Alzheimer’s disease: pathological mechanisms and the beneficial role of melatonin

Abstract

Abstract:  Alzheimer’s disease (AD) is a highly complex neurodegenerative disorder of the aged that has multiple factors which contribute to its etiology in terms of initiation and progression. This review summarizes these diverse aspects of this form of dementia. Several hypotheses, often with overlapping features, have been formulated to explain this debilitating condition. Perhaps the best-known hypothesis to explain AD is that which involves the role of the accumulation of amyloid-β peptide in the brain. Other theories that have been invoked to explain AD and summarized in this review include the cholinergic hypothesis, the role of neuroinflammation, the calcium hypothesis, the insulin resistance hypothesis, and the association of AD with peroxidation of brain lipids. In addition to summarizing each of the theories that have been used to explain the structural neural changes and the pathophysiology of AD, the potential role of melatonin in influencing each of the theoretical processes involved is discussed. Melatonin is an endogenously produced and multifunctioning molecule that could theoretically intervene at any of a number of sites to abate the changes associated with the development of AD. Production of this indoleamine diminishes with increasing age, coincident with the onset of AD. In addition to its potent antioxidant and anti-inflammatory activities, melatonin has a multitude of other functions that could assist in explaining each of the hypotheses summarized above. The intent of this review is to stimulate interest in melatonin as a potentially useful agent in attenuating and/or delaying AD.

Introduction

Alzheimer’s disease (AD) is a devastating disorder affecting around 35 million people worldwide [1]. Ten years ago, there were 4.5 million persons with AD in the US population alone; the number has increased to 5.3 million people in 2010, according to the Alzheimer’s Association [2]. It is estimated there will be 13.2 million people with this neurodegenerative disorder by 2050 [3].

Alzheimer’s disease is a primary, progressive neurological disease, which is of unknown etiology in more than 90% of the cases. Some characteristic neuropathological and neurochemical features lead to irreversible loss of neurons. Owing to the nature of the primarily affected neuronal circuits, the clinical hallmarks of AD are progressive impairment in memory, judgment, decision making, orientation to physical surroundings, and distorted language. This illness is the leading cause of dementia in older people [2].

There currently is no cure for AD. A recent meta-analysis of functional outcomes for commercially available acetylcholinesterase inhibitors and memantine in the treatment of patients with AD, the only FDA-approved drugs for AD, revealed only a modest trend favoring active treatment over placebo [4, 5]. Anti-inflammatory agents may reduce the risk of developing AD [6], but, on the contrary, according to results obtained from an elderly community–based cohort study, anti-inflammatory agents could even be dangerous for cognitive abilities [7]. It is also possible that anti-inflammatory drugs have no influence at all with the exemption of their well-known collateral effects [8, 9]. Vitamin E, estrogens, omega-3 fatty acids, and Ginkgo biloba have been tested in different studies, and they yielded contradictory results. And there is a long list of experimental therapies targeting different protagonists in the pathology of AD, such as tau protein, amyloid-β (Aβ targets: formation, aggregation, or toxicity), Aβ receptors or N-methyl-D-aspartate (NMDA) antagonists, serotonin receptors, loss of acetylcholine neurons, cholesterol, and antiaging drugs (reviewed in [10]; Fig. 1).

Figure 1.  Major targets for Alzheimer’s disease therapy.

Likewise, there is insufficient clinical evidence to support the effectiveness of melatonin by itself in managing the cognitive and noncognitive sequelae of people with dementia [11]. However, there are molecular and physiological bases that are worth analyzing, because melatonin may have an effective influence on several of the above-mentioned AD protagonists and the most prominent hypotheses to explain the cause of this disease, as reviewed below. There is also a growing body of evidence indicating the potential role of melatonin as an effective adjuvant in AD management [12–17]. On the contrary, there is one report indicating essentially negative results after using melatonin in patients with AD [18]. Even worse, there is a single recent publication which claims that melatonin may aggravate this neurodegenerative disorder [19].

The authors of the current review declare no conflict of interest related to this paper or financial relationships with commercial entities. The aim is to put together evidence on melatonin’s role in the best-known hypotheses that currently attempt to explain AD pathogenic mechanisms, starting with the fact that AD-related changes begin at the age when melatonin levels fall significantly. However, it is also clear from the beginning that more than 100 yr after the first clinical report of a case of AD, there is not yet a satisfactory hypothesis or a model capable of explaining or reproducing the pathogenic mechanisms of this devastating disease. Thus, all the proposed treatments for AD are groping for optimal experimental outcomes in regard to obviously incomplete hypotheses. This incompleteness may explain, at least in part, how different models yield widely different results. For example, long-term oral administration of melatonin in an amyloid precursor protein (APP) + PS1 double transgenic mice model of AD protects against cognitive deficits and markers of neurodegeneration [20], while it fails to protect animals expressing the Swedish AD mutant gene (Tg2576 mice) exposed to aluminum [21].

Melatonin levels in AD

Cerebral spinal fluid (CSF) melatonin levels are reportedly significantly decreased in aged individuals with early neuropathological AD-related changes in the temporal cortex [22]. In aged patients, melatonin levels in CSF have been found to be one-half those in young control subjects, but in patients with AD, the CSF melatonin levels are only one-fifth those in young subjects [23]. In fact, it is possible to replicate hippocampal CA1 and CA3 pyramidal neuron loss in rats by merely removing the pineal gland (which lowers melatonin levels) with this effect being reversed by melatonin replacement in the drinking water [24]. Also, constant light exposure, which decreases serum melatonin, is enough to cause Alzheimer-like damage, such as memory deficits, tau hyperphosphorylation at multiple sites, activation of glycogen synthase kinase-3 and protein kinase A, as well as suppression of protein phosphatase-1 and prominent oxidative stress [25].

Melatonin, mechanisms of action

Melatonin is derived from the aminoacid tryptophan in a multistep process involving the synthesis of serotonin, which is subsequently N-acetylated and O-methylated [26]. Melatonin is 5-methoxy-N-acetyltryptamine produced by the functional elements of the pineal gland; once released, it acts both as an endocrine product and as an antioxidant. Several precursors of melatonin including tryptophan and serotonin are reduced by aging, and their reduction may be linked to AD appearance [27, 28] (Fig. 2). Tryptophan deficiency is related to an accelerated degradation attributed, as reviewed below, to homocysteine, a risk factor for dementia and AD [29]. Serotonin deficiency, on the other hand, is linked to severe psychiatric symptoms in AD [28], although serotonin dysfunction may appear long before psychiatric symptoms; these symptoms are associated with altered brain serotonin transporter and glucose metabolism as identified using in vivo molecular imaging [30].

Figure 2.  Melatonin and its precursors, tryptophan and serotonin (*), appear reduced in aging, which is particularly significant in Alzheimer’s disease brain.

Besides the pineal gland, melatonin is also presumably produced in gastrointestinal tract, airway epithelium, pancreas, adrenal glands, thyroid gland, thymus, urogenital tract, placenta, and other organs [31]. Even nonendocrine cells, such as mast cells, natural killer cells, eosinophilic leukocytes, platelets, and endothelial cells, may produce melatonin [32]; this wide synthesis underlines its diverse physiological activities: from the control of biological rhythms [33, 34], metabolism of free radicals [35–39], immune responsiveness [40–42], monitoring of mood and sleep [43–45], cell proliferation and differentiation [46, 47], and control of seasonal reproduction [48, 49]. Importantly, melatonin production declines with age [50–53] because of dysfunction of the sympathetic regulation of pineal melatonin synthesis by the suprachiasmatic nucleus (SCN), a condition probably linked to early AD stages, once that the reactivation of the circadian system using light therapy and melatonin has shown promising positive results [54].

Firmly established as the key mediator controlling circadian rhythms [33, 34, 55], it has been discovered that melatonin, a small, lipophilic molecule, also had the capacity to directly scavenge the hydroxyl radical (˙OH) [56, 57]. Almost immediately, a link between melatonin’s hydroxyl radical-scavenging activity and aging was envisioned [58] as well as realizing that aging and Aβ-induced oxidative stress play a key role in AD as well.

Currently, melatonin is recognized both as a free radical scavenger and as an antioxidant [59]. Thanks to its electron-rich aromatic indole ring, melatonin directly donates an electron to free radicals at a potential of 715 mV and avoids redox recycling (reviewed in [60]), while it scavenges, with varying degrees of efficiency, the hydroxyl radical [56, 60–64], hydrogen peroxide [65], hypochlorous acid [66], singlet oxygen [67], superoxide anion radical [68], nitric oxide [69], and the peroxynitrite anion [70] (Fig. 3). Radicals may also be added in the C3 amide side chain of melatonin, which possesses an N–C=O functional group [71]. The rate constant for the scavenging of the ˙OH by melatonin is calculated to be on the order of 2.7 × 10¹⁰/M/s [72].

Figure 3.  Major oxidant pathways and the role of melatonin as an antioxidant (dashed blue arrows), promoting the activity of antioxidant enzymes. Derived from metabolic activity, particularly from mitochondria in aging, melatonin plays an important role as a free radical scavenger (blue balloons).

The free radical-scavenging capacity of melatonin also extends to its secondary, tertiary, and quaternary metabolites in a free radical-scavenging cascade that prolongs its useful life [73–75]. Thus, the interaction of melatonin with free radicals produces the oxidative pyrrole-ring cleavage, giving N1-acetyl-N2-formyl-5-methoxykynuramine (AFMK), and this substituted kynuramine may donate two electrons at different potentials (456 and 668 mV, respectively) to function as a reductive molecule capable to destroy reactive species and to protect macromolecules against oxidative damage [76]. Thus, via the AFMK pathway, a single melatonin molecule may scavenge up to 10 reactive oxygen and reactive nitrogen species (ROS/RNS) [73]. Further AFMK deformylation by the action of arylamineformamidase or hemoperoxidase enzymes produces N1-acetyl-5-methoxykynuramine (AMK) [77], which, in addition to its ability to react with various oxidizing and nitrosating free radical species, particularly singlet oxygen and nitrogen species, also may destroy carbonate and peroxyl radicals (reviewed in [78]) and function as an antioxidant [79]. AFMK and AMK metabolism may lead to other oxidation products, such as 3-indolinone, cinnolinone, and quinazoline compounds, for which no specific functions have been identified to date [78]. Interestingly, the parallel orientation of β-sheets, such as tau and Aβ filaments, generates channels extending along the length of the filament to which aromatic small molecules such as indolinones can bind via π–π interactions, stacked arrangement of aromatic molecules [80]. Using fluorescence spectroscopy, atomic force microscopy, and electron microscopy to screen 29 indole derivatives, Cohen et al. [81] identified three potent inhibitors of amyloid fibril formation and cytotoxicity, and the indole-3-carbinol was among them. The interaction of melatonin with Aβ will be reviewed below. Additionally, the 3-substituted indolinones have been identified as kinase inhibitors [82], which could be related to the anti-inflammatory actions of melatonin and its metabolites [83].

Melatonin may also prevent abnormal elevation of reactive nitrogen species, stimulate other antioxidant systems, and/or inhibit some pro-oxidant enzymes; these indirect actions of melatonin contribute to its potent antioxidant activity [84–86]. An evaluation in human diabetic skin fibroblasts demonstrated that melatonin increases the activity of superoxide dismutase (SOD), catalase (CAT), and glutathione peroxidase (GPx) and the level of glutathione (GSH) [87] (Fig. 3). Similar results have been obtained in fetal rat brain [88], in experimental brain trauma [89], as well as in cultured dopaminergic cells [90] and, of course, in AD transgenic mouse brain [91, 92]. These observations allow for the conclusion that melatonin exerts an antioxidant action by increasing the mRNA levels of the antioxidant enzymes SOD, CAT, and the GSH system, but it may depend somewhat on the model system investigated.

Finally, melatonin exhibits a lower-affinity binding site to a cytosolic quinine oxidoreductase 2 or QR2, also known as MT3 (reviewed in [93]). This enzyme, as any other quinone, has the ability to transform its substrates into more highly reactive compounds that are able to cause cellular damage. Once melatonin binds to its large active site, the enzyme produces fewer hydrogen bonds and hydrophobic contacts, which diminishes their reactivity [94] (Fig. 4).

Figure 4.  Receptor-mediated or acting directly on its substrates, melatonin exhibits a broad diversity of effects to reduce neurodegenerative changes in the central nervous system. It is a pleiotropic indoleamine, actually. All the protagonists in neurodegenerative diseases express melatonin receptors; when and how cells or their molecular effectors become activated or inhibited according to the expression of their melatonin receptors remains unclear. The scheme shows some of the published observations to date, all related to central nervous system and/or neurodegeneration. Thanks to its ability to transfer electrons, melatonin may repair damaged biomolecules derived from DNA oxidation, such as guanosine. LTP, long-term potentiation; MT, melatonin receptor (G-protein-coupled receptors); CaM, calmodulin; CRT, calreticulin; APP, amyloid b precursor protein; 5-LOX, 5-lipoxygenase; COX-2, cyclooxygenase; iNOS, inducible nitric oxide synthase; PLA2, phospholipase A2.

