Differential Modulation of GABAergic and Glutamatergic Neurons in the Ventral Pallidum by GABA and Neuropeptides

Abstract The ventral pallidum (VP) is an integral locus in the reward circuitry and a major target of GABAergic innervation of both D1-medium spiny neurons (MSNs) and D2-MSNs from the nucleus accumbens. The VP contains populations of GABAergic [VPGABA, GAD2(+), or VGluT(–)] and glutamatergic [VPGlutamate, GAD2(–), or VGluT(+)] cells that facilitate positive reinforcement and behavioral avoidance, respectively. MSN efferents to the VP exert opponent control over behavioral reinforcement with activation of D1-MSN afferents promoting and D2-MSN afferents inhibiting reward seeking. How this afferent-specific and cell type-specific control of reward seeking is integrated remains largely unknown. In addition to GABA, D1-MSNs corelease substance P to stimulate neurokinin 1 receptors (NK1Rs) and D2-MSNs corelease enkephalin to activate μ-opioid receptors (MORs) and δ-opioid receptors. These neuropeptides act in the VP to alter appetitive behavior and reward seeking. Using a combination of optogenetics and patch-clamp electrophysiology in mice, we found that GAD2(–) cells receive weaker GABA input from D1-MSN, but GAD2(+) cells receive comparable GABAergic input from both afferent types. Pharmacological activation of MORs induced an equally strong presynaptic inhibition of GABA and glutamate transmission on both cell types. Interestingly, MOR activation hyperpolarized VPGABA but not VGluT(+). NK1R activation inhibited glutamatergic transmission only on VGluT(+) cells. Our results indicate that the afferent-specific release of GABA and neuropeptides from D1-MSNs and D2-MSNs can differentially influence VP neuronal subtypes.


Introduction
The ventral pallidum (VP) is an integral component of the reward circuitry that translates motivation into action (Kalivas et al., 1999;Smith et al., 2009). In addition to strong GABAergic innervation from the nucleus accumbens (NAc), the VP also receives glutamatergic innervation from the basolateral amygdala and subthalamic nucleus, cholinergic innervation from local interneurons, as well as dopaminergic innervation from the ventral tegmental area (VTA; Root et al., 2015). VP neurons encode several integral components of reward processing and behavioral execution such as the valence of reward (Tindell et al., 2005;Ottenheimer et al., 2018), reward prediction errors (Ottenheimer et al., 2020;Stephenson-Jones et al., 2020), and the initiation of reward-seeking behavior (Richard et al., 2016).
At a cellular level, the VP is composed of heterogeneous subpopulations of neurons that differentially use GABA (VPGABA; ;70%) or glutamate (VPGlutamate; ;25%) as neurotransmitters (Heinsbroek et al., 2020). Both cell types drive opponent behaviors, with activation of VPGABA neurons supporting the seeking of positive reinforcement and activation of VPGlutamate facilitating behavioral avoidance (Faget et al., 2018;Tooley et al., 2018;Heinsbroek et al., 2020;Stephenson-Jones et al., 2020). Additionally, stressinduced adaptations in these two subpopulations of VP neurons differentially contribute to behavioral despair and social withdrawal, two key symptoms of depressive disorder (Knowland et al., 2017).
Both D1-medium spiny neurons (MSNs) and D2-MSNs from the NAc form functional synapses in the VP and exert opponent control on reward-related behaviors (Kupchik et al., 2015;Creed et al., 2016). Chemogenetic manipulations demonstrate that D1-MSN to VP afferents promote and D2-MSN to VP afferents inhibit reward seeking (Creed et al., 2016;Heinsbroek et al., 2017;Pardo-Garcia et al., 2019). How this antagonistic behavioral modulation is achieved on a microcircuit level in the VP remains unclear. However, a rabies tracing study revealed that significantly more D1-MSNs than D2-MSNs in the NAc innervate VPGlutamate cells (Heinsbroek et al., 2020). Nonetheless, it remains unknown whether the apparently biased connectivity between NAc afferents and VP neuronal subpopulations is reflected in GABA synaptic strength and modulation by cotransmitters differentially released with GABA from D1-MSN and D2-MSN axon terminals.
