Keywords

1 Introduction

Several subtypes of striatal neurons were described during the 1970s and 1980s using the Golgi labeling method or electron microscopy (e.g., Kemp and Powell 1971; Fox et al. 1971; Danner and Pfister 1979; Dimova et al. 1980; Preston et al. 1980; Wilson and Groves 1980; Bishop et al. 1982; Bolam et al. 1981b; Chang and Kitai 1982; Chang et al. 1982; Tanaka 1980; DiFiglia et al. 1976; Graveland and DiFiglia 1985; Graveland et al. 1985). In a series of detailed studies carried out in the monkey, DiFiglia and co-workers identified up to six types of neurons in the striatum: type I and type II spiny neurons , type I, type II, and type III aspiny neurons , and a very small cell apparently devoid of an axon that could be a glial cell (DiFiglia et al. 1976, 1979). Type I spiny neurons were relatively small in size (20–14 μm) and exhibited four to seven dendrites forming a spherical field around the cell body. The dendrites were described as smooth near the cell body, but they became heavily covered with dendritic spines more distally. The type I neuron has been also identified as the medium-sized spiny neuron (MSN or MSPN) or medium spiny I neuron in the cat and rodent (Kemp and Powell 1971; Dimova et al. 1980; Chang et al. 1982; Bishop et al. 1982). Type I neurons were considered to account for as much as 95–96 % of all striatal neurons (Kemp and Powell 1971). The monkey type II spiny neuron has a spindle-shaped cell body, thicker dendrites, less spines, and a more extensive dendritic field than type I and represents less than 1 % of all striatal neurons. This neuronal type has also been described in cats and rodents (Kemp and Powell 1971; Dimova et al. 1980; Chang et al. 1982; Bishop et al. 1982). The function of the type II spiny neuron is still not clear, but one study found a similar neuron containing neuropeptide Y (Kubota et al. 1991). Aspiny striatal neurons are considered to be local interneurons (Kawaguchi et al. 1997). This chapter will focus on the phenotype and function of projection medium spiny neurons (MSN) that constitute the majority of striatal neurons.

2 Phenotypic Diversity of Medium Spiny Striatal Neurons

Earlier immunohistochemical studies have established that MSN contain the GABA- synthetizing enzyme glutamic acid decarboxylase (Gad) and are releasing GABA as their primary neurotransmitter (Ribak et al. 1979; Vincent et al. 1982; Nagai et al. 1983; Ottersen and Storm-Mathisen 1984; Bolam et al. 1985; Smith et al. 1987). The presence of the mRNA encoding for Gad in most striatal neurons was confirmed later on using in situ hybridization histochemistry (Chesselet et al. 1987). Although all MSN are defined as GABAergic, they can be subdivided based on their connectivity and phenotype. This chapter will review the evidence supporting the notion that striatal MSN can be subdivided into two large populations based on connectivity, morphology, chemical phenotype, and physiology. These two subtypes contribute to the so-called direct and indirect pathway of the basal ganglia and their properties play a central role in current models of basal ganglia organization (Albin et al. 1989; Crossman 1987; DeLong 1983, 1990). The chapter will also review the experimental evidence that these two pathways play distinct roles in the control of motor and cognitive functions.

2.1 Co-expression of Peptides

Immunohistochemical studies carried out in the 1980s and 1990s have thoroughly documented that MSN co-express GABA with one or more than one of the three peptides met-enkephalin, substance P, and dynorphin. Immunohistochemical studies combined with tract-tracing methods found that substance P and dynorphin are co-expressed in specific populations of striatal projection neurons, whereas enkephalin is present in other populations of striatal projection neurons (Anderson and Reiner 1990). Such an organization was documented in different species including pigeons, turtles, and rats (Anderson and Reiner 1990). Co-expression of substance P and dynorphin immunoreactivities was found in both the striosome and matrix striatal compartments (Besson et al. 1990). However, the percentage of substance P- and dynorphin co-localization was slightly higher in striosomes than in the matrix. Conversely, about two-thirds of all neurons were identified as enkephalin-positive in both matrix and striosomes (Besson et al. 1990). In another study comparing cats and rats, Penny and colleagues found that neurons immunoreactive for dynorphin made up about half of the neurons in rat striatum and a little less than half in the cat. Labeling for enkephalin was found in a little less than half of the neurons in the rat and about half of the neurons in the cat (Penny et al. 1986). Substance P-immunoreactive neurons made up to 38 % of MSN in the rat and 39 % in the cat (Penny et al. 1986). An analysis using in situ hybridization histochemistry reported that the dynorphin mRNA was distributed in about half of patch and half of matrix neurons, while the enkephalin and the substance P mRNA were expressed in a little more than half of patch and about half of matrix neurons (Gerfen and Young 1988). Altogether, these immunohistochemical and in situ hybridization findings indicate that MSN can be subdivided into two major and numerically comparable populations, one that co-expresses substance P and dynorphin and one that co-expresses enkephalin. As discussed in the following paragraphs, there is evidence that a subpopulation of MSN co-expresses the three peptides, but the prevalence of these MSN remains controversial.

