Modeled Long-Term Effects of Psilocybin on Dynamic Activity and Effective Connectivity of Fronto-Striatal-Thalamic Circuits
This secondary analysis and computational modelling study of resting-state fMRI data from a within-subject longitudinal psilocybin trial in psychedelic-naïve healthy volunteers found that fronto-striatal-thalamic circuits became more dynamically active 4 weeks after a full dose. The changes were linked to reduced top-down and increased bottom-up information flow, with receptor maps suggesting roles for 5-HT2A and D2 receptors.
Authors
- Pasquini, L.
- Vohryzek, J.
- Escrichs, A.
Published
Abstract
Psilocybin has been shown to induce fast and sustained symptoms improvements across various psychiatric conditions, yet its long‐term mechanisms of action are not fully understood. Initial evidence suggests that longitudinal functional and structural brain changes implicate fronto‐striatal‐thalamic (FST) circuitry, a broad system involved in goal‐directed behavior and motivational states. Here, we performed secondary analyses and applied computational modeling to resting‐state fMRI data from a within‐subject longitudinal psilocybin trial in psychedelic‐naïve healthy volunteers. We first showed that dynamic FST activity increased 4 weeks after a full dose of psilocybin. We then proceeded to mechanistically account for these changes by providing tentative model‐based support that reductions in the structure–function coupling contribute to increased dynamic FST activity postpsilocybin. Finally, we used computational approaches to show that psilocybin induces longitudinal increases in bottom‐up and reduced top‐down modulation of FST circuits. We then used publicly available receptor maps to show that cortical reductions in top‐down modulation are linked to regional 5‐HT2A receptor availability, while increased information outflow via subcortical and limbic regions relates to local D2 receptor availability. Together, these findings suggest that increased FST flexibility weeks after a high dose of psilocybin is linked to serotonergic‐mediated decreases in top‐down information flow and dopaminergic‐mediated increases in bottom‐up information flow. This long‐term functional re‐organization of FST circuits may represent a common mechanism contributing to the potential clinical efficacy of psilocybin across various neuropsychiatric disorders including substance abuse, major depression, and anorexia nervosa.
Research Summary of 'Modeled Long-Term Effects of Psilocybin on Dynamic Activity and Effective Connectivity of Fronto-Striatal-Thalamic Circuits'
βBlossom's Take
Introduction
Psilocybin is increasingly studied as a treatment for several psychiatric conditions, and earlier work has suggested that its clinical effects may be accompanied by both acute and longer-lasting changes in brain organisation. Prior neuroimaging studies have reported increases in global brain network integration days to weeks after psychedelic administration, with some changes linked to improved well-being and to altered activity in medial frontal regions. However, the long-term neural mechanisms remain uncertain. In particular, the authors note that fronto-striatal-thalamic (FST) circuitry may be important because it sits at the intersection of cognitive control, motivation, reward, and several neuromodulatory systems, but its post-psilocybin reorganisation had not been well characterised. Pasquini and colleagues set out to examine whether a full dose of psilocybin produces longer-term changes in FST dynamic activity and information flow in psychedelic-naïve healthy volunteers. They aimed to test whether FST activity becomes more dynamic 4 weeks after 25 mg psilocybin, whether such changes relate to mental well-being, and whether computational models could explain the findings in terms of altered structure-function coupling and effective connectivity. They also explored whether these network changes aligned spatially with regional 5-HT2A receptor availability and D2 receptor availability. The study is a secondary analysis of a longitudinal within-subject trial, using resting-state fMRI and questionnaire data from the same participants across a control dose and a full psilocybin dose.
Methods
The researchers analysed longitudinal neuroimaging and questionnaire data from 25 healthy adults with no prior psychedelic experience in a fixed-order within-subject design. All participants completed a low, subthreshold dose of 1 mg psilocybin first, then 4 weeks later received a full dose of 25 mg psilocybin. Resting-state fMRI and surveys were collected at baseline, 4 weeks after the 1 mg session, and 4 weeks after the 25 mg session. Mental well-being was measured using the 14-item Warwick-Edinburgh Mental Wellbeing Scale (WEMWBS). Resting-state imaging was acquired on a 3T Siemens Tim Trio scanner. After standard preprocessing, the authors used the Automated Anatomical Labelling 90 atlas to define 28 FST regions of interest spanning frontal cortex, thalamus, and basal ganglia. From these regions they derived intrinsic functional connectivity matrices using Pearson correlations between regional time series. They also computed dynamic functional homogeneity (DFH), a measure of how similar regional activity is across time within a scan; higher DFH indicates more homogeneous, less dynamically variable activity. To explore mechanisms, the authors used two modelling approaches. First, they fitted partial-brain Hopf models to the FST network, using a weighted structural connectivity matrix derived from diffusion tensor imaging in 16 healthy young adults. These models were fit separately to the post-1 mg and post-25 mg conditions across a range of global coupling values, allowing them to identify the coupling parameter that best reproduced empirical DFH and, in separate analyses, empirical intrinsic functional connectivity. They repeated the modelling stochastically 30 times to assess robustness and also used 1000 within-subject permutations of DFH to build a null distribution for testing the coupling differences. Second, they estimated effective connectivity using a time-shifted correlation framework that compares forward and time-reversed signal trajectories to infer directionality of information flow, again constrained by structural connectivity. Separate whole-brain models were fitted for the post-1 mg and post-25 mg scans, with 30 stochastic simulations used to assess stability. They then summarised directional flow as inflowing and outflowing effective connectivity for each node. Finally, they compared nodal changes in effective connectivity with publicly available PET maps of 5-HT2A receptor availability and D2 receptor availability, parcellated to the same atlas. Cortical and subcortical analyses were treated separately because receptor distributions differ markedly between these tissue types. Statistical testing included repeated-measures ANOVA, paired t-tests, Pearson and Spearman correlations, Wilcoxon tests, permutation testing, and false discovery rate correction where appropriate.
