Iron is an essential transition metal for many fundamental neuronal functions in the brain, such as oxygen transport, mitochondrial respiration and myelin synthesis, as well as acting as a cofactor for a large number of metalloproteins involved in metabolism and signal transduction (Castellani et al., 2012; Biasiotto et al., 2016). Iron deficiency affects numerous neurological peculiarities, especially during babyhood, causing brain development stagnation. Iron levels in the brain increase with aging; however, excessive levels of iron lead to an over-abundant production of reactive oxygen species through Fenton chemistry, which eventually causes cell damage (Droge and Schipper, 2007; Muhoberac and Vidal, 2013). Moreover, a growing number of studies have reported that disruption of iron homeostasis may be a hallmark of many neurodegenerative disorders (Ward et al., 2014; Jiang et al., 2017).

Many studies, using a variety of different techniques, have identified that iron levels are increased in AD brains (Bartzokis et al., 1994; LeVine, 1997), specifically in the globus pallidus and putamen (Wang et al., 2014; Moon et al., 2016). However, meta-analyses have revealed unchanged or reduced serum iron levels in AD patients with respect to healthy subjects (Tao et al., 2014; Wang et al., 2015). Although the explanation for this imbalance is still unknown, studies on the role of iron in AD have shown that excessive iron stimulates hydroxyl radical formation via the Fenton reaction, which may contribute to increased oxidative stress levels in AD. Oxidative damage to AD-linked protein aggregations by excessive levels of iron can cause synaptic dysfunction and neuronal cell death, a detrimental process which can be alleviated by the administration of iron chelators (Salkovic-Petrisic et al., 2015). High iron content is loaded around senile plaques and elevates the production of Aβ by increasing the expression of amyloid-β precursor protein (APP). Iron modulates APP transcription through iron-responsive elements (IREs) present in the 5′-untranslated region of APP mRNA (Rogers et al., 2002). Excessive iron levels result in the dissociation of iron regulatory proteins (IRPs) from their IRE binding sites and abolish their repression of APP mRNA, leading to increased APP translation, whereas lead (Pb) represses the IRE-mediated enhancement of APP translation (Rogers et al., 2016). Furthermore, iron has a high affinity for binding with Aβ in vitro, which could promote its aggregation and accelerate the formation of oligomers (Honda et al., 2004; Liu et al., 2011). Previous studies have reported the presence of Aβ reduced reactive oxygen species (ROS) at the beginning of neurodegeneration, which probably benefit from the alleviation of oxidative stress by binding iron (Nunomura et al., 2001; Jolivet-Gougeon and Mallet, 2016). However, a high iron concentration gives rise to the antioxidant system becoming overburdened and thus results in cellular damage over time, leading to a vicious cycle involving a higher production of Aβ as a result of enhanced oxidative stress (Tamagno et al., 2008). In addition to Aβ, iron also binds to tau, which promotes its aggregation in iron-enriched regions (Sayre et al., 2000). Similarly, iron has also been shown to induce tau phosphorylation, which might be caused by the activation of cyclin-dependent kinase 5 (CDK5) and glycogen synthase kinase 3β (GSK3β) pathways (Lovell et al., 2004). The administration of iron chelators, such as desferrioxamine, could reduce iron-induced tau phosphorylation in AD transgenic mice (Guo et al., 2013).

