Crosstalk between the mTOR and Nrf2/ARE signaling pathways as a target in the improvement of long-term potentiation

Abstract

In recent years, a significant progress was made in understanding molecular mechanisms of long-term memory. Long-term memory formation requires strengthening of neuronal connections (LTP, long-term potentiation) associated with structural rearrangement of neurons. The key role in the synthesis of proteins essential for these rearrangements belong to mTOR (mammalian target of rapamycin) complexes and signaling pathways involved in mTOR regulation. Suppression of mTOR activity may impair synaptic plasticity and long-term memory, while mTOR activation inhibits autophagy, thereby potentiating amyloidosis and development of Alzheimer’s disease (AD) accompanied by irreversible memory loss. Because of this, suppression/inhibition of mTOR might have unpredictable consequences on memory. The Nrf2/ARE signaling pathway affects almost all mitochondrial processes. The activation of this pathway improves memory and exhibits therapeutic effect in AD. In this review, we discuss the crosstalk between the Nrf2/ARE signaling and mTOR in the maintenance of synaptic plasticity. Nrf2 pathway can be activated by pharmacological agents and by changes in mitochondria functioning accompanying various neuronal dysfunctions.

Introduction

Memory is a set of brain functions involved in the accumulation, storage, and reproduction of information and skills. Memory is inherent to animals with developed central nervous system. It allows these animals to adapt their behavior to environmental stimuli and depends on the functional plasticity of brain structures, cells, and systems of neurotransmitters (Giovannini and Lana, 2016). In classical neurobiology, there are two major types of memory: short-term (STM) and long-term (LTM). STM ensures formation of temporary neuronal connections in the frontal (especially, dorsolateral and prefrontal) and parietal cortices (Christophel et al., 2012). STM is limited and stores information for several seconds only. Its capacity is very low and does not exceed 7±2 objects, as suggested by Miller (1956) in his highly cited paper “The Magical Number Seven, Plus or Minus Two”. Recent studies have demonstrated that the number of objects an average human can hold in short-term memory is only 4 to 5 (Cowan, 2001). One of the STM manifestations is working memory (WM), which is as an operant component of STM used for the temporary storage of information during its active processing. In other words, WM is necessary for thinking. WM had often been used synonymously with STM, but now these two forms of memory are believed to be distinct (Aben et al., 2012). Brain regions involved in the WM are frontal and posterior cortices and subcortical structures (Eriksson et al., 2015).

LTM can store much larger volumes of information, presumably, for the entire lifetime. It is supported by stronger and more stable neuronal connections not associated with any particular brain region (Wood, 2011). The process of strengthening these connections that results in the LTM formation is long-term potentiation (LTP), which takes place mostly in the hippocampus (Bliss and Lomo, 1973). Although LTP has been studied for more than half a century, new molecular mechanisms involved in all stages of the LTP regulations are constantly discovered, thus opening new opportunities for the pharmacological modulation of LTP in various pathologies associated with the deterioration of cognitive functions, including dementias, neurodegenerative diseases, and age-related senility.

1. LTP mechanisms

LTP involves different mechanisms depending on the synapse shape. The most studied of them is LTP in synapses formed by Schaffer collaterals and CA1 pyramidal neurons that depends on N-methyl-D-aspartate glutamate receptors (NMDARs) (Baltaci et al., 2018).

There are early and late LTP phases (E-LTP and L-LTP, respectively). E-LTP lasts for 1–2 h. Membrane depolarization and glutamate binding to the postsynaptic receptors results in the activation of NMDARs, leading to the Ca²⁺ entry into the postsynaptic neuron. This entails activation of Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) and protein kinase C (PKC). At the same time, protein kinase M zeta (PKMζ) is activated via the Ca²⁺-independent mechanism. These kinases catalyze phosphorylation reactions required for the E-LTP induction. They phosphorylate AMPARs (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors) causing their activation, and mediate additional incorporation of these receptors in the postsynaptic membrane. After binding glutamate, AMPARs open, leading to flow of Na⁺ ions across the membrane. This causes strong membrane depolarization providing stable NMDAR activation. To note, this stage of E-LTP does not require the synthesis of new proteins and AMPAR delivery to the synapse from its nonsynaptic pool close to the postsynaptic membrane.

