The role of modulation of antioxidant enzyme systems in the treatment of neurodegenerative diseases

Abstract

Oxidative stress is a much-appreciated phenomenon associated with the progression of neurodegenerative diseases (NDDs) due to imbalances in redox homeostasis. The poor correlations between the in vitro benefits and clinical trials of direct radical scavengers have prompted research into indirect antioxidant enzymes such as Nrf2. Activation of Nrf2 leads to the upregulation of a myriad of cytoprotective and antioxidant enzymes/proteins. Traditionally, early Nrf2-activators were studied as chemoprotective agents. There is a consequential lack of clinical trials testing Nrf2 activation in NDDs. However, there is abundant evidence of their utility in pre-clinical studies. Herein, we review the endogenous Nrf2 regulatory pathway and avenues for targeting this pathway. Furthermore, we provide updated information on pre-clinical studies for natural and synthetic Nrf2 activators. On the basis of our findings, we posit that successful therapeutics for NDDs rely on the design of potent synthetic Nrf2 activators with a careful combination of other neuroprotective activities.

Keywords:

• Keap 1
• neurodegeneration
• Nrf2
• oxidative stress
• ROS

Introduction

To date, there has been a myriad of theories mapping the molecular mechanisms that underlie the pathology of NDDs. Mitochondrial dysfunction, oxidative stress, chronic inflammation and accumulation of protein aggregates have been identified in unison as pathological factors associated with a broad spectrum of NDDs such as Alzheimer’s disease (AD), Parkinson’s disease (PD), Huntington’s disease (HD) and amyotrophic lateral sclerosis (ALS)1,2. Although mitochondrial dysfunction has been implicated as a common feature among these heterogeneous diseases3–6, the role of oxidative stress has been identified as a major hallmark and the key mitochondrial insult consistently linked to the development of NDDs7. An imbalance between reactive oxygen species (ROS) production and the detoxification of reactive intermediates is associated with oxidative stress2 and subsequent attack of free radical species on neuronal cells.

Having failed in clinical trials, free radical scavengers have been explored extensively as means to reduce the severity of ROS onslaught on neuronal cells. Transcription factor NF-E2-related factor 2 (Nrf2) is generally believed to be an adaptive cellular response to endogenous and environmental stress8. Thus its emerging role in oxidative stress is that of a regulatory function against oxidants9. Activated by a variety of electrophiles and oxidants, Nrf2 is a basic leucine zipper/Cap ‘n’ collar protein, which regulates antioxidant response elements (ARE) containing genes10,11. The pharmacological activation and genetic up-regulation of Nrf2 has been shown to be fundamentally neuroprotective in a number of experimental models12.

In relevant Nrf2 knock out mice, the deletion of the Nrf2 gene in mice resulted in neurodegeneration, widespread astrogliosis13, increased sensitivity to neuroinflammation, severe dopaminergic dysfunction14, lack of ARE activation15 and show the neuroprotective function of Nrf2. Nrf2 knockout mice displayed increased vulnerability to common neurotoxins16,17 and the subsequent transplantation of Nrf2-overexpressing astrocytes protected against malonate18 and 6-OHDA16 insults. Further to these reports, a number of natural and synthetic compounds have been shown to serve as Nrf2 activators and of benefit in several in vivo models. In this review, we posit that the induction of Nrf2 is an exciting therapeutic target and the design of synthetics targeting this pathway may provide effective tratments in NDDs.

ROS/RNS-mediated oxidative stress

Based on its high oxygen consumption and the relatively reduced regeneration of neuronal cells, the human brain is predisposed to oxidative stress19. There is a paradoxical influence of oxygen due to its rather toxic nature via the accumulation of reactive oxygen species (ROS), reactive nitrogen species (RNS) and the disturbance of pro-oxidant equilibrium20,21. In neurodegeneration, reduced levels of antioxidant enzymes along with increased ROS/RNS in specific areas of the brain have been shown to damage several key macromolecules such as DNA, proteins and lipids22–26.

