Astrocyte alterations in neurodegenerative pathologies and their modeling in human induced pluripotent stem cell platforms

The best evidence for synaptic defects in neurodegenerative diseases comes from AD studies. In AD, synapse dysfunction is one of the most striking early pathological features. There is a progressive loss of synapses during the disease course, and in fact, the extent of synapse loss is the best correlate with cognitive decline [14]. The exact mechanisms behind the synapse loss and synaptic defects still remain unclear, although several theories involving beta-amyloid (Aβ) plaques, complement cascade pathway, and microglia have been proposed [15] [16]. In addition, the importance of astrocytes in AD and synaptic defects is currently strongly emerging and this review will address some of the recent advances supporting it. For example, astrocyte-derived cholesterol has been shown to be a crucial factor in promoting synaptogenesis [9]. Apolipoprotein E (apoE), the major cholesterol carrier in the brain, is mainly synthesized and secreted by astrocytes. APOE polymorphisms are the main genetic determinants for AD with the APOE4 allele being associated with increased AD risk [17]. In mouse studies, APOE4 astrocytes showed decreased rate of synapse pruning and turnover in the brain, leading to the accumulation of senescent, non-functional synapses, when compared to APOE3 astrocytes [18]. This could be linked to the increased synaptic vulnerability in AD. Zhao and colleagues used human iPSC-derived neurons and astrocytes in a transwell-based co-culture system to study the effects of APOE4 astrocytes on control neurons [19]. They found that APOE4 astrocytes were less capable of promoting synapse formation and the expression of synaptic proteins was decreased in comparison with APOE3 astrocytes. Furthermore, APOE4 astrocytes reduced the survival of neurons and eventually led to neurodegeneration. These data support the importance of astrocytes and APOE genotype in synaptic deficiencies and neurodegeneration linked to AD. Another astrocyte-secreted synaptogenic molecule, thrombospondin-1 (TSP-1), has decreased expression level in the brain of AD patients [20]. On the other hand, treatment with TSP-1 was shown to reverse the Aβ-induced synaptic dysfunction in vivo.

Intriguingly, astrocytes were very recently found to engulf and phagocytose dystrophic neurons and pre-synaptic elements associated with amyloid plaques in APP/PS1 mutant AD mice as well as in post-mortem human brains [21]. This result supports the role of astrocytes in synaptic regulation and phagocytosis also in the human brain, although the exact meaning for disease pathogenesis remains to be elucidated. It is hypothesized that the astrocytic clearance of dystrophic neurites and dysfunctional synapses is inadequate in AD, resulting in their accumulation and impaired neural networks. In mice, the rate of synapse pruning decreases upon aging leading to the accumulation of senescent synapses, a possible explanation behind the cognitive decline [4, 13].

Reactive astrocytes in post-mortem AD patient’s brain show markedly increased expression of GABA, the inhibitory gliotransmitter. The enhanced GABA signaling has been linked with memory and learning deficits in mouse models of AD [22]. In addition, a specific type of reactive astrocytes shows diminished potential to promote synaptogenesis and synaptic function, leading to a decreased number of synapses with only weak signals and, finally, neurodegeneration [23]. This kind of reactive astrocytes seems to be abundant in post-mortem AD brain. In mouse primary hippocampal cultures, astrocytes internalized oligomeric tau, which was followed by reduced gliotransmitter release, especially ATP [24]. The reduced ATP release from astrocytes reduced the synaptic transmission of neurons and expression of both pre- and post-synaptic proteins in them.

In other neurodegenerative diseases, there is so far less evidence supporting the role of astrocytes in synaptic defects. In ALS, astrocytes have a well-established role in the degeneration of motor neurons and astrocytes with both sporadic (sALS) and familial (fALS) background seem to have similar effects when studied either in vitro or in vivo [25–28]. ALS is characterized by synaptic disconnections at the neuromuscular junction leading to axonal degeneration and, finally, death of neurons through either cell-autonomous or non-cell-autonomous mechanisms [29]. When transplanted into the mouse spinal cord, human sALS astrocytes have been shown to lead to reduced pre-synaptic connections and denervation at the neuromuscular junctions [28]. There is also a significant loss and degeneration of synapses in the frontal cortex in ALS, but this has not been clearly linked to astrocytes or astrogliosis in the post-mortem human brain [30].

