Cholinergic modulation of striatal microcircuits

Abstract The purpose of this review is to bridge the gap between earlier literature on striatal cholinergic interneurons and mechanisms of microcircuit interaction demonstrated with the use of newly available tools. It is well known that the main source of the high level of acetylcholine in the striatum, compared to other brain regions, is the cholinergic interneurons. These interneurons provide an extensive local innervation that suggests they may be a key modulator of striatal microcircuits. Supporting this idea requires the consideration of functional properties of these interneurons, their influence on medium spiny neurons, other interneurons, and interactions with other synaptic regulators. Here, we underline the effects of intrastriatal and extrastriatal afferents onto cholinergic interneurons and discuss the activation of pre‐ and postsynaptic muscarinic and nicotinic receptors that participate in the modulation of intrastriatal neuronal interactions. We further address recent findings about corelease of other transmitters in cholinergic interneurons and actions of these interneurons in striosome and matrix compartments. In addition, we summarize recent evidence on acetylcholine‐mediated striatal synaptic plasticity and propose roles for cholinergic interneurons in normal striatal physiology. A short examination of their role in neurological disorders such as Parkinson's, Huntington's, and Tourette's pathologies and dystonia is also included.


Introduction
Cholinergic interneurons (ChIs) contribute to give striatum its place among structures with the highest levels of acetylcholine (ACh) in the brain (Zhou et al., 2002). Without a doubt, these interneurons exert a strong and complex modulation of striatal microcircuits. These large interneurons form synapses with medium size spiny neurons (MSNs) and other numerous smaller GABAergic interneurons of which there are 10 subtypes and counting . ChIs can be identified by their electrophysiological characteristics (Goldberg & Wilson et al., 2017) and by immunoreactivity of their enzymatic profile (Mesulam et al., 1984). The morphology of ChIs, the richness of their synaptic contacts as well as the expression of a variety of receptors has attracted the attention of neuroscientists. More than 1000 research articles on ChIs, published during the last two decades, have enriched the understanding of their function.

Striatal acetylcholine receptors
An early study indicated that destroying possible afferent pathways to striatum 'cortex, thalamus, globus pallidus or ventrotegmental area' did not affect the activity of choline acetylase nor acetylcholinesterase (AChE) or the histochemical staining within the nucleus (McGeer et al., 1971;Lynch et al., 1972). This led to the proposal that interneurons were the main intrinsic source of striatal ACh. We now know of external sources of ACh that arrives from the pedunculopontine and laterodorsal tegmental nuclei (Dautan et al., 2014), but the main source of striatal ACh still is the spontaneously active ChIs (Kitai & Surmeier, 1993;Pisani et al., 2007;English et al., 2012;Goldberg et al., 2012). At the cellular level, ACh exerts its actions through the activation of two families of receptors, muscarinic (mAChR) and nicotinic (nAChR). The mAChRs belong to the G-protein-coupled receptor (GPCR) family (Caulfield, 1993). These receptors are divided into group I (M 1 , M 3 , and M 5 ) and group II (M 2 and M 4 ). Group I receptors are coupled to G q/11 proteins via a subunits that activate protein kinase C (PKC) and phospholipase C (PLC) leading to the production of inositol triphosphate and diacylglycerol that results in an increase in intracellular calcium. Group I receptors are found in striatal MSNs of both the direct (dMSN) and indirect (iMSN) pathways. In MSNs, these receptors are postsynaptically in dendritic spine necks and extrasynaptically locations (Hersch & Levey, 1995;Yan et al., 2001). Group II receptors are coupled to G i/o proteins, inhibit adenyl cyclase (AC) activity and close voltage-activated calcium (Ca v ) Ca V 2 channels while opening inwardly rectifying potassium channels (Kir3) following GPCR activation (Caulfield, 1993;Nathanson, 2000;Eglen, 2006;Haga, 2013). Muscarinic M 2 receptors act as autoreceptors on ChIs and are located mostly extrasynaptically suggesting a role in volume neurotransmission (Bernard et al., 1998). M 2 receptors act as inhibitory heteroreceptors on striatal neuropeptide Y-somatostatin expressing (NPY-SOM) GABAergic interneurons and on corticostriatal glutamatergic terminals (Hersch et al., 1994;Bernard et al., 1998).
The high degree of similarity of the orthosteric ligand-binding site in all five types of muscarinic receptors is the main reason it has been difficult to identify subtype-selective ligands (Eglen, 2006;Dencker et al., 2012) and a reason why the dissection of specific cholinergic effects on neuronal activity and release has been difficult to achieve. Nevertheless, new pharmacological tools such as the highly specific antagonist peptide isolated from the green mamba snake venom are now being used (Jerusalinsky et al., 2000;Karlsson et al., 2000;Rowan & Harvey, 2011;Servent et al., 2011). Similarly, positive allosteric modulators and allosteric agonists are becoming promising tools, even providing some therapeutic potential for several central nervous system diseases (Digby et al., 2010;Bock et al., 2017).
Acetylcholine release is regulated by presynaptically located heteroand autoreceptors. Muscarinic autoreceptors M 2 /M 4 (Hersch et al., 1994;Ding et al., 2006), via direct G i/o -mediated inhibition of presynaptic Ca V 2.2 and Ca V 2.1 channels linked to exocytosis. Another presynaptic control of release is regulated by the M 4 auto-and heteroreceptor activation of the barium-sensitive potassium currents carried through K ir 3 potassium channels in ChIs (Yan & Surmeier, 1996;Ding et al., 2006) and corticostriatal terminals (Calabresi et al., 1998a).

