Contributions of VPS35 Mutations to Parkinson’s Disease

Parkinson's Disease (PD) is a multi-system neurodegenerative disease where approximately 90% of cases are idiopathic. The remaining 10% of the cases can be traced to a genetic origin and research has largely focused on these associated genes to gain a better understanding of the molecular and cellular pathogenesis for PD. The gene encoding vacuolar protein sorting protein 35 (VPS35) has been definitively linked to late onset familial PD following the identification of a point mutation (D620N) as the causal agent in a Swiss family. Since its discovery, numerous studies have been undertaken to characterize the role of VPS35 in cellular processes and efforts have been directed toward understanding the perturbations caused by the D620N mutation. In this review, we examine what is currently known about VPS35, which has pleiotropic effects, as well as proposed mechanisms of pathogenesis by the D620N mutation. A brief survey of other VPS35 polymorphisms is also provided. Lastly, model systems that are being utilized for these investigations and possible directions for future research are discussed.


INTRODUCTION
Parkinson's disease (PD) is a complex neurological disorder involving both motor and non-motor symptoms. Affecting 1% of the population over 60, it is the second most common neurodegenerative disorder (Lau and Breteler, 2006;Williams-Gray et al., 2013). The preva lence of the disease rises with age. According to a meta-analysis report by Pringsheim, Jette, Frolkis, & Steeves, 2014, the prevalence is 41 per 100,000 between ages 40 and 49, rising up to 428 between ages 60 and 69 and 1087 between 70 and 79. Another report states that there is a greater incidence in men compared to women, at a ratio of approximately 1.5:1 (Elbaz et al., 2002;Taylor et al., 2007;Wooten et al., 2004). The case burden in US alone is estimated to rise to more than one million by 2030 (Marras et al., 2018). The deterioration in motor function occurs primarily due to a loss of dopamine signaling in the basal ganglia. This loss of dopamine is due to the pro gressive loss of dopaminergic neurons in the substantia nigra. However, it is not only the basal ganglia that is affected in PD. Research suggests that PD involves mul tiple areas of the brain and results from a complex inter play between genetic and environmental factors. Clinically described motor symptoms include resting tre mors, bradykinesia, rigidity of limbs and gait defects.
Moreover, there are numerous non-motor symptoms such as cognitive defects and dementia, mood disorders, sleep disorders that add to the suffering of the patients (Pringsheim et al., 2014;Ascherio and Schwarzschild, 2016;Tysnes and Storstein, 2017). According to the cen ters for disease control and prevention (CDC), medical complications due to PD rank as the 14th leading cause of death in the United States (Beard et al., 2017). With no known cure for this disease, research has been focused on understanding the cellular and molecular pro cesses that may be responsible for the neurodegenera tion ( Fig. 1).
Hallmark histopathological features of PD are the loss of dopamine-producing neurons in the substantia nigra and the formation of large protein aggregates in surviving neurons termed Lewy bodies. Lewy bodies are composed primarily of a-synuclein protein in these neurons. The pathology of the a-synuclein aggregates shares many features with that of prion disease thereby spurring investigation into that hypothesis (Brundin and Melki, 2017;Surmeier et al., 2017). Mitochondrial dys function is also a widely reported feature of PD (Goswami et al., 2017). Familial PD-linked mutations have been identified in a number of genes (SNCA, LRRK2, VPS35, CHCHD2, GBA, Parkin, PINK1, DJ-1, ATP13A2, FBXO7 and PLA2G6) that have been shown to participate in mitochondrial function and biogenesis which further underscores the association with mitochon drial defects as a potential driver of the disease (Helley et al., 2017). Protein homeostasis, particularly synucleinopathy, and mitochondrial health have therefore become major areas of focus for research into character izing PD pathogenesis and developing therapeutics (Table 1).
In 2005, a study identified VPS35 levels to be reduced in affected brain regions in Alzheimer's disease (Small et al., 2005). In addition, another study reported increased Amyloid b depositions in mice heterozygous for VPS35 knockout (Vps35 +/m ) (Wen et al., 2011). Similarly, other work has linked abnormal microglial activity and abnormal hippocampal development, as observed in Alzheimer's Disease, to VPS35 depletion, but no polymorphisms in the VPS35 gene have been associated Appel et al., 2018).

