The Striatal Balancing Act in Drug Addiction: Distinct Roles of Direct and Indirect Pathway Medium Spiny Neurons

The striatum plays a key role in mediating the acute and chronic effects of addictive drugs, with drugs of abuse causing long-lasting molecular and cellular alterations in both dorsal striatum and nucleus accumbens (ventral striatum). Despite the wealth of research on the biological actions of abused drugs in striatum, until recently, the distinct roles of the striatum’s two major subtypes of medium spiny neurons (MSNs) in drug addiction remained elusive. Recent advances in cell-type-specific technologies, including fluorescent reporter mice, transgenic, or knockout mice, and viral-mediated gene transfer, have advanced the field toward a more comprehensive understanding of the two MSN subtypes in the long-term actions of drugs of abuse. Here we review progress in defining the distinct molecular and functional contributions of the two MSN subtypes in mediating addiction.


INTRODUCTION
Drugs of abuse exert potent molecular and cellular alterations in both dorsal striatum (dStr) and ventral striatum (nucleus accumbens, NAc), and many of these changes occur in medium spiny neurons (MSNs), the principal projection neurons in dStr and NAc, which account for 90-95% of all neurons in these regions. However, researchers have until recently been unable to clearly define the differential role of the two MSN subtypes in addictionrelated phenomena. The two MSN subtypes are differentiated by their enrichment of dopamine receptor 1 (D 1 ) or dopamine receptor 2 (D 2 ) as well as several other genes (Gerfen and Young, 1988;Gerfen et al., 1990;Le Moine et al., 1990Bernard et al., 1992;Ince et al., 1997;Lobo et al., 2006Lobo et al., , 2007Heiman et al., 2008;gensat.org) and by their distinct projections through the cortico-basal ganglia pathway (the direct vs. indirect pathways; Gerfen, 1984Gerfen, , 1992. Early work suggested that drugs of abuse exert most influence on the D 1 + MSNs, with the use of numerous dopamine receptor agonists and antagonists providing important insight into the functional and molecular roles of each MSN in drug reward behaviors (Self, 2010). However, current cell-typespecific methodologies, including fluorescent reporter mice that express GFP under D 1 or D 2 bacterial artificial chromosomes (BACs; Gong et al., 2003;Valjent et al., 2009; gensat.org), conditional mouse models such as the use of tetracycline-regulated inducible transgenic mice (Chen et al., 1998;Kelz et al., 1999), and transgenic mice expressing Cre-recombinase using D 1 or D 2 BACs, yeast artificial chromosomes (YACs), or knock-in mice (Gong et al., 2007;Lemberger et al., 2007;Heusner et al., 2008;Parkitna et al., 2009;Valjent et al., 2009;Bateup et al., 2010;Lobo et al., 2010; gensat.org) as well as cell-type-specific viral-mediated gene transfer (Cardin et al., 2010;Hikida et al., 2010;Lobo et al., 2010;Ferguson et al., 2011), have provided profound new insight into the precise molecular underpinnings of each MSN subtype and their regulation by drugs of abuse (Table 1).
Recent findings support the conclusion of a more predominate role for D 1 + MSNs in producing the reinforcing and sensitizing effect of drugs of abuse, with most robust molecular changes occurring in these MSNs. For instance, acute exposure to psychostimulants potently induces numerous signaling molecules including FosB, ERK, c-Fos, and Zif268 in the D 1 + MSNs, while repeated cocaine preferentially induces FosB and alters GABA receptor and other ion channel subunits in this cell-type as well (Robertson et al., 1991;Young et al., 1991;Berretta et al., 1992;Cenci et al., 1992;Moratalla et al., 1992;Hope et al., 1994;Bertran-Gonzalez et al., 2008;Heiman et al., 2008). Furthermore, disrupting or overexpressing specific molecules, such as FosB, DARPP-32, or Nr3c1 (the glucocorticoid receptor), in D 1 + MSNs typically mimics the drug-related behaviors observed when these alterations are made in a non-cell-type-specific manner, while disrupting such genes in D 2 + MSNs often causes an opposite response (Fienberg et al., 1998;Kelz et al., 1999;Deroche-Gamonet et al., 2003;Zachariou et al., 2006;Ambroggi et al., 2009;Bateup et al., 2010). Nonetheless, we cannot rule out an important contribution of the D 2 + MSNs in adaptations to drugs of abuse, because cocaine exposure alters gene expression in both MSN subtypes (Heiman et al., 2008) and D 2 -receptor agonists and antagonists exert potent effects in behavioral assays (Self, 2010). Indeed, recent findings show that molecular signaling adaptations in D 2 + MSNs potently modify an animal's behavioral response to drugs of abuse (Lobo et al., 2010). The latter findings showed that loss of TrkB (the receptor for BDNF) in D 2 + MSNs results in similar behavioral responses to cocaine as total TrkB knockout from the NAc, showing for the first time a selective dominant role for a molecular pathway in D 2 + MSNs in mediating the effects of drugs of abuse.  Durieux et al. (2009) Finally, recent literature reveals that the two MSNs exert antagonistic effects in drug-related behaviors, where activation of D 1 + MSNs or inhibition of D 2 + MSNs enhances an animal's sensitivity to a drug of abuse (Hikida et al., 2010;Lobo et al., 2010;Ferguson et al., 2011). These findings are consistent with opposing roles of the two MSNs and their direct vs. indirect pathways in the basal ganglia in motor behaviors (Alexander et al., 1986;Albin et al., 1989;Graybiel, 2000;Kravitz et al., 2010). This recent Frontiers in Neuroanatomy www.frontiersin.org literature is in accord with the general idea that dopaminergic neurotransmission, which is activated by all drugs of abuse, facilitates glutamatergic activation of D 1 + MSNs while inhibiting glutamatergic activation of D 2 + MSNs through its actions on D 1 vs. D 2 dopamine receptors (Figure 1). In this review, we address the current knowledge of the distinct molecular signaling exhibited by these two MSN subtypes in relation to their functional roles and responses to drugs of abuse.