Melatonin’s best-known receptors, MT1 and MT2, are transmembrane G-protein-coupled heterodimers whose signaling pathways lead to downstream effects on Ca²⁺ channels, Ca²⁺ signaling, and changes in MAP and ERK kinases and/or PI3K/Akt pathways [95] (Fig. 4). This means a broad spectrum of possibilities for melatonin is one factor that gives the indoleamine a pleiotropic nature [96, 97].

Once again, it is worth noting that the pathogenic mechanisms in AD are not well understood, and there are many hypotheses regarding this major cause of dementia. The three major hypotheses as well as their derivatives will be the common thread throughout this review. What has been published regarding melatonin and its potential role in each proposed mechanism will be added herein, where appropriate.

Pathogeny of AD

There is a failure of the intercommunication between neuronal circuits in Alzheimer’s disease resulting from synaptic loss and the destruction of neurons. As a consequence, working memory is not transferred through the hippocampus to long-term memory circuits. The disconnection progresses over time and affects some other functions in addition to memory, so that behavior, executive functioning, judgment, movement coordination, and pattern recognition may become eventually affected.

Three major hypotheses have been primarily explored in an attempt to explain AD: (i) cholinergic hypothesis, (ii) amyloid cascade hypothesis, and (iii) mitochondrial cascade hypothesis. Even though the amyloid cascade is the most extended hypothesis [98], the pathogenic role for Aβ is under debate because of reports showing a poor relationship between Aβ accumulation and cell death in the brain, in addition to other results demonstrating a weak correlation between Aβ and cognitive decline [99]. Furthermore, people with Aβ deposits do not necessarily suffer AD [100]. Even more importantly, there are some published data demonstrating that Aβ may be protective in brain disease [101, 102]. Thus, the pathogenic role of Aβ in AD deserves further scrutiny because all of the hypotheses mentioned include a pathological role of Aβ in AD.

Amyloid-β processing

Although far from conclusive, Aβ is the most studied factor related to pathogenic mechanisms of AD. Aβ is derived from the catalytic cleavage of an integral membrane protein, the APP, a ubiquitously expressed type I transmembrane protein whose primary function is not well known. This is the first obstacle in Alzheimer’s research, actually. It is known that the lack of APP in hippocampal neurons enhances neuritic growth, which influences only the synapse number at an early age but not in adult animals [103]. The conserved APP intracellular domain, genetically uncoupled from APP processing and Aβ pathogenesis, has a key role in survival, proper high-affinity choline transporter (ChT) targeting, and neuromuscular synapse development [104]. ChT is responsible for choline uptake from the synaptic cleft, a rate-limiting step in acetylcholine synthesis [105]. An extensive review [106] tackles in detail the possible role of APP in synaptic transmission and neural plasticity, with its respective implications for learning and memory.

Closely associated with lipid rafts in membranes, composed mostly of sphingolipids and cholesterol, the enzymes in charge of APP cleavage are proteases generating soluble isoforms of membrane proteins, a process first related to the secretion of angiotensin-converting enzyme in the kidney [107]. Thus, the APP chain of 695, 751, or 770 amino acids may suffer consecutive cleavage events. The large extramembranous N-terminal region may undergo proteolysis by the α-secretase that cleaves the molecule between Lys687 and Leu688, releasing a large (105–125 kDa), soluble ectodomain known as sAPPα [108]. This sAPPα carries a portion of the Aβ sequence [109] that normally includes 28 amino acids of the extracellular and 12–15 residues of the membrane-spanning region of APP; thus, the subsequent formation of amyloidogenic peptides may be precluded. This α-secretase may be modulated by metal ions and metalloprotease inhibitors and three related disintegrin and metalloprotease enzymes, ADAM9 [110], ADAM10 [108], and ADAM17 [111], which seem to exert an α-secretase activity. It has been speculated that the activation of these proteases, representing a nonamyloidogenic pathway, may offer a therapeutic method in AD [112, 113]. As a matter of fact, it is remarkable that levels of α-secretase ADAM-10 and sAPPα are reduced in the CSF of patients with AD compared to that of controls [114].

The amyloidogenic pathway is established by the concerted action of two secretases, the β-secretase, which cleaves the APP-N terminus, and the γ-secretase, which cleaves the APP-C terminus in the secondary transmembrane region. β-secretase is an aspartyl protease of 501 amino acids with two aspartic protease active site motifs, which is known as the β-site APP cleaving enzyme I or BACE-1 [115] and considered a prime drug target for lowering cerebral Aβ levels in the treatment for and/or prevention of AD. It is the initiating and rate-limiting enzyme in Aβ generation. BACE-1 activity on APP is related to the accessibility of APP within a lipid raft zone of the membrane [116, 117]. Once APP escapes from processing at the α-site, it is cleaved at the luminal domain, resulting in a 12 COOH-terminal fragment (C99), which remains membrane bound, and the soluble APPβ NH2-terminal fragment (sAPPβ) [118, 119]. Then, C99 is cleaved from the membrane by the γ-secretase, a multisubunit protease complex composed of a presenilin catalytic subunit in addition to nicastrin, the anterior pharynx-defective 1 (APH-1), and the presenilin enhancer 2 (PEN-2) [120, 121]. Binding of cholesterol to C99 appears to favor the amyloidogenic pathway in cells by promoting localization of C99 in lipid rafts [122]. The resultant peptide of 39–43 amino acid residues, Aβ, is delivered to the extracellular milieu where it forms insoluble aggregates and becomes the major component of senile plaques. Aβ_1–40 and Aβ_1–42 are the most common Aβ isoforms. Aberrant Aβ_1–42 accumulation within distal neurites and synapses is directly associated with subcellular pathology and neurotransmitters [123–125], while Aβ_1–40 is the predominant form of the Aβ peptides but less prone to form fibrils [126]. As a consequence of re-internalization from the extracellular space [127, 128] or directly by the cleavage of APP in endosomes generated from the endoplasmic reticulum (ER) or the Golgi apparatus, Aβ peptide accumulation also occurs inside neurons leading to trafficking problems, early axonopathy, synaptic loss, and neuron death [129–131].

What determines which enzyme will gain access to APP, thus determining the course of events? There are clues indicating that cholesterol in lipid rafts directly binds the C-terminal transmembrane domain of APP, and this interaction may be a determinant in favor of the amyloidogenic pathway [122] (Fig. 5). Cholesterol decreases the secretion of soluble amyloid precursor protein (sAPP) by interfering with APP maturation and inhibiting glycosylation in the protein secretory pathway, in such a manner that APP cannot be cleaved by α-secretase [132]. Processing APP at the β-site also requires proper orientation to be accessed by BACE-1 [133], which in turn localizes largely within cholesterol-rich lipid rafts [117, 134]. It is proposed that APP is actually a cholesterol sensor [117]. The AD brain shows significant cholesterol retention and high β- and γ-secretase activities as compared to age-matched nondemented controls, while cholesterol depletion may be associated with reduced cellular cholesterol, β-secretase activity, and Aβ secretion [116].

Figure 5.  Several routes to prevent the formation of Aβ neurotoxic aggregates are used by melatonin. It directly intervenes, affecting the stability of amyloid β-sheets by disrupting Asp⁻ -His+ salt bridges or affecting the synthesis and maturation of APP, where its ability to suppress cAMP activity may have a role, because of the cAMP-responsive regions on the APP promoter gene. However, its indirect actions are significant as well because melatonin may reduce the activity of GSK3 required for the amyloidogenic APP processing, by activating and/or enhancing the activity of PKC, or by inducing Akt. Both PKC and Akt may turn off GSK-3 through phosphorylation. COX-2, related to APP synthesis in astrocytes, is controlled by melatonin and its metabolites. Finally, melatonin has a key role in cholesterol and fatty acid distribution in membranes, as reviewed later. This is important because, as illustrated, amyloidogenic APP processing seems to be favored by cholesterol/sphingomyelin-enriched lipid rafts. APP, amyloid precursor protein; βs, β secretase; γs, γ-secretase; Aβ, amyloid β; MT, melatonin receptors; PLC, phospholipase C; DAG, diacyl-glycerol; PKC, protein kinase C; PI3K, phosphatidylinositol-3-kinase; Akt, a serine/threonine protein kinase; PKA, cAMP-dependent protein kinase.

Another factor influencing the access to APP in cell membranes seems to be insulin, a significant association that links AD to diabetes mellitus. Insulin accelerates APP trafficking from the trans-Golgi network to the plasma membrane [135], while the insulin-degrading enzyme (IDE) degrades not only insulin but also Aβ and the intracellular domain of APP [135, 136]. Thus, insulin reduces intracellular levels of amyloid and increases amyloid secretion in a process that probably involves the activation of the MAPK cascade [137], although it might also use the phosphatidyl inositol 3 kinase (PI3K)-pathway to release sAPP, the nonamyloidogenic secreted form of APP [138]. Importantly, this insulin–PI3K pathway locates and halts glycogen synthase kinase-3 (GSK-3) to promote glucose storage as glycogen, as part of intermediary metabolism. However, GSK-3 plays another role. Its α-isoform may interact directly with presenilins within the γ-secretase complex, and it is required for the amyloidogenic APP processing. This is the reason why GSK-3 has become a target for the treatment for AD [139]. Insulin deficiency in brain leads to enhancement of GSK3α/β activation, increases cerebral amyloidosis, and exacerbates behavioral deficits, as demonstrated in APP/PS1 transgenic mouse model of AD by impairing insulin downstream GSK3 and JNK pathways [140].

Not only fibrillar Aβ but a variety of Aβ oligomers may cause cellular damage. Soluble oligomers, referred to as amorphous aggregates, micelles, protofibrils, prefibrillar aggregates, amyloid β-derived diffusible legends (ADDLs), Aβ*56, globulomers, amylospheroids, toxic soluble Aβ, ‘paranuclei’, and annular protofibrils [141], appear within neuronal processes and synapses rather than within the extracellular space. They are neurotoxic rather than amyloid fibrils found in amyloid plaques [142] and may inhibit critical neuronal functions including long-term potentiation [143], a classic experimental paradigm for memory and synaptic plasticity [143, 144]. Even more, as a consequence of inhibition of the proteasome function, soluble oligomers may cause cell death [124]. Extracellular soluble Aβ species, on the other hand, are deposited around neuronal cell bodies and may interact with the lipid bilayer within dendritic arbors at discrete points, appearing co-localized with the postsynaptic density protein 95 (PSD-95) [145], which is related to synapse stabilization and plasticity [146].

Melatonin’s role on amyloid-β processing

There are some clues indicating an interaction between melatonin and Aβ. By using a thioflavin T (Th T) fluorescence assay, which measures the binding abilities of different compounds with Aβ, it is possible to demonstrate that melatonin directly interacts with Aβ and prevents its aggregation [147]. This fact has been well known since 1997, when it was documented using circular dichroism, electron microscopy, and nuclear magnetic resonance spectroscopy [148]. This phenomenon is not related to the antioxidant properties of melatonin [149] and involves the disruption of the His⁺-Asp⁻ salt bridges in Aβ peptide, which are determinants for the formation and stabilization of β-sheet structures [150]. Thus, 24 hr after the incubation with melatonin, Pappolla et al. [148] showed that original β-sheet content of Aβ was significantly diminished in opposition to the increase in β-sheet content when Aβ was incubated alone (Fig. 5).

A direct interaction between the 5-methoxy group on melatonin and His-13 of Aβ may occur as well. This eventuality may be attributed to the higher binding energies in the 5-methoxyindole group, according to single-point energy calculations [151]. Further investigations by electrospray ionization mass spectrometry (ESI-MS), the hydrophobic nature of Aβ, and melatonin interaction has been unveiled, and the proteolytic assessment suggests that the interaction takes place on the 29–40 Aβ-peptide segment [152]. As compared with some other antiamyloidogenic agents, such as daunomycin or the melatonin analogue 3-indolepropionic acid, melatonin exhibits a moderate degree of inhibition of aggregation, as evaluated by ESI-MS [153].

It is also possible that melatonin could regulate the synthesis and full maturation of APP, as it was demonstrated in melatonin-treated PC12 cells, which responded by decreasing its mRNA encoding β-APP. According to Lahiri and Song [154, 155], melatonin accomplishes this while potentiating the nerve growth factor–mediated differentiation.

Because APP gene promoter contains c-AMP-responsive regions, it is possible that c-AMP signaling pathways may induce APP synthesis, and this eventuality could be a link between neuroinflammation and neurodegeneration, as explored by Lee et al. [156]. They found that prostaglandins produced by brain injury or inflammation increases cAMP formation and stimulates overexpression of APP mRNA and holoprotein in primary cultures of cortical astrocytes. On the other hand, the relationship between melatonin or its metabolites and neuroinflammation relies importantly on its ability to inhibit prostaglandins by interfering with the COX-2/PGE pathway [157], which implies the participation of cAMP as a second messenger. In fact, acting through its membrane receptors, melatonin may block cAMP production, protecting white matter against a neonatal excitotoxic challenge. This neuroprotective effect may be prevented by luzindole, a well-known membrane melatonin receptor antagonist, or by forskolin, an adenylate cyclase activator [158]. Thus, in a receptor-mediated manner and by inhibiting adenylyl cyclase, melatonin may impair cAMP signaling, which is probably involved in the activation of the APP gene promoter. Therefore, melatonin could interfere with APP synthesis (Fig. 5).

