Psilocybin induces spatially constrained alterations in thalamic functional organization and connectivity

Classic serotonergic psychedelics such as psilocybin produce profound alterations in perception and cognition that are thought to be mediated largely via serotonin 2A receptors (5-HT2AR). While most human neuroimaging studies to date have emphasised cortical 5-HT2AR-rich regions, the thalamus also expresses 5-HT2AR and is a central hub in cortico‑striato‑thalamo‑cortical circuits implicated in both psychosis and altered perceptual states. Previous functional MRI studies of psychedelics have reported mixed findings regarding thalamic function and thalamocortical connectivity, and those studies typically treated the thalamus as a single unit despite well‑known nuclei‑specific structure and function. Gaddis and colleagues set out to examine whether acute psilocybin alters the internal functional organisation of the thalamus and nucleus‑specific thalamocortical connectivity. To do so they applied a data‑sparing, template Independent Component Analysis (tICA) approach to resting‑state fMRI that leverages population priors to generate reliable, subject‑level intrathalamic components, then compared intrathalamic spatial engagement and thalamocortical connectivity between placebo and an acute 10 mg/70 kg psilocybin condition. The study also tested associations between these neural changes and concurrent subjective effects reported by participants.

Participants were recruited from a larger multi‑phase study of psilocybin in long‑term meditation practitioners. Inclusion required medical and psychiatric health and substantial meditation experience (mean lifetime meditation ≈ 4,883 hours). From an initial Phase 1 cohort of 40, resting‑state baseline data from 38 participants were used to build tICA templates. Phase 2 comprised a single‑blind, within‑subject, placebo‑controlled comparison of placebo versus 10 mg/70 kg oral psilocybin; usable paired resting‑state scans were available for 18 participants (13 male, 5 female; mean age 54.4, SD 13.2). Participants had previously received 25 mg/70 kg psilocybin in Phase 1 several months earlier. Drug administration in Phase 2 was single‑blind with the first capsule always placebo and the second always 10 mg psilocybin, administered in a controlled clinical setting. Resting‑state fMRI scans were collected approximately 100 minutes after capsule administration (scan time 7 min 42 s; 3 mm isotropic voxels; TR = 2.2 s) on 3T scanners. Participants rated subjective effects immediately after each scan using brief items including overall psilocybin effect, facets of mindfulness (nowness, letting go, equanimity), selected items from the Mystical Experience Questionnaire aggregated into a brief MEQ (bMEQ), and positive/negative valence. Template ICA generation used the 38 baseline (Phase 1) scans masked to the thalamus (Harvard‑Oxford atlas thresholded ≥50%; 2,268 voxels). Group spatially constrained ICA (model order = 7) produced seven thalamic independent components (TIC01–TIC07). Reliability was assessed by repeating group ICA 100 times. Subject‑level variance estimates were obtained by splitting each subject's time series into two pseudo‑sessions and applying dual regression; between‑subject variance and mean component loadings formed the empirical priors for tICA. The tICA model was then fit to the 18 subjects' placebo and psilocybin scans to obtain posterior component spatial maps and standard errors. Intrathalamic "engagement" was defined voxelwise as component loadings that significantly differed from the template mean based on posterior estimates; tests were FDR‑corrected at 0.01 for voxelwise tests and at 0.05 for comparisons of engaged voxel counts. Two engagement metrics were used: counts of significantly engaged voxels across the whole thalamus, and counts restricted to voxels with robust baseline contribution to a component (z ≥ 2). Probabilistic maps were generated to identify voxels in which at least 25% of subjects showed greater engagement in placebo versus psilocybin. Spatial localisation used overlap with Morel atlas nuclei (percent overlap). For connectivity analyses, dual regression produced TIC timecourses and cortical ICA‑derived network (CIN) timecourses; 7×10 partial correlation matrices (ridge regression rho = 1) covering intra‑thalamic, thalamocortical and intra‑cortical edges were Fisher z‑transformed and compared between sessions with paired t‑tests (FDR = 0.05). Associations between neural changes and subjective effects used Spearman correlations, with an a priori threshold of R2 ≥ 0.2 (about r ≥ 0.45) reported as meaningful. For comparison with prior literature, a whole‑thalamus seed timecourse (mean across all thalamic voxels) was also correlated with cortical networks and tested for between‑session differences.

