Progress in LRRK2-Associated Parkinson’s Disease Animal Models

Mutations in the leucine-rich repeat kinase 2 (LRRK2) gene are the most frequent cause of familial Parkinson’s disease (PD). Several genetic manipulations of the LRRK2 gene have been developed in animal models such as rodents, Drosophila, Caenorhabditis elegans, and zebrafish. These models can help us further understand the biological function and derive potential pathological mechanisms for LRRK2. Here we discuss common phenotypic themes found in LRRK2-associated PD animal models, highlight several issues that should be addressed in future models, and discuss emerging areas to guide their future development.


INTRODUCTION
Parkinson's disease (PD) is a common, age-dependent, neurodegenerative disease with a lifetime risk of approximately 1.5% (Lees et al., 2009). The defining diagnostic features of PD include bradykinesia, resting tremor, rigidity, cognitive impairment, and psychiatric dysfunction (Goetz et al., 2008). Neuropathologically, PD is characterized by degeneration of dopaminergic (DA) neurons in the substania nigra pars compacta (SNpc) and the presence of Lewy Bodies (LBs) throughout the central nervous system (CNS) (Gelb et al., 1999;Dickson et al., 2009). Extensive human genetic studies have led to the discovery of several PD-causing genes, including leucinerich repeat kinase 2 (LRRK2) (Nalls et al., 2019). PD-causing mutations in LRRK2 occur in up to 41% of select patient populations and, as such, they represent the greatest known cause of heritable PD (Khan et al., 2005;Lesage and Brice, 2009). Mutations in LRRK2 are also found in sporadic PD, occurring at a rate of 1-3% (Gilks et al., 2005;Lesage et al., 2007;Paisán-Ruíz et al., 2008). Importantly, the clinical presentation of Parkinsonism in LRRK2 mutation carriers has been described as indistinguishable from sporadic PD patients (Adams et al., 2005;Kay et al., 2006). Given its importance in both sporadic and familial PD, understanding LRRK2 biology will assist in elucidation of common mechanisms of disease pathogenesis (for a more thorough review, see Kluss et al., 2019).
The LRRK2 gene encodes a large, 2,527-amino acid, 286-kDa, multi-domain protein belonging to the ROCO family (Zimprich et al., 2004). All ROCO proteins are characterized by a GTPase Ras-like G domain (Roc), followed in tandem by a C-terminal of Roc domain (COR) (Bosgraaf and Van Haastert, 2003;Marín, 2006). LRRK2 also contains a serine-threonine kinase domain, capable of phosphorylating both itself and a small group of substrates (West et al., 2005;Sheng et al., 2012;Steger et al., 2016). To date, most of the pathogenic mutations are clustered within the Roc, COR, or kinase domains and are found to alter LRRK2's biochemical activity (Chen and Wu, 2018).
Although LRRK2 activity has been linked to a diverse range of cellular processes (reviewed by Berwick et al., 2019), a large body of work suggests LRRK2 plays a key role in membrane trafficking along the endo-lysosomal pathway. These functions include synaptic vesicle endocytosis, degradation, and recycling of transmembrane receptors, anterograde trafficking of receptors from the trans-golgi network to lysosomes, and retromer-mediated transmembrane recycling. LRRK2-dependent regulation of these cellular processes may be associated with LRRK2's ability to bind and phosphorylate a cluster of Rab GTPases (Shin et al., 2008;Piccoli et al., 2011;MacLeod et al., 2013;Gómez-Suaga et al., 2014;Schreij et al., 2014;Pan et al., 2017;Steger et al., 2017;Pellegrini et al., 2018;Rivero-Ríos et al., 2019;Sheehan and Yue, 2019).
A diverse range of animal models have been developed which overexpress, knock-out, knock-down, or knock-in mutated and wildtype forms of the LRRK2 gene, for the characterization of LRRK2 biological and pathophysiological functions. In this review, we discuss common phenotypic themes found in LRRK2-associated rodent, Drosophila, Caenorhabditis elegans, and zebrafish animal models and highlight several new avenues toward development of future LRRK2 animal models.