DOPAMINE RECEPTOR SIGNALING IN D 1 VS. D 2 MSNs
As already noted, all drugs of abuse activate dopaminergic input to the NAc and related limbic brain regions Wise, 2004;Nestler, 2005). For instance, psychostimulants such as cocaine or amphetamine act directly on the dopaminergic reward pathway by interfering with the dopamine transporter: cocaine blocks the transporter and amphetamine reverses the transporter, both actions resulting in a build up of dopamine in the synapse which can activate downstream dopamine receptors on target neurons (Figure 1). The two MSNs are most notably differentiated by their enrichment of D 1 vs. D 2 -receptors although single-cell RT-PCR studies reveal that D 1 + MSNs express low levels of the D 2 -like receptor, D 3 and D 2 + MSNs express low levels of the D 1 -like receptor, D 5 (Surmeier et al., 1996). The two MSNs require glutamatergic innervation to drive neural activity; dopamine oppositely modulates these functional responses via stimulation of distinct dopamine receptor subtypes: by positively modulating excitatory glutamatergic input through D 1 receptor signaling via G s or G olf , which stimulates adenylyl cyclase leading to increased PKA activity, whereas dopamine negatively modulates this input through D 2 -receptor signaling via G i and G o which inhibit adenylyl cyclase causing decreased PKA activity (Surmeier et al., 2007;Gerfen and Surmeier, 2011). In reality, each receptor exerts complex effects on many additional downstream signaling pathways. At rest, the two MSN subtypes are generally inhibited, they are in what researchers have termed the down-state. Excitatory glutamatergic synaptic activity can release the MSNs from this down-state and shift them into a more depolarized state (the up-state). Dopamine oppositely modulates the excitatory glutamatergic shift to the up-state. D 1 activation of PKA enhances Cav1 L-type Ca 2+ channel activity, decreases somatic K + channel activity, and downregulates Cav2 Ca 2+ channels that control activation of Ca 2+ dependent, small-conductance K + (SK) channels, resulting in increased spiking in these MSNs (Surmeier et al., 2007;Gerfen and Surmeier, 2011). In contrast, D 2 signaling inhibits the up-state transition, thereby preventing increased spiking, via reduction of Cav1 Ltype Ca 2+ channel activity and Nav1 Na + channel activity while increasing K + channel currents (Surmeier et al., 2007;Gerfen and Surmeier, 2011; Figure 1). Such opposite alterations in the two MSNs suggest that increased dopamine signaling elicited by drugs of abuse should enhance glutamatergic activation of D 1 + MSNs and reduce glutamatergic activation of D 2 + MSNs. In reality, such responses are far more varied and complex for reasons that remain poorly understood. This topic will be addressed further below.