Acting through its MT2 receptor, melatonin stimulates phospholipase C (PLC) and, via diacylglycerol (DAG), activates protein kinase C (PKC) [159], which in turn phosphorylates and inactivates GSK-3, whose participation in APP synthesis is key, as mentioned before. However, PKC is also capable to directly promote α-secretase-mediated cleavage of APP favoring the nonamyloidogenic pathway [160]. Thus, by activating PKC, melatonin might impair Aβ overproduction (Fig. 5). Moreover, acting through its membrane receptors, melatonin uses a PI3K-dependent pathway to activate Akt, a serine/threonine protein kinase, which, besides participating in multiple survival pathways, phosphorylates and inactivates GSK-3 [161]. PI3K/Akt is the same pathway employed by insulin and by the insulin growth factor-1 receptor (IGF-1R) to interrupt GSK-3β activity under oxidative conditions [162] (Fig. 5).

Because the JNK pathway also could be involved in GSK-3 activation [140], we speculate that melatonin, which prevents JNK activation under oxidative stress conditions [163], may also employ this mechanism to prevent the activation of GSK-3. It is also feasible that melatonin, in its role as antioxidant, might enhance PKC anti-GSK3 activity by avoiding PKC inactivation. This may occur because PKC is redox sensitive and may be S-glutathiolated and inactivated during oxidative stress in the brain. Oxidative stress is a well-documented phenomenon in Alzheimer’s [164].

Stopping GSK-3 activity could be important not only in interrupting APP synthesis but also to reduce tau hyperphosphorylation, because GSK-3 phosphorylates tau [165, 166]. It is worth remembering that the microtubule-associated protein tau is the other key protagonist in AD pathology, being responsible, once hyperphosphorylated, for paired helical filament (PHF) formation.

An indirect interaction between melatonin and Aβ processing has been also proposed involving the hypoxia-inducible factor-1 (HIF-1), which upregulates BACE-1, facilitating the generation of cytotoxic Aβ peptide [167]. In fact, at a pharmacological dose, melatonin may prevent the generation of Aβ peptides by reducing both the BACE-1 protein and its mRNA, as demonstrated in a rat model employed to evaluate the effect of chronic intermittent hypoxia (CIH) on the Aβ generation in the hippocampus. Being a redox-sensitive transcription factor, HIF-1 is susceptible to melatonin redox modulation [164, 168] and, via this indirect manner, melatonin might prevent the formation of Aβ. HIF-1 has been observed to be abundant in AD microvessels where it regulates proinflammatory gene expression [169]. On the contrary, it is also known that the accumulation of Aβ by using a nonhypoxic mechanism may induce the accumulation and nuclear translocation of HIF-1, which in turn mediates a neuroprotective response, presumably by regulating glucose metabolism [170]. HIF-1 has a half-life of approximately 5 min in normoxic conditions and less than a minute under hypoxic conditions. Thus, the role of melatonin in these HIF-1 dependent mechanisms is currently only a matter of speculation.

Conformational changes in Aβ occur in minutes after addition of melatonin. In fact, the ability of melatonin to induce conformational changes in Aβ has been used to investigate the conformation and topology of Aβ peptides interacting with peptide-tethered planar lipid bilayers [171, 172]. Similarly, it has been also demonstrated that lipid composition of membrane bilayers plays a dominant role in mediating conformational changes and in AD pathogeny, as reviewed below.

Levels of Aβ aggregates in the brain were reduced by melatonin in aging mice [173], and in 8-month-old APP 695 transgenic [174] mice or in APP + PS1 double-transgenic mice. The latter were supplemented with melatonin from 2 to 2.5 months to age 7.5 months [20]. However, in old, amyloid plaque–bearing Tg2576 mice, which started melatonin treatment as late as 14 months of age (5 months later from the onset of the pathology [175]), melatonin failed to reproduce its antiamyloid effects (it seems to even fail to prevent oxidative stress [176]).

The melatonin/Aβ interaction could be an inconvenience according to the ‘bioflocculant hypothesis’. Even though the investigation related to inhibitors of Aβ aggregation as a real promise for many investigators [177], blocking or inhibiting Aβ could be a mistake owing to the soluble forms of this peptide, according to this hypothesis. This is because Aβ could have a primordial function: binding to unwanted solutes in the extracellular fluid, which then precipitates to build deposits or aggregates. Thus, Aβ plaques would be an efficient means of presenting neurotoxins to phagocytes [178].

Extracellular Aβ may suppress synaptic plasticity or inhibit long-term potentiation (LTP) [179]. There is, in fact, an odds ratio homocysteine/AD of 4.5 for histopathologically confirmed AD in a case–control study [180], which also demonstrated that patients with AD and high homocysteine levels showed a more rapid progression over the following 3 yr. Even more so, high homocysteine levels, considered a risk factor for dementia and AD [29], are related to apoptosis [181]. This methionine derivative has the ability to induce proapoptotic caspases (caspase 3 and caspase 9), DNA fragmentation, and the Bcl-2–associated X protein (Bax) while reducing the antiapoptotic Bcl-2 protein; these effects may be inhibited in the presence of melatonin [182].

Cholinergic hypothesis

The cholinergic hypothesis asserts that degeneration of cholinergic neurons in the basal forebrain and the associated loss of cholinergic neurotransmission in the cerebral cortex cause deterioration in cognitive function as seen in patients with AD [183, 184]. This theory was introduced in 1971 when it was demonstrated that cholinergic synapses were modified as a result of learning and that loss of sensitivity to acetylcholine was related to forgetfulness [140]. Years later by damaging cholinergic input to the neocortex or hippocampus from the basal forebrain, it was possible to reproduce a memory impairment as observed in AD [141]. Based on this old hypothesis, increasing the synaptic levels of acetylcholine (Ach) with the use of acetylcholinesterase (AchE) inhibitors has been employed as a treatment and is considered the standard of care for the treatment for mild-to-moderate AD; meanwhile, the search for new AchE inhibitors continues [185]. However, although it remains as a rational approach, nowadays the real efficacy of this treatment is under debate [142].

Both Aβ and oxidative stress may reduce Ach synthesis by reducing choline acetyltransferase activity [186, 187]. However, acetylcholine depletion in the AD brain is also related to free cytosolic ionic calcium and oxidative stress. It is well known that Aβ induces elevations of intracellular free Ca²⁺ by increasing calcium entry through L-type voltage-dependent calcium channels [188], and AchE release is a Ca²⁺-dependent phenomenon [189]. In this manner, Aβ may elevate AChE activity, as demonstrated in P19 cells [190]. Furthermore, oxidative stress, which is a key protagonist in AD, also plays a role in the enhancement of acetylcholinesterase activity induced by Aβ peptide [191].

Melatonin’s role on the cholinergic hypothesis

The Aβ-induced AchE activity may be significantly reduced by melatonin, protecting mice from Aβ-induced acetylcholine degradation [147]. Even in LPS injected mice, which exhibit AchE overactivity, melatonin inhibits this effect as demonstrated in the neocortical and hippocampal regions in vivo [192].

Aβ selectively interacts with the potentially neurotoxic NMDA receptor via a postsynaptic site [193], leading to dysregulation of Ca²⁺, which is explained because an intense stimulation of NMDA-type glutamate receptors results in a sustained elevation of cytosolic free calcium ([Ca²⁺]c) and its consequential dysregulation [194]. The mechanism of AchE control may be related to the stabilization of Ca²⁺ levels, because it seems that the increase in AChE expression around amyloid plaques is a consequence of a disturbance in calcium homeostasis; in fact, intracellular calcium mobilization upregulates AChE expression by modulating promoter activity and mRNA stability and, on the contrary, chelation of intracellular Ca²⁺ may inhibit acetylcholinesterase expression [195]. Thus, by controlling [Ca²⁺]c, it is possible to control AchE activity. Melatonin controls Ca²⁺ influx through different pathways, and by these means, it could control acetylcholinesterase expression as reviewed below. Hence, melatonin is important as an acetylcholinesterase and butyrylcholinesterase inhibitor, especially the cyclic 3-hydroxymelatonin analogues, which exhibit structural similarity to cholinesterase inhibitor drugs [196, 197] (Fig. 6). Melatonin, like insulin and the anticholinesterase drug, donepezil, exhibits an antiamnesic effect in amnesic mice mediated by enhancement of cholinergic activity at the expense of decreasing AChE activity [198].

Figure 6.  Melatonin may have a role as an acetylcholine enhancer by blocking the Ca²⁺-dependent release of anticholinesterase enzyme (red cross #1) or allowing the proper reentry of choline (red cross #2) by avoiding ChT oxidation. It is possible that melatonin restores ChAT activity under oxidative stress conditions (red cross #3) as observed in APP transgenic mice. However, the role of melatonin in cholinergic hypothesis remains to be clarified because oxidative stress along with calcium dysregulation, as observed in chronic cellular stress, is relevant in acetylcholine expression and its metabolism as well as Ach receptor activity (red cross #4). Mit, mitochondria; ER, endoplasmic reticulum; ChAT, choline acetyl transferase; Ach, acetylcholine; ChT, choline transporter; AchE, acetylcholinesterase; AchR, acetylcholine receptor.

Considering that oxidative stress and particularly peroxynitrite (ONOO⁻) overproduction is significant in AD brain, where the latter mediates neurotoxicity of Aβ [199], it is possible that S-nitrosylation of metabolic Ach intermediaries has a significant role. Among the multiple targets of OONO⁻ in AD brain, choline acetyltransferase (ChAT) may be a good candidate. ChAT, whose activity has been shown to decrease in the AD brain [200], may undergo S-nitrosylation followed by lysis and oligomerization, as demonstrated in cholinergic nerve endings and synaptic vesicles from Torpedo marmorata electroneurons [186]. Even though the mechanism is not fully understood, the high-affinity ChT, which provides choline for acetylcholine synthesis in neurons, seems to be regulated by OONO⁻ as well [201]. Thus, we speculate that melatonin may promote choline transport [200, 202] (Fig. 6).

There are some other factors also involving melatonin in the cholinergic hypothesis even though the mechanism is not fully understood. For example, ChAT, which binds acetyl coenzyme A to choline in Ach synthesis, has been found decreased in the frontal cortex and hippocampus of APP 695 transgenic mouse model of AD, and melatonin, chronically administered for 4 months, restored ChAT levels as observed in the transgenic animals [174]. It is possible to find ChAT oxidatively modified by the lipid peroxidation product, 4-hydroxy-2-nonenal (HNE) [203], a diffusible electrophile that covalently binds to proteins via Michael addition to Cys, His, and Lys residues [164, 204, 205]. Thus, owing to its free radical-scavenging activity as well as its indirect antioxidant actions, melatonin may reduce ChAT nitrosylation and/or oxidation [186] (Fig. 6). However, in rats infused intracerebroventricularly with amyloid-beta for 14 days, where ChAT activity was significantly reduced, melatonin was unable to restore the activity of this enzyme [205].

Oxidative stress and neuroinflammation in the pathology of AD

It is well known that the accumulation of Aβ in plaques as well as Aβ oligomers may produce sequential inflammatory/oxidative events and excitotoxicity, causing neurodegeneration and cognitive impairment [206]. In one way or another, all the proposed mechanisms for explaining AD pathogenic mechanisms are connected to oxidative stress and neuroinflammation, widely known hallmarks not only for AD but in general for all neurodegenerative diseases and obviously linked to the amyloid cascade [207–209].

Metabolic signs of oxidative stress in AD are always evident in neocortex and hippocampus, related to alterations in synaptic density. In response to elevated brain peroxide metabolism, AD brains show increased cerebral glucose-6-phosphate dehydrogenase activity [210], which is the first and rate-limiting enzyme of the pentose phosphate pathway, central to maintenance of the cytosolic pool of NADPH and thus the cellular redox balance. Even the brains of preclinical AD individuals, with normal antemortem neuropsychological test scores but abundant AD pathology at autopsy, may exhibit increased levels of the major product of lipid peroxidation, 4-hydroxynonenal, and acrolein, a powerful marker of oxidative damage to protein [211]. Inferior parietal lobule samples from early AD patients compared to age-matched controls have been examined for proteomic identification of nitrated brain proteins that revealed significant alterations in antioxidant defense proteins and energy metabolism enzymes, with all of them being directly or indirectly linked to AD pathology [212].

Aβ neurotoxic properties depend heavily on free radicals. The overproduction of free radicals in the pathogeny of AD may come from the microglial respiratory burst in response to Aβ-induced neuroinflammatory events [213–216]. The microglial respiratory burst in AD may result from (i) the interaction of Aβ with specialized receptors, (ii) the astrocyte/microglia intercommunication, or (iii) detection of damage-associated molecular patterns (DAMPs) through their corresponding receptors, leading to the activation of the phagocytic nicotinamide adenine dinucleotide phosphate (NADPH)-oxidase (PHOX). The activation of the NADPH oxidase probably both in neurons and in glia [217–219] links redox control and neuroinflammatory signaling pathways [220].

