Antipsychotic drug use complicates assessment of gene expression changes associated with schizophrenia

Abstract

Recent postmortem transcriptomic studies of schizophrenia (SCZ) have shown hundreds of differentially expressed genes. However, the extent to which these gene expression changes reflect antipsychotic drug (APD) exposure remains uncertain. We compared differential gene expression in the prefrontal cortex of SCZ patients who tested positive for APDs at the time of death with SCZ patients who did not. APD exposure was associated with numerous changes in the brain transcriptome, especially among SCZ patients on atypical APDs. Brain transcriptome data from macaques chronically treated with APDs showed that APDs affect the expression of many functionally relevant genes, some of which show expression changes in the same directions as those observed in SCZ. Co-expression modules enriched for synaptic function showed convergent patterns between SCZ and some of the APD effects, while those associated with inflammation and glucose metabolism exhibited predominantly divergent patterns between SCZ and APD effects. In contrast, major cell-type shifts inferred in SCZ were primarily unaffected by APD use. These results show that APDs may confound SCZ-associated gene expression changes in postmortem brain tissue. Disentangling these effects will help identify causal genes and improve our neurobiological understanding of SCZ.

Fig. 1. SCZ-associated differential gene expression between toxicological subgroups.

a Aggregated scores of differential gene expression (SCZ expression signatures) are shown based on toxicological findings (Neg.: negative for any APD; Pos.: positive for atypical, typical, or both APD [mixed] classes). Kruskal–Wallis test (KWt) on all groups: p = 3.83e-10; KWt within SCZ subgroups only: p = 0.014. For details and post hoc Dunn tests, see Supplementary Table S2. b As in a but SCZ samples were divided into groups of 3 or more cases positive for a single APD (atypical APDs: clozapine [n = 3], olanzapine [n = 8], risperidone [n = 6]; typical APDs: haloperidol [n = 14], fluphenazine [n = 5]). SCZ expression signatures were based on gene expression values residualized for major influential covariates, such as sex, age, RNA integrity, and postmortem interval (see “Methods”).

Fig. 2. Consensus WGCNA between human and macaque DLPFC transcriptome data.

a Relationship of module eigengenes to APD use in macaque (CLZ = clozapine, HAL.lo = low-dose haloperidol; HAL.hi = high-dose haloperidol) and humans with SCZ (data from HBCC). Heat map shows t-statistic from linear regression illustrating module eigengene directionality (same color indicates concordance; two-sided t test significance level. ***p < 0.001, **p < 0.01, *p < 0.05. Hierarchical clustering of rows with Euclidean distance cut at the level of four branches was used for better visibility. Heat map annotation shows the number of genes in each module and the most over-represented gene ontology term (based on Fisher’s exact test; excluding terms with ≤3 hits). SCZ.GWAS: Enrichment (Fisher’s exact test) of prioritized genes from the 2022 SCZ GWAS [15]. For details on the statistical tests and FDR-corrected p values, see Supplementary Table S2. b Relationships of individual genes to APD exposure in macaque (x-axis) and humans with SCZ (y-axis) for five example modules with predominantly divergent (M16, M9) or convergent (M11, M25, M31) gene expression patterns (shown as t-statistics). For visualization purposes, the top genes with nominal p < 0.05 for SCZ and p < 0.25 for each APD in macaques are highlighted in red. For a full list of genes and enriched gene ontology terms, see Supplementary Table S4.

Fig. 3. Estimated cell-type proportions for human and macaque DLPFC samples.

a Cell-type proportions of excitatory neurons (ExN), inhibitory neurons (InN), and astrocytes (Astro) in bulk DLPFC tissue of SCZ cases with different toxicological profiles (groups as in Fig. 1a) estimated via corresponding single-nucleus RNA-seq profiles. Kruskal–Wallis test (KWt) on all groups: p_ExN = 0.908; p_InN = 0.0318; p_Astro = 0.0689; KWt within SCZ subgroups only: p_ExN = 0.798; p_InN = 0.875; p_Astro = 0.565. Mann–Whitney U-test (MWU) between SCZ and controls was significant for InN (p = 0.00159) and Astro (p = 0.0176). Only InN remained significant after FDR correction. b Estimated proportions of ExN, InN, and Astro in bulk DLPFC of monkeys treated with clozapine (CLZ), haloperidol (low dose: HAL.lo; high dose: HAL.hi), or placebo. KWt: p_ExN = 0.0433; p_InN = 0.547; p_Astro = 0.336. MWU between APDs and placebo was significant only for HAL.hi in ExN (p = 0.0379). For detailed test statistics, see Supplementary Table S2. For the other cell types, see Supplementary Fig. S5.