Astrocytes had long been considered passive members of the CNS without electrical activity. It wasn’t until the 1990s when new techniques in Ca²⁺ imaging revealed them as excitable, albeit in ways different from the neuron. The earliest studies demonstrated how cultured hippocampal astrocytes responded to glutamate with increases in intracellular calcium [11], or that mechanical stimulation of a single astrocyte in a primary glial culture could increase intracellular calcium concentration [12]. In both instances, specific increases in intracellular calcium were propagated to cells in the surrounding cultures providing us with some of the earliest evidence of communication between astrocytes. Gap junctions between local astrocytes in addition to extracellular adenosine triphosphate (ATP) link this activity to the surrounding glia and onward in an electrically coupled syncytium [13,14]. Relatively slow Ca²⁺ waves, as compared to the neuronal action potential, are now recognized as hallmark features of astrocytes.

Astrocytes undergo both neuron-dependent and spontaneous excitation [15]. The neuron-dependent excitation of astrocytes is a well-documented way in which synaptic neurotransmitters directly or indirectly through the connected astrocyte syncytium lead to astrocyte excitation. Astrocytes respond to a wide range of neurotransmitters including glutamate, gamma-aminobutyric acid (GABA), acetylcholine, ATP, nitric oxide, and brain-derived neurotrophic factor (BDNF) [15,16]. Further, there is evidence that astrocytes are able to discriminate synapses belonging to different axon pathways. In rat hippocampal slices, Perea and Araque [17] showed that astrocytes not only respond differently to glutamate and acetylcholine, but also to glutamate released from the Schaeffer collaterals compared to glutamate from the alveus terminals. Beyond neuron-dependent excitation, astrocytes are capable of generating spontaneous intracellular calcium oscillations that trigger astrocytic glutamate release causing N-methyl-D-aspartate (NMDA)-receptor dependent neuronal activation [18]. Similarly, hippocampal astrocytes in situ exhibit calcium oscillations that occur independent of neuronal activity or metabotropic glutamate and purinergic receptors and are mediated by inositol-phosphate-3 receptor activation [19].

Ca²⁺ signaling regulation varies by the anatomical region that the astrocyte is found in. A recent study shows that astrocytes along the hippocampal mossy fiber pathway respond to repeated mossy fiber action potentials, are activated by neurotransmitters glutamate and GABA, and cover a large brain region indicating unsuitability in mediating individual synapses [20]. Importantly, observed spontaneous astrocyte Ca²⁺ signaling was independent of this activity and its associated neurotransmitters. In contrast, previous study of hippocampal CA1 and dentate gyrus astrocytes show a high sensitivity towards spontaneous activity by glutamate and single action potentials at a local level [21,22]. Such differences in astrocyte Ca²⁺ response sensitivity, effect size, and overall stimulus suggest differential neuronal population effects. Interestingly, the anatomical placement of these astrocytes along this circuit hints at their function. As part of the trisynaptic circuit, hippocampal input follows the perforant path where entorhinal cortical inputs synapse within the dentate gyrus, which then via mossy fibers is relayed to CA3, which is ultimately relayed to CA1 via schaffer collaterals. Such astrocyte heterogeneity along this circuit suggests differential function based on factors such as synapse number, source of inputs and outputs, in addition to variation in plasticity within these anatomical regions. The extent and significance of this astrocyte variation has yet to be systematically explored.

Together, growing evidence on the excitation of astrocytes establishes them as integral elements in the information-processing pathways of the CNS capable of contributing to excitation of both glial and neuronal elements. It’s important to note that Ca²⁺ signaling and its role in astrocyte function is rapidly changing and its effects may be too fine to adequately determine without reassessment [23].

