The development of inhibitors of leucine‐rich repeat kinase 2 (LRRK2) as a therapeutic strategy for Parkinson's disease: the current state of play

Current therapeutic approaches for Parkinson's disease (PD) are based around treatments that alleviate symptoms but do not slow or prevent disease progression. As such, alternative strategies are needed. A promising approach is the use of molecules that reduce the function of leucine‐rich repeat kinase (LRRK2). Gain‐of‐function mutations in LRRK2 account for a notable proportion of familial Parkinson's disease cases, and significantly, elevated LRRK2 kinase activity is reported in idiopathic Parkinson's disease. Here, we describe progress in finding therapeutically effective LRRK2 inhibitors, summarising studies that range from in vitro experiments to clinical trials. LRRK2 is a complex protein with two enzymatic activities and a myriad of functions. This creates opportunities for a rich variety of strategies and also increases the risk of unintended consequences. We comment on the strength and limitations of the different approaches and conclude that with two molecules under clinical trial and a diversity of alternative options in the pipeline, there is cause for optimism.


| PARKINSON'S DISEASE (PD)
Parkinson's disease (PD) is a movement disorder that has been observed in humans since ancient times but was first studied as a distinct condition by the English physician James Parkinson, who published his 'essay on the shaking palsy' in 1817 (Parkinson, 1817).
Today, PD is recognised as the second most common neurodegenerative disease, affecting up to 10 million individuals worldwide, with a lifetime risk estimated around 2% (Elbaz et al., 2002;Gan-Or et al., 2015). Despite so many years of study and so many affected individuals, PD is an incurable condition.
Of these, resting tremor is often the first symptom to present and is perhaps the most distinctive of the condition when compared with other motor disorders. It is now well established that these oftentimes debilitating symptoms are due to the degeneration of dopaminergic neurons within the substantia nigra pars compacta, which leads to a depletion of the neurotransmitter dopamine (Postuma et al., 2015).
However, PD is not limited to motor dysfunction (Pfeiffer, 2016). Several neuronal networks are affected, causing patients to also suffer from psychiatric and dysautonomic problems, and as the disease progresses, cognitive problems can develop (Pfeiffer, 2016). More common nonmotor symptoms include anosmia, depression and constipation. Because these nonmotor symptoms can occur more than 10 years prior to the onset of motor symptoms and diagnosis, prodomal PD is currently a research area attracting great attention.
From a pharmacological perspective, it is most important to emphasise that existing PD treatments only alleviate symptoms and do not delay or halt progression of the disease (AlDakheel et al., 2014). Current therapeutic approaches for PD are summarised in Table 1. The majority of these treatments aim to restore motor function by re-establishing dopamine signalling (AlDakheel et al., 2014) and are derived from pioneering studies by Arvid Carlsson who demonstrated that injection of the dopamine biosynthesis precursor L-dehydroxyphenylalanine (L-DOPA) rescued motor function in animal models (Carlsson et al., 1957). It took the best part of a decade to find a suitable dosing and delivery regimen for human PD patients (Cotzias et al., 1967), but oral L-DOPA and conceptually similar dopamine receptor agonists have been the mainstay of PD treatment ever since. For any drug, such enduring popularity would usually indicate that it is a highly efficacious and problem-free compound, and it is certainly true that after starting L-DOPA PD patients often experience a honeymoon period lasting several years, where motor symptoms may be in almost complete abeyance. But symptoms inevitably return as the disease progresses and the dopamine replacement therapies themselves can cause some fairly serious side effects, not least L-DOPA-induced dyskinesias (Tran et al., 2018). Furthermore, these treatments do not address the majority of nonmotor symptoms, which typically do not involve loss of dopamine neurones (Chaudhuri et al., 2006;Postuma et al., 2015). PD patients will require additional medications to treat these symptoms, such as selective 5-HT (serotonin) reuptake inhibitors for the treatment of depression (AlDakheel et al., 2014). But perhaps the most fundamental limitation of dopamine receptor agonists is that these compounds do not affect the progression of the underlying neurodegeneration.
The need for new therapeutic strategies for PD is clear. Moreover, because the world's population is ageing, it is also clear that this need is growing (Reeve et al., 2014). Over recent years, a great deal of research has therefore been directed towards the genetic causes of PD, in the hope that these will point the way to novel drug targets.
Therapeutic strategies that target proteins involved in PD aetiology offer great potential. In principle, such approaches may not only arrest PD progression but also may even be sufficient to prevent disease in the first place.