RODENT LRRK2 MODELS
Rodent models have been widely used in the study of LRRK2 biology due to their genetic and neuroanatomical similarities to humans. Both mice and rats possess a mammalian homolog of LRRK2 which shares approximately 86-88% sequence identity to human LRRK2. Importantly, all residues affected by pathogenic mutations in humans are conserved in rodent LRRK2 (Langston et al., 2016). Rodents also possess a LRRK1 homolog, which shares ankyrin repeats (ANK), leucine-rich repeats (LRR), and Roc, COR, and kinase domains with LRRK2, but may also contain a WD40 domain that is still contested in the literature (Biskup et al., 2007;Civiero et al., 2012;Sejwal et al., 2017;Xing et al., 2017). The dopaminergic neuroanatomical pathways of rodents and humans are also highly similar, leading to the development of an array of sensitive behavioral tests in rodents that may correlate to dopamine loss in human PD (Meredith and Kang, 2006;Redgrave et al., 2010). Therefore, rodents represent an ideal candidate for genetic manipulations to investigate LRRK2 biology toward investigation of PD pathogenesis.

Rodent LRRK2 KO Models
LRRK2 knock-out (KO) models have been chiefly employed to investigate the physiological function of endogenous LRRK2 (Tables 1, 2). An emerging theme of LRRK2 KO rodent models is that the resulting phenotypes do not mimic LRRK2-assocated PD. Rodents born without LRRK2 exhibit no loss of dopaminergic (DA) neurons, demonstrate mild to no behavioral or locomotor defects, have limited neuropathology, and have unchanged dopamine synthesis (as measured by DOPAC and HVA) (Andres-Mateos et al., 2009;Lin et al., 2009;Tong et al., 2010b;Herzig et al., 2011;Hinkle et al., 2012;Daher et al., 2014).
Although no clear neuronal phenotypes have emerged, several studies have suggested that LRRK2 may fulfill key functions in peripheral tissues. Most notably, LRRK2 KO kidneys exhibit striking age-dependent changes in color and weight, along with ultrastructural abnormalities (Tong et al., 2010a(Tong et al., , 2012Herzig et al., 2011;Hinkle et al., 2012;Baptista et al., 2013). A darker coloration of LRRK2 KO kidneys has been observed as early as 2-3 months, while increases in weight are reported as early as 1-5 months in a sex-specific manner (Herzig et al., 2011;Hinkle et al., 2012;Tong et al., 2012;Ness et al., 2013;Boddu et al., 2015). Curiously, one group has reported a decrease in the weight of LRRK2 KO kidneys at 20-27 months in age (Tong et al., 2010b(Tong et al., , 2012. Electron microscopy (EM) analysis of KO kidneys has further revealed an increase in the number and size of lysosomes in epithelial cells of the renal cortex, starting at 4 months of age (Herzig et al., 2011;Hinkle et al., 2012;Tong et al., 2012;Baptista et al., 2013). These studies also found increases in lipid-containing vesicles or droplets and lipofuscin (potentially contained within lysosomes). Alongside ultrastructural abnormalities, multiple groups have reported changes in autophagic markers, LC3, and p62 and lysosomal markers, LAMP-1 and Cathepsin D, consistent with alterations in the autophagy-lysosomal pathway (Tong et al., 2010a(Tong et al., , 2012Hinkle et al., 2012). Changes in homeostatic parameters regulated by the kidneys have also been reported in LRRK2 KO animals, including increased diastolic blood pressure, decreased plasma and serum chloride, and decreased specific gravity of urine (Herzig et al., 2011;Baptista et al., 2013;Ness et al., 2013;Boddu et al., 2015;Fuji et al., 2015). Kidney dysfunction may be detected using the biomarker of kidney health known as lipocalin-2 (NGAL), which has also been observed to be reduced in plasma and urine (Ness et al., 2013;Fuji et al., 2015).
Several additional abnormalities have been reported in other peripheral organs in LRRK2 KO animals. Lungs of LRRK2 KO rats exhibit an increased number and size of type II alveolar cell lamellar bodies, a lysosome-derived secretory vesicle that stores surfactant (Herzig et al., 2011;Baptista et al., 2013;Miklavc et al., 2014). Indications of liver dysfunction such as increased AST, ALT, and cholesterol levels have also been reported (Ness et al., 2013;Baptista et al., 2013). To a lesser degree, abnormalities in the spleen, such as decreased size and changes in cellular composition, were also found (Ness et al., 2013).
A less-clear phenotype seen has been the disruption of α-synuclein homeostasis in LRRK2 KO rodent models. While several studies have shown that LRRK2 KO or kinase inhibition is protective against α-synuclein aggregation in the brain Herzig et al., 2011;Daher et al., 2014;Bae et al., 2018), others have reported that LRRK2 KO causes accumulation of α-synuclein in the kidneys (Tong et al., 2010a(Tong et al., , 2012. These seemingly conflicting findings may indicate tissue or model specific differences in α-synuclein metabolism. An exception to phenotypes seen in single LRRK2 KO models has been a recently developed LRRK1 and LRRK2 double-KO mouse that shows α-synuclein pathology, disruption of the autophagy-lysosomal pathway, and DA neurodegeneration in the CNS (Giaime et al., 2017). While the result suggests the possibility that LRRK1 compensates for the loss of LRRK2 in No changes in open field and rotarod (Lin).
Frontiers in Neuroscience | www.frontiersin.org LRRK2 KO rats are protective from α-synuclein injection (Daher). Increased body weight (Ness). LRRK2 KO primary neurons have slowed clathrin-mediated endocytosis and decreased SVs in vitro (Arranz). the brain, it raises the question as to whether PD mutations of LRRK2 could be implicated in the loss of LRRK1/LRRK2 functions. Future studies will have to further verify these findings and elucidate potential mechanisms.