A dynamic balance between iron influx and efflux is critical for intracellular iron homeostasis, in which multiple transporter proteins play key roles. The dysregulation of iron importers, including transferrin (Tf), divalent metal transporter 1 (DMT1), lactoferrin (Lf) and melanotransferrin (MTf), as well as iron exporter ferroportin (Fpn), may account for iron accumulation in the affected brain regions of AD patients (Figure 1). DMT1, also known as divalent cation transporter 1, is expressed on neurons, astrocytes and microglia but not oligodendrocytes, and is relevant to the pathway involved in the influx of Fe²⁺ (Song et al., 2007). Two isoforms of DMT1, DMT1-IRE and DMT1+IRE, have been shown to colocalize with Aβ in the plaques of the AD brain and the levels of both DMT1 isoforms are significantly increased in the frontal cortex and hippocampus in a APP/PS1 transgenic mouse model (Zheng et al., 2009), which was accompanied by a reduction in Fpn expression (Xian-hui et al., 2015), suggesting that the deregulation of iron metabolism-related protein DMT1 and Fpn plays a critical role in the iron-mediated neuropathogenesis of AD. Hepcidin, an iron-homeostatic peptide, is colocalized with Fpn in neurons and astrocytes showed a reduced expression of Fpn in AD brains. Downregulation of hepcidin exhibits impairment in the iron export pathway which results in cellular iron retention (Raha et al., 2013). APP is a transmembrane protein which was found to catalytically oxidize Fe²⁺ to Fe³⁺ by ferroxidase activity and then interact with Fpn to promote iron export (McCarthy et al., 2014); however, this process is inhibited by extracellular zinc, which originates from zinc-Aβ complexes (Duce et al., 2010). From the same laboratory, these authors also showed that loss of soluble tau could cause iron retention by impairing APP-mediated iron export (Lei et al., 2012), and that such suppression can be caused by an iron chelator (Lei et al., 2015) or by lithium treatment (Lei et al., 2017). Furthermore, sirtuin 2 was reported to regulate cellular iron homeostasis via the deacetylation of nuclear factor erythroid-derived 2-related factor 2, which acts as a transcription factor to control Fpn expression (Yang et al., 2017). Interestingly, a recent report showed that compounds extracted from Chinese herbs could down-regulate DMT1 expression and up-regulate Fpn expression, thus providing a new strategy for reducing iron overload-induced impairment in AD (Dong et al., 2015). The Tf-transferrin receptor (TfR) complex is responsible for iron uptake in endothelial cells of the BBB. The transport of iron across this barrier is most likely the result of the receptor-mediated endocytosis of Tf-bound iron (Moos and Morgan, 2000). A previous proteomics study found that Tf levels were significantly different in the cerebrospinal fluid (CSF) of familial AD when compared between those who carried mutations and related non-carriers (Ringman et al., 2012). The structure of Lf is similar to Tf, in that they are both composed of two lobes, each having one binding site for Fe³⁺ (Baker et al., 1994). Lf is highly expressed in monocytes/macrophages and fibrillar-type senile plaques in the cerebral cortex of AD patients (An et al., 2009). In addition, senile plaque formation promotes Lf deposition with age (Wang et al., 2010). Lipoprotein receptor-related protein (LRP) is a cell surface receptor involved in Aβ clearance via an endocytic process; Lf can bind to LRP and substantially enhance the clearance of soluble Aβ rather than the production of Aβ (Qiu et al., 1999). Recently, a liposomal system with surface Lf was developed in order to deliver neuron growth factor across the BBB. This method was shown to be beneficial in controlling the progression of AD (Kuo and Wang, 2014; Meng et al., 2015).

Another iron-binding protein, MTf, specifically localized in a subset of reactive microglia cells, and associated with the formation of senile plaques in the AD brain, was reported previously (Jefferies et al., 1996; Yamada et al., 1999). Although serum MTf levels are 3–4 fold higher in AD patients than in normal controls (Kim et al., 2001), other studies argued that serum MTf levels remained unchanged in AD patients and that MTf does not play a role in the transport of iron in the brain (Desrosiers et al., 2003; Suryo Rahmanto et al., 2012). Thus, further studies are now needed to investigate the role of MTf in the development of AD. In addition, an increase in heme oxygenase-1 level has been identified in astrocytes and neurons, resulting in co-localization with NFTs, senile plaques and corpora amylacea (Smith et al., 1994; Schipper, 2011; Gupta et al., 2014), suggesting that pathological iron deposition occurs due to the release of iron from heme proteins in AD-affected neural tissues.

