Emerging therapies in Parkinson disease — repurposed drugs and new approaches

Since the publication of An Essay on the Shaking Palsy by James Parkinson two centuries ago, our understanding of the genetics, pathogenesis and clinical heterogeneity of Parkinson disease (PD) has evolved substantially. PD involves more than degeneration of dopaminergic neurons of the substantia nigra pars compacta (SNc), and its manifestations are recognizable beyond the car- dinal motor features of tremor, rigidity, bradykinesia and postural instability.

A complex interplay between genes and the environ- ment shapes the development of PD¹, and the existence of multiple pathways involving genes, proteins, cellular organelles and neural networks could contribute to the heterogeneity of symptom manifestation. Individual pathogenic pathways have been targeted for sympto- matic improvement and disease modification with var- iable success. Although dopaminergic therapies remain the gold standard for symptomatic management of PD, several unmet needs remain regarding treatment of dopaminergic-resistant motor and nonmotor symp- toms and for interventions that modify the natural clinical course of the disease. Many pharmacological agents specifically designed for disease modification have failed to meet the primary end point in multiple clinical trials over the past several decades. Challenges in these studies included, but were not limited to, the clinical heterogeneity of the population, patient selec- tion (for example, inclusion of patients with concurrent other neurodegenerative or vascular disease that might influence the outcome of the drug studied), lack of an adequate preclinical model of sporadic PD, lack of a disease biomarker and failure to identify preclinical PD. Another major challenge has been identification of the ideal time frame to apply disease-modifying thera- pies². The heterogeneity of PD has led to its classifica- tion into clinical subtypes^3,4 that reflect biological and pathophysiological differences among individuals with PD⁴. These differences suggest that individualization and precision treatment might be necessary to achieve disease modification.

In this Review, we discuss the emerging non- dopaminergic therapies for PD with a focus on dis- ease modification and treatment of motor symptoms. We examine pharmacological approaches, including small-molecule inhibitors, calcium channel blockers, iron chelators, anti-inflammatories and immunother- apies. We provide a summary of the success of these approaches in animal models, which have led to human clinical trials. We also examine non-pharmacological approaches, including gene therapies, neurotrophic fac- tors, cell restoration therapies and electrical modulation of neural circuits with deep brain stimulation (DBS). Although we recognize the important unmet needs in the treatment of nonmotor symptoms for PD, they are beyond the scope of this Review and have been covered extensively elsewhere^5–7.

Disease-modifying pharmacotherapies

Agents for specific molecular pathways

A mutation in SNCA, the gene encoding α-synuclein, was the first to be described in hereditary PD⁸. Since this seminal description, many other genetic muta- tions have been attributed to PD, including mutations in LRRK2 (encoding leucine-rich repeat kinase 2) and PRKN (encoding parkin) the most common auto- somal dominant and recessive mutations, respectively. Advances in understanding the genetic pathways in hereditary forms of PD have driven the development of potential disease-modifying therapies, which have been designed to target genes and/or proteins specific to α-synuclein pathology or other PD-related pathways (such as LRRK2-related pathways). Table 1 summarizes common genetic forms of PD, mechanisms of pathology and the latest advances in targeted therapies as reflected by either active clinical trials or drugs in the pipeline. Dysfunction in a single pathway identified by genetic studies might affect other pathways and involve mul- tiple cellular organelles. Targeting multiple pathways simultaneously might therefore be necessary to achieve disease modification.

Table 1 |.

Mutations and pathways implicated in PD and targeted therapies under development

Gene (inheritance)	Disease entity	Mechanism of pathology in mutation	Targeted therapy (in the pipeline and in clinical trials)
SNCA (autosomal dominant)	PARK1	Encodes α-synuclein Toxic oligomers of α-synuclein cause organelle dysfunction and impair interorganelle contacts, autophagy and synaptic vesicle trafficking14–18 Potential prion-like spread of α-synuclein oligomers199	Immunotherapies Strategies to decrease production, inhibit misfolding and enhance clearance of α-synuclein
LRRK2 (autosomal dominant)	PARK8	Encodes leucine-rich repeat kinase 2 Increases kinase activity in neurons and promotes cell death55 Mitochondrial DNA dysfunction Increases inflammatory response and apoptosis	Ursocholanic acid200 Ursodeoxycholic acid LRRK2 small-molecule kinase inhibitors51
PINK1 (autosomal recessive)	PARK6	Encodes PTEN-induced kinase 1 Mitochondrial dysfunction and mitochondrial calcium mishandling201	Ursocholanic acid200 Ursodeoxycholic acid Rapamycin202 Mitochondrial rescue therapies Silencing by siRNAs (tested in cultured neurons)
PRKN (autosomal recessive)	PARK2	Encodes parkin Dysregulation in mitophagy (in concordance with PINK1)203,204 Work in Drosophila suggests that parkin and PINK1 might function in a common genetic pathway and play a role in mitochondrial quality control and mitophagy203 Increases expression of PARIS and consequently represses PGC1α205
PARK7 (autosomal recessive)	PARK7	Encodes protein deglycase DJ1 Increases sensitivity of neurons towards oxidative stress DJ1 might contribute to the endoplasmic reticulum–mitochondria tethering, and reduced levels cause mitochondrial fragmentation and mitochondrial calcium mishandling206	Small-molecule chaperones that specifically bind to DJ1 are in the pipeline Phenylbutyrate upregulates DJ1 gene expression 11-Dehydrosinulariolide (neuroprotective effect in vitro and in vivo in animal models)207 Flavonoid extract208
VPS35 (autosomal dominant)	PARK17	Encodes vacuolar protein sorting- associated protein 35 Respiratory defects in complex I and II Increases mitochondrial fragmentation through increased interaction with DLP1 (ref.111) Impairs WASH complex and resulting macroautophagy209 AMPA receptor trafficking defects210	siRNA Mitochondrial rescue therapies
GBA (autosomal recessive)	PARK18	Encodes glucocerebrosidase Inhibition of macroautophagy and chaperone-mediated autophagy Decreases proportion of α-synuclein tetramers and related multimers compared with monomers47,69,211	Modulation of glucosylceramidase activity with ambroxol (AiM-PD trial; NCT02941822), LTI-291 or AT3375 Glucosylceramide synthase inhibitors: GZ/SAR402671 (MOVES-PD trial; NCT02906020)

AMPA, α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid; DLP1, dynamin-1-like protein; PARIS; parkin-interacting substrate; PD, Parkinson disease; PGC1α, peroxisome proliferator-activated receptor-γ co-activator 1α; siRNA, short interfering RNA.

α-Synuclein.

Friedrich Heinrich Lewy first described intracytoplasmic inclusions in the brains of patients with PD in 1912. This pathological hallmark was later named a ‘Lewy body’ and is formed by abnormal fibril- lary aggregations of α-synuclein protein⁹. α-Synuclein is a natively unfolded, 140-amino-acid protein that has been postulated to modulate synaptic activity and intracellular trafficking¹⁰. The exact physiological role of the protein is yet to be fully determined. α-Synuclein transitions between different conformations, including native monomers, tetramers and potentially toxic oli- gomers and fibrils¹¹. These conformations hypothetically coexist in a dynamic equilibrium governed by factors that accelerate and inhibit fibrillation, such as genetic mutations^12,13. Toxic α-synuclein oligomers impair deg- radation pathways^14,15, affect mitochondrial function^16,17, influence endoplasmic reticulum (ER) trafficking¹⁷ and loosen the association between ER and mitochondria, which might disrupt calcium homeostasis between the organelles¹⁸. Comprehension of the structure, genetics, pathology and spread of α-synuclein has revolutionized targeted therapies for PD.

Reduction of α-synuclein burden can be achieved by two means: reducing its synthesis or increasing its clearance (Table 2). With regards to reduction of syn- thesis, silencing of SNCA using small hairpin RNA and antisense oligonucleotides has provided mixed results in preclinical studies and, as a result, these approaches have not yet been explored in clinical trials^12,19. The major concern is that these silencing approaches might disrupt the as-yet unknown physiological role of the protein. Moreover, the degree of knockdown needed for a ther- apeutic benefit remains unknown. One report showed that β2-adrenergic receptors regulate SNCA expression and therefore β2-adrenergic receptor agonists have potential for disease modification by reducing SNCA expression and by promoting the health of dopaminer- gic neurons²⁰. However, in a large epidemiological study, β2-adrenergic receptor agonists and antagonists were not consistently associated with increased or decreased risk of PD²¹. Further animal and human studies are needed to fully comprehend the effect of β2-adrenergic receptors on PD risk.

Table 2 |.