The rest of this article deals with the current state of play regarding pharmacological inhibitors of leucine-rich repeat kinase 2 (LRRK2), the product of a gene that is strongly implicated as both a T A B L E 1 Overview of the current treatments for Parkinson's disease

Target Strategy Treatment
Motor symptoms (loss of dopaminergic neurons and striatal dopamine depletion) Dopamine replacement Dopamine precursor (e.g. L-DOPA), dopamine agonists (e.g. ropinirole) Inhibition of dopamine degradation MAO-B inhibitors (e.g. rasagiline) and catecholamine-O-methyltransferase inhibitors (e.g. entacapone) Glutamatergic activity decrease (promote dopamine synthesis) and other neurotransmitters cause of familial PD and an important risk factor contributing towards the more common sporadic form of the condition (Kluss et al., 2019;Nalls et al., 2014;Paisan-Ruiz et al., 2004;Zimprich et al., 2004). For reasons we outline below, targeting LRRK2 may be fruitful not only for individuals with pathogenic LRRK2 mutations but potentially also for the majority of PD patients whose LRRK2 status is likely to be nor-  (Maegawa et al., 2009;Migdalska-Richards et al., 2017). Direct targeting of α-synuclein is also being pursued. For example, the aggregation inhibitor anle138b (Wagner et al., 2013) is in clinical trials (NCT04685265; clinicaltrials.gov), as is the c-Abl inhibitor nilotinib (NCT02281474; http://clinicaltrials.gov), which is expected to remove α-synuclein by promoting autophagy (Karuppagounder et al., 2014). It is hoped that these compounds-in particular when allied with methods to identify at-risk individuals, such as genetic screens and research into prodromal PD, to allow earlier intervention-will become the next generation of PD treatments that are so desperately needed.

| LEUCINE-RICH REPEAT KINASE 2
Leucine-rich repeat kinase 2 protein (LRRK2) is the product of the LRRK2 gene. LRRK2 was originally discovered independently by two groups as the gene located within the so-called PARK8 locus-a region on chromosome 12 associated with autosomal dominant late-onset parkinsonism (Paisan-Ruiz et al., 2004;Zimprich et al., 2004). Mutations in LRRK2 are now accepted as one of the most common causes of familial PD (Kluss et al., 2019;Nalls et al., 2014). Pathogenic LRRK2 mutations give rise to phenotypes that, at the individual level, are clinically indistinguishable from those of sporadic PD, including age of onset, symptoms and patterns of neurodegeneration in nigrostriatal dopaminergic neurons. Post-mortem LRRK2 brains also display accumulation of Lewy bodies-proteinaceous aggregates typically enriched in α-synuclein and ubiquitin-that are a pathological hallmark of PD (Kalia et al., 2015). Curiously, LRRK2 brains also frequently display Tau protein pathology; although the gene encoding Tau, MAPT has been strongly implicated in sporadic PD via genome-wide association studies, the accumulation of Tau is usually associated with other neurodegenerative conditions including Alzheimer's disease (Iqbal et al., 2010;Nalls et al., 2014). Thus, LRRK2 may sit at the top of neurodegenerative cascades involving both α-synuclein and Tau. Complicating matters further, TAR DNA-binding protein 43 kDa (TDP-43) pathology, more usually associated with amyotrophic lateral sclerosis, has been reported in a small number of LRRK2 brains (Ling et al., 2013;Wider et al., 2010), as have post-mortem brains with no apparent proteinopathies (Gaig et al., 2007;Takanashi et al., 2018).
Importantly, genome-wide association studies (GWAS) have also shown that LRRK2 has a major influence on an individual's risk of developing sporadic PD (Do et al., 2011;Lill et al., 2012;Satake et al., 2009;Simon-Sanchez et al., 2009). Thus, with gene variants of different severity either causing familial PD or contributing risk of sporadic PD, LRRK2 is considered a pleiomorphic risk locus for this condition (Cookson, 2015) and it is clear that the study of LRRK2 may have wider implications for other forms of neurodegenerative disease.
Adding an extra layer of complexity, genome-wide association studies have also implicated LRRK2 as a factor in the genetic risk of Crohn's disease (Barrett et al., 2008;Franke et al., 2010) and leprosy (Wang et al., 2015;Zhang et al., 2009), whereas other studies have linked LRRK2 to tuberculosis (Hartlova et al., 2018). LRRK2 has also been associated with the risk of cancer but with conflicting results.
Some studies find that the pathogenic LRRK2 variant G2019S increases the risk of cancer (Agalliu et al., 2015;Inzelberg et al., 2012;Saunders-Pullman et al., 2010), but others do not (Allegra et al., 2014;Ruiz-Martinez et al., 2014). Interestingly, loss of LRRK2 appears to increase risk of lung cancer (Lebovitz et al., 2021), an observation that may be germane to the long-term use of LRRK2 inhibitors (see later in this article). In any case, the observations linking LRRK2 to Crohn's disease, tuberculosis and leprosy are curious, because these conditions share some clear similarities with each other (e.g. they are all chronic inflammatory conditions with other genetic loci in common), but similarities with PD are less obvious. Whether LRRK2 dysfunction has a common role in all four conditions or whether the molecular mechanism in PD is distinct remains to be seen, but it is interesting to note that although that no shared pathogenic variants have been identified, the protective R1398H variant appears to be common to risk of Crohn's disease and PD (Hui et al., 2018;Ross et al., 2011;Tan et al., 2010). As such, it is easy to imagine that, at the very least, research into LRRK2 in PD could also have important implications for these non-neurodegenerative conditions. LRRK2 itself is a large (2527 amino acid) and very unusual protein that belongs to the ROCO family of GTPases and contains multiple functional domains (Gotthardt et al., 2008;Paisan-Ruiz et al., 2004;Zimprich et al., 2004). The structure of LRRK2 is depicted in Figure 1.
