CHROMATIN ACCESSIBILITY IN NEUROPSYCHIATRIC DISORDERS

Abstract

Epigenetic mechanisms have recently emerged as critical regulators of brain function in health and disease. By controlling the accessibility and the expression of specific genes, these pathways can mediate transient and long-lasting changes in neuronal function in both physiological and pathological contexts. Due to the extreme complexity of the epigenetic regulatory landscape, emerging methods that directly assay chromatin accessibility are of particular interest. Here, I review recent insights gained on open and closed chromatin states in the brain, with emphasis on neuropsychiatric disorders. These advances generated an invaluable wealth of information that can help us better understand gene regulation in the brain and its impairments that contribute to the development of disease.

Introduction

Epigenetic mechanisms in the brain play a critical role in the regulation of complex behaviors (1). In part due to their dynamic nature and rapid turnover, these pathways regulate gene expression in response to environmental stimuli, enabling rapid adaptations and survival. Fluctuations of nutrient availability, for example, directly influence epigenetic regulation via the altered levels of metabolites that serve as substrates and cofactors for epigenetic enzymes (2). Further, histone acetylation and other epigenetic modifications have long been implicated in driving transcriptional changes that underlie learning and memory in physiological and pathological contexts (3). Considering the central role of these pathways in behavioral adaptations and neuronal function in general, a better understanding of epigenetic regulation in the central nervous system is of critical interest for the field of neuroscience.

In recent years, tremendous progress has been made in the genome-wide mapping of the epigenetic landscape in the brain and other tissues. This includes long-standing multi-site collaborations, such as the first, second and third phases of the ENCODE (Encyclopedia of DNA Elements) Project (4–6), the NIH Roadmap Epigenomics Program (7, 8) or the Genotype-Tissue Expression (GTEx) Project (9, 10). These efforts characterized RNA transcription, chromatin structure and histone post-translational modifications, DNA methylation, chromatin looping and occupancy by transcription factors and RNA-binding proteins in an unprecedented detail.

An important challenge in this endeavor is the extremely complex nature of epigenetic regulation. During the past decade, exploration of epigenetic mechanisms revealed that transcription occurs in the presence of an exceedingly complex mixture of modifications, which likely serve different roles in different contexts (11). Consequently, studying single or even multiple post-translational histone modifications and other epigenetic pathways can give only limited insight with respect to functional outcomes. Increasing focus has thus been given to methods geared towards directly assessing the open or closed state of chromatin. Combined with assays of histone and DNA modifications, chromatin accessibility mapping could provide additional insights and generate a more complete picture of gene regulation in the brain.

Various techniques to profile chromatin accessibility have existed for decades. A robust and well-established early method to assign putative functional elements in the genome was the identification of loci that were hypersensitive to DNase I (12, 13) or micrococcal nuclease (MNase) digestion (14). The basis for these assays is that due to spatial constraints, closed chromatin regions are less accessible for these enzymes compared to open chromatin regions. Subsequently, DNase and MNase digestion have been combined with high-throughput sequencing to map open chromatin-regions genome-wide and with high resolution (15, 16). A related method, FAIRE (formaldehyde-assisted isolation of regulatory elements), relies on differences in crosslinking efficiency between open and closed regions (17). DNA in open, nucleosome-depleted regions is cross-linked less efficiently to protein compared to closed regions. Following shearing and subsequent phenol-chloroform extraction, un-crosslinked DNA will segregate to the aqueous phase, while DNA cross-linked to protein will be found between the aqueous and organic phases (17). DNase-seq, MNase-seq and FAIREseq proved invaluable tools to understand specific aspects of gene regulation such as open or closed state of regulatory elements, promoters, enhancers and insulators, or nucleosome positioning and transcription factor binding. ATACseq (assay for transposase accessible chromatin coupled with high-throughput sequencing) is a relatively recent but perhaps the most widely used method (18, 19). Similarly to the previous assays, ATACseq relies on the increased accessibility of open chromatin regions to an enzyme. Specifically, ATACseq uses hyperactive Tn5 transposase with no known sequence specificity, which cuts DNA at all regions accessible for the enzyme and simultaneously ligates adapters for high-throughput sequencing. Consequently, sequencing adapters are preferentially inserted at open chromatin regions, generating DNA fragments that can be amplified for high-throughput sequencing. This dual role of Tn5 enables a relatively simple and streamlined protocol, making this technique particularly rapid, sensitive and reliable (18). Compared to previous methods, ATACseq can be completed in relatively few steps and is thus less prone to loss of material. This allows genome-wide assaying of open or closed state of chromatin within hours and using limited starting material. Consequently, the widespread adaptation of ATACseq has significantly facilitated the efforts to map chromatin accessibility in the brain.

