Learning & Memory; the unexpected role of astrocytes

(1)

Possibly the first image of an astrocyte - Otto Deiters (1856).

Josien E. Visser (10213791) | Literature Thesis | August 8, 2016 MSc in Brain and Cognitive Sciences (UvA), behavioural neuroscience

Supervisor: Prof. Dr. E.M. Hol Co-assessor: L.M. Beex-Osborn (MSc)

Learning & Memory

THE UNEXPECTED ROLE OF ASTROCYTES

(2)

Abstract

Astrocytes have been considered to be passive cells that are mainly important for neuronal support. However, during the past few decades, a wide range of studies indicated that these glial cells could also play an active role in cognitive functions like learning and memory. For a long time, memory research mainly focussed on neurons, but it has been established that astrocytes are in close connection with neural cells and influence long-term potentiation. Many studies showed that astrocytes can modulate neural activity by releasing various gliotransmitter, including glutamate and

D-serine. But, there is also still much debate about the involvement of astrocytes in memory

processes. In this review, novel developments concerning astrocytic contribution to processes related to memory are discussed. Behavioural research is described, but also studies that investigate the molecular principles of astrocytes and glial networks are considered. It is concluded that a memory trace is conserved in activity patterns of neurons, but astrocytes have a prominent role in the consolidation process.

(3)

Introduction

Until this day, no consensus has been reach about the exact molecular and cellular mechanisms of one of the most studied cognitive functions, learning and memory. For many decades, neurons were the focus in this field of research. Neuronal cells communicate via action potentials, which is a fast electrical alteration in the membrane potential. Due to these electrical potentials, neurons can be studied fairly easy with different electrophysiological techniques (Scanziani & Häusser, 2009). However, research suggests that glia, including astrocytes, also could play a major role in establishing learning and memory (Brockett, et al., 2015; Lee, et al., 2014).

The prevalence of astrocytes in the brain varies across species, but in humans, astrocytes outnumber neurons with a ratio of approximately 3:2 (Banaclocha, 2007). A long time it was thought that supporting neurons by nutrient provision (Brown & Ransom, 2007), blood brain barrier formation (Abbott, et al., 2006) and maintaining homeostasis (Simard & Nedergaard, 2004) were the main functions of astrocytes. At this time, years after their discovery, multiple studies revealed that the role of astroglia in learning and memory is not passive, but that these cells could play a crucial role in many cognitive functions (Lee, et al., 2014; Suzuki, et al., 2011). The reason that the importance of astrocytes in cognition was discovered only recently, is because these cells communicate via calcium waves (McCarthy & Salm, 1991). Only in the 80s, a technique called calcium imaging was developed, which gave scientists the ability to study Ca2+ _{concentration variations in cells of the central nervous}

system (Cannell, et al., 1987)

The purpose of this paper is providing an overview of how astrocytes contribute to processes regarding learning and memory. First, briefly, neural mechanisms behind learning and memory will be described. Secondly, the classical view of astrocytes passively supporting neurons will be considered. Furthermore, most importantly, novel developments concerning astrocytic contribution to processes related to memory are discussed.

Classical view of learning and memory

Memory consolidation is a key term in learning and memory research (Bontempi, et al., 1999; Milner, et al., 1968), which refers to the mechanisms that lead to permanent storage of a memory trace in the brain. Memory consolidation is assumed to be established by two different processes; the reorganization of the trace within other brain areas (system consolidation) (Bontempi, et al., 1999) and alterations in synapse strength (synaptic consolidation) (Kelleher, et al., 2004).

Systems consolidation

Multiple brain areas are involved in memory formation, including the hippocampus. This is a major brain structure in the human and vertebrate brain, which is situated in the medial temporal lobe (MLT) (Squire & Zola-Morgan, 1991). The hippocampus plays a key role in episodic, spatial and working memory (Squire, 1992). This brain structure is also important for acquiring short-term memories (Clark, et al., 2001; Zola-Morgan & Squire, 1986). On the other hand, it is thought that long-term memory consolidation occurs in the prefrontal cortex (Bontempi, et al., 1999; Graham & Hodges, 1997). Scientists still debate whether the hippocampus remains involved in long-term memory storage and acquisition (for excellent reviews see Kryukov, 2008; Winocur & Moscovitch, 2011).

(5)

Synaptic consolidation

Memory consolidation in the hippocampus and cortex is established by alterations in strength of synaptic connections. which is referred to as synaptic consolidation. Repetitive patterns of neuronal activity induce strengthening of synapses through a process that is termed long-term potentiation (LTP) (Cooke & Bliss, 2006) or reduction of synaptic efficacy called long-term depression (LTD) (Massey & Bashir, 2007). Synaptic efficacy can be assessed by electrophysiologically quantifying excitatory postsynaptic potentials (EPSPs). Increased EPSPs is an indication of LTP (Bliss & Lomo, 1973; Malenka & Bear, 2004), whereas LTD is characterized by prolonged reduction of this potential (Massey & Bashir, 2007).

LTP can be induced by high-frequency tetanic stimulation. This results in neuronal release of glutamate. If the concentration of glutamate is sufficient, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPA-receptors) and N-methyl-D-aspartate receptors (NMDA-receptors) will open. These events lead to a rapid Ca2+_{-influx, which acts as a secondary messenger.}

As can be observed in figure 1, in the early-phase of LTP, intercellular AMPA receptors are transported to the postsynaptic membrane due to the activation of Ca2+/calmodulin-dependent protein kinase II (CaMKII) and Protein Kinase C (PKC). When the calcium-concentration elevations persist, mitogen-activated protein kinases (MAPK) are phosphorylated. This results in gene transcription and ultimately the formation of proteins, such as novel AMPA-receptors. In addition, growth factor synthesis is increased, which is involved in the formation of new synapses (Sweatt, 1999). These events promote synaptic strengthening.

If synaptic strength could only rise, synapses could reach their peak, leading to an inability to consolidate new memories. As LTP can be brought about by high frequency trains of repetitive stimulation, hippocampal LTD is caused by prolonged low frequency stimulation (Kirkwood, et al., 1993). Furthermore, both LTD and LTP rely on a voltage-gated Ca2+_{influx. The difference between}

LTP and LTD lies within the velocity of calcium entering the cell; fast and large increases generate enhanced EPSPs, but after small and gradual elevations of Ca2+_{, LTD is initiated. These slow rises}

activate phosphatases, which removes phosphate groups and subsequently deactivates these proteins which, for example, could result in a decrease of AMPA-receptors in the synaptic membrane (Connor, et al., 1999).

Figure 1. Molecular mechanism for Long-Term Potentiation (LTP).

Glutamate release into the synaptic cleft after high-frequency stimulation leads to opening of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA-receptors) and N-methyl-D-aspartate receptors (NMDA-receptors). Ca2+_-influx via NMDA-receptors acts as a second messenger. This induces gene

transcription, resulting in formation of proteins and synthesis of growth factors (adapted from

http://bigtone.zone/category/long-term-potentiation/).

(6)

Although, research on systems and synaptic consolidation has provided much knowledge about the mechanisms behind memory, still many questions remain unanswered. For example, electrical signalling between neurons happens within a time scale of milliseconds, while consolidating a short-term memory trace into long-short-term memory requires a larger temporal window of hours or longer. It is possible that neurons cannot fulfil all requirements for memory consolidation, but that other cells in the brain are responsible for these processes. Astrocytes form large networks and glial communication have much slower temporal dynamics (Hirase, et al., 2004; Halassa & Haydon, 2010), suggesting that glial cells could be important for these aspects of memory storage. Despite being situated around neural synapses, astrocytic responses and anatomical changes have not been investigated in memory consolidation studies for many years. However, since the last couple of decades more research is carried out regarding this matter and the idea of astrocytes having an active role in learning and memory gets more widely accepted (Brockett, et al., 2015; Lee, et al., 2014; Nedergaard, 1994; Suzuki, et al., 2011).