Here we used circuit-specific optogenetics in transgenic mouse lines expressing a reporter in VPGABA neurons [VPGAD(1)] combined with whole-cell patch-clamp recordings to characterize the functional connectivity between the D1/D2-MSNs and VPGABA/VPGAD2(-) neurons and to test an hypothesis that the apparent bias in anatomic connectivity (Heinsbroek et al., 2020) is recapitulated in biased functional connectivity. We furthermore used MOR and NK1R agonists to evaluate the potential cell type-specific modulation by two of the coreleased neuropeptides. Given the previous literature that the stimulation of MOR or NK1R in the VP can promote motivated behavior, we hypothesized that agonists at these receptors would either stimulate VPGABA or inhibit VPGlutamate.
We found that while VPGABA cells receive comparable input from D1-MSN and D2-MSN afferents, VPGAD2(-) receives significantly lower GABA A receptor (GABA A R)-mediated transmission from D1-MSNs compared with D2-MSNs. MOR activation inhibited both GABA and glutamate presynaptic transmission onto both cell types to a similar extent. However, MOR stimulation had a postsynaptic effect (hyperpolarization and reduced firing) only in VPGAD2(1) but not VPGAD2(-). NK1R activation inhibited glutamatergic transmission only onto VPGAD2(-) cells. Hence, the neuropeptidergic control via the D2 coreleased neuropeptide enkephalin more tightly controlled VPGABA neurons, and the D1 coreleased neuropeptide only controls synaptic input onto VPGlutamate neurons.
Animals were group housed, fed ad libitum, and maintained in a humidity-controlled and temperature-controlled environment with 12 h light/dark cycle. Experiments were performed during the dark cycle. All experiments were conducted by the National Institute of Health Guidelines for the Care and Use of Laboratory Animals, and all procedures were approved by the Institutional Animal Care and Use Committee at our institute.
Research Article: New Research vertical puller (Narishige) and filled with internal solution according to distinct experiments (see below). Data were acquired at 10 kHz and filtered at 2 kHz using AxoGraph X software (AxoGraph Scientific). Recordings with unstable series resistance or with a series resistance .20 MX were discarded. Raw data were analyzed using AxoGraph X software.

Experiments to quantify GABA A R-mediated transmission from D1-MSN and D2-MSN terminals
Recordings of synaptic currents were performed in voltage clamp using a high-chloride internal solution to record GABA A R currents (in mM: 110 CsCl, 30 potassium gluconate, 10 HEPES potassium, 1 EGTA, 3 QX 314-Cl, 2.0 MgATP and 0.2 NaGTP, pH 7.2-7.3, 290 mOsm; Figs. 1, 2). Both, GAD2 (1) (VPGABA, identified by their fluorescent soma) and GAD2(-) cells (presumed VPGlutamate, identified by the lack of fluorescence) were identified at 40Â under a microscope. Recordings of synaptic transmission were started 10 min after invading the cell to allow for internal solution diffusion to remote dendrites. Excitatory synaptic transmission was blocked with 6-cyano-7-nitroquinoxaline-2,3-dione (10 mM) in the aCSF. IPSCs were evoked optically. The light pulse was produced by a 460 nm LED (Mightex) that was transmitted on the slice through the microscope objective. For input-output recordings, stimulation consisted of two pulses separated by 100 ms every 20 s. Stimulus intensity was increased from (3 ms, 0.5-8 mW) and at least six repetitions for each intensity recorded. For extended paired-pulse ratio, varying interstimulus intervals (25, 50, 100, and 200 ms) were tested every 20 s with six repetitions each. The stimulation intensity was set to evoke an IPSC with amplitudes within the dynamic range of the cells (i.e., 30-70% of maximal IPSCs; 200-900 pA).

