Droplet Paired-Tag: Describing the Epigenome and Transcriptome of the Brain at Single-Cell Resolution

A vast team of epigenetic researchers led by Bing Ren and Lei Chang recently employed Droplet Paired-Tag from Epigenome Technologies to simultaneously profile the transcriptome and active (H3K27ac) and repressive (H3K27me3) histone modifications across eighteen brain regions in three neurotypical male donors, which spanned the cerebral cortex (encompassing fifteen regions including frontal, temporal, parietal, and occipital lobes), cerebellum, and pons and took in 160 cell types.

Paired-Tag technology supports the generation of joint epigenetic and transcriptomic profiles at single-cell resolution and detects histone modifications and RNA transcripts in individual nuclei with efficiency comparable to single-nucleus RNA-seq/ChIP-seq assays. Excitingly, this new bioRxiv preprint article also reports the integration of Droplet Paired-Tag data with matched chromatin accessibility, DNA methylation, 3D genome architecture, and spatial transcriptome data to annotate over 500,000 candidate cis-regulatory elements (cCREs; such as gene enhancers), define dynamic chromatin states with cell-type- and regional specificity, connect chromatin states to gene regulatory networks, and much, much more (Xie et al.)! Overall, these analyses enabled the creation of a comprehensive single-cell multimodal atlas of the adult human brain that captured fine-grained heterogeneity and incorporated developmentally distinct populations. We now bring you some of the highlights of this exciting new study!

How Droplet-Paired Tag Helped to Create a Single-Cell Atlas of the Human Brain: The Highlights

An initial analysis of transcriptomic data alone resolved three major cellular classes and 35 cell subclasses in the brain, while subsequent analysis of histone modification profiles recapitulated the major cellular hierarchy and provided additional regulatory resolution, suggesting that epigenetic data may provide additional lineage-associated regulatory information beyond transcriptomic data. Of note, an integrative analysis of both datasets provided even further resolution, enabling the identification of rare, region-specific neuronal cell types. A focus on cluster-specific marker genes revealed that H3K27ac levels at gene transcription start sites positively correlated with transcription and that H3K27me3 levels at gene bodies correlated with gene repression and marked lineage-restricted loci.

Next, the authors employed chromatin-state annotation to resolve functional cCRE classes in the human brain by leveraging datasets from the same donor cohort (Li et al. and Tian et al.) and integrating single-nucleus H3K27ac and H3K27me3 profiles with matched single-cell chromatin accessibility and DNA methylation profiles. This strategy grouped inferred chromatin states into categories such as open (primarily structural features), active (three forms reflecting proximity to the peak signal, mostly promoters and enhancers), and Polycomb-repressed regions with concurrently low or high DNA methylation levels. An examination of loci across transcriptional/epigenetic subclasses highlighted the specificity of chromatin-state segmentation, supporting a model in which lineage marker activation and alternative-fate program repression function together to encode cell identity. Applying the developed framework to known human brain cCREs enabled the functional stratification of over 500,000 cCREs, annotating a large proportion of accessible cCREs within each subclass as "Active" or "Open". Furthermore, consistent with functional relevance, Active (and not Open) cCREs displayed a substantial enrichment for neuropsychiatric disorder-associated variants in genome-wide association studies (GWAS). While H3K27me3-associated repressive elements had remained under-characterized in the human brain, a new study identified 150,372 H3K27me3-modified elements across 26 cell subclasses and suggested that Polycomb-mediated repression helps maintain cell identity and represses transposable element expression/activity in terminally differentiated cells. Overall, the chromatin-state annotation helped to describe the functional heterogeneity of accessible cCREs, revealing how activation and Polycomb-mediated repression jointly regulate cell-type-specific regulatory programs in the adult human brain

