Abstract
Coordinated multijoint limb and digit movements—“manual dexterity”—underlie both specialized skills (e.g., playing the piano) and more mundane tasks (e.g., tying shoelaces). Impairments in dexterous skill cause significant disability, as occurs with motor cortical injury, Parkinson's disease, and a range of other pathologies. Clinical observations, as well as basic investigations, suggest that corticostriatal circuits play a critical role in learning and performing dexterous skills. Furthermore, dopaminergic signaling in these regions is implicated in synaptic plasticity and motor learning. Nonetheless, the role of striatal dopamine signaling in skilled motor learning remains poorly understood. Here, we use fiber photometry paired with a genetically encoded dopamine sensor to investigate striatal dopamine release in both male and female mice as they learn and perform a skilled reaching task. Dopamine rapidly increases during a skilled reach and peaks near pellet consumption. In the dorsolateral striatum, dopamine dynamics are faster than in the dorsomedial and ventral striatum. Across training, as reaching performance improves, dopamine signaling shifts from pellet consumption to cues that predict pellet availability, particularly in medial and ventral areas of the striatum. Furthermore, performance prediction errors are present across the striatum, with reduced dopamine release after an unsuccessful reach. These findings show that dopamine dynamics during skilled motor behaviors change with learning and are differentially regulated across striatal subregions.
Significance Statement
Dexterity is central to everyday life but impaired in Parkinson's disease, which is characterized by midbrain dopamine neuron degeneration. These neurons project to the striatum, where dopamine release plays a role in motor learning and execution. Endogenous dopamine dynamics in striatal subregions and how they relate to dexterous skill remain unclear. We used a fluorescence-based dopamine biosensor to elucidate the pattern of dopamine release in striatal subregions as mice learn and execute single-pellet skilled reaching. Dopamine release varies across dorsolateral, dorsomedial, and ventral striatum as mice learn the task and demonstrates “performance prediction errors” when mice fail to retrieve the pellet. These results suggest that phasic dopamine dynamics are important for learning and executing dexterous skills.
Introduction
Manual dexterity is a critical evolutionary adaptation for human survival. From the perspective of modern humans, dexterity is essential for everyday tasks such as fastening buttons and tying shoelaces; however, stroke, multiple sclerosis, spinal cord injury, and several movement disorders impair dexterity (Alberts and Wolf, 2009; Benedict et al., 2011; Kamm et al., 2012; Nardone et al., 2013; Térémetz et al., 2015). In particular, people with Parkinson's disease (PD) experience “coordinative” deficits that significantly impair quality of life (Vanbellingen et al., 2018). Unlike bradykinesia and hypometria, coordinative deficits respond only partially and inconsistently to levodopa (Melvin et al., 2005; Negrotti et al., 2005; Doan et al., 2008; Lukos et al., 2013; Lee et al., 2018; Fasano et al., 2022). This suggests that aspects of dopamine signaling not restored by levodopa influence dexterous skill.
Rodent single-pellet reaching (“skilled reaching”) is a commonly used model of human dexterity in which rodents reach with a single forepaw toward a sugar pellet, grasp it, and retrieve it. Skilled reaching requires precise multijoint coordination, depends on the motor cortex for optimal performance (Whishaw et al., 1986; Guo et al., 2015), is homologous to human reaching (Sacrey et al., 2009), and recapitulates many features of human neurological disorders (Klein et al., 2012). Given the importance of the motor cortex to human motor control, skilled reaching is likely to have translational relevance to human dexterous skill in health and disease.
Rodent skilled reaching is sensitive to dopamine manipulations. Rats with large 6-OHDA lesions make hypometric reaches similar to humans with PD (Whishaw et al., 1986, 2002). Coordination between the proximal forelimb and distal digit movements is altered by acute dopamine manipulations specifically during reaches. Repeated optogenetic nigral stimulation or inhibition gradually accelerated or delayed transitions between reach submovements, respectively (Bova et al., 2020). These artificial and perhaps nonphysiologic manipulations demonstrate that altered nigrostriatal signaling influences not only reward processing but also forelimb–digit coordination. It remains unknown, however, how striatal dopamine dynamics normally contribute to the acquisition and maintenance of fine motor skills.
We used fiber photometry recordings of the dopamine-sensitive biosensor dLight to determine striatal dopamine dynamics as mice learned and executed a head-fixed single–pellet reaching task. Dynamic dopamine signals emerged throughout the striatum, with a sharp increase at the reach onset that peaked after the mouse made contact with the pellet. There were subtle but clear differences across striatal subregions, with broader peaks in the ventral striatum (VS) and narrower peaks in the dorsolateral striatum (DLS). Furthermore, ventral dopamine increased more than dorsal dopamine during cues predicting target availability. Surprisingly, these signals were nearly identical for the striatum ipsi- and contralateral to the reaching paw. Finally, there was a brief postreach time window in which dopamine signals reflected successful versus failed grasping. These data show that striatal dopamine is modulated in specific ways across striatal subregions during skilled forelimb movements and suggest a mechanism by which disrupted dopamine signaling could affect multijoint coordination independently of movement “vigor.”
Materials and Methods
Mice
All animal care and experimental procedures were approved by the National Institute of Health and University of Michigan Institutional Animal Care and Use Committee. C57/B6 mice (https://www.jax.org/strain/000664, RRID:IMSR_JAX:000664) of both sexes (n = 24, aged 10–15 weeks at surgery) were housed at 22–24°C with a 12 h light:12 h dark cycle with standard mouse chow and water provided ad libitum, unless otherwise stated.
