Role of posterior medial thalamus in the modulation of striatal circuitry and choice behavior

Abstract

The posterior medial (POm) thalamus is heavily interconnected with sensory and motor circuitry and is likely involved in behavioral modulation and sensorimotor integration. POm provides axonal projections to the dorsal striatum, a hotspot of sensorimotor processing, yet the role of POm-striatal projections has remained undetermined. Using optogenetics with mouse brain slice electrophysiology, we found that POm provides robust synaptic input to direct and indirect pathway striatal spiny projection neurons (D1- and D2-SPNs, respectively) and parvalbumin-expressing fast spiking interneurons (PVs). During the performance of a whisker-based tactile discrimination task in head-restrained mice, POm-striatal projections displayed learning-related activation correlating with anticipatory, but not reward-related, pupil dilation. Inhibition of POm-striatal axons across learning caused slower reaction times and an increase in the number of training sessions for expert performance. Our data indicate that POm-striatal inputs provide a behaviorally relevant arousal-related signal, which may prime striatal circuitry for efficient integration of subsequent choice-related inputs.

Keywords: electrophysiology; fluorophotometry; mouse; neural circuitry; neuroscience; pupil; striatum; thalamus.

Figure 1.. POm equally innervates striatal cell types with faster latency in PV interneurons.

(A) Schematic detailing pAAV-ChR2-EYFP injection unilaterally into POm (Right), and optogenetic stimulation of POm-striatal afferents whilst recording from identified and unidentified neurons via ex vivo slice of posterior DLS (AP range: –0.34 to –1.22 relative to Bregma; Left). See Figure 1—figure supplement 1. Illumination (2.5ms pulses of 470 nm light,~0.6 mW intensity) was delivered through the 40 x objective. (B) Representative injection site (orange) in POm (Left), and viral spread of all electrophysiology injections within highlighted POm (purple; Right). S1BF = S1 Barrel Field. Scale = 1 mm. (C) Red box inset from panel (B) highlighting stereotypical POm-cortical projection pattern to S1BF L1 and L5a (Zhang and Bruno, 2019; La Terra et al., 2022; Ohno et al., 2012). Right: POm-striatal axons within posterior DLS. CC = corpus callosum. Scale = 200 µm. (D) Representative cell type-specific PSPs to SP stimulation. Colored lines = average PSP of 20 sweeps. Gray lines = 20 individual traces. Solid vertical and dashed horizontal lines = latency and amplitude, respectively. Red dashed line = 0 mV. Blue tick = photostimulation (PS). Time scale = 10ms. Voltage scale = 4 mV. (E) Amplitudes evoked by each cell type were similar (D1-SPNs=20 cells from 6 mice, D2-SPNs=11 cells from 5 mice, PVs = 17 cells from 7 mice, unidentified SPNs = 7 cells from 4 mice). Inset shows grand average PSPs. Time scale = 10ms. Voltage scale = 2 mV. (F) Latency to maximum PSP amplitude is significantly quicker in PVs than all other cell types. (G–H) Representative responses of (G) D1-SPN (Top) and D2-SPN (Bottom), and (H) PV (Top) and putative SPN (Bottom) to PPR stimulation (Assous and Tepper, 2019). PPR is defined as the ratio of PSP amplitude of pulse 2 over the ratio of PSP amplitude of pulse 1. PPR PS parameters = five 2.5ms pulses with 50ms interpulse intervals (20 Hz). Time scale = 100ms. Voltage scale = 2 mV. See Figure 1—figure supplement 2I Stimulation of POm-striatal afferents evokes similar PPR responses. (J–K) Representative responses of (J) D1-SPN (Top) and D2-SPN (Bottom), and (K) PV (Top) and putative SPN (Bottom) to train stimulation (Landisman and Connors, 2007). Colored lines = average of five individual gray traces. Train PS parameters = thirty 2.5ms pulses with 64.2ms interpulse intervals (15 Hz). Time scale = 1000ms. Voltage scale = 2 mV. (L) Relative PSP amplitude (average of pulses 5–15 compared to pulse 1) is significantly larger than both SPNs. Data are mean ± SEM. *p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001.

