Is inhibition of kinase activity the only therapeutic strategy for LRRK2-associated Parkinson's disease?

Parkinson's disease (PD) is a relatively common age-related neurodegenerative disorder [1]. In the past few years, it has become recognized that there are single-gene disorders that are clinically and pathologically similar to sporadic PD [2]. Furthermore, a series of genome-wide association studies (GWAS) have identified a number of loci that contribute to the risk for sporadic PD [3‐8].

The most common single genetic cause of autosomal dominant PD is missense mutations in the leucine-rich repeat kinase 2 (LRRK2) gene [9, 10]. Variation around the LRRK2 locus also appears to be a risk factor for sporadic PD [4, 11]. Many single-nucleotide substitutions have been reported in the LRRK2 gene, but only six have been verified to be pathogenic [12, 13]. These mutations are located predominantly in the enzymatic domains of LRRK2, including the kinase domain for which LRRK2 is named. There is a particularly common mutation, G2019S, in the kinase domain that increases the kinase activity of the protein (reviewed in [14]).

For many LRRK2 cases, the clinical signs in affected carriers of the mutation are remarkably similar to idiopathic PD and include tremor, rigidity, postural instability and bradykinesia [15‐17]. There is some variation, as individual patients may have developed amyotrophy [18], whereas in one of the first described families, there is more frequent gait disturbance and less frequent tremor compared to sporadic PD [19]. However, the general picture is that LRRK2 causes a disease that overlaps substantially with sporadic PD.

Because LRRK2 is a kinase, it has been suggested that targeting kinase activity might be a therapeutic strategy for familial PD. Furthermore, because LRRK2-related PD and sporadic PD are clinically similar, it might be extrapolated that targeting kinase activity might also be helpful in treating the more common idiopathic form of the disease.

In this review, we discuss the evidence underlying the idea that LRRK2's kinase activity might be modified to protect against PD. We also focus on the alternative idea that there are other aspects of LRRK2 function that might equally be addressed therapeutically. To understand these two ideas, it is important first to identify the mutations in LRRK2 that are associated with PD and how these affect protein function.

LRRK2 has a complex multidomain structure. The central part of the protein contains three independent domains. The Ras of complex protein (ROC) and C terminus of ROC (COR) bidomain is characteristic of the ROCO family of proteins [20]. The ROC domain binds GTP, and the COR domain is thought to be a regulator of ROC GTPase activity [21]. LRRK2 is an active protein kinase [22, 23], although the true physiological target of this kinase activity has not been resolved.

It has been proposed that the ROC-COR and kinase domains of LRRK2 may control each other. Mutations in the ROC domain that prevent GTP binding are kinase-inactive, leading to the suggestion that the GTPase activity regulates kinase activity (reviewed in [14]). Recent evidence presented by two independent groups, however, showed that kinase activity of LRRK2 is not increased by adding GTP as would be predicted with the use of this model [20, 24]. An alternative interpretation is that the GTP-binding mutants disrupt or destabilize the ROC domain, resulting in a loss of kinase activity as a secondary output. It is possible that instead the kinase domain of LRRK2 regulates the ROC domain and thus that the kinase domain is the modulator of activity rather than the output of the protein [12, 25]. However, whether this regulation actually happens in vivo is unknown [14]. Currently, it is thought that there is some form of regulation between the enzymatic domains of LRRK2, although the details of the mechanisms still need to be resolved.

The flanking ankyrin/heat repeats and leucine-rich repeat (LRR) domains at the N terminus and WD40 domain at the C terminus are thought to participate in protein-protein interactions. However, the physiological role or identification of interactors of these domains and how these domains dictate LRRK2 function remain unclear. There may be a role for some of these regions in the control of LRRK2 kinase activity, as the removal of sequences from the C terminus has been shown to abolish the phosphorylation ability of LRRK2 [26, 27]. Overall, the available data suggest that LRRK2 is a protein that has kinase activity, binds GTP and could play a scaffolding role via its protein-protein interaction domains.

Several groups have shown that the G2091S mutation augments in vitro kinase activity of LRRK2 by two- to fourfold [22, 23, 26, 28‐35]. This might suggest that all mutants have a simple gain of kinase function mechanism. I2020T is located at the next residue of LRRK2 as G2019S, but it has either a modest increase [36, 37] or a reduced [26] or no effect [35] on kinase activity levels compared to wild type (WT). The simplest interpretation is that the I2020T mutation does not increase kinase activity, at least in the way it is currently assayed, which contrasts with G2019S. Therefore, whether all mutations, even those in the kinase domain, increase the kinase activity of LRRK2 is uncertain. Some variants seem to consistently decrease kinase activity, including the G2385R variant [26] that is a genetic risk factor for PD in some populations. Therefore, these variants would seem to be pathogenic by a non-kinase-dependent mechanism.