Classical view of astrocytes

Rudolf Virchow, a German pathologist, was the first person who referred to glial cells as ‘nervenkit’ or ‘neuroglia’ (Greek γλία (glia) means "glue"), since he thought that it served as a form of connective tissue (Virchow, 1856). Otto Deiters published the book "Untersuchungen über Gehirn und Rückenmark", (Investigations on brain and spinal cord). This book contained a drawing of stellate glial cells, which most likely were the first images of astrocytes (see figure 2) (Deiters, 1865). Furthermore, in 1895 the term ‘astrocyte’ (in Greek, astron means star and cyte comes from “kyttaron", which means cell) was assigned to the star shaped glial cells in the central nervous system by Mihály Lenhossék (Lenhossék, 1895). Camillo Golgi was the first person who proposed that astrocytes facilitate nutrition provision for neurons. He based his ideas on the observation that astrocytes are in close contact with both blood vessels and neuronal cell bodies. Therefore, he hypothesized that these glial cells could be essential for metabolic exchanges (Golgi, 1871). Cajal criticised this hypothesis by arguing that not all endfeet of astrocytic processes are near arteries. However, in that time there was no clear distinction between glial cells yet. He might have mistaken oligodendrocytes for astrocytes and he probably underestimated the number of glial processes adjacent to capillaries due to limitations in techniques during that period (Ramón y Cajal, 1899). The belief that astrocytes carry out processes in a non-active manner originates from electrophysiological recordings. During these measurements only passive currents were detected in astroglia (Kuffler & Potter, 1964). In this chapter the classical view of astrocytes as primarily passive supporters of synaptic efficacy will be described, including their role in nutrition-provision. Furthermore, the importance of glia in blood-brain barrier formation and homeostasis will be considered.

Homeostasis

Maintaining homeostasis is one major role of astrocytes. These glial cells regulate the extracellular concentration of ions including potassium, sodium, chloride and hydrogen. For example, neuronal activity leads to an efflux of potassium into the extracellular matrix. Potassium is in large concentrations toxic for neurons. Astrocytes have three ways to transport these ions from the environment in order to maintain internal stability. Firstly, co-transporters are responsible for passive uptake where either sodium or chloride ions are exchanged for potassium molecules. Secondly, active buffering can be carried out by opening inwardly rectifying K+-channels or

(7)

activation of the Na+/K+-ATPase pump. Lastly, astrocytes form networks via gap junctions (also see Chapter ‘Communication via calcium waves and astrocytic networks’), which are channels between cells that directly connect the cytoplasm of glia. Gap junctions allow the transportation of ions to neighbouring cells resulting in higher buffer capacities (Simard & Nedergaard, 2004).

Figure 2. Possibly the first image of an astrocyte. Drawing

from Untersuchungen über Gehirn und Rückenmark – Otto

Deiters (1856) (no higher resolution available).

Blood-brain barrier

The blood-brain barrier (BBB) is a selective permeable barrier that protects the central nervous system. The BBB mainly consist of endothelial cells, which are connected with tight junctions. These cells are enclosed by various cellular elements, including the perivascular endfeet of astrocytes. These astrocytes are the link between the BBB and neurons in order to regulate the transportation of nutrients and waste products. Moreover, astrocytes are important for both the development and maintenance of the BBB. The presence of astrocytes induces tighter tight-junctions, which increases selectivity of this barrier (Abbott, et al., 2006).

Nutrition provision

As previously mentioned, many astrocytes are wrapped around blood vessels (Golgi, 1871). These glial cells are often also in close connection with neuronal dendritic and somatic parts (Halassa, et al., 2007). This makes these cells excellent candidates for supplying neural cells with nourishment. Glucose is considered to be the main energy source of neurons and is transported by the blood. Many techniques within the field of neuroscience depend on the interpretation that blood-flow increase is correlated with neuronal activation due to higher glucose and oxygen demands (Raichle & Mintum, 2006). The first evidence in mammals indicating that glial cells are important for glucose metabolism, comes from a study with an ingenious method to study glucose metabolism using [3H]2-Deoxy-o-glucose ([3H]2DG). [3H]2DG is a radiolabelled counterpart of [3H]2-Deoxy-o-glucose, which cannot be further metabolised. Therefore, this substance accumulates within the cell, which allows glucose uptake quantification. In the study by Poitry-Yamate and Tsacopoulus (1990) it was found that in the retina of a guinea pig [3H]2DG is mainly present in glia (Müller cells). Furthermore, the substance was phosphorylated, implying that glucose is metabolised in glial cells (Poitry-Yamate & Tsacopoulos, 1990). Yet, a similar study that examined glucose metabolism in cultured cells and in vivo in rats, concluded that glucose can be phosphorylated in both neurons and astrocytes (Itoh, et al., 2004). In

(8)

order to gain more spatial precision, a study with fluorescent labelled glucose combined with 2-photon microscopy was carried out in vivo. The results indicated that in a resting state, glucose accumulates both in neurons and astrocytes of the barrel cortex. However, during whisker stimulation, which led to increased neural activation in this part of the brain, glucose concentration rose more in astrocytes (Chuquet, et al., 2011). This can be explained by the observation that astrocytes possess 6-phosphofructo-2-kinase/fructose-2, 6-bisphosphatase-3 (Pfkfb3), whereas neurons lack this glycolytic enzyme. Glycolysis describes the degradation of glucose and results in the formation of ATP, an intracellular energy source. Furthermore, glucose directly taken up by neurons is mainly degraded by the pentose phosphate pathway, which is related to oxidative stress reduction and not to energy supply (Herrero-Mendez, et al., 2009) This indicates that glucose transported via astrocytes plays a more prominent role during enhanced activity and, despite neural uptake, is the main energy source for neurons. However, these studies do not provide any indications about whether phosphorylated glucose is transported to neighbouring neurons or if this energy source is subjected to further metabolic steps. This question can be explained by the lactate-shuttle hypothesis, which is described in the chapter “astro-neuronal chemical communication and the effect on learning and memory”.

To conclude, astrocytes have been thought to mainly support neuronal cells, due to their relevance in homeostasis (Simard & Nedergaard, 2004), blood-brain barrier formation and nutrition transport (Poitry-Yamate & Tsacopoulos, 1990; Herrero-Mendez, et al., 2009). However, due to technological advances, it is discovered that astrocytes could also play an active role in cognitive processes (Lee, et al., 2014; Suzuki, et al., 2011), which will be discussed in the next chapter.

Astrocytes and Cognition

Human beings often consider themselves as the organisms with the greatest mental competency. Researchers have tried to find possible reasons for this intelligence gain during evolution. Since for a long period, neuroscientific research has mainly focused on neurons, many studies have tried to find the explanation by comparing neural cells of different species. However, the increase in cognitive abilities could not be fully addressed to neuronal quantity, shape, diversity or brain size in comparison to body size (Roth & Dicke, 2005; DeFelipe, et al., 2002). In contrast, recent studies indicate that not neurons, but astrocytes have transformed substantially during mammalian evolution. Especially protoplasmic astrocytes have evolved considerably. Protoplasmic glia can be found in the grey matter and are the most prevalent astrocytes within the human brain. These glial cells are much larger and complex compared to rodent protoplasmic astrocytes. Furthermore, human astrocytes are highly symmetrical, whereas protoplasmic glia in rodents are not (Oberheim, et al., 2006). In addition, calcium waves propagate faster in hominid astrocytes compared to murine glial cells (Han, et al., 2013). These observations indicate that increase in human intellectual capabilities are not due to the evolution of neurons, but could be addressed to the transformation of astrocytes into more complex glial cells.