Reactive oxygen species (ROS) are oxygen-derived radicals such as superoxide (O^•-₂), hydroxyl (HO^•), alkoxyl (RO^•) and peroxyl (ROO^•), as well as non-radicals such as hydrogen peroxide (H₂O₂) and hypochlorous acid (HOCl)27,28. This list of reactive species can be expanded to include nitric oxide (NO^•), nitrogen dioxide (NO^•₂) and the more recently recognized peroxynitrite (ONOO⁻), which is classified as reactive nitrogen species (RNS)29. The production of oxidants following specific physiological cues serves as important signalling molecules in immune function, during redox signalling, in autophagy and cell division9,30. In neurodegeneration, the uncontrolled production of ROS/RNS impairs neuronal function and contributes to neuronal death31 particularly in the brain, where ROS-mediated redox signalling is actively involved in neurogenesis, neuronal differentiation, plasticity and memory consolidation32. Under physiological conditions, ROS production and elimination is tightly regulated under “redox homeostasis” with the maintenance of low ROS concentration33. The tight regulation of ROS production via redox homeostasis and the disturbance thereof causes overproduction of ROS and deficiency in enzymatic and non-enzymatic antioxidants. The generation of ROS/RNS in the brain is mainly due to premature “electron leaks” in the electron transport chain of the mitochondria34 and the production of superoxide during energy transduction in the mitochondrial has been linked with ageing and mitochondrial damage35.

The mitochondrion is described as a dynamic organelle, playing a crucial role in promoting the intact regulation of cellular function by controlling energy metabolism, mitochondrial shape/number as well as its DNA content1. It has been shown to be of great importance in energy production and cell physiology and impairment thereof has been implicated in a number of NDDs36–39. Dysfunctional mitochondria result in the reduction in the cellular supply of energy and failure in the maintenance of cellular homeostasis with the activation of cellular death pathways40. The pathology of several other NDDs displays a vast spectrum of mitochondrial dysfunction including respiratory chain dysfunction and oxidative stress, reduced ATP production, calcium dysregulation, mitochondrial permeability transition pore opening, perturbation in mitochondrial dynamics and deregulated mitochondrial clearance41. PD has received the greatest attention in relation to mitochondrial dysfunction and mitochondrial ROS production with reductions in complex I activity as demonstrated in the substantia nigra, lymphocytes and platelets41. Typically, the overproduction of amyloid beta (Aβ) in the postmortem brains of AD patients was indicated to induce imbalances in mitochondrial fission and fusion processes42,43. Finally, imbalances in mitochondrial dynamics were also observed in ALS models44,45.

Several enzymes in the brain have been shown to contribute to the production of ROS/RNS. As part of physiological function, NADPH oxidase (NOX)46, monoamine oxidase B (MAO-B)47, nitric oxidase synthase (NOS)48 and xanthine oxidase (XO)49,50 generate ROS/RNS. Typically, NOX-mediated production of ROS has been shown to play a critical role in neuronal apoptosis51 with evidence of microglial and neuronal NOX involvement in NDDs52,53. Furthermore, the expression of monoamine oxidase (MAO) in patients with AD was reported to be significantly higher than in the healthy controls of the same age54. MAO-B elevation in murine brain astrocytes resulted in several hallmarks of PD pathology such as decreased mitochondrial complex I activity, increased mitochondrial oxidative stress, activation of microglia in substantia nigra and finally progressive loss of dopaminergic neurons55.

Amidst the failure of several antioxidants with free radical scavenging roles in human trials, the call for targeting innate cellular protective mechanisms have drawn much attention to the Nrf2 pathway and its activators. As the master regulator of cellular response to oxidative stress, Nrf2 appeared as a promising target as modulation of this target could provide an alternative to the drawbacks associated with exogenous antioxidant therapy. Therefore, Nrf2 activating compounds have been explored as a means of inducing these endogenous antioxidant enzymes. This review thus summarizes the current information available on these compounds highlighting the need to delve into synthetics with improved bioavailability as Nrf2 modulators for the treatment of NDDs.

Nrf2/ARE pathway

The transcriptional response to reactive chemical stress is via the induction of a number of cytoprotective enzymes, which is mediated by the cis-acting Antioxidant-Response-Element (ARE)56. Nrf2 was identified as transcriptional factor acting on the ARE with in vivo studies implicating severely impaired ARE-dependent genes in knockout (Nrf^−/−) mice10,57. Nrf2 is known to function independently and dependently of Keap-1 to control the inducible and constitutive expression of ARE.