Results
Mean FST DFH did not differ significantly between baseline and 4 weeks after the 1 mg control dose. In contrast, DFH 4 weeks after 25 mg psilocybin was significantly lower than DFH after 1 mg psilocybin (t(24) = 2.87, p = 0.008), and it showed a trend towards being lower than baseline (t(24) = 1.92, p = 0.06). Because lower DFH indicates greater temporal heterogeneity, these results were interpreted as increased dynamic activity in FST circuits after the full psilocybin dose. At the whole-brain level, DFH decreased after both the 1 mg and 25 mg sessions relative to baseline, suggesting possible order effects related to the baseline scan. Head motion did not differ significantly across sessions. The reduction in FST DFH from the post-1 mg to the post-25 mg scan showed a trend-level association with an increase in WEMWBS mental well-being scores over the same interval (Rho(22) = -0.42, p = 0.06), after correction for baseline well-being, baseline DFH, and head movement. In the partial-brain modelling, the best-fitting global coupling parameter was lower after 25 mg psilocybin than after 1 mg. Across 30 stochastic simulations, the mean optimal G was 0.72 (SD 0.06) in the post-25 mg condition and 0.91 (SD 0.06) in the post-1 mg condition, with a reported difference of z(59) = 6.56, p < 0.0005. However, when the authors repeated the analysis using 1000 within-subject permutations of DFH, the difference in optimal G between conditions was not significant (z(999) = 0.03, p = 0.976), and the observed effect was smaller than 90% of the permuted differences (p1000 = 0.064). Fitting the models to static intrinsic functional connectivity did not show significant condition differences, suggesting DFH was more sensitive to the changes than static connectivity. Effective connectivity analyses showed a shift in information flow after 25 mg psilocybin. Outflowing and inflowing effective connectivity decreased in dorsal cortical regions, including dorsolateral prefrontal and medial frontal areas, but increased in limbic and subcortical regions, including effects that were especially evident in the left hemisphere for inflow. Changes in outflowing and inflowing effective connectivity were strongly correlated across nodes (Rho(88) = 0.87, p < 0.0005). Spatially, cortical changes in effective connectivity were negatively associated with 5-HT2A receptor availability: 5-HT2A maps correlated negatively with cortical inflowing changes (Rho(18) = -0.56, p < 0.05) and cortical outflowing changes (Rho(18) = -0.47, p < 0.05). No significant relationships were found in subcortical regions, and the cortical-subcortical correlations differed significantly. By contrast, D2 receptor availability correlated positively with subcortical changes in inflowing connectivity (Rho(8) = 0.88, FDR-corrected p < 0.05) and outflowing connectivity (Rho(8) = 0.95, FDR-corrected p < 0.005), but not with cortical changes.
Discussion
The authors interpret their findings as evidence that a full dose of psilocybin is followed, 1 month later, by more dynamic FST activity and altered directional communication within these circuits. They emphasise that the decrease in DFH after 25 mg indicates greater temporal heterogeneity in FST activity, and they link this to the trend towards improved mental well-being over the same period. They note that the absence of a clear baseline-to-25 mg difference may reflect the fixed-order within-subject design, with possible order effects such as scanner habituation, expectancy, or physiological adaptation. They argue that the lower optimal coupling parameter in the model suggests that structural connectivity exerts less constraint on the emergence of FST dynamics after psilocybin, which could permit more flexible and heterogeneous activity. In their account, this provides a preliminary mechanistic explanation for the empirical increase in dynamic FST activity. They also present the effective connectivity findings as showing decreased top-down influence from higher-order cortical regions and increased bottom-up and subcortical information flow. These patterns are said to align more closely with the REBUS account of psychedelic action, which proposes a relaxation of high-level priors and weaker top-down control, than with a simple thalamic gating model derived from acute effects. The authors further suggest that the cortical effective connectivity changes map onto regional 5-HT2A receptor availability, whereas the subcortical increases align with D2 receptor availability. They interpret this as consistent with serotonergic involvement in cortical top-down changes and possible indirect dopaminergic involvement in subcortical information flow, while noting that psilocybin is not a direct dopaminergic agonist. Several limitations are acknowledged. The empirical effects were weak to moderate statistically, and the permutation-based modelling result for global coupling only reached trend level, which the authors say may reflect limited statistical power. They also stress that the repeated stochastic simulations were used to test stability rather than to generate independent inferential observations. Additional limitations include the coarse FST parcellation, the use of group-level structural connectomes and normative receptor maps, and the inability to examine other relevant circuits such as anterior hippocampus-medial parietal pathways or other receptor systems such as glutamate and GABA. The authors conclude that their results open a mechanistic avenue for understanding how psilocybin may produce longer-term behavioural changes, including increased openness and well-being and reduced anhedonia, apathy, and substance craving. They suggest that future work should test whether this FST reorganisation underlies psilocybin’s apparent clinical effects across disorders.