The role of dopamine receptors in drug abuse is complex and often elusive (Self, 2010). There is an abundance of literature on the role of D 1 and D 2 -receptor agonists and antagonists in modulating rewarding properties and self-administration of drugs of abuse, however, the results differ depending on the type of agonist/antagonist used, the type of delivery (systemic vs. brain region-specific), and the timing of the treatment (Self, 2010). Such results are further confounded by non-striatal specific effects, such as the contribution of pre-synaptic D 2 -receptors from the VTA or presence of D 1 receptors in many other limbic regions, and the lack of specificity of the agonists/antagonists utilized as well as the expression of D 1 -like and D 2 -like receptors in both MSN subtypes as noted earlier. In general, it is thought that D 1 receptors play a more predominant role in the FIGURE 1 | All drugs of abuse increase dopamine signaling in striatum, which can differentially modulate glutamatergic activity in the two MSN subtypes. In particular, cocaine binds to the dopamine transporter preventing dopamine reuptake into the terminals of VTA dopamine neurons. Activation of G s / olf coupled D 1 receptors enhances PKA activity and alters Ca 2+ and K + conductances to enhance the glutamate mediated "up-state" in these MSNs. In contrast, activation of G i /G o D 2 -receptors diminishes PKA activity and alters Ca 2+ , Na + , and K + conductances to diminish the glutamate mediated "up-state." This shifts these MSNs back to their resting "down-state." primary rewarding properties of drugs of abuse, whereas D 2receptors play a role in drug seeking mechanisms (Self et al., 1996;Self, 2010). Studies with D 1 receptor and D 2 -receptor knockout mice provide some insight into the role of these receptors in the two MSNs. D 1 knockout mice show a blunted induction of immediate early genes (IEGs) c-Fos and Zif268 in response to cocaine, a diminished response to psychostimulant-induced locomotor activity but with no alterations in cocaine-conditioned place preference (CPP) -an indirect measure of drug reward, and diminished cocaine self-administration and ethanol consumption (Miner et al., 1995;Drago et al., 1996;Crawford et al., 1997;El-Ghundi et al., 1998;Caine et al., 2007). D 2 knockout mice display diminished rewarding effects to opiates and cocaine as well as decreased ethanol consumption but no reduction in cocaine taking (Maldonado et al., 1997;Cunningham et al., 2000;Risinger et al., 2000;Caine et al., 2002;Chausmer et al., 2002;Elmer et al., 2002;Welter et al., 2007). Such data support important roles for D 1 and D 2 -receptors in the two MSNs in multiple aspects of drug abuse, however, the knockouts lack striatal specificity and occur early in development, thus one cannot rule out other brain regions and cell-types and developmental factors in mediating these behaviors. Finally, decreased levels of D 2 /D 3 receptors in striatum, as visualized by brain imaging, has become a common marker of addiction in human patients especially during periods of withdrawal (Volkow et al., 2009). Rodents receiving viral-mediated gene transfer of D 2 -receptors to the NAc display attenuated cocaine self-administration and ethanol consumption (Thanos et al., 2004(Thanos et al., , 2008. These studies were not performed in a cell-type-specific manner, so we cannot rule out the possible effect of D 2 -receptor overexpression influencing D 1 + MSNs. This collection of data emphasizes the need to move to more selective approaches, including cell-type-specific, region-specific, and even temporally specific manipulations of the dopamine receptors to better elucidate their functional roles in the two MSN subtypes in drug addiction. Finally, it has been reported recently that D 2 -GFP homozygote BAC transgenic mice display increased expression levels of the D 2 -receptor in striatum and enhanced behavioral sensitivity and dopamine signaling to D 2 agonists. Moreover, both homozygotes and hemizygotes exhibit blunted behavioral responses to cocaine (Kramer et al., 2011). This study highlights the need to perform thorough characterization of D 1 and D 2 fluorescent reporter and Cre driver lines. However, the majority of the data collected in this study used homozygotes, which is not the ideal experimental genotype since 5-10% of transgene integrations result in insertional mutations (Meisler, 1992); therefore, the hemizygote genotype is the more reliable experimental genotype. Additionally, this study did not use littermate wildtype controls but used controls on a similar background (Swiss Webster) obtained from Taconic, while their transgenic lines were obtained from GENSAT and MMRRC. Finally, another group has shown normal cocaine locomotor behavioral responses in D 2 -GFP hemizygotes (Kim et al., 2011). Thus, future studies using proper controls and proper genotypes must be performed to fully characterize the various cell-type-specific transgenic lines available.

GLUTAMATE AND GABA SIGNALING IN D 1 VS. D 2 MSNs
Medium spiny neurons receive glutamatergic input from multiple brain regions including prefrontal cortex, amygdala, and hippocampus, and GABAergic input from local interneurons and perhaps collateral inputs from other MSNs. Net excitatory and inhibitory regulation of MSNs is no doubt crucial in regulating the drug-addicted state, and there is now a growing literature on the complex ways in which drugs of abuse alter glutamatergic neurotransmission in particular in the NAc (Pierce et al., 1996;Thomas et al., 2001;Beurrier and Malenka, 2002;Kourrich et al., 2007;Conrad et al., 2008;Kalivas, 2009;Wolf, 2010). Although MSNs are thought to primarily exist in an inhibited down-state under basal conditions with glutamate driving activity of both cell-types, there remains limited information with respect to distinct regulation occurring in D 1 vs. D 2 MSNs.
