Direct and Indirect Pathways of Basal Ganglia: A Critical Reappraisal

Abstract

The basal ganglia are subcortical nuclei controlling voluntary actions and have been implicated in Parkinson's disease (PD). The prevailing model of basal ganglia function states that two circuits, the direct and indirect pathways, originate from distinct populations of striatal medium spiny neurons (MSNs) and project to different output structures. These circuits are believed to have opposite effects on movement. Specifically, the activity of direct pathway MSNs is postulated to promote movement, whereas the activation of indirect pathway MSNs is hypothesized to inhibit it. Recent findings have revealed that this model might not fully account for the concurrent activation of both pathways during movement. Accordingly, we propose a model in which intrastriatal connections are critical and the two pathways are structurally and functionally intertwined. Thus, all MSNs might either facilitate or inhibit movement depending on the form of synaptic plasticity expressed at a certain moment. In PD, alterations of dopamine-dependent synaptic plasticity could alter this coordinated activity.

Full text

1022 VOLUME 17 | NUMBER 8 | AUGUST 2014 nature neuroscience
review
“Homines dum docent discunt” (men learn while they teach). This
sentence is from the seventh letter of Seneca to Lucilius1. Seneca indi-
cates that learning is a mutual process. For example, when preparing
or giving a lecture on an established scientific theory, one is forced
to weigh the evidence for and against this theory. Often, lecturing
can require greater clarity than is necessary when discussing experi-
ments in the laboratory. Thus, when teaching, we repeatedly analyze a
theory, learning through this didactic process. For instance, teaching
in a course of neuroscience and neurology, the functional anatomy
of several systems and its clinical consequences seems relatively easy.
Conversely, the description of the functional role of the basal ganglia
(BG) in motor control and the pathophysiology of related conditions
such as PD appears much more complex. Thus, when explaining the
canonical model of BG organization, the teacher must either provide
an oversimplification or a too complex explanation of the circuits
underlying movement and related disorders.
Is this problem a simple and limited didactic dilemma, addressable
with better chapters in textbooks, or does it involve a more general
problem with our understanding of the real organization of the BG?
Bearing out Seneca’s wisdom, while teaching BG physiology and patho-
physiology according to the model of direct/indirect pathways, we came
to believe that the current view has some limitations. In particular, we
feel that this model seems too rigid to account for recent experimental
and clinical findings that have revealed the complexity of the system
while simultaneously being too complex to explain the organization of
the BG to a class of students. Below, we attempt to address this issue,
discussing the implications of recent findings for the direct/indirect
pathway model.
Before the model
Several critical anatomical and physiological descriptions of BG
networks and the striatal complex were made before the elaboration of the
direct/indirect model2. From these studies, a simple network emerged,
linking cortex, striatum and output structures. Briefly, the striatal complex
receives glutamatergic excitatory inputs from cortical and thalamic struc-
tures. These inputs converge in the striatum to establish synapses with
both MSNs, the GABAergic output cells representing about 95% of striatal
neurons3, and aspiny GABAergic and large cholinergic interneurons4.
Other neuromodulatory inputs5, most prominently dopaminergic inputs
from the substantia nigra pars compacta (SNpc)5, reach the striatum.
These dual glutamatergic and dopaminergic projections converge onto
dendritic spines of the same MSN6. In addition, striatal interneurons
receive both glutamatergic and dopaminergic inputs, and most of them
synapse onto MSNs, representing a link between neuronal inputs from
striatal afferents and striatal projecting neurons7.
The classical model of direct and indirect pathways
The canonical view of the interaction between glutamatergic and
dopaminergic neurotransmission in the striatum, which lies at the
heart of the direct/indirect pathway model, originated in seminal
papers that hypothesized a dual organization of the striatum and
of BG outputs8,9. According to this model, cortical activation pro-
duces a release of glutamate that activates MSNs projecting to the
substantia nigra pars reticulata (SNpr) and the globus pallidus pars
interna (GPi) (the striato-nigral output neurons representing the
direct pathway; Fig. 1). MSNs are GABAergic cells; thus, they exert
an inhibitory action on neurons of the SNpr that are also GABAergic.
This inhibition of the SNpr leads to a disinhibition of the thalamic
glutamatergic neurons, which receive SNpr input and project to the
cortex. The behavioral result of this chain of events is locomotor
activation/movements.
Conversely, activation of striato-pallidal MSNs, which project
indirectly to the SNpr via the globus pallidus pars externa (GPe)
and the subthalamic nucleus (STN) (indirect pathway), inhibits the
1Clinica Neurologica, Dipartimento di Medicina, Università degli Studi di Perugia,
Ospedale Santa Maria della Misericordia, S. Andrea delle Fratte, Perugia, Italy.
2Fondazione Santa Lucia, IRCCS, via del Fosso di Fiorano 64, Rome, Italy.
Correspondence should be addressed to P.C. (paolo.calabresi@unipg.it).
Received 14 February; accepted 21 May; published online 28 July 2014;
doi:10.1038/nn.3743
Direct and indirect pathways of basal
ganglia: a critical reappraisal
Paolo Calabresi1,2, Barbara Picconi2, Alessandro Tozzi1,2, Veronica Ghiglieri2 & Massimiliano Di Filippo1
The basal ganglia are subcortical nuclei controlling voluntary actions and have been implicated in Parkinson’s disease (PD).
The prevailing model of basal ganglia function states that two circuits, the direct and indirect pathways, originate from distinct
populations of striatal medium spiny neurons (MSNs) and project to different output structures. These circuits are believed
to have opposite effects on movement. Specifically, the activity of direct pathway MSNs is postulated to promote movement,
whereas the activation of indirect pathway MSNs is hypothesized to inhibit it. Recent findings have revealed that this model
might not fully account for the concurrent activation of both pathways during movement. Accordingly, we propose a model in
which intrastriatal connections are critical and the two pathways are structurally and functionally intertwined. Thus, all MSNs
might either facilitate or inhibit movement depending on the form of synaptic plasticity expressed at a certain moment. In PD,
alterations of dopamine-dependent synaptic plasticity could alter this coordinated activity.
npg © 2014 Nature America, Inc. All rights reserved.

nature neuroscience VOLUME 17 | NUMBER 8 | AUGUST 2014 1023
review
GABAergic neurons of the GPe, leading to a disinhibition of the
glutamatergic neurons of the STN. The increased discharge of these
excitatory STN neurons in turn activates the SNpr GABAergic neu-
rons projecting to the thalamus. Ultimately, this effect results in the
reduction of locomotor activity and movement (Fig. 1).
In addition to their distinct projections, MSNs of the direct and
indirect pathway are characterized by the differential expression of
dopamine (DA) receptors. D1 DA receptors are expressed by direct
pathway MSNs, whereas D2 receptors are expressed by indirect path-
way MSNs. These two receptors are associated with distinct G proteins
that are linked to different intracellular signaling pathways and lead
to different biochemical responses following DA receptor activation.
This neurochemical segregation is considered to be further support
for a dichotomous effect of the activation of the direct and indirect
pathways10,11.
Experimental consequences of the model
The direct/indirect pathways model has been widely used to explain
experimental findings, build models of BG disorders, and explain
therapeutic effects of both pharmacological and neurosurgical treat-
ments. Distinct, and even opposite, roles of these two pathways in
regulating several physiological functions involving the BG, such as
basal locomotor activity and motor responses to drugs of abuse and
antipsychotic agents, have even been postulated12,13. In particular,
the selective loss of the striatal signaling protein DA- and cAMP-regulated
phosphoprotein 32 kDa (DARPP-32) in direct pathway MSNs reduces
basal and cocaine-induced locomotion, and, in a rodent model of
PD, abolishes dyskinetic behavior in response to levodopa (L-DOPA), a DA
precursor that is widely used as a form of therapy for PD. Conversely,
a loss of DARPP-32 in indirect pathway MSNs augments locomo-
tor activity and reduces cataleptic response to the antipsychotic
drug haloperidol12.
D1 striatonigral and D2 striatopallidal MSNs have typically been
considered as homogeneous populations regarding their somatoden-
dritic morphology, although neurochemical differences have been
found in these neuronal subtypes2,14. In fact, SNpr-projecting MSNs
express substance P, dynorphin and D1 DA receptors, whereas GPe
projecting MSNs express enkephalin and D2 DA receptors. The intro-
duction and use of D1 and D2 bacterial artificial chromosome (BAC)
transgenic mice13 to distinguish MSNs without post hoc analyses led
to the identification of further possible differences between these two
populations. Electrophysiological studies15,16 found differences in the
excitability of striatal D1 and D2 MSNs, with D2 MSNs consistently
firing at higher frequencies, as well as differences in resting membrane
potential, input resistance and rheobase current. Moreover, experi-
ments using two-photon laser-scanning microscopy in identified
MSNs showed that the dendrites of D2 MSNs are more excitable than
those of D1 MSNs, and that DA depletion augments this asymmetry17.