View full paper sections
| SAMPLE AND STUDY DESIGN
We used longitudinal neuroimaging data and questionnaires acquired as part of a fixed-order, within-subjects design to investigate the effects of psilocybin in 25 healthy human adults with no prior psychedelic experience (Table). All participants received: (i) a control dose of 1 mg psilocybin on the first dosing day, considered to be a subthreshold dose insufficient to induce a psychedelic experience; and (ii) a fully active dose of 25 mg psilocybin 4 weeks later, considered to be a high dose inducing profound psychedelic effects with a substantially higher occupancy of 5-HT2A receptors when compared to 1 mg (Figure). Mental well-being was measured using the 14-item Warwick-Edinburgh Mental Wellbeing Scale (WEMWBS, population mean [range] = 51 [14-70])). Participants' WEMWBS was assessed at baseline and 4 weeks after each dosing visit. For further information see Supporting Information Methods and previous work.
| FMRI ACQUISITION, PREPROCESSING, AND EMPIRICAL ANALYSIS
Imaging was performed on a 3T Siemens Tim Trio using a 12-channel head coil at Imperial College London. Imaging procedures, including the acquisition of resting-state fMRI data, were conducted at baseline and 4 weeks after each dosing session. Preprocessing was conducted using standard procedures. Further methodological details can be found in Supporting Information Methods and in previous work. The Automated Anatomical Labelling 90 (AAL-90) atlaswas then used to derive regional estimates of blood-oxygenlevel-dependent (BOLD) activity for 28 FST regions-of-interest (Figure). The AAL was used for the following reasons: (i) the AAL has been extensively used in the whole-brain modeling literature); (ii) the relative low number of parcels in the AAL is well suited for the demanding computational approaches implemented in this study; (iii) the AAL atlas provides clear regional parcellations for the frontal cortex, thalamus, and basal ganglia. These standardized anatomical labels were used to select the FST regions-of-interest for the extraction of time series data and subsequent analyses. Intrinsic functional connectivity (iFC) matrices were derived for each participant by correlating regional FST activity time series using Pearson's correlation (Figure). DFH was estimated at the whole-brain level as well as only for FST regions by computing the cosine similarity in regional activity levels across every timepoint, separately for each subject at each scan assessment (Figuresand). DFH was chosen as a measure of interest, since it provides a simple metric of brain dynamics reflecting homogeneity in brain activity across the duration of the entire fMRI scan. Higher values of DFH reflect higher homogeneity of brain activity in time. Mean DFH was derived by averaging the lower triangular elements of the matrix.
| STRUCTURAL CONNECTIVITY
The analyses leveraged a weighted structural connectivity matrix validated and used in previous studies, derived from diffusion tensor imaging data acquired in 16 healthy young adults (5 females, mean ± standard deviation age: 24.7 ± 2.5 years). Methodological details can be found in Supporting Information Methods and in previous work.
| PARTIAL-BRAIN MODEL
Whole-brain Hopf models are used to model the dependency of emerging functional activity dynamics from the underlying structural brain connectivity. In these models, each brain region is represented as a nonlinear oscillator operating near a supercritical Hopf bifurcation, allowing the system to transition between asynchronous and oscillatory activity states. This approach has been widely used to link structural connectivity with collective phase and amplitude dynamics across different brain states. Here, we used the supercritical Hop-bifurcation model for a single uncoupled region of interest j to derive partial-brain models, once for the post-1 mg and post-25 mg condition, which, instead of focusing on the whole brain, consist of coupled dynamical units of FST regions only. The dynamics of a brain region n is given by the following set of coupled dynamical equations: where j (t) is additive Gaussian noise with standard deviation = 0.01. This normal form has a supercritical bifurcation a j = 0 , so that if a n > 0, the system engages in a stable limit cycle with frequency f j = ω j ∕ 2π. On the other hand, when a j < 0, the local dynamics are in a stable fixed point representing a low activity noisy state. Within this model, the intrinsic frequency is estimated from the empirical data as the peak of the power spectrum between 0.04 and 0.08 Hz (Figure). (1) dx j dt = a jx 2 jy 2 j x j -ω j y j + j (t) (2) dy j dt = a jx 2 jy 2 j y j + ω j x j + j (t) FIGURE 1 | Decreased FST dynamic functional homogeneity following 25 mg psilocybin. (A) Within-subject study design. Twenty-five psychedelic-naïve healthy volunteers were dosed with 1 mg psilocybin 1 day after receiving an fMRI and surveys at baseline (B1). Volunteers received a second fMRI and surveys (B2) 4 weeks after this first dosing session, followed by 25 mg psilocybin the day after. A last fMRI assessment and key endpoint surveys (KE) were completed 4 weeks after being dosed with 25 mg psilocybin. (B) FST regions-of-interest (ROIs) from the AAL-90 atlas. (C) Time-series of resting-state BOLD activity extracted from FST regions for an individual participant j at the KE scan. (D) FST activity time-series were used to generate a time-resolved cosine similarity matrix of FST activity for participant j at the KE scan, reflecting FST dynamic functional homogeneity (DFH). Lower values of DFH reflect higher heterogeneity of FST activity across a scanning session. The lower triangular matrix was then averaged to derive a mean estimate of FST