Research by Han and colleagues provides evidence for the hypothesis that astrocytic transformations have highly influenced the cognitive competence of human beings. In this study the impact of hominid glial cells in rodents was investigated. Chimeric mice were developed in which human glial progenitor cells were xenografted to incorporate human astrocytes into the rodent brain. It was observed that these progenitor cells could develop into protoplasmic astrocytes. The xenografted

(9)

hominid glial cells seemed to have the large and complex form similar to astrocytes observed in the human brain (Han, et al., 2013). It was also recognized that the xenografted astrocytes were dispersed along many different brain areas in the rodent brain (e.g. hippocampus, neocortex, amygdala). Over time, hominid glia replaced most of the murine astrocyte cell population and often entirely took over the rodent glial cell population (Windrem, et al., 2014). Strikingly, it appears that the incorporation of human astrocytes enhances LTP in the hippocampus of chimeric rodents and this increase in EPSPs can be observed for a longer period in hominid glia mice compared to mouse glia progenitor engrafted mice. Since LTP is a mechanism that is associated with learning and memory, various learning tasks were carried out. Rodents with human glia performed better on different fear conditioning tasks (auditory and contextual fear conditioning). Spatial learning and memory, investigated by a ‘Barnes maze’, was also improved in chimeric mice. Lastly, a novel object location test was conducted in order to test spatial memory. Again, human glia chimeric mice executed this task better than the controls (Han, et al., 2013). To conclude, this study suggests that the evolution of astrocytes, rather than the transformation of neurons, could be the cause of enhanced intellectual capabilities of human beings compared to other species. This could be due to the fact that larger hominid glial cells enwrap more synapses compared to smaller rodent glial cells (Oberheim, et al., 2006). Therefore, the human cells could be able to integrate and propagate more information. It would be interesting to investigate whether glia of other species than homo sapiens have a similar effect on memory in chimeric mice in order to investigate the evolutionary development of glial cells and its influence on cognition. Although, the human glia engrafted mouse model is an intriguing new tool to study human astrocytes with promising results, the influence of the micro-environment cannot be ruled out. As also considered in this review, brain cells are highly interactive and can bidirectionally influence cellular processes (Araque, et al., 1999; Nedergaard, 1994). Many cell types in mice are both genetically and phenotypically strikingly different from the human counterparts. Therefore, hominid astroglia could behave differently in the environment of a mouse brain. Furthermore, future research is necessary to confirm the results obtained by Han and colleagues (2013).

Results obtained by Jahanshahi et al., (2008) implied that spatial learning induced by the Morris water maze increases the number of astrocytes in hippocampal area CA3. They hypothesised that this rise in glial cells could be due to the close interaction between neurons and astrocytes (Jahanshahi, et al., 2008). More specifically, during learning a process called neurogenesis occurs. Neurogenesis refers to the integration of new neurons within the hippocampal formation (Shors, et al., 2001). Furthermore, neurogenesis enhances memory performance (Bruel-Jungerman, et al., 2007). The origination of new astrocytes during learning could be a consequence of this phenomenon, since astrocytes are in close connection with neuronal synapses (Araque, et al., 1999), which could indicate that neuronal quantities positively correlate with astrocytes. A rise in glial quantity can be caused either by migration of astrocytes to this area or by glial proliferation. However, astrocytic proliferation is mainly a response observed during brain injury and disease. In order to propagate, astroglia experience a sequence of events that will induce conversion into reactive astrocytes. This glial state is highly different from the healthy condition (Sofroniew, 2009). Furthermore, Jahanshahi and colleagues visualized astrocytes by a phosphotanguestic acid haematoxylin (PTAH) staining. They declare that this is ‘the special staining method for astrocyte cells and their processes’ (Jahanshahi, et al., 2008), but it is not mentioned that this method is especially suitable for reactive glia (Bancroft & Stevens, 1990). Therefore, it seems that this is not an appropriate method to investigate mechanisms in the healthy central nervous system.

(10)

Compelling evidence for alterations of astrocytes upon memory processes comes from research on exercise (Brockett, et al., 2015). Physical activity is known to enhance both hippocampal neurogenesis and cognitive performance (Brown, et al., 2003). Investigators from Princeton University demonstrated that running, except from neurons, affects astrocytes as well. In this study, one group of rats was provided with a running wheel. The control group did not have this activity opportunity. Subjects were tested on object memory, including object in place and novel object preference. The rodents were also exposed to a cognitive flexibility task. The hippocampus and medial prefrontal cortex are important for the object in place tasks. Object preference depends on the perirhinal cortex and cognitive flexibility is mediated by the orbitofrontal cortex. The results revealed that rats who had the possibility to run performed better on the cognitive flexibility and object recognition tasks. However, there was no significant difference in object preference. This indicates that running has a positive influence on cognitive tasks involving the hippocampus, orbitofrontal cortex and medial frontal cortex, but not on tests that rely on the perirhinal cortex. Physical exercise led to an elevated number of neurons in all mentioned brain areas. The elevation in dendritic spine and synaptic density in the perirhinal cortex without enhanced object preference implies that the occurrence of neurogenesis does not always improve memory, but that other cellular mechanisms should also be considered. Therefore, cell body size of astrocytes in these areas was quantified by S100 labelling. It was observed that physical exercise caused astrocytic enlargement in the medial prefrontal cortex, orbitofrontal cortex and hippocampus. However, glial volume remained similar in the perirhinal cortex (Brockett, et al., 2015). These results reveal that glial cells, just like neurons, can have a form of plasticity. But, it also implies that, at least in these four areas, predictions of cognitive performance should be better based on astrocytic alterations than neuronal density. However, from this study it is not clear if the increased number of neurons induces astrocytic transformations or that glial cell body increase stimulates neurogenesis. Astrocytes can secrete growth factors (such as interleukins) that induce neural differentiation (Barkho, et al., 2006). Accordingly, it could be hypothesized that glial enlargement enhances growth factor production, which causes elevations in neurogenesis, but this idea remains to be verified. In conclusion, these studies suggest physical exercise induces morphological changes in astrocytes. Cognitive performance can be better derived from astrocytic changes than from synaptic and dendritic spine density. (Brockett, et al., 2015). Future research is required to investigate whether there is a causal connection between the formation of novel neurons and astrocytes.

Lee et al., (2014) provided more direct evidence for the importance of normal astrocyte functioning in learning and memory. In this study the influence of glial vesicle release on gamma oscillations was investigated. Gamma oscillations (25-100Hz) in the brain are closely associated with a wide range of cognitive processes, including learning and memory consolidation (Jensen, et al., 2007). It is hypothesized that gamma oscillations promote neural synchronization. Synchronized activity summates into the receiving neuron, which increases the chance of action potentials. Gamma waves are fast oscillations, which creates a small time window for neuronal firing. Decreased activity opportunity of neurons results in increased synchronization (Basar-Erogl, et al., 1996). In the study by Lee and his colleagues, cholinergic agonist carbachol (CCH) was applied to hippocampal slices. This drug binds to the acetylcholine receptor, which causes persistent neural oscillations in the gamma range. Calcium imaging revealed that concurrent with these fluctuations, Ca2+

concentrations increased in astrocytes (Lee, et al., 2014). Glial calcium waves are known to cause gliotransmitter release (Nedergaard, 1994) (see also the paragraph ‘communication via calcium waves’). In order to disrupt gliotransmitter release by astrocytes, a genetic construct was developed that induced tetanus toxin (TeNT) expression in astroglia. TeNT was transported into astrocytes by

