KEAP1 is a Zn metalloprotein with 625 amino acid residues and it has a branched stem dimeric structure, formed by 5 domains [67]. The N-terminal region (NTR), the BTB/POZ (Bric-a-brac, tramtrac, broad-complex/proxvirus zinc fingers) domain, implicated in KEAP1 homodimerization and Cul3 association, the intervening region (IVR), which acts as a linker between BTB and DGR domains, the double glycine repeat (DGR) domain, important for NRF2 regulation and actin interaction and the C-terminal domain (CTR) [67] (Figure 6).

KEAP1 is a cysteine-rich protein, being redox-active and reactive to the local environment [68]. The majority of KEAP1 cysteine residues are surrounded by basic amino acids that increase their reactivity by lowering their pKa values [69]. Thus, the 27 cysteine residues of KEAP1 are proposed to be the main sites through which the protein is able to sense electrophilic or oxidative stress [70]. Even though cysteines are distributed along the whole KEAP1 sequence, it has been proved that the C²⁷³, C²⁸⁸, and C²⁹⁷ residues in the BTB/POZ domain [71] and C¹⁵¹ in the IVR domain [72] are more reactive and serve as redox sensors for electrophiles, reactive oxygen species (ROS), and metals such as Cd²⁺, As³⁺, and Se⁴⁺ [73]. The chemical modification of their sulfhydryl groups decreases NRF2 ubiquitination, causes KEAP1–NRF2 dissociation, and promotes the nuclear translocation of NRF2 [74].

The ‘hinge and latch’ model [75] is the most widely accepted mechanism for KEAP1-dependent regulation. In this model, two different binding sites of NRF2 in the Neh2 domain interact with different affinities with a single overlapping site in the DGR domain of KEAP1, with the ETGE motif presenting higher affinity than DLG [76]. Under homeostatic conditions, the β-hairpin of the ETGE motif (high affinity) interacts with one of the KEAP1 subunits in a first step, which is called the open conformation [76,77]. This open conformation represents the “hinge”, in which NRF2 can move in space relatively freely. In a second step, the β-hairpin of the DLG motif (low affinity) interacts similarly with a second protomer of KEAP1 in the so-called closed conformation, representing the “latch” [76,77]. This closed conformation restricts the ability of NRF2 to move and enables an optimal positioning of a Lys-rich α-helix present in the Neh2 region of NRF2; this helix is the target for ubiquitin conjugation, consequently promoting NRF2 for proteasomal degradation [77].

Under oxidative stress conditions, NRF2 inducers oxidize the Cys residues present in KEAP1, promoting a conformational change [77]. This structural change in KEAP1 causes an improper spatial disposition of the target lysines, and thus NRF2 is no longer ubiquitinated nor degraded [78]. This leads to the saturation of KEAP1, and consequently any newly synthesized NRF2 can evade repression and directly accumulate into the nucleus, promoting its target gene expression [79].

While NRF2 can be physiologically activated by an increase of oxidative stress, it can also be exogenously induced by chemical agents [76]. The majority of known NRF2 inducers are electrophiles that modify the cysteine residues present in the thiol-rich domain of KEAP1 [80,81] in a covalent way, either by oxidation or alkylation. These electrophile adducts can inhibit KEAP1 in two different ways, namely: (1) by promoting a conformational change in KEAP1 that will inhibit his binding with NRF2, or (2) by blocking the interaction between KEAP1 and Cul3/Rbx1, inhibiting NRF2 ubiquitination and promoting its sequestration and the subsequent stabilization of the newly synthetized NRF2 [82,83].

In the past years, many natural and synthetic compounds have been described to potentially activate the NRF2 pathway through this mechanism, and these are summarized below.

KEAP1 is the main regulator of NRF2, by promoting its degradation [8]. The NRF2 recognition by the Kelch domain included in KEAP1 is a classic example of a protein–protein interaction as a drug target, which, generally speaking, remains a challenge for drug discovery due to the large surface of the targeted area and the lack of pockets for the interaction with inhibitors. However, in the case of KEAP1, the target surface is about 300–1000 Ų in size and has well-defined pockets with residues capable of stably recognizing small molecules [135]. These topological characteristics, together with the availability of KEAP1 3D structures, allow the development of small molecules that are able to recognize KEAP1 and disrupt its binding with other proteins [136].

