Cortical dopaminergic signaling mediates planning of directional movements Authors: John Chen, Alfredo Fontanini Published: 2025 ABSTRACT The anterior-lateral motor cortex (ALM) has been extensively studied as a model for understanding the cortical mechanisms involved in planning directional movements. ALM neurons show preparatory activity that predict whether the mouse will produce a left- or right-directed lick. While the neural mechanisms underlying ALM dynamics and their links to behavior are beginning to be elucidated, the sources of this directional activity remain unclear. Here, we examined the role of ALM dopaminergic circuits in mice performing a directional licking task. We report that dopamine (DA) signals and the activity of neurons expressing D1 receptors (D1R+) tracked the preparation and execution of licking and demonstrated a bias for planning of licks aimed at the direction contralateral to the recording site. Unilateral optogenetic manipulations of D1R+ neurons or DA afferents affected lateral licking in a way consistent with their role in planning contralateral movements. Together, these results show that cortical DA modulation plays an important role in the motor preparation of directional movements. INTRODUCTION Cortical motor circuits are known to be contralaterally organized, with activity in one hemisphere driving movement on or toward the opposite side of the body (1–4). In recent years, the mouse anterior-lateral motor cortex (ALM) has emerged as a model for studying the planning and execution of directional movement in the context of delayed-response tasks, where sensing a cue and acting upon it are separated by a delay period. Neural activity in ALM links sensory and motor encoding through preparatory firing that begins during stimulus presentation or during the delay period and persists until movement onset (5–9). This preparatory activity can be selective for the direction of the planned movement, as revealed in tasks where mice are required to perform directional licking to laterally positioned lick spouts (“lick left/lick right”). ALM also shows a contralateral (contra) bias, with populations of neurons firing selectively and more strongly in preparation of contra compared to ipsilateral (ipsi) licking (7–9). The behavioral relevance of this bias has been validated by optogenetic experiments, demonstrating that unilateral perturbations of ALM during motor preparation bias licking toward the contra side of stimulation and toward the ipsi side of inhibition (6, 8, 10, 11). In recent years, a great deal of attention has been devoted to studying the neural mechanisms responsible for generating ALM preparatory activity, its direction selectivity, and its bias. While a variety of multiregion, large-scale loops have been identified as important for the genesis of ALM preparatory activity (7, 8, 12–16), the possible contribution of neuromodulators has been largely overlooked. Here, we focus on understanding the contribution of dopamine (DA) to the preparation and execution of directional movements in mice performing a taste-based directional licking task. DA is an ideal candidate due to its well-described role in sensorimotor processes (17–25). Within ALM, DA signaling through DA receptor 1 (D1R) is required during motor preparation in the context of a cued-licking task (26). Antagonism of D1Rs in ALM delays licking initiation, although not execution, and affects preparatory activity for licking. Evidence from the basal ganglia also suggests the involvement of DA in directional bias. DA release and binding in the striatum can be lateralized and directionally selective (25, 27–29). Furthermore, unilateral DA depletion induces asymmetric motor deficits (30, 31). Whether the role of DA in directional movement is also mediated through its direct action on the cortex and whether dopaminergic signaling can contribute to ALM directional bias are unknown. To address this gap, we used a combination of behavioral testing, anatomical methods, in vivo imaging, and optogenetic manipulations. We examined local DA signals and the activity of neurons expressing D1Rs (e.g., D1R+ neurons) in mice performing a taste-based, two–alternative choice (2-AC) task. Mice were trained to discriminate two different taste cues by licking a central spout and associate each cue with directional licking toward one of two lateral spouts after a delay period. Using fiber photometry to monitor DA levels in ALM, we observed a contra bias in DA signals related to preparation and execution of correct licking responses. Cell type–specific two-photon calcium imaging revealed that responses of D1R+ neurons within ALM tracked these DA dynamics and were distinct from those of non-D1R–expressing neurons (D1R−). Specifically, D1R+ neurons exhibited earlier preparatory responses compared to D1R− neurons during lick preparation and encoded a stronger bias for contra licking during both delay and choice periods of correct trials. Optogenetic inhibition of D1R+ neurons and dopaminergic afferents during the delay reduced contra lick performance. Conversely, optogenetic stimulation of dopaminergic inputs to ALM induced a deficit in ipsi lick performance by promoting licks in the contra direction. Together, our results reveal a neuromodulatory mechanism contributing to the planning and execution of directional movements. This mechanism relies on the lateralization of DA release and engagement of D1R-related circuits involved in the control of movement planning. RESULTS Taste-guided, delayed-response sensorimotor task For this study, we trained a total of 40 head-restrained mice in a taste-guided, 2-AC task (15 mice for imaging experiments and 25 mice for optogenetic experiments). Mice were required to make an instrumental response of five licks to a central spout (termed “dry licks”) to receive gustatory stimuli from a central spout [sucrose (suc), 100 mM; salt (NaCl), 100 mM] and report the discrimination of the stimuli with directional licking to lateral spouts positioned on the left and right sides of the mouth (Fig. 1A and fig. S1A). Associations between tastants and lick direction were counterbalanced across mice. Each trial began with a central spout approaching the mouth, with mice licking the spout five times to trigger the delivery of a taste stimulus (pseudorandomly chosen from suc and NaCl). The central spout retracted away from the mouse 1 s after the delivery of the taste stimulus. Similar to other delayed-response 2-AC tasks commonly used to study sensorimotor planning in ALM (6–9), a delay period (3 s) separates taste sampling and lateral lick decisions. After the delay, two lateral spouts approached the mouth, signaling the mouse to report their lateral lick choice. Licking the correct lateral spout led to a water reward. Licking the incorrect lateral spout triggered a timeout (10 s) before the onset of the intertrial interval. In contrast to previous paradigms where mice only prepare a directional licking response, here, they must also plan the initial unidirectional licking to sample the taste stimulus. This design allowed us to study preparation for both central and lateral licking. DA dynamics in ALM during task performance. (A) Schematic of taste-guided, 2-AC task. (B) Task performance (fraction of correct trials). Error bars indicate means ± SEM. (C) Average lick rate during sampling for suc (blue) and NaCl (red) trials. (D) Average lick rate during lateral licks for right (blue) and left (red) trials. Data in (B) to (D) are from n = 15 mice, correct trials only. (E) Representative confocal image of TH immunostaining in ALM and dopaminergic cells in ventral tegmental area (VTA)/substantia nigra pars compacta (SNc; green). Scale bars, 500 μm (left), 100 μm (top right), and 500 μm (bottom right). (F) Top: Schematic for fiber photometry monitoring of DA in ALM with representative post hoc image of fiber placement and GRAB_DA2h expression (green), Scale bar, 500 μm. Bottom: Example GRAB_DA2h trace showing five consecutive trials. (G) Top: Normalized DA z-ΔF/F across trials from a session aligned to the first lateral lick. Colored traces represent trial types (blue, lick right/contra trials; red, lick left/ipsi trials). Bottom: Mean normalized DA z-ΔF/F across each trial type. (H) Left: Mean normalized DA z-ΔF/F aligned to the first dry lick for suc (blue) and NaCl (red) trials. Right: Same as left for contra (blue) and ipsi (red) lick trials aligned to first lateral lick. Shaded area indicates SEM. Gray-shaded rectangles represent 1.5-s time windows for sample, delay, and choice periods. (I) Mean AUC during each behavioral epoch for each trial type (sampling: blue, suc; red, NaCl; delay and choice: blue, contra; red, ipsi) across sessions. **P < 0.01 and ***P < 0.001. (J) Left: Mean taste selectivity of DA signals aligned to first dry lick. Right: Same as left for lick direction selectivity aligned to first lateral lick. (K) Mean direction selectivity during each behavioral epoch compared to baseline selectivity. ***P < 0.001 and ****P < 0.0001. Data in (H) to (K) are from n = 31 sessions, correct trials only. To quantify the baseline behavior in the task, we analyzed behavioral parameters from mice that did not undergo optogenetic perturbations (n = 15 mice). Once trained to criteria (see Methods), mice showed comparable performance for each tastant, indicative of learning the specific contingencies [n = 15 mice; suc performance, 86.7 ± 1.5%; NaCl performance, 87.4 ± 1.5%; paired t test, t(14) = −0.55, P = 0.59; Fig. 1B]. Mice did not show biases in lick rates for sampling either tastant [suc lick rate, 8.3 ± 0.1 Hz; NaCl lick rate, 8.4 ± 0.1 Hz; paired t test, t(14) = 0.34, P = 0.74; Fig. 1C] or toward either lateral spout [right, 8.0 ± 0.2 Hz; left, 8.4 ± 0.2 Hz; paired t test, t(14) = −2.01, P = 0.065; Fig. 1D]. Lick rates for central and directional licking were comparable [central, 8.4 ± 0.12 s; lateral, 8.2 ± 0.17 s; paired t test, t(14) = 0.64, P = 0.53; fig. S1B]. In addition, mice showed similar reaction times to lateral licks for both trial types [right, 0.76 ± 0.04 s; left, 0.76 ± 0.05 s; paired t test, t(14) = 0.20, P = 0.84; fig. S1, C and D]. Comparable behavioral performance between tastants and lack of lateral lick bias indicate that trained mice can discriminate tastants to prepare and execute symmetric lateral lick responses. DA dynamics in ALM during task performance Disruption of DA signaling through D1Rs in ALM slows lick initiation and delays preparatory activity of ALM neurons (26). We confirmed the presence of dopaminergic inputs within ALM using tyrosine hydroxylase (TH) staining (Fig. 1E). To track local DA release, we used fiber photometry and unilaterally recorded fluorescence emitted by the genetically encoded DA sensor GRAB_DA2h (Fig. 1F) (32). The fiber was implanted in the superficial layers. We observed reliable and robust changes in extracellular DA levels within ALM of seven well-trained mice performing the task (Fig. 1, F and G). When DA signals were aligned to the first dry lick, we observed phasic responses with onsets that preceded licking of the central spout and decay near termination of central licking (n = 31 sessions, 7 mice; time to half maximum, 1.38 ± 0.48 s aligned to first dry lick; Fig. 1H, left). Realigning DA signals to the lateral lick revealed that DA levels slowly increased from the delay into the choice periods, reaching a maximum in the choice period (time to half maximum, 1.38 ± 0.29 s aligned to first lateral lick; Fig. 1H, right). DA release was observed during the preparation to central licking and lateral licking but with significantly different onset times [n = 31; central lick, −0.22 ± 0.09 s; lateral lick, −1.27 ± 0.45 s; paired t test, t(30) = 12.0, P < 0.001; fig. S2A]. No significant transients were observed in a control group of mice injected with the DA-insensitive control sensor GRAB_DAmut (n = 4 mice, 11 sessions; Fig. 1H and fig. S2, B and C). We examined differences in the area under the curve (AUC) of DA