Baseline parcellation: Group ICA of baseline resting data produced seven spatially clustered and bilaterally symmetric thalamic independent components (TIC01–TIC07). Voxels with robust baseline loadings (z ≥ 2) showed complete segregation across components. Each TIC demonstrated a distinct pattern of overlap with Morel atlas nuclei, with best fits reported as TIC1: LD/LP, TIC2: VP, TIC3: PumM, TIC4: VA, TIC5: MD, TIC6: PuA and TIC7: VL. Intrathalamic engagement: Compared with placebo, psilocybin sessions showed significant reductions in engaged voxel counts for TIC03 (p = 0.006) and TIC05 (p = 0.01) after FDR correction (FDR = 0.05). When restricting to voxels with robust baseline loading (z ≥ 2), decreases during psilocybin were also significant for TIC02 and TIC07 in addition to TIC03 and TIC05. Probabilistic effect maps localized these between‑session decreases primarily to the mediodorsal (MD) and medial pulvinar nuclei. Associations with subjective effects: Six associations met the pre‑specified large‑effect threshold (R2 ≥ 0.2). Greater decreases in TIC03 engagement on psilocybin correlated with larger increases in "letting go" (r = -0.47). Increased engagement of TIC07 during psilocybin correlated with higher ratings of equanimity (r = 0.55), pure being (r = 0.52), joy (r = 0.52) and the aggregated bMEQ mystical experience score (r = 0.54). Decreases in TIC06 engagement were associated with larger increases in joy (r = -0.48). Functional connectivity: Across intrathalamic, thalamocortical and intra‑cortical edges, psilocybin produced more decreases than increases: 13 edges decreased and 5 increased significantly. Intrathalamically, two edges decreased and one increased. Among thalamocortical edges, five showed decreased connectivity and one showed increased connectivity on psilocybin. Intra‑cortical networks had six decreased and three increased edges. Specific between‑network associations with subjective effects included: intra‑thalamic decreases between TIC02–TIC03 correlated with reductions in "fusion" (r = 0.47), "sacredness" (r = 0.53) and "peace" (r = 0.49); decreases between TIC05–TIC07 correlated with decreases in "timelessness" (r = 0.54) and "joy" (r = 0.48). For thalamocortical pairs, decreases in TIC02–IC08 related to positive‑valence changes (r = -0.57), decreases in TIC03–IC09 related to overall effect (r = -0.47), and decreases in TIC03–IC17 associated with reductions in negative valence (r = 0.45) and increases in joy (r = -0.55). A decrease in TIC07–IC11 connectivity related to reduced letting go (r = 0.50) and greater negative valence (r = -0.50). Intra‑cortical changes also showed sizeable associations, for example increased IC08–IC17 connectivity correlated with reduced timelessness (r = -0.61). Whole‑thalamus seed results: Seed‑based whole‑thalamus analyses detected no significant between‑session thalamocortical edges after FDR correction. The largest, non‑significant change was greater connectivity with the occipital‑pole (IC09) during psilocybin (t = -2.68; FDR‑corrected p = 0.07); DMN (IC11) and left FPN (IC17) showed marginal non‑significant increases (FDR p ≈ 0.08). Notably, directionality from the seed‑based approach sometimes opposed ICA‑based findings: regions localized to specific TICs (for example TIC07 with DMN, TIC02/TIC03 with occipital‑pole) showed decreased connectivity during psilocybin, whereas other thalamic voxels showed opposite effects. The authors interpret this as focal decreases being masked when averaging across the whole thalamus.

Gaddis and colleagues interpret their findings to indicate that acute psilocybin induces spatially constrained alterations in thalamic functional organisation and in thalamocortical connectivity that are not apparent when the thalamus is treated as a single unit. The primary effect observed with their tICA approach was focal reductions in component engagement and in connectivity centred on intrathalamic regions mapping preferentially to the mediodorsal and pulvinar nuclei. Several of those nucleus‑specific changes correlated with concurrent subjective experiences that have been linked to psilocybin's therapeutic relevance. These results are positioned relative to prior literature showing mixed thalamic effects with classic psychedelics: whereas earlier whole‑thalamus studies often reported overall increases in thalamic activation or thalamocortical connectivity, the present component‑level analysis reveals that modest widespread increases may coexist with strong, anatomically focal decreases in nuclei that express 5‑HT2AR. The mediodorsal and pulvinar localisation is consistent with receptor expression data and with animal findings implicating the mediodorsal nucleus in psychedelic effects. The authors therefore argue that ICA‑based parcellation offers a more nuanced evaluation of thalamic involvement in the psychedelic state than whole‑thalamus masking. The investigators note methodological strengths of their data‑sparing tICA pipeline, which enabled reliable intrathalamic parcellation from standard‑length 3T resting‑state scans. They also acknowledge that applying the approach to higher‑resolution imaging, other modalities, or other altered states would be informative future work. The Discussion emphasises that whole‑thalamus averaging may obscure nuclei‑specific phenomena and that component‑level analyses can reveal relations between specific thalamic subdivisions, cortical networks (notably visual and default mode networks) and subjective experience. The authors frame these findings as a first report on intrathalamic functional organisation during acute psilocybin in humans and suggest that the observed nucleus‑specific changes merit further study in the context of therapeutic mechanisms and broader altered‑state neuroscience.