Brain chromatin accessibility in preclinical models

Much of what we know about the neural circuits and molecular mechanisms underlying neuropsychiatric disorders has emerged from preclinical research. Not surprisingly, most studies of genome-wide chromatin accessibility to date have also focused on animal models. Comprehensive atlases of open chromatin regions and gene expression have been generated with high resolution across hippocampal and cortical regions in various model organisms, for example rhesus macaque brains (20). In the mouse visual cortex, Gray et al. identified chromatin regions that were accessible in a cortical layer-specific manner (21). This was achieved using transgenic lines expressing a reporter in cortical layer-specific cells. By performing ATACseq on sorted populations, the authors were able to identify regions that were differentially accessible in specific layers. Genes and regulatory elements located in such regions are likely play an important role in the maintenance of layer-specific cell identity (21).

To further illustrate the potential power of this approach, an important question recently considered was how transient fluctuations in neuronal activity lead to long-lasting changes of gene expression, which underlie synaptic plasticity, learning and memory. To address this, Su et al. compared open chromatin regions in mouse dentate gyrus neurons before and after stimulation of neuronal activity and found dynamic regulation of chromatin accessibility. Stimulation led to early opening of active enhancers as well as binding sites for known immediate early genes including c-Fos (22). Using RNAseq, the authors showed that changes in accessibility were also accompanied by increased or decreased gene expression. Open chromatin regions were highly dynamic following stimulation, with 60% to 95% of activity-induced accessibility changes returning to baseline within 4 and 24 hours, respectively (22). Genes associated with more open chromatin were significantly enriched in pathways related to synaptic transmission, suggesting an important role in long-lasting synaptic adaptations in response to stimulation (22). Indeed, Koberstein and colleagues identified over 2000 differentially accessible regions, mostly in promoters, that were regulated during learning in the mouse hippocampus (23). The affected regions were primarily intronic, influenced gene expression and alternative splicing during memory consolidation and retrieval in a complex manner, and were enriched in known autism spectrum disorder risk genes (23). Further characterizing these dynamic chromatin changes could help identify master regulators that translate transient changes in neuronal activity into persistent alterations of gene expression and neuronal function and thus contribute to learning and memory.

Chromatin accessibility profiling could also advance our understanding of how regulatory networks are affected by disease. Recently, Song and colleagues integrated multi-omic datasets encompassing chromatin interactions (Hi-C), open chromatin regions (ATACseq) and transcriptomes (RNAseq) in four functionally distinct cell types: induced pluripotent stem cell (iPSC)-derived excitatory and lower motor neurons, iPSC-derived hippocampal dentate gyrus-like neurons and primary astrocytes (24). Using this strategy, the authors identified hundreds of thousands of long-range interactions between promoters and distal regulatory regions, linking distal elements to their target genes and pinpointing promoter interacting regions that contain putative regulatory SNPs (single nucleotide polymorphisms) for key genes associated with disease (24).

