ERG activates a stem-like proliferation-differentiation program in prostate epithelial cells with mixed basal-luminal identity.

To gain insight into how ERG translocations cause prostate cancer, we performed single cell transcriptional profiling of an autochthonous mouse model at an early stage of disease initiation. Despite broad expression of ERG in all prostate epithelial cells, proliferation was enriched in a small, stem-like population with mixed-luminal basal identity (called intermediate cells). Through a series of lineage tracing and primary prostate tissue transplantation experiments, we find that tumor initiating activity resides in a subpopulation of basal cells that co-express the luminal genes Tmprss2 and Nkx3.1 (called BasalLum) but not in the larger population of classical Krt8+ luminal cells. Upon ERG activation, BasalLum cells give rise to the highly proliferative intermediate state, which subsequently transitions to the larger population of Krt8+ luminal cells characteristic of ERG-positive human cancers. Furthermore, this proliferative population is characterized by an ERG-specific chromatin state enriched for NFkB, AP-1, STAT and NFAT binding, with implications for TF cooperativity. The fact that the proliferative potential of ERG is enriched in a small stem-like population implicates the chromatin context of these cells as a critical variable for unmasking its oncogenic activity.

Single-cell multi-ome regression models identify functional and disease-associated enhancers and enable chromatin potential analysis.

We present a gene-level regulatory model, single-cell ATAC + RNA linking (SCARlink), which predicts single-cell gene expression and links enhancers to target genes using multi-ome (scRNA-seq and scATAC-seq co-assay) sequencing data. The approach uses regularized Poisson regression on tile-level accessibility data to jointly model all regulatory effects at a gene locus, avoiding the limitations of pairwise gene-peak correlations and dependence on peak calling. SCARlink outperformed existing gene scoring methods for imputing gene expression from chromatin accessibility across high-coverage multi-ome datasets while giving comparable to improved performance on low-coverage datasets. Shapley value analysis on trained models identified cell-type-specific gene enhancers that are validated by promoter capture Hi-C and are 11× to 15× and 5× to 12× enriched in fine-mapped eQTLs and fine-mapped genome-wide association study (GWAS) variants, respectively. We further show that SCARlink-predicted and observed gene expression vectors provide a robust way to compute a chromatin potential vector field to enable developmental trajectory analysis.

An epigenetic barrier sets the timing of human neuronal maturation.

The pace of human brain development is highly protracted compared with most other species1-7. The maturation of cortical neurons is particularly slow, taking months to years to develop adult functions3-5. Remarkably, such protracted timing is retained in cortical neurons derived from human pluripotent stem cells (hPSCs) during in vitro differentiation or upon transplantation into the mouse brain4,8,9. Those findings suggest the presence of a cell-intrinsic clock setting the pace of neuronal maturation, although the molecular nature of this clock remains unknown. Here we identify an epigenetic developmental programme that sets the timing of human neuronal maturation. First, we developed a hPSC-based approach to synchronize the birth of cortical neurons in vitro which enabled us to define an atlas of morphological, functional and molecular maturation. We observed a slow unfolding of maturation programmes, limited by the retention of specific epigenetic factors. Loss of function of several of those factors in cortical neurons enables precocious maturation. Transient inhibition of EZH2, EHMT1 and EHMT2 or DOT1L, at progenitor stage primes newly born neurons to rapidly acquire mature properties upon differentiation. Thus our findings reveal that the rate at which human neurons mature is set well before neurogenesis through the establishment of an epigenetic barrier in progenitor cells. Mechanistically, this barrier holds transcriptional maturation programmes in a poised state that is gradually released to ensure the prolonged timeline of human cortical neuron maturation.

An encyclopedia of enhancer-gene regulatory interactions in the human genome.

Identifying transcriptional enhancers and their target genes is essential for understanding gene regulation and the impact of human genetic variation on disease1-6. Here we create and evaluate a resource of >13 million enhancer-gene regulatory interactions across 352 cell types and tissues, by integrating predictive models, measurements of chromatin state and 3D contacts, and largescale genetic perturbations generated by the ENCODE Consortium7. We first create a systematic benchmarking pipeline to compare predictive models, assembling a dataset of 10,411 elementgene pairs measured in CRISPR perturbation experiments, >30,000 fine-mapped eQTLs, and 569 fine-mapped GWAS variants linked to a likely causal gene. Using this framework, we develop a new predictive model, ENCODE-rE2G, that achieves state-of-the-art performance across multiple prediction tasks, demonstrating a strategy involving iterative perturbations and supervised machine learning to build increasingly accurate predictive models of enhancer regulation. Using the ENCODE-rE2G model, we build an encyclopedia of enhancer-gene regulatory interactions in the human genome, which reveals global properties of enhancer networks, identifies differences in the functions of genes that have more or less complex regulatory landscapes, and improves analyses to link noncoding variants to target genes and cell types for common, complex diseases. By interpreting the model, we find evidence that, beyond enhancer activity and 3D enhancer-promoter contacts, additional features guide enhancerpromoter communication including promoter class and enhancer-enhancer synergy. Altogether, these genome-wide maps of enhancer-gene regulatory interactions, benchmarking software, predictive models, and insights about enhancer function provide a valuable resource for future studies of gene regulation and human genetics.

Reprogramming tumor-associated macrophages to outcompete endovascular endothelial progenitor cells and suppress tumor neoangiogenesis.