Keap1-dependent regulation of Nrf2

When first discovered, Keap1 was identified as a cytoplasmic protein that binds Nrf2 at the N-terminal Neh2 domain (Figure 1)58. Afterwards, the Keap1 dimerisation via its BTB was found to be required for Nrf2 repression and its ETGE motif in the Neh2 domain identified as critical for binding Keap 159. At this point, Keap1 was viewed simply as a passive repressor of Nrf2 by virtue of sequestering it in cytoplasm and, therefore, preventing its nuclear translocation. This notion was supported by the fact that Nrf2 constitutively accumulates in the nuclei of Keap1-knockout mice60. Therefore, the first plausible model of Nrf2 activation comprised the disruption of the Keap1–Nrf2 complex resulting in nuclear translocation of Nrf2.

Figure 1. The domains of Nrf2 and Keap1. Both Nrf2 and Keap1 are highly conserved between species. In Nrf2 six homology domains, Neh1–6, were identified. The domains of interest in Nrf2 are Neh1, containing DNA-binding basic leucine zipper domain; Neh2, containing Keap1-binding redox-sensitive degron; and Neh6, containing redox-insensitive phosphodegron. The domains of interest in Keap1 are the Broad complex, Tramtrack and Bric-a-brac (BTB), the intervening linker region (IVR or simply linker) and Kelch repeats domain (Kelch). Keap1 homodimerizes via its BTB domain and binds to Nrf2 via its Kelch domain. Electrophilic inducers react with Cys 151 in the BTB domain abolishing Keap1 repressor activity. The linker region contains a number of cysteine residues critical to constitutive Keap1 repressor activity.

Figure 2. Modes of Nrf2 activation. Under unstressed conditions Nrf2 is constantly recruited for ubiquitination by Keap1–Cul3–Rbx complex resulting in low constitutive level of activity. Electrophilic compounds or ROS covalently bind to reactive cysteines interfering with the normal function of the complex and resulting in accumulation of Nrf2. Nrf2 then translocates into nucleus and drives transcription of antioxidant/detoxifying enzymes. PPI inhibitors achieve the same goal by preventing Nrf2 binding to Keap1.

Indeed, treatment with electrophilic inducers such as sulphoraphane61, diethyl maleate62 and β-naphthoflavone63 caused nuclear accumulation of Nrf2 and transcription of its target genes62. However, a number of electrophilic chemicals were shown to be unable to disrupt the Nrf2–Keap1 complex, suggesting that the Keap1–Nrf2 dissociation model is incorrect64–66. Furthermore, the accumulation of Nrf2 requires its de novo synthesis62, but the classic electrophilic inducers do not affect the rate of transcription of Nrf263,67,68. Indeed, Nrf2 was identified as a substrate of the ubiquitin–proteasome pathway69,70. Furthermore, the binding of Keap1 to Nrf2 was found to actively target Nrf2 for proteasomal degradation at a higher rate than unbound Nrf2, suggesting a more active role of Keap1 in the repression of Nrf2 activity61,71. This was a breaking point, as soon after Keap1 was found to function as an adaptor bridging Nrf2 and Cul3–Rbx1 E3 ligase and targeting Nrf2 for ubiquitination64,72–74.

The Nrf2 activation by electrophilic inducers has been thought to result from direct alkylation or oxidation of reactive cysteine residues on Keap1 resulting in dissociation of the Keap1–Nrf2 complex. Indeed, a number of reactive cysteine residues have been identified71,75. Particularly, Keap1–C151S mutant was insensitive to tert-butylhydroquinone (tBHQ) and SFN treatments and continued to repress Nrf2 activity under both basal and induced conditions, while both Keap1–C273S and Keap1–C288S mutants completely lost the ability to repress Nrf2 under any conditions71. The critical role of C273 and C288 for constitutive repression of Nrf2 was later confirmed in vivo76. Keap1 was reported to have a zinc atom coordinated to cysteine thiols, possibly in the linker region, making them more reactive77. This prompted another model, where electrophilic inducers replaced zinc and resulted in conformational change and release of Nrf277. Later, C151 was identified as the most reactive target towards Michael acceptors78. However, the activation of Nrf2 by arsenic was found to be independent of C15179. In view of reports of multiple reactive cysteines, it was proposed that Keap1–Nrf2–Cul3 complex constitutes a multiple-sensing mechanism80. The reactivity of Keap1 cysteine residues with various inducers has been reviewed81.