transients to taste stimuli (NaCl versus suc) during the sampling period (1.5 s after taste delivery) and to lick direction (left versus right) during the delay (1.5 s before lateral lick) and choice periods (1.5 s after lateral lick). DA responses to NaCl and suc were similar [n = 31 sessions; suc, 4919 ± 301 arbitrary units (au); NaCl, 4771 ± 305 au; paired t test, t(30) = 1.66, P = 0.11; Fig. 1I]. Realignment of DA signals to the time of taste delivery confirmed the lack of significant difference between responses to tastants during sampling (fig. S2D). Recategorization of DA signals based on whether the tastant instructs a contra versus ipsi lick during the sampling period confirmed the lack of directional differences during the sampling period (fig. S2E). In contrast, the DA signal was significantly larger for licking contra to the recorded hemisphere (contra) compared to ipsi (ipsi) licking during the delay period [contra, 3504 ± 247 au; ipsi, 2920 ± 209 au; paired t test, t(30) = 4.37, P < 0.001; Fig. 1I] and choice period [contra, 5707 ± 277 au; ipsi, 4731 ± 250 au; paired t test, t(30) = 6.24, P < 0.001; Fig. 1I]. We further quantified trial-averaged differences in DA signals between suc and NaCl trials as taste selectivity and between contra and ipsi trials as direction selectivity. Taste selectivity during the sample period did not show a bias of DA response for a particular tastant (Fig. 1J, left), suggesting that similar levels of DA are released in response to either taste stimulus and/or DA signals represent mainly licking during this period. In contrast, an enhanced DA response for contra lick trials emerged in the delay period and reached a maximum in the choice period (Fig. 1J, right). DA direction selectivity was only significant during the delay and choice periods [compared to baseline, 0.01 ± 0.01; sample selectivity, 0.02 ± 0.03 DA z-ΔF/F: paired t test, t(30) = −0.17, P = 0.87; delay selectivity, 0.19 ± 0.04 DA z-ΔF/F: paired t test, t(30) = −4.28, P < 0.001; choice selectivity, 0.32 ± 0.05 DA z-ΔF/F: paired t test, t(30) = −6.07, P < 0.0001; Fig. 1K] and only for correct trials. In error trials, a lack of bias in DA direction selectivity during both the delay and choice periods was observed (fig. S2, F to H). To determine the specificity of DA signals in ALM and rule out potential contamination of GRAB_DA2h responses to norepinephrine (NE), we performed fiber photometry recordings using GRAB_NE2h in a separate group of trained mice (fig. S3, A and B). GRAB_NE2h responses were observed during sampling of tastants but did not show taste-specific preferences (fig. S3, C and D). NE signals during the delay did not exhibit a contra direction bias as found in DA signals (fig. S3, C and D). In contrast, NE signals were enhanced for ipsi lick trials during the choice period [direction selectivity compared to baseline; sample selectivity, 0.02 ± 0.02 NE z-ΔF/F: paired t test, t(10) =1.46, P = 0.17; delay selectivity, 0 ± 0.05 NE z-ΔF/F: paired t test, t(10) = 0.17, P = 0.87; choice selectivity, −0.22 ± 0.05 NE z-ΔF/F: paired t test, t(10) = 4.23, P < 0.01; fig. S3E]. Comparison in direction selectivity between DA and NE signals during the delay and choice shows significant differences in directional representation between the two neuromodulators [delay selectivity, 0.19 ± 0.04 DA z-ΔF/F and 0 ± 0.05 NE z-ΔF/F: two-sample t test, t(40) = 2.40, P < 0.05; choice selectivity, 0.32 ± 0.05 DA z-ΔF/F and −0.22 ± 0.05 NE z-ΔF/F: two-sample t test, t(40) = 5.92, P < 0.001; fig. S3F]. These findings demonstrate the specificity of the DA signals recorded in ALM and suggest specific direction-dependent modulation of cortical DA signals during the delay period. Together, these data indicate that DA levels increase during the preparation and execution of licking to both the central and lateral spouts. Both anticipatory and perimovement DA signals showed a contra bias in correct trials. Preparatory and task-related activity in D1R+ neurons We next examined whether the activity of ALM D1R+ neurons follows patterns similar to those observed from DA release. D1R+ neurons were found throughout all laminar layers of ALM (fig. S4A). Anatomical tracing of their projections revealed targets consistent with those of both intratelencephalic (IT) and pyramidal tract (PT) neurons (fig. S4B). We performed dual channel two-photon calcium imaging in D1R-Cre mice crossed with Cre Ai14 reporter mice to express tdTomato in D1R-expressing neurons. RNAscope in situ hybridization probing for D1R mRNA was performed in these mice and confirmed the specificity of expression of tdTomato fluorescence upon Cre recombination under the D1R promoter (fraction of tdTomato-positive cells with D1R mRNA fluorescence, 97.1%; fig. S5, A and B). These mice were injected with pan-neuronal GCaMP7f, and ensembles of neurons were imaged in superficial layers of ALM (image depth, 150 to 350 μm; Fig. 2A). This approach allowed for the differentiation between recorded D1R+ and D1R− neurons (Fig. 2, A and B, and fig. S5, C and D). In total, we monitored activity from 963 D1R+ neurons and 4382 D1R− neurons imaged in randomly selected fields of view of ALM from mice performing the task [n = 26 sessions from eight mice; means ± SD, 38.5 ± 15.7 (number of D1R+ neurons per session) and 168.5 ± 71.1 (number of D1R− neurons per session)]. The fraction of D1R+ neurons imaged per session was 19.3 ± 10.0%, comparable to the overall percentage of cortical neurons that express the receptor in the cortex (33–35). A large number of D1R+ and D1R− neurons in ALM showed task-related responses (task-responsive: D1R+, 47%, 455 of 964; D1R−, 48%, 2101 of 4382). We categorized activity related to lick preparation (central and directional licks) and presentation of taste stimuli. Consistent with previous findings (7–9), neurons in ALM showed responses during the preparation of licking regardless of the type of lick (Fig. 2, C and D, left and middle). Preparatory activity in ALM D1R+ neurons. (A) Schematic for simultaneous calcium imaging of D1R+ and D1R− cells in ALM. (B) Left: Image of an example field of view from two-photon calcium imaging with magnification showing cells expressing only GCaMP7f (D1R− neurons, green) and cells expressing both GCaMP7f and tdTomato (D1R+ neurons, yellow). Right: Example traces of neurons identified in left (D1R+ neurons, magenta; D1R− neurons, green). Scale bar: 50 μm. (C) Example of three D1R+ cells with responses aligned to different behavioral events (taste responses to both suc and NaCl are combined). Top: Heatmap for trials. Bottom: Trial-averaged ΔF/F for trials shown above. Shaded region around traces represents SEM. (D) Same as (C) for D1R− cells. (E) Preparatory responses for central licks aligned to time of first dry lick for D1R+ (133 neurons, magenta) and D1R− (587 neurons, green) neurons. Trial-averaged responses for each neuron were calculated using all trials. (F) Top: Mean onset of preparatory response for central licking (time = 0) exhibited by D1R+ (magenta) and D1R− (green) neurons (*P < 0.05). Bottom: Mean time to peak of preparatory response (**P < 0.01). Error bars indicate SD of bootstrap resampling (n = 1000 iterations). (G) Same as (E) but for preparatory responses to lateral licks aligned to time of first lateral lick for D1R+ (150 neurons, magenta) and D1R− (621 neurons, green) neurons. (H) Top: Mean onset of preparatory response for lateral licking (time = 0) exhibited by D1R+ (magenta) and D1R− (green) neurons (*P < 0.05). Bottom: Mean time to peak of preparatory response (*P < 0.05). Having observed anticipatory DA release that emerged before the first dry lick (Fig. 1), we first examined preparatory activity in anticipation of licking the central spout. We analyzed neural responses aligned to the first dry lick and identified individual D1R+ and D1R− neurons exhibiting preparatory activity before central licking (task-responsive neurons: D1R+, 29%, 133 of 455; D1R−, 28%, 587 of 2101; Fig. 2E). To identify differences in the timing of preparatory activity, we computed the onset and time to peak of responses (see Methods). Preparatory activity emerged significantly earlier in D1R+ neurons compared to D1R− neurons (onset for D1R+, −0.55 ± 0.04 s; onset for D1R−, −0.45 ± 0.02 s; bootstrap, P < 0.05; Fig. 2F, top). On average, D1R+ neurons also showed a significantly shorter time to peak in preparation of central licks relative to D1R− neurons (time to peak for D1R+, 1.3 ± 0.05 s; time to peak for D1R−, 1.45 ± 0.02 s; bootstrap, P < 0.01; Fig. 2F, bottom). To examine preparatory activity specific to lateral licks, we analyzed only neurons exhibiting preparatory activity during the delay (task-responsive neurons: D1R+, 33%, 150 of 455; D1R−, 30%, 621 of 2101). Average responses of these cells were aligned to the first lateral lick (both left and right licks are used; Fig. 2G) in correct trials. Similar to the central lick preparation, D1R+ neurons exhibited significantly earlier onset of preparatory activity when compared to D1R− neurons (D1R+, −1.81 ± 0.07 s; D1R−, −1.63 ± 0.03 s; bootstrap, P < 0.05; Fig. 2H, top). In addition, preparatory activity of D1R+ neurons showed a significantly shorter time to peak during the delay period (D1R+, 2.24 ± 0.08 s; D1R−, 2.44 ± 0.04 s; bootstrap, P < 0.05; Fig. 2H, bottom). In addition to lick-related responses, we identified neurons responsive to taste stimuli (see Methods and Fig. 2, C and D, right). Taste-related responses beginning after the initiation to licking during the sampling period were sparse in both types of neurons (task-responsive neurons: D1R+, 14%, 65 of 455; D1R−, 11%, 242 of 2101). No differences between D1R+ and D1R− neurons were observed in onset and time to peak for taste responses (onset time: D1R+, −0.80 ± 0.07 s; D1R−, 0.78 ± 0.03 s; bootstrap, P = 0.59; time to peak: D1R+, 1.80 ± 0.07 s; D1R−, 1.69 ± 0.04 s; bootstrap, P = 0.88; fig. S5, E and F). No differences were observed in the magnitude of responses to suc and NaCl in D1R+ neurons (suc, 2.96 ± 0.49 ΔF/F; NaCl, 3.15 ± 0.59 ΔF/F; bootstrap, P = 0.42). Average responses to each tastant during the sampling period were comparable between the two cell groups (suc: D1R+, 2.96 ± 0.49 ΔF/F; D1R−, 2.66 ± 0.30 ΔF/F; bootstrap, P = 0.30; NaCl: D1R+, 3.15 ± 0.59 ΔF/F; D1R−, 2.64 ± 0.59 ΔF/F; bootstrap, P = 0.20; fig. S5, G and H). These data support D1R+-specific representation of preparatory activity related to licking, regardless of the type of lick (centrally versus laterally targeted) and tastant. Earlier onset and faster ramping in preparatory activity are specific to neurons expressing D1Rs in ALM, suggesting a cell type–specific modulation related to DA signaling during motor planning within ALM. Lick direction preference in D1R+ neurons Recording of DA release in ALM showed larger transients when mice prepare and execute correct contra licking than for ipsi licking (Fig. 1). To begin assessing whether D1R+ neurons also showed similar direction bias, we aligned responses to the first lateral lick and compared each neuron’s contra and ipsi correct lick trials. Both D1R+ and D1R− neurons can show selectivity for lick direction during the delay and choice period of the task (Fig. 3A). However, we observed significantly more contra-preferring than ipsi-preferring neurons among D1R+ direction-selective neurons, suggesting an overall contra preference within the D1R+ population (contra-preferring: D1R+, 59.1%, 65 of 110; χ2(1) = 7.27, P < 0.01; Fig. 3B). In contrast, there were similar numbers of contra-preferring and ipsi-preferring in the D1R− population (contra-preferring: D1R−, 52.4%, 281 of 536; χ2(1) = 2.51, P = 0.11; Fig. 3B). Additional analyses were performed to determine direction selectivity during different behavioral epochs. Contra bias in D1R+ neurons was observed during both the delay and choice periods, whereas a weaker contra bias was only seen during the delay period and not for the choice period for the D1R− population (D1R+: delay, 66.7% contra-preferring, 