Surgery
Mice were anesthetized with isoflurane (2–3%) and placed into a stereotaxic apparatus (KOPF Model 963). For postoperative care, mice were injected intraperitoneally with meloxicam (5 mg/kg). After exposing the skull via small incision, a small hole was drilled through the skull for injection. A pulled-glass pipette with ∼20 μm tip diameter was inserted into the brain, and the virus was injected by an air pressure system. The pipette was kept in place for 5 min after injection before withdrawal. For in vivo photometry experiments, pAAV-CAG-dLight1.1 (titer 1.3 × 1012 genome copies per ml; Addgene viral prep 111067-AAV1; http://n2t.net/addgene:111067; RRID:Addgene_111067) or pAAV-hSyn-EGFP (titer ≥ 7 × 1012 genome copies per ml; Addgene viral prep 50465-AAV1; http://n2t.net/addgene:50465; RRID:Addgene_50465) was injected into the DLS (100–200 nl, A/P, 0.9 mm; D/V, −2.9 mm; M/L, 2.1 mm from the bregma), dorsomedial striatum (DMS; 100–200 nl, A/P, 0.9 mm; D/V, −2.9 mm; M/L, 1.1 mm from the bregma), or VS (100–200 nl, A/P, 0.9 mm; D/V, −4.4 mm; M/L, 1.1 mm from the bregma).
Optic fiber implantations were performed during the same surgery as viral injection. Metal ferrule optical fibers (400-μm-diameter core, NA, 0.50; Thorlabs) were implanted. In a subset of mice, injections and fibers were implanted bilaterally (n = 9). Fibers were fixed to the skull using dental acrylic. Mice were given a minimum of 3 weeks of recovery and 1 week of acclimation before undergoing any experiments. After the completion of the experiments, mice were perfused, and the approximate locations of fiber tips were identified based on the coordinates of Paxinos and Franklin (2012).
Experimental design and statistical analysis
Skilled reaching training
Mice were first habituated by handling them for 5–10 min for 5 d. Mice were then food restricted to ∼85% of their ad libitum body weight. Mice were acclimated to pellets (20 mg dustless precision pellets; BioServe) used in subsequent experiments. Mice were also habituated to head fixation by increasing the duration of head fixation each day across 5 d (10 min on Day 1, 20 min on Day 2, etc.). Paw preference was identified by holding food pellets in front of the head-fixed mouse with forceps, allowing them to reach out and grab the pellet. Each mouse was then head-fixed in the reaching apparatus with the pellet platform (Fig. 1) close enough to their mouth that they could retrieve the pellet with their tongue or by scooping it into their mouth with their paw. Once the mouse was comfortable with this, the mouse was moved further from the pellet platform such that they had to reach out and grab the pellet. In a subset of mice that were reluctant to reach, we allowed a small number (<10) of reaches for pellets in a freely moving context before transitioning them again to the head-fixed setup.
Mice learn and perform head-fixed skilled reach-to-grasp movements. A,B, Head-fixed mice perform a reach-to-grasp movement to obtain a food reward. An LED turned on for 1 s at the trial onset; then a motor turned to place a food pellet within reach of the mouse. High-speed cameras acquired frames at 240 fps to track reach timing. A 30 s intertrial interval followed each trial. C,D, To record dopamine dynamics in the striatum, dLight was expressed in striatal neurons, and an optic fiber was placed to perform fiber photometry. E, Across training, mice improved the proportion of trials where they attempted a reach (top left) and successfully reached (top right). They also reduced their latency from the LED offset to the reaching onset (bottom left) and their latency from the reach onset to pellet contact (bottom right).
Trial start was signified by an LED turning on for 1 s, followed by a stepper motor turning to place a pellet in front of the mouse (1.2 s turn duration; controlled via custom Arduino code). As the stepper motor turned, two high-speed cameras (NaturalPoint, Optitrack PrimeX 13) were triggered to begin video acquisition at 240 fps. for 6 s. After a 30 s intertrial interval, the next trial started regardless of whether the mouse had reached for the pellet. The 20 trials were performed per run, and up to four runs occurred per day. Videos were manually scored to identify the trial outcome: “nonattempt,” where the mouse did not attempt a reach; “attempt,” where the mouse attempted a reach but did not successfully grasp the pellet; “miss,” where the mouse successfully grasped the pellet but did not successfully consume the pellet; and “hit,” where the mouse successfully consumed the pellet. In a subset of experiments (n = 3), mice that were successfully reaching with their preferred paw had a physical blockade put in place to stop them from reaching, forcing them instead to reach-to-grasp with their other paw.
dLight photometry
A fiber-optic patch cable (1 m long, 400 μm diameter; Doric Lenses) was firmly attached to the implanted fiber-optic cannula with a zirconia sleeve (Doric Lenses). LEDs (473 nm; Plexon) were set such that a light intensity of <0.1 mW entered the brain; light intensity was kept constant across sessions for each mouse. Emission light was passed through a filter cube (Doric Lenses) before being focused onto a sensitive photodetector (2151, Newport). The signal was digitized with a National Instruments data acquisition card and collected using a custom MATLAB script. Fluorescent traces were bleaching corrected by subtracting a double exponential fit and then adding back the mean of the trace prior to calculating ΔF/F. ΔF/F = (F − F0)/F0, where F0 was calculated as the 10th percentile of the entire fluorescence trace. These traces were z-scored to facilitate comparisons across days and mice. To establish differences in fluorescent time courses across groups, we downsampled the data from 1,000 to 20 Hz. To quantify discrete changes locked to a given behavior in bar plots, we established a baseline as the 1 s prior to the trial onset cue and compared it to the period immediately after the trial onset (2 s), reach onset (1 s), pellet contact (1 s), or consumption (1 s).
To investigate dopamine responses in dorsal striatum (DS) and VS during running and in response to liquid rewards, we head-fixed a subset of mice on a running wheel, and we have given them access to a lickspout with 10% sucrose. Wheel running was quantified by infrared beam breaks, and licking behavior was quantified by a capacitive lickspout. Fluorescent signals were averaged at the onset of each running or licking bout and then averaged across mice. We defined the baseline as the 2 s prior to the behavior onset, running behavior as the 8 s after the running onset, and licking behavior as the 2 s after the licking onset.
Software accessibility
Analysis was performed using custom MATLAB scripts. Analysis code is available via online repository (https://github.com/cburgess23/Mouse_dLight_skilledreaching.git).
