ABSTRACT
Parkinson’s disease (PD) is the second most common neurodegenerative disorder after Alzheimer’s disease. The etiology of PD is currently not fully understood, but strong evidences point to gene-environmental interactions that induce various aspects of neurodegeneration: oxidative stress, neuroinflammation, mitochondrial dysfunction, and epigenetic modulation, over several decades of disease progression. PD-like symptoms have been described in ancient medical texts, and since the characterization of motor symptoms by James Parkinson in 1817, it has evaded all curative attempts for stopping or reversing its progressive neurodegenerative nature. Numerous agents are evaluated for therapeutic potential against PD, most notably glial-cell line derived neurotrophic factor (GDNF). Despite efficacy in animal models, all human trials involving GDNF and related neurotrophic factors in PD patients have not achieved efficacy, suggesting a need for a rethinking in GDNF delivery to the human brain. The greatest challenge facing GDNF-mediated neuroprotection, as concluded from trials that failed to achieve efficacy, is the upregulation of GDNF by means other than genetic manipulation. Pharmacologically modulated signaling pathways which are in crosstalk with GDNF or could induce its endogenous upregulation therefore represent opportunities to fully harness the clinical benefits of GDNF, without the side effects associated with current methods for delivery of GDNF.
In this collection of studies, we showed that Prokineticin-2 (PK2), a secreted neuropeptide capable of binding to receptors PKR1 and PKR2, could protect against classic Parkinsonian toxicants MPP+ and Mn in MN9D mouse dopaminergic neurons in culture. We characterized the transcriptional regulation of PK2 by analyzing the PK2 promoter, which revealed hints for its role in cellular growth, maintenance, determination of neuronal genes, as well as during neurodegeneration. We validated some of the in silico findings using cell culture models of PD as well as the Mn overexposure model of PD in the mouse. Guided by these findings, we tested and found a fundamental relationship between PK2 signaling and GDNF signaling in astrocytes. The protective role and therapeutic potential of the PK2-PKR1-GDNF signaling axis was confirmed using a PKR1 agonist, IS20, in MPTP mouse model of neurodegeneration as well as in the MitoPark mouse model of PD. Collectively we had shown that PK2 signaling activation is relevant during neurodegeneration; its pharmacological modulation could induce GDNF upregulation and neuroprotective effects in multiple models of PD.

CHAPTER 1. GENERAL INTRODUCTION
Introduction
Parkinson’s disease (PD) is a complex neurological disorder. It is the most common neurodegenerative disorder after Alzheimer’s disease, affecting about 1.5 million people in the United States and over 10 million worldwide. With an aging world population, it is projected to increase by more than 50% by 2030 (Dorsey et al., 2007), and exceed $13 billion in national economic burden by 2030, $18 billion by 2050 (Kowal, Dall, Chakrabarti, Storm, & Jain, 2013).
In 1817, James Parkinson described a disease that is characterized today by akinesia, bradykinesia, rigidity, postural instability and fatigue (Barzilai & Melamed, 2003; Jarraya et al., 2009; Nassif & Pereira, 2018), a set of conditions with the earliest reference dating back to 600 B.C. in the ancient Indian medical literature (Manyam, 1990; Ovallath & Deepa, 2013). Current understanding of PD indicates that the early death of dopaminergic neurons in the ventrolateral substantia nigra pars compacta (SNpc), and the resulting severe dopamine depletion in the dorsal caudate putamen sit at the core of the disease. Postmortem analysis of PD patient brains has found another hallmark of PD: the accumulation of Lewy bodies in the neuron bodies and Lewy neurites in neuron processes, which comprised of mainly alpha-synuclein aggregates. The role of Lewy bodies and PD pathogenesis still stands unclear (Recasens et al., 2014; Wakabayashi et al., 2012).
PD diagnosis by clinicians is aided by medical imagining using PET scans, which could assess the functional integrity of the nigrostriatal dopaminergic system (Niccolini, Su, & Politis, 2014; Poewe, 1993). However, PD is diagnosed incorrectly in about 25% of patients (Tolosa, Wenning, & Poewe, 2006), in part due to a lack of laboratory diagnostic tests for definitive diagnosis (Kalia & Lang, 2015). The definitive diagnosis only comes after post-mortem pathological examination of degenerated dopaminergic neurons in the patient brain. Motor deficits occur when an estimated 60% of the dopaminergic neurons in the SNpc are lost with 80% reduction in striatal dopamine levels, attesting to the brain’s immense compensatory capacity for reduced dopamine production during early stage of the disease, while underscoring a need for early diagnosis independent of motor symptoms. Non-motor symptoms associated with PD have been increasing recognized and used in the differential diagnosis of PD from other similar conditions during the last decade (Richard L. Doty, 2012; Rodríguez-Violante, Zerón-Martínez, Cervantes-Arriaga, ; Corona, 2017). Because of prevalence of non-motor symptoms (over 90%) in PD patients, non-motor symptoms can be taken into consideration during differential diagnosis, together with motor symptoms and medical imagine data, to lower the rate of misdiagnosis. Because non-motor dysfunctions appear before motor dysfunctions, the progressive loss of dopaminergic neurons has presumably already begun, during the manifestation of non-motor symptoms, with its effects unseen until various compensatory mechanisms are overwhelmed. Therefore, the prodromal period preceding massive dopaminergic cell death represents a critical window of opportunity for a potential therapeutic intervention to alter the course of disease progression (Siderowf ; Lang, 2012).
The first mutations in SNCA (PARK1) were found to be responsible for PD in 1997 (M H Polymeropoulos et al., 1997; Mihael H. Polymeropoulos et al., 1996). Since then, with help from large scale genome-wide association studies (GWAS) enabled by advances in sequencing technology, a total of 28 mutations in various genomic loci have since been linked to increased risks of PD (Klein et al., 2012; Simón-Sánchez et al., 2009). However, PD patients who have mutations in these genes comprise of only roughly 5% of clinical cases (Pankratz et al., 2012; Shulman, De Jager, ; Feany, 2011), indicating that rather than being purely hereditary, the etiology of PD is multifactorial, involving an interplay of PD-susceptibility genes that are modified or influenced by environmental factors, leading to a cascade of events which result in PD pathology.
A meta-analysis had found 11 environmental factors that have positive association with PD, and the top risk factors are pesticide exposure, well-water drinking, beta-blocker use, prior head injury, agricultural occupation, and rural living (Foubert-Samier et al., 2012). Exposure to pesticides such as rotenone, paraquat, dieldrin, organophosphates and Mn (in the form of pesticide Maneb) are implicated in development of PD (A. G. Kanthasamy, Kitazawa, Yang, Anantharam, ; Kanthasamy, 2008; Roede ; Miller, 2014; Semchuk, Love, ; Lee, 1992; Singh et al., 2018; Tanner et al., 2011). In addition, aerosolized metals exposure has been reported as an environmental risk factor for PD (Brown, Lockwood, ; Sonawane, 2005). In particular, exposure to manganese in an occupational setting could increase risks for developing Parkinsonian symptoms by 3 – 10 fold (Gorell et al., 1999). Therefore, studying both the environmental and genetic factors of PD could provide insight into possible points of failure within the cell caused by mutations in the PARK genes, that can be exacerbated by environmental factors, leading to neurodegeneration.
Collectively, the consensus might be that PD is a multifactorial disease with contributions from genetic and environmental sources, which converge on the vulnerabilities of the dopaminergic neurons in the SNpc. Looking beyond the risk factors of the disease, some of which are inherently unavoidable, several common themes appearing during the neurodegenerative process have emerged. Oxidative stress, neuroinflammation, and mitochondrial dysfunction are seen from in vitro, in vivo studies, and post mortem analyses of PD patient brains. Until recently, these processes that accompany nigral dopaminergic degeneration was presumed to be consequences of cell death. However, over the past decades, evidence had pointed to their active participation in neurodegeneration and had implicated them in PD pathogenesis and disease progression.
The mainstay of PD symptomatic management remains to be therapies that increase dopamine concentrations or receptor activity in the caudate putamen. L-DOPA, the precursor for dopamine, was discovered by Arvid Carlsson in the 1950s to have effects on Parkinsonian animals, experiments for which the Nobel Prize in Physiology or Medicine was awarded in 2000. Decades later, L-DOPA remains today the strongest arsenal against disease symptoms. However, common therapies such as L-DOPA, dopamine agonists, monoamine type B inhibitors, induce side effects such as dyskinesia with long-term treatments. Without disease-modifying treatments, current therapies could manage motor symptoms but do not treat non-motor symptoms, and do little to slow disease progression.
One of the most promising candidate for neuroprotection and neuro-restoration of SNpc dopaminergic neurons are members of the glial-cell line derived neurotrophic factor (GDNF) family, consisting of GDNF, artemin, persephin, and neurturin. Since the discovery of GDNF in 1992 (O’Malley, Sieber, Black, & Dreyfus, 1992), extensive evaluation in cell culture, rodent, and primate models of PD had demonstrated its exceptional neuroprotective effects on the dopaminergic system (Bilang-Bleuel et al., 1997; Björklund, Rosenblad, Winkler, & Kirik, 1997; J. H. Kordower et al., 2000; L. F. Lin, Doherty, Lile, Bektesh, & Collins, 1993; Tseng, Baetge, Zurn, & Aebischer, 1997). Nerve injury induces upregulation of GDNF (Araujo & Hilt, 1997), and GDNF seem to be upregulated in post-mortem brains of PD patients (Backman et al., 2006). Therefore, GDNF could be a protective compensatory mechanism against neurodegeneration. Several landmark GDNF clinical trials had initially showed relative safety and efficacy in open-label studies, but ultimately, GDNF delivery using currently employed adeno-associated virus (AAV) showed no efficacy in phase II clinical trial (Blits & Petry, 2016; Remy, 2014; Tenenbaum & Humbert-Claude, 2017). Hence, the greatest challenge in meeting unmet medical need for PD remains to be the development of disease course-modifying therapies that could slow or stop the progression of the disease (Kalia & Lang, 2015).
We had recently discovered that a neuropeptide, prokineticin-2, is upregulated in response to neurotoxic stress. We found that its upregulation is a protective, compensatory response during neurodegeneration in animal models of PD. We had also found that it protects against neurodegeneration by activating pro-survival ERK1/2, AKT pathways, countering pro-apoptotic signals, and bolstering mitochondrial biogenesis (Gordon et al., 2016). We recent had also found that PK2 reduces neuroinflammation by activating pro-survival astrocytic phenotype. Yet, little is known about the regulation of PK2 during neurotoxic stress in dopaminergic neurons, or the relationship of PK2 with other neurotrophic factors co-regulated with it during neurotoxic stress. The focus of the following chapters will center around PK2’s role during neurotoxic stress—PK2’s transcriptional regulation, PK2’s role in manganese model of Parkinsonism, and the role of PK2 in regulation of GDNF expression.

Background and Literature Review I
The goal of this section is to summarize the pathophysiology of Parkinson’s disease and current understanding of its etiology through discussion of environmental risk factors as well as genetic risk factors for Parkinson’s disease. The proposed mechanisms and hypotheses for disease progression will be presented. Parkinson’s disease animal models relevant to the research presented in this work will also be discussed.