The ETGE motif contained in the Neh2 domain of NRF2 adopts a β-hairpin conformation to interact with the KEAP1 Kelch domain-binding pocket, which can be artificially subdivided into five subpockets (Figure 11a) [136]. The P1 and P2 subpockets are characterized by the presence of multiple positively charged arginine residues (Arg415, Arg483, and Arg485), enabling these sites to recognize negatively charged amino acid residues of substrates [137]. In particular, the carboxyl side chains of Glu79 and Glu82 at the ETGE motif of NRF2 (Figure 11b) promote binding by their interaction with the P1 and P2 subpockets. The central P3 subpocket is formed by small polar and nonpolar residues that stabilize the ETGE motif. The external P4 and P5 subpockets are formed by aliphatic chain and tyrosine-rich residues and improve the substrate recognition through hydrophobic interactions and hydrogen bonds [136,137].

The Kelch domain in KEAP1 is able to interact with several additional proteins containing ETGE-like sequences in a similar manner to NRF2 [138]. Some of them, such as SQSTM1/p62 and prothymosin α (ProTα), are also related to the regulation or the NRF2–ARE pathway, or they are implicated in the crosstalk with other hallmarks of neurodegeneration, such as IKK-β (inhibitor of the nuclear factor kappa-B kinase subunit beta) in neuroinflammation [138].

The KEAP1 Kelch domain recognizes p62 through a KEAP1-interacting region (KIR) and this disrupts the KEAP1–NRF2 system, leading to a crosstalk between the NRF2–ARE pathway and p62 [139] (Figure 11c). The mTORC1-dependent phosphorylation of this Ser349 elongates the side chain and adds negative charges to it, improving its recognition by KEAP1 (Figure 11d) [140]. The p62 protein is a cargo receptor for selective autophagy, and its strong interaction with KEAP1 allows it to sequester the latter into early autophagosomes, promoting its degradation through the autophagy–lysosome pathway [141]. Both the p62-mediated disruption of the KEAP1–NRF2 interaction and the p62-dependent degradation of KEAP1 increase the stability of NRF2, leading to neuroprotection [142]. Another ETGE-like substrate that can bind KEAP1 is ProTα, whose ENGE motif mediates its interaction with the KEAP1 Kelch domain (Figure 11e). The crystal structure of the KEAP1–ProTα complex shows a similar network of interactions to the NRF2 ETGE motif [143]. ProTα exhibits a neuroprotective profile against ischemia–reperfusion injury, which can be explained by its capacity to disrupt the NRF2–KEAP1 complex when present in a large excess [144,145].

KEAP1 also interacts with IKKβ, which plays a central role in the control of the pro-inflammatory NF-κB pathway after phosphorylation and several downstream steps [146]. The recognition between KEAP1 and IKKβ promotes the degradation of the latter and prevents the phosphorylation of IKKβ, reducing neuroinflammation [147]. The interplay between the NRF2–ARE and NF-ĸB pathways seems especially interesting in neurodegenerative and cerebrovascular diseases and can be explained in part by the interaction of KEAP1 with NRF2 and IKKβ [147]. Although a crystal structure of the KEAP1–IKKβ complex is unavailable [148], it has been studied through protein–protein docking and molecular dynamic analysis, which show that the ETGE motif of IKKβ adopts a β-turn conformation [148], occupying the substrate-binding pocket in a similar fashion to that of the NRF2 ETGE motif (Figure 11f) [149].

