Figure 1. Schematic overview of the biophysical modeling framework.

(A) Each node in the large-scale model represents a cortical microcircuit comprised of recurrently coupled excitatory (E) and inhibitory (I) neuronal populations. The model includes one node for each of the 180 left-hemispheric parcels in the Human Connectome Project’s Multi-Modal Parcellation (MMP1.0). Nodes interact through structured long-range excitatory projections, the strengths of which are constrained by a diffusion magnetic resonance imaging (MRI)-derived structural connectivity (SC) matrix. (B) Simulated synaptic activity in each node is transformed to a simulated blood oxygen level-dependent (BOLD) signal using the Balloon-Windkessel model of the hemodynamic response. (C) Lysergic acid diethylamide (LSD)’s effect on cortical microcircuitry is modeled as a modulation of neural gain due to serotonin-2A (5-HT_2A) receptor agonism by the LSD molecule. The degree to which neural gain is modulated within an area is scaled in proportion to the regional expression level of HTR2A, the gene which encodes the 5-HT_2A receptor protein. Gain curves of the excitatory and inhibitory neuronal populations are modulated independently, permitting cell-type specific effects. (D) Global brain connectivity (GBC), a graph-theoretic functional measure, is dramatically altered following LSD administration. The functional MRI (fMRI)-derived map of the change in GBC (ΔGBC) under LSD, relative to placebo, specifies the target model output. To simulate brain function in the LSD and placebo drug conditions, we simulate GBC maps with and without gain modulation, respectively. We compute the difference between the model GBC maps to construct a simulated ΔGBC map. Quantitative comparisons between empirical and model ΔGBC maps determine how well the model captures the topography of LSD-induced functional disruptions.

Figure 1—figure supplement 1. In the human brain, HTR2A is predominately expressed in cortical pyramidal neurons.

In the human brain, HTR2A is predominately expressed in cortical pyramidal neurons. (A) Cortical topography of the HTR2A expression map. (B) Whole-brain topography of the HTR2A expression map. For subcortex, we use the 358 subcortical parcels in the Cole-Anticevic Brain Network Parcellation (CAB-NP) (Ji et al., 2019). Expression levels are linearly rescaled such that the minimum value is zero, and the cortical parcel-wise average is one. The large difference between expression levels in cortex and subcortex is much greater than the variance across parcels within the cortex. Note that the Allen Human Brain Atlas has unilateral sampling of gene expression in the left hemisphere (Hawrylycz et al., 2015), and therefore the map is made bilaterally symmetric at the parcel level for cortex and coordinate level for subcortex. Gene expression mapping follows the method of Burt et al., 2018. (C) HTR2A expression levels grouped by gross anatomical structure. Box plots mark the median and inner quartile ranges for expression levels across parcels within each anatomical structure, and whiskers indicate the 95% confidence interval. ‘Subcortex (aggregated)’ comprises parcel expression levels for all subcortical structures (i.e., all 358 subcortical parcels). Expression of HTR2A is significantly higher in cortex than in subcortex (${\displaystyle W=57}$; p < machine precision; Wilcoxon signed-rank test). (D) The distribution of HTR2A expression levels across excitatory (red) and inhibitory (blue) human cortical cell types. HTR2A is significantly more expressed in excitatory neurons than in inhibitory neurons (W = 141,943; p < machine precision; Wilcoxon signed-rank test).