Parkinson’s Disease
Parkinson’s disease (PD) is the second most common neurodegenerative diseases behind Alzheimer’s disease. The mean onset for PD is 60 years of age, and it affects 10 million people worldwide, or 4-5% of people over 85, for whom mortality rate is 2-3 times that of the general population (de Lau et al., 2006; Lang ; Espay, 2018). Its prevalence and incidence nearly exponentially increase by age, and peak after 80 years of age (Callesen, Scheel-Krüger, Kringelbach, ; Møller, 2013; Driver, Logroscino, Gaziano, ; Kurth, 2009). Each year, 50,000 new cases of PD are diagnosed in the US alone, and the more than 1 million PD patients in the US exact a yearly national economic burden exceeding $14.4 billion in 2010 and is projected to exceed $ 18 billion by 2050 (Kowal et al., 2013). Due to relative complex etiologies of neurodegenerative diseases such as PD, they remain largely incurable, and are projected by the World Health Organization to overtake cancer as a leading cause of death by 2050 (Menken, Munsat, ; Toole, 2000). Due to its progressively debilitating nature, PD is one of the most feared diseases of mankind (Fan et al., 2015).
In his famous monograph “An Essay on the Shaking Palsy” in 1817, James Parkinson described a condition characterized by “Involuntary tremulous motion, with lessened muscular power, in parts not in action and even when supported; with a propensity to bend the trunk forwards…” (Parkinson et al., 2002), a condition with the earliest reference dating back to 600 B.C. in the ancient Indian medical literature by the name of Kampavata (Manyam, 1990; Ovallath ; Deepa, 2013). Today, PD is characterized clinically by akinesia, bradykinesia, rigidity, postural instability and fatigue (Barzilai ; Melamed, 2003; Jarraya et al., 2009; Nassif ; Pereira, 2018). The degree and severity of the motor disability is measured clinically by the Unified Parkinson’s Disease Rating Scale (UPDRS) that measure motor functions of activities of daily living (Fahn & Elton, 1987). The pathophysiology of PD is characterized by the death of dopaminergic neurons in the SNpc resulting in dopamine deficiency in the caudate putamen and leading to motor symptoms. Post-mortem analysis of PD patient brains also revealed accumulation of Lewy bodies in dopaminergic neurons, consist largely of insoluble alpha synuclein, which seem to accompany dying neurons and is therefore the second hallmark of PD. Although is found to be protective during early stages of PD (Harischandra, Jin, Anantharam, Kanthasamy, & Kanthasamy, 2015), during later stage of PD, the misfolding of ?-synuclein and accumulation of Lewy bodies are thought to be an essential mechanism causing the lesions seen in PD and dementia with Lewy body (Chartier & Duyckaerts, 2018). PD is also associated with numerous non-motor symptoms, include: hyposmia (Richard L. Doty, 2017; Masala et al., 2018), taste perception deficit (Cecchini et al., 2014), psychosis(Friedman, 2013), depression (Thobois, Prange, Sgambato-Faure, Tremblay, & Broussolle, 2017), constipation (Rossi, Merello, & Perez-Lloret, 2015), sleep disturbances (Videnovic & Golombek, 2013) and sexual dysfunctions (Varanda et al., 2016), all of which further reduce quality of life for PD patients (Martinez-Martin, 2011; Pfeiffer, 2016; Santos-García & De La Fuente-Fernández, 2013). Interestingly, these non-motor symptoms generally do not correlate with UPDRS scores or disease duration, and some symptoms, such as olfactory function, as measured by University of Pennsylvania Smell Identification Test (UPSIT), is correlated with age (Haehner, Hummel, & Reichmann, 2009). Together with other symptoms such as sleep disturbances, these non-motor symptoms could precede the development of motor symptoms by a decade or more (Postuma et al., 2012). Although the etiology of the disease is still largely not understood, our understanding of the molecular mechanisms behind PD has greatly advanced in the last two decades.

Genetic Risk Factors in PD Pathogenesis

Until two decades ago, PD was not considered to have a genetic basis (Billingsley, Bandres-Ciga, Saez-Atienzar, & Singleton, 2018). Advances in molecular genetics and sequencing technology had enabled the discovery of underlying causes of PD for several families with increased PD occurrence. SNCA mutation discovered in 1997 was the first mutation found to cause monogenic PD (M H Polymeropoulos et al., 1997), where a single mutation is sufficient to cause the disease. LRRK2 mutation was discovered in 2004 (Zimprich et al., 2004), the most common cause of monogenic PD. To date, six genes, SNCA (PARK1), LRRK2 (PARK8), Parkin (PARK2), PINK1 (PARK6), DJ-1 (PARK7), and ATP13A2 (PARK9) are found to cause monogenic PD. Mutations in these genes comprise of 30% of familial PD cases, and 3–5% of sporadic PD cases. Beyond mutations in genes that cause monogenic PD, mutations in other genes have been found to contribute as risk factor for PD. PRKN, PINK1, and DJ-1 mutations are associated with early-onset PD, and altogether, 28 distinct chromosomal regions have been found to be associated with PD (Klein et al., 2012). Yet, mutations in these genes accounts for only 10% of all cases of PD, while most cases of PD remain idiopathic. It had therefore come to light that PD is most likely multifactorial, resulting from gene-environment interactions that act on mutated susceptibility alleles present in the patient, to impact the aging brain.

Environmental Risk Factors and Role of Mn Exposure in development of PD and PD-Like Pathologies
A large-scale meta-analysis analyzed 30 potential risk factors for developing PD (Noyce et al., 2012). Interestingly, tobacco smoking was negatively associated with PD, but the association is confounded by the fact that early-stage PD patients experienced decreased responsiveness to nicotine and therefore quit tobacco use more often (Shahi & Moochhala, 1991). Although several other studies had found protective effects of nicotine in tobacco smoke and suggested its role in PD therapy (Barreto, Iarkov, & Moran, 2015; Ma, Liu, Neumann, & Gao, 2017), a history of smoking increased the risk of dementia in PD by almost two fold (Yaqian Xu, Yang, & Shang, 2016) and the nature of its effect on PD remain controversial (Ascherio & Schwarzschild, 2016; Shahi & Moochhala, 1991). The meta-analysis had found 11 environmental factors considered as risk factors for PD. Among them, the top risk factors are pesticide exposure, well-water drinking, beta-blocker use, prior head injury, agricultural occupation, and rural living (Foubert-Samier et al., 2012). Exposure to pesticides such as rotenone, paraquat, dieldrin, organophosphates and Mn (in the form of pesticide Maneb) are implicated in development of PD (A. G. Kanthasamy et al., 2008; Roede & Miller, 2014; Semchuk et al., 1992; Singh et al., 2018; Tanner et al., 2011). Additionally, exposure to Mn in an occupational setting could increase risks for developing Parkinsonian symptoms by 3 – 10 fold (Gorell et al., 1999). Longitudinal cohort study of 886 welding-exposed workers had found yearly changes in UPDRS scores dependent on dose of Mn exposed, which was especially severe in flux core arc welders who work in a confined space (Racette et al., 2017). Farmers exposed to Mn-containing pesticides Maneb and Mancozeb also showed adverse neurological effects (Thrash, Uthayathas, Karuppagounder, Suppiramaniam, & Dhanasekaran, 2007).

Mn exposure, target organ, clinical features, pathophysiology

Mn is a trace element found in all life on earth. It is an essential cofactor needed for such fundamental cellular processes as metabolism of fats and carbohydrates, regulation of blood sugar, and calcium absorption (Erikson, Syversen, Aschner, & Aschner, 2005). Mn is a cofactor in the reactive catalytic centers of essential enzymes manganese catalase and MnSOD, both of which convert and reduce oxidants (Species, Finkel, & Species, 2001). Mn exists in various chemical forms including several oxidation states (Mn2+, Mn3+, Mn4+) (Rask, Miner, & Buseck, 1987; Reaney & Smith, 2005), a property which enabled its wide industrial uses. Current safety measures may not adequately protect welders from aerosolized metals, and consequently, risks of overexposure to Mn for workers such as welders and miners in occupational settings are increased, usually via dermal absorption and inhalation routes. Excessive absorbed Mn can accumulate in the brain, a major organ of Mn toxicity during overexposure, where it preferentially concentrate in globus pallidus and striatum in monkeys (Dastur DK, Manghani DK, 1971; Fujii, 1975) and humans (Aschner, 2006; Aschner, Guilarte, Schneider, & Zheng, 2007; Sarkar et al., 2018; Wooten, Aweda, Lewis, Gross, & Lapi, 2017; Yokel, 2009), to cause cell death in the basal ganglia and disruption of the nigrostriatal pathway. Disruption of several neurotransmitter systems, particularly the dopaminergic system, results in pathology that manifests itself in a set of extrapyramidal symptoms similar to Parkinson’s disease, called manganism: postural instability, bradykinesia, micrographia, and a characteristic cock-walk caused by dystonia of the legs (Cersosimo ; Koller, 2006; C. W. Olanow, 2004; Perl ; Olanow, 2007). Due to these similarities, Mn exposure has been used as a mouse model for PD since 1973 (Villalobos et al., 2009). Furthermore, mitochondrial dysfunction and oxidative stress processes occurring in PD are strikingly similar to the neurodegenerative processes occurring also in manganism, suggesting that these dysregulated processes could be common in both manganism and PD.

Mn mechanisms of toxicity: transport, mitochondrial accumulation, induction of epigenetic changes
In the body, Mn can be transported into the cell by a few metal transporters: divalent metal transporter 1 (DMT1), bicarbonate ion symporters ZIP8 and ZIP14, transferrin receptor (TFR), solute carrier-39 (SLC39) family of zinc transporters, ATP13A2, and magnesium transporter HIP14. Among these, DMT1 and TFR are most studied in their function to transport Mn. In the cell, Mn is sequestered to the mitochondria by Ca2+ uniporter (Gunter, Gavin, Aschner, ; Gunter, 2006), and gradually increase in concentration due to extremely slow efflux (Gavin, Gunter, ; Gunter, 1990; Martinez-Finley, Gavin, Aschner, ; Gunter, 2013). Mn treatment in cells has been shown to increase oxidative stress, and some studies had shown that Mn could inhibit mitochondrial efflux of Ca2+, thus increasing Ca2+ concentration and leading to overproduction of ROS (Tjalkens, Zoran, Mohl, ; Barhoumi, 2006).
In dopaminergic neuronal cells, Mn could cause caspase-3-dependent PKC? cleavage of an active PKC? product (41?kDa) capable of translocating into the nucleus for expression of proapoptotic genes (Masashi Kitazawa et al., 2005; Latchoumycandane et al., 2005; D. Zhang, Kanthasamy, Anantharam, ; Kanthasamy, 2011). Meanwhile, PKC? also negatively regulates tyrosine hydroxylase (TH), the rate-limiting enzyme in dopamine synthesis, by enhancing protein phosphatase-2A activity in dopaminergic neurons (D. Zhang et al., 2011). Since Mn preferentially accumulates in the basal ganglia, it could profoundly affect dopamine release from dopaminergic neuronal terminals in the striatum (Guilarte, 2010). Furthermore, Mn exposure is capable of inhibiting the acetylation of core histones in SH-SY5Y cell models of PD (Guo et al., 2018) and inducing epigenetic changes in several PD-associated PARK genes (Tarale et al., 2017). Maternal developmental Mn exposure in mice could affect neurogenesis of newborn mice and increased promoter hypermethylation and transcript down-regulation of large number of genes even through postnatal day 77 (L. Wang et al., 2013).
Currently, Mn overexposure is treated with chelation therapy for immediate sequestration of remaining Mn in body fluids, most commonly with edetate calcium disodium (EDTA), and more recently, with para-aminosalicylic acid (PAS), for the primary goal of removing the patient from further exposure (J. Lee, 2000). The extrapyramidal symptoms of manganism patients initially respond to L-DOPA treatment after short-term observation (Mena, Court, Fuenzalida, Papavasiliou, ; Cotzias, 1970), but failed to show significant benefit from L-DOPA treatment in patients with chronic manganese poisoning. Furthermore, no treatment is available for the neurorestoration of degenerated neurons affected by Mn overexposure. Since cellar mechanisms of toxicity overlaps with PD, the common pathways of neuronal degeneration are similar between manganism and PD, the differences of manganism vs PD arises out of preferential accumulation of Mn in basal ganglia. It is therefore speculated that protective strategies against neurodegeneration caused by Mn overexposure could also have similar protective effects against neurodegeneration in PD.