Another group of LRRK2 mutations is clustered in the ROC-COR bidomain. The R1441 position in the ROC/GTPase is mutated to cysteine, glycine or histidine (that is, R1441C/G/H) in families with PD [10, 38‐42]. Several families who have a Y1699C mutation in the adjacent COR domain have been described [10, 43]. A recent addition to this mutation cluster is N1437H, which was found in a large Norwegian family with autosomal dominant parkinsonism [13]. Overall, the balance of the evidence is that R1441 and Y1699 mutations raise the kinase activity of LRRK2 only very modestly. In contrast, the R1441C/G [44‐46] and Y1699C mutations consistently lower GTPase activity [46]. The GTPase activity for N1437H has not been determined, but it has been shown that this mutant has increased GTP binding [13], which is consistent with lower GTPase activity. The reported GTP-binding capacity of R1441C/G and Y1699C is less consistent, with some reports of an increase and others finding no change in GTP binding [45, 47, 48]. These effects are probably related, as it has been proposed that the ROC and COR domains may interact with each other. R144C destabilizes the ROC domain [49], whereas Y1699C substitution strengthens the intramolecular ROC-COR interaction [46]. Although the precise details still need to be worked out, this suggests that interactions in the ROC-COR bidomain are important in controlling the GTPase activity of LRRK2 and that this may be dysregulated by mutation.

Although the molecular mechanisms by which mutations affect LRRK2 seem to differ, there are phenotypes reported at the cellular level that appear to be consistent between mutations. At least under the conditions of high-level expression of LRRK2 in vitro, mutant forms of LRRK2 are acutely toxic [23, 27, 29, 47, 50]. For G2019S LRRK2, there is evidence that cell death also occurs in vivo with viral vector-mediated expression [51, 52].

In cultured neurons, neurite shortening is also a rather consistent phenotype associated with mutant LRRK2, even in models where cell death is not apparent [28, 53‐58]. Importantly, all mutations tested to date have one or both of these effects in cells, suggesting that these are common outputs of the same signaling pathways that are affected by mutant LRRK2.

Although not all mutations the have been found to increase kinase activity of LRRK2 using assays that have been tried to date, kinase-inactive versions of the same mutations are not toxic [23, 29, 55]. Furthermore, a kinase-dead G2019S is not toxic to dopaminergic (DA) neurons in vivo [51]. This implies that pathogenesis is dependent on having functional kinase activity. This has led to the idea that the enzyme activity of LRRK2 might be targeted in developing a new therapeutic drug for patients with PD.

The facts that G2019S, the most frequent mutation in LRRK2-associated PD, causes augmentation of LRRK2 kinase activity and that cytotoxicity is dependent on kinase activity lead to the suggestion that inhibition of LRRK2 kinase activity might be therapeutically beneficial. Given the complexity of the LRRK2 molecule, however, it is also possible that there are other aspects of function that might be targeted. Herein we discuss the possibilities of LRRK2 therapeutic agents from the standpoint of kinase inhibition, GTP binding inhibition and other possible alternative avenues for LRRK2-associated PD treatment (Figure 1).

Given the number of potential ways in which to interfere with LRRK2 dysfunction, it is rather early to discuss which drugs might be best developed to treat LRRK2-related PD. There are some obvious issues that may stand in the way of the assessment and development of any drug, however, and it is worth discussing those issues with the goal of identifying key experiments in the field that should be performed with a long-term view that LRRK2 is "druggable."

With the development of any drug, the first step in assessing the efficacy of the drug is to test it in a suitable animal model. To date, there are more than ten mouse models carrying different LRRK2-transgenes driven by different promoters (reviewed in [68]). Disappointingly, these models have not yet fully recapitulated all the clinical features seen in a PD patient. Although some DA neuronal death is observed in some of these models, the loss is not progressive and is rather limited with mild motor deficits. However, it has been reported that acute expression of mutant LRRK2 using viral vectors does replicate the DA neuronal cell loss seen in PD [51, 52]. These models might be helpful in that they provide an opportunity to assess the acute application of inhibitors and might be the first level of testing novel agents with transgenic animals used for longer-term experiments, possibly using phenotypes other than DA neuronal cell loss if these can be shown to be reliable.