Histology
Mice were anesthetized with pentobarbital and transcardially perfused with phosphate-buffered saline (PBS) followed by 10% formalin. Then, brains were postfixed in 10% formalin for 24 h. Next, brains were transferred into 20% sucrose for ∼48 h. Brains were then frozen, and coronal sections were cut at 40 μm by a freezing microtome. Slices were then washed in PBS and incubated in a conjugated GFP rabbit polyclonal antibody as a primary antibody (1:5000 dilution, Novus Biologicals; catalog #NB600-308AF488) at 4°C overnight. The next day, slices were washed in PBS and mounted on glass slides using VECTASHIELD Antifade Mounting Medium with DAPI (Vector Laboratories). Images were then captured by a fluorescence microscope (Olympus IX83) at 10× magnification.
Statistics
Statistical analyses were performed in GraphPad Prism or MATLAB with an alpha level of 0.05. The data presented met the assumptions of the statistical test employed. Comparisons between the baseline and response within cohorts were evaluated using paired t tests. Comparisons between average responses across cohorts were evaluated using a two-way ANOVA. Differences in time courses across cohorts were evaluated using two-way ANOVAs on data downsampled by 50 from the original acquisition rate (from 1,000 to 20 Hz). Exclusion criteria for experimental animals were (1) sickness or death during the testing period or (2) if histological validation of the injection site demonstrated an absence of dLight or GFP expression. The N numbers represent the final number of healthy/validated animals.
Results
Striatal dopamine release is elevated during execution of a skilled reach
To identify dopamine release dynamics across skilled reaching, we head-fixed the food-restricted mice in front of a pellet platform and trained them to reach for the available pellets. The trial onset was signified by illumination of an LED for 1 s, followed by rotation of the platform to place a pellet in front of the mouse (Fig. 1A,B). On each trial, the mouse may or may not reach for the pellet; therefore trial and reach counts are not necessarily the same. High-speed cameras captured video at 240 fps to allow for scoring of mouse behavior. dLight1.1 was expressed in the striatum, and an optic fiber was placed above the area of expression to allow for fiber photometry recordings of dopamine release (Fig. 1C,D). Head-fixed mice (n = 18) consistently performed a skilled reach-to-grasp movement. Over many trials (range, 140–440; mean, 290 trials/mouse), mice increased their attempt and success rates, culminating in a >60% average success rate once well trained (Fig. 1E; all trials split into five bins of equal trial numbers across learning for each mouse). Throughout training, the latency to start a reach decreased, with subtle decreases in the time from start of reach to pellet contact (mean reach-to-contact latency, 0.212 ± 0.010 ms; Fig. 1E).
For all mice with dLight expression and a fiber implanted in the striatum on the contralateral side to the reaching paw, we investigated striatal dopamine dynamics in response to the trial onset (i.e., the LED), the start of the reach, contact with pellet, and start of consumption for successful hit trials (Fig. 2A). There was a robust increase in dopamine-dependent fluorescence at the trial onset; dopamine release increased at the start of the reach, peaking just prior to pellet consumption (Fig. 2B; p < 0.05; two-way ANOVA/Sidak multiple-comparison test). Dopamine significantly increased at both the trial onset (baseline, −0.09 ± 0.02; response, 0.32 ± 0.09; p < 0.001; two-way ANOVA/Sidak multiple-comparison test; event, p = 0.005; F(1,46) = 8.6; group, p = 0.1; F(1,46) = 2.7; interaction, p = 0.006; F(1,46) = 8.3) and reach onset (baseline, −0.09 ± 0.02; response, 1.43 ± 0.14; p < 0.0001; two-way ANOVA/Sidak multiple-comparison test; event, p < 0.0001; F(1,46) = 45; group, p < 0.0001; F(1,46) = 26; interaction, p < 0.0001; F(1,46) = 35), when compared with the baseline period before the trial start (Fig. 2C,D). No increase was observed in control mice expressing GFP instead of dLight (n = 6).
dLight recordings show dynamics during task performance while GFP recordings do not. A, Using high-speed videography, task-specific events were scored, including the trial onset, the start of the mouse's reach, the mouse grasping the pellet, and the onset of pellet consumption. B, Dopamine increased in response to the cues predicting the trial onset and in response to the reach onset, peaking around the time of pellet consumption. Lines under time courses denote a significant difference from GFP at that time, p < 0.05, two-way ANOVA/Sidak multiple-comparison test. C,D, dLight recordings demonstrated significant increases in response to the trial onset and the reach onset compared with the baseline and compared with GFP control recordings. These data are averaged across all 18 mice with striatal dLight recordings, independent of fiber placement.
We hypothesized that the reach-related dopamine response would be more prominent and sustained in the striatum contralateral to the reaching paw. In a subset of mice, we recorded bilateral striatal dopamine release throughout behavioral training (n = 8). Dopamine levels increased in both the contralateral (trial onset: baseline, −0.06 ± 0.03; response, 0.27 ± 0.10; p = 0.02; two-way ANOVA/Sidak multiple-comparison test; event, p = 0.0005; F(1,32) = 14; group, p = 0.6; F(1,32) = 0.26; interaction, p = 0.88; F(1,32) = 0.02; reach: baseline, −0.06 ± 0.03; response, 1.09 ± 0.17; p < 0.01; two-way ANOVA/Sidak multiple-comparison test; event, p < 0.0001; F(1,32) = 66; group, p = 0.36; F(1,32) = 0.8; interaction, p = 0.45; F(1,32) = 0.55) and ipsilateral (trial onset: baseline, −0.04 ± 0.04; response, 0.20 ± 0.10; p = 0.05; two-way ANOVA/Sidak multiple-comparison test; reach: baseline, −0.04 ± 0.04; response, 1.35 ± 0.25; p = 0.001; two-way ANOVA/Sidak multiple-comparison test) striatum in response to the trial onset and to the start of the reach (Fig. 3A,B). There was no significant difference between contralateral and ipsilateral dopamine levels during either the trial onset or reach onset (trial onset, p = 0.88; F(1,32) = 0.02; two-way ANOVA; reach, p = 0.36; F(1,32) = 0.84; two-way ANOVA). In a further subset of mice, we blocked their preferred reaching paw after they had become proficient in reaching, forcing them to learn to reach with their nonpreferred paw (n = 3; Fig. 3C,D). These mice achieved similar hit rates with their less preferred paw. During both normal and paw-blocked reaching, mice showed a significant increase in dopamine release at the reach onset in both the contralateral and ipsilateral striatum relative to the reaching paw (normal reaching, p < 0.01; two-way ANOVA/Sidak multiple-comparison test; event, p < 0.0001; F(1,8) = 95; group, p = 0.64; F(1,8) = 0.22; interaction, p = 0.74; F(1,8) = 0.11; paw-blocked reaching, p < 0.01; two-way ANOVA/Sidak multiple-comparison test; event, p < 0.0001; F(1,8) = 92; group, p = 0.94; F(1,8) = 0.004; interaction, p = 0.66; F(1,8) = 0.02).
