Sensitive dLight3 for imaging broad-spectrum dopamine events across brain regions

Abstract

Dopaminergic neurons modulate movement, motivation, and learning by dynamically regulating dopamine release across distributed neural circuits. However, existing genetically encoded dopamine sensors lack the sensitivity and resolution to capture the full amplitude and temporal complexity of in vivo dopamine signaling, limiting insight into its functions across behavioral contexts. Here, we present dLight3.8, a fluorescence-intensity and lifetime-based sensor with a substantially expanded dynamic range compared to existing dopamine sensors, enabling transformative detection and differentiation of dopamine release across brain regions and behaviors. Specifically, the enhanced sensitivity of dLight3.8 permits robust, single-trial recording of dopamine release spanning a wide concentration range in response to electrical, optogenetic, and behavioral stimuli, in multiple species and circuits. Using dLight3.8, we uncover a region-specific, gradual shift in dopamine encoding across motor learning, from tracking lick timing to signaling reward prediction. Our findings demonstrate that dLigth3.8 provides quantitatively reliable, highly sensitive measurements of graded dopamine release, which is essential for elucidating diverse roles of dopamine signaling in shaping animal behavior.

Figure 1.. dLight3 Engineering and benchmarking to FSCV in acute brain slice.

a, Simulated structure using Alphafold of dLight 3.6 consisting of DRD1 (blue), linkers (grey), and cpGFP module (green). Sequence alignment and highlighted amino acid residue changes of dLight 3.6 and 3.8 from parent dLight 1.3b. b, Representative images of dLight 3.6 and dLight 3.8 in vitro hippocampal neuronal culture. Fluorescence intensity at 15mM DA saturation state and signal-to-noise ratio of apo and saturation state shown. Scale bars = 50 μm. c, in vitro hippocampal neuron culture dopamine dose-response curves. Data were fitted with Hill Equation, n=4. All data shown as mean ΔF/F₀ ± SEM. d, Mean Tau and Delta Tau lifetime measurements of dLight3.6 and dLight3.8 using HEK T-Rex stable cell line under Dopamine, DRD1 agonists Dyhydrexidine and SKF-81297, and DRD1 antagonist SCH-23390 n=5. e, (Left) Pharmacological specificity of dLight 3.6 and dLight 3.8 in HEK T-REx 293 stable cell lines relative to Dopamine response (1.00 ±0.01 ΔF/F₀, n=8). DRD1 full agonist (Dihydrexidine, 1.43 ± 0.01 ΔF/F₀, n=8) DRD1 partial agonists (SKF-81297, 1.12 ± 0.01, n=8, A77636, 0.63 ± 0.01, n=8, Apomorphine, 0.17 ± 0.002, n=8), Norepinephrine (0.05 ± 0.02, n=8), DRD1 antagonists (SCH-23390, 0.018 ± 0.001, n=8, SKF-83566, 0.002 ± 0.001, n=8), DA addition following DRD1 antagonists (DA+SCH-23390, 0.020 ± 0.002, n=8, DA+SKF83566, 0.06 ± 0.01, n=8), DRD2 antagonists (Sulpiride, 0.87 ± 0.02, n=8, Haloperidol, 0.60 ± 0.01, n=8). All data shown as mean ΔF/F₀ ± SEM relative to ΔF/F of dopamine response. All additions at 1μM final concentration. (Right) Combined Schild regression with SKF-83566 and SCH-23390 on dLight 3.6 and dLight 3.8, n=6. f, (Top Left) Schematic for DLS slices with most fiber tracks intact. (Right) FSCV setup for electrical stimulation and epifluorescence imaging. g, Time-lapse of fluorescent intensity changes (ΔF/F) in response to a pseudo-one-pulse (p1P) or a 20 pulses electrical stimuli in DLS acute slice when expressing dLight at DAT terminals (DAT-Cre expression, dotted line, n = 16, 9, 12 slices from ≥ 3 animals) and postsynaptic cells expressing DRD1 (DRD1-Cre expression, solid line n = 7, 4, 14 slices from ≥ 3 animals). All data shown as mean ΔF/F₀ ± SEM. h, ΔF/F fold change normalized to dLight1.3b. Box plot center line shows the median; box limits show upper and lower quartiles. i, Off-rate kinetic Tau_1/2 at 36.5°C±0.5°C. Epifluorescence imaging at 10 Hz. Violin plot shows the full range of the data, with each line denoting each quartile. j, Left: FSCV measurements in DLS acute brain slices without methylphenidate (MPD) (solid black line) and following bath application of MPD (30μM) (dotted blue line). Right: Example traces of time-lapse imaging across the experiment.