In addition to the four protein-protein interaction domains, LRRK2 contains domains conferring both serine-threonine kinase activity and GTPase activity, the latter via the combination of a Ras of complex proteins (Roc) domain and a C-terminal of Roc (COR) domain, which are often collectively referred to as a Roc-COR tandem domain (Gotthardt et al., 2008). A number of pathogenic mutations have been found, all of which encode single amino acid substitutions with the Roc, COR and kinase domains, although a relatively common variant associated with increased risk encodes a G2385R amino acid substitution with the WD40 domain, at the C-terminus of the protein (Berwick et al., 2019;Kluss et al., 2019). Typically, LRRK2 exists in cells as both homodimers and monomers, with dimerisation suggested to regulate LRRK2 subcellular location and function (Ho et al., 2018).
Thus, LRRK2 dimerisation and membrane recruitment have been suggested to play a role in kinase activation, perhaps in a mechanism governed by the membrane-bound small GTPase Rab29, which has been shown to bind LRRK2 and increase its kinase activity in cells (Liu et al., 2018;Purlyte et al., 2019).
The biological function of LRRK2 within cells remains unclear, largely because this protein has been linked to a multitude of different cell biological processes and a clear front runner has yet to emerge from the pack. For example, LRRK2 has been linked to effects on cell signalling, microtubule and other cytoskeletal dynamics, ciliogenesis, autophagy, endocytosis, synaptic vesicle trafficking and the biology of numerous cellular organelles such as mitochondria, the trans-Golgi network and lysosomes (reviewed by Berwick et al., 2019). Complicating things further, there is no real agreement over the type of cells in which LRRK2 dysfunction is most important for PD, with published data supporting many neural cell types and also immune cells (Berwick et al., 2019). In principle, therefore, LRRK2 dysfunction could cause either degeneration of dopaminergic neurons via a cell autonomous process or via a noncell autonomous neuroinflammatory mechanism-or potentially via both.
Nonetheless, within the context of this review, the uncertainty regarding LRRK2 function is not of great importance. This is because when it comes to the effects of LRRK2 pathogenic mutations on LRRK2 itself, there is clear consensus on two key points: 1. All pathogenic LRRK2 mutations increase both LRRK2 autophosphorylation and the LRRK2-mediated phosphorylation of certain Rab small GTPases in cells. Whether all these mutations technically increase LRRK2 kinase activity is debatable, because only the common G2019S kinase domain mutation reproducibly increases kinase activity in biochemical assays performed in vitro (Greggio & Cookson, 2009). Furthermore, the G2385R risk variant is actually reported to reduce kinase activity in these experiments (Rudenko et al., 2012). However, this argument is ultimately a semantic one-the effect of all PD-causing or PD risk mutations in LRRK2 is to increase substrate phosphorylation in vivo (Berwick et al., 2019;Kluss et al., 2019). To most intents and purposes, therefore, they can be considered to increase kinase activity; at the very least, they are gains of kinase function.   progressed to such an extent that clinical trials of one such compound in sporadic PD patients were announced only a matter of weeks after the first evidence that LRRK2 kinase function is actually increased in this cohort (Denali Therapeutics, 2018;Di Maio et al., 2018). In short, the progress of LRRK2 drug development relative to basic research into this protein is quite remarkable.
In the subsequent three sections of this article, we summarise the development of conventional LRRK2 kinase inhibitors, while also identifying potential risks and challenges in the road ahead. For more detailed descriptions of the chemistry of these molecules, we point the reader to the following more specialised reviews (Domingos et al., 2019;Hatcher et al., 2017). Later in the manuscript, we branch out to alternative pharmacological strategies to reduce the function of this protein, which together create a rich palette of potential therapeutic options.

| ATP-COMPETITIVE LRRK2 KINASE INHIBITORS
The majority of molecules targeting LRRK2 have been conventional or 'Type I' kinase inhibitors, which compete with ATP for binding to the ATP-binding pocket within the kinase domain (Hatcher et al., 2017;Lu et al., 2020). This story inevitably began with compounds that were a long way from candidate drugs ready to enter clinical trials, but over the years, these molecules have evolved to have superior pharmacological profiles and one compound is now in clinical trials.
The first compounds reported to inhibit LRRK2 kinase activity were either broad-range kinase inhibitors able to inhibit LRRK2 with high potency but very low selectivity, for example, staurosporine, which has IC 50 values for LRRK2 of 1-2 nM (Anand et al., 2009;Covy & Giasson, 2009), or a series of Rho kinase inhibitors that were found to inhibit LRRK2 and Rho kinases with similar high nanomolar potencies (Nichols et al., 2009). Fortunately, these compounds were swiftly superseded by two more selective molecules, LRRK2-IN-1 (Deng et al., 2011) and CZC-25146 (Ramsden et al., 2011), which were identified specifically as LRRK2 inhibitors via kinase inhibitor library screens. Both these compounds are reported to inhibit LRRK2 with IC 50 s in the low nanomolar range and exhibit decent selectivity; for example, LRRK2-IN-1 was found to display greater than 90% inhibition towards only 12 out of 440 kinases using an Ambit KINOMEscan that measures the binding of drugs to a panel of kinases as a proxy for kinase inhibition (Deng et al., 2011;Karaman et al., 2008). had the critical failing of being unable to cross the blood-brain barrier, which is clearly a fundamental requirement of any pharmaceutical aimed at treating neurodegenerative disease. Somewhat surprisingly, the structurally similar but larger compound TAE684 did enter the brain in rodents, but it was found to be ineffective as a LRRK2 inhibitor in this organ. GSK2578215A, which is appreciably smaller than all three aminopyrimidine molecules, also crossed the rodent blood-brain barrier with little effect on LRRK2.
