A pathological increase in the amount of Aβ can induce functional and morphological changes in glial cells, including calcium dysregulation. In fact, microglia and astrocytes are activated close to senile plaques to internalize and break down Aβ (Mohamed and Posse de Chaves, 2011). This cellular activation may result in an inflammatory response and OS, playing a dual role in the pathophysiology of AD with both detrimental and neuroprotective results. Inflammatory mediators (i.e., bradykinin) may increase intracellular calcium concentration via nicotinic receptors and PI3K–Akt pathway in cultured astrocytes (Makitani et al., 2017). Tau has also been connected with astrocytes in AD, as Aβ was shown to bind the CaSR in human astrocytes, activating intracellular signaling which induced the production and release of phosphorylated Tau (Chiarini et al., 2017).

Amyloid beta has been shown to disrupt gliotransmission by enhancing calcium signaling in astrocytes (Lee et al., 2014). This calcium/gliotransmission alteration could underlie an important role of astrocytes in AD pathology. Actually, astrocytic calcium levels are abnormal in several models of AD as both acute and chronic exposure to Aβ elevates baseline calcium levels in cultured astrocytes (Haughey and Mattson, 2003; Alberdi et al., 2013; Lim et al., 2013). This calcium is partially released from intracellular sources such as the endoplasmic reticulum (Toivari et al., 2011). In addition, Aβ interacts with several types of surface receptors in astrocytes which leads to calcium entry, including purinergic receptor P2Y1 (Delekate et al., 2014), nicotinic receptors (α7-nAChRs) (Xiu et al., 2005; Lee et al., 2014), and glutamate metabotropic receptor mGluR5 (Grolla et al., 2013; Ronco et al., 2014). For instance, hippocampal astrocytes exposed to Aβ increased the frequency of NMDA receptor-mediated slow inward currents, together with calcium elevations mediated through α7nAChR activation (Pirttimaki et al., 2013). Aβ-induced dysfunction of NMDA receptors in astrocytes disrupts neuron–glial signal transmission with dramatic consequences on neuronal homeostasis, synaptic transmission, and plasticity. Therefore, neurotoxicity and selective neurodegeneration may be explained by Aβ simultaneous interaction with several receptors and neurotransmitter systems in the context of astrocyte calcium dysregulation.

In AD, it has been shown that Aβ can interrupt glutamate uptake capacity and astrocytic calcium signaling (Vincent et al., 2010; Matos et al., 2012). Also, an increase in the expression of astrocytic Tau from aged transgenic animals leads to a decline in GLT activity and therefore in subsequent neurodegeneration (Komori, 1999; Dabir et al., 2004). Some studies have demonstrated in ex vivo astrocyte preparations that Aβ_1-42 decreases the expression of GLT-1 and GLAST, two major GLTs in astroglia, via adenosine A2A receptors (de Vivo et al., 2010; Matos et al., 2012). Therefore, disruption in the clearance of excitatory neurotransmitters and increased levels of Aβ and Tau from astrocytes seem to be involved in the neuronal excitotoxicity observed in AD.

Glutamate NMDA and AMPA receptors have been related to the physiopathology of AD (Parameshwaran et al., 2008). Different studies have identified the expression of functional NMDA receptors in astrocytes (Kommers et al., 2002; Lalo et al., 2006) involved in cerebral vasodilatation, synaptic transmission, and neuronal–glial signaling (Verkhratsky and Kirchhoff, 2007; Palygin et al., 2010; Parfenova et al., 2012). Hence, Aβ-induced dysfunction in the expression and function of glutamate receptors in astrocytes, mainly in NMDA receptors, can interfere with neuronal–glial communication (Mota et al., 2014). The cellular excitotoxicity produced by the excessive stimulation of NMDA receptors in neurons and astrocytes has been shown to be reduced with the use of MK801 and memantine (NMDA receptor antagonists) (Lee et al., 2010). Furthermore, due its possible therapeutic role in neurodegenerative diseases including AD, a recent antagonist (UBP141) with preferential effects on astroglial NMDA receptors has been developed (Palygin et al., 2010, 2011). A better comprehension of the differences between neuronal and glial NMDA receptors may provide key elements for the development of novel therapeutics which primarily or selectively target astrocytic function. As well, Aβ can induce glutamate release from astrocytes resulting in an extrasynaptic activation of NMDA receptors. In this case, nitromemantine, which selectively inhibits extrasynaptic NMDA receptors, may protect against Aβ-induced synaptic dysfunction in the hippocampus (Talantova et al., 2013). Additionally, nitromemantine may prevent the synapse-destroying effects of Aβ/α7-nAChR signaling (Dal Prà et al., 2015).