Figure 1—figure supplement 1.. Recorded cell location, intrinsic electrophysiological parameters, and sequentially patched PSP amplitude and latency.

(A–D) Representative responses of (A) D1-SPN (salmon), (B) D2-SPN (green), (C) PV interneuron (dark gray), and (D) unidentified SPN (orange) to hyperpolarizing and depolarizing current injections. Vertical voltage scale = 40 mV. Horizontal current scale = 100ms. Vertical current scale = 50 pA. (E) Recording location schematic of all recorded cells including cells excluded due to action potentials (lightning bolt). Note = all recordings took place within posterior DLS as it is the only striatal region innervated by POm axonal projections [S1, S2]. (F) Small but significant correlation (r²=0.08493; p=0.0293) between PSP amplitude and increasing distance from the injection site (AP = –2.05) excluding cells with action potentials. (G–L) Parameters are consistent with literature concerning both SPNs and PV interneurons: (G) Mean resting membrane potential (RMP; D1-SPN: –86.80±0.58 mV; D2-SPN: –86.00±0.69 mV; PV: –82.55±0.76 mV; SPN: –87.88±0.52 mV) (F_(4,56) = 11.24, p<0.0001, D1 vs. PV p<0.0001, D2 vs. PV p=0.0079, SPN vs. PV p=0.0001), (H) Mean input resistance (Rin; D1: 74.97±3.60 MΩ, D2: 89.25±8.25 MΩ, PV: 105.00±6.51 MΩ, SPN: 84.58±6.16 MΩ) (F_(4,56) = 14.27, p=0.0026, D1 vs. PV p=0.0011, (I) Mean maximum frequency (D1: 26.34±2.95 Hz, D2: 38.38±3.46 Hz, PV: 134.00±10.07 Hz, SPN: 38.00±7.25 Hz) (F_(4,56) = 42.64, p<0.0001, D1 vs. PV p<0.0001, D2 vs. PV p=0.0032, SPN vs. PV p=0.0027), (J) Mean half-height width (HHW; D1: 0.99±0.02ms, D2: 1.01±0.04ms, PV: 0.38±0.02ms, SPN: 1.26±0.14ms) (F_(4,56) = 42.06, p<0.0001, D1 vs. PV p<0.0001, D2 vs. PV p<0.0001, SPN vs. PV p<0.0001), (K) Mean 90% rise time (D1: 2.42±0.08ms, D2: 2.56±0.12ms, PV: 1.14±0.05ms, SPN: 3.00±0.32ms) (F_(4,56) = 40.52, p<0.0001, D1 vs. PV p<0.0001, D2 vs. PV p<0.0001, SPN vs. PV p<0.0001), and (L) Mean afterhyperpolarization (AHP) amplitude (D1: –11.02±0.91 mV, D2: –9.79±1.05 mV, PV: –20.39±0.73 mV, SPN: –7.64±1.05 mV) (F_(4,56) = 38.28, p<0.0001, D1 vs. PV p<0.0001, D2 vs. PV p<0.0001, SPN vs. PV p<0.0001). (M–O) Identified and unidentified neurons were patched sequentially and tested using the same protocols (SP, PPR, and Train) to control for injection site variability. (M) Representative PSP responses of a sequentially patched D1- and D2-SPN pairing (n=8 D1-D2-SPN sequential recordings from five mice). Time scale = 10ms. Voltage scale = 2 mV. (N) No significant differences in PSP amplitude between sequentially patched D1-SPN and D2-SPN pairs. (O) No significant differences in latency between sequentially patched D1-SPN and D2-SPN pairs. (P) Same as in M) for recording acquired from sequentially patched PV and SPN pairings (n=9 PV-SPN sequential recordings from seven mice). Time scale = 10ms. Voltage scale = 2 mV. (Q) No significant difference in PSP amplitude between sequentially patched PV and SPN pairs. (R) Significant differences in latency between sequentially patched PV and SPN pairs (6.85±0.42 for PV vs. 10.75±0.87 for SPN, p=0.0006, n=9 pairs). Data are mean ± SEM. ** p<0.01, *** p<0.001, **** p<0.0001.

