Striosomes control dopamine via dual pathways paralleling canonical basal ganglia circuits

Abstract

Balanced activity of canonical direct D1 and indirect D2 basal ganglia pathways is considered a core requirement for normal movement, and their imbalance is an etiologic factor in movement and neuropsychiatric disorders. We present evidence for a conceptually equivalent pair of direct D1 and indirect D2 pathways that arise from striatal projection neurons (SPNs) of the striosome compartment rather than from SPNs of the matrix, as do the canonical pathways. These striosomal D1 (S-D1) and D2 (S-D2) pathways target substantia nigra dopamine-containing neurons instead of basal ganglia motor output nuclei. They modulate movement with net effects opposite to those exerted by the canonical pathways: S-D1 is net inhibitory and S-D2 is net excitatory. The S-D1 and S-D2 circuits likely influence motivation for learning and action, complementing and reorienting canonical pathway modulation. A major conceptual reformulation of the classic direct-indirect pathway model of basal ganglia function is needed, as well as reconsideration of the effects of D2-targeting therapeutic drugs.

Keywords: SNpc; SNpr; SPNs; central zone of globus pallidus external; czGPe; direct D1 pathway; dopamine control; indirect D2 pathway; limbic system; movement modulation; striatal dopamine release; striatal projection neurons; striosomes; substantia nigra pars compacta; substantia nigra pars reticulata.

Figure 1.. Striosomal S-D1 and S-D2 populations form parallel pathways to dopamine-containing nigral neurons

A. Sagittal section from Pnoc-Cre;Ai14-tdTomato (magenta) transgenic mouse brain immunolabeled for CalDAG-GEFI (cyan). CPu: caudoputamen. B. Section from Nts-Cre;Ai14-tdTomato (magenta) transgenic mouse brain immunolabeled for CalDAG-GEFI (cyan). C. Pnoc-Cre;Ai14 mouse section double-labeled for Pnoc (magenta) and tyrosine hydroxylase (TH, cyan) to identify dopaminergic cell bodies and dendrites. Labeling of striosome-dendron bouquets was found in all 4 mice used for cell counts. D. Nts-Cre;Ai14 line F mouse double-labeled for Nts (magenta) and TH (cyan) to identify dopamine cell bodies and descending dendrites. Lack of Nts-Cre;Ai14-tdTomato labeling of the striosome-dendron bouquet was found in all 3 mice used for cell counts. Cell nuclei stained with DAPI (blue). E. Percentage of striosome-marker-positive SPNs double-labeled for either D1 or D2/A2a (Pnoc-Cre;Ai14 n = 4 mice, Nts-Cre;Ai14;D1-GFP n = 2 mice and Nts-Cre;Ai14;D2-GFP n = 1 mice). For each mouse, three coronal sections (anterior, mid-level and caudal) were evaluated. Counts made only for cells within striosomes identified by MOR1. Error bars represent SEM See also Figures S1, S2 and, S5, and Tables S1 and S2.

Figure 2.. Distribution of RVΔG-labeled neurons targeting PV-positive neurons in SNpr or dopamine-containing neurons in SNpc

A-F. DAT-Cre RV tracing. A. Injection site and protocol. B. Sagittal section. RV-EYFP-labeled neurons counterstained with anti-MOR1 (striosome marker, red), anti-CalDAG-GEFI (CDGI, matrix marker, blue) and DAPI (gray). RV-EYFP-labeled presynaptic neurons (green) are located preferentially in striosomes and czGPe. CPu: caudoputamen. STN: subthalamic nucleus. C. SNpc injection site (orange box in B), with starter neurons co-expressing the RV (green), V5 (red) and tyrosine hydroxylase (TH, blue). D. Striatal region (red box in B). Presynaptic neurons projecting to dopamine cells are localized preferentially in striosomes. E. GPe region (blue box in B) showing that presynaptic neurons projecting to dopamine cells are distributed preferentially in the central MOR1-rich, CalDAG-GEFI-poor zone. F. Proportions of RV-labeled striatal (total n = 1544 neurons in 15 sections from 4 DAT-Cre mice) and GPe (total n = 658 neurons in 6 sections from 3 DAT-Cre mice) neurons targeting dopamine neurons in the SNpc that were identified in striosomes (S) and czGPe, respectively. G-L. PV-Cre RV tracing. G. Injection site and protocol. H. Sagittal section depicting RVΔG-mCherry-labeled neurons (green) counterstained with anti-MOR1 (striosome marker, red), anti-CalDAG-GEFI (matrix marker, blue) and DAPI (gray). RVΔG-mCherry-labeled presynaptic neurons (green) are located preferentially in the matrix and the pzGPe. I. SNpr injection site (green box in H), depicting starter neurons co-expressing the RVΔG-mCherry (green) and PV (red) together with EBFP and EGFP from the two helper viruses. J. Striatal region (purple box in H), showing presynaptic neurons projecting to SNpr-PV cells localized preferentially in matrix. K. GPe region (yellow box in H), showing presynaptic neurons projecting to SNpr-PV cells localized preferentially in pzGPe, MOR1-poor/CalDAG-GEFI-rich zone. L. Proportions of RV-labeled striatal (total n = 1434 neurons in 12 sections from 3 PV-Cre mice) and GPe (total n = 305 neurons in 11 sections from 3 PV-Cre mice) neurons targeting PV neurons in SNpr located in striosomes and czGPe, respectively. See also Figures S3 and S5, and Tables S1 and S2.