Aβ causes microglial proliferation mediated by PHOX, which is demonstrated by a marked translocation of the cytosolic factors p47phox and p67phox to the microglial membrane in brains of patients with AD, and this is correlated with proinflammatory events, such as TNF-α and IL-1β overproduction [221, 222]. The synergy between oxidative and nitrosative stress plus neuroinflammation may increase the overproduction of OONO⁻ by a 1,000,000-fold [223]. PHOX is a multicomponent enzyme system composed of two integral membrane proteins, p22phox and gp91phox, integrated as cytochrome b558, three essential cytosolic components, p47phox, p67phox, p40phox, and the above-mentioned GTPase Rac1, of the Rho family of small G proteins. In general terms, the complex begins its integration when the cytosolic p47phox subunit becomes phosphorylated and transports the total cytosolic components to the docking site where they assemble to the flavocytochrome b558 (reviewed in [224]). GTP-bound Rac coordinates the translocation of the p47phox/p67phox/p40phox complex and its dissociation from GTP permits the subsequent inactivation of the PHOX complex, a crucial step where SOD plays a key role acting as a stabilizer of Rac [225]. Once integrated, PHOX transfers electrons from NADPH to molecular oxygen generating .

Because Aβ induces oxidative stress that is related to mitochondrial damage, a mechanism closely linked to apoptosis is established [226–229]. Reciprocally, oxidative stress may induce intracellular accumulation of Aβ, enhancing the amyloidogenic pathway [226, 230, 231] (Fig. 7).

Figure 7.  Aβ plaques and oligomers are in the middle of a complex set of interactions among astrocytes, microglia, and neurons originating a neuroinflammatory response. This is linked to reactive oxygen species (ROS) and NOS overproduction (gray clouds) culminating in oxidative stress, which in turn feeds back on neuroinflammation. Organelle dysfunction, particularly mitochondria, adds more free radicals and aggravates the situation. Even worse, oxidative stress and Aβ are interdependent phenomena; thus, the more the oxidation, the more the amyloid accumulation. Newly formed Aβ contributes to more neuroinflammation and oxidative stress, closing the vicious cycle. In fact, Aβ can be an oxidant by itself, as shown. During inflammatory and oxidative stress, communication between cellular protagonists is importantly mediated by calcium waves (blue waves) apart from cytokines. Melatonin (green) and its major metabolites AFMK and AMK play key roles by scavenging free radicals directly, while they enhance endogenous antioxidant systems, as shown in Fig. 3. AMK is relevant particularly in mitochondria, where it takes ETC components as electron donors or acceptors. Going further, melatonin and its metabolites have a role in neuroinflammation by regulating both proinflammatory signals and oxidative stress mediators, such as COX2 and iNOS by avoiding NF-κΒ full integration. ctk, cytokines; Aβ, amyloid-beta; AFMK, N1-acetyl-N2-formyl-5-methoxykynuramine; AMK, N1-acetyl-5-methoxykynuramine; 3-OHM, cyclic 3-hydroxymelatonin; ETC, electron transport chain; NF-κB, nuclear factor kappa B; COX-2, cyclooxygenase 2; iNOS, inducible nitric oxide synthase; PGE2, prostaglandin E2.

Additionally, Aβ_1–42 may initiate free radical chain reactions by itself. It has a critical methionine residue at position 35, which is highly hydrophobic and possesses a sulfur atom sensitive to oxidation (:S: → O=S: → O=S=O) [231], or if the lone pair of electrons on the S atom undergoes one-electron oxidation, it produces a positively charged sulfuranyl radical (MetS+) [232] (Fig. 7). In this manner, S–O bonded MetS+ may initiate free radical chain reactions with allylic H atoms on unsaturated acyl chains of lipids making the lipid hydroperoxide and propagating the chain reaction [233].

Aβ can also directly trap molecular oxygen, reducing it to H₂O₂ in the presence of iron (Fenton reaction), as it has been demonstrated by spectrochemistry in AD brain [234]. Fe²⁺ ions are generated via a redox cycling of iron (Fe²⁺↔ Fe³⁺), and in the presence of a metal chelator, such as clioquinol, this Aβ neurotoxicity is reduced [235]. The matter is relevant because significant alterations in Cu, Zn, and Fe have been found in AD brain in those areas showing severe histopathologic alterations [236, 237]. In general, drugs that prevent oxidative stress include antioxidants, modifiers of the enzymes involved in ROS generation and metabolism, metal-chelating agents and agents, such as anti-inflammatory drugs, that can remove the stimulus for ROS generation.

H₂O₂ is a well-known uncoupler of the mitochondrial respiratory activity, producing a concentration-dependent inhibition of state 3 (ADP-stimulated) respiration and reducing substantially the ADP:O ratio [238]. An evaluation of electron transport chain complexes and Krebs cycle enzymes revealed that α-ketoglutarate dehydrogenase, succinate dehydrogenase, and aconitase are susceptible to H₂O₂ inactivation, which is a reversible process [239].

Under normal conditions, excessive ROS are neutralized by the action of endogenous and exogenous antioxidant defense systems. In addition to the above-mentioned oxidant-generating properties, Aβ may bind the peroxidase enzyme, CAT with high affinity, inhibiting H₂O₂ breakdown [240] and thus worsening redox conditions. However, all the antioxidant mechanisms play roles in the AD brain. Thus, the overexpression of superoxide dismutase-2 (SOD-2), which is localized to mitochondria, scavenges hippocampal superoxide and prevents memory deficits in Tg2576 AD mice [241], which carry both mutant APP and presenilin 1 transgenes [242]. In another AD mouse model, 3xTg-AD, there are significant rises in the activities of SOD and GPx, compared with the controls, whereas levels of reduced GSH are significantly decreased with a concomitant rise in oxidized glutathione (GSSG). This set of events implicates a high oxidative state and depletion of proton donors [243]. The 3xTg-AD mouse harbors PS1_M146 V, APP_Swe, and tau_P301L mutations and progressively develops extracellular senile plaques and intracellular neurofibrillary tangles (NFTs) as well as cognitive impairments [244]. Interestingly, even the exogenous antioxidant systems seem to fail in AD, which are apparently not related to under-nourishment because, as demonstrated in 79 patients where the plasma chain-breaking antioxidants α-carotene, β-carotene, lycopene, vitamin A, vitamin C, and vitamin E were measured by HPLC in addition to a total antioxidant capacity assay, a tool for measuring the inhibitory effect of antioxidants [245]; all of the measured parameters were below the normal range.

Microglia activation and neuroinflammatory response

A common factor in AD pathogeny is the overactivation of microglia with the consequent overexpression of proinflammatory cytokines and a significant increase in ROS [246–248]. ROS, in turn, may come from the innate immune response promoted by danger signals [249, 250] or from the damaged mitochondria [251, 252].

Aβ peptides may activate microglia through (i) Toll-like receptors 2 (TLR2), (ii) scavenger receptor (SR), (iii) receptor for advanced glycation end products (RAGE), (iv) a cell surface receptor complex, and (v) TNFR1, whose deletion, as observed in APP23 transgenic mice (APP23/TNFR1^(−/−)), may inhibit Aβ generation and diminishes Aβ plaque formation in the brain [253]. Aβ aggregates as foreign protein particles are recognized by TLRs, and these become important Aβ innate immune receptors, as demonstrated in antisense knockdown of TLR2 or using functional blocking antibodies against TLR2, which may suppress Aβ-induced expression of proinflammatory molecules and integrin markers in microglia [254]. Even TLR4 could also play a role, as demonstrated in mouse models homozygous for a destructive mutation of TLR4; these show significant increases in diffuse and fibrillar Aβ deposits [255]. However, it is not clear whether TLR signaling pathways involve the clearance of Aβ deposits in the brain or they initiate a neuroinflammatory response, responsible for the synaptic impairment observed in AD pathology [256]. Important receptors, such as the class B scavenger receptor CD36 and the LPS-binding molecule CD14, signal through TLR2. CD36 recognizes a variety of ischemic by-products acting as ligands, including oxidized low-density lipoprotein (LDL), long-chain fatty acids, thrombospondin-1, and, again, Aβ. In microglia and in other tissue macrophages, Aβ initiates a CD36-dependent signaling cascade involving the Src kinase family members, Lyn and Fyn, as well as the mitogen-activated protein kinase, p44/42. Aβ also causes the blockade of Src kinases downstream of CD36 and inhibits macrophage inflammatory responses to β-amyloid [257].

Another scavenger receptor, the macrophage receptor with collagenous structure (MARCO) along with the chemotactic G-protein-coupled receptor formyl-peptide-receptor-like-1 (FPRL1) has been documented to be essential in the amyloid β-induced signal transduction in glial cells [258]. Neurons, microglia, and endothelial cells, which surround the senile plaques in the AD brain, express higher levels of RAGE, which may trigger oxidative stress and NF-κB activation [259]. The interaction of Aβ with RAGE may be a direct interaction [216], or it may involve damaged molecular patterns, such as the S100B protein. In primary cortical neurons, the transcription factor Sp1 mediates IL-1β induction by S100B without evidence of a role for NF-κB, whereas in microglia, S100B stimulates NF-κB or AP-1 transcriptional activity and upregulates Cox-2, IL-1β, IL-6, and TNF-α expression through RAGE engagement [260, 261]. Finally, a cell surface receptor complex for fibrillar Aβ, linked to the small GTPase Rac1 and critical in signaling to PHOX, has been described. This molecular complex mediates microglial activation through the stimulation of intracellular tyrosine kinase–based signaling cascades, and it is integrated by the B-class scavenger receptor CD36, the integrin-associated protein/CD47, and the _α6β₁-integrin [262].

NF-κB may be activated from a variety of pathways, from the canonical pathway where the proinflammatory TNF-α, IL-1, and LPS exert their action in addition to DAMPS, to the noncanonical pathway where CD40 and lymphotoxin receptors activate a p52/relB complex. Moreover, there are other atypical pathways, where genotoxic stress, hypoxia, UV light, H₂O₂, or the epidermal growth factor receptor 2, among others, may intervene (reviewed in [263]). The link between NF-κB and neurodegenerative disorders, particularly AD, is an old one [264, 265].

In rat primary cultures of microglial cells and human neutrophils and monocytes, Aβ activates PHOX, and this effect may be potentiated by the proinflammatory stimulus, such as interferon-gamma or TNF-α, but blocked by tyrosine kinase inhibitors [266]. Mediated by PHOX, oligomeric Aβ may induce ROS production, possibly through N-methyl-D-aspartate receptors (NMDAR), and these PHOX-related ROS, in turn, release the prostanoid precursor arachidonic acid through the activation of ERKs, which phosphorylate cytosolic phospholipase A2α [219].

The rate-limiting enzyme, COX-2, can be induced by multiple cellular factors such as growth factors or the proinflammatory cytokines IL-1β and TNF-α in neurons, astrocytes, and microglia. COX-2 in turn regulates PGE2 signaling in neurons [267] and can activate APP transcription in astrocytes [156], as well as glutamate release from astrocytes, which is responsible for excitotoxic damage in AD [193, 268, 269]. PGE2 and COX-2 feedback each other and modulate neuroinflammation, regulating the production of multiple inflammatory molecules.

Melatonin’s role in neuroinflammation and oxidative stress

Importantly, melatonin has a key role in Aβ-induced assembly of PHOX and the subsequent production of ROS, as demonstrated in cultures of microglia incubated in the presence of fibrillar Aβ. According to Zhou et al. [270], melatonin may impair the assembly of PHOX by inhibiting the translocation of p47phox and p67phox subunits of PHOX from the cytosol to the plasma membrane. This becomes feasible owing to blockade of the phosphorylation of p47phox, a PI3K-dependent phenomenon, and consequently impairing the binding of p47phox to gp91phox. This mechanism is related to melatonin’s capacity to inhibit Akt (protein kinase B) activity in microglia, which is the Ser/Thr kinase downstream of PI3K in these cells [270]. It is worth mentioning that the activation of the PI3K/Akt pathway may be mediated by H₂O₂, acting as an intracellular messenger [271]; melatonin is known to directly scavenge H₂O₂ [234, 239] (Fig. 7).

Melatonin directly detoxifies H₂O₂ and produces the biogenic amine AFMK and a potent free radical scavenger, which in turn may suffer deformylation, giving rise AMK.This latter antioxidant and free radical scavenger is particularly relevant in mitochondria [78, 157, 239] (Fig. 7). Melatonin and/or its metabolites function as antioxidants [91, 272], free radical scavengers, and antiapoptosis agents and prevent abnormal nitric oxide (NO) elevation [273] in the cerebral cortex.

Between microglia and astrocytes, a fluid communication exists. Several astrocyte factors released including transforming growth factor β (TGF-β), macrophage colony-stimulating factor, granulocyte/macrophage colony-stimulating factor, IL-10, IL-β, and ApoE modulate microglia activity [274–276]. Glial Ca²⁺ waves can trigger responses in microglial cells, and the calcium waves arise, in vitro, in response to Aβ administration [277]. Extracellular ATP, in its role as a DAMP and as part of the innate immune receptor surveillance behind the Aβ-induced inflammasome activation [278], may also elicit Ca²⁺ waves and activate a microglial inflammatory response [279]. As will be explained below, melatonin also has a role in this process.

