Molecular and circuit determinants in the globus pallidus mediating control of cocaine-induced behavioral plasticity

Abstract

The globus pallidus externus (GPe) is a central component of the basal ganglia circuit that acts as a gatekeeper of cocaine-induced behavioral plasticity. However, the molecular and circuit mechanisms underlying this function are unknown. Here, we show that GPe parvalbumin-positive (GPe^PV) cells mediate cocaine responses by selectively modulating ventral tegmental area dopamine (VTA^DA) cells projecting to the dorsomedial striatum (DMS). Interestingly, GPe^PV cell activity in cocaine-naive mice is correlated with behavioral responses following cocaine, effectively predicting cocaine sensitivity. Expression of the voltage-gated potassium channels KCNQ3 and KCNQ5 that control intrinsic cellular excitability following cocaine was downregulated, contributing to the elevation in GPe^PV cell excitability. Acutely activating channels containing KCNQ3 and/or KCNQ5 using the small molecule carnosic acid, a key psychoactive component of Salvia rosmarinus (rosemary) extract, reduced GPe^PV cell excitability and impaired cocaine reward, sensitization, and volitional cocaine intake, indicating its therapeutic potential to counteract psychostimulant use disorder.

Keywords: behavioral vulnerability; carnosic acid; chemogenetics; cocaine; dopamine; drug abuse; globus pallidus; intrinsic excitability; rabies virus; ventral tegmental area; voltage-gated potassium channels.

Figure 1:. Linking the GPe and VTA^DA→NAcLat cells.

(A) Schematic of DREADD inhibition experiments of five projection-defined ventral midbrain DA neuron subpopulations. 1) CAV-FLEx^loxP-Flp was injected into one of five DA neuron target sites. Then, 2) AAV-FLEx^FRT-hM4Di-mCherry, or AAV-FLEx^FRT-YFP as a control was injected into the VTA/SNc. Two weeks later, mice were tested for cocaine CPP, sensitization, and withdrawal anxiety-related behavior (15 mg/kg cocaine). CNO 5 mg/kg was given i.p. 30 minutes before each cocaine administration. (B) Image of the ventral midbrain of a mouse where CAV-FLEx^loxP-Flp was injected into the NAcLat and AAV-FLEx^FRT-hM4Di-mCherry into the VTA. Green = tyrosine hydroxylase (TH), a marker for DA neurons. Red = mCherry. Bottom, high magnification images of overlap between mCherry and TH. (C-D) CPP scores and time spent in each chamber for each group of animals with saline or cocaine, and hM4Di or YFP treatments. One-way ANOVA p = 0.0065; pairwise t-tests: YFP saline vs. YFP cocaine p = 0.0054; YFP cocaine vs. hM4Di cocaine p = 0.049. Note that one data point (YFP saline) is below the y-axis cutoff. Numbers of mice for each group are reported below the x-axis. (E-F) Sensitization scores for each group of animals with saline or cocaine, and hM4Di or YFP treatments. One-way ANOVA p = 0.016; pairwise t-tests: YFP cocaine vs. hM4Di cocaine p = 0.014. (G) Schematic of slice electrophysiology experiments testing the effect of GPe^PV cell stimulation on VTA^DA→NAcLat cell activity. AAV-FLEx^FRT-ChR2-YFP was injected into the GPe, AAV-FLEx^loxP-GFP was injected into the VTA, and red retrobeads were injected into the NAcLat of DAT-Cre::PV-Flp mice. Acute slice preparations were prepared one month later, and recordings were performed in current clamp from GFP+, retrobead+ cells in the VTA, with or without 20 Hz stimulation of 473 nm light. Shown below are representative traces from the same VTA^DA→NAcLat cell with (blue) and without (black) 20 Hz optogenetic stimulation of GPe^PV fibers in slice. (H) Cumulative probability plot for the latency to first spike. (I) Latency to first spike from the same cells, with light off and light on, p = 0.0007. (J) Schematic of experiments of output mapping of five projection-defined ventral midbrain DA neuron subpopulations. CAV-FLEx^loxP-Flp was injected into one of five ventral midbrain DA neuron target sites, and AAV-FLEx^FRT-mGFP was injected into the VTA/SNc. Experiments were terminated two months later. (K) Sample images of the DMS for each target injection, indicating that injections targeting the NAcLat resulted in the most GFP-expressing axons terminating in the DMS, consistent with the highly unique projections of each DA cell population and demonstrating specificity of injections. (L) The percentage of each brain region covered by axons from the corresponding ventral midbrain DA neuron subtype. n=4 for each condition. (M) Experimental schematic of chemogenetic inhibition of VTA^DA→NAcLat collaterals in the DMS. CAV-FLEx^loxP-Flp was injected into the NAcLat, and AAV-FLEx^FRT-hM4Di-mCherry or AAV-FLEx^FRT-hM4Di-YFP was injected into the VTA. One month later, experiments were initiated, including an injection of CNO microspheres into the DMS, following by cocaine injections testing the effects on CPP and locomotor sensitization. The shortened behavioral protocol was used as in Figure S1B. (N) Sample image of the VTA indicating hM4Di expression from VTA^DA→NAcLat cells. (O-P) Inhibition of VTA^DA→NAcLat collaterals in the DMS prevented cocaine CPP, p = 0.018. (Q-R) Inhibition of VTA^DA→NAcLat collaterals in the DMS prevented cocaine-induced locomotor sensitization, p = 0.027.