Striatal dopamine dynamics are comparable in the contra- and ipsilateral striatum. A, Dopamine dynamics were similar in the contralateral striatum and ipsilateral striatum relative to the reaching paw. B, dLight recordings demonstrated significant increases in response to the trial onset and reach onset compared with the baseline in both the contralateral and ipsilateral striatum. C, Mice initially learned to reach with their preferred reaching paw and perform the task well. Dopamine significantly increased during reaching relative to the baseline in both the contralateral and ipsilateral striatum relative to the reaching paw. D, Mice in which their preferred reaching paw was physically blocked could relearn the task and perform well. Dopamine significantly increased during reaching relative to the baseline in both the contralateral and ipsilateral striatum relative to the new, nonpreferred, reaching paw. These data are averaged across mice independent of fiber placement.
Dopamine levels increased in both the female (n = 10; trial onset: baseline, −0.07 ± 0.03; response, 0.26 ± 0.11; p = 0.03; two-way ANOVA/Sidak multiple-comparison test; event, p < 0.0001; F(1,32) = 24; group, p = 0.6; F(1,32) = 0.26; interaction, p = 0.33; F(1,32) = 0.94) and male (n = 8; trial onset: baseline, −0.11 ± 0.03; response, 0.40 ± 0.16; p = 0.05) mice in response to the trial onset (Fig. 4A,B). Similarly, both female (reach: baseline, −0.07 ± 0.03; response, 1.49 ± 0.22; p < 0.0001; two-way ANOVA/Sidak multiple-comparison test; event, p < 0.0001; F(1,32) = 20; group, p = 0.5; F(1,32) = 0.08; interaction, p = 0.72; F(1,32) = 0.02) and male (reach: baseline, −0.11 ± 0.03; response, 1.35 ± 0.15; p < 0.0001) mice had significantly elevated dopamine responses at the reach onset. There was no significant difference between male and female mice during either the trial or reach onset (trial onset, p = 0.61; F(1,32) = 0.26; two-way ANOVA; reach, p = 0.51; F(1,32) = 0.45; two-way ANOVA).
Striatal dopamine dynamics are similar in male and female mice. A, Dopamine increased in response to the cues predicting the trial onset and in response to the reach onset similarly in both male and female mice. B, Recordings showed no significant difference between male and female mice in response to the trial onset and reach onset. These data are averaged across mice independent of fiber placement, though we found in subsequent analysis that dopamine release differs across the dorsal–ventral striatal axis (Extended Data Fig. 4-1).
Figure 4-1
Dopamine release differs across the dorsal-ventral axis of the striatum. A-B. Dopamine increases in response to locomotion and liquid rewards, with recordings in the dorsomedial striatum responding more strongly to running, and ventral striatum (VS) recordings responding more strongly to rewards. C. GFP controls showed no response to either behavior. Download Figure 4-1, TIF file.
Previous recordings of dopamine activity suggest that dopamine dynamics may vary across striatal subregions, particularly along the dorsal–ventral axis (Howe and Dombeck, 2016; Parker et al., 2016). We confirmed this by quantifying dopamine release in response to spontaneous bouts of running or licking for a sucrose reward. While there were some dynamics to both events in DS (n = 6; running: baseline, 0.21 ± 0.03; response, 0.32 ± 0.04; t(5) = −2.42; p = 0.05; paired t test; licking: baseline, 0.35 ± 0.07; response, 0.45 ± 0.11; t(5) = −0.87; p = 0.41; paired t test) and VS (n = 4; running: baseline, 0.26 ± 0.06; response, 0.36 ± 0.02; t(3) = −1.3; p = 0.2; paired t test; licking: baseline, 0.02 ± 0.07; response, 0.89 ± 0.12; t(3) = −15; p < 0.005; paired t test), there was a bias toward reward-type responses in VS and motor responses in DS (Extended Data Fig. 4-1A,B). No dynamics were seen in response to running or licking in GFP controls (n = 5; running, t(4) = 1.4; p = 0.27; licking, t(4) = 0.05; p = 0.96; paired t test; Extended Data Fig. 4-1C).
Striatal subregion-specific dopamine dynamics during skilled reaching behavior
As dopamine release is differentially regulated in striatal subregions, averaging across all recordings independent of fiber location may not give an accurate sense of dopamine dynamics. We therefore separated our recordings into mice with fibers in the DLS (n = 4), DMS (n = 6), and VS (n = 8) on the hemisphere contralateral to the reaching paw (Fig. 5A). We investigated dopamine responses in each subregion during trials that were successful (hit trials) or unsuccessful (miss trials), in addition to trials where the mouse attempted to reach but failed to grasp the pellet (attempt trials) or did not attempt a reach (nonattempt trials). In the DLS, dopamine was elevated slightly during trial onset cues and sharply at the reach onset, with the fast onset and offset dynamics (Fig. 5C). There was a significant increase relative to the baseline during both the trial onset response (baseline, −0.07 ± 0.04; response, 0.11 ± 0.03; t(3) = −3.7; p = 0.03; paired t test) and the reach response (baseline, −0.07 ± 0.04; response, 1.79 ± 0.28; t(3) = −6.05; p = 0.009; paired t test; Fig. 5D,E). When comparing hit and miss trials, there was evidence of performance prediction error, with reduced dopamine in the miss trials at the time when pellet consumption would have likely occurred in hit trials [hit, 0.62 ± 0.25; miss, −0.48 ± 0.08; t(3) = 5.7; p = 0.01; paired t test; Figs. 2C, 5F, pink-shaded area (250–1,250 ms after pellet contact); average grasp-to-consumption duration, 0.320 ± 0.022 ms].
