Astrocyte-Neuron Interaction through Cannabinoid Signalling in Plasticity

(1)

Literature thesis 12 ECTS

06/2017 – 01/2018

Rick van Dorp

Student ID: 10893725

Supervisor: Dr. Pascal J. P. Chameau Co-assessor: Dr. Taco R. Werkman

SILS, Faculty of Science, University of Amsterdam

MSc in Brain and Cognitive Sciences, University of Amsterdam Track Behavioural Neuroscience

(2)

Introduction

Neurons do not just communicate with other neurons and they are definitely not the only information-bearing cells in the central nervous system. In recent years people have come to agree that astrocytes have an important role to play in neural functioning, not just maintaining the synapse environment, but taking part in signalling too. An important part of being an astrocyte is partaking in the establishment and maintenance of neuronal plasticity, modulating nearly every part in the process of plasticity (Barker & Ullian, 2010). This astrocytic function is most obvious in synapses that have astrocyte processes surrounding them: the tripartite synapse, where astrocytes, pre- and postsynaptic neurons communicate and exchange signals to each other. Additionally, it has become quite established that cannabinoids have an important function in plasticity as well, filling the role of intermolecular messenger molecules travelling retrogradely from postsynaptic cells to presynaptic cells. More recent discoveries have highlighted endocannabinoids as a player in astrocyte-neuron communication, starting a cascade of astrocyte-specific signalling like calcium waves, and gliotransmitters that are detected by and acted on by neurons.

The recent discoveries in this area might be very important for all research regarding astrocytes, cannabinoids or plasticity and therefore big gains can be made by looking into the involved pathways and trying to determine their mechanism of action.

In this thesis I combine the ideas of plasticity mediated by astrocytes and plasticity mediated by cannabinoids in order to understand their interplay and discuss the regulation of astrocyte-neuron interaction through cannabinoid signalling and the implications of this interaction for research in plasticity and cannabinoid signalling.

(4)

Basic plasticity

The word plasticity literally means the extent to which something can be moulded and changed and the word is used in a lot of disciplines, ranging from physics to linguistics. In neuroscience, neuroplasticity is a term used broadly to describe a lot of different processes leading to a lasting change in the neuronal structure or functionality. Examples are the functional and morphological change to brain matter on a large scale as a result of injury and the change of susceptibility of a single neuron or group of neurons to signals from another neuron. In this particular thesis, the focus will be synaptic plasticity plasticity: the modulation of susceptibility to signals between neurons and the strength of synaptic connections.

This property of plasticity is essential to the functioning of the nervous system and is one of its core principles, because it is responsible for maintaining and evaluating connections.

Due to synaptic plasticity, important connections are strengthened while irrelevant connections are weakened. Thus, synaptic plasticity increases the accuracy of the involved neurons and improves the relevance of the transmitted information. The process takes place during an individuals whole life and lies at the basis of important processes, like sensitization, learning and association. Synaptic plasticity can occur on different scales, often on the single synapse level, but large-scale simultaneous changes on numerous different synapses and neurons are not uncommon.

There are many pathways through which synaptic plasticity can be established, usually the plasticity occurs as a result of a deviation in the firing speed or pattern of the presynaptic neuron. This might trigger a specific response in the postsynaptic cell, that leads to a cascade of pathways. The response pathway can affect the synaptic signalling of the presynaptic neuron, through, for example the inhibition of neurotransmitter release or it can affect the susceptibility of the postsynaptic neuron by, for example, increasing the amount of receptors. In the most extreme cases this can cause a change in the connection of neurons, from complete abolishment of specific neural connections to the formation of completely new connections.

Plasticity can occur on many timescales, ranging from milliseconds to tens of seconds to much longer, but generally a distinction is made between long-term and short-term plasticity, as the two types can occur independently of each other at the same synapse and the pathways are slightly different. Additionally, plasticity can both enhance or diminish signalling between neurons. In this matter it is important to realize the difference between excitatory and inhibitory synapses, the former mainly being mediated by glutamate (Meldrum 2000; abbreviated as glu or glut, while neurons depending on them are called glutamatergic) as neurotransmitter, while the main neurotransmitter of latter synapse type is generally γ-aminobutyric acid (GABA, while cells depending on them are called GABAergic). Several other neurotransmitters used in this process are adenosine, D-serine and dopamine, while many other exist.

Though this is the general principle in its simplest way, there are many exceptions to this process, among them processes involving non-neuronal cells, like astrocytes which will be discussed later in a later chapter.

Short-term plasticity

On small end of the different timescales of plasticity, there is short-term plasticity, usually being defined as lasting anywhere from tens of milliseconds to several minutes (Zucker & Regehr, 2002). A very simple form of plasticity on this scale is the exhaustion of relevant neurotransmitters in the presynaptic terminal, due to strong activity for a short while, called synaptic fatigue (White et al., 1979). In this form of plasticity no more of the specific neurotransmitters can travel through the

(5)

synaptic cleft until the stores of neurotransmitters are replenished again and vesicles containing them are present near the synapse. This is the case for all neurotransmitters, on inhibitory and excitatory synapses and rarely takes more than several seconds, ranging from less than a second (Ryan et al., 1997) to tens of seconds (Armbruster & Ryan, 2011). Synaptic fatigue is a type of plasticity that does not necessarily require the participation of the postsynaptic neuron and finds its origin mainly in the presynaptic terminal (Simons-Weidenmaier et al., 2006).

A more complicated kind of short-term plasticity is depolarization-induced suppression of inhibition (DSI), in which a short sequence of quick bursts from the presynaptic axon leads to a depolarization and the influx of calcium ions to a dendrite inhibited through GABA (Vincent et al., 1992). Calcium ions initiate the production of retrograde messengers in the postsynaptic cell, among them endocannabinoids, which inhibit the release of GABA from the presynaptic cell, effectively bypassing the inhibition for several hundreds of milliseconds to multiple seconds. In a situation without DSI happening, the neurotransmitter GABA would inhibit the signalling in the postsynaptic neuron by opening ion channels for chloride ions. The increase in ion concentration causes a hyperpolarization of the postsynaptic neuron, making it less susceptable to a signal in the form of depolarization (Obata et al., 1978). In DSI however, the train of action potentials causes the cell to depolarize through a calcium influx and retrograde messengers anyway, leading to a reduction of inhibitory currents, drastically reducing the polarization caused by the GABA-mediated influx of chloride ions. The result is a postsynaptic cell that is no longer inhibited for a short while and open to new signals (See figure 1A).

The same can happen the other way around with an excited dendrite, the bursts causing it to be hyperpolarized instead, reducing the excitation (depolarization-induced suppression of excitation, DSE), a process first described by Kreitzer & Regehr (2001b). Resulting from the postsynaptic depolarization, the release of excitatory glutamate is inhibited, causing the postsynaptic cell to no longer respond to the signals from the presynaptic cell. This phenomenon is not as well understood as DSI, but pathways and mechanisms of actions of DSI and DSE are expected to be comparable. The most important difference as of the current consensus being that DSI is mediated to a very large extent by endocannabinoids, while DSE was not as much affected in a cannabinoid receptor knock-out model (Ohno-Shosaku et al. 2002, Katona et al. 2001). This might indicate the involvement of another type of receptor or different pathways. However, plenty more about cannabinoids will be discussed in a later chapter about cannabinoid signalling in general.