Rodent Transgenic LRRK2 Models
Several rodent transgenic models have been generated to overexpress (OE) WT or pathogenic variants of LRRK2. These transgenic models have been developed using either insertions of LRRK2 cDNA or bacterial artificial chromosomes (BACs) (Tables 3-5).
An emerging theme in transgenic mouse models is that overexpression of pathogenic LRRK2 mutants, such as G2019S or R1441C, can induce PD-like phenotypes. Reported phenotypes include DA neuronal loss, disruption of dopamine homeostasis, ultrastructural changes, L-DOPA responsive locomotor defects, and pathological accumulations of tau and α-synuclein (Ramonet et al., 2011;Chen et al., 2012;Weng et al., 2016;Xiong et al., 2018). Several of these phenotypes have also been reproduced in rats, though without any obvious DA neuronal loss (Zhou et al., 2011;West et al., 2014;Sloan et al., 2016). In mice, DA neuron degeneration is typically observed at mid-to older age, occurring typically at 15-20 months in the SNpc (Ramonet et al., 2011;Weng et al., 2016;Xiong et al., 2018). Prior to cell death, DA neurons are frequently observed to exhibit abnormal morphology and reduced synaptic vesicles (Burke and O'Malley, 2013;Tsika et al., 2014;Liu et al., 2015;Weng et al., 2016). Decreased striatal dopamine content, dopamine metabolites, and evoked dopamine release also frequently occur in conjunction with DA neuronal death (Ramonet et al., 2011;Zhou et al., 2011;Liu et al., 2015;Sloan et al., 2016;Lim et al., 2018;Xiong et al., 2018). Consequently, these rodents exhibit locomotor defects, which are partially rescued with L-DOPA (Sloan et al., 2016;Weng et al., 2016;Xiong et al., 2018). To date, two studies using cDNA LRRK2-G2019S overexpression have been able to show both DA neurodegeneration and pathological inclusions (Chen et al., 2012;Xiong et al., 2018). In one such model, phosphorylated tau was increased at 12 months in the SNpc, in parallel with DA neuronal loss (Chen et al., 2012). The other reported an increase in phosphorylated and high molecular weight α-synuclein in the striatum and ventral medial body at 24 months of age (Xiong et al., 2018).
Although transgenic OE rodent models can capture many of the cardinal features of PD patients to various extents, there are several key caveats to bear in mind. For example, the question of whether DA neuronal loss occurs in a cell-autonomous or non-cell-autonomous manner has arisen, due to cell type-specific expression. Only mouse models which employ a DA neuron specific (TH) or neuronal transgene (CMV enhancer/PDGFβ) promoter has been able to induce DA neurodegeneration (Ramonet et al., 2011;Chen et al., 2012;Weng et al., 2016;Xiong et al., 2018). Of note, other neuronal transgene promoters, such as Thy1.2 and CaMKII, have failed to induce DA neuronal death, possibly owing to a lack of sufficient expression in midbrain DA neurons (Tsika et al., 2014;Garcia-Miralles et al., 2015). Alternative rodent models generated using a BAC approach, which use endogenous LRRK2 promoter to drive the expression No loss in SNpc up to 12-13 months (limited expression in SNpc).
Not assessed.
Decreased ambulatory movement, distance moved, and increased time on pole test at 12 months (Chen). Decreased ambulatory movement at 8-9 months (Chou).
Not assessed (limited expression in SNpc).
Trend for increased rotarod and distance traveled, but not significant (data not shown).
No obvious pathology up to 19 months due to lack of expression in SN (data not shown).
Increased rotarod performance at 3-4 months and increased distance traveled in first 30 min at 7 months (males only). These effects waned later in age (data not shown for rotarod).