Small peptides based on structural similarity to KEAP1 substrates were the earliest examples of molecules designed to act as KEAP1 inhibitors [150]. However, they have several limitations, including their inability to cross the BBB and their poor stability. Thus, the development of small molecules as KEAP1 binding disruptors remains the main approach, overcoming the bioavailability and stability issues associated to peptides [151]. HTS of a library of more than 300,000 compounds led to tetrahydroisoquinoline 19 (Figure 12) as the first-in-class small-molecule KEAP1–NRF2 PPI inhibitor [151]. Further research efforts on this scaffold led to new structure–activity relationships [152]. The naphthalene sulfonamide derivative 20a (Figure 12) was also identified as a KEAP1–NRF2 inhibitor in a fluorescence anisotropy assay [153]. Structural determination showed an interaction between 20a and the P3–P5 subpockets in the Kelch domain of KEAP1. Based on this finding, compound 20b (Figure 12) was designed [153] with its two carboxyl groups forming multiple hydrogen bonds and electrostatic interactions with the basic Arg residues present in P1 and P2, thus improving the affinity with the KEAP1–NRF2 interface [153]. In the development process, the naphthalene core was replaced by an isoquinoline core to give 20c (Figure 12), improving the solubility, metabolic stability, and reducing the intrinsic toxicity of naphthalene scaffold [154]. Recently, the exploration of bioisosteric replacements of both carboxylic acids afforded the ditetrazole analogue 20d (Figure 12), which maintains the potent PPI inhibition activity and improves the cellular potency thanks to a better pharmacodynamic profile [155]. Another study demonstrated that oxadiazole-urea based compounds 21a-c (NK-252, Figure 12) interact with the KEAP1–Kelch domain [156], with the carboxylic group and the urea–oxadiazole structure interacting via hydrogen bonds and the phenyl moiety by π-stacking with the P1, P2 and P3 subpockets, respectively.

Virtual screening of a 300,000-compound library afforded several hit structures that are able to interact with the KEAP1–NRF2 interface, including compounds 22, 23, and 24 [157]. Compound 25, structurally related to 20, exhibits a very interesting profile as a neuroprotector in a PD animal model by inhibiting the KEAP1–NRF2 interaction [158]. Fragment-based drug discovery was successfully employed in the development of 26, which is a novel phenylpropanoic acid-based KEAP1–NRF2 PPI inhibitor [159]. Crystallographic studies revealed the overlapping between the original fragments and 26 in the binding pocket [159]. This compound showed potency in cell-based assays in the nM range and was shown to activate the NRF2 pathway in vivo [99]. Similarly, simpler five-membered heterocycles such as 1,4-diphenyl-1,2,3-triazoles 27a-b were designed in silico to mimic the Glu residues in NRF2 and were shown to interfere in the KEAP1–NRF2 recognition process and induce the expression of NRF2 target genes in live cells [160]. Isomeric compounds containing the 1-phenyl-1,3,4-triazole scaffold were also described in the patent literature [161]. Compounds 28a and 28b (Figure 12) inhibit the KEAP1–NRF2 interaction with moderate inhibitory potency and induce the expression of NRF2 downstream target genes, leading to neuroprotection in an in vitro model of PD [161]. In particular, compounds 28 are particularly promising as anti-neurodegenerative agents, since in vivo studies using a HD model revealed that they inhibit the neurodegeneration of medium spiny neurons [162].

Research into KEAP1–NRF2 PPI inhibition is rapidly becoming a hot topic in medicinal chemistry, and it is evolving rapidly. In this context, the availability of crystal structures and a broad range of structural analyses related to recognition processes at the KEAP1–NRF2 complex have contributed to the development of several lead compounds by modern drug discovery strategies and the exploration of new chemical sources [152,157,161]. However, these molecules are often polar and possess a relatively large molecular weight, with a limited BBB penetration, leading to poor pharmacodynamic and pharmacokinetic profiles in NDDs [136]. Therefore, PPI inhibitor design with high potential activity and improved physicochemical properties is crucial to develop a new generation of drugs with a better safety profile, compared to KEAP1 covalent inhibitors.

Histone deacetylases (HDACs) remove acetyl groups in histones associated to the KEAP1 promoter region, inducing an increase in KEAP1 transcription [163]. Consequently, HDAC inhibitors, which counteract this effect, improve the redox balance and attenuate neuronal degeneration in some NDDs such as HD [163], ALS [164], and in animal models of stroke [165]. In a model of transient cerebral ischemia, treatment with suberolhydroxamic acid (SAHA) (29, Figure 13), also known as vorinostat, reduced infarct volume by 30–40% [166]. Trichostatin A (TSA) (30, Figure 12) was identified as an inhibitor of KEAP1 expression and leads to an improvement of NRF2 activity and protection against cerebral ischemia [167].