FosB overexpression in D 1 + MSNs (see below for more details) enhances the rewarding effects of cocaine and increases levels of the Ca 2+ -impermeable glutamate receptor subunit, GluR2, in NAc. Furthermore, viral-mediated gene transfer of GluR2 to the NAc similarly enhances the rewarding effects of cocaine (Kelz et al., 1999). However, it is not known whether the induction of GluR2 seen in response to FosB overexpression in D 1 + MSNs is also specific to these neurons, and the viral overexpression of GluR2 is not cell-type-specific, therefore we cannot infer direct conclusions about GluR2 function in these two MSNs in drug reward. Heusner and Palmiter (2005) assessed the role of NMDA glutamatergic conductance in cocaine behaviors by expressing an NR1 subunit, which contains a mutation in the pore that reduces calcium flux, selectively in D 1 + MSNs. This group showed that lack of NMDA conductance in D 1 + MSNs prevents cocaine-induced CPP and cocaine locomotor sensitization, highlighting the necessity for NMDA signaling in D 1 + MSNs for the rewarding and sensitizing effects of cocaine (Heusner and Palmiter, 2005). Furthermore, recently it was found that knocking out the NR1 subunit in D 1 + MSNs attenuates amphetamine sensitization and this phenotype was rescued by resupplying the NR1 subunit to D 1 + MSNs specifically in the NAc (Beutler et al., 2011). Finally, knockdown of the mGluR5 subunit, using RNA interference, in D 1 + MSNs has no effect on the initial rewarding properties of cocaine but diminishes the cue-induced reinstatement of cocaine seeking (Novak et al., 2010). While these data reveal compelling roles for glutamatergic signaling in D 1 + MSNs, future work is needed to study glutamatergic systems in D 2 + MSNs. Future research should also evaluate how modulation of these glutamate receptor subunits in the two MSN subtypes affects the structural synaptic changes observed in NAc after drugs of abuse (Dietz et al., 2009;Russo et al., 2010), particularly the dendritic alterations observed after cocaine exposure selectively in the D 1 + MSNs (Lee et al., 2006;Kim et al., 2011) which may be associated with the increase in miniature excitatory postsynaptic currents observed in D 1 + MSNs (Kim et al., 2011). Interestingly, FosB induction in D 1 + MSNs has been related directly to such dendritic adaptations after chronic cocaine (Maze et al., 2010).
In contrast to glutamate, there is a lack of research on GABA function in the two MSNs in addiction models, which is surprising Frontiers in Neuroanatomy www.frontiersin.org considering both ethanol and benzodiazepines enhance the effects of GABA and the two MSNs receive dense GABAergic inputs as stated above. There is also considerable evidence pointing to enhanced inhibition in the NAc at least after chronic cocaine exposure (White et al., 1995;Peoples et al., 1998;Zhang et al., 1998;Thomas et al., 2001;Beurrier and Malenka, 2002). Heiman et al. (2008) performed high throughput genetic screening in the two MSNs after chronic cocaine exposure and, interestingly, the most altered biological process in the D 1 + MSNs was GABA signaling. In particular, there was potent upregulation of GABA A receptor subunits Gabra1 and Gabra4 as well as the GABA B receptor subunit Gabrb3, and this group found that chronic cocaine increases the frequency of small-amplitude GABAergic mini inhibitory postsynaptic currents (mIPSCs) in D 1 + MSNs (Heiman et al., 2008). On the other hand, another group recently showed that chronic cocaine results in an opposite response with decreased frequency and amplitude of mIPSCs in the D1+ MSNs (Kim et al., 2011). However, the latter group did show diminished membrane excitability in the D 1 + MSNs after chronic cocaine, which could be a reflection of enhanced GABA tone and is consistent with the field's assessment of enhanced inhibition in the NAc after exposure to chronic cocaine. Furthermore, such differences between the two groups could simply be due to the timing of cocaine exposure and withdrawal. In general, there is a need to study glutamatergic and GABAergic function in the two MSNs in response to drugs of abuse and the field is now equipped with the resources that make such a cell-type-and region-specific study possible.

OTHER RECEPTOR SIGNALING IN D 1 VS. D 2 MSN SUBTYPES
The two MSNs are differentially enriched in other G-proteincoupled receptors in addition to dopamine receptors. D 1 + MSNs express higher levels of the acetylcholine muscarinic receptor 4 (M 4 ; Bernard et al., 1992;Ince et al., 1997) and D 2 + MSNs are enriched in both adenosine receptor 2A (A 2A ; Schiffmann et al., 1991;Schiffmann and Vanderhaeghen, 1993) and G-proteincoupled receptor 6 (Gpr6; Lobo et al., 2007;gensat.org). M 4 is coupled to G i/o , which would produce an opposite response, compared to D 1 receptors, in D 1 + MSNs by inhibiting cAMP/PKA activity. Indeed, a D 1 + MSN selective M 4 knockout displayed enhanced behavioral sensitization to cocaine and amphetamine (Jeon et al., 2010). Furthermore, recent studies using a designer receptor exclusively activated by a synthetic drug (DREADDs) showed that activation of the DREADD Gi/o-coupled human M 4 receptor (hM 4 D) in D 1 + MSNs diminished behavioral sensitization to amphetamine, with the opposite response seen in D 2 + MSNs (Ferguson et al., 2011). Such data reveal the antagonizing role of M 4 receptors in D 1 + MSNs in drug abuse. As well, since the hM 4 D receptor potently inhibits these MSNs, the data provide insight into the effect of altered activity of these two MSNs in drug abuse, which will be discussed further below. Both A 2A and Gpr6 are positively coupled to G s /G olf proteins, implicating their role in antagonizing the D 2 -receptor in D 2 + MSNs. Indeed, stimulation of A 2A receptors has been shown to reduce both the development and expression of cocaine sensitization (Filip et al., 2006), impair the initiation of cocaine self-administration (Knapp et al., 2001), and antagonize the reinstatement of cocaine seeking elicited by cocaine, D 2 -receptor stimulation, or cocaine-conditioned cues ). As Gpr6 is also enriched in D 2 + MSNs (Lobo et al., 2007), its role in behavioral functions of the striatum should be evaluated. To date, it has been shown to influence instrumental learning (Lobo et al., 2007) but its role in drug abuse models is yet unknown.
