Effective-connectivity of thalamocortical interactions following d-amphetamine, LSD, and MDMA administration

The renewed clinical interest in serotonergic psychedelics is driven by promising therapeutic signals for disorders such as depression, anxiety, and addiction, yet the neurobiological mechanisms mediating these effects remain incompletely understood. One influential account, the disrupted thalamic filter model (TFM), proposes that changes in neurotransmission (for example increases in dopaminergic tone) can impair thalamic filtering and cause a ‘‘cortical flooding’’ that contributes to altered conscious states. Prior human neuroimaging studies using resting-state functional connectivity (iFC) have reported substance-induced increases in thalamocortical iFC, often localised to thalamic nuclei rich in 5-HT2A receptors, but iFC does not indicate directionality of influence. To address whether thalamocortical dysconnectivity is driven primarily by the thalamus or cortex and whether effects differ by cortical type, Avram and colleagues used spectral dynamic causal modelling (DCM) to estimate directed (effective) connectivity between the thalamus and two cortical sets: unimodal (sensorimotor and visual) and SAL-derived transmodal regions. Using a double-blind, placebo-controlled, crossover design with LSD, MDMA, d-amphetamine, and placebo in the same participants, the study tested whether these substances increase thalamus-to-cortex (bottom-up) effective connectivity for unimodal regions, and whether LSD in particular increases thalamus-to-transmodal connectivity while the amphetamines produce distinct effects, consistent with prior iFC results.

The study used a double-blind, placebo-controlled, within-subject crossover design in 25 healthy volunteers (12 female; mean age 28.2 ± 4.35 years). Each participant completed four sessions in a randomised, counterbalanced order receiving 0.1 mg LSD, 125 mg MDMA, 40 mg d-amphetamine, and placebo (ethanol/mannitol). Structural and resting-state functional MRI data were collected on a 3T Siemens Magnetom Prisma; imaging parameters and further participant details are reported in the supplement (not reproduced here). Regions of interest (ROIs) were selected from prior thalamic seed-based iFC analyses to capture cortical areas implicated across substances. Two separate fully connected DCM models were specified: one with unimodal cortical ROIs (auditory cortex, postcentral gyrus, lingual gyrus, cuneus) and the thalamus, and a second covering SAL-derived transmodal regions (including insula, ACC, supramarginal gyri) with thalamic ROIs defined per substance from SAL-thalamic iFC peaks. ROIs were masked with 8 mm spheres and each ROI time series was extracted as the first principal component. Time series were preprocessed for motion (six parameters), physiological noise (CSF and white matter signals from small spheres), and filtered with a 128-s high-pass filter. Spectral DCM (as implemented in SPM12) was used to invert fully connected models for each subject and each condition, fitting cross-spectral densities using a power-law model of endogenous fluctuations. Model fits explained a high proportion of variance across conditions (approx. 89%–91% depending on substance and model), with no subject showing less than 75% explained variance. Between-subject effects were analysed with a parametric empirical Bayes (PEB) hierarchical Bayesian GLM and Bayesian model reduction to identify plausible reduced models. The authors report both strong effects (posterior probability >0.99 and effect size >0.1 Hz) and weaker effects (posterior probability >0.05 and effect size <0.1 Hz). Self-connections (log-scaled) were reported as indicators of inhibitory decay; more negative values indicate reduced self-inhibition (disinhibition). To explore behavioural relevance, multiple linear regressions tested whether VAS items (including ‘‘any drug effect,’’ ‘‘good drug effect,’’ ‘‘drug liking,’’ ‘‘drug high,’’ ‘‘stimulated,’’ ‘‘ego dissolution,’’ ‘‘speed of thinking,’’ ‘‘concentration,’’ etc.) were predicted by corticothalamic (all cortical→thalamus) or thalamocortical (thalamus→cortex) effective connectivity parameters that showed strong effects. These exploratory regressions were not corrected for multiple comparisons.