Anatomical
In general, anatomical studies have revealed that ChIs immunoreactive for choline acetyltransferase (ChAT), with a large multipolar cell body of 23-50 lm in diameter and widespread aspiny dendrites that arborize up to 1 mm (Kimura et al., 1981;Bolam et al., 1984b;Wilson et al., 1990) with 3-6 primary dendrites that extend in a radial pattern (Doig et al., 2014). Electron microscopy of rat striatal tissue performed by Doig et al., 2010Doig et al., , 2014 indicates that ChIs receive a prominent inhibitory input and that most of excitatory input is from thalamic afferents; a single ChI receives 8450 AE 694 connections of which the majority are symmetric. Moreover, there are approximately three times more vesicular glutamate transporter type 2 (vGLUT2)positive thalamic terminals than vesicular glutamate transporter type 1 (vGLUT1)-positive cortical terminals in an individual ChI (Doig et al., 2014). It is important to mention that boutons expressing vGLUT1 and vGLUT2 are the highest in the dorsal one-third in the rat striatum (Wouterlood et al., 2012). However, since vGLUT2 is also expressed in some dopamine terminals in ventral striatum (Stuber et al., 2010), it is harder to isolate thalamic inputs.
In spite of the comparative small number of ChIs (Lehmann et al., 1979;Bolam et al., 1984a;Bennett & Wilson, 1999;Bennett et al., 2000;Kreitzer, 2009;Girasole & Nelson, 2015), their long and many branched axons allow a widespread release of ACh (Bolam et al., 1984a;Contant et al., 1996;Calabresi et al., 2000). Initially, ChIs were described as homogeneously dispersed; however, in mice, a greater concentration of ChIs in the dorsomedial compared to ventrolateral areas was observed following a stereological reconstruction (Matamales et al., 2016). A correlation between this distribution and the presence of vGLUT1 and vGLUT2 contribute to a possible segregation of function.
Similar to dopaminergic axon varicosities, cholinergic ones, form few structurally defined synaptic connections, therefore favoring a slow cholinergic volume transmission (Descarries et al., 1997;Zhou et al., 2001;Aznavour et al., 2003;Coppola et al., 2016;Ovsepian et al., 2016;Dunant & Gisiger, 2017). The integration of a striatal cholinergic tone established by volume and synaptic transmission is considered to act within neuronal networks to change their balance of activity to possibly initiate neuronal ensembles with specific functions (Fuxe et al., 2012).

Electrophysiological
The spontaneously active firing characteristic of ChIs ensures the basal cholinergic tone (Kawaguchi et al., 1995;Lee et al., 1998;Wilson, 2005). These neurons have high input resistance, a broad action potential duration (Wilson et al., 1990;Tubert et al., 2016), a depolarized, and often changing, resting membrane potential that is usually fixed at À60 mV with a low holding current (Threlfell et al., 2012). These interneurons also called 'tonically active neurons or TANs' and 'autonomous pacemakers' are able to produce action potentials at 2-10 Hz in the absence of synaptic input (Bolam et al., 1984a;Wilson et al., 1990). Behind this tonic or pacemaking mechanism, it is an interplay of several ionic conductances (Wilson et al., 1990;Pisani et al., 2007). Their pacemaker cycle begins with an initial tetrodotoxin-sensitive sodium current-induced depolarization that leads to calcium influx from Ca V 2 channels. This first calcium influx in turn activates the calcium and voltage-activated big potassium currents (BK). This potassium influx contributes to membrane repolarization and the activation of the Ca V 2.2 current that, in turn, activates the small-conductance calcium-activated potassium current (SK). This second potassium current induces a medium duration after-hyperpolarization (mAHP) of 100-200 ms that defines the spike pattern and spike width (Kawaguchi, 1992;Bennett et al., 2000;Goldberg & Wilson, 2005). A decrease in intracellular calcium levels reduces the SK current and consequently the mAHP. The I h inward cyclic nucleotide-gated cation current (HCN) repolarizes the membrane to about À60 mV, with a resulting inactivation of the outward potassium A-type K V 4 current. At the end of the cycle, depolarization is slowed down, the persistent sodium current is activated, and the threshold for an action potential is reached, beginning a new sequence (Bennett et al., 2000;Goldberg & Wilson, 2005;Deng et al., 2007;Pisani et al., 2007).
Another feature of ChIs is a long pause in the tonic firing that follows bursts of action potentials. Their intrinsic properties allow ChIs to fire in regular, irregular, and in burst fashion interspersed with long pauses (Bennett et al., 2000;Goldberg & Wilson, 2005, 2017Wilson, 2005;Sanchez et al., 2011). During a burst, a subthreshold accumulation of calcium through Ca V 1 channels recruits an additional potassium current that, in turn, produces a long-lasting (several seconds) hyperpolarization (sAHP) Tubert et al., 2016).
It is considered that the delta frequency activity of these interneurons results from the combination of synaptic inputs and intrinsic mechanisms (Beatty et al., 2015). A muscarinic-dependent coherence between motor cortex and ChIs can be established following optogenetic stimulation at both beta and low gamma frequencies (Kondabolu et al., 2016). The reports on striatal oscillatory activity at different frequencies and the synchronization with other brain regions have been the topic of several recent publications (Brittain & Brown, 2014;Feingold et al., 2015;Sharott et al., 2017).
The fact that the firing properties of TANs are similar to some GABAergic interneurons has created confusion in the proper neuronal differentiation (Berke, 2008;Beatty et al., 2012;Gonzales et al., 2013;Gonzales & Smith, 2015;Apicella, 2017). It would be best to identify all interneurons, including cholinergic, not only associated with their extracellular electrophysiological characteristics but also with other criteria. The systematic approach to interneuron research being developed (Kepecs & Fishell, 2014;Wamsley & Fishell, 2017) will provide a database of properly classified interneurons (e.g., mRNA-expression profile). The future will likely bring further determination of their individual electrophysiological characteristics and integrative properties.

