Neuron–astrocyte interactions in neurodegenerative diseases: Role of neuroinflammation

Abstract

Selective neuron loss in discrete brain regions is a hallmark of various neurodegenerative disorders, although the mechanisms responsible for this regional vulnerability of neurons remain largely unknown. Earlier studies attributed neuron dysfunction and eventual loss during neurodegenerative diseases as exclusively cell autonomous. Although cell-intrinsic factors are one critical aspect in dictating neuron death, recent evidence also supports the involvement of other central nervous system cell types in propagating non-cell autonomous neuronal injury during neurodegenerative diseases. One such example is astrocytes, which support neuronal and synaptic function, but can also contribute to neuroinflammatory processes through robust chemokine secretion. Indeed, aberrations in astrocyte function have been shown to negatively impact neuronal integrity in several neurological diseases. The present review focuses on neuroinflammatory paradigms influenced by neuron–astrocyte cross-talk in the context of select neurodegenerative diseases.

Introduction

Many neurodegenerative diseases, including Alzheimer’s disease (AD), Parkinson’s disease (PD), amyotrophic lateral sclerosis (ALS) and neuronal ceroid lipofuscinosis, a family of inherited lysosomal storage disorders collectively referred to as Batten disease, are associated with neurodegeneration in select brain regions.^1–3 For decades, neuron loss in many neurodegenerative diseases has been traditionally viewed as a cell-autonomous process. However, recent discoveries have shown that other central nervous system (CNS) cells can contribute to neuronal injury/death in these disorders in a non-cell autonomous manner, which has challenged the neuro-centric view of physiopathology in neurodegenerative diseases.^4–9 Accordingly, intercellular communication between neurons and other CNS parenchymal populations, including astrocytes, microglia and oligodendrocytes, have been identified and the significance of these cell–cell communications in the context of neuroprotection or neurodegeneration has been extensively investigated.^4–9 The present review focuses on the role of astrocytes in dictating neuronal health versus pathology, as other contributions in this special issue are dedicated to neuron–microglia and neuron–oligodendrocyte cross-talk.

Astrocytes are indispensable for neuron homeostasis in the CNS, and carry out a variety of functions ranging from bioenergetic and metabolic support of neurons, neurotransmitter clearance from the synaptic cleft after neuronal excitation, spatio-temporal buffering of extracellular K⁺, scavenging reactive oxygen species (ROS), regulation of cerebral blood flow, and providing trophic support to neurons by regulating axonal guidance and synaptogenesis.^10–16 Astrocytes are endowed with a complex machinery to execute these diverse functions by expressing neurotransmitter receptors and transporters, gap junctions/hemichannels, cell-specific enzymes for metabolic support (i.e. glutamine synthetase) as well as the capacity to secrete numerous factors into the extracellular milieu that exert paracrine effects on neurons.^14,17–22 Any perturbation in one or more of these astrocytic functions can profoundly affect neuronal homeostasis, as evidenced in acute and chronic neurodegenerative diseases.^8,23–28

Astrocyte activation and neuroinflammation

Reactive astrocytosis (astrocyte activation) is observed in nearly all neurodegenerative diseases, and is broadly defined as a defensive reaction by astrocytes in response to a variety of external stimuli, which accompanies either acute (i.e. traumatic brain injury, cerebral ischemia) or chronic neurological conditions (i.e. AD, PD, ALS, Batten disease).^27,29–33 On activation, astrocytes undergo morphological changes and alter their gene expression profiles. A major hallmark of reactive astrocytosis is increased expression of the intermediate filament proteins glial fibrillary acidic protein (GFAP) and vimentin.³⁴ Although the functional implications of augmenting intermediate filament protein expression are unclear, this likely participates in limiting bystander damage to surrounding healthy parenchyma, as has been shown in stroke and spinal cord injury.^35–38

A number of signals can induce reactive astrocytosis that are likely context-dependent, including pathogen-associated molecular patterns (PAMP; such as lipopolysaccharide and peptidoglycan)^39–41 and danger-associated molecular patterns (DAMP), which are typically intracellular molecules that are sensed as danger signals when released from damaged/dying cells.⁴² In addition, emerging evidence supports the concept that astrocytes can also sense danger from within, triggered by mitochondrial damage or disruptions in autophagic pathways, which elicits a specialized intracellular response mediated by Nod-like receptors (NLR).^43–46 Typically, intracellular NLR engagement converges on inflammasome activation; however, the functional impact of NLR signaling in astrocytes remains a conundrum, as evidence suggests that highly purified astrocytes do not produce interleukin (IL)-1β,^40,47 the best characterized inflammasome substrate.

