Mutations in the leucine-rich repeat kinase 2 (LRRK2) gene, associated with Parkinson's disease, have been shown to affect intracellular trafficking pathways in a variety of cells and organisms. An emerging theme is that LRRK2 can bind to multiple membranous structures in cells, and several recent studies have suggested that the Rab family of small GTPases might be important in controlling the recruitment of LRRK2 to specific cellular compartments. Once localized to membranes, LRRK2 then influences downstream events, evidenced by changes in the autophagy–lysosome pathway. Here, I will discuss available evidence that supports or challenges this outline, with a specific emphasis on those aspects of LRRK2 function that have been controversial or remain to be fully clarified.
In 2004, mutations were found in the leucine-rich repeat kinase 2 (LRRK2) gene that caused an autosomal dominant form of Parkinson's disease (PD) in many different families throughout the world [1–3]. These early observations were soon supplemented by reports of relatively common mutations in different populations . Additionally, it was shown that LRRK2 is a risk factor gene for sporadic PD by virtue of the LRRK2 locus being identified in large genome-wide association studies (GWAS) .
When mutations in LRRK2 were first cloned, the function of the LRRK2 protein was unclear as there were no publications reporting specific properties or cellular roles, although the initial reports suggested that LRRK2 was expressed in many tissues, implicating a function relevant across multiple organs. Additionally, the primary sequence of LRRK2 protein showed that it contained both kinase and GTPase domains surrounded by multiple potential protein–protein interaction domains, suggesting a role in cellular signaling pathways.
Over a decade later, several aspects of LRRK2 function have been clarified and a general framework has been elaborated that suggests how LRRK2 fits with other genes for PD to promote neurodegeneration. There is probably too much information to review adequately, so here I will focus on one specific facet of LRRK2 function, the interaction(s) with proteins involved in vesicular trafficking. I will discuss, separately, those aspects that have been confirmed by multiple studies and are therefore relatively well accepted compared with those that are more controversial and indicate where aspects of LRRK2 biology still need to be clarified.
LRRK2 is distributed across multiple cellular membranes via protein interactions
Early studies of LRRK2 noted that it was distributed in cytosolic and vesicular compartments in cells, but was largely excluded from the nucleus . Using a GFP-reporter construct expressed at low levels, Alegre-Abarrategui et al.  mapped LRRK2 expression in cells in culture to a variety of different cellular compartments, including the neck of caveolae, microvilli, multivesicular bodies and autophagic vacuoles. Additionally, localization to the centrosomal marker γ-tubulin was noted, suggesting that LRRK2 is also associated with nonmembrane bound organelles (Figure 1). Several studies also nominated that, in neurons, LRRK2 is present in synaptosomal fractions, presumably representing synaptic vesicles [7,8].
The localization of LRRK2 in cells supports a role in vesicular trafficking.
As many of these structures represent lipid membranes, a reasonable question is how LRRK2 manages to bind each of them. While direct binding of LRRK2 to lipids themselves has not been excluded, an alternate mechanism is that LRRK2 interacts with other proteins that are themselves membrane-associated. Such an indirect association might be more likely than direct binding if LRRK2 has the ability to populate multiple membranes. A good set of candidate proteins for directing other proteins to membranes are the Rab family of small GTPases. Rabs are lipidated to insert them into subcellular membranes and their activity is controlled by GTP-binding and hydrolysis using variety of regulatory mechanisms, including interactions with guanosine exchange factors and GTPase accessory proteins . Importantly, there are >60 Rabs in most mammalian species that are directed to many different cellular membranes  (Figure 1). Several independent observations have supported a physical or genetic interaction between Rabs and LRRK2, often using high-throughput screening approaches.
Using a yeast two-hybrid approach, Shin et al.  first nominated Rab5b as an interactor of LRRK2. In neurons, Rab5 proteins play important roles in the synaptic vesicle exo-endocytic cycle at both presynaptic and synaptodendritic compartments, where they target clathrin-coated endocytic vesicles to endosomes . However, Rab5 is not a neuronally restricted protein and plays important roles in multiple processes across different cell types, including the control of autophagy . This suggests that the LRRK2–Rab5 interaction might be important in nonneuronal systems. Dodson et al.  have that the Drosophila melanogaster homolog lrrk interacts with Rab5 and Rab7. The mammalian LRRK1 protein, a close homolog of LRRK2, can also interact genetically with Rab7 .