Figure 2.. Two-Photon imaging of dLight3 in acute brain slices and in vivo in superior colliculus.

a-f,Two-photon imaging of dLight3 in acute slice. a, Spontaneous dLight3.8 activity revealed by two-photon imaging displayed spatial heterogeneity, timelapse of fluorescence intensity from the identified ROIs in DLS acute slice. Scale bars: 10μm. b, Time-lapse of fluorescent intensity changes (ΔF/F) in response to electrical stimuli (1, 2, 5, 10 or 20 pulses, 40Hz) in DLS (n=13 for 3.6, n=18 for 3.8) and PFC (n=12 for 3.6, n=6 for 3.8), following stereotaxic injection. Mean ΔF/F₀ ± SEM. c, (Top) Maximum peak amplitude in response to 1, 2, 5, 10 or 20 pulses for each sensor. (Bottom) Off-rate kinetic Tau_1/2 at 29°C. Box plot center line shows the median; box limits show upper and lower quartile. d, (Top Left) Cleared brain showing dLight3.8 expression across brain regions following systemic injection of CAP-B10 viral capsid. Scale bar: 3000μm. (Bottom Left) Representative images of DLS, PFC, and HPC of dLight3.8 expression. Scale bar: 100μm. (Right) Time-lapse of fluorescent intensity changes (ΔF/F) in response to electrical stimuli in acute slices following 6 weeks expression of systemic injection using either PhP.eB or CAP-B10 viral capsid. DLS (PhP.eB n=12, B10 n=31), PFC (PhP.eB n=3, B10 n=14) and HPC (PhP.eB n=3, B10 n=18). Mean ΔF/F₀ ± SEM. e, (Top) SNR of peak amplitude increased with the number of stimuli, comparing CAB-B10 and PhP.eB. (Bottom) Peak ΔF/F and Tau_1/2 at 20Hz comparing systemic delivery and stereotaxic injection. Box plot center line shows the median; box limits show upper and lower quartile. f-i, In vivo two-photon imaging of dLight3 benchmarked to dLight1.3b. f, Schematic of tail shock and looming behavior paradigm g, (Left) Trial-averaged peak responses for the first 10 shock trials versus the last 10 shock trials. Response amplitude is defined as the average ΔF/F between 3.2–5.5 s after shock onset across trials. Each symbol represents one ROI within a field of view. Black bars: mean. ***P<0.001 (paired t test). (Right) Trial-averaged peak responses for the first 10 looming trials versus the last 10 looming trials. Each symbol represents one ROI within a field of view. Black bars: mean. ***P<0.001 (paired t test). h, (Top) averaged two-photon images of dLight3.6, dLight3.8, and dLight1.3b expressed in SC. Grid lines superimposed on the dLight1.3b image divide the field of view into 9 ROIs. Scale bar: 50μm. (Middle) Heatmap of trials averaged ΔF/F₀ (20 trials) to tail shock for dLight3.6, dLight3.8, and dLight1.3b for each ROI. Dashed lines: onset of 0.5-s-long shock. Scale bar: 5sec. (Bottom) Trial-averaged ΔF/F₀ (mean ± SEM) for the entire field of view. Black lines with gray shading: trial-average without shock. Scale bar: 5sec. i, (Top) averaged two-photon images of dLight1.3b, dLight3.6, and dLight3.8 expressed in SC. Scale bar: 50μm. (Middle) Heatmap of trials averaged ΔF/F₀ (40 trials) to looming stimuli of dLight1.3b, dLight3.6, and dLight3.8 for ROIs. Dashed lines: onset of 1-s-long looming. Scale bar: 5sec. (Bottom) Trial-averaged ΔF/F0 (mean ± SEM) for the entire field of view. Black lines with gray shading: Trial-average without looming. Scale bar: 5sec.

