THC exposure of human iPSC neurons impacts genes associated with neuropsychiatric disorders
Human-induced pluripotent stem cells (hiPSCs) serve as a tool for the study of developmental processes and disease-relevant models. This has been especially valuable for the study of the human brain where primary tissue for study has been the most difficult to obtain. hiPSCs have provided mechanistic insights into both neurodevelopmental disorders1 and neurodegenerative diseases2,3. Research into psychiatric disorders such as autism4, bipolar disease5, and schizophrenia6 have greatly benefited from the insights afforded by hiPSCs, as these are largely considered human- specific disorders. hiPSC-based models facilitate isogenic investigations into molecular and environmental factors that may exacerbate or ameliorate disease predisposition.
The widespread use of cannabis calls for a concerted effort into increased understanding of both the positive and negative effects of the drug. Brain imaging studies of the primary psychoactive component of cannabis, Δ9-tetrahydrocannabinol (THC), demonstrated structural and functional changes following regular cannabis use7, while molecular studies uncovered signaling pathways downstream of the two cannabinoid (CB) receptors, CB1, and CB2. Depression of glutamate signaling is a common feature of THC-induced effects via the CB1 receptor in both humans and in animal models8,9.
There is a significant association between cannabis use and schizophrenia in human subjects10,11,12,13,14, however, whether this reflects patient self-medication of prodromal symptoms or an environmental modulation of genetic susceptibility remains an ongoing discussion15,16. We recently reported molecular abnormalities in schizophrenia patient hiPSC-derived neurons in response to neural activity6; here we describe a distinct overlap in hypo-excitability, particularly in the glutamate system, between schizophrenia patient-derived neurons and those treated with THC. THC exposure seems to deregulate glutamate receptors and other genes involved in synaptic function. We observe significant THC-dependent changes in postsynaptic density, ion channel and WNT pathway genes, and epigenetic regulators; and molecular connections to autism and intellectual disability. Although the molecular mechanisms may not be precisely the same, the convergence of glutamatergic hypo-function may partially explain the increased risk for psychiatric disorders amongst those exposed to cannabis.
Materials and methods
Generation of hiPSC neurons and RNA sequencing
Human fibroblasts were obtained from ATCC (CRL-2522) and Coriell (control: GM03440, GM03651, GM04506, AG09319, AG09429; SZ: GM01792, GM02038, GM01835, and GM02497). Limited phenotypic information for each donor is available from the Coriell Cell Repository, and summarized in the methods of Topol et al17. Unfortunately, THC exposure status for each donor is unknown. hiPSCs were reprogrammed using tetracycline-inducible lentiviral vectors and differentiated to neural precursor cells (NPCs) as previously described18. NPCs were differentiated on poly- ornithine/laminin coated plates for 6 weeks. Passage-matched NPCs were used for all experiments. All hiPSC and NPCs used were mycoplasma-free. Forebrain neural progenitor cells were generated from five control and four case hiPSCs as previously reported6,18,19 and neurons were differentiated according to a 6-week maturation protocol. Samples used in RNA sequencing or quantitative RT–PCR can be found in Supplementary Table S1. THC was dissolved in DMSO to 1 mg/ml and prepared as previously described;20 in all experiments, an equivalent volume of DMSO was used as a vehicle control. Acute (1 μM THC for 24 h) and chronic (50 nM THC; five treatments over 7 days) THC exposure (and DMSO-vehicle control) occurred immediately prior to collection. KCl was dissolved in PBS as previously described6; in all experiments, an equivalent volume of PBS was used as a vehicle control. 50 mM KCl treatment occurred for the final three hours prior to collection; consistent with our previous molecular6 and neurotransmitter release21 studies. For RNA-seq experiments, two wells per individual were treated. The RNA Integrity Number (RIN) was determined using an RNA Nano chip (Agilent Technologies) on the Agilent 2100 Bioanalyzer. All samples have high RIN (mean ± SD: 9.54 ± 0.21). 500 ng of total RNA was used as input material for library preparation using the TruSeq Stranded Total RNA Kit (Illumina, USA).