Glutamatergic
Presynaptic regulation of ACh release has an important function in control of the excitability in striatal microcircuits (Fig. 2). The regulation of dopamine release mediated by a glutamate-ACh link has become important, and metabotropic glutamate (mGlu) receptors are being explored as potential targets for the treatment of neurodegenerative diseases (Ribeiro, 2005). As indicated before, glutamatergic fibers from both cortex and intralaminar thalamus form asymmetric synaptic contacts on striatal ChIs but with a higher proportion of synaptic contacts from thalamic inputs (Doig et al., 2014). Cortical axons contact distal striatal dendrites, and thalamic axons contact striatal somas and dendritic shafts (Lapper & Bolam, 1992). In primates, approximately 20% of synaptic connections to ChIs are presumed glutamatergic and localized on the distal dendrites (Gonzales et al., 2013;Gonzales & Smith, 2015), and in rodents, the soma and proximal dendrites of ChIs are the targets of glutamatergic input (Doig et al., 2014). However, both cortical and thalamic stimulation induces short latency responses in ChIs and effects of the different afferent synaptic locations have been explored. Compared to responses induced by thalamic stimulation, cortical responses are less robust and attenuate if the stimulation is repeated (Doig et al., 2014). These differences could mediate the length of the pause and strength of the rebound; sustained thalamic input seems to keep cholinergic firing followed by long pauses with no rebound. Moreover, the variable intrinsic activity of ChIs seems more important than the location of the afferents in the moment-to-moment variability in the size of neuronal recruitment (Kosillo et al., 2016). The section 'Influence of cholinergic interneurons within the striatal microcircuits: dopaminergic terminals' describes other experiments that have contributed to clarify the role of glutamate receptors selectively activated by cortical or thalamic afferents.
The dopamine-ACh interaction is mediated by D 2 and D 1/5 receptors. D 1 /D 5 subtypes are expressed in dendrites (Bergson et al., 1995; and D 2 receptors are located in soma, dendrites, and axons (Alcantara et al., 2003). The activation of D 1 /D 5 receptors in slice preparations enhances ChIs excitability (Centonze et al., 2003b;Ding et al., 2011). Apparently, a cAMPdependent mechanism allows the closure of potassium channels and promotes the opening of nonselective cation channels (Aosaki et al., 1998). Cholinergic receptors expressed in the dopaminergic axon terminal fields modulate dopamine release; nAChRs increase dopamine release (Imperato et al., 1986;Calabresi et al., 1989) whereas presynaptic M 5 mAChRs reduce it (Foster et al., 2014). At the somatodedritic level, both nAChRs and M 5 mAChR increase spontaneous activity (Foster et al., 2014). Other effects on dopamine release mediated by other mAChR subtypes appear related to the stimulation of receptors located in non-dopaminergic neurons (Zhang et al., 2002a).
Using optogenetic stimulation of dopaminergic terminals in vitro, a biphasic modulatory action on ChIs was similar to the pauserebound response of putative ChIs recorded in vivo. This consisted in a decrease in spike rate and a delayed excitatory response that peaked 0.4-0.6 s after stimulation (Straub et al., 2014).
Although presynaptic D 2 receptors on ChIs limit ACh release through voltage-gated Ca V 2 channels, an important control of downstream processes is also provided by the regulators of G-protein signals (RGS) (Anderson et al., 2009). Ding et al. (2006) observed that following dopamine depletion, M 4 rather than D 2 receptors alter signaling in ChI. In the absence of dopamine, M 4 autoreceptors suffer the attenuation of Ca v 2 channel opening and pacemaking by upregulation of the expression of RGS9. Consistently, significant decreases of RGS9 protein concentration and mRNA were observed in dopamine depleted animals following L-DOPA treatment (Yin et al., 2011).

Other afferents
Axon terminals releasing serotonin, histamine, or adenosine are known to modulate the activity of ChIs. Serotonin afferents from the dorsal raphe nucleus (Miguelez et al., 2014) induce a direct excitatory effect on ChIs through 5-HT 2 (Blomeley & Bracci, 2005) and 5-HT 6 receptors (Bonsi et al., 2007). Similarly, histamine-containing afferents from the hypothalamic tuberomamillary nucleus (Bolam & Ellender, 2016) depolarize ChIs by the activation of GPCR histamine receptor type 1 (H 1 ) (Bell et al., 2000). In nucleus accumbens, the activation of ChI H 3 receptors decreases their spontaneous activity, but this effect can only be observed in accumbens since striatum does not seem to express this histamine receptor subtype (Varaschin et al., 2018). The purine nucleoside, adenosine, is released by neurons and glia. Of the four subtypes of GPCR adenosine receptors in brain, the A 2A subtype is mostly expressed in striatum (Dunwiddie & Masino,  (1992) (Preston et al., 2000;. Concomitant dopamine D 2 and A 2A receptor stimulation inhibits ACh release (Song & Haber, 2000;Tozzi et al., 2011). Moreover, adenosine reverses N-type calcium currents in ChIs and both MSNs through membrane G-protein pathways Hernandez-Gonzalez et al., 2014).

Influence of cholinergic interneurons within striatal microcircuits
In spite of their relative small number, ChIs within the striatal microcircuits form enmeshed axonal projections with an extensive neuromodulatory presynaptic and postsynaptic effect (Descarries et al., 1997;Descarries & Mechawar, 2000) and most likely, interact with all neuronal elements through synaptic and volume transmission (Threlfell & Cragg, 2011). The modulation of striatal microcircuits by ChIs is exemplified in studies involving neuronal excitability and neurotransmitter release (Figs 2 and 3).