Copper is another essential transition metal that plays a critical role in several cellular functions; for example, as a structural component of enzymes required for energy metabolism and antioxidant defense. Furthermore, copper is involved in cellular respiration, free radical defense, and neurotransmitter synthesis (Zatta and Frank, 2007; Desai and Kaler, 2008; Scheiber and Dringen, 2013). Copper deficiency in the brain has an adverse effect on the development and maintenance of myelin and can induce degeneration of the nervous system. In contrast, excessive levels of copper also augment the Fenton reaction causing the generation of free radicals, as has been extensively reported for iron (Brewer, 2008; Gybina et al., 2009). Current evidence indicates that alteration in copper levels occurs in neurodegenerative diseases, such as AD.

The involvement of copper in the pathophysiology of AD is complex. High concentrations of copper have been detected in senile plaques (Lovell et al., 1998). In contrast, some studies have reported a deficiency of total copper brain levels in the AD brain (Deibel et al., 1996), and a recent meta-analysis showed that although the combined level of plasma and serum copper was higher in AD patients (Klevay, 2008), the total levels of copper in CSF were no different when compared between healthy subjects and AD patients (Ventriglia et al., 2012; Vaz et al., 2017). The rational explanation for this heterogeneity is that a significant quantity of copper precipitates with senile plaques in AD-affected regions, leading to copper deficiency in other regions. There is a common agreement that copper interacts with both Aβ and tau, and exacerbates their pathological consequences (Sparks and Schreurs, 2003; Kitazawa et al., 2009). Copper directly binds to Aβ with a high affinity and facilitates its oligomer formation (Tougu et al., 2008; Jin et al., 2011). The mechanism underlying copper-mediated Aβ oligomer cytotoxicity might involve oxidative stress, because copper and Aβ can catalytically generate hydrogen peroxide in vitro. The toxicity of the Cu-Aβ complex could be reversed by copper chelators, such as clioquinol (Matlack et al., 2014) and PBT2 (Adlard et al., 2008). In addition, APP and Aβ precursor-like protein 2 (APLP2) also have a copper binding site (Barnham et al., 2003). It has been reported that APP may function as a copper transporter since APP or APLP2 knockout mice showed elevated copper levels in the cerebral cortex (White et al., 1999b). However, the same author also published another study in which APP knockout in cortical neurons did not affect copper uptake (White et al., 1999a), indicating that APP may not be a copper-carrier, but rather shows inappropriate interactions with copper. Copper promotes the redistribution of APP at the cell membrane by enhancing exocytosis and reducing endocytosis (Acevedo et al., 2011). Moreover, copper increases the phosphorylation of endogenous APP via GSK3β and facilitates its proteolytic cleavage to generate Aβ (Acevedo et al., 2014). It was found that the microtubule-binding domain of tau can also bind with copper. Copper binds to tau and causes its aggregation in vitro (Su et al., 2007). In an in vivo mouse model of AD, copper exposure induced tau hyper-phosphorylation and generated hydrogen peroxide (Kitazawa et al., 2009). The mechanism of copper-mediated tau phosphorylation is thought to occur via the activation of CDK5 and GSK3β pathways (Crouch et al., 2009a).