Ca²⁺ astrocyte activation is required for the release of astrocyte gliotransmitters [21,24]. Volterra and Meldolesi have previously suggested a precise classification criterion for gliotransmitters. This definition includes synthesis and/or storage in astrocytes, a regulated release stimulated by physiological stimuli, activation of rapid responses in neighboring cells, and finally a role of the gliotransmitters in physiological processes [15]. Even within this growing list of gliotransmitters, molecules released by the astrocyte may not all hold the same breadth of function or effect; some including D-serine may act as strong mediators of synaptic plasticity and LTP, while others such as ATP and tumor necrosis factor (TNF)-α may act as modulators and regulate plasticity in more subtle ways [25].

By releasing gliotransmitters to communicate over short distances with neighboring neurons, astrocytes actively contribute to functions spanning synapse development/transmission to controlling blood vessel flow. There is now a growing list of molecules secreted by astrocytes in response to stimulation. These include glutamate, ATP, adenosine, D-serine, prostaglandins and TNF-α among many others that are secreted from the astrocyte and act at specific receptors with specific outcomes. Glutamate is perhaps one of the most well-known gliotransmitter released by astrocytes. It acts at the alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and NMDA-receptors, and metabotropic glutamate receptors (mGluRs) and contributes to LTP by increasing excitatory postsynaptic potentials and upregulating AMPA receptors [5]. As discussed later in this review, other secreted gliotransmitters such as D-serine, ATP and TNF-α significantly contribute to LTP [25].

Mechanistically, gliotransmitter release from astrocytes is attributed to both exocytosis-mediated and more recently non-exocytosis pathways. Glutamate release by astrocytes occurs classically through exocytosis in a Ca²⁺-dependent manner previously thought specific only to synapses. This may occur through a synaptic-like vesicular (SLMV) compartment that contains crucial elements required for glutamate exocytosis, including the vesicular glutamate transporter, vesicular SNARE protein, and cellubrevin [15,26]. Interestingly, the gliotransmitter D-serine that regulates NMDA-receptor mediated LTP may be coreleased with these vesicles or SLMVs carrying glutamate [27]. As more evidence accumulates, it is possible for gliotransmitters to be released from Ca²⁺-independent nonexocytotic mechanisms. These include hemichannels such as the Cx43 prevalent in astrocytes, volume-regulated anion channels or VRACs, purinergic channels such as P2X₇ or the cysteine-glutamate exchanger [28].

With expanding knowledge on astrocyte signaling and gliotransmission as it pertains to the tripartite synapse, research has expanded on the neuronal functional outcomes of this phenomenon on LTP. LTP is the process by which varied biological processes may be reinforced. This popularly includes memory formation but may also include formation of addiction behavior with drugs. Instead of considering the presynaptic and postsynaptic neurons in isolation, the astrocyte has become increasingly recognized as part of a tripartite model where it acts as an essential 3rd element actively involved in how a signal may be conducted at the neuronal synapse.

LTP mechanisms have been extensively studied in the past two decades. The most well studied mechanisms are those involving NMDA-dependent glutamate signaling and a cAMP-dependent presynaptic form of plasticity that we will elaborate on here. For LTP induction, both the presynaptic and postsynaptic neurons must be depolarized, following which there much be high frequency stimulation of the NMDA-type glutamate receptors that results in a large amount of calcium influx into the post-synaptic neuron [29]. This influx leads to activation of calmodulin kinase II, phosphorylation of AMPA receptors, and their exocytosis to the postsynaptic membrane [29,30]. The requirement of AMPA receptors for NMDA-dependent LTP induction was first understood from experiments involving activation of postsynaptically silent synapses in the hippocampus [31,32]. It is now a well evidenced in the literature as the mechanism of LTP induction [29].

The corollary to LTP is long-term depression (LTD) that as the name suggests is a reduction in activity-dependent reduction in synaptic strength. It has gained increasing attention in recent years with its functional importance studied in monocular deprivation studies that lead to blindness in the unused eye. Of the several mechanisms that have been posited, NMDA-receptor dependent LTD is a leading one. Experimentally, LTD may be induced through low-frequency stimulation (1–3 Hz for 5–15 minutes) of the presynaptic neurons [29]. Repetitive stimulation such as this results in a prolonged increase in calcium in the postsynaptic neuron due to repeated stimulation of the NMDA receptors [29]. The calcium influx selectively activates phosphatases that cause endocytosis of, and reduce the AMPA receptors at the synaptic cleft.