Given that LRRK2-IN-1, CZC-25146 and TAE684 are all similar in structure, it is unsurprising that a great many of the next generation of LRRK2 kinase inhibitors were smaller aminopyrimidine molecules that were developed with the hope of retaining potency and selectivity for LRRK2, while facilitating access to the brain. A number of such compounds were published in the next few years including HG-10-102-01 Choi et al., 2012) and a series of compounds-G1023, GNE-7915, GNE-0877 and GNE-9605-made by Genentech (Estrada et al., , 2014. All these compounds have potencies in the low (or sub-molar) nanomolar range and excellent selectivity and are able to enter the brains of mice and inhibit LRRK2 in vivo. The Genentech compounds were also tested in rats and cynomolgus monkeys, where they were found to have good pharmacokinetic and drug metabolism profiles (Estrada et al., , 2014. Thus, LRRK2 kinase inhibitors moved from purely in vitro tools to compounds that could be used to study LRRK2 in the brains of animal models. Over the next few years, a number of other structurally distinct LRRK2 kinase inhibitors (i.e. neither aminopyridines nor arylbenzamides) have been reported in the scientific literature, as these compounds continue to evolve and improve. Structural categories include pyrrolopyrimidines such as Pfizer's PF-06447475 (Henderson et al., 2015), indazoles such as Merck's MLi-2 (Fell et al., 2015) and thiophenes, also from Merck (Greshock et al., 2016).
In addition, further types of organic molecules that have not been described in the scientific literature are the subject of patents for use as LRRK2 inhibitors, such as macrocyclic derivatives from GlaxoSmithKline and Ipsen Pharma (Ding & Ren, 2020). For a detailed review of the chemistry, we refer the reader to existing work in this area (Domingos et al., 2019;Hatcher et al., 2017).
One recent development that is worthy of mention, however, is the discovery of ATP-competitive LRRK2 inhibitors that exhibit selectivity for the pathogenic G2019S LRRK2 variant over wild-type LRRK2 (Garofalo et al., 2020). Via compound library screening followed by in silico docking analysis and systematic chemical modifications, Garofalo et al. (2020) identified a number of indazole molecules displaying selectivity for G2019S. The most selective compound was reported to have a >300Â greater potency in cellular assays and >200Â greater potency in vitro. As the authors themselves rightly observe, this 'is remarkable, given that G2019S-LRRK2 differs from WT-LRRK2 only by a single amino acid' (Garofalo et al., 2020). Unfortunately, this molecule is unlikely to be developed further as it is unable to cross the blood-brain barrier, but we note that another G2019S-selective indazole compound identified in the same work has since been shown to inhibit LRRK2 kinase activity in peripheral blood mononuclear cells from homozygous G2019S LRRK2 patients, but not in cells from wild-type controls (Bright et al., 2021). Thus, even though these molecules may only be suitable for G2019S LRRK2 PD patients, there is clearly a lot of potential in this project and we await further developments.
Taking this section as a whole, it is evident that there are a multitude of promising ATP-competitive LRRK2 kinase inhibitors in the F I G U R E 2 Chemical structures of the leucine rich repeat kinase 2 (LRRK2) inhibitors described in this review. The chemical structures of the selective LRRK2 inhibitors mentioned by name in this article are shown. Structures were generated using ACD/ChemSketch (ACD Labs) with structures cleaned and most likely tautomeric forms selected pipeline. As therapeutics for the treatment of PD, most of these compounds will inevitably fall by the wayside, but with a diversity of chemical structures available, there are good grounds for optimism. A variety of chemical structures means a greater range of pharmacological properties and a better chance of finding a drug that is potent, selective and free from adverse effects. Later in this article, we will focus on LRRK2 inhibitors that are or have been used in human clinical trials, but before we get to that point, we will first comment on certain areas of concern that have been raised in preclinical studies.  (West, 2017). Although this can in principle be addressed by keeping the therapeutic dose within a tight window, this is potentially a serious problem, as LRRK2 has hundreds of reported interacting partners and is believed also to act as a scaffolding protein, not just as a kinase (Berwick et al., 2019;Lewis & Manzoni, 2012).
The greatest worry perhaps inevitably comes from in vivo studies of Lrrk2 knockout mice and rats. Mice that are double knockout for both Lrrk2 and its paralog Lrrk1 display neurodegeneration (Giaime et al., 2017), yet because animals that only lack Lrrk2 have no overt neurological phenotypes (for a detailed review of LRRK2 animal models, see Seegobin et al., 2020), this concern does not relate to the brain. Instead, Lrrk2 knockout mice and rats display an unusual vacuolated phenotype in their kidneys, liver and lung (Baptista et al., 2013;Ness et al., 2013;Seegobin et al., 2020;Tong et al., 2010). None of these alterations appear to have any significant impacts on the health of these laboratory animals, but the lung phenotype is a particular concern given that respiratory problems are very common in PD (van de Wetering-van Dongen et al., 2020), so any further compromise to lung function could ultimately lead to a worsened clinical outcome.