L-LTP lasts for more than 3 h and requires both synthesis of new proteins and initiation of mRNA transcription (Kelleher et al., 2004). Constant activation of transcriptional and translational processes is ensured by the MAP kinase cascade that phosphorylates and positively regulates ERK (extracellular signal-regulated kinase). ERK activation marks the transition from E-LTP to L-LTP. ERK phosphorylates transcription factors (e.g., CREB), MAP-2, CaMKII (Garner et al., 1988; Burgin et al., 1990), GluR1 and GluR2, PKMζ, AMPAR, and other proteins (Kelleher et al., 2004). Activation of synthesis of these proteins results in an increase in the surface of the dendritic postsynaptic membrane and its sensitivity to neurotransmitters. These synaptic changes are preserved by the scaffold proteins PSD-95 and Homer1c (Meyer et al., 2014), while translation of proteins involved in the synaptic changes is regulated by protein complexes, the key component of which is mTOR (mammalian target of rapamycin) (Giovannini and Lana, 2016)

2. mTOR and memory

2.1. Major properties of mTOR

The mTOR (mammalian or mechanistic target of rapamycin) is a relatively large (259 kDa) highly conserved serine/threonine protein kinase ubiquitously expressed in all types of cells. This protein was named after rapamycin, a compound isolated from Streptomyces hygroscopicus in 1972 (Sehgal et al., 1975). It was found that rapamycin inhibits mTOR and exhibits the antibacterial, antifungal, immunosuppressive, antitumor, and gerontoprotective activities (Seto, 2012). In 1999, rapamycin was approved by the Federal Drug Administration (FDA) as immunosuppressant for preventing kidney rejection. Two water-soluble rapamycin derivatives, temsirolimus and everolimus, were approved by the FDA in 2007 and 2009, respectively, for the treatment of kidney cancer. In 2011, FDA approved everolimus for the treatment of progressive neuroendocrine pancreatic tumors. In 2012, everolimus was approved for the treatment of patients with subependymal giant cell astrocytoma (Li et al., 2014).

The mTOR functions as a component of two large protein complexes – mTORC1 и mTORC2 (Lipton and Sahin, 2014). Beside mTOR, both complexes contain mLST8 (mammalian lethal with sec13 protein 8) and DEPTOR (DEP domain containing mTOR-interacting protein), a negative mTOR regulator. DEPTOR overexpression results in the suppression of mTORC1 and TORC2 activities (Peterson et al., 2009). Other protein components of mTORC1 and mTORC2 are Tel2 (telomere maintenance 2) and Tti1 (Tel2interacting protein 1) that stabilize these complexes; the knockdown of either Tti1 or Tel2 caused disassembly of mTORC1 and mTORC2 (Kaizuka et al., 2010). mTORC1 includes two specific proteins: Raptor (regulator-associated protein of the mTOR), which is essential for the complex activation, and PRAS40 (proline-rich Akt substrate of 40 kDa) that acts as a negative regulator of mTORC1 (Laplante and Sabatini, 2013). mTORC2 contains three unique proteins essential for the complex functioning: Rictor (rapamycin-insensitive companion of mTOR), mSin1 (mammalian stress-activated map kinase-interacting protein 1), and Protor (protein observed with Rictor) (Lipton and Sahin, 2014).

2.2. Upstream and downstream regulation of mTORC1

The mTORC1 is regulated mainly by the PI3K/Akt signaling pathway. PI3K (phosphoinositide 3-kinase) catalyzes phosphatidylinositol 4,5-bisphosphate (PIP2) conversion into phosphatidylinositol 3,4,5-trisphosphate (PIP3), which is essential for the downstream phosphorylation of PDK1 (phosphoinositide-dependent kinase 1) and Akt (protein kinase B, PKB) (Ersahin et al., 2015). PI3K is activated by GPCRs (G protein-coupled receptors) and RTKs (receptors tyrosine kinases) (Houslay et al., 2016). PI3K can directly activate mTORC2 (Gan et al., 2011), while mTORC1 activation requires inactivation of TSC1/2 (tuberous sclerosis proteins 1 and 2) complexes. TSC2 contains GAP (GTPase activating protein) domain, which inhibits the activity of Rheb. When bound to GTP, Rheb activates mTORC1 (Inoki et al., 2003). TSC1/2 proteins are inactivated by the PI3K/AKT and MAPK signaling pathways. The major activators of TSC1/2 are GSK3β (glycogen synthase kinase 3β) and AMPK (AMP-dependent protein kinase) (Huang and Manning, 2008)