In Huntington’s disease patients, reduced glutamate uptake has been detected and this is linked to astrocytes, since glutamate is cleared from the perisynaptic area mostly by astrocytes [32]. In Down’s syndrome, there is reduced level of astrocyte-secreted TSP-1 leading to decreased synaptic density and spine malformations [33]. Taken together, these findings support the importance of astrocytes, astrocyte-secreted gliotransmitters, and other factors in synapse defects and neurodegeneration, although more human studies are needed for confirmation.

Reactive astrogliosis is a common feature of astrocytes and compromises neuronal survival. This process involves both molecular and morphological changes in astrocytes and highly depends on the events triggering it. Therefore, it may have both beneficial and detrimental effects on surrounding cells [50]. Astrocytes communicate with each other through gap junction proteins, mainly connexin (Cx) 43 and Cx 30. Mice lacking Cx43 and Cx30 expression on astrocyte endfeet manifest a significant loss of AQP4 and β-dystroglycan, a transmembrane receptor anchoring astrocyte endfeet to the perivascular basal lamina. The absence of astroglial connexins weakens the BBB that opens upon increased hydrostatic vascular pressure and shear stress [51].

Accumulating data support the importance of vascular astrocytes in AD pathology [52, 53]. Astrocyte endfeet surrounding vascular Aβ deposits showed morphological changes including retraction and swelling, and reduced expression of GLUT1 and lactate transporters in transgenic arctic β-amyloid (arcAβ) mice expressing human amyloid precursor protein (APP) [54]. These changes occur at early stages of the disease and are consistent with NVU uncoupling, suggesting that astrocyte dysfunction contributes to the early behavioral and cognitive impairments. AQP4 deficient mice are unable to clear soluble Aβ efficiently, proposing that this pathway may remove Aβ from the CNS [38]. In humans, the loss of perivascular AQP4 localization was associated with increased Aβ burden and increased AD pathology (Braak stage) [55]. Furthermore, primary human astrocytes obtained from non-demented human brain tissue upregulate the expression of neprilysin and scavenger receptor B1 as a response to Aβ–APOE complexes. This function appears to be defective in astrocytes derived from AD brain [56]. APOE4, which in CNS is mainly secreted by astrocytes, accelerates pericyte loss and microvascular reduction by enhancing pro-inflammatory pathway both in pericytes and ECs [57]. These findings suggest that Aβ clearance of astrocytes is impaired in AD due to astrocyte polarization induced by brain Aβ deposition [58, 59]. Nevertheless, the contribution of Aβ-induced changes to BBB is still unclear.

AD patients’ brains are characterized by high prevalence of degenerated endothelium and impairment of BBB transport system [34, 60]. BBB breakdown in the hippocampus precedes neurodegeneration [61, 62] and follow-up studies in AD patients have confirmed BBB leakage in both gray and white matter regions due to a loss of tight junction barrier function [61, 63, 64]. Moreover, cognitive decline was associated with stronger BBB leakage in patients with an early AD, suggesting that BBB damage might be part of a cascade of pathologic events that eventually leads to dementia [61]. In aging brain, altered AQP4 expression and its redistribution from astrocytic endfoot membranes to non-end-foot membranes compromise Aβ clearance through the glymphatic system and likely contribute to AD pathology [55].