α-Synuclein-targeting therapies in clinical trials and potential future targets and methods

Stage	Inhibition of α-synuclein production	Inhibition of misfolding or aggregation of α-synuclein	Activation of cellular turnover of α-synuclein	Immunotherapies
In clinical trials	None	Misfolding: NPT200–11 (completed phase I) Aggregation: NPT088 (completed phase I in patients with AD and will be tested in PD on the basis of AD study results); glycerol phenylbutyrate (phase I); nilotinib (phase II); PBT434 (phase I)	In patients with PD carrying a GBA mutation: ambroxol (phase II); GZ/SAR40261 (phase II; MOVES-PD)	Passive immunization: RO7046015, also known as PRX002 (PASADENA; phase II); BIIB054 (SPARK; phase II); MEDI1341 (phase I); BAN0805 (expected to start phase I in 2019) Active immunization: affitope PD01 (completed phase I); affitope PD03 (completed phase I)
Future targets or methods	siRNAs ASO β2-AR agonists	Small-molecule inhibitors: squalamine; epigallocatechin gallate anle138b; CLR01; KYP-2047 (prolyl oligopeptidase inhibitor)	IU1 (small molecule)	LAG3 (anti-LAG3 monoclonal antibody BMS-986016 is in early clinical trial phases for patients with cancer and can be tested in PD if deemed safe)

β₂-AR, β₂-adrenergic receptor; AD, Alzheimer disease; ASO, antisense oligonucleotide; IU1, 1-[1-(4-fluorophenyl)-2,5-dimethylpyrrol- 3-yl]-2-pyrrolidin-1-ylethanone; LAG3, lymphocyte-activation gene 3; PD, Parkinson disease; siRNA, small interfering RNA.

Regarding clearance strategies, α-synuclein is degraded both by the ubiquitin–proteasome system and by the autophagy–lysosomal pathway^14,22. Increasing clearance of α-synuclein via enhancement of protea- somal activity has been studied using small molecules including IU1, a small-molecule inhibitor of USP14 (a proteasome-associated deubiquitinase) with potential for PD treatment²³. Successful neuroprotection has also been reported following an increase in autophagy in both in vitro and in vivo models of PD²⁴. Overexpression of transcription factors such as transcription factor EB (TFEB), lysosome-associated membrane receptor pro- tein 2a (LAMP2a) or beclin 1, regulators of autophagy pathways, might have neuroprotective effects^25–27. The mTOR pathway has been implicated in the pathogenesis of PD, and small FDA-approved molecules such as rapa- mycin (an mTOR inhibitor) have been shown to protect against neuronal death in animal models of PD²⁸ and could be used in future clinical trials. These molecules remain at the preclinical stage, and the safety of such compounds for long-term use in patients with PD needs to be thoroughly investigated before any translation into clinical trials.

Inhibition of the misfolding and/or aggregation of α-synuclein has been an appealing approach for disease modification. Several compounds are being tested in clinical trials and are summarized in Table 2. Tyrosine-protein kinase ABL is activated in the brains of patients with PD and in mouse models of PD gen- erated via administration of the neurotoxin MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine). ABL activation leads to inhibition of parkin through tyros- ine phosphorylation and results in neural degeneration from accumulation of toxic parkin substrates (such as parkin-interacting substrate (PARIS))²⁹. Nilotinib is an FDA-approved ABL inhibitor for the treatment of chronic myelogenous leukaemia. Evidence from an MPTP animal model of PD suggests that nilotinib protects dopaminergic neurons²⁹. An initial pilot study involving 12 patients with PD showed safety and toler- ability of nilotinib, with good CNS penetration and tar- get engagement³⁰. A multicentre phase II double-blind placebo randomized controlled trial (RCT) is under- way to evaluate the safety and tolerability of nilotinib in PD (NILO-PD). The study will enrol 135 patients and the primary outcome will be safety and tolera- bility of the drug. The secondary outcome will assess motor and cognitive effects of nilotinib on patients with PD³¹.

Screening of drug compound libraries has led to the suggestion that some small-molecule inhibitors of protein aggregation are promising drug candidates^32,33 (Table 2). For example, squalamine, a naturally occurring compound, has been shown to inhibit the aggregation and reduce the toxicity of α-synuclein oligomers in neu- rons and in animal models of PD³⁴. Another molecule, CLR01, prevents α-synuclein fibrillation by binding to exposed lysine residues. CLR01 was able to ameliorate motor deficits in a mouse model of PD³⁵. Encouraging results from preclinical studies with such inhibitors have demonstrated that a better understanding of α-synuclein aggregates could open up new avenues for therapeutic development.

The hypothesis that misfolded α-synuclein spreads by a prion-like mechanism emerged from observed similarities between α-synuclein and classic prion pro- tein structure¹⁹. α-Synuclein fibrils are known to induce pathology in animal models, resulting in synaptic dys- function and neuronal death^36–39. Further support of this hypothesis came after the discovery of Lewy body inclusions in grafted embryonic mesencephalic cells years after implantation in brains of patients with PD³⁷. Immunization against α-synuclein hypothetically could neutralize the extracellular protein and prevent trans- mission of toxic fibrils. Lymphocyte-activation gene 3 (LAG3) also has been identified as a receptor (and a potential therapeutic target) that could facilitate cell- to-cell transmission of toxic α-synuclein fibrils⁴⁰ (fig. 1; Table 2).

Fig. 1 |. Mechanisms of potential therapies for Parkinson disease.

Numerous potential therapies for Parkinson disease (PD) are at various stages of preclinical and clinical testing. These agents target many aspects of PD pathogenesis. Therapies in bold are at advanced stages of testing. Several basic pathophysiological pathways are depicted, including those that lead to Lewy bodies, and those involving inflammatory factors, mitochondrial dysfunction and oxidative stress. A synapse to the right illustrates the potential spread of pathological α-synuclein to nearby cells. Bars indicate either inhibition or decrease and arrows indicate activation or increase. Arrows are also used for a sequence of events in a pathway. The exclamation marks inside the mitochondrion represent a dysfunctional mitochondrion. AKT, protein kinase B; β2-AR, β2-adrenergic receptor; DJ1, protein deglycase DJ1; DRP1, dynamin-1-like protein; ER, endoplasmic reticulum; GBA, glucocerebrosidase; GDNF, glial cell-derived neurotrophic factor; LAG3, lymphocyte-activation gene 3; LRRK2, leucine- rich repeat kinase 2; Mdivi-1, mitochondrial division inhibitor 1; MIRO, mitochondrial Rho GTPase 1; MitoQ, mitoquinone; PINK1, PTEN-induced kinase 1; PPARγ, peroxisome proliferator-activated receptor-γ; siRNA, small interfering RNA.

An exciting new approach to target oligomeric α-synuclein is the use of antibodies. This approach has been attempted in Alzheimer disease but is new to PD. Several studies have reported neuroprotection after either passive (using an antibody-based approach) or active (using full-length protein or short peptides) immunization in preclinical models^41–44. Success from these studies has prompted several clinical trials. A phase I clinical trial using PRX002, a monoclonal antibody against α-synuclein, reported that the treatment was safe and tolerable and demonstrated antibody binding to peripheral α-synuclein and cerebrospinal fluid (CSF) penetration⁴⁵. A phase II clinical study is in progress (Table 2). The development of engineered intrabodies (antibodies adapted for intracellular localization) and conformational antibodies could also be ways to target oligomeric α-synuclein. Affitope PD01 and PD03 are active immunotherapies that use a synthetically pro- duced α-synuclein-mimicking peptide. Both agents were shown to be safe and tolerable in initial pilot studies⁴⁶. The SYMPATH consortium, a European collaborative effort, was launched to further enhance the clinical development of α-synuclein vaccines.

α-Synuclein remains the most compelling target for disease modification in PD, but several questions need to be answered before further progress is made. Despite successful disease modification via targeting of α-synuclein in an animal model of PD¹¹, α-synuclein tox- icity and the potential for disease modification remain to be proved in sporadic PD. Furthermore, the absence of α-synuclein on autopsy in a subset of patients clini- cally diagnosed with PD (such as individuals with PD related to parkin or LRRK2 (refs^1,47,48)) and the presence of α-synuclein in the brains of people with multiple sys- tem atrophy, Lewy body dementia or lysosomal storage disease⁴⁹, and in healthy elderly patients⁵⁰, have raised concerns about the specificity of targeting misfolded α-synuclein protein in PD. Patient selection, lack of biomarkers for target engagement and safety and toler- ability of the targeted therapies against α-synuclein have all been major challenges that will need to be overcome in order to achieve successful disease modification.

LRRK2.

