Mesolimbic dopamine ramps reflect environmental timescales
- Neuroscience Graduate Program, University of California, San Francisco, United States
- Department of Neurology, University of California, San Francisco, United States
- Department of Psychology, University of Wisconsin–Madison, United States
- Neuroscience Training Program, University of Wisconsin–Madison, United States
- Weill Institute for Neurosciences, Kavli Institute for Fundamental Neuroscience, Center for Integrative Neuroscience, University of California, San Francisco, United States
Figures
Figure 1 with 2 supplements
Pavlovian conditioning dopamine ramps depend on ITI.
(A) Top, fiber photometry approach schematic for nucleus accumbens core (NAcC) dLight recordings. Bottom, head-fixed mouse. (B) Pavlovian conditioning experimental setup. Trials consisted of an 8 s auditory cue followed by sucrose reward delivery 1 s later. (C) Cumulative Distribution Function (CDF) of ITI duration for long (solid line, mean 55 s) and short ITI (dashed line, mean 8 s) conditions. (D) Experimental timeline. Mice were divided into groups receiving either a 3 kHz fixed and dynamic up↑ tone or a 12 kHz fixed and dynamic down↓ tone. (E) Tone frequency over time. (F) Peri-stimulus time histogram (PSTH) showing average licking behaviors for the last 3 days of each condition (n=9 mice). (G) Average anticipatory lick rate (baseline subtracted) for 1 s preceding reward delivery (long ITI vs short ITI: **p=0.0046). (H) ANCCR simulation results from an 8 s dynamic cue followed by reward 1 s later for long ITI (teal) and short ITI (pink) conditions. Bold lines show the average of 20 iterations. (I) Left, average dLight dopamine signals. Vertical dashed lines represent the ramp window from 3 to 8 s after cue onset, thereby excluding the influence of the cue onset and offset responses. Solid black lines show linear regression fit during the window. Right, closeup of dopamine signal during window. (J) Average peak dLight response to cue onset for LD and SD conditions (***p=1.9 x 10–4). (K) dLight dopamine signal with linear regression fit during ramp window for example SD trials. Reported m is slope. (L) Session average per-trial slope during ramp window for the first day and last 3 days of each condition (last day LD to first day SD: *p=0.036, last day SD to first day SF: *p=0.023). (M) Average per-trial slope for last 3 days of each condition (LD vs SD: **p=0.0026, SD vs SF: *p=0.011). All data presented as mean ± SEM. See Supplementary file 1 for full statistical details.
Figure 1—figure supplement 1
Dependence of ANCCR on eligibility trace time constant.
(A) Schematic showing exponential decay of cue eligibility traces for two-cue sequential conditioning (left) and multi-cue conditioning (right) with a long inter-trial interval (ITI). In this case, a long ITI results in a proportionally large eligibility trace time constant, T, producing slow eligibility trace decay (Appendix 1). Reward delivery time indicated by vertical dashed line. (B) Schematized ANCCR magnitudes (arbitrary units) for cues in the two-cue (left) and multi-cue (right) conditioning paradigms with a long ITI. Since the eligibility trace for the first cue is still high at reward time, there is a large ANCCR at this cue. The remaining cues are preceded consistently by earlier cues associated with the reward, thereby reducing their ANCCR. (C) Same conditioning trial structure as in A, but with a short ITI and smaller T, producing rapid eligibility trace decay. (D) Schematized ANCCR magnitudes for cues in both conditioning paradigms with a short ITI. Since the eligibility trace for the first cue is low at reward time, there is a small ANCCR at this cue. Though the remaining cues are preceded consistently by earlier cues associated with the reward, the eligibility traces of these earlier cues decay quickly, thereby resulting in a higher ANCCR for the later cues.
Figure 1—figure supplement 2
Pavlovian conditioning cohort 1 histology and dopamine responses.