Although the complete crystal structure of the Keap1–Nrf2–Cul3 complex has not been resolved yet, some individual domains and their interactions have been revealed. The crystal structure of the Kelch motif was identified as a six β-propeller structure resembling a “doughnut” shape82. In addition to the high affinity ETGE motif, a secondary low affinity DLG motif was identified in the Neh2 domain83. Soon after, the crystal structures of Kelch motif bound to ETGE and DLG motifs revealed their binding at the centre groove of Kelch motif84,85. Keap1 was found to exist as a homodimer via its BTB motif in the Cul3 ligase complex, which led to the widely accepted “hinge and latch” model86,87 According to this model, Nrf2 binds the Keap1 homodimer via its ETGE and DLG motifs positioning the middle lysine rich alpha helix of the Neh2 domain for ubiquitination by Cul3 ligase complex. This model explains the faster rates of Keap1-mediated ubiquitination and degradation of Nrf2 under unstressed conditions. In support of this model, a low-resolution Keap1 dimer structure revealed two large spheres attached by short linker arms to the sides of a small forked-stem structure, resembling a “cherry-bob”88. The crystal structure of BTB domain of Keap1 has recently been elucidated, revealing a number of basic amino acids in close proximity of Cys-151, explaining its highly reactive nature89. The authors further revealed the structure of bardoxolone, a potent Keap1 antagonist, covalently bound to Cys151. They further demonstrated that this interaction was sufficient to prevent Cul3 binding, and thus inhibit the proteasomal degradation of Nrf2. The latest structural findings on the interactions within the Nrf2–Keap1–Cul3 complex have been reviewed in detail by Canning and Bullock90.

Keap1-independent regulation of Nrf2

Various canonical kinases have long since been implicated in the regulation of Nrf263,91–93. Activation of PKC directly phosphorylates Nrf2 at Ser-40 and leads to its translocation into nucleus93–95. Subsequently, this phosphorylation at Ser-40 was shown to be necessary for its dissociation from Keap1, but not required for Nrf2 accumulation in the nucleus96.

GSK-3β was later identified as a common downstream effector of PI3K/Akt97,98, PKC99 and MAPK/ERK100. Additionally, glycogen synthase kinase-3 beta (GSK-3β) was shown to negatively regulate Nrf2 by reducing its nuclear localisation14,97 and acting to phosphorylate a number of Src subfamily kinases, which translocate into the nucleus and phosphorylate Nrf2 leading to its nuclear export101,102. All these results suggest that GSK-3β could be a potential target. The utility of GSK-3β/Nrf2 regulatory pathway was proven in vivo by the synergistic use of a GSK-3β inhibitor and an Nrf2 activator to protect the mice brain against a neurotoxin103.

More recently another regulatory pathway was reported. The Neh6 domain of Nrf2 contains a cluster of serine residues that are phosphorylated by GSK-3104,105. The β-transducin repeat containing protein (β-TrCP) adapter protein recognizes phosphorylated Nrf2 and targets it for ubiquitination by SKP1–CUL1–F-box (SCF), a Cul1 E3 ubiquitin ligase104. The inhibition of GSK-3 in vivo and in vitro was shown to lead to activation of Nrf2, possibly via disruption of β-TrCP–Nrf2105,106. Within the Neh6 domain, two short β-TrCP recognition motifs have been identified, of which one requires phosphorylation by GSK-3107. These findings prompted an updated “dual degradation” model of Nrf2 regulation, which would explain the varying levels of Nrf2 activity under stressed and unstressed conditions104. The increased Nrf2 activity caused by downregulation of the two repressors Keap1 and β-TrCP has been correlated to longevity in mice108.

Nrf2 activators

The ability of Nrf2 activators to mediate a global antioxidant response presents a highly attractive strategy to counteract chronic oxidative stress at the cellular level. With the myriad of in vitro and in vivo models showcasing its regulatory role and impact in neuroprotection, natural and synthetic Nrf2 activators have been reported with varying modes of Nrf2 activation (Figure 2). These activators function as indirect inhibitors of the Keap1–Nrf2 complex by interacting with the cysteine sulphydryl groups. In theory, these compounds can promote an innate control of genes encoding a vast array of antioxidants and proteins overcoming drawbacks associated with exogenous antioxidants.