Initially, it was assumed that the BBB stays unaltered during the development of PD, but several studies with human and animal models have linked BBB dysfunction with the development of PD pathology. In studies with PD patients, the increased permeability of the BBB was associated with decreased expression of P-glycoprotein demonstrated by increased uptake of [¹¹C]-verapamil in the midbrain [65] and frontal white matter regions in PD patients [66]. Taking into account that P-glycoprotein is an active drug transporter, it is possible that the reduction of P-glycoprotein may be associated with accumulation of neurotoxins and aggregated proteins in the PD brain affecting directly dopaminergic neuron survival [67]. In addition to human studies, several toxin-induced animal models, including 6-OHDA treated rats and MPTP-treated mice, have shown a disruption of the BBB that is linked to the loss of dopaminergic neurons [68, 69]. It has also been shown that activation of microglia and astrocytes in PD leads to the release of pro-inflammatory cytokines, including IL-6, IL-1B, and TNFα, which can increase neuronal death and rearrange tight junction protein expression on ECs [70]. Moreover, the loss of signaling interaction between astrocytic perivascular endfeet and ECs leads to hypertrophy and upregulation of GFAP and vimentin in astrocytes triggering reactive astrogliosis and weakens BBB tightness [41].

In ALS, neurodegeneration of upper and lower motor neurons in the brain and spinal cord is not completely understood. The pathogenesis of the disease is complex with recent evidence showing BBB disturbances in the brain of fALS patients that carry SOD1 mutation and in post-mortem brains of sALS patients [71]. Primary degeneration of ECs and pericytes compromising vascular barrier integrity both in the brain and spinal cord of ALS patients has been described. The BBB disruption could be the main obstacle for efficient drug delivery to the CNS due to the impairment of P-glycoprotein efflux transporter in the brain and spinal cord microvessels [72, 73]. Therefore, restoration of BBB integrity could promote delivery of therapeutics to the CNS as well as help to remove waste products from the brain, providing an improved environment for motor neuron survival. The astrocyte contribution to BBB disruption in ALS has been studied only in rodents. In end-stage SOD1, rats increased expression of AQP4 in the spinal cord gray matter and around blood vessels promoted perivascular edema-linked BBB opening in ALS [74]. As elevated AQP4 was also detected near motor neurons, it is possible that dysfunctional astrocytes contribute to further motor neuron degeneration [75]. In parallel Kir4.1 expression was decreased in the brainstem and cortex of SOD1 rats. Thus, ability of astrocytes to properly maintain water and potassium homeostasis in CNS is reduced, not only affecting the BBB but also impeding motor neuron survival in ALS. In pre-symptomatic SOD1 mutant mice, BBB breakdown has been demonstrated prior to motor neuron degeneration and inflammatory changes in the spinal cord [76]. Whether connexin-mediated changes are directly affecting BBB function in ALS, is not known yet. However, although BBB damage might not be a causative event in ALS, it is still a significant, and possibly critical, pathologic effector for the disease progression.

To better understand the role of human astrocytes in BBB functionality in neurodegenerative diseases and to develop new treatment strategies with predictive drug permeability, there is a demand for development of improved BBB models which include astrocytes and other components of human origin. Current in vitro BBB models mostly use brain microvessels and astrocytes of bovine, porcine, or rodent origin [77]. Even though the models are well established, the characteristics and functionality of the BBB in these species differ from that in humans especially in the expression level of important transporters such as P-glycoprotein [78–81]. The first human BBB models were generated from primary and immortalized human cells [82, 83]. However, the limited availability of primary human cells and lack of barrier tightness in immortalized cells underlines the need for more accessible and relevant cell source for BBB modeling. Recent advances in stem cell technology have created the opportunity to generate cells important for building human in vitro BBB [84, 85]. Stem cell sources have demonstrated a significant advantage over other cell types in respect of human origin, scalability, self-renewal, and potential to generate all BBB cell components from the same patient.

Astrocytic metabolic dysfunction has been suggested be a triggering factor for amyloid beta (Aβ) deposition [97]. While amyloid deposition has been mostly seen as a consequence of reduced Aβ clearance, increased production of amyloidogenic species may take place through metabolic regulation of beta and gamma secretase levels [98, 99], or aberrant post-translational modification of amyloid precursor protein APP [100]. It has been demonstrated that exposure of cultured astrocytes to amyloid peptides alters their glucose uptake and its downstream metabolism in parallel with increased hydroperoxide and glutathione release [101]. The pro-inflammatory environment typically present in AD brain may alter astrocytic metabolism causing increased glutamate and ATP release through connexin 43 hemichannels, which in turn could result in microglial activation and neuronal vulnerability [102, 103].