Figure 3.. Striosomal SPNs target dopamine-projecting czGPe neurons

A. Terminals from N172-tdTomato-labeled striosomal neurons (red) on SNpc-dopamine-targeting czGPe neurons (RV-EGFP-labeled neurons, green). N = 6 mice. B. Close-up of the czGPe, showing SNpc-targeting czGPe neurons (RV-EGFP-labeled, green) surrounded by dense processes from N172-tdTomato-labeled striosomal neurons (red). C. Magnification of two SNpc-targeting czGPe neurons with contacts from N172 striosomal neuron processes. White lines indicate the plane of the Z-stack projection for yz and xy image coordinates. D. SYP-positive (blue) and N172 striosomal neuron (yellow) terminals on SNpc-targeting czGPe neurons (magenta). E. Magnification of cell indicated by square in D. F. Left: Representative images (top: same cells as in D, bottom: magnification of one cell) demonstrating the identification of boutons based on N172-positive/SYP-positive markers, as well as their sphericity and volume (see methods). Right: Quantification of colocalized terminals (4 sections) demonstrating that the majority of identified boutons are associated with RV-labeled czGPe neurons and that from these the majority of them were identified on their dendritic processes. RV: The percentage of N172-positive/SYP-positive boutons found on RV-labeled cells. S: The percentage of N172-positive/SYP-positive/RV-positive terminals found on the cell body of RV-labeled cells. D: The percentage of N172-positive/SYP-positive/RV-positive terminals found on the dendrites of RV-labeled cells. See also Figures S4 and S5, and Tables S1 and S2.

Figure 4.. S-D1 and S-D2 effects on dopamine release and motor kinematics

A. Protocol for targeting S-D1 (Pnoc) SPNs for optogenetic stimulation. B. Average dopamine response to optogenetic stimulation (8-sec train at 40 Hz, 10 trials) C. Average response before (Pre) and during stimulation. D. Trajectories of mice freely moving in open field box (30 × 30cm) before (green), during (red) and after (Post, blue) optogenetic stimulations of S-D1 SPNs. E. Trajectories before, during and after optogenetic stimulation for 10 trials reoriented to the axis defined by the base of the tail to the neck one frame before the stimulation. F. Distance traveled during and after the optogenetic stimulation relative to equivalent time-period before stimulation. G. Schematic of protocol for targeting S-D2 SPNs. H. Average response to optogenetic stimulation (8-sec train at 40 Hz, 10 trials) of S-D2 (Nts) SPNs. I. Average response before and during the stimulation. J. Behavioral effects of optogenetic manipulation of S-D2 SPNs in self-paced actions. Conventions as in D. K. Traces around the optogenetic stimulation, as in E. L. Difference of distance traveled as in F. M. Behavioral motif clusters in 3D space using UMAP. Twenty unique motifs were identified by the B-SOiD clustering analysis. Each color represents a different behavioral motif cluster. N. Box plot illustrating the ratio of motif usage between stimulation-on and stimulation-off periods for 8 Pnoc and 11 Nts mice. Average motif usage rates were measured in frames per sec, calculated separately for each condition and analyzed for statistical differences with two-tailed t-test across different conditions. O-Q. Left: Distribution density of head twist angular velocity, defined by the angle between the direction from the center of the body to the neck and from the neck to the snout (O), body twist angular velocity (P) and moving speed, quantified by tracking the center of the body (Q) from 8 Pnoc and 11 Nts mice. Data were aggregated from all mice and analyzed separately for periods of stimulation-on and stimulation-off. Right: Box plots illustrating the ratio of average angular velocities and moving speed during stimulation-on versus stimulation-off periods, calculated for each mouse and statistically compared. In all plots, *p < 0.05, **p < 0.01 and ***p < 0.001. See also Figure S7.