Induction of Neuroinflammation from Exposure to Environmental Neurotoxicants

Inflammation of the brain, or neuroinflammation, triggered by environmental toxicants, is recognized as a major contributor of PD progression during neurodegeneration. Although divergent molecular events characterize the progression of Parkinson’s disease (PD), Alzheimer’s disease, and amyotrophic lateral sclerosis, neuroinflammation emerges as one common theme among the cellular events underlying these neurodegenerative diseases (Block ; Hong, 2005; Heneka et al., 2015; J. K. Lee, Tran, ; Tansey, 2009; McGeer ; McGeer, 2004; Mrak ; Griffin, 2005). Recent work from animal studies and epidemical studies have provided evidence that neuroinflammation contributes to progressive cell death events (Y. S. Kim ; Joh, 2006; McGeer ; McGeer, 2008).
Neuroinflammation is mediated primarily by two types of glial cells: microglia and astrocytes. normally quiescent in a healthy brain, glia are immune effector cells that provide critical support for the neurons which they surround. Resting microglia secrete low amounts of pro-inflammatory cytokines and exhibit a ramified morphology, a characteristic of “surveilling” microglia. Similarly, astrocytes normally participate in the glutamine-glutamate cycle, maintain glial-neuron contact, and secrete neurotrophic factors. However, upon toxicant insult, both can be activated. Microglia undergo morphological changes into an amoeba-like morphology, and dramatically increase the release of pro-inflammatory cytokines. Astrocytes upregulate surface expression of GFAP, and secrete pro-inflammatory cytokines. Activation of glia is one of the earliest response to injury that is intended to be a compensatory, neuroprotective response to neurodegeneration as evident by glial release of neurotrophic factors during early phase of injury, and is intended to facilitate phagocytic removal of dead cells or debris or preserve neuronal survival.
However, chronic injury, as can be caused by repeated exposure to environmental toxicants, trigger glial activation that could establish a self-sustaining cycle of neuroinflammation (Orr, Rowe, ; Halliday, 2002), which manifests itself as proliferation of activated microglia and astrocytes, activation and nuclear translocation of NF-?B and elevation of cytotoxic cytokines, including tumor necrosis factor-? (TNF-?), inducible nitric oxide synthase (iNOS), nitric oxide (NO), interleukin (IL)-1beta, IL-6, cyclooxygenase-2(COX-2), and prostaglandins E2 (PGE2).
Potent inducers of neuroinflammation, such as MPTP, could elevate brain inflammatory factors years after a single administration in nonhuman primates, long after initial injury and cell death (Orr et al., 2002; Tansey, McCoy, ; Frank-Cannon, 2007), suggesting that a cascade of events involving neuronal cell death could aggravate glia activation and produce a vicious cycle of a self-propelling and self-sustaining neurotoxicity. In the case of PD, sustained neuroinflammation in the substantia nigra and elsewhere contribute to disease progression, as it is detrimental to dopaminergic neuron survival. Postmortem brains of AD, PD, and HD patients have revealed extensive neuroinflammation with distinct elevation of pro-inflammatory cytokines TNF-?, IL-12, IL-6, and IL-1?. While each of the chemicals described below (manganese, MPTP, rotenone, and dieldrin) may impact cellular processes through unique mechanisms, a common pathway of toxic neuroinflammation emerges.

Mn
Manganese (Mn) is an essential trace element in all known living organisms for such fundamental cellular processes as fat and carbohydrate metabolism, regulation of blood sugar, and calcium absorption (Erikson et al., 2005). Mn has a wide range of industrial uses, and is present in welding fumes, mining dust, and fungicides. Risks of its occupational overexposure are therefore increased for welders (Ngwa, Kanthasamy, Jin, Anantharam, ; Kanthasamy, 2014), miners (Gendelman et al., 2015), or famers who are exposed to Mn-based pesticides Maneb (Mn ethylene-bis-dithiocarbamate) and Mancozeb. Mn is transported into neuronal cells by divalent metal transporter 1 (DMT1), and chronic overexposure to Mn could directly cause neuronal cell death (Roth, Horbinski, Higgins, Lein, ; Garrick, 2002). Mn3+ participate in the oxidation of dopamine to aminochrome (dopaminochrome), the precursor of neuromelanin. However, aminochrome accumulation in cell culture medium induces acute cell death (Paris ; Segura-Aguilar, 2011). In neuronal cell cultures, Mn treatment induced cytochrome C release, caspase-3 activation, and DNA fragmentation. More recently, mechanistic studies had shown that caspase-3 activation following Mn treatment could proteolytically activate protein kinase Cdelta (PKC?), a noncanonical member of protein kinase C family of kinases(Masashi Kitazawa et al., 2005). Cultured neurons expressing a dominant negative form of PKC? protein were found to be resistant to Mn-induced apoptosis. Small interfering RNA suppression of PKC? expression or cotreatment with the PKC? inhibitor rottlerin significantly blocked Mn-induced DNA fragmentation, suggesting that activation of PKC? signaling represents a major mechanism in Mn-induced apoptotic cell death (Latchoumycandane et al., 2005). Mn could directly cause neuronal cell death, while at the same time also induce neuroinflammation (Sarkar et al., 2018). In astrocyte cultures, Mn exposure increased expression of GFAP, a marker for activated astrocytes, and stimulated the release of proinflammatory cytokines. Mn also disrupts astrocytic regulation of glutamate by increasing the promoter activity of Ying Yang 1, (YY1), a negative regulator of glutamate transporter (GLT-1) (Karki et al., 2014). Microglial activation in response to Mn is even more pronounced. In primary and immortalized microglia cultures, manganese could induce much higher excessive expression of TNF-?, IL-12, IL-6, IL-10, and IL-1? and stimulate iNOS protein levels through the activation of the NF-?B and p50/p65 nuclear translocation to induce expression of inflammatory genes (Filipov, Seegal, ; Lawrence, 2005; Verina, Kiihl, Schneider, ; Guilarte, 2011). These toxic inflammatory factors are directly cytotoxic to neurons (Filipov et al., 2005)

MPTP/MPP(+)
MPTP is a contaminant during the synthesis of meperidine (pethidine), an opioid drug used illicitly. Users who accidently injected MPTP developed irreversible PD-like symptoms of rigidity and bradykinesia (J. Langston, Ballard, Tetrud, ; Irwin, 1983). Postmortem analysis of these patients showed selective dopaminergic cell death and marked increase in glial activation in the substantia nigra (J. W. Langston et al., 1999). As a lipophilic pro-toxicant, MPTP readily crosses the blood brain barrier, where it is metabolized by monoamine oxidase B (MAO-B) in astrocytes into active toxicant MPP(+) (Heikkila, Manzino, Cabbat, ; Duvoisin, 1984, 1985). The exact mechanism of MPP(+) release from astrocytes is not clear, but nonetheless, it can be taken up by dopaminergic neurons via the dopamine transporter (DAT) (Gainetdinov, Fumagalli, Jones, ; Caron, 2002). MPP(+) accumulates in the mitochondria to inhibit complex I of the electron transport chain, thereby causing a dissipation of mitochondrial potential, a collapse in ATP production, and neuronal cell death. Surface expression of signals by dying neurons to stimulate phagocytosis by microglia leads to greater infiltration of the region by activated microglia to remove dead neurons(Tansey et al., 2007).
In microglia, MPP(+) can be taken up via organic cation transporter 3 (OCT3) and potentiates LPS-induced TNF-? expression(Qian He, Wang, Yuan, ; Wang, 2017). MPP(+) downregulates miR-7116-5p in microglia, which normally suppresses overproduction of TNF-?, suggesting that MPP(+)-induced dysregulation of this suppression sends TNF-? production in microglia into overdrive (Qian He et al., 2017).
MPP(+) in astrocytes activates inflammasomes, further aggravates neuroinflammation (Qiao et al., 2016). Mechanistically, MPP(+) downregulates ATP13A2, thereby increasing lysosomal membrane permeabilization and cathepsin B release. This induces activation of nod-like receptor protein 3 (NLRP3) inflammasome, to produce excess IL-1? from astrocytes (Qiao et al., 2016). Furthermore, MPTP treatment reduced IGF-1 levels in the substantia nigra of aged rats (Labandeira-Garcia, Costa-Besada, Labandeira, Villar-Cheda, ; Rodríguez-Perez, 2017), which reduced the stimulation AKT phosphorylation and activation (Madathil et al., 2013). When optimally activated, AKT signaling leads to a robust anti-inflammatory response by inducing DJ-1 and HIF1alpha levels (Jha, Jha, Kar, Ambasta, ; Kumar, 2015). Disruption of this pro-survival, anti-inflammatory signaling pathway compounds the inflammatory effects produced by glial cells.

Rotenone
Rotenone is a broad-spectrum insecticide, piscicide, and pesticide, and exposure to it has been implicated in the development of neuroinflammation and the progression of PD (Betarbet ; Greenamyre, 2008; Cannon et al., 2009; Dranka, Zielonka, Kanthasamy, ; Kalyanaraman, 2012; Sherer, Betarbet, Kim, ; Greenamyre, 2003; Tanner et al., 2011). Rotenone is a known inhibitor of mitochondrial complex I; its exposure in neuronal cells causes the collapse of mitochondrial potential and release of mitochondrial cytochrome C, initiating the intrinsic apoptotic pathway. In microglia, rotenone exposure directly induces the phosphorylation of p38 and activation of p38 MAPK pathway, a stress activated protein kinase pathway that leads to NF-?B activation and p50/p65 transcription factor nuclear translocation for expression of pro-inflammatory genes IL-1beta and TNF-? (Bachstetter et al., 2011). Although the role of NF-?B remains controversial, activation of NF-?B in microglia and astrocytes generally results in production of proinflammatory cytokines TNF-? and IL-6, which are produced in particularly high amounts by glial cells (Mattson ; Camandola, 2001). In addition, transcription factor p65 can bind to the NF-?B consensus sequence on the COX-2 promoter, leading to expression of COX-2, a major pro-inflammatory mediator (Minghetti, 2004).
Rotenone exposure also induces the activity of GSK3? (Han, Casson, Chidlow, ; Wood, 2014), a crucial regulator of the inflammatory response. GSK3? has an inhibitory effect on CREB nuclear translocation, thereby allowing for transcription of pro-inflammatory genes such as interleukin-1-Beta (IL-1?) and TNF-?(Maixner ; Weng, 2013). Inversely, inhibition of GSK3? increases CREB DNA binding activity and increases transcription of anti-inflammatory IL-10 (Maixner ; Weng, 2013). Rotenone-induced cytotoxicity in cultured dopaminergic neuronal cells could be attenuated by the GSK3? inhibitor SB216763 (Hongo et al., 2012), suggesting that GSK3? signaling could mediate rotenone-induced neuronal toxicity. In astrocytes, GSK3? induction generally increases iNOS, nitric oxide (NO), cyclooxygenase-2(COX-2), prostaglandins E2 (PGE2), and TNF-? expression (H.-M. Wang et al., 2013).
In microglia, GSK3? mediates the increased release of pro-inflammatory cytokines (Q. Cao, Karthikeyan, Dheen, Kaur, ; Ling, 2017). A general mechanism could be mediated through activation of mixed lineage kinases (MLK). Both rotenone and MPP(+) can activate GSK3?, and decreasing GSK3? activity blocked MLK3 signaling cascades through disruption of MLK3 dimerization-induced autophosphorylation, inhibiting the downstream stress activated JNK pathway, ultimately leading to a decrease in TNF-? secretion(L. H. Wang, Besirli, ; Johnson, 2004; M.-J. Wang, Huang, Chen, Chang, ; Kuo, 2010). Furthermore, the pan-MLKs inhibitor, CEP-1347, was shown to prevent dopaminergic neuronal loss in pre-clinical animal models of PD (A. Kanthasamy et al., 2012).

Dieldrin
Dieldrin is an organochlorine pesticide that was used in the US in the late 1980s for control of insects. Despite its discontinuation of use in early 1990s, it persists heavily in the environment. Dieldrin could stimulate pro-inflammatory IL-8 and TNF-? responses in Jurkat cells(Saliha, Eric, Frederique, Etinne, ; Christian, 2018), and can activate peripheral neutrophils to promote the production of IL-8 (Pelletier et al., 2001). The highly lipophilic dieldrin readily crosses the blood brain barrier (Moretto ; Colosio, 2011) and its exposure has been implicated in PD disease progression (Cowie et al., 2017; A. G. Kanthasamy et al., 2008). Presence of dieldrin could be seen in postmortem analysis of brains in some PD patients but not in control brains (Corrigan, Murray, Wyatt, ; Shore, 1998; Fleming, Mann, Bean, Briggle, ; Sanchez-Ramos, 1994; A. G. Kanthasamy, Kitazawa, Kanthasamy, ; Anantharam, 2005), together with high levels of glial activation, thus correlating dieldrin exposure with neuroinflammation. While less is known about dieldrin’s role in induction of neuroinflammation, evidence from in vitro studies so far has shown that dieldrin could activate microglia via NADPH oxidase 2 (NOX2) activation (Taetzsch & Block, 2013) to increase ROS, thereby causing a pro-inflammatory state in microglia. Additionally, dieldrin could activate the non-receptor tyrosine kinase, Fyn kinase (Saminathan, Asaithambi, Anantharam, Kanthasamy, & Kanthasamy, 2011), which, could phosphorylate PKC? at the Y311 site, resulting in increased PKC? kinase activity (Panicker et al., 2015). Similar to Mn, dieldrin could also induce neuronal cell death by activating caspase-3, then caspase-3-dependent proteolytic activation of PKC? (M. Kitazawa, Anantharam, & Kanthasamy, 2003). Furthermore, dieldrin exposure has been found to induce aberrant acetylation of core histone H3 and H4 within minutes of exposure in dopaminergic neuronal cultures (C. Song, Kanthasamy, Anantharam, Sun, & Kanthasamy, 2010). Alternatively, dieldrin treatment could also induce caspase-dependent proteolytic cleavage and inactivation of poly(ADP-ribose) polymerase (PARP), a cellular pathway for DNA damage repair that can be inactivated by extensive DNA damage and activated by pro-survival BCL-2 overexpression (Masashi Kitazawa, Anantharam, Kanthasamy, & Kanthasamy, 2004). Interestingly, dieldrin exposure-induced epigenetic changes in Jurkat T cells could cause increased transcription of human endogenous retroviruses, vestiges of ancient retroviral infections of the germline normally kept in check by heterochromatin (Saliha et al., 2018). Expression of these degenerated copies of viral genes nonetheless could induce inflammation and is implicated in multiple sclerosis.