Open chromatin profiling of the human brain

By utilizing preclinical models, tremendous advances have been made in our understanding of epigenetic processes and their contribution to neuronal function. Most of these advances, however, have not been effectively translated into improvements of clinical care for various reasons. Among others, there is still an overwhelming need for the development of translational animal models that closely reflect the human condition. Emphasizing this need, recent critics have pointed to a high rate of failure of neuropsychiatric drug development due to lack of efficacy in clinical studies (25, 26). Some even argued that current animal models of disease are not to be relied upon to predict treatment efficacy in humans (27). This lack of translational validity is due to a number of factors including the polygenicity of human disease, the reductive nature of most currently available animal models and the evolutionary distance of critical brain regions. In addition, there is a dearth of direct knowledge regarding the epigenetic landscape of the human brain, which necessitates further studies aimed at the molecular profiling of the post-mortem human brain (28). Such efforts are of critical importance for the field as they significantly advance our biological understanding in an unbiased manner and have the potential to guide the development of improved preclinical models as well as to identify novel therapeutic targets without the need for a priori hypotheses. Hence, genome-wide chromatin accessibility mapping of the human brain is of primary interest. These studies are inherently challenging as flash freezing of tissue can adversely impact nuclear integrity and chromatin structure (29). To address this, protocols for ATAC-compatible cryopreservation of neuronal cells (29) as well as improved ATAC protocols that enable profiling of frozen tissue with reduced background (30) have been developed.

Recently, ATACseq was used to generate a comprehensive map of open chromatin regions in the healthy human brain (31). Neuronal and non-neuronal nuclei were isolated from over a dozen distinct brain regions. Chromatin accessibility showed marked differences between neuronal and non-neuronal populations, with enrichment for known cell type-specific markers. Neuronal chromatin accessibility also exhibited substantial variability across brain regions while open regions in non-neuronal nuclei were more uniform across the brain. The most pronounced differences were observed between neurons of the neocortex, hippocampus, thalamus, and striatum. Numerous cell type- and region-specific open chromatin regions were identified that could inform about unique transcriptional regulatory networks. Indeed, transcription factor footprinting analysis pointed to cell type differences in the regulation of gene expression and identified protein-coding and long non-coding RNAs with cell type and brain region specificity. Intriguingly, cell- and region-specific differentially accessible chromatin regions were enriched for genetic variants associated with neuropsychiatric traits. These findings emphasize the importance of conducting cell type- and region-specific epigenetic studies to elucidate regulatory and disease-associated mechanisms in the human brain (31).

Chromatin accessibility profiling was also used to define non-coding regions regulating gene expression in the developing human cortex. Torre-Ubieta and colleagues (32) generated a high-resolution map of non-coding elements by integrating chromatin accessibility and gene expression in the developing germinal zone and cortical plate. Target genes of human-specific enhancers were found to be enriched in outer radial glia, a cell type linked to human cortical development. Common genetic variants associated with neuropsychiatric disease risk were significantly enriched within these regions, emphasizing their relevance for cognitive functions in the adult (32). More recently, Markenscoff-Papadimitriou et al. generated an atlas of open chromatin from nine regions of the mid-gestation human telencephalon, as well as upper and deep layers of the prefrontal cortex (33). By defining open chromatin regions with temporal, regional and laminar specificity, the authors were able to identify numerous developmentally regulated functional elements. Chromatin accessibility at these loci correlated with transcriptional activity across brain regions at different gestational stages and enhancer function was further validated in vitro as well as using transgenic mice (33).