The downstream components of the mTORC1 signaling pathway are proteins involved in the ribosome biogenesis and mRNA translation (Thoreen, 2017). The eIF4F (eukaryotic initiation factor 4F) complex binds 5’-cap of mRNAs and initiates translations. This complex consists of three subunits: DEAD-box RNA helicase eIF4A, cap-binding eIF4E protein, and scaffold eIF4G protein. mTOR activates eIF4E by phosphorylating and inactivating 4E-BP (eIF4E-binding protein), which negatively regulates eIF4E. mTOR activates eIF4B that interacts with eIF4A. The resulting complex provides interaction of the 40S ribosomal subunit with mRNA (Merrick, 2015). However, the effect of mTORC1 in this process is mediated by the S6 kinase (S6K). Besides, S6K regulates activity of the ribosomal protein S6 (RP6S), a component of the 40S ribosomal subunit (Chauvin et al., 2014).

It should be noted that mTOR plays an important role in maintaining mitochondrial homeostasis. A large amount of mTOR is associates with the mitochondrial outer membrane, where it coordinates mitochondrial activity and growth regulation (Desai et al., 2012). This localization of mTOR at the mitochondrial outer membrane can directly increase the sensitivity of mTOR activity to the cell redox status. It was shown that mTOR co-immunoprecipitates with the mitochondrial transmembrane proteins such as VDAC1 (Voltage-dependent anion-selective channel 1) and Bcl-xl (B-cell lymphoma-extra-large) (Ramanathan and Schreiber, 2009). VDAC1 is believed to form an ion channel that facilitates ATP release from mitochondria (Okada et al., 2004). Bcl-xl prevents the cytochrome c release from mitochondria, blocks caspase activation, and acts as an anti-apoptotic protein (Kharbanda et al., 1997). It was also shown that mTORC1 is required to maintain mitochondrial biogenesis in resting muscle (Carter and Hood, 2012). Also, Cunningham et al. (2007) have demonstrated that mTOR controls the activity of YY1 (Yin-Yang 1) and PGC-1α (Peroxisome-proliferator activated receptor gamma coactivator1α) transcriptional factors. PGC-1α is the “master regulator of mitochondrial biogenesis” which regulates the activity of a large number of transcription factors (Fernandez-Marcos and Auwerx, 2011). YY1 is known to regulate the expression of more 700 genes, such as nuclear-encoded mitochondrial genes, ribosomal protein genes, and genes involved in RNA processing (Xi et al., 2007). The importance of mTORC1 in the regulation of mitochondrial biogenesis is supported by the fact that ablation of Raptor and loss of mTOR induced impairs mitochondrial biogenesis (Bentzinger et al., 2008; Risson et al., 2009)

2.3. mTOR role in LTP regulation

As mentioned above, increasing the surface of the postsynaptic membrane and its sensitivity to neurotransmitters requires neuronal growth, which, in turn, requires an active protein synthesis during L-LTP. The involvement of mTORC1 in this process was for the first time demonstrated by Tang et al. (2002), who found that mTORC1 activation in dendrites promotes protein synthesis. It was also found that PI3K-mTORC1 signaling is activated via BDNF (brain-derived neurotrophic factor) binding. BDNF belongs to the family of neurotrophin growth factors interacting with the tyrosine kinase receptor TrkB (Tejeda and Díaz-Guerra, 2017). Schratt et al. (2004) showed that activation of the BDNF-PI3K-mTORC1 pathway regulates translation of CaMKIIα, NMDAR subunit, and Homer2. Later, it was demonstrated that mTORC1 is involved in the regulation via the cap-dependent mRNA translation of multiple proteins participating in L-LTP (Lenz and Avruch, 2005; Tsokas et al., 2005; Liao et al., 2007; Kelly et al., 2007; Jaworski and Sheng, 2006, Swiech et al., 2008, Switon et al., 2017).

The functions of mTORC2 are less studied. It is known that mTORC2 positively regulates the activity of mTORC1 through Akt phosphorylation via mSin (Yang et al., 2015). mTORC2 also regulates cytoskeleton architecture (Jacinto et al., 2004) by causing actin polymerization/depolymerization via control of the PKC and Tiam1 signaling (Angliker and Rüegg, 2013). Actin polymerization and cytoskeleton rearrangements are some of structural changes that take place in the dendrites during LTP (Huang et al., 2013).