Natural Nrf2 activators

A number of natural Nrf2 activators (Figure 3) have been reported yet, despite strong rationale and evidence backing their use, their development has been met with a number of hurdles amidst their poor bioavailability, solubility issues, high cost of human trials and stability problems.

Figure 3. Structures of natural of Nrf2-activators.

The natural polyphenolic drug curcumin is a compound extracted from the turmeric plant with a distinct structure, in that it possesses two methoxyphenol groups linked to a β-diketone bridge109. This confers both Michael acceptor and metal chelation properties110. Curcumin has been known as an antioxidant for over a decade and has thus been evaluated in NDDs for its role as a free radical scavenger of the superoxide and hydroxyl ions.

Recently, its selective role on the Nrf2–Keap1 ARE pathway has been the focus of study. In rotenone-treated mice, an Nrf2-specific short pin RNA or phosphoinositide 3-kinase inhibitor demonstrated a marked attenuation in tyrosine hydroxylase and glutathione111. As a result, the neuroprotective effects of curcumin against oxidative damage were eliminated. In a likewise manner, resveratrol was used as a polyphenol and examined for its neuroprotective behaviour112. Resveratrol was shown in this study to function as an Nrf2 activator. Furthermore, the authors indicated that due to the depletion of Nrf2 in astrocytes, there was a reduction in neuroprotection. In Nrf2^−/− astrocytes, a decreased mitochondrial antioxidant expression and an inability to up-regulate cellular antioxidants following preconditioned treatment were also recorded. There is much interest in utilizing curcumin and resveratrol for AD therapy due to their influence on amyloid beta, antioxidant and anti-inflammatory properties113,114. On the molecular level, as a Michael acceptor, the electrophilic α,β-unsaturated carbonyl groups of curcumin is believed to selectively interact with cysteines on Keap1 enabling the release of Nrf2 for action115.

Several clinical trials proceeded to evaluate the efficacy of curcumin administration in various pathological forms of NDDs but these results were not encouraging. Typical examples are the trials coded NCT00164749 and ACT00099710 for AD116,117. In the first study, on one hand, patients with AD were evaluated using a combination of curcumin and ginko in the treatment of AD (NCT00164749). Results indicated no differences between the treatment groups in Aβ levels and mini-mental stage examination (MMSE) scores. On the other hand, patients with mild to moderate AD were treated with curcumin C3 complex (ACT00099710). This study also indicated no difference between treatment groups and although the doses were tolerated (2 g and 4 g daily), the authors reported low bioavailability. A number of trials have been conducted evaluating reservatrol in the treatment of AD (NCT01504854 and NCT00678431). The study, NCT01504854, reported no difference between groups when AD duration was measured in years with no alteration in results following a post-hoc re-analysis118. The diverse off-target effects of reservatrol were discussed which only engaged the central molecular target in the nM range. This raises concerns generally about the lack of specificity of the natural activators, thus calling for the design of synthetics with potent activity and reduced off-target effects.

Another natural drug of interest is puerarin, which is an antioxidant of kudzu roots, which was previously evaluated and reported to possess neuroprotective role. Li et al.119 recently showed that puerarin treatment of rat substantia nigra neurons attenuated 6-OHDA caused dopaminergic neuronal degeneration, precipitated as a result of oxidative stress. In this study, glutathione and catalase levels in the substantia nigra were gradually elevated and Keap1 Nrf-2 was shown to be progressively elevated following treatment. The authors suggested that puerarin activated the nuclear translocation within neurons via the Nrf2/ARE pathway which then triggered the synthesis of antioxidant enzymes to mediate ROS attack/lesions.

Gastrodin, on the contrary, was evaluated for its neuroprotective effects in PD models. In an MPTP mice model of PD, gastrodin pretreatment was shown to forestall neurotoxicity120. Induced oxidative stress was prevented and gastrodin treatment in MPTP intoxicated mice resulted in a robust increase in heme oxidase 1, superoxide dismutase, glutathione levels as well as the nuclear translocation of Nrf2 with the implication of the ERK1/2 phosphorylation pathway.

