Psilocybin restrains activity-based anorexia in female rats by enhancing cognitive flexibility: contributions from 5-HT1A and 5-HT2A receptor mechanisms

Abstract

Psilocybin has shown promise for alleviating symptoms of depression and is currently in clinical trials for the treatment of anorexia nervosa (AN), a condition that is characterised by persistent cognitive inflexibility. Considering that enhanced cognitive flexibility after psilocybin treatment is reported to occur in individuals with depression, it is plausible that psilocybin could improve symptoms of AN by breaking down cognitive inflexibility. A mechanistic understanding of the actions of psilocybin is required to tailor the clinical application of psilocybin to individuals most likely to respond with positive outcomes. This can only be achieved using incisive neurobiological approaches in animal models. Here, we use the activity-based anorexia (ABA) rat model and comprehensively assess aspects of reinforcement learning to show that psilocybin (post-acutely) improves body weight maintenance in female rats and facilitates cognitive flexibility, specifically via improved adaptation to the initial reversal of reward contingencies. Further, we reveal the involvement of signalling through the serotonin (5-HT) 1 A and 5-HT2A receptor subtypes in specific aspects of learning, demonstrating that 5-HT1A antagonism negates the cognitive enhancing effects of psilocybin. Moreover, we show that psilocybin elicits a transient increase and decrease in cortical transcription of these receptors (Htr2a and Htr1a, respectively), and a further reduction in the abundance of Htr2a transcripts in rats exposed to the ABA model. Together, these findings support the hypothesis that psilocybin could ameliorate cognitive inflexibility in the context of AN and highlight a need to better understand the therapeutic mechanisms independent of 5-HT2A receptor binding.

Fig. 1. Effects of psilocybin on body weight maintenance in ABA.

Weight loss trajectories of individual rats (n = 16 saline; n = 19 psilocybin) over the 10-day ABA period (A) and proportion resistant to weight loss (B). Psilocybin facilitated body weight maintenance over 85% for more days (C, t(33) = 2.508, p = 0.0172), with a trend toward lower body weight % loss per day (D, t(33) = 1.918, p = 0.0638) that resulted in attenuation of severe weight loss (E, t(33) = 2.146, p = 0.0394). Total daily wheel revolutions (F) increased as expected in ABA (G, ABA Phase F(1, 33) = 126.5, p < 0.0001) but were similar between groups (Treatment F(1, 33) = 1.159, p = 0.2985) across both baseline (p > 0.9999) and ABA (p = 0.3089; Interaction F(1, 33) = 1.033, p = 0.3169) with no difference in the change in proportional running wheel activity in the penultimate hour before food access (H, t(33) = 0.1.098, p = 0.2800). Ninety-minute food intake (I) increased similarly across the ABA phase with no difference in mean daily intake (J, t(33) = 0.9908, p = 0.3290). Comparison of only psilocybin treated rats that were susceptible (PSI-S) versus resistant (PSI-R) to ABA highlights the characteristic starvation-induced hyperactivity displayed by PSI-S (K) during the first 7 days of exposure to ABA conditions (L, ABA Phase F(1, 17) = 36.48, p < 0.0001; ABA Outcome F(1, 17) = 19.01, p < 0.0001; Interaction F(1, 17) = 17.09, p = 0.0047; ABA PSI-S > PSI-R p < 0.0001), in contrast to the selective increase of running in anticipation of food access displayed by PSI-R (M, t(17) = 6.203, p < 0.0001), accompanied by diverging food intake trajectories (N) with greater mean 7-day intake by PSI-R (O, t(17) = 2.577, p = 0.0196). Comparison of ABA susceptible rats that received psilocybin (PSI-S) or saline (SAL-S) revealed no differences in the development of starvation-induced hyperactivity (P) following the onset of ABA conditions (Q; ABA Phase F(1, 24) = 126.5, p < 0.0001; Treatment F(1, 24) = 1.159, p = 0.2895; Interaction F(1, 24) = 1.033, p = 0.3269), selective running in anticipation of food (R; t(24) = 0.6452, p = 05252) or food intake over time (S) or on average (T; t(24) = 0.4783, p = 0.6368). Grouped data show mean ± SEM, with individual data points on bar graphs. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. SAL saline, PSI psilocybin, BW body weight, ABA activity-based anorexia, FAA food anticipatory activity, PSI-S psilocybin treated ABA susceptible, PSI-R psilocybin treated ABA resistant. For full statistical analysis details see Fig. 1 Statistics Table.