Experiments to quantify the modulation of transmitter release by neuropeptides
Recordings of synaptic currents were performed in voltage-clamp configuration using a cesium-methanesulfonatebased internal solution (in mM: 115 Cs1methanesulfonate, 10 HEPES, 1 EGTA, 1 MgCl2, 1 NaCl, 2 Mg21-ATP, 0.3 Na 1 -GTP, 3 QX 314-Cl, and 10 BAPTA-tetracesium, with 290 mOsm, at pH 7.4; Figs. 3, 4). We used a high concentration of BAPTA in the internal solution to chelate calcium and to prevent calcium signaling and potential synaptic plasticity induced by depolarization (Francis et al., 2019). Afferents were stimulated electrically via a bipolar stimulation electrode positioned 200-300 mm dorsomedial to the recorded cell. The stimulation intensity was set to evoke an EPSC with amplitudes within the dynamic range of the cells (i.e., 30-70% of maximal IPSCs; 200-900 pA). Biophysically isolated basal AMPAR evoked EPSCs (eEPSCs) at -65 mV and basal GABA A evoked IPSCs (eIPSCs) at 0 mV were recorded for a stable baseline. We then applied DAMGO (1 mM) or GR73632, (0.2 mm). After 10 min of drug application, we subsequently recorded AMPAR eEPSCs at À65 mV. eEPSCs and eIPSCs are regularly blocked by DNQX and picrotoxin (PTX), respectively.
Statistical analyses of data were performed with GraphPad Prism (GraphPad Software) using tests indicated in the main text. Nonparametric tests were used if data distributions for comparisons of groups did not pass the Kolmogorov-Smirnov normality test (Extended Data Table 1-1).

Results
GABA A R-mediated transmission from D1-MSNs on GAD2(1) and GAD2(-) cells First, we aimed to probe the synaptic strength between NAc D1-MSNs on VPGABA and VPGlutamate cells. We microinjected a cre-dependent virus carrying a ChETA variant of channelrhodopsin into the NAc core of Drd1a-Cre::Gad2-T2a-NLS-mCherry (Fig. 1A). Adult mice were killed 4 weeks after virus injection and either GAD2(1) cells (presumed VPGABA cells) or GAD2(-) (presumed VPGlutamate cells) were recorded within the terminal field A, Schematic of the experimental protocol. VGluT2-IRES-Cre mice were crossed with Ai14 mice. Ai14 mice express robust tdTomato fluorescence in VPGlutamate cells following Cre-mediated recombination. VPGlutamate cells were recorded, and synaptic afferents were stimulated electrically. EPSCs and IPSCs were isolated biophysically by clamping the cells at -65 and 0 mV, respectively. eEPSCs and eIPSCs are were regularly blocked by DNQX and PTX, respectively (see example traces). For control experiments without agonists, see Extended Data Figure 3-1. B, Example of serial recordings of EPSCs and IPSCs in a representative VPGlutamate cell. Both, EPSCs measured at À65 mV and IPSCs at 0 mV were decreased by the application MOR agonist DAMGO. C, DAMGO inhibited elicited EPSCs and IPSCs (left) and significantly increased the paired-pulse ratio of EPSCs (right; pairwise comparison between preapplication and postapplication; 6 cells, 5 animals). D, Example of serial recordings of EPSCs and IPSCs in a representative VPGlutamate cell. EPSCs measured at À65 mV but not IPSCs at 0 mV were decreased by application of NK1R agonist GR73632. E, GR73632 application selectively inhibited elicited EPSCs but not IPSCs (left) and selectively increased the paired-pulse ratio of elicited EPSC (right; 9 cells, 5 animals). F, Change in E/I ratio after DAMGO or GR73632 indicates a divergent modulation of inputs onto VPGlutamate cells. *p , 0.05; **p , 0.01.