Another comparable process of plasticity can happen in both inhibitory and excitatory cells, without the involvement of a depolarization: metabotropic suppression of inhibition (MSI) or excitation (MSE). In this form of plasticity the activation of a metabotropic receptor is the starting point of the signalling cascade leading to the suppression of inhibition or excitation (Straiker & Mackie, 2007). The resulting pathways are slightly different, but comparable: G-proteins activate pathways resulting in the production of PLC, leading indirectly to the production of retrograde messengers that cause an inhibition of neurotransmitter release from the presynaptic neuron, negating the inhibition or excitation caused (See figure 1B).

Another difference between the metabotropic suppression or depolarization-induced suppression is the required amount of calcium ions, being far larger in the latter (Ohno-Shosaku et al., 2005). However, both pathways are promoted in a calcium-dependent way and therefore, the two ways are hypothesized to be a coincidence detector, combining depolarizations and the release of neurotransmitter to facilitate plasticity (Hashimotodani et al., 2005).

(6)

Long-term plasticity

On larger timescales long term plasticity can last from several seconds to much longer periods, essentially the whole lifespan of an individual, but it is often researched in the range of several minutes to days. In a much investigated and common form of long-term plasticity a signalling cascade starts in the postsynaptic cell as a response to consecutive signals of a presynaptic excitatory neuron to cause long-term potentiation (eLTP). Normally, synaptic transmission leads to an increase of sodium ion concentration in the postsynaptic cell. These Na+_{-ions pass through}

AMPA receptors into the postsynaptic neuron contributing to a possible action potential. If more sodium ions pass into the AMPA receptor than are typically released during a single neuronal spike, the larger increase of Na+_{concentration in the postsynaptic neuron leads to a strong depolarization}

and the opening of NMDA receptors, by removing a blocking magnesium ion, when these receptors are also activated by the presence of glutamate (Nowak et al., 1984; Collinridge et al., 1983). The NMDA receptors, in turn, allow for an influx of calcium ions (Kleckner & Dingledine,1988; Calabresi et al., 1992), that serve as a second messenger. The calcium ions can lead to a number of reactions that changes the susceptibility of the of the synapse to signals (Alberini et al., 1995). An important target for Ca2+_{ions in this case is the kinase CaMKII (Ca}2+_{/calmodulin-dependent protein}

kinase II), which is involved in several pathways that facilitate plasticity (Otmakhov et al., 1997). Among these facilitated pathways are incorporating AMPA receptors in the synapse trough exocytosis (Takahashi et al., 2003), an increase in channel conductance (Benke et al., 1998) or an increase of the number of synapses (Engert et al., 1999), further increasing the probability to start an action potential. This process is called long-term potentiation or LTP (several effects are displayed in figure 2).

A variation of the same process can also lead to a decrease of susceptibility to signalling (depression) and is called long-term depression or LTD. In this case the opposite happens: the postsynaptic cell shows reduced reaction to the presynaptic activity following a small increase of calcium ion concentrations or low frequency stimulation (Dudek & Bear 1992). In this process the calcium ions lead to a different signalling pathway that results in the phosphorylation of AMPA receptors (Ogasawara et al., 2008) and the removal of these receptors through exocytosis and trafficking (Beattie et al., 2000), changes that lead to a decrease in susceptibility to signals from a preceding neuron.

Figure 1. Two types of short-term plasticity, From Mackie 2008.

A) Depolarization-induced suppression of inhibition or excitation. A depolarization of the postsynaptic bouton leads to a rise of calcium ion concentration and the production of cannabinoids in this neuron.

B) Metabotropic suppression of inhibition. The detection of specific neurotransmitters (Ach or Glu) leads to the production of PLC and endocannabinoids in the postsynaptic neuron.

After the production of endocannabinoids or other retrograde messengers and the activation of the associated presynaptic receptors, the release of neurotransmitter from the axon terminal is inhibited.

(7)

Additionally, plasticity can affect the release of neurotransmitters from the presynaptic cell, not unlike depolarization-induced or metabotropic suppression of neurotransmitter release, but in this case causing a long-term change in plasticity. To facilitate this, retrograde messenger molecules are produced in the postsynaptic cell, the molecules reach the presynaptic cell through facilitated diffusion or transport, where several signalling pathways lead to the inhibition of neurotransmitter release. An important group of these retrograde messengers in this type of plasticity is molecule group of endocannabinoids, but others like NO (nitric oxide) and BDNF (brain-derived neurotrophic factor) have been shown to be involved as well.

An important type of plasticity that can have both long- and short-term effects on synaptic connections is spike timing-dependent plasticity. In this process of plasticity the timing of presynaptic and postsynaptic action is compared and the connection is changed accordingly: if a postsynaptic action takes place before an action potential reaches the presynaptic cell, the connection is weakened, if a postsynaptic action potential follows a presynaptic one, the synapse is strengthened (Sjöström et al., 2010).

Heterosynaptic plasticity

There is a possibility for synaptic plasticity to be established in nearby synapses that were not involved in the original synaptic transmission. The effects of this method of establishment of synaptic plasticity can be both long-term and short-term and the affected synapses can be both depressed and potentiated as a consequence. In fact, a change in synaptic plasticity, often has an effect on nearby synapses that follows the so-called 'Mexican hat' pattern (Royer & Paré, 2004): synapses in close vicinity of the original altered synapse have a comparable potentiation or

Figure 2. Schematic representation of a synapse and several types of plasticity (From Sanes & Lichtman, 1999). After intense simulation with the opening of both AMPA and NMDA receptors, calcium ions serve as a second messenger initiating several pathways leading to the establishment of LTP in this particular figure. Several different pathways result in a change in the pre- or postsynaptic neuron. These include the increase of available neurotransmitter, the increase of synapses between the involved neurons or the increase of receptor expression at the postsynaptic neuron.

(8)

depression, while the effect weakens gradually further away from this synapse (See figure 3 A. After a sufficient distance the effect has reduced enough to become inverted, causing an inhibition when the original synapse is potentiated and a potentiation when the original synapse is inhibited (see figure 3A).

It has been described quite extensively (Chistiakova et al., 2014;Kempter et al., 2001) that this type of plasticity is essential for the upkeep of brain networks, as a network exclusively using homosynaptic plasticity would inevitably lead to a positive feedback system with synapses involved in signalling getting more probability to be potentiated, while weaker connections getting less, with little possibility for change or to reverse process. Heterosynaptic plasticity serves as an additional mediating system to introduce a higher variety of information on synaptic changes and a possibility to maintain weaker connections, without letting them fully dissipate.

The most widely accepted way to spread information for heterosynaptic plasticity is through intercellular calcium signalling within neurons. Ca2+_{has been shown to be involved several times}

(Yang et al., 1999; Chistiakova et al., 2014) and this is in line with the way heterosynaptic LTP spreads a large increase in Ca2+_{-ions leading to the induction of LTP while a smaller increase leads}

to the induction of LTD (See figure 3B). More recently other molecules have been shown to be mediating in heterosynaptic plasticity, like ATP released from astrocytes (Chen et al., 2013), or even directly causing synaptic changes, like dopamine (Ishikawa et al., 2013). Therefore it is not unlikely that other inter- or intracellular messenger molecules (like cannabinoids) could at least be mediating the process of heterosynaptic plasticity as well.