(11)

a lentiviral vector combined with human Glial Fibrillary Acidic Protein (hGFAP). hGFAP is considered to be an astrocyte-specific promotor. The lentivirus was also coupled with green fluorescent protein (GFP) in order to verify that TeNT was only present in glial cells and did not affect neurons. It was confirmed that lenti-hGFAP-TeNT -GFP inhibited glutamate release from cultured astrocytes. Furthermore, immunohistochemistry in brain slices indicated that GFP co-localized with GFAP, a major component of the astrocytic cytoskeleton, and not with NeuN, a neuronal marker. The results indicated that if astrocytic vesicular fusion was suppressed by tetanus neurotoxin presentation, in vitro gamma oscillations were of shorter duration. These observations suggest that astrocytes are not important for the generation of gamma oscillations, but glial cells could play a role in the maintenance of these fluctuations. In order to investigate these results in vivo, transgenic mice were developed in which tetanus neurotoxin expression in astrocytes could be modulated by doxycycline administration. In vivo inhibition of astrocytic gliotransmitter release resulted in a low-gamma (20-40Hz) power decrease. Finally, the influence of astrocytic transmitter release impairments on cognition was examined. The data revealed that vesicle fusion prevention in vivo in astrocytes did not influence fear conditioning or performance on a Y maze, which is a maze that allows testing the tendency of mice to explore novel environments. Yet, loss of astrocytic functioning resulted in impaired novel object recognition performance. Specifically, TeNT expression in transgenic mice decreased time spend with the novel object compared to control animals. Furthermore, inhibition of TeNT rescued this effect. This research indicated that impaired astrocytic functioning changes the properties of gamma oscillations, which leads to deficits in recognition memory (Lee, et al., 2014). However, brain waves were not recorded while the mice were engaged in the task, so it would be interesting to investigate whether there is a correlation between gamma wave duration and task performance. Furthermore, astrocytes are highly heterogeneous and although GFAP is the most widely astrocytic marker, not all astroglia express this protein (Kimelberg, 2009). For that reason, it is probable that groups of glial cells did not express TeNT, which implies that the suppression of gamma oscillations could be larger.

In brief, human intelligence could not be explained by the evolution of neurons. Astrocytes have been subjected to greater transformations over time, being larger and more symmetrical in homo

sapiens (Han, et al., 2013; Windrem, et al., 2014). Research indicates that size of astrocytes increases

by physical activity. Exercise is known to induce neurogenesis and enhances performance on memory task (Brown & Ransom, 2007). Furthermore, study suggests that glial alteration are better predictors for cognitive performance than the occurrence of neurogenesis (Brockett, et al., 2015)., which could imply that astrocytes play a major role in memory. Yet, it is not clear whether task performance is better due to astrocytic plasticity or that these cells change after improved memory. Lastly, gamma oscillations are associated with memory consolidation. Research revealed that impairing astrocytic functioning reduces duration and power of these waves. Moreover, recognition memory was worse in mice that lacked vesicle release form astrocytes in the hippocampus (Lee, et al., 2014). These studies indicate that interfering with astrocytic functioning ultimately leads to changes in cognition. However, the mechanisms behind these changes still remain largely unidentified. This paper will continue discussing what is known about glial cells and how this could influence memory.

Communication via calcium waves and astrocytic networks

The observation that astrocytes convey ion channels and membrane receptors that can interact with neurotransmitters led to the hypothesis that astroglia are able to respond to neural activity (Barres,

(12)

et al., 1990; Chaban, et al., 2004; McCarthy & Salm, 1991). This assumption was studied by McCarthy and Salm (1991). In this research, astrocytic calcium levels, after administration of various neuroligands, were determined. Neuroligand application to glial cultures caused intercellular Ca2+

level increases in these cells, which indicates that astrocytes have a form of excitability by Ca2+

concentration alterations. Moreover, the height of the calcium response differed between cells and neuroligands. Occasionally, neuroligands induced Ca2+_{oscillations in distinct astrocytic populations}

(McCarthy & Salm, 1991). This discrepancy in calcium response between glial populations indicates that there are major differences between astrocytes and their receptor composition. However, often is the cultured condition not similar to the situation in brain slices or in vivo. Accordingly, future studies should elucidate that this heterogeneity of glia can also be observed under circumstances alike the natural situation. Interestingly, Shao and McCarthy (1994) revealed that subpopulation of astrocytes express different receptors and this receptor composition changes during development, but can also alter during adulthood, for example, due to changes in neuronal firing (Shao & McCarthy, 1994). These observations suggest that besides neuronal plasticity, an ‘astrocytic plasticity’ mechanism influences glial properties.

Various papers suggest that astrocytes have a specific Ca2+_{mediated communication system in which}

calcium can propagate into different astrocytes that form an astrocytic network (Hirase, et al., 2004; Finkbeiner, 1992). There is still debate whether astrocytic Ca2+_{signalling influences LTP. Several}

studies from the lab of Prof. Dr. Ken McCarthy did not find any alterations of neuronal Ca2+

-concentrations or EPSPs after uncaging or inhibiting astrocytic Ca2+_{release in transgenic mice}

affecting Gq-G protein-coupled receptors coupled receptors (Gq GPCRs), which are receptors that

mediate Ca2+_{transients in astroglia (Agulhon, et al., 2010; Fiacco, et al., 2007). On the contrary, there}

is also compelling evidence that glial Ca2+_{concentrations influence LTP. For example, Prof. Dr.}

Nedergaard showed neuronal Ca2+_{concentrations elevations after stimulating glial cells, which also}

induced a calcium increase in astrocytes (Nedergaard, 1994). Another study carried out by using various in vivo and in vitro techniques revealed that cholinergic induced LTP is influenced by intercellular astrocytic Ca2+_{concentrations (Navarette, et al., 2012). The contradicting results could}

be caused by differences in methodology between these studies.

In order to study Ca2+_{spread in astrocytes, Hirase and his colleagues conducted an in vivo 2-photon}

calcium study in the cortex of rats. The results indicated that increased neural activity caused elevated astrocytic calcium concentrations. Moreover, these calcium elevations were transmitted to neighbouring cells after neuronal bursts. Spontaneous astrocytic calcium concentration elevations did not cause propagation into adjacent glial cells, which could imply that glial network activation is more prevalent after neuronal activity (Hirase, et al., 2004). Remarkably, astrocytes can transport glucose from distal blood vessels towards sides with increased neural activity patterns (Rouach, et al., 2008). This implies that astrocytes can modify their network for direct delivery of neuronal energy sources.

Finkbeiner (1992) was the first to observe that gap junctions are important for intracellular astrocytic Ca2+_{wave propagation. In this research, firstly, the effect of extracellular glutamate flow on the}

velocity and movement of intercellular calcium waves was examined, which revealed that intercellular calcium waves, caused by glutamate administration, were not affected by the extracellular solution flow. Subsequently, the effect of two gap junction channel blockers, octanol and halothane, on calcium wave propagation was investigated. The results indicated that adding a gap junction blocker impaired intercellular calcium wave propagation. However, the calcium spread within cells remained intact (Finkbeiner, 1992). The presence of gap junctions between astrocytes,

(13)

also in other brain areas, has been verified by numerous other studies (Blomstrand, et al., 1999; Giaume, et al., 1991; Venance, et al., 1995). The mechanisms behind intracellular calcium waves are mediated by inositol trisphosphate (IP3). IP3 can be transported to the cytosol of neighbouring cells,

since it can cross gap junctions (Giaume & Venance, 1998). IP3 receptors (IP3Rs) are situated on the

endoplasmic reticulum (ER), which is a major component that is associated with calcium ion concentration regulation. Upon IP3R activation, calcium is released within the cell (Scemes, 2000).