LRRK2 is a member of the Ras-of-complex (ROC) family of proteins⁵¹. Genome-wide association studies (GWAS) and a meta-analysis of GWAS have strengthened the evidence for a link between increased LRRK2 kinase activity and PD^52–54. The LRRK2 Gly2019Ser mutation, which localizes to the kinase domain of the protein, is the most frequent mutational cause of late-onset PD. Mutations in the GTPase domain of LRRK2 increase kinase activity and can induce par- kinsonism. How LRRK2 mutations cause PD remains unknown, but the mutations seem to impair autophagy and lysosomal function and induce mitochondrial dysfunction^55,56. The interplay between LRRK2 and α-synuclein is an area of growing research. A functional interaction between LRRK2 and α-synuclein has been proposed and has opened the possibility that LRRK2- targeted therapies could be beneficial for patients with idiopathic PD⁵⁷. The mechanism of this interaction is not fully clear, and new evidence suggests that LRRK2 activ- ity has a minimal effect on the burden of α-synuclein pathology in neurons, indicating that LRRK2 leads to pathogenesis via a different mechanism than α-synuclein⁵⁸. Additional studies will be needed to deter- mine the clinical and therapeutic implications of any interaction between LRRK2 and α-synuclein. Patients with parkinsonism caused by the LRRK2 Gly2019Ser mutation are phenotypically similar to patients with idiopathic PD but have a higher propensity to develop tremor and dystonia and have a slower disease pro- gression than patients with idiopathic PD⁵⁹. LRRK2 interacts with several key proteins associated with PD, which makes it an attractive therapeutic target⁶⁰.

Preclinical development and testing of small-molecule inhibitors has shown promise in animal models includ- ing non-human primates. DNL201, a small-molecule LRRK2 inhibitor, achieved 90% inhibition of LRRK2 kinase activity at peak drug blood levels in a phase I study of healthy volunteers. Testing of DNL151, another LRRK2 inhibitor, is in progress in a phase I trial. LRRK2 is highly expressed in kidneys, lungs and the immune system, and several animal studies that have manipulated LRRK2 at the gene or protein level have reported adverse effects, particularly in these systems. LRRK2-knockout mice show structural changes in the renal tubules of the cortex and medulla, but their renal function remains normal. These structural changes did not seem to occur in the context of pharmacological inhibition of LRRK2 (refs^61,62). In non-human primates, LRRK2 kinase inhi- bition induced morphological changes in the lungs, consisting of an increased size and number of lamellar bodies in type II pneumocytes, raising a potential safety issue in humans⁶³. The Michael J. Fox Foundation has initiated an LRRK2 safety initiative to address these concerns, and LRRK2 kinase inhibitors are proceeding towards clinical trials.

The main challenges facing the development of LRRK2-targeted therapies have been the absence of preclinical models to accurately reflect LRRK2-induced PD and the absence of reliable biomarkers for disease progression. Resolution of these issues will be impor- tant to address safety concerns and to confirm target engagement⁶⁴. New research has suggested a poten- tial role of wild-type LRRK2 kinase activity in idio- pathic PD pathogenesis⁶⁵. The hope is that therapies targeted against LRRK2 might be useful in treating idiopathic PD.

Glucocerebrosidase.

Although homozygous mutation of GBA (encoding glucocerebrosidase) results in Gaucher disease, people with heterozygous GBA mutations have an increased risk of parkinsonism or dementia⁶⁶, and a large multicentre analysis confirmed that the GBA muta- tion is a risk factor for PD⁶⁷. Low GBA enzymatic activ- ity has also been implicated in disease acceleration and worsened prognosis in PD⁶⁸. GBA-linked parkinsonism is associated strongly with impairment of the α-synuclein lysosomal degradation process. Conversely, increased levels of toxic soluble oligomers are associated with a depletion of lysosomal GBA. These findings are thought to indicate a neurotoxic cycle in patients with GBA muta- tion^69,70. Therapeutically increasing GBA activity might therefore enhance α-synuclein degradation in neurons⁷⁰. Small-molecule chaperones for GBA are being developed for the treatment of Gaucher disease under the premise that the mutant protein is trapped in the ER and that trafficking of the protein to the lysosome will decrease ER-associated degradation and increase lysosomal func- tion^71,72. Treatment with the GBA chaperone ambroxol in cell culture and subsequently in patients with GBA mutation demonstrated that increasing GBA activity pre- sents a feasible therapeutic approach⁷³. A clinical trial of ambroxol in PD (AiM-PD) is currently in progress (Table 1). Another chaperone, LTI-291, is also currently undergoing testing for GBA-associated PD. A next- generation small-molecule GBA chaperone, AT3375, is also under development for both Gaucher disease and PD.

Glucosylceramide synthase inhibitors reduce the levels of glucosylceramide and glucosylsphingosine and have been shown to reverse cognitive impairment and decrease α-synuclein aggregation in a GBA mouse model of PD⁷⁴. A phase II clinical trial (MOVES-PD) is underway to test a glucosylceramide synthase inhibitor, venglustat (also known as GZ/SAR402671), in individ- uals with early PD with GBA mutation. This study is anticipated to enrol 243 participants and is estimated to conclude in 2022.

Despite the advances made in linking GBA to PD, not all patients with GBA mutations will develop parkin- sonism. Hence, improvement of our understanding of the pathological mechanisms underlying GBA mutation is important before patient selection for future clinical trials. In addition, the level of GBA activation needed to achieve disease modification remains unclear, and no specific biomarker is currently available to confirm target engagement with small-molecule chaperones or glucosylceramide synthase inhibitors. Furthermore, the generalizability of similar therapies to idiopathic PD remains unclear.

Neuronal rescue

Neuronal vulnerability and calcium as a therapeutic target.

The concept of differing neuronal vulnerabil- ity within different cell populations emerged from the observation of selective dopaminergic neuronal loss in the SNc relative to other areas of the parkinsonian brain⁷⁵. Little is known about why these neurons are particularly sensitive, but they possess several properties that probably render them more susceptible to age, spe- cific mutations and/or environmental toxins than other neuronal populations⁷⁶. For example, the long and highly branched axons of dopaminergic neurons, which contain numerous transmitter release sites, increase mitochon- drial oxidative stress⁷⁷. In vivo, these neurons have slow tonic activity⁷⁸. Slow Ca²⁺ oscillations help to maintain slow tonic spiking by generating a membrane potential oscillation that facilitates Ca²⁺ entry into mitochondria, which stimulates oxidative phosphorylation and ATP production^79,80. This process results in high levels of cyto- solic and mitochondrial Ca²⁺ (fig. 1). Dihydropyridine, an agent that inhibits voltage-dependent Ca²⁺ channels and reduces cytosolic Ca²⁺ levels^79–81, was shown to reduce the risk of PD in some epidemiological studies⁸². Data from preclinical and clinical studies support a role for calcium channels in PD and have resulted in a 5-year, phase III, disease-modification clinical trial in patients with early- stage PD — STEADY-PD III. This trial uses isradipine, a calcium channel blocker approved by the FDA for the treatment of hypertension. An estimated 336 participants will be enrolled, and the trial completion is anticipated in 2019. The primary outcome measures will include Movement Disorder Society–Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) total scores at baseline and 36 months, and secondary measures will include MDS- UPDRS motor scores in the ‘off ‘ state, cognitive functions and quality of life. The inclusion criteria for this trial focus on early PD (Hoehn and Yahr ≤ 2 and participants off dopaminergic therapy at enrolment).

Iron-targeting agents.

Iron deposition in the neurons of the substantia nigra (SN) of patients with PD has been demonstrated via electron probe X-ray micro- analysis and is thought to be a disease-causing process⁸³. Iron can elicit oxidative damage, which results in the generation of reactive oxygen species⁸⁴. Removal of iron from the SN might, therefore, slow disease pro- gression. Promising results from preclinical studies suggest that iron chelators can cross the blood–brain barrier and remove excess iron⁸⁵ (fig. 1). These results have prompted the initiation of a double-blinded, ran- domized, placebo-controlled, pilot clinical trial of the iron chelator deferiprone in PD⁸⁶. This study evaluated drug safety, brain iron content changes (by MRI) and PD clinical status (by UPDRS scores) and demonstrated that short-term deferiprone therapy for participants with PD was safe and could decrease iron in specific brain regions.

FAIRPARKII, a randomized phase II study for conserva- tive iron chelation therapy in PD, is currently recruiting. Many unresolved issues remain with the iron chelation approach, including drug dose, lack of target engage- ment biomarkers and the stage of disease chosen for patient enrolment. Whether clearing iron will result in any clinical or disease-modifying benefit also remains to be determined.

Neuroinflammation-targeting agents.

Many observa- tions from epidemiological^87,88, post-mortem^89,90, ani- mal⁹¹, serum and CSF studies^92,93 support a therapeutic strategy of immune system alteration in PD⁹⁴. Substantial data also suggest that neuroinflammation plays a part in cell death in PD⁹¹. A growing body of evidence points to the neuroprotective effect of NSAIDs in PD. A study of 293 participants with PD and 286 matched controls found that NSAID users who took more than two pills per week for at least 1 month had a reduced risk of PD compared with controls⁹⁵. These results strengthened the idea of a beneficial role of NSAIDs for PD. To date, however, meta-analyses have failed to confirm the link between NSAIDs and reduced risk of PD^88,96, but multi- ple studies have shown an association between ibuprofen and a decreased risk of PD⁹⁷.