Medium spiny neurons
ChIs synapse onto dendritic spines (Hersch & Levey, 1995;Alcantara et al., 2001) of iMSN and dMSNs (Izzo & Bolam, 1988;Bernard et al., 1992;Yan et al., 2001;Goldberg et al., 2012 Fig. 2. Influence of afferents on cholinergic activity and release. As mentioned in the text, pre-and postsynaptic auto-and heteroreceptors to ChIs and their afferents can selectively affect the spatial and temporal release of ACh with important functional consequences. The participation of different types of glutamate receptors not only modulates ChI activity and ACh release but also exerts a fine control over dopamine release and other interneuronal and MSN activity. Coincident afferent striatal activation can induce short-and long-term changes in ACh release important in the expression of striatal functions; in this way, ChIs, although few in number, are centrally positioned to likely control neuronal activity using wired and volume transmission. See Table 2 for the letters associated to the references of postsynaptic and presynaptic auto-and heteroreceptors. excitatory actions by direct modulation of several ionic currents, mainly potassium, sodium, and calcium (Pineda et al., 1995;Perez-Rosello et al., 2005;Shen et al., 2007;Carrillo-Reid et al., 2009). Both dMSNs and iMSNs express M 1 receptors, and their activation increases neuronal excitability by the enhancement of the persistent sodium conductance and by directly or indirectly depressing potassium currents (Akins et al., 1990;Galarraga et al., 1999;Figueroa et al., 2002;Perez-Rosello et al., 2005;Shen et al., 2005Shen et al., , 2007Carrillo-Reid et al., 2009;Goldberg et al., 2012;Perez-Ramirez et al., 2015). Both M 1 , M 4 receptors are expressed in dMSNs (Santiago & Potter, 2001;Yan et al., 2001;Goldberg et al., 2012), and the activation of M 4 with muscarine increases MSN excitability by enhancing Ca V 1 channels (Hernandez-Flores et al., 2015). A strong depolarization induced by glutamatergic striatal afferents triggers a postsynaptic release of endocannabinoids (eCB). CB 1 receptors are one of the most abundant GPCRs in the central nervous system and are located at excitatory and inhibitory presynaptic and axonal compartments. CB 2 receptors are primarily localized in microglia (Kendall & Yudowski, 2016). CB 1 receptors are coupled to pertussis toxin-sensitive G i/o type G-proteins, and their striatal activation results in a presynaptic long-term depression in corticostriatal synapses (Adermark & Lovinger, 2007).
ChIs are also important regulators of striatal eCB. ACh produces an indirect modulatory effect in the regulation of striatal plasticity through the eCB system (Oldenburg & Ding, 2011). At inhibitory synapses, M 1 receptor stimulation promotes eCB production and retrograde activation of CB 1 R that suppresses the inhibitory synaptic transmission. In contrast, at excitatory glutamatergic synapses, an M 1 agonist reduces postsynaptic Ca v 1.3 currents that, in turn, decrease eCB production and activation of presynaptic CB 1 R (Wang et al., 2006;Narushima et al., 2007). Low to moderate activation of corticostriatal afferents in vitro (5 Hz/60 s) produces a long-lasting disinhibition of synaptic input that complies with all the requisites for the induction of striatal high-frequency stimulation-induced LTD (Calabresi et al., 1992;Adermark & Lovinger, 2007;Kreitzer & Malenka, 2008).
The in vitro long-lasting disinhibition of synaptic input induced by corticostriatal afferents can be prevented with an antagonist to the non-a7 nAChR; moreover, a nicotine-induced facilitation of eCB-LTD is occluded by the dopamine receptor agonist quinpirole and by the mAChR antagonist scopolamine (Adermark et al., 2018). Using a slightly different paradigm to induce LTD in MSNs (i.e., direct activation of mGlu 1 with the agonist (S)-3,5-Dihydroxy-phenylglycine (50 lM) plus postsynaptic depolarization to À50 mV), selective optogenetic stimulation of cortical or thalamic afferents revealed that cortical, but not thalamic afferent stimulation, induces a significant eCB-LTD accompanied by a decreased probability of presynaptic release. Double immunohistochemistry of CB 1 R and vGLUT1 or vGLUT2 indicates cortical vGLUT1 terminals colocalize %4 times more with CB 1 (Wu et al., 2015).
Long-term changes in striatal excitability by cortical and thalamic axonal stimulation could be related to their different proposed functions: goal-directed behavior for cortical afferents (Graybiel, 1995) and attention and arousal for thalamic afferents (Alloway et al., 2017).    Fig. 3. Presynaptic muscarinic and nicotinic control of striatal glutamate release. Illustrated are the effects of ACh release within striatal microcircuits as discussed in the sections 'dopaminergic terminals' and 'glutamatergic terminals'. The cartoon depicts Right: An increase in glutamate release mediated by presynaptic a7 nAChR on glutamate terminals. Left: A decrease in glutamate release mediated by two mechanisms: (i) a direct effect of ACh on presynaptic mAChRs (M 2 , M 3 , and M 4 ), or (ii) an indirect effect of ACh mediated by an increase in dopamine due to activation of a4b2* nAChRs on dopamine terminals. Dopamine action on inhibitory D 2 receptors on glutamate terminals reduces glutamate release. Such a complex action on the same terminal as depicted in (Fig. 2) [if indeed the receptors are coexpressed on single terminals] suggests either that fine control of the concentration of glutamate or the precise timing of it is important for MSN activity. The second, more indirect, inhibition by a4b2* nAChRs on dopamine terminals may be an important source of the increased activity in striatum in the absence of dopamine when such inhibition would be removed. The symbol code depicts the receptor types and their location.
The interrelation between MSNs, glutamatergic cortical afferents, ChIs, and presynaptic action on dopamine terminals opens deliberation as to whether other receptors located in these microcircuits have a direct or indirect effect on MSNs (see Fig. 3).
Within striatal microcircuits, there is a neuronal chain that follows glutamatergic input to ChIs, then inputs to NPY-NGF interneurons, and finally GABAergic input to MSNs evidenced in vitro following multicellular recordings and calcium imaging. The activation of nAChRs on GABAergic interneurons induces a global decrease in neuronal activity indicating a general activation of inhibitory GABAergic interneurons (Plata et al., 2013). Similarly following synchronized activation of ChIs, the GABAergic NPY-NGF subtype produces the inhibition of MSNs mediated by ChI to GABAeric interneuronal synapses and then to MSN (Faust et al., 2015(Faust et al., , 2016. The recurrent inhibition of ChIs is sensitive to nicotinic antagonists therefore not mediated by the GABAergic interneuron (Sullivan et al., 2008;English et al., 2012). Optogenetic activation of glutamatergic thalamic afferents to ChIs provides a nicotinic excitatory input to NPY-NGF interneurons that in turn modulate MSN activity (Assous et al., 2017). Within this neuronal chain, it is still unknown if other interneurons such as the FA subtype also participate.
Additionally, persistent low-threshold spiking (PLTS) interneurons are highly excited by cortical afferents (Assous et al., 2017) and are directly and indirectly modulated by both nACh and mACh receptors. The amplitude of striatal intracellular responses mediated by GABA decreases in the presence of muscarine and ACh (Sugita et al., 1991).
A mutual excitatory interaction exits between ChIs and PLTS: ChIs acting on nAChR directly excite PLTS interneurons and indirectly through mAChR on unidentified GABAergic terminals. The net effect of a tonic cholinergic action on the GABAergic interneurons is inhibitory as both nicotinic and muscarinic antagonists reverse the inhibition (Elghaba et al., 2016). This evidence suggests that interconnected ChIs and GABAergic interneurons form a subcircuit that could allow flow of information independent of classical inputs such as MSNs to FSI (Luo et al., 2013;Faust et al., 2015Faust et al., , 2016.