MicroRNAs (miRNAs or miRs) are small noncoding RNAs with 18–25 nucleotides in length that are able to bind to the 3′-untranslated region (UTR) of the target mRNAs [168]. After miRs and target mRNA binding, and depending on the complementarity of both sequences, the target mRNA is degraded or its transcription is suppressed [168]. The miR-7 reduce the expression of KEAP1 after the interaction with its 3′-UTR in the SH-SY5Y cell line [169]. This downregulation of KEAP1 expression leads to NRF2 stabilization and increases the expression of several cytoprotective proteins. The effect of miR-7 appears to be especially interesting in PD models, where the overexpression of miR-7 protects neuronal cells against oxidative stress [169]. It is also relevant that the brain regions more affected in PD, such as substantia nigra and striatum, show an increased level of miR-7 compared with non-affected regions such as the cortex or cerebellum [170]. Furthermore, miR-7 expression downregulation observed in a PD mice model [170] suggests its implication in PD development.

Based on these precedents, KEAP1 expression regulation by epigenetics or miRs can provide an effective strategy to improve neuroprotection through the NRF2–ARE pathway.

GSK-3β is a Ser/Thr kinase involved in glycogen metabolism, cell proliferation, apoptosis, and other functions within the cell [171]. Regarding NDDs, GSK-3β activity has been reported to be upregulated in AD and has been considered to play a key role on tau hyperphosphorylation [172]. Therefore, GSK-3β has been one of the most important targets in drug development toward AD. In 2006, Rojo et al. described the ability of this kinase to modulate the NRF2–ARE response [173], promoting NRF2 phosphorylation and nuclear exclusion mediated by Fyn [174]. GSK-3β phosphorylates Fyn, to increase its activity; in turn, Fyn phosphorylates NRF2 at Tyr568, leading to its extranucleation [175] and degradation.

Additionally, another line of research has demonstrated that GSK-3β phosphorylates the DSGIS [17] degron motif at Neh6 of NRF2 [176], as described above. This post-translational modification increases the affinity of β-TrCP for NRF2 to induce its proteasomal degradation [17], and it is further discussed in next paragraph. Moreover, GSK-3β its activity is regulated by several kinases: the PI3K/Akt pathway downregulates GSK-3β by the phosphorylation of Ser9 [177] to induce its inactivation [178], and p38-MAPK also regulates its activity by phosphorylation at Thr390 [179].

Several GSK-3β inhibitors have reached clinical trials toward different diseases including AD, mild cognitive impairment, autism spectrum disorders, and cancer. Initially, NRF2 activation via GSK-3β inhibition was demonstrated for LiCl, a GSK-3β inhibitor used for the treatment of bipolar disorder and compound TDZD-8 [180], which is able to increased HO-1 expression. Recently, LiCl-induced NRF2 activation has been reported in different in vitro and in vivo models, confirming the regulation of the latter by GSK-3β [181]. TDZD-8-induced NRF2 activation has been further demonstrated in an in vivo model of renal ischemia/reperfusion injury. It reduced oxidative stress and cellular apoptosis via NRF2 activation [182].

A clinically advanced example is tideglusib, a thiadiazolidinone derivative, which was developed as a non-ATP competitive GSK-3β inhibitor for AD treatment, which reached phase II clinical trial (ARGO study) [183], but it failed to produce clinical benefit. Researchers suggested a potential negative effect of the non-linear dose response and propose further dose-finding studies in early AD cases. Tideglusib and pioglitazone, also a thiadiazolidinone derivative, showed neuroprotection against MPPT toxicity mediated by NRF2 induction and phase II response activation. In this model, authors demonstrated that the activation of the antioxidant response was mediated by GSK-3β inhibition [184], although the neuroprotective effect was observed at a higher concentration (5 μM) compared to its GSK-3β IC₅₀ (5 nM).

