Impulsive Choices Emerge When the Anterior Cingulate Cortex Fails to Encode Deliberative Strategies

Abstract

Impulsive individuals excessively discount the value of delayed rewards, and this is thought to reflect deficits in brain regions critical for impulse control such as the anterior cingulate cortex (ACC). Delay discounting (DD) is an established measure of cognitive impulsivity, referring to the devaluation of rewards delayed in time. This study used male Wistar rats performing a DD task to test the hypothesis that neural activity states in ACC ensembles encode strategies that guide decision-making. Optogenetic silencing of ACC neurons exclusively increased impulsive choices at the 8 s delay by increasing the number of consecutive low-value, immediate choices. In contrast to shorter delays where animals preferred the delay option, no immediate or delay preference was detected at 8 s. These data suggest that ACC was critical for decisions requiring more deliberation between choice options. To address the role of ACC in this process, large-scale multiple single-unit recordings were performed and revealed that 4 and 8 s delays were associated with procedural versus deliberative neural encoding mechanisms, respectively. The 4 and 8 s delay differed in encoding of strategy corresponding to immediate and delay run termination. Specifically, neural ensemble states at 4 s were relatively stable throughout the choice but exhibited temporal evolution in state space during the choice epoch that resembled ramping during the 8 s delay. Collectively, these findings indicate that ensemble states in ACC facilitate strategies that guide decision-making, and impulsivity increases with disruptions of deliberative encoding mechanisms.

Keywords: decision-making; delay discounting; electrophysiology; impulsivity; optogenetics; prefrontal cortex.

Figure 1.

Description of optogenetics and electrophysiology experiments and evidence of discounting. A, Description of a single trial during DD depicting optogenetic inhibition of ACC (A, top) and epochs selected for analysis in awake-behaving recordings from ACC (A, bottom). The green highlighted portion of the trial depicts the points during a single trial where laser was turned ON (A, top) while the blue highlighted region (A, bottom) depicts the portion of the trial analyzed for the electrophysiology experiment. B, Example single-session depicting how an animal makes choices during the 4 s delay. Choice trials are depicted in red (immediate choices) and blue (delay choices) while forced trials are shown in white. An example of IM-Change and DEL-Fail-to-Change run is depicted by the red and blue horizontal lines, respectively, at the bottom. i-value refers to the number of pellets dispensed by the adjusting (Immediate) lever on a given trial. C, Mazur hyperbolic DD curve fit to indifference points across delays (0, 1, 2, 4, 8, 16 s) during Laser OFF sessions for optogenetic animals (n = 8) indicating the rate of discounting (k = −0.081). D, Probability of choosing an immediate choice at each delay. A shift away from preference toward the delay lever (0, 1, 2, and 4 s) as delays increase and an equal probability of choosing between immediate and delay levers at the 8 s delay. The horizontal dashed line indicates where animals were equally likely to choose between the immeidate and delay lever.

Figure 2.

Optogenetic inhibition of ACC prior to choice increases impulsivity measures and is disadvantageous. A, ArchT expression spread and optic fiber placements (left) for all animals (n = 8). Representative image of viral spread and optic fiber placements (right). B, Indifference points decrease as delay increases (ANOVA: F_(1.48,10.35) = 27.53, p = 0.0001). Indifference points were decreased on Laser ON (green) sessions compared with Laser OFF (black) sessions (ANOVA: F_(1,7) = 27.53, p = 0.003). Specifically, at the 8 s delay (Holm–Šídák test, p = 0.001), indifference points were decreased for Laser ON (green) compared with Laser OFF (black) sessions but not the 4 s (Holm–Šídák tests, p = 0.84) or 16 s delays (Holm–Šídák tests, p = 0.09). Optogenetic manipulation occurred at the 4, 8, and 16 s delays. C, The average number of pellets earned during choice trials for Laser ON sessions was lower than that during Laser OFF sessions for the 8 and 16 s delays. *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 3.

Optogenetic inhibition of ACC increases impulsivity by disrupting ability to deviate from low-value immediate choices at the 8 s delay. Choices were stratified by i-value (low, i-value <4 and high, i-value >3) and choice lever (immediate or delay). Consecutive number of immediate low i-value (A), immediate high i-value (B), delay low i-value (C), and delay high i-value (D) choices for Laser ON (green) versus Laser OFF (black) conditions were analyzed using a probability density functions (PDF). Optogenetic inhibition increases the consecutive number low i-value immediate choices (A) at the 8 s delay (Kolmogorov–Smirnov, p = 0.022). *p < 0.05.