The flavonoid, baicalein, was also tested in a 6-OHDA neurotoxic model and shown to display neuroprotection in vitro121. The flavonoid was useful only as a prophylactic, as it failed to exhibit restoration after 6-OHDA induced cellular damage. Treatment with baicalein showed elevated Nrf2/hemo oxygenase and reduced Keap1 expression in a time and dose–response manner with the activation of the PKCα and PI3K/AKT pathway.

Sulphoraphane (SFN) is an isothiocyanate and sulphoxythiocarbamate that has been recognized for its antioxidant properties. It was evaluated in several in vivo models of AD, PD and cell models of oxidative and nitrosative stress. Previously, SFN and allyl disulphide were shown to enhance glutathione and suppress the loss of neurons in both parkin and α-synuclein Drosophila models12. This raised the possibility of their use as therapeutics in PD. The potency of SFN in stimulating the Nrf2/ARE pathway has been shown using behavioural tests that rely on the Y-maze122. Regrettably, they were found to be of no use on Aβ aggregation, a principal pathophysiological characteristic of neuronal dysfunction in AD122. Interestingly, SFN protected neurotoxicity following 6-hydroxydopamine treatments in PD with its neuroprotective effect linked to its modulatory role on neuronal pathways and enhancement on glutathione-S-transferase and glutathione reductase123. Conversely, using allicin (diallyl thiosulphide) Aβ-memory impairment was ameliorated and activated Nrf2 hippocampal level in aging mice was observed124,125. Allicin has been shown to rescue tunicamycin-induced neuronal stress following pretreatment126. In this study, allicin acted by moderately elevating RNA-dependent protein kinase (PRK)-like ER-resident kinase (PERK) and its downstream substrate, Nrf2.

Although earlier epidemiological studies revealed reduced risk between tobacco user and coffee drinkers, Trinh et al.127 demonstrated that caffeine and nicotine were not neuroprotective in Drosophila strains as measured by the number of dopamine neurons in 20-day old transgenic flies. Rather, the results revealed that decaffeinated coffee and nicotine-free tobacco provided neuroprotection in the α-synuclein and parkin Drosophila model. In this model, 0.15% decaffeinated coffee provided more neuroprotection than 0.03% nicotine tobacco; the converse was observed in the parkin PD model. The Nrf2 activating component was identified as cafestol in coffee and results implicated its neuroprotective effects with the up-regulation of glutathione. The serum cholesterol elevation effects of cafestol in this work highlight the need to design structural synthetics that lacked this side effect. The Nrf-2 activating compounds in coffee and tobacco were presumed capable of crossing the blood–brain barrier and thus highly attractive therapeutic agents.

Synthetic Nrf2 activators

Synthetic oleanane triterpenoids are the most studied synthetic activators of Nrf2 (Figure 4). They have been shown to possess anti-inflammatory, cytoprotective, antiproliferative and pro-apoptotic properties making them potential candidates for the treatment of a wide range of diseases128.

Figure 4. Structures of the synthetic oleanane triterpenoids.

Initially, 2-cyano-3,12-dioxooleana-1,9(11)-dien-28-oic acid (CDDO) was developed as a highly potent anti-inflammatory agent based on the structure of oleanolic acid129,130. However, the poor oral bioavailability of CDDO led to the development of many other derivatives. The details of their development and structure-activity relationship studies have been recently reviewed by Sporn et al.131. Of note are the two Michael acceptor enone motifs in the rings that are critical for their anti-inflammatory activity132. Mechanistically, CDDO and its derivatives are highly potent Nrf2 activators77,133 and NFκB pathway inhibitors134,135.