Another example of the NRF2 induction capacity of GSK-3β inhibitors is compound SB216763, which is a highly potent and selective GSK-3β inhibitor. SB216763 showed an interesting cytoprotective effect against the toxicity induced by doxorubicin in primary podocytes via the pathological activation of GSK-3β and oxidative stress [185]. SB216763 protected podocytes through NRF2 induction and phase II antioxidant response activation, and this response was directly associated to its GSK-3β inhibition capacity [185]. Finally, NRF2 induction via GSK-3β inhibition has been demonstrated for many other GSK-3β inhibitors such us YQ138 [186], obacunone [187], or corilagin [188]; however, in the latter cases, the NRF2 mechanism of action can also be related to a direct interaction with KEAP1 or other activation pathways due to the structural features and multi-target activity of the compounds.

Besides KEAP1, there are other mechanisms that control the stability of NRF2. The most important ones are the β-transducin repeat containing protein (β-TrCP), connected to the PI3K/AKT pathway, and Hrd1 (Figure 14).

(a) PI3K/Akt. The PI3K/Akt pathway regulates metabolism, growth, proliferation, cell survival, and gene transcription. Akt is a Ser/Thr kinase activated by several signals that produce phosphatidylinositol-(3,4,5)-trisphosphate (PIP3) catalyzed by phosphatidylinositol-3-kinase (PI3K) [189]. PIP3 serves as anchor for Akt and its activator kinase, PDK1, which activates Akt by phosphorylation at Thr308. Phosphorylated Akt is further activated by mTORC2 by phosphorylation at Ser473 [189]. A negative regulation of Akt is exerted by protein phosphatase 2A (PP2A), PHLPP1/2, and the tumor suppressor phosphatase and tensin homolog (PTEN) [189].

The PI3K/Akt pathway contributes to Nrf2 induction in a KEAP1-independent mode, as described by Cuadrado et al. [190]. By using the PI3K inhibitor LY294002, they observed a substantial inhibition of carnosol-mediated Nrf2 induction. Further evidence to support the regulation of Nrf2 induction by PI3K activation was obtained by using a biotinylated analogue of CDDO-Im, which is a potent Nrf2 inducer belonging to the cyanoenone triterpenoid family that activates the PI3K/Akt pathway because it reacts covalently with Cys124 inside the active site of PTEN [191], inhibiting its activity. Similarly, 4-hidroxynonenal (an Nrf2 inducer) modifies PTEN Cys71 [192]. The Cys71 and Cys124 residues of PTEN act as redox sensors and, when oxidized, they form a disulfide bridge inactivating the PIP3 3-phosphatase activity of PTEN and increasing the activity of Akt [193]. Therefore, PTEN is considered a negative repressor of Nrf2, acting as a redox sensor inside the cell; once it is inactivated, Nrf2 is activated via the PI3K/Akt pathway activation.

It was demonstrated that PI3K activation leads to Nrf2 induction [173,194] by increasing Akt activity that, in turn, inhibits the activity of GSK-3β by phosphorylation at Ser9 [195] and Ser21 [196]. Furthermore, GSK-3β can also be inhibited by PKC [197], P70S6K, and P90RSK [198], all of which are regulated by PI3K through PDK1 [199].

(b) β-TrCP. The PI3K/AKT pathway activates NRF2 via the inhibition of GSK-3β, which phosphorylates NRF2 [190], allowing its recognition by β-TrCP, which in turn marks NRF2 for ubiquitinylation and subsequent proteasomal degradation [200]. β-TrCP is an F-Box-containing protein of the WD40 subfamily that acts as substrate adaptor/receptor within the Skp1-Cul1-F-box (SCF) ubiquitin ligase SCF^β-TrCP [201] and participates in the ubiquination of IκB, β-catenin, CDC25, and APC [70], among others, besides NRF2. β-TrCP recognizes its substrates by direct binding to phosphorylated destruction motifs; for instance, it binds to phosphorylated NRF2 at its ³⁴³DSGIS³⁴⁷ and ³⁸²DSAPGS³⁸⁷ sequences [202]. Interestingly, this sequence mimics the GSK-3β target motifs, and thus many of the GSK-3β target proteins are known to be processed by β-TrCP for degradation [203].