Research Article: New Research of transfected MSNs within the VP (Fig. 1B). Optical stimulation of D1-MSN terminals was used to evoke GABAergic inhibitory postsynaptic responses [optically evoked IPSCs (oIPSCs)]. We found that a higher percentage of recorded GAD2(1) cells than GAD2(-) cells received input from D1-MSN afferents ( Fig. 1B; x 2 ¼ 16.26, p , 0.01). Figure 1C, left, shows the oIPSCs of responsive GAD2(1) and GAD2(-) cells in response to increasing laser intensity. The strength of D1-MSN input on GAD2(1) cells was significantly stronger than that for D1-MSN input on GAD2(-) cells (Fig. 1C, left; F (1,132) ¼ 10.33, p , 0.01; cell type, two-way ANOVA) and the maximum oIPSC amplitudes were higher for GAD2(1) cells (Fig. 1C, right; p , 0.01, Mann-Whitney test). To estimate whether the weaker synaptic strength of D1-MSNs onto GAD2(-) could be explained by lower GABA release probability, we quantified paired-pulse ratios in response to four different interstimulus intervals (ISIs) and found a strong trend toward higher ISIs for D1-MSN to GAD2(-) synapses (Fig. 1D, left; F (1,81) ¼ 3.486, p ¼ 0.065, cell type, two-way ANOVA). An additional set of cells from a previous experiment that was only tested for a fixed ISI of 100 ms also showed significantly higher paired-pulse ratios for D1-VPGlutamate responses (Fig. 1D, right; t (35) ¼ 3.187; p , 0.01, unpaired t test). These results indicate that the weaker synaptic strength of D1-VPGlutamate is likely caused by a lower GABA release probability from these synapses.

GABA A R-mediated transmission from D2-MSN on GAD2(1) and GAD2(-) cells
Similar to the previous section, we next probed the functional connectivity between NAc D2-MSNs on VPGABA and VPGlutamate cells via optogenetics in Drd2a-Cre:: Gad2-mCherry mice ( Fig. 2A). GABAergic oIPSCs were recorded from GAD2(1) and GAD2(-) cells in the fluorescent terminal fields of VP slices (Fig. 2B). We found that a higher percentage of VPGABA responded to D2-MSN terminal stimulation ( Fig. 2B; x 2 ¼ 8.866, p , 0.001). Input-output curves testing the relationship between laser power and oIPSC showed no significant difference in the strength of GABAergic transmission of D2-MSN inputs onto GAD2(1) Figure 4. MOR activation but not NK1R activation inhibits both GABA A R and glutamate transmission onto VPGABA cells. A, Schematic of the experimental protocol. VGluT2-IRES-Cre mice were crossed with Ai14 mice for robust tdTomato fluorescence in VPGlut(-) cells. VPGlut(-) cells (presumably VPGABA) were recorded, and synaptic afferents were stimulated electrically. EPSCs and IPSC were isolated biophysically by clamping the cells at -65 and 0 mV, respectively. EPSCs and IPSCs were regularly blocked by DNQX and PTX, respectively (see example traces). B, Representative example of serial recordings of EPSCs and IPSCs in a representative VPGlut(-) cell. Both EPSCs measured at À65 mV and IPSCs measured at 0 mV were decreased by DAMGO. C, DAMGO application inhibited both EPSCs and IPSCs (left) without significantly changing the paired-pulse ratio (right; 8 cells 6 animals). D, Representative example of serial recordings of EPSCs and IPSCs in a representative VPGlut(-) cell. Neither EPSCs measured at À65 mV nor IPSCs measured at 0 mV were modulated by the NK1R agonist GR73632. E, GR73632 application did not affect IPSC or EPSC amplitude (left) or paired-pulse ratios before and after drug application (right; 7 cells 4 animals). F, No significant change in E/I ratio after DAMGO. and GAD2(-) (Fig. 2C, left; F (1,161) ¼ 0.0539, p ¼ 0.871; cell type, two-way ANOVA). The maximum oIPSC amplitudes were also similar for responses to stimulation of D1-terminals and D2-terminals (Fig. 2C, right; p . 0.05, Mann-Whitney test). The paired-pulse ratios were similar as well, indicating no difference in release probability between D2-VPGABA and D2-VPGlutamate synapses (Fig. 2D, left; F (1,116) ¼ 2.568, p ¼ 0.118; afferent type, two-way ANOVA; Fig. 2D, right; p ¼ 0.567, Mann-Whitney test).