Synergistic Effects
The cause of neuroinflammation and the resulting neuronal cell death is generally multifactorial, involving exposure to chemicals that could initiate such toxic conditions, or several such initiators that could potentiate neuroinflammatory effects, as well as cellular responses to disturbances of signaling pathways. Bacterial inflammogen LPS has been shown to potentiate neurotoxic effects induced by environmental toxicant rotenone (Gao, Hong, Zhang, & Liu, 2002). Similarly, Mn-containing pesticide Maneb could act synergistically with paraquat, another pesticide with structural similarity with MPTP, to produce compounded toxicity on the dopaminergic system (Thiruchelvam, Brockel, Richfield, Baggs, & Cory-Slechta, 2000). Such multifactorial mode of exposure more realistically reflects the condition experienced in patients during neuroinflammation induced by neurodegeneration.
As mechanisms of action for environmental toxicants to cause neuroinflammation are elucidated, therapeutic strategies are devised against these target pathways. For example, the Fyn inhibitor saracatinib is used to inhibit Fyn-PKC? signaling in status epilepticus-induced neuroinflammation (Sharma et al., 2018), and rotenone-induced cytotoxicity in cultured dopaminergic neuronal cells could be attenuated by the GSK3? inhibitor SB216763 (Hongo et al., 2012). Epidemiology studies and animal studies had suggested that non-steroidal anti-inflammatory drugs (NSAIDs) could lower risk of PD (McGeer & McGeer, 2004). Lastly, neuroinflammation could also aggravate tissue damage by inducing oxidative stress in glia, through production of high levels of ROS, and through arachidonic acid signaling by activating cyclooxygenase (COX) and lipoxygenase (LOX) pathways.

Hypotheses Regarding PD Pathogenesis

Several hypotheses have attempted to correlate environmental exposures with the development of PD motor and non-motor pathologies.

Braak staging
The Braak staging hypothesis attempts to explain PD pathology in terms of PD temporal and spatial progression; it correlates the exposure of environmental agents (chemicals, viruses, bacteria) in the peripheries (nose, gut) with initial appearance of Lewy bodies in the olfactory system and enteric nervous system (Braak et al., 2003; Jellinger, 2009), resulting in gut constipation and lessened sense of smell. Interestingly, a 1918 pandemic flu that became strongly associated with post-encephalitic parkinsonism also had lend strength to this hypothesis (Billingsley et al., 2018). During stage II-III of the Braak staging, Lewy body pathology subsequently enters the central nervous system, and by spreading in a caudal-to-rostral direction, affects the brain stem, followed by substantia nigra, (Dickson, Uchikado, Fujishiro, & Tsuboi, 2010), damaging particularly the non-myelinated dopaminergic neurons. In later stages of PD, Lewy pathology have been found in the cortex, corresponding to the later Braak stages (Del Tredici & Braak, 2016).

Oxidative stress and mitochondrial dysfunction
SNpc dopaminergic neurons are uniquely vulnerable to damage. These non-myelinated neurons, number approximately 300,000-600,000 in humans (Chinta & Andersen, 2005; Schultz, 2007), have extensive innervations extending into the striatum (each neuron has an upwards of 150,000 presynaptic terminals in the striatum (Roberts, Force, & Kung, 2002)), resulting in neuronal soma accounting for only 1% of cell volume (Sulzer, 2007). The extensive networks of innervations require relatively higher energy demands and are thus more susceptible to defects in organelle trafficking. Further, since neuromelanin is a major iron storage in the brain, the highest concentrations of iron are found in the SN and the striatum (Fernandez, Ferrer, Gil, & Hilfiker, 2017). The accumulation of excess iron is further substantially increased in brains of PD patients (Jiang, Luan, Wang, & Xie, 2006). Excessive iron can cause hydroxyl radical production via the Fenton reaction, leading to oxidative stress accompanied by the oxidation and modification of proteins, lipids, carbohydrates and DNA. The oxidative stress is compounded by the spontaneous auto-oxidation of dopamine to produce O2-, which is converted to H2O2 (Miyazaki & Asanuma, 2008) in mitochondria of dopaminergic neurons, inducing mitochondrial damage (Surmeier & Schumacker, 2013). The convergence of oxidative stress and mitochondrial dysfunction make SNpc dopaminergic neurons particularly vulnerable. Borrowing a concept from cancer biology where multiple hits by initiator carcinogens, promoter carcinogens, and loss of apoptotic signal balances are required for cancer progression and metastasis, the multiple hit hypothesis for PD etiology focuses on environmental exposure as “primary hits” that initiate neuronal stress in combination with loss of protective pathways in neurons as “secondary hits”, resulting in dopaminergic neurodegeneration (Sulzer, 2007).

Olfactory vector hypothesis
Another hypothesis, the olfactory vector hypothesis for PD, attempts to correlate olfactory deficits, a symptom manifested in almost 90% of both Alzheimer’s disease (AD) and PD during early stages of these diseases (Richard L. Doty, 2008), with the initiation of AD and PD pathologies (R L Doty, Reyes, ; Gregor, 1987; Ward, 1986; Wattendorf et al., 2009), with some proponents even suggesting that PD is primarily an olfactory disorder with accompanying motor dysfunctions during later stages (Hawkes, 1999). This hypothesis for PD pathogenesis takes into consideration that 1) clinical observation of PD found olfactory deficits “rivals or exceeds the prevalence rate of the defining motor signs of the disorder”; 2) findings that suggest pesticide exposure is a top risk factor for PD; 3) a number of viruses, aerosolized metals, chemicals, could be taken up by olfactory sensory neurons and enter the brain through the olfactory bulb, a circumventricular organ that allow toxicants to bypass much of the blood brain barrier. A number of studies indicating that MPTP treatment or LPS treatment administered through the nasal cavity better recapitulate motor and non-motor symptoms compared to intraperitoneal administration seem to support this hypothesis (D. S. Prediger et al., 2011; Qing He et al., 2013).

Animal Models of PD for Studying Gene-Environment Interactions

Several animal models of PD have been developed in order to test hypotheses regarding PD pathogenesis and to evaluate therapeutic strategies devised against PD progression. MPTP is the most frequently used Parkinsonian toxicant applied in the generation of animal models of PD (Beal, 2001; Przedborski et al., 2001), with the obvious advantage that of MPTP was clinically observed in 1979 to produce a human model of the disease upon accidental injection (Davis et al., 1979; J. Langston et al., 1983).
More recently, a genetic model of PD called MitoPark mice has been developed which recapitulate most of the characteristic behavioral symptoms of PD, including the slow progressive dopaminergic degeneration that takes place over a time course of months. This causes a gradual onset of motor function impairment seen in PD patients (Ekstrand et al., 2007). The animal model is constructed by conditionally knockout of mitochondrial transcription factor A (TFAM) in dopaminergic neurons using the Cre-loxP system, leading to reduced mitochondrial DNA expression, and respiratory chain deficiency (Ekstrand et al., 2007). Interestingly, using MitoPark mice, it was found that Mn exposure worsened depletion of striatal dopamine and accelerated the progressive nature of motor deficits already taking place. Importantly, using the MitoPark mice, Mn was found to aggravate the neuroinflammatory processes as indicated by increased IBA-1-immunoreactive microglia cells in the SN, strengthening the experimental evidence the interaction between PD susceptibility genes and environmental exposures. The MitoPark mouse genetic model of PD therefore represent a unique model for studying the interaction of genetic and environmental factors, and their contributions to the neurodegenerative process, during disease progression (Langley et al., 2018).

Background and Literature Review II
Current PD Experimental Therapy Landscape and the Potential of GDNF

PD remain an incurable disease condition and the state-of-the art PD therapy remain focused on disease symptomatic management (Pires et al., 2017). L-DOPA, the precursor for dopamine, discovered by Swedish scientist Arvid Carlsson in the 1950’s to mitigate motor symptoms in Parkinsonian mice, to which the Nobel Prize in physiology or medicine was awarded in 2000, remains the strongest arsenal against disease symptoms (Mercuri & Bernardi, 2005). Currently, pharmacological treatment for enhancing dopamine concentrations or receptor activity in the caudate putamen include dopamine agonists and monoamine oxidase B (MAO-B) inhibitors and Catechol-O-methyl transferase (COMT) inhibitors, both which inhibit degradation of dopamine, which could be used in combination with L-DOPA.
Although more choices in medications and combinations of medications have been recently made available, the fundamental treatment strategy of symptomatic management has remained the same. However, despite an ever more complex combination of therapies, once the “honeymoon” of L-DOPA period has waned, PD patients derive increasingly less efficacy (Rascol et al., 2003) and increasingly more side-effects (Smith, Wichmann, Factor, & Delong, 2012) from the available treatments. Chronic use of L-DOPA leads to dyskinesia induced by stimulation of dopamine receptor 1 expressed by medium spiny neurons in the striatum xxx. To mitigate dyskinesia induced by L-DOPA treatment, dopamine receptor agonists apomorphine, pramipexole and ropinirole are prescribed, either in place of L-DOPA, or more likely during later stages of PD, in combination with L-DOPA (Holloway et al., 2004). However, although they reduce dyskinesia, they also increase orthostatic hypotension and psychiatric symptoms such as psychosis, depression. In the case of orthostatic hypotension, it affects 30% of the PD patients and is treated with L-Threo-dihydroxyphenylserine (L-DOPS) (S., M.V., A., & O., 2013). In the case of PD-induced psychosis, it affects around half of advanced PD patients, and limited options include pimavanserin (Holloway et al., 2004). Surgical interventions for PD symptomatic control involves the implantation of deep brain stimulation (DBS) device into the brain of PD patients, commonly the internal pallidum or the subthalamic nucleus (Fang & Tolleson, 2017). An electrical wire is planted inside the target area, which applies intermittent direct electrical current, with frequencies above the native firing rates, thus generating an inhibitory effect on the target (Benazzouz & Hallett, 2000). However, major drawbacks, aside from being only available to selective patients, include surgical complications, occasional hardware failure, and worsening of neuropsychiatric side effects (Ughratdar, Samuel, & Ashkan, 2015). In late stage PD, treatment-resistant motor symptoms contribute to the burden of complications facing PD patients, for whom falls or accidental choking could be common (Hely, Morris, Reid, & Trafficante, 2005). Clearly, the current challenge is a fundamental treatment therapy that could slow or stop the progressive nature of the disease and restore dopaminergic innervation and function.

Current Clinical Trials for Experimental Therapies
There are three main strategies of current disease modifying-therapies, which are, first, to compensate for dopamine deficiency or other neurodegenerative changes, second, to provide trophic support to neurons and compensate for metabolic abnormalities, and third, to replace lost dopaminergic neurons using cell-based therapies (Lang & Espay, 2018). A summary of clinical trials currently underway is listed in Table 1 (Athauda & Foltynie, 2016; Bergstrom, Kallunki, & Fog, 2016; Brundin, Dave, & Kordower, 2017; Karuppagounder et al., 2014; Kingwell, 2017; Mahul-Mellier et al., 2014; Schneeberger, Tierney, & Mandler, 2016; Surmeier, Obeso, & Halliday, 2017).
Drug name Main Mechanism of Action Status of Trial
NPT200-11 Inhibition of alpha-synuclein misfolding Phase I
PD01A/PD03A Immunization against alpha-synuclein Phase I
R07046015 Immunization against alpha-synuclein Phase II
BIIB054 Immunization against alpha-synuclein Phase II
Nilotinib Inhibition of C-Abl /alpha-synuclein aggregation Phase II
Isradipine Inhibition of Cav1 channels Phase III
Inosine Elevation of urate levels in plasma or cerebrospinal fluid Phase III
Deferiprone Iron chelator/crosses BBB to reduce iron levels in brain Phase III
Exenatide Restoration of protein synthesis processes, autophagy, and mitochondrial biogenesis Completed
AZD3241 irreversible inhibitor of myeloperoxidase Phase II
Nicotine patches Protection of apoptosis for dopaminergic neurons Phase II
EPI-589 (R-troloxamide quinone) Increase in GSH levels in cells Phase II
Stromal stem cells Release of pro-survival factors Phase I/II combined
AAV-2 GDNF Neurotrophic factor for dopaminergic neuronal survival Recruiting for phase I

As a hallmark during PD neurodegenerative process, alpha-synuclein aggregation has been the target of a number of clinical trials, with goals for reduction of its protein synthesis, misfolding, cell-to-cell transmission, or increase in its degradation (Kalia, Kalia, & Lang, 2015). However, a critical downside lies in that these trials focus on the formation of alpha-synuclein aggregation and Lewy body, a late-stage event in the pathophysiology of PD, instead of early events. Moreover, since alpha-synuclein has been found to protective during early stage of disease, more concrete understanding of the role of alpha-synuclein is needed to ensure successes of therapies targeting alpha-synuclein. Some studies focus on specific aspects of the neurodegenerative process as clinical endpoints, rather than improvements in UPDRS, for demonstration of efficacy. For example, AZD3241 is being evaluated in a phase IIa randomized placebo controlled multicenter positron emission tomography (PET) study for its effects on activated microglia in PD patients (Jucaite et al., 2015). EPI-589 (R-troloxamide quinone) is being evaluated in phase II trial using the increases in cellular levels of antioxidant GSH as a biomarker and endpoint (clinical trial NCT02462603). Aside from ones listed in the table, Epigallocatechin gallate (EGCG) is also close to clinical trial stage (Perni et al., 2017; Yan Xu et al., 2016). Lastly, glial cell-derived neurotrophic factor (GDNF) which has been evaluated in a number of high profile clinical trials but did not realize hoped-for results xxx, is currently under a phase II trial in association with the Michael J. Fox Foundation using an improved delivery method. The potential of GDNF as a therapeutic candidate is a topic which will receive special attention in the following sections.