By using selective GSK-3β inhibitors or siRNAs targeting GSK-3β, a stabilization and nuclear localization of NRF2 was initially observed. Similar results were obtained by using siRNAs toward β-TrCP, demonstrating the regulation of NRF2 by both proteins. The Neh6 domain of NRF2 presents two motifs, DSGIS and DSAPGS, that can be recognized by SCF^β-TrCP once they are phosphorylated [204]. Thus, the GSK-3β/β-TrCP system induces NRF2 ubiqination via β-TrCP/Cul1in1/Rbx1 E3 ligase complex after phosphorylation, being a KEAP1-independent mechanism activated by receptor signal transduction [205]. Considering this regulatory mechanism, a hypothesis of “dual modulation” by KEAP1 and β-TrCP was proposed by Cuadrado et al. [77]. Therefore, β-TrCP is a fine tune regulator of NRF2 for cell signaling, considering the high number of proteins regulated by Nr2 with a high impact on metabolic adaptation [78].

(c) DJ-1. In another connection to NDDs, a direct link between DJ-1, a protein involved in PD, and Nrf2 has been described. DJ-1 inactivation in PD is related to a higher susceptibility of neurons to oxidative stress [206]. Compelling evidence demonstrates that DJ-1 acts as a negative regulator of PTEN [207]. Therefore, DJ-1 downregulation reduces Nrf2 activity due to the increased activity of PTEN. Furthermore, DJ-1 presents several ROS-sensitive Cys residues that can act as oxidative stress sensors. Finally, it has been proposed that DJ-1 can act as a nitric oxide-transferring protein to PTEN Cys residues to inhibit its activity [208].

(a) AHR. The aryl hydrocarbon receptor (AHR) is a ligand-activated transcription factor of the bHLH/PAS family that mediates the effects of many xenobiotics (e.g., polyaromatic hydrocarbons, dioxins) and endogenous compounds [209]. It is widespread in the nervous system and, among others, regulates neuronal functions [209]. Miao et al. showed that the transcription of NRF2 is directly regulated by AHR [210], and in fact, they exhibit both ARE and XRE response elements in their promoter region. Moreover, they share several common antioxidant target genes [211].

(b) P53. The P53 protein, known as the guardian of the genome, is a transcription factor that is activated by DNA damage, regulating the pathways for cell-cycle arrest, DNA repair, senescence, and apoptosis. Indeed, it has been proposed that p53 may contribute to neuronal death processes common to all NDDs [212]. There are several connections between the NRF2 and p53 pathways [213]. First, the activation of p53 induces the expression of p21, which competes with KEAP1 for binding to NRF2 and prevents NRF2 degradation, rising its levels [214]. Moreover, NRF2 induces the expression of NQO1, which hinders the degradation of p53 by the 20S proteasome [215].

(c) Epigenetic regulation. As in the case of KEAP1, the transcription of NRF2 is subject to epigenetic regulation via methylation of the NRF2 promoter in CpG islands, H3 histone methylation, and H4 histone acetylation [216]. There is some evidence showing that DNA-methyltransferase inhibition during brain development increases susceptibility to oxidative DNA damage in the aged brain via the hypomethylation of promoters of AD-associated genes such as the β-amyloid precursor protein [217]. Interestingly, DNA demethylation by treatment with 5-azacytidine (31, Figure 15) was shown to upregulate NRF2 expression in an AD cellular model. [218] In the same context, it has been proven that one of the mechanisms of neuroprotection by sulforaphane (1, Figure 15) in mouse neuroblastoma cells is an upregulated NRF2 expression and promoted NRF2 nuclear translocation by decreasing the methylation levels of the NRF2 promoter gene [219].

