Lasting dynamic effects of the psychedelic 2,5-dimethoxy-4-iodoamphetamine ((±)-DOI) on cognitive flexibility

Abstract

Psychedelic drugs can aid fast and lasting remission from various neuropsychiatric disorders, though the underlying mechanisms remain unclear. Preclinical studies suggest serotonergic psychedelics enhance neuronal plasticity, but whether neuroplastic changes can also be seen at cognitive and behavioural levels is unexplored. Here we show that a single dose of the psychedelic 2,5-dimethoxy-4-iodoamphetamine ((±)-DOI) affects structural brain plasticity and cognitive flexibility in young adult mice beyond the acute drug experience. Using ex vivo magnetic resonance imaging, we show increased volumes of several sensory and association areas one day after systemic administration of 2 mgkg^-1 (±)-DOI. We then demonstrate lasting effects of (±)-DOI on cognitive flexibility in a two-step probabilistic reversal learning task where 2 mgkg^-1 (±)-DOI improved the rate of adaptation to a novel reversal in task structure occurring one-week post-treatment. Strikingly, (±)-DOI-treated mice started learning from reward omissions, a unique strategy not typically seen in mice in this task, suggesting heightened sensitivity to previously overlooked cues. Crucially, further experiments revealed that (±)-DOI's effects on cognitive flexibility were contingent on the timing between drug treatment and the novel reversal, as well as on the nature of the intervening experience. (±)-DOI's facilitation of both cognitive adaptation and novel thinking strategies may contribute to the clinical benefits of psychedelic-assisted therapy, particularly in cases of perseverative behaviours and a resistance to change seen in depression, anxiety, or addiction. Furthermore, our findings highlight the crucial role of time-dependent neuroplasticity and the influence of experiential factors in shaping the therapeutic potential of psychedelic interventions for impaired cognitive flexibility.

Fig. 1. (±)-DOI increased the volume of several sensory and association cortical areas within one day.

A Experiment timeline. The brains were collected one day after an injection of either 2 mgkg⁻¹ (±)-DOI or saline vehicle. B–E The False Discovery Rate (FDR) correction at an explorative 20% resulted in the marginal t-statistic = 3.49 (linear model F_1,14 = 12.19). Significant increases were in primary somatosensory (S1) shoulder (t = 4.25) and trunk (t = 4.15) areas, retrosplenial agranular area (RSA, t = 3.68), lateral parietal association area (LPtA, t = 3.61), ventral secondary auditory area (AuV, t = 3.53), and lateral secondary visual cortex (V2L, t = 3.49). Volume increases in the primary visual area (V1, t = 5.75) and temporal association area (TeA, t = 4.83) survived the strict 5% FDR (F_1,14 = 23.30, marginal t-statistic =4.83). n_group = 8. D: dorsal. V: ventral. L: left. R: right.

Fig. 2. Choice performance on the two-step task prior to any drug treatment reflects the combined influence of outcomes and transition types.

