Abstract

A B S T R AC T : The burden that Parkinson's disease (PD) exacts on the population continues to increase year after year. Though refinement of symptomatic treatments continues at a reasonable pace, no accepted therapies are available to slow or prevent disease progression. The leucine-rich repeat kinase 2 (LRRK2) gene was identified in PD genetic studies and offers new hope for novel therapeutic approaches. The evidence linking LRRK2 kinase activity to PD susceptibility is presented, as well as seminal discoveries relevant to the prosecution of LRRK2 kinase inhibition. Finally, suggestions are made for predictive preclinical modeling and successful first-in-human trials. The upcoming tenth anniversary of the discovery of mutations in the leucine-rich repeat kinase 2 (LRRK2) gene in Parkinson's disease (PD) highlights numerous achievements in discovery and innovation Optimism also exists in industry and academia alike for targeting the LRRK2 protein for therapeutic intervention in neurodegeneration. Yet, lessons from Huntington's disease and Alzheimer's disease show a long and convoluted road between gene discovery and drug development. Rational strategies that rely on accurate comprehension of pathobiological mechanisms are likely required to identify efficacious therapeutics. Focusing on the last 10 years of work, this perspective article provides a context for exploring the critical issues related to LRRK2 in PD susceptibility and therapeutic development. Specific suggestions for the advancement of LRRK2-targeting small-molecule kinase inhibitors to successful first-in-human studies are proposed. Linking LRRK2 to PD Initial excitement with LRRK2 was not that another locus was found to be linked to another familial version of PD. In that case, LRRK2, localized to the PARK8 locus, is promptly eighth in line and many more candidates having since followed (e.g., PARK9-PARK20). Rather, the first wave of excitement came from the descriptions of the families linked to the PARK8 locus. Usually, familial PD-linked loci are confined to a few families, but many unique families across the globe were linked to PARK8. surprisingly, the discovery of mutations in the LRRK2 gene in late 2004 was disclosed by several genetic groups collaborating into two independent articles published at the same time. S C I E N T I F I C P E R S P E C T I V E S 8-10 The second source of excitement was the nature of the disease linked to PARK8. Usually, familial parkinsonism involves early-onset forms of disease, often in concert with neurological symptoms not usually associated with late-onset typical disease. The importance of these Mendelian-inherited genes in idiopathic PD then becomes reliant on downstream pathological, functional, or therapeutic approaches. PARK8 needs no additional studies to demonstrate importance in late-onset PD. One of the largest, best described families linked to PARK8 was reported in 1995 by Ronald Pfeiffer and Zbigniew Wszolek who concluded that "This large kindred appears to represent a neurodegenerative disorder closely resembling, if not identical to, idiopathic PD." 11 This prescient observation has borne out in the last decade remarkably unscathed, even in the face of issues that commonly fog coherent genotype-phenotype linkages, such as clinic bias in subject ascertainment and publication bias of outlier families and cases. There are dozens of common nonsynonymous variants scattered throughout the LRRK2 gene in various populations and individuals (http://www.uniprot.org/ uniprot/Q5S007) and, possibly, hundreds of rare or idiosyncratic variants. Only a minority of these variants are linked to PD. As yet, there is no biochemical assay, no definitive molecular biology test, to conclusively demonstrate the pathogenicity of a particular variant. Pathogenic mutations in LRRK2 (listed in [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] 181 Although pathogenic variation in LRRK2 is rare in humans, common genetic variants (e.g., minor allele frequencies of greater than 1% in a particular population) in the LRRK2 gene are well established to affect susceptibility to disease. Some of these susceptibility variants are listed in 12 In consideration of both familial and population studies, apart from a-synuclein (PARK1/4), no other gene shares as strong a relationship to late-onset PD. In late-onset PD genetics, frequencies of pathogenic mutations are usually incredibly low in clinical populations, and associated presentation of the inherited disease in carriers is highly variable. The remarkably high frequencies of LRRK2 mutations in late-onset PD have allowed unprecedented insight into LRRK2-linked phenotypes. Two salient features have emerged: First, there are no reliable clinical measures or tests to identify a LRRK2 mutation carrier from idiopathic late-onset PD, short of genetic testing. 