Figure 2:. Cocaine alters RABV-mediated input labeling onto GPe^PV cells and reduces spontaneous inhibitory input onto GPe^PV cells.

(A) Experimental schematic of RABV input mapping from GPe^PV cells. A mixture of AAV-FLEx^loxP-TC66T and AAV-FLEx^loxP-RABV-G was injected into the GPe of PV-Cre mice. Thirteen days later, mice received a single injection of 15 mg/kg cocaine, or saline. EnvA-pseudotyped rabies virus (RABV) was injected into the GPe on Day 15. Experiments were terminated on Day 20. (B) Percentage of total RABV-labeled inputs from 19 different input sites to GPe^PV cells. Two-way ANOVA drug effect p >0.99, interaction effect p = 0.006. n = 4 saline, 6 cocaine. (C) Sample images from the DMS from saline- and cocaine-treated mice. (D) The ratio of RABV-labeled inputs from putative excitatory cells (cortex/thalamus/subthalamic nucleus (STN)) to putative inhibitory inputs (DMS/DLS) was higher in cocaine-treated mice, p = 0.04. Two points were excluded from the cocaine group for this analysis as outliers (more than 2 standard deviations outside of the mean, here one above and one below the mean). (E) Experimental schematic of chemogenetic activation of DMS^D2 cells. AAV-FLEx^loxP-hM3Dq-mCherry, or AAV-DIO-YFP as a control, was injected into the DMS. Two weeks later, cocaine CPP and sensitization were tested, with CNO being injected i.p. 30 minutes before cocaine. (F) Sample image of hM3Dq-mCherry expression in the DMS of A2a-Cre mice, p = 0.0049. (G-H) Chemogenetic activation of DMS^D2 cells prevented cocaine CPP, p = 0.025. (I-J) Chemogenetic activation of DMS^D2 cells prevented cocaine-induced locomotor sensitization. (K) Schematic of experiment. AAV-DIO-YFP was injected into the GPe of PV-Cre mice. Four weeks later, mice received a single administration of either saline or 15 mg/kg cocaine. The following day, acute slices were cut, and whole-cell patch clamp recordings were conducted from YFP+ (GPe^PV) cells, in voltage clamp mode. (L) Cumulative probability plot of the amplitude of individual spontaneous IPSCs in saline- vs. cocaine-treated mice. (M) Comparison of IPSC amplitude in saline- vs. cocaine-treated mice, p = 0.040. (N) Comparison of IPSC frequency in saline- vs. cocaine-treated mice, p = 0.016. (O) Cumulative probability plot of the amplitude of individual spontaneous EPSCs in saline- vs. cocaine-treated mice. (P) Comparison of EPSC amplitude in saline- vs. cocaine-treated mice, p = 0.08. (Q) Comparison of EPSC frequency in saline- vs. cocaine-treated mice, p = 0.14.

Figure 3:. Recording activity of each node of the DMS^D2→GPe^PV→SNr^GABA→VTA^DA→NAcLat pathway.