An earlier immunohistochemical study by Besson and colleagues (Besson et al. 1990) found that a majority of MSN expressing substance P/dynorphin also expressed enkephalin. In a more recent combined immunohistochemical and retrograde transport study in the monkey, about half of striatal neurons were found to co-express dynorphin and enkephalin (Nadjar et al. 2006). In a combined patch-clamp and PCR study, it was confirmed that some MSN co-express detectable levels of substance P and enkephalin mRNAs, but the frequency of these neurons could not be assessed (Surmeier et al. 1996). On the other hand, a more recent RT-PCR study by Wang and colleagues found that in 4-week-old rats, 11 % of MSN contained both substance P and enkephalin, while in 4-month-old rats, co-localization was only 3 % (Wang et al. 2006). In another study, the same group found that 32.3 % of MSN that contain both substance P and enkephalin are localized in the striosomal compartment (Wang et al. 2007). An immunohistochemical study in the rat nucleus accumbens found than less than 30 % of neurons co-express enkephalin and substance P, whereas more than 69 % co-express substance P and dynorphin (Furuta et al. 2002). Altogether, these findings support the likelihood that some MSN neurons co-express the three peptides, but the extent of co-localization varies between studies. Such differences could be partly explained by methodological (i.e., immunohistochemical versus gene expression studies) or species differences. As discussed above, it is also possible that the reported variability is due to developmental factors and/or differences between striatal compartments (Wang et al. 2006, 2007; Furuta et al. 2002).

2.2 Medium Spiny Neurons Connectivity

Early tract-tracing studies determined that the striatum contains projection neurons sending axons to the ipsilateral GP and/or to the SNr (Szabo 1967; 1970). Later on, it was reported that most striatal neurons are projection neurons (Bolam et al. 1981a, b; Graybiel and Ragsdale 1979) and combined Golgi and retrograde labeling methods identified them as MSN (Somogyi and Smith 1979). Retrograde axonal transport studies in primates further indicated that striatal projection neurons could be subdivided on the basis of their projections to the Gpe or to the Gpi and SNr (Parent et al. 1984a, b) and at least three types of neurons were distinguished based on the fact that they projected either to the Gpe alone, to the SNr alone, or to both structures (Feger and Crossman 1984). Single cell-tracing methods provided further insights into the connectivity of striatal neurons and confirmed that MSN projected to more than one structure (Chan et al. 1981; Wilson and Phelan 1982; Parent et al. 1995a, b; Wu et al. 2000). In the primate, at least three types of striatal neurons were identified based on their target region (Parent et al. 1984a,1995a, b). One type projected to the GPe alone, a second type projected to the GPe and GPi, and a third type projected to the GPi, GPe, and SNr (Parent et al. 1995a, b). In the rat striatum, neurons were similarly subdivided into three types. Type I neurons projected to the GP only, type IIa neurons projected primarily to the SNr and EP, but also sent a small projection to the GP and type IIb neurons projected to the GP and SNr but not to the EP (Kawaguchi et al. 1990). The proportion of striatal neurons projecting to these different structures was not documented in this later study. A retrograde labeling study found that about one third of MSN that project to the GP have axon collaterals to the SNr (Castle et al. 2005).

In summary, tract-tracing combined with single-cell labeling studies have revealed that some MSN preferentially project to the GP (primate Gpe), while others preferentially project to the EP (or the primate Gpi) and to the SNr. In the literature, MSN that project to the GP (or primate GPe) are known as striatopallidal or indirect pathway neurons, while MSN that project to the SNr and/or EP (primate Gpi) are known as striatonigral or direct pathway neurons (Fig. 3.1). However, it is apparent that this subdivision is a simplification and that some MSN do not fit this strict classification. Using viral gene transfer strategies in transgenic mice, it has been shown that the density of axon collaterals in the GP made by MSN that primarily project to the SNr (direct pathway) increases when the excitability of striatopallidal neurons is increased (Cazorla et al. 2014 and 2015). This indicates that the specificity of axonal projections from MSN can be modulated and further calls into questions the notion that striatal projections can be rigidly subdivided into a direct and indirect pathway.

Fig. 3.1
figure 1

Illustrates the major connections between basal ganglia structures and the organization of the direct and indirect pathway in a sagittal view of the rat brain. As discussed in this chapter, the subdivision into a direct and indirect pathway is a simplification since indirect pathway MSN can send axon collaterals to the SNr/EP, while direct pathway MSN can send axon collaterals to the GP. GP globus pallidus, EP entopeduncular nucleus, STN subthalamic nucleus, SNc substantia nigra pars compacta, SNr substantia nigra, pars reticulata