30 of 45: χ2(1) = 10.0, P < 0.01; choice, 60.9% contra-preferring, 56 of 92: χ2(1) = 8.70, P < 0.01; D1R−: delay, 57.5% contra-preferring, 131 of 228: χ2(1) = 10.1, P < 0.01; choice, 52.1% contra-preferring, 232 of 445: χ2(1) = 1.62, P = 0.20; fig. S6A). Lick direction preference in D1R+ neurons. (A) Top: Examples of four D1R+ neurons with lick direction selectivity during delay and choice periods aligned to the first lateral lick event (dashed line; time = 0 s). Bottom: Examples of four D1R− neurons. Correct contra (blue) and correct ipsi (red) trials are shown. Green circles represent the time of first central lick. Shaded area indicates SEM. Gray rectangles indicate delay and/or choice epochs. (B) Overall fraction of contra-preferring (blue) and ipsi-preferring (red) neurons for D1R+ and D1R− direction-selective neurons (**P < 0.01). (C) Top: Heatmap of mean selectivity normalized to peak for each D1R+ and D1R− selective neurons aligned to the first lateral lick event. Neurons are sorted by their preference; contra-preferring, blue; ipsi-preferring, red. Bottom: Mean selectivity for contra-preferring (blue) and ipsi-preferring (red) D1R+ and D1R− selective neurons (correct trials only). (D) Left: Mean selectivity during delay and choice epochs for contra-preferring (blue) and ipsi-preferring (red) neurons of D1R+ neurons (*P < 0.05). Right: Same for D1R− neurons. Error bars indicate SD of bootstrap resampling (n = 1000 iterations). (E) Left: Mean selectivity during delay and choice epochs for contra-preferring D1R+ (magenta) and D1R− (green) neurons (*P < 0.05). Right: Same for ipsi-preferring D1R+ and D1R− neurons. (F) Left: Mean population selectivity normalized to the peak of direction-preferring neurons (both contra- and ipsi-preferring neurons) for each cell population (D1R+, magenta; D1R−, green). Right: Quantification of normalized selectivity during the delay and choice periods (*P < 0.05). (G) Left: Distribution of onset times in direction selectivity for contra-preferring of each cell population (D1R+, magenta; D1R−, green; *P < 0.05). Right: Same for ipsi-preferring D1R+ and D1R− neurons. Data in (B) to (G) from direction-selective cells (D1R+, 65 contra-preferring and 45 ipsi-preferring; D1R−, 281 contra-preferring and 255 ipsi-preferring). We next examined the magnitude of direction selectivity within each population by computing the difference between trial-averaged contra-lick and ipsi-lick activity for each direction-selective neuron (correct trials only; Fig. 3C). Within the D1R+ direction-selective population, the average selectivity in contra-preferring D1R+ neurons was significantly larger than ipsi-preferring D1R+ neurons during both the delay and choice periods (D1R+: delay, contra: 2.60 ± 0.57 ΔF/F, ipsi: 1.22 ± 0.23 ΔF/F; bootstrap, P < 0.05; choice, contra: 4.41 ± 0.82 ΔF/F, ipsi: 2.62 ± 0.33 ΔF/F; bootstrap, P < 0.05; Fig. 3D, left, and fig. S6, B and Bb). In contrast, contra-preferring and ipsi-preferring D1R− neurons showed similar strengths in selectivity (D1R−: delay, contra: 1.80 ± 0.14 ΔF/F, ipsi: 1.69 ± 0.15 ΔF/F; bootstrap, P = 0.29; choice, contra: 3.62 ± 0.25 ΔF/F, ipsi: 3.55 ± 0.29 ΔF/F; bootstrap, P = 0.45; Fig. 3D, right, and fig. S6, B and Bb). Activity biased for contra preparation is thus enhanced specifically within D1R+ neurons. Comparison between cell populations showed stronger selectivity in D1R+ contra-preferring cells than that in D1R− contra-preferring cells only during the delay period (contra-preferring: delay, D1R+: 2.60 ± 0.57 ΔF/F, D1R−: 1.80 ± 0.14 ΔF/F; bootstrap, P < 0.05; choice, D1R+: 4.41 ± 0.82 ΔF/F, D1R−: 3.62 ± 0.25 ΔF/F; bootstrap, P = 0.12; Fig. 3E, left, and fig. S6, C and Cc). Conversely, a trend of weaker selectivity in D1R+ ipsi-preferring cells compared to D1R− ipsi-preferring cells was observed (ipsi-preferring: delay, D1R+: 1.22 ± 0.23 ΔF/F, D1R−: 1.69 ± 0.15 ΔF/F; bootstrap, P = 0.09; choice, D1R+: 2.62 ± 0.33 ΔF/F, D1R−: 3.55 ± 0.29 ΔF/F; bootstrap, P = 0.08; Fig. 3E, right, and fig. S6, C and Cc). To further compare the overall representation of direction preference between the two cell populations, we normalized the selectivity of each neuron to compute a selectivity index (see Methods). Similar to the patterns of direction preference found in DA signals in the task (Fig. 1), the overall selectivity index of D1R+ neurons was biased toward contra licks (Fig. 3F, left) and greater than that of D1R− neurons throughout the delay and choice periods (delay, D1R+: 0.12 ± 0.04, D1R−: 0.04 ± 0.02; bootstrap, P < 0.05; choice, D1R+: 0.14 ± 0.06, D1R−: 0.02 ± 0.03; bootstrap, P < 0.05; Fig. 3F, right). Separate analyses of error trials revealed that overall selectivity in D1R+ neurons was reduced in error trials and became similar in magnitude to that in D1R− neurons (fig. S6D). Because of the limited number of error trials performed in the task and the limited bias under this condition, all the following analyses will focus exclusively on ALM activity in correct trials. Preparatory responses for licking appear earlier in the delay period for D1R+ neurons (Fig. 2). We analyzed the onset of selectivity in contra- and ipsi-preferring neurons for both cell populations. contra selectivity emerged earlier in D1R+ neurons than in D1R− neurons during the delay period (onset of selectivity; contra-preferring: D1R+, −0.69 ± 0.05 s; D1R−, −0.53 ± 0.03 s; bootstrap, P < 0.05; Fig. 3G, left). In contrast, ipsi selectivity of D1R+ neurons began closer to the time of the lick response and after D1R− neurons (onset of selectivity; ipsi-preferring: D1R+, −0.41 ± 0.03 s; D1R−, −0.50 ± 0.06 s; bootstrap, P = 0.08; Fig. 3G, right). Thus, direction selectivity emerges the earliest in contra preferring D1R+ neurons. These results demonstrate that D1R+ neurons show enhanced contra preference, a behavior consistent with local DA signals and distinct from cells that do not express D1Rs in ALM. Strength and stability of direction coding in the population of D1R+ neurons Upon demonstrating differences in encoding of direction with single-neuron analyses, we examined overall direction selectivity at the population level, pooling all the recorded neurons from each group (D1R+ and D1R−) and computing population selectivity during the delay period (WTdelay, contra−ipsi; see Methods). To determine whether one population of neurons encoded stronger direction selectivity than the other, we performed a permutation test by shuffling the cell identity for each D1R+ and D1R− neuron and computing the difference in the population selectivity for each shuffle (1000 iterations). The true mean of the difference in selectivity was positive and significantly larger than the shuffled distribution [null: 95% confidence interval (CI), −0.101 to 0.103; observed mean difference in population selectivity, 0.104 ΔF/F; permutation test, P < 0.05; Fig. 4A], indicating an enhanced representation of population selectivity by the D1R+ neurons. Strength and stability of direction coding in the population of D1R+ neurons. (A) True and shuffled distributions of the mean difference in population selectivity between D1R+ and D1R− neurons during the delay and choice periods. Solid vertical lines represent the mean of each distribution, and dashed lines represent the 95% CI of shuffled distribution. (B) Left: Population activity of D1R+ neurons along WTdelay that differentiates lick direction during the delay period (contra, blue; ipsi, red). Right: Same as left for D1R− neurons. Shaded region around traces represents SD of bootstrap resampling (n = 1000 iterations). Gray rectangles indicate the delay epoch. (C) Average difference between contra and ipsi population activity along WTdelay for D1R+ (magenta) and D1R− (green) populations showing stronger population selectivity in the D1R+ population (*P < 0.05). Data in (A) to (C) are from all recorded D1R+ (964 neurons) and D1R− neurons (4382 neurons) (D) Pairwise Pearson’s correlation of selectivity between individual time points during the delay and choice periods for direction-selective D1R+ and D1R− neurons (correct trials only). (E) Left: Pearson’s correlation of population selectivity between each time point on the x axis and the mean population selectivity of the choice period (1 s after lateral lick onset) for contra-preferring D1R+ (magenta) and D1R− neurons (green). Right: Quantification of average correlation in population selectivity during the delay period (*P < 0.05). Error bars indicate SD of bootstrap resampling (n = 1000 iterations). (F) Left: Same as (E) for ipsi-preferring D1R+ and D1R− neurons (bootstrap difference in mean correlation, P = 0.64). Data in (D) to (F) are from direction-selective neurons (D1R+, 65 contra-preferring and 45 ipsi-preferring; D1R−, 281 contra-preferring and 255 ipsi-preferring). To further investigate and visualize population selectivity, we analyzed neural activity in an activity space, where each dimension represented activity of an individual neuron (10, 12, 36). We defined a population selectivity vector by computing distinct axes in the activity space that best discriminated trial types based on lick direction during the delay period for each group of D1R+ and D1R− neurons separately (see Methods). Trial-averaged activity for all neurons was then projected along the axes in activity space that maximally discriminated lick direction for each population (both correct contra and ipsi trials). Activity projected onto WTdelay contained direction-selective preparatory activity for D1R+ and D1R− populations (Fig. 4B). The difference in contra and ipsi activity (selectivity) was significantly larger in the D1R+ population compared to D1R− population (D1R+, 3.7 × 10−3 ± 0.7 × 10−3 au; D1R−, 2.1 × 10−3 ± 0.3 × 10−3 au; bootstrap difference in mean selectivity between populations, P < 0.05; Fig. 4C). This result is consistent with our population permutation analysis in supporting stronger representation of direction selectivity by D1R+ neurons. We further confirmed the differences in population activity with principal components analysis (PCA; see Methods). Visual inspection of contra and ipsi trial activity represented in the first three principal components (variance; D1R+: PC 1, 65.0%; PC 2, 11.4%; PC 3, 8.7%; D1R−: PC 1, 56.4%; PC 2, 14.5%; PC 3, 9.7%; fig. S7, A and B), and activity trajectories in PC space confirmed a strong difference in activity related to lick direction by the D1R+ population (fig. S7C). We quantified this difference as the Euclidean distance in PC space between contra and ipsi activity. D1R+ neurons showed a stronger increase in the Euclidean distance during the delay period in comparison to their D1R− counterpart (D1R+, 57.1 ± 10.9 au; D1R−, 34.9 ± 10.4 au; bootstrap difference in mean distance, P < 0.05; fig. S7D), supporting a stronger representation in direction coding by this population of ALM neurons. We next investigated the maintenance of selectivity in the populations of D1R+ and D1R− neurons between the delay and choice periods. If direction selectivity encoded during motor planning persists into lick choice, then selectivity should be similar throughout the delay and choice periods. To examine the stability of selectivity for both D1R+ and D1R− direction-selective neurons (including contra- and ipsi-preferring neurons), we computed the Pearson’s correlation between the population vectors for pairs of time points during the two behavioral periods using matched numbers of cells. Visual inspection of Fig. 4D suggested that the pairwise correlation of selectivity was temporally broader for D1R+ neurons compared to D1R−. That is, selectivity remained more similar across the delay and choice periods for D1R+ than D1R− neurons, suggesting persistent encoding of lick direction. We further quantified the correlation in population selectivity for each time point during the delay with the average population selectivity during the choice period (1 s after lateral lick onset). Selectivity across the delay period showed higher correlation to