A Trial events. A mouse initiated the trial by poking the central port. In step 1, either both left/right ports lit up for the animals to choose which one they poke (free choice), or only the left or right port lit up to force the animal to explore that choice (forced choice, max. 25% of trials per session). Distinct auditory cues signalled active up or down port. A tone (identical to the up/down cues) or white noise would cue water reward delivery or omission, respectively. B The probabilistic structure of the task and the types of reversals. Reward probabilities of step 2 states reversed in blocks which could be non-neutral (reward probabilities switch between 80% and 20%) or neutral (both reward probabilities 50%). Reward reversals were triggered based on a behavioural criterion for non-neutral blocks (random interval of 5–15 trials after the exponential moving average across 8 previous free choices >75% correct) or after a random 20–30 trial interval for the neutral block. Animals were trained on serial reward reversals (“learned” adaptations). The transition structure was initially fixed until a single experimenter-directed transition reversal that occurred after drug treatment (“novel” adaptation). C Example sessions. Top: Transition and reward probabilities. Bottom: exponential moving average of choices (tau = 8 trials). Green lines represent reward blocks, and their y-position represents the correct choice (left, right, or neutral). D Stay probabilities for the step 1 choice were a function of subsequent common (C) and rare (R) transitions and reward (+), or omission (−), trial outcomes, as well as their interaction. RM two-way ANOVA (all P < 0.001, BF_incl ≫ 100): Transition F_1,81 = 351.7, η_p² = 0.81; Outcome F_1,81 = 43.9, η_p² = 0.35; Transition X Outcome F_1,81 = 541.1, η_p² = 0.87. Stars represent Bonferroni-corrected post-hoc comparisons. E Logistic regression analysis quantifying how transitions, outcomes, and their interaction predict repeating the same step 1 choice on the next trial. Correct and Choice predictors correct for cross-trial correlations and capture any side bias. One sample t-test or Wilcoxon signed rank tests against zero (all P < 0.001, BF₁₀ ≫ 100): Correct t₈₁ = 19.8, Cohen’s d = 2.18; Choice V = 3402, Rank-Biserial Corr. = 1.00; Outcome V = 3286, Rank-Biserial Corr. = 0.93; Transition V = 3403, Rank-Biserial Corr. = 1.00; Transition X Outcome V = 3403, Rank-Biserial Corr. = 1.00. F Lagged regression analysis shows how the repeated step 1 choices were influenced by the trial history. Stars represent family-wise Bonferroni-corrected one sample t-tests or Wilcoxon signed rank tests against zero. G Mice exhibit asymmetry in how they learn from positive and negative feedback. The reinforcing value of rewards was modulated by the type of transition, but reward omissions did not contribute to trial-by-trial learning. One sample t-tests or Wilcoxon signed rank tests against zero: Reward by transition V = 3403, P < 0.001, Rank-Biserial Corr. = 1.00, BF₁₀ ≫ 100; Omission by transition t₈₁ = 0.83, P = 0.411, BF₁₀ = 0.17. Data shown as mean ± SEM. n = 82. *P < 0.05. **P < 0.01. ***P < 0.001.

Fig. 3. (±)-DOI administered one week before a novel transition reversal improved learning via a new choice strategy.

A Two-step Experiment 1 timeline. After the animals were fully trained and had one week of stable performance on the two-step task (Pre-drug period), they were treated with either saline vehicle or 2 mgkg⁻¹ (±)-DOI at the end of the daily testing. The day after treatment, testing was continued on the same version of the task for one week (Post-drug period). A single novel transition reversal (TR) was then initiated at the start of the second post-drug week. Animals’ performance was tracked for a further two weeks (Post-TR period). B (±)-DOI induced a high frequency head-twitch response (unpaired Welch’s t-test, t_14.9 = 16.58, P < 0.001, Cohen’s d = 6.5, BF₁₀ » 100) and ear-scratch response (unpaired Welch’s t-test, t_17.8 = 4.11, P < 0.001, Cohen’s d = 1.6, BF₁₀ = 65.2). C The number of correct choices per session did not differ across treatment groups pre- or post-drug (two-way RM ANOVA: Time F_1,24 = 0.42, P = 0.521, BF_incl = 0.33; Drug F_1,24 = 0.34, P = 0.566, BF_incl = 0.50; Time X Drug F_1,24 = 0.002, P = 0.967, BF_incl = 0.36). Post-TR adaptation was not influenced by the drug (nonlinear one-phase association regression fits, F_3,96 = 0.63, P = 0.599, AICc₁₀ = 0.09). D The number of reward reversals per session did not differ pre-drug, but it increased post-drug irrespective of the treatment type (two-way RM ANOVA: Time F_1,24 = 4.3, P = 0.048, η_p² = 0.15, BF_incl = 1.51; Drug F_1,24 = 2.1, P = 0.160, BF_incl = 0.95; Time X Drug F_1,24 = 0.46, P = 0.503, BF_incl = 0.41). Post-TR, the treatment groups remained comparable (nonlinear line regression fits, F_2,100 = 0.110, P = 0.896, AICc₁₀ = 0.13, slope_global = 1.9, CI [1.4, 2.3]). E The pre-drug reward reversal performance was comparable across treatment groups (double exponential fit permutation test, tau_fast P = 0.671, tau_slow P = 0.851, tau_{fast/slow mix} P = 0.900). Post-drug, (±)-DOI-treated mice were quicker to reverse their choices (tau_fast P = 0.070, tau_slow P = 0.036, tau_{fast/slow mix} P = 0.046). In the late post-TR period (post-TR sessions 7–12, when the animals restarted doing many reward reversals per session), the trend for faster adaptation in (±)-DOI-treated mice persisted, but the difference was not statistically significant (tau_fast P = 0.130, tau_slow P = 0.085, tau_{fast/slow mix} P = 0.152). F Trial-to-trial learning of reward and transition probabilities was not different pre- and post-drug (two-way RM ANOVA: Time F_1,24 = 1.4, P = 0.246, BF_incl = 0.49; Drug F_1,24 = 1.3, P = 0.261, BF_incl = 0.66; Time X Drug F_1,24 = 0.4, P = 0.528, BF_incl = 0.41). (±)-DOI-treated mice were faster during the TR adaptation (nonlinear one-phase association regression fits, F_3,98 = 4.76, P = 0.004, AICc₁₀ = 40.10). G Trial-to-trial reward learning was not different pre- and post-drug (two-way RM ANOVA: Time F_1,24 = 3.4, P = 0.079, BF_incl = 1.04; Drug F_1,24 = 0.8, P = 0.387, BF_incl = 0.61; Time X Drug F_1,24 = 0.6, P = 0.444, BF_incl = 0.44). (±)-DOI-treated mice were faster during the TR adaptation (nonlinear one-phase association regression fits, F_3,98 = 2.82, P = 0.043, AICc₁₀ = 2.53). H Trial-to-trial reward omission learning was comparable pre- and post-drug (two-way RM ANOVA: Time F_1,24 = 0.13, P = 0.725, BF_incl = 0.29; Drug F_1,24 = 0.76, P = 0.392, BF_incl = 0.47; Time X Drug F_1,24 = 0.01, P = 0.907, BF_incl = 0.36). Post-TR, the best-fit of (±)-DOI-treated mice (horizontal line vs. line fit P = 0.007, AICc₁₀ = 14.81, slope_(±)-DOI = -0.112, CI [−0.192, −0.032]) was different to that of vehicle-treated mice (horizontal line vs. line fit P = 0.868, AICc₁₀ = 0.33, mean_Veh = 0.010, CI [−0.066, 0.086]). Trial-to-trial reward omission learning was not significant for either treatment group pre-TR (one sample t-tests, all P > 0.452, BF₁₀ < 0.36). Post-TR, vehicle-treated mice remained uninformed by omissions (one sample t-tests, all P > 0.739, BF₁₀ < 0.29), but (±)-DOI-treated mice started incorporating omissions in their choice strategy by the second post-TR week (one sample t-tests: [–3] t₁₂ = 0.87, P = 0.403, BF₁₀ = 0.38; [–6] t₁₂ = -0.87, P = 0.406, BF₁₀ = 0.38; [–9] t₁₂ = −2.82, P = 0.015, BF₁₀ = 4.08; [–12] t₁₂ = −2.48, P = 0.029, BF₁₀ = 2.46). Data shown as mean ± SEM. n_Veh = 13. n_(±)-DOI = 13. *P < 0.05. **P < 0.01.