(Continued)
Frontiers in Neuroscience | www.frontiersin.org has presented with more subtle phenotypes, such as changes in dopamine homeostasis and mild behavioral or locomotor dysfunction (Tables 4, 5).
One clear advantage of rodent OE models is that they have enabled the study of neurodegenerative mechanisms. One potential mechanism of action is LRRK2-mediated phosphorylation of synaptic proteins with known functions in vesicle endocytosis. Specifically, LRRK2 has been shown to phosphorylate the synaptic proteins synaptojanin 1 (SYNJ1), Endophilin A1 (SH3GL2) and auxilin (DNAJC6) (Matta et al., 2012;Arranz et al., 2015;Pan et al., 2017;Nguyen and Krainc, 2018). Broadly speaking, these studies have proposed that increased LRRK2-kinase activity disrupts the physiological function of these presynaptic trafficking proteins, resulting in defects in synaptic vesicle endocytosis (Pan et al., 2019).

Mouse LRRK2 KI Models
Currently, the disease mutations of LRRK2 knock-in (KI) models have only been developed in mice ( Table 6). Unlike OE models, KI models do not suffer from potential overexpression artifacts or interspecies differences. However, LRRK2 KI mouse models have failed to show DA neuron degeneration or α-synuclein pathology. Rather, these models exhibit neurophysiological changes, altered DA homeostasis, and modest behavioral abnormalities.
Two independent groups have reported that the LRRK2-G2019S KI mice exhibit increased frequency of spontaneous excitatory post-synaptic potentials (sEPSCs) in spiny projection neurons (SPNs) of the striatum (Matikainen-Ankney et al., Volta et al., 2017). Young (1-3 months) G2019S KI mice also exhibit elevated dopamine release upon repeated stimulation (Volta et al., 2017). These changes are mirrored by increases in total distance moved and rearings indicative of hyperkinesia at a young age (Longo et al., 2014;Yue et al., 2015;Volta et al., 2017). However, in older mice (>12 months), extracellular levels of dopamine appear decreased and hyperkinetic behavior is reduced (Yue et al., 2015;Volta et al., 2017). Perhaps surprisingly, a recent study has reported that young G2019S KI mice are more resilient to chronic social defeat stress (CSDS) (Matikainen-Ankney et al., 2018). The authors suggest that this may be due to the inability for calcium-permeable AMPA receptors (CP-AMPARs) to integrate into the synaptic membrane, blocking the formation of long-term potentiation (LTP). However, whether the synaptic changes at young adulthood in rodents confer altered non-motor or motor phenotypes in late-onset PD remains unclear.

Drosophila LRRK2 MODELS
Drosophila LRRK2 models can offer several advantages over rodents. Firstly, the relatively short lifespan of Drosophila (∼2 to 3 months) allows age-dependent changes in phenotypic variability to manifest quicker than in rodent models (e.g., synuclein pathology; Feany and Bender, 2000). In addition, the presence of the UAS-GAL4 system in several Drosophila lines can create a diverse range of genetic manipulations. Finally, No changes (Daher). Decreased latency to fall on rotarod at 6 months. No difference in cylinder and beam walking test (Walker). Postural instability at 8 months but was recovered at 12 months. Rearings increased at 12 months (Lee).
No changes in open field, rotarod and acoustic startle response. No change in AMPH injection compared (Tong).
Loss of TH staining is decreased in DM and PM clusters at 3 and 10 days (Lee). No loss in PPM1/2, PPL1, PPM3 (Imai). No change in number or distribution (Wang). No change in PPL1 cluster (Tain).
Climbing deficits at 3 days and loss of fertility in females (Lee). Smaller abdomen and decreased fertility (Imai). No behavioral abnormality or decrease in fertility (Wang). Tendency for decreased climbing ability (Tain). Not assessed. Climbing deficits at 3 days and loss of fertility in females (data not shown).
the dopaminergic system in Drosophila consists of six welldefined neuronal clusters that can easily be quantified to assay DA-specific neuronal death.