Almost any disturbance in the signaling pathway from BDNF to mTORC1 negatively affects LTP. Thus, viral-mediated ablation of BDNF or TrkB impaired LTP not only in C1, but also in C3 neurons (Lin et al., 2018). Activation of PTEN (phosphatase and tensin homolog), which is a negative PI3K regulator, by propofol decreased synaptic plasticity, while PTEN inhibition improved LTP (Wang et al., 2015). Mice deficient by Akt3 exhibited disturbed mTOR activity, microcephaly, and impaired LTP and cognitive functions. (Zhang et al., 2019). Activation of GSK3β (positive regulator of TSC1/2) suppressed LTP (Zhu et al., 2007). Administration of low concentrations of rapamycin prevented L-LTP in mTOR^+/− mice, but not in the wild-type animals (Stoica et al., 2011). Knocking-out Rictor, a component of mTORC2, also impaired LTP (Huang et al., 2013).

These data convincingly demonstrate that mTORC1 and mTORC2 play an essential role in LTP. mTORC1 (Hou et al., 2004; Banko et al., 2006) and mTORC2 (Zhu et al., 2018) also directly regulate long-term depression (LTD). LTD is a process opposite to LTP that involves selective weakening of neural connections established in the synapses during LTP. LTD prevents the synapses from reaching their maximal efficiency, which would block retaining of new information (Massey et al., 2007). mTORC1 plays an important role in the memory retrieval from the LTM, when the earlier obtained information is reactivated from the latent state to a state that permits its expression (Lopez et al., 2015; Pereyra et al., 2018). Therefore, the BDNF-PI3K-Akt-mTORC1 axis and its regulators contribute to the LTM formation and functioning, including LTP, LTD, and memory retrieval.

3. Effect of Nrf2/ARE signaling on memory

3.1. Major characteristics of the Nrf2/ARE signaling pathway

AREs (antioxidant response elements) are fragments of gene promoter sequences that were discovered in the studies of antioxidant properties of polyphenol compounds. The regulatory cis-activating element of ARE is the 5’-A(G)TGAC(T)nnnGCA(G)-3’ sequence (Wasserman and Fahl, 1997). In most cases, this sequence was found in the genes coding for proteins with cytoprotective properties, e.g., antioxidant enzymes, proteins of xenobiotic detoxification phase II, anti-inflammatory factors, and metabolic enzymes and regulators involved in the maintenance of redox homeostasis (Dinkova-Kostova and Abramov, 2015), mitochondrial biogenesis, and mitophagy (Gureev and Popov, 2019).

Six NF-E2 (nuclear factor erythroid-derived) proteins have been identified so far, the most studied of which is Nrf2 (NF-E2 p45-related factor 2). Nrf2 is a short-lived protein (half-lifetime, ~15 min); in the absence of activating factors, Nrf2 is ubiquitinated and undergoes proteasomal degradation. The ubiquitin ligase systems that provide Nrf2 degradation include Keap1 (Kelch-like ECH-associated protein 1) which acts as an adaptor protein ensuring Nrf2 interaction with the E3 ubiquitin ligase complex Cul3 (Cullin 3) and Rbx1 (RING-box protein 1). The latter is required for the Nrf2 interaction with the ubiquitin ligase system (Tkachev et al., 2011). Another negative Nrf2 regulator is GSK3β that phosphorylates Nrf2 at serine and threonine residues, after which this protein is recognized by SCF/β-TrCP (SCF, Skp1/ Cul1/ F-box proteins; β-TrCP, β-transducin repeat containing protein). The complex formed with SCF/βTrCP binds Cul1 (Cullin 1), leading to the generation of the ubiquitin ligase complex and further Keap1-independent degradation of Nrf2 (Rada et al., 2011). Another mechanism for the negative Nrf2 regulation involves E3 ubiquitin ligase HRD1 that degrades misfolded proteins during endoplasmic reticulum stress after their transport to the cytoplasm (Wu et al., 2014).

Nrf2 can interact with AREs in the cell nucleus via several different mechanisms. The canonical activation pathway involves Nrf2 binding to the bZip (basic leucine zipper) transcription factors (most often, small MAF proteins) and CBP coactivator (histone acetyltransferase). This results in the changes in the chromatin structure that provide relaxation of the ARE-containing promoter and facilitate the binding of RNA polymerase. This mechanism of expression regulation is typical for most target genes of Nrf2 (René et al., 2010).