Astrocytes can also perturb neuron homeostasis in the context of neurodegenerative diseases by exerting inflammatory effector functions.^48–57 Although astrocytes are not traditionally considered an inflammatory cell type, they are recognized as a robust source of chemokines in response to diverse stimuli,⁵⁸ and by extension, play a critical role in peripheral leukocyte recruitment into the inflamed/injured CNS. In terms of sensing PAMP and DAMP, astrocytes express pattern recognition receptors, including various Toll-like receptors (TLR; TLR2, TLR3, TLR4 TLR5 and TLR9), although their repertoire is less extensive compared with microglia, the resident innate immune cell in the CNS parenchyma. Among the TLR detected in astrocytes, TLR3 is the most highly induced in response to pro-inflammatory cytokines (tumor necrosis factor-α, interleukin [IL]-1β and interferon-γ) as well as TLR agonists, including poly I:C, LPS, flagellin and CpG oligonucleotides (ODN),^59,60 whereas TLR2 plays a pivotal role in recognizing Gram-positive bacteria, such as Staphylococcus aureus, as well as microbial lipoproteins.^39,61,62 Astrocytes can also sense bacterial/viral DNA (as modeled by non-methylated CpG ODN) through TLR9, which also triggers an inflammatory cascade. In astrocytes, as in other cell types, all TLR transduce activation signals through the adaptor protein, MyD88, with the exception of TLR3, which utilizes TRIF, and TLR4, which can signal through either a MyD88-dependent or -independent pathway. TLR signaling leads to nuclear factor kappa-B and mitogen activated protein kinase activation, which in astrocytes primarily elicits the production of type I interferon (i.e. interferon-α and interferon-β) as well as a wide array of chemokines,^63–65 including CCL2, CCL5, CCL20, CXCL1, CXCL2, CXCL10 and CXCL12.^59,66 These chemokines serve to promote immune cell recruitment from the systemic circulation and amply immune circuits within the CNS.⁶⁶ Typically, the desired result is a swift and efficient inflammatory response to neutralize the offending pathogen/insult and promote rapid resolution/healing to restore CNS homeostasis. However, if inflammatory responses are perpetuated, this can lead to heightened or chronic inflammation, which can be detrimental to neuron survival. This highlights the need to achieve a delicate balance with regard to inflammation in the CNS, which can be influenced by astrocytes based on their central role in controlling neuron health and strategic positioning as sentinels at the blood–brain barrier.

While chemokine synthesis by astrocytes is widely accepted, whether they are a major source of pro-inflammatory cytokines remains a subject of debate. Although astrocytes have been shown to produce pro-inflammatory cytokines in vitro, including tumor necrosis factor-α and IL-1β,^67,68 it remains possible that microglial contamination, which is a common occurrence in astrocyte cultures, could be influencing these responses. In support of this possibility, FACS-purified astrocytes did not show any evidence of pro-inflammatory cytokine production in response to the Gram-negative pathogen, Citrobacter koseri.⁴⁰ Likewise, astrocyte responses to TLR2 and TLR3 agonists were greatly enhanced by microglia, and responses to TLR4 agonists were completely dependent on the presence of functional microglia in cultures.⁴⁷ However, it should be noted that astrocytes are responsive to pro-inflammatory cytokines derived from external sources acting in a paracrine manner, as astrocytes express cell surface receptors for various pro-inflammatory cytokines. These cytokines, in turn, trigger the synthesis of a variety of chemokines and other inflammatory mediators from astrocytes, including prostaglandins, arachidonic acid, and ROS and reactive nitrogen species (RNS).^37,48 The impact of reactive astrocytosis in the context of neuroinflammation will be discussed in the following sections in select neurodegenerative diseases, including AD, PD, ALS and forms of neuronal ceroid lipofuscinosis, collectively referred to as Batten disease. The rationale for selecting these neurodegenerative diseases is based on their shared characteristics of aggregate or storage material accumulation in different neural cell types.⁶⁹

AD

AD represents the most common neurodegenerative disorder, with approximately 44 million people suffering from this disease worldwide and over 5 million people afflicted in the USA.⁷⁰ Although most cases of AD are sporadic in etiology, familial forms of AD are inherited as an autosomal dominant trait with mutations in amyloid precursor protein, presenilin 1 and presenilin 2 identified as pathogenic factors.^71–74 Clinical symptoms of AD include short-term memory loss, difficulty in judgment, progressive cognitive decline, tactile agnosia and apraxia.⁷¹ AD is typified by the intracellular accumulation of neurofibrillary tangles (NFT) composed of hyperphosphorylated tau in neurons within the entorhinal cortex, hippocampus and neocortex,^75–78 as well as senile plaques composed of amyloid-β (Aβ).^79–81 Structural changes occurring in the AD brain include cortical atrophy with shrinkage of gyri and widening of sulci with enlarged ventricles in the frontal and temporal lobes.^79,80