the mean selectivity in the choice for contra-preferring D1R+ neurons compared to contra-preferring D1R− neurons (D1R+, 0.76 ± 0.09; D1R−, 0.52 ± 0.04; bootstrap difference in mean correlation, P < 0.05; Fig. 4E). In contrast, correlation in selectivity between the delay and choice periods was comparable for ipsi-preferring D1R+ neurons and ipsi-preferring D1R− neurons (D1R+, 0.45 ± 0.09; D1R−, 0.49 ± 0.07; bootstrap difference in mean correlation, P = 0.64; Fig. 4F). Within D1R+ neurons, the correlation of selectivity for contra-preferring neurons was significantly higher than that for ipsi-preferring neurons (contra-preferring, 0.76 ± 0.09; ipsi-preferring, 0.49 ± 0.07; bootstrap difference in mean correlation, P < 0.05; fig. S7E). This was not the case for D1R− neurons (fig. S7F). This suggests that contra selectivity arising earlier during the delay is maintained into the choice period by similar groups of D1R+ neurons, while different groups of D1R− neurons represent contra selectivity during different behavioral epochs. In summary, these results suggest that populations of D1R+ and D1R− neurons in ALM exhibit differences in direction-selective activity, with stronger and persistent contra bias found in the D1R+ population only. Epoch-specific perturbations of DA circuits in ALM To test the functional role of D1R+ neurons in guiding directional behaviors, we optogenetically inhibited their activity during different behavioral periods of the task. To specifically target D1R+ neurons in ALM, we injected D1R-Cre mice with Cre-dependent soma-targeted (st) GtACR2 (Fig. 5A) (37, 38). Unilateral inhibition of left or right ALM occurred either during the sampling or the delay period for 30 to 40% of trials (Fig. 5B). Each condition (sample/delay and right/left hemisphere) was tested for each subject in distinct sessions. Unilateral inhibition of D1R+ neurons in ALM during the sampling period did not significantly perturb performance in either contra- or ipsi-instructed trials [n = 5 mice; contra: control, 85.8 ± 3.8%; inhibition, 82.8 ± 5.5%; paired t test, t(4) = −1.45, P = 0.22; ipsi: control, 87.1 ± 3.4%; inhibition, 86.9 ± 5.5%; paired t test, t(4) = −0.15, P = 0.88; Fig. 5C and fig. S8A]. However, silencing during the delay period significantly reduced performance only in contra trials, while performance was maintained in ipsi trials [n = 5 mice; contra: control, 81.4 ± 2.5%; inhibition, 63.7 ± 5.0%; paired t test, t(4) = −12.89, P < 0.001; ipsi: control, 83.7 ± 2.9%; inhibition, 85.8 ± 5.2%; paired t test, t(4) = 1.56, P = 0.19; Fig. 5C and fig. S8A]. Thus, inhibition of D1R+ neurons in ALM is sufficient to cause an ipsi lick bias. To assess changes in latency of lateral lick response, we computed the change in reaction times between trials with and without inhibition for both contra and ipsi trials. Differences in reaction times between the two trial types were similar during the sampling period, suggesting no directional effect on response latency due to inhibition [contra, 0.00 ± 0.05 s; ipsi, 0.01 ± 0.02; paired t test, t(4) = −0.31, P = 0.77; fig. S8B]. However, we observed an increase in reaction times for contra trials that was significantly greater than those for ipsi trials during delay inhibition [contra, 0.14 ± 0.05 s; ipsi, 0.04 ± 0.03, paired t test, t(4) = 2.94, P < 0.05; fig. S8B]. In a different group of D1R-Cre mice injected with control virus (FLEX-tdTomato), no change in performance following light stimulation was observed for either contra or ipsi trials during the sampling or delay periods (fig. S9, A and B). We further compared contra and ipsi performance between experimental and control groups and observed a significant difference only in contra performance during delay inhibition (fig. S9, C and D). Epoch-specific optogenetic manipulations of DA circuits in ALM. (A) Left: Schematic for optogenetic inhibition of D1R+ neurons in ALM. Right: Representative post hoc image of SIO-stGtACR2 (red) expression in D1R-Cre mice with nuclear counterstain (blue). Scale bars, 500 μm (top) and 100 μm (bottom). (B) Top: Unilateral inhibition was performed on either left or right hemisphere of ALM. Contra (blue) and ipsi (red) licks are determined relative to the hemisphere perturbed. Bottom: Schematic of trial structure and windows of optoinhibition. (C) Behavioral performance with (stim trials) and without (control trials) unilateral inhibition of D1R+ neurons in ALM during the sampling and delay periods (n = 5 mice; **P < 0.001). Thick lines represent mean, and thin lines represent individual mice. (D) Left: Schematic for optogenetic stimulation of dopaminergic afferents in ALM. Right: Representative post hoc image of DIO-ChR2 (green) expression in DAT-Cre mice with nuclear counterstain (blue). Scale bars, 100 μm (top), 200 μm (bottom left), and 50 μm (bottom right). (E) Same as (B) for unilateral ChR2 activation of dopaminergic afferents. (F) Behavioral performance with (stim trials) and without (control trials) unilateral DA stimulation in ALM during the sampling and delay periods (sampling, n = 4 mice; delay, n = 5 mice; **P < 0.01). (G) Left: Schematic for optogenetic inhibition of dopaminergic afferents in ALM. Right: Representative post hoc image of FLEX-JAWS (red) expression in DAT-Cre mice with nuclear counterstain (blue). Scale bars, 100 μm (top), 100 μm (bottom left), and 20 μm (bottom right). (H) Same as (B) and (E) for unilateral JAWS inhibition of dopaminergic afferents. (I) Behavioral performance with (stim trials) and without (control trials) unilateral DA inhibition in ALM during the sampling and delay periods (n = 6 mice; **P < 0.01). To confirm the functional role of ALM DA in biasing directional licking, we unilaterally stimulated DA terminals in ALM during different behavioral periods of the task and tested effects on behavioral performance. If DA plays a role in biasing toward contra licking, then activation of DA axons in ALM during ipsi trials should result in augmented contra responses and reduced performance. To specifically target and stimulate DA inputs in ALM, we injected dopamine transporter (DAT)–Cre mice were injected with Cre-dependent ChR2 in ventral tegmental area (VTA) and locally targeted the light stimulation to ALM (Fig. 5D). Similar to the unilateral optoinhibition experiments, light stimulation occurred during the sampling or delay period in a random subset of trials on either left or right ALM (Fig. 5E). Each condition (sample/delay and right/left hemisphere) was tested for each subject in separate sessions. Unilateral stimulation of DA afferents in ALM during the sampling period did not significantly affect performance of contra or ipsi trials [n = 4 mice; contra: control, 89.1 ± 2.8%; stimulation, 88.0 ± 7.4%; paired t test, t(4) = −0.39, P = 0.72; ipsi: control, 90.5 ± 7.5%; stim, 87.3 ± 7.1%; paired t test, t(4) = −1.79, P = 0.17; Fig. 5F and fig. S8C]. However, stimulation of DA terminals during the delay period significantly reduced performance in ipsi trials by inducing erroneous contra licking, while performance was maintained in contra trials [n = 5 mice; contra: control, 85.9 ± 8.2%; stimulation, 89.1 ± 8.4%; paired t test, t(4) = 1.96, P = 0.12; ipsi: control, 79.4 ± 4.9%; stimulation, 66.3 ± 5.3%; paired t test, t(4) = −5.79, P < 0.01; Fig. 5F and fig. S8C]. Reaction times were not altered with activation of DA inputs in ALM (fig. S8D). DAT-Cre mice injected with control virus [DIO–enhanced yellow fluorescent protein (EYFP)] did not show changes in performance following light stimulation for either left or right hemispheres during the sampling or delay periods (fig. S9, E and F). Additional comparison of contra and ipsi performance between experimental and control groups confirmed a significant difference only in ipsi performance during delay inhibition (fig. S9, G and H). To further confirm the role of DA in promoting contra licking, we unilaterally inhibited DA afferents in ALM during the sampling and delay periods and assessed behavioral effects. To specifically target and inhibit DA inputs in ALM, we injected DAT-Cre mice with Cre-dependent (FLEX) JAWS in VTA and targeted the red light above ALM (Fig. 5, G and H). As for the experiments above, each condition (sample/delay and right/left hemisphere) was tested for each subject in separate sessions. Unilateral inhibition of DA afferents in ALM during the sampling period did not significantly affect performance of contra or ipsi trials [n = 6 mice; contra: control, 88.8 ± 6.6%; inhibition, 90.1 ± 8.6%; paired t test, t(5) = 1.30, P = 0.25; ipsi: control, 87.7 ± 8.7%; inhibition, 84.0 ± 13.5%; paired t test, t(5) = −1.08, P = 0.33; Fig. 5I and fig. S8E]. However, similar to the effects observed under optoinhibition of D1R+ neurons, silencing dopaminergic terminals during the delay period disrupted contra lick performance, while ipsi performance remained intact [n = 6 mice; contra: control, 82.8 ± 6.2%; inhibition, 62.5 ± 13.3%; paired t test, t(5) = −4.11, P < 0.01; ipsi: control, 93.2 ± 4.2%; inhibition, 94.1 ± 3.6%; paired t test, t(5) = 2.01, P = 0.10; Fig. 5I and fig. S8E]. The performance for unperturbed contra trials decreased compared to the performance for unperturbed ipsi trials [n = 6 mice; control: contra, 82.8 ± 6.2%; ipsi, 93.2 ± 4.2%; paired t test, t(5) = 5.30, P = 0.003]. This effect may be ascribed to a carryover of DA terminals perturbation for contra trials within a single session. No significant change in reaction times were observed during the sampling period [contra, 0.00 ± 0.03 s; ipsi, 0.03 ± 0.11; paired t test, t(4) = −0.80, P = 0.46; fig. S8F]. However, we observed an increase in reaction times for contra trials compared to ipsi trials during delay inhibition [contra, 0.12 ± 0.03 s; ipsi, 0.03 ± 0.03; paired t test, t(4) = −4.34, P < 0.01; fig. S8F]. Therefore, inhibition of DA afferents in ALM disrupts contra lick performance. These findings demonstrate that dopaminergic input is required in the motor preparation of contra licking. Together, these data demonstrate that activity of the D1R+ population in ALM is required during the preparation of contra lick decisions and correct timing of licking. Furthermore, optogenetic manipulation of DA fibers in ALM is sufficient to bias licking in a way that is consistent with the role of DA in driving D1R+ neurons toward preparation of contra movements (Figs. 1 and 3). DISCUSSION The findings presented in this study demonstrate the involvement of cortical dopaminergic circuits in planning and executing directional movements. More specifically, the experiments show that DA and D1R+ neurons play a role in shaping preparatory activity in the ALM of mice performing a delayed-response task guided by gustatory stimuli. In this task, mice were trained to discriminate tastants (suc or NaCl) sampled from a central spout, plan their response during a delay period, and report the identity of the sampled tastant by licking one of two lateral spouts for water reward. Using fiber photometry in combination with a DA sensor, GRAB_DA, we observed changes in DA signals within ALM that tracked the preparation and execution of both central and lateral licking. The experiments demonstrated a directional bias in DA levels in the form of enhanced activity in correct contra lick trials during both the delay and choice periods. To assess the effects of this bias on cellular activity, we imaged calcium signals of D1R+ neurons and compared them to those of D1R− neurons. Calcium imaging showed earlier preparatory activity for central and lateral licking in D1R+ neurons when compared to D1R− neurons. Furthermore, D1R+ neurons were predominantly selective for contra licking, while D1R− neurons showed similar preference for contra and ipsi licking. At the population level, directional selectivity and preference were found to be enhanced in the D1R+ population during preparation of lateral lick choice. To confirm the functional significance of the contra bias observed for ALM dopaminergic circuits, we performed both