Notably, three-dimensional reconstructions revealed a significantly
greater total dendritic length of D1 versus D2 MSNs, suggesting that
dendritic anatomy might contribute to differences in MSN excitabil-
ity18. Finally, striatal DA denervation has been found to reduce spines
and glutamatergic synapses, potentially via dysregulation of a Ca2+
channel implicated in the pathophysiology of PD, on striatopallidal,
but not striatonigral, MSNs19.
The introduction of D1 and D2 BAC transgenic mice has also pro-
vided new results regarding features of striatal synaptic plasticity and
its possible functional implications. Although previous studies have
demonstrated the expression of activity-dependent long-term depres-
sion (LTD) in the large majority of MSNs, suggesting an absence of
neuronal segregation between the two pathways20–23, BAC-mediated
targeting of direct and indirect pathways yields a different result.
Kreitzer and Malenka16 showed a selective D2 receptor activation–
dependent LTD in MSNs of the indirect pathway. Notably, this form
of synaptic plasticity was absent in PD rodent models and was res-
cued by D2 receptor stimulation. To determine whether synaptic
plasticity could be unidirectional in D1 and D2 receptor–expressing
MSNs, spike timing–dependent plasticity was also investigated in
brain preparations from DA receptor BAC transgenic mice. This
study showed that, although DA has a complex and complementary
role in these two types of MSNs to ensure bidirectional plasticity in
physiological conditions, this role is altered in mouse models of PD
and only unidirectional changes in plasticity occur24.
Cortex
Putamen
GPe
GPi
DA
SNpc
SNpr
Thalamus
Caudate
STN
Physiological condition
Direct
Indirect
a
Cortex
Putamen
GPe
GPi
DA
SNpc
SNpr
Thalamus
Caudate
STN
Parkinson’s disease
Kim Caesar/Nature Publishing Group
b
Figure 1 Schematic representation of the direct/indirect pathway classical
model in the physiological condition and in Parkinson’s disease. (a) In the
physiological condition, DA arising from the SNpc is thought to activate
D1-expressing striatal MSNs of the direct pathway (red lines) and to inhibit
D2-expressing striatal neurons of the indirect pathway (blue lines). The
output nuclei GPi and SNpr project to the thalamus, which in turn sends
efferents that complete the cortico-basal ganglia-thalamo-cortical loop.
(b) In Parkinson’s disease, degeneration of nigral neurons reduces DA
receptor stimulation in striatal MSNs. The imbalance between direct and
indirect pathways results into abnormal activation of output nuclei and over-
inhibition of thalamic neurons projecting to the cortex.
npg © 2014 Nature America, Inc. All rights reserved.

1024 VOLUME 17 | NUMBER 8 | AUGUST 2014 nature neuroscience
review
More recently, optogenetic techniques allowed the identification of
specific cell types in vivo through optical stimulation of recorded cells
or optical monitoring of the activity of neurons of a specific cell type.
This approach has permitted the measurement of neural activity-
dependent fluorescence changes from specific types of neurons of the
BG in behaving animals. A study using in vivo optogenetic methods
showed that excitation of D1 and D2 receptor–expressing MSNs
acts bidirectionally on locomotion25. Notably, bilateral excitation of
indirect pathway MSNs elicited a parkinsonian-like state, character-
ized by freezing, bradykinesia and decreased locomotor initiation.
Conversely, activation of direct pathway MSNs reduced freezing and
increased locomotion. Moreover, in a rodent model of PD, activa-
tion of direct pathway reversed freezing, bradykinesia and deficits of
locomotor initiation25.
An interaction between the direct and indirect pathways in action
selection has been postulated, suggesting an integration between these
pathways in producing coordinated behavior, both in terms of motor
output and in fine temporal patterning of neural activity26,27. Indeed,
these interpretations of the classic direct/indirect model predict that
both pathways are active during movement selection and that the inte-
gration of the two outputs arbitrates between the ultimate selection of
motor programs based on outcomes obtained from specific actions.
This was highlighted in an interesting optogenetic study showing that
the stimulation of the direct pathway in one hemisphere increases the
likelihood of choosing the contralateral action according to action
value, whereas indirect pathway stimulation has the opposite effect,
effectively decreasing the action value28.
The selective activation of the direct/indirect pathway is a
possible critical aspect of the synaptic mechanisms implicated in
reinforcement and punishment. Given that reinforcement maintains
or increases, whereas punishment decreases, the probability of spe-
cific behaviors, alterations of these processes could be involved in
psychiatric disorders29 and drug addiction30. The striatum, in fact,
has been highly implicated in both reinforcement and punishment
processes. A recent study investigated the hypothesis that the direct
(indirect) pathway mediates reinforcement (punishment). According
to this hypothesis, optogenetic activation of D1-expressing neurons
causes persistent reinforcement, whereas activation of D2-expressing
neurons induces transient punishment31. Striatal DA levels, by acting
in a distinct manner on either the direct or the indirect pathway, are
critical in rewarding and aversive learning, as well as in drug addic-
tion. The direct pathway seems to be involved in reward learning and
cocaine sensitization, whereas the indirect pathway is implicated in
aversive behavior32. Consistent with this latter observation, it has also
been reported that a transient disruption of indirect pathway MSNs
activity facilitates behavioral sensitization, whereas the decrease of
excitability of direct pathway MSNs impairs the persistence of behav-
ioral sensitization following drug exposure33. Findings suggesting a
distinct role of the direct/indirect pathways in reward, punishment
and behavioral sensitization might be relevant both for psychiatric
diseases, such as depression and obsessive-compulsive disorder, and
impulse control disorders in PD patients. This latter condition is
a DA dysregulation syndrome, frequently observed in PD patients
assuming DA replacement therapy, and causing pathologic gambling,
hypersexuality, compulsive shopping, compulsive eating, excessive
engagement in hobbies and punding34.
Clinical consequences of the model
The direct/indirect pathway model has been of great importance in the
interpretation of the experimental and clinical findings obtained in ani-
mal models and PD patients following pallidotomy, subthalamotomy
and deep brain stimulation (DBS) of the STN, as well as of other BG
nuclei (Fig. 1). Unilateral GPi pallidotomy improved all of the cardinal
motor signs of PD, including tremor, rigidity, bradykinesia, abnormal
gait and balance. In addition, L-DOPA–induced dyskinesias were
markedly improved. Although the greatest improvement occurred on
the side contralateral to the lesion, significant ipsilateral improvement
was also observed for bradykinesia, rigidity and dyskinesias35.
The important clinical observation that lesions of the STN, a key
node of the indirect pathway, abolish the cardinal features of PD
contributed to a renaissance in the use of surgical approaches in
the treatment of PD. Although bilateral subthalamotomy improves
symptoms in advanced PD36, this clinical effect seems to be vari-
able, probably depending on the location and volume of the lesions.
Conversely, several clinical studies have clearly shown that the admin-
istration of high-frequency continuous electrical stimulation to the
STN through a surgically implanted device reduces motor symptoms
and L-DOPA–related motor complications in PD patients37. Clinical
trials demonstrated the superior efficacy of neurostimulation over
best pharmacological management both in patients with advanced
PD38 and in patients with early motor complications39. The successful
application of DBS in PD, as a consequence of the pathophysiological
theory of the direct/indirect pathway in BG disorders, led to the appli-
cation of this technique in several medication-refractory hyperki-
netic movement disorders, such as tremor and dystonia, as well as in
psychiatric diseases, such as obsessive-compulsive disorder40.
An imbalance in the activity of striatal direct and indirect path-
way MSNs has also been postulated in Huntington’s disease (HD),
a progressive fatal neurological condition caused by an expansion
of trinucleotide CAG repeats that leads to striatal degeneration41.
An electrophysiological study in mouse genetic models of HD has
suggested that there are differential and complex imbalances in
glutamate and DA modulation in direct and indirect pathway MSNs
during HD progression42. In particular, hyperactive behavior at the
early stage could be explained by augmented glutamate activity and
DA tone in direct pathway MSNs, whereas the hypokinesia obser ved
during advanced stages could be explained by a reduced input to
these neurons42.