Glial-cell line Derived Neurotrophic Factor
GDNF is the founding member of GDNF family of neurotrophic factors artemin, persephin and neurturin, which are all members of the transforming growth factor-? superfamily (Airaksinen & Saarma, 2002). GDNF is a glycosylated homodimer, with a molecular weight of 33-45 kDa as a dimer and 16 kDa after deglycosylation (L. F. Lin et al., 1993). Synthesized as a 211-amino acid pre-proGDNF, it is cleaved by furin proteases prior to secretion as a 134-amino acid mature protein.
Each of the four members of the GDNF family signal through multicomponent receptor complexes, by preferential binding of GDNF, neurturin, artemin, or persephin to glycosylphosphatidylinositol-anchored cell surface protein GFR?1, GFR?2, GFR?3, GFR?4, respectively. The GDNF-GFR?1 complex binds to the Ret tyrosine kinase, the GDNF co-receptor, for transduction of intracellular signaling (Figure 1).

Figure 1. Signaling mechanisms of GDNF family neurotrophic factors. GFR? receptors are preferentially located on lipid rafts, and the pairs GDNF-GFR?1, neurturin-GFR?2, artemin-GFR?3, persephin-GFR?4 activate Ret tyrosine kinase for signal transduction. Adapted from (Airaksinen & Saarma, 2002)

Regulation of GDNF expression

A number of factors could induce expression of GDNF, such as docosahexaenoic acid, imipramine, adenosine, apomorphine, dopamine, and riluzole (Y. Kim et al., 2011; L. Zhang et al., 2017). During development, GDNF support the growth of midbrain dopamine neurons and motoneurons, as well as peripheral neurons, including sympathetic, parasympathetic, sensory and enteric neurons (Sariola, 2003). During kidney development, a number of transcription factors, Pax2, Eya1, Six1, Six2, Sall1, Foxc1, Wt1, and Hox11 could induce GDNF expression (Saavedra, Baltazar, & Duarte, 2008).
During neuronal injury, a predominant amount of GDNF is secreted by glial cells for neuronal maintenance and survival (Saavedra et al., 2008). Compensatory induction of GDNF is observed in response to lipopolysaccharide (LPS) treatment in astroglioma cells (Appel, Kolman, Kazimirsky, Blumberg, & Brodie, 1997). In the dopaminergic system, Ret expression is required for modulation of protective effects, whereas the absence of Ret abolishes GDNF’s neuroprotective and regenerative effect (Drinkut et al., 2016). In hippocampal and cortical neurons lacking Ret tyrosine kinase, the neuronal cell adhesion molecule (NCAM) is an alternative signaling receptor for GDNF family neurotrophic factors. GDNF-GFRa1 association with NCAM, instead of Ret, induces axonal growth in hippocampal and cortical neurons via binding to NCAM, leading to its interaction with focal adhesion kinase Fak (Paratcha, Ledda, ; Ibáñez, 2003). Transcriptionally, GDNF can be upregulated by Nurr1, Pitx3 in cultured dopaminergic neurons. Activation of p44/42 ERK and PI3-K/AKT pathways have also been show to induce GDNF expression (Tanabe, Matsushima-Nishiwaki, Iida, Kozawa, ; Iida, 2012). GDNF can also be regulated by FGF2, via induction of EGR1 (Shin et al., 2009). Epigenetically, chronic stress could affect the histone modifications and DNA methylation of the Gdnf promoter, leading to changes in GDNF expression (Uchida et al., 2011).
GDNF is currently being assessed for its therapeutic potential in clinical trials, as detailed in the next sections. As one of the most studied neurotrophic factors in the brain, much knowledge has accumulated about its biology and its mechanisms of action. However, transcription factors that drives its expression are not completely known, and even less is known about pharmacological agents that could modulate its expression. Yet such agents could prove immensely helpful in activating neurotrophic factors such as GDNF in brains of PD patients.

GDNF: a promising candidate for disease modifying therapy

Discovery and early studies
With early studies in 1993 that discovered GDNF in glioma cell lines, researchers quickly realized its potential as a potent and selective growth factor for mesencephalic dopaminergic neurons and a stimulator of neurite growth (L. F. Lin et al., 1993). By 1994, direct injection of GDNF had been used in unilateral 6-OHDA rat model of PD for its effects on amphetamine-induced contralateral rotational behavior, and showed a 4-fold decrease in contralateral rotational behavioral in rats that received ipsilateral dopaminergic neuron degeneration (Hoffer et al., 1994). The next year, in 1995, studies injecting GDNF in the SN or in striatum of mice before MPTP injection found potent protection against MPTP-induced nigrostriatal lesion, and observed higher dopamine nerve terminal densities and dopamine levels (Tomac et al., 1995). By 1996, monkeys treated with MPTP were used to successfully show the potential of intraputamental inject of GDNF in rescuing motor deficits as well as restoring striatal dopamine levels (Gash et al., 1996).

First-in-human safety and tolerability studies (Rush Medical Center Chicago)
Perhaps boldened by the non-human primate study, the first-in-human study was done at Rush Medical Center in Chicago in 1999, led by Jeffrey Kordower, in a 65-year-old PD patient, who received monthly intracerebroventricular injections of GDNF (J H Kordower et al., 1999). However, no improvements on motor symptoms were seen, and side effects included nausea, loss of appetite, tingling, intermittent hallucinations, and depression. Importantly, GDNF was thought to not have diffused into the deep layers of the brain (intraparenchymal diffusion). The results suggested that an alternative GDNF delivery system is needed. A second study in monkeys was reported in 2000, where lentivirus coding for GDNF was injected into the striatum and SN of young adult rhesus monkeys treated prior with MPTP. The lenti-GDNF was able to reverse functional deficits and completely prevented nigrostriatal degeneration (J. H. Kordower et al., 2000). The study had also found extensive anterograde and retrograde transport of GDNF, although it was realized much later that GDNF retrograde transport to the SN was required for its protective effects (Tenenbaum ; Humbert-Claude, 2017). A third study in 2002 subjecting MPTP-treated rhesus monkeys with chronic infusion of 5 or 15 µg/day of GDNF using programmable pumps into the lateral ventricle or the striatum, had found restoration of the nigrostriatal dopaminergic system and significantly improved motor functions (Grondin et al., 2002).

Intracerebroventricular GDNF injection (Phase I/II, open-label, led by Gary Nutt)
Prompted by new efficacy data in primates, in the same year in 2002, a prominent Swedish neuroscientist, Patrick Brundin, called for a clinical trial evaluating GDNF as a PD therapy (Brundin, 2002). In a perhaps still premature attempt, the first controlled trial was conducted in 2003, led by Gary Nutt, to test the safety, tolerability, and biological activity of GDNF administered by an implanted intracerebroventricular (ICV) catheter in a dose-escalation study in 50 PD patients for 8 months (Nutt et al., 2003). After the 8 months, some patients became part of the open-label study extending exposure up to an additional 20 months, and some patients were in groups that received the maximum of 4,000 µg GDNF. Unfortunately, the study did not achieve the efficacy that was hoped: motor UPDRS scores were not improved, and moreover, significant adverse effects, such as nausea and vomiting were common hours to several days after treatment with GDNF, and one person died after three weeks, from events unrelated to GDNF treatment.

Direct putamenal GDNF injection (Phase I, open-label, Frenchay Hospital)
Since the first-in-human study reported in 1999 had speculated that ICV-administered GDNF did not penetrate deep into the layers of brain into the SN and striatum, another study reported in 2003, led by Steve Gill, tested safety and feasibility of direct putamenal injection of GDNF protein using catheters, in 5 PD patients (Gill et al., 2003). The recombinant methionine human GDNF was produced by Amgen as liatermin or r-metHuGDNF, and had the sequence of mature GDNF with an N-terminal-tagged methionine. The results from this open label, phase I safety study were positive, with no serious clinical side effects and drastic improvements in motor symptoms. Positron emission tomography (PET) scans of (18)Fdopamine uptake also showed improvements in putamen dopamine storage after 18 months. However, since this study was an open label study, a randomized, controlled study is needed.

Intraputamenal GDNF infusion (Phase II, randomized controlled trial, Amgen)

Encouraged by results from previous open-label study, a phase II trial started in 2003 and reported in 2006, led by Anthony E Lang at the Toronto Western Hospital, in association with Amgen. A randomized controlled trial of 34 PD patients receiving continuous bilateral intraputamenal infusion of liatermin 15 at µ/day, or placebo, in each putamen, was conducted (Lang et al., 2006). Although patients did not develop severe side effects, no change in primary endpoint was seen and no change in UPDRS ratings were found. Additionally, some patients developed antibodies against GDNF. Although are no direct adverse effect from production of antibodies against GDNF during the treatment course, the neutralizing antibodies could render the treatment less effective, and depending on the epitopes that the antibodies targeted, could further decrease endogenous GDNF after treatment cessation.

Unilateral intraputaminal GDNF infusion (Phase I, open-label, led by John Slevin)
Another study was done in parallel for testing the safety and tolerability of a dose-escalation regimen of administration of Amgen’s GDNF protein, increasing from 3 µg/day, to 30µg/day, over a 6-month period. The delivery method used was unilateral intraputaminal GDNF infusion using a multiport catheter that delivered continuous infusion at 2µl/h, into the most affected putamen in the 10 patients with advanced PD (Slevin et al., 2005). At 24 weeks of treatment, UPDRS scores significantly improved, as did balance and gait. The only observed side effects were transient Lhermitte symptoms, a tingling electric shock sensation. The open-label study demonstrated the safety and potential efficacy of unilateral intraputaminal GDNF infusion and showed sustained effects after the withdrawal of treatment. But before a phase II trial could be conducted, Amgen halted all trials using GDNF, since the earlier phase II trial led by Anthony Lang did not meet primary endpoint of improvement in UPDRS, and toxicity in higher doses of GDNF was seen in non-human primates. A follow-up study of patients in the phase I open-label trial confirmed the previous phase II trial; all benefits seen during treatment were lost after 1 year of GDNF infusions, UPDRS returned to patient baseline, and motor symptoms required higher levels of conventional antiparkinsonian drugs (Slevin et al., 2007). Further, GDNF neutralizing antibodies were also detected.