optogenetic stimulation and silencing experiments. We relied on silencing of D1R+ neurons and inactivation of dopaminergic afferents in ALM to test whether they could reduce contra licking. Conversely, we tested whether the activation of dopaminergic afferents could promote contra licking. Optogenetic inhibition of both D1R+ neurons and DA inputs in ALM specifically reduced contra lick performance. Optostimulation of dopaminergic afferents provided results consistent with the role of DA in driving this contra bias. Activation of DA afferents produced a significant bias of ipsi trials toward contra licking. While we did observe a small directional trend toward increased performance in contra lick trials in the case of stimulation of dopaminergic afferents (Fig. 5F), this change was not significant, possibly attributable to a ceiling effect imposed by the high task performance (~87%). The optogenetic manipulations of DA afferents presented here relied on the DAT-Cre transgenic mice. Previous studies have shown higher ectopic expression upon Cre-mediated recombination in TH-Cre mouse line compared to DAT-Cre transgenic lines, suggesting higher selectivity in DAT-Cre mice in labeling midbrain DA neurons (39–41). However, DAT-Cre lines exhibit weaker penetrance in targeting dopaminergic cells due to significant heterogeneity in DAT expression throughout the midbrain, with some subpopulations of VTA neurons expressing lower DAT levels (39, 42). Although we cannot exclude additional effects from cells with reduced DAT expression, the experiments presented here demonstrate a critical role for cortical DA in mediating directional movement. Over the past decade, ALM has been extensively studied for its role in planning, initiation and execution of movements, in particular with regard to lateralized tongue movements. Two modes of ALM neural activity have been identified as important for tongue movements. The first mode is represented by nondirectional ramping activity preceding the initiation of licking and is believed to represent a commitment to move (43–46). The second mode involves direction-selective preparatory firing and represents planning for a specific lateral lick (9, 12, 47). Optogenetic manipulations of neural activity either locally in ALM or across multiple regions connected to it have been used to identify the sources of nondirectional and directional preparatory activity. Nondirectional, ramping activity is likely to originate externally to ALM, with long-range, multiregion loops involving areas such as the thalamus and basal ganglia (10, 45, 48). Unlike nondirectional ramping, directional preparatory activity requires intracortical activity to be maintained during the planning phase of the task (10). Here, we provide evidence for the existence of a neuromodulatory mechanism for directionality that relies on dopaminergic inputs from the VTA and intra-ALM circuits characterized by the expression of D1Rs. DA is known to be involved in movement, including licking and tongue movements (49–53). Dopaminergic neurons in the VTA and substantia nigra pars compacta (SNc) encode a series of processes and variables associated with movement, such as planning, directionality, timing, vigor, and speed (23, 54, 55). With regard to movement preparation, DA neurons were shown to fire in anticipation of intentional movements. DA neurons are activated before the onset of self-initiated movements and exhibit slow ramping on the second-long timescale in both rodent and primate studies (17, 24, 56–58). As for directionality, dopaminergic neurons show a contra bias with lateralized and directionally selective DA release (25, 27–29). For instance, in the context of a reversal learning task in which mice must press one of two lateral levers following a central nose poke, VTA and SNc neurons projecting to the dorsomedial striatum show stronger responses for trials involving contra lever pressing compared to ipsi trials (27). Similar results have been observed for dopaminergic axons and DA release in the basal ganglia (28, 29). In this study, we focused on VTA as the source of dopaminergic inputs for ALM, given its role in modulating prefrontal circuits during delayed-response tasks requiring movement timing (19) and motor cortex in reward-related motor learning (59). Our findings demonstrate a critical function of the VTA-ALM connection in motor planning of directional licking. However, the role of SNc should not be ruled out, and more work is needed to elucidate the function of its inputs in ALM. Compared to what we know about dopaminergic circuits in the basal ganglia, less is known on the function of cortical DA in movement and action. Studies on the medial prefrontal cortex (mPFC) provide clues on the role of dopaminergic inputs from the VTA. mPFC DA is likely involved in modulating ramping activity and plays a role in action control as reported in the context of an interval timing task in rats (19, 60). This effect on action timing appears to be mediated by DA acting on D1R+ neurons. To our knowledge, only one report of DA in ALM has been published. In that study (26), antagonism of D1R in ALM of mice performing a cued-licking task delayed preparatory activity and the onset of licking, suggesting that DA signaling through D1R is involved in the planning and initiation of centrally directed licking (26). As for the directionality of movement, unlike for the basal ganglia, the role of cortical DA in contra bias has not been established yet. The results presented here are the first demonstration of DA signals and engagement of D1R+ neurons in ALM. Furthermore, the data demonstrate that cortical elevations of DA occur before and during central and lateral licks and are stronger for correct contra licks. Analysis of D1R+ neurons further shows that their activity parallels DA dynamics. D1R+ neurons in ALM not only show activity preceding lick initiation, but their patterns of preparatory activity emerge earlier than those of D1R− neurons. D1R+ neurons also show stronger directional selectivity than D1R− and feature a clear contra bias during both the delay and choice periods. Direction selectivity is strongly correlated between the delay and choice periods for D1R+ neurons, suggesting persistent coding for contra licking sustained by dopaminergic inputs. In contrast, selectivity between the delay and choice periods for D1R− neurons is not as strongly correlated, suggesting that different groups of neurons may fire in different behavioral epochs, independently of direct DA modulation. Overall, the similarity in dynamics between DA and D1R+ signals in ALM suggests that DA transmission through D1Rs may drive and sustain contra bias of preparatory activity in the neurons expressing the receptor. Furthermore, the results from optogenetic manipulations, which show a coherent role of DA and D1R+ neurons in guiding contra licking in correct trials, are consistent with the idea that DA may be one of the drivers of D1R+ neurons’ functional specialization. While our data are suggestive of a link between DA dynamics and the contra bias of D1R+ neurons, establishing a causal relationship between these two phenomena will require recording D1R+ neuron spiking activity in response to optogenetic manipulations of dopaminergic terminals. It is worth emphasizing that the selectivity in DA signals and distinct representations found in D1R+ neurons are only present correct trials and not in error trials. This result suggests a potential reward-related component of DA signals in the planning of directional movement. Future studies will be necessary to establish the causal link between DA dynamics to D1R+ neuron firing and understand the mechanisms involved in the interaction between reward anticipation and motor planning in ALM. While DA signals and D1R+ activity show a contra bias and are involved in driving contra licking, they should not be thought of as a circuit operating on an “all-or-none”–labeled line logic. DA signals also occurred for ipsi licks, a sizeable amount of D1R+ neurons preferred the ipsi direction, and D1R− neurons also had a response preference to the contra side. We view the contra function of ALM dopaminergic circuits as a phenomenon emerging from the imbalance in activity between contra- and ipsi-preferring cells. According to this model, the movement bias toward the contra side would emerge when contralaterally preferring DA signals and cellular activity exceed ipsi preferring signals. As for the general elevation of DA and the engagement of D1R+ cells for both trial types, one could speculate that the direction-independent, common pattern may represent a general preparatory signal (26, 45). A full exploration of the circuit-level computational mechanism responsible for controlling the directional bias was beyond the goal of the present study and will be the subject of future investigations. These investigations will further rely on optogenetic perturbations at finer temporal scale to explore DA’s role on ALM dynamics (10). How does the activity of D1R+ neurons relate to what is known of ALM circuits? Of the cell types found in ALM (7, 8, 12), long-range projection neurons have been the best studied and are classically divided into two categories: IT and PT neurons. In the context of motor planning, preparatory activity is observed in both IT and PT neurons, but direction selectivity and contra bias arise predominantly in the PT neurons (7, 8, 61). This projection-specific coding maps onto laminar location of projection neurons in ALM, as IT neurons are found throughout the superficial and deep layers and PT neurons are found only in deep layers. Accordingly, contra bias has been found mostly in the deep layers of ALM. Given what is known from other cortical areas (i.e., mPFC, where 80% of the D1R+ neurons are pyramidal) (33), it is very likely that ALM D1R+ neurons are, for the majority, projection neurons. Our anterograde tracing confirms their projection status and further establishes that they have projection patterns consistent with both IT and PT. While D1R+ neurons represent a mixed population of IT and PT, the ones that we imaged are likely to be mostly IT, since we focused exclusively on the superficial layers of ALM. Hence, our results provide evidence of a D1R+ type of IT neurons that, even in the superficial layers, can distinctively encode direction selectivity with a contra bias. Given that superficial layer IT neurons receive input from sensory cortical areas and thalamus and connect to PT neurons, it is possible that D1R+ IT neurons may be directly involved in sensorimotor transformations and initiate contralaterally selective activity during planning and execution of movement (13, 62–64). Evidence from mPFC showing that D1R+ neurons are widely connected to D1R− neurons supports the idea that once activated by DA, D1R+ neurons may be able to spark and broadly propagate excitatory activity, despite their limited prevalence in ALM (65). Of course, additional mechanisms involving disinhibitory circuits may play a role in the propagation of the preparatory signal. Electrophysiological recordings in ALM during spontaneous licking found inhibitory responses preceding excitatory ones (66). DA has been found to selectively enhance the excitability of vasoactive intestinal peptide interneurons in the superficial layers of the mPFC, suggesting modulation of disinhibitory microcircuits through D1Rs (33). While layer-specific expression of D1Rs by subtypes of interneurons is beginning to be explored (D1Rs are enriched in superficial-layer parvalbumin (PV) and somatostatin (SOM) interneurons, see single-cell sequencing of ALM performed by the Allen Institute), it remains unknown how DA modulates inhibitory microcircuits within ALM through D1 activation. Our results do not exclude a potential role for neurons expressing D2Rs in ALM. Although inactivation of this receptor in this region does not affect the initiation of licking behavior (26), D2Rs’ function is involved in other behavior-related