Fig. 2. Effects of psilocybin on reversal learning, effortful responding and response suppression.

Psilocybin administered after training the day prior to reversal of reward contingencies (A) significantly improved accuracy of responding during the initial 3 h reversal session (B, Treatment F(1, 29) = 5.128, p = .0312; 6 x 30 min time bins) and increased the number of rats (C) able to reach performance criterion (D, 80 target pokes in a 100-poke moving window) on the first day of reversed reward contingencies. While there was no difference in the time (from first poke to poke that achieved criterion; (E), t(18) = 1.514, p = .1474), total pokes (F, t(18) = 1.536, p = 0.1420) or target pokes (G, t(18) = 0.5614, p = 0.5815) required to reach criterion, psilocybin-treated rats required fewer non-target pokes to reach criterion (H, t(18) = 2.425, p = 0.0260), tended to earn rewards faster (I, t(18) = 1.759, p = 0.0956), and were both faster to first engage with the task (time from device access to first poke; (J), t(18) = 2.307, p = 0.0332) and to earn their first reward (time from first poke to earning first pellet; (K), t(18) = 2.291, p = 0.0343). Psilocybin treatment had no effect on breakpoint (pokes required to earn final pellet before 10 min of inactivity) on a classic progressive ratio task (L, t(23) = 0.7795, p = 0.4436), extinction following fixed ratio training (M, Treatment F(1, 17) = 0.3212, p = 0.5783; Interaction F(179, 3043) = 0.2623, p > 0.9999), goal directed engagement on increasingly uncertain schedules of reinforcement (N, Treatment F(1, 21) = 1.741, p = 0.2013; Interaction F(2, 42) = 0.7244, p = 0.4906), or extinction following variable ratio training (O, Treatment F(1, 21) = 0.2681, p = 0.6100; Interaction F(179, 3759) = 0.3509, p > 0.9999). Grouped data show mean ± SEM, with individual data points on bar graphs. *p < .05. SAL saline, PSI psilocybin, FR5 fixed ratio 5, VR variable ratio. For full statistical analysis details see Fig. 2 Statistics Table.

Fig. 3. Effects of 5-HT2A and 5-HT1A antagonism on reversal learning in control and psilocybin-treated rats.

Following the final training session pre-treatment with either saline (vehicle control), the 5-HT2AR antagonist MDL100907, or the 5-HT1AR antagonist WAY100635, was followed 30 min later by treatment with either saline (A) or psilocybin (L) before reversal of reward contingencies the following day, with first reversal day performance accuracy highlighted (B, M, respectively). While 53/5% (8/15) of Saline+SAL rats reached reversal day 1 criterion, 0% (0/9) of MDL + SAL treated rats did so (C), showing global impairment compared to Saline+SAL across nearly all outcome measures, achieving significantly lower session accuracy (D, SAL + > MDL+ p = .0221), earning fewer pellets (E, SAL + > MDL+ p = .0324), and making fewer target (F, SAL + > MDL+ p = .0292) and non-target (G, SAL + > MDL+ p = 0.0048) pokes in the session. While there was no delay in task engagement (time from device access to first poke; H, SAL+ vs MDL+ p = 0.4572), target poke latency was increased (I, SAL + < MDL+ p = 0.0502) in the 6/9 rats that made a target poke, and only 3/9 rats earned a single pellet (i.e. made at least 5 target pokes; (J) time from first poke to earning first pellet, SAL+ vs MDL+ p = 0.1507), even though task engagement duration did not differ (time from first to final poke; (K), SAL+ vs MDL+ p = 0.6048). Conversely, WAY + SAL resulted in 53.8% (7/13) of rats reaching criterion, nearly identical to Saline+SAL, with these groups being similar across most measures except WAY + SAL having fewer non-target pokes (G, SAL + > WAY+ p = 0.0041) despite an elongated target poke latency (I, SAL + < WAY+ p = 0.0244). With 75% (12/16) of Saline+PSI rats reaching reversal day 1 criterion, MDL + PSI treatment produced a moderate decrease to 55.6% (5/9) reaching criterion (N), although only a non-significant decrease in accuracy (O, SAL+ vs MDL+ p = 0.2837), while there was a trend toward fewer pellets (P, SAL + > MDL+ p = 0.0698) and target pokes (Q, SAL + > MDL+ p = 0.0542), and a significant reduction in non-target pokes (R, SAL + > MDL+ p = 0.0210) across the session, with no differences in any latency measures (S–U, all SAL+ vs MDL+ ps > .2991) or session duration (V, SAL+ vs MDL+ p = 0.4345). In contrast, WAY + PSI produced severe impairment with only 15.4% (2/13) of rats reaching criterion, with significantly reduced session accuracy (O, SAL + > WAY+ p = .0024), pellets earned (P, SAL + > WAY+ p = 0.0015), and target pokes (Q, SAL + > WAY+ p = 0.0013), and delayed target poke latency (T, SAL + < WAY+ p = 0.0212, with only 10/13 rats achieving a target poke) compared to Saline+PSI, whilst there were no differences for non-target pokes (R, SAL+ vs WAY+ p = 0.9497), first poke latency (S, SAL+ vs WAY+ p = 0.1636), relative first pellet latency (U, SAL+ vs WAY+ p = 0.2588, although only 7/13 earned a pellet), nor session duration (V, SAL+ vs WAY+ p = 0.7309). Bar graphs show mean ± SEM with individual data points. *p < 0.05, **p < 0.01. SAL saline, PSI psilocybin, SAL+ saline pre-treatment, MDL + MDL100907 pre-treatment; WAY + WAY100635 pre-treatment. For main ANOVA results and full statistical analysis details see Fig. 3 Statistics Table.