Figure 3.. Benchmarking dLight3.0 to GRAB-DA using optogenetic stimuli in mice and flies.

a, Schematic showing dLight expression in NAc and ChrimsonR expression in VTA of DAT-Cre mice as well as fiber implantation for simultaneous optogenetic stimulation and recording. (GRAB-DA3m, n=7; dLight 3.6, n=7; dLight3.8, n=5). b, Schematic of the experimental design consisting of simultaneous two-photon imaging and widefield optogenetic stimulation of TH-Gal4 neurons co-expressing a dopamine sensor and ChrimsonR::mCherry in explant fly brains. Yellow square shows mushroom body neuropil. Immunochemistry images show co-staining of dLight3.8 (green), mCherry (red) and neuropils (blue). White arrowheads show TH-labeled neurons innervating the mushroom body vertical lobes. Scale bar: 50 μm. (GRAB-DA2m, n=12; dLight3.8, n=9). c, ΔF/F₀ mean ± SEM traces time-series of GRAB-DA3m, dLight3.6, and dLight3.8 to 40 trials of repeated optogenetic stimulation of ChrimsonR at 5, 10, and 20Hz stimulation frequencies. d, Average peak amplitude of the dopamine sensors to the three different optogenetic stimulation frequencies. e,f, Tau on and off of the dopamine sensors to optogenetic stimulation. Tau-on (s) DA3m vs 3.6: p = 0.037, Tau-off (s) DA3m vs 3.6: p = 0.018 determined by t-test. g, Dopamine signal from TH-labelled neurons innervating the mushroom body vertical lobe expressing either dLight3.8 only (left), GRAB-DA2m with ChrimsonR::mCherry (middle) or dLight3.8 with ChrimsonR::mCherry (right). Grey bars show 2s optogenetic stimulation delivered at different frequencies. Traces show ΔF/F averaged across brains, mean ± SEM. h, Peak ΔF/F at different stimulation frequencies, calculated from g. Grey lines show individual brains; colored dots show mean ± SEM n=9–12 brains per genotypes. i, Fold change to GRAB-DA2m at different stimulation frequencies, calculated from h. Asterisks indicate significant differences between groups based on paired t-tests with Bonferroni correction (p < 0.0083). j-n, Two-photon imaging of dopamine in behaving flies j, Schematic of a head-fixed fly walking on an air-suspended ball while receiving 1 M sucrose solution, with dLight3.8 signal imaged in the mushroom bodies using two-photon microscopy. k, Two-photon maximum intensity projection of dLight3.8 fluorescence in the mushroom bodies (white dashed outlines); γ2 and γ5 compartments are shaded in blue and magenta, respectively. Scale bar: 10 μm. l, dLight3.8 signal (zΔF) during sucrose consumption, averaged across sucrose presentations in an individual fly; white arrowheads indicate γ5 compartments. m, Example traces of dLight3.8 signal from γ2 and γ5 compartments during sucrose consumption in three individual flies; dashed lines indicate feeding onset. n, dLight3.8 signal (ΔF/F: mean ± SEM) in γ2 and γ5 compartments during sucrose consumption, averaged across sucrose presentation and flies (n = 6); red bar indicates feeding onset.