Figure 3. Approximation of the spread of heterosynaptic LTP. (Own work)

A) Through an axon. Location of LTP is displayed in green, being strongest closest to where the spread started, slowly fading in strength until there is no change in synaptic strength. After sufficient distance, there is a region where LTD is induced, displayed in red, until that fades too.

B) This can be explained by a rise in calcium ions starting at the synapse that was active. A high increase of calcium ions leads to the induction of LTP while a smaller increase leads to the induction of LTD..Relative Ca2+_{-level is displayed in blue, relative effect is displayed in red and the Ca}2+_{threshold between LTD and}

LTP is dsiplayed as a gray line.

Depression Potentiation Depression A

(9)

The role of the endocannabinoid system in synaptic plasticity

An important part of plasticity is the feedback between the involved neurons often done through messenger molecules and G-protein-coupled receptors. The cannabinoid system (eCB) is a system of retrograde messengers extensively involved in cellular feedback. The eCB system has exactly this function in plasticity, which makes it a very important player.

Basics of Cannabinoids

The cannabinoid system is a widespread signalling system that has a presence in almost every cell in the body. Its main components are the native fatty acid-derived cannabinoid molecules anandamide (N-arachidonoylethanolamine, AEA, Devane et al., 1992) and 2-arachidonoylglycerol (2-AG, Sugiura et al., 1995) and the receptors of these molecules called cannabinoid receptor 1 and 2 (CB1R, CB2R; Devane et al., 1988; Matsuda et al., 1990). The high level of expression of these receptors in the nervous system gives an indication of their importance in the nervous system. Although the localization is a bit different between the two, with CB1R being expressed mainly in the CNS, while CB2R is expressed mainly outside the CNS and in much lower numbers in the CNS. Within the CNS, CB1R is clearly present almost everywhere in the brain, being very numerous in the cortex and subcortical structures (Herkenham et al., 1990; 1991a;b;c; Pagotto et al., 2006), located presynaptically in neurons, besides being expressed in astrocytes and other glial cells. The presence of CB2R in the CNS has been debated for a very long time but recent papers have shown an expression in neurons, at least in the rat brainstem (Van Sickle et al., 2005) and mPFC (Den Boon et al., 2012), it is quite established that it also is highly expressed in immune cells like microglia (Onaivi 2006; Pertwee 2006; Cabral et al., 2008).

There are several enzymes involved in the production of the primary CBR agonists: though present in smaller amounts, AEA is produced through several specific phospholipases like NAPE-PLD and PLC (N-acyl phosphatidylethanolamine-specific phospholipase D and phospholipase C; Devane & Axelrod 1994; See figure 4 step 1&2) that are localized both pre- and postsynaptically, with a preference for dendrites (Cristino et al., 2008). Diacylglycerol lipases (DGLα and DGLβ) are responsible for the production of 2-AG (Mechoulam et al., 1995; see figure 4 steps 4&5), which also takes place postsynaptically in dendrites, but there is some evidence for presence in microglia as well (Murataeva et al., 2014). Functionally speaking, the main difference between AEA and 2-AG seems to be the in process and cause of their synthesis. While 2-2-AG is classically said to be synthesized on demand in large amounts, to be broken down shortly afterwards (Kondo et al., 1998), AEA seems to be involved in more long-term changes in behaviour like depression (Gobbi et al., 2005) and analgesia (Hohmann et al., 2005). The so-called ‘on-demand’ synthesis of 2-AG has been challenged recently though and the presence of a neuronal 2-AG pool has been suggested (Alger & Kim, 2011).

Additionally the cannabinoid agonists have a different binding affinity and efficacy for binding to the cannabinoid receptors. Generally, 2-AG has a higher binding affinity and efficacy than AEA for CB1R (Sugiura, 2009), making it a full agonist, while AEA is a partial agonist. Both agonists have the same binding affinities for CB2R, but 2-AG has a larger intrinsic activity, making it a stronger agonist compared to AEA on this receptor (Gonsiorek et al., 2000).

The effect of cannabinoids is strongly regulated by the time before hydrolysis to arachidonic acid, inactivating them as an agonist of the CBRs. In the case of 2-AG the main enzyme responsible for the breakdown is monoacylglycerol lipase (MAGL; Dinh et al., 2002; see figure 4, step 6), which is present in abundance in presynaptic terminals (Labar et al., 2010), contributing strongly to the active time window and spread through diffusion of the 2-AG molecule and its effects. AEA is broken down mainly by the enzyme fatty acid amine hydrolase (FAAH; Deutsch & Chin, 1993; see

(10)

figure 4, step 3), which is present in high concentrations throughout the CNS (Lever et al., 2009), which makes AEA very short-lived molecule in the body.

The enzymes breaking down 2-AG and AEA are not 100% specific to the molecules described above and FAAH is responsible in part for hydrolysing 2-AG and vice versa, MAGL is partly responsible for the hydrolysis of AEA. Additionally there are several other enzymes involved in the deactivation of cannabinoid molecules, but for the sake of clarity those will not be extensively discussed. However, it is important to note that, due to their important role in the functioning of cannabinoids, basically determining their window of action, different enzymes that break down cannabinoid molecules are a target of current medical research (Ulugöl, 2014; Ghosh et al., 2015). An inhibitor of FAAH for example can increase the amount of AEA in the synaptic cleft, which has a large effect on nociception, making inhibitors of this kind act as an indirect anaesthetic agent (Patel et al., 2003).

There are several other endogenous cannabinoids like 2-arachidonyl glyceryl ether (Hanuš et al., 2001) and N-acyl-dopamine (Bisogno et al., 2000) and there might be other undiscovered agonists for the CBRs present in the body. Besides that, there are vast numbers of exogenous molecules that can serve as an agonist for both CB1R and CB2R. A large number of different exo- or phytocannabinoids is present in cannabis and other plants while plenty of synthetic cannabinoids are being produced anew as well. However, this thesis will mostly handle endogenous cannabinoid signalling and because 2-AG is the most numerous in the CNS and is a stronger agonist than AEA on CB1R, it will be the main cannabinoid in this thesis.

Other than the numerous agonists for the CB receptors, there are other receptors that can be activated by the cannabinoid messengers. Most notably, AEA can activate the transient receptor potential cation channel, subfamily V, member 1 (or transient receptor potential vanilloid subtype 1; TRPV1), which has different pathways resulting from its activation

Cannabinoids in Plasticity

Cannabinoids serve as an essential part in synaptic plasticity, having the role of a retrograde messenger, passing from postsynaptic cells to presynaptic terminals through diffusion. At the axon terminal cannabinoid receptors are activated by the endocannabinoids and the resulting pathways change the firing characteristics of the presynaptic neuron, resulting in a change in plasticity.

Figure 4. Scheme of the prodcution and breakdown of 2-AG and AEA (adapted from El Manira & Kyriakatos, 2010)

AEA: 1.) NAT creates NAPE form phosphatidylethanolamine.

2.) NAPE-PLD creates AEA from NAPE. 3.) AEA is broken down to arachidonic acid and ethanolamine by FAAH.

2-AG: 4.) PLC creates DAG from phosphatidylinositol.

5.) DGL (or DAGL) converts DAG to 2-AG. 6.) 2-AG is broken down to archidonic acid and glycerol by MAGL (or MGL).