Nrf2 inducers are well studied and classified. They include natural quinones (flavonoids, resveratrol, curcumin, estradiol and its metabolites, dopamine), synthetic quinones (tert-butylhydroquinone, tert-butylhydroxyanisole, probucol), diphenols, phenylenediamines, Michael acceptors (crotonaldehyde), isothiocyanates (sulforaphane), thiocarbamates, 1,2-dithiole-3thiones, oxathiolane oxides, alkene polysulfides, hydroperoxides, arsenic (III) compounds, heavy metal ions (Cd, Co, Cu, Au, Hg, Pb), carotenoids, terpenoids and related compounds, and selenium-containing compounds (especially, diselenides and selenols) (Tkachev et al., 2011). Other strong Nrf2 activators are reactive oxygen species (ROS) and chemical substances promoting their generations (Erlank et al., 2014). Despite the structural diversity, a common feature of all these compounds is their electrophilic nature and the ability to reduce the disulfide bond between Cys273 and Cys288 in Keap1, thus impairing Keap1 interaction with Nrf2 and facilitating Nrf2 translocation to the nucleus, where it binds to AREs of the target genes (Tkachev et al., 2011).

3.2., Effect of Nrf2/ARE on LTP

During the last decade, numerous data have been accumulated highlighting the importance of Nrf2/ARE signaling in the LTM maintenance and LTP formation. Nrf2^−/−mice exhibited LTP impairment, as was revealed by examining the synaptic transmission of granule cells of the dentate gyrus in response to perforant path stimulation. High frequency stimulation increased the field excitatory post-synaptic potential in Nrf2+/+ mice, but these values were strongly diminished in Nrf2−/− mice. According to the authors, this was related to the disruption of neurogenesis in the hippocampus (Robledinos-Antón et al., 2017). Nrf2^−/− neurons exhibited decreased synaptic density and dendritic arborization (Zweig et al., 2019). Compared to the wild-type animals, mice deficient in Nrf2 were more sensitive to the LTP suppression in induced obesity (Tarantini et al., 2018), hypoglycemia, and type I diabetes mellitus (McNeilly et al., 2016). Jo et al. (2001) first demonstrated LTP inhibition by lipopolysaccharide (LPS) injection. Nrf2 activators, such as dimethyl fumarate (FDA-approved preparation for treatment of amyotrophic lateral sclerosis) and naringenin (citrus flavanone), decreased LPS toxicity and promoted LTP (Paraiso et al., 2018; Khajevand-Khazaei et al., 2018). High concentration of D-galactose produced the toxic effects in the hippocampus and suppressed LTP (Krug et al., 1991), while Nrf2 activation improved LTM parameters in mice with the D-galactose-induced memory impairments. Similar effect was demonstrated for quercetin (Dong et al., 2017), Maltol (ginseng extract) (Sha et al., 2019), and a mixture of probiotics containing Lactobacillus paracasei ssp. paracasei BCRC 12188, L. plantarum BCRC 12251, and Streptococcus. thermophiles BCRC13869 (Ho et al., 2019). Nrf2 activation by melatonin prevented LTP impairment during carbon ionic anticancer therapy but caused serious neurodegenerative changes due to the oxidative damage of the hippocampus (Liu et al., 2018). Tannins from the amla (Emblica officinalis) fruit activated Nrf2 in the cortex and hippocampus and prevented LTP impairments caused by fat- and salt-rich diet (Husain et al., 2018). Another factor causing memory impairment is hypobaric hypoxia resulting from the low oxygen partial pressure at high altitudes. It was demonstrated that the synthetic peptide NAP (NAPVSIPQ) improved memory damage caused by hypobaric hypoxia, (Sharma et al., 2011).

In all the above-mentioned studies, the Nrf2/ARE signaling pathway was considered as a mechanism for the regulation of cell antioxidant defense that ameliorates the oxidative stress, thereby promoting LTP. Undoubtfully, oxidative stress negatively affects the functioning of various components of the central nervous system (Patel, 2016). However, there are reasons to believe that the role of Nrf2/ARE is not limited to the antioxidant defense. Habas et al. (2013) demonstrated that an increase in the neuronal activity can activate the Nrf2/ARE signaling. Intracellular Ca²⁺ (which is involved in the LTP activation) can promote the activity of such strong ROS producer as NADPH oxidase (Abramov et al., 2005). ROS, in their turn, can act as signaling molecules activating Nrf2 (Covas et al., 2013).