CDDO imidazolide (CDDO-Im) induced Nrf2-dependent antioxidant genes, attenuated LPS-induced proinflammatory cytokine expression and decreased mortality in mice136. CDDO-methylamide (CDDO-MA) was able to cross the blood–brain barrier and significantly improved spatial memory retention and reduced plaque burden, Aβ42 levels, microgliosis, and oxidative stress in Aβ-mice model of AD137. In another study, CDDO-MA demonstrated significant neuroprotection against MPTP- and 3-nitropropionic acid (3-NP)-induced nigrostriatal dopaminergic neurodegeneration in mice and rats138. Oral administration of either CDDO ethyl amide (CDDO-EA) or CDDO trifluoroethyl amide (CDDO-TFEA) blocked MPTP-induced dopaminergic neurotoxicity in mice139. Both CDDO-EA and CDDO-TFEA penetrated blood–brain barrier, enhanced motor performance and extended the survival of G93A SOD1 transgenic ALS mice140. When orally administered to N171-82Q Huntington transgenic HD mice, both CDDO-EA and CDDO-TFEA upregulated Nrf2/ARE-induced genes in the brain and peripheral tissues, reduced oxidative stress, improved motor impairment and survival141. CDDO methyl (CDDO-Me or bardoxolone methyl) administration to mice following ischemia–reperfusion injury upregulated Nrf2 and HO-1, decreased infarct volume and improved neurological symptoms142. As these results indicate, without a doubt, CDDO derivatives are highly potent activators of Nrf2 and inducers of its target genes both in vitro and in vivo.

To date, there are no clinical trials testing synthetic oleanane triterpenoids for the treatment of NDDs. However, there are a number of clinical trials that tested bardoxolone methyl (CDDO-Me) for the treatment of cancer and chronic kidney disease (CKD). Although bardoxolone methyl was found to be generally safe in earlier trials, a phase 3 trial (NCT01351675) evaluating its use for the treatment of chronic kidney disease was terminated in 2012 due to increased rates of heart-related adverse events143. There is a concern, however, of whether late stage CKD was the right clinical indication to pursue144. It is feasible that drugs with improved specificity and better toxicity profile could prove beneficial in exploiting the protective and preventive functions of Nrf2 pathway in treating patients with early stages of degenerative diseases instead.

Dimethyl fumarate (DMF) is an orally bioavailable chemically simple compound with a Michael acceptor functionality. In two randomized, double blind, placebo-controlled phase 3 studies (NCT00420212 and NCT00451451), dimethyl fumarate (BG-12) significantly reduced the proportion of patients who had a relapse145,146. The United States Food and Drug Administration (FDA) and the European Medicines Agency (EMA) have subsequently approved Dimethyl fumarate in 2013 and 2014, respectively, for adult patients with relapsing-remitting MS. Although its mechanism of action in humans is not fully clear, there is a plenty of animal and cell-based reports demonstrating the involvement of Nrf2 activation147–149.

DMF protected rat neural progenitor cells and motor neurons against H₂O₂-induced oxidative stress149. Both DMF and its primary metabolite, monomethyl fumarate, significantly protected neuronal and astrocyte cultures against H₂O₂-induced toxicity in an Nrf2-dependent manner147,150. In two HD mice models, the treatment with DMF preserved the motor function and the morphology of neurons in the striatum and the motor cortex151.

Carbon monoxide (CO), the product of HO-1 metabolism, mediates potent anti-inflammatory and anti-apoptotic effects152,153. For this reason, a number of synthetic hybrids (Figure 5) combining Nrf2-activating and CO-releasing functionalities, named “HYCOs”, have been developed154,155.

Figure 5. Structures of Nrf2-activating carbon monoxide synthetic hybrids.

The Nrf2-activating component is the well-known Michael acceptor enone motif, while the CO liberating component is a cobalt carbonyl complex. Interestingly, CO is only liberated when cobalt is oxidized under conditions of oxidative stress. HYCOs (Figure 5) markedly increased Nrf2/HO-1 expression, liberated CO and exerted anti-inflammatory activity in vitro and in vivo. At least in vitro, HYCO-1 reduces inflammation and NO production more effectively than dimethyl fumarate, suggesting the synergistic/additive effect of the two functions.

Chalcones are natural compounds abundant in edible plants and possess anti-inflammatory and anti-cancer activities among other biological activities156,157. The chalcone scaffold is a well-known Michael acceptor having enone functionality sandwiched between two aromatic rings. Notably, the α,β-double bond and the carbonyl group are critical for the induction of HO-1158,159.