Experimental evidence beyond the model
Despite the experimental and clinical findings described, which sup-
port the direct/indirect pathway model, the functional relevance of
these two pathways in motor generation and control is still a matter of
debate. In fact, in contrast with the classical model suggesting oppos-
ing roles of the two pathways, it has recently been demonstrated that
optogenetic activation of striatal direct and indirect pathway projec-
tion neurons produces different cellular responses in SNr neurons,
with stimulation of each pathway eliciting both excitations and inhi-
bitions43. Moreover, experimental findings have suggested a coor-
dinated activation of both pathways during action selection. Thus,
it is possible that the coordinated activity of the direct and indirect
pathways is critical for the appropriate timing and synchrony of BG
circuits during movement.
In particular, a recent study from Cui and colleagues44 challenges
the classical view of BG function, providing an alternative explanation
for understanding the origin of motor symptoms in BG disorders.
The authors developed an in vivo method to measure direct/indirect
pathway MSN activity that uses Cre-dependent viral expression of the
genetically encoded Ca2+ indicator GCaMP3 in the dorsal striatum
of D1-Cre (direct pathway specific) and A2A-Cre (indirect pathway
specific) mice and uses fiber optics and time-correlated single photon
counting in mice performing an operant task. Taking advantage of
npg © 2014 Nature America, Inc. All rights reserved.

nature neuroscience VOLUME 17 | NUMBER 8 | AUGUST 2014 1025
review
this innovative approach, the authors found that neural activity in
both direct and indirect pathway MSNs was transiently increased
when animals initiated actions. Conversely, this concomitant firing
increase in both pathways was never observed when the animals were
inactive. Thus, it could be assumed that activation of MSNs from both
pathways in one hemisphere preceded the initiation of contraversive
movements. These data on direct/indirect pathways are in apparent
contrast with recent studies showing that optical activation of indi-
rect pathway MSNs decreases locomotion25 and that either ablation
or disruption of the function of indirect pathway MSNs increases
locomotion12,45. Thus, although optogenetic techniques have permit-
ted the identification and the activation of specific cell types in vivo,
these approaches might still have some technical limitations that bias
functional interpretation of the obtained results. In particular, in vivo
optogenetic imaging might not be ideal for exploring the activity of
subcortical structures in freely moving animals, as there are limita-
tions in penetration depth.
Another important issue concerning the model of motor control
is how activity in cortical networks regulates direct and indirect
pathways. Recently, it has been found that cortical information
about motor planning and choice, conducted by intratelencephalic
and pyramidal tract neurons of the motor cortex, is directed to both
direct and indirect pathway MSNs46. This latter observation strongly
supports the possibility that the two pathways act in conjunction to
initiate movements as postulated by Cui and colleagues44.
The use of D1 and D2 BAC transgenic mice to distinguish between
direct and indirect pathway MSNs has also been recently questioned
by experimental findings showing that these animals might show
some functional alteration in comparison with wild-type mice, pos-
sibly leading to phenotypic alterations. In particular, a study using
homozygous D2 eGFP mice, and based on behavioral, electrophysio-
logical and molecular characterization, found that mice expressing
eGFP through the BAC vector are not comparable with the wild-type
littermates, as they overexpress D2 receptors47. Support for caution
in the interpretation of data resulting from the use of eGFP trans-
genic mice is also provided by a study that combined substance P
and adenosine A2A receptor immunohistochemistry (selectively
expressed in direct and indirect pathways, respectively)48,49 to iden-
tify neurons of the two pathways in both eGFP transgenic mice and
control animals. Using this technique, Bagetta and colleagues found
DA-dependent LTD in MSNs of both pathways in control mice, sup-
porting the results of the original studies showing that this form of
plasticity is expressed in both pathways. Surprisingly, D1 eGFP trans-
genic mice showed a lack of LTD in D1-expressing MSNs and showed
behavioral alterations48. These findings suggest caution in the use of
BAC mice targeting DA receptors, as genetic manipulation in these
animals might result per se in behavioral and electrophysiological
phenotypic abnormalities. Nevertheless, two other recent studies have
analyzed BAC transgenic mice, suggesting that, although it is impor-
tant to screen new transgenic mouse lines for abnormal behavior and
physiology, these BAC transgenic lines still represent useful tools for
analyzing behavior and synaptic plasticity50,51. Thus, this question
remains open.
The direct and indirect pathways are often described not only as
functionally opposing, but also as anatomically segregated. However,
a recent study reviewing single-cell tracing studies in rats found that
about one-third of MSNs projected exclusively to the GPe (pure
indirect pathway), whereas a small minority (3%) projected only to
the SNr or entopeduncular nucleus (EN) (pure direct pathway)52.
60% of labeled neurons projected to the SNr/EN and possessed col-
lateral terminal fields in the GPe. Given that these GPe collaterals
may have the ability to bridge the direct and the indirect pathways,
they have been named bridging collaterals. Regulating the extent of
bridging collaterals could be a mechanism by which the direct path-
way modulates the indirect pathway, thereby affecting the behavioral
balance maintained in concert by both pathways. The demonstration
of these bridging collaterals might also have a strong clinical impli-
cation, as they are regulated not only by endogenous DA via the
activation of D2-like DA receptors, but also by antipsychotic agents
blocking D2-like receptors, such as haloperidol, a drug widely used in
schizophrenia52. Consistent with this study, the occurrence of bridging
collaterals of striatal outputs has already been hypothesized both in
the rat53 and in the monkey54,55.
An anatomical study has reported axonal collateralization of
striatofugal cells in non-human primates showing coexpression of
D1-like and D2-like receptors, as well as of the different opioid
peptides56. In particular, the authors demonstrated that neurons
TrkB
receptor
D1-D2
heteromer
D1
receptor
NMDA
receptor
BDNF
Ras-GRF1
cAMP
PKA
PKA
Ca2+
Ca2+
PLC
IP3
P-DARPP32 PP1
DARPP32
AC
ATP
Gs
Gβγ
Gβ
Gγ
Gαq
MEK
ERK
ERK
Arc
c-Fos
Nucleus Gene
expression
CREB
PKC
CaMKII
CaMKII
DAG
GSK3
?
ER
PI3k
Akt
Targets
IP3R
Kim Caesar/Nature Publishing Group
Figure 2 Signaling pathways downstream of D1-like and D1-D2
heteromer receptor activation. Increased intracellular calcium levels,
activation of calcium/calmodulin-dependent protein kinase type II
(CaMKII) are mechanisms that are triggered by the activation of D1/D2
heteromers. Activation of the G-coupled D1-D2 heteromer induces a
phospholipase C (PLC)-dependent calcium release, resulting in the
activation of CaMKII and its translocation to the nucleus. CaMKII then
induces cAMP responsive element binding protein (CREB) phosphorylation
and gene expression. In addition, dopamine-induced D1-D2 heteromer
activation can phosphorylate and inactivate glycogen synthase kinase-3
(GSK3). The phosphorylation state of GSK3 can be also regulated by
BDNF-induced activation of TrkB receptor. Dopamine D1-like receptor can
induce two different signaling pathways. Dopamine D1 receptor directly
modulates NMDA receptor activation by Gβγ proteins. Moreover, by
acting on cAMP level increase, the D1 receptor activation induces the
activation of protein kinase A (PKA), which can translocate to the
nucleus and act directly on CREB. PKA phosphorylates DARPP-32,
inducing the disinhibition of the NMDA/Ras-GRF1/ERK pathway, which
finally targets intranucleous CREB protein. All these downstream
pathways result in the transcriptional activation of several genes. AC,
adenylyl cyclase; Arc, activity-regulated cytoskeleton-associated protein;
DAG, diacylglycerol; ER, endoplasmic reticulum; ERK, extracellular
signal-regulated kinases; IP3, inositol 1,4,5-trisphosphate; MEK,
mitogen-activated protein kinase; PKC, protein kinase C; P-DARPP-32,
phosphorylated dopamine- and cAMP-regulated phosphoprotein 32 kDa;
PP1, protein phosphatase 1; PI3k, phosphatidylinositide 3-kinases;
Ras-GRF1, Ras-guanine nucleotide-releasing factor 1.
npg © 2014 Nature America, Inc. All rights reserved.