Dorsolateral striatal dopamine dynamics during skilled reaching. A, Fiber locations for photometry recordings across subregions of the striatum. B, Using high-speed videography, task-specific events were scored, including the trial onset, the start of the mouse's reach, the mouse grasping the pellet, and the onset of pellet consumption. C, Dopamine dynamics in DLS in response to nonattempt trial, attempt trial, hit trial, and miss trial outcomes. Top right, Representative hit trials from one mouse. Middle right, Distribution of pellet contact-to-consumption latencies in hit trials. D,E, DLS dopamine increased in response to the trial onset and reach onset when compared with the baseline. F, Performance prediction error is evident in DLS when comparing dopamine dynamics in hit versus miss trials (pink-shaded area in A).
In the DMS, dopamine was elevated during trial onset cues and the onset of a reach, with fast onset and slower offset dynamics (Fig. 6A). There was a significant increase relative to the baseline during the reach response (baseline, −0.06 ± 0.03; response, 1.08 ± 0.19; t(5) = −5.45; p = 0.002; paired t test; Fig. 6B,C). When comparing hit and miss trials, there was clear evidence of performance prediction error, with a reduced dopamine response in the miss trials at the time when pellet consumption would have likely occurred in hit trials (hit, 0.78 ± 0.19; miss, −0.13 ± 0.13; t(5) = 3.97; p = 0.01; paired t test; Fig. 6A,D, pink-shaded area).
Dorsomedial dopamine dynamics during skilled reaching. A, Dopamine dynamics in DMS in response to nonattempt trial, attempt trial, hit trial, and miss trial outcomes. Top right, Representative hit trials from one mouse. Middle right, Distribution of pellet contact-to-consumption latencies in hit trials. B,C, DMS dopamine did not increase significantly in response to the trial onset, showing variability across recordings, but did increase at the reach onset when compared with the baseline. D, Performance prediction error is evident in DMS when comparing dopamine dynamics in hit versus miss trials (pink-shaded area in A).
In the VS, dopamine was elevated during trial onset cues and the onset of a reach, with slower onset and offset dynamics (Fig. 7A). There was a significant increase relative to the baseline during both the trial onset (baseline, −0.12 ± 0.03; response, 0.32 ± 0.10; t(7) = −3.67; p = 0.007; paired t test) and the reach response (baseline, −0.12 ± 0.03; response, 1.51 ± 0.20; t(7) = −7.76; p = 0.0001; paired t test; Fig. 7B,C). When comparing hit and miss trials, there was a clear performance prediction error, with a reduced dopamine response in the miss trials at the time when pellet consumption would have likely occurred in hit trials (hit, 1.08 ± 0.2; miss, 0.11 ± 0.16; t(7) = 7.4; p < 0.001; paired t test; Fig. 7A,D, pink-shaded area). In general, reach responses for all striatal regions were present on even the first few reaches that mice made, though trial onset responses were absent on early reaches. Reach-related dopamine release was greater in miss and hit than attempt trials, which would have occurred later in training than the attempt trials (Extended Data Fig. 7-1).
VS dopamine dynamics during skilled reaching. A, Dopamine dynamics in VS in response to nonattempt trial, attempt trial, hit trial, and miss trial outcomes. Top right, Representative hit trials from one mouse. Middle right, Distribution of pellet contact-to-consumption latencies in hit trials. Reach responses were evident even on the first few reaches (Extended Data Fig. 7-1B,C). B,C, VS dopamine increased significantly in response to the trial onset and at the reach onset when compared with the baseline. D, The performance prediction error is evident in VS when comparing dopamine dynamics in hit versus miss trials (pink-shaded area in A).
Figure 7-1
Dopamine release dynamics in the first few reaches. A-C. Heatmaps showing dopamine dynamics across the first reach (Left) and average of the first three reaches (Right) for each mouse in DLS (A), DMS (B), and VS (C). Download Figure 7-1, TIF file.
When comparing dopamine release during execution of skilled reaching across DLS, DMS, and VS, there were different release dynamics across areas. VS dopamine was increased during the trial onset cues, specifically in response to the stepper motor turning, when compared with both DLS and DMS (Fig. 8A; two-way ANOVA/Sidak multiple-comparison test; F(120,900) = 2.604; p < 0.001). VS dopamine generally had slower dynamics, while DLS dopamine peaked at a higher level than VS and DMS and had generally faster dynamics. DLS had a significantly greater slope at the reach onset than DMS and VS (Fig. 8B,C; ANOVA/Sidak multiple-comparison test; F(2,15) = 5.73; p = 0.014). Reach-to-contact latencies reduced as mice became more proficient in the task; this was true across mice with recordings in each striatal region (Extended Data Fig. 8-1A) and did not explain peak dopamine response, dopamine ascent slope, or dopamine decay slope in any of the three striatal regions (Fig. 8D,E; Extended Data Fig. 8-1B,C). Even for the few mice in which there was a significant association between dopamine dynamics and reach-to-contact latency, features of phasic dopamine release explained very little of the variance in reach-to-contact latency.
Differences in dopamine dynamics across striatal areas. A, Comparison of dopamine dynamics across subregions of the striatum. Lines under time courses denote a significant difference at that time, p < 0.05, two-way ANOVA/Sidak multiple-comparison test. B,C, Quantification of the rise and decay slopes across striatal regions demonstrated that the DLS ascent slope was significantly greater than VS ascent slope. *Denotes a significant difference, p < 0.05, ANOVA. D, Examples of reach-related dopamine release regressed against reach-to-contact latency in example mice. E, Coefficients of determination and p values for regressions of dopamine amplitude against reach-to-contact latency for each mouse. Reach-to-contact latency is not predictive of peak dopamine release or slope (Extended Data Fig. 8-1) across mice in DLS, DMS, or VS. Red outlines in panel E denote significant correlations for individual mice.