Figure 4.

Neural activity of ACC neural populations during immediate and delay choices with low and high i-values during the 4 and 8 s delay. A, Electrophysiology placements of silicone probes (A, left). Representative image of sagittal slice with probe placement in the right ACC (left) and for all animals (A, right). B, C, Grand average mean firing rate for the 4 s (B; n = 2,120 neurons) and 8 s (C) delay (n = 2,078 neurons) separated by immediate low and high i-value (IM-Low, light blue; IM-High, dark blue) and delay low and high i-value (DEL-Low, light red; DEL-High, dark red) choices aligned to the time that the animal presses the choice lever (dashed line at time = 0; low, i-value <4 and high, i-value >3). Individual timepoints where IM-Low and IM-High firing rates differ (+) or DEL-High and DEL-Low differ (*) as indicated by FDR-corrected t tests are marked at the bottom of the graph (B, C). Scheffe multiple-comparison tests indicate whether overall firing rates differ between immediate and delay high and low i-value conditions within the figure legends (B, C). *p < 0.05, DEL-Low versus DEL-High i-value; ⁺⁺⁺⁺p < 0.0001, IM-Low versus IM-High i-value (B, C).

Figure 5.

PCA for runs during the 4- and 8 s delays reveal shift from procedural (4 s) to deliberative decision-making (8 s). A–F, Population activity during the third and fourth trial for each run (IM-Change, IM-Fail-to-Change, DEL-Change, DEL-Fail-to-Change, see key) from the 4 s (A, C, E) or 8 s (B, D, F) delay were analyzed using PCA. Runs consisted of an initial consecutive three choices on either the immediate (IM) or delay (DEL) lever, and the fourth trial consisted of either the “Change” (solid lines; end run) or “Fail-to-Change” (dashed lines; continue run) trial (see key). A, B, Trajectories (101 time bins/points) for the third and fourth trial of the four runs (8 trial types total) were analyzed and plotted in 3D space for each of the top 3 PCs for the 4 s (A) and 8 s (B) delays. The X denotes the choice point in each trajectory. C, D, Firing rates for each of the eight trial types were plotted for each individual PC for the 4 s (C_1–3) and 8 s (D_1–3) reveal that the individual PCs encode different dimensions of the decision-making process such as whether the run was IM or DEL (C₁, D₁) for PC1, run transition (C₂, D₂) in PC2 for the 4 s (C) and 8 s delay (D); however, the encoding of run transition is less clear at the 8 s delay (D₂). PC3 encoded trial 4 choice during the 4 s delay (C₃) and run transitions for DEL runs at the 8 s delay (D₃). E, F, Change in Euclidean distance (mean ± SEM) between the third and fourth trial across time (E, F, choice point indicated by vertical gray line) and average distance (±SEM) between the third and fourth trial trajectories (E, F, inset) for each of the four runs during the 4 s (E) and 8 s (F) delays. Individual timepoints where IM-Change and IM-Fail-to-Change distances differ (red, *) or DEL-Change and DEL-Fail-to-Change (blue, +) as indicated by FDR-corrected t tests are marked at the bottom of the graph (E, F). ⁺⁺p < 0.01, DEL-Change versus DEL-Fail-to-Change; **p < 0.01, IM-Change versus IM-Fail-to-Change.

Figure 6.

Choice latencies between the third and fourth trial of runs provide evidence of a shift from procedural (4 s) to deliberative decision-making (8 s). A, B, Choice latencies (mean rank ± SEM) compared for third and fourth trial for each of the four runs at the 4 s (A) and 8 s (B) delays. A, Choice latencies increase on the 4th compared with 3rd trial during the DEL-Change run (A₃, Wilcoxon signed-rank test: Z = −3.74, p = 0.0002). No differences in choice latencies were observed between the third and fourth trial for any other run (A₁, A₂, A₄, Wilcoxon signed-rank test: Z's < 1.72, p's > 0.09). B, Choice latencies significantly differed between the third and fourth trial for the IM-Change (B₁, Wilcoxon signed-rank test: Z = 2.74, p = 0.006) and DEL-Change (B₃, Wilcoxon signed-rank test: Z = −3.82, p = 0.0001) runs. No differences in choice latencies were observed between the third and fourth trial for the Fail-to-Change runs (B₂, B₄, Wilcoxon signed-rank test: Z's < 0.12, p's > 0.90). **p < 0.01, ***p < 0.001.