The cannabinoid receptor 1 (CB1) is expressed ubiquitously throughout the central nervous system (Mackie, 2008), hence it is difficult to dissect the precise role of specific brain regions and cell-types in mediating 9-tetrahydrocannabinol (THC) addiction. Recently, deletion of CB1 from D 1 + MSNs was found to modestly affect behavioral responses to THC, including blunted effects in THC-induced hypolocomotion, hypothermia, and analgesia (Monory et al., 2007). It would be interesting to evaluate cannabinoid receptor function in D 2 + MSNs since these MSNs express endocannabinoid-mediated long-term depression (eCB-LTD), which requires dopamine D 2 -receptor activation (Kreitzer and Malenka, 2007).
The glucocorticoid receptor, Nr3c1, is also broadly expressed in the CNS and periphery. Stress-induced glucocorticoid secretion can potentiate maladaptive behaviors including drug addiction (Frank et al., 2011). In particular, disrupting glucocorticoid signaling in D 1 + MSNs by deleting Nr3c1 diminished the motivation these mice display to self-administer cocaine, and this is consistent with previous data where Nr3c1 was deleted from the entire brain (Ambroggi et al., 2009). These data are consistent with other findings described in this review, showing a predominant role for D 1 + MSNs in mediating many of the effects of drugs of abuse.
Finally, we recently disrupted BDNF signaling in the two MSNs by deleting its TrkB receptor selectively from each MSN subtype. We observed opposite effects on cocaine-elicited behaviors: cocaine-induced locomotor activity and the induction of cocaine CPP were enhanced after TrkB deletion from D 1 + MSNs, but attenuated after deletion from D 2 + MSNs (Lobo et al., 2010). Interestingly, the deletion of TrkB from D 2 + MSNs mimics the effects of total deletion of TrkB from the NAc as well as disruption of BDNF signaling from the VTA (Horger et al., 1999;Graham et al., 2007Graham et al., , 2009Bahi et al., 2008;Crooks et al., 2010). These findings thus show for the first time a predominant role of a signaling cascade in D 2 + MSNs in mediating the effects of a drug of abuse. The predominant role of D 2 + MSNs in mediating BDNF's effects on cocaine-elicited behaviors is not surprising considering both TrkB mRNA and protein are enriched in D 2 + MSNs (Lobo et al., 2010;Baydyuk et al., 2011). The behavioral changes observed in these mice were accompanied by enhanced neuronal activity in the D 2 + MSNs upon a selective knockout of TrkB. These findings prompted us to use optogenetic technology to selectively manipulate MSN activity in cocaine reward (see below).

TRANSCRIPTION FACTORS IN D 1 VS. D 2 MSNs
The most compelling evidence for the more robust role of D 1 + MSNs in drug abuse comes from literature evaluating induction of intracellular signaling molecules. As stated above, acute doses of psychostimulants induce IEG expression, including c-Fos, Zif268 (Egr1), and FosB primarily in D 1 + MSNs in NAc Frontiers in Neuroanatomy www.frontiersin.org and dStr (Robertson et al., 1991;Young et al., 1991;Berretta et al., 1992;Cenci et al., 1992;Moratalla et al., 1992;Bertran-Gonzalez et al., 2008). This induction requires activation of D 1 receptors, and the cell-type-specificity of the IEG induction in response to acute cocaine was recently confirmed using D 1 -GFP and D 2 -GFP reporter mice (Bertran-Gonzalez et al., 2008). Interestingly, the confirmation of cocaine's induction of c-Fos primarily in D 1 -GFP throughout striatum with a small induction in D 2 -GFP MSNs only in dStr was confirmed using a context-dependent paradigm (mice were injected in a novel environment outside of their home cage). Furthermore, a previous study using in situ hybridization in mice also showed induction of c-Fos in D 1 + and D 2 + MSNs in dStr, although in this study representative bar graphs show greater number of D 1 + c-Fos positive neurons (Ferguson et al., 2006). Interestingly, this study reveals significantly enhanced c-Fos induction in D 2 + MSNs in the dStr after loss of ERK1, which parallels our findings of enhanced c-Fos induction in D 2 + MSNs specifically in the NAc shell after disruption of BDNF signaling which is known to enhance ERK activity (Lobo et al., 2010). However, opposite behavioral responses to cocaine were observed in each study, which may reflect induction of c-Fos in D 2 + MSNs in dStr vs. NAc shell. Finally, previous literature using in situ hybridization/immunohistochemistry in rats has shown acute psychostimulants can induce c-Fos equally in both MSNs when the drug is given in a novel environment (Badiani et al., 1999;Uslaner et al., 2001a,b;Ferguson and Robinson, 2004) and chronic administration of amphetamine is reported to selectively induce c-Fos in D 2 + MSNs (Mattson et al., 2007). These different results could be a reflection of the experimental procedures used (in situ hybridization vs. GFP reporter mice) or even be due to the animal species used as the latter experiments used rats.