Fig. 4. Effects of 5-HT2AR and 5-HT1AR antagonism on psilocybin-induced improvements in reversal learning.

Reversal learning following 5-HT2AR antagonism via pre-treatment with MDL100907 (A) was completely impaired in saline treated animals (0/9 [0%] reached reversal day 1 criterion) whereas psilocybin treatment prevented this impairment (5/9 [55.6%] reached criterion). Psilocybin treatment following MDL-mediated 5-HT2AR antagonism resulted in significantly greater session accuracy (B, t(16) = 3.034, p = 0.0079), pellets earned (C, t(16) = 2.255, p = 0.385), and target pokes made (D, t(16) = 2.191, p = 0.0436) compared to saline treatment, with no differences for non-target pokes (E, t(16) = 0.7609, p = 0.4578), first poke latency (time from device access to first poke; F, t(16) = 1.034, p = 0.3163), target poke latency (G, t(11) = 0.5098, p = 0.6202), relative first pellet latency (time from first poke to earning first pellet;(H), t(6) = 1.295, p = 0.2428), or session duration (time from first to final poke; (I), t(16) = 1.133, p = 0.2741). The opposite performance pattern was observed following 5-HT1AR antagonism via WAY100635 pre-treatment (J), with 7/13 (53.8%) saline treated rats reaching criterion compared with only 2/13 (15.4%) psilocybin treated rats. Although not significant, psilocybin treatment produced a trend toward lower session accuracy (K, t(24) = 2.011, p = 0.0556), fewer pellets earned (L, t(24) = 1.806, p = 0.0834), and target pokes made (M, t(24) = 1.771, p = 0.0893), whilst there was no difference between groups for non-target pokes (N, t(24) = 1.317, p = 0.2001), first poke (O, t(24) = 0.7925, p = 0.4538), target poke (P, t(19) = 0.1996, p = 0.8440), or relative first pellet (Q, t(14) = 0.3062, p = 0.7639) latency, or session duration (R, t(24) = 0.1337, p = 0.8948). Bar graphs show mean ± SEM with individual data points. *p < 0.05, **p < 0.01. SAL saline, PSI psilocybin. For full statistical analysis details see Fig. 4 Statistics Table.

Fig. 5. Effects of psilocybin on the expression of Htr1a and Htr2a transcripts in the mPFC.