Drosophila dLRRK KO Models
Unlike rodents and humans, which possess both LRRK1 and LRRK2 genes, Drosophila only possesses a single LRRK2 ortholog, dLRRK. dLRRK is 2,445 amino acids in length and shares an overall sequence identity of 24% with human LRRK2 (Wang D. et al., 2008). Like LRRK2 KO rodent models, loss of dLRRK does not cause DA neurodegeneration ( Table 7) (Imai et al., 2008;Wang D. et al., 2008;Tain et al., 2009). Several studies have, however, reported locomotor deficits and loss of fertility in dLRRK KOs (Lee et al., 2007;Imai et al., 2008;Tain et al., 2009). An unclear point of investigation is whether dLRRK KOs are sensitive or protective to treatment with reactive oxygen species (ROS) with two studies observing conflicting evidence for H 2 O 2 on mortality (Imai et al., 2008;Wang D. et al., 2008). Future studies should evaluate reasons for this discrepancy and probe potential mechanisms for ROS sensitivity or protection in dLRRK KOs.

Drosophila Transgenic LRRK2 Models
As with rodent OE models, Drosophila OE of human LRRK2 G2019S and R1441C has resulted in DA neuronal loss, disruptions of dopamine homeostasis, and L-DOPA responsive locomotor defects (Table 8). Interestingly, other variants such as I1122V, Y1699C, I2020T, and G2385R have also been shown to cause DA neuronal loss in Drosophila (Imai et al., 2008;Liu et al., 2008;Ng et al., 2009;Venderova et al., 2009;Lin et al., 2010;Godena et al., 2014). This loss occurs at mid-to older age (35-60 days) and is commonly seen in the PPL1 DA neuronal cluster (Imai et al., 2008;Liu et al., 2008;Ng et al., 2009;Venderova et al., 2009). In addition to cell loss, transgenic models also exhibit decreased DA content and reduced climbing ability starting at birth or at 10 days, respectively (Imai et al., 2008;Ng et al., 2009).
Several studies have also reported sensitivity to ROS in flies expressing pathogenic LRRK2 mutations (Imai et al., 2008;Ng et al., 2009). Unlike rodent models, OE of human WT LRRK2 has also been reported to induce DA neurodegeneration (Liu et al., 2008). In addition, OE of Drosophila WT dLRRK (under a TH promoter) has also been shown to cause neurodegeneration (Imai et al., 2008). These models raise several questions, including whether DA neuronal loss may be caused by expression of a different species' LRRK2 and whether cell-type specific expression can cause DA neuronal loss.

Caenorhabditis elegans LRRK2 MODELS
Caenorhabditis elegans have a well-characterized CNS with a total of 302 neurons, 8 of which are dopaminergic. Unlike rodents, the simplicity of the C. elegans's nervous system has allowed for quantification and visualization of DA neuron morphology in vivo (Yao et al., 2010). In addition, its shorter lifespan, small size and cost-effectiveness have facilitated highthroughput drug screening (Maulik et al., 2017). Like Drosophila, C. elegans has one single ortholog of LRRK2, Lrk-1, which is broadly expressed in head and tail neurons, the hypodermis, intestine, and muscles (Sakaguchi-Nakashima et al., 2007). Both Lrk-1 KO and transgenic OE models of LRRK2 have been reported in C. elegans.