GDNF family – Neurturin phase I/II trials

Another GDNF family member, neurturin, was undergoing a number of clinical trials in parallel with GDNF. In 2008, the first safety and tolerability of intraputaminal delivery of adeno-associated virus serotype 2-neurturin (AAV2-neuturin/CERE-120) was conducted in PD patients by the pharmaceutical company Ceregene (Marks et al., 2008). In this open-label trial, the delivery method was found safe and feasible. But two years later, in 2010, Ceregene reported that the a double-blind, randomized, controlled trial of CERE-120 largely did not meet its primary endpoint (Marks et al., 2008), and importantly, brain tumors were found in both the AAV-neuturin treatment group and the AAV vector group, suggesting that the vector could be associated with tumorigenesis.
Since the report also suggested that impaired axonal transport of neurturin from putamen to SN reduced the effectiveness of neurturin injection in striatum, SN-injection of AAV-neuturin was therefore assessed for functional efficacy in rats (Herzog et al., 2013), then in human (Bartus et al., 2013) for safety and feasibility of direct delivery into the SN, which reported in 2013. For a second double-blind, randomized, controlled trial reported in 2015 (Warren Olanow et al., 2015), AAV-neurturin or placebo were delivered into both putamen and SN in 51 PD patients. Although there were no clinically meaning for adverse events, no significant differences were found between the groups in the evaluation of primary endpoint.
Despite the promising potential of GDNF and GDNF family member neurturin, side effects and efficiency in delivery remain big hurdles. To date, all attempts to develop GDNF or any other therapy as a cure for PD have failed. Candidate therapies shown to work in current animal models of PD had failed to modify the progression of disease and show improvements in human clinical endpoints. Two major aspects that undermined the efficacy of GDNF in humans have been attributed to difficulty in candidate drug evaluation in animal models and delivery of the drug into patients.

Reasons for trial failures: animal model selection

Currently, no single animal model of PD exists that is ideal for all types of studies. Administration of Parkinsonian toxicants such as 6-hydroxydopamine (6-OHDA) and MPTP in rodents and primates have been widely-used to elicit dopaminergic neuronal degeneration and motor symptoms, and have been tremendously useful in evaluating dopamine-based therapies that could potentially provide symptomatic management (Bezard, Yue, Kirik, & Spillantini, 2013; C. Warren Olanow & Kordower, 2009). However, dopamine-based therapies are themselves stopgap solution in finding a disease-modifying therapy.
MPTP is the most commonly used model, and the most commonly used method employs acute administration of MPTP to kill dopaminergic neurons. However, in PD patients, dopaminergic neuronal death occurs in a backdrop of a slowly evolving neurodegenerative processes, with competing compensatory responses, and may not reflect the time-dependent, complex cascade of events relevance to what occurs in the brain of a PD patients. Instead, MPTP treatment in mice could have its own compensatory response different from that of PD patients. For example, acute MPTP treatment could induce TH+ neurons to increase their branching as a compensatory response (D. D. Song & Haber, 2000).
To reduce attrition of drug candidates, more accurate animal models would be needed to evaluate new neuroprotective therapies. To this end, transgenic models based on genes involved in the pathogenesis of PD found through human genetic studies are receiving more attention.

Reasons for trial failures: delivery methods and side effects

Purified GDNF protein was directly injected into the putamen of PD patients in the trials conducted by Amgen. In a separate study using the same protocol and procedures, Salvatore et al had reported that unilaterally infused GDNF in the putamen of adult rhesus monkeys was concentrated around the delivery catheter, with concentration dropping exponentially in the tissues away from the catheter (Salvatore et al., 2006). The volume of distribution of GDNF around the catheter ranged between 87 to 369 mm3, which covers 2%-9% of the total 3-4000 mm3 of the human putamen (Salvatore et al., 2006). Since GDNF protein has a limited diffusion range, an ideal delivery method, aside from pharmacological modulation of endogenous GDNF, would carry GDNF or GDNF-coding cDNA further away from the delivery site, into larger areas of the putamen and SN, for efficient protective effects. Although AAV2-mediated delivery of GDNF cDNA allow GDNF expression in wider areas of the putamen, as seen in the AAV2-neuturin trial, both AAV2-neuturin and vector were associated with tumorigenesis (Marks et al., 2008). Lastly, some patients had developed GDNF-neutralizing antibodies. Little is known about this process, but it is likely that excess GDNF injected into the brain is leaked from the CNS into the peripheries, where the exogenous GDNF protein encounters immune cells in the peripheries. In summary, for the above reasons, pharmacological modulation of endogenous GDNF levels that could be tightly controlled might resolve some of these issues associated with protein diffusion, leakage, and transgene tumorigenicity.

Background and Literature Review III
Prokineticin-2

Structure of Prokineticins

Prokineticin-1 and prokineticin-2 are a pair of secreted signaling neuropeptides discovered around thirty years ago. From the venom of the black mamba snake, a non-toxic cysteine-rich protein was isolated and named mamba intestinal toxin 1 (MIT-1), for its ability to potently contract guinea pig ileum (Schweitz, Pacaud, Diochot, Moinier, & Lazdunski, 1999). Later it was found to be homologous with another protein isolated from the skin of the frog Bombina variegata named Bv8 (Mollay et al., 1999). Since both could potently induce gut motility, when the human homologs were cloned, the MIT-1 homolog was named Prokineticin-1 (PK1) maps to chromosome 1p21. The Bv8 homolog was named Prokineticin-2 (PK2), which maps to chromosome 3p21.1, an unstable chromosomal synteny breakpoint region . PK1 (86 amino acids), and PK2 (81 amino acids) share 45% amino acid identity. Contained within the protein sequences are 19 peptides serving as secretion signal in the mature proteins.
Prominent structural features and determinants include 10 crosslinked cysteine residues forming five disulfide bridges, which is important for the structural integrity of mature PK1 and PK2, as highlighted in purple in Figure 2. The mature PK1 and PK2 proteins are predicted to have an approximate ellipsoid shape. Receptor binding are thought to involve the exposed ends of N-terminus and C-terminus, of which the conserved N-terminus sequence of 6 amino acids AVITGA, and the C-terminal cysteine-rich domain, are indispensable for functional activation of the receptors (Figure 2). Mutations in the AVITGA sequence could abolish its activity, and substitution of alanine for methionine at the N-terminal produce PKR1 and PKR2 antagonists. However, the AVITG peptides alone cannot activate the receptors (Bullock, Li, & Zhou, 2004). Mutations in cysteine residues of the C-terminal domain also result in PK1 and PK2 without biological activity (Bullock et al., 2004).
A PK2 splice variant containing an extra 21 amino acids between exons 2 and 4, named PK2L, could undergo furin proteolytic cleavage of to produce PK2?, which has a 10-fold lower potency as compared to PK2, and showed selectivity for PKR1 (J. Chen et al., 2005). The function of the splice variant in PK2 signaling is still unclear. No splice forms of PK1 are currently found. The isoelectric point (pI) of PK2 is a relative basic 8.85, and pI of PK2L is 10.68, due to its extra 21 amino acids that are rich in lysine and arginine.

Figure 2. Multiple sequence alignment of mature protein sequences for human PK1 (red), human PK2 (yellow), frog PK2 homolog Bv8 (green), and snake PK1 (blue) reveals 10 conserved cysteines in PK1 and PK2. PK1 and PK2 have 44% amino acid identity. The N-terminal sequences (yellow highlighted region) before the first cysteine (AVITGA) is conserved among all species.

PK1 and PK2 Signaling Through PKR1 and PKR2

PK1 and PK2 bind to two cognate receptors prokineticin receptor 1 (PKR1), and prokineticin receptor 2 (PKR2) with similar efficiency (D. C. Lin et al., 2002), although the nonmammals homolog of PK1, MIT-1, is a PKR2-prefering agonist (L. Negri & Ferrara, 2018). In humans, PKR1 and PKR2 are located at chromosome 2p13.3, and 20p13, respectively. PKR1 and PKR2 are G protein-coupled receptors (GPCR), belonging to the family-A of GPCR, and are related to the neuropeptide Y receptor family, which has been found to stimulate migration of neuroprogenitor cells in the SVZ (Decressac et al., 2009). The sequences of both receptors are highly conserved, with nearly 85% identity, with most of the sequence variation appearing in the extracellular N-terminal region (D. C.-H. Lin et al., 2002). The structure of PKR1/PKR2 is shown in Figure 3. The endogenous peptide ligands PK1 and PK2 make contacts on the extracellular surface of the receptors with the second extracellular loop, which is stabilized by a disulfide bond formed between two cysteine residues located on the first and second extracellular loops (L. Negri & Ferrara, 2018).

Figure 3. PKR1 and PKR2 have 85% sequence similarity. The predicted structure of PKR1 and PKR2 (differences shaded in black). The extracellular domain (N-terminus) and intracellular (C-terminus) are labelled. Seven transmembrane regions are shaded in grey boxes.
The binding of PK1 or PK2 to either receptors may be coupled to Gq to induce calcium mobilization (Q Y Zhou, 2006), or Gs to induce cAMP accumulation (J. Chen et al., 2005), or Gi to activate p44/42 MAPK signaling (Ngan & Tam, 2008), , indicating that multiple pathways are involved in prokineticin signaling depending on cell type.
PK1 and PKR1 are more widely distributed in peripheral tissues such as reproductive, gastrointestinal, and blood systems than PK2 and PKR2, who are expressed more widely in the central nervous system (Michelle Y Cheng, Leslie, & Zhou, 2006). High levels of both PKR1 and PKR2 are seen in steroidogenic glands (ovary, testis, adrenal gland and placenta), and PKR2 knockout mice exhibit arrested spermatogenesis (Masumoto et al., 2006). Interestingly, in the fallopian tube, PKR1 expression is stimulated by nicotine exposure, via nicotinic AChRalpha-7 (Shaw et al., 2010).
In the brain, neurons express both PKR1 and PKR2, but overall, PKR2 expression is predominant in the brain. Yet in astrocytes, PKR1 expression is predominant, with little expression of PKR2 (Puverel, Nakatani, Parras, & Soussi-Yanicostas, 2009). The non-selectivity of the PKs and differential expression of PKR1 vs PKR2 suggests that the effect of the ligands PK1 and PK2 are mediated depending on the receptor that are present at the target tissues. For instance, corpus luteum-derived endothelial cells (LEC) express both PKR1 and PKR2, whereas aorta endothelial cells (BAEC) and brain capillaries endothelial cells (BCEC) express only PROKR1. In these tissue types, PROK1 is mitogenic for LEC and BAEC endothelial cells, but has no effect on BCEC endothelial cells (Monnier & Samson, 2010).
PK2 is expressed in the bone marrow, and responds to granulocyte colony-stimulating factor for inducing hematopoietic cell proliferation (Shojaei et al., 2007). Importantly, PK2 expression can also be stimulated by hypoxia through HIF1? to induce growth and angiogenesis by endothelial cells. Similarly, PK2 can be activated by STAT3, which induces proliferative effects (Kujawski et al., 2008). Due to these mitogenic effects, PK2 is implicated in tumorigenesis and survival of tumors in hypoxic conditions. The PK2/PKR1 signaling axis promotes cardiomyocyte survival and angiogenesis, and in the mouse model of myocardial infarction, transient PKR1 transfection in the heart could protect its structure and function (Boulberdaa, Urayama, & Nebigil, 2011). Mechanistically, overexpressing PK2 or PKR1 activates Akt to protect cardiomyocytes against oxidative stress, and siRNA-PKR1 completely reversed the protective effects (Urayama et al., 2007). Further, loss of PKR1 in mice leads to heart and kidney abnormalities due to mitochondrial defects (Nebigil, 2010). One of the earliest discovered properties of PK2 is its ability to induce hyperalgesia to pain stimulus when injected into paws of rats. The dorsal root ganglion cells responding to stimuli express PKR1 as well as the transient receptor potential vanilloid 1 (TRPV1), and PKR1 receptor activation potentiates the activation of TRPV1, significantly lowering the nociceptor threshold to physical and chemical stimuli (Lucia Negri et al., 2002). PKR1-null mice showed impaired acute nociception, compromising what is an essentially protective response against further injury.
PK1 could induce differentiation of bone marrow cells into monocyte/macrophages (Dorsch et al., 2005), and is highly expressed in the endothelial cells of blood vessels and the ovaries to induce angiogenesis, giving it the alternative name endocrine gland vascular endothelial growth factor (EG-VEGF) (Fraser et al., 2005). A number of PK1, PKR1, PKR2 polymorphisms are associated with recurrent pregnancy loss (Y. Cao et al., 2016; M.-T. Su et al., 2010; M. T. Su, Lin, Chen, & Kuo, 2014; Mei-Tsz Su et al., 2010). These and other studies had found that during pregnancy, the expression of EG-VEGF/PK1 and PKR1, PKR2 are controlled temporally in human placenta of the first and third trimester period. This expression pattern suggests their important roles, especially during human early pregnancy, when vascularization of the chorionic villi in the placenta is essential for successful pregnancy. The dynamic expression of PK1 may be regulated by estrogen, progesterone and human chorionic gonadotrophin, and has also been proposed to be regulated by HIF-1? via hypoxia-response element in the promoter region of PK1. Several polymorphisms, PROKR1 (I379V) and PROKR2 (V331M) conferred protection in humans via decreased intracellular calcium influx and increased cell invasiveness (M. T. Su, Lin, Chen, Wu, & Kuo, 2013). Cell proliferation, cell–cell adhesion, and tube organization are not affected, and the mechanisms of protection owing to the polymorphisms are not completely understood.