1026 VOLUME 17 | NUMBER 8 | AUGUST 2014 nature neuroscience
review
projecting to the GPi show immunolabeling for D1 and D2 receptors,
rather than just D1 receptors, as classically viewed56. Similarly, striatal
neurons projecting to the GPe show immunolabeling for both D1 and
D2 receptors, rather than just D2 receptors56.
Heteromers of dopamine receptors as a possible molecular
cross-talk between direct and indirect pathways
In recent years, multiple levels of cross-talk between direct and
indirect pathways have been revealed. Accordingly, a first level of
interaction is represented by the molecular cross-talk between hetero-
meric D1-like and D2-like DA receptors57,58. Moreover, interactions
between these pathways, represented by retrograde messengers and
nitric oxide (NO), mediate a biochemical cross-talk, whereas the
synaptic cross-talk is exerted by distinct classes of striatal interneu-
rons. These multiple levels of interactions alter the rigid rule of the
separation between the two systems.
Increased intracellular Ca2+ levels, activation of CaMKII and
release of brain-derived neurotrophic factor (BDNF) are mechanisms
triggered by the activation of D1/D2 heteromers (Fig. 2)57,58. These
mechanisms are also required for striatal physiological and activity-
dependent forms of synaptic plasticity59–61, as well as for those
obser ved following the onset of L-DOPA–induced dyskinesia in
models of PD62 and chronic drug abuse63. However, the possible
involvement of D1/D2 heteromers in these events is still unclear, as
these heteromers occur more frequently in the ventral than in the
dorsal striatum57,58. Future studies are necessary to convincingly
demonstrate the function of heteromers in native expression systems
and their distinct signal transduction coupling.
Endocannabinoid system as a biochemical cross-talk between
direct and indirect pathways
Endogenous molecules such as endocannabinoids (eCBs) and NO
modulate the activity of MSNs by non-canonical modes, as well as that
of their afferent and efferent connections, and represent an additional
biochemical substrate for the cross-talk between direct and indirect
pathways. Retrograde signaling is the principal mode by which eCBs
mediate short- and long-term forms of plasticity both at excitatory
and inhibitory synapses and interacts with dopaminergic system64–66.
The role of eCBs in the control of LTD of MSNs is another major
issue of discussion of great relevance to the direct/indirect pathway
model. Lovinger’s group showed for the first time that the induction
of striatal LTD is dependent on activation of the CB1 cannabinoid
receptor. In fact, LTD was facilitated by blocking cellular eCB uptake.
The endocannabinoid necessary for striatal LTD is thus likely to be
released postsynaptically as a retrograde messenger demonstrating
a new role for eCBs in the induction of LTD in a circuit necessary
for habit formation and motor control65,67. Although this effect was
observed in most of the neurons, suggesting a lack of segregation in
a specific pathway, the inhibition of glutamate release by retrograde
endocannabinoid signaling was frequency dependent and D2 receptor
mediated68. Notably, postsynaptic blockade of eCB membrane trans-
port altered eCB release and LTD in the large majority of MSNs69.
In a seminal study, Wang and colleagues70 investigated how the
induction of striatal LTD, which in the original studies20–23 was
observed in most MSNs, could depend on D2 dopamine receptors
localized only in the postsynaptic membrane of a single subclass
of MSNs. In fact, if this was true, LTD should be inducible in neu-
rons from only one of the two projection systems of the striatum, as
reported in the study by Kreitzer and Malenka16. Using transgenic
mice in which neurons that contribute to these two systems were
identified, Wang and colleagues demonstrated that this was not the
case. They raised the idea that cholinergic interneurons, also defined
as tonically active interneurons, are critically involved in this D2
dependence of LTD (Fig. 3). In fact, activation of D2 receptors
induces pauses in the activity of these interneurons, reducing the
release of acetylcholine (ACh) and relieving the inhibitory cholin-
ergic tone on MSNs expressing M1 muscarinic receptors70. Given
that activation of M1 receptors suppresses L-type Ca2+ currents, the
reduced cholinergic tone disinhibits MSNs, promoting production of
eCBs and LTD. Consistent with this hypothesis, application of the M1
antagonist pirenzepine reduces baseline corticostriatal glutamatergic
transmission70,71. Moreover, this effect, induced by the activation of M1
receptors, is blocked by a CB1 receptor antagonist70. These observations
suggest that, although the eCBs system is critical for the biochemical
cross-talk between direct and indirect pathways, the choliner-
gic interneuron represents the cellular substrate for the synaptic
cross-talk between the two classes of MSNs (Fig. 3). A more recent
study, using immunohistochemical characterization of substance P–
positive (direct pathway) and A2A receptor–positive (indirect
pathway) MSNs, confirmed that D2-dependent LTD is present in both
classes of MSNs48.
Profound modifications in eCB signaling after DA depletion
occur in experimental models of PD and in patients suffering from
the disease72,73. In a PD model, striatal levels of anandamide (AEA),
an endogenous cannabinoid neurotransmitter, are increased74.
This molecular change is associated with increased spontaneous
Ca2+
Ca2+ Ca2+
PLC
PLC
Ca2+
store
Ca2+
store
Ca2+
Cortical/thalamic inputs
Glutamatergic
terminals
Cholinergic
interneuron
MSN
Glu
Glu
D2R
DA
DA
D1R
D1R
D2R
A2AR
eCBs
eCBs
ACh
ACh
DAG DAG
Cav1.3
M1R
M1R
CB1R
CB-1RNMDAR
A2AR
Dopaminergic
terminals
NMDAR
Kim Caesar/Nature Publishing Group
Figure 3 Role of MSNs and cholinergic interneurons in the production
and functions of endocannabinoids in the striatum. Striatal cholinergic
interneurons project to both MSNs expressing D1-like and MSNs expressing
D2-like dopamine receptors. The combined activation of both A2A and
D2 receptors on cholinergic interneurons decreases the release of ACh.
The decreased levels of ACh on the M1 muscarinic receptors located on
the synaptic sites of D1 and D2 receptor–expressing MSNs reverses the
blockade of the L-type calcium channels. The increase in intracellular
calcium concentration might in turn trigger endocannabinoid release at the
postsynaptic sites of both D1 and D2 receptor–expressing MSNs, thereby
depressing glutamatergic synaptic transmission of both the direct and
indirect pathways. The NMDA receptors contribute in both D1- and
D2-expressing MSNs to the intracellular calcium increase and to the
resulting endocannabinoid release. A2AR, adenosine 2A receptor; Cav1.3,
L-type calcium channel; CB1R, endocannabinoid receptor.
npg © 2014 Nature America, Inc. All rights reserved.

nature neuroscience VOLUME 17 | NUMBER 8 | AUGUST 2014 1027
review
glutamatergic activity recorded from the large majority of MSNs and
is reversed by L-DOPA treatment, making a clear segregation in the
parkinsonian state also unlikely74,75.
eCB-dependent synaptic plasticity of MSNs could represent a syn-
aptic mechanism for the formation of persistent drug-related habits.
In particular, the dorsal striatum might be implicated in the shift from
casual drug use to compulsive drug use and addiction76. Consistent
with this view, it has been observed in a mouse model of cannabinoid
tolerance that persistent activation of the eCB pathway impairs LTD
in MSNs77.
NOS-positive and cholinergic interneurons: a synaptic cross-talk
between direct and indirect pathways
NO was identified as a biological intercellular messenger more than
20 years ago and has been implicated in synaptic transmission and
plasticity, as well as in neurodegeneration78. In the striatum, NO is
produced by a subclass of GABAergic interneurons. In fact, in this
structure, three neurochemically distinct subtypes of GABAergic
interneurons have been distinguished: fast-spiking interneurons
expressing the calcium-binding protein par valbumin, interneurons
expressing the calcium binding protein calretinin, and a third class
of interneurons showing low-threshold spikes and coexpressing neu-
ropeptide Y, somatostatin and nitric oxide synthase (NOS) (Fig. 4)7.
Striatal NO-producing interneurons are important for the regulation
of corticostriatal synaptic transmission and motor behavior (Fig. 3).
Moreover, these interneurons can also show distinct forms of synaptic
plasticity in response to different patterns of stimulation79. Striatal
NO synthesis is stimulated by concomitant activation of glutamate
and D1-like DA receptors, and this gas diffuses into the dendrites of
MSNs containing high levels of NO receptors called soluble guanylyl
cyclases (sGC). NO-mediated activation of sGC leads to the synthesis
of the second messenger cGMP80. Electrophysiological experiments
have shown that NOS inhibitors prevent LTD induction81,82. One
prominent molecular target of NO is the striatally enriched sGC, sug-
gesting the possibility that activation of this enzyme and subsequent
cGMP formation is sufficient to induce LTD. Accordingly, it has been
shown that the cGMP phosphodiesterase inhibitor zaprinast and the
intracellular application of cGMP itself can induce LTD during low-
frequency synaptic activation81.