Figure 1—figure supplement 2.. Representative responses of identified striatal cells to PPR and train stimulation.

(A) Box-and-whisker plot of the normalization of the second PSP to the first PSP of all recorded D1-SPNs (salmon), D2-SPNs (green), PV interneurons (dark gray), and unidentified SPNs (orange). (B) Summary plots of all PSP amplitudes normalized to the first pulse for all recorded D1-SPNs, D2-SPNs, PV interneurons, and unidentified SPNs. Vertical black lines indicate the averaging within from pulse 5 to pulse 15. Data are mean ± SEM.

Figure 2.. Mice rapidly learn to discriminate between two textures, and all three activity parameters markedly increase across learning.

(A) Schematic detailing pAAV-hSynapsin1-axon-jGCaMP8s-P2A-mRuby3 injection unilaterally into POm (Right), and a 400 µm cannula implanted in the left posterior DLS (Left). (B) Representative injection site in POm (Left), and cannula placement in the posterior DLS along with ascending POm axons (Right). Scale = 1 mm. (C) Top: Stimulating timing and texture movement representation during a trial. Note = both LEDs (isosbestic = 405 nm; axon-jGCaMP8s=470 nm) were constantly on for every session. Bottom: Outcomes for each stimulus-response pair. (D) Schematic representing texture movement and potential outcomes during a single trial of the Go/NoGo whisker discrimination paradigm. (E) Changes in Hit Rate, FA Rate, Sensitivity (d’) and Bias of the FP cohort (n=5 mice) as they transition from the Learning to the Expert phase. Note that mice are classified as Expert when they achieve a Hit Rate ≥0.80 and a FA Rate ≤0.30 for two consecutive sessions. Red line = 0. See Figure 2—figure supplement 1. (F) Average number of sessions required for expert discrimination of the FP cohort. (G–K) Three activity parameters (licking, axonal calcium, and pupil activity) from a representative (G) Shaping (session 1), (H) Early Learning (first two sessions after Shaping), (I) Late Learning (last two sessions before Expert), (J) Expert, and (K) Reward sessions from the same mouse (FPOm-18). Top: licking activity within a session (150 trials). Colored ticks = lick. Vertical black line = sound cue representing trial start as the texture moves toward the whisker field. Vertical red lines = start (texture arrival at endpoint in whisker field) and end (texture departure toward starting point) of the PT window (time where mice can respond by licking). Vertical brown line = average reaction time (RT; time of first lick that triggers an outcome) across all trials in each session. Colored boxes = 500ms grace period (licking does not trigger any outcomes). Note = no response line is present in the Reward session (K) as licking does not trigger any outcomes, and water was automatically delivered at PT end. Top Middle: Lick histogram. Middle: Heatmap sorted by trial outcome (to the Right of heatmap) highlighting axonal ZMAD calcium activity for each trial. Trial outcome is color coded (blue = Hit, yellow = Miss, orange = FA, brown = CR). Bottom Middle: Average axonal calcium activity of 150 trials for each session. Bottom: average pupil area (as a percentage) of 150 trials for each session. Data are mean ± SEM. Time scale = 1 s.

Figure 2—figure supplement 1.. Individual longitudinal learning-related changes in behavioral parameters, licking activity, and calcium activity, and methodology of measuring pupil dynamics.