The degenerating brain areas in PD, such as SN, contain relatively less astrocytes than other brain areas, making dopaminergic neurons potentially more vulnerable to the metabolic and energy disturbances [104]. In line with this assumption, it has been demonstrated that glucose deprivation promotes α-synuclein aggregation [105], whereas increased lactate supply inhibits it [106], providing a rationale for involvement of astrocytic metabolic dysfunction in PD pathogenesis. Moreover, astrocytes isolated from PD animal models expressing mutant α-synuclein or deficient for Parkin showed functional and morphological deficits of mitochondria, which was linked to astrocytes’ inability to support neuronal differentiation [107]. In addition, expression of PD-related A53T mutant α-synuclein in astrocytes was found to result in downregulation of glutamate transporters and wide-spread gliosis [108]. In the context of PD treatment, astrocytes have been shown to convert l-DOPA to dopamine [109], and in the striatum, astrocytes serve for l-DOPA storage [110]. It is, therefore, likely that metabolic and energy disturbances in astrocytes contribute to not only development of PD, but also in deteriorating treatment responses.

Several lines of research indicate astrocytic contribution also in HD. In a mouse model of HD, a deficit of astrocytic GABA release through GAT-3 has been thought to cause reduced tonic inhibition neuronal output in the striatum [111].

Furthermore, in HD mouse models astrocytes have shown increased Ca²⁺-dependent exocytotic glutamate release which correlates with elevated biosynthesis of glutamate due to upregulation of astrocytic pyruvate carboxylase [112].

Both non-cell-autonomous and cell-autonomous mechanisms are implicated in glial cell contribution to ALS [113–115], and astrocytes appear to be major determinants of ALS progression. Animal studies indicate a toxic gain of function in ALS astrocytes, which may converge into impaired supportive capacity of astrocytes to motoneurons [26, 116]. Subsequently, the initial astrocyte cell-autonomous dysfunction may lead to the non-cell-autonomous neuronal toxicity [117, 118].

Application of metabolomics, transcriptomics, and lipidomics on ALS models indicates significant dysregulation of a number of metabolic pathways in ALS astrocytes [119–125]. Metabolomics studies on models composed from motoneurons and astrocytes expressing mutant human SOD1 revealed disruption of the TCA cycle and glutamate metabolism under oxidative stress induced by menadione [126]. Furthermore, dysregulation of purine, pyrimidine, lysine, and glycerophospholipid metabolism pathways was demonstrated in ALS astrocytes co-cultured with motoneurons. In this model, glutamate exposure caused significant alterations in the astrocyte metabolic fingerprint and associated with decreased lactate levels by affecting lactate efflux from astrocytes [127]. Moreover, these findings correlate with the data from CSF analyses of human ALS patients [120]. Importantly, ALS in human patients has been associated with hypermetabolism, possibly involving aberrant astrocytic mitochondrial metabolism [128], which is present already at the pre-symptomatic state in SOD1 mutant mouse models and may thus have a key role of in early disease pathogenesis [129].

Analysis of ALS models expressing mutant SOD1 reveals prominent astrocytic alterations linked to metabolic dysregulation due to impairment of the astrocyte lactate transport and pro-nerve growth factor-p75 receptor signaling [129]. In addition, aberrant activation of metabotropic receptor 5 mGluR5 has been demonstrated in ALS astrocytes causing persistent elevation of intracellular calcium [130]. More recently, a specific chronic activation of the α2-Na/K ATPase/α-adducin complex in ALS astrocytes has been demonstrated, resulting in biased energy metabolism and acquisition of pro-inflammatory phenotype [131].

It has been demonstrated that astrocytes also potentially regulate neuronal proteostasis by affecting autophagic degradation pathways. The ability of conditioned medium from ALS-patient iPSC-derived astrocytes to modulate the autophagy pathway neuronal cells has been linked to the decreased expression of LC3-II and concomitant increase in the expression of SOD1 [132].