Figure 8-1
Reach-to-contact latency is reduced across learning. A. Reach-to-contact latency was reduced with more experience with the task. Individual mice denoted by different opacity of fill. DLS: p < 6.5 × 10−6; DMS: p < 5.0 × 10−9; VS: p < 5.2 × 10−7. B. Examples of dopamine ascent (Top) or decay (Bottom) slope correlated with reach-to-contact latency in individual mice. C. Coefficients of determination and p-values for regressions of dopamine ascent and decay slopes against reach-to-contact latency for each mouse Reach-to-contact latency is not predictive of dopamine ascent or decay across mice in DLS, DMS, or VS. Red outlines in panel E denote significant correlations for individual mice. Download Figure 8-1, TIF file.
Dopamine dynamics change across motor learning
We next investigated how dopamine dynamics change across learning of skilled reaching behavior in different areas of the striatum. For each mouse, the first 20 hit trials (early learning) were compared with the last 20 hit trials (late learning). In the DLS, peak reach dopamine response, half-peak width of the reach response, and the timing of the peak dopamine response relative to the start of the reach were all unchanged (peak, t(3) = 0.09; width, t(3) = −0.68; timing, t(3) = 0.78; p > 0.4; paired t tests; Fig. 9A). Example heatmaps from a representative mouse show that dopamine dynamics were largely similar across learning (Fig. 9B). The response to the trial onset increased slightly (Fig. 9C; early, −0.05 ± 0.9; late, 0.26 ± 0.04; t(3) = −3.4; p = 0.04; paired t test), but reach response and consumption response were similar across learning, resulting in only a subtle shift in dopamine release from the reach to the trial onset cues (Fig. 9D).
Dopamine dynamics change across learning. A, Peak dopamine reach response, half-peak width of the dopamine reach response, and the timing of the peak dopamine reach response across learning in DLS. B, Representative heatmap from hit trials across learning in one mouse with a DLS recording. C, Differences in dopamine dynamics across learning in DLS. D, Relationship between peak ΔF / F at the trial and reach onset for early and late reaches. Dopamine responses tend to shift from the reach onset to trial onset with training. E–H, Same as A–D for DMS. I–L. Same as A–D for VS. *Denotes a significant difference, p < 0.05, paired t test. Lines under time courses denote a significant difference at that time, p < 0.05, two-way ANOVA/Sidak multiple-comparison test.
In the DMS, peak dopamine reach response was decreased in late versus early learning (Fig. 9E; early, 3.2 ± 0.54; late, 1.9 ± 0.25; t(5) = 2.72; p = 0.04; paired t test), as was the timing of the peak dopamine response (early, 0.66 ± 0.05; late, 0.53 ± 0.03; t(5) = 3.35; p = 0.02; paired t test). Dopamine dynamics show subtle differences between early and late learning. Example heatmaps from a representative mouse show that dopamine dynamics change across learning (Fig. 9F). Specifically, brief periods around the LED onset show an increased response during late learning, and pellet consumption shows an increased response during early learning (Fig. 9G); these changes were not significant when averaging over larger time periods (LED; t(5) = −2.15; p = 0.08; reach, t(5) = 2.3; p = 0.06; consumption, t(5) = 2.4; p = 0.06; paired t tests; Fig. 9G insets) but overall resulted in a shift in relative response from the reach to the trial onset cues (Fig. 9H).
In the VS, peak dopamine reach response is decreased in late versus early learning (Fig. 9I; early, 3.3 ± 0.45; late, 2.2 ± 0.15; t(7) = 2.5; p = 0.04; paired t test), while there is no change in half-peak width and the timing of the peak dopamine response (width, t(7) = 0.38; p = 0.7; timing, t(7) = −0.33; p = 0.75; paired t tests). Dopamine dynamics show clear differences across learning in line with reward prediction errors (RPEs). Example heatmaps from a representative mouse show that dopamine dynamics shift from the reach response to the cue predicting trial onset (Fig. 9J). Trial onset cue responses increased during late learning, and pellet consumption responses decreased during late learning (Fig. 9K), resulting in a shift in relative response from the reach to the trial onset cues (Fig. 9L). These changes are reflected when averaging over larger time periods for both trial onset (early, 0.21 ± 0.05; late, 0.88 ± 0.22; t(7) = −2.66; p = 0.03; paired t test) and reach (early, 1.9 ± 0.34; late, 1.1 ± 0.17; t(7) = 2.58; p = 0.03; paired t test; Fig. 9K insets).
We then also compared early (first 20) and late (last 20) miss trials. Due to limited miss trials in some mice, if there were fewer than 40 miss trials total, we compared the first half and last half of miss trials rather than the first 20 and last 20. In the DLS, half-peak width of the reach response was significantly reduced across learning (early, 0.28 ± 0.02; late, 0.20 ± 0.01; t(3) = 3.23; p = 0.04; paired t test; Fig. 10A), while peak response and timing of peak response were unchanged. Example heatmaps from a representative mouse show that dopamine dynamics in miss trials were largely similar across learning (Fig. 10B). The response to the trial onset and the reach response were similar across learning, though there was a significant difference during the period after reach onset perhaps corresponding to a larger performance prediction error later in learning (Fig. 10C). In the DMS, the peak response, half-peak width, and peak timing of dopamine during miss trials were not significantly different across learning (Fig. 10E,F). Neither the trial onset nor reach response showed significantly different dynamics across learning (Fig. 10G), and there was a trend toward a larger dopamine dip after unsuccessful reaches (early, −0.69 ± 0.1; late, −1.14 ± 0.18; t(5) = 2.23; p = 0.07; paired t test; Fig. 10H). In the VS, there was no significant difference in peak dopamine, half-peak width, or peak time across learning (Fig. 10I). An individual example heatmap across miss trials shows the emergence of a trial onset response across learning (Fig. 10J). The mean time course also shows an increase in trial onset cue responses with learning, though there is no change in the reach onset response or the dopamine dip resulting from the unsuccessful trial (Fig. 10K,L).