(A) Mouse coronal brain sections showing reconstructed locations of optic fiber tips (red circles) in NAcC for Pavlovian conditioning cohort 1. (B) Example average dLight traces for the last three days of all conditions. Vertical dashed lines at 3 and 8 s represent the ramp window period. Black lines display the linear regression fit during this period. (C) Same as in B but for the average dLight traces across all animals (n=9 mice).
Figure 2 with 4 supplements
Pavlovian conditioning dopamine ramps do not depend on training order.
(A) Experimental timeline in which the SD condition occurs before the LD condition. (B) Left, average dLight dopamine signals for SD and LD conditions. Vertical dashed lines represent the ramp window from 3 to 8 s after cue onset. Solid black lines show linear regression fit during the window. Right, close-up of dopamine signal during window (n=9 mice). (C) Session average per-trial slope during ramp window for the first day and last 3 days of each condition (last day LF to first day SD: **p=0.0045, last day SD to first day LD: **p=0.0067). (D) Average per-trial slope for last 3 days of each condition (LF vs SD: ***p=9.1 x 10–4, SD vs LD: *p=0.010).
Figure 2—figure supplement 1
Pavlovian conditioning licking behavior data.
(A) PSTH showing average licking behavior for the last 3 days of each condition for Pavlovian conditioning cohort 2 (n=9 mice). (B) Average anticipatory lick rate (baseline subtracted) for 1 s trace preceding reward delivery for cohort 2 (p=0.77). (C) Comparison of average baseline subtracted lick rate during the ramp window (3–8 s after cue onset) across all conditions for both cohorts (p=0.093, n=18 mice). (D) Comparison of average lick slope during the ramp window across all conditions (fixed tone vs dynamic tone: ***p=9.8 x 10–4).
Figure 2—figure supplement 2
Pavlovian conditioning cohort 2 histology and dopamine responses.
(A) Mouse coronal brain sections showing reconstructed locations of optic fiber tips (red circles) in NAcC for Pavlovian conditioning cohort 2. (B) Example average dLight traces for the last three days of all conditions. Vertical dashed lines at 3 and 8 s represent the ramp window period. Black lines display the linear regression fit during this period. (C) Same as in B but for the average dLight traces across all animals (n=9 mice).
Figure 2—figure supplement 3
Pavlovian conditioning cumulative dopamine data.
(A) Individual plots for each mouse from Pavlovian conditioning cohort 1 displaying the cumulative distribution of per-trial slopes for the last three days in all conditions. Vertical dashed lines indicate the average trial slope for LF (gray), LD (teal), SD (pink), and SF (purple) conditions. (B) Same as in A but for Pavlovian conditioning cohort 2 mice.
Figure 2—figure supplement 4
Pavlovian conditioning dopamine ramps do not correlate with pre-cue dopamine activity.
(A) Average dLight dopamine signal for Pavlovian conditioning cohort 1. Black lines represent linear regression fit during the pre-cue window and ramp window, each marked with gray shaded regions. (B) Average per-trial dLight slope during the pre-cue window for the last 3 days of each condition (p=0.37). (C) Linear regression β coefficients for per-trial ramp dLight slope vs. pre-cue dLight slope in SD condition calculated per animal (p=0.082). (D) Scatter plot with linear regression fit (black line) of Z-scored ramp slope vs pre-cue slope pooled across mice for all trials in the last 3 days of SD condition (p=0.17).
Figure 3 with 2 supplements
Per-trial dopamine ramps correlate with previous ITI.