This concern was intensified following an investigation into the effect of two of the Genentech aminopyridine inhibitors, GNE-7915 and GNE-0877, in cynomolgus monkeys (Fuji et al., 2015). Worryingly, treatment of these animals with either compound caused vacuolated Type II pneumocytes, which was very consistent in appearance with the lung phenotype of Lrrk2 knockout mice and rats. Subsequently, a similar effect was observed in mice treated with the indazole LRRK2 inhibitor MLi-2 developed by Merck (Fell et al., 2015). However, a recent dedicated toxicological study using GNE-7915, MLi-2 and the Pfizer LRRK2 inhibitor PFE-360 in cynomolgus has indicated that these fears may be exaggerated  reported no increased frequency of 'lung, liver, kidney, cardiovascular system, nervous system, immunity or cancer' phenotypes (Whiffin et al., 2020). However, these results are to some extent contradicted by a more recent report associating LRRK2 loss-of-function with a type of lung cancer that appears to be Type II pneumocyte in origin (Lebovitz et al., 2021). This group had previously reported lowered LRRK2 expression in lung adenocarcinoma (Lebovitz et al., 2015) but have since looked at the connection in greater detail. They report that lower levels of LRRK2 are associated with decreased patient survival and that Lrrk2 KO mice are more prone to developing lung adenomas when exposed to urethane, a cigarette smoke carcinogen (Lebovitz et al., 2021). These results are concerning, but it is probably pertinent to observe that lower LRRK2 expression is most strongly correlated with lung cancer in patients who are smokers and that there is currently little evidence to suggest decreased LRRK2 expression is sufficient to cause lung cancer (Lebovitz et al., 2021). Thus, LRRK2 inhibitors may not be suitable for current or recent smokers, but otherwise, the risk of lung cancer appears small.
A further complication is how to reliably measure the extent of LRRK2 kinase inhibition within the brains of patients, such that clinicians can establish an optimal dose and monitor treatment progression in their patients. Existing biomarkers, such as Rab protein phosphorylation (Eyers, 2018;Steger et al., 2016) or LRRK2 autophosphorylation on serine-1,292 (Sheng et al., 2012), which can be measured with phospho-specific antibodies, appear to be faithful proxies for LRRK2 kinase function in brain tissue from experimental animals, but these assays cannot easily be replicated on humans. Less invasive procedures will be needed, for example, proteins or metabolites present in patient blood or urine samples, although correlating these readouts with LRRK2 kinase inhibition in brain tissue is going to be a tough ask. With this in mind, we note that certain bis(monoacylglycerol)phosphates (BMP), phospholipids that are considered specific to lysosomes, have been reported to be increased in the urine of LRRK2 PD patients (Alcalay et al., 2020) and also decreased in the urine of both cynomolgus monkeys treated with LRRK2 kinase inhibitors and Lrrk2 knockout mice (Fuji et al., 2015). As such, urine BMP levels may be an ideal non-invasive biomarker. It should be noted that sporadic PD patients do not appear to have elevated urine BMP levels compared with healthy controls (Alcalay et al., 2020). However, the same study found that in PD patients urine BMP levels appear to be inversely correlated with cognitive performance irrespective of LRRK2 status.
Thus, the overall difference in urine BMP levels between sporadic PD and control cohorts is likely a consequence of heterogeneity within each group and that at an individual level at least, urine BMP levels may be a faithful biomarker of disease progression (Alcalay et al., 2020). The full potential of BMP remains to be established, but unsurprisingly, this biomarker is already being used in clinical trials of LRRK2 inhibitors (Denali Therapeutics, 2020).

| LRRK2 KINASE INHIBITORS IN CLINICAL TRIALS
Although the field is very crowded, to date just two LRRK2 kinase inhibi- So what is the situation with other LRRK2 kinase inhibitors? At present, with Denali over 2 years into testing, it can be assumed that other companies are waiting for the full results of DNL-151's Phase 1 and 2 trials, before deciding whether to press ahead with their own compounds.
Watching keenly will be Cerevel, who have a LRRK2 inhibitor listed as one of their pipeline drugs on their website (Cerevel, 2021) and have acquired the rights to Pfizer's neurological disease compounds. Presumably, this includes the pyrrolopyrimidine compound PFE-360, which, as mentioned above, was found to have a no-effect dose at which it inhibited LRRK2 in monkey brains but did not induce any lung phenotype . Also paying close attention will be Merck, whose MLi-2 compound also showed signs of a no-effect dose, albeit within a much narrower window than was observed for PFE-360 . Perhaps importantly, this drug has an especially high potency, with an IC 50 in the high picomolar range (Fell et al., 2015).
Clearly, Cerevel and Merck will not be the only interested parties. Thus, although the world plays wait and see, we can at least do so safe in the knowledge that if DNL-151 fails, there are plenty of alternative options.