Glutamate is the primary excitatory neurotransmitter in the brain and also the primary neurotransmitter for LTP and LTD induction, acting through the ionotropic NMDA and AMPA-type glutamate receptors. Astrocytes play significant roles in regulating LTP induction. They are the primary recyclers of glutamate and sequester up to 90% of the neurotransmitter at the synaptic cleft [33-35]. In fact, early experiments that quantified glutamate-uptake by astrocytes (astrocyte transporter currents) played a pivotal role in our understanding of LTP induction, namely that LTP induction is not related to further glutamate release, but rather an intracellular response involving AMPARs as described above [36,37]. Astrocytes also release glutamate through a calcium-regulated mechanism that allow for astrocyte-neuron communication [26,38,39]. Triggers include neuron-dependent excitation of astrocytes through neurotransmitters including synaptic glutamate itself that can act through astrocytic metabotropic glutamate receptors [15]. In the hippocampus, glutamate release from astrocytes also synchronizes neuronal activity and control synaptic strength [40].

Depression is a common disorder that is characterized by anhedonia, loss of interest, psychomotor retardation, and suicidality. It is a significant cause of disability worldwide with 16.2% lifetime and 6.6% 12-month prevalence in the United States population [49]. Several studies in humans have recorded pathological glial loss in the brains of major depressive disorder (MDD) patients. These include glial reduction in the subgenual prefrontal cortex, anterior cingulate gyrus, and prefrontal cortex [50-52]. In the hippocampus, several magnetic resonance imaging studies reveal a decreased volume [53]. Interestingly, while another study found depth of hippocampal slices from post-mortem brains of MDD patients, at the cellular level there was a 30% increase in glial density across hippocampal pyramidal subfields and the granular cell layer. Compared to other regions in the brain with glial loss, in the hippocampus however there was a unique reduction in neuropil that consists of glial cells, their processes, dendrites and axonal processes that may explain the volume loss [54]. These studies show that astrocytes are pathological hallmarks of MDD.

Several studies have attempted to delineate how astrocytes may mediate MDD behaviors. Previous research establishes that L-alpha-aminoadipic acid (L-AA) selectively ablates astrocytes. Targeting the prefrontal cortex, studies show that L-AA astrocyte ablation leads to anhedonia in sucrose preference test, anxiety in novelty suppressed feeding test, and overall increased helplessness in forced swimming and two-way active avoidance tests in vivo [52,55]. Additionally, the role of astrocytes in mediating depressive-live behavior may be directly related to its release of the gliotransmitter ATP. Mice that are susceptible to chronic social defeat, a model that mimics depression, may have lower levels of ATP in the prefrontal cortex and the hippocampus. In their study, Cao et al. [56] showed that administration of ATP produces rapid antidepressant effects while blocking ATP release in mice lacking IP3 receptor type 2 or by inhibiting vesicular gliotransmission produces depressive-like behavior. Furthermore, using a transgenic Gq GPCR astrocyte mouse model to selectively activate astrocytes through Ca²⁺ signaling and ATP release, Cao et al. [56] were able to rescue some depressive behaviors in mouse models of depression. Supporting the role of astrocyte purinergic signaling, at least one study looking at the ventral PFC in patients with suicide showed altered expression of genes involving ATP biosynthesis in addition to GABA transmission [57].