The role of the NO/cGMP pathway in corticostriatal LTD induc-
tion has also been investigated in a rat model of parkinsonism and
L-DOPA–induced dyskinesia to test the possibility of targeting stri-
atal phosphodiesterases to reduce involuntary movements caused by
chronic treatment with this drug82. L-DOPA–induced dyskinesia was
associated with the loss of LTD expression at glutamatergic striatal
synapses onto both classes of MSNs. Inhibitors of phosphodieste-
rases rescued the induction of this form of synaptic plasticity via a
mechanism requiring the modulation of intracellular cGMP levels.
Cortical/thalamic inputs
LTP LTD
D2R
DA
M1R
GABA GABA
NO NO
NO
D1R
D1-D2R
Cholinergic
interneuron
NOS
interneuron
Fast spiking
interneuron
MSN
Glu
D1/D5R
D1/D5R
ACh
M2/4R
CB1R
Dopaminergic
terminals
Physiological conditiona
LTP LTD
Locomotor
activation
Locomotor
inhibition
Output nuclei
Cortical/thalamic inputs
No
LTP
No
LTD
D2R
M1R
GABA GABA
NO NO
NO
D1R
D1-D2R
Cholinergic
interneuron
NOS
interneuron
Fast spiking
interneuron
MSN
Glu
D1/D5R
D1/D5R
ACh
M2/4R
CB1R
Dopaminergic
terminals
M2/4R
Parkinson’s disease
b
No change in
motor state
No change in
motor state
Output nuclei
No
LTP
No
LTD
Glu DA
D1R
D2R
A2AR
eCBs
ACh
M1R CB-1RNMDAR
Kim Caesar/Nature Publishing Group
Figure 4 Integrative hypothesis for the role of striatal circuits in controlling motor activity in the physiological condition and in Parkinson’s disease.
(a) Glutamatergic inputs originating from both the cortex and the thalamus release glutamate onto striatal neurons. Dopaminergic terminals, originating from
the substantia nigra pars compacta, release dopamine onto MSNs and different subtypes of striatal interneurons. In particular, three main subtypes of striatal
interneurons are implicated in the feedforward and parallel control of striatal circuits. Cholinergic interneurons release ACh acting on both presynaptic
glutamatergic terminals and postsynaptic MSNs; these interneurons also respond to dopamine via D1/D5 and D2 receptors. NOS-positive interneurons
produce NO, acting as a retrograde messenger as well as on MSNs facilitating LTD at the postsynaptic level. Fast-spiking interneurons release GABA on
MSNs, providing a parallel inhibitory system that controls both direct and indirect pathway MSNs. MSNs can express either D1-like or D2-like receptors, as
well as D1-D2 heteromeric receptors. eCBs released from MSNs can act as retrograde messengers on CB1 cannabinoid receptors located on glutamatergic
terminals. The induction of either LTP or LTD in MSNs regulates the striatal control on output structures and motor activation/inhibition. (b) The advanced
phase of Parkinson’s disease is caused by a severe dopamine denervation that leads to the complete loss of striatal synaptic plasticity. Under this condition,
both LTD and LTP of MSNs are lost. As a consequence of the loss of these forms of plasticity, variations of output signals from the striatum are absent and no
change in the motor state can be induced. Dopamine denervation also alters the physiological activity of striatal interneurons as well as the neurochemical
signals that originate from these cells and influence the activity of MSNs. A2AR, adenosine 2A receptor; CB1R, endocannabinoid receptor; DA, dopamine;
D1R and D2R, dopaminergic receptors; M1R and M2/4R, muscarinic receptors.
npg © 2014 Nature America, Inc. All rights reserved.

1028 VOLUME 17 | NUMBER 8 | AUGUST 2014 nature neuroscience
review
This effect on synaptic plasticity was associated with a significant
reduction of abnormal dyskinesias following intrastriatal injection
of phosphodiesterase inhibitors82. Thus, drugs selectively targeting
phosphodiesterases can ameliorate L-DOPA–induced dyskinesia,
possibly by restoring physiological synaptic plasticity in MSNs of both
direct and indirect pathways.
NO is also implicated in the pathophysiology of brain ischemia,
and endogenous DA, via the activation of D1/D5 receptors
expressed in striatal NOS-positive interneurons, seems to amplify
this event. The D1-like receptor antagonist SCH-23390 prevented
post-ischemic long-term potentiation (LTP) in all recorded MSNs.
Immunofluorescence analysis confirmed the induction of post-
ischemic LTP in both substance P–positive, (putative D1 receptor
expressing) and adenosine A2A receptor–positive (putative D2
receptor expressing) MSNs83. Thus, in conjunction with the synaptic
cross-talk between the direct and indirect pathways, represented by
striatal GABAergic and cholinergic interneurons7,14, endogenous
striatal eCBs and NO constitute two systems that influence both path-
ways in parallel (Fig. 3).
Clinical observations beyond the model
The hypothesis that GPi and STN neurons are hyperactive in the
parkinsonian state is supported by studies showing a reduction of
parkinsonian symptoms following lesions of the GPi or STN in mon-
keys and patients with PD84–86. However, another study found that,
in PD patients, apomorphine, a nonselective D1- and D2-dopamine
receptor agonist, significantly decreased the firing rates of GPi neu-
rons at doses sufficient to produce an ON state, but did not change
the overall firing rate of STN neurons87. This latter finding suggests
that the apomorphine-induced reduction of parkinsonian symptoms
is not solely the result of a decrease in overall activity in the GPi or
STN neurons, as predicted by the direct/indirect model of BG, but
requires alternative interpretations.
There is also a simple observation arising from clinical experience
with PD patients that casts doubt on a strict interpretation of the
direct/indirect pathway model. In the last few decades, a great effort
has been undertaken to find a better treatment for PD than L-DOPA.
This drug, however, remains the gold standard in the therapy of this
neurodegenerative disease88. In fact, although long-term treatment
with L-DOPA induces dyskinetic movements62,89, it can be considered
to be the most effective option in almost all phases of the disease.
How does L-DOPA work? This drug is a precursor of endogenous DA
that activates both D1-like and D2-like receptors. Although multiple
mechanisms have been linked to this drug, only this dual pharmaco-
logical effect seems to provide the potent motor activation generated
by L-DOPA in PD patients90. Selective D2/D3 receptor agonists are
currently available and offer interesting therapeutic options91, espe-
cially in the early phase of the disease. However, none of them are
able to generate a therapeutic response similar to that achieved with
L-DOPA. In fact, although it has been recently shown that D2 receptor
activation can reduce motor disability in rodents, thereby reducing
the risk of dyskinesia, high doses of these agonists might also inter-
act with D1 receptors in producing both therapeutic and dyskinetic
actions through heteromeric receptors9 2. More importantly, clinical
therapeutic strategies, selectively activating the D1 receptors in PD,
are neither feasible nor testable at present. Conversely, we also have
to assume that it would be difficult, if not impossible, to observe the
therapeutic effects of L-DOPA in PD patients in the absence of D2-like
receptor stimulation. Accordingly, motor disabilities in PD patients
treated with L-DOPA are markedly worsened by the use of classical
neuroleptic agents (mainly antagonizing D2-like receptors) to control
behavioral alterations93. Thus, we might conclude that the therapeutic
efficacy of L-DOPA results from the activation of both D1-like and
D2-like striatal receptors. Accordingly, it has been recently shown that
experimental parkinsonism induced by MPTP leads to a decrease in
dendritic spine density in both D1 and D2 receptor–containing MSNs
and that intensive exercise leads to increased dendritic spine density
and arborization in MSNs of both the pathways94.
Oscillatory activity in PD: a link with direct/indirect pathways?
DA levels can rapidly modulate the synchronicity and oscillatory
behavior of cortical and striatal circuits95. Moreover, electrophysio-
logical studies in rodent and primate animal models of PD and in PD
patients have discovered abnormally synchronized oscillatory activity
at multiple levels of the basal ganglia–cortical loop96,97. This patho-
logical synchronization correlates with akinesia and is suppressed by
either dopaminergic therapies or DBS. In a rodent model of PD, it
has been shown that striatal NMDA receptors gate cortico-pallidal
synchronization98 , suggesting an interesting similarity with the
NMDA-dependent forms of corticostriatal plasticity. Although the
specific contribution of direct and indirect pathways to this oscillatory
behavior remains to be established, one could speculate that the two
pathways are abnormally coordinated in PD.