Full size image

As discussed in previous paragraphs, MSN can express various combinations of peptides (Besson et al. 1990; Surmeier et al. 1996; Reiner et al. 1999; Nadjar et al. 2006; Wang et al. 2006, 2007). In the rat, it was reported that striatal neurons labeled after an injection of retrograde tracer in the SNr were labeled with dynorphin and substance P, but only 1 % co-expressed enkephalin immunoreactivity (Lee et al. 1997). In contrast, neurons labeled after an injection into the GP were labeled with enkephalin, but only 17 % and 10 % were, respectively, labeled for dynorphin and substance P (Lee et al. 1997). An in situ hybridization study in the rat has shown that the majority of neurons expressing enkephalin project to the GP while a few project to the SNr, whereas neurons expressing dynorphin and substance P project mainly to the SNr but a few also project to the GP (Gerfen and Young 1988). In the monkey, 70 % and 50 % for neurons labeled after an injection of retrograde tracer, respectively, into the GPe or into the GPi co-expressed the three peptides (Nadjar et al. 2006). It is unclear if the discrepancy in co-expression between rodent and primate studies is due to species and/or methodological differences. In any case, current evidence suggests that MSN that co-express the three peptides may be those that do not fit the strict definition of direct and indirect pathway neuron.

2.3 Segregated Expression of Dopamine Receptors

Early neurochemical studies have shown that the dopamine D1 and D2 receptors are the two major subtypes of dopamine receptors expressed in the striatum and that they exert opposite effects on the activation of adenylyl cyclase, with D1 receptors being stimulatory and D2 receptors inhibitory (Kebabian and Calne 1979; Stoof and Kababian 1981, 1984). Molecular cloning studies have determined that the family of dopamine D1 receptors includes the Drd1a and Drd5 receptors and that the family of dopamine D2 receptors includes the Drd2, Drd3, and Drd4 receptors (review in Beaulieu and Gainetdinov 2011). Although all dopamine receptors are expressed in the striatum (Surmeier et al. 1996), the Drd1a and the Drd2 receptors are the most abundant. The following paragraphs discuss the notion that the expression of the Drd1a and Drd2 receptors also contributes to define two different populations of MSN. This is an important notion since most studies carried out in genetically engineered mice or using viral delivery methods are based on it.

In a combined immunohistochemical and electron microscope study in the rat striatum, no co-localization of the D1 and D2 receptor was seen (Hersch et al. 1995). On the other hand, a study found that all striatal neurons co-expressed the D1 and D2 receptor (Aizman et al. 2000). However, between these two extreme outcomes, most other studies support the idea that only a subset of MSN co-expresses the D1 and D2 receptor (e.g. Meador-Woodruff et al. 1991; Weiner et al. 1991; Lester et al. 1993; Larson and Ariano 1994; Deng et al. 2006). The possibility that only some MSN co-express the Drd1a and Drd2 receptors was confirmed using PCR combined with path-clamp (Surmeier et al. 1996). The development of the Bacterial Artificial Chromosome (BAC) technology and genetically engineered mice has confirmed, at least in rodents, the limited co-expression of dopamine D1 and D2 receptors (e.g. Valjent et al. 2009). In mice expressing the marker tdTomato under the control of the Drd1a promotor and green fluorescent protein under the control of the Drd2 promotor, at embryonic day 18 only about 10 % of MSN were double-labeled. This proportion decreased at post-natal day 1 and 14 (Thibault et al. 2013). Similar evidence for limited co-expression was found in the neonatal mouse (Biezonski et al. 2015).

There is evidence that the segregation or co-expression of dopamine Drd1a and Drd2 receptors may correlate with the pattern of expression of specific peptides. Using a combination of patch-clamp and single-cell qPCR analysis , it was found that MSN having detectable levels of enkephalin, but not substance P mRNA, expressed high levels of the Drd2 receptor mRNA, while MSN with detectable levels of substance P but not enkephalin mRNA expressed high levels of the Drd1a receptor mRNA (Surmeier et al. 1996). The mRNAs for other dopamine receptor subtypes were rarely detected in MSN expressing enkephalin, but some co-expressed the D1b receptor (Surmeier et al. 1996). Conversely, the Drd3 receptor mRNA was detected in one-half of MSN expressing substance P, but other dopamine receptors were rarely detected (Surmeier et al. 1996). Finally, most MSN that co-expressed detectable levels of substance P and enkephalin mRNAs also co-expressed the Drd1a and Drd2 mRNAs (Surmeier et al. 1996).