VPS35 AS PART OF THE RETROMER COMPLEX AND RELATED FUNCTIONS
VPS35 was originally identified in yeast as a member of the retromer complex. This complex is involved in the intracellular trafficking of proteins (Seaman et al., 1998). Its role in endosome to Golgi retrograde transport has been very well characterized (Seaman, 2012;Follett et al., 2014a;Trousdale and Kim, 2015). Retromer dys function has been implicated in Alzheimer's disease, Parkinson's disease and several other neurodegenerative diseases, along with various developmental processes, including wing and eye development in Drosophila, and bone development (Belenkaya et al., 2008;Zhang et al., 2011;Chan et al., 2016). In this context, the interaction of the retromer complex and the wnt signaling pathway has been extensively studied (Belenkaya et al., 2008;Zhang et al., 2011Zhang et al., , 2018Small and Petsko, 2015;Wang and Bellen, 2015;Chan et al., 2016). Structurally, VPS35 forms a trimer with VPS26 and VPS29, to form the cargo recognition complex (CRC). The CRC then associates with a dimer of sorting nexins, belonging to the SNX-BAR family of proteins. This association is fur ther facilitated by RAB7. Together, the retromer complex plays important roles in vesicular sorting (Bonifacino and Hurley, 2008;Seaman et al., 2009;McGough and Cullen, 2011;Seaman, 2012;Lucas et al., 2016;Kovtun et al., 2018). Deficiency of either VPS35 or VPS29 leads to the degradation of the other two CRC components (Fuse et al., 2015). Transmembrane receptors such as the cation-independent mannose 6 phosphate receptor (CIMPR) and the membrane iron transporter DMT1 are well-characterized targets of retromer-mediated sorting (Arighi et al., 2004;Tabuchi et al., 2010;Lucas et al., 2016). In addition, there have been several studies on the role of the retromer in recycling of beta-2 adrenergic receptors (b2AR) in dendritic cells and trafficking of G protein-coupled receptors (Bunnett and Cottrell, 2010;Feinstein et al., 2011;Choy et al., 2014;Bowman et al., 2016;McGarvey et al., 2016;Pavlos and Friedman, 2017;Bahouth and Nooh, 2017;Sposini et al., 2017). The retromer is known to carry cargo from endosomes to the trans-Golgi network (TGN), performing retrograde transport, as well as carrying cargo from endosomes to the plasma membrane, allowing recycling of membrane bound receptors (Trousdale and Kim, 2015). Moreover, included in the list of retromer cargo, are degradative enzymes such as Cathepsin D, which are important for lysosomal function and could potentially influence autop hagic flux. Furthermore, amyloid precursor protein (APP) and a-synuclein (SNCA) are also reported to be sorted by the retromer complex (Miura et al., 2014;Li et al., 2016;Gallon and Cullen, 2015;Aufschnaiter et al., 2017;Follett et al., 2017Follett et al., , 2014aReitz, 2018). These findings have sparked considerable interest in the contributions of the retromer to human diseases.

PROPOSED MECHANISMS OF VPS35 IN NEURONAL DYSFUNCTION
The importance of the roles played by the retromer complex in signal transduction and receptor trafficking points to the dysfunction of the retromer as the major potential mechanism of neuronal degeneration caused by the VPS35 D620N mutation (Follett et al., 2017(Follett et al., , 2014aSmall and Petsko, 2015;Wang and Bellen, 2015;Reitz, 2018). Knockdown of VPS35 leads to the degradation of VPS29 and vice-versa (Fuse et al., 2015). Therefore, given that any free VPS35, existing out side the retromer complex, is rapidly degraded, the role of VPS35 D620N in PD is likely to involve altered retromer function. Trafficking defects of AMPA receptors was reported in a VPS35 heterozygote mouse model as well as cultured mouse hippocampal and cortical neurons treated with VPS35 shRNA (Munsie et al., 2015;Tian et al., 2015). Additionally, dopamine receptor D1 (DRD1) is also reported to be recycled with the involve ment of VPS35 (Wang et al., 2016a). Studies have reported reduced a-synuclein degradation in VPS35-deficient cells (Braschi et al., 2010;Miura et al., 2014;Sugiura et al., 2014;Tang et al., 2015a,b). Formation of toxic a-synuclein aggre gates and fibril formation are hallmarks of PD (Patel and Witt, 2018). A study conducted by Mensˇ ı´kova´ et al.