(A) Scatter plot for an example animal showing the relationship between dopamine response slope within a trial and previous ITI for all trials in the last 3 days of SD condition. Plotted with linear regression fit (black line) used to find this animal’s β coefficient of –0.045. (B) Linear regression β coefficients for previous ITI vs. trial slope calculated per animal (***p=5.6 x 10–4). (C) Scatter plot of Z-scored trial slope vs. previous ITI pooled across mice for all trials in the last 3 days of SD condition (***p=6.6 x 10–10). The Z-scoring per animal removes the effect of variable means across animals on the slope of the pooled data. (D) dLight dopamine signal for two consecutive example SD trials showing the change in ITI and change in slope. The gray shaded regions indicate ITIs, and the vertical dashed lines mark the ramp window period. Reported m is slope. (E) Scatter plot for the same example animal in A showing the relationship between the change in dopamine slopes and the change in ITI across all trials in the last 3 days of SD condition. Plotted with linear regression fit (black line). Dot colors indicate magnitude of Δ ITI: light pink for Δ ITI below –1 s; gray for Δ ITI between –1 s and 1 s; dark pink for Δ ITI above 1 s. (F) Linear regression β coefficients for Δ ITI vs. Δ slope calculated per animal (***p=3.2 x 10–4). (G) Comparison of the average Δ slope for Δ ITI below –1 s vs above 1 s (*** p=2.3 x 10–4).
Figure 3—figure supplement 1
Pavlovian conditioning dopamine responses do not correlate with broader estimates of ITI.
(A) Linear regression β coefficients for trial dLight slope vs. average previous ITI for the past 1 through 10 ITIs calculated per animal for all trials in the last 3 days of the SD condition (**p=0.0056, ns p>0.05; using Benjamini-Hochberg Procedure). (B) Same as in A but for the LD condition (*p=0.019, ns p>0.05).
Figure 3—figure supplement 2
No significant correlations exist between additional dopamine and behavior measurements.
(A) Left, linear regression β coefficients for dLight slope vs. dLight onset peak calculated per animal in SD condition (p=0.54). Right, scatter plot with linear regression fit (black line) of Z-scored dLight slope vs dLight onset peak pooled across mice for all trials in the last 3 days of SD condition (p=0.84). (B) Same as in (A) but for LD condition (left p=0.54, right p=0.84). (C) Left, linear regression β coefficients for dLight slope vs. lick slope calculated per animal in SD condition (p=0.98). Right, scatter plot with linear regression fit (black line) of Z-scored dLight slope vs lick slope pooled across mice for all trials in the last 3 days of SD condition (p=0.84). (D) Same as in (C) but for LD condition (left p=0.52, right p=0.091). (E) Left, linear regression β coefficients for dLight slope vs. dLight onset peak calculated per animal in SD condition (p=0.52). Right, scatter plot with linear regression fit (black line) of Z-scored dLight slope vs dLight onset peak pooled across mice for all trials in the last 3 days of SD condition (p=0.15). (F) Same as in (E) but for LD condition (left p=0.54, right p=0.86). (G) Left, linear regression β coefficients for dLight slope vs. dLight onset peak calculated per animal in SD condition (p=0.52). Right, scatter plot with linear regression fit (black line) of Z-scored dLight slope vs dLight onset peak pooled across mice for all trials in the last 3 days of SD condition (p=0.44). (H) Same as in (G) but for LD condition (left p=0.54, right p=0.84). The Benjamini-Hochberg procedure is used for p values from all t-tests and p values from all linear regression separately for all comparisons in this figure.
Figure 4 with 2 supplements
VR navigation dopamine ramps depend on ITI.
(A) Head-fixed VR approach schematic. (B) VR navigation task experimental setup. Trials consisted of running down a patterned virtual hallway to receive sucrose reward. VR monitor remained black during the ITI. (C) Experimental timeline. Following training, mice were assigned to either long or short ITI conditions for 8 days before switching. (D) Velocity PSTH aligned to trial onset for long (teal) and short (pink) ITI conditions (n=9 mice). (E) Average change in velocity at trial onset. Bottom asterisks indicate both conditions significantly differ from zero (long: ***p=1.0 x 10–4, short: ***p=2.3 x 10–5). Top asterisks indicate significant difference between conditions (**p=0.0028). (F) Velocity PSTH aligned to reward delivery. (G) Average velocity during 1 s preceding reward (p=0.50). (H) PSTH showing average dLight dopamine signal aligned to trial onset. (I) Comparison of peak dLight onset response (***p=6.3 x 10–5). (J) Left, average dLight dopamine signal across distances spanning the entire virtual corridor. Vertical dashed lines represent the ramp window from 20 to 57 cm (10 cm before end of track). Solid black lines show linear regression fit during window. Right, close-up of dopamine signal during window. (K) dLight dopamine signal with linear regression fit during ramp window for example short ITI trials. Reported m is slope. (L) Session average per-trial slope during ramp window for all days of each condition. (M) Comparison of average per-trial slope during ramp window for last 3 days of both conditions (*p=0.035).