Neuronal loss in AD is moderate in cortical structures (26–30%), with up to 45% pyramidal neuron loss reported in AD patients, which correlates with the density of NFT and senile plaques.^71,82,83 Neuronal loss is associated with reactive astrocytosis concomitant with a wide range of alterations in normal astrocyte functions (discussed in detail in the following sections).^32,84–86 The extent to which reactive astrocytes affect neuron function in AD remains unclear. However, Aβ treatment of neuron–astrocyte co-cultures caused a transient increase in astrocyte [Ca²⁺]i that elicited mitochondrial depolarization and ROS production, resulting in neuronal death.^24,87,88 Likewise, Aβ augments glucose uptake in astrocytes, which increases glycogen deposition, H₂O₂ production and glutathione release. These changes in astrocyte metabolism elicited by Aβ exposure have been shown to impair neuronal viability and reduce synaptic proteins in neuron–astrocyte co-cultures⁸⁹. Additionally, Aβ augmented glutamate release from astrocytes, leading to neuronal NMDA receptor stimulation and subsequent synaptic loss.⁹⁰ Aβ has been shown to activate poly(ADP-ribose) polymerase in astrocytes, and cause neuronal death through oxidative stress.⁹¹ Furthermore, astrocytes participate in Aβ clearance,⁹² and Aβ degradation is impaired in reactive astrocytes, which might increase the risk for direct toxic effects of Aβ on neurons,⁹³ although microglia are a major effector cell responsible for Aβ clearance.^94,95

PD

PD is the second most common neurodegenerative disorder with an incidence of 120 per 100 000 persons by the age 70 years. Greater than 95% of PD is sporadic, without any genetic linkage or predisposition identified, although α-synuclein mutations have been linked to familial forms of PD.^96,97 Clinical symptoms of PD range from tremors, rigidity, absence of voluntary movements, stooped posture and abnormal facial expressions as a result of muscle spasticity.^98–101

Prominent neuropathological features of PD include intraneuronal deposition of cytoplasmic inclusions called Lewy bodies and nigrostriatal dopaminergic neuron loss^102–104, which results in dopamine depletion in the brain.^105–107 Besides striatal dopaminergic neuron loss, PD is also occasionally associated with the death of noradrenergic and serotonergic neurons in the raphe nucleus, as well as cholinergic neurons in the entorhinal cortex.^108,109 These changes are typically associated with robust astrocyte activation.^{27,31,110–112} Multiple lines of evidence have shown that reactive astrocytes adversely affect neuronal function in the context of PD. For example, the engraftment of astrocytes expressing human brain derived neurotrophic factor into the striatum of a rat parkinsonism model reduced neurobehavioral deficits.¹¹³ Likewise, the astrocyte modulating agent 2-propyloctanoic acid not only reduced striatal dopaminergic neuron loss, but also prevented reactive astrocytosis in a rat model of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-induced parkinsonism,¹¹⁴ suggesting that astrocyte activation contributes to neuron loss in PD. 2-propyloctanoic acid is known to inhibit S100β expression and augment glutamate transporter expression, the latter of which might be envisioned to limit neurotoxicity by enhancing glutamate uptake. Additionally, 5-lipoxygenase activity was increased in reactive astrocytes after MPTP administration, which was suggested to inhibit dopamine uptake in striatal neurons, thereby rendering these cells sensitive to MPTP-mediated death.¹¹⁵ Likewise, suppression of inducible nitric oxide synthase and associated inflammatory events in astrocytes after MPTP injection protected striatal neurons from apoptosis.¹¹⁶ Collectively, these studies suggest that astrocytes can negatively impact neuronal survival in the context of PD. However, it should be noted that many of these findings were obtained with MPTP injection, which is an aggressive model that elicits rapid and robust dopaminergic neuron loss. Whether astrocytes negatively impact neuron integrity in other PD models where disease progression is more protracted in nature remains to be determined.

ALS

ALS is a progressive neurodegenerative disease that affects motor neurons in the cerebral cortex, brain stem and spinal cord, resulting in fatal paralysis.^6,117 Clinical features of ALS include progressive peripheral muscle weakness, atrophy, and spasticity as a result of upper and lower motor neuron degeneration. The majority of ALS cases are sporadic (>90%), whereas a minor fraction are associated with mutations in Cu/Zn superoxide dismutase (SOD1). In ALS, SOD1 mutations result in protein misfolding and aggregate formation, which is thought to cause motor neuron death in both cell-autonomous and non-cell autonomous manners.⁵⁶ With regard to the latter, neuroinflammation has recently gained traction as a prominent factor in ALS pathogenesis, with both microglia and astrocytes producing inflammatory mediators that impact motor neuron survival. For example, mutant SOD1 has been shown to trigger chemokine and cytokine production by astrocytes that are neurotoxic.¹¹⁸ Interestingly, expression of mutant human SOD1 in mouse astrocytes, but not motor neurons, elicited neuronal death by releasing soluble factors that induced BAX-dependent death.¹¹⁸ Likewise, co-cultures of motor neurons expressing wild-type human SOD1 with astrocytes transduced with mutant human SOD1 also caused significant motor neuron death¹¹⁹ typified by the loss of mitochondrial function and induction of mitochondrial permeability transition.¹²⁰ Collectively, these studies highlight an important non-cell autonomous role for astrocytes in mediating motor neuron death in ALS.