(2018) identified a VPS35 mutation and a FBXO7 muta tion in a patient that displayed profuse Lewy body pathol ogy in various brain regions, including the substantia nigra and other midbrain regions (Mensˇ ı´kova´ et al., 2018). This accumulation has classically been attributed to abnormal sorting of degradative enzymes that are normally targeted to the lysosome. However, this accumulation could also be partially explained by autophagy defects, similar to what has been reported in HeLa cells expressing the VPS35 D620N mutation (Mcgough et al., 2014;Zavodszky et al., 2014). The mechanism proposed in this study, for the autophagy disruption, was impaired WASH (Wiskott-Aldrich syndrome protein and SCAR homolog) complex association. Given that autophagy is a major process responsible for removing proteins, macro molecules and organelles (Gatica et al., 2018), this is another potential mechanism for the neurodegeneration caused by VPS35 (Barth et al., 2010;Lynch-Day et al., 2012;Tofaris, 2012).
Furthermore, defects in mitochondrial fusion and function have also been reported in some studies (Tang et al., 2015a,b;Wang et al., 2016b). Mitochondrial dys function has been shown to affect neuronal function severely and is believed to be a driving force of neurode generation (Hauser and Hastings, 2013;Mullin and Schapira, 2013;Subramaniam and Chesselet, 2013;Haelterman et al., 2014;Winklhofer, 2014). Also, it is noteworthy that autophagy is the only means by which damaged mitochondria are turned over (Ashrafi and Schwarz, 2013). Therefore, disruption in autophagy due to VPS35 D620N mutation may also result in, if not exac erbate, any mitochondrial dysfunction that may be direc ted by the VPS35 D620N mutation.

CURRENT MODEL ORGANISMS AND SYSTEMS FOR STUDYING VPS35
Cell Culture A variety of tissue culture model systems are available for studying the structure and function of VPS35, and the pathogenic effects of its mutations. SH-SY5Y neuroblastoma cells can easily be induced to adopt a dopaminergic state by treatment with retinoic acid (Korecka et al., 2013;Krishna et al., 2014;Shipley et al., 2016). This system was used by Tang et al. (2015a,b) to demonstrate mitochondrial impairment due to VPS35 deficiency as well as due to the D620N muta tion (Tang et al., 2015a,b). This system was also used to characterize the role of VPS35 in lysosomal clearance of AIMP2, which is a substrate of the PD associated gene, Parkin (Yun et al., 2017). Microglial BV2 cells have been used to understand the role of VPS35 in Alzheimer's Dis ease, and microglial physiology (Lucin et al., 2013;Yin et al., 2016). In addition, HEK 293T cells also present a popular choice for understanding cellular pathways affected by VPS35 and retromer function due to their high plasmid transfection efficiency (Yang et al., 2008;Follett et al., 2014b;Williams et al., 2018 (Nayerossadat et al., 2012). Korolchuk et al. (2007) knocked down VPS35 in Drosophila S2 cells using an RNAi approach, in order to identify novel proteins important for endocytosis, where candidate proteins were selected based on binding pre dictions to either a adaptin or clathrin heavy chain, or on having a predicted membrane bending domain (such as the BAR domain). These candidate proteins were then knocked down to screen for subsequent endocytosis defects. VPS35 was among the short list of proteins that were identified as exhibiting a significant reduction in endocytosis, where the effect due to VPS35 was the most severe (Korolchuk et al., 2007).
In addition to immortalized cell lines, primary neuronal and glial cultures have also been used to investigate VPS35 functions and mutations (Wen et al., 2011;Tsika et al., 2014;Tang et al., 2015a,b;Williams et al., 2018). Tang et al. (2015a,b) isolated and cultured dopaminergic neurons from brains of mice expressing microRNA direc ted against VPS35 (Tang et al., 2015a,b). These cells were found to have a reduced expression of MFN-2 or Mitofusin, and exhibited mitochondrial fragmentation, as well as impaired mitochondrial function. Similar results were obtained using SH-SY5Y and NLT neuroblastoma cultures (Tang et al., 2015a,b). In the study conducted by Tsika et al. (2014), rat primary cortical cultures were used to show that overexpression of human VPS35 resulted in neuronal cell death and higher sensitivity to cellular stress factors associated with PD (Tsika et al., 2014). Wen et al. (2011) demonstrated using mouse hip pocampal slice cultures that VPS35 haploinsufficiency exacerbates long-term potentiation impairment already present in mice expressing Swedish mutant form of the amyloid precursor protein (Wen et al., 2011). Williams et al. (2018) discovered a novel link between the PD gene, Parkin, and VPS35, wherein Parkin mediates ubiq uitination of VPS35, which does not affect VPS35 turn over, but may have a role to play in the retromer mediated endosomal sorting (Williams et al., 2018). Using primary cortical neurons, they identified ATG9A, a WASH-dependent retromer cargo, to be missorted upon silencing of the Parkin gene. Using cell lines and primary cultures can therefore be very powerful tools for decipher ing molecular mechanisms under precisely controlled conditions.