NRF2 activity can also be regulated by microRNAs, which repress protein production by interacting with complementary seed sequences in the UTR of target mRNAs. Several miRs are involved in direct NRF2 regulation and are related with neurological disorders associated with oxidative stress [168]. Thus, miR-27a, miR-142-5p, miR-144, and miR-153 were identified as NRF2 suppressors by in silico analysis. Studies in neuronal SH-SY5Y cells showed that the overexpression of these miRs suppresses NRF2 activity by targeting its mRNA, leading to uncontrolled redox homeostasis [220]. Accordingly, schisandrin B (32, Figure 15), a dibenzocyclooctadiene lignan found in the traditional Chinese medicinal herb Schisandra chinensis, has shown neuroprotective effects in 6-OHDA PD models by acting as an inhibitor of miR-34a action [221]. Many miRs are involved in the pathological events after cerebral stroke, and among them, miR 93 has been found to be a negative regulator of Nfr2 expression that contributes significantly to deregulation of the redox balance in ischemia [222]. On the other hand, miR-424 upregulates NRF2, and its overexpression has an antioxidant effect in focal cerebral ischemia/reperfusion in mice that is suppressed by NRF2 knockdown and treatment with SOD inhibitors [223].

Beyond the canonical pathway, the tight control of the phase II antioxidant response inside the cells encompasses a plethora of pathways that regulate NRF2 by different post-translational mechanisms [224,225]. Different positive and negative NRF2 regulation proteins have been described including the previously discussed GSK-3β [173], PI3K/Akt [190], MAPKs [226,227], β-TrCP [190], and PKC [228,229], among others. This fine-tuned control over NRF2 through non-canonical mechanisms opens up the opportunity to design drugs aimed at specific targets.

(a) MAPKs. These enzymes are serine/threonine kinases that modulate crucial biological processes including cell survival [226,230], apoptosis [228,231], and gene expression [227,230]. There are three main MAPK pathways composed by the ERKs: the c-jun N-terminal kinases (JNKs) and the p38 kinases directed to phosphorylation of Ser or Thr residues near to Pro residues [227]. NRF2 regulation by different MAPK pathways has been widely described by the use of selective inhibitors and transfection studies, which demonstrated that ERK signaling increases NRF2 activity [232] and P38 reduces NRF2 signaling [233]. Sun et al. demonstrated that a number of phosphorylation sites at NRF2, including Ser215, Ser408, Ser558, Thr559 and S577, are post-transcriptionally modified by the members of this family of kinases [234].

(b) P38. The P38 MAPK family comprises several isoforms (P38α, β, γ and δ) with specific distribution and substrates. Once activated by phosphorylation, they in turn phosphorylate different proteins and transcription factors to regulate processes including inflammation, cell division, and differentiation or apoptosis [235]. P38 kinases are mainly activated by oxidative stress [236], which is a pathological hallmark of NDDs. P38 has been related to AD, since it is activated by the Aβ peptide and hyperphosphorylated tau through oxidative stress [237]. In this line, the P38 pathway was described to be overactive in AD brains at initial stages [238]. The general consensus points to a negative regulation of NRF2 by P38 MAPK as shown by Keum et al., who described the active p38 isoforms that are able to phosphorylate the NRF2 protein to reduce its nuclear accumulation [239]. P38 MAPK phosphorylates NRF2 at three Ser residues (Ser215, Ser408, and Ser577), improving its interaction with KEAP1 and thus promoting its degradation [9]. Conversely, p38 activation activated the NRF2–ARE pathway [240,241] after the administration of several xenobiotics, which is an observation that was correlated with the inhibitory activity of P38 on GSK-3β [239]. Most probably, the activity of P38 on NRF2 depends of small modifications of the homeostatic equilibrium, and the final response depends on the stimuli, timing, and cellular type.

(c) JNK. The c-jun-NH₂-terminal kinase (JNK) pathway is activated mainly by stress stimuli and is an important member of the MAPK signaling pathways [242], playing important roles in development, cell growth, inflammation, and apoptosis [243]. JNKs are a family of threonine protein kinases that includes three different isoforms (JNK1, JNK2, and JNK3), the latter of which is specifically expressed in the central nervous system [244] and is considered a key target for NDDs. JNK3 has been found to be activated in AD [245], PD [246], ALS, and stroke [247], among other NDDs. Sustained JNK activation has been widely related to pathological conditions promoting cell damage and death and has also been correlated with decreased NRF2 signaling. Recently, a direct crosstalk between JNK and NRF2 has been revealed in the context of protection against liver injury induced by acetaminophen, which is closely associated to JNK activation [248]. Active JNK (p-JNK) increased NRF2 turnover by direct interaction with the Neh1 domain of NRF2. This interaction allows the phosphorylation of Ser residues at Neh6 degron domain of NRF2. Thus, it was demonstrated that p-JNK phosphorylates Ser335 at DSGIS motif of NRF2, codifying its degradation through a KEAP1-independent mechanism [249].