Lewy bodies that are released into the extraneuronal environment have been shown to activate macrophages and microglia, which subsequently alters effector T cell populations^94,98. Data from animal models of PD demon- strate that regulatory T (Treg) cells can help with immune tolerance and can have disease-modifying effects. Sargramostim, an FDA-approved human recombinant granulocyte–macrophage colony-stimulating factor (GM-CSF) for use in patients with cancer⁹⁹, is being tested as an anti-neuroinflammatory therapy in PD. A clinical trial using sargramostim demonstrated that it was well tolerated and that the adverse events asso- ciated with this therapy were not different in patients with PD than in those observed in other patients¹⁰⁰. Sargramostim improved Treg cell function and UPDRS motor scores did not worsen over time. Results from this study came from a small population and will need to be confirmed in future larger clinical trials (fig. 1).

Several groups have demonstrated the expression of the enzyme myeloperoxidase (MPO; a known player in inflammation and degeneration) by microglial cells in brains of individuals with PD^101–103. AZD3241 is a selective and irreversible MPO inhibitor that was able to suppress microglia and protect dopaminergic neu- rons in preclinical studies and was shown to be safe and well tolerated in phase I/II trials^104,105 Administration of AZD3241 for 8 weeks resulted in a statistically signifi- cant reduction in binding of a radioligand to the trans- locator protein (TSPO; a marker of activated microglia), which indicates that the drug was able to modulate the oxidative cellular environment in brains of individu- als with PD¹⁰⁵. This therapy warrants further trials to validate clinical efficacy.

Data from preclinical studies have shown that exenatide, an approved glucagon- like peptide 1 (GLP1) agonist, can normalize dopaminergic func- tion. The agent is thought to be capable of crossing the blood–brain barrier and might provide neuroprotective effects via the GLP1 receptor¹⁰⁶. Exenatide might exert its positive effects by decreasing neuroinflammation through the mitogen-associated protein kinase (MAPK) and phosphoinositide 3-kinase (PI3K)–protein kinase B (AKT) pathways¹⁰⁷ and might have beneficial effects on mitochondrial function (fig. 1). In an RCT¹⁰⁸ in which patients with PD were administered exenatide for 48 weeks, participants showed significant improvement in UPDRS III scores (off medication), compared with the placebo group, but no significant differences in the UPDRS IV fluctuation scores (on medication) or other metrics such as cognition, mood, dyskinesia and quality of life. The difference in motor UPDRS score at 60 weeks between the treatment and placebo arms did not change from the difference observed at 12 weeks, which suggests a symptomatic effect rather than disease modification. Two other GLP1 agonists, liraglutide and lixisenatide, are currently being assessed for disease-modifying potential in the LixiPark RCT.

Mitochondria-targeting agents.

Mitochondria are dynamic organelles that have an important role in energy metabolism and redox homeostasis¹⁰⁹ and undergo fis- sion and fusion in response to changes in metabolism and environmental stress¹¹⁰. Excessive mitochondrial fragmentation has been implicated in the pathology of sporadic and familial PD, and many genes associated with familial PD have been shown to affect key com- ponents of the mitochondrial respiratory chain and to contribute to neuronal death¹¹¹ (Table 1).

Therapies that target mitochondria can be specific to their fission and/or fusion or target oxidative stress pathways (Table 3). Peroxisome proliferator-activated receptor-γ (PPARγ) is a ligand-activated transcription factor that is expressed on neurons, astrocytes and microglia. This protein belongs to the nuclear hormone receptor superfamily and has a role in redox balance, mitochondrial function and immune regulation. PPARγ agonists such as pioglitazone inhibit neuroinflammation and decrease the levels of COX2 (also known as PTGS2), suppress inducible nitric oxide synthase and cytokines such as tumour necrosis factor, reduce apoptosis and inhibit proteasomal dysfunction¹¹² (fig. 1). In animal models, pioglitazone has been shown to increase mito- chondrial biogenesis, leading to a reduction in neuro- degeneration¹¹³. US National Institute of Neurological Disorders and Stroke (NINDS) Exploratory Trials in Parkinson’s Disease (NET-PD) FS-ZONE investiga- tors performed a phase II, multicentre, double-blinded RCT using pioglitazone in early PD¹¹⁴. The results revealed that pioglitazone was not effective in slowing progression of PD.

Table 3 |.

Neuronal mitochondria rescue strategies

Agents	Initial study results	Current clinical trials and future directions
Therapies targeting the mitochondrial oxidative stress pathway
Coenzyme Q10 (CoQ10) and GPI 1485	CoQ10 is an antioxidant that supports mitochondrial function. CoQ10 was found to be safe and well tolerated in phase I trials and had possible clinical benefit in a phase II trial (QE2 study)212. The QE3 study, which used a higher dosage of CoQ10, was a randomized, double-blinded phase III study that failed to reveal disease-modification potential for CoQ10 (ref.213) GPI 1485 is a neurotrophic immunophilin-ligand that was shown to reverse neuronal degeneration and prevents cell death in animal models. In a 6-month phase II clinical trial of GPI 1485, there was a trend towards favourable impact on imaging but not clinical measures in patients with PD214 The NINDS NET-PD trial, which enrolled and randomly assigned 213 participants, 71 to the CoQ10, 71 to the GPI 1485 and 71 to the placebo arm, was not conclusive regarding the disease-modifying effects of the tested compounds and recommended that further studies should be done to confirm efficacy215	CoQ10 trials failed to reveal disease-modification potential for PD; therefore, no further clinical trials are planned Multiple antioxidants failed in disease-modification trials in the past owing to a multitude of factors, including patient selection, bias from additional therapies, lack of a reliable biomarker and poorly defined patient selection and outcomes
Glutathione	Reduced glutathione was found early in the course of PD, and glutathione augmentation has been proposed as a therapeutic strategy in PD In a phase IIb study, intranasal glutathione was not superior to placebo216	Glutathione/NAC or EPI-589: safety and biomarker study in PD (NCT02462603)
N-Acetylcysteine (NAC)	NAC may restore the level of glutathione in neurons by functioning as a source of cysteine; hence, NAC has the potential to lower oxidative stress. Some data suggest a positive impact of NAC on dopaminergic neurons217	NAC for Neuroprotection in PD (NAC for PD) (NCT01470027)
Inosine	Inosine increases levels of urate, a potential antioxidant, and was proved safe and tolerable in patients with early PD (SURE-PD trial)118	SURE-PD III is in progress and plans to enrol 270 patients with early untreated PD
Statins	Different studies reported increased and decreased risk of PD with use of statins218,219: in Taiwanese patients with diabetes mellitus, statin use was associated with reduced risk of PD whereas a retrospective case–control analysis conducted in the USA suggested increased odds	Whether statins increase, decrease or have no effect on the risk of PD remains a controversial issue; larger multicentre trials are needed to confirm the findings
Glitazones	The use of glitazones has been associated with a decreased risk of incident PD in populations with diabetes The NET-PD FS zone investigators found that pioglitazone did not modify progression of PD114	Although pioglitazone was deemed futile, other glitazones are yet to be evaluated for their disease-modification potential in PD Larger trials with careful patient selection and better-defined primary outcomes are needed
Minocycline and creatine	Both minocycline and creatine have been shown to protect against MPTP-induced dopamine depletion in mice The NINDS NET-PD investigators found that neither agent was futile in phase II trials, but larger efficacy trials are needed220	Combination therapies using multiple antioxidants with better-defined inclusion criteria and outcome measured are being planned
Mitochondria-specific therapies
Mitochondria-targeted antioxidant therapy	Examples include MitoQ, MitoVitE, MitoApocynin and MitoTEMPOL221 Evidence of neuroprotective effects from animal models These conjugated compounds are postulated to bind to the mitochondria more easily than the parent antioxidant221 MitoQ did not slow progression of PD in a double-blinded, placebo-controlled trial222	Conjugated compounds other than MitoQ have yet to be tested for disease-modification potential in clinical trials Failures of parent and conjugated antioxidant compounds to slow disease progression in PD raise questions about the role of the oxidative stress pathway in PD
Dynamin-1-like protein (DRP1)-targeting compounds	Mdivi-1 (mitochondrial division inhibitor) is a DRP1 inhibitor that has been shown to block the excessive mitochondrial fission implicated in PD pathogenesis223	No clinical trials of DRP-targeting compounds and MIRO are planned Safety issues remain a concern and extensive testing in preclinical model is recommended prior to clinical application
Mitochondrial Rho GTPase 1 (MIRO)	MIRO is implicated in mitochondrial motility and abnormally accumulates on the mitochondrial surface, leading to delayed mitophagy Partial reduction of MIRO rescues mitophagy224

MitoVitE, mitotocopherol; MitoQ, mitoquinone; MPTP, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine; NINDS, US National Institute of Neurological Disorders and Stroke; PD, Parkinson disease.

