Cortical structural differences following repeated ayahuasca use hold molecular signatures

Classical serotonergic psychedelics including psilocybin, LSD and DMT are thought to promote both acute subjective effects and longer-term neuroplastic adaptations via partial agonism at the 5-HT2A receptor. Preclinical work has shown 5-HT2A agonists can trigger immediate early genes (IEGs), and later synaptogenesis, dendritogenesis and other markers of structural plasticity; human neuroimaging studies have likewise reported persistent changes to large-scale functional networks after single doses. Less is known, however, about how repeated psychedelic exposure affects brain anatomy at a macroscale level, or how any structural differences map onto molecular signatures associated with psychedelic pharmacology. Mallaroni and colleagues set out to examine whether sustained ayahuasca use is associated with altered whole-cortex anatomical organisation and whether such alterations spatially co-locate with cortical gene expression relevant to ayahuasca's pharmacology and plasticity-related pathways. Using high-field 7T MRI from 24 long-term Santo Daime members and 24 matched controls, the study employed morphometric similarity network (MSN) analysis to characterise cortical structural network topography and related regional MSN differences to transcriptional maps from the Allen Human Brain Atlas, testing hypotheses about group differences in MSNs, their clustering within functionally relevant networks, and co-localisation with 5-HT2A expression and other candidate genes.

This was a cross-sectional, fixed-order observational study comparing 24 experienced Santo Daime members (10 female, mean age 55.2, membership mean 14.2 years, mean ceremony attendance 563 times) with 24 healthy controls matched on age and sex drawn from an external 7T MRI repository. Inclusion criteria for Santo Daime participants included absence of MRI contraindications and no use of medicinal substances in the previous 24 hours; controls had no history of neurological, psychiatric or somatic disease. Ethical approval and written consent were obtained; the extracted text does not report additional behavioural matching variables (for example IQ) beyond age and sex. Structural MRI data were acquired at 7 Tesla using MP2RAGE sequences with slightly different acquisition parameters between groups (voxel size 0.9 mm isotropic for Santo Daime group, 0.7 mm isotropic for controls). Images underwent intensity non-uniformity correction, skull-stripping, tissue segmentation and surface reconstruction using sMRIPrep and FreeSurfer pipelines; multiple quality metrics (CJV, CNR, FWHM) were inspected. Cortical surfaces were parcellated into 308 approximately equal-sized nodes derived from a subparcellation of the Desikan–Killiany atlas. For each node seven T1-weighted morphometric features were extracted: cortical thickness, surface area, mean curvature, Gaussian curvature, folding index, curvature index and grey matter volume. Features were z-scaled across regions per subject and pairwise Pearson correlations between regional morphometric feature vectors produced a 308 × 308 MSN per participant. Regional morphometric similarity (MS) was computed as the sum of weighted correlations between a region and all others. Case-control differences were assessed by fitting linear regression models to regional MS values while regressing out age, sex and an age × sex interaction; two-sided t-statistics (ayahuasca minus healthy controls) were extracted and corrected across 308 regions using Benjamini–Hochberg FDR (p < 0.05). MSN results were contextualised against the Yeo 7 functional networks and von Economo cytoarchitectonic classes. Sensitivity analyses included constructing MSNs using Spearman correlations and adding total intracranial volume (TIV) as a nuisance regressor, with Jaccard coefficients used to compare main and replication results. Regional cortical thickness (CT) was also analysed with comparable regressors to compare with prior findings. To relate structural differences to cortical transcriptomics, the investigators used left-hemisphere expression maps from the Allen Human Brain Atlas mapped to the DK-308 parcellation (analysis restricted to 152 left-hemisphere regions because only two AHBA donors sampled the right hemisphere). An a priori list of 66 gene targets was assembled covering receptors, channels and transporters pertinent to ayahuasca's binding profile and candidate neuroplasticity genes reported after acute 5-HT2A agonist administration. Partial least squares (PLS) regression was used to relate regional case-control MSN t-values to the 66-gene expression matrix; significance of the first PLS component (PLS1) was assessed via 10,000 permutations and spin-permutation testing to account for spatial autocorrelation. Contribution of individual genes to PLS1 was evaluated using 10,000 bootstrap resamples, with a conservative 99% confidence threshold for selecting positively or negatively contributing genes.