The mechanisms involved in copper dislocation in the AD brain remain clear. Copper transporter 1 (CTR1) and the copper transporting P-type ATPases, including ATP7A and ATP7B, are the major transporters involved in the cellular regulation of monovalent copper (Kuo et al., 2006; Yu et al., 2017) (Figure 2). DMT1 may participate in delivering divalent copper into cells for the synthesis of copper containing enzymes (Zheng and Monnot, 2012). Nevertheless, in the context of copper overload, ATP7A and ATP7B play dual-functions to export excess copper out of cells in a manner which is dependent on ATP hydrolysis. In addition to the transporters, a cluster of intracellular proteins, recognized as molecule chaperones such as antioxidant protein-1, cytochrome oxidase enzyme complex and copper chaperone for superoxide dismutase (SOD), also take part in delivering copper to specific targets (Harris, 2001; Zheng and Monnot, 2012). In a Drosophila model of AD, genetic knockdown of CTR1C, one of the CTR1 family members homologous to the human gene, significantly reduced copper accumulation in the brain (Lang et al., 2013). These consistent results were also observed in flies when CTR1B, another copper importer, was inhibited, or when DmATP7, a copper exporter, was enhanced in AD flies. These flies exhibited increased levels of Aβ production but a reduction in Cu-Aβ complex-induced oxidative stress, suggesting that Aβ oligomers, or the increased Aβ aggregates, were less toxic in a reduced copper influx mediated by CTR1 knockdown (Lang et al., 2013). In a mouse model of AD, ATP7A is upregulated in activated microglial cells where the amyloid plaques are clustered, resulting in a significant change of microglia copper trafficking. This depicts a neuromechanism in which inflammation-induced copper dyshomeostasis in microglia is associated with AD (Zheng et al., 2010). Genetic analysis of the genome of AD patients revealed that a cohort of single nucleotide polymorphisms in ATP7B account for imbalances in circulating nonceruloplasmin-bound copper which increase the risk of AD, supporting the notion that changes in copper homeostasis may accelerate the neurodegeneration that lead to AD (Bucossi et al., 2011; Bucossi et al., 2013; Squitti et al., 2013).

Zinc is the second most abundant metal in the body after iron and is a vital component of 100s of enzymes and proteins. Zinc is also involved in cell signaling in a much more extensive way than other metals, particularly because it can act as a neurotransmitter (Barr and Burdette, 2017). During neuronal activity, synaptic transmission releases zinc into the synaptic cleft, which features a major zinc-efflux mechanism in neurons. Zinc interacts with various post-synaptic receptors and transporters. For example, zinc inhibits N-methyl-D-aspartate receptors (NMDARs) and activates GPR39; collectively these mechanisms serve to modulate cell signaling and synaptic plasticity (Paoletti et al., 1997; Khan, 2016). Synaptic zinc deficiency not only affects the proper development of the brain, but also causes depression-like symptoms and dysfunctions in learning and memory (Swardfager et al., 2013). Furthermore, zinc deficiency has been reported to increase the overload of other metals, such as copper, nickel and possibly toxic metals, by upregulating zinc transporters (Antala and Dempski, 2012). However, excessive zinc can be sequestered by the mitochondria and triggers the generation of ROS by interfering with the activity of complex III of the electron transport chain, thereby promoting mitochondrial dysfunction and neuronal death in a manner which is relevant to a variety of disorders (Frazzini et al., 2006).

The role of zinc in the pathogenesis of AD has been intensively studied since a previous study described the redistribution of zinc into extracellular senile plaques (Bush et al., 1994). However, there is an inconsistent conclusion with regard to zinc levels in AD patients. Many studies have shown high levels of zinc in the AD brain and CSF (Religa et al., 2006; Hozumi et al., 2011). Some other reports reported no difference or even reduced zinc levels in the affected brain and serum of AD patients compared to controls (Panayi et al., 2002; Ventriglia et al., 2015; Wang et al., 2015). The reason for such heterogeneity in zinc levels remains unknown. Unfortunately, this is likely to hinder the development of ways to supplement or obliterate zinc as a form of AD treatment. Despite such conflict, the binding of zinc to histidine residues in the C-terminus of Aβ, and the promotion of aggregates, should be noted. Relative to iron and copper, zinc binds to Aβ with a greater affinity upon a wide range of pH (Yoshiike et al., 2001). On the one hand, this binding conceals the proteolytic cleavage site where Aβ is degraded by metalloproteases (Crouch et al., 2009b); on the other hand, binding induces the loss of zinc bioavailability in the synaptic cleft, thus contributing to changes in synaptic plasticity and a cognitive decline in AD (Deshpande et al., 2009). Furthermore, zinc deficiency may also have deleterious effects on the maturation of brain-derived neurotrophic factor (Khan, 2016). Although excessive zinc enhances Aβ toxicity, zinc also changes the conformation of Aβ and prevents copper from interacting with Aβ, which, in turn, ameliorates the oxidative stress burden (Cuajungco et al., 2000; Pedersen et al., 2012). In addition, zinc increases the expression of presenilin 1 in order to facilitate zinc uptake; at the same time, zinc also inhibits the activity of γ-secretase, which is involved in the generation of Aβ from APP (Hoke et al., 2005). Furthermore, zinc may also able to bind to tau proteins with moderate affinity and modulate phosphorylation by activation of GSK3β, extracellular regulated protein kinase 1/2 (ERK1/2), and c-Jun N-terminal kinase (JNK) (Pei et al., 2006; Lei et al., 2011).