Mice
Homozygous knockout of VPS35 is embryonically lethal in mice. To overcome this challenge several mice models have been developed, including a hemizygous deletion mutant (Wen et al., 2011). This mutant was used to demonstrate that VPS35 interacts with the protease BACE-1, predominantly responsible for Amyloid beta pro duction, and that VPS35 reduction increases BACE-1 activity in the mouse hippocampus. In addition, A VPS35 D620N knock-in strain of mice has been devel oped recently by Cataldi et al. (2018), in collaboration with Jackson Laboratories, to further characterize dopamine release and monoamine transporters in a VPS35 D620N mutant background (Cataldi et al., 2018). These mice were generated by Cre-recombinase-mediated recombi nation of the VPS35 exon 15, on chromosome 8. Subse quent assessment revealed no detectable movement disorder in these transgenic mice compared with non transgenic controls. However, the dopamine turnover was reported to be increased. Dopamine transporter levels were reported to have been reduced while vesicular monoamine transporter levels were increased (Cataldi et al., 2018). Mir and colleagues (2018) used these same mice to study the interaction between LRRK2 and VPS35 (Mir et al., 2018). Using this model, they were able to demonstrate that the VPS35 D620N mutation increases LRRK2-catalyzed phosphorylation of three different RAB proteins, RAB 8a, 10 and 12. They concluded that the VPS35 D620N mutation causes a gain of function with respect to the kinase activity of LRRK2 and is an upstream regulator (Mir et al., 2018). A similar mouse model, generated using a CRISPR/Cas-9-mediated knock-in strategy, was used previously by Ishizu et al. (2016) to characterize reduced dopamine release in the striatum of heterozygous mice with one gene copy of VPS35 D620N knocked in (VPS35 D620N KI ), while the other copy had one base pair deleted in exon 15, creating a premature stop codon, referred to as VPS35 Del1 (Ishizu et al., 2016). These VPS35 D620NKI /VPS35 Del1 heterozy gous mice also displayed premature death and significant neurodegeneration throughout life, indicating that this model could be of considerable value for Parkinson's Dis ease research.
Rat A transgenic rat model generated via lentivirus-mediated gene transfer has been described and utilized by Tsika et al. (2014). These rats overexpressed one of two human VPS35 constructs, one of wild-type and the other contain ing the D620N mutation. However, the researchers con cluded that the human D620N construct did not alter vesicular sorting of retromer cargo proteins in primary cor tical neurons from these animals. They went on to confirm this finding in yeast and human (patient donated) fibrob lasts. Contrary to this finding, the human VPS35 D620N overexpression was found to induce neurodegeneration in the substantia nigra of these rats suggesting that, in this model, VPS35 D620N might be primarily causing neu ronal loss without altering retromer function (Tsika et al., 2014).
Drosophila melanogaster. MacLeod et al. (2013) used an overexpression vector to demonstrate that overex pression of wild-type VPS35 reduced the defects caused by LRRK2 G2019S mutation, and those caused by RAB7L1 knockdown. This was shown in LRRK2 mutant flies and primary rat neuronal cultures (MacLeod et al., 2013). Similarly, a VPS35 overexpressing strain of flies was used by Linhart et al. in 2014, to demonstrate the res cue of LRRK mutation derived eye defects, by eye speci fic VPS35 overexpression in Drosophila (Linhart et al., 2014). In another study, flies with the Drosophila VPS35 (dVps35) knocked out, were shown to be embryonically lethal (Inoshita et al., 2017). This lethality could not be rescued by knock-in of human VPS35 wild-type or human VPS35 D620N. However, it was rescued by knock-in of dVps35 WT and by knocking in dVps35 D647N (the Dro sophila analog of the D620N mutation in humans). Using transgenic flies expressing wild-type dVps35 or dVps35 D647N, in all three possible dVps35 heterozygote and homozygote backgrounds (+/+, +/-, -/-), they were then able to demonstrate that Drosophila LRRK (dLRRK) and dVps35 affect synaptic architecture and endocytosis via the same pathways (Inoshita et al., 2017).