Two groups identified an interaction between LRRK2 and Rab7L1 (also known as Rab29). MacLeod et al.  suggested a physical interaction based on examination of genetic interactions in human brain expression data, while my laboratory nominated this interaction using high-throughput screens using recombinant proteins . Rab7L1/Rab29 is potentially important because it is a candidate gene within a GWAS locus for sporadic PD ; therefore, a LRRK2:Rab7L1 interaction may represent a link between different forms of PD. Finally, LRRK2 may also interact physically with Rab32 and Rab38 , which are particularly highly expressed in melanocytes.
Very recently, it was suggested that a subset of Rabs act as substrates for the kinase activity of LRRK2 . Specifically, T73 of Rab10 and equivalent threonine residues in the switch II region of other Rabs (Rab1b, Rab8a and Rab10) were directly phosphorylated by LRRK2 in vitro, whereas Rabs that have a serine at this position, including Rab7L1, are less efficient LRRK2 substrates . However, inhibition of LRRK2 diminished the phosphorylation of serine-containing Rabs, such as Ser105 of Rab12, despite these being substrates in vivo. A second site of phosphorylation, Thr6, has been proposed to be LRRK2-dependent in Rab5b .
Collectively, these data suggest that Rabs are important interactors and kinase substrates of LRRK2. Because a subset of Rabs are targets and/or interactors of LRRK2, then it would follow that LRRK2 would be present at multiple distinct membrane compartments. If these hypotheses are correct, it can be reasonably inferred that Rabs confer some specificity of membrane targeting for LRRK2.
Vesicular trafficking and effects of LRRK2 in cells
Based on the above distribution of LRRK2, we might predict that this protein is therefore involved in multiple cellular processes that are reflective of alterations in vesicular trafficking.
The area of localization of LRRK2 includes portions of the major vesicular endocytic uptake systems, including those important at synapses. Consistent with this idea, Shin et al.  suggested that knockdown or overexpression of LRRK2 impairs synaptic endocytosis and affects neurite outgrowth  perhaps by a Rab5-dependent mechanism. Other groups have also proposed that too much or too little LRRK2 affects synaptic endocytosis via phosphorylation either of the BAR-domain protein endophilinA [20,21] or the ATPase, N-ethylmaleimide-sensitive fusion protein . At the time of writing, these results have been reported from single laboratories, and so confirmation is needed, but they collectively suggest that a very specific regulation of LRRK2 is important for synaptic function.
Away from the cell surface, the localization of LRRK2 at multivesicular bodies and autophagic vacuoles  and lysosomes  suggests a role in processes downstream of endocytosis, potentially including the autophagy–lysosome system. Autophagy is a highly conserved set of pathways that promotes protein and organelle turnover through the lysosome using a series of membrane-based engulfment and fusion events. Importantly, alterations in function of the autophagy–lysosome system are thought be a convergent path for multiple forms of PD .
Cell-based assays strongly indicate that LRRK2 plays a role in regulating the autophagy–lysosome system. For example, in the original description of LRRK2 localization to multivesicular bodies, siRNA against LRRK2 was reported to increase the turnover of lipidated LC3, or LC3-II, a marker of the nascent autophagophore and autophagosome prior to fusion to the lysosome . In contrast, expression of a mutant form of LRRK2, R1441C, blocked autophagic turnover of multivesicular bodies. Consistent with this idea, it was shown that several pathogenic mutations in different domains of LRRK2 (R1441C, Y1699C and G2019S) in patient fibroblasts, and hence expressed at the endogenous level, blunted the autophagic response to starvation . Conversely, inhibition of the kinase activity of LRRK2 increases LC3 turnover , consistent with the previously discussed data using siRNA . However, some experiments have indicated that shRNA can diminish LC3 turnover in cells that endogenously express LRRK2, indicating that some caution should be applied before trying to determine the precise effects of LRRK2 across an organism .
As well as this cellular data, there is evidence that LRRK2 influences the autophagy–lysosome system in vivo. Knockout of the Drosophila lrrk gene results in sterility of female flies and is correlated with altered positioning of lysosomes [12,28], potentially consistent with the above observations of increased autophagosome turnover. Overexpression of wild-type or mutant LRRK2 in Caenorhabditis elegans enhances normal age-related impairments in autophagy, increasing autophagic mass and again potentially consistent with loss of turnover of autophagosomes .
Elegant experiments in mice have confirmed that LRRK2 deficiency is associated with alterations in the autophagy–lysosome system [30,31], although there are still some areas that need to be clarified (discussed below).