In neurons, zinc homeostasis is principally controlled by three groups of transporters, comprising the zinc transporters (ZnTs), zinc-regulated transporter-like and iron-regulated transporter-like proteins (ZIPs), and metallothioneins (MTs) (Figure 3). ZnTs facilitate zinc efflux from cells or mediate excessive zinc from the cytoplasm into organelles and intracellular vesicles (Huang and Tepaamorndech, 2013). ZIPs play roles which are almost opposite to those of the ZnTs, and drive zinc import into cells or promote the movement of zinc from intracellular vesicles into the cytoplasm (Cousins et al., 2006; Huang and Tepaamorndech, 2013). MTs are zinc homeostasis-regulating proteins that control cellular zinc levels and related signaling pathways (Krezel et al., 2007). Immunofluorescent evidence demonstrated that ZnTs (ZnT1, 3, 4, 5, 6, and 7) are extensively present in Aβ plaques in the cortex of human AD brains (Zhang et al., 2008). Of these, ZnT3 is predominantly present on the synaptic vesicles of zinc-containing glutamatergic neurons (Palmiter et al., 1996). With aging, decreased levels of ZnT3 were found in elderly people, particularly in people with AD (Whitfield et al., 2014). ZnT3 knockout mice displayed an age-dependent deficit in cognition (Adlard et al., 2010), while mice overexpressing APP, but lacking ZnT3, appeared to have lower levels of synaptic zinc and a reduced plaque burden (Lee et al., 2002), implicating the contribution of synaptic zinc to the deposition of amyloid plaques in AD. In contrast, increased intra-neuronal zinc, caused by ZnT3 knockout, exacerbated neuronal damage in AD (Lovell et al., 1998). The reuptake of zinc into the presynapse still, however, requires characterization. The ZIPs have a fundamental role in mediating zinc influx, and other mechanisms are likely to indirectly participate, such as presenilins, mutations of which are known to cause familial AD (Greenough et al., 2011). In addition, MTs are major zinc-buffering peptides that maintain the cytosolic zinc balance. Four main MT isoforms (MT-1 to -4) are also expressed in brains. MT-1 and MT-2 are upregulated in AD patients, whereas MT-3 is downregulated (Hidalgo et al., 2006). In an animal model of AD, the lack of MT1/2 results in a reduction in the amyloid plaque burden, and therefore recovers the APP-induced changes in mortality (Manso et al., 2012a). MT3 is correlated to Aβ aggregation by cysteine oxidation. MT3 deficiency partially rescues the APP-induced mortality of females and leads to changes in APP-induced behavioral phenotypes of mice (Manso et al., 2012b). More recently, it was found that MT3 modulates Aβ uptake in astrocytes through its positive effect upon actin polymerization (Lee et al., 2015). Furthermore, a combination of microscopy and spectroscopy studies demonstrated that MT3 prevents Cu-Aβ-mediated neurotoxic effects probably via a metal exchange between Zn-MT3 and the aggregated Cu-Aβ (1-40), thus leading to the suppression of ROS production (Pedersen et al., 2012). In contrast, another study investigated the protective role of astrocyte-derived MTs on primary cortical neurons against Aβ toxicity by reducing ROS content, upregulating Bcl-2 and inhibiting proinflammatory cytokine production (Kim et al., 2012). To better understand AD-related zinc dyshomeostasis, it is necessary to investigate the role of zinc transporters in the etiology of this disease.