(11)

The synthesis of 2-AG starts with the activation of metabotropic glutamate receptors, specifically mGluR5, by synaptic glutamate (Izumi & Zorumski, 2012; figure 5, step 1). The activation of this G-protein coupled receptor releases several G-proteins, of which one activates the membrane-associated enzyme phospholipase C (PLC; Putney & Tomita, 2011). PLC then catalyses the production of inositol triphosphate (IP3; Putney & Tomita, 2011) which, in turn, leads to an influx of

calcium ions (Michell et al., 1981). The production of IP3 also synthesizes diacylglycerol, which is

hydrolysed to 2-AG by DGL (Jung et al., 2005; figure 5, step 2). The synthesis of 2-AG in this way, requires DGL to be activated by calcium, making this process very dependent of the calcium ion concentration (Kano, 2014).

In a comparable pathway AEA is produced after N-acyltransferase (NAT; Cadas et al., 1997) creates the main product for NAPE-PLD, called N-arachidonoyl Phosphatidylethanolamine (NAPE). NAPE-PLD then creates anandamide from NAPE (Okamoto et al., 2007). Like in the synthesis of 2-AG, this pathway requires the activation of NAT by binding to calcium ions (Okamoto et al., 2004).

According to Lovinger & Mathur (2012), the cue for the cannabinoids to be produced as a result of synaptic firing is repetitive activation of the synapse at 1-5Hz for GABAergic cells and 100Hz for glutamatergic cells.

After synthesis, the endocannabinoids diffuse towards and through the cell membrane. The exact process of intracellular endocannabinoid transport is currently under debate, but it is generally agreed upon that endocannabinoids move mainly through the extracellular space by facilitated diffusion. Through membranes, the molecules possibly get some help of a transmembrane transporter (Hillard & Jarrahian, 2000; Chicca et al., 2012), though most cannabinoids can diffuse freely through the membrane, being non-charged fatty acids.

After passing through the membrane the molecules then diffuse toward the presynaptic axon terminal, where they activate the CB1 receptors. The activation of the receptors generally leads to the inhibition of the release of neurotransmitter-containing vesicles (Elphick & Egertova, 2001). Additional spillover to other synapses is partly blocked by astrocytes, while it can activate cannabinoid receptors at other synapses nearby and facilitate synaptic plasticity distal from the original synapse.

Figure 5: Schematic of cannabinoid-induced plasticity in a glutamatergic neuron (adapted from Cachope, 2012). Receptors in the postsynaptic cell are activated as a result of neuronal firing, either NMDARs open and allow a calcium ion influx or mGluRs activate, in this case due to the presence of glutamate (1), starting a G-protein pathway leading to the PLC-induced synthesis of DG, from which 2-AG is synthesised by DGL (or DAGL, 2). 2-AG then moves through the cell membrane and toward the presynaptic cell, where it binds to CB1R (3). The g-proteins coupled to the CB1R lead to the activation of several pathways that inhibit the release of neurotransmitter from the presynaptic axon (4).

(12)

The activation of the cannabinoid receptors can start several pathways resulting from different parts of the G-protein coupled receptor (see figure 5, step 3&4). Most pathways lead to a reduction of neurotransmitter vesicle release, which can have different effects on the postsynaptic cell, depending on the type of presynaptic cell (be that GABAergic, cholinergic, glutamatergic or otherwise). The basic forms are suppression of potentiation or of inhibition, as described in the chapter about general plasticity. There are also several mechanisms organising the timescale of the plasticity, which mainly appears to be regulated by the different G-proteins (summarized in figure 6).

The Gβγ dimer combination of G-protein subunits can lead to the facilitation of short-term plasticity

by inhibiting voltage-dependent calcium channels (VDCC or voltage-gated calcium channels, VGCC) for a short time (Herlitze et al., 1996; Wilson et al., 2001). When not inhibited, the VDCCs can allow an influx of calcium ions near the synapses. The calcium ions that enter near the presynaptic terminal are required for the release of neurotransmitter molecules from vesicles into the synaptic cleft, by facilitating the fusion of nearby neurotransmitter-containing vesicles and the membrane at the synaptic terminal (Catterall et al., 2013). The inhibition of these channels thus leads to a decrease in neurotransmitter release in a direct manner that does not take a lot of steps or time. In this way, short-term plasticity only requires a small timescale of endocannabinoid production. This same pathway seems involved in all types of short-term plasticity that are regulated by (endo-)cannabinoids, both depolarization-induced and metabotropic, meaning the type of induction does not change the pathways following activation of the cannabinoid receptor. In the aspect of inhibition or excitation, the difference between the suppression of the two seem to be the threshold of induction, which appears to be higher in DSE (in the hippocampus at least) than in DSI. This is likely because a lower number of CB1Rs is present on these axons (Ohno-Shosaku et al., 2002).

Figure 6. More detailed schematic view of the suspected presynaptic effects following CB1R activation (Adapted from Atwood et al., 2014). The activation of CB1R leads G-protein subunits to inhibit several pathways promoting and facilitating neurotransmitter release. Most importantly, the βγ subunit blocks VDCCs at the synapse, which are responsible for the CaMKII and calcineurin pathway involved in the release machinery and the αi/o subunits block the pathway leading to the production of RIM1α, which is also involved in the release machinery.

(13)

Another G-protein subunit, Gαi/o, meanwhile can lead to the facilitation of long-term plasticity

(Atwood et al., 2014). This pathway is classically suggested to be more involved in long-term as opposed to short-term plasticity (Chevaleyre et al., 2007; Huang et al., 2005). In this pathway the G-protein first leads to an inhibition of the cAMP/PKA activity (Childers & Deadwyler, 1996), which, in an uninhibited situation, leads to phosphorylation of RIM1α (Rab3-interacting molecule 1α). A decrease in RIM1α concentration has been shown to reduce the frequency of miniature inhibitory postsynaptic currents (mIPSCs; Chevaleyre et al., 2007), a measurement at the dendrite level for the presence of inhibition of a neuron. Additionally the RIM1α molecule itself has an involvement in the release of neurotransmitter into the synaptic cleft, being part of the large group of proteins responsible for the docking and fusion of vesicles (Kaeser & Südhof, 2005; summarized in figure 6). This all leads to a theory that, through this pathway, the release of neurotransmitters is inhibited by the activation of CB1R on a longer timescale.

It has been shown that the presence of cannabinoid molecules is not required for the maintenance of long-term plasticity after it has been established. The removal of the cannabinoid molecules from the synaptic environment after the establishment of plasticity does not facilitate a change in plasticity (Chevaleyre & Castillo, 2003).

Though stated here with a clear distinction between short- and long-term, the involvement of RIM1α in long-term plasticity and VDCCs in short-term plasticity is not set in stone. It appears to differ for different cell types and brain structures which pathway and molecules are involved in the facilitation maintenance of plasticity. For example, in the CA1-region of the hippocampus, specifically in the Schaffer-collateral excitatory synapses and GABAergic synapses, RIM1α is involved in short-term plasticity (Calakos et al., 2004), while a role for the VDCCs is suggested in the facilitation of long-term plasticity, in a pathway different from the short-term Gβγ pathway in the

nucleus accumbens (Robbe at al., 2002). The pathways described above however seem to be the most likely to result from CB1R activation in the majority of the CNS (Atwood et al., 2014; Reggio 2010).