Different mechanisms have been put forth to describe the role of astrocytes in epilepsy. Human studies of patients with temporal lobe epilepsy have shown evidence of reactive astrogliosis. One manner in which astrocytes may mediate epilepsy includes dysfunction of the glutamate-glutamine cycle. The increase in GFAP and vimentin that occur with reactive astrogliosis has been associated with a concurrent decrease in glutamine synthetase, an enzyme that converts glutamate to glutamine which is a precursor to the neurotransmitter GABA. It has been hypothesized that this results in a lack of inhibitory synaptic transmission in the surrounding neurons. In their study, Ortinski et al. [58] showed how adeno-associated virus induction of reactive astrogliosis in mouse CA1 pyramidal neurons reduced the inhibitory input on the surrounding neurons. Reduced inhibitory postsynaptic current was observed that could be increased with exogenous administration of glutamine.

Astrocytes may also result in epileptiform activity through their effects on adenosine levels. This comprises the adenosine kinase (ADK) hypothesis of epileptogenesis. Adenosine plays neuroprotective roles in seizure regulation with its levels typically elevated following a seizure. It may mediate seizure arrest and postseizure refractoriness. Adenosine is converted to adenosine monophosphate via the enzyme ADK which is predominantly found in the astrocytes. Reactive astrogliosis causes an increased amount of available ADK that can potentially act on adenosine and reduce its levels leading to seizure aggravation [59]. Finally, increasing evidence is demonstrating that calcium oscillations in, and glutamate release from astrocytes could also lead to epileptiform discharges. Cultured astrocytes from human epileptic foci show increased spontaneous calcium-oscillation frequency [60,61]. Inducing cortical epileptiform activity may increase calcium-oscillations in astrocytes that can be suppressed though the use of anticonvulsants. Further, astrocytes may also cause excitations in nearby neurons that resemble interictal paroxysmal depolarizations. Other studies have posited that glutamate release from astrocytes instead of being necessary to cause epileptiform discharges could play more of a modulatory role [62,63]. Other studies have shown calciumdependent glutamate release from astrocytes may also contribute to neuronal death seen in status epilepticus [64].

Astrocyte role in the pathogenesis of schizophrenia comes from recent studies involving the gliotransmitter D-serine and NMDA neurotransmission that bear on the glutamate hypothesis of schizophrenia. According to this hypothesis, a relative deficiency of glutamate and consequent hypofunction at the NMDA receptor may contribute to schizophrenia [65]. Early evidence for the hypothesis came from studies involving NMDA receptor antagonists, such as phencyclidine that induced schizophrenia-like positive and negative symptoms in subjects [66]. Studies have shown that activation of the NMDA receptor requires binding of both glutamate and a coagonist D-serine that is a predominant endogenous ligand that binds at the ‘glycine’ site of the receptor [67,68]. Using D-amino oxidase to selectively degrade D-serine is enough to significantly reduce NMDA receptor-mediated neurotransmission [67].

More importantly, D-serine is also a gliotransmitter that is localized to the astrocytes that contain the enzyme serine racemase to convert L-serine to D-serine [69]. Accumulating evidence has demonstrated decreased levels of D-serine in schizophrenia patients, or polymorphisms of the serine racemase gene [70]. In fact, D-serine has proven an efficacious therapeutic agent in the treatment of schizophrenia [70-72]. Recent evidence has put forth a more compelling case for astrocyte function. DISC1 is a gene that was found to be altered in a Scottish family and associated with schizophrenia, depression and bipolar disorder [73]. DISC1 functions in neurodevelopment and regulation of hippocampal adult neurogenesis [73-75]. One recent study has shown how mutant DISC1 may physiologically impair the function of D-serine by binding to serine racemase and causing its degradation and thus resulting in a diminished availability of D-serine at the NMDA receptor [76]. Furthermore, astrocytic DISC1 expression influences adult hippocampal neurogenesis and hippocampal-dependent behaviors through abnormal production of D-serine [77]. Taken together, these studies suggest that astrocyte derived D-serine mediated by DISC1, in addition to NMDA receptor antagonism and reduced glutamate activity may be involved in schizophrenia pathogenesis.