DFH for participant j at the KE scan. This procedure was repeated for each participant at each scan assessment. (E) Mean cosine similarity of FST DFH was significantly reduced at KE 4 weeks after dosing with 25 mg psilocybin, when compared to B2 assessed 4 weeks after a control dosing with 1 mg psilocybin, while only trending decreases were found when comparing FST DFH at the KE to B1 assessed before dosing. (F) Reduction in FST DFH at KE compared to B2 showed a trending correlation with increases in mental well-being, assessed through the Warwick-Edinburgh Mental Well-being Scale. The influence of well-being at baseline and change in mean head frame-wise displacement was regressed out before performing the correlation. (A) Adapted with permission from. FST = fronto-striatal-thalamic. To model the partial-brain dynamics, we added an additive coupling term representing the input received in node j from every other node i, which is weighted by the corresponding structural connectivity represented by C ij . The partialbrain dynamics were defined by the following set of coupled equations: where the noise standard deviation = 0.01. The local bifurcation parameters, a j = -0.02, are at the brink of the local bifurcations which is where the best fitting was demonstrated to be achieved in previous work. The variable x j emulates the BOLD signal of each FST region j. The term G denotes the global coupling weight, scaling equally the total input received in each brain area. Two separate partial-brain models were independently estimated for the post-1 mg and post-25 mg conditions for varying values of the global coupling parameter G by fitting simulated and empirical DFH at each step. Therefore, the estimation of the optimal coupling parameter G was independently estimated for the post-1 mg and post-25 mg conditions. To assess the stability and reproducibility of the global coupling parameter estimates across sessions under stochastic model optimization, we ran 30 separate simulations each with 1500 iterations, which allowed us to compare model fits as well as estimates of optimal coupling parameters G across the post-1 mg and post-25 mg conditions. Repeated simulations are essential in generative modeling to assess robustness of inferred parameters under intrinsic noise; however, they do not represent independent biological samples and therefore are not used as inferential units for group-level hypothesis testing. To assess for inter-individual variation in DFH across the post-1 mg and post-25 mg conditions, two separate partialbrain models were independently estimated for the post-1 mg and post-25 mg by running 1000 permutations resampling DFH within participants across both conditions, with each simulation involving 1500 iterations. Differences in G across permutations were estimated and used as a null distribution to test the significance of G differences derived from modeling DFH in the post-1 mg and post-25 mg conditions. This approach allowed us to assess the influence of individual variability on model fits as well as on estimates of optimal coupling parameters G across the post-1 mg and post-25 mg conditions. Finally, two additional separate partial-brain models were independently estimated for the post-1 mg and post-25 mg conditions for varying values of the global coupling parameter G by fitting simulated and empirical iFC at each step of the fitting procedure. This analysis involved 30 separate simulations each with 1500 iterations.
| EFFECTIVE CONNECTIVITY
To derive effective connectivity, we capitalized on a recently developed approach that computes the time-shifted correlations between: (i) forward time series of two brain regions and (ii) reversed time series of two brain regions. This framework was integrated within two separate whole-brain models, one for post-1 mg and another one for post-25 mg psilocybin data. In both models, we optimized effective connectivity, reflecting local constraints of structural connectivity on information flow between pairs of regions). To assess the stability of the effective connectivity estimation procedure under stochastic initialization and model noise, each model was repeated across 30 stochastic simulations each with 1500 iterations. The statistics applied to these repeated runs were used to evaluate robustness of the estimated effective connectivity patterns rather than to provide independent subject-level observations. The entry point for these analyses was the structural connectivity matrix obtained with probabilistic tractography from diffusion tensor imaging data in healthy volunteers. The following procedure was used to update effective connectivity: where EC ij denotes the effective connectivity between pairs of regions, iFC ij is based on the non-shifted iFC matrices defined through the mutual information measure obtained by: We used ε = 0.0005 and ε′ = 0.0001 and continued the iteration until the algorithm converged. FS ij reflect the time-shifted forward, x i (t), and reversed, x (r) i , functional connectivity matrices derived with a time lag τ of 2 TRs):
| 5-HT2A RECEPTOR AND D2 RECEPTOR AVAILABILITY MAPS
Publicly available PET availability maps of the serotonin 5-HT2A receptor and of the dopaminergic D2 receptor were derived from a previously published study. Briefly, for each radiotracer, group-averaged maps were derived by only using the scans of healthy participants. Normalized PET images were parcellated to the AAL atlas and z-scored receptor availability maps were derived for the 28 FST regions analyzed in this study. Detailed information on each study, the associated receptor/transporter, tracer, number of healthy participants, age, and reference with full (3) ) (5) methodological details has been previously published).