Therefore, using astrocytic signaling as a possible target for drug development may have a therapeutic function in AD’s prevention and control. The antiepileptic drug levetiracetam has shown to reverse synaptic dysfunction as well as memory and learning deficits in human APP (hAPP) transgenic mice (Sanchez et al., 2012). Moreover, a retrospective observational study has shown clinical benefits of levetiracetam in early AD (Vossel et al., 2013). One way this drug may act is increasing glutamate and GABA transporters in astrocytes (Ueda et al., 2007). Chronic administration of levetiracetam may attenuate glutamate excitotoxicity and increase inhibitory neurotransmission. This molecular mechanism involving astrocytes may result in a reduction of cognitive abnormalities in AD.

Astrocytes also participate in the degradation and removal of Aβ as they express different types of proteases involved in the enzymatic cleaving of Aβ. The metalloendopeptidases NEP, IDE, and ECE1 and ECE2 have been reported to be expressed in astrocytes, and are involved in the degradation of monomeric Aβ species (although NEP also hydrolyze oligomeric forms) (Mulder et al., 2012; Ries and Sastre, 2016). It has been proposed that the modification from “natively folded-active” to “aggregated-inactive” form of IDE and NEP may be a relevant pathological mechanism in late-onset AD (Dorfman et al., 2010). Astrocytes also express and secrete several MMPs, including MMP-2 and MMP-9, which degrade both monomeric and fibrillar extracellular forms of Aβ (Ries and Sastre, 2016). Furthermore, it was found in APP/presenilin 1 transgenic mice that astrocytes surrounding Aβ plaques increased the expression of both MMP-2 and MMP-9 (Yan et al., 2006; Yin et al., 2006).

Apolipoprotein E is primarily produced by astrocytes in the CNS and has been proposed to play a major role in AD. In mice, ApoE(^-/-) astrocytes have been shown to fail to respond or internalize Aβ deposits to the same extent as do wild-type astrocytes (Koistinaho et al., 2004). As well, mice astrocytes expressing the ApoE 𝜀4 allele were less effective eliminating Aβ plaques than those astrocytes expressing the ApoE 𝜀3 allele (Simonovitch et al., 2016). Astrocytes derived from human induced pluripotent stem cells (iPSC) which expressed the ApoE 𝜀4 allele failed to support neuronal neurotrophic functions such as survival and synaptogenesis (Zhao et al., 2017). As the presence of ApoE 𝜀4 allele is considered a major risk factor in AD while the presence of ApoE 𝜀2 allele is considered a protective factor, a differential regulation of these isoforms regarding the presence of Aβ and associated responses such as neuroinflammation has been proposed (Dorey et al., 2017).

One of the most important research areas in AD is related to the development of biomarkers. Although several types of biomarkers have been explored, there is still not one that specifically diagnose, differentiate, and predict the rate of decline between populations of cognitively healthy/preclinical dementia, mild cognitive impaired and AD individuals (Fiandaca et al., 2014; Salvatore et al., 2015; Huynh and Mohan, 2017). Due to their fundamental role in CNS homeostasis, astrocytes could be considered as possible targets for tracking and studying in vivo changes in AD as well as serving as a biomarker for the disease.

L-Deprenyl is a selective inhibitor of the enzyme MAO-B, predominantly found on astrocytes (Levitt et al., 1982). This compound has been successfully used in vitro and in vivo to study the distribution and activity of MAO-B through different techniques including quantitative autoradiography and positron emission tomography (PET) (Kumlien et al., 1992; Arakawa et al., 2017). In postmortem samples from individuals with AD, the activity of both MAO-B and the binding of [3H]L-deprenyl was found to be increased in many brain regions (Jossan et al., 1991). In addition, MAO-B has been found to be increased during reactive astrocytosis in neurodegenerative conditions (Ekblom et al., 1993, 1994). As astrocytes produce MAO-B and this enzyme is increased during reactive astrocytosis, which is a process observed in AD, it seemed plausible to use L-dyprenyl as a marker of astrocytosis in this condition.