PK1-GDNF Signaling in the Enteric Nervous System

In the enteric nervous system, PK1 is indispensable for the development of enteric neurons. PK1, but not PK2, is expressed in mouse embryonic gut during enteric nervous system development (Ngan et al., 2007). During this critical period, dysregulation in GDNF or PK1 signaling in the enteric system causes in defects in proliferation and differentiation of neural crest stem cells, and cause a condition in humans known as Hirschsprung disease (Iwashita, Kruger, Pardal, Kiel, & Morrison, 2003). Mechanistically, the proliferative and differentiation effects of PK1 is potentiated by GDNF via upregulating PKR1 expression in enteric neural crest stem cells. Further, overexpression of PKR1 could rescue a lack of RET signaling, suggesting that PK1 and GDNF signaling pathways share some common downstream targets (Ngan et al., 2007).
However, little is known about the involvement of prokineticin signaling with GDNF in the central nervous system during development, and even less during neurodegenerative processes. Since PK2, and not PK1, participates as key regulator of biological processes in the brain, the following sections will focus on the function of PK2 in the brain and during neurodegenerative processes.

PK2 Signaling in the Brain

PK2 transcriptional regulation by circadian rhythm genes

In the brain, PK2 and PKR2 are both highly expressed in the hypothalamus, the limbic system, and olfactory bulbs. PK2 is also highly expressed in the suprachiasmatic nucleus (SCN), the regulator of circadian rhythm in mammals. Levels of PK2 expression oscillates from the highest during the day to almost undetectable at night, and is entrainable by light (Michelle Y. Cheng et al., 2002). In vitro studies revealed that PK2 is under the control of first-order clock-controlled genes, particularly CLOCK-BMAL1, master regulator of circadian rhythm. CLOCK-BMAL1, basic helix-loop-helix transcription factors (bHLH) could transcriptionally regulate PK2 gene expression by binding to enhancer-box (E-Box) element sequence (CACGTG) in the PK2 proximal promoter region. Soon it was identified that other bHLH transcription factors Ngn1 and MASH1/ASCL1 could also bind to the PK2 promoter to regulate its effects on neurogenesis (C. Zhang et al., 2007), HIF1?, as a bHLH transcription factor, was also postulated to bind to the PK2 promoter.
In vivo studies indicated that mice lacking the clock gene also have drastically reduced oscillation of PK2 expression in the SCN. Moreover, intracerebroventricular injection of PK2 protein disrupts normal nocturnal locomotor activity in nocturnal mice. Interestingly, injection of PK2 into the arcuate nucleus, a target of SCN, could suppress feeding without affecting drinking. When injected into the subfornical organ, an indirect target of SCN via the paraventricular nuclei (PVN), PK2 stimulated drinking without affecting feeding (Lucia Negri et al., 2004). Therefore, PK2 output from the SCN could transmit the circadian rhythm of feeding and water regulation in the body. Studies in diurnal rats revealed similar temporal expression pattern as nocturnal mice, which seems to indicate that diurnality lies downstream of the SCN for rodents (Lambert, Machida, Smale, Nunez, & Weaver, 2005). However, PK2 antagonist could produce opposite effects of the PK2 signaling on the arousal levels in the nocturnal mice and diurnal monkey, which could be attributable to the differential expression of receptors for PK2 in the intrinsically photosensitive retinal ganglion cells (ipRGC) that transmit photonic information to SCN for the two animals, therefore suggesting that diurnality could lie upstream of the SCN starting from the response of ipRGC to light (Qun Yong Zhou et al., 2016).

PK2 induces neurogenesis
PK2 signaling is necessary for olfactory bulb neurogenesis and the continuing migration of neuroprogenitor cells (NPC) from the subventricular zone (SVZ) through the rostral migratory stream (RMS) during adulthood. PK2 is highly expressed in the olfactory bulbs, where it acts as a chemoattracant for SVZ-born neuroprogenitor cells migrating towards the olfactory bulb 4. In the SVZ, accumulation of pkr2 transcripts was detected in almost all migrating neuroblasts (Puverel et al., 2009). In the absence of PK2 signaling, it seems that the neuroprogenitors cells do not detach from the rostral migratory stream properly, or they are disoriented about the direction of migration 4 In Pk2-/- mice, the olfactory bulbs do not develop normally and are less than half the size of wildtype controls, and exhibit multiple abnormalities in the various layers of the olfactory bulbs 4. Interestingly, the lack of organization of the olfactory bulb hinders the migration of a group of neurons, gonadotrophin-releasing hormone (GnRH) neurons, from the olfactory placode in the peripheries, through the olfactory bulb, during early development, to the hypothalamus where they extend processes to the median eminence and secrete gonadotrophin-releasing hormone during puberty (Schwarting, Wierman, & Tobet, 2007). The precise mechanisms that guide these small group of neurons through such long distances are unclear, but nonetheless, the lack of organization of the olfactory bulb obfuscates their migratory path. Thus, due to its crucial involvement in the morphogenesis of the mature olfactory bulb, loss of PK2 signaling arising from defects in either PK2 or PKR2 in humans causes a severe form of Kallmann syndrome, characterized by both anosmia and lack of puberty 58. Similar to the case with pregnancy, PKR2 polymorphisms could strongly influence disease severity. In three disease-associated mutations of the PKR2, W178S, G234D, and P290S, the mutant receptors are trapped in the cellular secretary pathway within the cell, never transported to the cell surface or properly integrated into the cell membrane (D. N. Chen, Ma, Liu, Zhou, & Li, 2014).

PKR1/PKR2 Agonists and Antagonists
While PK1 and PK2 make are endogenous and convenient agonists to the cognate PKR1 and PKR2 receptors, efforts have been made to synthesize smaller or more efficient agonists, and to look for antagonists of the receptors. Early studies had found that changes in the conserved N-terminal sequences result in the loss of agonist activity. The first antagonists were made such a way – substitution of alanine with methionine at position 1 created A1MPK1, and addition of a methionine to the N terminus created MetPK1 (Bullock et al., 2004).
A chemical Prokineticin receptor antagonist, 6?(2?Amino?pyridin?3?ylmethyl)?amino?1,3?bis?(4?methoxy?benzyl)?1H?1,3,5 triazine?2,4?dione, was able to block PK2-induced increases in intracellular Ca2+ mobilization CHO cells transfected with human PKR1 or PKR2 (Watson et al., 2012). No off-target effects were detected in a Novartis safety panel of 60 different receptors, suggesting a high-selectivity for prokineticin receptors.
From patented chemical structures release by (at the time) Janssen Pharmaceuticals, 3 related compounds, with general structure 1,3,5-triazin-4,6-diones, were synthesized and were shown to inhibit Bv8/PK2-induced intracellular Ca2+ mobilization (Balboni et al., 2008). The most effective compound, triazine Compound 1, at 300 nM, was able to inhibit close to 96% of cells from responding to 1 nM of Bv8/PK2-induced intracellular Ca2+ mobilization, suggesting that these triazine compounds are potential pharmacological prokineticin receptor antagonists (Balboni et al., 2008).
Recently, three more compounds, PKR-A, PKRA7, and A547 were synthesized and found to have antagonistic activities against either or both of the receptors. PKRA7, a PKR2-preferring antagonist, has been found to have anti-tumor activity in glioblastoma and pancreatic cancer xenograft tumor models (Curtis et al., 2013). PKR-A is a PK2 receptor antagonist that blocks both PKR1 and PKR2. It belongs to a group of morpholine carboxamide prokineticin antagonists (Patent US7855201), and was found to inhibit PKR2 at a IC50 of 48.1 ± 4.6 nM, in CHO cells stably expressing PKR2 (M Y Cheng et al., 2012). An interesting case lies with A457, another morpholine carboxamide prokineticin antagonist. As previously mentioned, three disease-associated mutations of the PKR2 (W178S, G234D, and P290S), cause deficiency in trafficking, resulting in retention of PKR2 within the cellular secretory pathways (D. N. Chen et al., 2014). Using a different modeling methodology employing the use of Phyre2 (Protein Homology/Analog Y Recognition Engine), the PKR2 W178S, G234D, and P290S sequences were modelled. Interestingly, A457 dramatically increased cell surface expression and rescued the function of PKR2 with P290S mutation in a dose- and time-dependent manner, with no de novo protein synthesis. Data showed that A457 could affect the conformation only around its binding site. It is therefore thought that P290S mutation causes a subtle distortion in transmembrane domain VI, and binding of A457 allosterically to the mutant receptor enables it to act as a pharmacological chaperone, to affect its conformation and correct the misfolding (D. N. Chen et al., 2014).
In the recent few years, the first PKR1 agonists were designed, synthesized, and characterized. A group based in France used a homology model computational screen method to screen a library of 250,000 compounds and found 10 potential hits, which were tested in vitro (Gasser et al., 2015). One compound was found to mobilize intracellular Ca2+, activate AKT and ERK signaling. Thirty more compounds, named IS1 through IS30, were then synthesized based on this parent compound. The most potent derivative, IS20, was confirmed for its selectivity and specificity to PKR1. Using this compound, the group showed that IS20 prevented cardiac lesion formation after myocardial infarction, and improved cardiac function. In line with numerous previously published in vitro studies using endogenous PKR1 agonists which confirmed PKR1’s proliferative effects in cardiomyocytes (Boulberdaa, Turkeri, et al., 2011; Boulberdaa, Urayama, et al., 2011; Nebigil, 2010; Urayama et al., 2007), IS20 was shown to promote proliferation of cardiac progenitor cells and neovasculogenesis (Gasser et al., 2015).
We previously showed that PK2 mRNA was highly induced by TNF? in dopaminergic cell culture, and by neurodegeneration in animal models of PD. In MPP+ and MPTP classic Parkinsonian toxicant models, upregulation of PK2 in dopaminergic cells rescued MPP+-induced cell death in cell culture and gene delivery of PK2 rescued MPTP-induced cell death in vivo 6. Our data therefore supported the hypothesis that the upregulation of PK2 is a neuroprotective compensatory response relevant in PD-related neurodegeneration. Given the availability of PKR1 agonists, we aim to evaluate the therapeutic potential of PKR1 agonist in activating protective prokineticin signaling in animal model of PD. Clearly, prokineticin signaling is a complex network involving multiple modes of signal transduction in various cell types, and it is our goal to further elucidate its mechanism of action, especially during neurodegenerative stress events.

CHAPTER 2. CONCLUSIONS AND FUTURE DIRECTIONS
This section serves as a general summary of the research chapters presented previously, and to discuss possible directions for future work on this topic. In this work, we showed that during MPP+–induced neurodegeneration, NRF1, EGR1 and HIF1? are potential regulators of PK2 gene expression. We then demonstrated the mechanism of PK2-mediated protection on Mn-treated dopaminergic neuronal cell culture model of PD via mitochondrial biogenesis and BCL-2 upregulation, and showed its in vivo relevance in Mn mouse model of PD. We lastly demonstrated the clinical relevance and potential of PK2 signaling via pharmacological activation of GDNF in MPTP and MitoPark model of PD.