Neuronal ceroid lipofuscinosis

Neuronal ceroid lipofuscinosis (NCL) encompasses a family of inherited disorders caused by mutations in CLN genes collectively referred to as Batten disease.^121,122 To date, mutations in 14 different CLN genes have been identified that are broadly classified into infantile, late infantile, juvenile and adult onset forms.^122,123 The childhood forms of Batten disease are characterized by blindness, behavioral deficits, seizures, and progressive cognitive and motor impairment that leads to premature death, with the infantile and late infantile forms showing the most aggressive progression.^124,125 A histopathological hallmark of all NCL is the lysosomal accumulation of autofluorescent lipopigments and proteins,¹²⁶ including subunit C of mitochondrial adenosine triphosphate (ATP) synthase (primarily in juvenile NCL) and sphingolipid pigments (in other forms of NCL).^127–129

Infantile NCL (INCL) is the most aggressive NCL form, with an average life expectancy of 2–6 years.¹³⁰ INCL is caused by a mutation in CLN1, which encodes for palmitoyl protein thioesterase, an enzyme responsible for the cleavage of long-chain fatty acid residues on proteins containing cysteine moieties.^131,132 Clinical symptoms of INCL can manifest as early as 6 months-of-age, which rapidly progress to severe motor and cognitive deficits, seizures, and premature death.¹²⁵ Late infantile NCL (LINCL) is caused by mutations in CLN2 that encode the lysosomal enzyme tripeptidyl peptidase I, which cleaves tripeptides from the terminal amine groups of partially unfolded proteins.^133,134 LINCL onset is typically delayed compared with the infantile form, with affected children generally succumbing to disease between the ages of 8–12 years. Other CLN mutations have been associated with variant late infantile NCL, including CLN5, CLN6 and CLN8, where comparatively less is known regarding gene function or associated pathophysiology. Juvenile NCL (JNCL) results from mutations in the CLN3 gene, for review, see¹²⁵. The precise function of CLN3 remains unknown; however, based on functional analysis in yeast models and neuronal cell lines, CLN3 has been implicated in lysosomal acidification, endocytic and vesicle trafficking, and proper maintenance of mitochondrial function.^135–137 Similar to INCL and LINCL, JNCL also presents with visual impairment, seizures, and progressive cognitive and motor decline, but with a later onset, typically between 5–10 years-of-age,¹²⁵ and the disease course is more protracted, with an average life expectancy between the late teens to early-mid 20s.

NCL are associated with neuronal loss in select brain regions, including the hippocampus, thalamus and various cortical areas.^138–140 Additionally, astrocyte activation is evident in each NCL form.¹⁴¹ In JNCL mouse models, studies have shown early prominent astrocyte activation, as indicated by increased GFAP expression, that precedes neuronal loss in these mice that does not occur until around 12 months-of-age.^142,143 Astrocyte activation is also apparent in INCL, and in this setting astrocytes appear to play a protective role in disease, as CLN1/GFAP/vimentin triple mutant mice experienced an accelerated disease course typified by heightened cytokine/chemokine production.¹⁴⁴ It remains unclear why disease progression was exacerbated when GFAP and vimentin were absent, as astrocytes themselves remained intact. This finding suggests that GFAP and vimentin loss alters astrocyte physiology or function, making them less adept at providing what appear to be beneficial roles, as disease severity was increased in CLN1/GFAP/vimentin mutant animals. Reactive astrocytes have also been described in LINCL and variant late infantile NCL forms (i.e. CLN5 and CLN6), which often precedes neurodegeneration.^145–149

Although a direct effect of dysfunctional astrocytes on neuronal loss in the various NCL forms is not yet established, a recent study from our group has shown a progressive decline in the expression of key molecules involved in glutamate homeostasis (i.e. GLAST and glutamine synthetase) in activated astrocytes in JNCL.¹⁴³ By extension, this is expected to have a negative impact on neuron homeostasis and survival in the context of CLN3 mutation.

Potential factors that disrupt neuron-astrocyte interactions in neurodegenerative diseases

As aforementioned, astrocytes are indispensable for neuron homeostasis, as they carry out a variety of critical functions, including glutamate clearance at the synaptic cleft, release of gliotransmitters that provide trophic support to neurons, and supplying precursors for neuronal energy metabolism and anti-oxidant defense. Alterations in one or more of these astrocyte functions have been shown in several neurodegenerative diseases, which likely contribute to neuronal loss in these disorders. A few examples are discussed below.

Altered gliotransmission

Astrocytes express numerous neurotransmitter receptors, including glutamate, gamma-aminobutyric acid, acetylcholine, norepinephrine and adenosine,¹⁵⁰ and constantly communicate with neighboring neurons through a variety of second messengers, such as Ca²⁺ and IP3.⁴² During intense synaptic activity, neurotransmitter release (in particular glutamate) can elicit increased [Ca²⁺]i in astrocytes that in turn, triggers glutamate release from astrocytes primarily through hemichannels. Hemichannel-mediated glutamate release feeds back to activate neurons to regulate synaptic transmission, thus maintaining a type of neurotransmission “synchrony” through Ca²⁺-glutamate oscillations between astrocytes and neurons.^151–153 Some of these events are reportedly altered during neurodegenerative disease. For example, Aβ was shown to specifically augment [Ca²⁺]i in astrocytes, but not neurons, and this rise in astrocytic [Ca²⁺]i induced neuronal mitochondrial permeability transition (mPT) by a mechanism involving ROS production.⁸⁷ Likewise, Aβ was shown to induce glutamate release from astrocytes, leading to neuronal NMDA receptor activation and synapse loss.⁹⁰