Considering the extreme cellular heterogeneity of the brain, a particularly exciting development is the advent of single cell genomics, which enables chromatin mapping at cellular resolution. These methods can also be used to identify cell types in heterogenous tissues de novo. For example, in the developing mouse cortex, single-nuclei ATACseq delineated over 20 distinct neuronal and non-neuronal cell populations, defined cell type-specific regulatory sequences and outlined potential master regulators of gene expression that govern developmental changes in cortical cell composition (34). More recently, single cell ATACseq was used to generate chromatin accessibility profiles of close to a million single cells from over 50 human fetal samples in different tissues and at different developmental stages (35). Domcke and colleagues observed over 1 million open regions spanning close to 20% of the human genome and identified cell type-specific links between candidate regulatory elements and genes based on co-accessibility. Further, the enrichment of transcription factor motifs at open chromatin regions revealed both known and unknown regulators of cell fate specification and maintenance. These transcription factors were putatively assigned as activators or repressors depending on whether gene expression and the accessibility of their cognate motifs were positively or negatively correlated across cell types (35). To identify variants that affect local chromatin accessibility, the authors also assessed allelic imbalance of ATACseq reads at heterozygous loci. Variants that affect chromatin accessibility in cis lead to significant imbalance with most sequencing reads coming from the more accessible allele. This study identified over 500 genetic variants with significant allelic imbalance, which could help pinpoint functional variants in non-coding regions. Further, Domcke et al. reported cell type-specific enrichment of heritability for complex traits including common human diseases (35). In a similar study, Song et al. characterized cell type-specific open chromatin peaks, chromatin-chromatin interactions and transcriptomes for radial glia, intermediate progenitor cells, excitatory neurons, and interneurons isolated from mid-gestational samples of the human cortex (36). Promoters of lineage-specific genes tended to exhibit increased openness and increased number of chromatin-chromatin interactions, features that might contribute to the fine-tuning of lineage-specific transcription and development (36). Lineage-specific enrichment of transcription factor motifs in open chromatin regions can also be used to identify potential master regulators of neuronal development. Using this strategy, POU2F1 (POU class 2 homeobox 1) was proposed as a putative regulator of excitatory neuron development (35).

While most developmental studies to date have focused on the neocortex, other regions, such as the hippocampus are also of critical interest considering its role in the consolidation of long-term memories. Zhong and colleagues used single-cell RNAseq and ATACseq to characterize the cell types, lineage, chromatin features and transcriptional regulation of the developing human hippocampus (37). Chromatin accessibility and gene expression displayed marked spatial specification in this brain region. The authors defined close to 50 cell types and their developmental trajectories, and identified the migrating paths and cell lineages of hippocampal progenitors, as well as regional markers of Cornu Ammonis and dentate gyrus neurons (37). This study found important similarities between the chromatin accessibility and gene expression features of the developing human hippocampus and that of the early postnatal mouse hippocampus, supporting the translational value of carefully selected rodent models. On the other hand, the authors also identified numerous loci that were open and expressed in a primate-specific manner and may thus have played an important evolutionary role. These included, for example, members of the neuroblastoma breakpoint protein family (NBPF). NBPF proteins contain a repeated domain that has been previously linked to brain evolution and structural complexity (38). In fact, transient over-expression of NBPF1 in the developing mouse hippocampus resulted in a marked increase of the number of dentate gyrus neurons (37).

These datasets provide a valuable resource for understanding neurodevelopment in humans and in model organisms. Further, similar strategies can and have been used in the human brain to study pathways that contribute to disease. This is particularly exciting for neuropsychiatric disorders, where efficacious treatment options remain severely lacking.

Chromatin landscapes in neuropsychiatric disorders

While genome-wide association studies have identified hundreds of loci associated with disease risk in neuropsychiatric disorders, very few of these fall in coding regions of the genome. Consequently, the question remains how non-coding variants contribute to the development of disease. Describing chromatin state at these loci could prove particularly useful to address this (Figure 1). Specifically, chromatin accessibility profiles of the brain can provide insight regarding the functional state of promoters, enhancers and other regulatory elements in different cell types, which could help illuminate their role in neuropsychiatric disease.

Figure 1. Chromatin accessibility mapping can identify functional variants that contribute to neuropsychiatric disease.

A large fraction of genetic variants (stars) that were linked to neuropsychiatric disease risk are located outside of protein-coding regions. Nevertheless, non-coding variants can contribute to disease by altering the activity of enhancers and other regulatory elements, as well as by disrupting chromatin-chromatin interactions between genes and their regulatory regions. Variants located in open chromatin regions are more likely to be functional variants affecting gene expression compared to those in closed regions. Hence, profiling open chromatin regions in the brain can help identify loci of interest that might play a significant role in gene regulation and the etiology of neuropsychiatric disorders.

Early reports show that most open chromatin regions are differentially accessible between neurons and non-neuronal cell populations, and are enriched for known cell type-specific genes, promoters and enhancers (39). Fullard and colleagues also reported that, compared to non-neuronal cells, accessible regions in neurons are highly evolutionarily conserved and enriched in distal regulatory elements (39).