(A) Longitudinal learning-related changes as mice progress through each time point (Shaping, Early Learning, Late Learning, and Expert). The Shaping time point permitted mice to learn to lick the spout for a water reward during the first three sessions without the presence of the textures. During the last two Shaping sessions, the Go and NoGo textures were presented simultaneously with a Go texture probability of 0.90 for the first session and 0.75 for the second session. For the first Learning session, and all subsequent sessions, the Go texture probability was set to 0.50. Mice were considered Learning at this time point. The Early Learning time point was considered the first two sessions after Shaping, while the Late Learning was considered the last two sessions before achieving Expert status. Mice were considered Expert when they had a Hit Rate ≥0.80 and a FA Rate ≤0.30 for two consecutive sessions. After discrimination training was completed, mice were subjected to a Reward session. During the Reward session, the lick thresholds were unobtainable, the textures were oriented so they could not be contacted by the whiskers, and water was automatically delivered at the end of the PT window (second vertical red line). Left: Hit Rate (black) and FA Rate (gray) across learning. Left Middle: Sensitivity (d’; black) and Bias (gray) across learning. Red dashed line indicates 0. Note that a strict d’ threshold was not used due to artificially increased d’ values as Hit Rate and/or FA Rate approached their extremes (0 or 1) as in FPOm-18 sessions 8–15. Right Middle: Probability density function for licking activity at each time point. Right: Average axonal ZMAD calcium activity at each time point. Scale bar = 1 s. (B) To measure pupil dynamics during the Go/NoGo discrimination task, orofacial video was synchronously recorded with calcium activity. This video was cropped in DeepLabCut [S3, S4], and nine markers were manually placed within the video: eight circumscribing the pupil (top, top right, right, bottom right, bottom, bottom left, left, and top left) and one on the spout. Once the model was trained, an ellipse was fitted through the eight pupil markers in sci-kit [S5]. Pupil values were converted from pixels to percentage, and the baseline was normalized to 0. Notably, pupil-related dynamics (e.g. Shaping, Early Learning and Expert) during the trial outcomes (e.g. Hit, Miss, False Alarm, and Correct Rejection) are similar to previously observed dynamics [S6] despite using a deep learning-based methodology [S7]. Data are mean ± SEM. Scale bar = 1 s.

Figure 3.. All three activity parameters exhibit marked increases across learning, but only axonal calcium activity remains unchanged, irrespective of trial type or outcome segmentation.

(A) Grand average axonal calcium activity at each behavioral time point. Data are mean ± SD. (B) Average of maximal axonal calcium amplitude markedly increases across learning before regressing to Shaping levels during the Reward session. ᴨ p<0.01 Shaping vs. Early, ß p<0.001 Shaping vs. Late, # p<0.001 Shaping vs. Expert, ¶ p<0.05 Early/Late vs. Reward, † p<0.001 Expert vs. Reward. (C) Average area under the curve of the receiver-operator characteristic (auROC) also markedly increases across learning before regressing to Shaping levels during the Reward session. ^ p<0.05, Shaping vs. Early;+p < 0.01, Shaping vs. Late; £ p<0.001, Shaping vs. Expert; Ω p<0.05, Early vs. Expert; @ p<0.01, Late/Expert vs. Reward. (D) Grand average probability density function for licking-related activity at each time point. (E) Axonal calcium activity at Learning and Expert time points 2 s pre and 2 s post grand average RT. Data are mean ± SD. (F) Pre-RT axonal calcium activity is significantly larger than post-RT axonal calcium activity. (G) Grand average of normalized pupil area at each behavioral time point. (H) Representative cross-correlation of pupil area and axonal calcium activityI Cross-correlation of pupil area and calcium activity plotted for each behavioral time point for each mouse. (J) Grand average of all licking (Top), calcium (Middle), and normalized pupil (Bottom) activity segmented by trial type: Go texture (Left) and NoGo texture (Right) presentation. (K) Average of maximal axonal calcium amplitude markedly increases across learning for both Go and NoGo texture presentation before regressing to Shaping levels during the Reward session. ¶ p=0.0016 Go Shaping vs. Go Early, ᴨ p<0.0001 Go Shaping vs. Go Late/Expert, @ p=0.0265 Go Early vs. Go Late, ß p=0.0154 Go Early vs. Go Expert, # p=0.0005 Go Early vs. Go Reward, † p<0.0001 Go Late/Expert vs. Go Reward. ^ p=0.0128 NoGo Shape vs. NoGo Early,+p<0.0001, NoGo Shape vs. NoGo Late/Expert, £ p=0.0031 NoGo Early vs. NoGo Reward, Ω p=0.0085 NoGo Early vs. NoGo Late,? p=0.0295 NoGo Early vs. NoGo Expert,! p<0.0001, NoGo Late/Expert vs. NoGo Reward. (L) Average auROC markedly increases across learning for both Go and NoGo texture presentation before regressing to Shaping levels during the Reward session. p=0.0041 Go Shaping vs. Go Early, p<0.0001 Go Shaping vs. Go Late/Expert, ∆ p=0.0041 Go Early vs. Go Late, > p=0.0003 Go Early vs. Go Expert, $ p=0.0014 Go Early vs. Go Reward, and p<0.0001 Go Late/Expert vs. Go Reward. ∞ p=0.0462 NoGo Shaping vs. NoGo Early, Ø p<0.0001 NoGo Shaping vs. NoGo Late/Expert, ∑ p=0.0017 NoGo Early vs. NoGo Late, ¥ p=0.0005 NoGo Early vs. NoGo Expert, ¢ p=0.0142 NoGo Early vs. NoGo Reward, € p<0.0001 NoGo Late/Expert vs. NoGo Reward. (M–P) Grand average of licking (Top), calcium (Middle), and normalized pupil (Bottom) activity segmented by trial outcomes: (M) Hit, (N) Miss, (O) FA, and (P) CR. (Q) Average of maximal axonal calcium amplitude of each mouse markedly increases across learning for all trials outcomes before regressing to Shaping levels during the Reward session. Data are mean ± SEM unless noted otherwise. * p<0.05, ** p<0.01, **** p<0.0001. See also Figure 3—figure supplement 1.