In another form of plasticity regulated by cannabinoids, there is a way of neurons to inhibit their own signalling after intense stimulation (Bacci et al., 2004). The stimulation leads to a synthesis of cannabinoids and the activation of CB1R on the same neuron. Next inward-rectifying potassium channels are recruited to change potassium conductance in this slow-self inhibition (Marinelli et al., 2008).

Other Actors and Actions

The functional presence of CB2R in neurons has been debated for a long time, but recently it has become clear that the receptor is definitely present in the CNS and neurons. The expression of this receptor is a lot more inducible than CB1R (Miller & Devi, 2011) and seems to be related to addiction behaviour, as frequent substance abuse leads to a clear increase in the expression of this receptor (Onaivi et al., 2008). In the absence of CB1R, CB2R appears to be able to mimic or take over its function, restoring DSE in CB1 knock-out mice (Atwood et al., 2012). However, given the inducible nature of CB2R expression, it is hard to say with this setup, whether CB2R would be having the same function in an animal with CB1R present.

Relatively recent it has been shown that CB2R has a functional presence in synapses of rats without the CB1R knockout mutation, particularly as an intracellular receptor in the postsynaptic neuron (Den Boon et al., 2012). In this context CB2R seems to have an effect on the excitability of the postsynaptic cell, thus having a possible contribution in synaptic plasticity.

It has been long debated whether this is the case, while it seemed that this receptor had its main functional presence outside the CNS, or at least not in synapses. Newer research has been looking

(14)

into the actual function of CB2R at the synapse environment and whether there is a functional difference between CB2R and CB1R. Within plasticity, the role for the CB2 receptor seems to be less widespread, as would be evident from its smaller expression pattern. The full range of functions and pathways of CB2R has not clearly been established yet, but it has been shown that they can activate different synaptic ion channels in the hippocampus (Stempel et al., 2016) and prefrontal cortex (Den Boon et al., 2014).

The diffusion of cannabinoids through the intercellular space makes them a candidate for partaking in heterosynaptic plasticity. Indeed the local addition of cannabinoids in brain slices has been performed (Chevaleyre & Castillo, 2006) and seems to affect multiple synapses within the diffusible range of the injection point (Kreitzer et al. 2002, Vincent & Marty 1993) resulting in spreading DSI. For long-term plasticity, the evidence of spread is rather sparse, in fact pointing to the fact that LTD can only be invoked in (or close to) the original synapse (Gerdeman et al., 2002). Intracellularly the spread of cannabinoid molecules and the accompanying signals seem to be equally contributing, not even being able to induce a short-term depression within the same purkinje cell 20 μm away (Brown et al., 2003). Another player in the diffusion of cannabinoids and the spread of cannabinoid-induced plasticity are the non-neuronal cells surrounding the neurons: glial cells, in particular the astrocytes. Their close proximity, proven involvement in several forms of plasticity and the presence of CB1R in the cells, makes them a very interesting player in the establishment of plasticity and they have a bigger role in cannabinoid- mediated plasticity than it would seem at a glance.

(15)

Glial cells and plasticity

Besides the many different types of neurons, there numerous other types of cells in the brain, with a plethora of functions, collectively called glial cells. Among these glial cells are the myelin-forming oligodendrocytes, that increase the efficacy of signal movement through axons and dendrites, microglia that serve an immune function and ependymal cells making up the tight lining along capillaries and the ventricles, basically making up the doorway between the brain and the rest of the body (called the blood brain barrier, BBB). Another type of glial cell in the CNS is the ubiquitous and widely present astrocyte. These cells have a large range of functions broadly described as maintaining the delicate ion concentration ratios of, for example, potassium around synapses (Walz, 2000; Benfenati & Ferroni, 2010), supplying the neurons with the necessary nutrients like lactate (Simard & Nedergaard, 2004; Brown & Ransom 2007), besides giving structural support and possibly having a role in the regulation of axonal repair after brain damage (Anderson et al., 2016). There is however, clear evidence that astrocytes are involved in synaptic transmission too. The astrocytes are often connected to synapses as a third player, the astrocytic processes surrounding the synapse environment in a construction called the tripartite synapse (Visible in figure 7 through electron microscopy; schematically in figure 8) and appear almost everywhere in the brain in large numbers (Wolff 1970; Araque et al., 1999). The presence of reuptake channels for neurotransmitters and receptors for glutamate in astrocytes (Flott & Seifert, 1991) already indicates a role in maintaining the signalling environment, indirectly influencing signalling efficacy. However, the astrocytes do not simply break down the neurotransmitters they take up, but they have their own pathways resulting from the detection of neurotransmitters like glutamate.

Figure 7 (left). Electron microscopy image from a clear tripartite synapse (adapted from Hines and Haydon, 2014). Presynatic cell with clear vesicles in blue, postsynaptic cell in red and astrocyte processes in green. The synaptic cleft is clearly visible as a dark line between the neurons.

Figure 8 (right). Classic schematic view of the tripartite synapse (From Perea et al., 2009), with a post- and presynaptic neuron in blue and astrocyte in red. Showing the interactions of the neurons and astrocyte, neurotransmitters leading to an increase in astrocytic calcium concentration and the subsequent release of gliotransmitters that affect both neurons.

(16)

Gliotransmitters

The astrocytic pathways resulting from the detection of neurotransmitters are consistently mediated by calcium ion elevations, that ultimately result in the production of one of several messenger molecules, collectively called gliotransmitters. Among these gliotransmitters are glutamate, D-serine and ATP, which is converted to adenosine in the intracellular space, before reaching a neuronal target (summarized in figure 9).

The detection of the synaptic neurotransmitters or other neuromodulatory molecules, by receptors like G-protein-coupled receptors in the astrocyte lead to an increase in astrocytic calcium ion levels through the Gq-IP3 pathway (Di Castro et al., 2011), which lies at the foundation of many types of

inter- and intracellular calcium signalling. In this pathway an activated G-protein coupled receptor leads to the activation of IP3 which, in turn leads to the release of calcium ions mostly from the

internal stores (Sheppard et al., 1997) of the astrocyte, located in the endoplasmatic reticulum (ER).

A theory about the process of recruitment of calcium in astrocytes and the temporal restriction of the rise in calcium ion concentration involves a feedback loop between IP3 and calcium ions: IP3

recruits calcium ions from the ER or nearby stores, calcium ions are then responsible for a signalling pathway leading to the release of gliotransmitters. At the same time, high cytosolic levels of calcium ions lead to the activation of IP33K, which breaks down IP3 and, with that, stops the rise

in cytosolic calcium ion concentration (summarized by de Pittà et al., 2012). Because of the delay in the latter process, this system allows for significant peaks in calcium ion levels, which last only a short while before diminishing again.

The exact system in which the raise of calcium ion levels caused by specific modes of synaptic transmission lead to the release of specific gliotransmitters is not exactly known and is intensively researched. One theory is that the spatial organization within the astrocyte plays an important role in the determination of the produced gliotransmitter (De Pittà et al., 2012). A fast breakdown of either IP3 or a removal if the involved Ca2+ would then only allow the pathway to continue to a nearby

Figure 9. A more detailed schematic of astrocyte-neuron interaction (adapted from Bazargani & Attwell, 2016). It is visible that several neurotransmitters produced in neurons (like GABA, ATP and glutamate) are able to activate receptors at the astrocyte membrane, which leads to an increase in cytosolic Ca2+_levels.