Across both DCM models, all three active substances produced directed connectivity changes involving the thalamus and cortical ROIs, with patterns that depended on cortex type. Unimodal cortices: Compared to placebo, d-amphetamine and MDMA reduced effective connectivity from several unimodal cortical regions (left auditory cortex, right postcentral cortex, left cuneus; with some weaker substance-specific differences) to the thalamus, while the thalamus increased its effective influence on the left auditory cortex and right lingual gyrus. Both amphetamines increased self-inhibition in the right lingual gyrus and decreased self-inhibition in the right postcentral gyrus (weaker for MDMA). D-amphetamine specifically increased thalamic self-inhibition (weaker effect) and reduced connectivity from right postcentral and right lingual to left auditory cortex. MDMA additionally increased thalamus→right cuneus and thalamus→right postcentral (weaker effects) and decreased left cuneus→left auditory cortex. LSD also decreased effective connectivity from left auditory and right postcentral cortices to the thalamus, increased effective connectivity from right lingual gyrus to thalamus, and increased thalamic self-inhibition; LSD produced weaker increases in thalamus→left auditory cortex and thalamus→right lingual gyrus and reduced self-inhibition in right postcentral cortex. Overall, the unimodal pattern across substances was characterised by decreased corticothalamic (top-down) and increased thalamocortical (bottom-up) effective connectivity for specific unimodal regions. Transmodal (SAL-derived) cortices: D-amphetamine and MDMA increased effective connectivity from bilateral insulae and left supramarginal gyrus to the thalamus (weaker effects for MDMA), while the thalamus decreased its influence on left and right insula (weaker for right insula). Both amphetamines increased insula→ACC and insula→right supramarginal connectivity (weaker effects) and left supramarginal→ACC (weaker). Thalamic self-inhibition was increased following the amphetamines (weaker for MDMA). Substance-specific findings included d-amphetamine’s reduction of thalamus→left supramarginal and increased left insula self-inhibition (weaker), and MDMA-related decreases in ACC→insula effective connectivity. LSD produced increased thalamus→ACC and thalamus→supramarginal gyrus connectivity (weaker for left supramarginal), increased thalamic self-inhibition, and a weak decrease in right insula self-inhibition; no corticothalamic increases were observed for LSD. Thus, amphetamines tended to increase top-down cortical influence on the thalamus while reducing bottom-up thalamic influence to SAL regions, whereas LSD increased thalamus→transmodal connectivity. Associations with subjective effects: Multiple linear regressions linked some effective-connectivity changes to VAS items. For unimodal regions, d-amphetamine corticothalamic changes predicted ‘‘speed of thinking’’ (F3,21=4.57, p=.01), and d-amphetamine thalamocortical changes predicted ‘‘good drug effect’’ (F2,22=3.96, p=.03) and ‘‘drug liking’’ (F2,22=5.54, p=.01). LSD corticothalamic changes predicted ‘‘concentration’’ (F3,21=4.48, p=.01). MDMA-induced effective-connectivity changes did not predict VAS items. Effective-connectivity changes involving transmodal regions did not significantly predict any VAS items. Pearson correlations across all VAS items and coupling parameters were reported in supplementary material (not reproduced here).

Using spectral DCM, Avram and colleagues report that LSD, d-amphetamine, and MDMA all altered directed thalamocortical and corticothalamic interactions, but the patterns depended on cortical type. For unimodal cortices (auditory and visual/sensorimotor areas), all three substances tended to increase thalamus→cortex (bottom-up) effective connectivity and reduce cortex→thalamus (top-down) connectivity, consistent with the idea of reduced top-down control and enhanced bottom-up signal flow posited by the disrupted thalamic filter model. The authors note that this ‘‘cortical flooding’’ effect is not unique to classical psychedelics, since structurally related amphetamines with different subjective profiles produced similar unimodal thalamocortical changes. By contrast, interactions with SAL-derived transmodal cortices differed across substances. D-amphetamine and MDMA increased cortical influence on the thalamus while decreasing thalamic influence on those cortical areas, a pattern interpreted as increased top-down control over sensorimotor processing and possibly underlying enhanced cognitive, emotional, or prosocial effects. LSD instead increased thalamus→transmodal effective connectivity, suggesting a breach or flattening of hierarchical organisation and greater unimodal–transmodal integration. The authors relate these directed-connectivity findings to prior iFC and effective-connectivity studies, noting both convergences and discrepancies and attributing some differences to methodological choices (for example spectral DCM here versus regression DCM in other work), ROI definitions, and sample size. The investigators also report that some directed connectivity changes predicted subjective experiences: d-amphetamine thalamocortical increases related to ‘‘good drug effect’’ and ‘‘drug liking,’’ and LSD corticothalamic reductions related to poorer ‘‘concentration.’’ No transmodal connectivity changes predicted VAS items. The authors interpret these associations cautiously, noting they are exploratory and uncorrected for multiple comparisons. Key limitations acknowledged by the study team include methodological differences with prior studies that complicate direct comparisons, potential power limitations given the modest sample size, and the restricted set of ROIs analysed due to the computational demands of spectral DCM. The authors also note regional specificity in effects (for example stronger LSD effects in transmodal than unimodal cortices) that may relate to receptor distributions such as 5-HT2A density. Overall, the discussion frames the results as refining TFM by showing cortex-type dependent directed changes and by demonstrating that increased bottom-up thalamocortical signalling can arise from multiple pharmacological mechanisms.

LSD, d-amphetamine, and MDMA increased excitatory thalamus→unimodal cortex effective connectivity while reducing the influence of those unimodal cortices on the thalamus. The amphetamines produced opposing directed effects with SAL-derived transmodal cortices (increased cortical→thalamus and decreased thalamus→cortex), whereas LSD increased thalamus→transmodal connectivity as well, indicating a disturbance of hierarchical thalamocortical organisation and greater unimodal–transmodal integration.