Figure 3—figure supplement 1.. Establishment of control and target windows, and representative example of all three activity parameters segmented by trial type and outcome.

(A) Establishment of the Control (1 s following trial start, encompassing the pre-trial intertrial interval) and Target (2 s before and after PT window start) windows. Task-related events are directly above the heatmap. Trial outcomes are color coded (blue = Hit, yellow = Miss, orange = False Alarm (FA), brown = Correct Rejection (CR). (B) Representative axonal calcium activity during a single trial (outcome = hit). Black arrowheads above positive deflections indicate a calcium peak that was greater than or equal to the 90th percentile of all calcium peaks. The yellow circle encompassing the largest calcium peak was selected as the maximal calcium amplitude. (C) Longitudinal representation of the average maximum calcium amplitude (Top) and the average trial % with a detectable calcium peak (Bottom) within the Control (Left) and Target (Right) windows. (D) Average maximum calcium amplitude Top, ᴨ p=0.0005, Control Shaping vs. Control Learning; ß p<0.0001, Control Shaping vs. Control Expert; # p<0.0001, Target Shaping vs. Target Learning/Expert; ¢ p=0.0073, Target Learning vs. Target Reward;+p = 0.0091 Target Expert vs. Target Reward; Mixed-Effects Analysis with Tukey’s correction for multiple comparisons). Average trial % with a detectable calcium peak (Bottom, ¶ p<0.0001 Control Shaping vs. Control Learning/Expert; † p<0.0001, Target Shaping vs. Target Learning/Expert; € p=0.0073, Target Learning vs. Target Reward; ¥ p=0.0091, Target Expert vs. Target Reward; Mixed-Effects Analysis with Tukey’s correction for multiple comparisons). (E–F) Three activity parameters (licking, axonal calcium, and pupil) from a representative session segmented by trial type: (E) Go texture or (F) NoGo texture presentation. Top: Licking activity. Licks are denoted as colored tick marks. The vertical black line represents a sound cue indicating trial start as the presented texture moves toward the whisker field. The vertical red lines denote the start (texture arrival at its endpoint in the whisker field) and end of the PT window. The vertical brown line denotes average RT. The colored boxes denote a 500ms grace period wherein the mouse can lick freely without triggering any outcomes. Top Middle: Lick histogram. Middle: Heatmap sorted by trial outcome (Right of heatmap) highlighting axonal calcium activity for each trial. Trial outcome is color coded (Blue = Hit, yellow = Miss, Orange = FA, Brown = CR). Bottom Middle: Average axonal calcium activity. Bottom: Average pupil area (as a normalized percentage). (G–J) The representative session from E-F was further segmented by trial outcome: (G) Hit, (H) Miss, (I) FA, and (J) CR. Underneath the Bottom panel, texture movement that is dependent on trial outcome is illustrated. Data are mean ± SEM. **** p<0.0001.