Overall, the current research demonstrates that astrocytic non-cell-autonomous and cell-autonomous mechanisms including metabolic dysfunction and related changes in gene expression play a central role in pathogenesis of neurodegenerative diseases, and identification of molecular targets in astrocytes can serve for designing therapy that could prevent or slow down the disease progression.

Neuroinflammation is a major contributor to AD, driven by the activation of microglia and astrocytes. A recent study of CSF markers from several hundred individuals with mild cognitive impairment and patients with AD dementia revealed increased levels of YKL-40, ICAM-1, VCAM-1, IL-15, and Flt-1 during the pre-clinical, prodromal, and dementia stages of AD [140]. There was a correlation of these biomarkers with increased CSF levels of total tau, especially for YKL-40, a glycoprotein secreted by microglia and astrocytes. Interestingly, high astrocytic expression of IL-15 has also been shown to promote tissue damage in multiple sclerosis and following brain ischemia [141].

Aβ, the main pathological hallmark of AD, activates astrocytes to release chemokines and cytokines, and in an interaction with the increased levels of pro-inflammatory cytokines even increases the production of Aβ in neurons and astrocytes [142]. In fact, hypertrophic reactive astrocytes are strongly associated with Aβ pathology, as they accumulate around Aβ plaques, and are seen [143]. Especially, the pro-inflammatory A1 phenotype of astrocytes is seen in post-mortem tissue from AD patients. While Aβ has been demonstrated to induce a pro-inflammatory profile, and even astrogliosis in AD, it is not clear whether it is Aβ that induces this phenotype [23] [144].

In AD, neuroinflammation is characterized by the accumulation of cytokines such as IL-1β, IL-6, TNF-α, or TGF-β, which can, through activation of β- and γ-secretase activity, further contribute to cerebral amyloid deposition, augmentation of APP expression, Aβ formation, and finally to recruitment and activation of microglial cells [145].

Pro-inflammatory astrocyte activation appears to be linked to the release of synaptotoxic factors and loss of glutamate regulation. Dysregulation of Ca²⁺ homeostasis is thought to play the central role in these events. Due to the complexity of the Ca²⁺ balancing mechanisms, any malfunction in the regulatory checkpoints may contribute to the overall pathological alterations in neurodegenerative diseases. The Ca²⁺/calmodulin-dependent protein phosphatase, calcineurin has been suggested to provide a l link between Ca²⁺ dysregulation and astrocyte activation [146]. Calcineurin is highly expressed by subsets of activated astrocytes in both humans and animal models, resulting in the consecutive NFAT activation and contributing to the AD neuropathology [147].

One of the key aspects in PD pathophysiology is neuroinflammation, including presence of reactive astrocytes, in the substantia nigra pars compacta [148]. This is thought to be a downstream response to the death of dopaminergic neurons, even though there is also increasing amount of evidence, suggesting that astrocytes have an initiating role in PD pathophysiology [149].

Several genes implicated in PD pathology, including SNCA, DJ-1, PLA2G6, ATP13A2, LRRK2, PINK, and Parkin, have been demonstrated to play a role in the origination and regulation of astrocyte activation in response to inflammatory stimuli [149]. Both α-synuclein and DJ-1 have been shown to affect the astrocytic TLR4 signaling [150–152]. Furthermore, if looking at the secreted pro-inflammatory mediators, Parkin expression is known to be modified by IL-1β and TNF-α stimulation [153] and DJ-1 has been shown to regulate astrocyte activation by IFN-γ [154]. Finally, functional ATP13A2 expression prevents NLRP3 inflammasome activation [155].

Interestingly, a recent study demonstrated that post-mortem brain samples from PD patients display increased astrocytic senescence. Furthermore, paraquat, a known chemical to cause parkinsonism, induced astrocytic senescence and senescence-associated secretory phenotype with robust secretion of numerous inflammatory cytokines, chemokines, growth factors, and proteases both in in vitro and in in vivo settings [156]. Another recent report studying transplantation of neural progenitors in PD utilized co-transplantation of rat astrocytes and showed increased anti-inflammatory as well as decreased pro-inflammatory profile associated with behavioral improvement in the animals [235]. This further emphasizes the dual role of astrocytes in the neurodegeneration as astrocytes can either support the neurons or cause their degeneration.

