Spatial correspondence of LSD-induced variations of brain functioning at rest with serotonin receptor expression

Delli Pizzi and colleagues situate their work within a recent resurgence of experimental research on LSD, noting that psychedelic compounds illuminate the neural bases of perception and selfhood and hold therapeutic promise for mood disorders. While partial agonism at 5-HT2A receptors is widely regarded as central to psychedelic effects, LSD is pharmacologically atypical because it also engages 5-HT1A receptors; how these pleiotropic actions map onto large-scale changes in brain function remains incompletely understood. Previous fMRI studies have reported increased between-network cortical connectivity under LSD and some voxel-wise work has suggested spatial overlap between LSD-induced connectivity changes and cortical 5-HT receptor topography, but questions remain about whether single or multiple receptor subtypes explain bidirectional connectivity effects and how local neuronal activity is affected in parallel. The present study set out to test whether voxel-wise changes in intrinsic functional connectivity and local signal amplitude induced by acute LSD map onto the cortical distribution of 5-HT1A and 5-HT2A receptors. To do so, the investigators analysed resting-state fMRI from a placebo-controlled, within-subject design in healthy volunteers, computing Intrinsic Connectivity Contrast (ICC) as a whole-brain centrality metric and fractional Amplitude of Low-Frequency Fluctuations (fALFF) as a proxy for local signal amplitude. The authors predicted both increases and decreases in connectivity that might parallel increases and decreases in local amplitude, and that these spatial patterns would correspond to receptor density maps obtained from in-vivo PET atlases; they also planned exploratory tests of relationships between rs-fMRI changes and subjective visual hallucinations.

The study used a within-subject, placebo-controlled design. Twenty healthy volunteers attended two 3T MRI sessions at least fourteen days apart; on one session participants received placebo (10 mL saline) and on the other they received LSD 75 μg in 10 mL saline infused over two minutes. Session order was counterbalanced and each session included three 7:20-minute eyes-closed resting-state fMRI runs; only the first resting scan (no stimulation) was analysed here. Five participants were excluded (one for anxiety, four for excessive head motion), yielding a final sample of 15 subjects (four women; mean age 30.5 ± 8.0 years). Anatomical T1-weighted images were processed with FreeSurfer 6.0 to reconstruct cortical surfaces and labels; surface-based rs-fMRI preprocessing in the CONN toolbox included slice-timing correction, motion correction, outlier detection/scrubbing, and resampling to surface and volumetric resolutions. Instead of global signal regression, the investigators removed multiple signals from CSF and white matter to mitigate non-neuronal confounds while avoiding the potential artefacts introduced by GSR; mean frame-wise displacement was included as a covariate. The two primary rs-fMRI metrics were fALFF, defined as the ratio of BOLD power in 0.01–0.1 Hz to the total accessible frequency range (0–0.25 Hz with TR=2s), and ICC, a voxel-wise centrality measure equal to the root-mean-square connectivity strength between a voxel and all other voxels. Receptor density maps for 5-HT1A and 5-HT2A were taken from an assembled collection of in-vivo PET atlases pooling over 1200 subjects. These PET maps were projected onto the fsaverage5 surface and co-registered with the fALFF and ICC maps. Spatial correspondence between maps was tested using a spin-test spatial permutation framework (1000 permutations) that preserves the spatial covariance of cortical topology; Spearman correlations were used and resulting p-values were FDR-corrected (reported as P_spin). Network-level analyses averaged z-scored fALFF and ICC values within each of Yeo’s 17 networks adapted to individual anatomy; receptor density within networks was normalised to whole-cortex averages to produce m-values. Statistical thresholds: whole-brain cluster-wise inference used a voxel-wise cluster-forming threshold of p<0.001 and cluster-wide p<0.05. Network analyses used one-way ANOVA with Bonferroni correction across 17 networks × 2 hemispheres (corrected p set at p<0.001). Behavioural associations were tested using linear regression models (rs-fMRI indices as independent variables, visual analogue scale ratings for simple and complex visual hallucinations as dependent variables), with FDR correction applied to those tests.