Figure 4.. Photoinactivation increases number of sessions required For expert discrimination.

(A) Schematic detailing pAAV-hSyn-JAWS-KGC-GFP-ER2 (JAWS) injection unilaterally into POm (Right) and a 200 µm cannula implanted in the left posterior DLS (Left). For the No Stim cohort, only the cannula was implanted in the left posterior DLS. Activation of the inhibitory JAWS opsin was performed constantly on for 1 s before and after texture arrival in the whisker field. JAWS activation probability per trial = 0.50. (B) Representative injection site in POm (Left), and the cannula placement in the posterior DLS along with ascending POm axons (Right). Scale = 1 mm. Red inset shows ascending POm axons underneath the optic cannula. Inset scale = 200 µm. (C) Top: Stimulation timing (constant illumination for 2 s, centered around texture arriving at its endpoint) and texture movement representation during a trial. Note that no light is presented for the No Stim cohort as no stimulation occurred. Bottom: Outcomes for each stimulus-response pair. (D) Schematic representing texture movement and potential outcomes during a single trial. (E) Changes in Hit Rate, FA Rate, Sensitivity (d’), and Bias of all JAWS cohort mice (n=4) as they transition from the Learning to the Expert phase in box-and-whisker plots. Note that mice are classified as Expert when they achieve a Hit Rate ≥0.80 and a FA Rate ≤0.30 for two consecutive sessions. Red line = 0. See Figure 4—figure supplement 1. (F) Probability density function for overall licking-related activity at each behavioral time point for the JAWS cohort. Vertical black line = sound cue representing trial start as the texture moves toward the whisker field. Vertical red lines = start (texture arrival at endpoint in whisker field) and end (texture departure toward starting point) of the PT window (time where mice can respond by licking). Colored boxes = 500ms grace period (licking does not trigger any outcomes). (G–H) Same as in (E, F) for the No Stim cohort. (I) JAWS cohort requires significantly more training sessions for expert discrimination compared to the FP and No Stim cohorts. (J) Longitudinal representation of sessions required for expert discrimination. (K) Comparison of Hit Rate, FA Rate, Sensitivity (d’), and Bias during the Learning and Expert phases. (L) Average RT is slower during photoinactivation than non-photoinactivated trials. Data are mean ± SEM. * p<0.05.

Figure 4—figure supplement 1.. Individual longitudinal learning-related changes in behavioral parameters and licking activity for the JAWS and no stim cohorts, and whisker trim.

(A–B) Longitudinal learning-related changes as mice from the (A) JAWS and the No Stim (B) cohorts progress through each behavioral time point (Shaping, Early Learning, Late Learning, and Expert). The Shaping time point permitted mice to learn to lick the spout for a water reward during the first three sessions without the presence of the textures. During the last two Shaping sessions, the Go and NoGo textures were presented simultaneously with a Go texture probability of 0.90 for the first session and 0.75 for the second session. For the first Learning session, and all subsequent sessions, the Go texture probability was set to 0.50. Mice were considered Learning at this time point. The Early Learning time point was considered the first two sessions after Shaping, while the Late Learning was considered the last two sessions before achieving Expert status. Mice were considered Expert when they had a Hit Rate ≥0.80 and a FA Rate ≤0.30 for two consecutive sessions. Left: Hit Rate (black) and FA Rate (gray) across learning. Middle: Sensitivity (d’; black) and Bias (gray) across learning. Red dashed line indicates 0. Note that a strict d’ threshold was not used due to artificially increased d’ values as Hit Rate and/or FA Rate approached their extremes (0 or 1). Right: Probability density function for licking activity at each time point. Scale = 1 s. (C) Following the attainment of expert status, a group of mice (n=4) underwent a whisker trim session. In this session, mice performed 75 trials as normal with no optogenetic stimulation. After, the right whiskers contacting the textures were trimmed, and another 75 trials were performed. Behavioral performance, based on the four parameters (Hit Rate, FA Rate, Sensitivity, and Bias), was compared. Data are mean ± SEM. ** p<0.01, *** p<0.001.