14 This is owing, in part, to the second feature critical for understanding LRRK2 in PD: Pathogenic mutations are not fully penetrant. In Ashkenazi Jewish cohorts of PD, lifetime penetrance is estimated at less than 30% for developing PD. 14 Penetrance in typical Caucasian populations is not clear, but is the subject of scrutiny by 23andme.com and other active consortia. 18 Nevertheless, other factors besides LRRK2 mutations are necessary for the development of PD. LRRK2 in the Kinome Genetic studies have a habit of identifying proteins in neurodegenerative disease that make terrible targets for traditional therapeutic interventions. Of the 7,668 unique genes associated with known or potential druggability, frustratingly few of them are associated with PD. 19 Indeed, many of the PARK loci highlight loss-of-function recessive forms of disease (e.g., parkin, PINK1, and DJ-1). Because most therapies in the clinic, particularly small-molecule based, tend to attenuate or ablate the activity of a protein target, restoring complex function that is lost can be much more challenging. As a protein kinase and prominent card-bearing member of the druggable proteome, LRRK2, in many respects, is the most exciting drug target identified in modern PD research. Human DNA encodes 518 protein kinases, and this collection of protein, known as the kinome, is included in the druggable proteome. However, LRRK2 bears little resemblance to other protein kinases. LRRK2 is awkwardly nestled with other problematic proteins in the so-named "tyrosinekinase-like family," more by virtue of the nonspecific fact that LRRK2 is a multidomain protein versus anything known about function or expression profiles. 20 Within superfamilies in the kinome, the encoded kinase domains are often so inbred in sequence similarity that it becomes difficult to find small molecules that interact with only one class of kinase or an individual protein. Specificity, not potency or other druglike properties, is the first fundamental problem with targeting proteins such as LRRK2. To tackle specificity issues, exploitation of LRRK2-specific sequences and structures in the adenosine triphosphate (ATP) pocket present a way to navigate through the usual quagmire of off-target interactions. LRRK2 harbors unusual amino acids in kinase subdomains that are otherwise highly conserved across the kinome. For example, the DFG hinge motif, critical to "in" versus "out" conformations in kinase activation, is DYG in LRRK2. The LRRK2 DYG motif is further altered with the pathogenic LRRK2 mutation, G2019S, to a DYS motif. These LRRK2-specific sequences encode the very amino acids that form the ATP pocket that many kinase inhibitors interact with for therapeutic gain. 21 Besides unique kinase-domain sequences, there are other features specific to LRRK2. LRRK2 is regrettably named because several other human protein kinases have leucine-rich repeat domains (e.g., leucine-rich repeat receptor kinases), but no other protein has a tandem encoded GTPase domain with proven enzymatic function. This defining enzymatic duet is conserved across >500 million years of evolution between humans to single-celled organisms such as Dictyostelium, 22 demonstrating obvious essentiality to the arrangement. As with the LRRK2 kinase domain, the LRRK2 GTPase domain also diverges from other G-protein families (guanine nucleotide-binding proteins). 23 Given that the LRRK2 GTPase domain cannot reasonably be assigned to any of the main G-protein families (e.g., Ras, Rho, Rab, Arf, or Ran), a new family called "Ras-like proteins in complex with other domains" (ROC) was created. In humans, the family is comprised of LRRK1 and LRRK2 and substantiates the overall evolutionary distance of LRRK2 from other well-characterized kinases. Alignments of the most intrinsically conserved LRRK2 GTPase residues against the prototypical H-Ras protein suggest that the amino acids commonly used in biochemical studies to inactivate GTPases (e.g., H-Ras sequence glycine 12 and glutamine 61) are already substituted in the LRRK2 GTPase domain. Furthermore, the typical phenylalanine-to-leucine mutation, useful for studying many G proteins (e.g., position 28 in H-Ras), is also natively a leucine in LRRK2. These two features are indicative of low affinity for nucleotides, compared to other G proteins, and a mostly inactive enzyme in cells. The notion that the vast majority of the LRRK2 enzyme lay enzymatically dormant in cells has, to the chagrin of LRRK2 biologists, largely borne out experimentally. Nevertheless, a recent first study of its kind suggests therapeutic potential for molecules that may bind to the GTPase domain. 