Current evidence supports the notion that the segregation of MSN based on the expression of specific dopamine receptors and/or peptides correlates with a pattern of projection. This possibility is supported by several immunohistochemical and gene expression studies that have shown that Drd1a receptors are expressed in MSNs that primarily project to the SNr and/or the EP (or primate Gpi), while Drd2 receptors are expressed in MSNs that primarily project to the GP (or primate Gpe) (Aubert et al. 2000; Beckstead et al. 1988; Gerfen et al. 1990; Harrison et al. 1990; Le Moine et al. 1991; Harrison et al. 1992; Herve et al. 1993; Le Moine and Bloch 1995; Yung et al. 1995). Several studies in genetically engineered mice have confirmed that fluorescence induced by the activity of the Drdr1a receptor promotor in the striatum is high in the SNr, while fluorescence induced by the activity of the Drd2 receptor in the striatum is high in the GP (Gong et al. 2003; Lobo et al. 2006; Gertler et al. 2008; Bertran-Gonzalez et al. 2008; Shuen et al. 2008; Matamales et al. 2009). However, a combined confocal and retrograde labeling study in the rat found that although a large majority of neurons projecting to the SNr and EP also expressed the D1 receptor, 23 % of neurons projecting to the GP also expressed the D1 receptor (Deng et al. 2006). Conversely, although the vast majority of MSN projecting to the GP were labeled for the D2 receptor, 40 % of MSN projecting to the SNr and EP were also labeled for the D2 receptor (Deng et al. 2006). Another study confirmed that although MSN neurons projecting to the SNr mainly express the Drd1a receptor, some also expressed the Drd2 receptor (Matamales et al. 2009). In another study, however, MSN labeled with a retrograde marker injected in the SNr did not express the D2 receptor (Gertler et al. 2008). A combined retrograde and immunohistochemical study in the monkey from Nadjar and colleagues has shown that MSN projecting to the Gpi or to the Gpe are immunolabeled for both dynorphin and enkephalin and for both the D1 or D2 receptor (Nadjar et al. 2006).

In conclusion, most experimental studies support the notion that MSN can be subdivided based on their expression of the Drd1a and Drd2 receptors, of the peptides enkephalin, substance P, and dynorphin, and on their area of projection. One consensus that emerges is that most MSN that project to the SNr and the GPi (or rodent EP) also express substance P and dynorphin and the Drd1a receptor, while most MSN that project to the GPe (or rodent GP) also express enkephalin and the Drd2 receptor. Gene expression studies support this dichotomy since drugs acting on D1 receptors or on D2 receptors differentially modulate gene expression of peptides preferentially expressed by direct or indirect pathway neurons (e.g. Bertran-Gonzalez et al. 2008; Gerfen et al. 1990; Cenci et al. 1992; Cole et al. 1992; Dragunow et al. 1990; Laprade and Soghomonian 1995; Robertson et al. 1992). However, based on the data discussed above, it is also clear that some MSN can co-express both the Drd1 and Drd2 receptors and can co-express the peptides enkephalin and substance P/dynorphin. The possibility that those MSN that co-express all markers are those that project to the SNr, EP (or Gpi), and GP (or Gpe) is supported by some studies. Interestingly, it has been recently shown that the activation of Drd2-expressing MSN in genetically modified mice increases the density of axon collaterals from direct pathway neurons to the GP (Cazorla et al. 2014). These striatonigral axon collaterals are functional and able to inhibit the firing rate of GP neurons (Cazorla et al. 2014). In contrast, the density of axon collaterals from striatonigral neurons to the GP did not change when the excitability of Drd1-expressing striatonigral neurons was modulated (Cazorla et al. 2014). This pioneering study indicates that the connectivity of MSN is not static, but can be modulated in different physiological conditions and it further emphasizes the notion that the subdivision of MSN into a direct and indirect pathway is a simplification. The possibility that MSN that do not fit the strict classification of direct and indirect pathway neuron play a distinct role in the physiology of the basal ganglia remains to be determined. With this caveat in mind, the following paragraphs will present and discuss evidence that the so-called direct and indirect MSN have different physiological properties and functional roles.

2.4 Membrane Properties of Direct and Indirect Pathway Neurons

The heterogeneous connectivity and chemical phenotype of striatal projection neurons are paralleled by heterogeneous electrophysiological and morphological properties . The organization of MSN into Drd1a and Drd2-expressing subsets may be determined in part by cortical inputs because striatal neurons expressing the Drd1 receptor receive a majority of inputs from cortical neurons whose projections are restricted to the telencephalon, whereas striatal neurons expressing the Drd2 receptor receive more input from cortical neurons that contribute to the pyramidal tract (Lei et al. 2004). Using RT-PCR and confocal microscopy in slice preparations from mutant mice expressing eGFP under the activity of the dopamine Drd1 or Drd2 receptor, it was reported that Drd1-expressing striatal neurons are less excitable than Drd2-expressing neurons (Gertler et al. 2008). In addition, Drd1-expressing neurons have more primary dendrites than Drd2-expressing neurons (Gertler et al. 2008). Such a difference in excitability was also documented in another study showing that the threshold for firing action potentials is lower in Drd2-expressing than in Drd1-expressing MSN (Cepeda et al. 2008). Whole-cell and outside-out patch recordings in slices from bacterial artificial chromosome (BAC) transgenic mice were used to examine the role of GABAA receptor-mediated currents in dopamine receptor Drd1- and Drd2-expressing neurons (Ade et al. 2008). Although inhibitory synaptic currents were similar between the two neuronal populations, D2-expressing neurons had greater GABAA receptor-mediated tonic currents . Low GABA concentrations produced larger whole-cell responses and longer GABA channel openings in Drd2- than in Drd1-expressing neurons (Ade et al. 2008). It has been reported that the loss of dopamine innervation to the striatum differentially affects the excitability of Drd1- and Drd2-expressing neurons (Fieblinger et al. 2014). In parkinsonian mice, intrinsic excitability of Drd2-expressing neurons was depressed. High-dose l-DOPA treatment normalized intrinsic excitability. In contrast, the intrinsic excitability of Drd1-expressing neurons was significantly elevated and high-dose l-DOPA partially normalized this effect (Fieblinger et al. 2014). Altogether, these studies reinforce the notion that the different connectivity and chemical phenotype of Drd1 and Drd2-expressing striatal neurons is paralleled by different functional properties. The factors contributing to these differences remain unclear, but could involve cortical inputs because an electron microscopy study has shown that cortical synapses are smaller on Drd1- than on Drd2-expressing neurons (Lei et al. 2004).

