Transcriptome Analysis of Human Induced Excitatory Neurons Supports a Strong Effect of Clozapine on Cholesterol Biosynthesis

Antipsychotics modulate dopamine and other neurotransmitters which is often thought to underly their therapeutic effects. Nevertheless, less studied consequences of antipsychotics on neuronal function may contribute to their efficacy. Revealing the complete picture behind their action is of paramount importance for precision medicine and accurate drug selection. Progress in cell engineering allows the generation of induced pluripotent stem cells (iPSCs) and their differentiation to a variety of neuronal types, recently reviewed by Das et al. (Das et al., 2020), providing new tools to study antipsychotics. Although many studies have examined their effects on a variety of cell types (Bae and Paik, 1997; Ferno et al., 2005) there is not yet been published study doing so with human neurons. We used a publicly available control iPSC line and a protocol for rapid 21-day induction using neurogenin 2 overexpression to generate excitatory cortical neurons (Zhang et al., 2013) (see detailed materials and methods in the supplementary material). Immunofluorescence (supplementary figure 1C) and transcriptomics (supplementary table 1) validated their glutamatergic neuronal identity. At differentiation day 18 we exposed 6 of the 12 replicate cultures of these excitatory neurons to clozapine at 300 ng/ml (920 nmol/L), the standard therapeutic plasma levels, and the remaining 6 to the clozapine solvent alone (methanol) for the 3 last days (suppl. Figure 1). We harvested the two sets of cultures, extracted total RNA, depleted ribosomal RNA and measured their transcriptomes by RNA sequencing. From the replicates exposed only to solvent 2 failed RNA quality control so 6 exposed and 4 control cultures were used for analysis. (detailed methods in the supplementary material). Genes with low read counts (< 10) were removed to decrease noise, leaving 19, 296 for analysis. Differential expression analysis between clozapine exposed and unexposed neurons showed 51 genes changing at false discovery rate (FDR) <0.05, 122 genes at FDR <0.1 and 286 genes at FDR<0.2. The list of all genes and the analysis results are shown in Supplementary Table 2. We report results for genes at FDR<0.1 although results were similar at all 3 significance levels. Pathway enrichment analysis (www.pantherdb.org), to our surprise but in agreement with the results of many investigators studying glial-like cells (Bae and Paik, 1997; Ferno et al., 2005), showed a very strong effect on cholesterol metabolism genes. More than a quarter (12) of all annotated cholesterol genes (46) in the genome that were expressed in these cells were significantly changed at FDR<0.1, all upregulated. This is a 35-fold enrichment with an adjusted enrichment FDR = 1.70E⁻¹⁰. In strong contrast no other functional category showed any evidence of enrichment.

All 12 cholesterol genes where up-regulated (Figure 1C, 1D). A heatmap of all 44 cholesterol genes is shown in Figure 1B. Most differences were observed in the genes with higher expression and the corresponding dendrogram perfectly separated exposed from unexposed cells. Given this observation we ran a principal component (PC) analysis using the entire set of 46 annotated cholesterol genes. This showed that PC1, capturing the largest fraction of variance in cholesterol genes, perfectly separated exposed from unexposed cells (Figure 1A), underlining the profound effect of clozapine on neuronal cell cholesterol metabolism. On Supplementary Table 3 listing the results for all cholesterol genes, even most of the non-significant genes were upregulated by clozapine.

Figure 1:

(A) Principal component Analysis of RNA-seq of clozapine-exposed IPSC-derived neurons. Control samples are represented in Orange. clozapine-exposed samples are in Blue. (B) Dendrogram of Top 46 genes. Asterisks denotes significantly differentially expressed genes after clozapine treatment according to FDR levels: *** for FDR 0.05, ** for FDR 0.1, * for FDR 0.2. Analysis with FDR cut-off of 0.05, 0.1, and 0.2 contains 6 genes, 12 genes, and 14 genes, respectively. (C) Top 12 significantly differentially expressed genes after clozapine treatment. All top 12 genes are key components of Cholesterol Biosynthesis pathway with FDR<0.1. AC1255257.1 and AC008897.3 are lncRNAs associated with cholesterol genes. (D) Expressivity level of top 12 genes. Bar graph shows expression level of each gene in pairs of treated (Blue) vs untreated IPSC-derived samples (orange). Error bars with SEM.

Cholesterol is essential for neuronal physiology. The brain is the most cholesterol-rich organ and possesses its own independent cholesterol metabolism machinery (Zhang and Liu, 2015) separate from peripheral tissues due to stringent diffusive properties of the blood-brain-barrier. Alterations in brain cholesterol metabolism have been associated with Alzheimer’s, Parkinson’s, Huntington’s and amyotrophic lateral sclerosis (Jin et al., 2019). Cholesterol has profound effects on the biophysical properties of neuronal membranes, affecting the function of membrane-resident signaling components including ion channels, transporters and receptors (Lijnen, 1997). It is an essential element of the exocytosis apparatus and crucial in the biogenesis and transport of synaptic vesicles, whose membrane contains more cholesterol than other intracellular organelles (Pfrieger, 2003; Schmitz and Orso, 2001). Although neurons produce sufficient cholesterol to survive, generate axons and dendrites and form a few inefficient synapses, it is the glial cells and astrocytes that supplement them with the required additional cholesterol needed for massive synapse formation (Pfrieger, 2003). It has also been shown that developing neurons produce higher amounts of cholesterol per cell than astrocytes (Genaro-Mattos et al., 2019). It has been proposed that the effects of antipsychotics consistently identified in glia and retinal pigment may reflect a mechanism of their action (Polymeropoulos et al., 2009; Vik-Mo et al., 2009). Importantly, Smith-Lemli-Opitz syndrome (SLOS) is caused by an inborn error of cholesterol biosynthesis and ~50% of individuals meet criteria for autism (Tierney et al., 2001). Finally, the gene encoding HMCGR, the rate-controlling enzyme for the production of cholesterol and targeted by statins (Sirtori, 2014), was upregulated in our data along with an associated lncRNA (AC008897.3). HMCGR has been found to carry excess loss of function mutations in schizophrenia (meta-analysis p = 4.5×10⁻⁴, (schema.broadinstitute.org/results), directly implicating cholesterol metabolism in the disease. It must be noted that clozapine has a well-established effect on blood lipid levels, so the transcriptional effects we describe might not be confined to the brain and may underlie some of its side effects.

While the glutamatergic system may be particularly relevant in schizophrenia, since drug-resistant schizophrenia involves such abnormalities (Aringhieri et al., 2018), studying one cell type in isolation cannot capture the complexity of interactions between neuronal populations in the brain and could miss effects on dopamine or serotonin receptors. This is an unavoidable limitation of our study. Care must also be takes as our study only examines one cell line. Cell lines from different individuals might show different effects, which might be an important observation for precision medicine. Nevertheless, by singling out cholesterol metabolism as a major strong response, our work provides support that this may be a significant mechanism in the therapeutic benefits of clozapine. Understanding all possible mechanism that can contribute to disease improvement is important for future tailored treatments guided by precision medicine.