This often happens through PLC and IP₃. The increase in calcium ion concentration then leads to the prodcution and release of gliotransmitters (like GABA, D-serine, glutamate and ATP) that have the capacity to activate receptors on the neurons and affect neuronal signalling.

Additionally the increase in calcium can propagate as a calcium wave through the astrocyte.

(17)

enzyme responsible for the synthesis of only one type of gliotransmitter, while the average distance between the ER stores and the membrane connecting to the tripartite synapse is indeed small: about 200-600nm (Bezzi et al., 2004; Bergersen et al., 2012). This theory seems likely, but is not definitively proven.

Several of the pathways resulting from the astrocytic release of gliotransmitters can result in the activation of numerous receptors of neurons, among them NMDARs, mGluRs and P2X besides playing a role in the synchronization of large scale neuronal properties like sleep (Fellin et al., 2012) and possibly disorders like epilepsy (Halassa et al., 2007). Most importantly for this thesis, gliotransmitters are clearly capable of establishing or abolishing the plasticity of synapses. ATP released from astrocytes can, for example lead to the insertion of AMPA receptors in the postsynaptic cell, in part by activating P2X receptors, thus leading to an increased susceptibility of the neuron to excitation from that specific synapse, a form of LTP (Gordon et al., 2005). Another way of influencing the plasticity of synapses, astrocytic glutamate released near the presynaptic cell can activate the mGluRs there and consequently increase the probability of neurotransmitter release, leading in another way to the establishment of LTP (Perea & Araque, 2007). Among other pathways resulting in plasticity, a role of astrocytes in the establishment of LTD has been proven too, specifically in spike timing-dependent LTD (tLTD), which is mainly dependent on the timing of signalling between the pre- and postsynaptic cells. In this form of plasticity, the activation of NMDARs in the presynaptic cell, by astrocytic glutamate is essential (Rodríguez-Moreno & Paulsen, 2008). Interestingly, the release of glutamate in the last case, appears influenced by the activation of astrocytic CB1Rs, activated by postsynaptically-produced endocannabinoids (Min & Nevian, 2012). Several other pathways involving astrocytic TNFα (Jourdain et al., 2007), D-Serine (Henneberger et al., 2010), glutamate (Parri et al., 2001; Fellin & Carmignoto, 2004) and ATP (Lalo et al., 2014) have been described and they all indicate that the release of gliotransmitters can lead to, or at least mediate in the establishment of plasticity in synapses connected to astrocytes in a tripartite synapse configuration.

Astrocytes in heterosynaptic plasticity

Besides homosynaptic plasticity, astrocytes, being connected to numerous neurons, can mediate in plasticity changes in different synapses or neurons that are relatively far away from from the synapse in which the signal originated. This process of heterosynaptic plasticity is essential in the integration of neurons in networks and astrocytes seem to be in the perfect position to mediate in the establishment of heterosynaptic plasticity: one astrocyte can be connected to roughly 1 000 000 synapses (Oberheim et al., 2009) in a tripartite synapse configuration, besides the connection to neurons astrocytes are connected to other astrocytes and allow free diffusion through gap junctions. This connection allows the spread of calcium signalling elicited from neural transmission (Schipke et al., 2008). Indeed, widespread propagation of calcium waves have been observed in astrocytes (Hoogland et al., 2009).

Calcium waves vary in duration and size and it is suggested that theses calcium waves can elicit gliotransmitter synthesis relatively far away from the original synapse that started the cascade, like they would do closer to the synapse that the wave originated from. The gliotransmitters can continue the cascade where they are released, affecting synaptic transmission and signalling in that area. This means that calcium waves elicited by neurotransmitter release should be able to establish synaptic plasticity far away from their origin, and establish heterosynaptic plasticity.

The actual large scale effect of calcium waves on signalling and plasticity is debated: Agulhon and his colleagues show no relation between the astrocytic calcium ion increase and neuronal plasticity (2010) while Navarrete and colleagues claim opposite (2012), though this comparison might be complicated by the differences in scale and setup of the experiments.

(18)

The possibility of astrocytes to spread signalling on a larger scale, either through synaptic transmission-derived release of gliotransmitters or large-scale calcium waves, also allows astrocytes to contribute to balancing neuronal networks, inhibiting strong synapses and promoting weak ones. This process maintains stability in neural networks and dysfunctional astrocytes have been shown to contribute to large-scale network balance disorders like epilepsy (Bedner et al., 2015; Coulter & Steinhäuser, 2015).

Lastly, another process worth considering when talking about synaptic plasticity and the influence of astrocytes on it, is the change in astrocyte morphology as a result of neuronal signalling (Reichenbach et al., 2010). The perisynaptic astrocytic processes can retract from and extend to the synaptic cleft and this most notably happens during lactation, when the processes retract from synapses in the hypothalamus on a large scale, which drastically increases synaptic glutamate spillover (Oliet et al., 2001). When talking about the involvement of astrocytes on plasticity and the additional supporting effect they have on synapses, it is not unlikely to think that the morphological change of astrocytes can affect signal transmission and thus can be considered a form of plasticity.

(19)

Astrocytes and cannabinoid-induced plasticity

Because of the involvement of both astrocytes and endocannabinoids in plasticity, the presence of astrocytes near synapses and the large amount of endocannabinoids produced in healthy neural functioning, the idea of cannabinoid receptors on astrocytes, near tripartite synapses formed. The validity of this idea was first tested by Navarrete & Araque (2008), where they showed the expression of CB1R on astrocytes, though not expressed in very high numbers

Further investigation on the cannabinoid receptors in astrocytes suggested that the receptors are particularly numerous near synapses, which obviously makes them a likely candidate for participation in the tripartite synapse. Additionally, considering the important role of cannabinoids in plasticity, it was hypothesized that astrocytes might play an important role in cannabinoid-induced or cannabinoid-regulated plasticity.

Indeed, more recently, Han and colleagues have shown that selectively removing CB1R from astrocytes impairs the LTD characterizing exogenous cannabinoid intoxication to a large extent in mice (Han et al., 2012), besides showing that the removal of astrocytic CB1R leads to the abolition of spatial working memory impairment as a result of exogenous cannabinoid application. This shows that astrocytic CB1Rs are definitely involved in cannabinoid signalling, at least for exogenous cannabinoids.

More specifically, the activation of CB1R on astrocytes, likely through attached the Gq/11 protein

(Navarete & Araque, 2008), can lead to the same activation of IP3 that several neurotransmitters

can, leading in this case to one of several possible outcomes. Currently three main processes following the activation of CB1Rs in astrocytes by endocannabinoids have been described.

Firstly, the facilitation of tLTD, by astrocytic glutamate binding to the NMDARs in the presynaptic region (see figure 10a). The CB1Rs in the astrocyte are activated by endocannabinoids produced in the postsynaptic cell, in turn triggered by both a calcium ion influx in the postsynaptic cell resulting from backpropagation of a previous action potential and activation of postsynaptic mGluRs (Min & Nevian, 2012), indicating current synaptic activity.

Second, promoting neurotransmitter release and thus facilitating potentiation, in synapses far away from the synapse creating cannabinoids (see figure 10b). In this case too, the astrocyte produces glutamate as a result from an increase of Ca2+_{levels, but the calcium travels a bit further down the}

astrocyte to reach synapses that were not involved in the process that originally produced the cannabinoids to incite the rise in calcium levels in the astrocyte. At these other synapses, the astrocyte releases glutamate that binds to mGluRs in the perisynaptic region of the presynaptic cell (Navarrete & Araque, 2010). The activation of presynaptic mGluRs can and will lead in this case to an increased probability of neurotransmitter release, therefore potentiating the synapse (Perea & Araque, 2007).