| STATISTICAL ANALYSES
Repeated measures ANOVA were used to compare mean frame-wise head displacement in the scanner and changes in WEMWBS across scanning and dosing sessions. Repeated measures ANOVA and associated paired t-tests were used to assess changes in mean DFH cosine similarity of FST regions across the three time points. Kolmogorov-Smirnov statistics were used to estimate partial-model fits by comparing the distributions of modeled DFH cosine similarity to empirical DFH cosine similarity. Pearson's correlation coefficients were used to estimate partial-model fits by comparing the distributions of modeled iFC to empirical iFC. Partial Pearson's correlation was used to associate changes in WEMWBS scores between the third and second scan to change in mean DFH cosine similarity in the same time interval. This analysis was corrected for baseline well-being, baseline DFH, and head movement in the scanner. The correspondence of modeled and empirical iFC matrices was estimated through Pearson's correlation. Wilcoxon rank sum tests were used to compare fit estimates from the partial-brain model applied to DFH estimates from the second and third scan. Wilcoxon rank sum tests were also used to compare the magnitude of G values yielding the lowest fit estimates across the 1000 within-subject permutations applied to DFH estimates from the second and third scan. To assess the significance of the observed change in structurefunction coupling, we performed a within-subject permutation test. Because this analysis was designed to evaluate the specific decrease identified by the stochastic modeling, significance was assessed using a directional (one-sided) test. Paired t-tests across simulations were used to assess the robustness of model-derived differences in effective connectivity between conditions. Only pairs of nodes showing significant differences across stochastic models were considered when computing outflowing and inflowing effective connectivity, derived by averaging values across rows and columns, respectively. Spearman correlation coefficients were further used to spatially associate regional estimates of the 5-HT2A receptor and of the D2 receptor with nodal estimates of changes in incoming and outcoming effective connectivity following 25 mg psilocybin. Here, separate analyses were performed for cortical and subcortical FST regions due to the marked differences in receptor and transporter densities when comparing subcortical and cortical areas. Control analyses were performed by associating whole-brain changes in effective connectivity to whole-brain receptor maps. Fisher's r-z transformation was used to significantly compare the strength in correlation between subcortical and cortical areas. Repeated measures ANOVA were used to compare mean frame-wise head displacement in the scanner and changes in WEMWBS across scanning and dosing sessions. A p-value < 0.05 was considered statistically significant. False discovery rate (FDR) correction was used to adjust p-values, unless specified otherwise. MATLAB R2021a (. mathw orks. com/ produ cts/ matlab. html) was used for statistical and computational analyses.
| REDUCED STRUCTURAL CONSTRAINTS UNDERLY INCREASED DYNAMIC FST ACTIVITY FOLLOWING 25 MG PSILOCYBIN
We assessed the long-term effects of psilocybin on FST functional dynamics by capitalizing on resting-state fMRI data acquired in psychedelic-naïve healthy adults (Table) at three times: at baseline (B1), 4 weeks after a control dose of 1 mg psilocybin (B2), and 4 weeks after a full active dose of 25 mg psilocybin (key endpoint KE, Figuresand). BOLD activity time-series were extracted from 28 FST regions-of-interest (Figure,C) and used to measure DFH (Figure) for each individual at each scan. DFH was computed by assessing the similarity of regional activity over different timepoints and provides an index of FST activity homogeneity over time. We then investigated whether DFH of FST circuits differed across the three scanning time points (Tableand Figure). Mean FST DFH did not significantly differ when comparing the baseline to the post-1 mg psilocybin scan (t(24) = 0.26, p = 0.79). Mean FST DFH acquired post-25 mg psilocybin was significantly lower than DFH post-1 mg psilocybin (t(24) = 2.87, p = 0.008) and trending lower than DFH at baseline (t(24) = 1.92, p = 0.06). Comparing changes in DFH at the whole-brain level revealed decreases at both the B2 and KE time points when compared to B1, suggesting the presence of order effects related to the baseline assessment when assessing whole-brain changes in DFH (Figure). No significant differences in mean head framewise displacement were detected across sessions (Table). Reductions in FST DFH between the second and third scan correlated with mental well-being increases assessed in the same time interval, that is, 1-month post-25 mg psilocybin (Figure, Rho(22) = -0.42, p = 0.06). Correlations were corrected for baseline mental well-being, baseline FST DFH, and differences in mean head frame-wise displacement between KE and B2. These findings suggest longitudinal increases in dynamic FST activity following psilocybin administration, with such increases being potentially linked to mental well-being improvements after dosing. Finally, we used a computational modeling approach to investigate whether psilocybin mechanistically alters the coupling between structural connectivity and FST dynamic activity. We used a Hopf model; Kringelbach and Deco 2020) (Figure) to generate simulated BOLD activity time series reproducing (Figure) the statistical properties of empirical brain dynamics (Figure). We found a lower optimal G value for FST dynamics post-25 mg (G KE = 0.7) psilocybin when compared to post-1 mg psilocybin (G B2 = 0.9) (Figure). For each of the 30 simulations assessing model stochasticity and separately fitting the post-25 mg and post-1 mg brain dynamics, we identified the G value yielding the lowest fitting, indicating lower coupling parameter estimates for the post-25 mg condition when compared to the post-1 mg condition (z(59) = 6.56, p < 0.0005, mean G KE (SD) = 0.72 (0.06); mean G B2 (SD) = 0.91 (0.06); Figure). This statistical comparison across simulation-derived parameter estimates quantified model-implied differences in the expected optimal coupling parameter under intrinsic dynamical noise. Importantly, these comparisons reflect variability of the generative model rather than inter-individual biological variability. Formal group-level inference was independently evaluated using the permutation-based framework described below. We ran 1000 simulations involving within-subject permutations of DFH across the post-1 mg and post-25 mg conditions. G values yielding the lowest fitting did not significantly differ across permuted post-1 mg and permuted post-25 mg DFH (z(999) = 0.03, p = 0.976, mean G KE (SD) = 0.8191 (0.079); mean G B2 (SD) = 0.8192 (0.078); Figure). The observed difference in G between conditions was smaller than 90% of the differences obtained under permutation (p 1000 = 0.064; Figure). Coupling parameters and fitting metrics did not significantly differ across the KE and B2 conditions when fitting repeated simulation models to regular iFC (Figure), suggesting that fitting DFH is more sensitive to changes in structure-function coupling than static iFC. Overall, these findings suggest a reduced constraining effect of structural connectivity on emerging functional dynamics.