| OTHER THERAPEUTIC APPROACHES TO REDUCE LRRK2 FUNCTION
Although there are plenty of alternative LRRK2 kinase inhibitors to DNL-151, there are also plenty of therapeutic strategies for reducing the function of LRRK2 that do not involve classical ATP-competitive kinase inhibition (Figure 3). This smorgasbord of approaches broadens the targets of potential therapeutic drugs far beyond the kinase domain and includes molecules that affect LRRK2 conformation, protein-protein interactions, GTPase activity and even LRRK2 mRNA.
We describe these molecules in the following sections.  (Li et al., 2014). In addition to blocking LRRK2 GTP binding in cells, these compounds were also able to reduce LRRK2 autophosphorylation, thereby confirming the secondary effect of inhibiting LRRK2 kinase activity (Li et al., 2014). Demonstrating their potential, both compounds promoted the survival of cultured neurons, whereas compound 68 also reduced LPS-induced neuroinflammation in a mouse model (Li et al., 2014). In subsequent work, this group reported an optimised analogue of compound 68, FX2149, which is more brain penetrant and correspondingly more potent in the same mouse model . To the best of our knowledge, only one other group has published with these compounds. Fascinatingly, Blanca Ramirez et al. (2017) (Biogen, 2021).

| LRRK2 GTP-BINDING INHIBITORS
Antisense oligonucleotides are single-stranded DNA molecules of varying lengths that are designed to bind specifically to an mRNA molecule of interest via base pair complementarity, typically causing protein synthesis from that mRNA to be blocked (Chery, 2016). The nucleotides within antisense oligonucleotides are typically chemically modified to prevent degradation by nucleases, while also facilitating uptake across F I G U R E 3 Sites of action of the leucine rich repeat kinase 2 (LRRK2) inhibitors described in this review. The various types of therapeutic strategy that are under development or have been proposed for targeting LRRK2 are summarised in the figure. Proposed strategies are written in italics. For simplicity, non-selective kinase inhibitors (e.g. staurosporine and ponatinib) are omitted. ATP-competitive kinase inhibitors are listed by chemical class; note that the categorisation of DNL-151 and DNL-201 is presumed on the basis of patents held by the manufacturer the plasma membrane and enhancing binding to RNA (Silva et al., 2020). Importantly, because the mode of delivery used by Biogen and Ionis is intrathecal injection, that is, directly into the CSF, this strategy obviates the need for BIIB094 to cross the blood-brain barrier. The use of antisense oligonucleotides to target LRRK2 still carries many of the same concerns as use of LRRK2 kinase inhibitors and intrathecal delivery is a fairly invasive means of drug delivery, so there are evidently downsides to this approach. Nonetheless, it is important to note that the blood-brain barrier works in both directions, so intrathecal delivery will greatly reduce the chance of adverse effects in the periphery, not least the now familiar lung phenotype.
Fascinatingly, antisense oligonucleotides are already being used in the clinic for the treatment of spinal muscular atrophy, a rare neurodegenerative disease of motor neurons (Kolb & Kissel, 2015). In this case, the antisense oligonucleotide, which is referred to as nusinersen, is not acting to repress its target mRNA but to prevent the binding of a splicing factor, thereby promoting alternative splicing (Chiriboga et al., 2016). It is also worth observing that other antisense oligonucleotides are under investigation for use in Huntington's disease, which, unlike ISIS3139, specifically target mRNA produced by the disease-causing allele and not the wild type (Silva et al., 2020). This is potentially significant, because this strategy should in principle avoid reducing all huntingtin expression, thereby protecting the essential functions of this protein. Theoretically, PD patients with LRRK2 mutations (although not sporadic PD patients) could be treated in a similar way. However, the ability to select between mutant and wild type that differ only by a single base change is going to be considerably more challenging than mutations like the expanded CAG repeats causing Huntington's disease.

| ALLOSTERIC INHIBITORS OF LRRK2 KINASE ACTIVITY
Allosteric kinase inhibitors can be defined as compounds that inhibit the enzymatic activity of a kinase by binding the kinase outside its ATP-binding site and thus reducing kinase activity in an ATP non-competitive manner . In principle, the GTPase inhibitors described above, which bind the LRRK2 Roc domain and indirectly reduce LRRK2 kinase activity, meet this definition, but we have chosen to classify them separately, because kinase inhibition is not their primary objective. In practice, allosteric inhibitors are more likely to bind the kinase domain. There are advantages and disadvantages with any therapeutic strategy, but allosteric compounds have the potential to increase selectivity, by targeting regions that are less conserved than the ATP-binding pocket, although designing such molecules is more challenging.
Recently, Schaffner et al. (2019)  AdoCbl is considerably less potent than conventional LRRK2 inhibitors, although the authors suggest that AdoCbl is likely to be highly selective for LRRK2 as this molecule is structurally dissimilar to any other known kinase inhibitor. This last point clearly needs to be tested empirically. In any case, this molecule was able inhibit LRRK2 kinase activity and neurotoxicity in both in vitro and in vivo models carrying pathogenic LRRK2 variants (Schaffner et al., 2019). The exact location within LRRK2 to which AdoCbl binds was not identified but based on enzymological data can be assumed to be distinct site(s) within the LRRK2 kinase domain but outside the ATP-binding pocket.