4. Interaction between Nrf/ARE and PI3K/Akt/mTOR signaling pathways

Nrf/ARE signaling is closely associated with the PI3K/Akt signaling pathway, as it was shown by Sha et al. (2019), who demonstrated that Maltol (a naturally occurring aromatic compound) activates PI3K/Akt and Nrf/ARE, thereby providing protection against oxidative stress. The role of PI3K/Akt in the activation of Nrf2 had also been demonstrated in earlier studies (Wang et al., 2008). Akt phosphorylates and deactivates GSK3β (Rojo et al., 2008). Also, Akt stabilizes p21, which disrupts Nrf2 interaction with Keap1 (Chen et al., 2009).

Also, Nrf2/ARE positively regulates mTOR. This was first demonstrated by Shibata et al. (2010). These authors showed that mutant cell lines with constitutively activated Nrf2 display upregulated expression of the RagD protein belonging to small G-proteins that activate mTORС1. Later, Sasaki et al. (2012) confirmed the involvement of Nrf2 in the regulation of RagD expression and, therefore, mTORC1 activity. Jia et al. (2016), who showed that the downregulation of the Nrf2 expression by shRNA resulted in the suppression of mTOR, indirectly demonstrated the relationship between Nrf2 and mTOR in glioma cells. The authors suggested that the downregulation of Nrf2 caused the ATP deficit leading to the the activation of AMPK, which is a negative mTOR regulator (Jia et al., 2016). The most obvious mechanism of Nrf2 action on mTOR was proposed by Bendavit et al. (2016), who identified the ARE sequence in the promoter of the MTOR gene and demonstrated that Nrf2 can directly regulate expression of this gene. To a certain extent, this regulation also depends on the activity of PI3K/Akt.

Such close involvement of the Nrf2/ARE signaling pathway in the functioning of mTOR suggests that Nrf2 activators may directly improve cognitive functions not only by counteracting the oxidative stress, but also by directly participating in the LTM formation. This opens new prospects for the use of Nrf2 activators in the treatment of numerous disorders associated with impairment of cognitive functions. Although no evidence of direct Nrf2 involvement in the regulation of LTP and other memory-related process has yet been obtained, a growing interest in the Nrf2 activators gives hope that this problem will be comprehensively studied in the nearest future.

5. Nrf/ARE and PI3K/Akt/mTOR signaling pathways as targets in the treatment of Alzheimer’s disease

Alzheimer’s disease (AD) is the most severe form of dementia accompanied by irreversible changes in the synaptic plasticity characterized by pronounced LTP impairment (Di Lorenzo et al., 2019). The mainstream thinking is that the major causes of AD are tauopathy - the hyperphosphorylation of tau protein, that causes its aggregation in the neurofibrillary tangles (Kim et al., 2016), and accumulation of beta-amyloid (Aβ) plaques formed by the cleavage of APP (amyloid-β precursor protein) by BACE1 (β site APP cleaving enzyme 1) and γ-secretase (Selivanova et al., 2018). Accumulation of Aβ is associated with massive oxidative stress and neuroinflammation, as well as LTP impairment (Chen et al., 2000). Tau hyperphosphorylation also causes LTP dysfunction (Shipton et al., 2011).

One of the factors in the AD development is the disturbance in the Nrf2 functioning. It was found that the activity of Nrf2 in the brain of AD patients or animal AD models was significantly decreased, which was manifested as downregulation of expression of its target genes (Kanninen et al., 2009). Immunohistochemical analysis revealed that in the hippocampus of AD patients, Nrf2 was located mainly in the cytoplasm, while in healthy individuals it was evenly distributed between the nucleus and the cytoplasm (Ramsey et al., 2007). These changes in the Nrf2 localization might be one of the factors in the AD development. Transgenic APP/PS1 mice with the mutant APP gene demonstrate elevated levels of Aβ, while vector-mediated delivery of human NFE2L2 gene to the hippocampus of these mice reduced the level of toxic Aβ and enhanced the cognitive functions of the animals (Kanninen et al., 2008). Nrf2 deficit exacerbates the severity of cognitive impairments in the APP/PS1 mice via disturbing p62-dependent autophagy (Joshi et al., 2015).