Recently, researchers genetically profiled the cocaine contextdependent, c-Fos activated neurons in rats using immunolabeled fluorescence activated cell sorting (FACS) and showed that the c-Fos+ neurons are enriched in a D 1 + MSN gene, prodynorphin (Pdyn), but have lower levels of D 2 and A 2A , both D 2 + MSN genes (Guez-Barber et al., 2011), suggesting that the c-Fos+ activated neurons consist primarily of D 1 + MSNs. Furthermore, this group previously showed that c-Fos expressing MSNs are important for this context-dependent sensitization, as ablation of these neurons abolishes this behavioral phenotype (Koya et al., 2009). Although previous data showed that the cocaine context-dependent induction of c-Fos occurs in both D 1 + and D 2 + MSNs in rats, the more recent results correspond to findings in which deletion of c-Fos selectively from D 1 + MSNs blunts cocaine-induced locomotor sensitization in mice (Zhang et al., 2006). Furthermore, this group found that deletion of c-Fos in D 1 + MSNs blunts the dendritic spine changes normally induced by cocaine in the NAc, indicating a role for c-Fos in mediating these synaptic plasticity changes. Finally, the group observed no change in the induction of cocaine CPP, but found that loss of c-Fos in D 1 + MSNs prevented extinction of cocaine CPP. Such data illustrate a dynamic role for c-Fos induction in D 1 + MSNs, however, one cannot rule out the differential effects at the behavioral level as being mediated by any of several other limbic brain regions that express the D 1 receptor.
Another IEG that has been extensively studied in the two MSN subtypes is FosB. Acute exposure to cocaine induces FosB in D 1 + MSNs (Berretta et al., 1992), whereas chronic exposure induces FosB, a stable product of the FosB gene generated by alternative splicing (Hope et al., 1994;Nestler et al., 2001;Nestler, 2008), in D 1 + MSNs (Nye et al., 1995;Moratalla et al., 1996;Lee et al., 2006). Similar findings are observed with many other drugs of abuse as well as with natural rewards such as food, sex, and wheel running. For example, chronic wheel running, which is a natural reward (Iversen, 1993;Belke, 1997;Lett et al., 2000), induces FosB in D 1 + MSNs but not D 2 + MSNs (Werme et al., 2002). To gain functional insight into the role of FosB in the two MSNs, our group generated NSE-tTa lines, termed 11A and 11B, which direct transgene expression to either D 1 + or D 2 + MSNs, respectively (Chen et al., 1998;Kelz et al., 1999;Werme et al., 2002). Line 11A mice crossed with a Tet-Op FosB line show increased responses to the rewarding and locomotor effects of cocaine (Kelz et al., 1999), which is consistent with FosB induction in D 1 + MSNs (Nye et al., 1995;Moratalla et al., 1996). Furthermore, these same mice display increased morphine reward (evaluated by CPP) as well as diminished morphine analgesia and enhanced morphine tolerance, while the 11B Tet-Op FosB mice show no change in morphine reward. Overexpression of a dominant negative antagonist of FosB exerts effects opposite to those seen with FosB, although this mouse model does not distinguish D 1 vs. D 2 MSNs (Peakman et al., 2003). Together, these data further supports the role of FosB induction in D 1 + MSNs as an important molecular player in the rewarding properties of drugs of abuse (Zachariou et al., 2006). This phenomenon is also observed in other reward behaviors, in particular, wheel running: 11A Tet-Op FosB mice display increased wheel running behavior, whereas 11B Tet-Op FosB mice display diminished wheel running (Werme et al., 2002). The finding that FosB induction in D 1 MSNs promotes reward is consistent with recent findings that such cell-typeselective induction also promotes resilience responses to chronic stress (Vialou et al., 2010). Finally, chronic cocaine induction of FosB in D 1 + MSNs was shown to be accompanied by robust long-lasting increases in dendritic spine densities (Lee et al., 2006) and recently FosB in the NAc was shown to be both necessary and sufficient in mediating the increased density of dendritic spines in this brain region (Maze et al., 2010). Such data support a role for FosB in D 1 + MSNs in mediating the rewarding aspects of drugs of abuse and natural rewards as well as the accompanying structural plasticity changes. The data also suggest that induction of FosB in D 2 + MSNs confers negative consequences to rewarding stimuli. Since FosB induction in D 2 + MSNs is seen in response to chronic stress and antipsychotic drug exposure (Hiroi and Graybiel, 1996;Perrotti et al., 2004), further studies of the latter actions are needed.