(A) Proposed pathway of basal ganglia circuit being studied. Cocaine evokes DA release from VTA^DA→NAcLat collaterals in the DMS. This reduces activity in DMS^D2 cells that project to GPe^PV cells, increasing their activity. This increase then elevates inhibition onto SNr^GABA cells, reducing their inhibition. This in turn disinhibits VTA^DA→NAcLat cells. (B) Schematic of experiment. GCaMP was expressed in each of the four targeted cell populations. GCaMP activity was recorded for 30 minutes in the open field on Day 1 following a saline injection, Day 2 following an injection of 15 mg/kg cocaine, and on day 3 following a saline injection. Five days later, mice were run through CPP and sensitization protocols. (C-N) Schematic, GCaMP injection/fiber implantation sites for each of the four conditions, sample traces from Days 1–3, and percentage of time spent above the activity threshold for each population on Days 1 and 3. Shown are data from DMS^D2 cells (C-E; p = 0.025), GPe^PV cells (combined GPe^PV→ventral midbrain and GPe^PV, (F-H; p = 0.005), SNr^GABA cells (I-K; p = 0.033), and VTA^DA→NAcLat terminals (L-N; p = 0.004). Individual comparison for GPe^PV and GPe^PV→ventral midbrain cells are shown in Figure S9. (O-R) Peri-stimulus time histograms (PSTH) for DMS^D2 cells (O, n = 9), GPe^PV→ventral midbrain cells (P, n = 9), SNr^GABA cells (Q, n = 10), and VTA^DA→NAcLat terminals (R, n = 10) during chamber crossings in the CPP pre-test and post-test, as well as locomotor initiation and cessation in the open field. Locomotion recordings were taken from the second day of habituation (Day 13, panel B). A comparison of GPe^PV and GPe^PV→ventral midbrain cells is shown in S4N-O.

Figure 4:. Relationship of GPe^PV cell activity relates to subsequent CPP and sensitization, and cocaine-induced downregulation of Kcnq3 and Kcnq5 in GPe^PV cells.

(A) Schematic of fiber photometry experiments. Mice were the same as those used in Figure 3. (B-C) The spontaneous activity of GPe^PV cells one day preceding and one day following cocaine administration was highly correlated with subsequent CPP (D; D1 p = 0.006; D3 p = 0.016, n = 14) and sensitization (E; D1 p = 0.004, D3 p = 0.002, n = 18). Data from GPe^PV→ventral midbrain and GPe^PV cells generally were combined in panels D and E as all data (e.g., Figure 3, S9, and our previous work) indicated these were largely the same cell populations. Data from GPe^PV→ventral midbrain cells are shown as large dots, and GPe^PV cells as small dots. (D) UMAP plot of cells identified in snRNA-seq experiments from the GPe. Four distinct clusters of GABAergic neurons were identified along with astrocytes, oligodendrocytes, microglia, and oligodendrocyte precursor cells (OPCs). (E) Heatmap of gene expression for known marker genes that define cell types. (F) Volcano plot of differentially expressed genes in cocaine- vs. saline-treated mice. Many more genes were downregulated after cocaine administration than were upregulated. This includes the voltage-gated potassium channel Kcnq3. Both Kcnq3 and Kcnq5 are shown. (G) Violin plots of Kcnq3 and Kcnq5 expression in each GABAergic cluster identified in panel D. Expression of Kcnq3 was significantly downregulated in GABAergic cluster 2 following cocaine (log₂FC = −0.22, p-val adj = 0.0013); all other comparisons were not statistically significant.

Figure 5:. Carnosic acid reduces GPe^PV cell excitability through opening KCNQ3/5 channels.

(A) Schematic of dual voltage clamp setup in Xenopus oocytes to test carnosic acid effects on select members of the Kcnq gene family. (B) Mean traces showing KCNQ currents as indicated, +/− carnosic acid (voltage protocol inset), expressed in Xenopus oocytes and measured by two-electrode voltage clamp; n = 8 per group for all panels. KCNQ3* = KCNQ3-A315T, a mutant that permits homomeric KCNQ3 to pass robust currents. KCNQ2 was recording starting at −80mV, and KCNQ3* starting at −120, given the large effect of carnosic acid on KCNQ3*. (C) Peak current for KCNQ2 and KCNQ3*. (D) Conductance (G/Gmax) for KCNQ2 and KCNQ3*. (E) Membrane voltage for each condition with and without carnosic acid application. KCNQ2, p-value = 0.55; KCNQ3*, p < 0.0001. (F) Schematic of experiments to test the effects of carnosic acid on GPe^PV cell excitability. (G) Carnosic acid caused GPe^PV cells to fire fewer action potentials across a range of current injections. Two-way repeat measures ANOVA condition effect, vehicle vs. carnosic acid p = 0.033). n = 7 cells for vehicle, 10 for carnosic acid. Shown are sample traces from each condition following a 50 pA current injection. (H) The rheobase, defined as the minimum current required to elicit a single action potential, was higher in carnosic acid-treated cells, p = 0.020. (I) Sample image of a recorded, biocytin-filled cell co-staining with HA, which was tagged to Cas9. (J) Carnosic acid application to slices where Kcnq3 and Kcnq5 were deleted from GPe cells using CRISPR had no significant effect on GPe^PV cell excitability, and in fact, slightly increased it. Two-way repeat measures ANOVA condition effect, vehicle vs. carnosic acid p = 0.12, n = 7 cells for vehicle, 10 for carnosic acid. Shown are sample traces from each condition following a 50 pA current injection. (K) Carnosic acid application did not alter the rheobase of cells lacking Kcnq3 and Kcnq5, p = 0.54, n = 6 cells each.