Subsequently, calcium is released from multiple release sites in an astrocyte. In addition, Ca2+_wave

frequency and amplitude is higher in release regions (Yagodin, et al., 1995). In order to provide evidence for the importance of IP3Rs in astrocytic calcium wave generation, Sheppard and his

colleagues (1997) conducted a study in which calcium release locations were compared with IP3R

occurrence sites. The results indicate that IP3R concentration is higher in regions where calcium is

released (Sheppard, et al., 1997).

Severely limiting gap junctional conductance did not completely suppress Ca2+_{propagation between}

glial cells (Scemes, et al., 1998), which implies that there is also another calcium propagation mechanism between astrocytes. Besides an intercellular calcium wave spread, also an extracellular calcium propagation mechanism has been discovered. Hassinger and colleagues obtained cortical astrocytes from mice. Subsequently, cell-free regions, with various widths, were constructed between these cells, disabling the formation of cytoplasm connecting channels. Using electrical stimulation, calcium waves were generated in target astrocytes. It was shown that Ca2+_{waves spread to}

neighbouring astrocytes when the possibility of junction formation was eliminated. In addition, the waves propagated with the same velocity between non-coupled glial cells as between confluent astrocytes. Extracellular calcium spread likelihood decreased with increasing cell-free area width and was possible up to 120µm cell separation (Hassinger, et al., 1996). In order to understand the mechanism behind this extracellular communication, similar experiments were conducted in which a cell-free area was created between astrocytes. Electrical stimulation induced calcium waves in astrocytes and samples of the extracellular fluid were obtained during calcium propagation. Subsequently, it was demonstrated that these samples contained ATP. In order to verify that calcium spread is ATP dependent, purinergic receptor agonists were administered to the solution, which led to impaired astrocytic extracellular communication (Guthrie, et al., 1999). ATP is released from glia via connexin hemichannels that connect astrocytic cytosol with the extracellular space (Stout, et al., 2002). ATP binds to purinergic P2Y receptors (P2YR), a class of GPCRs located in the astrocytic membrane. This will initiate the phosphatidylinositol pathway. Phospholipase C-β (PLCβ) is activated, causing IP3 production. IP3 will induce Ca2+_{within neighbouring glia (Sheppard, et al.,}

1997). In summary, gap junctions are important connections between astrocytes that enable Ca2+

wave spread between these cells (Finkbeiner, 1992). However, removing these direct links between astrocytes still results calcium wave propagation via the release of glial ATP (Hassinger, et al., 1996; Scemes, 2000).

The importance of glial network formation for learning and memory is studied by inhibiting astrocytic gap junction formation in mice. This can be established by developing knock-out mice for astrocytic gap junction proteins, connexin30 (Cx30) and Cx43. These animals were tested by the object-location memory task, which is based on the tendency of mice to explore novel situations. In this paradigm the rodents could inspect the arena with two identical objects for 3min followed by a 10min recollection period. After 10min the mice were placed back in the testing arena where one object was relocated. The knock-out mice did not spend more time with the moved item, suggesting deficient memory recall. Interestingly, synaptic density remained unchanged in knock-out mice

(14)

(Lutz, et al., 2009), indicating that the number of neurons was not responsible for this result. Furthermore, this outcome suggests that normal astrocytic functioning is not required for neuronal survival.

In conclusion, astrocytes respond to the release of neurotransmitters with Ca2+_{waves. These waves}

can propagate to other glial cells via gap junctions due to cytoplasmic exchange of IP3. IP3 binds to IP3R on the ER, which induces calcium release in the cell (Finkbeiner, 1992; Hassinger, et al., 1996). Disabling junction formation still results in Ca2+_{wave diffusion, implying the existence of}

extracellular communication between glial cells. This is accomplished by glial ATP release, which acts upon astrocytic GPCRs via PLCβ promotes IP3 production. Gap junctions allow the formation of large astrocytic networks. Transport through these glial networks can be highly precise (Rouach, et al., 2008). Knocking down network formation decreases spatial memory performance (Lutz, et al., 2009). Additionally, the vast majority of research done in the field of astrocytic calcium waves is done

in vitro, so it is important to replicate these data in living animals. Moreover, many studies regarding

Ca2+_{waves require pharmaceutical applications, which does not exclude interference with other}

cellular processes. It would be interesting to investigate whether calcium waves that are induced by neuronal activation in one area affect neuronal behaviour in another part of the brain. As described in the first chapter, is it still a not fully resolved how a memory trace can relocate from the hippocampus to the cortex (Bontempi, et al., 1999). Although neurons are able to transmit signals over large distances (Li, et al., 2014), it is a possibility that astrocytic networks facilitate systems consolidation, because glial systems are widespread and can direct signals highly precise that, for example, information from the hippocampus can be directed to cortical areas. However, still a lot of research needs to be performed to proof any validity of this hypothesis.

Astrocytes in ensemble with neurons: sinapsis tripartita

Pfrieger and Barres were the first to discover that neurons are more reactive when these cells are surrounded by glial cells. In this in vitro study, retinal ganglion cells were cultured with or without glia. Neuronal reactivity was quantified by measuring spontaneous EPSPs from the ganglion cells. It was observed that culturing ganglion cells with either oligodendrocytes or astrocytes increased frequency and amplitude of EPSPs. In addition, it seems that these changes are due to enhanced synapse efficacy, because culturing ganglion cells without any glia results in the same synapse density (Pfrieger & Barres, 1997). Further evidence for astrocytic influences on neuronal behaviour was found by electrically stimulating one glial cell. As expected, electrical stimulation led to Ca2+_{elevations in}

the target cell and propagated to neighbouring glia, but neurons that were connected to these cells also responded with Ca2+_{elevations (Nedergaard, 1994). Furthermore, astrocytes are just like neurons}

capable of chemical communication, in this case referred to as gliotransmission (see chapter Gliotransmitters and their effect on learning and memory). Similar to neuronal transmission, depends the release of gliotransmitters on vesicle fusion in the membrane of astrocytes by the formation of SNARE-complexes (Araque, et al., 2000). Previous research indicates that one of the transmitters released by glia is glutamate, which can bind to neuronal NMDA-receptors (Nedergaard, 1994). As explained in the first chapter about the classical view of learning and memory, binding of glutamate induces Ca2+_{responses in neurons, which is able to activate intercellular pathways in}

neural cells (Sweatt, 1999). This illustrates that gliotransmitters could have an effect on neuronal mechanisms (also see chapter ‘Gliotransmitters’, paragraph glutamate). These findings, together with the observations that astrocytes respond to neurotransmitter (McCarthy & Salm, 1991), contributed

(15)

to the hypothesized that there is tri-directional signal transmission between glial cells and neurons. Alfonso Araque was the first to introduce the term ‘tripartite synapse’. As can be observed in figure 3, a ‘tripartite synapse’ is a synapse in which not only pre- and post- synaptic communication is possible, but there is also information flow between astrocytes and neurons (Araque, et al., 1999).

Figure 3. The tripartite synapse. Neuronal activity induces the release of neurotransmitters into the synaptic

cleft. These transmitters will bind to receptors on the post-synaptic neuron, but will also trigger astrocytic receptors. Activation of glial receptors results in gliotransmitters release, which in turn can interact with neuronal receptors. This mechanism causes a tri-directional chemical transmitter exchange termed tripartite synapse (figure adapted from Araque, et al., 1999).