ALS is considered to be a non-cell-autonomous disease with substantial involvement of astrocytes [157]. Post-mortem studies of ALS patients have revealed increased GFAP expression in the motor cortex, motor nuclei of the brain stem, the corticospinal tract, and the ventral horn of the spinal tract [158]. Furthermore, astrocytic production of pro-inflammatory mediators, such as type 1 interferons, TGF-β1, and prostaglandin D₂, has been demonstrated to be upregulated in ALS-patient post-mortem tissues and in ALS models based on embryonic stem cell-derived human astrocytes [113, 159, 160]. In addition, the astrocytic NLRP3 inflammasome expression has been shown to be increased, leading to augmented IL-1β production in the post-mortem tissues of sALS patients [161]. Most recently, the advances in the patient neuroimaging have allowed direct visualization of neuroinflammation in ALS. Although in the post-mortem tissues from ALS patients, astrocytosis is seen in parts of the corticospinal tract other than pons [162], PET imaging with [11C]-DED, a commonly used marker of activated astrocytes, was able to demonstrate increased uptake rate in pons and white matter [163], which may reflect corticospinal tract astrocytosis. Finally, a strong astrocytic A1-phenotype has been discovered in the motor cortex of ALS patients, further highlighting the pro-inflammatory phenotype of astrocytes in the areas affected by ALS [23].

Increased susceptibility to pro-inflammatory stimulation has been demonstrated in several mouse models of HD [164].

The enhanced inflammatory response in the brain was observed only in models with expression of mHTT including glial cells. The studies on primary astrocytes revealed increased IkB kinase activity, resulting in prolongation of NF-kB activation and upregulating pro-inflammatory factors, particularly TNF-α. Importantly, enhanced activation NF-kB-p65 was observed in astrocytes of patients and mice with HD [164]. Further attempts to block soluble TNF-α signaling using a dominant negative inhibitor XPro1595 caused decrease of TNF-α in the cortex and striatum, resulting in increased neuronal density and improved motor function. In addition, diminished amount of mutant HTT aggregates, and decreased gliosis was evident in brains of R6/2 mice [165].

Astrocytes are glycolytic in their metabolism, to the extent that they survive without active respiratory chain [175]; thus, neurons have been considered as the major ROS producers in the brain. However, recent reports have shown that astrocytes can contribute to the redox status of the brain both directly by producing ROS and indirectly by affecting neuronal metabolism. Neurons rely on astrocytes for lactate supply and diminished lactate supply from astrocytes is likely to alter the oxidative microenvironment in the brain by increasing neuronal respiratory activity and reducing neurovascular coupling [176]. In addition, upon inflammation ROS is produced by both activated microglia and astrocytes to serve in the immune response and to trigger signaling cascades [177, 178]. Several antioxidant response pathway genes are regulated by the transcription factor Nrf2, which binds to the antioxidant response element (ARE) present in these genes. In normal physiological conditions, Keap1 targets Nrf2 for ubiquitin-mediated degradation; however, in oxidative conditions, Keap1 is not able to bind Nrf2 leading to accumulation of Nrf2 in the nucleus and induction of ARE-containing genes [179, 180]. This Nrf2–ARE pathway is particularly weak in neurons, whereas astrocytes both show strong response to Nrf2 activation and have higher expression of ARE-containing genes than neurons (Fig. 1e) [181]. Isolated neurons are known to possess very limited antioxidant defenses, including low catalase and GSH levels, both regulated by Nrf2, and Nrf2 deficiency does not affect their expression [182]. Astrocytes on the other hand are much more resistant to oxidative insults. They have high capacity for production and storage of GSH and oxidative stress increases GSH release from astrocytes. The GSH released from astrocytes is degraded and the cysteine-rich products are taken up by neurons to be used as building blocks for neuronal GSH synthesis in a process regulated by Nrf2 [183, 184]. Nrf2 deficiency also increases astrocyte vulnerability to oxidative stress [182].