Effect of stimulating NK1 and l-opioid receptors on VGluT2(1) cells In addition to GABA, D1-MSNs corelease the neuropeptides substance P and dynorphin (activating NK1Rs and k opioid receptors, respectively), and D2-MSNs corelease enkephalin and neurotensin (activating m/d -opioid receptors and neurotensin receptors, respectively). The experimental procedure to determine whether the pharmacological activation of these receptors can differentially regulate inhibitory and excitatory neurotransmission onto VGluT2(1) is illustrated in Figure 3A. Fluorescent cells in VGluT2::Ai14 reporter mice were recorded, and afferents were stimulated electrically. AMPAR-mediated currents were biophysically isolated by clamping cells at the reversal potential for GABA A Rs (À65 mV), and GABA A R-mediated transmission was isolated by clamping cells at the reversal potential for AMPAR and NMDAR (0 mV). eEPSCs and eIPSCs were regularly blocked by DNQX and PTX, respectively (see example traces). The time course and response of one representative experiment with the MOR agonist DAMGO is illustrated in Figure 3B. The VGluT2(1) neuron was clamped at À65 mV to record EPSCs, and, after a stable baseline of 5 min, cells were slowly depolarized to 0 mV to record IPSCs. DAMGO was applied after a baseline recording of 5 min. After 10 additional minutes of recording at 0 mV, cells were clamped at -65 mV for at least 5 more minutes. With a set of control experiments, which followed the time course of depolarization and repolarization, we verified that the EPSCs amplitude of VPGlutamate cells was not significantly changed by prolonged depolarization (Extended Data Fig. 3-1A). Figure 3C shows that DAMGO application significantly depressed evoked IPSCs and EPSCs onto VGluT2(1) cells (Fig. 3C, left; t (5) ¼ 8.801, p , 0.01; t (5) ¼ 2.963, p , 0.05; Student's t test). The paired-pulse ratio of responses was evaluated with two paired pulses separated by a 100 ms interstimulus interval. Evoked IPSCs showed a trend toward increase in paired-pulse ratio after DAMGO application, and evoked EPSCs showed a significant increase (Fig. 3C, right; t (5) ¼ 2.023, p ¼ 0.1; t (5) ¼ 13.06, p , 0.01; paired t test), indicating that the inhibition of EPSCs was based on the decrease of glutamate release probability. Figure 3D shows the time course and the response of one representative experiment with the NK1R agonist GR73632, which was applied after a baseline recording of 5 min. IPSCs were not affected by GR73632 application while EPSCs were significantly decreased (Fig.  3E, left; t (8) ¼ 1.188, p ¼ 0.269; t (8) ¼ 3.758, p , 0.01; Student's t test). The significant increase in paired-pulse ratio indicates a decrease glutamate release probability after GR73632 application (Fig. 3E, right; t (5) ¼ 0.418, p ¼ 0.687; t (8) ¼ 2.753 p , 0.05; paired t test).
Next, we calculated the ratio of excitation to inhibition (E/I ratio) by dividing the average amplitude of EPSC (at À65 mV) by the average amplitude of IPSC (at 0 mV). Figure 3F illustrates the change in E/I ratio after neuropeptide application. Overall, GR73632 showed a strong trend toward a significant decrease of the E/I ratio compared with the DAMGO application ( Fig. 3F; p ¼ 0.06, Mann-Whitney test). These results imply that substance P released from D1 terminals could shift the balance from neurotransmission toward inhibition.