Accruing evidence seems to support the use of L-deprenyl in AD. A comparative study using PET between one of the currently accepted biomarkers for AD, the 11C-Pittsburgh Compound B, and (11)C-deuterium-L-deprenyl, concluded that the latter provided non-redundant information on both functional and pathologic aspects of the disease (Rodriguez-Vieitez et al., 2016a). Furthermore, L-deprenyl has provided valuable information about the stage of progression of AD. In a human study, the highest binding for (11)C-deuterium-L-deprenyl was observed in Braak I–II (initial AD stages), whereas it decreased with the most advanced Braak stages (Gulyás et al., 2011). Similar results have been obtained in other studies using (11)C-deuterium-L-deprenyl, where astrocytosis is prominent at the initial phases, even at preclinical stages, and then declines as the disease progresses (Carter et al., 2012; Schöll et al., 2015; Rodriguez-Vieitez et al., 2016b). In addition, a similar laminar binding pattern for tau and [3H]L-deprenyl at the temporal lobe was recently demonstrated, suggesting tau deposits and astrocytic inflammatory processes are closely related in AD (Lemoine et al., 2017). All these results point to an early contribution of astrocytes in AD pathology.

Neurons and astrocytes work in a coordinated way throughout different metabolic pathways to synthesize and release glutamate and GABA (Bak et al., 2006). At inhibitory synapses this pathway is called the GABA–glutamine cycle and it depends on GABA transporters and a multi-enzyme machinery that coordinates this process (i.e., GABA transaminase, glutamate decarboxylase, and glutamine synthetase) (Bak et al., 2006; Hertz, 2013). In AD, the processes related to GABA–glutamine cycle and GABA release from astrocytes seem to be altered. The glutamine–glutamate/GABA cycle consists of the transfer of glutamine from astrocytes to glutamatergic and GABAergic neurons. This process depends on glutamine synthetase and the tricarboxylic acid cycle (Walls et al., 2015). A reduction in pyruvate carboxylation, glutamine levels, and tricarboxylic acid cycle turnover in GABAergic neurons and astrocytes was shown in the transgenic rat AD model, McGill-R-Thy1-APP (Nilsen et al., 2014). Similarly, reduced expression of glutamine synthetase in postmortem AD brain samples indicates a profound alteration in neurotransmitter and protein synthesis, as well as metabolic dysfunction (Robinson, 2000). Astrocytes may produce and release GABA, with a main role on hippocampal synaptic plasticity function during memory processing. Increased activity of glutamate decarboxylase enzyme was found in glial synaptosomes obtained from the cortex of APP/TS1 transgenic mice, suggesting Aβ plaques stimulate GABA synthesis from astrocytes (Mitew et al., 2013). Reactive astrocytes from APP/PS1 transgenic mice have also been shown to produce GABA involving MAO-B, and release it through the bestrophin 1 channel, in an aberrant manner (Jo et al., 2014). In the same study, the suppression of GABA production or release from astrocytes completely restored the cognitive deficits and impairments in synaptic plasticity observed in the mice. Under physiological conditions, astrocytic GABA exerts a disinhibitory action at the perforant path to dentate gyrus neurons via GABA_B receptors on interneurons. However, in the APPswe/PSEN1dE9 mice, it has been shown an inhibitory action of astrocytic GABA by targeting GABA_A receptors in glutamatergic terminals (Yarishkin et al., 2015). These results provide a useful specific GABAergic target aimed at memory impairment reduction in AD. Alterations in the metabolic functions of astrocytes and consequently in glutamate and GABA–glutamine cycles may help explain cognitive disorders in AD (Le Prince et al., 1995; Robinson, 2000; Nilsen et al., 2014). Neurotransmitter transporters and effectors together with GABA-metabolizing enzymes are of special interest in drug development regarding therapeutical options for GABA-related neurological dysfunctions such as AD (Sarup et al., 2003; Nava-Mesa et al., 2014; Mutis et al., 2017; Sánchez-Rodríguez et al., 2017). Although special attention should be taken regarding the differential functional roles of neuronal and glial neurotransmitter transporters and overlying GABA/glutamate metabolic pathways in the development of high selective cell-specific drugs, in order to avert pharmacological interactions and unexpected side effects.