In large prospective epidemiological studies, high plasma levels of urate were associated with a low risk of PD^115,116. Additionally, in a rat model of PD, urate pre- vented dopaminergic neuronal loss in the SN¹¹⁷. These findings facilitated the initiation of a safety and tolera- bility study for inosine, a urate precursor and potential antioxidant, in patients with PD. Inosine was shown to be effective in raising serum and CSF levels of urate¹¹⁸. SURE-PD3, an ongoing phase III RCT, was designed to investigate the role of inosine in early PD. The pri- mary outcome is change in total UPDRS scores over 24 months. Oxidative stress pathways remain an appeal- ing target despite the failure of multiple previous clin- ical trials. Table 3 summarizes major trials targeting oxidative stress and mitochondrial pathways, treatment challenges and future directions.

Symptomatic non-dopaminergic agents

Neurotransmitters such as acetylcholine, serotonin (5-hydroxytryptamine (5-HT)), noradrenaline, GABA, adenosine and glutamate play an essential part in PD symptomatology and might be involved in dopamine- resistant motor symptoms. Table 4 summarizes the dif- ferent neurotransmitter systems and their involvement in tremors, disorders of gait (particularly freezing of gait and falls), motor fluctuations and dyskinesias. figure 2 illustrates the different networks and neurotransmitters involved in motor symptoms of PD and the therapies targeted to these systems.

Table 4 |.

Major non-dopaminergic neurotransmitter networks in current and future PD motor treatment

Neurotransmitter	Target and/or receptor	Medications	Effect on motor symptoms	Clinical trials updates	Future directions and challenges
Acetylcholine	Cholinesterase inhibitors: affect balance and gait circuitry in mesencephalic locomotor region (that is, cholinergic pedunculopontine nucleus) Anticholinergics act on muscarinic receptors	Cholinesterase inhibitors: donepezil and rivastigmine Nicotinic acetylcholine receptor-α7 agonist: AQW051 Anticholinergics such as trihexyphenidyl	Cholinesterase inhibitors might improve gait and reduce falls in patients with PD Nicotinic acetylcholine receptor-α7 agonist was hypothesized to improve duration of action of levodopa and decrease LID on the basis of MPTP-lesioned monkeys225 Anticholinergics have a limited role in controlling parkinsonian tremor	Effect of donepezil on gait and balance in PD (NCT02206620): ongoing AQW051 did not improve dyskinesias or parkinsonian severity in a phase II clinical trial (NCT01474421)	Larger double- blinded placebo- controlled trials are needed to evaluate the role of cholinesterase inhibitors on gait and balance in patients with PD Despite the potential for anticholinergics to control parkinsonian tremor, they are poorly tolerated owing to their cognitive and systemic adverse effects
Adenosine	Adenosine (A2A) receptor antagonist	Istradefylline Preladenant Tozadenant	Reduction of ‘off’ time when added to levodopa Possible role for treating postural abnormalities	Istradefylline proved efficient in reducing motor fluctuations in randomized trials Preladenant was deemed futile in three phase III trials (NCT01155479) Tozadenant trial terminated after reports of agranulocytosis (NCT03051607)	Further studies are needed on istradefylline as an add-on therapy to levodopa to reduce ‘off’ time and test its potential effect on postural abnormalities Istradefylline is approved in Japan but is not approved by the FDA A2A receptor antagonist role is becoming more controversial in PD
Serotonin	Variable effects, mostly on 5-HT1A and 5-HT2A	5-HT1A agonists: buspirone, eltoprazine, sarizotan and NLX-112 5-HT2A antagonists and inverse agonists Clozapine	Improve dyskinesias (5-HT1A agonists) Helps rest tremor (clozapine)	Buspirone for LID: seems to be more effective in milder cases of LID Eltoprazine reduced peak-dose dyskinesia in a phase IIa study Sarizotan: improved dyskinesias but worsened motor symptoms in a phase III study NLX-112: no clinical trial results available yet, but effective at reducing LID in parkinsonian rats	Promising approach, as many agents are already approved by the FDA and can be tested in clinical trials for treatment of motor symptoms, especially LID Buspirone for dyskinesia: BUS- PD and BUSPARK (NCT02589340, NCT02617017) are ongoing and will help determine the efficacy of buspirone for LID Eltoprazine phase IIb study for LID is ongoing (NCT02439125) NLX-112 study results are pending More organized and larger-scale trials will be needed Agranulocytosis is the main risk with clozapine use
Glutamate	NMDA receptor antagonist mGluR5 antagonists Positive allosteric modulator of mGluR4 Glutamate release inhibitor	Non-selective: amantadine and amantadine extended-release formulation, dextromethorphan, memantine, safinamide (glutamate release inhibitor) Selective to mGluR5: mavoglurant (AFQ056) and dipraglurant (ADX48621) Selective to mGluR4: foliglurax (PXT002331)	Improve dyskinesias Improve duration of action of levodopa (e.g. safinamide)	Amantadine extended-release formulations: GOCOVRI approved after results showed dose-dependent decrease in dyskinesia (2017); OSMOLEX extended-release formulation approved for parkinsonism (2018) Safinamide (XADAGO) approved in 2017 as adjunctive to levodopa to increase ‘on’ time Mavoglurant (AFQ056): ceased (failed in phase II trial and nonmotor adverse effects were reported) Dextromethorphan/ quinidine helps minimally with LID (NCT01767129) Dipraglurant (ADX48621) appears to be safe and well tolerated in phase IIa trial226, with some evidence that it reduces peak-dose dyskinesia	Amantadine has been associated with self-reported improvement in FOG, but larger randomized trials are needed to confirm efficacy Non-selective glutamatergic antagonists (e.g. amantadine) carry potential systemic and cognitive adverse effects More trials are needed to evaluate effects of dipraglurant (ADX48621) on PD dyskinesias Results of the foliglurax study on LID and end-of-dose wearing-off are still pending
Noradrenaline	Locus coeruleus brainstem noradrenergic circuits might affect gait and balance Improvement of attention might have an indirect effect on preventing falls	Methylphenidate Atomoxetine	Possible effect on FOG although evidence not strong	Methylphenidate and atomoxetine to enhance gait and balance in PD (NCT02879136) TAME-PD	Larger clinical trials are needed with more specific analysis of outcome measures

5-HT, 5-hydroxytryptamine (serotonin); FOG, freezing of gait; LID, levodopa-induced dyskinesia; mGluR, metabotropic glutamate receptor; MPTP, 1-methyl-4- phenyl-1,2,3,6-tetrahydropyridine; NMDA, N-methyl-d-aspartate; PD, Parkinson disease.

Fig. 2 |. Basal ganglia neurotransmitter network.

A classic model of the basal ganglia is shown with a striatonigral direct pathway through the globus pallidus internus (GPi), an indirect pathway reaching the globus pallidus externus (GPe) and subthalamic nucleus (STN) before the GPi and a hyperdirect pathway to the STN. Dopaminergic input to the striatum is supplied by the substantia nigra (SN). Other neurotransmitter foci are also depicted, including the pedunculopontine nucleus (PPN) and nucleus basalis of Meynert (NbM) for acetylcholine (ACh), the raphe nucleus (RN) for serotonin (5-hydroxytryptamine (5-HT)) and the locus coeruleus (LC) for noradrenaline (NE). These neurotransmitters travel throughout the basal ganglia and related structures and play an important part in Parkinson disease motor symptoms. Several existing and investigational therapies that aim to address abnormal levels of neurotransmitters are shown.

DA, dopamine; mGluR, metabotropic glutamate receptor; NMDA, N-methyl-d-aspartate.

Tremors.

The classic tremor in PD occurs at rest and can re-emerge at posture. Tremor is considered a clin- ically unique feature of PD as it seems to correlate less with dopamine deficit than do bradykinesia and rigidity. The origin and the circuitry of tremor in the brain represent an area of active research. Basal ganglia dys- function is thought to trigger the tremor, whereas the cerebellothalamocortical circuit might act to control the tremor amplitude¹¹⁹. Tremor in patients with PD may or may not respond to dopamine treatment, and in some patients tremor can worsen after treatment with levodopa. These findings suggest that other neuro- transmitters might be involved, including acetylcholine or 5-HT¹²⁰. Anticholinergics such as trihexyphenidyl are useful in a subset of patients, but neuropsychiatric and cognitive adverse effects have been common reasons for withdrawal¹²¹. Clozapine, a potent antipsychotic used to treat PD hallucinations, can also be effective in con- trolling PD tremors^122,123. Clozapine is a 5-HT1A receptor agonist and a 5-HT_2A/2C receptor antagonist in addition to having antidopaminergic, anticholinergic and anti- histaminergic effects¹²⁴. The small but serious risk of agranulocytosis remains the major drawback of using clozapine for treatment of dopamine-resistant tremors. β-Blockers have also been used with variable success in controlling parkinsonian tremor^125,126.