Effect of stimulating NK1Rs and l-opioid receptors on VGluT2(1) excitability
In previous experiments, we interrogated the effect of neuropeptides on GABA and glutamate transmission. Next, we wanted to quantify potential effects on VGluT2 (1) such as RMP and firing rate. To this aim, cells were recorded at resting membrane potential in the current clamp and every 20 s were depolarized with a ramp pulse of current increasing from 0 to 50 pA over 5 s (Fig. 5A,B, top, depolarizing ramp, middle panels, one representative baseline response). After a 5 min baseline, DAMGO or GR73632 was applied, and the impact on RMP and firing rate in response to the ramp depolarization was quantified for a 5 min baseline and 2-6 min after drug application ( Fig. 5A,B, bottom panels, a representative response 5 min after drug application). Neither drug changed RMP (Fig.  5C,D; DAMGO, p ¼ 0.250; GR73632, p ¼ 0.688; Wilcoxon test) or firing rate (Fig. 5E,F; DAMGO, p ¼ 0.109; GR73632, p ¼ 0.982; Wilcoxon test) of recorded VGluT2(1) cells.

Discussion
Cell type-specific GABAAR signaling in the ventral pallidum We found that the D1-MSN GABA release onto GABA A Rs of GAD2(-) cells was significantly weaker than on GAD2(1) cells. D2-MSN GABA A Rs mediated transmission was equally strong on GAD2(1) and GAD2(-) cells. The GABA release probability of D1-GAD2(-) synapses was significantly lower compared with D1-GAD2(1) synapses, indicating a presynaptic restriction of GABA release. These results are surprising because they demonstrate that, contrary to our expectations, D1-MSNs seem to exert higher inhibitory control over VPGABA cells than over VPGlutamate cells, at least at levels of low activity when neuropeptide release would be expected to be low (Francis et al., 2019; see Figure 6. MOR but not NK1R activation hyperpolarizes VPGABA cells. A, Representative voltage traces of a VPGlut(-) cell in response to a 50 pA current ramp before and after the DAMGO application illustrates the inhibition of firing. B, Representative voltage traces of a firing VPGlut(-) cell before and after GR73632 application show no change. C, Time course of change in RMP before, during, and after drug application. D, DAMGO but not GR73632 significantly hyperpolarized the RMP of VPGlut(-) cells. DAMGO, 9 cells from 5 mice; GR73632, 10 cells from 6 mice. E, Time course of change in firing rate before during ramp-like depolarization, before and after drug application. F, DAMGO but not GR73632 significantly decreased the elicited firing of VPGlut(-) cells. *p , 0.05. below). Since the strength of synapses is not static but can be modulated by experience, the weak D1-GAD2(-) synapses may serve as a substrate for future synaptic potentiation. Indeed, we and others have recently shown that withdrawal from chronic cocaine exposure selectively depresses D2-MSN synapses in the ventral pallidum via presynaptic MOR-dependent mechanisms (Creed et al., 2016;Heinsbroek et al., 2017). Furthermore, Creed et al. (2016) found that cocaine selectively potentiates D1-MSN synapses in the VP and that rectification of this plasticity inhibits cocaine sensitization. These results indicate that while the inhibitory control of D1-MSNS over VPGlutamate cells might be significantly lower in physiological conditions, it might become more important in disease states such as addiction and depression.