Even though the underlying mechanisms and their scopes in neuroinflammation remain to be unveiled, it has been suggested that melatonin could have modulatory effects on ATP-dependent gliotransmission or glial calcium waves derived from different brain regions and species, regulating astroglial function [280]. Studied in the context of rhythmic circadian outputs to pervasive neurobehavioral states, a functional shift in the mode of intercellular communication between junctional coupling and calcium waves in glial cells was found to be induced by melatonin [281]. However, it is well known that melatonin modulates intracellular free Ca²⁺, and by this means, melatonin may protect cells from calcium-dependent pathways of death, such as calpain and caspase-3 in cells undergoing excitotoxicity and oxidative stress, as demonstrated in vitro in rat C6 astroglial cells [282]. By controlling Ca²⁺ influx, melatonin attenuates glutamate-mediated excitotoxicity, which is responsible for NMDAR-mediated damage of neurons. This in turn is one of the postulated Aβ-mechanisms of damage, as mentioned above [219]. In in vitro experiments with a hippocampal cell line challenged with H₂O₂, Aβ, or glutamate, cell death was prevented by the melatonin derivative, AFMK, which is formed by the interaction of melatonin with H₂O₂ or [76] (Fig. 7). Additionally, melatonin may inhibit not only glutamate-induced ion currents but also ion currents from the other ionotropic glutamate receptors, kainate, and AMPA [283]. Also, it is possible that melatonin antagonizes glutamate release, as observed in cortical synaptosomes in old mice and in neurotoxicity induced by KCl [284].

We reported in vivo that melatonin significantly reduced the proinflammatory response, decreasing by nearly 50% the Aβ-induced levels of proinflammatory cytokines IL1-β, IL6, and TNF-α [285]. We speculated that melatonin affected NF-κB DNA binding activity based on a previous report by Natarajan et al. [286]and Chuang et al. [287], who found that NF-κB DNA binding activity was inhibited by melatonin and was lower at night when endogenous melatonin levels are high. Furthermore, 60 min after an intraperitoneal injection of melatonin, a reduction in NF-κB DNA binding activity was replicated. More recently, it has been demonstrated in Aβ-treated brain slices that melatonin reduces NF-κB-induced IL-6 in a concentration-dependent manner [288]. By administering melatonin, it is also possible to reduce Aβ-induced impairment in learning and memory in rats along with a significant decrease in positive glial cells expressing NF-κB-induced IL-1β in addition to C1q in hippocampus [289], both of which are involved in glial activation (Fig. 7). The critical complement component C1q, in turn, may induce the translocation of NF-κB p50p50 homodimers, at least as observed in human monocytes [290], and it is always related to AD pathology usually linked to fibrillar β-amyloid [291].

The pleiotropic transcription factor NF-κB, composed of homo- and heterodimers of five members of the Rel family including NF-κB1 (p50), NF-κB2 (p52), RelA (p65), RelB, and c-Rel (Rel) plays a key role in inflammatory processes but also it is a protagonist in plasticity and neuronal development (reviewed in [292]). Thus, it is complicated to point out that NF-κB inhibition may be a therapeutic target in AD. Nonetheless, by using an immunological assay, it is possible to demonstrate how melatonin prevents the Aβ-induced expression of NF-κB [147]; more specifically, according to Deng et al. [293], melatonin inhibits p52 NF-κB binding as demonstrated by examining the expression of LPS-induced iNOS and COX-2. The latter action involves a promiscuous histone acetyltransferase (HAT), within the nuclear cofactor p300, which is essential for COX-2 transcriptional activation by proinflammatory mediators. By inhibiting p300 HAT activity, melatonin may suppress p52 acetylation, binding, and transactivation [293]. In this manner, it is possible to block the rate-limiting enzyme COX-2 (Fig. 7). This enzyme is induced by multiple cellular factors, such as growth factors or the proinflammatory cytokines IL-1β and TNF-α in neurons, astrocytes, and microglia. COX-2 in turn regulates PGE₂ signaling in neurons [267] and can activate APP transcription in astrocytes [156], as well as glutamate release from astrocytes, which is responsible for excitotoxic damage in AD [268, 269, 293]. PGE₂ and COX-2 feedback on each other and modulate neuroinflammation, regulating the production of multiple inflammatory molecules.

Experiments using transformed lymphatic-derived endothelial cell line demonstrated the ability of melatonin to prevent TNF-α induced phosphorylation of NF-κB p65, although the mechanism is unclear [294]. The administration of melatonin 1 hr after closed head injury also may inhibit the activation of NF-κB during the late phase (8 days), an effect attributed to its prolonged antioxidant effect at the site of injury. However, melatonin did not alter early phase (24 hr after closed head injury), which implies a selective mechanism of neuroprotection [295]. One could expect that such an interference with NF-κB would also affect the role of this transcription factor in plasticity and neurogenesis. Thus, even though not linked to these NF-κB-dependent mechanisms, there are reports indicating a significant depression as well as instability of synaptic transmission in the central nervous system (CNS), although melatonin-dependent fluctuations in synaptic potentials were apparent only when the involved circuit was tetanized [296]. Such depressive effects of melatonin in synaptic transmission would be expected to influence epileptic seizure activity [297]. Nonetheless, other results indicate that, instead of depressing synaptic transmission, melatonin modulates neuronal excitability in the hippocampus, and this modulatory activity depends on its receptors [298, 299]. In fact, melatonin may modulate specific forms of plasticity in hippocampal pyramidal neurons, as demonstrated by electrophysiological methods [300] where neurons exposed to melatonin were found to change their excitability in response to repetitive stimulation, which reveals melatonin as an activity-dependent modulator of subsequent synaptic plasticity (metaplasticity; Fig. 4).

It seems that melatonin may thus regulate neuroinflammation through free radical control and modulation of important proinflammatory transcription factors and their signaling pathways while reducing glutamate excitotoxicity, whether it be by inhibiting glutamate-induced ion currents or by controlling the glutamate delivery. On the other hand, even though functional cytoplasmic membrane melatonin receptors have been described in astrocytes derived from chick brain [301], which could suggest a role for melatonin as a metabolism regulator in astrocytes, these receptors have not been corroborated in human glia.

The nuclear hormone retinoid z receptor/retinoid orphan receptor (RZR/ROR), from the retinoid-related orphan receptors family, are likely associated with melatonin signaling and have been identified in the promoter region of 5-lipoxygenase (5-LOX), a key protagonist in neuroinflammation. By repressing the expression of 5-LOX mRNA in human B lymphocytes, melatonin may reduce the proinflammatory response via nuclear receptor RZR/RORα [302]. Furthermore, the transcriptional activation of RZR/RORα by melatonin is possible even in the nanomolar range [303, 304]. There are no reports confirming this effect of melatonin in the CNS, but 5-LOX is widely expressed in the brain [305] where it has neuromodulatory and neuroendocrine functions and plays an important role in aging and AD, as we will review later. It is worth mentioning that, in addition to AA-derived leukotrienes, 5-LOX also modulates the γ-secretase activity in membranes, favoring Aβ formation [306].

Although there is an isolated report indicating that melatonin is not an important modulator of macrophage and microglia function [307], melatonin’s role controlling the primarily microglia-guided neuroinflammatory response is demonstrated in multiple reports. This is a consequence of its regulation of the NF-κB overexpression [308], the amount of LPS-induced proinflammatory cytokines [192], or prevention of GSK-3β activation and neuroinflammation in response to Aβ, as observed in astrocytes and microglial cells [166]. Meanwhile, a cumulative dose of 10 mg/kg melatonin may attenuate kainic acid-induced neuronal death, lipid peroxidation, and microglial activation, reducing the number of DNA breaks in vivo, as demonstrated in adult male Sprague–Dawley rats [309].

The antioxidant and immunomodulatory effects have inserted melatonin into the two-hit hypothesis [310], which states that although either oxidative stress or abnormalities in mitotic signaling can, independently, serve as initiators in AD, both processes are necessary to propagate the pathological features of the disease.

The mitochondrial cascade hypothesis

According to Swerdlow and Khan [311], the mitochondrial cascade hypothesis asserts that inheritance determines mitochondrial baseline function and durability, and mitochondrial durability influences how mitochondria change with age. Thus, according to this hypothesis, once a threshold of mitochondrial changes is reached, AD histopathology and symptoms ensue. Even though it was formulated in 2004 as a formal hypothesis, some strong evidence linking AD to mitochondrial damage in brain cells had been reported many years earlier. For example, in 1980, by investigating the mechanism for the production of acetyl-CoA used in acetylcholine synthesis, a small, but significant, reduction in the activity of the pyruvate dehydrogenase (besides ATP-citrate lyase and acetoacetyl-CoA thiolase) was found in postmortem brain tissue from cases of AD [312]. In 1985, the activity of the pyruvate dehydrogenase complex was reported to be reduced to about 30% of control values in histologically unaffected occipital cortex as well as in histologically affected frontal cortex of the brains of patients with AD [313]. Likewise, AD as ‘a primary defect in cytochrome oxidase’ was proposed years later [314]. However, it currently is debatable whether Aβ is a downstream product of the mitochondrial functional decline or whether Aβ-induced mitochondrial damage is an extension of the amyloid cascade hypothesis. The amyloid cascade, proposed 20 yr ago, suggested that faulty metabolism of APP was the initiating event in AD pathogenesis, leading subsequently to the aggregation of Aβ, specifically Aβ_1–42 [315, 316]. However, long before the appearance of extracellular Aβ deposits, they are detectable within mitochondria [317].

Beyond hypotheses, some other important features have been found since those early years. It has been demonstrated that Aβ_1–42 uncouple the mitochondrial respiratory chain, and this event plays a key role in pathology of AD [311]. Structurally, Aβ induces swelling of isolated mitochondria [318, 319] and, functionally, decreases ATP synthesis and the activity of various mitochondrial enzymes, as demonstrated in vivo [320] and in vitro in cultured neuronal cells or in Aβ-exposed astrocytes [319–321]. Later, different neurotoxic mechanisms for Aβ were proposed, including disruption of mitochondrial function via binding of the Aβ-binding alcohol dehydrogenase (ABAD) protein [322]or the formation of ion channels allowing calcium uptake, which induces neuritic abnormality in a dose- and time-dependent manner [323], or the opening of the mitochondrial permeability transition pore coupled to inhibition of respiratory complexes [324, 325]. We have found (Rosales-Corral et al., unpublished data) that following the intracerebral injection of fibrillar Aβ, the peptide is revealed both intracellularly and intramitochondrially, deep in the cristae, coinciding with other reports which demonstrate that Aβ progressively accumulates in mitochondria where it is associated with diminished enzymatic activity of the respiratory chain complexes III and IV [317]. Presence of Aβ in mitochondria is related to a reduction in the rate of oxygen consumption by the electron transport chain [317]. The enzymes in charge of importing Aβ to mitochondria have been identified as a complex of translocases, i.e., translocases of the outer membrane (TOM) and the translocase of the inner membrane (TIM) [326] (Fig. 8).

Figure 8.  Both Aβ and abnormal phosphorylated tau play key roles in mitochondrial dysfunction long before amyloid plaques appear. Hyperphosphorylated tau may cause ETC dysfunction by impairing complex I activity, although its major capacity for damaging may come from its ability to interact with ANT from MtPTP, which leads to swelling, mitochondrial dysfunction, and cell death, ultimately. Amyloid-β, whose mitochondrial receptors (ABAD and Hsp) have been identified, causes ETC dysfunction by interrupting the activity of complex III and complex IV, and possibly it may disrupt ion-motive ATPase. Moreover, Aβ impairs energy metabolism by inhibiting directly the activity of the α-ketoglutarate enzyme during the tricarboxylic acid cycle and the pyruvate dehydrogenase before the cycle. Aβ also disrupts Ca²⁺ homeostasis, which overloads mitochondrial matrix and may lead to complex II deficiency, membrane potential loss, ATP reduction, and ROS overproduction in addition to MtPTP disturbance. Aβ-induced oxidative stress has been related to membrane dysfunction, oxidation of ETC components, free radical leak, and MtDNA damage because MtDNA is particularly vulnerable to these events. The way CytC escapes from mitochondria is not completely clear, but once released, it initiates a chain of events leading to apoptosis. Melatonin (represented as red crosses) tends to accumulate inside mitochondria, where (1) it may reduce oxidative stress and its deleterious consequences on MtDNA, proteins, and membrane lipids, such as cardiolipin; (2) it strongly inhibits MtPTP currents and prevents cytochrome c release in a dose-dependent manner; (3) it may recycle electron carriers, such as NADH; (4) it may prevent apoptosis by impairing Cytc release from mitochondria; (5) Ca²⁺ regulation by melatonin may protect mitochondrial functioning. ROS/RNS, reactive oxygen/nitrogen species; Ub, ubiquitin; ABAD, amyloid-β binding alcohol dehydrogenase; Hsp, heat-shock protein; Cytc, cytochrome c; ETC, electron transport chain; MtPTP, mitochondrial permeability transition pore; MtDNA, mitochondrial DNA; APAF-1, apoptosis-activating factor 1.

In addition to direct effects caused by Aβ in mitochondria, there are severe changes attributed to the Aβ-induced oxidative stress. A disturbance of mitochondrial dynamics, a term that includes fission, fusion, movement, and mitochondrial architecture, seems to be implicated in AD pathogeny. It has been demonstrated in human brains of patients with AD where mitochondrial distribution tends to be predominantly perinuclear and fission or fragmentation prevails over fusion, a phenomenon related to a low metabolic capability [327, 328]. These events involve large dynamin-related GTPases, such as the dynamin-related protein (Drp1). Localized to mitochondria, Drp1 is a key factor in mitochondrial division and particularly sensitive to redox regulation [328]. It has been reported that NO overproduced in response to Aβ protein could be responsible for the impairment of Drp1 via S-nitrosylation [329], and this eventuality may lead to an imbalance of fission/fusion in mitochondria, which in turn is correlated with neuronal damage and synaptic loss [330].