Third, the internalization of postsynaptic AMPARs following the activation of postsynaptic NMDARs by glutamate released from astrocytes, leading to the establishment of LTD (see figure 10c; Han et al., 2012). The use of exogenous cannabinoids in this experiment makes it impossible to determine whether this is hetreosynaptic plasticity or homosynaptic plasticity. Although the experiment shows that the involvement of astrocytes in the establishment of LTD as a result of cannabinoid intoxication is undeniable, it is hard to say how this affects LTD in non-intoxicated individuals. This is even more important when considering the difference in binding and activation strength of different cannabinoids. The cannabinoids used in this particular experiment (HU210 and Δ9_{-THC) are generally considered to be unlike both AEA and 2-AG (Devane et al., 1992; Pertwee}

2008), being less selective, yet having lower affinity and efficacy. Nevertheless the importance of CB1R in astrocytes is proven in this research.

(20)

In all these cases it looks as if endocannabinoids can start several of the same pathways that other neurotransmitters do when they are detected by astrocytic receptors: they facilitate the IP3-induced

recruitment of Ca2+_{, which leads to the release of gliotransmitters. Interestingly, the only type of}

gliotransmitter to be produced after activation of CB1R in astrocytes in all the papers mentioned above, is glutamate. This is the case both when the gliotransmitter release is observed near the synapse that originally produced endocannabinoids, as well as when the astrocyte facilitates heterosynaptic plasticity.

It is interesting to see that the release of cannabinoids, classically seen as a messenger that facilitates homosynaptic or small scale inhibition, can do the exact opposite in combination with astrocytes: they can facilitate potentiation a considerable range away from the site of original production, around 4 times as far in, for example the hippocampus (Navarrete et al., 2014). This is

Figure 10. Different types of plasticity as a result of endocannabinoid-induced astrocyte-mediation (From Navarette et al., 2014).

A) Spike timing-dependent depression: the depression of a synapse following unexpected spike-timing. Endocannainoids are produced by the neuron that activate the CB1Rs on the astrocyte, leading to a increase in calcium (1). The rise in calcium leads to the production of astrocytic calcium, that in this case binds to presynaptic NMDARs (2), inhibiting the signalling through this synapse (3).

B) Lateral potentiation: A synapse produces endocannabinoids as a result of stimulation and the synaptic connection lesssens. CB1Rs on a nearby astrocyte are activated (1) and lead to an increase in Ca2_{+ levels (2).}

The increase in astrocytic calcium leads to the production of glutamate further away from the original synapse, which is released near another synapse and causes potentiation by binding to mGluRs there.

In C) the application of exogenous cannabinoids lead to the same rise in CA2+ levels (1) and glutamate production (2) as endogenous cannabinoids, but in this case the glutamate causes LTD by causing an internalization of AMPA receptors (3).

(21)

likely a way for astrocytes to help maintain a balance in the ratio of excitation and inhibition in local networks and avoid the overactivity of a single connection. In this specific case of cannabinoid-induced heterosynaptic plasticity, balancing inhibition and excitation seems very clear: in the local synapse, where the cannabinoids in this case are produced, neurotransmission is inhibited after intense stimulation, reducing the strength of the connection. Meanwhile, the endocannabinoids produced in this synapse activate CB1Rs at the astrocyte, where Ca2+_{levels rise}

and glutamate is produced that promotes potentiation in synapses that are connected to this astrocyte. When considering the possibility of calcium waves spreading through multiple astrocytes the potentiating effect might even spread further and thus facilitate the balancing of inhibition and excitation in large networks.

This also seems in line with the way heterosynaptic plasticity spreads, the ‘Mexican hat’ pattern. Chistiakova and colleagues (2014) argued the spread of LTD in this pattern is likely caused by an intracellular Ca2+_{rise, spreading through the neuron. However, this could very well be intracellular}

Ca2+_{-waves through astrocytes instead of, or in addition to neurons. Both potentiation and}

depression of synapses have been shown to be among the hetereosynaptic effects of astrocytes on synaptic plasticity and the existence of large-scale calcium waves as a signal in and between astrocytes has been established, therefore we could argue that astrocytes could be, for a large part, responsible for the facilitation of heterosynaptic plasticity. In the meanwhile cannabinoids could serve as the prime messenger between astrocytes and neurons in this process, as it is established that cannabinoids are important, if not vital, for the heterosynaptic effects of astrocytes.

Additional research into astrocyte-neuron interaction through cannabinoid signalling, will greatly enhance the understanding of the workings and function of astrocytes and at the same time will increase our understanding of plasticity and will possibly offer us ways to intervene in the process for the for treatment of related neurological disorders and diseases.

A potential clinical use for cannabinoids and astrocytes

An interesting phenomenon and a research field where cannabinoids and astrocytes have already been shown to be involved is research into epilepsy. It has been shown relatively recently, that dysfunctional astrocytes are present in brain tissue of individuals with epilepsy (Coulter & Steinhäuser, 2015). The dysfunctions in these astrocytes generally generally include faulty K+

-channels and water -channels. Cannabinoids, on the other hand have been proposed as an anti- and proconvulsive agents for many years (Gowers, 1881; Rosenberg et al., 2015), the effect varying, depending on the type of cannabinoid involved. Since cannabinoid-activated G-proteins have been shown to act on K+_{-channels in neurons (McAllister et al., 1999; Guo & Ikeda, 2004) and K}+

-channels are important in the origination of epileptic seizures (D’adamo et al., 2013; Kaufmann et al., 2013) the same could be true in astrocytes and when combined with the knowledge that astrocytes are widely connected to neurons and can facilitate widespread effects through calcium waves, cannabinoids could have an important role in the mediation of large-scale effects like epilepsy.

In the future this might offer a mechanism that could mediate in the genesis of epileptic seizures and possibly a way of alleviating or decreasing seizures in people that currently have trouble finding reliable anticonvulsant medicines.

More research would definitely help in establishing the pathways involving astrocytes that contribute to epilepsy and when combined with looking into cannabinoids and their effects and pathways in astrocytes, I estimate it very likely that a usable and reliable medication for this condition can be developed.

(22)

Discussion

It is clear that the endocannabinoid system is a widespread signalling system, capable of eliciting plasticity to a large extent. With the discovery of CB receptors in astrocytes the cannabinoid system proves to be even more widespread and likely an important player in balancing the inhibition and excitation of large networks. There are definitely a lot of aspects in this system yet to be discovered, especially considering that large parts, like the presence of CB1R in astrocytes or the presence of CB2R in neurons and glial cells, are found relatively recent and therefore haven not been investigated fully. Some questions came to mind while writing this theses, that I sum up below. One of the first things I thought of while writing was the following: could CB2R have a presence on astrocytes? And if so, could their function be the same as in neurons? It would, however, be very hard to actively look for CB2R in astrocytes as it is established that there are few ways to specifically detect CB2R and not CB1R. A way to circumvent this would be using the technique formulated by Schmöle (2015) using a GFP-CB2 transgenic mouse. Although this group found no presence of CB2R in astrocytes, it might be interesting to see whether the expression of CB2R is inducible like it is in other cell types as a result of frequent cannabinoid stimulation or perhaps after the specific removal of CB1R in astrocytes as used by Han and colleagues (2012). The last statement requires answer to another question: whether the inducibility of the expression of CB2R can be elicited by the lack of CB1R, which could be tested by using a CB1R knock-out mouse and measuring parameters of CB2R signalling.