Figure 4.. In vivo dual-fiber photometry recordings across brain regions in mice and rats.

a, Schematic of viral injection of a single dLight variant and dual-fiber-optic implant in NAc and PFC of a mouse brain for sensor benchmarking (dLight3.6, n=9; dLight3.8, n=9; dLight1.3b, n=3). b-d, dLight benchmarking during operant reward foraging in mice. b, Schematic of the operant reward foraging protocol. c, Average dLight response to pokes at the reward port when rewards are collected (left) and when no reward is delivered (right) in NAc (top) and mPFC (bottom). Signals are shifted so the baseline is equal for every trial at the time of the reward delivery. Signals are averaged over all trials and subjects. Shaded areas indicate SEM. d, Peak of dLight transient in response to reward collection for all three dLight variants normalized to the average peak reward response of dLight1.3b. Differences between variants were characterized using linear mixed-effects models with a fixed effect for each variant group and a random effect for each subject (NAc, 3.8 vs 1.3b ***: p = 1.3e-5, 3.8 vs 3.6 ***: p = 2.2e-4. PFC, 3.8 vs 1.3b *: p = 0.03, 3.6 vs 1.3b **: p = 5e-3). Box plot center line shows the median; box limits show upper and lower quartiles. Dots show individual trial responses. e-g, dLight benchmarking during classical fear conditioning in mice. e, Average dLight response on day 2 of fear conditioning in NAc (top) and mPFC (bottom). Inset: response to tone initiation. Signals are shifted so the baseline is equal for every trial at the time of the tone initiation. Signals are averaged over all trials and subjects. Shaded areas indicate SEM. f, Peak of dLight transient in response to the shock for all three dLight variants normalized to the average peak shock response of dLight1.3b. In NAc (top) dLight3.8 has 14.3x and dLight3.6 has 9.2x higher response than 1.3b. In PFC (bottom) dLight3.8 and dLight3.6 both have 5.6x higher response than 1.3b. Differences between variants were characterized using linear mixed-effects models with a fixed effect for each variant group and a random effect for each subject (NAc, 3.8 vs 1.3b **: p = 8e-3. PFC, 3.8 vs 1.3b *: p = 0.037, 3.6 vs 1.3b *: p = 0.037). Box plot center line shows the mean; box limits show upper and lower quartiles. Dots show individual trial responses. g, dLights 3.8 and 3.6 measure a decreased shock peak transient between trials 1–3 and 18–20 of conditioning (left). dLight3.6 measures an increase in tone response in PFC between the first three trials after the animals experience shock during conditioning and the last three trials of extinction (right). Differences between shock and tone responses were characterized using a linear mixed-effects model with a fixed interaction effect for each variant and trial group and a random effect for each subject. (Shock - NAc, 3.8 ***: p = 3.8e-4, 3.6 *: p = 0.01. PFC, 3.8 *: p = 0.015, 3.6 **: p = 4e-3. Tone - PFC, 3.6 *: p = 0.039). h-l, dual-site dLight3.8 recordings in rats during a probabilistic two-armed bandit reward choice task (n=4). h, Schematic showing expression of dLight3.8 and dualfiber-optic implant in PFC and DMS of rat brain. i, Schematic of the probabilistic two-armed bandit reward choice task. j, Average z-scored ΔF/F dLight traces at the time of reward delivery grouped by trial outcome and number of rewards received over the last 3 trials. Signal baselines are shifted for each session based on the average of the signal from the lowest reward condition in the 100ms before reward delivery. Signals are averaged over all sessions and subjects. Shaded areas indicate SEM. k, Distributions of reward peak amplitudes of dLight transients on individual trials for each reward history condition. Peak amplitudes were calculated as the difference between peak values within a window of time after reward and the deepest trough on either side of the peak. In both regions, the peak amplitudes decrease as the number of prior rewards increases. Differences between reward history conditions were characterized using linear mixed-effects models with a fixed effect for each reward history group and a random effect for each subject (*: p < 8e-3, **: p < 1e-4, Bonferroni correction at p < 0.05). Box plot center line shows the median, and the box limits are the 1st and 3rd quartile. Connected dots show averages for each subject. l, Independent effect of rewards received at each trial-back latency on the reward-related transient of the current trial. Each trace indicates the linear regression coefficient for a reward received at each trial-back latency predicting the z-scored ΔF/F values at each aligned timepoint across all sessions and subjects. The ‘0’ trial-back trace is the regression intercept and captures the average reward transient when there were no rewards in the last 3 trials. Dots above the traces indicate timepoints where the associated coefficient was significantly different than 0 (two-tailed t-test, Bonferroni correction at p < 0.05). Shaded areas indicate SEM.