Conclusion
Although some recent findings raise possible doubts concerning a too
rigid application of the direct/indirect pathway model, at present, the
literature does not provide compelling evidence against it. However, we
feel that this model, which tries to explain the entire complex activity
of BG function in both physiological and pathological conditions,
needs to be revised to integrate more recent scientific findings.
At this stage, it is possible to postulate a representation of the BG that,
although not depicting all circuits, provides an interpretation of the
‘filtering’ function of the striatum in the BG activity as an integrative
system of cortical glutamatergic and nigral dopaminergic inputs
(Fig. 4). In this scenario, the direct and indirect pathways should not
be seen as separate, parallel systems, as hypothesized in the classical
interpretation of the model. On the contrary, as we describe here, the
two pathways are structurally and functionally intertwined at least
at two distinct levels: in the striatum, where the direct and indirect
pathways communicate via the complex interneuronal network and
the biochemical links between the two MSN subtypes, and outside
of the striatum, where GPe collaterals may bridge the two pathways,
potentially allowing the direct pathway to modulate the indirect
pathway52. The existence of these bridging collaterals provides further
support for the evidence obtained in non-human primates that striatal
neurons projecting to either the GPi or the GPe show immunolabeling
for both D1 and D2 DA receptors56.
The interaction between the two pathways would therefore be
dynamic in both physiological and pathological conditions, with DA
and dopaminergic agents not being able to control a single pathway
in isolation without influencing the other. In this context, the precise
direction of striatal synaptic plastic changes would be driven by both
the intensity of cortical/thalamic glutamatergic activation and the
amount and the precise timing of DA release59.
The coordinated action of the direct and indirect pathways during
action initiation44 and the concomitant activation of SNc make a circuit
to signal start and stop of action sequences99,100 (Fig. 4). Conversely, it is
possible that, in PD, the alterations of the main forms of DA-dependent
synaptic plasticity in MSNs could make the striatum unable to filter neu-
ronal signals. Thus, the lack of coordinated activity between these two
pathways will result in the inability of the PD patient to physiologically
npg © 2014 Nature America, Inc. All rights reserved.

nature neuroscience VOLUME 17 | NUMBER 8 | AUGUST 2014 1029
review
start and stop action sequences and in a global slowing of motor func-
tion, causing bradykinesia, freezing and gait festination (Fig. 4).
This interpretation integrates the classical direct/indirect hypoth-
esis, as it considers the importance of striatal interneurons in striatal
physiology and suggests that all MSNs might either facilitate or inhibit
movement depending on the form of synaptic plasticity expressed in a
certain moment. Future experimental and clinical studies will provide
answers to the outstanding questions on the complex functions of the
direct/indirect pathways (Box 1).
ACKNOWLEDGMENTS
We thank A. Pisani for reading the manuscript and for critical discussion.
This work was supported by Progetto di Ricerca di Interesse Nazionale (PRIN)
2011 2010AHHP5H (to P.C.) and Progetto del Ministero della Salute, Giovani
Ricercatori (GR-2008-1142336 to B.P.; GR-2010-2316671 to V.G.).
COMPETING FINANCIAL INTERESTS
The authors declare competing financial interests: details are available in the online
version of the paper.
Reprints and permissions information is available online at http://www.nature.com/
reprints/index.html.
1. Seneca, L.A. Moral letters to Lucilius (Epistulae morales ad Lucilium): a Loeb
Classical Library edition, Vol. 1 (ed. Gummere, R.M.) (W. Heinemann, 1917).
2. DiFiglia, M., Pasik, P. & Pasik, T. A Golgi study of neuronal types in the neostriatum
of monkeys. Brain Res. 114, 245–256 (1976).
3. Dubé, L., Smith, A.D. & Bolam, J.P. Identification of synaptic terminals of thalamic
or cortical origin in contact with distinct medium-size spiny neurons in the rat
neostriatum. J. Comp. Neurol. 267, 455–471 (1988).
4. Lapper, S.R. & Bolam, J.P. Input from the frontal cortex and the parafascicular
nucleus to cholinergic interneurons in the dorsal striatum of the rat. Neuroscience
51, 533–545 (1992).
5. Pickel, V.M., Chan, J. & Sesack, S.R. Cellular basis for interactions between
catecholaminergic afferents and neurons containing Leu-enkephalin-like
immunoreactivity in rat caudate-putamen nuclei. J. Neurosci. Res. 31, 212–230
(1992).
6. Bouyer, J.J., Park, D.H., Joh, T.H. & Pickel, V.M. Chemical and structural analysis
of the relation between cortical inputs and tyrosine hydroxylase-containing
terminals in rat neostriatum. Brain Res. 302, 267–275 (1984).
7. Kawaguchi, Y., Wilson, C.J., Augood, S.J. & Emson, P.C. Striatal interneurones:
chemical, physiological and morphological characterization. Trends Neurosci. 18,
527–535 (1995).
8. Albin, R.L., Young, A.B. & Penney, J.B. The functional anatomy of basal ganglia
disorders. Trends Neurosci. 12, 366–375 (1989).
9. DeLong, M.R. Primate models of movement disorders of basal ganglia origin.
Trends Neurosci. 13, 281–285 (1990).
10. Gerfen, C.R. et al. D1 and D2 dopamine receptor–regulated gene expression of
striatonigral and striatopallidal neurons. Science 250, 1429–1432 (1990).
11. Gerfen, C.R. & Surmeier, D.J. Modulation of striatal projection systems by
dopamine. Annu. Rev. Neurosci. 34, 441–466 (2011).
12. Bateup, H.S. et al. Distinct subclasses of medium spiny neurons differentially regulate
striatal motor behaviors. Proc. Natl. Acad. Sci. USA 107, 14845–14850 (2010).
13. Valjent, E., Bertran-Gonzalez, J., Herve, D., Fisone, G. & Girault, J.A.
Looking BAC at striatal signaling: cell-specific analysis in new transgenic mice.
Trends Neurosci. 32, 538–547 (2009).
14. Kawaguchi, Y., Wilson, C.J. & Emson, P.C. Intracellular recording of identified
neostriatal patch and matrix spiny cells in a slice preparation preserving cortical
inputs. J. Neurophysiol. 62, 1052–1068 (1989).
15. Cepeda, C. et al. Differential electrophysiological properties of dopamine D1 and
D2 receptor–containing striatal medium-sized spiny neurons. Eur. J. Neurosci.
27, 671–682 (2008).
16. Kreitzer, A.C. & Malenka, R.C. Endocannabinoid-mediated rescue of striatal LTD
and motor deficits in Parkinson’s disease models. Nature 445, 643–647 (2007).
17. Day, M., Wokosin, D., Plotkin, J.L., Tian, X. & Surmeier, D.J. Differential excitability
and modulation of striatal medium spiny neuron dendrites. J. Neurosci. 28,
11603–11614 (2008).
18. Gertler, T.S., Chan, C.S. & Surmeier, D.J. Dichotomous anatomical properties of
adult striatal medium spiny neurons. J. Neurosci. 28, 10814–10824 (2008).
19. Day, M. et al. Selective elimination of glutamatergic synapses on striatopallidal
neurons in Parkinson disease models. Nat. Neurosci. 9, 251–259 (2006).
20. Calabresi, P. et al. Dopamine and cAMP-regulated phosphoprotein 32 kDa controls
both striatal long-term depression and long-term potentiation, opposing forms of
synaptic plasticity. J. Neurosci. 20, 8443–8451 (2000).
21. Calabresi, P., Maj, R., Mercuri, N.B. & Bernardi, G. Coactivation of D1 and D2
dopamine receptors is required for long-term synaptic depression in the striatum.
Neurosci. Lett. 142, 95–99 (1992).
22. Calabresi, P., Maj, R., Pisani, A., Mercuri, N.B. & Bernardi, G. Long-term synaptic
depression in the striatum: physiological and pharmacological characterization.
J. Neurosci. 12, 4224–4233 (1992).
23. Lovinger, D.M., Tyler, E.C. & Merritt, A. Short- and long-term synaptic depression
in rat neostriatum. J. Neurophysiol. 70, 1937–1949 (1993).