Structure–activity relationship studies around the chalcone scaffold yielded 1 (Figure 6) as a potent activator of Nrf2 in vitro and in mice160. Recently, its heterocyclic analogue 2 demonstrated improved aqueous solubility and oral bioavailability, which translated into higher activity in vivo161. Synthetic chalcones 3a and 3b at 10 μM were non-toxic and upregulated g-Glutamylcysteine Ligase Catalytic Subunit (GCLC) and haeme oxygenase-1 (HO-1) up to 3- and 25-folds, respectively, and attenuated H₂O₂-induced apoptosis of in vitro150. E-α-(4-methoxyphenyl)-2′,3,4,4′-tetramethoxychalcone4 at 30 μM caused translocation of Nrf2 into nucleus, potently induced HO-1 and downregulated NFκB activity with no observed toxicity at higher concentrations151.

Figure 6. Structures of chalcone scaffolds with Nrf2 activity.

The high electrophilicity of traditional Nrf2-activators poses the risk of “off-target” effects due to their ability to react with the cysteines of other proteins. Therefore, there has been an increased interest in the direct targeting of Nrf2/Keap1 protein–protein interaction (PPI) as an alternative strategy (recently reviewed by Wells162 and Abed et al.163). High-throughput screening using fluorescence polarisation (FP) assay yielded several micromolar hits164,165. Remarkably, the structure-based design by Jiang et al.166 led to the discovery of the first nanomolar inhibitor of Keap1–Nrf2 PPI in FP assay. However, all these compounds have only micromolar activities in cell-based assays. Very recently a fragment-based approach yielded KI-696 (Figure 7), a highly active Nrf2 activator in vitro and in vivo167. Targeting Keap1–Nrf2 PPI is certainly feasible and we have yet to see the full potential of this approach.

Figure 7. Chemical structure of KI-696.

Conclusion and perspective

Following the failure of radical scavengers in NDDs, recent efforts have been aimed at the identification of alternative targets of which the master regulator Nrf2 took the centre stage. The activation of Nrf2 by any of the discussed approaches in this review can up-regulate key antioxidant/detoxifying enzymes and restore the redox homeostasis within a diseased neuron.

In this review, we highlight a number of reported Nrf2-activators with the emphasis on pre-clinical evidence. The most active and well-studied activators possess an electrophilic Michael acceptor motif that covalently binds to reactive cysteines in Keap1 disrupting its Nrf2-repressor activity. This property alone is not sufficient, as such molecules could possess off-target effects as discussed. A logical approach would be to design a molecule around the Michael acceptor motif to couple high reactivity and selectivity obtained from its crystal structure information. However, this approach is hindered by the lack of complete crystal structures of Nrf2 and its repressor proteins. Nonetheless, we are starting to see first rational designs come to light based on partial crystal structures. In the last couple of years, PPI inhibitors are gaining momentum and the latest reports show very promising lead compounds. We are seeing an increased number of rational drug design approaches that could lead to novel synthetic molecules.

The utility of Nrf2 as a therapeutic target has been clearly established in animal models of neurodegeneration. Traditionally, clinical trials evaluating natural Nrf2 activators, with the exception of curcumin, have focused on other diseases such as cancer. There is a striking lack of clinical trials evaluating these compounds for treatment of NDDs. However, clinical trials utilizing curcumin in both PD and AD demonstrated limited efficacy, likely due to its known poor bioavailability. DMF on the contrary, has been recently approved for the treatment of MS and its beneficial action is at least in part due to the activation of Nrf2168. Conversely bardoxolone was pulled out of a human trial due to its cardiovascular risk further stressing the need to design highly selective Nrf2 activators.

We posit that the utility of Nrf2-activators may lack the ability to fully restore neuronal function but may be useful for preserving surviving neurons. Therefore, testing Nrf2 activation in early stage patients could prove more relevant to this pathway. It is logical to conclude that even a marginal improvement in late-stage patients could imply a substantial benefit in the form of delayed disease progression of neurodegeneration.

In view of the poor bioavailability and lack of specificity of the natural activators, the development of highly potent rationally designed synthetic Nrf2 activators could serve as promising leads in the treatment of NDDs. A highly potent and selective Nrf2 activator could be of high value to address some of the hallmarks of neurodegeneration, such as oxidative stress. However, a more complete and effective treatment may require a careful combination of Nrf2 activating component with other neuroprotective action(s) to better address the complexity of neurodegeneration.

Declaration of interest

The authors report that they have no conflict of interest.