| PSILOCYBIN ALTERS EFFECTIVE CONNECTIVITY BETWEEN FST NODES
While in the previous section we used empirical and computational methods to investigate longitudinal changes in dynamic FST activity following a high dose of psilocybin, we next aimed to investigate longitudinal changes in effective connectivity. We leveraged a whole-brain model that capitalizes on the underlying structural connectivity (Figure) and the inherent time-dependent asymmetry of brain signals to estimate the directionality of information flow between brain regions. We used this approach to estimate effective connectivity matrices separately for the post-1 mg and the post-25 mg psilocybin scans, as well as a subtraction matrix of both post-dosing sessions. The computed EC patterns were consistently reproduced across stochastic model runs (Figure, stability statistics FDR-corrected across stochastic realizations), suggesting stability of the estimated solutions under stochastic perturbation, yet should not be interpreted as individual changes in effective connectivity. Averaging across rows provides an estimate of outflowing effective connectivity (EC-out), or total information flow from the source region to target regions, while averaging across columns provides an estimate of total inflowing effective connectivity (EC-in), or total information flow from source regions to the target region. Outflowing effective connectivity following 25 mg psilocybin was reduced among dorsal cortical regions spanning the dorsolateral prefrontal and medial frontal cortices, while it was increased in limbic and subcortical areas (Figure). Inflowing effective connectivity following 25 mg psilocybin showed a similar pattern, with reductions among the dorsolateral prefrontal and medial frontal cortices, yet, it was increased among limbic and subcortical areas, particularly, on the left hemisphere (Figure). Nodal changes in outflowing effective connectivity following 25 mg psilocybin correlated with changes in inflowing effective connectivity (Figure, Rho(88) = 0.87, p < 0.0005).
| FST EFFECTIVE CONNECTIVITY CHANGES CORRELATE WITH LOCAL 5-HT2A RECEPTOR AND D2 TRANSPORTER AVAILABILITY
The acute effects of psilocybin are primarily mediated through its agonist action on the serotonergic 5-HT2A receptor. We wondered whether there were also long-term changes in causal interactions between brain regions related to the regional availability of 5-HT2A receptors. We derived regional FST maps of 5-HT2A receptor densities by leveraging publicly available maps from standardized radiotracer positron emission tomography (PET) studies). The FST 5-HT2A receptor availability map (Figure) was correlated with nodal maps of outflowing and inflowing effective connectivity change following 25 mg psilocybin administration (Figure). These analyses were conducted separately for cortical and subcortical areas due to these structures showing marked differences in PET signal intensity values and due to opposing changes in inflowing and outflowing effective connectivity between subcortical and cortical areas. The distribution of the 5-HT2A receptor correlated negatively with both cortical inflowing (Rho(18) = -0.56, p < 0.05) and outflowing (Rho(18) = -0.47, p < 0.05) effective connectivity changes. No significant correlations were found between 5-HT2A receptor availability and subcortical inflowing (Rho(6) = 0.33, p = 0.42) or outflowing (Rho(6) = 0.21, p = 0.62) effective connectivity changes, with subcortical correlations being significantly different from cortical ones (Figure). Crucially, whole-brain 5-HT2A receptor availability did not correlate with whole-brain cortical inflowing (Rho(80) = -0.11, p = 0.33) or outflowing (Rho(80) = -0.11, p = 0.33) effective connectivity changes, suggesting, at the group level, a spatial overlap between FST effective connectivity changes following 25 mg psilocybin and 5-HT2A receptor availability. Given that our analyses did not find a subcortical relationship between effective connectivity changes and 5-HT2A receptor availability, we turned our attention to the dopaminergic system, an important modulator of FST circuits. FST D2 receptor availability (Figure) correlated positively with subcortical inflowing (Figure, Rho(8) = 0.88, FDR corrected p < 0.05) and outflowing (Figure, Rho(8) = 0.95, FDR corrected p < 0.005) effective connectivity changes but not with cortical inflowing (Rho(18) = -0.13, p = 0.57) and outflowing (Rho(18) = -0.12, p = 0.61) effective connectivity, with cortical correlations being significantly different from subcortical ones (Figure,F).