Another set of vesicular functions that relate to LRRK2 include transit of vesicles to and from the ER–Golgi, which represents a protein sorting pathway. Many proteins that are synthesized at the rough ER are then trafficked through the Golgi apparatus and sorted towards their final destination using accessory proteins. For example, cathepsin proteases bind to the cation-independent mannose-6-phosphate receptor (CI-M6PR) in the trans-Golgi network (TGN) and then help sort towards their lysosomal destination. CI-M6PR is then recycled back to the TGN using retromer protein complexes  that includes a known PD gene, VPS35 , or other retrograde protein complexes . LRRK2 has been proposed to be involved in several aspects of the TGN–retromer system. Two independent reports have suggested that increased expression of VPS35 will prevent toxic effects of mutant LRRK2 in cells and flies [14,35]. Additionally, a constitutively active form of the VPS35 interactor Rab9 is reported to reverse loss-of-function phenotypes in lrrk knockout flies .
LRRK2 can be recruited to the TGN by protein interaction partners, including Rab7L1 in mammalian cells [14,15], and by a Bcl2 athanogene/heat shock protein complex in C. elegans . These observations in different species are very likely to be related to each other as BAG5–HSP70 stabilizes the interaction of LRRK2 with Rab7L1 in mammalian cells . Expression of the LRRK2:Rab7L1 complex is associated with fragmentation of the TGN and loss of immunoreactivity for TGN markers, a process that requires the autophagy–lysosome system . These observations need to be confirmed or refuted by others, but they may be related to fragmentation of TGN seen in some mouse models , or the retrograde transport of Golgi-derived vesicles promoted by LRRK2 overexpression in C. elegans .
It has been reported that LRRK2 may interact with a component of the ER exit site, Sec16a, and affect ER to Golgi transport as well as the above retrograde effects. Because the trafficking of lysosomal proteins such as cathepsins requires anterograde sorting from the ER through the TGN, such observations of an effect of LRRK2 on ER exit might explain the accumulation of lysosomal proteins in LRRK2 knockout mice [30,31,39], although equally a lack of adequate retrograde sorting and retrieval of CI-M6PR may also contribute to lysosomal dysfunction.
It is important to note that some of the effects of LRRK2 on vesicular transport may be mediated by protein interactions that are not at membranes. LRRK2 is known to be present at tubulin-enriched structures [6,40], to interact with a variety of tubulin isoforms [41,42] and influence the acetylation of a subset of tubulin molecules [43,44]. Microtubules, which are formed of tubulin and accessory proteins, are important for intracellular transport within cells, including secretory pathways from the Golgi . It is reasonable to speculate that the ability of LRRK2 to simultaneously bind to tubulin isoforms and Rabs is important for its function in vesicular trafficking (Figure 1).
Overall, the above examples suggest that LRRK2 is not only present at several membranous structures in a variety of cell types but also plays functional roles in controlling vesicular trafficking. There are, however, many important mechanistic questions that need to be asked about how, precisely, LRRK2 affects these important phenomena. I will briefly discuss an area that I find particularly difficult to be certain about, which relates to the direction of effect of mutations in LRRK2.
Uncertainties about direction of effects lead to new questions about LRRK2 mutations
When considering a gene like LRRK2 where multiple mutations have been shown to promote pathogenic events, an important consideration for harnessing the gene for therapeutic development is what properties those mutations share. Across all human genes, some mutations clearly cause a loss of normal function, whereas others are more complex and can cause gain of normal or acquisition of novel functions. In a few cases, loss of function alleles can be demonstrated to be neutral or even protective, leading to a strong inference that inhibiting normal protein function would be therapeutically beneficial . Therefore, it is critical to understand the direction of effect of PD-associated mutations — do they disrupt the normal function of LRRK2 or enhance it, creates a novel function, or even have some complex combination of all of the above?
The balance of evidence in cell-based models is that most of the best-characterized LRRK2 mutations are active. For example, cells transfected with any of the known pathogenic mutations show more autophosphorylation of serine 1292 in LRRK2 [47,48], more phosphorylation of Rab proteins  and more retention at the TGN  than the wild-type version. However, a full explanation for these observations remains elusive as only some LRRK2 mutations, principally the G2019S mutant of the kinase domain, have a substantially enhanced kinase activity when purified in vitro [17,49]. Likewise, direct binding to Rabs is not notably different between mutations, at least when assayed using co-immunoprecipitation . A possible explanation of some of these discrepant data is that there are factor(s) in the cell that are missing from simple in vitro reconstitutions that perhaps influence signaling of LRRK2, but then this begs the question as to what those factor(s) are.
