prone to assemble in dimers rather than presenting small structural perturbations, which reflect on the antibody-antigen affinity. Interestingly, purified GTP-dead LRRK2 (T1348N) retains ability to form dimers, indicating that loss of GTP-binding capacity, although likely to perturb the three dimensional structure, is not sufficient to abolish dimerization. This is consistent with previous reports showing that LRRK2 self-interacts via multiple different domains 
We have also been able to visualize dimeric LRRK2 particles from chromatographic fractions corresponding to an apparent molecular weight of 600 kDa, which have been probed for endogenous LRRK2. There are now several proteins that have been shown to interact with LRRK2, including 14-3-3, HSP-90 and tubulin and whether these contribute to the ~600 kDa peak has not been addressed. Given that immunogold EM was performed on enriched protein fractions rather than highly pure sample due to technical limitations on purifying endogenous material, we cannot rule out whether the chromatographic peak contains both dimeric and monomeric LRRK2 with binding partners. A recent paper by Ito and Iwatsubo presents intriguing observations supporting the conclusion that LRRK2 from lysates is predominantly monomeric 
. As the SEC profiles and immunogold particle distance distributions are comparable in the stable lines compared to endogenous LRRK2 in NIH-3T3 lysates, we can conclude that LRRK2 forms dimers in cells, probably co-existing with monomers.
The next key steps will be to pursue in more detail the structure of LRRK proteins and its links to their biology. The validated LRRK proteins presented here can be used to investigate their conformation by cryo-electron microscopy, in order to obtain more detailed structural information to model their tridimensional structure. The correlation of more detailed structural information of LRRK1 and LRRK2 to their cell biological properties such as how dimerization is linked to cellular complex formation will allow a better understanding of how LRRK biochemical properties impact LRRK mediated cellular functions. This is of particular interest for LRRK2 whose biochemical properties, such as its kinase activity, have been proposed as potential targets for disease modifying Parkinson's disease therapy.