Dopaminergic terminals
It is clear that synaptic release modulated at the terminal level, independent of the cell body, is a major component of the striatal microcircuits (Rice & Cragg, 2008). It has been calculated that within a sphere of striatal tissue of 20 lm in diameter, point-to-point synaptic communication for dopamine and ACh terminals takes place. Axons of dopamine and cholinergic neurons contribute each % 400 terminals that are intermingled with other 2000-4000 unidentified terminals (Descarries et al., 1997). Such observations led Agnati et al. (1986), as quoted by Fuxe et al. (2013), to propose the concept of volume transmission as a non-junctional mode of intercellular communication. By modeling striatal dopamine spillover after quantal release, Rice & Cragg (2008) concluded that uptake does not limit the initial overflow from an extrasynaptic or synaptic release site, resulting in the formation of a cloud of dopamine that can reach extrasynaptic dopamine receptors which are more abundant than the synaptic receptors.
Studies of cholinergic modulation of dopaminergic terminals suggest that ACh diminishes dopamine release via nAChRs located on dopamine terminals (Rice et al., 2011); however, when dopamine release and the activity of ChIs could be simultaneously monitored with fast scan voltammetry, a synchronous activation of ChIs increased striatal dopamine release; for references, see Cachope & Cheer (2014). Therefore, endogenous release of ACh directly triggers striatal dopamine release (Cachope et al., 2012) and ChIs synchronized by their thalamic input promote dopamine release (Threlfell et al., 2012). The prolonged debate about the interrelation between dopamine and ACh release has been slowly resolving, as more data are gathered. We now know that presynaptic nAChRs are highly expressed on striatal dopaminergic terminals (Jones et al., 2001;Zhou et al., 2001;Zoli et al., 2002;Salminen et al., 2004;Gotti et al., 2009;Livingstone & Wonnacott, 2009;Garcao et al., 2014;Wang et al., 2014;Gonzales & Smith, 2015;Howe et al., 2016), and that their activation facilitates dopamine release (Exley & Cragg, 2008).
Combined light activation of dopamine terminals and chemogenetic stimulation of ChI potentiates dopamine release (Aldrin-Kirk et al., 2018). Moreover, a neurotoxic dopamine depletion plus chemogenetic activation of ChIs in vivo increases the use of previously akinetic forelimbs induced by a low dose of L-DOPA; however, the activation of ChI combined with a D 2 agonist (quinpirole), but not a D 1 agonist, increases the L-DOPA-induced abnormal involuntary movements (Aldrin-Kirk et al., 2018). This is congruent with other observations of exacerbation of dyskinesias by D 2 agonists in mice (Alcacer et al., 2017) and increases in dyskinesias seen by the activation of M 1 receptors on dMSN in combination with presynaptic M 2 blockade (Bernard et al., 1992;Yan et al., 2001).
When considering microcircuits, different affinities or the complete absence of ACh (in knockout mice) can produce different modulatory effects. For example, a low affinity a7-containing nAChR will quickly become desensitized with a resulting decrease in cholinergic modulation; on the contrary, a high affinity a4b2*containing nAChR will desensitize more slowly, with a resulting increase in modulatory effect of ACh. Moreover, a ChAT knockout results in mice with no ChIs and produces increased phasic-to-tonic dopamine signal with altered dopaminergic and glutamatergic tone (Patel et al., 2012).
The participation of corticostriatal and thalamostriatal afferents on dopamine release has been clarified using selective optogenetic activation; increases in dopamine release by the corticostriatal terminal field are mediated by nAChR but modulated by mAChR. Moreover, the increase in dopamine release results from the action of AMPA receptors on ChIs that promote short-latency action potentials. Dopamine release driven by thalamostriatal afferents involves additional activation of NMDA receptors and action potential generation over longer timescales (Kosillo et al., 2016).
If the presence of NMDA receptors in thalamic afferents is observed, it would be interesting to know if they act as 'sniffers' of spillover glutamate release, have neurotrophic/neuroprotective function, or are involved in the modulation of postsynaptic responses.

Glutamatergic terminals
As mentioned before, striatal glutamatergic afferents arrive from cortex and thalamus (Ding et al., 2010;Doig et al., 2014), and presynaptic mAChRs (subtypes M 1 , M 2 , M 3 , M 4 ) are located on axon terminals (Hersch et al., 1994). Electrophysiological in vitro recordings of striatal slices have been useful to clarify their inhibitory role in the modulation of presynaptic release from excitatory terminals to MSNs and their participation in striatal microcircuits.
Stimulated release of glutamate reduces responses to field pairpulse stimulation (Barral et al., 1999) and random synaptic events (Hernandez-Echeagaray et al., 1998). Furthermore, pair recordings of interactions between ChIs and MSNs indicate that spontaneous activity of ChIs decreases the amplitude of the MSN intracellularly induced EPSC and that M 2 /M 4 antagonists prevents the decrease (Pakhotin & Bracci, 2007). Similarly, the activation of M 4 presynaptic receptors with a positive allosteric modulator decreases glutamate release with a consequent reduction in postsynaptic excitatory currents in both types of MSNs (Pancani et al., 2014). Moreover, mAChR induce presynaptic inhibition of striatal glutamatergic terminals through an action on Ca v 2 channels (Barral et al., 1999) with a consequent decrease in glutamate release at both corticostriatal (Hernandez-Echeagaray et al., 1998;Barral et al., 1999;Higley et al., 2009) and thalamostriatal terminals on dMSN and iMSN (Ding et al., 2010), (Fig. 3).
Apart from the muscarinic action, nAChRs play a bidirectional modulation on corticostriatal glutamate release due to the presynaptic location of a7-containing nicotinic heteroreceptors on corticostriatal afferents and presynaptic a4b2*-containing nicotinic heteroreceptors on dopamine afferents that in turn contact corticostriatal terminals. The activation of a7-containing nicotinic heteroreptors on cortical afferents increases glutamate release (Campos et al., 2010;Howe et al., 2016), whereas the activation of presynaptic a4b2-containing nicotinic heteroreceptors on dopamine afferents produces a two-link chain reaction: first, enhanced dopamine release stimulates presynaptic D 2 heteroreceptors that in turn produce a decrease in glutamate or 'brake' effect (Campos et al., 2010;Howe et al., 2016). Certainly the affinity of nAChRs and mAChR, their location, physiological properties, and activation state of the terminal field have already begun to explain the spectrum of pre-and postsynaptic responses to ACh and for that matter to other neurotransmitter receptors.
The variety of auto-and heteroreceptors located presynaptically at synaptic and non-synaptic locations can selectively affect the spatial and temporal control of spontaneous and action potential-driven neurotransmitter release, depending on the terminal subtype and their intrinsic activity (Banerjee et al., 2016;Pittaluga, 2016). After coincident presynaptic activation, short-and long-term changes in neurotransmitter release can also occur (Atwood et al., 2014), but most importantly, the controls on release described in this section reflect a precise receptor-mediated regulation (Fig. 3).