Dopamine dynamics change across learning in miss trials. A, Peak dopamine reach response, half-peak width of the dopamine reach response, and the timing of the peak dopamine reach response during miss trials across learning in DLS. B, Example heatmap from miss trials across learning in one mouse. C, Differences in dopamine dynamics during miss trials across learning in DLS. D, Quantification of the reduced dopamine release after the reach in miss trials in DLS. E–H, Same as A–D for DMS. I–L, Same as A–D for VS. *Denotes a significant difference, p < 0.05, paired t test. Lines under time courses denote a significant difference at that time, p < 0.05, two-way ANOVA/Sidak multiple-comparison test.
Discussion
We found sharp bilateral striatal dopamine transients time-locked to reach-to-grasp movements in mice. Dopamine increased sharply at the reach onset, with peaks locked most tightly to pellet contact and consumption. These phasic dopamine signals varied in subtle but consistent ways across striatal subregions, with narrower peaks in DLS. Over time, these responses migrated to the cue signaling impending pellet availability, primarily in VS and DMS. Finally, dopamine levels remained relatively elevated in successful trials, but dropped below the baseline with failed pellet retrieval. These results suggest that phasic striatal dopamine signaling plays an important role in learning and maintaining dexterous skills.
Recent experiments suggest that phasic dopamine in DS is associated with movement, while phasic dopamine in VS is associated with reward. DMS dopamine axons activate with contralateral movement or while initiating locomotion, while VS dopamine axons respond to reward or reward-predicting cues (Howe and Dombeck, 2016; Parker et al., 2016). Pairing dopamine activation with a cue induced conditioned approach with ventral tegmental area–VS stimulation but nondirected locomotion with SNc–DS stimulation (Saunders et al., 2018). It has been suggested that these apparently distinct regional functions actually reflect specialized RPE-like signals, with movement-locked dopamine release indicating proximity to the reward rather than driving movement (Tsutsui-Kimura et al., 2020). However, others have been unable to unify movement- and reward-linked dopamine signals into a single framework (Lee et al., 2019; Jørgensen et al., 2023). The functions of phasic dopamine across striatal subregions therefore remain controversial.
In our experiments, phasic dopamine increases were linked to reach initiation in all subregions, and there was a brief period shortly after pellet contact (near the time of pellet consumption) in which dynamics differed between successful and failed reaches. Phasic dopamine events were shorter in DLS, consistent with previous findings (Garris et al., 1994; Brown et al., 2011; Calipari et al., 2012) and likely attributable to differences in dopamine transporter density (Walters et al., 2020). In DMS and VS, but not DLS, dopamine peaks migrate to the trial onset cue with experience, consistent with RPE signals in temporal difference models. These subregional changes acquired through experience are similar to those observed during pavlovian conditioning in rats (Mohebi et al., 2024), though DLS-conditioned stimulus–evoked dopamine increased with training in mice (Salinas et al., 2023). The lack of migration of DLS dopamine signals could reflect a fundamental difference in the “meaning” of dopamine across subregions. That is, DLS dopamine could signal movement, while DMS and VS dopamine could signal prediction errors. Alternatively, it has been suggested that phasic dopamine signals RPEs on progressively shorter timescales moving from VS to DLS. In this framework, the DLS peak would not migrate because the expected reward would occur later than the timescale over which DLS dopamine predicts reward (Mohebi et al., 2024). This would suggest that DLS dopamine is uniquely important for refining rapidly executed motor skills. Based on the accepted views of striatal subregion functions described above, we would expect VS dopamine signals to motivate task engagement, while DMS dopamine signals are important for deciding to reach (action selection). Subregion-specific manipulations of dopamine signaling during reaching are needed to test this hypothesis.
Striatal dopamine is generally considered to subserve two functions: regulating movement vigor and mediating reinforcement learning. Here, we use “vigor” to represent the amplitude, speed, or frequency of movement, which is intentionally broad to capture the range of deficits observed in dopamine-deficient states (Dudman and Krakauer, 2016). Over minutes to hours, tonic dopamine levels clearly influence one or more of these aspects of vigor (Mazzoni et al., 2007; Niv et al., 2007; Beeler et al., 2010b; Leventhal et al., 2014; Panigrahi et al., 2015). On shorter timescales corresponding to phasic signaling, however, the evidence is mixed. Several studies have found that dopamine neurons increase their activity at or prior to the movement onset and/or dopamine neuron stimulation causes rapid movement initiation (Howe and Dombeck, 2016; Coddington and Dudman, 2018; da Silva et al., 2018; Saunders et al., 2018; Hamilos et al., 2021; Hunter et al., 2022). Others have found that the primary dopamine neuron response to movement initiation is a firing pause (Dodson et al., 2016) or decrease in striatal dopamine or terminal activation (Tsutsui-Kimura et al., 2020; Markowitz et al., 2023). It therefore remains unclear if, or under what conditions, phasic dopamine release regulates instantaneous vigor.
If phasic peri-reach striatal dopamine signaling “invigorates” reaching, two things should be true. First, some aspects of phasic dopamine signaling should be correlated with kinematics. Second, manipulating nigrostriatal dopamine should immediately alter reach kinematics. Regarding the first prediction, phasic peri-reach dopamine release in DMS and VS decreased with experience as reach accuracy improved and reach duration shortened. Furthermore, measures of phasic dopamine signaling were uncorrelated with reach duration on a trial-by-trial basis (Fig. 8D,E; Extended Data Fig. 8-1B,C). Because we did not measure detailed three-dimensional kinematics, we cannot definitively say whether phasic dopamine signals are correlated with “vigor” on a moment-by-moment basis. Nonetheless, it seems unlikely that a neuromodulatory signal that operates over 100 s of milliseconds could fine-tune kinematics on a very short timescale (Sippy and Tritsch, 2023). Additionally, previous optogenetic experiments also argue against an immediate effect of phasic dopamine signaling on forelimb kinematics. Repeated dopamine neuron stimulation or inhibition progressively alters forelimb kinematics across many reaches, but does not instantaneously affect reaching or grasping (Bova et al., 2020). While strong, prolonged dopamine neuron stimulation can initiate movement (Howe and Dombeck, 2016; Coddington and Dudman, 2018; da Silva et al., 2018; Saunders et al., 2018; Hunter et al., 2022), shorter pulses reinforce or alter behavior without an immediate change in movement kinematics (Hamilos et al., 2021; Markowitz et al., 2023). This is consistent with a recent report that restoring tonic striatal dopamine in dopamine-depleted mice permits nondexterous forelimb movements but that phasic dopamine release is not required (Liu et al., 2022). Therefore, it is less likely that phasic dopamine signaling under normal physiologic conditions modulates instantaneous reach “vigor.” However, we cannot exclude the possibility that the dopamine increase near the reach onset is necessary for movement initiation. Perturbations to dopamine signaling at specific reach phases will be required to address this question.