3 Functions of the Direct and Indirect Pathway

3.1 Movement Control

The classical functional models of the basal ganglia are based on the notion that activation of the striatal direct pathway facilitates movement, while activation of the indirect pathway inhibits movement (Alexander et al. 1986; Alexander and Crutcher 1990; DeLong 1990). These models are supported by anatomical and physiological data and propose that paucity or loss of movement in Parkinson’s disease results from an increased activation of indirect pathway neurons and a decreased activation of direct pathway neurons (Albin et al. 1989). This dual effect would result in an increased basal ganglia output and an increased inhibition of thalamo-cortical projections to the frontal and prefrontal cortex ultimately leading to a lesser activation of cortical motor and premotor regions. Gene expression studies are consistent with an opposite role of the direct and indirect pathway on movement because in experimental models of Parkinson’s disease, enkephalin gene expression in the indirect pathway is increased and preprodynorphin and preprotachykinin expression is decreased in the direct pathway (Reviewed in Soghomonian and Chesselet 2000). These changes in peptide gene expression have been considered to parallel changes in neuronal activity. A complementary role of the direct and indirect pathway in movement control was proposed in another model in which the direct pathway would contribute to the selection of motor programs, while the indirect pathway would inhibit competing motor programs (Mink 1996). The idea that the direct and indirect pathways have opposite and/or complementary roles on movement has been tested in transgenic mice models and using viral targeting methods. For instance, optogenics has been used in mice expressing channelrhodopsin-2 under the activity of the dopamine Drd1a or Drd2 receptors with the objective of independently manipulating direct or indirect pathway MSN. Using this approach, it was found that the bilateral excitation of striatal neurons expressing the dopamine Drd2 gene elicited a Parkinsonian state in mice, characterized by increased freezing, bradykinesia, and decreased locomotor initiation (Kravitz et al. 2010). In contrast, activation of striatal neurons expressing the Drd1a gene reduced freezing episodes and increased locomotion (Kravitz et al. 2010). In addition, activation of Drd1a-expressing neurons completely rescued freezing, bradykinesia, and deficits in locomotor initiation observed in a 6-hydroxydopamine-lesioned mouse model of Parkinson’s disease (Kravitz et al. 2010). Conversely, other evidence has shown that the experimental ablation or disruption of the indirect pathway increases motor activity (Durieux et al. 2009; Bateup et al. 2010). Although the studies described above are consistent with the hypothesis that the direct and indirect pathways play an opposite role in the activation of movement, they do not clarify their respective role in various aspects of movement performance such as movement selection, initiation, termination, or in instrumental learning. The following paragraphs review and discuss studies that have attempted to address these questions.

Using a Cre-dependent viral expression of the genetically encoded calcium indicator GCaMP3 in Drd1a receptor- or A2a receptor-expressing (respectively direct and indirect pathway neurons) neurons in the striatum, Cui and co-workers were able to study the pattern of activation of direct and indirect pathway neurons during the execution of movement in mice performing an operant task (Cui et al. 2013). They found that both pathways were co-activated during the initiation of movement and that their concurrent activation preceded the initiation of contraversive movements and predicted the occurrence of movement (Cui et al. 2013). These findings suggest that the initiation and execution of normal movements requires a co-activation of direct and indirect striatal circuits. The finding of a co-activation of direct and indirect pathway MSN is consistent with the model proposing that these pathways could contribute to concomitantly activate selected movements and inhibit competing movements. In a study combining optogenetic identification of direct and indirect pathway MSN with electrophysiological recordings in mice that were trained to learn a rapid motor sequence, Jin and colleagues (Jin et al. 2014) found that similar percentages of direct and indirect pathway MSN responded during the start or the end of the sequence. However, while direct pathway neurons responded similarly at the start and end of the sequence, indirect pathway neurons preferentially responded at the start of the sequence (Jin et al. 2014). Jin and colleagues interpreted this result as evidence that the direct pathway plays a preferential role in the initiation of movement, while the indirect pathway plays a preferential role in the inhibition of competing motor programs (Jin et al. 2014). The finding that the majority of changes in MSN activity occurred at the start and end of a motor sequence rather than during the sequence itself was interpreted as evidence that the basal ganglia control sequences of movements (chunking), rather than individual movements (Jin et al. 2014). In another study, genetically engineered mice were trained to execute two distinct and sequential responses to get a reward in an operant chamber (Rothwell et al. 2015). Using selective manipulations of direct and indirect pathway neurons, the study reported that serial order learning strengthened cortical synapses on direct pathway neurons (Rothwell et al. 2015).