Manganese (Mn) is a critical trace element present in human tissues and exerts important physiological roles for growth and intracellular homeostasis (Prakash et al., 2016). Manganese is also used as a cofactor for key enzymes involved in normal cell function, such as SOD and glutamine synthetase. An increasing number of reports have provided evidence that Mn overload is associated with neurodegenerative diseases, and that even a small excess of Mn can induce symptoms that are consistent with manganism (Park, 2013). The mechanisms of Mn-mediated cytotoxicity involve the over-production of ROS, mitochondrial dysfunction, abnormal energy metabolism, accumulation of intracellular toxic metabolites, the depletion of cellular antioxidant defense and autophagy (Guilarte, 2013; Martinez-Finley et al., 2013).

Manganese was detected in significantly higher levels in the brain of AD patients with dementia compared to healthy subjects, while the highest level of Mn was detected in the parietal cortex (Srivastava and Jain, 2002; Tong et al., 2014). This suggests that Mn overload may be involved in the pathology of AD and cognitive dysfunction. Chronic Mn exposure to non-human primates was shown to alter gene expression which dispersed Aβ plaques. Most of the altered genes were targeted by p53; one such gene was amyloid-beta precursor-like protein 1 (APLP1), the most highly upregulated gene in the frontal cortex (Guilarte, 2010). Mn exposure appears to particularly target the frontal cortex leading to incipient dementia (Schneider et al., 2013). In addition, Mn treatment increases levels of Aβ peptides both in vitro and in vivo; the mechanism involved is probably related to the disruption of Aβ degradation (Tong et al., 2014). More recently, a study showed that Mn could weakly bind to the specific site of Aβ, as described for other biometals (Wallin et al., 2016). However, the effects of such binding by Mn to Aβ in the promotion of Aβ aggregation need to be further demonstrated. Mn is a component of Mn superoxide dismutase (Mn-SOD), which is an antioxidant enzyme that plays an important role in maintaining mitochondria vitality. Increased Mn obstructs oxidative respiration, thus increasing ROS production and eventually leading to mitochondrial dysfunction (Gunter et al., 2006). Partial deficiency of Mn-SOD increased Aβ plaque deposition and tau phosphorylation in a transgenic mouse model of AD (Li et al., 2004; Melov et al., 2007). In contrast, the overexpression of Mn-SOD showed some benefit against AD pathology by reducing the burden associated with cortical plaques (Dumont et al., 2009), which further demonstrated the links between mitochondrial oxidative stress and the pathophysiology of AD. In addition, Mn toxicity results in a cognitive decline in humans, and it has been postulated that excessive Mn uptake causes iron deficiency in the Golgi apparatus, consistent with fact that Mn and iron compete with the same binding sites and transport mechanisms, at least to some extent (Carmona et al., 2010).