| DISCUSSION
In this study, we investigated long-term changes in dynamic FST activity and effective connectivity before and 1 month after psilocybin administration. Longitudinal resting-state fMRI data from a within-subject psilocybin trial in psychedelic naïve healthy controls were examined using both empirical and computational approaches. DFH was used to index functional homogeneity of dynamic activity in FST circuits, assessed across the duration of an entire fMRI session. While higher levels of DFH suggest reduced dynamics in FST activity, lower levels of DFH indicate increased dynamics in FST activity. FST DFH did not differ when comparing baseline to 1 month after a control dose of 1 mg psilocybin. In contrast, FST DFH 4 weeks after 25 mg was significantly lower when compared to post-1 mg psilocybin, although only a trending decrease was found when compared to the baseline. This lack of significant decreases when comparing the key endpoint to the baseline could reflect an inherent limitation of the trial's within-subject design, with several order effects, including scanner habituation, expectancy shifts, or physiological adaptation, acting as confounding factors. Longterm increases in dynamic FST activity were linked to longitudinal improvements in mental well-being following 25 mg psilocybin. Although trending for significance, this association was corrected for baseline well-being, baseline DFH, and longitudinal changes in head movement while scanned. Overall, this finding suggests longitudinal increases in dynamic FST activity a month after a full dose of psilocybin, with these increases potentially accounting for longitudinal improvements in mental well-being. Although caution is granted given the weak statistical significance, our findings are in line with previous reports showing longitudinal anterior cingulate and prefrontal cortex functional reorganization days to weeks after the administration of a psychedelic substance, with functional brain changes correlating with longitudinal improvements in well-being. Which mechanisms may account for such empirical increases in dynamic FST activity? We used a partial-brain model, a variant of the regular whole-brain Hopf model, to characterize how anatomical constraints shape FST dynamics 4 weeks after 25 mg psilocybin when compared to the earlier 1 mg visit. Within this modeling framework, the optimal global coupling parameter (G)-which governs the strength of coupling between structural connectivity and emergent functional dynamics-was consistently lower in the post-25 mg condition compared to the post-1 mg condition. Repeated stochastic simulations indicated that this shift in optimal G is stable under intrinsic model noise, suggesting that the reduction is not driven by a single realization of the dynamical system. However, permutation-based group-level inference revealed a trend-level signal, indicating that this reduction should be interpreted cautiously. Within the context of the model, a lower G implies a reduced constraining influence of anatomical connectivity on functional dynamics. This model-informed decrease in structure-function coupling provides a preliminary mechanistic account for the observed increase in dynamic FST activity, potentially enabling more heterogeneous and flexible activity patterns in FST circuits. Further computational models leveraging the concept of nonreversibility revealed long-term effects of psilocybin on effective connectivity between FST nodes. Here, effective connectivity is referring to the directional flow of information between brain regions, as commonly derived through Granger causality or dynamic causal modeling). Yet, it differs from these methods by accounting for the constraining effects of local structural weights on causal interactions between pairs of nodes. Decreases in both inflowing and outflowing effective connectivity following a high dose of psilocybin were found among higher-order brain regions exerting top-down information flow overlapping with the dorsolateral prefrontal, frontal, and insular cortices. Conversely, lower-order subcortical and limbic regions-such as the thalamus, putamen, and anterior cingulate-showed marked increases in both inflowing and outflowing effective connectivity. These findings possibly suggest increased information exchange between lower-order brain regions spanning subcortical and primary cortical brain areas. Average decreases in top-down effective connectivity spatially mapped on the cortical distribution of the 5-HT2A receptor, while average increases in bottom-up information outflow spatially mapped on the cortical distribution of the D2 receptor. How are these findings in keeping within major theoretical models of psilocybin's action? One model posits that psilocybin and related psychedelics acutely disrupt cortico-striatal-thalamic circuits through 5-HT2A activation in the prefrontal cortex, which in turn disrupt GABAergic pallido-thalamic neurons, leading to thalamic disinhibition and inundation of the cortex with sensory information. Evidence for this model has been provided by directional modeling of resting-state fMRI data, showing that lysergic acid diethylamide, a classic psychedelic akin to psilocybin, acutely decreases striatal input to the thalamus. Our findings suggest longitudinal subcortical increases in inflowing and outflowing information, which are in contrast with a thalamic gating model proposed to underlie acute effects. Another model, known as "Relaxed Beliefs Under Psychedelics" (REBUS), proposes that psychedelics disrupt the hierarchical organization of brain function by dysregulating statistical regularities in the spontaneous activity of higher-level cortical areas-causing less efficient top-down control of activity in functionally lower-level regions (Carhart-Harris and Friston 2019). In Bayesian terms, the REBUS model states that psychedelics cause a relaxation of precision weightings encoding prior beliefs, resulting in less confident assumptions. Relaxing the precision-weighting of priors encoded in higher-level, cortical areas, would in turn catalyze an increase in information outflow from lower-level regions. Overall, our findings support this model, extending it beyond the originally proposed acute action to a more enduring subacute action (i.e., 1 month post-25 mg psilocybin). Our findings also show that long-term increases in subcortical information outflow and inflow correlate with the regional expression of the