The authors also report that AdoCbl disrupts LRRK2 dimerisation and promotes the proteolytic degradation of this protein. How these observations fit together mechanistically is not yet known, but because LRRK2 dimers have greater kinase activity than LRRK2 monomers, it is tempting to speculate that AdoCbl may be inhibiting LRRK2 kinase by preferentially binding to dimeric LRRK2 and inducing monomerisation. This story is clearly at its infancy and therapeutic potential of AdoCbl is unclear, but its identification as a new allosteric inhibitor of LRRK2 may at least open the door to the discovery of other similar molecules.
One strategy for the development of allosteric kinase inhibitors of LRRK2 that may be slightly less useful is the development of 'Type II' kinase inhibitors that target the DFGψ motif (sometimes abbreviated to DFG) of the kinase domain activation loop Schmidt et al., 2019). (Type II allosteric inhibitors in general bind to kinase domain activation loops; DFGψ is a conserved sequence of amino acids with the activation loop, where D, F and G are aspartic acid, phenylalanine and glycine and ψ represents a hydrophobic residue.) Importantly, Type II kinase inhibitors are only able to bind kinase domain activation loops that are in an inactive 'DFG-out' conformation. In doing so, these compounds prevent the conformational change to an active 'DFG-in' conformation that is required to allow access of ATP to the adjacent ATP-binding pocket (Schmidt et al., 2019). As such, Type II kinase inhibitors block kinase activity by gating access to ATP, rather than by competing with ATP. The limitation of developing Type II kinase inhibitors that target the DFGψ motif for LRRK2 is that this is the location of the amino acid substitutions encoded by two pathogenic mutations, G2019S and I2020T. Significantly, both G2019S and I2020T stabilise the LRRK2 activation loop in the active DFG-in conformation (Liu et al., 2013;Ray et al., 2014). Given this, Type II kinase inhibitors that target the DFGψ motif of LRRK2 can be expected to be less effective against these pathogenic variants than against wild-type LRRK2.  (Liu et al., 2013;Ray & Liu, 2012). Two of these inhibitors target the activation loop hinge region, rather than the DFGψ motif, and these compounds displayed little difference in potency when tested against wild-type LRRK2 and the G2019S and I2020T variants (Ray & Liu, 2012   and rats (Daher et al., 2015;Nguyen et al., 2020). These data are supported by observations of an equivalent effect of introducing an artificial kinase inactivating mutation into the overexpressed LRRK2 protein Nguyen et al., 2020;Tsika et al., 2015). Of note, mutations that perturb LRRK2 GTPase function, by either blocking binding to GTP or elevating the rate of GTP hydrolysis, cause a similar reduction in neurodegeneration, thus supporting the use of LRRK2 GTPase inhibitors as an alternative strategy . So, what about other models of neurodegeneration? An alternative experimental system is viral transduction of α-synuclein. Unfortunately, results using these technologies are mixed. In the first study, Daher et al. (2015) found that PF-06447475 reduced α-synucleininduced neurodegeneration in rats. By contrast, Andersen et al. (2018) were unable to show neuroprotection or rescue of motor symptoms using PFE-360, although they did report a reversal of an aberrant subthalamic firing pattern that the authors describe as resembling alterations seen in PD. However, in a follow-up report describing the effects of longer term PFE-360 treatment, the same group found no improvement in subthalamic nucleus firing, although a mild improvement in motor function was observed (Andersen et al., 2019). Taking their observations together, the authors conclude that their findings 'do not strongly support LRRK2 inhibition for the treatment of PD' (Andersen et al., 2019).
There are many potential reasons for the disagreement, such as the different viral systems used, the degree of α-synuclein expression achieved and the treatment times used. Although structurally very similar (Figure 2), the LRRK2 inhibitors used could also be a factor. It is also worth observing that even if the results were in complete agreement, they would still need to be treated with caution, because the rate at which α-synuclein accumulates in the brains of PD patients cannot be anything like as fast as that achieved by viral transduction.
With this in mind, recent observations made using the pesticide rotenone to induce dopaminergic neurodegeneration in mice may be insightful (Rocha et al., 2020). This study measured both rotenoneinduced dopaminergic cell death and the preceding changes to endolysosomal and autophagic systems that mirror defects seen in the surviving neurons from patients with idiopathic PD (Rocha et al., 2020). Remarkably, the Type I LRRK2 kinase inhibitor PFE-360 reduced the effects of rotenone on all these measures. It must be stressed that these data are only directly transferable to humans who develop PD following acute rotenone exposure, but these results are nonetheless encouraging. Chronic exposure to pesticides including rotenone is an established environmental cause of increased PD risk and the endolysosomal and autophagic defects that were rescued are consistent with idiopathic PD. As such, these data point to LRRK2 inhibitors having use in idiopathic PD and LRRK2 PD.