processes such as decision flexibility (67) and cortical plasticity required in motor learning (68–71). In summary, the results presented here provide insights on the function of motor cortical dopaminergic D1R transmission in the planning and execution of directional movements. These data significantly advance our understanding of how DA modulation in the cortex controls movement lateralization and support neuromodulation as a key process in the control of motor actions. METHODS Mice (experimental subjects) This study is based on 49 mice adult mice (10 to 20 weeks old), 40 for imaging and optogenetic experiments and 9 for anatomical experiments. Twenty-five D1R-Ai14 mice were used for fiber photometry, RNA in situ hybridization, and two-photon calcium imaging experiments. D1R-Ai14 mice were bred by crossing D1R-Cre mice [B6.FVB(Cg)-Tg(Drd1-Cre)EY262Gsat/Mmucd, MMRRC, stock no. 030989-UCD] with Ai14 [B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J, the Jackson Laboratory, stock no. 007914]. Ten D1R-Cre mice and 14 DAT-Cre mice [B6.SJL-Slc6a3tm1.1(cre)Bkmn/J, the Jackson Laboratory, stock no. 006660] were used for optogenetic experiments. All mice were maintained under a 12-hour light/dark cycle with ad libitum access to water and chow unless otherwise specified. Both males (n = 22) and females (n = 27) were used, and data were pooled. All experimental protocols were approved by the Institutional Animal Care and Use Committee at Stony Brook University and complied with university, state, and federal regulations on the use and care of laboratory animals (protocol no. IACUC2020–00217). Method details Surgical procedures All mice were anesthetized with an intraperitoneal injection of a cocktail of ketamine (70 mg/kg) and dexmedetomidine (1 mg/kg). Depth of anesthesia was monitored regularly by visual inspection of breathing rate and assessment of tail reflex. The body temperature was maintained at ~35°C using a heating pad (DC temperature control system, FHC, Bowdoin) placed on the surgical stereotaxic apparatus. Once depth of anesthesia was stable, the animal’s head was shaved, disinfected (with iodine solution and 70% alcohol), and fixed on a stereotaxic apparatus. Carprofen (5 mg/kg) and penicillin (150,000 UI/kg) were injected subcutaneously for postoperative reduction of inflammation and antibiotic treatment. For in vivo fiber photometry, a circular 0.5-mm craniotomy was performed above left ALM [centered at 2.4-mm anterior-posterior (AP) and 1.5-mm medial-lateral (ML) to bregma]. Glass pipettes were pulled and beveled to a sharp tip (diameter of ~30 μm) and front-loaded with viral suspension before injection. AAV9-hSyn-GRAB_DA2h (dilution, 1:2 with 0.9% saline solution; titer, 2.5 × 1013; Addgene, catalog no. 140554-AAV9) or AAV5–EF1α–enhanced green fluorescent protein (dilution, 1:2 with 0.9% saline solution; titer, 1.9 × 1013; Addgene, catalog no. 105547-AAV5) was unilaterally injected in the left ALM at a depth of ~200 μm (200 nl, 0.5 nl/s) using a Nanoject syringe pump (UMP3T-1, World Precision Instruments). In a different cohort of mice, AAV9-hSyn-GRAB_NE1h (dilution, 1:1 with 0.9% saline solution; titer, 1.9 × 1013; volume, 200 nl; Addgene, catalog no. 123309-AAV9) was unilaterally injected in left ALM at a depth of ~200 um. The pipette was left in place for at least 5 min before being slowly retracted. After AAV injection, an optical fiber [400-μm core; numerical aperture (NA) = 0.66; Doric Lenses] was slowly lowered into ALM at a depth of ~150 μm and fixed in place using dental adhesive (C&B Metabond, Parkell). For calcium imaging experiments, a circular 3-mm craniotomy was performed above left ALM (centered at 2.4-mm AP and 1.5-mm ML to bregma) in D1R-Ai14 mice. AAV1-Syn-jGCaMP7f-WPRE.SV40 (dilution, 1:4 with 0.9% saline solution; titer, 3.0 × 1013; Addgene, catalog no. 104488-AAV1) was injected at four locations centered around ALM (spaced 400 μm apart) at a depth of 150 μm for each location (50 nl each, at 0.5 nl/s). After each injection, the pipette was left in place for at least 5 min before being slowly retracted. Once all injections were completed, a glass imaging window was implanted over the craniotomy using cyanoacrylate and dental adhesive. The glass imaging window was constructed from three layers of microscope cover glass. For optogenetic inhibition of D1R+ cells, AAV1-hSyn1-SIO-stGtACR2 (dilution, 1:15 with 0.9% saline solution; titer, 1.9 × 1013; Addgene, catalog no. 105677-AAV1) was bilaterally injected in ALM at a depth of 150 μm (150 nl each, 0.5 nl/s). In a different group of mice, control virus AAV9-CAG-FLEX-tdTomato (dilution, 1:15; titer, 2.1 × 1013; volume, 150 nl; Addgene, catalog no. 28306-AAV9) was injected instead. For optogenetic inhibition of DA afferents in ALM, AAV1-CAG-FLEX-JAWS-tdTomato (dilution, 1:1 with 0.9% saline solution; titer, 1.9 × 1013; volume, 300 nl; Addgene, catalog no. 84446-AAV1) was bilaterally injected in VTA [AP, −3.1 mm; ML, ±0.5 mm; dorsoventral (DV), −4.1 and −4.3 mm below brain surface]. For optogenetic stimulation experiments, AAV9-Ef1a-DIO hChR2(E123T/T159C)-EYFP (dilution, 1:5 with 0.9% saline solution; titer, 2.1 × 1013; volume, 300 nl; Addgene, catalog no. 35509-AAV9) was bilaterally injected in VTA (AP, −3.1 mm; ML, ±0.5 mm; DV, −4.1 and −4.3 mm below brain surface). In a different group of mice, control virus AAV9-Ef1a-DIO-EYFP (dilution, 1:5 with 0.9% saline solution; titer, 2.3 × 1013; volume, 150 nl; Addgene, catalog no. 27056-AAV9) was injected instead. After injections, a thin layer of cyanoacrylate was applied on the intact skull. Clear dental acrylic (Jet Repair Acrylic, Lang Dental, part no. 1223-clear) was then applied over the cyanoacrylate. A final layer of clear nail polish was applied (Electron Microscopy Sciences, 72180). For all mice, a customized headpost for head fixation was cemented on the skull (approximately over the cerebellum) using dental adhesive. Animals were allowed to recover for a minimum of 7 days with ad libitum access to water before undergoing water restriction and behavioral training. During water restriction, mice were maintained at ~85% of their body weight and given 1.5 ml of water on days they were not trained or tested. During experimental sessions lasting 1 to 2 hours, mice received all their water or were given supplementary water to maintain body weight. Taste-based, 2-AC task Water-restricted mice were head fixed and trained to perform a taste-based, 2-AC task in a custom-built apparatus. The behavioral apparatus consisted of one central spout and two lateral spouts, all mounted on Zaber motors (X-LSM, Zaber). The positions and speeds of the spouts were controlled by these motors via LabView software. The central spout was constructed with three independent metal tubes, each delivering separate tastants (100 mM suc, 100 mM NaCl, and water) in ~2-μl droplets. Each lateral spout consisted of one metal tube that delivered a drop of water (~3 μl) as a reward. All tastants and rewards were delivered by a gravity-based taste delivery system using solenoid valves (The Lee Company, LHDA123115H). The distance between the center of the lateral spouts was 5 mm. All spouts were electrically coupled to custom-built lick detectors that detect licks via completion of an electrical circuit when the tongue contacts the spouts. After each trial, a rinse with two drops of water was used to clean the central spout and then aspirated using a movable vacuum. Calibration of volumes delivered from each taste tube and lateral spout was performed before each session. To minimize the potential influence of nongustatory cues (odors from tastants and auditory clicks from valves) on task performance, a fan was constantly operating near the central spout to remove possible odors, and constant white noise was played to mask the sound of valve clicks. Mice were trained to associate suc delivered from the central spout with water reward at the left lateral spout and NaCl delivered from the central spout with water reward at the right lateral spout. Taste-lick direction association was counterbalanced across mice. At each trial, the central spout approached the mouth, and mice were required to lick the dry spout five times to trigger the delivery of taste stimulus (pseudorandomly chosen from suc and NaCl). The central spout retracted away from the mouse 1 s after the delivery of taste stimulus. After a delay period of 3 s, two lateral spouts approached the mouth, signaling the mouse to report their lateral lick choice. Licking the correct lateral spout led to a water reward. Therefore, the first correct lateral lick represents reward delivery. Licking the incorrect lateral spout triggered a timeout (10 s) before the onset of the intertrial interval. The intertrial interval was 8 ± 2 s. Trials in which the mouse did not lick the central spout and/or the lateral spouts were rare and occurred near the end of the session (~3% of all trials). Mice were trained until they achieved the performance criterion of ≥75% correct choices for more than three sessions. Mice reached performance criterion after 2 to 3 weeks of training. Recordings or manipulations were performed once the performance criterion was achieved, typically 3 to 4 weeks after surgical procedures. Fiber photometry recordings Unilateral photometry recordings were conducted after mice learned the task (typically 3 to 4 weeks after virus injection). We recorded either GRAB_DA2h or GRAB_NE2h with isosbestic (control) signals simultaneously using a real-time processor (RZ5, Tucker-Davis Technologies) sampling at 1 kHz. The excitation beam for the GRAB_DA2h and GRAB_NE2h signal was produced by a 472/30-nm filtered light-emitting diode (LED) (Thorlabs) at a modulation frequency of 211 Hz. The excitation beam for the isosbestic signal was produced by a 405/30-nm filtered LED (Thorlabs) at a modulation frequency of 310 Hz. Fluorescence excitation and detection were both transmitted through one multimode optical fiber, the output of which was directed through lock-in amplifier (Doric). Fluorescence signals were demodulated at light driver frequencies before data processing. Light intensity emitted at the tip of the patch cord was measured using an optical power meter (Thorlabs, PM100D) and maintained at 30 to 40 μW across sessions. Recordings began ~30 s before the first trial to allow both fluorescent signals to stabilize, and recording frames were synchronized with behavioral events recorded on RZ5P with Synapse software (Tucker-Davis Technologies). Two-photon calcium imaging Calcium imaging experiments took place after mice learned the task (typically 3 to 4 weeks after virus injection). Images were acquired using a movable objective microscope (Sutter) with a resonant scanning module controlled by MScan (Sutter). The objective was a 16× water immersion lens (NA = 0.8, 3-mm working distance; Nikon). jGCaMP7f was excited at 940 nm with a laser power of 50 to 80 mW at the front of the objective from a Ti:sapphire femtosecond laser (Coherent). Images (512 pixels by 512 pixels) were acquired at 31 Hz with a 450 μm–by–450 μm field of view. At the start and end of each session, the laser wavelength was temporally switched to 1010 nm to image tdTomato fluorescence (red channel). A small image stack (±20 μm) was acquired around a particular z plane to capture cells that were out of focus and to account for potential out of focus cells due to chromatic aberration. Imaging started 4 s before each trial start and frame signals were synchronized with behavioral events using a RHD2000 recording system (Intan Technologies). For volumetric imaging of different planes, we used a single-axis piezo nanopositioner (LC400, nPoint Inc.) to collect volumes of two planes spaced by 200-μm distance (at ~150 and ~300 μm; maximum, 400 μm) at 7.75 Hz. Alternating frames were used to avoid flyback and allow settling time of the piezo-controlled objective. Similar to single-plane imaging, laser wavelength was temporally switched to 1010 nm to capture a small