An alternative interpretation is that phasic dopamine facilitates corticostriatal plasticity to establish effective reach-to-grasp kinematics with repetition. Striatal dopamine is believed to signal differences between expected and actual outcomes. This is considered an “RPE” in the setting of extrinsic reward (Schultz et al., 1997; Mohebi et al., 2019), but the concept has been extended recently to sensory/perceptual prediction errors and performance prediction errors. Perceptual prediction errors, the difference between anticipated and real percepts, have been identified in the tail of the striatum and are suggested to underlie hallucinations (Schmack et al., 2021). “Performance prediction errors” differ from RPEs in that accurate motor performance is the outcome of interest rather than extrinsic reward. For example, dopamine neuron activity in songbirds reflects song quality (Gadagkar et al., 2016; Duffy et al., 2022). In our experiments, it is difficult to distinguish reward from performance prediction errors because successful movement is defined by reward acquisition. Regardless of the precise interpretation of peri-reach dopamine transients, however, our data suggest that phasic dopamine signaling helps to refine skilled movements. This is consistent with previous findings that dopamine receptor blockade and dopamine replacement cause progressive rather than immediate changes in rotarod performance in mice (Beeler et al., 2010a, 2012) and precisely timed optogenetic manipulations cause progressive as opposed to immediate changes in behavior (Bova et al., 2020; Markowitz et al., 2023; Tang et al., 2023).
Still, a pure “performance prediction error” (or RPE) signal is unlikely to explain the role of phasic striatal dopamine in motor learning for two reasons. First, there are more degrees of freedom in reach-to-grasp movements than instrumental tasks. Given a choice between discrete alternatives, a dopamine dip in the context of reinforcement learning has a clear interpretation: select a different option on subsequent trials. However, it is not obvious how a mouse should adjust its reach-to-grasp movement given a dopamine dip after a failed reach. Secondly, optogenetic manipulation of nigrostriatal dopamine signaling during reaches causes predictable changes in forelimb kinematics (Bova et al., 2020). If phasic dopamine release reflects reach outcomes, reach kinematics should be more variable after peri-reach dopamine suppression and more consistent after peri-reach dopamine stimulation. Instead, dopamine neuron inhibition and stimulation have consistent but opposite effects on reach kinematics: stimulation shortens reaches and accelerates transitions between reach submovements, while inhibition lengthens reaches and delays the transition from paw transport to grasping. These interventional experiments coupled with our new results indicate that phasic dopamine signaling in the striatum plays a role in learning and adapting skilled movements. The precise meaning of these signals remains unclear, however.
Mechanistically, phasic dopamine release likely regulates corticostriatal plasticity at synapses active during reaches. D1 dopamine receptors along the striatal “direct” pathway have a lower binding affinity than indirect pathway D2 receptors. This suggests that phasic dopamine increases facilitate plasticity along the direct pathway, while phasic dopamine dips facilitate plasticity along the indirect pathway (Surmeier et al., 2007; Dreyer et al., 2010). Recent work, however, suggests that phasic changes in dopamine strongly affect both D1 and D2 receptors (Marcott et al., 2014; Yapo et al., 2017), with effects lasting much longer than the release event itself (Hunger et al., 2020). A dopamine pulse followed by a dip is predicted to quickly restore baseline D1 and D2 receptor occupancy, while a phasic increase followed by a return to baseline dopamine levels is predicted to cause a prolonged increase in D1 and D2 receptor occupancy. The phasic increase we observed at the reach onset may prime dopamine receptors to facilitate corticostriatal plasticity when information regarding the outcome becomes available (Coddington and Dudman, 2019). Changes in corticostriatal synapses may underlie the increase in corticostriatal synchrony that occurs as skilled reaching is acquired (Lemke et al., 2019).
Reach-to-grasp movements recruit a complex network of cortical and subcortical structures. The motor cortex is important for accurate paw transport as well as fine digit movements necessary for grasping (Guo et al., 2015; Wang et al., 2017; Lemke et al., 2019; Sauerbrei et al., 2020; Park et al., 2022). Interestingly, motor cortical dopamine is necessary to learn, but not perform, single-pellet reach-to-grasp (Hosp et al., 2011). Cortical control of paw transport depends on thalamic input (Sauerbrei et al., 2020), which in turn receives input from the basal ganglia and cerebellum. Basal ganglia, thalamus, and cerebellum are implicated in the control of paw transport, as opposed to fine digit control for grasping (Lemke et al., 2019; Guo et al., 2021; Lopez-Huerta et al., 2021; Calame et al., 2023). However, this may be because these studies tracked paw position, but not individual digits. The cerebellum seems to be important specifically for online adjustment of reach trajectories. Collectively, these studies suggest a model in which corticostriatal–thalamocortical circuits initiate a sequence of motor commands to transport the paw to the pellet and initiate a well-timed grasp. The phasic dopamine signals observed here presumably provide feedback to adjust kinematic parameters across reaches, while the cerebellum provides online adjustments during individual reaches. Further work will be required to validate this model and identify the specific aspects of reach-to-grasp movements modulated by dopamine, as well as the neural mechanisms by which they are implemented.
Footnotes
We would like to thank Elizabeth Pappas for her help in training mice and Brandon Toth as well as other members of the Burgess lab for their helpful discussion and feedback. This work was supported by a National Institute of Neurological Disorders and Stroke (T32NS007222; A.T.H.), a Brain Research Foundation Seed Grant, and National Institutes of Health (R01DK129366 to C.R.B.).
The authors declare no competing financial interests.
- Correspondence should be addressed to Christian R. Burgess at crburge{at}umich.edu.