Manganese transport is mediated by multiple importers, including DMT1, Tf/TfR, ZIP8 and ZIP14, and dopamine transporter (DAT), as well as by various exporters including park9/ATP13A2, SLC30A10, Fpn, and the secretory pathway Ca²⁺-ATPase 1 (SPCA1) (Figure 4). Of these, DMT1 is not only a transporter for iron influx, but also the first mammalian transporter for cellular Mn uptake. DMT1 facilitates the movement of Mn across the BBB, particularly under in conditions of iron scarcity (Garrick et al., 2006). DMT1 transports divalent Mn, while Tf transports trivalent Mn into cells through a ligand-receptor endocytosis mechanism (Subramaniam et al., 2002). ZIP8 and ZIP14 have high binding ability for zinc, but several studies have found that they are also involved in the absorption of body Mn from the liver and lungs (Fujishiro et al., 2012; Bekerydemir et al., 2017; Lin et al., 2017). In recent years, efforts have focused upon the role of exporter proteins in maintaining Mn levels. Genome analysis identified SLC30A10, a cell surface-localized efflux exporter, which may transport both zinc and Mn. Patients who suffer from Parkinson’s disease, and carry mutations in SLC30A10, show accumulation of Mn in the brain (Tuschl et al., 2016). SLC30A10 has consistently shown to be significantly reduced in the frontal cortex of AD patients, and in APP/PS1 transgenic mice, suggesting that its dysregulation contributes to the pathology of AD (Bosomworth et al., 2013). A number of studies have also shown that the iron exporter Fpn could also function as a cellular exporter of Mn in a pH-dependent manner in order to attenuate Mn accumulation and cytotoxicity (Madejczyk and Ballatori, 2012; Seo and Wessling-Resnick, 2015). Furthermore, ATP13A2 acts as a cation transporter in Mn and zinc transportation. Studies have shown that the over-expression of ATP13A2 reduces intracellular Mn concentration, which in turn, alleviates Mn-provoked lethality; loss-of-function mutations in ATP13A2 are correlated with increases in both α-synuclein and Aβ plaques in Lewy body disease (Murphy et al., 2013). Moreover, plasma membrane ATPase-related 1 (PMR1), a SPCA1 homolog in yeast, has been shown to mediate Ca and Mn transport, and the ectopic expression of SPCA1 in yeast increases sensitivity to Mn toxicity (Ton et al., 2002). Thus, it has been suggested that SPCA1 represents another regulator for cellular Mn homeostasis, although the affinity between SPCA1 and Mn, as well as the roles of SPCA1 in AD pathogenesis, require further investigation.

Magnesium (Mg) is one of the most abundant divalent cations in cells and has been known to be required for multiple enzymatic synthesis and cellular processes including energy metabolism, ion channels and synaptic plasticity (Misra and Draper, 1998; Bairoch, 2000). Calcium (Ca) acts as a ubiquitous second messenger that has been studied extensively with regard to its role in controlling cellular function (Komuro and Kumada, 2005). Intracellular Ca concentrations are tightly regulated by various Ca channels, pumps and Ca binding proteins, and are also modulated by other metal ions, such as Mg. In neurons, Mg and Ca collectively play a vital role in neuronal growth and signal transmission. It is noteworthy that Mg has been identified as an antagonist to Ca. Under physiological conditions, Mg maintains intracellular Ca levels and protects neurons from excitatory responses induced by Ca overload (Levitsky and Takahashi, 2013). However, disturbances in Mg and Ca homeostasis alter a cascade of events which lead to a variety of diseases including diabetes, cancer, and neurodegeneration (Volpe, 2013).

There is compelling evidence that brain and serum Mg levels are significantly lower, and Ca levels significantly higher, in AD patients than those in age-matched healthy subjects (Andrasi et al., 2005; Cilliler et al., 2007; Stutzmann, 2007). Chronically increased Ca levels increase the expression of APP and ApoE and facilitate the formation of Aβ aggregation through a mechanism involving the stabilization of γ-secretase (Brzyska and Elbaum, 2003; Ho et al., 2010), and reciprocally, Aβ aggregation alters membrane Ca permeability that further worsens AD (Kelly and Ferreira, 2006). Few studies have investigated the roles of Mg in the pathogenesis of AD. An in vitro study showed that both Ca and Mg can trigger the process of hyper-phosphorylated tau aggregation (Yang and Ksiezak-Reding, 1999). The administration of magnesium-l-threonate increases Mg levels in the brain, which decreases levels of β-secretase (BACE1) thus reducing the levels of soluble APP and β c-terminal fragments, thus alleviating synaptic loss and the cognitive deficits associated with Alzheimer’s symptoms (Li et al., 2014). Furthermore, treatment with Mg sulfate reduces levels of hyper-phosphorylated tau by inhibiting GSK-3β phosphorylation and increasing the activity of protein kinase B (Akt) and phosphatidylinositol 3 kinase (PI3K) (Gomez-Ramos et al., 2006; Xu et al., 2014), thus suggesting that Mg acts as a neuroprotective factor in the development of AD. It is though that the underlying mechanism involves Mg blocking the long term activation of NMDAR-mediated Ca influx and thus reducing Ca-induced neuroinflammation. NMDARs are cationic channels activated by glutamate, possessing a high permeability to Ca ions upon synaptic activity, such as learning and memory. It has been shown that the over-activation of NMDAR by Aβ aggregation may occur during the early stages of AD (Parameshwaran et al., 2008). Continuous Ca influx increases intracellular Ca levels, which initiates a number of enzymatic processes that result in protein destruction, peroxidation, and neuronal death (Mota et al., 2014). Under normal conditions, Mg, as an endogenous blocker, binds to NMDAR subtypes, NR1/2A and NR1/2B, which are constituent parts of NMDARs present in AD-affected brain regions (Kotermanski and Johnson, 2009). Blocking channels by adding Mg reduces Ca influx into post-synaptic neurons, so as to alleviate excitotoxic cell death during dementia. In addition, the activation of ATP-gated P2X purinergic receptors (P2XRs) associated with neuroinflammation has been identified in neurodegenerative disease (Witting et al., 2004). One subtype, P2X7R, can form an oligomer to create membrane pores in microglia, thus facilitating Ca influx (North, 2002). Using tissue culture, it has been shown that Mg can reduce levels of intracellular Ca through P2X7R, and that ameliorated purinergic receptor stimulation activated neuroinflammation, suggesting that elevated Mg is a broad spectrum inhibitor of Ca entry through cell surface channels (Lee et al., 2011).