Our findings that a similar percentage of GAD2(-) cells receive D1-MSN and D2-MSN input seemingly contradicts a recent rabies tracing study from our laboratory that showed more D1-MSNs than D2-MSNs innervating VPGlutamate cells (Heinsbroek et al., 2020). It cannot be excluded that some rabies tropism toward D1-MSN existed, overestimating D1-MSN input onto VPGlutamate cells. Furthermore, while rabies tracing allows quantification of direct synaptic connections between two cells, it does not evaluate the strength of these connections. Some of these D1-MSN synapses may be weak or silent and not detected by whole-cell patch-clamp recordings. It could also be that some of these synapses are purely neuropeptidergic (Cifuentes and Morales, 2021). A similar noncanonical segregation between glutamate and dopamine release has also been demonstrated in the VTA (Cifuentes and Morales, 2021).

Cell type-specific neuropeptide signaling in the ventral pallidum
To examine how coreleased neuropeptides modulate neurotransmission and excitability of VPGABA and VPGlutamate, we focused on activating either MOR (via the MOR agonist DAMGO) or NK1R (via the NK1R agonist GR73632) because they are major targets of enkephalin and substance P in the VP (Kupchik et al., 2014;Vanderah and Sandweiss, 2015) and their action within the VP is implicated in motivated behaviors (Hasenöhrl et al., 2000;Smith and Berridge, 2005;Tang et al., 2005). We found that the administration of DAMGO inhibited GABAergic and glutamatergic afferents on both VPGABA and VPGlutamate to a similar extent. Interestingly, DAMGO hyperpolarized and inhibited action potential firing only in VPGABA cells. The DAMGO-induced hyperpolarization was accompanied by a significant decrease in input resistance, which indicates that the hyperpolarization is mediated by a canonical increase of potassium conductance (North et al., 1987). We furthermore found that GR73632 has the potential to downregulate the activity of VPGlutamate cells by selectively inhibiting glutamate transmission on this cell type (Fig. 7C,  D). The significant increase in paired-pulse ratio indicates a presynaptic locus of the NK1R-mediated inhibition of glutamate transmission. NK1R have been reported to be Gs-protein coupled and mainly expressed on cholinergic neurons in the VP (Mengual et al., 2008). The observed presynaptic depression of glutamatergic terminals could be accomplished via a heterosynaptic process involving acetylcholine release and presynaptically expressed Giprotein-coupled M2 receptors. Alternatively, presynaptic NK1R activation could trigger a nitric oxide synthasedependent LTD, as described for the cerebral cortex (Matsumura et al., 2021). Neuropeptides are stored in dense core vesicles, and neuropeptide release usually requires higher synaptic activity compared with GABA release (Francis et al., 2019). Hence the corelease of GABA and neuropeptides from the same synaptic terminals can multiplex the synaptic control over their postsynaptic target in an activity-dependent manner with GABAergic inhibition acting at low presynaptic activity and neuropeptidergic modulation acting at higher or sustained presynaptic activity. The potential of activity-dependent multiplexing was illustrated in an elegant study by Soares-Cunha et al. (2020), who found that D2-MSN stimulation elicits opponent behavioral responses, depending on the stimulation pattern. A brief optogenetic simulation of D2-MSNs inhibited VP neurons and conditioned place preference, whereas a prolonged stimulation elicited conditioned place aversion (CPA). CPA could be blocked by the administration of the d opioid receptor antagonist naltrindole (Soares-Cunha et al., 2020).

Limitations of this study and potential future directions
One caveat of this study is that for the afferent and cell type-specific experiments (Figs. 1, 2), we used an indirect approach using tomatocells GAD2(-) to identify VPGlutamate, which could result in not all GAD2(-) cells recorded being glutamatergic neurons. Cholinergic neurons constitute a small (,5%) group of neurons in the dorsolateral VP, where our recordings were made, and, similar to the striatum (Tepper and Bolam, 2004), cholinergic neurons in the VP can also be discriminated from other cell types because of their significantly larger soma size (Bengtson and Osborne, 2000). It is also possible that GABA cells expressing low levels of GAD2 or the genetic reporter could be mischaracterized as VPGlutamate. However, a recent RNA scope analysis by Stephenson-Jones et al. (2020) reveals negligible overlap between GAD2-expressing and VGluT2-expressing cells. Although we cannot completely rule out some mislabeling, similar to the common practice of identifying D1-MSNs or D2-MSNs based on a lack of reporter presence in mice expressing a reporter in one or the other subpopulation, it seems appropriate to assume that GAD2(-) cells are largely VPGlutamate.