Defects in astrocyte glutamate uptake by either reduced glutamate transporter expression or activity could also cause neuronal death by excitotoxicity. For example, Aβ reduces astrocyte GLAST and GLT-1 levels,¹⁵⁴ and decreased EAAT2 glutamate transporter expression has been observed in post-mortem AD brains.¹⁵⁵ Astrocyte glutamate clearance was also impaired after MPTP exposure,^156,157 as well as in a mouse model of PD.¹⁵⁸ Loss of astrocyte glutamate transporters was also shown in ALS.^159–161 Astrocytes in a mouse model of JNCL show reduced GLAST and glutamine synthetase expression, revealing glutamate recycling defects in this disorder.¹⁴³ Therefore, although these neurodegenerative diseases have distinct etiologies, impaired glutamate regulatory mechanisms appear to represent a unifying theme that might contribute to neuron loss, although this remains to be definitively determined.

Astrocyte hemichannels

Recent studies suggest that gliotransmitters can also be released by astrocytes through hemichannel (HC) opening.^162,163 HC are composed of a hexameric ring of connexin or pannexin proteins; however, only the former are capable of forming gap junction channels when they align with adjacent HC on neighboring cells.¹⁶⁴ HC have been reported to open under both physiological and pathological conditions to facilitate the release of various small m.w. metabolites and gliotransmitters from astrocytes, including ATP, glucose, glutathione, glutamate, gamma-aminobutyric acid and D-serine.¹⁶³ Transient HC opening could be beneficial by maintaining molecular gradients across the astrocyte membrane; however, chronic HC activity can have deleterious effects not only on astrocytes by dissipating chemical gradients, but also on adjacent neurons thorough sustained ATP and D-Serine release, which can cause neuronal death.^162,165 Astrocytes in a mouse model of JNCL showed a transient increase in HC opening at an early postnatal age (1 month) that significantly preceded neuronal loss, which occurs around 12 months in this model.¹⁴³ HC blockade using the novel carbenoxolone-based inhibitor, INI-0602, enhanced gap junction communication, and led to significant reductions in lysosomal storage material in the brains of CLN3 mutant mice.¹⁴³ Increased astrocytic HC activity has also been observed in a mouse model of AD,¹⁶⁶ and exogenous Aβ application in acute brain slices augmented astrocyte HC activity, resulting in increased ATP and glutamate release and neuronal death.¹⁶⁷ Furthermore, inflammatory cytokines that are elevated in various neurodegenerative diseases (i.e. IL-1β, tumor necrosis factor-α) significantly increase astrocyte HC activity, again reinforcing the interplay between neuroinflammation and astrocytic dysfunction and downstream implications on neurodegeneration.^{168,169–172}

Purinergic receptors

Purinergic receptors play a pivotal role in neuron–astrocyte communication, as ATP and glutamate release mediated by astrocyte HC opening and metabotropic purinergic receptors (P2YR subtypes) trigger neuronal activity, thereby completing the glutamate neurotransmission loop between neurons and astrocytes.^173–175 However, excessive ATP release from astrocytes could exert autocrine effects by stimulating purinergic receptors to induce prostaglandin and nitric oxide production,^176,177 which would disrupt astrocyte–neuron interactions. Alternatively, sustained ATP release can stimulate microglial purinergic receptors to elicit pro-inflammatory responses.¹⁷⁸ In the light of these events, purinergic receptor activation has been observed in various neurodegenerative diseases including AD, PD and ALS,^179–182 and purinergic receptor antagonists have generally showed beneficial effects in counteracting neuroinflammation in these diseases.¹⁸³ Furthermore, synaptophysin loss and memory dysfunction were significantly reduced in adenosine receptor 2A (A2A) KO mice after intracerebral Aβ administration.¹⁸⁴ Likewise, microglia isolated from P2X(7) receptor KO mice produced less IL-1β after Aβ treatment.¹⁸⁵ Interestingly, conditional A2A receptor deletion in forebrain neurons improved motor incoordination in a mouse MPTP model of Parkinson’s disease, whereas striatal dopamine levels remained low. Intracerebral administration of KW-6002, an A2A receptor antagonist, significantly attenuated the MPTP-induced loss of striatal dopamine levels as well as microglial and astrocyte activation, suggesting that glial purinergic receptors are critical for exerting toxic effects in PD.¹⁸⁶ In contrast to these findings, a recent study reported a somewhat unexpected result in the SOD1-G93A model of ALS, where P2X7 receptor ablation exacerbated reactive gliosis and motor neuron death, suggesting that purinergic receptors can also exert neuroprotective effects in select neurodegenerative disorders, at least during some stages of disease progression.¹⁸⁷