(d) ERKs. The extracellular signal-regulated kinases (ERKs) are a major MAPK subfamily involved in a large number of biological processes. ERK activation has been highly related to cytoprotection, specifically to neuroprotection, linking this kinase pathway to potential targets against NDDs [250]. NRF2 regulation by the ERK pathway has been reported in many different cellular types, and it has been implicated in the mechanism of action of several NRF2 inducers. tBHQ, a known NRF2 inducer [132], increases NRF2 stability inducing the activation of the phase II response in a ERK1/2-dependent mechanism, since its NRF2 induction capacity was abolished in the presence of ERK inhibitors [251]. ERK activation was demonstrated to be necessary for NRF2–ARE signaling in response to several other NRF2 inducers such as pyrrolidine dithiocarbamate [252], caffeic acid [253], sulfuretin [254], gallic acid [254], luteolin [255], and butylated hydroxyanisole [256] in HepG2 cells; lico A in RAW 264.7 cells (26576227), or phenylethanoid glycoside in PC12 cells. Some reports suggest a potential direct crosstalk between NRF2 and ERK1/2 by which activated ERK1/2 might be able to directly phosphorylate NRF2 at Ser40, codifying its nuclear stabilization to increase NRF2–ARE pathway activation [257]. This hypothesis is supported by an increase of p-Ser40-NRF2 protein levels measured by Western blot. However, a direct interaction of ERK1/2 and NRF2 has not been fully proven, and other pathways might be contributing such as PKC, JNK, or Akt [257].

(e) PKC. This protein is part of a family of serine/threonine kinases related to many cellular signaling responses including apoptosis, survival, and differentiation. Although PKC activation is traditionally considered receptor-dependent, there is also a ROS-dependent activation mechanism [258]. The discovery that phorbol esters (PKC activators) induce the expression of NRF2-related proteins led to hypothesize the regulation of NRF2-phase II response by PKC [228]. It was proved that, together with other kinases [229], PKC phosphorylates Ser40 at the NRF2Neh2 domain, disrupting the KEAP1–NRF2 interaction [259]. Ser40 phosphorylation is necessary for its nuclear translocation, but it does not facilitate its nuclear accumulation [260]. A second NRF2 regulation mechanism by PKC is related to the capacity of several PKC isoforms (α, β₁, γ > β₂; not ε) to phosphorylate GSK-3β at inactivation sites [261]. Focusing on NDDs, PKC protein levels and activity was found to be decreased in AD brains [262]. Considering the positive modulation of NRF2 by PKC and other potentially beneficial properties, several PKC activators have been proposed as potential drugs for AD treatment [263].

(f) CK2. Casein kinase II (CK2) is a Ser/Thr kinase that acts on more than 300 substrates, participating in the regulation of many different processes including signal transduction, gene transcription, replication processes, and survival pathways, among others. Among the processes modulated by CK2, it regulates several key regulators against cell stress [264]. CK2 is also related to NRF2 modulation, since its sequence presents at least 13 potential phosphorylation positions targeted by CK2 [265]. By using deletional analyses of NRF2, Neh4 and Neh5 domains were identified as CK2 target regions where this kinase is able to phosphorylate several residues, inducing NRF2 nuclear accumulation and pathway activation [265].

(g) Fyn kinase. Fyn kinase, a member of the Src family, is closely related to the negative regulation of NRF2. Fyn kinase directly phosphorylates NRF2 at Tyr568 to codify its nuclear export [266]. Tyr568 phosphorylation facilitates NRF2 recognition by exportin1, which is the protein related to NRF2 nuclear elimination [266].