Although there is only a small subpopulation of VP neurons that are not VPGlutamate and VPGABA cells, there are several distinct subpopulations of GABA neurons that can be subdivided based on molecular markers such as preproenkephalin or parvalbumin (Root et al., 2015;Knowland et al., 2017;Heinsbroek et al., 2020), or are based on their projection targets, and drive distinct affective and reward behaviors (Knowland et al., 2017;Vachez et al., 2021). While the investigation of these multiple subpopulations is beyond the scope of this article, the results from such studies will allow for a more complete understanding of how the accumbens-VP subcircuits modulate reward and aversion.
Another caveat of this study is that only synthetic analogs of the coreleased neuropeptides were used in relatively high doses. Accordingly, our experiments may not accurately reflect the in vivo release of enkephalin and substance P. Different affinities of MORs expressed on different synaptic terminals could allow for an even more nuanced tuning of the modulation of transmitter release. Emerging techniques to visualize in vivo neuropeptide release (Jullié et al., 2022) in combination with cell type-specific activity monitoring should be useful in more precisely understanding the effects of endogenous peptide release. While we only looked at the activation of MORs, enkephalin does also act on d opioid receptors, which are exclusively expressed on D2-MSN synapses in the ventral pallidum (Creed et al., 2016;Soares-Cunha et al., 2020) and likely also control information flow through the VP.
Our study focused on fast ionotropic GABA transmission via GABA A receptors, but metabotropic GABA B receptors are also present in the VP. However, GABA B expression is low and constrained to cell bodies of non-GABAergic neurons (Margeta-Mitrovic et al., 1999). Nonetheless, our study cannot exclude that GABA B receptors may play a cell type-specific role in modulating neurotransmission and excitability.
Finally, it should be noted that the action of different neuropeptides strongly depends on the VP subcompartment being studied. For example, MOR activation via DAMGO in the caudal VP elicits appetitive responses, whereas MOR activation in the more rostral regions elicits aversive responses (Smith and Berridge, 2005). Also, VP inactivation of the rostral VP inhibits cue-induced reinstatement, whereas inhibition of the caudal VP had no effect (Mahler et al., 2014). Because our recordings strictly focused on the subcommisural region (bregma between 0.3 and 0.14 mm), it will be important in future studies to investigate neuropeptidergic effects on neurotransmission and excitability along the rostrocaudal axis.

Conclusions
Overall, our findings show that D1-MSN and D2-MSN afferents in the VP have the potential to exert bidirectional control over VPGABA and VPGlutamate neurons with enkephalin released from D2-MSNS potentially exerting stronger inhibition and VPGABA (via an additional postsynaptic inhibition that is lacking on VPGlutamate cells; Fig.  7). Our data indicate that the effect of substance P on both cell types could also be activity dependent where GABA transmission is stronger on VPGABA under low activity and glutamate transmission onto VPGlutamate can be inhibited via substance P release under high activity.
Understanding the nuanced subcircuit regulation between the accumbens and VP is critical for integrating the emerging understanding that the two distinct subpopulations within each nucleus (accumbens and VP) differentially control whether an animal will approach or avoid a stimulus. Further experimentation based on our initial studies (as outlined above), will ultimately parse subcircuits based on neurotransmitter expression and projection patterns; thereby, potentially providing a rationale for therapeutic regulation in neuropsychiatric disorders where symptoms include maladaptive reward or avoidance behaviors, such as substance use disorders or post-traumatic stress disorder, respectively.