The metabolic cooperation between astrocytes and neurons is essential to the brain functioning. The energy metabolism of neurons depends on blood oxygen supply but also on astrocytic glucose transporters, mainly GLUT1 (Morgello et al., 1995). In addition, astrocytes may convert glycogen to lactate during periods of higher activity of the nervous system (Falkowska et al., 2015). Both in vivo and in vitro studies indicate that astrocytes participate in the regulation of cerebral blood flow according to neuronal activity and metabolic demand (Magistretti and Pellerin, 1999; Magistretti, 2006). Therefore, astrocytes are key to guarantee an adequate coupling between brain activity and metabolic supply. Several studies have shown reduced cerebral glucose metabolism in early stages of AD and correlation with symptoms severity (Desgranges et al., 1998; Mosconi et al., 2005, 2006, 2008). As mentioned early, Aβ affects neuronal excitability and it also may reduce astrocytic glycolytic capacity (Soucek et al., 2003; Schubert et al., 2009) and reduce the neurovascular unit function (Acosta et al., 2017; Kisler et al., 2017). Moreover, reductions in GLUT1 and lactate transporters in astrocyte cultures derived from transgenic AD mice have been reported (Merlini et al., 2011). In AD, the resulting metabolic dysfunction may alter the overall oxidative neuronal microenvironment (Mosconi et al., 2008). The chronic sustained effect of diminished lactate supply, increased neuronal activity, and reduced neurovascular coupling, underlines the OS increase during AD. Therefore, astrocytes are crucial players either acting as protectors against OS or participating in the progression of AD. The specific role of astrocytes on inflammatory response and OS damage will be reviewed in the next sections.

Amyloid beta has been reported to interact with numerous cellular receptors and astrocytes express a large amount of them, including TLRs, scavenger receptors, glycoprotein receptors, lipoprotein receptors, RAGE, acetylcholine receptors, complement and chemokine receptors, T-cell receptors, and mannose receptor, among others (Farfara et al., 2008). Binding of Aβ to different types of receptors seems to depend on the Aβ peptide form (monomer or fibrillar). For example, IR- and SEC-R bind monomeric forms of Aβ, scavenger receptor CD36 and glycoprotein receptors lactadherin, and CD47 prefers fibrillary Aβ, while RAGE, ApoE, and nAChR α7nAChR bind both monomer and fibrillar forms (Verdier et al., 2004). The specific outcome of all these complex Aβ-astrocytic interactions is still under research, as the precise intracellular and intercellular communication changes prompted by the different types of Aβ acting on these receptors is yet to be elucidated. Despite the gaps in knowledge, the activation of some receptors, in particular RAGE, has been reported to induce proinflammatory changes in astrocytes when exposed to Aβ (González-Reyes and Graciela Rubiano, 2016).

Advanced glycation end products receptor is a multiligand pattern-recognition receptor, member of the immunoglobulin superfamily with a variety of isoforms present in brain cells (Ding and Keller, 2005). In addition to Aβ and several AGE, RAGE can bind DNA-binding protein HMGB1/amphoterin (Hori et al., 1995) and S100/calgranulins (Hofmann et al., 1999). The main intracellular pathway activated through RAGE is the NF-κB pathway (Tóbon-Velasco et al., 2014), although it can also activate other downstream pathways including Cdc42-Rac, p21ras and MAPK, JNK, and ERK (González-Reyes and Graciela Rubiano, 2016). Furthermore, astrocytes have been reported to adopt a phagocytic profile capable of engulfing Aβ, mediated by CD36, CD47, and RAGE receptors (Jones et al., 2013). Apart from Aβ, the interaction of astrocytic RAGE with other ligands, such as S100B, may also be involved in AD neuroinflammation (Cirillo et al., 2015). These findings point to the interaction of RAGE/NF-κB pathway in astrocytes as an important factor in the development or maintenance of inflammation in AD.