Over recent decades, a large body of transporters, channels, exchangers, and buffering proteins have been found to be involved in the maintenance of cellular Mg and Ca homeostasis. As illustrated in Figure 5, a range of channels are known to mediate Mg influx into cells, including magnesium transporter 1 (MagT1) and transient receptor potential melastatins 6 and 7 (TRPM6/TRPM7), as well as cyclin M (CNNM) transporter. Correspondingly, sodium-independent Mg exchanger and solute carrier family 41 member 1 (SLC41A1) are required to favor Mg extrusion (Romani, 2011; de Baaij et al., 2015). Furthermore, intracellular Ca balance can be achieved through a large number of Ca transporters. Voltage-gated Ca channels, store-operated Ca channels, NMDAR, and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) are responsible for increases in Ca. Ca-binding buffering proteins, such as calbindin, facilitate the storage of Ca in the endoplasmic reticulum (ER), whereas the action of the Na/Ca exchanger, as well as the Ca-ATPase pump, mediates the export of Ca out of cells. In the AD brain, mutant presenilins are known to activate two types of Ca receptors and that plasma membrane Ca-permeable channels allow the leakage of Ca ions from the ER into the cytoplasm, thus causing significant effect upon ER-Ca dynamics (Tu et al., 2006; Cheung et al., 2008). Moreover, studies have indicated that Aβ oligomers can either stimulate the formation of Ca permeable channels, or bind to NMDARs, thus promoting Ca entry through the plasma membrane (Diaz et al., 2009; Arbel-Ornath et al., 2017). However, our knowledge about the roles of Mg transporters in AD pathology is very limited. Recently, the physiological role of TRPM7 was found to be coordinated by presenilins, the mutation of which can lead to Familial Alzheimer’s disease (Oh et al., 2012). Elimination of TRPM2 in APP/PS1 mice reduced ER stress and improved age-dependent memory deficits, while in vitro studies showed that the knockdown of TRPM2 blocked the Aβ-mediated increase in whole cell current magnitude, thus indicating the importance of TRPM2 activity in Aβ neuronal toxicity (Ostapchenko et al., 2015). In this study, the authors reported that TRPM2 alteration leads to Ca imbalance, although its role in the regulation of Mg associated with AD was ignored (Ostapchenko et al., 2015). In addition, Amyotrophic Lateral Sclerosis and Parkinsonism Dementia has also been shown to be associated with lower Ca and Mg levels than healthy subjects, thus implicating the dysfunction of TRPM2 and TRPM7 channels (Hermosura and Garruto, 2007).