Despite their low antioxidant defenses but high mitochondrial respiratory activity and consequently relatively high ROS production, post-mitotic neurons survive and remain functional for decades. This is thought to be achieved through antioxidant support from surrounding astrocytes [185].

Aβ plays a key role in oxidative stress induction in AD and astrocytes seem to be involved in this. Aβ exposure in neuron–astrocyte co-cultures leads to a loss of mitochondrial membrane potential and increased ROS production in astrocytes but not in neurons [186]. The increase in ROS is due to the activation of astrocytic NADPH oxidase, as inhibition of it blocks the ROS production and protects neurons [186]. The increased ROS due to activated NADPH oxidase further activates PARP, which leads to decreased NAD levels and metabolic failure in astrocytes further leading to neuronal death [187]. These data from rodent models strongly suggest that astrocytes have a central role in oxidative stress induction in AD. Human data are still scarce; however, our recent report from PSEN1 mutant patient astrocytes showed that also human AD patient astrocytes manifest Aβ deposits and altered metabolism with increased ROS production and these result in reduced neuronal support [188].

Oxidative stress directly affects Ca²⁺-homeostasis [189], which is one of the hallmarks of astrocytes and impaired in AD. In astrocytes, increased ROS production together with Ca²⁺ leakage from ER leads to opening of the mitochondrial permeability transition pore and transient depolarization of mitochondria, resulting in loading of cellular Ca²⁺ from ER to mitochondria [190]. The transient events were not detrimental; however, they eventually lead to complete loss of membrane potential and necrotic cell death and further increase in ROS production [190].

Astrocytic activation of transcription factor Nrf2 results in upregulation of antioxidant responses spreading to surrounded neurons [191, 192]. It has been demonstrated that lentiviral Nrf2 gene transfer causes a robust reduction in astrocytic activation and induction of Nrf2 target gene heme oxygenase 1 [193]. Finally, activation of the Nrf2 pathway by small molecules has been shown to confer protection in several neurodegenerative disorders, including AD and PD [194], and studies using mice with astrocyte-specific overexpression of Nrf2 prove that this neuronal protection is mediated by astrocytes, further signaling the importance of astrocytes in oxidative stress and neurodegeneration [184, 195].

RC dysfunction in an SOD1 mutant rat model for ALS has been shown to lead to increased ROS production by astrocytes and motor neuron loss, which can be prevented by treating the astrocytes with antioxidants, suggesting a role for astrocytes in induction of oxidative stress also in ALS [196].

NF-κB transcription factors are central in inflammation and activated mainly by pro-inflammatory stimulation, which is present in many neurodegenerative diseases. However, also ROS has a strong role in regulating NF-κB signaling, and again, certain NF-κB-regulated genes are involved in regulating cellular ROS levels [197]. The most direct way for NF-κB to regulate cellular ROS levels is through activating antioxidant genes such as SOD2, SOD1, and thioredoxins [198–200]. However, as inflammation stimulates ROS production, NF-κB can activate also genes that lead to increased ROS. These include NADPH oxidase and inducible nitric oxide synthase (iNOS) [201, 202].

How ROS regulates NF-κB signaling in astrocytes is more complex. First of all, ROS has been reported to stimulate NF-κB in the cytoplasm, but inhibit it in the nucleus and many of the interactions of ROS and NF-κB are cell type and even astrocyte-specific [203]. Oxidation of cysteines is the most well characterized way that ROS affects NF-κB pathway. The initial reactions to sulfenic acid are usually reversible by antioxidants, but further oxidation leads to irreversible changes [203]. ROS can inhibit NF-κB DNA binding by direct oxidation of a cysteine of p50 [204]. The same cysteine can also be s-nitrosylated by NO produced by iNOS, which is activated by NF-κB, thus creating a negative feedback loop [205]. NF-κB activation may also further induce COX-2 and phospholipase A2 expression and promote inflammation [206].