Processing of RNA sequencing data and analyses
RNA sequencing data has been deposited into Sequence Read Archive (SRA; PRNJA419702, “RNA-Seq of iPSC-derived neurons”). Reads were mapped to GRCh38.p5 reference genome using STAR (version 2.5.1a). Known Gencode gene levels (version 24) were quantified by RSEM (version 1.3.0). To facilitate inter-dataset comparisons, we performed ranked (Spearman) and unranked correlation (Pearson) analysis of the controls in both the ±KCl and ±THC datasets, and confirmed that the samples are highly comparable (all control comparisons are ≥ 97%). Differentially expressed genes were identified with edgeR in R after TMM normalization and filtering. p-values and false discovery rate (FDR) were calculated and differentially expressed genes (DEG) were determined as those with an estimated p-value ≤ 0.05 and FDR ≤ 0.01.
Gene sets for enrichment analyses
To further characterize the DEGs we performed enrichment analysis, using a group of gene sets for known molecular pathways and biological processes, including: Gene Ontology (GO) sets of molecular functions (MF), biological processes (BP), and cellular components (CC) (http://www.geneontology.org); the KEGG dataset (http://www.genome.jp/kegg/pathway.html); and the HUGO Gene Nomenclature Committee (HGNC) gene families (http://www.genenames.org). The genes in each gene set were tested for overlap using Fisher’s exact test and FDR correction. Differential expressed genes were (i) separated in upregulated and downregulated genes; (ii) analyzed for full GO overrepresentation according to hypergeometric testing with a significance cutoff FDR = 0.05 in BiNGO (version 3.0.3); (iii) processed with the enrichment map pipeline (https://f1000research.com/posters/5-1235) a p-value cutoff = 0.001, q-value cutoff = 0.05 and Jaccard coefficient cutoff = 0.25 and (iv) visualized in Cytoscape (version 3.5.1).
For qPCR experiments, three wells per individual were treated with either DMSO- vehicle control for 7 days, acute THC exposure (1 μM THC for 24 h) or chronic THC treatment (five treatments with 50 nM THC over 7 days) immediately prior to collection at 6 weeks. Candidate genes were validated for THC-treated and activity-treated alterations using quantitative RT–PCR. cDNA synthesis was performed using the SuperScript III First-Strand Synthesis System (ThermoFisher Scientific, USA). Briefly, 500 ng of total RNA was used and random hexamer primed protocol was followed. Each cDNA sample was amplified in triplicate using SYBR Green PCR Master Mix (ThermoFisher Scientific, USA). Primer pairs used for this analysis are described in Supplementary Table S2.
Generation of gene datasets
As no up to date datasets for associated genes were available for autism, intellectual disability or schizophrenia, we generated our own through extensive literature and database searches. Specific details are available in the Supplemental Information ‘Generation of Gene Databases’.
hiPSC-derived neurons as a model for THC biology
To gain further insight into THC-related molecular mechanisms we utilized hiPSC-derived neurons from four controls as previously reported6. THC (or vehicle control) was added to hiPSC-derived neurons from each individual as acute (1 μM THC for 24 h) or chronic (50 nM THC; five treatments over 7 days) treatments. Acute and chronic THC concentrations were rationally selected from studies of primary mouse neurons22 and experimentally validated in hiPSC neurons23. RNA was extracted and subjected to RNA sequencing (RNA-seq) using the Illumina platform. Our bioinformatic analyses pipeline combined integrated genome/transcriptome alignment using STAR, quantification using RSEM and differential expression using EdgeR. Relative to vehicle treatment, acute THC exposure resulted in 497 genes significantly altered in hiPSC- derived neurons compared to untreated controls, while chronic THC exposure perturbed 810 genes (Fig. 1a; Supplementary Table S3; Supplementary Figure 1). The overlap between acute and chronic exposures was highly significant (421 genes; p– value = 0e + 00, odds ratio = 586.5, Fisher’s exact test). Specific subsets of genes involved in the glutamate receptor pathway and mitochondrial function were altered in response to acute or chronic THC exposure (Supplementary Table S4; Supplementary Figure 2; Fig. 1b–d) and have previously been implicated in THC biology8,24,25. These results provide data to support the use of hiPSC-derived neurons as a model for investigating THC responses in an in vitro human neuronal system.