24. Shen, W., Flajolet, M., Greengard, P. & Surmeier, D.J. Dichotomous dopaminergic
control of striatal synaptic plasticity. Science 321, 848–851 (2008).
25. Kravitz, A.V. et al. Regulation of parkinsonian motor behaviours by optogenetic
control of basal ganglia circuitry. Nature 466, 622–626 (2010).
26. Hikosaka, O., Nakamura, K. & Nakahara, H. Basal ganglia orient eyes to reward.
J. Neurophysiol. 95, 567–584 (2006).
27. Redgrave, P. et al. Goal-directed and habitual control in the basal ganglia:
implications for Parkinson’s disease. Nat. Rev. Neurosci. 11, 760–772
(2010).
28. Tai, L.H., Lee, A.M., Benavidez, N., Bonci, A. & Wilbrecht, L. Transient stimulation
of distinct subpopulations of striatal neurons mimics changes in action value.
Nat. Neurosci. 15, 1281–1289 (2012).
29. Eshel, N. & Roiser, J.P. Reward and punishment processing in depression.
Biol. Psychiatry 68, 118–124 (2010).
30. Koob, G.F. & Volkow, N.D. Neurocircuitry of addiction. Neuropsychopharmacology
35, 217–238 (2010).
31. Kravitz, A.V., Tye, L.D. & Kreitzer, A.C. Distinct roles for direct and indirect pathway
striatal neurons in reinforcement. Nat. Neurosci. 15, 816–818 (2012).
32. Hikida, T., Kimura, K., Wada, N., Funabiki, K. & Nakanishi, S. Distinct roles of
synaptic transmission in direct and indirect striatal pathways to reward and
aversive behavior. Neuron 66, 896–907 (2010).
33. Ferguson, S.M. et al. Transient neuronal inhibition reveals opposing roles of indirect
and direct pathways in sensitization. Nat. Neurosci. 14, 22–24 (2011).
34. Voon, V. et al. Chronic dopaminergic stimulation in Parkinson’s disease: from
dyskinesias to impulse control disorders. Lancet Neurol. 8, 1140–1149 (2009).
35. Vitek, J.L. et al. Randomized trial of pallidotomy versus medical therapy for
Parkinson’s disease. Ann. Neurol. 53, 558–569 (2003).
36. Alvarez, L. et al. Bilateral subthalamotomy in Parkinson’s disease: initial and
long-term response. Brain 128, 570–583 (2005).
37. Fasano, A., Daniele, A. & Albanese, A. Treatment of motor and non-motor features
of Parkinson’s disease with deep brain stimulation. Lancet Neurol. 11, 429–442
(2012).
38. Odekerken, V.J. et al. Subthalamic nucleus versus globus pallidus bilateral deep
brain stimulation for advanced Parkinson’s disease (NSTAPS study): a randomized
controlled trial. Lancet Neurol. 12, 37–44 (2013).
Box 1 Outstanding questions
• Do specific striatal MSNs express in vivo a single form of synaptic plasticity or can these neurons undergo different, and even opposite, forms of
plasticity depending on their functional and metabolic state (membrane potential, energetic condition, endogenous striatal levels of DA)?
• Can striatal interneurons modulate the activity of MSN and, in turn, facilitate or inhibit motor activity depending on the form of synaptic plasticity
expressed in a specific functional state?
• What are the specific physiological effects resulting from the activation of D1/D2 heteromers as well as of other DA heteromers?
• How do distinct levels of DA denervation, as observed during the evolution of the natural history of PD, differentially affect plasticity in direct/
indirect pathway MSNs as well as in various subtypes of striatal interneurons?
• How is the activity of direct/indirect pathway MSNs altered by the pathological processes implicated in the different phases of nigral neurodegenera-
tion in PD?
• Most animal investigations dealing with potential therapeutic interventions in the direct/indirect pathways have not been performed as blinded
studies. Can similar results be obtained using a blind protocol as in many clinical trials?
• Will a selective activation of MSNs of either direct or indirect pathways in isolation and in conjunction be possible in the near future in humans?
• How would these specific activations affect the symptoms and the natural history of disabling neurodegenerative diseases of the BG such as PD and HD?
npg © 2014 Nature America, Inc. All rights reserved.

1030 VOLUME 17 | NUMBER 8 | AUGUST 2014 nature neuroscience
review
39. Schuepbach, W.M. et al. Neurostimulation for Parkinson’s disease with early motor
complications. N. Engl. J. Med. 368, 610–622 (2013).
40. Krack, P., Hariz, M.I., Baunez, C., Guridi, J. & Obeso, J.A. Deep brain stimulation:
from neurology to psychiatry? Trends Neurosci. 33, 474–484 (2010).
41. Vonsattel, J.P. et al. Neuropathological classification of Huntington’s disease.
J. Neuropathol. Exp. Neurol. 44, 559–577 (1985).
42. André, V.M. et al. Differential electrophysiological changes in striatal output
neurons in Huntington’s disease. J. Neurosci. 31, 1170–1182 (2011).
43. Freeze, B.S., Kravitz, A.V., Hammack, N., Berke, J.D. & Kreitzer, A.C. Control of
basal ganglia output by direct and indirect pathway projection neurons.
J. Neurosci. 33, 18531–18539 (2013).
44. Cui, G. et al. Concurrent activation of striatal direct and indirect pathways during
action initiation. Nature 494, 238–242 (2013).
45. Durieux, P.F. et al. D2R striatopallidal neurons inhibit both locomotor and drug
reward processes. Nat. Neurosci. 12, 393–395 (2009).
46. Kress, G.J. et al. Convergent cortical innervation of striatal projection neurons.
Nat. Neurosci. 16, 665–667 (2013).
47. Kramer, P.F. et al. Dopamine D2 receptor overexpression alters behavior and
physiology in Drd2-EGFP mice. J. Neurosci. 31, 126–132 (2011).
48. Bagetta, V. et al. Dopamine-dependent long-term depression is expressed in
striatal spiny neurons of both direct and indirect pathways: implications for
Parkinson’s disease. J. Neurosci. 31, 12513–12522 (2011).
49. Deng, Y.P., Lei, W.L. & Reiner, A. Differential perikaryal localization in rats of D1
and D2 dopamine receptors on striatal projection neuron types identified by
retrograde labeling. J. Chem. Neuroanat. 32, 101–116 (2006).
50. Chan, C.S. et al. Strain-specific regulation of striatal phenotype in Drd2-eGFP
BAC transgenic mice. J. Neurosci. 32, 9124–9132 (2012).
51. Nelson, A.B. et al. A comparison of striatal-dependent behaviors in wild-type and
hemizygous Drd1a and Drd2 BAC transgenic mice. J. Neurosci. 32, 9119–9123
(2012).
52. Cazorla, M. et al. Dopamine D2 receptors regulate the anatomical and functional
balance of basal ganglia circuitry. Neuron 81, 153–164 (2014).
53. Kawaguchi, Y., Wilson, C.J. & Emson, P.C. Projection subtypes of rat neostriatal
matrix cells revealed by intracellular injection of biocytin. J. Neurosci. 10,
3421–3438 (1990).
54. Lévesque, M. & Parent, A. The striatofugal fiber system in primates: a reevaluation
of its organization based on single-axon tracing studies. Proc. Natl. Acad. Sci.
USA 102, 11888–11893 (2005).
55. Parent, A., Charara, A. & Pinault, D. Single striatofugal axons arborizing in both
pallidal segments and in the substantia nigra in primates. Brain Res. 698,
280–284 (1995).
56. Nadjar, A. et al. Phenotype of striatofugal medium spiny neurons in parkinsonian
and dyskinetic nonhuman primates: a call for a reappraisal of the functional
organization of the basal ganglia. J. Neurosci. 26, 8653–8661 (2006).
57. Hasbi, A. et al. Calcium signaling cascade links dopamine D1–D2 receptor
heteromer to striatal BDNF production and neuronal growth. Proc. Natl. Acad.
Sci. USA 106, 21377–21382 (2009).
58. Perreault, M.L., Hasbi, A., O’Dowd, B.F. & George, S.R. Heteromeric dopamine
receptor signaling complexes: emerging neurobiology and disease relevance.
Neuropsychopharmacology 39, 156–168 (2013).
59. Calabresi, P., Picconi, B., Tozzi, A. & Di Filippo, M. Dopamine-mediated regulation
of corticostriatal synaptic plasticity. Trends Neurosci. 30, 211–219 (2007).