Co-release from cholinergic interneurons
Although it goes against the Dale's principle of one neurotransmitter per neuron, the concept of corelease is now more accepted (Hnasko & Edwards, 2012). The presence of the glutamate type 3 vesicular transporter (vGLUT3) in neurons typically indicates the possibility of corelease (Kljakic et al., 2017). In striatum, a high expression of the glutamate transporter vGLUT3 is seen in a population of vesicles that express both vGLUT3 and vesicular acetylcholine transporter (vAChT) (Gras et al., 2002;Amilhon et al., 2010;Kljakic et al., 2017). Striatal corelease of ACh and glutamate has been determined following two main strategies: electrophysiological and genetic manipulation. Following the electrophysiological approach, there are two studies: one reports that optical stimulation of ChIs induces in MSNs two glutamate-dependent responses (Higley et al., 2011) and another reports that ACh release following synchronous ChIs triggers an action potential-independent presynaptic release of GABA colocalized in dopaminergic terminals (Nelson et al., 2014). With the genetic approach, it was observed that following the deletion of the vAChT gene and subsequent elimination of ACh release, alterations in gross motor skills and in performance attributed to ACh, are still present most likely as a consequence of coreleased glutamate (Guzman et al., 2011).
Several questions must be answered regarding this topic: Does corelease for both neurotransmitters occur at the same time? Is release differentially regulated? Is release spatially coupled? How does the presence of two neurotransmitters contribute to microcircuits function? Does the ratio neurotransmitters change?

Striosome and matrix compartments
Almost 40 years ago, Graybiel & Ragsdale (1978) reported two distinct densities or compartments in the distribution of AChE in the striatum of primates and cats. These two compartments are called striosomes or patches, and matrix. Striosomes receive dopamine afferents from SNc and glutamatergic afferents from medial prefrontal, anterior cingulate, orbitofrontal, and anterior insular cortices (Benarroch, 2016). Stereological analysis in humans finds a differential distribution of ChIs with most of them located in the periphery of the striosomes (Bernacer et al., 2007). Similarly in rodents, ChIs are found in the border of striosomes (Kubota & Kawaguchi, 1993) with extended processes into both compartments (Kubota & Kawaguchi, 1993). In recent reviews, ChIs are described as preferentially located in the matrix (Crittenden & Graybiel, 2011;Crittenden et al., 2017). Using new tools, attempts to exclusively stimulate one compartment in vitro are clarifying the location of ChIs. Whole-cell patch recordings of ChIs with a posteriori identification of their compartment location revealed that GABAergic currents mediated by nAChRs are more frequently observed in the matrix than the striosome (Inoue et al., 2016), and the photoactivation of the matrix compartment with independent local stimulation and patch-clamp recordings revealed lack of synaptic connectivity between matrix and striosomes (Lopez-Huerta et al., 2016). The presence of ChIs in the areas high in calbindin-D28K and ChAT (Prensa et al., 1999) referred to as the 'peristriosomal boundary' reaffirm the location of ChIs between as well as within matrix and striosome compartments (Brimblecombe & Cragg, 2017).
A separation between matrix and striosomes has been established in rats by their different thalamic afferents. Unzai et al. (2017) reported that striatum and nucleus accumbens receive afferents to the striosome compartment mostly from thalamic midline nuclei, whereas the intralaminar nuclei innervate the matrix compartment. Moreover, whereas most terminal fields form en passant boutons, clusters or plexus containing many boutons are observed on terminal fields of the parafascicular nucleus. From the functional point of view, information from these two thalamic areas support the function previously inferred (Vertes et al., 2015): limbic (emotional) control for the striosomes and sensorimotor associative for the matrix (White & Hiroi, 1998;Crittenden & Graybiel, 2011;Buot & Yelnik, 2012).

Participation of cholinergic interneurons in striatal plasticity
It is broadly believed that long-lasting changes in synaptic efficiency at corticostriatal synapses are the cellular basis of motor learning (Pisani et al., 2007;Fino & Venance, 2011;Deffains & Bergman, 2015). These plastic changes have been shown as LTD or as long-term potentiation (LTP). Early reports of striatal long-term changes indicated that either LTP or LTD could be produced by high frequency stimulation of cortical or thalamic glutamatergic inputs along with postsynaptic depolarization (Calabresi et al., 1992;Lovinger et al., 1993;Wickens et al., 1996;Centonze et al., 2001).
Further studies revealed that the precise timing and order between presynaptic and postsynaptic action potentials dictate the occurrence of either LTP or LTD in the paradigm of spike-timing-dependent plasticity (STDP) (Markram et al., 2011). As in the case of longterm changes induced by high-frequency stimulation, STDP-induced LTD and LTP was also induced in corticostriatal synapses (Fino et al., 2008;Pawlak & Kerr, 2008;Shen et al., 2008;Fino & Venance, 2011;Shindou et al., 2011;Jedrzejewska-Szmek et al., 2017). Two variables are important for corticostriatal STDP: the frequent in vivo bombardment of pre-and postsynaptic inputs onto striatal neurons, and the presence of modulators like ACh, dopamine, or serotonin. Extracellular ACh and the level of M 1 receptor stimulation control the direction of LTP or LTD (Calabresi et al., 1999a;Centonze et al., 1999). Additionally, cholinergic modulation of eCB synthesis has been linked to these long-lasting processes (Wang et al., 2006;Narushima et al., 2007).
The interaction between dopamine and ACh is important in the regulation of MSN excitability and plasticity. It appears that in vitro cortical inputs first activate striatal GABAergic FS interneurons, then ChIs, and finally MSNs (Fino et al., 2008). This order of events provides a facilitating effect on the MSNs while they receive cortical information and so define the direction of the plasticity (Deffains & Bergman, 2015).
High-frequency stimulation of cortical or thalamic afferents that synapse onto ChIs leads to an early monosynaptic glutamate-dependent depolarization (EPSP) followed by an intrastriatal disynaptic GABAergic hyperpolarization (IPSP). In the presence of a GABAergic antagonist, induction of LTP depends on a rise in intracellular calcium and the activation of dopamine D 1 /D 5 but not D 2 receptors (Suzuki et al., 2001;Bonsi et al., 2004;Oswald et al., 2015). Moreover, in the absence of a GABAergic antagonist, the LTP of IPSPs recorded in ChIs is presynaptically mediated. The amplitude of each unitary induced IPSP is the same whereas their frequency increases (Suzuki et al., 2001;Miura et al., 2002). Other experiments suggest that the direction of STDP is determined by the rheobase of the ChIs. If the minimal current amplitude to evoke an action potential is low, LTD is observed in the recorded ChI, whereas LTP is induced if the ChI has a high rheobase (Fino et al., 2008;Fino & Venance, 2011).
The study of plasticity of cortical input to striatal GABAergic interneurons is limited due to their low population prevalence and cellular variability. So far, there are a few studies describing STDP on FS or PLTS-NOS expressing interneurons (Fino et al., 2008(Fino et al., , 2009). However, with the help of transgenic mice targeting specific interneurons, in the near future, the knowledge in this field will grow.