Studies carried out by electron microscopy visualized tripartite synapses in the hippocampus of fixed brain slices by three-dimensional reconstructions. The processes regarding astrocytes within these synapses are referred to as perisynaptic astrocyte processes (PAPs). In the slices it was observed that PAPs were more often observed at bigger synapses, which are considered more stable neural connections (Witcher, et al., 2007). Interestingly, not all synapses are enclosed by PAPs. In the hippocampus 60% of neurons are in contact with perisynaptic astrocytes (Witcher, et al., 2007) and in the barrel cortex this ratio is approximately 90% (Genoud, et al., 2006), which indicates that this composition is not requisite for synapse formation. Although this study showed that perisynaptic astrocytes can have different conformations, the static slices do not give any information whether PAPs can be altered by neuronal activation. In order to answer this question, whisker stimulation on mice was performed. After 24hours astrocytes in the barrel cortex of stimulated animals were compared with control. The results revealed that neuronal activation increased enclosement of neurons by PAPs (Witcher, et al., 2007). Furthermore, two-photon time lapse imaging of astrocytes

(16)

that are in contact with dendritic spines illustrated that these processes facilitate stabilisation of spines and occur with a high motility (Nishida & Okabe, 2007).

In conclusion, previous research revealed that except chemical communication between the pre- and post-synapse there is also bidirectional transmission from neurons to astrocytes (McCarthy & Salm, 1991; Nedergaard, 1994; Pfrieger & Barres, 1997; Sweatt, 1999). The combination of these findings formed the basis of the tripartite synapse hypotheses (Araque, et al., 2000). Imaging of these processes revealed that enwrapment of astrocytes results in larger synapses and that neuronal stimulation causes the glial cells to increase enclosement of the synaptic cleft (Witcher, et al., 2007). Moreover, these processes are largely motile (Nishida & Okabe, 2007). These studies demonstrated that astrocyte confirmations are plastic and it implicates that astrocytes could guide synaptic efficiency. However, all these observations were acquired post-mortem. Further studies will need to record these processes real time in vivo to visualize the exact structural changes of astroglia after neuronal activation.

Astro-neuronal chemical communication and the effect on

learning and memory

As described earlier in this paper, upon calcium release in the cell, astrocytes can secrete neuroactive molecules called gliotransmitters (Nedergaard, 1994). Astrocytes can secrete the same chemical transmitters as neurons, for example ATP (Guthrie, et al., 1999; Stout, et al., 2002), D-serine (Mothet, et al., 2000) and glutamate (Nedergaard, 1994). In this chapter astrocytic chemical transmitters and their influence on LTP and memory, will be addressed. Table 2 shows gliotransmitters that have been investigated in association with learning and memory.

Adenosine Triphosphate (ATP)

As explained in the chapter about calcium waves in astrocytes, ATP can be secreted from glial cells to induce non- gap junctional calcium wave propagation (Guthrie, et al., 1999; Stout, et al., 2002). Additionally, astrocytic ATP can have an effect on neurons. Neuronal glutamate release induces ATP secretion from glia. Additionally, electrophysiological recordings of hippocampal neurons and astrocytes in co-cultures and hippocampal slices has revealed that ATP application supresses post-synaptic EPSPs of excitatory synapses by binding to P2Y receptors. Furthermore, ATP inhibits Ca2+

-receptor opening on the pre-synapse, which leads to a reduction in neurotransmitter release into the synaptic cleft (Zhang, et al., 2003). This event could lead to reduced LTP establishment.

Evidence that astrocytic ATP could have an effect on memory was provided by a study with domestic chicks. Memory was tested by a discriminative avoidance task, which tests reinforced learning. In this task chicks were taught to peck at red and blue beads. Subsequently, the red bead was covered with an aversive substance. Reinforcement learning was measured by counting the pecks at the red and blue bead. Chicks that remembered the task would peck less at the red ones, since the animals remembered an uncomfortable experience. In order to discriminate between memory stages, the birds were examined 10min (short-term memory), 30min (intermediate) or more than 60min (long-term memory) after acquisition. The results indicated that ATP injection in the hippocampus improves long-term memory. Furthermore, these effects are mediated by the P2Y1 receptor of hippocampal interneurons, since applying an antagonist for this receptor causes a reduction of

(17)

performance on the discrimination task. In order to investigate whether the learning effect is caused by gliotransmission, thrombin and fluoracetate (FAc) and thrombin (Gibbs, et al., 2011). Thrombin stimulates Ca2+_{release in glial cells and FAc inhibits metabolism in these cells. Application of}

thrombin causes memory improvement at the same times as ATP enhances task performance. These effects are reversed by FAc administration. Although this study suggests that ATP influencing memory comes from astrocytes, other mechanisms cannot be ruled out. Firstly, neurons are also able to secrete this adenosine, so neuronal ATP can contribute to this learning effect. Furthermore, next to ATP, astrocytes also excrete other gliotransmitters. To specify, applying these drugs does not only inhibit ATP secretion, but influences whole astrocytic functioning and, as further described in this chapter, other gliotransmitters also play a role in learning and memory.

In summary, Zhang et al., (2013) found a reduction of LTP after astrocytic secreted ATP elevations. On the other hand, Gibbs and his colleagues (2011) observed memory improvement after ATP injection, which is caused by binding of ATP to the P2Y1 receptor. Furthermore, it was implied that gliotransmission is responsible for these effects. Although these results seem to be contradictory, this does not necessarily have to be the case, since P2Y1 receptors are present on hippocampal interneurons (Gibbs, et al., 2011). Reduction of interneuronal LTP decreases GABAergic signalling, causing an increase in pyramidal neuron activity (Kullmann & Lamsa, 2007).

D-serine

D-serine is derived from L-serine by the serine racemase enzyme and co-activates NMDA-receptors

in the presence of glutamate. This neuromodulator is primarily synthesized in protoplasmic astrocytes around synaptic clefts with a high density of NMDA receptors (Mothet, et al., 2000). Furthermore, an ingenious study found a correlation between astrocytic enwrapment around the synapse and D-serine concentration. This research was based on the observation that in lactating animals PAPs withdraw from the nerve terminal which caused D-serine reduction. In order to investigate the effects of this amino acid on neuronal processes, electrophysiological recordings were carried out. The results indicate that D-serine concentration decrease causes LTD, which implies that this molecule sustains LTP (Panatier, et al., 2006). However, serine racemase has also been observed in neurons, which suggest that the effects on LTP can also be mediated by neuronal D-serine. Therefore, a study was carried out that involved astrocytic or neuronal selective knock-out mice targeting this enzyme. The data gathered in this study suggests that LTP in CA1 is more dependent on neural than on astrocytic D-serine coactivation (Benneyworth, et al., 2012).

Research on the effects of nicotine contradicts the finding that neural D-serine mediates LTP by demonstrating that glial D-serine plays a role in long-term memory. Nicotine, mainly known as the addictive substance in tobacco, can increase synaptic potentiation and is also known to induce Ca2+

concentration elevations in astrocytes. This alkaloid can act upon astrocytes due to the expression of nicotinic acetylcholine receptors (nAChRs) on glial cells causing a reduction in K+_-currents

(Hernández-Morales & García-Colunga, 2014). Impairing astrocytic functioning reduced the effect of nicotine and impaired long-term memory assessed by the inhibitory avoidance task. However, this effect was rescued by D-serine application (López-Hidalgo, et al., 2012).

To summarise, D-serine is mainly synthesized by astrocytes and (Mothet, et al., 2000) astrocytic enwrapment of synapses positively correlates with the concentration of this neuromodulator (Panatier, et al., 2006). Reduction of D-serine levels induces LTD. However, there is still debate

(18)

whether the D-serine affecting LTP is from neural or astrocytic origin (Benneyworth, et al., 2012; López-Hidalgo, et al., 2012).