Nrf2 activators, such as puerarin, CDDO-MA, orientin, , NaHS, carnosic acid, vanillic acid, polysaccharides from Amanita caesarea, and CART (cocaine- and amphetamine-regulated transcript), decreased the level of Aβ and enhanced cognitive parameters in rodents. Benfothiamine and dimethyl fumarate reduced the extent of tau hyperphosphorylation in a Nrf2-dependent manner. Methylene blue, allicin, and Mini-GAGR (cleavage product of low-acyl gellan gum) were able to decrease simultaneously the levels of Aβ and hyperphosphorylated tau (Bahn and Jo, 2019).

So far, none of the Nrf2 activators had been approved by FDA for the treatment of AD. Some of these compounds had been tested in clinical trials, including curcumin (Phase II; NCT00164749) and resveratrol (Phase III; NCT00743743) (Robledinos-Antón et al., 2019); however, no desired effect has been observed, probably, due to their inefficient delivery to the brain (Fão et al., 2019). Clinical trials of Tideglusib (NCT01350362) had been cancelled at phase II because of the lack of clinical effect (Lovestone et al., 2015). At present, DL-3-n-butylphthalide (NCT02711683, recruiting phase) is being clinically tested because of its ability to activate the PI3k/Akt/Nrf2/ARE signaling pathway (Qiu et al., 2018). TRx0237 (methylene blue) undergoes clinical trial (Phase III; NCT03446001), although as a compound inhibiting tau aggregation. However, there are data demonstrating that methylene blue activates the Nrf2/ARE pathway as well (Stack et al., 2014; Gureev et al., 2016).

There is no commonly accepted opinion on the role of mTOR in AD pathogenesis. On one hand, LTP impairment is an important clinical manifestation of AD (Di Lorenzo et al., 2019); hence, pharmacological activation of the PI3K/Akt/mTORC1 is an obvious approach for improving the memory in AD patients. Inhibition of the mTOR pathway associated with the LTP impairment was demonstrated in the hippocampus of Tg2576 mice and mice treated with exogenous Aβ1–42 (Ma et al., 2010). Aβ1–42 suppressed mTOR phosphorylation in the mouse neuroblastoma cells. The level of mTOR phosphorylation was reduced in the cortex of APP/PS1 transgenic mice and lymphocytes of AD patients (Lafay-Chebassier et al., 2005). Aβ25–35 inhibited mTOR activity in the HT22 cells and mouse hippocampus and impaired learning ability in mice (Fan et al., 2015). It was also demonstrated that acute administration of Aβ1–42 promoted Akt phosphorylation; however, chronic exposure to Aβ1–42 resulted in the downregulation of Akt phosphorylation consistent with abnormalities in excitatory neurotransmission (Abbott et al., 2007). Pharmacological modulation of mTOR might also improve cognitive abilities during AD. Thus, tripchorolide (T₄) enhanced cognitive parameters and reduced the content of Aβ in 5XFAD mice via activation of the PI3K/Akt/mTOR signaling (Zeng et al., 2015).

However, there is also an opposite opinion on the role of mTOR in AD pathogenesis. The content of phosphorylated mTOR was elevated in post-mortem tissues of AD patients (Griffin et al., 2005; Li et al., 2005). However, it was also shown that mTORC1, but not mTORC2, was activated in the AD brains (Sun et al., 2014). mTORC1 inhibition with rapamycin downregulated Aβ and prevented cognitive deficit in transgenic mice (Spilman et al., 2010; Caccamo et al., 2010; Ozcelik et al., 2013; Van Skike et al., 2018). Removing one mTOR allele in Tg2576 mice improved cognition and reduced the levels of Aβ (Caccamo et al., 2018).

Most likely, the link between mTOR and AD is autophagy (Jung et al., 2010). Autophagy is one of the essential processes controlling Aβ accumulation. Aβ peptides are generated by the APP cleavage in autophagosomes that later fuse with lysosomes, with the formation of autophago-lysosomes involved in further clearance of Aβ. Autophago-lysosome formation is suppressed in AD, resulting in Aβ accumulation (Nixon, 2007). Therefore, autophagy dysfunction is an important pathological manifestation of AD that leads to the exacerbation of amyloidosis and memory deficit (Nilsson et al., 2013). mTORC1 activation not only promotes cell survival but also suppresses autophagy (Li et al., 2017). mTORC1 inhibits the ULK1/2 (Unc51-like kinase) complex, an important activator of autophagy (Russell et al., 2013). mTORC1 suppression (e.g., by rapamycin) promotes autophagy, which protects hippocampal and cortical neurons from Aβ accumulation (Cai and Yanb, 2013).