Fig. 4. (±)-DOI administered one day before a novel transition reversal did not lead to improved learning.

A Two-step Experiment 2 timeline. The animals were treated with either saline vehicle or 2 mgkg⁻¹ (±)-DOI after they were fully trained and had one week of stable performance on the two-step task (Pre-drug period). A transition reversal (TR) was initiated the day after drug treatment. Animals’ performance was tracked for a further two weeks (Post-TR period). B (±)-DOI induced a high frequency head-twitch response (unpaired Welch’s t-test, t_13.3 = 15.61, P < 0.001, Cohen’s d = 5.9, BF₁₀ » 100) and ear-scratch response (unpaired Welch’s t-test, t_19.5 = 7.96, P < 0.001, Cohen’s d = 3.0, BF₁₀ » 100). C The number of correct choices per session was not significantly different across treatment groups pre-drug (unpaired t-test, t₂₇ = 0.096, P = 0.924, BF₁₀ = 0.35) or post-TR (nonlinear line regression fits, F_2,112 = 0.22, P = 0.804, AICc₁₀ = 0.14, slope_global = 0.067, CI [0.057, 0.076]). D The number of reward reversals per session was not different pre-drug (Mann–Whitney test, W = 86.5, P = 0.431, BF₁₀ = 0.42) or post-TR (nonlinear line regression fits, F_2,109 = 0.00008, P > 0.999, AICc₁₀ = 0.11, slope_global = 1.6, CI [1.3, 1.9]). E The pre-drug reward reversal performance was not different across treatment groups (double exponential fit permutation test, tau_fast P = 0.102, tau_slow P = 0.100, tau_{fast/slow mix} P = 0.244). In the late post-TR period, the two treatment groups remained comparable (tau_fast P = 0.169, tau_slow P = 0.520, tau_{fast/slow mix} P = 0.508). F Trial-to-trial learning of reward and transition probabilities was not different pre-drug (Mann–Whitney test, W = 120, P = 0.533, BF₁₀ = 0.46) or post-TR (nonlinear line regression fits, F_2,111 = 0.17, P = 0.837, AICc₁₀ = 0.14, slope_global = 0.35, CI [0.28, 0.43]). G Trial-to-trial reward learning was not significantly different pre-drug (Mann–Whitney test, W = 123, P = 0.451, BF₁₀ = 0.48) or post-TR (nonlinear line regression fits, F_2,112 = 0.30, P = 0.738, AICc₁₀ = 0.16, slope_global = 0.79, CI [0.65, 0.93]). H Trial-to-trial reward omission learning was not significantly different pre-drug (unpaired t-test t₂₇ = 0.15, P = 0.880, BF₁₀ = 0.35) or post-TR (nonlinear horizontal line regression fits, F_1,114 = 0.033, P = 0.856, AICc₁₀ = 0.35, mean_global = -0.055, CI [−0.116, 0.006]). Data shown as mean ± SEM. n_Veh = 15. n_(±)-DOI = 14.