ACh and striatal microcircuits
Tonically active ChIs are central in any analysis of the striatal microcircuits and perhaps should be considered within a functional relevant microcircuit. In order to be able to clearly isolate neuronal microcircuits in behaving animals, technical advances are needed. The study of neuronal ensembles was originated by the analysis of the spatiotemporal organization of groups of neurons. To perform the mathematical analyses to reveal interacting neuronal ensembles as multidimensional microcircuits, many neurons should be recorded at once (Yuste, 2015;Carrillo-Reid et al., 2017). Although single cell studies have been valuable revealing direct postsynaptic actions, sometimes conflicting interpretations can occur using the recordings of many interacting cells (Carrillo-Reid et al., 2011). In recent years, these calcium-imaging techniques have provided the most powerful tool to study spontaneous or drug-induced neuronal modulation of %60-80 striatal neurons for at least 20 min without losing the single cell resolution (Carrillo-Reid et al., 2008).
In the section 'Influence of cholinergic interneurons. . .GABAergic interneurons', we described that the stimulation of striatal ChIs through nAChRs activation excites GABAergic interneurons that in turn induce recurrent inhibition in themselves and nearby ChIs (Sullivan et al., 2008). This effect could conceivably impact the activity in the whole population of striatal neurons. To study this possibility, Plata et al. (2013) artificially increased activity in the whole population of striatal neurons by bath application of NMDA or a previous chronic dopamine depletion. Under these conditions, it is clear that bath application of 1 lM nicotine clearly inhibits the hyperactive microcircuits.
Excitatory striatal activation of MSNs mediated by mAChRs has also been reported (Lv et al., 2017). The activation of M 1 receptors enhances a persistent sodium current that can synchronize a large population of MSNs (Carrillo-Reid et al., 2009). Moreover, M 1 receptor activation inhibits the persistent K V 7-potassium or the Mcurrent in the dendritic/spine compartment of MSNs  and as expected, a specific antagonist of M 1 receptors also decreases striatal neuronal activity (Hernandez-Flores et al., 2015). The influence of ChI on Kv7 channels is relevant, since these channels are widely expressed and are known to control neuronal excitability, the resting membrane potential, the spiking threshold, and to set the firing frequency within the burst and the subsequent hyperpolarization that follows a burst (Greene & Hoshi, 2017).
Parkinson's disease is a common neurological disorder characterized by a decreased dopamine level. Early clinical and experimental studies revealed that PD was also characterized by increased striatal extracellular levels of ACh (Barbeau, 1962;Cachope & Cheer, 2014). Indeed, the earliest pharmacological treatment of PD consisted of administration of anti-cholinergic agents (e.g., weak antimuscarinic diphenylhydramine, benztropine, orphenadrine; Fahn, 2014). However, the cumulative effect of anti-cholinergic medication 'anti-cholinergic burden', and the 'anti-cholinergic risk' associated with a decrease in the use of anti-cholinergic in old hospitalized patients. In a study of databases reporting side effects of anti-cholinergics, Salahudeen et al. (2015) compiled a list of those anti-cholinergics frequently prescribed and indicated that medicated patients suffer more frequent falls and hip fractures, increased dyskinesias, and suffer from hallucinations, blurry vision, and memory impairment than non-medicated patients.
The elevation of cholinergic signaling in PD is directly related to the alterations in ChI spiking (Tanimura et al., 2018). As described before, M 4 autoreceptors in ChIs slow firing rate and ACh release (Zhang et al., 2002b). In the rodent model of PD, dopamine depletion induces an upregulation of RGS4-dependent processes that result in decreased M 4 signaling in ChI . Alternative RGS modulation of ACh release might aid future treatment of patients. Experiments using the same animal model of PD report that halorhodopsin photoinhibition of ChIs in mice reduces akinesia, bradykinesia, and sensory motor neglect; however, in wild-type mice, the specific striatal blockade of M 1 and M 4 receptors has a similar effect. This suggests that the main participants in the absence of ACh are likely the M 1 and M 4 receptors since specific striatal blockade of M 1 and M 4 receptors has a similar effect (Ztaou et al., 2016). These results agree with the electrophysiological studies of muscarinic and dopaminergic interactions described in (Hernandez-Flores et al., 2015).
Recently Burbulla et al. (2017), using long-term cultures of human-induced pluripotent stem cells-derived dopamine neurons, has demonstrated a toxic cascade triggered by dysfunctional mitochondria that can induce neuronal pathological changes and cellular dysfunctions observed in PD. Now, research is centered on whether the same toxic mitochondrial intracellular cascade is present in the genetic and idiopathic forms of the disease. More work may eventually demonstrate the primary cause of SNc dopamine neuron death.
Dystonia involves intermittent or sustained abnormal involuntary muscle contractions that produce twisting postures in the absence of other neurological signs. Repetitive movement and uncontrolled muscle contractions can start early in childhood (Valente et al., 1998;Klein & Fahn, 2013). Early onset of dystonia is a genetically determined mutation in the gene TOR1A (Sciamanna et al., 2012). As in PD, the reciprocal modulation between dopamine and ACh is at the center of dystonia. For instance, high doses of anti-cholinergics (trihexyphenidyl) are used in the treatment of this disease (Burke et al., 1986). Electrophysiological experiments in ChIs of mice overexpressing mutant torsin A show that the sensitivity of a D 2 agonist-mediated inhibition of Ca v 2.2 N-type current is increased. Following D 2 agonists, a reduction in mAHP and threshold for action potentials is expected (Sciamanna et al., 2011). In mice with a conditional knockout of the dystonia 1 protein, the activation of thalamostriatal inputs induces a short pause and increased rebound activity in ChIs that could result from a postsynaptic increase and a presynaptic decrease in M 1 and M 2 -dependent currents (Sciamanna et al., 2012).
Gilles de la Tourette's syndrome is a neurodevelopmental disorder characterized by motor and phonic tics, usually measured by the Yale Global Tic Severity Scale (Leckman et al., 1989). In the last few years, several advances have been achieved toward the understanding of the neuropathology of this syndrome.
The participation of ChIs in this syndrome is supported by postmortem findings of a significant 49% loss of cholinergic and 42% loss of parvalbumin-positive FS interneurons with a no significant change in % 20% in DARPP-32 expression in MSNs (Kataoka et al., 2010); however, targeted toxin lesion of ChIs in the dorsolateral striatum of adult mice fails to show any abnormal stereotypes . Moreover, the radiotracer [ 18 F] fluoroethoxy-benzovesamicol that is successfully used to image overexpressed vAChT in mice  failed to detect changes in the number of ChIs in Tourette's syndrome patients (Albin et al., 2017), perhaps obscured by the pedunculopontine cholinergic afferents.
Since stereotypy is regarded as a predominant aspect of this syndrome, using cocaine-induced stereotyped behaviors to test the function of ChIs, it is observed that a lesion of ChI or blockade of mAChR (scopolamine) prolongs the time course of the stereotypy, whereas blockade of dopamine D 2 receptors (raclopride) stops the stereotypy presumably by increasing the extracellular cholinergic concentration (Aliane et al., 2011). These results suggest that a restoration of cholinergic transmission may have important consequences in the arrest of stereotypy. This is supported by a decrease in stereotyped behaviors in children following the administration of a cholinesterase inhibitor (donepezil) (Cubo et al., 2008).
Pharmacological animal models of the syndrome have been produced following blockade of striatal GABA A receptors. In mice, rats, and monkeys, intrastriatal administration of specific GABA A antagonists (picrotoxin or bicuculine) induces increased activity in striatum and its outputs (i.e., subthalamic nucleus and thalamus) and motor abnormalities similar to tics (McCairn et al., 2009;Bronfeld et al., 2013), for review, see Yael et al. (2015).
Huntington's is a progressive late-onset neurodegenerative disease characterized by psychiatric symptoms and cognitive deficit. It is caused by a CAG trinucleotide repeat in the gene encoding huntingtin. The resulting huntingtin accumulates forming inclusion bodies with other proteins, initially in neurons of striatal and cortical motor and prefrontal areas (Shepherd, 2013). In postmortem human tissue and rodent models of the disease, there is a striatal pre-and postsynaptic loss of GABA, glutamate, dopamine, and muscarinic acetylcholine receptors (Penney & Young, 1982;Dure et al., 1991) and a preferential degeneration of MSNs (Reiner et al., 1988) with a faster loss in iMSNs (Cha et al., 1998;Deng et al., 2004;Starr et al., 2008). Although the number of ChIs is relatively normal (Ferrante et al., 1987), these interneurons have decreased the levels of vAChT and ChAT (Smith et al., 2006). In an animal model of the disease (Q140 huntington-like mice), Deng & Reiner (2016) studied the specific vGLUT2 thalamic inputs to ChIs. They observed a reduction in the extension of the dendritic trees, with a subsequent loss of synapses, as also reported before (Deng et al., 2013). The authors propose that a reduced thalamic excitatory drive onto iMSNs could be responsible for an initial observed hyperkinesia in mice. Then, a subsequent loss of dMSNs could lead to the permanent hypokinesia in this animal model.
In recent years, interest has shifted in somewhat different directions. Two examples: (i) attention to the posttranslational modifications of huntingtin by the covalent attachment of a small ubiquitin modifier (SUMO) protein (PIAS1). PIAS1 participates in the huntingtin accumulation of inclusion bodies and as expected, a reduction in PIAS1 prevents the formation of inclusion bodies and reduces inflammation (Ochaba et al., 2016). (ii) Attention to the participation of NMDA receptors in neuronal degeneration pointing to the molecular link between mutant huntingtin and the synaptic retrieval of the GluN3A subunit of the NMDA receptors. Mutant huntingtin redirects an intracellular store of juvenile NMDA+GluN3A to the surface of the neurons favoring neuronal loss. Overexpression of GluN3A in normal mice induced synapse loss. Moreover, as expected, the genetic ablation of GluN3A subunits improves motor performance and decreases cell loss in mutant mice (Marco et al., 2013).