Another important feature related to AD pathogeny is mtDNA damage. Perhaps because it is not protected by histones, mtDNA with its 37 genes is more susceptible to oxidative stress-induced deletions and point mutations than nuclear DNA. Even though the consequences of these alterations remain to be clarified [331], mtDNA damage is usually linked to dysfunction (decrement of mitochondrial electron transport chain efficiency) and apoptosis [332, 333].

Also related to Aβ-induced oxidative stress, mitochondrial proteins and lipids become disturbed leading to dysfunction. We have found significant alterations in cholesterol and fatty acids content in mitochondrial membranes following the injection of Aβ (Rosales-Corral et al., unpublished data), associated with functional impairment; as a consequence of increased membrane permeability and changes in lipid polarity owing to oxidative injury, cytochrome c is released from the intermembrane space of mitochondria [334], behaving as an important intermediate in apoptosis and associated with impaired mitochondrial respiration, as observed in brain, platelets, and fibroblasts of patients with AD [335].

An important feature related to the direct interaction between Aβ and cyclophilin D (CypD) has been found [336]. Ca²⁺-associated CypD is part of the mitochondrial permeability transition pore (MtPTP) and translocates from the matrix to the inner membrane where it appears linked to oxidative stress, and by facilitating the opening of mPTP, CypD causes mitochondrial swelling with cellular and synaptic perturbations [337, 338]. The importance of the association of Aβ/CypD is underlined by the fact that a deficiency in CypD may attenuate Aβ-induced mitochondrial oxidative stress, an effect accompanied by improved synaptic function and an improved cognitive performance, as observed in APP transgenic/CypD double-mutant mice [336] (Fig. 8).

As a result of Aβ entrance and mitochondrial damage, energy demands of cells become impaired. We have found functional disorders of F0F1-ATPase in submitochondrial particles obtained from platelets of patients with Alzheimer’s-type dementia [339], but the impairment of ion-motive ATPases in response to Aβ is reproducible in hippocampal neurons in culture [340]. Nonetheless, another report on F0F1-ATPase, searching in isolated mitochondria from platelets and postmortem motor cortex and hippocampus from patients with AD, did not find abnormalities in F0F1-ATPase functioning [341].

Melatonin’s role in mitochondrial hypothesis

Melatonin’s role in this context is mostly related to its ability to scavenge free radicals in addition to its indirect antioxidant properties because it enhances endogenous antioxidant systems in mitochondria [342–348]. It does this basically by maintaining and regenerating the GSH content, which is an important antioxidant mechanism in mitochondria [346]. Similarly, melatonin reduces peroxidation of lipids in mitochondrial membranes and free radical leakage from this organelle. Thus, it is possible to reproduce in vitro mtDNA damage by adding Aβ_1–42 to neurons in culture, but the addition of melatonin prevents significantly mtDNA damage [14, 347]. Added to drinking water and chronically administered, melatonin prevents mitochondrial impairment, maintaining or even increasing ATP production in senescent-prone mice suffering age-dependent mitochondrial dysfunction accompanied by an important oxidative/nitrosative stress [348]. It is worth noting that, additionally, melatonin seems to accumulate in the mitochondria, in such a manner that mitochondrial melatonin levels could be even 100 times higher than melatonin levels in plasma [346]; this claim, however, requires confirmation.

Indole propionamide, similar to melatonin but with a longer half-life has been proven to protect against mitochondrial toxins capable of collapsing the mitochondrial proton potential, causing severe mitochondrial dysfunction and ATP deprivation. Under those circumstances, this recently discovered endogenous indole may act as a recyclable electron and proton carrier, restoring the proton gradient and mitochondrial ATP synthesis [349].

In reference to MtPTP, the antiapoptotic effects of melatonin have been explained by its ability to inhibit the opening of the protein channels responsible for calcium and cytochrome c (cyt c) release from mitochondria. Also, it may function by reducing the loss of the mitochondrial membrane potential in the presence of glucose deprivation-related events [350] (Fig. 8).

Cardiolipin is an important component (20%) of the inner mitochondria membrane. Being particularly susceptible to oxidative stress, cardiolipin becomes implicated in cyt c release during apoptotic events [351], in part because it sensitizes mitochondria to Ca²⁺ mPTP [352]. One of the probable mechanisms of mitochondrial protection by melatonin relates to the prevention of cardiolipin oxidation, avoiding MtPTP opening and restoring Ca²⁺ balance (as reviewed in [353]). However, it is also possible that melatonin directly inhibits MtPTP at single-channel and cellular levels, as demonstrated in patch-clamp recordings on the inner mitochondrial membrane [350].

Three other important proapoptotic factors related to mitochondrial functioning and signaling have been demonstrated to be modulated by melatonin in brain. The executory caspase-3, which is known to be directly linked to cyt c release and widely linked to cell death in AD [99, 320, 354], can be downregulated by melatonin [92]. On the contrary, melatonin may enhance bcl-2 expression as demonstrated in AD transgenic mice [92] and in ischemic brain [355]. Bcl-2 is recognized as an antiapoptotic protective factor, and its relation to Aβ is also widely known, because Aβ may deplete bcl-2 as demonstrated in human primary neuron cultures [356], in microglia [357], or in human neurons from patients with AD [358]. Furthermore, the proapoptotic bcl-2–associated X protein (Bax), which moves from the cytosol to the mitochondria, binds to bcl-2, and promotes cyt c release, is increased in the presence of Aβ in human neurons [356]. However, under a variety of experimental conditions, melatonin has demonstrated its utility in diminishing bax [174, 359–361]. Thus, melatonin modulates mitochondrial pathways to apoptosis.

Mitochondrial damage is linked not only to energy dysmetabolism and leakage of free radicals, which in turn feeds back to induce oxidative stress, but also to increased leakage of Ca²⁺ currents, besides the above-mentioned apoptosis-inducing factors.

Calcium hypothesis

Other hypotheses have been proposed, complementing the amyloid cascade theory. For example, it has been proposed that a calcium-signaling deficit causes accumulation of APP because APP α-processing is a Ca²⁺-dependent process, and this phenomenon provides excessive substrate for β- and γ-secretases, the enzymes responsible for APP processing and Aβ overproduction (Fig. 1) [362]. That Aβ increases calcium uptake has been demonstrated in PC12 cells [363], in human cortical neurons linked to glutamate excitotoxicity [364], in AD brain frontal cortex, and in plasma membrane vesicles from both rat and human brain [365]. This occurs via stimulation of L voltage-sensitive calcium channels, as demonstrated in cultured neurons [366], but Aβ also may increase calcium uptake via potassium channels, the NMDA receptor, the nicotinic receptor, or even by its own calcium-conducting pores (reviewed in [367]). Conversely, calcium accelerates Aβ aggregation, even at physiological concentrations [368], and it is exacerbated by synthetic calcium ionophores [369]. At the same time, Aβ-induced calcium waves feed the neuroinflammatory response, and this increases Aβ aggregation and calcium waves, as mentioned previously [277–279].

During the events leading to oxidative stress and neuroinflammation in AD pathogeny, the glutamate-override of the glutamate/cystine-antiporter system, which controls the levels of glutamate by exchanging cystine in cells for the neurotransmitter glutamate, may lead to an excessive glutamate activity and consequently excessive influx of cations, Ca²⁺ in particular [268, 269]. Glutamate receptors, especially NMDA, are deeply involved in AD pathology [370]by controlling Ca²⁺ influx. Ionic calcium, in turn, activates a number of enzymes, including phospholipases, endonucleases, xanthine oxidase, neuronal nitric oxide synthase, as well as proteases, such as the calcium-dependent cysteine protease, calpains, among others [190, 269, 338, 367, 369]. Thus, the glutamate-induced overestimation of NMDA receptors becomes neurotoxic; this process is common to several neurodegenerative diseases, and it is well known as glutamate excitotoxicity (reviewed in [269]).

Melatonin’s role in calcium hypothesis

Melatonin may reduce NMDA-induced high [Ca²⁺]c levels in addition to its ability to directly inhibit the mitochondrial permeability transition pores, a mechanism linked particularly to oxidative stress, as mentioned before [350]. Several other mechanisms have been postulated to explain the regulation of intracellular Ca²⁺ by melatonin. As an example, it is possible that melatonin acting on its MT2 receptor may inhibit adenylyl cyclase, and with this, it decreases cAMP formation, blocking the cAMP-dependent protein kinase (PKA), which would activate calcium release channels [370–372]. The reader is reminded that pineal and cortical melatonin receptors, MT1 and MT2, are significantly decreased in AD brain [373]. Conversely, it has been known from many years that calcium influx regulates melatonin production in the pineal gland [374] (Fig. 9).

Figure 9.  Some of the most relevant issues in the calcium hypothesis are related to the following: (1) its functional association with γ-and β-secretase activity; thus, APP amyloidogenic processing seems to be a Ca²⁺-dependent process; (2) Aβ, in turn, facilitates Ca²⁺ uptake using a variety of calcium channels; (3) Ca²⁺ accelerates Aβ aggregation, which in turn causes neuroinflammation, oxidative stress, and glutamate release; (4) calcium massive entrance causes excessive accumulation of Ca²⁺ within ER and mitochondria, one dysfunctional and apoptosis-related phenomenon; (5) dysregulated intracellular Ca²⁺ activates a number of enzymes including CaM, which in turn activates CaM-dependent kinases responsible for tau phosphorylation; (6) cytosolic PLA2 is also a Ca-dependent enzyme and, when activated, causes arachidonic acid release and the subsequent activation of the neuroinflammatory pathway via COX-2. Reports on melatonin (red crosses) reveal a role for it: (1) as a free radical scavenger and antioxidant, (2) as an antiamyloidogenic, (3) as an inhibitor of Ca²⁺ influx, (4) as a CaM antagonist, and melatonin also may impair Ca²⁺ leaking from ER, and (5) by regulating calreticulin, it is tempting to speculate that melatonin may delay ER stress. APP, amyloid precursor protein; ROS/RNS, reactive oxygen and reactive nitrogen species; βs, beta secretase; γs, gamma secretase; NMDA-R, n-methyl-daspartate receptor; L-Type R, voltage dependent L-type calcium receptor; AC, adenylyl cyclase; cPLA2, cytoplasmic phospholipase A2; AA, arachidonic acid; ER, endoplasmic reticulum; UPR, unfolded protein response; MIT, mitochondria; CaM, calmodulin.

Melatonin also may inhibit the mobilization of Ca²⁺ from ER as well as Ca²⁺ influx through voltage-sensitive channels [375]. Importantly, melatonin controls the NMDA receptor whose activation comprises multiple regulatory sites controlling Ca²⁺ influx into the cell. On the contrary, in the presence of the Ca²⁺ ionophore A-23187, the inhibitory effect on Ca²⁺ by melatonin is suppressed [376], returning to a glutamate-derived excitatory state. The mechanism involved in NMDA-R control may also imply redox modulation [350, 377]. However, it is also possible that melatonin increases the concentration of the NMDA receptor subunits 2A and 2B, as demonstrated in rat hippocampus, in a dose-dependent manner [378].

Another melatonin mechanism of protection related to calcium influx control is the multifunctional calcium-modulated protein (calmodulin, or CaM) [379], which mediates the calcium requirement for retrograde axonal transport of AchE [380]. Melatonin interaction with CaM is so avid that CaM had been considered a receptor for melatonin [381, 382] (Fig. 4). Even though more recent NMR and molecular dynamic studies suggest a lower affinity [383], it has been demonstrated that melatonin decreases, in a specific manner, the activity and autophosphorylation of CaM kinase II, a key protein kinase involved in neurite maturation [384], and causes neurite enlargement through an increase in tubulin polymerization derived from its CaM antagonism [385, 386]. In this manner, apart from its antioxidant and antiapoptotic effects as well as its anti-AchE actions related to Ca²⁺ and CaM modulatory effects, melatonin may preclude microfilament and microtubule collapse, as demonstrated in N1E-115 cells [372]. Even more, by activating the Ca²⁺-dependent α isoform of PKC [387], melatonin may restore neurite formation, microtubule enlargement, and microfilament organization in microspikes and growth cones in cells damaged with H₂O₂. On the contrary, the PKC inhibitor, bisindolylmaleimide, blocks neurite formation and microfilament reorganization elicited by melatonin. Thus, by regulating Ca²⁺ and CaM, melatonin possesses modulatory actions on cytoskeletal protein phosphorylation, suggesting that it may re-establish neurite formation and the basal levels of phosphorylated tau, at least in N1E-115 cells treated with okadaic acid [388, 389].