C. elegans Lrk-1 KO Models
Like other LRRK2 KO models, loss of Lrk-1 does not lead to the degeneration of DA neurons or striking locomotor phenotypes (Sakaguchi-Nakashima et al., 2007;Saha et al., 2009;Sämann et al., 2009;Yao et al., 2010). C. elegans do, however, demonstrate a mis-localization of synaptic vesicles in DA neurons in vivo (Sakaguchi-Nakashima et al., 2007;Sämann et al., 2009). In addition, Lrk-1 KOs have been found to be more sensitive to ER stressors, such as tunicamycin, which has been recently observed to induce the recruitment Decreased locomotion starting at 10 days (time required to recover from tapping).
Decreased locomotion starting at 10 days (time required to recover from tapping).
Decreased locomotion starting at 10 days (time required to recover from tapping).
Decreased locomotion starting at 10 days (time required to recover from tapping). Climbing is impaired at 60 days with ddC promoter.
Climbing is impaired at 60 days with ddC promoter.
No climbing deficits.
Climbing impaired at 4 weeks with ddC promoter and 6 weeks with elav.
Loss at 35 days using ddC promoter and 35 days with elav promoter (Liu).
Climbing impaired at 4 weeks with ddC promoter and 6 weeks with elav (more severe than human LRRK2 WT OE model) (Liu).
Loss at 60 days old using ddC and elav promoter in PPM1/2 and PPL1 (Imai).
Loss at 60 days old using ddC and elav4 promoter in PPM1/2 and PPL1 (Imai).
Decreased fertility in females and climbing deficits in 45-day old flies (Imai).
No climbing deficits (Imai). of LRRK2 from the trans-Golgi to the lysosome in a Rab29dependent manner (Sämann et al., 2009;Kuwahara et al., 2016;Eguchi et al., 2018). Both of these processes depend on the phosphorylation of Rab GTPases by LRRK2, and future studies in C. elegans may reveal more mechanistic insight in vivo.

C. elegans LRRK2 Transgenic Models
The OE of human LRRK2 G2019S and R1441C in C. elegans has been shown to cause DA neurodegeneration, reduced dopamine levels, and locomotor dysfunction (Sämann et al., 2009;Yao et al., 2010;Liu et al., 2011). Interestingly, two conflicting studies suggest that LRRK2-mediated neurodegeneration may depend on either GTPase or kinase activity. One study suggests that OE of human LRRK2 K1347A (a defective GTP binding mutant) does not produce Parkinsonian phenotypes in C. Elegans, inferring that GTPase activity is essential for toxicity (Yao et al., 2010), while another indicates that the kinase dead human LRRK2 OE also does not produce loss of DA neurons . Furthermore, OE of human LRRK2 WT has also been shown to cause DA neurodegeneration (Yao et al., 2010). Future C. elegans transgenic OE models will have to clarify these discrepancies and use better GTPase inactive mutants in C. elegans to assess for toxicity.

ZEBRAFISH LRRK2 MODELS
Zebrafish (Danio rerio) is an attractive model due to its wellcharacterized neuronal circuitry, conserved neurobiochemical mechanisms between humans, optical transparency, small size, and ease of drug administration (Vaz et al., 2018). The lifespan of zebrafish is relatively long compared to other rodent models, with 71% survivability at 5 years, making it ideal for aging research (Vaz et al., 2018). Like Drosophila and C. elegans, zebrafish possess a sole ortholog to human LRRK2, called zLRRK2, sharing a 47% amino acid sequence identity with humans (Sheng et al., 2010).

Zebrafish zLRRK2 KD Models
Since an initial study reported that zLRRK2 KO is embryonically lethal, a targeted knock-down (KD) approach has been used as an alternative to reduce the expression of zLRRK2 (Sheng et al., 2010). This initial study saw that targeting the WD40 domain of zLRRK2 using morpholino oligonucleotides resulted in the loss of diencephalon DA neurons and locomotor defects (Sheng et al., 2010). A later study supported this finding and also observed α-synuclein aggregates in multiple brain regions (Prabhudesai et al., 2016). However, one study using a dosage of morpholinos in between these two studies achieved a greater KD of zLRRK2, but could not replicate these phenotypes (Ren et al., 2011). These studies may show conflicting results due to the reported off-target effects of morpholinos, and thus different approaches should be considered (Kok et al., 2015). As such, further study is required to establish a consistent theme in zebrafish zLRRK2 KD models.

Zebrafish LRRK2 Mutant Models
Limited zebrafish mutant models have been developed, but a recent report using zinc finger nucleases (ZFNs) was able to introduce a mutation in the WD40 domain of zLRRK2 to generate a KI model (Sheng et al., 2018). They observed increases in locomotion in the adult stage and a weakened antibacterial response. Another report using transient overexpression of human LRRK2 WT and G2019S in zebrafish embryos was able to see disruptions in the ubiquitin-proteasome system (UPS) (Lichtenberg et al., 2011). Future studies will have to develop more mutant models of zebrafish models in order to gain more phenotypic insights.