RNA sequencing implicates synaptic function, demethylation and ion channel function in THC-treated hiPSC neurons
Closer inspection of functional gene clusters associated with THC treatment revealed the potential contribution of genes involved at the postsynaptic density such as HOMER1, GRID2, GRIK1, and SIPA1L1 (Fig. 2a, b). Moreover, chronic THC treatments resulted in the alteration of additional synaptic related genes such as SYNGAP1 and SHANK1. Ion channel genes, especially potassium voltage-gated channel genes (KCNE4, KCNA4, KCNJ10, and KCNN3) are also responsive to both THC treatments with further ion channel genes (KCNJ2, KCNA2, and KCNT2) associated following chronic THC exposure (Fig. 2a). These results strongly implicate synaptic function as a key target of THC-mediated responses. Interestingly, we found epigenetic related transcriptional responses evident in both acute and chronic THC exposures that included the alterations of genes involved in the dynamic methylation/demethylation process (DNMT1, GADD45B, and APOBEC3C); chronic THC exposure resulted in further decreases of histone modification-related proteins such as SETD1A, SETD5, CBX6, KMT2A, KMT2C, and NCOA6 and methyl binding proteins MECP2 and MBD5 (Supplementary Table S5). Network analyses (Enrichment map pipeline in Cytoscape) using the genes altered in response to THC exposure reinforce the involvement of pathways linked to developmental, chromatin regulation and mitochondrial biology (Fig. 2c; Supplementary Table S6).
THC exposure significantly alters genes implicated in autism and intellectual disability
We noticed that many genes implicated in psychiatric disease coincided with genes altered in response to THC treatments. In order to calculate statistical relevance we needed to first update the numbers of genes associated with these disorders and found genes related to autism spectrum disorder (1037 genes), intellectual disability (2461 genes) and schizophrenia (723 genes; see Supplementary Information ‘Generation of Gene Databases’ for details; Supplementary Table S7). Included in our list of significantly altered transcripts following THC exposure is a substantial number of genes linked to autism (80 genes) and intellectual disability (167 genes), with fewer overlapping with schizophrenia (Fig. 3a); autism and intellectual disability associated genes are significant for both p-value and odds ratio using the Fisher’s exact test (Fig. 3b). These data suggest that endogenous THC responsive pathways include many psychiatric disease-associated genes and that changes in these genes, either genetically or epigenetically, may contribute to cannabis-related adverse reactions such as psychosis in some users.
Overlapping signaling pathways between THC and schizophrenia
We compared the bioinformatic results from our current THC studies to data obtained from our previous studies of schizophrenia patient hiPSC-derived neurons6,18 to ensure that the quality of the differentiations were comparable across experiments (Supplementary Figure 3). Raw data from all subjects from Roussos et al6. was applied to our bioinformatic pipeline. WNT and mitochondrial pathways (Supplementary Table S8) were significantly altered in both our current THC and previous schizophrenia studies18,26. Genes related to ion channel function were also highly represented in both the THC and schizophrenia gene lists (Supplementary Table S8). Although these pathways were conserved, specific genes related to altered ion channel, WNT or mitochondrial function did not overlap.
Blunted activity-dependent transcriptional response shared between THC and schizophrenia
In our previous study6, we demonstrated that schizophrenia-associated hiPSC-derived neurons had a blunted transcriptomic response to KCl relative to controls. We repeated this experimental design on control hiPSC-derived neurons from four individuals, providing either acute (1 μM THC, 24 h), chronic (50 nM THC, 7 days) or vehicle treatment, after which cells were activated using 50 mM KCl (or vehicle) for 3 h as before. We saw a significantly blunted transcriptomic response, more prominent with the acute (75% reduction compared to KCl-activated control neurons; p-value = 1.3e−73, odds ratio = 278.6, Fisher’s exact test) than the chronic exposure (60% reduction compared to KCl-activated control neurons; p-value = 4.4e−83, odds ratio = 181.6, Fisher’s exact test) of THC (Fig. 4a; Supplementary Table S9; Supplementary Figure 4).
After re-running the raw schizophrenia-associated data from activity-dependent experiments conducted in Roussos et al6., we again saw a dramatic reduction (~93%; p-value = 4.3e−27, odds ratio = 605.7, Fisher’s exact test) in the schizophrenia-associated transcriptomic response (Fig. 4b; Supplementary Table S10). We tested candidate genes on a cohort of schizophrenia-associated hiPSC-derived neurons (Fig. 4c) and found blunted expression profiles for COX7A2, GRID2 and HOMER1 (Fig. 4d) using quantitative PCR. Quantitative PCR further confirmed this blunting effect of THC exposure; significantly reduced expression of immediate early genes such as NR4A1 and FOSB was observed following KCl treatment (Fig. 4e, f), consistent with what we found previously for these genes in schizophrenia-associated hiPSC-derived neurons6.