OTHER INTRACELLULAR SIGNALING MOLECULES IN D 1 VS. D 2 MSNs
One signaling molecule that has been well studied in the two MSNs in the context of drug abuse is the protein kinase, ERK (extracellular signal related kinase). Acute or chronic exposure to cocaine induces phosphorylated ERK (pERK), the activated form of the protein, in the NAc and dStr in D 1 + MSNs using D 1 -GFP and Frontiers in Neuroanatomy www.frontiersin.org D 2 -GFP BAC transgenic reporter mice (Bertran-Gonzalez et al., 2008) and this response is mediated through D 1 receptors (Valjent et al., 2000;Lu et al., 2006). This group also showed that pMSK-1 (phospho-MAP and stress activated kinase-1) and histone H3, both targets of pERK signaling, are robustly induced in pERK containing D 1 + MSNs after acute cocaine exposure and modestly increased after chronic cocaine (Bertran-Gonzalez et al., 2008). pERK is also induced is response to chronic morphine, in particular, pERK is robustly induced in D 1 + MSNs and modestly induced in D 2 + MSNs in the NAc shell after withdrawal in response to the context-specific association with morphine (Borgkvist et al., 2008). The precise functional role of pERK in drug addiction remains to be determined. Pharmacological treatment with ERK inhibitors has been shown to decrease cocaine reward, however, a knockout of ERK1 potentiates cocaine reward, suggesting that ERK inhibitors may preferentially be affecting ERK2.
Recently, we showed that optogenetic activation of D 1 + MSNs in the NAc, which increases an animal's rewarding responses to cocaine, potently reduces both pERK1 and pERK2. Future studies manipulating ERK expression in a cell-type-specific manner are necessary to fully address the functional role of ERK signaling in the two MSNs in drug abuse. DARPP-32 is another signaling molecule that has been extensively studied in response to drugs of abuse. It is well known that acute psychostimulants lead to PKA phosphorylation of DARPP-32 at threonine 34 (T34), causing it to become a potent inhibitor of protein phosphatase 1 (PP-1), which regulates the phosphorylation state of many effector proteins, including transcription factors, ionotropic receptors, and ion channels (Greengard et al., 1999). However, until recently, it was unclear which MSN subtype mediates this biochemical change. Greengard et al. (1999) generated BAC transgenic mouse models that enable the evaluation of DARPP-32 phosphorylation in D 1 + or D 2 + MSNs by expressing tagged versions of DARPP-32 using D 1 or D 2 BACs allowing for immunoprecipitation of DARPP-32 from each MSN subtype. These studies demonstrated that acute cocaine treatment increases T34 phosphorylation in D 1 + MSNs and induces phosphorylation of threonine 75 (T75) by Cdk5, which inhibits PKA signaling, selectively in D 2 + MSNs (Bateup et al., 2008). Finally this group showed that deletion of DARPP-32 from each MSN subtype using D 1 -Cre and D 2 -Cre BAC transgenic mice results in opposite regulation of cocaine-induced locomotor activity (Bateup et al., 2010). Loss of DARPP-32 from D 1 + MSNs diminished the locomotor effects of cocaine, which mimics previous data evaluating a total DARPP-32 knockout (Fienberg et al., 1998), whereas loss of DARPP-32 from D 2 + MSNs enhanced cocaine locomotor responses. Such data provide concrete evidence for differential roles of DARPP-32 in the two MSNs in response to drugs of abuse and illustrate the importance of cell-type-specific methods to fully understand the contribution of these two neuronal types in drug addiction.