Conclusions and future directions
There is an emerging idea that like dopamine, ACh is necessary at a minimum concentration to maintain striatal function. The complex distribution of the receptors for ACh and the tonic activity in the cells themselves suggests a 'maintenance' role. The input to these interneurons from cortex and thalamus allows them access to goaldirected behavioral contexts (from cortex?) and to attentional and arousal internal signals (from thalamus?). The pause in firing that accompanies newly learned cues is similar in timing with the burst of dopamine activity that itself may generate the later burst of activity in the ChIs.
It is easy to imagine that these temporary changes in extracellular transmitter concentrations are a mechanism to remodel striatal functional microcircuits to adjust to the change in circumstances that initiated the pause. The intimate involvement of ACh in the long-term changes in excitability in striatal cells in vitro is also an indication that such a scheme might be involved in the response to novel cues that are recognized as significant by the animal. In this scenario, the distribution of receptors on both cells and terminals suggests that the organization of synaptic microcircuits in the striatum might underlie the changes in functional assemblies that result in changes in behavior.
Methods to identify these functional assemblies and demonstrate their sensitivity to local transmitter concentrations are being developed. They will provide information about the detailed physiology of such changes in function and perhaps begin to make sense of the detailed receptor localizations in the striatal microcircuitry. Work on optogenetic manipulation of the 'maintenance transmitters' is already leading to direct tests of their role. Moreover, methods to image activity, at single cell resolution, in groups of related neurons in freely moving animals are developing. We are reaching a time when such ideas cease to be speculation and become testable hypotheses about the role of acetylcholine in animal behavior.