After image quality control, 24 Santo Daime participants and 24 matched controls were included. There were no significant between-group differences in image quality, age or sex. Overall, the Santo Daime group showed lower mean MS values compared to controls (t = 4.58, p < 0.0001), interpreted as a net increase in anatomical differentiation across the cortex. MSN distributions were comparable to prior multimodal healthy reference maps; construction using Spearman correlations produced highly concordant results (r (306) = 0.97, p spin < 0.0001). Regional analyses revealed a spatially distributed pattern: decreased morphometric similarity (architectonic differentiation) in sensorimotor cortices including inferior frontal gyrus, precuneus and pre/postcentral gyri, and increased morphometric similarity (architectonic de-differentiation) in midline, temporal and prefrontal structures such as orbitofrontal, entorhinal, cingulate and anterior insular cortices. The case-control t-map correlated strongly and negatively with the mean control regional MS (Pearson's r (306) = -0.84, p spin < 0.0001), indicating that regions normally high in MS tended to show greater reductions. Quantitatively, 29.55% of regions exhibited positive t-values with negative mean MS (de-differentiation) while 56.17% showed negative t-values with positive mean MS (differentiation). At the level of functional and cytoarchitectonic systems, ayahuasca users showed decreased MS in the Yeo somatomotor (SM), dorsal attention (DA) and default mode (DMN) networks (pFDR = 0.0165–0.0006) and increased MS in limbic networks (p < 0.0001). Von Economo classes showed decreased MS in granular association isocortical types 1 and 2 (pFDR = 0.0010, 0.0003) and increased MS in limbic and insular classes (pFDR = 0.0002, < 0.0001). Whole-brain modularity was unchanged, although functional community affiliations shifted across modules. A trend-level negative association between ceremony attendance frequency and MS was observed (maximum Spearman's rho (46) = -0.36, p = 0.0865). Replication analyses that added TIV as a covariate showed high concordance with the main results (Jaccard = 90%, t-map p spin < 0.0001, r = 0.997). Cortical thickness analyses contrasted with some prior reports: Santo Daime members showed cortical thickening in midline and superior frontal regions (including posterior cingulate cortex and medial frontal cortex) and sparse thinning in parietal/occipital regions; CT t-maps correlated with MSN t-maps (r (306) = 0.39, p spin < 0.0001). In the transcriptomic analysis, PLS1 explained 11% of the variance in the case-control MSN t-map (permutation p = 0.0181) and correlated positively with MSN t-values (Pearson's r (150) = 0.33, p spin = 0.0004). Regions with positive PLS1 gene weights were overexpressed in areas showing increased MS in ayahuasca users; negative weights aligned with regions of decreased MS. Of the 66 candidate genes, 18 made significant contributions to PLS1 (p < 0.01), with 11 positively weighted and 7 negatively weighted. 5-HT2A gene expression was among contributors, with factor loadings indicating relatively lower 5-HT2A expression in sensorimotor cortices that showed greater morphometric differentiation. The analysis also implicated a broader set of receptors and plasticity-related genes, including serotonergic, dopaminergic and endocannabinoid-related targets as well as transcription factors and immediate early genes.

Mallaroni and colleagues interpret their findings as early evidence that sustained ritualistic ayahuasca use is associated with a reorganisation of cortical structural network topography, characterised by architectonic differentiation in sensory and somatomotor regions and de-differentiation in midline, limbic and transmodal prefrontal nodes. They note that these macroscale patterns are partly consistent with hypotheses about 5-HT2A-mediated plasticity but extend beyond a single receptor: morphometric changes were more strongly associated with a combination of receptor genes implicated in ayahuasca's polypharmacology and with transcriptional regulators and immediate early genes previously linked to psychedelic-induced neuroplasticity in preclinical assays. The authors situate the structural results in relation to functional and behavioural literature. Decreased MS in interoceptive and somatosensory regions may reflect long-term compensatory adaptations in areas linked to self-referential and embodied processing, while de-differentiation of transmodal hubs could underpin previously reported functional tolerance and preserved or enhanced executive performance in experienced users. At a molecular level, they highlight evidence consistent with downregulation/desensitisation of 5-HT2A receptors after repeated agonist exposure and stress the likely contribution of downstream pathways (AMPA/TrkB/mTOR) and other receptor systems (endocannabinoid, adrenergic, 5-HT1A/2C) to the observed structural signatures. Key limitations are acknowledged: the cross-sectional design precludes causal inference and lifestyle or community factors (for example strong social bonding within Santo Daime) may contribute to differences; external controls were matched only on age and sex, limiting behavioural covariate matching; small sample size reduces reliability and generalisability; differences in acquisition protocols between cohorts may have influenced morphometry despite replication checks; and cortical gene expression mapping relies on the AHBA which is spatially sparse (six donors) and may covary with demographic variables. The authors call for larger, multi-centre, and longitudinal studies—ideally combining diverse syncretic groups and baseline measures—to better characterise neuroadaptive changes and to integrate structural-functional coupling in biophysically informed models.

The study provides initial evidence that repeated ayahuasca use is associated with alterations in cortical anatomical organisation that co-localise with molecular signatures relevant to psychedelic pharmacology and neuroplasticity. Using a pharmacologically informed transcriptomic approach, the findings suggest that molecular mechanisms implicated in psychedelic action may scale up to macroscale structural differences in vivo. The authors recommend larger, culturally sensitive consortium efforts and longitudinal designs to further investigate neuroadaptive effects of repeated psychedelic exposure.