Across most studies of neuropsychiatric disease to date, differential accessibility of specific regions compared to matched controls is rare. For example, in a large cohort of over 130 schizophrenia cases and over 130 controls, Bryois and colleagues reported only a few differentially accessible regions with genome-wide significance in the dorsolateral prefrontal cortex (40). This could in part be due to the complex, multifactorial nature of neuropsychiatric disorders. Increasing focus is thus given to initiatives such as RDoC (Research Domain Criteria), which aim to identify and describe deficient functional domains independently from DSM-5 diagnosis (41). The hope is that this new nosology will result in biologically-based disease definitions, which are more closely linked to underlying neurobiological impairments. Another potential reason for the lack of chromatin accessibility differences with genome-wide significance in many studies is the extreme cellular heterogeneity of the brain. As discussed, the recent emergence of single-cell genomics may facilitate the identification of regions that are differentially accessible in neuropsychiatric patients compared to healthy controls, but only in specific cell populations. These regions that were previously masked in homogenate tissues could have important functional contributions to disease.

Alternative strategies to assess the functional relevance of open chromatin regions in bulk ATACseq datasets have been utilized with some success. Identifying open chromatin regions enriched for disease risk variants seems particularly promising to pinpoint known and putative regulatory regions that contribute to the etiology of neuropsychiatric disorders. For example, in the above referenced schizophrenia study, open regions in neurons were highly enriched for schizophrenia SNPs. This enrichment was even more pronounced in accessible regions that overlapped with evolutionary conserved domains, suggesting that conservation of specific sequences can further emphasize the functional relevance of risk variants and putative regulatory elements (40). Transcription factor footprinting analysis is another promising approach. Using this strategy, Fullard and colleagues were able to assign putative functional roles to a number of non-coding schizophrenia risk variants. An interesting example is rs10750450, a SNP proximal to the SNX19 (Sorting Nexin 19) gene, increased expression of which was previously linked to schizophrenia (42). This SNP led to elevated transcriptional activity in vitro, emphasizing its potential functional contribution to gene expression and disease (39).

More recently, Hauberg et al. used ATACseq to profile the chromatin accessibility in four distinct populations (glutamatergic neurons, GABAergic neurons, oligodendrocytes, and microglia/astrocytes) in three different regions (anterior cingulate cortex, dorsolateral prefrontal cortex, and primary visual cortex) of the post-mortem human brain (43). As also reported by others, chromatin accessibility varied to a great extent by cell type and brain region, with, interestingly, the largest variability observed in glutamatergic neurons. Similarly to the Bryois et al. study, accessible chromatin regions were enriched for neuropsychiatric and in particular schizophrenia risk variants. The strongest enrichment was observed in glutamatergic neurons, emphasizing their important role in the etiology of this disease (43). A similar distinction between neurons and other cell types was also noted in other studies, where risk variants for psychiatric disorders were primarily enriched in neuron-specific transcriptional enhancers and promoters (44). This is not necessarily the case for all neurological disorders —risk variants for Alzheimer’s disease, for example, were shown to be most enriched in microglia-specific enhancers (44). These and other studies illustrate the exciting potential of chromatin accessibility profiling to better understand the genetic and epigenetic basis of neuropsychiatric disease. As part of the CommonMind Consortium, high quality curated functional genomic datasets including chromatin accessibility profiles are now publicly available from thousands of individuals with schizophrenia or bipolar depression (45). Such repositories are an exceptional resource for the field of neuropsychiatry in its ongoing quest to identify functional risk variants, pathways that contribute to disease, and novel targets that could be utilized in future therapeutic approaches.