60. Jia, Y., Gall, C.M. & Lynch, G. Presynaptic BDNF promotes postsynaptic long-term
potentiation in the dorsal striatum. J. Neurosci. 30, 14440–14445 (2010).
61. Lovinger, D.M. Neurotransmitter roles in synaptic modulation, plasticity and
learning in the dorsal striatum. Neuropharmacology 58, 951–961 (2010).
62. Calabresi, P., Di Filippo, M., Ghiglieri, V., Tambasco, N. & Picconi, B. Levodopa-
induced dyskinesias in patients with Parkinson’s disease: filling the bench-to-
bedside gap. Lancet Neurol. 9, 1106–1117 (2010).
63. Russo, S.J., Mazei-Robison, M.S., Ables, J.L. & Nestler, E.J. Neurotrophic factors and
structural plasticity in addiction. Neuropharmacology 56 (suppl. 1) 73–82 (2009).
64. Adermark, L., Talani, G. & Lovinger, D.M. Endocannabinoid-dependent plasticity
at GABAergic and glutamatergic synapses in the striatum is regulated by synaptic
activity. Eur. J. Neurosci. 29, 32–41 (2009).
65. Gerdeman, G.L., Ronesi, J. & Lovinger, D.M. Postsynaptic endocannabinoid
release is critical to long-term depression in the striatum. Nat. Neurosci. 5,
446–451 (2002).
66. Mathur, B.N. & Lovinger, D.M. Endocannabinoid-dopamine interactions in striatal
synaptic plasticity. Front Pharmacol 3, 66 (2012).
67. Adermark, L. & Lovinger, D.M. Retrograde endocannabinoid signaling at striatal
synapses requires a regulated postsynaptic release step. Proc. Natl. Acad. Sci.
USA 104, 20564–20569 (2007).
68. Yin, H.H. & Lovinger, D.M. Frequency-specific and D2 receptor–mediated
inhibition of glutamate release by retrograde endocannabinoid signaling.
Proc. Natl. Acad. Sci. USA 103, 8251–8256 (2006).
69. Ronesi, J., Gerdeman, G.L. & Lovinger, D.M. Disruption of endocannabinoid
release and striatal long-term depression by postsynaptic blockade of
endocannabinoid membrane transport. J. Neurosci. 24, 1673–1679 (2004).
70. Wang, Z. et al. Dopaminergic control of corticostriatal long-term synaptic
depression in medium spiny neurons is mediated by cholinergic interneurons.
Neuron 50, 443–452 (2006).
71. Tozzi, A. et al. The distinct role of medium spiny neurons and cholinergic
interneurons in the D2/A2A receptor interaction in the striatum: implications for
Parkinson’s disease. J. Neurosci. 31, 1850–1862 (2011).
72. Di Filippo, M. et al. The endocannabinoid system in Parkinson’s disease.
Curr. Pharm. Des. 14, 2337–2347 (2008).
73. Pisani, V. et al. Homeostatic changes of the endocannabinoid system in Parkinson’s
disease. Mov. Disord. 26, 216–222 (2011).
74. Gubellini, P. et al. Experimental parkinsonism alters endocannabinoid degradation:
implications for striatal glutamatergic transmission. J. Neurosci. 22, 6900–6907
(2002).
75. Maccarrone, M. et al. Levodopa treatment reverses endocannabinoid system
abnormalities in experimental parkinsonism. J. Neurochem. 85, 1018–1025 (2003).
76. Gerdeman, G.L., Partridge, J.G., Lupica, C.R. & Lovinger, D.M. It could be habit
forming: drugs of abuse and striatal synaptic plasticity. Trends Neurosci. 26,
184–192 (2003).
77. Nazzaro, C. et al. SK channel modulation rescues striatal plasticity and control
over habit in cannabinoid tolerance. Nat. Neurosci. 15, 284–293 (2012).
78. Vincent, S.R. Nitric oxide neurons and neurotransmission. Prog. Neurobiol. 90,
246–255 (2010).
79. Fino, E. & Venance, L. Spike-timing dependent plasticity in striatal interneurons.
Neuropharmacology 60, 780–788 (2011).
80. West, A.R. & Tseng, K.Y. Nitric oxide–soluble guanylyl cyclase–cyclic GMP
signaling in the striatum: new targets for the treatment of Parkinson’s disease?
Front. Syst. Neurosci. 5, 55 (2011).
81. Calabresi, P. et al. A critical role of the nitric oxide/cGMP pathway in corticostriatal
long-term depression. J. Neurosci. 19, 2489–2499 (1999).
82. Picconi, B. et al. Inhibition of phosphodiesterases rescues striatal long-term depression
and reduces levodopa-induced dyskinesia. Brain 134, 375–387 (2011).
83. Arcangeli, S. et al. Ischemic-LTP in striatal spiny neurons of both direct and indirect
pathway requires the activation of D1-like receptors and NO/soluble guanylate
cyclase/cGMP transmission. J. Cereb. Blood Flow Metab. 33, 278–286 (2013).
84. Aziz, T.Z., Peggs, D., Sambrook, M.A. & Crossman, A.R. Lesion of the subthalamic
nucleus for the alleviation of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-
induced parkinsonism in the primate. Mov. Disord. 6, 288–292 (1991).
85. Bergman, H., Wichmann, T. & DeLong, M.R. Reversal of experimental parkinsonism
by lesions of the subthalamic nucleus. Science 249, 1436–1438 (1990).
86. Lang, A.E. et al. Posteroventral medial pallidotomy in advanced Parkinson’s
disease. N. Engl. J. Med. 337, 1036–1042 (1997).
87. Levy, R. et al. Effects of apomorphine on subthalamic nucleus and globus pallidus
internus neurons in patients with Parkinson’s disease. J. Neurophysiol. 86,
249–260 (2001).
88. Lewitt, P.A. Levodopa for the treatment of Parkinson’s disease. N. Engl. J. Med.
359, 2468–2476 (2008).
89. Jenner, P. Molecular mechanisms of L-DOPA-induced dyskinesia. Nat. Rev.
Neurosci. 9, 665–677 (2008).
90. Mercuri, N.B. & Bernardi, G. The ‘magic’ of L-dopa: why is it the gold
standard Parkinson’s disease therapy? Trends Pharmacol. Sci. 26, 341–344
(2005).
91. Perez-Lloret, S. & Rascol, O. Dopamine receptor agonists for the treatment of
early or advanced Parkinson’s disease. CNS Drugs 24, 941–968 (2010).
92. Bagetta, V. et al. Rebalance of striatal NMDA/AMPA receptor ratio underlies the
reduced emergence of dyskinesia during D2-like dopamine agonist treatment in
experimental Parkinson’s disease. J. Neurosci. 32, 17921–17931 (2012).
93. Connolly, B.S. & Fox, S.H. Drug treatments for the neuropsychiatric complications
of Parkinson’s disease. Expert Rev. Neurother. 12, 1439–1449 (2012).
94. Toy, W.A. et al. Treadmill exercise reverses dendritic spine loss in direct and
indirect striatal medium spiny neurons in the 1-methyl-4-phenyl-1,2,3,6-
tetrahydropyridine (MPTP) mouse model of Parkinson’s disease. Neurobiol. Dis.
63, 201–209 (2014).
95. Costa, R.M. et al. Rapid alterations in corticostriatal ensemble coordination during
acute dopamine-dependent motor dysfunction. Neuron 52, 359–369 (2006).
96. Hammond, C., Bergman, H. & Brown, P. Pathological synchronization in
Parkinson’s disease: networks, models and treatments. Trends Neurosci. 30,
357–364 (2007).
97. Kasanetz, F., Riquelme, L.A., Della-Maggiore, V., O’Donnell, P. & Murer, M.G.
Functional integration across a gradient of corticostriatal channels controls UP
state transitions in the dorsal striatum. Proc. Natl. Acad. Sci. USA 105,
8124–8129 (2008).
98. Zold, C.L., Escande, M.V., Pomata, P.E., Riquelme, L.A. & Murer, M.G. Striatal
NMDA receptors gate cortico-pallidal synchronization in a rat model of Parkinson’s
disease. Neurobiol. Dis. 47, 38–48 (2012).
99. Jin, X. & Costa, R.M. Start/stop signals emerge in nigrostriatal circuits during
sequence learning. Nature 466, 457–462 (2010).
100. Calabresi, P. & Di Filippo, M. Neuroscience: brain’s traffic lights. Nature 466,
449 (2010).
npg © 2014 Nature America, Inc. All rights reserved.