image stack centered at each imaging plane to image tdTomato fluorescence. Optogenetic inhibition Unilateral optogenetic inhibition of D1R+ neurons took place after mice learned the task (typically 3 to 4 weeks after virus injection). D1R-Cre mice were bilaterally injected with AAV-SIO-stGtACR2 or AAV-FLEX-tdTomato. A 473-nm laser [100-mW diode-pumped solid-state (DPSS) laser system, Opto Engine LLC] was used to deliver the light from an optic fiber onto the surface of the clear skull above ALM (beam diameter, ~400 μm; centered at 2.4-mm AP and 1.5-mm ML to bregma). A 2-s, 40-Hz temporal profile with a linear ramp down during laser offset (200 ms) was used during the sample or delay period. Average power at skull surface was ~3 mW, and laser intensity was calibrated before each session with an optical power meter (Thorlabs, PM100D). Laser parameters were selected on the basis of published literature (38). Inhibition during the sample epoch began after mice performed the dry licks and ended when the central spout was fully retracted, lasting 2 s. Inhibition during the delay epoch began 2 s before the end of the period. Termination of the sampling inhibition occurred during the very early phase of the delay but did not overlap with the window of optogenetic perturbation for the delay period. Two-second windows were used to account for residual taste-evoked activity that may have persisted after sampling tastants and preparatory activity for lateral licking that may have begun earlier than our analysis window for the delay period (see below). Inhibition occurred during either the sample or delay periods for 30 to 40% of trials with each mouse receiving two to three sessions of inhibition for each behavioral period. Sessions inhibiting left and right ALM alternated across days. Similarly, sessions with inhibition occurring during the sample or delay period were performed separately and also alternated. No difference in performance for unperturbed trials was detected across successive sessions [one-way analysis of variance (ANOVA), F(2,26) = 0.64, P = 0.53]. Unilateral optogenetic inhibition of DA afferents in ALM was performed after mice learned the task. A 633-nm laser (80-mW Cobolt 06-MLD laser system, HÜBNER Photonics) was used to deliver the light from an optic fiber onto the surface of the clear skull above ALM (beam diameter, ~400 μm; centered at 2.4-mm AP and 1.5-mm ML to bregma). A 2-s, 20-Hz temporal profile with a linear ramp down during laser offset (400 ms) was used during the sample or delay period. Average power at skull surface was ~10 mW, and laser intensity was calibrated before each session with an optical power meter (Thorlabs, PM100D). Inhibition occurred during either the sample or delay periods (as defined above) for 30 to 40% of trials. Sessions inhibiting left and right ALM alternated across days. Inhibition occurring during the sample or delay period was also alternated between sessions. No difference in performance for unperturbed trials was detected across successive sessions [one-way ANOVA, F(2,10) = 0.63, P = 0.55]. Optogenetic stimulation Unilateral optogenetic stimulation of dopaminergic terminals in ALM took place after mice learned the task (typically 3 to 4 weeks after virus injection). DAT-Cre mice were bilaterally injected with AAV9-Ef1a-DIO-hChR2(E123T/T159C)-EYFP or AAV9-Ef1a-DIO-EYFP in VTA (AP, −3.1 mm; ML, ±0.5 mm; DV, −4.1 and −4.3 mm below brain surface). A 473-nm laser (100-mW DPSS laser system, Opto Engine LLC.) was used to deliver the light from an optic fiber onto the surface of the clear skull above ALM (beam diameter, ~400 μm; centered at 2.4-mm AP and 1.5-mm ML to bregma). A 2-s, 40-Hz temporal profile (5-ms pulses) was used during the sample or delay period. Average power at skull surface was ~5 mW, and laser intensity were calibrated before each session with an optical power meter (Thorlabs, PM100D). Stimulation parameters were selected on the basis of published literature on cortical stimulation of DA afferents (72–74). In contrast to striatal circuits that receive dense dopaminergic innervation, cortical circuits receive weaker DA release due to sparser innervation and require stronger stimulation parameters (75). Timing of stimulation during the sample and delay epochs were the same used for experiments inhibiting D1R+ cells and DA afferents. Stimulation occurred during either the sample or delay periods for 30 to 40% of trials with each mouse receiving two to three sessions of stimulation for each behavioral period. Stimulation of left and right ALM alternated between sessions. Experiments with stimulation occurring during the sample or delay period were also alternated across sessions. No difference in performance for unperturbed trials was detected across successive sessions [one-way ANOVA, F(2,16) = 0.24, P = 0.79]. Immunohistochemistry Mice were deeply anesthetized with an intraperitoneal injection of ketamine (140 mg/kg)/dexmedetomidine (2 mg/kg) and were intracardially perfused with phosphate-buffered saline (PBS), followed by 4% paraformaldehyde (PFA). The brain was postfixed with 4% PFA overnight, cryoprotected with 30% suc for 3 days, and then sectioned with a cryostat into 50-μm coronal slices. For visualizing optic fiber tracks and/or expression of AAV constructs, slices were counterstained with Hoechst 33342 (1:5000 dilution; Thermo Fisher Scientific, Waltham, MA, H3570) using standard techniques. Reconstruction of fiber tracks and expression of viruses for experiments relying on fiber photometry, two-photon calcium imaging, and optogenetic perturbations are illustrated in figs. S10 and S11. Tissue sections used for TH immunostaining were postfixed in 4% PFA [0.01 M PBS (pH 7.4)] for 24 hours and then rinsed in PBS (3× of 10-min rinses). Tissue was incubated in a preblock solution in PBS—5% bovine serum albumin (BSA; Sigma-Aldrich), 5% normal goat serum (Vector Laboratories), 0.2%/1% Triton-X (VWR)—for 1 hour. Following preblock incubation, tissue sections were incubated overnight at 4°C in an antibody incubation solution in PBS—1% BSA (Sigma-Aldrich), 1% normal goat serum (Vector Laboratories), 0.1% Triton X-100 (Sigma-Aldrich)—containing anti-TH antibody (1:1000; Abcam, ab112). After 24 hours, tissue was rinsed in 3× for 10 min in PBS and transferred into the antibody incubation solution containing secondary antibody (1:200; goat-anti rabbit Alexa Fluor 488, Vector Laboratories, BA-1000) and incubated for 3 hours in 4°C. Following a PBS rinse 3× for 10 min, sections were counterstained with Hoechst 33342 using standard techniques. RNAscope in situ hybridization assay Manual RNAscope assay was performed using the RNAscope Multiplex Fluorescent Reagent Kit v2 (ACDbio, 323100) with the RNAscope Probe-Mm-Drd1 (#461901). Mice were perfused with ribonuclease-free PBS and 4% PFA. Brains were postfixed in 4% PFA overnight at 4°C and then dehydrated in 30% suc for a day. Fifteen-micrometer sections were obtained using a cryostat and kept at −80°C before assays. The staining procedures followed the manufacturer’s protocols for fixed-frozen tissue samples with minor modification. Briefly, tissue samples were dehydrated for 1 hour and pretreated with hydrogen peroxide (10 mins) at room temperature before immersion in antigen retrieval solution (5 min). Hydrophobic barriers were applied, and then protease solution was applied for a 2-hour incubation period. Target probe was then applied and amplified (30 min) in a 40° oven for three iterations. Last, green opal dye and 4′,6-diamidino-2-phenylindole were applied, and coverslips were mounted. Quantification and statistical analysis Statistical tests for each main figure are detailed in tables S1 to S5. Behavioral analyses Performance in the task was calculated as the fraction of correct trials over the total number of correct and error trials. Performance for suc and NaCl trials were calculated separately. Trials with no licking to the central or lateral spouts were excluded from analyses (~3% of all trials). Reaction times were measured as the interval between initiation of lateral spout movement and first lick to a lateral spout. Paired t tests (two-tailed) were used to compare performance, lick rates, and reaction times. Data analyses for fiber photometry Demodulated fluorescent 405- and 470-nm signals were low-pass filtered at 6 Hz. To correct photobleaching of signals, each signal was fitted with a least-squares polynomial and mean-subtracted. To account for motion correction, linear regression was performed between 405- and 470-nm signal. ΔF/F was calculated as (470-nm signal − fitted 405-nm signal)/(fitted 405-nm signal). Responses were then aligned to two behavioral events: time of first dry lick to the central spout and time of first lateral lick (choice). Responses were z-score transformed relative to a baseline window of −4 to −2 s before the first dry lick for each recording session. We chose a 2-s wide window well before the first lick to ensure a stable baseline normalization. Similar results were observed when we used a 1-s window. Fiber photometry signals (DA or NE) from each session were sorted on the basis of taste trial types (suc and NaCl) and lick direction trial types (left lick and right lick) and quantified as AUC during each of the behavioral periods (baseline, −2- to −1-s window before first dry lick; sample, 1.5 s after first dry lick; delay, 1.5-s window before first lateral lick; choice, 1.5-s window after first lateral lick). A window width of −1.5 s was chosen for event-related activity based on visual inspection of the latency of direction-selective activity. Trial-type differences across sessions were analyzed using paired t tests (two-tailed). Analyses examining differences across mice yielded similar results. Trial-averaged differences in responses between each taste (suc versus NaCl) and lick direction (left versus right) were quantified as taste and direction selectivity, respectively, using correct trials or error trials separately. Significant selectivity during different behavioral periods was tested against selectivity during the baseline using paired t tests (two-tailed). Direction selectivity between DA and NE responses were compared using two-sample t tests. Data analyses for calcium imaging Image processing of calcium imaging data Images were downsampled from 31 to 6.2 Hz with ImageJ and concatenated across trials within a session. Nonrigid motion correction with NoRMCorre was performed to minimize brain motion. Regions of interest (ROIs) corresponding to cell bodies and calcium traces were automatically extracted with the constrained nonnegative matrix factorization algorithm (76). Fluorescent signals (F) were calculated by multiplying the spatial footprint of an ROI with its temporal trace (background subtracted of raw image) at each pixel and then averaged within pixels of the ROI. F was then detrended by subtraction of baseline fluorescence and neuropil activity using eighth percentile values over ~30 s to calculate ΔF/F. Extracted ROIs were identified on the basis of average fluorescent signals and by maximum-correlation maps (each pixel in map represents the average correlation across time between the pixel and its neighbors). ROIs were further manually inspected on the basis of ROIs’ shape and ΔF/F traces using custom scripts in MATLAB. Spatial footprints of ROIs were then applied as masks over stacks of images from the red channel (tdTomato fluorescence) recorded before and after the session to distinguish ROIs that expressed D1Rs from those that do not. For each ROI, a ratio of mean pixel intensity within the ROI over mean pixel intensity within an adjacent 4-pixel wide border was computed and sorted on the basis of magnitude. The ratio for each ROI, along with the overlap of the ROI’s spatial footprint with that of tdTomato+ cells recorded in the red channel, was inspected. A threshold of 1 was used to classify cell type; cells with ratios above this threshold were identified as D1R+, while cells with ratios less than