Gait.

Freezing of gait can be either levodopa sensitive or levodopa resistant, which suggests the involvement of non-dopaminergic networks and neurotransmitters in PD-related gait disorders. Physiology of freezing of gait involves a dorsal (cognitive) and a ventral (emo- tional) locomotor network in addition to multiple non-dopaminergic neurotransmitter systems. These sys- tems include the cholinergic, glutamatergic, noradrener- gic and GABAergic systems, among others¹²⁷. To explore the involvement of the cholinergic system in gait dys- function in PD, investigators from the UK initiated the ReSPonD trial, a phase II RCT that tested the efficacy of the cholinesterase inhibitor rivastigmine in patients with PD who had fallen. The study revealed that rivastigmine improved gait stability and decreased chances of falls in patients with PD¹²⁸. Another study has confirmed the role of the central cholinesterase inhibitor, donepezil, in reducing the frequency of falls in patients with PD¹²⁹. Larger controlled trials will be needed to confirm the efficacy of cholinesterase inhibitors. Serotonergic and noradrenergic networks are affected in PD owing to degeneration of the raphe nucleus and locus coeruleus, respectively, and are implicated in gait disorders. In a double-blind, multicentre, placebo-controlled trial, methylphenidate improved hypokinesia and freezing of gait in patients with advanced PD who received sub- thalamic nucleus (STN) stimulation¹³⁰. Methylphenidate and atomoxetine are currently being tested for effects on gait in the TAME-PD trial (fig. 2).

Motor fluctuations and dyskinesia.

Motor fluctua- tions and dyskinesia are troublesome adverse effects that result from a combination of PD progression and dopaminergic treatment. An estimated 40% of patients with PD who receive levodopa treatment develop motor fluctuations after 4–6 years of treatment¹³¹. Several options exist to treat these complications. The most common include adjustment of the dose, frequency of administration or type of the dopaminergic medica- tion administered. Progression of these symptoms and fluctuations can trigger consideration of other surgical options (including DBS or levodopa–carbidopa intesti- nal gel). Multiple non-dopaminergic compounds sum- marized in Table 4 have been, or are currently being, investigated for an effect on motor fluctuations.

Adenosine (A2A) receptor antagonists can improve motor functions in animal models of PD¹³². However, in two phase III RCTs, the A2A receptor antagonist prelade- nant, used as an adjunctive therapy, did not improve the motor function or reduce ‘off ‘ time in patients with PD compared with placebo or rasagiline^133,134. Preladenant has also failed as a monotherapy in patients with early PD in another phase III trial¹³⁵. Given the results of these phase III studies, Merck discontinued preladenant in 2013. In contrast to preladenant, the A2A receptor antag- onist istradefylline was tested in several RCTs^136–138, and a meta-analysis suggested that it can be useful to alleviate ‘off ‘ time and motor fluctuations, although the effect on dyskinesia was not clear¹³⁹. Development of tozadenant (another A_2A receptor antagonist) was halted in 2017 after reports of agranulocytosis.

Evidence suggests that 5-HT has a role in levodopa- induced dyskinesia (LID)¹⁴⁰. Findings from animal models indicate that serotonergic receptor agonists can reduce motor fluctuations and LID¹⁴¹. Multiple seroton- ergic receptor agonists or antagonists are being tested in clinical trials for their effects on motor fluctuations and dyskinesia (Table 4). Similarly, glutamate has been implicated in motor fluctuations in patients with PD, and N-methyl-d-aspartate (NMDA) receptor antagonists such as amantadine have been used with variable suc- cess for the treatment of dyskinesia¹⁴². However, systemic and nonmotor adverse effects (cognitive and psychiat- ric) limit their use. Therapies with increased selectivity against metabotropic glutamate receptors are being tested for their effects on motor complications¹⁴³ (Table 4).

Non-pharmacotherapeutic interventions

Gene therapy

The aim of gene therapy is to replace, silence or mod- ify mutated genes with the use of non-replicating viral vectors such as adeno-associated virus (AAV) or lenti- virus. Several studies in humans have shown that these viruses are safe and efficient for gene delivery^144–147. One of the advantages of this approach is the ability to deliver therapies to specific brain regions, thereby reducing off- target effects. Disadvantages include the difficulty in regulating the amount of therapy delivered and the fact that the therapy might be irreversible. Here, we discuss viral vector-mediated targeted delivery of genes encod- ing proteins involved in dopamine production (such as aromatic-l-amino acid decarboxylase (AADC)) or basal ganglia network modulation (such as glutamate decarboxylase (GAD)) for symptomatic motor therapy. We also discuss neurotrophic factors and gene therapy for potential disease modification. Table 5 summarizes clinical trials, major advances in the field, outcomes and challenges of these treatment strategies.

Table 5 |.

Enzymatic or neurotrophic factor therapies for PD using direct infusion or gene therapy

Therapy and trial(s)	Trial design	Outcome	Challenges and future directions
Symptomatic therapies
Human AADC gene therapy for PD: three trials145,147,227	Phase I open-label studies, n = 21 patients total	Favourable with evidence of radiological response and clinical improvement. Four-year clinical follow-up showed modest increase in UPDRS scores228	Small sample and absence of controls. Unclear what is the best vector dose and delivery method. Will benefit from a randomized trial and longer follow-up
Safety study of AADC gene therapy (VY-AADC01) (NCT01973543)	Phase I open-label study, aiming to recruit 15 participants	Initial report of the ongoing study confirmed safety and tolerability of VY-AADC01	Study is employing a high vector concentration and a different surgical technique using MRI guidance for precision in vector delivery. Estimated completion in December 2019
ProSavin, a lentiviral vector-based gene therapy (encoding AADC, TH and GCH1)229	Phase I/II open-label trial (n = 15)	Favourable with significant improvement of motor UPDRS III score (mean score decreased by 11 points) and medication reduction at 12-month follow-up	Small sample, increased on-medication dyskinesias and motor fluctuations. A second-generation gene therapy trial, OXB-102, is in progress
AAV-mediated expression of GAD for PD: two trials146,230	One phase I trial (n = 12) and one phase II trial (n = 37)	Favourable radiological and clinical outcomes with reduction in UPDRS motor scores at 6 months (23.1%) compared with sham surgery (12.7%). Benefit persisted at 12 months152	Safe and effective in phase II. No phase III trials planned yet
Disease-modifying therapies
GDNF infusion: two open-label pilot studies231,232 and two randomized double-blinded studies233	Phase I open-label study (n = 5), bilateral putamenal delivery Phase I open-label study (n = 10), unilateral putamenal delivery Phase I/II randomized double-blinded study (n = 34 with 1:1 randomization), bilateral putamenal delivery Phase I/II randomized double-blinded study (n = 41) followed by a 9-month open-label extension study	Open-label trials found favourable radiological (increased putamenal18 F-l-DOPA uptake) and clinical outcomes (39% decrease in UPDRS motor scores with bilateral putamenal delivery versus 30% decrease with unilateral putamenal delivery) Persistent clinical benefit without serious adverse events in a 2-year follow-up study234 RCT found favourable radiological benefit that did not translate into clinical benefit	Animal models revealed cerebellar injury with putamen liatermin infusion Three patients developed antiliatermin antibodies, which can cross-react with endogenous GDNF Novel method of administration of GDNF by the Bristol study group through convection-enhanced delivery rather than diffusion- dependent delivery. Despite the study not meeting the primary end point, the open-label extension study is still ongoing235
NRTN (vector delivery): four trials in total. AAV2-NRTN CERE-120 in bilateral putamen162,163 and dual targets (putamen and SN)236,237	In bilateral putamen: one open-label (n = 12) and one randomized study (n = 58 total, 38 patients were randomly assigned in treatment arm) Dual target putamen and SN: open-label phase I (n = 6 patients). RCT (n = 51 total, 24 patients were randomly assigned to the treatment arm)	In bilateral putamen: UPDRS motor scores improved at 12 months (reduction of 14 points) in phase I trial. No statistically significant difference was found in motor UPDRS scores at 12 months when both arms were compared in the randomized study Dual target putamen and SN: safe and feasible in phase I. In the RCT, no statistically significant difference was found when UPDRS motor scores were compared in the two arms	Putamen: in RCT, 5 patients developed tumours (n = 3 patients in treatment arm and n = 2 patients in sham-surgery arm) Failure to reveal clinical benefit in dual targeting (putamen and SNc) raised doubts about trophic factors, especially NRTN, and methods of factor delivery in the nigrostriatal system
GDNF (vector delivery): AAV2-GDNF for advanced PD (NCT01621581)	Phase I open-label study with aim to recruit 25 patients	NA	Estimated completion date January 2027. Status active but not recruiting now
CDNF infusion: safety of CDNF by brain infusion in patients with PD (NCT03295786)	Phase I/II randomized double-blinded study aiming to enrol 18 participants	NA	Estimated completion date in 2019

AADC, aromatic-l-amino acid decarboxylase; AAV, adeno-associated virus; CDNF, cerebral dopamine neurotrophic factor; GCH1, GTP cyclohydrolase 1; GDNF, glial cell-derived neurotrophic factor; l-DOPA, 3,4-dihydroxyphenylalanine; NA, not applicable; NRTN, neurturin; PD, Parkinson disease; RCT, randomized controlled trial; SN, substantia nigra; SNc, substantia nigra pars compacta; TH, tyrosine 3-monooxygenase; UPDRS, Unified Parkinson’s DiseaseRating Scale.

