No abstract

Genome-wide association studies (GWAS) have successfully identified 145 genomic regions that contribute to schizophrenia risk, but linkage disequilibrium (LD) makes it challenging to discern causal variants. Computational finemapping prioritized thousands of credible variants, ∼98% of which lie within poorly characterized non-coding regions. To functionally validate their regulatory effects, we performed a massively parallel reporter assay (MPRA) on 5,173 finemapped schizophrenia GWAS variants in primary human neural progenitors (HNPs). We identified 439 variants with allelic regulatory effects (MPRA-positive variants), with 71% of GWAS loci containing at least one MPRA-positive variant. Transcription factor binding had modest predictive power for predicting the allelic activity of MPRA-positive variants, while GWAS association, finemap posterior probability, enhancer overlap, and evolutionary conservation failed to predict MPRA-positive variants. Furthermore, 64% of MPRA-positive variants did not exhibit eQTL signature, suggesting that MPRA could identify yet unexplored variants with regulatory potentials. MPRA-positive variants differed from eQTLs, as they were more frequently located in distal neuronal enhancers. Therefore, we leveraged neuronal 3D chromatin architecture to identify 272 genes that physically interact with MPRA-positive variants. These genes annotated by chromatin interactome displayed higher mutational constraints and regulatory complexity than genes annotated by eQTLs, recapitulating a recent finding that eQTL- and GWAS-detected variants map to genes with different properties. Finally, we propose a model in which allelic activity of multiple variants within a GWAS locus can be aggregated to predict gene expression by taking chromatin contact frequency and accessibility into account. In conclusion, we demonstrate that MPRA can effectively identify functional regulatory variants and delineate previously unknown regulatory principles of schizophrenia.

Background Alzheimer’s disease (AD) is a progressive neurodegenerative disease that impacts nearly 400 million people worldwide. The accumulation of amyloid beta (Aβ) in the brain has historically been associated with AD, and recent evidence suggests that neuroinflammation plays a central role in its origin and progression. These observations have given rise to the theory that Aβ is the primary trigger of AD, and induces proinflammatory activation of immune brain cells (i.e., microglia), which culminates in neuronal damage and cognitive decline. To test this hypothesis, many in vitro systems have been established to study Aβ-mediated activation of innate immune cells. Nevertheless, the transcriptional resemblance of these models to the microglia in the AD brain has never been comprehensively studied on a genome-wide scale. Methods We used bulk RNA-seq to assess the transcriptional differences between in vitro cell types used to model neuroinflammation in AD, including several established, primary and iPSC-derived immune cell lines (macrophages, microglia and astrocytes) and their similarities to primary cells in the AD brain. We then analyzed the transcriptional response of these innate immune cells to synthetic Aβ or LPS and INFγ. Results We found that human induced pluripotent stem cell (hIPSC)-derived microglia (IMGL) are the in vitro cell model that best resembles primary microglia. Surprisingly, synthetic Aβ does not trigger a robust transcriptional response in any of the cellular models analyzed, despite testing a wide variety of Aβ formulations, concentrations, and treatment conditions. Finally, we found that bacterial LPS and INFγ activate microglia and induce transcriptional changes that resemble many, but not all, aspects of the transcriptomic profiles of disease associated microglia (DAM) present in the AD brain. Conclusions These results suggest that synthetic Aβ treatment of innate immune cell cultures does not recapitulate transcriptional profiles observed in microglia from AD brains. In contrast, treating IMGL with LPS and INFγ induces transcriptional changes similar to those observed in microglia detected in AD brains.

To infer potential causal relationships between 3D chromatin structure, enhancers, and gene transcription, we mapped each feature in a genome-wide fashion across eight narrowly-spaced timepoints of macrophage activation. Enhancers and genes connected by loops exhibited stronger correlations between histone H3K27 acetylation and expression than can be explained by genomic distance or physical proximity alone. Changes in acetylation at looped distal enhancers preceded changes in gene expression. Changes in gene expression exhibit a directional bias at differential loop anchors; gained loops are associated with increased expression of genes oriented away from the center of the loop, while lost loops were often accompanied by high levels of transcription with the loop boundaries themselves. Taken together, these results are consistent with a reciprocal relationship in which loops can facilitate increased transcription by connecting promoters to distal enhancers while high levels of transcription can impede loop formation.

Molecular quantitative trait loci (QTLs) generate mechanistic hypotheses explaining how genetic variation impacts brain traits, but gene regulatory effects in bulk-post mortem brain tissues are undetected at many brain trait-associated loci. We hypothesized that the function of genetic variants may only be apparent in specific contexts, such as during stimulation of developmental signaling pathways. We measured chromatin accessibility and gene expression after activation of the canonical Wnt pathway in primary human neural progenitor cells from 82 donors. Brain structure and neuropsychiatric disorder-associated variants were enriched within Wnt-responsive regulatory elements. Thousands of context-specific molecular QTLs were identified during Wnt stimulation. Context-specific molecular QTLs increased the number of colocalizations by up to 70% and led to the nomination of developmental gene regulatory mechanisms underlying complex brain traits.