Fig. 5. Effects of (±)-DOI on the adaptability to a novel reversal occurring one week after treatment were dependent on experience.

A Two-step Experiment 3 timeline. The animals were treated with either saline vehicle or 2 mgkg⁻¹ (±)-DOI when they were fully trained and had one week of stable performance on the two-step task (Pre-drug period). For the next week, the animals were kept under water deprivation but were not allowed any further experience on the two-step task. A novel transition reversal (TR) was then initiated at the start of the second post-drug week. Animals’ performance was tracked for a further two weeks (Post-TR period). B (±)-DOI induced a high frequency head-twitch response (unpaired Welch’s t-test, t_13.4 = 14.24, P < 0.001, Cohen’s d = 5.5, BF₁₀ » 100) and ear-scratch response (Mann–Whitney U test, W = 7, P < 0.001, Rank-Biserial Corr. = 0.92, BF₁₀ = 28.8). C The number of correct choices per session did not differ across treatment groups pre-drug (unpaired t-test, t₂₅ = -0.381, P = 0.707, BF₁₀ = 0.38) or post-TR (nonlinear line regression fits, F_2,104 = 2.24, P = 0.111, AICc₁₀ = 1.11, slope_global = 0.064, CI [0.051, 0.077]). D The number of reward reversals per session was not significantly different pre-drug (unpaired t-test, t₂₅ = −0.045, P = 0.965, BF₁₀ = 0.36) or post-TR (nonlinear line regression fits, F_2,102 = 2.54, P = 0.084, AICc₁₀ = 1.48, slope_global = 1.7, CI [1.2, 2.2]). E The pre-drug (double exponential fit permutation test, tau_fast P = 0.513, tau_slow P = 0.354, tau_{fast/slow mix} P = 0.610) and late post-TR (tau_fast P = 0.560, tau_slow P = 0.719, tau_{fast/slow mix} P = 0.368) reward reversal performance was comparable across treatment groups. F Trial-to-trial learning of reward and transition probabilities was not different pre-drug (unpaired t-test t₂₅ = 0.123, P = 0.903, BF₁₀ = 0.36). (±)-DOI group’s fit was significantly different post-TR. Nonlinear regression fits: (±)-DOI, line fit (one-phase association model failed to converge on a best-fit curve), slope = 0.064, CI [0.051, 0.077]; Vehicle, line vs. one-phase association fit: F_1,49 = 6.19, P = 0.016, AICc₁₀ = 6.79. G Trial-to-trial reward learning was not different pre-drug (unpaired t-test, t₂₅ = 0.449, P = 0.657, BF₁₀ = 0.39) or post-TR (nonlinear line regression fits, F_2,104 = 3.01, P = 0.054, AICc₁₀ = 2.36, slope_global = 0.47, CI [0.39, 0.56]). H Trial-to-trial reward omission learning was not significantly different pre-drug (unpaired t-test, t₂₅ = 0.563, P = 0.578, BF₁₀ = 0.40) or post-TR (nonlinear horizontal line regression fits, F_1,106 = 0.229, P = 0.633, AICc₁₀ = 0.39, mean_global = -0.064, CI [−0.125, −0.002]). Trial-to-trial reward omission learning was not contributing to the choice strategy of either treatment group (one sample t-tests, all P > 0.095, BF₁₀ < 0.97). Data shown as mean ± SEM. n_Veh = 13. n_(±)-DOI = 14.