D-serine

D-serine, an endogenous NMDA receptor agonist, has been identified as an important gliotransmitter in neuron–astrocyte interactions with respect to glutamatergic neurotransmission.^188,189. D-serine is synthesized from L-serine in a reaction mediated by serine racemase (SR), whose activity is highly enriched in neurons.¹⁹⁰ However, as neurons are unable to synthesize L-serine from glucose, L-serine is provided by astrocytes, which abundantly synthesize L-serine from glucose by a reaction catalyzed by D-3-phosphoglycerate dehydrogenase.¹⁹¹ Once released from astrocytes, L-serine is internalized by neurons and is converted to D-serine by SR for export back to astrocytes (through the serine shuttle) for storage.¹⁹² Therefore, astrocytes have an abundant D-serine pool and regulate its release to activate neuronal NMDA receptors in a controlled manner. While regulated D-serine release from astrocytes plays an important role in long-term potentiation, an important event in learning and memory,¹⁹³ excessive D-serine release has been implicated in neuronal excitotoxicity through a NMDA-dependent mechanism in neurodegenerative diseases.^194,195 An interesting study reported that Aβ-induced NMDA-mediated excitotoxicity was attenuated in SR knockout mice, which highlights the importance of D-serine in neuronal excitotoxic mechanisms under the influence of astrocytes. Furthermore, SR activity and D-serine levels were increased in SOD1 transgenic mice, resulting in enhanced NMDA-mediated excitotoxic death of neurons.^196,197

Neuron–astrocyte interactions in bioenergetic/metabolic regulation

The main bioenergetic/metabolic pathways utilized by astrocytes and neurons differ. Specifically, astrocytes rely heavily on glycolysis in response to neuron-derived glutamate and the process of glutamate reuptake is a highly energy demanding process, whereas neuronal energetic demands shift them more towards ATP-generating pathways, such as oxidative phosphorylation.¹⁹⁸ The energetic crosstalk between astrocytes and neurons will be discussed in the following section with examples provided from AD, PD, ALS and NCL where available.

Astrocyte–neuron lactate shuttle

Bioenergetic and metabolic communication exists between neurons and astrocytes through glucose/lactate exchange. Lactate is predominantly, if not exclusively, produced in astrocytes, and glutamate plays a central role in this process.¹⁹⁹ For example, during neurotransmission, neurons release glutamate into the synaptic cleft, which is rapidly cleared by astrocytes through selective glutamate transporters along with the release of 3 Na⁺ ions. Na⁺ release is mediated by Na⁺/K⁺ ATPase activity, which consumes 1 ATP to extrude 3 Na⁺ ions from astrocytes. The majority of glutamate internalized by astrocytes is converted to glutamine, while a portion is diverted to the TCA cycle through conversion to α-ketoglutarate.²⁰⁰ This process of glutamate–glutamine conversion is energy-demanding, and to meet these energy demands, astrocytes enhance glucose uptake and stimulate glycolysis to produce pyruvate, which is later converted to lactate and transported to neurons through neuron-specific monocarboxylate transporter 4.^201,187

Although the astrocyte neuron lactate shuttle has been historically accepted as the model whereby astrocyte-derived lactate functions as a major energy fuel for neurons, in recent years, the astrocyte neuron lactate shuttle mechanism has been a subject of debate.^202,203 Accordingly, it has been argued that stimulation of glycolysis to produce lactate during intense neuronal activity in response to glutamatergic neurotransmission is not necessarily operative in all types of astrocytes. Furthermore, neurons do possess an inherent capacity to increase glucose uptake and channel glucose to glycolysis to sustain their energy demands during neuronal excitation.^202,203 Based on the differential kinetic properties of glucose and lactate transporters in neurons (glucose transport capacity being higher in neurons),^204,205 it has been proposed that neurons actively take up glucose and release lactate for astrocytes,^202,203 a phenomenon that appears to oppose the astrocyte neuron lactate shuttle.¹⁹⁹ This possibility has been counter argued by Magistretti’s group in that the mere examination of glucose transporter capacity being higher in neurons might not translate into neurons having increased glucose utilization rates than astrocytes.²⁰⁶ This has also been supported by studies where depletion of the neuron-selective glucose transporter, GLUT3, in Tg mice did not show changes in glucose utilization,²⁰⁷ whereas, deletion of the astrocytic glucose transporter, GLUT1, resulted in a severe neurological phenotype.²⁰⁸ Perhaps this controversy could be resolved by extending studies integrating in vivo and in vitro metabolomics together with thorough investigations into inherited metabolic disorders associated with glucose metabolism.