Yeast and Caenorhabditis elegans
In 2006, Prasad and Clark identified the retromer as an important component in Wnt signaling-mediated neuronal polarity in Caenorhabditis elegans. In a mutagenesis screen, they identified a vps-35 deletion mutant that resulted in reversed polarity of the mechanosensory neurons ALM and PLM. This defect was rescued by overexpressing vps-35 specifically in Wnt-expressing muscle and epidermal cells, but not in the neurons themselves (Prasad and Clark, 2006). This presents an important mechanistic role of VPS35, because the role of Wnt signaling in PD has been described in literature (Berwick and Harvey, 2012;Salasˇ ova´ et al., 2017). Dhungel et al. (2015) discovered that in VPS35-deficient yeast cells, EIF4G1 upregulation was highly toxic, but the two mutations by themselves were not lethal (Dhungel et al., 2015). EIF4G1 is a trans lation initiation factor scaffold protein, responsible for the translation initiation complex formation (Villa et al., 2013). While they could not establish a direct interaction between the two proteins, a few common pathways were identified leading to the hypothesis that they were func tionally related. The authors also noted that in a VPS35 null background, overexpression of EIF4G1 resulted in fewer but larger synaptic boutons. This group then inves tigated proteotoxic stress in yeast cells and activation of the unfolded protein response (UPR). It was seen that there was increased proteotoxic stress and UPR activa tion in VPS35 deletion mutants, only upon overexpressing EIF4G1. This led to the conclusion that the EIF4G1 upregulation combined with retromer dysfunction was causing toxic protein aggregation. In addition, this group demonstrated that Sortillin functions downstream of VPS35 or in parallel and is able to suppress the EIF4G1 overexpression toxicity in VPS35 deletion mutants. Lastly, this group showed, using yeast, C. elegans and transgenic mice that VPS35 deletion increases alpha synuclein-driven toxicity. In transgenic mice, increased toxicity was also seen after overexpression of VPS35 D620N and VPS35 P316S (Dhungel et al., 2015). There fore, altered VPS35 function can facilitate protein aggre gation, a hallmark of PD. Furthermore, Zhang et al. (2018) demonstrated using a vps35 deletion mutant strain of C. elegans that VPS35 is required for propagating the mitochondrial unfolded protein response (UPR mt ) from cells undergoing mitochondrial stress to surrounding cells. Here, VPS35 and the retromer was proposed to be involved in the retrieval of a Wnt secretion factor, MIG-14, which in turn, played a role in secretion of the Wnt protein EGL-20. This led to the conclusion that EGL-20 was acting as a ''mitokine" signal that induced UPRmt in a cell-non-autonomous manner in the surround ing cells and required functional retromer activity .

FUTURE RESEARCH DIRECTIONS
With the advent of CRISPR/Cas-9-mediated genome editing technology, therapeutic applications targeting the VPS35 gene can now be conceived (Calatayud et al., 2017). This could potentially be done in inducible pluripo tent stem cells, obtained from a patient, genetically mod ified and transplanted into the affected individual's brain. This technique has recently been described and is await ing U.S Food and Drug Administration approval (Loring, 2018). A major obstacle in such an approach is the prion like properties of a-synuclein aggregates (Brundin and Melki, 2017). The pre-existing toxic a-synuclein aggre gates in the patient's existing brain environment may induce the formation of a-synuclein fibrils in the trans planted cells, thus negating any genetic corrections made in the transplanted cells. Supporting this concern, previ ous clinical trials involving the transplantation of embryon ically derived grafts have shown synucleinopathy in the transplanted cells upon autopsy (Kordower et al., 2008;Li et al., 2008).
In addition, further characterization of the VPS35 D620N knock-in mice is now required to properly investigate the cellular pathways that are perturbed. This in vivo model system will allow for a more direct comparison of a mouse model system and human disease without the artifacts association with overexpression and/or random integration of a transgene using virus conventional transgenic mouse generation. The creation of a PD-relevant cell line containing the D620N mutation in the endogenous VPS35 gene would add validity to findings already reported using cells that have a VPS35 D620N construct stably overexpressed. Finally, characterization of the pathways perturbed by mutations in VPS35, is far from complete. A better understanding of the interplay of VPS35 and these pathways is required, so that therapeutic interventions can be crafted.