The dual role of the direct and indirect pathways on movement is paralleled by a dual effect on neurons in the output regions of the basal ganglia. Indeed, the effectiveness of optogenetic stimulation of the direct pathway in producing movement significantly correlated with the extent of inhibition of a subpopulation of SNr neurons (Freeze et al. 2013). In contrast, motor suppression induced by activation of the indirect pathway seemed to be most strongly influenced by the population of excited SNr neurons (Freeze et al. 2013). Freeze and colleagues argued that the striatal direct and indirect pathways represent an inhibitory gate that can respectively open or close motor output from the basal ganglia (Freeze et al. 2013). This interpretation is consistent with other experimental evidence that signals through the striatopallidal indirect pathway inhibit movements through a phasic excitation of the SNr (Sano et al. 2013). In their study, Jin and colleagues found that the activity in the SNr correlated with that of direct pathway neurons, while activity in the GP correlated with that of indirect pathway neurons (Jin et al. 2014).

Most studies reviewed above are consistent with the hypothesis that activation of direct pathway neurons facilitates movement, while activation of indirect pathway neurons inhibits movement. A more complex theoretical model has been proposed in which activation of indirect pathway MSN would contribute to both the selection and concurrent inhibition of competing movements (Keeler et al. 2014). The model, which is based on evidence that dopamine D1 and D2 receptors have different biochemical properties and that their pharmacological manipulation differentially alter different phases of movement in an operant task, proposes that the direct and the indirect pathway are, respectively, involved in the preparation and the selection of movement (Keeler et al. 2014). In this model, the activation of a small subset of indirect pathway MSN would contribute to select movement and concurrently would exert a lateral inhibition on neighboring indirect pathway MSN to inhibit competing movements. In this model, the paucity of movement observed in Parkinson’s disease could be explained by an abnormal activation of large populations of indirect pathway MSN so that the mechanisms leading to movement selection via lateral inhibition would be disrupted (Keeler et al. 2014). The possibility that the indirect pathway is involved in movement selection appears consistent with experimental evidence that its selective elimination impairs the accuracy of response selection in the execution of an auditory discrimination task without influencing the response time (Nishizawa et al. 2012). Conversely, selective elimination of the striatonigral pathway lengthens the response time, but does not affect the accuracy of a response selection in a two-choice reaction time task dependent on a visual stimulus (Fukabor et al. 2012). In conclusion, the exact role of the direct and indirect pathway in the control of movement remains hypothetical and future studies using more refined methods should help settle the uncertainty about the role of these pathways in movement initiation and movement selection.

3.2 Associative Learning, Social Behavior, and Decision Making

The basal ganglia and the striatum play an important role in learning and executing a motor performance in response to a specific sensory or environmental context (Seger and Spiering 2011). In particular, the striatum is involved in action-outcome learning and in habit learning. In action-outcome learning, the performance of a specific behavior depends on a mental representation of the outcome. In habit learning, the performance depends on a particular context. Habits are less sensitive to reward devaluation, indicating a competition between action-outcome learning and habits. The ventromedial striatum may be preferentially involved in action-outcome learning, while the dorsolateral striatum may be preferentially involved in habit learning (Balleine et al. 2007). The reader is referred to Chaps. 5, 11, 12, 18, and 19 for more detailed discussions on the role of the striatum in learning. The objective in the following paragraphs will be to discuss the respective contribution of the striatal direct and indirect pathways to learning and learning-dependent behaviors such as social behavior and decision-making.