Largely left out in this thesis are most cannabinoids other than 2-AG, though AEA was mentioned to a small extent. Though it is true that 2-AG is a very important endocannabinoid molecule and has consistently been shown to be the most present in the CNS, I do think it is important not to underestimate the other endogenous cannabinoids and their role in plasticity. We know for a fact that different cannabinoids can have different binding affinity to the classic CB receptors and a different efficacy. Compared to 2-AG, AEA for example is a weak agonist. In a situation where both molecules are present in large amounts near CB1R, the competition of the two for the same receptors will lead to AEA ‘taking the spots’ where 2-AG could act. The result is a smaller effective activation of CB1Rs, because the efficacy of AEA binding to CB1R is lower, meaning it has a smaller response (Savinainen et al., 2001). Though most efficacies and binding affinities of most known endogenous and exogenous cannabinoids are already estimated and to an extent calculated, this is still something to keep in mind especially considering the medical potential of the this class of messenger molecules.

Second to that, since the involvement of astrocytes and CB1Rs is a recent idea, the effect of different endocannabinoids like AEA or exogenous cannabinoids, has not been fully investigated and might provide more insightful results about the influence of astrocytes on plasticity. I assume the technique recently used by Han and colleagues (2012), removing CB1R selectively from astrocytes, could help in this regard. This could be combined with the local introduction of different endogenous and exogenous cannabinoids to determine the role of astrocytes in cannabinoid signalling in a broader sense.

Another thing that might improve the general understanding of the combination of cannabinoids and astrocytes on plasticity is the involvement of the different G-proteins of CB1R in the signalling following from detection by CB1R, specifically in an astrocyte membrane. Like the way different G-proteins of CB1R in synaptic membranes lead to different pathways, likely having different effects or being a coincidence detector of each other. The same might be true for the CB1R in

(23)

astrocytes, several different G-proteins could lead to different pathways within the astrocyte.

Navarrete and Araque (2008) claim that, besides being bound to Gαi/o and Gβγ, like neuronal CB1Rs,

astrocytic CB1Rs are more likely to be bound to Gq/11 and considering the hallmark of G-protein

coupled receptors is the capability to start different pathways it could be very interesting to see what kind of other influences the activation of astrocytic CB1R could have.

Perhaps there is a comparable difference as in neurons, where the different pathways are in part responsible for the different timescales of plasticity. Or maybe there are more gliotransmitters released as a consequence of other G-proteins. When also considering the idea that astrocytes are important intermediaries between groups of neurons and capable of balancing inhibition and excitation, there could be important pathways discovered from this study.

In this case, protein-specific inhibitors, could be used, starting with the inhibitors for the G-proteins used by neuronal CB1R and CB2R. Care should be taken that these inhibitors only inhibit the G-proteins in astrocytes, to make sure they do not interfere in other non-astrocytic pathways. Additionally, I would be interested to see, or at least get a sense of the spread of cannabinoid signalling through astrocytes. It is a relatively new idea and therefore not a lot is known about it, but it would be nice to have a sense how widespread the results of a single synapse can be only through astrocytes. Navarrete and Araque (2010) did this to some extent when they measured the effective range of cannabinoid signalling through astrocytes. This single study was limited to the hippocampus though, while it is not far-fetched that the same process might be different throughout different brain areas where tripartite synapses and endocannabinoid signalling coexist, considering the different properties, cell sizes and densities throughout the brain.

A recurring question to me, regarding calcium-waves, is how an astrocyte can differentiate between a rise in calcium caused by a receptor activated by glutamate, by cannabinoids or any other neurotransmitters, yet there seem to be clearly different pathways resulting from the rise in calcium ions. De Pittà and colleagues (2012) described this could happen through spatial restriction: receptors for cannabinoids or other neurotransmitters would then always be near the site of astrocytic glutamate production and the calcium can be recruited relatively nearby. Additionally they state that different organelles or different types of vesicles contain the different gliotransmitters. This would make sense if the calcium rise in astrocytes would remain local and small scale. With the spread of calcium like in calcium waves, however, this spatial restriction seems a very unlikely to me. A calcium wave large enough to spread to other synapses or even other astrocytes would then inevitably lead to the production of all possible gliotransmitters produced by a rise in calcium levels.

Another answer could be the concentration of calcium recruited, that different calcium levels lead to the different pathways, like different levels of calcium can lead to the establishment of different types of plasticity. However, on a large scale this appears to be an unreliable method of distinction., especially because I cannot imagine calcium waves can be precise enough in calcium concentration, since they inherently come and go, meaning a relatively slow rise and slow drop afterwards.

More likely to me seems the participation of another messenger. This messenger could either be produced beside to the calcium recruitment within the astrocyte and travel with the calcium waves through gap-junctions to the relevant synapses, or it could be produced near the target synapse and determine the effect of the calcium wave there. In the latter case, the messenger could be produced either within or outside the astrocyte. In the case the messenger is produced outside the cell it could be one of the many neurotransmitters for which astrocytes already have receptors in their plasma membrane, like D-serine, adenosine or dopamine. The type of gliotransmitter produced and released in this case would depend on what type of neurotransmitter is most abundant in the synaptic area, and could therefore depend on the history of plasticity or signal transmimssion at the synapse.

(24)

Involvement of the plasticity happening in the synapse before could be very relevant, in for example the case of cannabinoid-induced establishment of heterosynaptic LTP through astrocytes, as discovered by Navarrete & Araque (2010). As it is now this form of cannabinoid induced, astrocyte-mediated plasticity could change the synaptic strength of a lot of synapses that are not supposed to be potentiated. A system combining the reaction of calcium waves and a local messenger would greatly validate the system and make the resulting plasticity reliable, causing synaptic plasticity only in synapses that have a (recent) history of changes in plasticity, while leaving inactive and unrelated synapses out unchanged.

An experiment could be done applying different neurotransmitters near tripartite synapses that are reached by calcium waves originating from other synapses. The postsynaptic currents could be determined through patch-clamp measurements and different types of neurotransmitters could be used in this synapse, both when a calcium wave is elicited in astrocyte from another synapse (or an artificially elicited calcium wave) and when the nearby astrocyte does not show a calcium wave. If there is any change in the postsynaptic current depending on the application of the neurotransmitter or the presence of a calcium wave, it would mean that there could be a local interaction of a local neurotransmitter and a calcium wave originating from another synapse.

Answering several of these questions through research would help in establishing a broader understanding of astrocytes and their role in the establishment of plasticity and the role of cannabinoids in that process. At the same time, the involvement of cannabinoids in this process gives us an opportunity to intervene in the process and possibly develop medication for disorders and diseases, as many different cannabinoids can be produced in different ways and the endogenous cannabinoid metabolism can be influenced in many ways.

In conclusion, important steps haven been made in understanding the interactions of neurons and astrocytes and plasticity on a larger scale. I personally think that it might not take to long before we see medication based of the interaction of cannabinoids and astrocytes to cure, or at least relieve the symptoms of mental disorders, caused by a dysfunction of this astrocyte-neuron interaction through cannabinoid signalling.