Figure 5.. Simultaneous imaging of dopamine release and S-D1 (Pnoc proxy) or S-D2 (Nts proxy) neuronal activity during a 2-choice probabilistic switching task

A. Protocol for simultaneous imaging of dopamine release and neuronal activity in DMS and DLS. B. Cross-section from a representative Nts;Flp mice illustrating fiber placements for simultaneous imaging of dopamine release and Nts neuronal activity in the right DLS and left DMS (left), and mapping of fiber placement for all Nts;Flp and Pnoc;Flp mice used (right). C. Diagram of the 2-choice probabilistic switching task. D. Frame from behavioral recording annotated with DeepLabCut. E. Example of actual mouse behavior in an expert session (blue crosses) versus the fitted trial-by-trial Q-learning model (green trace) during 140 trials with reward at right (no shading) or left (yellow shading) port. Dashed red lines mark the switch of rewarded ports. F. Fraction of choices for the right port plotted against the relative action value (Q-learning), illustrating decision-making strategies. G. Examples of photometry traces showing dopamine level and Pnoc or Nts activity recorded from the DMS and DLS during task performance. H. Average photometry traces (mean ± SEM) showing activity of dopamine, Pnoc and Nts recorded in the DMS (top) and DLS (bottom) during the task. Activity was averaged separately for ipsilateral and contralateral choice trials, and normalized for time. Shading indicates the initiation, turn and choice phases of the trial. I. Cross-correlation analysis between Pnoc or Nts neuronal activity and dopamine signaling in the DMS (top) or DLS (bottom). Pnoc (S-D1) and dopamine signals were anticorrelated with a prominent inverse peak for DMS and positive peak in DLS with a negative lag, suggesting dopamine fluctuations precede changes in S-D1 neuron firing rates by ~0.2 sec. By contrast, Nts (S-D2) neuronal activity and dopamine signaling showed a positive correlation for contralateral choice trials and for both ipsilateral and contralateral choice trials in DLS and a negative correlation with ipsilateral choice trials in DMS. Peaks in the Nts neuronal activity lead the increases in dopamine levels in DLS and lags in the DMS, with a lag time of ~0.4 sec. J. Task-bracketing index for Pnoc (S-D1), Nts (S-D2) and dopamine. Pnoc has a positive bracketing index in DMS and negative index in DLS. Dopamine has a negative bracketing index for both. No significant bracketing index was found for Nts neuronal activity ***p < 0.01. K. Pnoc and Nts activity do not dissociate rewarded and unrewarded outcomes. Dopamine shows a positive reward response only in the DMS. Dopamine release and Nts activity in DLS show increased activity in unrewarded trials (see Figure S7 for alignment to the port exit). Asterisk (*) indicates time points where the +R and −R difference was significant (p ≤ 0.05). See also Figure S7.

Figure 6.. Performance of neural activity pattern classification in predicting task behaviors, reward outcomes and action values across Pnoc (S-D1), Nts (S-D2) and dopamine release signals in DMS and DLS

A. Confusion matrices show the accuracy of an LSTM model in classifying task space events (initiation, left turn, right turn, left choice and right choice) based on Pnoc (top), Nts (middle) and dopamine (bottom) signals recorded in DMS (left) and DLS (right). Red and green squares indicate, respectively, ipsilateral and contralateral events relative to the recording site. Dashed yellow squares highlight task events identified with significantly higher accuracy than chance. Values are averaged across all mice. Except for dopamine release in the DMS, all signals show higher-than-chance accuracy for all task events (diagonal). Dopamine release in the DMS shows significantly higher-than-chance accuracy for initiation and only for contralateral turn and choice events. B. Confusion matrices showing the accuracy of an LSTM model in classifying outcomes at choice ports (air puff, left rewarded, left unrewarded, right rewarded and right unrewarded). The same color coding and highlighting are used as in A. Dopamine provides the best performance, but overall accuracy is lower compared to task events classification in panel A. C. Average signals of Pnoc, Nts and dopamine release during turn, contralateral choice and ipsilateral choice, split by high and low action value trials as calculated with a Q-learning model. Significant differences in amplitude were found at different time points between high and low value trials across all signals. Asterisk (*) indicates the time points where the difference between high-value and low-value signals was significant (p ≤ 0.05). D. Confusion matrices using the same method and color coding as A and B for classifying left high value choice, left low value choice, right high value choice and right low value choice. This panel shows that the Pnoc (S-D1) signal in both DMS and DLS carries action value information only for high value contralateral actions. Nts (S-D2) shows poor separation of high and low value trials, and although dopamine contains value information, its accuracy is relatively low. See also Figure S6.

Figure 7.. Dual direct-Indirect pathway model for basal ganglia output pathways modulating behavior

The striosome and matrix compartments are the sources of parallel D1-direct and D2-indirect pathways to the SNpc (striosomes) and to the SNpr (matrix). The matrix direct-indirect pathways modulate movement; the striosomal direct-indirect pathways, considering their links to the limbic system, could modulate mood state and motivation to act, collectively functions affected by basal ganglia disorders. SNpr-PV: substantia nigra pars reticulata parvalbumin-positive neurons; SNpc-dopamine: substantia nigra pars compacta dopamine-containing neurons.