Figure 6:. Carnosic acid reduces cocaine CPP and locomotor sensitization through KCNQ3/5 in the GPe.

(A) Schematic for behavioral experiments where the behavioral effects of carnosic acid were tested. (B-C) i.p. administration of 2.5 mg/kg carnosic acid 30 minutes prior to cocaine administration blocked cocaine CPP, p = 0.034. (D-E) i.p. administration of 2.5 mg/kg carnosic acid 30 minutes prior to cocaine administration blocked cocaine-induced locomotor sensitization, p < 0.0001. One point for the carnosic acid group is located below the axis. (F) i.p. administration of 2.5 mg/kg carnosic acid had no significant effect on locomotion as assessed in a 30-minute locomotor test with carnosic acid only (no cocaine), p = 0.06. (G-H) Local GPe infusion of 10 μM carnosic acid prior to cocaine administration blocked cocaine CPP, p = 0.045. (I) Local GPe infusion of 10 μM carnosic acid prior to cocaine administration had no significant effect on locomotion as assessed in a 30-minute locomotor test with carnosic acid only (no cocaine), p = 0.99. (J) Schematic of experiments testing the effects CRISPR-based knockdown of Kcnq3 and Kcnq5 on cocaine-induced behaviors. (K) Verification of Cas9 expression in the GPe, as assessed by immunostaining for HA. (L) Control animals receiving carnosic acid did not form a significant CPP (p = 0.23), whereas animals in which Kcnq3 and 5 were knocked out in GPe cells did form a significant CPP (p = 0.044). n = 13 and 14, respectively. (M-N) Animals in which Kcnq3 and 5 were knocked out in GPe cells showed an elevated locomotor response to cocaine, as well as enhanced sensitization relative to control mice, p = 0.046. n = 14 for each.

Figure 7:. Carnosic acid reduces volitional cocaine intake.

(A) Timeline of cocaine IVSA experiments. (B) Number of active lever presses over the twelve-day acquisition period for animals treated with vehicle, or one of two carnosic acid concentrations (Two-way ANOVA, group effect p < 0.0001; vehicle vs. low CA p < 0.0001; vehicle vs. high CA p = 0.0019). n = 15 for vehicle group, 8 for CA-treated groups. (C) Carnosic acid-treated mice showed a significantly lower escalation in cocaine intake over the course of the twelve-day self-administration session (One-way ANOVA p = 0.01; vehicle vs. low CA p = 0.023; vehicle vs. high CA p = 0.021). (D) Low concentrations of carnosic acid significantly reduced the preference for pressing the active lever, while higher concentrations of carnosic acid had no effect (Two-way ANOVA group effect p < 0.0001; vehicle vs. low CA p < 0.0001; vehicle vs. high CA p = 0.11). (E) No significant effects of carnosic acid on inactive lever pressing were observed (Two-way ANOVA group effect p = 0.23).

Figure 8:. Schematic of key findings from this study.

We implicated a closed loop circuit consisting of DMS^D2→GPe^PV→SNr^GABA→VTA^DA→NAcLat cells – the latter with collaterals to the DMS – in cocaine-induced behavioral plasticity. The DMS, GPe, and SNr connections are all GABAergic, and control DAergic output of VTA^DA→NAcLat cells, which mediate cocaine reward. Following cocaine, spontaneous activity in the GPe^PV and VTA^DA→NAcLat (DMS) cells increases, leading to elevations in DA signaling in the striatum. This elevation in GPe^PV cell activity is mediated both by decreases in inhibitory drive from DMS^D2 cells as well as a downregulation in the expression of Kcnq3 and Kcnq5 in GPe^PV cells. As baseline GPe^PV cell activity is linked to behavioral responses to subsequent cocaine administrations, GPe^PV cell activity can be viewed as a barometer of cocaine sensitivity. Carnosic acid, a KCNQ3/5 channel opener, can prevent cocaine reward when administered systemically or within the GPe, and can interfere with cocaine self-administration behaviors. The hypothesized mechanism of action is through inhibition of GPe^PV cells, which prevents cocaine-induced behavioral plasticity.