The transcription factor NF-κB is currently considered as an important agent related to neuroinflammation in AD (Shi et al., 2016). NF-κB is known to be mainly activated by two pathways, the canonical (or classical) and the non-canonical (or alternative) (Nakajima and Kitamura, 2013). The canonical pathway involves activation of various receptors including RAGE and cytokine receptors, such as TNF receptor, IL-1 receptor, and the TLR family. These will induce the further activation of many downstream agents, in special IKKs alpha (IKKα) and beta (IKKβ), and NEMO (in charge of the degradation of the cytoplasmic inhibitor IKBα), and the subsequent complexes (mainly RelA) that act as transcription factors in the nucleus (Marcu et al., 2010; Wan and Lenardo, 2010). The non-canonical pathway, also known as the NEMO-NF-κB-independent pathway, occurs when NF-κB is activated by specific recruitment of TRAF2 and TRAF3, and involves p52 and RelB (Morgan and Liu, 2011). Both canonical (Wang et al., 2013) and non-canonical (Akama and Van Eldik, 2000) activation of NF-κB has been observed in astrocytes stimulated with Aβ, but still is not clear which type predominates in AD or if a differential NF-κB activation is related to the stage of the disease. It has been reported that most of the cytokines and chemokines produced by non-stimulated and activated astrocytes are direct targets of the NF-κB pathway, suggesting a central role of this factor in the proinflammatory (neurotoxic) and immunoregulatory (neuroprotective) actions of astrocytes in the CNS (Choi et al., 2014). Also, NF-κB is involved in other functions such as neuronal survival, differentiation, apoptosis, neurite outgrowth, and synaptic plasticity, all found to be altered in AD (Mémet, 2006).

In astrocytes and microglia, the activation of NF-κB due to Aβ stimulation leads to the production of the pro-inflammatory cytokines IL-1β, IL-6, iNOS, and TNFα (Bales et al., 1998; Akama and Van Eldik, 2000; Hou et al., 2011). In rats treated with Aβ_1-42 oligomers, it was shown that COX-2, IL-1β, and TNFα were expressed in reactive astrocytes surrounding the Aβ-injection site and in nearby blood vessels, as well was found co-localization of NF-κB proteins with GFAP and COX-2 (Carrero et al., 2012). In primary astrocytic and mixed astrocytic-neuronal cell cultures from rats, the use of minocycline, an anti-inflammatory agent, reduced astrocytic inflammatory responses together with a decrease in neuronal loss, caspase-3 activation, and caspase-3-truncated Tau species in neurons (Garwood et al., 2011). Minocycline has been shown to inhibit the NF-κB signaling pathway in spinal rat astrocytes (Song et al., 2016). Other reports of beneficial outcomes due to regulation of the NF-κB signaling pathway in astrocytes were reviewed by Colombo and Farina (2016). Although an exaggerated neuroinflammatory response is observed in AD, an absolute suppression of the NF-κB signaling pathway may be undesirable and even worsen the pathological condition. In APPswe/PS1dE9 transgenic mice, the suppression of NF-κB attenuated astrogliosis in the hippocampus and cortex of the animals but increased the amount of Aβ_1-42, suggesting a role of astrocytic-mediated neuroinflammation in the clearance of Aβ (Zhang et al., 2009). As well, the clinical evidence for the use of non-steroidal anti-inflammatory drugs (NSAIDS) in AD patients has not proven to be of benefit (Aisen et al., 2008; Szekely et al., 2008; Beeri et al., 2012; Alzheimer’s Disease Anti-inflammatory Prevention Trial Research Group, 2013).

Astrocytes not only are activated and induced to release chemokines and cytokines in the presence of Aβ, these cells also react to the presence of pro-inflammatory cytokines and even increase the production of Aβ in response. Therefore, the presence of inflammation is capable of increasing the production of Aβ. In addition, the development of neuroinflammation has also been related to cognitive changes in AD (Westin et al., 2012; Echeverria et al., 2016; Laurent et al., 2017).

Neuroinflammation in AD is characterized by the accumulation of cytokines such as IL-1β, IL-6, TNF-α, or TGF-β, which can contribute with cerebral amyloid deposition, augmentation of APP expression, Aβ formation, and subsequent recruitment and activation of microglial cells (Esler and Wolfe, 2001). In general, TNF-α, IL-1β, IFN-γ, L-6, and TGF-β are able to stimulate β-secretase and γ-secretase enzymatic activity through a JNK-dependent MAPK pathway, which cleaves APP and initiates Aβ formation (Liao et al., 2004). Astrocytes express and respond to a large scope of cytokines and chemokines suggesting a central role in the inflammatory-induced production of Aβ.