PK2 Transcriptional Regulation during Neurodegeneration
Despite recent advances in understanding of PK2’s functions in the brain, its role and its transcriptional regulation during neurotoxic stress remain incompletely understood. It has been shown that in the CNS, PK2 can be transcriptionally regulated via Enhancer-box (E-box) binding by basic helix-loop-helix (bHLH) transcription factors. CLOCK/BMAL1, the circadian rhythm master regulators, and Ngn1/MASH1, proneural genes expressed during olfactory bulb neurogenesis, have been determined to regulate PK2 transcription. Further, HIF1?, another bHLH transcription factor, is speculated to regulate PK2 transcription via binding to several putative HIF1? binding sites on the PK2 proximal promoter. Our studies confirmed that overexpressed HIF1? could induce PK2 promoter activity and gene expression, while also confirming the involvement of EGR1 in regulating PK2 in a similar manner. Importantly, we found that except in the striatum, PK2 expression generally co-regulated with EGR1 and HIF1? in Mn overexposure model of neurotoxic stress. Further mechanistic studies conducted in dopaminergic cell cultures using CRISPR cas9-based knockdown and chemical inhibition of HIF1?, demonstrated that found that EGR1 and HIF1? are major regulators of PK2 gene expression during MPP+-induced PK2 upregulation. It has been reported that treatment with DHB, a HIF1? stabilizer, could rescue mitochondrial dysfunction, a major mechanism of DHB-mediated protective effects (D. W. Lee et al., 2009). Since PK2 could induce proliferator-activated receptor-? coactivator 1? (PGC-1?) in the activation of mitochondrial biogenesis and respiration (Gordon et al., 2016), it therefore could be speculated that HIF1? activation of PK2 could help in mediating PGC-1?’s protective effects. Further studies need to be conducted to confirm the mechanism of a putative HIF1?-PK2-PGC-1? axis.
EGR1 activation of PK2 transcription leading to neurotrophic factor upregulation represents another potential direction for future work. EGR1 is activated by MAPK/ERK signaling pathway to stimulate neurite outgrowth in a traumatic brain injury model (Chasseigneaux et al., 2011; Plummer, Van den Heuvel, Thornton, Corrigan, ; Cappai, 2016). EGR1, EGR2, EGR3, and EGR4 are co-upregulated with brain-derived neurotrophic factor (BDNF) after treatment with pridopidine in Huntington disease mouse models (Kusko et al., 2018). EGR1, EGR2, EGR4 expression are induced by erythropoietin (EPO), a neuroprotective cytokine in models of ischemic injury (Mengozzi et al., 2012), while PK2 is also co-upregulated during ischemic injury (M Y Cheng et al., 2012; Landucci et al., 2016). Further, EGR1 could induce angiogenesis via induction of PK1 in the peripheries (Sheng et al., 2018). In light of our current findings that suggest GDNF induction by PK2, further efforts could be put on confirming a putative EGR1-PK2-AKT-PI3K-GDNF signaling axis.
An alternative pathway could also be proposed. EGR1 could also activate high levels of GDNF by directly binding to the Gdnf promoter (Shin et al., 2009). It has also been shown that PK2 could reciprocally activate EGR1 expression via ERK1/2 activation. Together with our findings currently presented here, this represent a reciprocal, positive-feedback loop that induce expression of both PK2 and EGR1. Therefore PK2-ERK-EGR1-GDNF signaling could represent an alternative signaling axis that deserves future effort for elucidation and validation. The contribution from each putative signaling pathways would also need to be evaluated.
We have also found, in our present studies, that PK2 expression is induced early during neurodegenerative process induced by Mn overexposure, suggesting that PK2 could serve as a potential marker neurodegeneration. Elucidating the transcriptional mechanism underlying PK2 upregulation in dopaminergic neurons allow us to identify events upstream and players involved in PK2 upregulation, and begin to find other similarly co-regulated pathways. Understanding the perturbations in these complex pathways due to environmental factors or neurodegenerative events is crucial for targeting any pathway, such as the PK2 signaling pathway, to achieve neuroprotection.

Neuroprotective Effects of PK2 signaling
We had previously reported that PK2 is induced during neurodegeneration in multiple cell culture, animal model, and clinical cases of PD. These data revealed a perhaps new function of PK2 in the brain, especially in relevance to PD. Following logical next steps, we had also reported the effects of AAV-mediated PK2 gene delivery in the striatum, a concept similar to delivery of AAV-neurturin in the mouse striatum (Jeffrey H. Kordower et al., 2006; Marks et al., 2010). We reported protection against MitoPark and MPTP-induced neurodegeneration and reduced inflammatory microglial activation in mice treated with AAV-PK2 (Gordon et al., 2016). However, due to requirements of AAV-based gene delivery to be stereotaxically injected into the striatum, and the possibility of tumorigenesis as seen in the AAV-neurturin studies, which had found tumors in both mice injected with AAV vector as well as AAV-neurturin, long-term gene delivery via AAV is not currently optimal.
Therefore, in the studies presented here, we showed the potential for a small-molecule PKR1 receptor agonist to activate PKR1 signaling and induce GDNF expression in cultured astrocytes and organotypic slices. Treatment with IS20 demonstrated neuroprotective and neurorestorative effects, suggesting its clinical relevance in PD therapy. In C57B/6 mice, intranasally administered IS20 successfully crosses the blood brain barrier to reach maximum concentration in around 30 minutes. Using C57B/6 mice injected with MPTP as mouse model of PD, we had found IS20 to protect dopaminergic neurons against MPTP-induced neurodegeneration and increases dopamine levels in the striatum of MPTP-treated mice. Furthermore, we showed that, in transgenic MitoPark mice, 4-week intranasal IS20 administration induces GDNF and TH levels in the SN, reduces microglial activation and protects against dopaminergic neurodegeneration, and these protective effects translate to functional improvements in locomotor activity in MitoPark mice. These findings suggest that pharmacological modulation of GDNF signaling could be achieved with minimum side effects using intranasal administration of a PKR1 receptor agonist, IS20. Furthermore, no exogenous proteins are injected, nor does IS20 persist in the brain. For the above reasons, pharmacological modulation of endogenous GDNF levels that could be tightly controlled might resolve issues associated with protein diffusion, leakage, and transgene tumorigenicity.
Further considerations and experimentation are needed to improve and optimize the efficacy of intranasal administration of IS20. Optimization of formulations for carrier solutions, such as pH, choice of buffer solutions, and total volume could drastically improve delivery of IS20 into the nasal cavity and bioavailability in the CNS. Diffusion models that take into account the drug transport through the paracellular space across the nasal epithelium and perineural space, to reach the subarachnoid space of the brain could immensely aid in design of delivery methods as well as determination of optimal dosing (Cowie et al., 2017).

Induction of Neurotrophic Factors by PK2
Current undergoing work is examining the effect of IS20 on other family members of the GDNF family, neurturin, artemin, and persephin, in cell culture models. Even though artemin co-receptor GFR?3 and persephin co-receptor GFR?4 have limited expression in the adult brain, neurturin and co-receptor GFR?2 are constitutively expressed in the adult brain, and neurturin has shown to approximate the neuroprotective effects demonstrated by GDNF. Since neurturin is being evaluated in phase II clinical trials, potential induction of neurturin by IS20 represents a new opportunity for study.
Our preliminary studies suggest that PK2 could also induce GDNF in bone marrow-derived stem cells, imply that IS20, as a PKR1 ligand, IS20 could also induce GDNF in these cells. Given the therapeutic potential of stem cells in PD therapy for their capacity to secrete neurotrophic factors, PKR1-overexpressing stem cells that are implanted in PD patients could exhibit heightened production of neurotrophic factors upon IS20 treatment. Therefore, IS20-induced GDNF upregulation in other cell types aside from astrocytes, if confirmed, could suggest larger, farther-reaching effects of IS20 in the brain.

Further Considerations: Use of Appropriate Animal Models of PD

It is worthwhile to emphasize that proper models of disease are needed to reduce pipeline attrition from a drug development perspective. Administration of Parkinsonian toxicant in rodents and primates has been widely employed for elicitation of dopaminergic neuronal cell death and PD-like motor symptoms. In preclinical studies for PD, they have been used in evaluating therapies that mitigate motor symptoms. However, in the narrow effort to obtain motor deficits and subsequently mitigate these motor symptoms using candidate therapies, less attention has been paid to assessing the animal models’ ability to recapitulate the chronic and progressive nature of the disease, or the specific processes associated with dopaminergic neurodegeneration. For example, subacute MPTP treatment (3-5 daily injections) in mice have been in widespread use to generate loss of dopaminergic cells in SNpc, but it does not follow the progressive nature of the disease in humans, and due to rodents’ higher capacity for compensatory response against dopamine loss, do not reliably reproduce behavioral deficits (Rommelfanger et al., 2007; Schober, 2004; Tillerson, Caudle, Reverón, ; Miller, 2002). Importantly, MPTP treatments do not induce complete loss of striatal dopaminergic neuron projections, while clinical data from observation of PD patient post-mortem brains showed complete degeneration of dopaminergic projections in the caudate-putamen at 5 years post-diagnosis (42). Therefore, while putamenal gene delivery of GDNF can induce substantial GDNF expression overall in the brain, the anticipated beneficial effects from anteriograde transport of GDNF into the SNpc might play a less prominent role. This might confound the results for studies delivering GDNF only in the striatum of advanced PD patients, as employed by clinical trials conducted by Amgen (Lang et al., 2006; Nutt et al., 2003; Tatarewicz et al., 2007) and others (Slevin et al., 2005, 2007). Instead, other neurotrophic signaling molecules that could diffuse and elicit GDNF expression beyond the caudate putamen might extend the beneficial effects of GDNF into the SNpc.
Studies on the effects of PK2, specifically loss-of-function mutations, during neurodegeneration in wildtype adult mice are made difficult due to the developmental defects caused by a complete knockout of PK2. Therefore, a Cre-LoxP system which could conditionally and selectively ablate PK2 expression will be immensely useful in further studying the effects of PK2 compensatory responses, or a lack there of, during neurotoxic stress.

Other Future Directions: Role of PK2 in Neurogenesis
Preliminary data generated recently from our lab have demonstrated that PK2 participates in CNS neurogenesis in the subventricular zone (SVZ) lining the lateral ventricles. In the adult brain, the SVZ-derived neuroprogenitor cells migrate through the rostral migratory stream (RMS) in a chain of cells, eventually arriving at the olfactory bulbs, where they tangentially migrate from the inner layers of the olfactory bulbs to the outer layers. PK2 is upregulated by Ngn/MASH1 for olfactory bulb neurogenesis during development . The secreted PK2 most likely acts as a chemoattractant, as PKR2 can be detected on almost all migrating neuroprogenitor cells in the RMS . However, less is known about PK2’s role in adult neurogenesis, and lesser yet is known regarding its role in neurodegenerative processes such as in PD.
Environmental risk factors such as Mn exposure may lead to dysregulation of adult neurogenesis by reducing survival of neuroprogenitor cells as they migrate towards the olfactory bulb. The reduction of new olfactory bulb neurons affects not only the olfactory bulb functions, but also affect other brain regions such as the striatum, which is dependent on a continuous supply of neuroprogenitor cells to replenish lost neurons throughout adulthood and especially during neurodegeneration.
The role that PK2 plays in neurogenesis and determination of neuronal fate is incompletely understood. Through our in silico analysis, we have found the presence of a putative binding site for neuron-restrictive silencer factor (NRSF), that could suggest PK2’s involvement in key processes of commitment of neuronal cell fate or axon sprouting. NRSF represses expression of neuronal genes in nonneuronal cells and in NPCs by negatively regulating the activity of neuropilin-1, a positive regulator of axon branching promoted by the actions of VEGF and EGF, thus representing a possible negative regulator of PK2 during neurogenesis (Huang, Myers, ; Dingledine, 1999; Jones ; Meech, 1999) (Kurschat, Bielenberg, Rossignol-Tallandier, Stahl, ; Klagsbrun, 2006). These lines of evidence suggest that this balancing control aims to prevent the expression of PK2 and other co-regulated neurogenesis-associated genes such as neuropilin-1 until the precise moment that they are needed. In the same vein, vertebrate homologues of enhancer of split complex (HES) is another bHLH TF that was predicted to bind to the PK2 promoter. HES1 is critically important for sustainable neurogenesis (Dearden, 2015); decreasing expression of HES1 promotes the expression of proneural genes, while reducing the pool of neural stem cells. Interestingly, loss of HES1 function leads to increased MASH1-positive NPCs in the olfactory placode, and results in excess neurogenesis (Cau, Gradwohl, Casarosa, Kageyama, ; Guillemot, 2000). Thus, by antagonizing the activity of MASH1, HES1 works to provide a delicate balance between neurogenesis of new neurons and maintenance of neural stem cell pool. Further, the expression of PK2 is differentially upregulated as NPC differentiate into functional neurons while they migrate within the SVZ, and the upstream events that result in the cell’s decision to precisely upregulate PK2 at the most opportune moment remains unclear. Factors such as NRSF and HES1 involved in suppressing neuronal genes could provide such fine control of PK2 expression. The positive regulation of neural genes such as PK2 during neurogenesis by MASH1, countered by possible negative regulation by HES1 to repressing proneural identity, represent an exciting area of research. The perturbation of this delicate balance caused by exposure to Parkinsonian toxicants such as Mn deserves more attention.
The decade-long neurodegenerative process is complex, involving toxicant exposure creating oxidative stress and mitochondrial dysfunction, exacerbated by self-propagating neuroinflammation. Little is yet known about epigenetic alternation, particularly DNA methylation, by environmental toxicants, such as Mn. Altered expression of PD-related genes such as alpha-synuclein, and PK2, by epigenetic mechanisms illustrate the process of environmental-gene interactions, and could account for differences in individual susceptibility and variability through the course of disease progression. This represent a new area of research that deserves more focus.