There is another Ca²⁺-related melatonin receptor, which may be relevant in AD, i.e., calreticulin (CRT) [390]. This ER resident protein is known to have a role in the folding and assembly of oligomeric membrane proteins. In fact, CRT may function as a molecular chaperone for the Aβ precursor protein, as it has been demonstrated in HEK 293 cells transfected with APP [391]. Calreticulin may initiate ER stress, so it is tempting to speculate that ER stress-induced CRT augments the folding capacity of the ER. By regulating calcium influx and release within the cytoplasm, melatonin regulates not only cellular homeostasis but apoptosis as well (Fig. 9).

It is obvious that in addition to the inhibition of ER and mitochondrial-related mechanisms of apoptosis, melatonin may also prevent spontaneous neuronal apoptosis by acting on downstream signal targets of the protein kinase B or Akt: the forkhead box protein O1 (FOXO-1) and the GSK-3β [392]. Akt is linked to the PI3-K/Akt survival pathway and is required for the expression of long-term potentiation in learning and memory. It is also linked to the insulin receptor substrate (IRS) for the purpose of regulating glucose uptake through a series of phosphorylation events. Moreover, PI3K is phosphorylated upon NMDA receptor-dependent CamKII activity [393], which is also a melatonin target, as mentioned above [386–388]. According to Tajes et al. [392], melatonin increases the activation of prosurvival PI3K/Akt, whereas it inhibits GSK-3β and causes an increase in FOXO-1 phosphorylation. This is a proapoptotic transcription factor involved in insulin activity and in the cellular response to oxidative stress. Finally, GSK-3β, primarily linked to the inactivation of the glycogen synthase, has been also associated with AD and frontotemporal dementia [394] where it is related to microtubule stability because it phosphorylates the microtubule-associated protein tau [395] whose hyperphosphorylation is an early event preceding the appearance of neurofibrillary tangles (NFT) in AD (Fig. 10).

Figure 10.  Aβ and other associated events, such as high cholesterol levels, have insulin resistance in common. Insulin resistance interrupts the neuroprotector PKB/Akt pathway, leading to activation of GSK3 and diminishing IDE levels. Once GSK3 becomes active, it phosphorylates tau while increasing Aβ production, because GSK3 is required for the amyloidogenic processing of APP. IDE, normally stimulated by Akt, reduces its function under insulin resistance conditions. In this manner, insulin degradation is avoided. However, because amyloid is also a target for IDE, Aβ degradation decreases in parallel with the former event, and the neurotoxic peptide tends to accumulate as a consequence and competes for insulin binding to the insulin receptor. Aβ may even inhibit IDE’s activation by interfering with the PI3K/PKB/Akt pathway. There are reports indicating that melatonin (red crosses) (1) may increase IGF-1 levels, (2) may activate the insulin receptor β-subunit tyrosine kinase, which then phosphorylates and activates IRS, (3) may activate directly the PI3K/PKB/Akt pathway through its MT receptors, and (4) may act on CaMKII melatonin inhibits tau phosphorylation, but if CaMKII were responsible for PI3K activation (above left), then it would inhibit the PI3K/PKB/Akt pathway. MT, melatonin receptors; IGF-1, insulin-like growth factor; IR, insulin receptor; PI3K, phosphoinositide 3 kinase; IRS, insulin receptor substrate; PIP, phosphatidylinositol bis- and triphosphate; PDK-1, Phosphoinositide-dependent kinase; CREB, cAMP response element binding; FOXO-1, Forkhead box protein O1; IDE, insulin-degrading enzyme; GSK3, Glycogen synthase kinase; ADDL, Aβ-Derived Diffusible Ligands.

Insulin resistance hypothesis

As can be seen, a number of glucose metabolism related factors begin to appear. GSK-3β is closely controlled by Akt and insulin signal transduction [162], while GSK-3α increases Aβ production [139]. Insulin-activated PI3K-Akt signaling pathway may be interfered with by Aβ [396], and it is well known that an impaired insulin signal transduction is involved in AD pathogeny [397]. On the other hand, the redox-sensitive insulin-degrading enzyme, IDE [398], the most important degradative system for insulin and insulin-like growth factor 1 (IGF-1), is also a major Aβ-degrading enzyme [399]. All of this leads us to another hypothesis, which is provocative because it links AD to diabetes, mitochondrial dysfunction, obesity, and even the cholinergic hypothesis (reviewed in [400]) (Fig. 10).

Also linked to Aβ aggregates and oxidative stress, this insulin resistance hypothesis involves IGF-1 and maintains that a reduced input of this hormone, by causing insulin resistance in the brain, may be the primary pathogenic event in late-onset AD [401, 402]. In fact, many years before the protective effect of IGF-1 in hippocampal neurons in vitro subjected to Aβ toxicity was described [403], epidemiological studies revealed a significant association between diabetes mellitus and AD [165, 404, 405]. Later, it was demonstrated that insulin deficiency and diabetes mellitus exacerbate cerebral amyloidosis and behavioral deficits in transgenic mice [406].

Insulin-like growth factor 1 is a natural molecule similar to insulin, which controls insulin action and is responsible for promoting growth effects, such as increased nitrogen retention and cell proliferation. Beyond glucose homeostasis, insulin is involved in cellular transport of Aβ, and a lack of an adequate insulin input has been related to intracellular Aβ accumulation [135]. In fact, by feeding Tg2576 mice with an insulin resistance-inducing diet, it is possible to promote Aβ aggregates, to increase γ-secretase activities, and to decrease IDE activity [406]. Aβ peptides, particularly ADDLs, compete for insulin binding to the insulin receptor, making the cell insulin resistant [407], very similar to IGF-1 receptor [408], and causing IGF-1/insulin dysfunction; this in turn will affect amyloid trafficking via the inappropriate MAPk and PI3K/Akt activation [401]. The PI3K/Akt pathway also allows for IGF-1 to inhibit GSK-3β [409], a kinase involved in the phosphorylation of tau, as mentioned before, which is another hallmark of AD pathology.

Melatonin’s role in the insulin resistance hypothesis

Male Syrian hamsters receiving melatonin for 10 wk show a significant increase in serum levels of IGF-1 [410], while the protective effect of melatonin on hepatocytes of CCl4-exposed rats [411]has been attributed to increased IGF-1 levels [412]. Evaluated in the context of hypoxia-induced periventricular white matter injury, melatonin reduces brain damage by enhancing IGF-1 activity, while it diminishes glutamate release and the proinflammatory response induced by hypoxia [413]. On the other hand, 6 months after the administration on daily basis of 2 mg of melatonin in elderly women, IGF-I was found to be slightly but significantly increased [414].

An explanation as to how melatonin raises IGF-1 levels is related to melatonin-induced sleep improvement [415]. Sleep disturbances, as observed in obesity or in sleep apnea syndrome, cause alterations in the growth hormone/IGF-1 system [416]. Sleep restriction, in fact, is related to increased insulin resistance, as demonstrated both in obese subjects and in those with AD.

By acting on its receptors, which are present also in the hippocampus [298, 417], melatonin may induce a rapid tyrosine phosphorylation and activation of the insulin receptor β-subunit tyrosine kinase (IR) and downstream Akt serine phosphorylation [418]. Thus, by using the PI3K-Akt signaling pathway, melatonin seems to activate Akt (although other receptors acting on the same pathway could also be involved [161]). The activation of Akt may have at least four consequences: (i) antiapoptotic activity, especially by reducing p53 transcriptional activity, which is particularly true in the hippocampus [419], in addition to the phosphorylative inactivation of proapoptotic factors such as FOXO-1 [392] and the Bcl-2-associated death promoter (BAD) [420]; (ii) inhibition of tau hyperphosphorylation and the consequent paired helical filament formation by inhibiting the tau kinase GSK-3β [392, 409]; (iii) activation in astrocytes of neurotrophic factors such as cAMP response element binding (CREB) and the glial cell-derived neurotrophic factor (GDNF), while the neuroinflammatory control and neuronal plasticity are approached through the astrocytic and neuronal overexpression of NF-κB; and (iv) the reactivation of IGF-1 could reduce insulin resistance [397, 421], ameliorate glucose transport [422], and additionally reduce Aβ accumulation from the cerebral parenchyma [401, 408, 409, 423] (Fig. 10).

Here, it is necessary to look at melatonin’s role in Akt signaling. In isolated microglial cells exposed to Aβ, melatonin has shown inhibitory, dose-dependent effects on the activity of Akt. This impairs the NADPH oxidase assembly [270] as we have seen before at the purpose of AD neuroinflammatory response. No apparent phospho-GSK-3 was detected in those experiments. On the contrary, in isolated astrocytes, melatonin may induce Akt, and by this means, it may inhibit GSK-3β [161]. These apparently contradictory effects may be related to MT1/MT2 dimerization as a mechanism determining the receptor-mediated biological effects of melatonin [424]. As an example, the ERK activity is increased by melatonin in mouse neuroblastoma cells that express only MT1 receptors [425], whereas in human umbilical vein endothelial cells expressing both MT1 and MT2 receptors, ERK is inhibited by melatonin [426]. However, differences related to types of melatonin receptors among astrocytes, microglia, and neurons have not been established yet, except in injured white matter of neonatal rats where MT1 and MT2 seem to be strongly expressed in both astrocytes and microglial cells and to a lesser extent in neurons and immature oligodendrocyte [427]. In the AD brain, an increased MT1 immunoreactivity in pyramidal hippocampal neurons has been documented [417], but in this case, differences between astrocytes, microglia, and neurons in reference to their melatonin receptors were not established (Fig. 4). Astrocytes express MT1 and MT2 receptors, and both seem to be necessary to activate the Akt/PI3k signaling pathway [161]. In microglial cells, what or how many melatonin receptors would be necessary to activate this same pathway, or even if they are necessary, remains to be clarified. By resolving that question, we might explain some differences between neurons, microglia, and astrocytes in response to melatonin. This is relevant in relation to insulin because melatonin may influence insulin secretion [428, 429].

Experimental evidence demonstrates that melatonin improves glucose tolerance and increases insulin receptors in muscle and the expression of GLUT-4, a glucose transporter, in addition to glucose clearance from the blood [430, 431]. A final explanation describes these phenomena as protective actions by melatonin to prevent hypoglycemia during winter dormancy in animals [432]. In the AD brain, where glucose uptake and metabolism are impaired even before the appearance of neuronal degeneration [433], melatonin effects might contribute to delay the progression of the disease, as has been experimentally demonstrated in Aβ-injected mice [147] as well as in high-fat-diet-fed insulin-resistant mice [434] or in vitro by using in C2C12 murine skeletal muscle cells where it was also shown that the glucose transport amelioration by melatonin may occur via insulin receptor substrate-1/PI3-kinase pathway [435].

Feeding rabbits a cholesterol-enriched diet also increases Aβ levels in the hippocampus. This is an intriguing phenomenon, because cholesterol metabolism in the brain has been widely known to be independent of cholesterol metabolism in the body. Even more so, cholesterol in the blood does not pass the blood brain barrier [436]. How hypercholesterolemia increases Aβ levels in the brain may be explained by the oxidized cholesterol metabolite 27-hydroxycholesterol, which, acting on IGF-1 signaling, activates Akt. This survival promoter may decrease IDE, while it increases GSK-3 (α and β), and both phenomena appear associated with elevated Aβ levels, as observed in organotypic hippocampal slices [423] (Fig. 10).

The lipid connection

The first question is how cholesterol in the blood increases cholesterol in brain and how cholesterol in membranes influences Aβ overproduction. Answers to these questions would help to explain why cholesterol may be considered an early risk factor for AD. There is a link between APP processing, Aβ production, and the lipid environment [437, 438]. APP is a type I membrane protein that undergoes proteolytic cleavage within its ectodomain, followed by intramembrane cleavage on its C-terminal fragment. These proteolytic pathways lead to the generation of Aβ, and both processes involve proteolytic enzymes known as secretases, as previously explained. The intramembrane cleavage site of the APP depends on the length of its transmembrane domain [439], while membrane composition and its physical state modulate the ratio between the APP derivatives Aβ_1–42 and Aβ_1–40 [440]. Aβ_1–42 is more pathogenic than Aβ_1–40 as described earlier in this review. The cleavage of APP may occur at the exact center of the lipid bilayer [441] where cholesterol may enhance γ-secretase-mediated Aβ production [442, 443] and, reciprocally, Aβ aggregates preferentially bind cholesterol within the lipidic rafts [444]. Thus, a drastic reduction in membrane cholesterol may result in decreased amyloid production [442, 443]. On the contrary, it has also been shown that neuronal membrane cholesterol loss may enhance amyloid peptide generation [445], which calls into question the population-based studies which have demonstrated that cholesterol is an early risk factor for the development of amyloid pathology [446].

Sphingomyelin (SM), another particularly significant lipid in signal transduction, when it accumulates, provides a favorable milieu for GM1 ganglioside-induced assembly of Aβ-protein [447]. According to Grimm et al. [438], while the control of cholesterol and SM metabolism involves APP processing via regulation of the activity of the γ-secretase enzyme, Aβ directly activates neutral sphingomyelinase, the enzyme responsible for metabolizing SM to phosphocholine and ceramide and downregulates SM levels or reduces de novo cholesterol synthesis by inhibiting the hydroxymethyl-glutaryl-CoA reductase (HMGR) activity. Thus, Aβ becomes a cholesterol and sphingomyelin regulator, while cholesterol and glycosphingolipids may elevate Aβ production and SM may reduce Aβ production by inhibition of γ-secretase [438]. This is particularly true in individuals carrying the ApoE4 genotype, because this isoform binds more avidly to Aβ [448] (Fig. 11).