24 The uniqueness of the LRRK2 enzyme thus presents a gift and a curse: a gift in that there are viable protein domains and interactions that are unique to LRRK2, so molecules should exist that interact only with LRRK2 and therefore subvert off-target interactions. The curse is that decades of research on how G proteins regulate protein kinases may not provide relevant insight into LRRK2 function, given that the LRRK2 enzyme diverged quite early from other better-characterized kinases and G proteins. Impact of Pathogenic Mutations on LRRK2 Activity In order to therapeutically target LRRK2, it would be useful to understand the effects of pathogenic mutations on LRRK2 function. LRRK2 may have dozens of different activities in hundreds of different kinds of cells, so narrowing down the property most clearly linked to PD would provide a reasonable foundation to pursue and validate targeted therapeutics. Less than 1 year after the discovery of mutations in LRRK2, it was possible to clone LRRK2, develop reasonable polyclonal antibodies (Abs), and create an initial assay to measure LRRK2 kinase activity. 25 During this time, some reports, based on homologous modifications made to other protein kinases, suggested that PD mutations would inhibit kinase function. 25 Kinase-activating effects of LRRK2 mutations could be caused by many factors. Based on the distribution of pathogenic mutations across the LRRK2 ROC, COR, and kinase domain ( First and foremost may be the assumption that a representative proportion of the active state of the LRRK2 enzyme is captured and preserved from protein purified from tissues or cells for use in a particular assay. For example, the I2020T alters the proportion of protein in an active DYG-in pocket conformation, but the increase in kinase activity can be negated depending on assay conditions and the nature of the kinase substrate. 27 Similar active-state stabilization mechanisms may be occurring in the ROC domain for other pathogenic mutations. 28 Active-state conformations can also be affected variably by at least seven other known factors, usually present in unknown stoichiometries with respect to enzyme, in published kinase-assay experiments: (1) LRRK2 protein cofactors and interactors such as 14-3-3 and heatshock proteins that coelute with LRRK2, as well as other less-abundant interacting factors such as ArfGap1, (2 and 3) metal such as Mg 11 bound to the GTPase domain, and kinase domain, (4) guanine nucleotide bound to the GTPase domain, (5) adenosine nucleotide bound to the kinase domain, (6) peptide substrate docked in the kinase domain, and M O V I N G L R R K 2 I N H I B I T O R S T O T H E C L I N I C A subsequent quantitative MS study provided evidence that the autophos pS1292 residue is particularly abundant and therefore be detected directly from lysates. In the first study directly measuring LRRK2 autophos activity from LRRK2 protein expressed in cell lines, pathogenic LRRK2 mutations increased the proportion of pS1292-LRRK2, relative to total LRRK2. 32 Still, outside of measuring autophos in cells, the activating effects of individual mutations outside of the kinase domain have variable or negligible effects on some aspects of kinetics in certain experimental paradigms. Though the identification of bone-fide LRRK2 kinase substrates important in LRRK2-linked cellular pathways might overshadow studies that measure the effects of pathogenic mutations through measuring autophos levels, for now, the results closest to cellular (and thus relevant) conditions support a kinase-activation hypothesis for LRRK2-linked pathogenesis Small-Molecule LRRK2 Inhibitors There is no longer debate as to whether smallmolecule kinase inhibitors can be highly selective and clinically efficacious. 