Dopamine restoration.

AADC is an enzyme that con- verts 3,4-dihydroxyphenylalanine (l-DOPA) to dopa- mine and thereby enables endogenous stimulation of dopamine production in surviving neurons in PD. AADC has been shown to enable decarboxylation of l-DOPA by an AAV delivery approach¹⁴⁸ and has resulted in behavioural recovery¹⁴⁹ in preclinical models. Studies that use vectors for single-enzyme AADC are summarized in Table 5. A safety study of VY-AADC01, an investigational agent administered via injection into the putamen of patients with PD, is anticipated to con- clude by the end of 2019. Initial reports have shown that VY-AADC01 is well tolerated and has no serious adverse events¹⁵⁰. The study is using a higher vector genome concentration than was used in previous AADC gene therapy trials and is employing real-time MRI to mon- itor vector delivery. Potential clinical and radiological responses will also be assessed.

ProSavin is a viral vector containing three genes encoding enzymes required for making dopamine: tyrosine 3-monooxygenase, AADC and GTP cyclohy- drolase 1. Results from the study showed an improve- ment in motor symptoms in patients with PD 12 months after treatment and prompted the launch of OXB-102, a second-generation gene therapy (an approach that uses a vector encoding the same enzymes but that enables greater dopamine production). Preliminary results from animal studies showed superior efficacy of OXB-102 compared with ProSavin and supported further clinical development. Success in the phase I study has led to the launch of the phase II Axo-Lenti-PD trial (also known as SUNRISE-PD trial). The trial has an open-label dose escalation phase of OXB-102 and a randomized double- blinded phase. The primary outcome remains the safety of the compound, but secondary outcome measures will focus on UPDRS scores ‘on’ and ‘off’ medica- tions and motor fluctuations at 6 months. Preliminary results from the first two enrolled patients are expected in March 2019.

Basal ganglia network modulation.

In patients with PD, nigrostriatal degeneration leads to an excessive inhibi- tory output from the globus pallidus internus (GPi) and substantia nigra reticulata (SNr) as a result of disinhibi- tion of the STN, which is thought to drive the GPi and SNr via release of glutamate. This pathway suggests that enhancement of GABA transmission from the STN and its terminal regions via GAD could be of therapeutic benefit. AAV-GAD administration to the STN resulted in improvements in bradykinesia, gross motor skills and tremor in non-human primate models of PD¹⁵¹. Following these promising animal studies, Neurologix announced a gene therapy trial, NLX-P101, that is based on the introduction of a gene encoding GAD into brain cells via AAV. Symptom improvement was reported in patients with PD on the basis of UPDRS scores 6 months after the administration of AAV-GAD (Table 5). The 12-month clinical effects of bilateral STN delivery of GAD revealed persistent improvement of the UPDRS motor scores in the treatment group compared with the sham group¹⁵². A significant reduction in the daily duration of LID was also observed in the treatment group.

Neurotrophic factors.

The mechanism by which neu- rotrophic factors elicit their neuroprotective effect is poorly understood. However, several studies have shown that neurotrophic factors can induce beneficial effects on dopaminergic neurons^153,154.

As a survival factor for dopaminergic neurons of the midbrain region¹⁵⁵, glial-derived neurotrophic factor (GDNF) is an attractive target for slowing the course of degeneration. The beneficial effects of GDNF have been demonstrated in animal models^156,157. A phase II double- blinded, placebo-controlled study of a continuous infu- sion of GDNF into the putamen of patients with PD was conducted by Amgen. Study results showed that the therapy was well tolerated. However, it did not meet the primary end point as no significant improvements in UPDRS scores were observed and the study was halted for long-term safety reasons. In 2013, a 5-year phase I trial of AAV2-GDNF was launched. The trial is currently active, but no results are available.

A close homologue of GDNF, neurturin (NRTN), is another potential therapy¹⁵⁸. Injections of recombinant NRTN have demonstrated variable results in animal models of PD^159–161. CERE-120 (also known as AAV2- neurturin; developed by Ceregene) was tested in patients with PD in the first trial of a gene therapy using a neuro- trophic factor. Phase I study results showed improve- ment in UPDRS scores in patients with PD¹⁶². A phase II study was subsequently launched, but no significant differences were found between treatment groups in the primary end point measured by the UPDRS scores¹⁶³ (fig. 1; Table 5).

Failure of GDNF and NRTN to show clinical benefits precipitated a search for alternative neurotrophic factors for PD. Cerebral dopamine neurotrophic factor (CDNF) was neuroprotective for midbrain dopaminergic neurons in a rat model of PD¹⁶⁴, and a phase I/II trial is currently evaluating the safety and efficacy of direct putamen infu- sion in patients with PD (Table 5). Other neurotrophic factors, such as brain-derived neurotrophic factor and vascular endothelial growth factor, are currently in the preclinical pipeline^165,166.

In summary, gene therapy is a promising avenue for treatment of PD, and successful advances have been made in dopamine restoration and sympto- matic therapies, although disease modification with trophic factors has not yet been shown to be feasible in humans. Challenges that limit these therapies include the potential need to deliver a combination of neuro- trophic factors, the determination of best targets and the improvement in delivery methods.

Cellular therapies

The field of cell transplantation did not start to gain momentum as a potential therapy for PD until after the development of the 6-hydroxydopamine-lesioned rat model of PD in the late 1970s¹⁶⁷. Assessment of the thera- peutic potential of cell transplantation in the lesioned rodent brain was made possible by the development of this model. Early experiments in the 1980s transplanted adrenal medullary cells (which can produce catechola- mines such as dopamine) into the striatum of patients with PD. However, these transplants were deemed futile and unsafe¹⁶⁷. Transplantation trials of mesencephalic dopaminergic neurons obtained from fetuses showed positive effects but had ethical and safety challenges. Alternative and potentially promising cell therapies have included dopaminergic cells derived from human pluripotent stem cells (hPSCs), which could be either human embryonic stem cells (hESCs) or induced pluri- potent stem cells (iPSCs). Despite all of the advances made in the field of cellular transplant in the past few decades, dopaminergic cell replacement therapy for PD is still considered a symptomatic therapy rather than a disease-modifying intervention¹⁶⁸.

Mesencephalic fetal cell transplantation.

Open-label studies showed that transplantation of mesencephalic fetal dopaminergic neurons into the striatum of patients with PD improved motor symptoms and reduced motor fluctuations, presumably by restoring striatal dopa- minergic transmission¹⁶⁹. Subsequent RCTs failed to demonstrate meaningful clinical improvement, and graft-induced ‘runaway’ dyskinesias (that is, post- transplantation dyskinesias that occur even after a prolonged absence of antiparkinsonian medication¹⁷⁰) were reported^171,172, thereby halting further trials. The mechanism of the runaway dyskinesias remains uncer- tain and is yet to be fully elucidated. Despite adverse effects, young patients¹⁷¹ with low disease burden¹⁷² and individuals free from pre-intervention dyskinesias had improved outcomes. This observation contributed to the launch the European consortium TRANSEURO to globally evaluate fetal mesencephalic cell transplant in PD. However, tissue availability and the ethical dilemma of using fetal tissue in research¹⁶⁷ are challenges that still need to be addressed (fig. 3).

Fig. 3 |. Extraction and induction of dopaminergic cells for neuronal restoration.

There are several emerging approaches for extracting dopaminergic cells. Fetal mesencephalic cell (FMC) transplantation initially seemed promising, but runaway dyskinesias halted further trials. The TRANSEURO consortium will comprehensively re-evaluate this therapy. Embryonic stem cells (ESCs) can be used for transplantation but carry the same ethical challenges as FMCs (the ethical challenge is visually depicted by the locks in the figure). Red blood cell (RBC) and fibroblast induced pluripotent stem cells (iPSCs) are an alternative approach that eliminates this ethical challenge, but, similar to other pluripotent stem cells, these cells carry neoplastic potential. Finally, neurons induced by direct programming from somatic cells might be a viable but technically challenging route. The figure depicts some transcription factors used in reprogramming cells. ASCL1, achaete– scute homologue 1; FOXA2, hepatocyte nuclear factor 3β; KLF4, Krüppel-like factor 4; LMX1A, LIM homeobox transcription factor 1α; MYC; Myc proto-oncogene protein; NURR1, nuclear receptor related-1 protein; OCT4, organic cation/carnitine transporter 4; SOX2, transcription factor SOX2; TH, tyrosine 3-monooxygenase.