A number of studies have used genetically engineered mice to selectively manipulate the direct or indirect pathways and assess the impact on operant learning. These studies suggest a differential role of the direct and indirect pathways in different aspects of associative and reward-based learning. In particular, these studies support the notion that the direct pathway is involved in reward-based learning, whereas the indirect pathway may be involved in avoidance behavior. For instance, in a place preference paradigm in an operant box, optogenic stimulation of Drd1a-expressing neurons induced a persistent reinforcement, whereas stimulation of Drd2-expressing neurons induced a transient punishment (Kravitz et al. 2012). Using another genetic approach to selectively inactivate with tetanus-toxin striatal neurons expressing substance P or enkephalin (direct and indirect neurons, respectively), Hikida and colleagues found that loss of the direct but not the indirect pathway impaired reward-based learning (Hikida et al. 2016). In contrast, the avoidance aversive behavior in a dark chamber associated with an electric shock was impaired after loss of the indirect but not the direct pathway, leading the authors to conclude that the indirect pathway is critical for evoking aversive behavior (Hikida et al. 2010, 2016). Using a similar experimental approach, it was shown that Drd1a receptors in the direct pathway are critical for the acquisition, but not for the expression of appetitive reward learning (Hikida et al. 2013). In contrast, activation of Drd2 receptors in indirect pathway neurons was critical for both the acquisition and expression of aversive behavior (Hikida et al. 2013). When the transmission of either direct or indirect pathway MSN was unilaterally blocked using tetanus toxin, infusion of protein kinase A inhibitors in the accumbens core abolished passive avoidance to an electric shock when the indirect pathway was blocked (Yamaguchi et al. 2015). In addition, protein kinase A activity was increased in indirect pathway and decreased in direct pathway neurons in both aversive memory formation and retrieval (Yamaguchi et al. 2015), indicating that the second messengers systems associated with dopamine receptors are involved in these effects. In another series of experiments, mice were trained to lick a spout in response to a whisker deflection (Sippy et al. 2015). Striatal projection neurons in the dorsolateral striatum showed a strong task-related modulation and increased their activity in successful trials (Sippy et al. 2015). However, direct but not indirect pathway neurons exhibited a prominent early sensory response and optogenetic stimulation of direct pathway neurons substituted for whisker stimulation in trained mice (Sippy et al. 2015). These data support the hypothesis that direct pathway neurons are permissive for the initiation of learned reward-based action (Sippy et al. 2015). Francis and colleagues documented a dual effect of the direct and indirect pathways in mood and motivated behavior. Specifically, the activity of Drd1a-expressing neurons was decreased, while the activity of Drd2-expressing neurons was increased in mice displaying depression-like behaviors after chronic social defeat stress (Francis et al. 2015). Stimulation of Drd1a-expressing neurons increased behavioral resilience to depression, while inhibition induced depressive-like behavior after chronic social defeat stress. In contrast, the repeated activation of indirect pathway neurons in stress naïve mice induced social avoidance following a subthreshold exposure to a social defeat stress (Francis et al. 2015). Another study has shown that stimulation of Drd2-expressing neurons of the nucleus accumbens converts risk-preferring rats to risk-averse rats (Zalocusky et al. 2016). This finding is consistent with a general role of the indirect pathway in avoidance behavior.

Other studies indicate that in addition to be involved in avoidance behavior, the indirect pathway may play an important role in mediating cognitive flexibility by preventing the execution of actions that used to be rewarded but that are not anymore. Using the transmission-blocking tetanus toxin approach in the mouse, it was documented that the direct pathway in the nucleus accumbens is required for learning the association between a visually cued task and a reward (Yawata et al. 2012). In contrast, inactivation of the indirect pathway did not impair learning acquisition, but it increased perseverative behavior in response to a strategy switch in which the reward was placed in another location (Yawata et al. 2012). In this study, the administration of the D2 receptor agonist quinpirole tended to increase perseverative errors, particularly during the switching task, thus confirming that a decreased inhibitory action of D2 receptors on indirect pathway neurons is necessary for learning a new strategy (Yawata et al. 2012; Nakanishi et al. 2014). These data are consistent with the model of “Go” and “No Go” in which the Go signal is provided by activation of the direct pathway and the “No Go” signal by activation of the indirect pathway (Frank et al. 2004; Frank 2011). In such a case, a decreased activation of indirect pathway neurons could lead to perseveration and enhance the expression of habits. This is otherwise supported by evidence that post-synaptic plasticity of Drd2-expressing striatopallidal neurons in the dorsolateral striatum correlates with habit learning (Shan et al. 2015). In addition, habitual behavior in mice was correlated with a strengthening of direct and indirect pathway neurons in the dorsolateral striatum (O’Hare et al. 2016), but neurons in the direct pathway had a tendency to fire before the indirect pathway and habit suppression correlated with a weakened direct pathway output while habit expression correlated with indirect pathway event amplitude (O’Hare et al. 2016).

3.3 Addiction and Obesity

It is well-established that the striatum and dopamine are involved in reward-mediated behaviors and in addiction (Schultz 2011 and 2013; Hyman et al. 2006). On the other hand, food intake is regulated via several mechanisms, among which the reward system plays an important role. Obesity can be the result of excessive food consumption and may involve mechanisms similar to those involved in drug abuse (Kenny et al. 2013). The following paragraphs discuss evidence that the direct and indirect pathway play a dual role in addiction and in obesity.