The authors next revealed how cell-type-resolved enhancer-gene links and gene regulatory networks connect chromatin state to transcription and genetic risk by first integrating H3K27ac profiles with single-cell/single-nucleus chromatin accessibility and 3D chromatin interaction data to infer functional enhancer-gene relationships. The results demonstrated that enhancer-gene links, anchored in chromatin activities and the 3D genome, captured main features of the cell-type-specific regulation that underlies human brain cell identity. They then integrated enhancer-gene links with transcription factor binding and transcriptomic data to construct active (H3K27ac-associated) and repressive (H3K27me3-associated) gene regulatory networks and generate an overview of transcriptional regulation. The results confirmed the previously described model, supporting a "dual regulatory logic" in which transcription factors promote certain lineage programs while simultaneously antagonizing alternative-fate programs. The integration of enhancer-gene links and gene regulatory networks also provided a means to interpret disease-associated non-coding variants; fascinatingly, the study identified significant associations between 24 cell subclasses and 17 major neuropsychiatric traits using H3K27ac-modified cCREs. The team then trained a sequence-based deep learning model to predict cell-type-specific epigenetic signals and the allelic impact of variants, thereby estimating effects at nucleotide resolution. Encouragingly, the model achieved high performance in predicting histone modifications, chromatin accessibility, and gene expression across cell subclasses. Together, these analyses demonstrated that joint transcriptomic and chromatin-state analyses helped construct cell-type-resolved regulatory maps that connect distal CREs to target genes, define transcription factor networks, and translate GWAS associations into cell-type-resolved regulatory mechanisms.

The subsequent analysis of epigenetic encoding of structural and areal positional identity in the adult human brain demonstrated that histone modification landscapes encode discrete structural identities across brain structures and continuous areal specialization within the cortex. The exploration of positional regulatory programs not fully captured by transcriptomic analysis provided a framework for understanding the recording and maintenance of spatial organization and developmental history in the adult human brain. Furthermore, their analyses revealed that spatially organized gene regulatory programs that integrate transcriptional activation with Polycomb-mediated repression support laminar identity in the adult human cortex. Overall, this projection of epigenetic heterogeneity onto physical space revealed how developmental history and positional information persist in the adult brain and provided a framework linking chromatin state to cortical architecture.

The authors next integrated chromatin loop and H3K27ac/H3K27me3 data to determine how chromatin topology reflected regulatory state; interestingly, this functional annotation of chromatin loops uncovered subclass-specific active and repressive 3D genome topologies. Their strategy permitted the classification of active and repressive loops, with active loops dominating the annotated interactome across neuronal and non-neuronal subclasses. Active loops mostly represented enhancer-promoter communication, with target genes associated with neuronal functions; meanwhile, repressive loops spanned significantly longer genomic distances, contacted promoters less frequently, and preferentially connected with transcriptionally silent loci. Interestingly, lineage-specific folding analysis revealed that H3K27me3-associated chromatin conformations sequestered inappropriate regulatory programs via repressive 3D contacts. The team also identified a highly cell-type-specific class of "super long-range" loops resembling Polycomb-associated long-range interactions. Annotation of these loops by the epigenetic state of their anchors in neurons suggested that these interactions encoded a durable "developmental memory"; overall, the sequestration of subsets of developmentally active genes in post-mitotic neurons into super-long-range silencing loops likely ensures robust long-term repression of early-stage regulatory programs that must remain silent in mature cell types.

Finally, the study integrated human data with matched Droplet Paired-Tag datasets from the mouse cortex to distinguish conserved and human-specific epigenetic programs; overall, this analysis revealed evolutionary conservation and divergence in active and repressive regulation. Active programs establishing core cellular identity exhibit strong conservation across mammals and are governed by a constrained level of transcription factor interactions. Meanwhile, the greater divergence in repressive programs suggested that species-specific evolution more often proceeds through the rewiring of repressive mechanisms than through changes in activation circuits, which may provide a flexible substrate for species-specific adaptations in brain development, plasticity, and disease susceptibility.

Conclusions: Highlighting the Power of Paired-Tag

The authors note that moving beyond mere catalogs of molecular features and defining how gene regulatory programs constrain cellular identity, plasticity, and disease vulnerability remains a central challenge in human brain genomics. The team hope to have taken a step in this direction by jointly profiling the transcriptome with active and repressive histone modification profiles at single-cell resolution with the help of Droplet Paired-Tag, thereby creating a functional reference for the regulatory architecture of the adult human brain, and then integrating this data with 3D genome organization, spatial information, and genetic variants to understand better the organizing principles that transcriptomes or chromatin accessibility cannot resolve on their own.

For more on this fascinating study, head over to bioRxiv, and for more on Droplet Paired-Tag and the related products and services provided by Epigenome Technologies, please explore our website!