Molecular, physiological and functional features underlying antipsychotic medication use related cortical thinning

Abstract

Use of antipsychotic medication is related to thinning of the cerebral cortex, but the underlying mechanisms of this effect remain largely unknown. Here, we investigated potential mechanisms across multiple levels of description by comparing antipsychotic medication related cortical thinning to atlases of normative neurotransmitter distributions, structural and functional organization of the brain, and meta-analyses of functional activation from the Neurosynth database. We first analyzed a single-site discovery sample of patients (N = 131) with early psychosis for whom antipsychotic related cortical thinning was estimated based on lifetime exposure to antipsychotics. Findings were replicated using data from a large (N ≥ 2168) ENIGMA meta-analysis on schizophrenia patients. We discovered that antipsychotic related cortical thinning is associated with a number of neurotransmitter systems, most notably the serotonin system, as well as physiological measures, functional networks and neural oscillatory power distributions typical for regions subserving higher cognition. At the functional level, antipsychotic related cortical thinning affects regions involved in executive function and motivation, but not perception. These results show how molecular, physiological, and large-scale functional patterns may underlie antipsychotic related cortical thinning.

Fig. 1. Higher lifetime antipsychotic exposure is associated with lower cortical thickness in the Turku sample (N = 131).

Figure on the left shows unthresholded results. Figure on the right shows permutation corrected map at p < 0.05. Model includes lifetime antipsychotic exposure, age, sex, and diagnostic group as predictors. Colorbar indicates p-value. There were no vertices where the thickness increased statistically significantly.

Fig. 2. Regional sensitivity to antipsychotic exposure and underlying brain features in the Turku sample.

Features associated with antipsychotic related cortical thinning include serotonergic, cholinergic, structural, functional, and metabolic features. Positive correlation indicates that regions that have a higher value of the measured brain feature are more susceptible to the effects of antipsychotics than regions that have a lower value and vice versa. The dots show Pearson’s correlation coefficient r between antipsychotics effects and a given brain feature, the color of the dot indicates statistical significance, all associations survived false discovery rate (FDR) correction for multiple comparisons The ends of the boxes represent the first and third quartiles and the center line represents the median of the null distribution (10,000 rotations), the whiskers represent the non-outlier end-points of the distribution.

Fig. 3. Parcel-wise associations between antipsychotic medication and cortical thickness in the Turku and ENIGMA samples.

In the Turku sample, mean cortical thickness in each parcel and lifetime antipsychotic exposure are correlated while controlling for age, sex, and group. In the ENIGMA sample, correlation between mean cortical thickness in each parcel and current antipsychotic dose is calculated while controlling for age and sex. Colorbar indicates partial correlation r. Spatial correlation of antipsychotic effects is similar between the two samples (Pearson’s r = 0.47, pspin = 0.0008). Two regions with high and low sensitivity to antipsychotics are labeled in the scatter plot.

Fig. 4. Associations between antipsychotic related cortical thinning and molecular, functional, structural, and metabolic features replicated in the ENIGMA sample.

The dots show Pearson’s correlation coefficient r between antipsychotics effects and a given brain feature, the color of the dot indicates statistical significance, all associations survived false discovery rate (FDR) correction for multiple comparisons. The ends of the boxes represent the first and third quartiles and the center line represents the median of the null distribution (10,000 rotations), the whiskers represent the non-outlier endpoints of the distribution.

Fig. 5. Antipsychotic related cortical thinning and cognitive functions.

Probabilistic maps for 123 different cognitive functions were derived from Neurosynth database and correlated with antipsychotic related cortical thinning. On the left, the cognitive functions were grouped into 11 categories and ranked based on average correlation. In the Turku sample, we found that functions related to ‘perception’ were negatively correlated and ‘reasoning / decision making’, ‘executive / cognitive control’, and ’motivation’ were positively correlated with antipsychotic related cortical thinning (two-sided permutation test over groupings). Similarly, in the ENIGMA sample, ‘perception’, ‘executive / cognitive control’, and ‘motivation’, but also ‘language’ and ‘social function’ were associated with antipsychotic related cortical thinning. Top and bottom 10% of the strongest individual term correlations in both samples are shown on the right.