CONCLUDING REMARKS AND FUTURE PERSPECTIVES
Here we highlight several phenotypic themes found in LRRK2associated PD animal models, with an emphasis on rodents. LRRK2 KO models in mice, rats, Drosophila, and C. elegans have not produced DA neurodegeneration. Rather, phenotypic changes have been observed primarily in peripheral tissues such as the kidneys and lungs. Interestingly, primates treated with LRRK2 kinase inhibitors have also show disruptions in the kidneys and lungs with no obvious neuropathological changes (Fuji et al., 2015). Future models should work to uncover other novel phenotypes associated with LRRK2 KO models, which can be indicative of LRRK2 kinase inhibitor treatment, thereby facilitating the investigation of drug-based toxicity or efficacy. By contrast, transgenic OE of pathogenic LRRK2 in mice and Drosophila has been shown to cause robust DA neurodegeneration, tau and α-synuclein pathology, locomotor/behavioral deficits, alterations in DA homeostasis, and L-DOPA-responsive behavior. A recently reported mouse model was found to have many of these features including DA neurodegeneration, locomotor changes and α-synuclein pathology, and it may be useful for determining drug efficacy (Xiong et al., 2018). Other phenotypes, such as tau pathology and cognitive, behavioral, and peripheral organ changes have not been assessed in this model and could provide other important clinical measures. Finally, LRRK2 KI models used in mice and zebrafish have produced neurophysiological changes and modest behavioral or locomotor abnormalities. These phenotypes seen can be indicative of early or prodromal PD, and with careful behavioral analysis, such as one done by Giesert et al. (2017) in a R1441C KI mouse model, it can reveal other important phenotypes.
An emerging area in LRRK2 biology that may help guide future LRRK2 models is the role of Rab proteins and whether they can be used as an in vivo biomarker for LRRK2 activity. LRRK2 is capable of phosphorylating Rab3A/B/C/D, Rab8A/B, Rab10, Rab12, Rab29, Rab35, and Rab43 (Steger et al., 2016(Steger et al., , 2017. Using LRRK2 G2019S, R1441C, kinase-dead, and phosphomimetic (S910A + S935A) KI rodent models, the endogenous phosphorylation of Rab proteins can be carefully evaluated in a tissue-specific manner. This should provide important clues for cell-type and region-specific readouts of LRRK2 activity Lis et al., 2018). Furthermore, a recent study has highlighted LRRK2's role at the lysosome, where LRRK2 is recruited onto stressed lysosomes by Rab29 and phosphorylates Rab8A and Rab10 (Eguchi et al., 2018). Interestingly, Rab8A KO mice have been developed, which may give important insights into LRRK2 biology at the lysosome (Sato et al., 2007). Furthermore, RAB29 was also observed to be a risk factor for the development of PD, and future studies should investigate a potential unifying mechanism for LRRK2 and Rab29 in PD pathogenesis (MacLeod et al., 2013;Beilina et al., 2014;Nalls et al., 2019).
Another emerging area in LRRK2 animal models is the role of LRRK2 in producing immune system abnormalities. A recent LRRK2 phosphomimetic (S910A + S935A) KI mouse model has shown a reduction in astrogliosis, while a transgenic OE of LRRK2-G2019S in mice exhibited an increase in astrogliosis (Xiong et al., 2018;Zhou et al., 2018). Another study using G2019S KI mice has exhibited astrogliosis using a sub-toxic dose of MPTP dose to cause DA neurodegeneration (Arbez et al., 2019). Investigation of LRRK2 and astrogliosis using these models may help elucidate potential pathogenic mechanisms for LRRK2 in the CNS.
LRRK2 animal models have thus facilitated our understanding of LRRK2 biology, have led us to determine PD pathogenic mechanisms, and have facilitated the discovery of novel phenotypes in LRRK2 pathogenesis in PD and therapeutic development.

AUTHOR CONTRIBUTIONS
SS, GH, and ZY wrote and edited the manuscript. DL had curated information on C. elegans and zebrafish models with assistance from BT. IC and MB provided topic specific knowledge on LRRK2 and guided sections in writing. All authors contributed to the article and approved the submitted version.

FUNDING
This work was supported by an NIH grant (R01NS06123, R01GM115844, and P50NS094733) to ZY. In addition, DL is supported by China Scholarship Council and MB is supported Fundación Ramón Areces.