A study reported that a systemic immune challenge in wild-type mice during late gestation induced the development of AD-like pathology during aging, with animals displaying increased levels of hippocampal APP and altered Tau phosphorylation, together with microglia and astrocytic activation (Krstic et al., 2012). Additionally, it was shown that LPS-induced systemic inflammation in mice could contribute to cognitive impairment and increased expression of APP and Aβ_1-42, associated with increased production of inflammatory mediators such as COX-2, IL-1, and iNOS (Lee et al., 2008). The same paper reported that these changes were accompanied with astrocytic activation. Other studies have found evidence, both in human and murine models, that inflammation induces the expression of Aβ. Primary astrocytes from mice, treated with a combination of TNFα and INF-γ, significantly increased levels of BACE1, APP, and Aβ_1-40 (Zhao et al., 2011). In human primary astrocytes, treatment with INF-γ in combination with either TNFα or IL-1β induced the secretion of Aβ_1-40 and Aβ_1-42 (Blasko et al., 2000). Furthermore, TGF-β1 was found to induce overexpression of APP in astrocytes but not in neurons (Lesné et al., 2003). Cytokines seem to act on the 5′-untranslated region (5′-UTR) of the APP gene in astrocytes (Lahiri et al., 2003).

On the one hand, it seems that the presence of Aβ is able to induce the production and release of pro-inflammatory cytokines and chemokines from astrocytes, which could as well act in an autocrine manner to further induce the production of Aβ from astrocytes and possibly other cells. On the other hand, these results suggest that inflammation may be present at early stages (pre-clinical) of the disease or even that inflammation may be responsible for the appearance of pathological Aβ production and accumulation. Under both circumstances, astrocytes appear to be deeply involved in inflammatory changes observed in AD.

Other factors related to astrocytes may contribute to the appearance or enhancement of neuroinflammation in AD. For example, the presence of elevated glucose levels (as found in diabetes) has been shown to increase neurotoxicity and the release of pro-inflammatory cytokines from primary human astrocytes (Bahniwal et al., 2017). A relation between AD and diabetes/metabolic syndrome has been explored previously (González-Reyes et al., 2016), as both Aβ and AGE bind to RAGE in astrocytes. As well, pro-inflammatory signaling in astrocytes may involve changes in the expression of the calcium-dependent phosphatase CaN, which has been shown to interact with another transcription factor involved in inflammatory responses, the NFAT (Acosta et al., 2017). Enhanced nuclear accumulation of CaN/NFAT was observed in human AD hippocampus and astrocytic cultures treated with Aβ (Abdul et al., 2009). In addition, it has been reported that Aβ deregulates calcium homeostasis via CaN and its downstream target NF-κB, as well as increasing NF-κB-dependent expression of mGluR5 and IP3R2 in astrocytes (Lim et al., 2013). Changes in mGluR5 and IP3 receptor expression have been reported in astrocytes surrounding amyloid plaques in a genetic mouse model of AD (Norris et al., 2005). Another possible factor contributing to the presence of neuroinflammation is the BBB, as in AD, it has been reported that the BBB occasionally loses its integrity (Chakraborty et al., 2017). This may be explained in part thanks to the accumulation of Aβ in brain blood vessels and also due to the associated vascular inflammation, allowing crossed communication between the peripheral immune system and the brain (Takeda et al., 2014). As astrocytes have a very important interaction with the BBB and its functional components are plausible to consider their involvement in BBB-associated neuroinflammatory changes in AD.

Uncontrolled neuroinflammation is a critical element in the progression of AD, impairing the normal function of the CNS. Astrocytes, together with microglia, are the main cells involved in the inflammation/immune responses of the CNS. The presence of Aβ activates different astrocytic cell receptors, mainly RAGE, inducing the activation of the inflammatory pathway NF-κB responsible for the transcription of numerous pro-inflammatory cytokines and chemokines in astrocytes. In addition, the presence of pro-inflammatory cytokines such as IL-1β can act on astrocytes stimulating the production of Aβ and perpetuating a pro-inflammatory profile in astrocytes. Astrocytes are key in the maintenance of the homeostatic balance of the CNS and use the mechanism of reactive astrogliosis as a defensive reaction (Pekny et al., 2016), therefore is fundamental to understand the pathophysiological process that causes astrocytes to convert from a protective agent into a cell that produces a maladaptive astrogliosis response in AD (Table 1).