35 Tremendous efforts have focused on the identification of small-molecule kinase inhibitors that selectively target LRRK2 kinase activity to bring enzymatic function back to normal (e.g., WT), or ablate activity altogether, in the hopes of a neuroprotection strategy for PD. Initial studies identified several classes of nonselective kinase inhibitors (e.g., molecules that inhibit >20 known kinases at 50% inhibitory concentration <1 mM) with excellent (low nanomolar) potency against LRRK2. These molecules include staurosporine, sunitinib, CZC 54252, and TAE684. Owing to their promiscuity, these compounds have limited utility when applied to cells and cannot provide information on the safety of selective LRRK2 inhibition. More recently, molecules with improved selectivity have been described on several distinct scaffolds. The first of these, aptly dubbed LRRK2-IN-1, has been widely deployed in numerous high-profile biological studies that attempt to define the role of LRRK2 in model systems and/or rescue pathological effects of G2019S-LRRK2. Two other LRRK2 inhibitor scaffolds were identified with improved specificity, pharmacokinetics, and distribution in vivo. GSK2578215A is a highly selective compound with minimal inhibition of other kinases, of 460 kinases tested, and shows evidence of brain penetration. Despite few options currently available for in vivo experiments, the limiting factor is not the ability to resolve efficacy. Abs directed to any of the phosphorylated residues identified in LRRK2 would theoretically track LRRK2 kinase inhibition, including the phosphoserine Abs that are not autophosphorylation sites. 34 Abundant N-terminal phosphorylation sites, such as pS935 and pS910, that do not directly measure LRRK2 activity, but faithfully correlate with LRRK2 inhibition, have proven useful in dozens of studies. However, these sites are not preferred over autophos sites because they would also track inhibition of other kinases (i.e., not LRRK2) that phosphorylate the LRRK2 sites. In addition, 14-3-3 proteins require pS935 and pS910 phosphorylation to bind to LRRK2, so factors that alter 14-3-3 function in cells may have indirect effects on pS935 and pS910 phosphorylation by allowing other kinases to interact with the sites that would normally be blocked by 14-3-3 protein bound to LRRK2. Because of patent-life vulnerability, it is reasonable to expect that the best LRRK2 inhibitor series currently remain undisclosed. Published inhibitor series likely harbor critical flaws that preclude consideration as strong clinical candidate molecules. Nevertheless, the inhibitor series that have been publicized show relatively good selectivity and potency toward LRRK2, so that better molecules should exist within the scope of reasonable amounts of effort. Preclinical Approaches for the Identification of Efficacious LRRK2 Kinase Inhibitors In PD research, there are no known neuroprotective treatments, so identification of a model system that predicts clinical success for neuroprotection cannot exist with certainty. However, several major advances in PD research that preceded the discovery of LRRK2 in PD by a few years have had resounding trickledown effects in preclinical approaches that should be considered in testing LRRK2 kinase inhibitors. First, Abs directed to abnormal a-synuclein, applied in a systematic manner to postmortem archived brain sections from PD, led to the identification of a staging system that followed a common progression for a-synuclein lesions. Most in vitro studies have concluded that the various toxicities caused by overexpression of LRRK2 are dependent on LRRK2 kinase activity. Another salient, but also highly variable, result of overexpression of LRRK2 in neurons and cell lines alike is the development of skein-like LRRK2 aggregates. There are no consensus assays emergent from the literature that have been easily reproduced across laboratories to assess LRRK2 toxicity. LRRK2 function, relevant to mechanisms important in PD, may be alternatively understood in combination with dysfunction elicited by other factors underlying PD. In the first large-scale study involving Tg mice that conditionally overexpress mutant (A53T) a-synuclein, deletion of the LRRK2 gene was found to provide protection from broad-sweeping damage to the forebrain. The effects of chronic inhibition of LRRK2 should be closely evaluated in preclinical models to identify issues that will have to be addressed in clinical trials. There have been no described phenotypes associated with heterozygous knockdown of LRRK2 in mice or rats. Homozygous LRRK2 KO animals might mimic the effects of a perfect drug that achieves complete ablation of the target at all times. However, protein kinases can also serve in important functions that are independent of kinase activity. For example, KO of the CamKII protein impairs presynaptic plasticity and vesicle docking at the synapse, whereas kinase inhibition of CamKII does not impair these functions.