Coronal section with brain atlas overlay (AP + 3.2 mm from bregma); (A) depicting regions of interest (PrL and IL). The proportion of Htr1/2a+ cells that were double labelled with Htr1a and Htr2a was not changed by psilocybin treatment in either the PrL (B, F(3, 15) = 0.7801, p = .5233) or IL (C, F(3, 15) = 0.2449, p = 0.8637). The proportion of Htr1/2a+ cells that were exclusively Htr1a labelled was increased following psilocybin administration in both the PrL (D, F(3, 15) = 2.443, p = 0.1043, SAL<PSI12h p = 0.0500) and the IL (E, F(3, 15) = 4.277, p = 0.0227, SAL<PSI6h p = 0.0525, SAL<PSI12h p = 0.0103), whilst those exclusively Htr2a labelled decreased following psilocybin treatment at a trend level in PrL (F, F(3, 15) = 2.192, p = 0.1314, SAL>PSI12h p = 0.0931) and significantly in IL (G, F(3, 15) = 4.426, p = 0.0203, SAL>PSI6h p = 0.0335, SAL>PSI12h p = 0.0129). The spatial distribution of IL Htr1/2a+ cells along the midline from Layer I (H) was significantly different for each uniquely labelled cell population (I₁ F(49, 300) = 10.34, p < 0.0001; J₁ F(49, 300) = 3.549, p < 0.0001; K₁ F(49, 300) = 3.436, p < 0.0001). In each case psilocybin treatment also had a significant effect, producing a significantly reduced overall proportion of double labelled cells (I₁, F(1, 300) = 9.214, p = 0.0026) and exclusively Htr2a labelled cells (K₁, F(1, 300) = 19.16, p < 0.0001), but a significantly increased overall proportion of exclusively Htr1a labelled cells (J₁, F(1, 300) = 22.38, p < 0.0001) accompanied by a significant Distance by Treatment interaction (J₁, F(49, 300) = 1.459, p = 0.0313). AUC was decreased at a trend level for double labelled cells (I₂, t(6) = 2.030, p = 0.0887), significantly increased for exclusively Htr1a labelled cells (J₂, t(6) = 3.102, p = 0.0211) and significantly reduced for exclusively Htr2a labelled cells (K₂, t(6) = 3.097, p = 0.0212). A separate cohort of animals underwent ABA induction, were administered either saline or psilocybin when they reached <85% baseline body weight, and culled ~6h later (when bodyweight had dropped to close to ~80% in most cases; L). The proportion of mPFC Htr1/2a+ cells that expressed both Htr1a and Htr2a was not effected by psilocybin administration nor ABA exposure (M, Treatment F(1, 17) = 0.5663, p = 0.4620; ABA Exposure F(1, 17) = 0.4685, p = 0.5029; Interaction F(1, 17) = 1.208, p = 0.2871), whereas psilocybin significantly increased or significantly decreased the proportion of exclusively Htr1a labelled (N, Treatment F(1, 17) = 15.50, p = 0.0011; Non-ABA SAL < PSI p = 0.0298, ABA SAL < PSI p = 0.0206) or Htr2a labelled (O, Treatment F(1, 17) = 9.038, p = 0.0079; Non-ABA SAL > PSI p = 0.0463) cells, respectively, in a generally consistent and ABA independent manner (all ABA exposure and interaction ps > 0.4607). Htr1a (green) and Htr2a (red) expression on distinct cell populations in the mPFC (P) identified through DAPI (blue). The absolute number of Htr2a transcripts associated with mPFC Htr1/2a+ cells (Q) was significantly altered by psilocybin (F(1, 17) = 4.587, p = 0.0470), ABA exposure (F(1, 17) = 5.098, p = 0.0374), and their interaction (F(1, 17) = 15.26, p = 0.0011), such that psilocybin treatment significantly reduced Htr2a copy number specifically in the ABA brain (ABA SAL > PSI p = 0.0005). This pattern was mostly replicated by the number of Htr2a copies per mPFC Htr1/2a+ cell (R, Treatment F(1, 17) = 12.12, p = 0.0029; ABA Exposure F(1, 17) = 2.048, p = 0.1705; Interaction F(1, 17) = 5.606, p = 0.0300), with a significant reduction in the density of Htr2a specifically following psilocybin treatment after ABA induction (ABA SAL > PSI p = 0.0007). Grouped data show mean ± SEM, with individual data points on bar graphs (except AUC). Values are the average of 4 (PrL and IL) or 8 (mPFC) sections per animal. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. AP anterior-posterior, SAL saline, PSI psilocybin, PrL prelimbic cortex, IL infralimbic cortex, AUC area under the curve, mPFC medial prefrontal cortex (PrL and IL combined); Htr1/2a+ cells expressing Htr1a and/or Htr2a; ABA activity-based anorexia. Scale bars for (A) 2 mm and (P) 30 µm. For full statistical analysis details see Fig. 5 Statistics Table.