γ -Aminobutyric acid (GABA)

γ-Aminobutyric acid, or GABA, is the principal inhibitory neurotransmitter in the central nervous

system. GABA operates in an inhibitory manner by inducing hyperpolarization in the post-synaptic neuron. Although most GABAergic neurons are believed to be inhibitory, some can also have an excitatory effect (Szabadics, et al., 2006).

Previous research indicates that GABA derived from astrocytes induces tonic inhibition in neurons in the cerebellum. GABA is released through Bestrophin 1 (Best1) channels in the glial membrane (Lee, et al., 2010). Furthermore, glial GABA concentrations correlates with these inhibitory currents. Besides in the cerebellum, the anion channel Best1 is situated on astrocytes in the hippocampus. However, in this memory-associated area, no astrocytic GABA is detected and tonic inhibition is also sparse (Yoon, et al., 2011). The small amount of research investigating the influence of tonic inhibition on memory suggests that tonic inhibition reduces LTP (Martlin, et al., 2010), which implies that glial GABA release inhibits LTP. However, due to the absence of this gliotransmitters in the hippocampus, there are no indications that this affects learning and memory.

Research on Alzheimer’s disease (AD) implies that GABA release from reactive astrocytes plays a role in the cognitive impairments observed in this condition. In a mouse model for AD it was established that, unlike healthy hippocampal glial cells, reactive astrocytes around β-amyloid plaques in the hippocampus produce GABA. Furthermore, a passive avoidance experiment and the Morris water maze test revealed that mice with induced AD had learning and memory deficits. However, feeding these mice selegiline, a Monoamine Oxidase-B (MAO-B), reversed the effects in the passive avoidance task. MAO-B synthesizes GABA in glial cells, therefore, inhibiting this protein reduces GABA production. Selegiline application partially restored the performance on the Morris water maze. It was hypothesized that selegiline had a stronger effect on the passive avoidance task, because the MAO-B inhibitor was applied for a shorter period during these experiments and this drug can lose its efficacy over time. Electrophysiological recordings showed that LTP in the dentate gyrus was decreased in AD mice, but selegiline application rescued this impairment (Jo, et al., 2014). To conclude, this elegant study reveals that GABA release from hippocampal reactive astrocytes can have an effect learning and memory. However, since GABA is not produced by normal astroglia in the hippocampus, it is highly unlikely that this astrocytic GABA in a healthy brain has any effect on hippocampus dependent tasks.

Multiple studies revealed that GABA derived from astrocytes can influence neural activity (Lee, et al., 2010; Yoon, et al., 2011). Furthermore, GABA from reactive astrocytes in the hippocampus play a role in the cognitive deficits that can be observed in AD (Jo, et al., 2014). However, it seems that there is insufficient evidence regarding this subject to draw any firm conclusion about astrocytic GABA in a healthy brain having an influence on learning and memory. The absence of GABA in the hippocampus implies that this gliotransmitter does not play a role in short-term memory. Future studies should investigate the role of astrocytic GABA in the cortex and long-term memory. Furthermore, cell and area-specific Best1 knock-out mice could give more information about the contribution of this channel to cognition.

(19)

Glutamate

As described in the chapter about the classical view of learning and memory, glutamate is an excitatory neurotransmitter, which is strongly linked to LTP and cognition (Zola-Morgan & Squire, 1986; Sweatt, 1999). Glutamate release by astrocytes has been mentioned repeatedly in this paper and is also the widest studied gliotransmitter (Navarette, et al., 2012; Parpura, et al., 1994).

Parpura and his colleagues were the first to demonstrate astrocytic glutamate release upon intercellular Ca2+_{-concentration increase by stimulating the glial cells with bradykinin in vitro.}

Subsequently, a Ca2+_{-increase could be observed in neighbouring neurons. This effect was mediated}

by the glutamatergic NMDA-receptor (Parpura, et al., 1994). In vivo evidence for the role of glutamate in LTP was provided by a study that performed calcium-imaging and electrophysiological recordings in alive rodents. Astrocytic Ca2+_{-concentrations were increased by inducing cholinergic activity,}

which caused LTP in synapses between hippocampal area CA3 and CA1. Antagonist of either muscarinic receptors or metabotropic glutamate receptors (mGluRs) diminished this effect. Subsequently, by modifying astrocytic intercellular calcium release, it was demonstrated that glial Ca2+_{elevations are necessary for this type of LTP induction and that this is probably because of the}

release of astrocytic glutamate, activating mGluRs (Navarette, et al., 2012). In conclusion, glutamate is the primary excitatory neurotransmitter and induces LTP. Astrocytes could enhance the chances of LTP by secreting additional glutamate into the synaptic cleft, which could increase the Ca2+

-concentration in the post-synaptic neuron.

Despite the evidence that astrocytic glutamate is the leading factor in LTP induction mediated by glutamatergic receptors, involvement of other gliotransmitters cannot be ruled out. Astrocytic vesicles contain various chemical molecules. For example, as mentioned before, D-serine is also a glutamate receptor agonist and could also contribute to the effects on LTP (Mothet, et al., 2000). It would be interesting to investigate to which extend glial glutamate is responsible for LTP induction, however, investigating the influence of this gliotransmitter on cognition is rather difficult due to astrocytic glutamate transporters causing rapid clearance (Anderson & Swanson, 2000), which makes it difficult to detect.

Lactate

There are two pathways that lead to lactate production. One way of lactate formation is through glycolysis, which is a process that degrades glucose to pyruvate. Subsequently, pyruvate is metabolised into lactate. Apart from glucose, energy can also be derived from glycogen, which functions as energy storage since it is multibranched polysaccharide of glucose. Stainings of glycogen acid–thiocarbohydrazide-silver protein revealed that the main storage site of glycogen is in astrocytes (Cataldo & Broadwell, 1986). Furthermore, lactate is derived from glycogen via a process referred to as glycogenesis (Bittar, et al., 1996; Gandhi, et al., 2009; Pierre & Pellering, 2005). There is growing support for the claim that lactate is transported from glial cells to neurons (Gandhi, et al., 2009), which is referred to as the lactate shuttle hypothesis (Brooks, 2009).

Research has provided support for the astro-neuronal lactate shuttle. Firstly, lactate accumulation and acceleration is 2-to-4 times higher in astroglia compared to neurons (Gandhi, et al., 2009). Furthermore, during glycolysis, glucose is degraded to pyruvate and subsequently processed into lactate. Previous research indicates that astrocytes contain the enzyme that facilitates this process, called lactate dehydrogenase 5 (LDH-5). On the contrary, LDH-1 is thought to oxidize lactate. Antibodies against LDH-1 can be detected in neurons and astrocytes (Bittar, et al., 1996). This implies

(20)

that lactate is produced in astrocytes and utilized in both neurons and glial cells. Moreover, lactate transporters, called monocarboxylate transporters (MCTs), can be found in both cell types (Pierre & Pellering, 2005).

Recently, research has discovered that lactate has an influence on memory. Suzuki et al., (2001) observed that glial glycogenesis and lactate release in the hippocampus are required for long-term consolidation evoked by inhibitory-avoidance training. Subsequently, the influence of glycogen phosphorylase inhibition on LTP was investigated. Application of a glycogen metabolism blocker, 1,4-dideoxy-1,4-imino-d-arabinitol (DAB), resulted only in a short induction of LTP. However, L-Lactate application administration recovered this phenomenon resulting in prolonged LTP. Knock-down of MCTs also disrupts long-term memory, which implies that lactate-shuttling promotes consolidation. Again, L-Lactate application could reverse these results (Suzuki, et al., 2011). Another study indicates that lactate is also important for spatial working memory, tested by a spontaneous alternation tasks (Newman, et al., 2011). Similar to the data gathered by Suzuki et al., (2011), inhibiting glycogenesis or lactate transport results in decreased performance on the behavioral task and this can be rescued by lactate administration. Hippocampal lactate injections before testing significantly improved task performance. Furthermore, recordings via bio probes carried out in this study demonstrated that extracellular glucose and lactase concentrations are negatively correlated. Lactate increase occurs a little ahead of glucose decrease, suggesting that astrocytes prepare for the reduction of glucose availability by metabolizing glycogen energy storages (Newman, et al., 2011).