Therefore, mTORC1 is a promising target for inhibiting Aβ generation. However, the classical mTORC1 inhibitor rapamycin was not tested in clinical trials for AD treatment. Kaeberlein and Galvan (2019) believe that the reason is its wide availability as a generic drug (temsirolimus and everolimus); hence, there is little incentive for large pharmaceutical companies to invest in its development and testing in this context. Another reason might be that only four compounds (donepezil, galantamine, memantine, rivastigmine) have been approved by FDA for the AD treatment, none of them after year 2002 (except memantine + donepezil combination) (Arti, 2019). It also should be kept in mind that mTORC1 inhibition might not only activate autophagy (required for the Aβ clearance), but also disturb synaptic plasticity by suppressing protein synthesis during L-LTP. Memory impairment by rapamycin had been demonstrated in numerous studies (Stoica et al., 2011; Jobim et al., 2012; Deli et al., 2012; Pereyra et al., 2018). In view of this, it seems more appropriate to fine-tune the mTORC1 activity via activation of other signaling pathways rather than selectively suppress it. The Nrf2/ARE signaling pathway might be a convenient target for such regulation.

Conclusion.

Despite the fact that the process of LTM formation is reasonably well studied, the molecular mechanisms of synaptic changes are not. The activity of mTORC1 is, on one hand, necessary for maintaining synaptic plasticity. On the other hand, it can result in a suppression of autophagy leading to accumulation of Aβ and AD development/progression. Regarding this, the pharmacological regulation of mTORC1 pathway should be a bit more balanced than merely inhibiting or activating. It is our opinion that Nrf2/ARE signaling pathway represents more favorable target for pharmacological intervention because it could provide fine tuning of synaptic plasticity, as well as amyloidosis in AD.

Figure 1. Mechanism of mTORC and Nrf2/ARE signaling crosstalk during long-term potentiation.

Membrane depolarization and glutamate binding to the postsynaptic membrane receptors induce NMDAR activation, which leads to the Ca2+ entry into the postsynaptic neuron. Ca2+ activates CaM and CaMKII, which is necessary for AMPAR phosphorylation. Upon binding to glutamate, AMPAR becomes permeable to Na+, which is necessary for strong depolarization, which is necessary for stable activation of NMDAR. Ca2+ activates MAP kinase pathways (Ras/Raf/MEK1/2-ERK1/2-Rsk), which phosphorylate CREB. CREB induces mRNA transcription for synapse growth. The activation of mTORC1 is due to BDNF, which binds to TrkB and activates PI3K. PI3K catalyzes the turnover of (PI(4,5)P2) into (PI(3,4,5)P3), which is necessary for PDK1 and Akt phosphorylation. PTEN is a negative regulator of PI3K; PI3K can directly regulate the mTORC2 complex, which regulates the actin polymerization process and cytoskeleton rearrangement. mTORC2 in turn, directly phosphorylates Akt. Akt phosphorylates and inactivates TSC1/2, which is a negative regulator of mTORC1 by downregulation of Rheb. mTORС1 activates eIF4E due to phosphorylation and inactivation of 4E-BP, which ensures translation initiation. mTORC1 activates S6K, which activates eIF4B and S6K, which are necessary for ribosomal biogenesis. ERK1/2 and Rsk can phosphorylate and inactivate TSC1/2, thereby activating mTORC1. Nrf2 translocates into nucleus and binds with ARE of mTOR promotor, and upregulates its expression, which leads to enhance of mTORC1 and mTORC2 activity. ERK1/2 can directly phosphorylate and activate Nrf2. Akt phosphorylates and inactivates GSK3β (negative regulator of Nrf2). Akt stabilizes p21, which inactivates Keap1 (negative regulator of Nrf2). ROS inactivate Keap1. NADPH oxidase is a potent ROS source that can be activated by Ca2+, which is translocated into neurons via NMDAR.