The development of OS in AD has been related to mitochondrial dysfunction, leading to superoxide overproduction ending in synaptic damage (Friedland-Leuner et al., 2014; Bhat et al., 2015). Mitochondrial dysfunction in AD seems to be linked to the increased presence of ROS and RNS (Islam, 2017). Müller et al. (2010) observed a decreased mitochondrial potential in transgenic Thy1-APP751SL mice. The same authors reported that increased intracellular Aβ production might trigger mitochondrial dysfunction quite early and independently of Aβ plaques and, that the accumulation of these alterations with aging lead to disruption of respiratory chain complexes (mainly III and IV) and significant reduction in the generation of NADH. The authors suggested that progressive increase in oxidant production together with a decrease in antioxidant components may conduce to the loss of brain homeostasis observed in AD. However, in AD, it has been demonstrated that prior to the appearance of senile plaques, brains present glucose hypometabolism due to abnormal oxidative metabolic routes in the mitochondria, which also induce increased ROS production and subsequent oxidative cell damage (Maruszak and Żekanowski, 2011). Additionally, variants of gene expression profiles in AD have shown downregulated expression of NeuroD6, which encodes a transcription factor involved in triggering antioxidant responses and the maintenance of the production of mitochondrial antioxidants (Uittenbogaard et al., 2010; Fowler et al., 2015). Also, RNA-Seq and microarray data analysis indicated a consistent downregulation of NeuroD6 in brains of individuals with AD, suggesting downregulation of NeuroD6 as a possible biomarker for AD (Satoh et al., 2014).

The role of ROS and OS in neurodegenerative diseases, including AD, is not entirely clear, although it has been observed that a modest level of oxidative RNA damage occurs during the process of aging in brain neurons, but a prominent level of oxidative RNA damage is present in vulnerable neurons which correspond to the earliest stage of cognitive decline in the transition from cognitively normal aging to AD (Nunomura et al., 2012). Furthermore, DNA bases are vulnerable to OS damage involving hydroxylation, protein carbonylation, and nitration in AD (García-Blanco et al., 2017). Changes in oxidative markers have been reported in brain regions such as hippocampus and inferior parietal cortex, which are also compromised in AD (Floyd and Hensley, 2002).

Brain is considered to be especially vulnerable to OS and susceptible to lipid peroxidation because of its high lipid and poly-unsaturated fatty acids content, and its low concentrations of antioxidants (Butterfield et al., 2013). In AD, neurotoxic effects of Aβ induce OS through lipid peroxidation, protein degradation, and amino acid oxidation, which in turn increase the production of ROS and RNS by positive feedback (Swomley and Butterfield, 2015). Alkenals, 4-hydroxynonenal (HNE), and 2-propenal (acrolein) are reactive products obtained from lipid peroxidation induced by Aβ. These agents can modify covalently some amino acids residues or change protein conformation, which in turn affects its function. Thus, a coupling between increased lipid peroxidation and structural modification of GLT-1 has been proposed, explained by increased HNE binding due to excessive Aβ_1-42 (Butterfield et al., 2002). These events can compromise astrocyte function, inducing glutamate transport inhibition and increasing excitotoxicity to neurons in AD.

Astrocytes seem to be involved in the processes leading to the appearance or maintenance of OS in AD. Abramov et al. (2004a), working in rat astrocytes, have shown that Aβ treatment increases intracellular-free calcium influx from the extracellular space, and induces changes in mitochondrial functions. These changes are associated with the activation of NOX due to Aβ interaction on the membrane, which, in turn, induces [Ca²⁺]i changes. The main changes leading to mitochondrial dysfunction in astrocytes are associated with mitochondrial depolarization, increased conductance, and presence of mitochondrial permeability transition pores (Duchen, 2000). A possible mechanism which explains why calcium influx is induced by Aβ into astrocytes could be related with the formation of calcium selective channels on the membrane, which seem capable of generating a different conductance (Abramov et al., 2004b). These channels have been shown to be formed by insertion of Aβ peptides in the membrane, and also are arranged in a structural configuration which requires a lesser content of cholesterol on the lipidic membrane (Kawahara and Kuroda, 2001; Arispe and Doh, 2002; Arispe et al., 2007).