Figure 11.  Cholesterol may interfere with α-secretase activity or may enhance γ-secretase (γs) activity, favoring the amyloidogenic pathway; also, it is possible that cholesterol contributes by relocalizing APP near to the β-secretase (βs) influence. Aβ, correspondingly, affects cholesterol influx and efflux because of its ability to bind cholesterol transporters, while it inhibits cholesterol esterification. Sphingomyelin, which makes up about 10% of the lipids of brain, has been related to γ-secretase (γs) activity and the amyloidogenic pathway. Aβ for its part may induce apoptosis by activating the cPLA2-dependent sphingomyelinase–ceramide pathway, which is a phenomenon related to caspase 3 activation, PI3K/Akt inhibition, and arachidonic acid release. Melatonin’s role (red crosses) in this hypothesis is related to the following: (1) functional stability of the membrane by avoiding the lipid peroxidation cascade, (2) interfering with lipid-derived proinflammatory signals, (3) it is also possible that melatonin may interfere with PLA2 activity, (4) melatonin reduces cholesterol absorption or augments endogenous cholesterol clearance mechanisms, and (5) melatonin may interfere with cholesterol transport as well or directly interact with cholesterol; even though the effects derived from this possible interaction remain to be clarified, it has been demonstrated that melatonin may favor the displacement of cholesterol from the phospholipid bilayer. Reciprocally, cholesterol might influence melatonin’s antioxidant capacity.

Other important relationships between lipids and AD involve the polyunsaturated fatty acids (PUFAs), particularly docosahexaenoic acid (DHA, 22:6x-3) and arachidonic acid (AA, 20:4x-6), a precursor in the production of eicosanoids. Arachidonic acid is a well-known neuroinflammation by-product, delivered from cellular membranes by phospholipases A2, cyclooxygenases, and lipoxygenases under the effect of cytokines and chemokines once the proinflammatory response to injury is established [164]. This is particularly true in the AD brain [449, 450]. Importantly, AD is associated with depletion of DHA in the brain [451, 452]. DHA comprises about 40% of the PUFAs in the brain [453], it is found predominantly in neuronal membranes in the gray matter, and it has an important role in cholinergic stimulated signal transduction at the synapse and phospholipid-mediated signal transduction involving activation of phospholipase A2 and/or PLC [454]. In fact, regional gray matter volumes in the anterior cingulate cortex, amygdala, and hippocampus, calculated using optimized voxel-based morphometry on high-resolution structural magnetic resonance images in adult humans, revealed a positive association between long-chain omega-3 fatty acid intake and corticolimbic gray matter increase in volume [455]. There is experimental evidence supporting the role of DHA in the formation of new memories; for example, in fish oil-deficient rats tested for learning ability related to reference memory and working memory, chronic administration of DHA led to an improvement in reference memory-related learning both in young [456] and in old rats [457]. The diet-induced increase in brain DHA levels has been related to enhanced reference and working memory performance as well [458]. Moreover, DHA may promote the differentiation of neural stem cells into neurons by promoting cell cycle exit and suppressing cell death as evaluated in neural stem cells obtained from 15.5-day-old rat embryos [459] probably by regulating basic helix–loop–helix transcription factors [460]. Aβ oligomer-induced effects, such as phosphorylation of tau and inactivation of insulin receptor substrate via c-Jun N-terminal kinase signaling, may be suppressed by the dietary intake of fish oil/DHA, and this may be accompanied by improvement in Y-maze performance, as demonstrated in 3xTg-AD mice [461].

Melatonin’s role in the lipid connection

Melatonin’s utility is attributable in great part to its amphipathic nature, which allows it to pass easily through all morphophysiological barriers, and this property involves an active interaction with membrane lipids. The most widely known effect of melatonin on lipids is to prevent their peroxidation [14, 462, 463]. In fact, there is an inverse correlation between melatonin levels and peroxidation of lipids following the intracerebral injection of Aβ [464]. By doing this, melatonin avoids some of the major problems derived from cell membrane dysfunction and subsequent neurotoxicity [462, 465]. However, the role of melatonin seems to be more complex (Fig. 11).

Melatonin may interfere with proinflammatory signals by blocking two principal enzymatic pathways: COX and LOX. Both of these insert molecular oxygen into molecules of free and esterified polyunsaturated fatty acids, such as arachidonic acid, and thereby synthesize several different biologically active eicosanoids. 12/15-LOX protein levels and enzyme activity, linked to mechanisms of oxidative stress and neurodegeneration, have been demonstrated to be increased in AD brain [466], the same as 5-LOX. This latter is also known as an active Aβ inducer [467], and 5-LOX gene expression may be suppressed by melatonin through its high-affinity nuclear receptors, as demonstrated in hippocampus [468]. It is noted that a drop in melatonin levels, as observed in elderly subjects with an aging-associated melatonin deficiency, has been related to 5-LOX overexpression [469]. On the contrary, a proradical effect has been described for melatonin in which it uses these pathways [470], which is a transitory, early melatonin effect observed in leukocytes, accompanied by strong liberation of arachidonic acid and explained as a consequence of calmodulin binding.

Lipid mobilization to obtain polyunsaturated fatty acids from the membranes for LOX and COX enzymes requires the intervention of an enzyme, the calcium-dependent phospholipase A2 (PLA2); melatonin may be a negative endogenous regulator of cytosolic PLA2 presumably through a melatonin receptor-mediated process as demonstrated in the rat pineal gland in vitro [471].

Melatonin may also significantly reduce cholesterol absorption causing significant decreases in total cholesterol, triacylglyceride, very low-density lipoprotein cholesterol, and low-density lipoprotein cholesterol in plasma and the concentrations of cholesterol and triacylglyceride in the liver; this was demonstrated in rats fed on high-cholesterol diet [472]. In a very particular study (hypercholesterolemic mice fed with an atherogenic diet), it has reported that melatonin induces atherosclerotic lesions in the proximal aorta as a consequence of exacerbated LDL oxidation [473]. On the contrary, in normolipidemic postmenopausal women, melatonin has shown to inhibit oxidative modification of low-density lipoprotein particles [474].

The cholesterol–melatonin relationship in the biological membranes is, from a functional point of view, particularly relevant. Evaluated in dry cholesterol/lecithin mixed reversed micelles, melatonin is mainly located in and oriented in the proximity of the polar heads where it appears to compete with cholesterol for the hydrophilic binding sites of lecithin; this is particularly true at high cholesterol concentrations. Thus, free radical displacement coming from the aqueous compartment to a high concentration of cholesterol could influence melatonin’s antioxidant activity. At the same time, the more concentrated cholesterol is in membranes, the greater the membrane rigidity. It is speculated that competition with melatonin may favor the displacement of cholesterol from the phospholipid bilayer [475]. This feature gains relevance because decreasing membrane cholesterol in mature neurons may reduce their susceptibility to Aβ, tau production, and cell death, whereas increasing membrane cholesterol in young neurons enhances the Aβ-mediated cellular processes, as demonstrated in hippocampal neurons [476] (Fig. 11).

Concluding remarks

Is it just a coincidence that while melatonin declines with age, the probability of experiencing AD grows? New findings on CSF flow, possibly moving from the choroid fissure into the ventricular system, could help to explain why melatonin is found in higher concentration in the CSF than in simultaneously sampled blood. Thus, neural tissue in contact with the ventricular system via hypothetical choroid plexus portals would have high levels of cellular melatonin [477]. A CSF deficiency of melatonin has been demonstrated to precede clinical symptoms of AD [22, 265] and, the loss of this lipophilic antioxidant normally concentrated in the ventricular CSF exposes highly active and vulnerable brain tissue to self-generated oxygen radicals.

Melatonin’s more common indication in patients with AD is sleep regulation because sleep disruptions, nightly restlessness, and sundowning are frequently observed in elderly and particularly in patients with AD [44]. This is related to decreased levels of both melatonin [415] and melatonin receptors in the SCN [478]. It has also a potential to treat mood disorders [30, 43], commonly associated with AD [28, 30].

By taking into account the most compelling hypotheses trying to explain the cellular and biomolecular alterations in Alzheimer’s disease, however, a growing body of evidence supports the protective role of melatonin, exceeding the above-mentioned conventional uses. Thus, melatonin has a role in each of the different reviewed hypotheses: (i) it prevents amyloid overproduction, (ii) it reduces hyperphosphorylation of tau [15, 166], (iii) it is an antioxidant and free radical scavenger, (iv) it modulates proinflammatory processes, (v) it works well as an anticholinesterase agent, (vi) it prevents mitochondrial damage and the apoptotic phenomena related with AD, (vii) it may impair calcium-dependent toxicity, (viii) it reduces insulin resistance as well as glucose transport, and finally (ix) it is able to maintain the integrity and functionality of cellular membranes, thanks to its ability to interact with lipids or against their neuroinflammatory or proapoptotic signals when the lipid balance becomes affected.

The functional translation of these biomolecular effects has been also well documented. MT2 receptor-deficient mice undergo impairment of synaptic plasticity and learning-dependent behavior, suggesting that MT2 receptors participate in hippocampal synaptic plasticity and in memory processes [299]. Also relevant are the protective effects of melatonin on cognition in a variety of tasks of working memory, spatial reference learning/memory, and basic mnemonic function, as observed in a transgenic model of AD [20]. It is worth remembering the increased melatonin 1a-receptor immunoreactivity in the hippocampus of patients with AD, which may be a compensatory response to impaired melatonin levels [417].

It is also true that in transgenic animals, additionally exposed to aluminum for months (aluminum has been circumstantially linked to AD [16, 479, 480]), melatonin did not ameliorate the behavioral effects [21] even though they did respond well to the antioxidant actions of [92]. Limited or even null results in memory performance or other high mental functions using melatonin have been observed in some clinical trials [11]. However, these changes are associated with the loss of brain cells. The greater the loss of brain cells, the more severe the deterioration in high mental functions, and no drug exists that is capable of regenerating lost neurons. Once the brain tissue has degenerated, there is just a little chance of recovering [176].

There are several concerns about melatonin. From a cellular, basic perspective, we find a single report where melatonin reportedly worsens the neurodegenerative pathology. It is a rotenone-induced Parkinson’s disease-specific model, where melatonin not only failed to impair neuronal degeneration but potentiated neurodegeneration [19]. However, synergistic effects of melatonin against MPTP-induced mitochondrial damage and dopamine depletion have been also reported [481]. Otherwise, the evidence indicating the protective role of melatonin on mitochondria within CNS is vast [239, 342–346, 348–350, 352, 353, 355, 359].

On the other hand, there are several clinical concerns. One refers to dose and side effects. In a few isolated studies, melatonin has been related to sleepiness, dizziness, headaches, nightmares, confusion, sleepwalking, daytime sleepiness, and abdominal discomfort, even though some results deserve a re-analysis. For example, using a high dose of melatonin (20 mg/kg) in mice undergoing electroconvulsive stimulation, a strong long-term memory deficit was attributed to melatonin [482]. However, it was not clear what caused the memory impairment actually, because the electroconvulsive stimulation has been often related to memory impairment by itself; and not only in rodents but in humans (reviewed in [483]). On the other hand, not in rats but in epileptic children, a randomized, double-blind, placebo-controlled trial demonstrated that melatonin improved cognitive and social function as well as emotional well-being and behavior [484]. It has been also reported that ramelteon, a synthetic melatonin derivative, administrated prior to a short (2 hr) evening nap, impairs significantly neurobehavioral performance for up to 12 hr after awakening [485].However, a 2-hr nap is not a short nap; naps longer than 30 min have been largely associated with a loss of productivity and sleep inertia [486]. Thus, probably ramelteon was not responsible for a low neurobehavioral performance, but the long nap by itself.

Melatonin doses ranging from 3 to 6.6 g/day for more than 30 days were administrated in patients with Parkinson’s disease, and the number of collateral effects, such as headache, somnolence, or abdominal cramps, were isolated, with melatonin being ‘remarkably well’ tolerated [487]. More severe collateral effects have never been observed. In fact, doses up to 800 mg/kg failed to produce death in mice [488]; indeed, no lethal dose of melatonin has been established overall; melatonin has been repeatedly shown, at any dose to be free of significant side effects.

Other concerns go in the opposite direction: melatonin’s extremely short half-life in the circulation, a question that have led to the development of synthetic melatonergic drugs with substantially longer half-life than melatonin [489]. It should be noted however that the short half-time of melatonin in the blood does not necessarily translate into a short half-life within the cells.

This review underlines the potential of melatonin to slow the progressive deterioration of AD brain, in light of the known hypotheses that attempt to explain the neurodegeneration. As observed in AD models and based on a multitude of experimental results, melatonin benefits may well stem from actions that exceed its well-known antioxidant properties. Unfortunately, while there are a number of well-founded hypotheses, the real cause of neurodegeneration in AD is still unknown. Very likely, there are many contributing causes to this highly complex disease.

Acknowledgements

A. Coto-Montes is an associate professor at Universidad de Oviedo (Spain). Jose A. Boga is a researcher of HUCA/FICYT (Oviedo, Spain). His stay at UTHSCSA has been subsidized by Instituto de Salud Carlos III (Madrid, Spain).