In summary, there are two lactate sources, glucose and glycogen. Glycogen is a form of glucose storage that is only observed in astrocytes. Glycogen can be processed in pyruvate which also leads to lactate concentration increases in glial cells. A lactate shuttle hypothesis between astrocytes and neurons has been proposed. Various studies altering enzyme functioning responsible for lactate processing and trafficking support this hypothesis (Bittar, et al., 1996; Gandhi, et al., 2009; Pierre & Pellering, 2005). However, so far there is no direct evidence that lactate is relocated from astrocytes to neurons. Future studies should investigate how this process takes place. There is compelling evidence that astrocytic lactate is required for memory consolidation and that shuttling via transporters is important for this process (Newman, et al., 2011; Suzuki, et al., 2011). During inhibition of glycogen metabolism LTP is not preserved (Suzuki, et al., 2011), suggesting that astrocytes are not important for the initiation of a memory trace, but are required for the maintenance of potentiation.

TNF-α

Another possible gliotransmitter is the cytokine tumour necrosis factor alpha (TNF-α). Presentation of TNF- α to cultured hippocampal neurons increases the expression of AMPA-receptors in the synaptic cleft, which is an indication of enhanced synaptic efficiency. Moreover, this was correlated with the frequency of mEPSCs (Beattie, et al., 2002). Both the application of astrocyte conditioned medium or the administration of tetrodotoxin (TTX), inhibiting neuronal functioning, in combination with TNF-α resulted in enhanced AMPA-receptor density in neurons. This implies that astrocytes were responsible for the receptor increase (Beattie, et al., 2002; Stellwagen & Malenka, 2006).

TNF- α could have contributed to the effect observed in the study by Han et al., (2013) (see chapter the role of astrocytes in cognition). In this article human glial progenitor cells were engrafted in mouse brains resulting in the development of human protoplasmic astrocytes. The results revealed that engrafted mice have increased LTP induction and memory performance. Moreover, the qPCR

(21)

results indicated that TNF- α was expressed in the engrafted mice, but this cytokine was not present in controls. Inhibition of TNF-α reduced the electrical potentiation in chimeric mice and suppressed AMPA-receptor expression (Han, et al., 2013).

Together, these results indicate that TNF- α can increase synaptic efficiency by promoting AMPA-receptor insertion in the neuronal membrane. Furthermore, this cytokine is provided by astrocytes (Beattie, et al., 2002; Stellwagen & Malenka, 2006). Interestingly, the effects on LTP and memory in human astrocyte engrafted mice could be (partially) mediated by TNF- α expression in these glial cells (Han, et al., 2013). Since it appears that TNF- α is less prevalent in rodent glial cells, it would be interesting to investigate the influence of TNF-α on LTP in human neurons.

Table 1. Summary of gliotransmitters and their effect on learning and memory.

Gliotransmitter Effect on memory Experimental method References

ATP 1. ↓ LTP of interneurons

2. ↑ Long-term memory

1. Electrophysiological recording of astro-neuronal co-cultures. FAc administration.

2. In vivo ATP and

thrombin/ FAc application

1. (Zhang, et al., 2003) 2. (Gibbs, et al., 2011)

D-serine 1. ↑ Glial enwrapment ↑ LTP 2.↑ Long-term memory 1. In vitro electrophysiological recordings of the hypothalamus of lactating animals. 2. Behavioural testing combined with microinjection of D-serine and pharmaceutical tools.

1. (Panatier, et al., 2006)

2. (López-Hidalgo, et al., 2012)

GABA 1. Glial GABA released

through Best1-channels and causes tonic inhibition. However, not in the hippocampus.

1. Electrophysiological recordings of slices. GABA concentration was measured by anti-bodies.

1. (Yoon, et al., 2011)

Glutamate 1. ↑ neuronal Ca2+

-concentration after glial stimulation. NMDA-receptor mediated. 2. ↑ LTP induced by glial activation. Mediated by mGluRs. 1. Astro-neuronal co-cultures calcium-imaging. Application of receptor blockers and bradykinin. 2. Electrophysiological and calcium imaging recordings in vivo. 1. (Parpura, et al., 1994) 2. (Navarette, et al., 2012) Lactate 1. ↑ LTP 2. ↑ Memory performance 1. In vivo hippocampal recordings and pharmaceutical applications. 1. (Suzuki, et al., 2011)

(22)

2. Memory retrieval task and bilateral drugs or L-lactate injections. 2. (Newman, et al., 2011) (Suzuki, et al., 2011) TNF-α 1. ↑ AMPA-receptor density and LTP.

1a. Recordings and AMPA-receptor quantification in cultured neurons. Application of astrocyte conditioned medium or TTX. 1b. Human astrocyte engrafted into mice. Inhibition of TNF-α combined with electrophysiological recordings in slices.

1a. (Beattie, et al., 2002)

1b. (Han, et al., 2013)

Concluding remarks

The goal of this literature thesis was to describe the role that astrocytes play in learning and memory. For many years, research on memory has focussed on neurons, due to the discovery of the electrical properties of these cells (Scanziani & Hausser, 2009). However, inserting research on astrocytes could contribute to unravelling the mechanism behind learning and memory, because glial cells appear to be involved in various aspects of consolidation, ranging from regulating extracellular neurotransmitter concentration on the synapse level (Gibbs, et al., 2011; Guthrie, et al., 1999; López-Hidalgo, et al., 2012; Nedergaard, 1994; Newman, et al., 2011; Panatier, et al., 2006; Simard & Nedergaard, 2004; Zhang, et al., 2003) to network formation throughout the brain (Finkbeiner, 1992; Hirase, et al., 2004; McCarthy & Salm, 1991; Rouach, et al., 2008).

Although, it seems that a memory trace is conserved in activity patterns of neurons, astrocytes have a prominent role in the consolidation process. This is illustrated by the notion that gamma-oscillations are induced, but not maintained when glial vesicle release in inhibited (Lee, et al., 2014). Further evidence for this hypothesis comes from the inhibition of glycogen metabolism in astrocytes. Similarly, this intervention results in LTP induction, but the time of electrical potentiation is shorter (Suzuki, et al., 2011). Furthermore, multiple studies indicated that altering astrocytic functioning affects long-term memory, but short-term memory remained intact (Gibbs, et al., 2011; López-Hidalgo, et al., 2012; Suzuki, et al., 2011).

The proposed mechanism for this astrocytic facilitation of memory trace stabilization is represented by the tripartite synapse (Araque, et al., 1999). Firstly, neuronal activity leads to neurotransmitter release (McCarthy & Salm, 1991). When neurotransmitter release is high, it will spill over the synaptic cleft and bind to glial receptors. In turn, Ca2+_{concentration will rise in astrocytes which results in}

gliotransmitters release (Araque, et al., 2000; Nedergaard, 1994). Subsequently, these gliotransmitters will bind to neuronal receptors, inducing different responses depending on the gliotransmitters (Beattie, et al., 2002; Gibbs, et al., 2011; López-Hidalgo, et al., 2012; Navarette, et al., 2012; Newman, et al., 2011; Panatier, et al., 2006; Parpura, et al., 1994; Yoon, et al., 2011; Zhang, et al.,