Despite controversies with regard to astrocytic and neuronal glucose/lactate metabolism, defective neuron–astrocyte metabolic coupling has been reported in various neurodegenerative diseases.^32,209–211 For example, a significant reduction in brain glucose uptake and cerebral glucose metabolic rate was reported in AD patients.²¹² Likewise, cerebral glucose uptake was reduced in an AD transgenic mouse model, which resulted in reduced cerebral lactate release, suggesting impaired astrocyte glycolysis in AD that can have profound effects on neuronal energy metabolism.²¹³ This is further supported by reduced monocarboxylic acid transporter expression in astrocytes²¹⁴, revealing multiple defects in neuron-astrocytic metabolic coupling in AD. Additionally, several glycolytic and citric acid cycle metabolites were reduced in an AD transgenic rat model,²¹⁵ and various energy metabolites, including glucose, lactate and other glycolytic intermediates, were also reduced in the CSF of PD patients.²¹⁶ Interestingly, in vitro exposure of dopamergic neurons to CSF from PD patients resulted in increased tyrosine hydroxylase expression and subsequent neuron death, as shown by increased LDH release,²¹⁷ which is also consistent with defective neuron–astrocyte metabolic cross-talk in PD. Metabolic dysfunction involving the lactate shuttle has also been reported in a transgenic ALS mouse model, as shown by reduced astrocytic lactate transporter expression and concomitant reduction in lactate levels, as well as a reduction in phosphoglycerate kinase, a rate-limiting enzyme of glycolysis.²¹⁸ Magnetic resonance spectroscopy in patients with various NCL forms showed a marked elevation of brain myo-inositol (astrocytic marker) and lactate levels,²¹⁹ the latter suggestive of a neuronal energy metabolic block. Additionally, increased levels of phosphofructo kinase, likely translating to increased glycolysis, was shown in the brains of CLN3^Δex7/8 mice, a model of JNCL, suggesting compensatory mechanisms in astrocytic energy metabolism.²²⁰ Furthermore, deletion of btn1, the yeast orthologue of CLN3, showed increased glycolysis.²²¹

Astrocytic glycogen metabolism

Astrocytes also exert metabolic crosstalk with neurons through glycogen metabolism. Glycogen is the storage form of glucose and is readily mobilized in astrocytes to produce glucose-6-phosphate, which is further channeled into glycolysis to yield lactate.²²² Glycogenolysis is regulated by neuronal vasoactive intestinal peptide and noradrenaline to provide lactate in response to increasing neuron metabolic demands.²²² Studies have shown that astrocytic glycogen contributes to learning and memory.²²³ Alterations in astrocyte glycogen metabolism have been reported in AD and multiple sclerosis, where fluoxetine, which improves neuronal energy metabolism by promoting astrocytic glycogenolysis, has been shown to reduce the number of tremors and significantly enhance cerebral blood flow in PD patients. Fluoxetine also significantly improved memory and cognitive deficits in AD patients.^224,225 It is notable that various inflammatory mediators, including cytokines, chemokines, prostaglandins and ROS, are known to alter neuron–astrocyte metabolic coupling.⁸⁶ Accordingly, astrocytes exposed to inflammatory cytokines failed to deliver anti-inflammatory precursors, such as cysteinylglycine for GSH synthesis in neurons.²²⁶ Likewise, cytokines were shown to augment glucose uptake in astrocytes, which was not reflected by enhanced lactate levels, suggesting that cytokines perturb downstream events of glucose metabolism in astrocytes, resulting in less lactate production and release that can negatively impact neuronal survival.^227,228 Inflammatory cytokines also enhance glutathione release from astrocytes and concomitantly induce superoxide production, representing a “double whammy” to impair astrocyte anti-oxidant activity.²²⁷

Oxidative stress (OS) is another deleterious factor in the setting of neuroinflammation that can compromise neuron–astrocyte metabolic interactions.^209,229,230 Exaggerated ROS production during aberrant metabolic reactions in either astrocytes or neurons can elicit DNA, RNA and protein damage.²³¹ Increased protein carbonyls were consistently found in AD patients in both neurons and astrocytes with pronounced accumulation in non-synaptic and synaptic mitochondria.²³² Additionally, increased carbonylation of metabolic enzymes in the AD brain, including creatine kinase, pyruvate kinase, enolase, phosphoglycerate mutase, lactate dehydrogenase and ATP synthase, supports the possibility of perturbed glycolysis in astrocytes and concomitant downstream impairments on ATP production.²³³ Such oxidative inactivation of LDH and other energy generating enzymes, such as glyceraldehyde 3-phosphate dehydrogenase and enolase, has also been observed in PD and ALS.^234–236 Collectively, these reports strongly support the concept that various inflammatory mechanisms elicited during neurodegenerative diseases can perturb neuron–astrocyte metabolic coupling and contribute to neurodegeneration.