dopamine transporter. While psilocybin does not exert direct agonist effects at dopamine receptors and has limited known dopamine-releasing properties, these findings may nonetheless be consistent with prior evidence suggesting indirect dopaminergic involvement following psychedelic administration. This finding could be viewed as consistent with recent preclinical work showing that psychedelics reopen a critical period for learning social reward, as well as with the well-established principle that dopamine neurotransmission in midbrainstriatum-orbitofrontal cortex circuits is especially involved in reinforcement learning) and possibly, belief updating. Importantly, given the low addiction potential of psilocybin and related psychedelics, and that diminished striatal connectivity has been associated with vulnerability to drug addiction) and alcohol dependence, this dopaminergic involvement is unlikely to reflect a drug-dependency risk of psilocybin. The following limitations need to be considered when interpreting the present findings. First, our empirical analyses revealed weak to moderate statistical evidence for longitudinal changes in dynamic FST activity and its association with well-being improvements. This, together with the finding that permutation testing of the modeled global coupling parameter yielded only a trend-level signal, may reflect limited statistical power given the sample size rather than definitive evidence against a group difference. In addition, the statistics of the stochastic EC simulations were primarily intended to assess robustness and stability of the inferred solutions under stochastic perturbation, rather than to provide independent biological observations for subjectlevel inference. Second, FST regions exhibit high functionalanatomic heterogeneity and future studies would benefit from the use of more fine-grained cortical and subcortical atlases as well as from the use of individualized FST parcellations and structural connectivity estimates, in particular given recent evidence from animal and human studies showing that psilocybin has a long-term effect, not only on brain function, but also on brain structure. The main objective of this study was to characterize changes in functional dynamics induced by psilocybin, not to test hypotheses about individual differences in white-matter architecture or how individual changes following psilocybin administration may impact structure-function models. Yet, the use of group-averaged structural connectomes as well as group-level receptor availability maps limits the generalizability of the findings, since normative cortical receptor maps may act as surrogates for cortical thickness or general brain health. Future work would likely benefit from more personalized approaches as recently implemented in psychedelic neuroimaging studies. Third, recent work suggests long-lasting functional connectivity disruptions between the anterior hippocampus and medial parietal areas, a circuit not investigated in the current work. Future computational work should explore the potential mechanistic underpinnings underlying these long-term functional changes induced by psilocybin. Finally, the current work does not investigate the effects of other receptor systems, including glutamatergic or GABAergic receptors, on effective connectivity changes. Nonetheless, our findings open new avenues for the application of mechanistic approaches to elucidate the long-term impact of pharmacological interventions on the brain. Specifically, our data suggest that psilocybin increases dynamic FST activity and subcortical information outflow through cortical serotonergic and subcortical dopaminergic systems. This FST functional reorganization, in turn, may underly cross-diagnostic longitudinal behavioral changes, including increased openness and well-being, and decreased anhedonia, apathy, and substance craving after psychedelics). Further research is necessary to determine whether this long-term functional reorganization of FST circuits is responsible for the clinical efficacy of psilocybin across neuropsychiatric conditions.
Full Text PDF
Full Paper PDF
Create a free account to open full-text PDFs.
Study Details
- Study Typeindividual
- Populationhumanssimulation
- Characteristicsbrain measures
- Journal
- Compound
- Topics
- APA Citation
References (28)
Papers cited by this study that are also in Blossom
Avram, M., Müller, F., Rogg, H. et al. · Biological Psychiatry (2022)
Bogenschutz, M. P., Ross, S., Bhatt, S. R. et al. · JAMA Psychiatry (2022)
Ly, C., Greb, A. C., Cameron, L. P. et al. · Cell Reports (2018)
Carhart-Harris, R. L., Giribaldi, B., Watts, R. et al. · New England Journal of Medicine (2021)
Carhart-Harris, R. L., Bolstridge, M., Rucker, J. et al. · Lancet Psychiatry (2016)
Carhart-Harris, R. L., Erritzoe, D., Williams, T. et al. · PNAS (2012)
Carhart-Harris, R. L., Friston, K. J. · Pharmacological Reviews (2019)
Davis, A. K., Barrett, F. S., May, D. G. et al. · JAMA Psychiatry (2021)
Daws, R. E., Timmermann, C., Giribaldi, B. et al. · Nature Medicine (2022)
De La, M., Revenga, F., Zhu, B. et al. · Cell Reports (2021)
Show all 28 referencesShow fewer
Deco, G., Cruzat, J., Cabral, J. et al. · Current Biology (2018)
Doss, M. K., Madden, M. B., Gaddis, A. et al. · Brain (2021)
Doss, M. K., Považan, M., Rosenberg, M. D. et al. · Translational Psychiatry (2021)
Goodwin, G. M., Aaronson, S. T., Alvarez, O. et al. · New England Journal of Medicine (2022)
Johnson, M. W., Garcia-Romeu, A., Griffiths, R. R. · The American Journal of Drug and Alcohol Abuse (2016)
Kringelbach, M. L., Cruzat, J., Cabral, J. et al. · PNAS (2020)
Lyons, T., Spriggs, M. J., Kerkelä, L. et al. · Nature Communications (2026)
Madsen, M. K., Stenbaek, D. S., Arvidsson, A. et al. · European Neuropsychopharmacology (2021)
McCulloch, D. E-W., Madsen, M. K., Stenbæk, D. S. et al. · Journal of Psychopharmacology (2021)
Nardou, R., Sawyer, E., Song, Y. J. et al. · Nature (2023)
Nichols, D. E. · Pharmacological Reviews (2016)
Nour, M. R., Evans, J., Carhart-Harris, R. L. · Journal of Psychoactive Drugs (2017)
Pasquini, L., Palhano-Fontes, F., Araújo, D. B. · Journal of Psychopharmacology (2020)
Preller, K. H., Razi, A., Zeidman, P. et al. · PNAS (2019)
Nicol, G. E. · Nature (2024)
Stoliker, D., Novelli, L., Vollenweider, F. X. et al. · Biological Psychiatry (2023)
Vargas, M. V., Dunlap, L. E., Dong, C. et al. · Science (2023)
Vollenweider, F. X., Preller, K. H. · Nature Reviews Neuroscience (2020)