There are some general limitations in the development of all LRRK2 drugs. First, the potential therapeutic agent should be able to cross the blood-brain barrier (BBB). As discussed in Targeting LRRK2 kinase and GTPase activity, the most promising candidate assessed biochemically and in in vitro models is LRRK2-IN1. When tested in vivo, however, it was shown that LRRK2-IN1 does not affect the phosphorylation of S910 and S935 of LRRK2 in the mouse brain, but that it does abolish phosphorylation of S910/S935 in the kidneys, indicating that LRRK2-IN1 has poor BBB permeability [62]. These data suggest that future LRRK2 inhibitors need to have distinct properties to overcome the current lack of BBB penetration.

A corollary of this issue is that poor penetration of LRRK2 therapeutic agents across the BBB could cause an inadvertent accumulation of the agents in peripheral tissues. This could raise a potentially important toxicity problem, as, for example, it was recently shown that knocking down LRRK2 in mice had deteriorating effects on kidneys [69, 70]. It would therefore be important to monitor the function of organs in which LRRK2 is highly expressed, particularly with regard to kidney and lung function. LRRK2 is also highly expressed in B cells, so immunological function should be assessed.

Another major problem in the development of LRRK2 drugs is the selectivity of the therapeutic agents. Even selective LRRK2 medications could have significant off-target effects if used at high concentrations for an extended period. For example, it is well-established that no kinase inhibitor is exclusively selective to a single kinase [71]. If the concentration of the inhibitor is high enough, it could inhibit other structurally similar kinases and cause multiple side effects, a particular problem for extended use.

Another reasonable question is whether therapeutics based on LRRK2 could also be beneficial for sporadic PD. Ultimately, this question cannot be answered without a therapeutic agent in hand, and in all likelihood this agent would be one that has proven benefit in LRRK2 cases. If one had an appropriate compound, however, it would be possible to address the hypothesis whether LRRK2 PD and sporadic PD are mechanistically linked, so it is worth thinking about the probability that this hypothesis would be supported.

On the one hand, LRRK2 is a defined subset of PD where, as we have discussed, there is even some question whether all mutations work in the same way. This would suggest that LRRK2 mechanisms would be relevant only for LRRK2 PD, perhaps even for single mutations. On the other hand, LRRK2 cases are similar clinically, and sometimes pathologically, to sporadic PD; therefore, it is possible that the underlying mechanisms are similar. In further support of the idea that LRRK2 may play a role in sporadic PD, there is a signal around the LRRK2 locus in GWAS in sporadic PD that is not accounted for by specific LRRK2 mutations [4]. Although the genetic basis of this association is unclear, one interpretation is that changes in the regulation of WT LRRK2 account for some of the lifetime risk of PD. By extrapolation, this would support the idea that LRRK2 is mechanistically linked to sporadic disease and would predict that LRRK2 therapies should be tried in idiopathic disease.

Recently, LRRK2 has been associated with an increased risk in some cancers, in particular melanoma [72, 73]. PD patients with G2019S mutations have a threefold increased risk of developing melanoma before the onset of PD [72]. Whether a kinase inhibitor of LRRK2 would have any impact on cancer is very difficult to predict, but might be considered a potential extra target for drug development.

A biomarker of LRRK2 is needed to answer several of the above questions. Demonstration that a given drug has the desired effect in humans, that is, shows engagement of the target, will be crucial for deciding whether a given treatment has actually tested the underlying hypothesis. Furthermore, having a proximate biomarker for pathological LRRK2 activity would also be helpful in deciding whether the protein is involved in sporadic PD. What such a biomarker would look like is uncertain, although the action of LRRK2 kinase on a target protein is an obvious possibility. Development of these types of markers is eagerly awaited.

We have discussed some therapeutic intervention points for LRRK2 PD based on what is currently known about LRRK2 function and LRRK2-induced cytotoxicity. Although LRRK2 has been implicated in many cellular pathways, including apoptosis, cytoskeleton dynamics, protein translation and other cell signaling cascades, current data are predominantly descriptive and do not explain much of the underlying molecular mechanisms. The lack of this knowledge makes the development of drugs for LRRK2-associated PD difficult. Therefore, elucidation of the exact molecular mechanisms of the neurodegenerative processes caused by different mutations in LRRK2 would be very valuable in the design of LRRK2-specific therapeutics. Of all the proven pathogenic LRRK2 mutations, only G2019S substitution has augmented phosphorylation activity. Data on other mutants do not support significant changes in their kinase function, but they do indicate that other mutants cause toxicity. This fact allows us to propose that the therapeutic strategy for PD caused by different LRRK2 mutations could be diverse and may be dependent on the unique pathogenic mechanism conferred by each substitution.