Earlier studies have documented that pharmacological antagonists of D1 receptors block conditioned place preference for cocaine (Hiroi and White 1991; Baker et al. 1998). Using a fluorescent calcium indicator as a marker of neuronal activity, it was found that cocaine intake shifts the balance between the direct and indirect pathway towards the direct pathway (Luo et al. 2011) and loss of the direct pathway reduces locomotor activity and attenuates locomotor sensitization to repeated cocaine (Hikida et al. 2016). Similarly, decreased excitability of the direct pathway impairs persistence of amphetamine-induced behavioral sensitization (Ferguson et al. 2011). In the conditioned place preference paradigm, blockade of the direct but not indirect pathway reduces cocaine-induced place preference (Hikida et al. 2016). Optical activation of nucleus accumbens Drd1a- but not Drd2-expressing MSN enhanced morphine-conditioned place preference (Koo et al. 2014). In another study, it was found that activation of dopamine D1 receptors on the direct pathway is important for inducing cocaine-dependent sensitization and cocaine-induced addictive behavior (Hikida et al. 2013). Lobo and colleagues subsequently reported that a targeted deletion of Tropomyosin-related kinase B (TrkB), the receptor for the brain-derived neurotropic factor (BDNF), in direct pathway MSN diminished the rewarding properties of cocaine (Lobo et al. 2006). Loss of the dopamine-receptor activated second messenger DARPP-32, in direct but not in indirect pathway MSN, prevented the stimulatory action of the psychomimetic phencyclidine on motor activity (Bonito-Oliva et al. 2016). Altogether, these findings are consistent with the notion that the activation of the direct pathway plays a key role in several addictive effects induced by psychostimulants. In contrast, activation of the indirect pathway seems to play an opposite role on several psychostimulant-induced behaviors. For instance, there is evidence that increasing the activity of the indirect pathway promotes resilience to compulsive cocaine seeking (Bock et al. 2013). Lobo and colleagues reported that the targeted deletion of the neurotrophic receptor TrkB in Drd2-expressing MSNs enhanced cocaine reward (Lobo et al. 2006). Moreover, TrkB deletion in Drd2-expressing MSN increased the excitability of indirect pathway neurons and optogenetic stimulation of these neurons decreased cocaine reward-seeking behavior (Lobo et al. 2006). Loss of the indirect pathway also leads to a delayed cocaine sensitization, although sensitization eventually re-emerges (Hikida et al. 2013). A decreased excitability of the indirect pathway facilitates behavioral sensitization (Ferguson et al. 2011). An increased synaptic strength of glutamatergic synapses on Drd2-expressing indirect pathway neurons in the nucleus accumbens was documented in mice with a history of intravenous cocaine self-administration (Bock et al. 2013). This synaptic strengthening was inversely correlated with the emergence of compulsive-like cocaine responding (Bock et al. 2013). Altogether, these data suggest that activation of the indirect pathway may oppose the addictive properties of drugs of abuse.

Adenosine A2a receptors are densely expressed in striatopallidal neurons (Svenningsson et al. 1997; Schiffmann et al. 2007). Pharmacological agonists that modulate adenosine A2a receptors and increase striatopallidal transmission reduced consumption of both highly palatable and standard chow in rats (Micioni Di Bonaventura et al. 2012) and reduced lever-pressing for food rewards (Jones-Cage et al. 2012). Conversely, pharmacological blockade of A2a receptors increased palatable food consumption when administered alone and enhanced palatable food intake triggered by intra-accumbens administration of an μ-opioid receptor agonist (DAMGO) (Pritchett et al. 2010). These findings are reminiscent of the inhibitory effects of indirect pathway stimulation on drug reward described in the previous paragraphs and suggest that Drd2-expressing indirect pathway neurons may regulate food intake in much the same way that they regulate drug rewards. A link between compulsive eating and indirect pathway neurons is supported by some studies. In particular, viral knockdown of Drd2 receptors in the striatum accelerates the development of compulsive food-seeking behavior in rats (Johnson and Kenny 2010), suggesting that the indirect pathway may control compulsive food-seeking.

4 Conclusions

The existence of a direct and indirect striatal pathway is supported by considerable experimental evidence, but there is also evidence that this segregation is not absolute. In addition, recent evidence indicates that both the density of MSN axonal projections to a specific target and the expression of phenotypic markers in MSN can change during the development of the brain and/or in response to physiological challenges. Studies in genetically engineered mice have documented that the manipulation of neurons that preferentially express phenotypic markers of direct or indirect pathway neurons (i.e., dopamine Drd1a versus Drd2 receptors or enkephalin versus substance P and dynorphin) has a different impact on behavior. It should be emphasized, however, that most mice studies manipulate subsets of MSN based on their expression of specific dopamine receptors rather than on their specific area of projection. Thus, the function of MSN that do not fit the strict definition of direct or indirect pathway neuron (i.e. neurons that project to all output regions of the basal ganglia) remains unclear.

Another major outcome of recent experimental studies in mice has been to support the notion that the direct and indirect pathways play an opposite and/or complementary role in the organization of movement, in associative and in reward-based learning. In particular, current evidence supports the notion that the direct pathway is involved in the facilitation of movement and reward-associated actions, while the indirect pathway is involved in the inhibition of competing motor actions and/or the inhibition of unrewarded actions. It is important to emphasize that most studies leading to these conclusions involved experimental conditions in which the activity of large numbers of MSN was homogeneously manipulated, a situation that most likely does not occur in physiological conditions. In fact, the temporal and spatial pattern of activation or deactivation of direct and indirect pathway neurons during the preparation, initiation, execution, and termination of actions is complex. This suggests that different subsets of direct and indirect MSN code for different variables associated with an action. In order to provide a better insight into the functions of MSN, future studies should aim at activating and/or deactivating more discrete subsets of direct or indirect pathway neurons and multi-synaptic neuronal circuits associated with different subsets of direct and/or indirect pathway neurons.