Figure 4—figure supplement 1
VR navigation task histology and responses.
(A) Mouse coronal brain sections showing reconstructed locations of optic fiber tips (red circles) in NAcC for VR navigation task. (B) Left, CDF of ITI duration for long (teal), medium (gray), and short (pink) ITI conditions. Middle, CDF plot of inter-reward interval (IRI) durations for each condition. Right, CDF plot of trial durations for each condition. (C) Comparison of average trial duration for long and short ITI conditions (p=0.34). (D) Lick rate PSTH aligned to reward delivery indicates minimal anticipatory licking behavior. (E) Scatter plot showing relationship between average per-session slope and inter-reward interval for the last 3 days in long (teal) and short (pink) ITI conditions. Black line indicates linear regression fit (*p=0.012). (F) CDF plots for each mouse separately showing the distribution of per-trial slopes for the last three days in both conditions. Vertical dashed lines indicate the average trial slope for long (teal) and short (pink) ITI conditions.
Figure 4—figure supplement 2
Trial-by-trial correlation of dopamine slope vs previous inter-reward interval (IRI) in the VR task.
(A) Scatter plot for an example animal showing the relationship between dopamine response slope within a trial and previous inter-reward interval (IRI) for all trials in the last 3 days of the short ITI condition. Plotted with linear regression fit (black line) used to find this animal’s β coefficient of –0.0014. Here, we are measuring the environmental timescale using IRI instead of ITI because the trial duration in this task (see Appendix 1) depends on the running speed of the animals, which varies trial to trial. Thus, IRI measures the net time interval between successive trial onsets. In the Pavlovian conditioning experiment, IRI and ITI differ by a constant since the trial duration is fixed. (B) Linear regression β coefficients for previous IRI vs trial slope calculated per animal (p=0.32). (C) Scatter plot of Z-scored trial slope vs. previous IRI pooled across mice for all trials in the last 3 days of the short ITI condition (*p=0.035).
Tables
Key resources table
| Reagent type (species) or resource | Designation | Source or reference | Identifiers | Additional information |
|---|---|---|---|---|
| Strain, strain background (Mus musculus, both sexes) | C57BL/6 J | Jackson Laboratory | RRID:IMSR_JAX:000664 | |
| Recombinant DNA reagent | AAVDJ-CAG-dLight1.3b | Patriarchi et al., 2018 | ||
| Software, algorithm | B-CALM | Zhou et al., 2024 | RRID:SCR_023884 | |
| Software, algorithm | Doric Neuroscience Studio | Doric Lenses | RRID:SCR_018569 | |
| Software, algorithm | pyPhotometry | Akam and Walton, 2019 | RRID:SCR_022940 | |
| Software, algorithm | BonVision | Lopes et al., 2021 | RRID:SCR_021534 | |
| Software, algorithm | Python | https://www.python.org/ | RRID:SCR_008394 |
Additional files
-
MDAR checklist
- https://cdn.elifesciences.org/articles/98666/elife-98666-mdarchecklist1-v1.docx
-
Supplementary file 1
Statistical details.
- https://cdn.elifesciences.org/articles/98666/elife-98666-supp1-v1.docx
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Mesolimbic dopamine ramps reflect environmental timescales
eLife 13:RP98666.
https://doi.org/10.7554/eLife.98666.3