Perhaps more importantly, there is not yet a full consensus on whether mutations have consistent effects on vesicular trafficking in vivo, although two particularly interesting observations have emerged. First, in LRRK2 knockout tissue, the major organ affected is not the brain but kidney, which happens to have a high level of LRRK2 relative to the homolog LRRK1 [30,50]. Second, the effects of LRRK2 deficiency are age-dependent . In 7-month-old cohorts, there was an increased ratio of LC3-II/LC3-I but the opposite was true in older animals. However, in an independent study, there was a generalized increase in LC3-II from 12 to 20 months rather than a biphasic response . One difficulty with interpreting these data, especially in light of the variable data in cells, is that LC3-II/LC3-I ratios can increase by both greater autophagosome generation and/or by diminished clearance. In cells, it is possible to address this directly by inhibiting the lysosome and thus estimating overall flux through the pathway, but this is not straightforward in animals. Therefore, although these data suggest, from multiple independent laboratories, that loss of LRRK2 impacts autophagy there are several specific aspects that need to be clarified in the future. Importantly, although there are mice where pathogenic LRRK2 mutations have been knocked in to the germline and should therefore be expressed at the endogenous level in the correct cell types [52–54]. If LRRK2 were a simple loss of function allele, then one would predict a similar kidney phenotype in knockin and knockout animals. One available study suggests that the G2019S knockin allele does not have any histopathological defects, arguing against a simple loss of function event , but this needs to be confirmed with additional animal models and using assays perhaps more directly targeted at immediate LRRK2 biology.
Along the same lines, there is some disagreement between the two available studies [14,15] in the expected direction of effect that Rab7L1 has on LRRK2 function and PD risk. Our own data suggest that, in cells, knockdown of Rab7L1 rescues effects of LRRK2 on Golgi integrity , whereas MacLeod et al. propose that a putative constitutively active Q67L mutant of Rab7L1 can rescue LRRK2 effects on neurite length, i.e. that more Rab7L1 is beneficial. One difficulty with this specific mutant is that while other Rabs have a GTP preference when a QL mutation is introduced, in our hands Rab7L1–Q67L fails to retain GTP or GDP after pre-loading with nucleotides . This correlates, in both available studies, with a loss of Rab7L1 at the trans-Golgi network [14,15]. Additionally, although MacLeod et al.  proposed that lower Rab7L1 protein levels in the human brain are associated with an increased PD risk, our own data using both microarray  and RNA-Seq  approaches found the opposite association. How these different observations will eventually be resolved is not yet clear, but it is of interest that a recent report of Rab7L1/Rab29 knockout mice shows that they mimic the phenotypes of LRRK2-deficient animals . This suggests, in contrast with previous models  but consistent with our predictions , that LRRK2 and Rab7L1 activity work to modify vesicular trafficking in the same direction. As for the comparison of knockout and knockin LRRK2 alleles discussed above, these are important data to clarify and confirm.
Several lines of evidence, gathered from different laboratories using multiple organisms, suggest that LRRK2 and its homologs play important roles in the control of vesicular trafficking in cells. A specific focus has emerged around Rab proteins as interactors and/or LRRK2 kinase substrates as these proteins likely explain why LRRK2 is present in multiple membranes and affects multiple protein uptake, sorting and recycling events. However, some aspects of the mechanism(s) involved and specifically relating to the direction of effect of LRRK2 mutations and interactions remain to be clarified.
One of the things that has been most exciting about working on LRRK2 is to have gone from having literally nothing but a sequence to at least some outline of basic biology in a relatively short space of time. Furthermore, the possibility that LRRK2 might be a reasonable drug target for PD, currently an incurable condition, has helped maintain interest in this molecule for many constituencies. However, we can also clearly state that until LRRK2-based approaches have been proved to have clinical utility, then there is a great deal of work to do, both on the fundamental biology and on the translational endpoints that will affect patients. My view is that these last two aspects are intertwined and that, by understanding how and why LRRK2 influences vesicular trafficking, we will find additional pathways that we can manipulate to improve PD outcomes.
This research was supported by the Intramural Research Program of the National Institutes of Health, National Institute on Aging.
The Author declares that there are no competing interests associated with this manuscript.