Despite the widespread use of the neurotoxins 6-hydroxydopamine (6-OHDA) and 1-methyl-4-phenyl-1,2,3,6tetrahydropyridine (MPTP) to induce dopaminergic damage in rodent PD models (Schober, 2004), we are aware of only one report of an attempt to rescue the phenotype of animals treated with these compounds using LRRK2 inhibition. This study used a low dose of MPTP that was not toxic in wild-type animals but induced significant impairments in transgenic mice that overexpress either wild-type LRRK2 or G2019S-LRRK2, with a greater effect seen in the G2019S animals (Arbez et al., 2020). Remarkably, treatment with LRRK2-IN-1 dramatically reduced dopaminergic cell death and also rescued motor deficits to a similar extent to L-DOPA (Arbez et al., 2020). There are clearly some caveats with these data, not least the choice of LRRK2 Research into the therapeutic potential of LRRK2 antisense oligonucleotides in preclinical models is also thin on the ground, although the use of LRRK2 antisense oligonucleotides is to some extent informed by the preclinical work using LRRK2 inhibitors described above (and vice versa). The first study using LRRK2 antisense oligonucleotides is that of Zhao et al. (2017), who investigated the effect of intracerebral ventricular injection of Lrrk2 antisense oligonucleotides in mice injected with synthetic α-synuclein fibrils. Somewhat surprisingly, given that the two different antisense oligonucleotides used only achieved a $50% reduction in Lrrk2 expression, both oligonucleotides markedly reduced α-synuclein aggregation and dopaminergic cell death and almost entirely rescued motor defects (Zhao et al., 2017). In the absence of independent replication, these observations must be treated with caution, but the idea that a modest knockdown of LRRK2 expression is sufficient to elicit a significant biological effect in the mouse brain is supported by the only other paper using antisense oligonucleotides targeting LRRK2 in animal models. In this report (Korecka et al., 2020), Korecka et al. (2019)  dimers are believed to have greater kinase activity that LRRK2 monomers, and the cycling of LRRK2 between these states is has been suggested to play a key role in the normal regulation of this protein (Berwick et al., 2019). Indeed, work performed in zebrafish models indicates that the deletion of the C-terminal WD40 domain is sufficient to prevent dimerisation, thereby reducing autophosphorylation and reducing the neurotoxicity of LRRK2 (Jorgensen et al., 2009).
Thus, disruption of LRRK2 dimerisation can be expected to, in turn, impair LRRK2 kinase activity, creating a mechanism that could be harnessed for the development of selective inhibitors of LRRK2 kinase. Such approaches have already been used successfully to identify compounds that prevent dimerisation of other proteins, for example, ST2825, a heptapeptide that blocks dimerisation of myeloid differentiation primary response 88 (MyD88) (Loiarro et al., 2007).
However, this may be a trickier task for LRRK2, because LRRK2 dimerisation appears to have several intramolecular interactions involved, with the WD40 and COR domains implicated in particular (Greggio et al., 2008;Guaitoli et al., 2016;Jorgensen et al., 2009). As such, efficient disruption of LRRK2 dimers may require a cocktail of molecules, each targeting a different protein interaction interface.
On a similar note, in addition to roles as a kinase and GTPase, LRRK2 has long been suggested to have a function as a scaffold protein, interacting with numerous other proteins to form dynamic signalling complexes at specific cellular localisations (Berwick et al., 2019;Lewis & Manzoni, 2012). As such, disruption of protein-protein interactions required for the pathogenic effects of LRRK2 may be fertile ground for drug discovery. Nonetheless, although many hundreds of LRRK2 interaction partners have been reported, when the expression levels of these proteins have been modulated in cellular models, very few appear to exert a robust influence on measures relevant to pathogenicity of PD-causing LRRK2 variants. An important exception, however, may be 14-3-3 proteins. 14-3-3 proteins bind LRRK2 in a manner dependent on the phosphorylation of LRRK2. Phosphorylation sites to which 14-3-3 proteins associate include Ser910 and Ser935, in the LRR domain (Nichols et al., 2010), Ser1444 in the Roc domain (Muda et al., 2014) and Thr2524 at the very C-terminus (Manschwetus et al., 2020). Although the interplay between these 14-3-3 binding events is complex and not yet fully understood, the functional consequences of LRRK2 interacting with these proteins are clear. Most notably, binding of 14-3-3 proteins to LRR domain residues stabilises LRRK2 in a monomeric and cytoplasmic form that can be assumed to have decreased kinase activity (depicted in Figure 3).
Consistently, pathogenic LRRK2 mutations have been reported to disrupt LRRK2-14-3-3 interactions (Lavalley et al., 2016;Nichols et al., 2010). Based on these and other observations, Soliman et al. (2020) have speculated that small molecule stabilisers of 14-3-3 protein-protein interactions, some of which have recently been described in the literature (Ballone et al., 2018), could be used to produce molecules that stabilise LRRK2-14-3-3 complexes. We await the next instalment of this story.
Finally, it is important to mention the impact two very recent breakthroughs in LRRK2 structural biology that have already provided data relevant to LRRK2 inhibitor development. In the first work, to prevent filament formation (Deniston et al., 2020). Interestingly, this latter observation mirrors those made for the LRRK2 GTPase inhibitors, compounds 68 and 60 (Blanca Ramirez et al., 2017).
In conclusion, therefore, with so many routes to achieving LRRK2 inhibition being taken, the prospects for developing new PD medications based around targeting LRRK2 are good. Where lead molecules and different strategies fall by the wayside, there are plenty of alternatives to take over. Nonetheless, with two molecules in clinical trials, the extent to which these treatments bring clinical benefits will undoubtedly be the hot topic over the next few years.
11.1 | Nomenclature of targets and ligands

CONFLICT OF INTEREST
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
Data sharing is not applicable to this article because no new data were created or analysed in this study.