Due to their dynamic nature, epigenetic pathways are particularly responsive to environmental influences. Following exposure to noxious stimuli, such mal-adaptations can play an important role in the development of neuropsychiatric disease. For example, exposure to abused substances can strongly impact epigenetic regulation in the brain, and thereby affect chromatin accessibility, gene expression and neuronal function. This has been shown for a plethora of substances including both prescription and illicit drugs (46–48). A striking example is alcohol, breakdown of which leads to an accumulation of circulating acetate that is rapidly incorporated into brain histone acetylation. This mechanism was recently shown to fuel the expression of genes that encode memories for drug-related environmental cues, and thereby contribute to the development of alcohol use disorder (49). Chronic heroin use also induces histone acetylation, resulting in a more open state of chromatin at genes related to glutamatergic signaling, which have been implicated in the development of substance use disorders (50). The functional importance of these epigenetic changes is further underlined by the fact that bromodomain inhibitors (which block the molecular read-out of acetylated histones) reduce heroin self-administration and cue-induced drug-seeking behavior in animal models (50).

More recently, ATACseq was used to assess dorsal striatal chromatin accessibility in human heroin users and controls (51). This brain region is involved in the regulation of reward, inhibitory control, motivation and goal-directed behaviors, which are strongly implicated in substance use disorders. Via fluorescence-assisted nuclei sorting, quantitative open chromatin profiles were generated in neuronal and non-neuronal populations from heroin users and matched controls. This strategy identified a putative regulatory region near the FYN tyrosine kinase gene in neurons, which was the most significantly affected by heroin in that the largest proportion of signal variability was attributable to disease (51). The changed accessibility of this region also resulted in increased expression and activity of the FYN kinase in heroin users and rats that were trained to self-administer heroin. FYN was shown to phosphorylate Tau and has been previously linked to tauopathies such as Alzheimer’s disease (52, 53). Increased phosphorylation of Tau was also described in the brain of human heroin users, and might play a role in cognitive impairments associated with the use of this drug. Heroin-induced activity of FYN could hence be an important mechanism by which these lesions are established. Indeed, Tau phosphorylation was elevated at a residue directly targeted by FYN both in heroin users and in primary neurons chronically treated with morphine. Saracatinib, a FYN inhibitor currently in clinical trials for Alzheimer’s disease (54), acutely attenuated addiction-like behaviors in rats self-administering heroin, similar to effects observed with siRNA-mediated knockdown of Fyn in the dorsal striatum. These findings suggest that opioid exposure induces neurodegenerative cellular processes associated with epigenetic disturbances and that the use of FYN inhibitors could be promising for heroin medication development (51).

Future directions

These are exciting times for the complex, multi-omic profiling of the brain in health and disease. ATACseq is now used routinely in many laboratories and novel iterations continue to further enhance its capabilities. For example, Omni-ATAC, a modified protocol with improved signal-to-noise ratio, can be used to generate robust and reproducible chromatin accessibility profiles from frozen tissues (30), enabling the future interrogation of archival brain specimens. Pi-ATAC (Protein-indexed ATAC) combines chromatin profiling with proteomics at the single cell resolution, which could be used to link the protein levels and post-translational modifications of specific transcription factors to the accessibility of their DNA binding motifs (55). Single-cell ATACseq will be used to interrogate chromatin accessibility in specific cell populations. Along with single-cell transcriptomic approaches, these methods will have the potential to uncover cell type-specific differences in chromatin accessibility and gene regulation, as well as to assess the contribution of specific cell types and subpopulations to neuropsychiatric disease. While this review focused on chromatin accessibility, additional tools and resources are being rapidly developed that can further inform about functional pathways contributing to disease. Methods to assay genome wide chromatin-chromatin interactions in an unbiased manner (Hi-C) (56) or in the context of transcription factor binding and histone modifications (HiChIP) (57) will help define interactomes between promoters, enhancers, gene bodies and intergenic open chromatin regions and thus link known and putative distal regulatory elements to the genes they regulate. Together, these efforts will significantly advance our understanding of gene regulation in the brain and its impairments in the context of neuropsychiatric disorders.

Highlights.

• Chromatin accessibility profiling can help understand gene regulation in the brain

• Recent advances enable high resolution profiling of human and preclinical samples

• An increasing amount of information is available in healthy and diseased brains

• Differentially accessible regions are of significant interest for disease etiology