MODULATING ACTIVITY OF D 1 OR D 2 MSNs
Directly modulating the activity of the two MSN subtypes has recently provided novel insight into the molecular and functional role of D 1 and D 2 MSNs in addiction. We used optogenetic tools combined with a conditional (i.e., Cre-dependent) adeno-associated viral (AAV) vector expressing the blue lightactivated cation channel, channelrhodopsin-2 (ChR2). We injected the vector, or a control, into the NAc of D 1 -Cre or D 2 -Cre BAC transgenic mice and then stimulated the injected region with blue light to selectively activate D 1 + vs. D 2 + MSNs in the context of cocaine CPP. We found that activation of D 1 + MSNs potentiates induction of cocaine CPP, whereas activation of D 2 + MSNs inhibits this induction (Lobo et al., 2010). As noted previously, we observed the same behavioral effects when TrkB was deleted selectively from these MSN subtypes: enhanced cocaine CPP and locomotor activity after TrkB deletion from D 1 + MSNs, and reduced cocaine CPP and locomotor activity after TrkB deletion from D 2 + MSNs. The likely common action of TrkB knockout and optogenetic stimulation in D 2 + MSNs is their increased activity, since deletion of TrkB from these cells increases their electrical excitability. As mentioned earlier, we also found a robust reduction of pERK after TrkB deletion from D 1 + MSNs. pERK is a known downstream target of BDNF signaling, therefore, the shared behavioral effects observed after TrkB deletion from D 1 + MSNs and from optogenetic activation of these cells might be due to converging effects on pERK activity. However, future work is needed to determine the precise, shared molecular underpinnings that govern the behavioral effects seen after disruption of BDNF signaling and optogenetic control of these two neuronal subtypes.
Other groups have used different tools to modulate activity of the two MSNs in drug abuse models. Hikida et al. (2010) used AAV vectors to express tetracycline-repressive transcription factor (tTa) using the substance P (a D 1 + MSN gene) or enkephalin (a D 2 + MSN gene) promoters. These vectors were injected into the NAc of mice, in which tetanus toxin light chain (TN) -a bacterial toxin that cleaves the synaptic vesicle-associated protein, VAMP2 -was controlled by the tetracycline-responsive element, to selectively abolish synaptic transmission in each MSN subtype. Consistent with our optogenetic approach, these data showed a role of D 1 + MSN activity in enhancing cocaine CPP as well as cocaine-induced locomotor activity, since abolishing synaptic transmission in D 1 + MSNs diminished both behavioral effects. In contrast to the optogenetic studies, the authors found no alterations in cocaine CPP after abolishing synaptic transmission in D 2 + MSNs, but did observe reduced cocaine-induced locomotor activity in response to the first two cocaine exposures. Interestingly, this group showed that inactivation of the D 2 + MSNs played a more profound role in mediating aversive behaviors.
As stated earlier, Ferguson et al. (2011) used herpes simplex virus (HSV) vectors to express an engineered GPCR (a G i/o -coupled human muscarinic M 4 designer receptor exclusively activated by a designer drug, hM 4 D) that is activated by an otherwise pharmacologically inert ligand using enkephalin and dynorphin promoters to selectively silence D 1 + or D 2 + MSNs in the dStr. The authors showed that transiently disrupting D 2 + MSN activity in dStr facilitated amphetamine sensitization, whereas decreasing excitability of D 1 + MSNs impaired the persistence of amphetamine-induced sensitization. Finally, abolishing D 2 + MSNs in the NAc at adult ages using diptheria toxin receptor enhances the rewarding effect of amphetamine (Durieux et al., 2009). Such data are in accordance with our optogenetic findings, Frontiers in Neuroanatomy www.frontiersin.org and together implicate opposite roles of D 1 + vs. D 2 + MSNs in drug addiction, with D 1 + MSNs promoting both reward and sensitizing responses to psychostimulants and D 2 + MSNs dampening these behaviors.

FUTURE DIRECTIONS
The field has made tremendous advances toward understanding the selective role of the D 1 + and D 2 + MSN subtypes in NAc and dStr in mediating the effects of drugs of abuse. In particular, recently developed tools that enable the selective manipulation of these cell-types have played a predominant role in obtaining the majority of this information. What are the next steps? Since the underlying molecular adaptations in drug addiction models are not static, but very dynamic, it is crucial to develop the capability to selectively manipulate signaling molecules of interest in D 1 + vs. D 2 + MSNs in a temporally precise way. DREADDs and optogenetic tools can help with this time scale manipulation. DREADD ligands can be administered at different time courses throughout drug behavioral paradigms to parcel out the selective role of signaling receptors in the two MSNs in drug models. Optogenetic tools in particular provide an extremely powerful means to temporally regulate not only neuronal activity but G-protein-coupled receptor signaling using OptoXRs (Airan et al., 2009), glutamatergic signaling (Volgraf et al., 2006;Numano et al., 2009), GABAergic signaling, and even certain intracellular signaling molecules (Wu et al., 2009;Hahn and Kuhlman, 2010). Ultimately, it may be possible to extend these capabilities to optogenetic regulation of transcriptional activity. Likewise, optogenetic tools are making it possible for the first time to study the influence of specific inputs to striatum and to determine whether such inputs impinge in selective ways on D 1 + vs. D 2 + MSNs (Higley and Sabatini, 2010). The ability to control such signaling and molecular properties with great temporal resolution will allow major steps to be made toward a more comprehensive understanding of the two MSN subtypes, and other cell subtypes in NAc and dStr, in mediating the time course and different phases of drug addiction.