The study found a strong spatial correspondence between voxel-wise changes in local signal amplitude (fALFF) and functional centrality (ICC) induced by LSD: these two maps were highly correlated (Spearman r=0.838, P_spin <0.0001). Both fALFF and ICC modulation maps showed positive spatial correlations with 5-HT2A receptor density (fALFF r=0.365, P_spin =0.013; ICC r=0.423, P_spin =0.001). Conversely, fALFF modulations were negatively correlated with 5-HT1A receptor density (r=-0.448, P_spin =0.02). A conjunction analysis identified two principal clusters with opposite effects. One cluster in left frontoparietal cortex exhibited increased fALFF and ICC under LSD; this region was relatively rich in 5-HT2A receptors. The other cluster comprised bilateral limbic regions showing decreases in both metrics and was relatively dense in 5-HT1A receptors. Network-parcellation analyses specified where increases and decreases occurred. LSD increased both fALFF and ICC in: part B of the salience/ventral attention network (Yeo component 8), parts A/B of the control network (components 11 and 12), and parts A/B of the default mode network (components 14 and 15). Additional fALFF increases were seen in part A of the dorsal attention network (component 5) and part C of the control network (component 13). Decreases of both fALFF and ICC were observed in limbic network components (9 and 10) and part C of the default mode network (component 16). Reduced fALFF appeared in part A of the salience/ventral attention network (component 7), and diminished ICC was observed in part B of somatosensory networks (component 4). Network components showing increased fALFF/ICC were characterised by high 5-HT2A and low 5-HT1A density, while components with decreased metrics showed the opposite receptor profile. In behavioural correlations, increased fALFF (but not ICC) in the left part B of the default mode network (Yeo component 15) was associated with higher visual analogue scale scores for simple visual hallucinations (β=0.551, FDR-corrected p=0.033) and for complex visual hallucinations (β=0.555, FDR-corrected p=0.033).

Delli Pizzi and colleagues interpret their findings as showing that LSD produces spatially patterned increases and decreases in local signal amplitude and functional centrality that correspond with the cortical distributions of 5-HT2A and 5-HT1A receptors, respectively. The high concordance between fALFF and ICC maps suggests that local amplitude changes and alterations in a voxel’s overall connectivity weight tend to co-occur, possibly reflecting a common neurovascular or metabolic substrate; the authors note prior reports linking these metrics to brain glucose metabolism. The pattern of increased ICC in default mode, salience, and frontoparietal control networks is reconciled with previous literature by distinguishing intra-network versus inter-network connectivity: ICC indexes global connectivity and likely captures increased inter-network coupling under LSD, whereas some prior reports of reduced DMN connectivity focused on intra-DMN decreases. The authors argue that regions high in 5-HT2A receptors show enhanced amplitude and cross-cortical integration consistent with the stimulatory action of 5-HT2A and its known necessity for hallucinations; this is supported by the observed correlation between increased fALFF in a DMN subcomponent and simple and complex visual hallucinations. Conversely, limbic regions and other areas rich in 5-HT1A receptors showed decreased fALFF and ICC, which the investigators link to the inhibitory signalling associated with 5-HT1A and to possible effects on emotional processing. The authors acknowledge limitations. First, some observed BOLD changes could reflect vascular rather than neuronal effects of LSD; although they argue a purely vascular explanation is unlikely to produce the specific topography observed, concurrent measures of cerebral blood flow and volume would be needed for a definitive dissociation. Second, the sample size was modest (n=15 after exclusions), reflecting practical constraints of pharmacological neuroimaging; this limits power and makes behavioural correlations exploratory. Finally, the authors suggest further research to determine whether similar cortical reconfigurations occur with other psychedelics and to explore implications for pharmacological treatment of mood disorders, but note that these implications remain to be tested empirically.