Figure 5.. Fiber-photometry recording of DA dynamics across a motor learning task in mice.

a, Schematic of a motor timing task. b, Learning curves for all animals with NAc photometry recording (n=4) across delay training and expert sessions. (Top) Delay duration. (Bottom) Mean time to first lick. Thin lines, individual animals, thick lines, average across animals. c, (Top) Schematic for recording NAc DA dynamics with fiber photometry. (Bottom) Average dLight responses aligned to trial onset in rewarded (blue), unrewarded (black), and nocue trials (green) in an example session with 1.5s fixed delay duration. Vertical dashed lines indicate timing of cue onset and lick. Magenta boxes indicate time windows for analysis of post-cue and post-lick DA responses. Responses are normalized to pre-cue activity. Shaded regions, bootstrap SEM. d, DA responses aligned to trial onset in three example sessions at different phases of learning. d₁, day 5 of delay training. d₂ and d₃, days 1 and 6 of expert session with 1.5s fixed delay, respectively. Top row, rewarded trials. Middle row, unrewarded trials. Individual colors indicate the average ΔF/F of trials with different lick times. Lick times are indicated by vertical dotted lines. Shaded region, SEM. Bottom row, heat map of DA responses in all trials sorted by lick time (white dots). e, Relationship between first lick time and single-trial post-cue NAc DA response (0.3 ~0.4 s after cue onset) in 3 example sessions for rewarded (blue) and unrewarded (black) trials. Each dot represents individual trials. Magenta line, least squares fit. Slope, slope of the least square fit line. f, Left, mean post-cue NAc DA response per session. Thin lines, individual animals, thick lines average across animals. Right, average post-cue NAc DA response in sessions grouped by early delay training (delay duration 0.1~0.7s), late delay training (delay duration 0.7~1.3), and expert (1.5s fixed delay). Cue response shows a U-shaped curve, decreasing during delay training and increasing again in expert mice. Box plot center line shows the mean; box limits show upper and lower quartiles. g, (Left) Least squares regression slope of single-trial post-cue NAc DA response on lick time per session. Thin lines, individual animals, thick lines average across animals. (Right) Regression slope in sessions grouped as in (f). Cue response becomes less predictive of upcoming lick timing as animals learn to lick later. Box plot center line shows the mean; box limits show upper and lower quartiles. h, Relationship between first lick time and single-trial post-lick DA response (0.35~0.5 s after first lick) in 3 example sessions for rewarded (blue) and unrewarded (black) trials. Each dot represents individual trials. Magenta line, least squares fit for rewarded trials. Slope, slope of the least square fit line. i, Least squares regression of single-trial post-lick DA response on first lick time and reward outcome in learning (L) and expert (E) sessions. (Top) Regression model. (Bottom) Regression coefficients for reward (left) and lick time (middle), and the relation between regression coefficient for lick time and mean lick time in individual training sessions (right). Shuffle indicates regression with DA responses shuffled across trials. For the regression analysis, only lick time in the rewarded trials was considered, as the no-rewarded response was less modulated by lick time. * p<0.05, ** p<0.005, bootstrap. Box plot center line shows the mean; box limits show upper and lower quartiles.