the threshold were identified as D1R− neurons (fig. S5D) Single-neuron responses To facilitate comparison between D1R+ and D1R− cells and account for potential differences in basal Ca2+ levels, extracted ΔF/F signals of individual cells were normalized to baseline time windows (trial by trial, detailed below). ΔF/F signals of each trial were aligned to three behavioral events: time of first dry lick to the central spout, time of taste delivery, and time of first lateral lick (choice). All ROIs (D1R+ and D1R−) were categorized as responsive during dry licking, taste delivery, delay, and choice periods. Only excitatory responses were analyzed. To determine responses to dry licking, we compared the mean baseline activity (−4 to −3 s before first dry lick) to the mean activity during dry licking across all trials (first five licks of ~0.55 s, responsive if P < 0.05, Wilcoxon rank sum test). For responses to tastants, we compared the mean baseline activity (−4 to −3 s before taste delivery) to the mean activity during taste delivery for suc and NaCl trial types (1.5-s window after five dry licks, responsive if P < 0.01, Wilcoxon rank sum test). For preparatory responses during the delay period, we compared the mean baseline activity (−9 to −7 s before first lateral lick) to the mean activity during the delay period for each left lick and right lick trial types (1.5-s window before first lateral lick, responsive if P < 0.01, Wilcoxon rank sum test). For perimovement responses during the choice period, we compared the mean baseline activity (−9 to −7 s before first lateral lick) to the mean activity during the choice period for each left lick and right lick trial types (1.5-s window after first lateral lick, responsive if P < 0.01, Wilcoxon rank sum test). For neurons responsive to both tastants and licking during the sampling period, we additionally compared the mean activity immediately before taste delivery (−1 s before taste delivery) to the mean activity during taste delivery for suc and NaCl trial types (1.5-s window after five dry licks, responsive if P < 0.05, Wilcoxon rank sum test). This additional test was used to further recategorize responses where observed taste responses might be a carryover of the dry-licking response. Task-responsive neurons were further categorized as taste-selective and direction-selective during the sample, delay, and choice periods using correct trials based on lick direction. Taste-selective cells were defined as neurons that showed significantly different response between suc and NaCl trials during the sampling period (1.5-s window after 5 dry licks, correct suc versus correct NaCl, P < 0.05, Wilcoxon rank sum test). This criterion identified <5% of taste-responsive cells as exhibiting significant taste preference during this period for both D1R+ (1.5%) and D1R− (1.0%) cells. Direction-selective cells were defined as neurons that showed significantly different response between ipsi and contra trials during any of the three behavioral periods (correct left versus correct right, P < 0.05, Wilcoxon rank sum test). To determine which taste or lick direction each selective neuron preferred, the mean activity between trial types was compared. Direction selectivity of contra-preferring and ipsi-preferring neurons was calculated as the trial-average difference between contra and ipsi trials (separately for correct and error trials). Direction selectivity index was computed by normalizing a neuron’s selectivity to its the peak selectivity. Response onset for each group of neurons were identified using change-point analysis (77) based on mean and slope of trial-averaged activity aligned to the first central or lateral lick event. Onsets measured with the change-point analysis appear to be later than those suggested by visual inspection of population averages in Fig. 2 (E and G). The cause for this apparent mismatch is to be found in the stringent criterion for detecting onsets with change-point analysis applied to single-neuron signals combined with the variability of onsets and firing rates across cells. Time to peak was calculated as time between 20 and 100% of normalized activity (each time point was normalized to the difference between the minimum and maximum of trial-averaged activity for each neuron). All trials were used to calculate response onsets and time to peak. Population selectivity coding Population selectivity vectors were calculated using all recorded neurons. Across n neurons, trial-averaged responses for each trial-type xcontra and xipsi were calculated to represent the population response at different time points of the trial (12). Population selectivity vector was calculated asWT=xcontra−xipsi A significant difference in selectivity strength between populations was computed using a permutation test, where cell identity was shuffled without replacement. This procedure was repeated 1000 times to create a null distribution of difference in mean selectivity between populations. P value was calculated as the proportion of times where shuffled data have an equal or bigger difference than the real population difference. PCA of population coding To visualize differences in population activity over time, we applied PCA. A pseudo-population was constructed from all recorded neurons from each population (randomly sampled with replacement and matched number of cells). For each neuron, correct contra and ipsi trial-averaged activities aligned to the first lateral lick (time = 0) were concatenated. We used PCA to find the PC coefficients of the matrix and applied the coefficients to contra and ipsi population activity and projected onto PC space. The first three PCs captured greater than 75% of variance in activity. Euclidean distance in PC space was calculated as the distance between contra and ipsi population activity summed across the first 3 PC axes at each time point. P value was calculated as the fraction of sampled D1R− Euclidean distance means greater than or equal to the observed mean Euclidean distance for D1R+ neurons at each time point (bootstrap difference in mean with null hypothesis: D1R− population shows larger distance). Analyses for population coding direction Coding direction (CD) was computed using population selectivity vector calculated during the delay (−1.5 s to 0 before first lateral lick; WTdelay). The resulting vector was normalized by its L1 norm to calculate CD for delay period (CDdelay). CDdelay was orthogonalized to activity during the sample period using the Gram-Schmidt procedure to isolate selectivity during the delay period (10). CDs were calculated for each D1R+ and D1R− population separately with random sampling (with replacement) to generate a bootstrap distribution with size-matched samples between populations. For each random pooling, population activity was projected along the CDs, as the inner product and selectivity for each CD were calculated as the difference in CD-projected activity (contra projected − ipsi projected). One thousand iterations of this procedure were performed to compare bootstrap distributions between D1R+ and D1R− populations. P value was calculated as the fraction of bootstrap differences in mean selectivity less than or equal to 0 (bootstrap difference in mean selectivity with null hypothesis: D1R− population show larger selectivity). To quantify the similarity of selectivity across time, Pearson’s correlation was calculated using direction-selective neurons from each cell population (matching the number of cells from each population, 1000 iterations). Pearson’s correlation was calculated between individual time points in the delay and choice periods. Correlation between population selectivity during the delay and the population selectivity at the beginning of the choice period (1 s after first lateral lick) was also computed using Pearson’s correlation. Quantification of the mean correlation during the delay period was performed for each group of neurons (D1R+ versus D1R− and contra-preferring versus ipsi-preferring) and compared using bootstrap sampling, where P value was calculated as the fraction of bootstrap differences in mean correlation greater than or equal to the observed difference in mean correlation. Behavioral analyses for optogenetic experiments Performance in the task was computed separately for control trials and inhibition/stimulation trials. Both control and inhibition/stimulation trials were further sorted by contra and ipsi trials, and performance was calculated as the fraction of correct trials over the total number of correct and error trials. Trials with no licking to the central or lateral spouts were excluded from analyses (~2% of all trials in optogenetic sessions). Manipulations produced similar effects across both ALM hemispheres; therefore, data from left and right ALM manipulations were pooled together. Reaction times were measured as the interval between the initiation of the lateral spout movement and the first lick to a lateral spout. Reaction times were pooled across left and right hemisphere inhibition/stimulation sessions since optogenetic perturbations produced similar behavioral effects. Change in reaction times was calculated as the difference in lick latencies between trials with and without perturbations for ipsi and contra lick trials separately. Paired t tests (two-tailed) were used to compare average performance between control and inhibition/stimulation trials for each lick direction. Paired t tests (two-tailed) were used to compare reaction times between ipsi and contra trials. Image analysis for RNAscope in situ hybridization Confocal images were preprocessed and analyzed in ImageJ. Two to three slices per mouse (n = 3 mice) with 1-μm z-stack planes were used for quantification of colocalization of endogenous tdTomato expression and Drd1 probe (green). tdTomato cell bodies within each image were identified as primary objects and related to secondary objects, Drd1 probe. Percent colocalization was calculated as the number of tdTomato-positive cell bodies with Drd1 probe (with at least one punctum) divided by number of tdTomato-positive cell bodies for each slice and then averaged for each mouse. Statistics and reproducibility Data are reported as means ± SEM. unless stated otherwise. Statistical analyses using t test, Wilcoxon rank sum test, chi-squared test, permutation test, Pearson’s correlation, and bootstrap sampling were performed using the Statistics toolbox and built-in functions in MATLAB. Two-sample bootstrap testing was performed by calculating an observed test statistic (mean differences) and comparing to a null distribution of the test statistic (resampled from pooled data with recentered mean) (12, 48, 78). P value was calculated as the fraction of bootstrap differences in mean less than or equal to 0 (if null hypothesis is that D1R− population shows a larger t statistic) or greater than or equal to 0 (if null hypothesis is that D1R− population shows a smaller t statistic). Permutation tests were performed by comparing a test statistic (mean) with a null distribution (sampled without replacement, 1000 iterations). Statistical significances were considered at *P < 0.05, **P < 0.01, and ***P < 0.001 and corrected for multiple comparisons. No statistical methods were used to predetermine sample sizes, and the study was not powered for sex differences; however, sample sizes are similar to those reported in previous publications (7, 8, 79). For all experiments, mice within a specific strain were selected randomly without bias but balanced for sex and age. Task contingency (taste-lick association) was counterbalanced, and trial sequences were randomly generated. Across mice, criterion for task acquisition was defined as >75% correct choices for more than 3 days. Data collection and analysis were not performed blind to the conditions of the experiments due to the surgical (administration of virus) and other procedures related to the experiments. All viral injections were verified by post hoc histology (figs. S10 and S11).