Mitochondrial dysfunction in neurodegenerative diseases

Mitochondrial dysfunction is a characteristic feature of numerous neurodegenerative diseases.^209,237,238 For example, Aβ has been shown to selectively enhance [Ca²⁺]_i in astrocytes, but not neurons, which induced mPT opening and activation of the DNA repair enzyme Poly(ADP ribose) polymerase. These astrocytic changes were shown to induce oxidative stress in neurons, and inhibition of astrocytic mPT with cyclosporine A significantly attenuated neuronal death,⁸⁷ suggesting that abnormal astrocytic function contributes to neuronal loss in AD. Similarly, astrocytes derived from DJ-1 knockout mice (a model of familial PD) showed severe mitochondrial abnormalities, including reduced mitochondrial motility, mPT induction and failure to protect neurons from oxidative injury and death,^239–241 showing that astrocytic mitochondrial abnormalities impair neuroprotective responses in PD. Likewise, astrocytes derived from mutant SOD1 mice showed mitochondrial dysfunction, which translated to less neuroprotective activity. Interestingly, overexpression of mitochondrial-targeted catalase significantly improved mitochondrial function in SOD1 mutant astrocytes and prevented motor neuron death.²⁴² Furthermore, Nrf2 activation and glutathione induction in astrocytes protected motor neurons from oxidative injury and apoptosis.²⁴³ Mitochondrial abnormalities have also been reported in neurons from a mouse model of JNCL (CLN3^−/−), reflected by decreased oxygen consumption and mitochondrial electron transport chain enzymes.²⁴⁴ Similarly, enlarged mitochondria were observed in a neuronal cerebellar granular cell line derived from JNCL mice.¹²⁹ Although astrocytic mitochondrial dysfunction in NCL has not yet been reported, it should be noted that astrocyte activation is observed in all NCL forms, and mitochondrial dysfunction is a common denominator in reactive astrocytes, suggesting a potential link between the two (see review Rama Rao and Kielian, 2015).

Astrocytes and neuronal trophic support

As aforementioned, a main function of astrocytes is to provide trophic support to neurons.^245,246 This is achieved, in part, by the synthesis and secretion of various soluble factors by astrocytes, including extracellular matrix proteins (matricellular proteins, such as thrombospondins, hevins and glypicans), cholesterol and tenascin, as well as growth factors (transforming growth factor-β, CNTF, S100B) that bind to neuronal receptors and engage signaling pathways to synthesize/recruit proteins involved in synapse modeling.^247–249 Although the involvement of astrocyte-derived soluble factors on neuronal synaptic integrity has not been extensively investigated, studies have shown that astrocyte matricellular proteins and growth factor production is impaired in neurodegenerative diseases where astrocytes have been implicated in the loss of synaptic integrity.^248,250,251 For example, AD post-mortem brains showed reduced thrombospondin-1 (TSP-1) levels,²⁵² and Aβ treatment of astrocytes in vivo augmented ADAMTS-4, which contains a TSP-1 motif.²⁵³ A recent study showed that Aβ_1–42 augmented intracellular TSP-1 levels in astrocytes concomitant with reduced extracellular TSP-1 protein, suggesting aberrant TSP-1 trafficking.²⁵¹ Furthermore, conditioned medium from Aβ-treated astrocytes caused a significant reduction in synaptophysin and PSD95 expression in primary neurons,²⁵¹ suggesting that impaired astrocytic trophic support in the context of AD might impact neuronal synaptic integrity. Similar to what has been observed after astrocyte exposure to Aβ, astrocytes from STAT3 KO mice released less TSP-1, which led to a reduction in synaptophysin and PSD95 levels in motor neurons, suggesting that astrocytic TSP-1 functions to promote motor neuron survival and synaptic integrity.²⁵⁴ It is interesting that various inflammatory signals, such as oxidative stress, prostaglandins and cytokines, also impair astrocyte TSP-1 synthesis.^255–257 Collectively, these findings suggest that the loss of critical astrocyte trophic support negatively impacts neuronal survival during neurodegenerative diseases.

Conclusions

Neurodegenerative diseases are debilitating neurological disorders typically accompanied by neuronal loss in select brain regions that can manifest as memory loss, movement disorders, paralysis and/or severe cognitive deficits. Both cell-autonomous and non-cell autonomous pathways have been implicated in mediating neuronal death in these disorders, and emerging evidence strongly supports a role for astrocytic dysfunction affecting neuronal loss in a non-cell autonomous manner by inducing neuroinflammatory responses (Fig. 1). Therefore, targeting astrocyte dysfunction in these diseases might have a therapeutic benefit to counteract neuropathology.

Figure 1.

Potential factors contributing to the disruption of neuron–astrocyte interactions in neurodegenerative diseases. Astrocyte dysfunction likely arising from reactive astrocytosis contributes to neuron death in a non-cell autonomous manner through the release of damaging molecules (glutamate [Glu], adenosine triphosphate [ATP], D-serine [D-ser]) as well as inflammatory mediators, such as chemokines. The majority of these molecules also act on astrocytes in an autocrine manner and exacerbate astrocyte dysfunction. Furthermore, aberrant synthesis/release of matricellular proteins, such as thrombospondins (TSP), as well as neuronal energy precursors (lactate) from astrocytes could severely curtail trophic support to neurons and disrupt energy metabolic coupling between neurons and astrocytes. In addition, accumulation of storage materials within neurons also directly regulates cell death in a cell autonomous manner. Aβ, amyloid-β; α-Syn, α-synuclein; EAAT2, glial excitatory amino acid transporter 2; Gln, glutamine, GS, glutamine synthetase; HC, hemichannels; m-SOD1, mutant superoxide dismutase; MCT1&2, monocarboxylic acid transporters 1 & 2; NFT, neurofibrillary tangles; NMDA, N-methyl-D-aspartate; P2Y & P2X2, purinergic receptors; ROS, reactive oxygen species; TSP-11/2, thrombospondins-1/2.