Additionally, it has been found that Aβ_1-42 oligomers are key factors on the induction of OS stress by astrocytes. Aβ_1-42 oligomers binding to RAGE on astrocytes induce ROS production via NOX complex activation (Askarova et al., 2011). However, astrocytes are also able to trigger ERK1/2 pathways and cytosolic phospholipase A2 phosphorylation, independent of NOX activation, which in turn causes mitochondrial dysfunction by decreasing mitochondrial membrane potential, enhancement of NOX activity, and overproduction of ROS (Zhu et al., 2006). Astrocytes seem to be a primary target of Aβ, as this peptide induces various effects related to OS such as altered intracellular calcium signaling and calcium-dependent reduction in astrocytic GSH (Abramov et al., 2004b). Although this GSH depletion affects astrocytes, neurons are the cells that show a higher rate of cell death, suggesting that neurotoxicity reflects the neuronal dependence on astrocytes for antioxidant support. This could better be explained by the fact that astrocytes are the producers of the primary elements required for the production of GSH in neurons, such as glycine and cysteine (Abramov et al., 2003; Gandhi and Abramov, 2012). Regarding GSH, it has also been shown that in cultured astrocytes, a prolonged incubation with Aβ reduces induction of the transporter ABCC1 which is the main pathway for GSH release (Ye et al., 2015). Astrocytes play a central role in maintaining the neuronal integrity, nevertheless cytokines and neurotransmitters released by damaged or activated astrocytes may increase the neurotoxicity and vulnerability of neurons. During chronic OS, as observed in AD, the crosstalk communication between astrocytes and neurons is impaired resulting in disrupted memory consolidation. This compromise in memory formation is probably due to calcium overload and activation of MAPK pathways in astrocytes, which involve as well the JNK/SAPK pathways, and may conduce to anomalous deleterious signaling, including autophagic astroglial signals and apoptosis (Ishii et al., 2017).

During neuroinflammation, increased concentrations of ROS/RNS may lead to the activation of the transcription factor NF-κB which induces the overexpression of NO synthases in astrocytes and microglia, in particular NOX2 and iNOS, resulting in peroxynitrite production by superoxide and NO reaction producing neuronal damage (Saha and Pahan, 2006; Brown, 2007; Morgan and Liu, 2011). Moreover, NF-κB activation induces the expression of COX-2 and cytosolic phospholipase A2, which in turn stimulate the generation of prostaglandins, promoting inflammation and OS (Hsieh and Yang, 2013). Castegna et al. (2003) reported that the formation of peroxynitrite ONOO^- leads to protein nitration in enzymes, such as alpha and gamma enolases, implicated in brain glucose metabolism. Thus, the signaling pathway NF-κB, which is also heavily involved in inflammatory reactions, has been proposed to be involved in OS, as a direct crosstalk between ROS and NF-κB has been reported (Turillazzi et al., 2016). In AD, the presence of chronic OS alters the protective physiological role of the NF-κB transcription pathway, which normally promotes cell survival and prevents apoptosis and necrosis, through modulation of the JNK signaling pathway (Morgan and Liu, 2011). In astrocytes, it was reported that under certain conditions, IL-1β may act stimulating astrocytic GSH production, and potentially, augmenting total antioxidant capacity in the brain, via an NF-κB-dependent process (He et al., 2015). In this way, NF-κB pathway has been associated with both pro-oxidant and antioxidant roles. In AD, an alteration of this pro- and anti-oxidant role of NF-κB in astrocytes seems to be present, tending toward a pervasive pro-oxidative and pro-inflammatory profile (Table 2).

Novel anti-Alzheimer’s drugs will need to consider the selective modulation of astrocyte activity in order to reduce pro-inflammatory signaling as well as to attenuate OS and diminish excitotoxicity (Figure 1). Taking into account the complex physiopathology of AD, a deep knowledge about dysfunctional astrocyte intracellular pathways evoked by Aβ opens the possibility for the design of new effective multi-target directed drugs.