Stem cell therapy.

In 2006, dopaminergic neurons derived from hESCs were implanted intrastriatally in an animal model of PD and resulted in behavioural improvements¹⁷³. Subsequently, several researchers have worked to refine the conversion process of hESCs to dopaminergic cells by improving specification¹⁷⁴. When grafted, these cells modulated neurotransmis- sion in a rodent model of PD, retained the ability to sur- vive long term and grew and functioned efficiently¹⁷⁵. Notwithstanding, the therapy has safety limitations that need to be addressed before human application (fig. 3).

Another approach to neuronal restoration is to use a cocktail of transcription factors to reprogramme skin or blood cells to become iPSCs or induced neurons¹⁷⁶. This technique would facilitate the generation of patient- specific transplantable cells, which could circumvent immune rejection as well as other ethical issues that surround hESCs. Successful iPSC transplantation in monkeys has led to plans for a clinical trial employing iPSCs in patients with PD that has already commenced in Kyoto, Japan¹⁷⁷.

Results from clinical trials have shown that grafted cells are capable of long-term survival and are able to integrate into appropriate functional brain networks. Advances in iPSC research have now made it feasible to move these applications towards the clinical environ- ment. The international GForce-PD initiative connected centres from the USA, Europe and Japan for the clinical application of hPSCs in PD. Many collaborative efforts are being planned for the next 2 years: the EUROPEAN STEM-PD, which comprises two European consortia (TRANSEURO, which will use fetal mesencephalic cells, and NeuroStemCellRepair, which will use hESCs); NYSTEM-PD (New York State Stem Cell Science), which will use hESCs; CiRA (Center for iPS Cell Research and Application, Kyoto University, Japan), which will use allogeneic iPSCs; and a Summit for Stem Cell trial, which will use autologous iPSCs¹⁷⁸. Outside of this initia- tive, a few open-label clinical trials are already recruiting, including a phase I/II trial to assess safety and efficacy of striatum transplantation of hESC-derived neural pre- cursor cells in patients with PD¹⁷⁹. An important consid- eration is that stem cell therapies remain experimental, and commercial stem cell tourism is a major challenge for the field as unapproved cell-based therapies are being administered peripherally or directly into the CNS and can lead to serious complications¹⁶⁸. Additionally, the stem cell transplantation field will probably be limited by many of the same challenges that previous transplan- tation studies have encountered. Perhaps the biggest challenge remains the failure to address degeneration in multiple motor and nonmotor pathways.

Deep brain stimulation

A new era for neural network modulation began in 1987, when Alim Benabid implanted the first DBS electrode for the treatment of a patient with tremor¹⁸⁰. Wide appli- cation of DBS to PD first began in the early 1990s, and since then the procedure has improved the treatment of PD. Multiple RCTs have demonstrated the benefits of DBS for improving motor symptoms, fluctuations (that is, ‘off ‘ time) and quality of life compared with medical therapy alone^181–184. The two main targets in PD, the GPi and STN, have provided similar motor benefits¹⁸⁵, with the choice of target often dictated by patient profile and goals of treatment. Despite the substantial advances in the field of neuromodulation, dopaminergic-resistant symptoms such as axial symp- toms (affecting gait, balance, posture and speech) poorly respond to current DBS technologies¹⁸⁶. Novel approaches include stimulation of different networks, use of alternative hardware (directional DBS leads), adaptive DBS and use of advanced imaging for more individualized targeting¹⁸⁷. Non-invasive DBS and opto- genetically inspired DBS are promising therapies on the horizon.

New targets.

The pedunculopontine nucleus (PPN) has emerged as a target for DBS after animal studies showed that stimulation yielded improvements in akinesia and gait¹⁸⁸. Initial positive reports on gait dysfunction and postural instability in patients with PD^189,190 drove clini- cal trials for further evaluation of the effects of PPN-DBS on gait and freezing. Results have been suboptimal owing to several factors, including a limited number of patients, variation in methodology, inconsistent definition of freezing in patients with PD, imprecise definition of the location of stimulation and poor choice of stimulation programming parameters. The SNr has also been eva- luated as a target to improve refractory axial symptoms and has shown promising results^191,192. Several other DBS targets have been emerging for treatment of motor symptoms (including the zona incerta, posterior subtha- lamic area and centromedian thalamus). Further analy- sis of results that establish common standard protocols for targeting and programming will be needed.

Adaptive DBS and individualized targeting.

Adaptive DBS is a responsive DBS system that is capable of sens- ing physiological signals and auto-adjusting stimula- tion, which could improve clinical benefit, limit adverse effects and reduce neurostimulation battery costs. The most challenging aspect of this therapy has been detec- tion of a reliable physiological signal to target. Beta band activity has been the most studied physiological signal, but more work will be needed to assess its reliability¹⁹³. Individualization of DBS therapy utilizing diffusion tensor imaging and functional MRI to identify appro- priate networks for planning targets has the potential to improve clinical DBS outcomes.

Non-invasive and optogenetically inspired DBS.

A method for non-invasive DBS has emerged through the application of temporal interference. Temporal interfer- ence consists of the application of two overlapping electric fields of different frequencies via an external device. The difference in frequencies in the overlap region of the two fields delivers stimulation deep in the brain, without inter- rupting the function of superficial structures¹⁹⁴. This tech- nique remains at a preclinical level but has the potential to reduce the risks associated with invasive DBS (fig. 4).

Fig. 4 |. Neuromodulation.

Deep brain stimulation is an established therapy for Parkinson disease, but there is growing interest in non-invasive electrical stimulation modalities and optogenetic stimulation modalities. Non-invasive electrical stimulation might be achieved with interference of two electric fields with slightly different frequencies (temporal interference), leading to localized stimulation. In optogenetic stimulation, an opsin is delivered to specific neurons that then express a light-sensitive channel on their membrane, permitting their stimulation with direct light source. Alternatively, certain nanoparticles, if delivered inside the brain, are able to convert infrared light from an extracranial source to visible light in the brain and hence permit the non-invasive stimulation of optogenetically modified neurons. Electrical stimulation approaches are generally non-specific, whereas optogenetic approaches have the potential for targeting specific neurons.

Optogenetics, a technique that uses light to control genetically modified neurons expressing light-sensitive ion channels and pumps, has gained interest in the past few years (fig. 4). Optogenetically defined DBS for the GPe has demonstrated the feasibility of exciting or inhibiting different cells types to enhance the treatment response in animals¹⁹⁵. Identification of suitable targets and precise light delivery with penetration to deep brain structures remain challenges for this therapy. One study successfully demonstrated that upconversion nanopar- ticles can absorb near-infrared light emitted outside the brain to convert it into local emission of visible light and stimulate deep brain structures (fig. 4). This approach could facilitate the process of less invasive optogenetic manipulation of neuronal activity¹⁹⁶. The safety of these nanoparticles in humans remains to be established.

Potential for disease modification.

Evidence from animal models has suggested the possibility of the neuroprotec- tive effects of STN-DBS. However, results from clinical trials have yet to support this notion. STN-DBS did not protect the nigrostriatal system against α-synuclein- mediated toxicity, according to one study¹⁹⁷. Delivery of autologous nerve grafts (as a source of neurotrophic factors) to the SNr during STN-DBS surgery has been another approach used in an open-label trial¹⁹⁸, but larger trials are needed to evaluate the efficacy and disease-modifying potential of this strategy. This new approach has been termed DBS-plus.

Conclusions

PD is associated with a range of pathophysiological processes, including α-synuclein aggregation, neu- roinflammation, mitochondrial dysfunction, neuronal vulnerability, iron deposition and neural network alter- ations. The complexity of these intertwined pathways and the heterogeneity in clinical phenotypes will require a targeted approach for therapy. Although current treat- ment options provide symptomatic relief, advances in high-throughput screening methods for small mol- ecules, improved disease modelling and progress in analytical technologies are likely to facilitate novel compounds and repurposed drugs. Immunotherapies could provide a novel mechanism for the body to boost its response to α-synuclein. Research in the field of cell-based therapies has provided an improved com- prehension of the disease, and some iPSC therapies could be used in personalized therapy. Adaptive DBS and optogenetically inspired DBS might aid the defi- nition of more specific targets. Taken together, these advancements suggest that the future of PD therapies is promising.