Tumors develop by invoking a supportive environment characterized by aberrant angiogenesis and infiltration of tumor-associated macrophages (TAMs). In a transgenic model of breast cancer, we found that TAMs localized to the tumor parenchyma and were smaller than mammary tissue macrophages. TAMs had low activity of the metabolic regulator mammalian/mechanistic target of rapamycin complex 1 (mTORC1), and depletion of negative regulator of mTORC1 signaling, tuberous sclerosis complex 1 (TSC1), in TAMs inhibited tumor growth in a manner independent of adaptive lymphocytes. Whereas wild-type TAMs exhibited inflammatory and angiogenic gene expression profiles, TSC1-deficient TAMs had a pro-resolving phenotype. TSC1-deficient TAMs relocated to a perivascular niche, depleted protein C receptor (PROCR)-expressing endovascular endothelial progenitor cells, and rectified the hyperpermeable blood vasculature, causing tumor tissue hypoxia and cancer cell death. TSC1-deficient TAMs were metabolically active and effectively eliminated PROCR-expressing endothelial cells in cell competition experiments. Thus, TAMs exhibit a TSC1-dependent mTORC1-low state, and increasing mTORC1 signaling promotes a pro-resolving state that suppresses tumor growth, defining an innate immune tumor suppression pathway that may be exploited for cancer immunotherapy.

ChromaFold predicts the 3D contact map from single-cell chromatin accessibility.

The identification of cell-type-specific 3D chromatin interactions between regulatory elements can help to decipher gene regulation and to interpret the function of disease-associated non-coding variants. However, current chromosome conformation capture (3C) technologies are unable to resolve interactions at this resolution when only small numbers of cells are available as input. We therefore present ChromaFold, a deep learning model that predicts 3D contact maps and regulatory interactions from single-cell ATAC sequencing (scATAC-seq) data alone. ChromaFold uses pseudobulk chromatin accessibility, co-accessibility profiles across metacells, and predicted CTCF motif tracks as input features and employs a lightweight architecture to enable training on standard GPUs. Once trained on paired scATAC-seq and Hi-C data in human cell lines and tissues, ChromaFold can accurately predict both the 3D contact map and peak-level interactions across diverse human and mouse test cell types. In benchmarking against a recent deep learning method that uses bulk ATAC-seq, DNA sequence, and CTCF ChIP-seq to make cell-type-specific predictions, ChromaFold yields superior prediction performance when including CTCF ChIP-seq data as an input and comparable performance without. Finally, fine-tuning ChromaFold on paired scATAC-seq and Hi-C in a complex tissue enables deconvolution of chromatin interactions across cell subpopulations. ChromaFold thus achieves state-of-the-art prediction of 3D contact maps and regulatory interactions using scATAC-seq alone as input data, enabling accurate inference of cell-type-specific interactions in settings where 3C-based assays are infeasible.

Discovery of Competent Chromatin Regions in Human Embryonic Stem Cells.

The mechanisms underlying the ability of embryonic stem cells (ESCs) to rapidly activate lineage-specific genes during differentiation remain largely unknown. Through multiple CRISPR-activation screens, we discovered human ESCs have pre-established transcriptionally competent chromatin regions (CCRs) that support lineage-specific gene expression at levels comparable to differentiated cells. CCRs reside in the same topological domains as their target genes. They lack typical enhancer-associated histone modifications but show enriched occupancy of pluripotent transcription factors, DNA demethylation factors, and histone deacetylases. TET1 and QSER1 protect CCRs from excessive DNA methylation, while HDAC1 family members prevent premature activation. This "push and pull" feature resembles bivalent domains at developmental gene promoters but involves distinct molecular mechanisms. Our study provides new insights into pluripotency regulation and cellular plasticity in development and disease. One sentence summary: We report a class of distal regulatory regions distinct from enhancers that confer human embryonic stem cells with the competence to rapidly activate the expression of lineage-specific genes.

Spatial and temporal organization of the genome: Current state and future aims of the 4D nucleome project.

The four-dimensional nucleome (4DN) consortium studies the architecture of the genome and the nucleus in space and time. We summarize progress by the consortium and highlight the development of technologies for (1) mapping genome folding and identifying roles of nuclear components and bodies, proteins, and RNA, (2) characterizing nuclear organization with time or single-cell resolution, and (3) imaging of nuclear organization. With these tools, the consortium has provided over 2,000 public datasets. Integrative computational models based on these data are starting to reveal connections between genome structure and function. We then present a forward-looking perspective and outline current aims to (1) delineate dynamics of nuclear architecture at different timescales, from minutes to weeks as cells differentiate, in populations and in single cells, (2) characterize cis-determinants and trans-modulators of genome organization, (3) test functional consequences of changes in cis- and trans-regulators, and (4) develop predictive models of genome structure and function.

Epiphany: predicting Hi-C contact maps from 1D epigenomic signals.

Recent deep learning models that predict the Hi-C contact map from DNA sequence achieve promising accuracy but cannot generalize to new cell types and or even capture differences among training cell types. We propose Epiphany, a neural network to predict cell-type-specific Hi-C contact maps from widely available epigenomic tracks. Epiphany uses bidirectional long short-term memory layers to capture long-range dependencies and optionally a generative adversarial network architecture to encourage contact map realism. Epiphany shows excellent generalization to held-out chromosomes within and across cell types, yields accurate TAD and interaction calls, and predicts structural changes caused by perturbations of epigenomic signals.

Drivers of heterogeneity in synovial fibroblasts in rheumatoid arthritis.

Inflammation of non-barrier immunologically quiescent tissues is associated with a massive influx of blood-borne innate and adaptive immune cells. Cues from the latter are likely to alter and expand activated states of the resident cells. However, local communications between immigrant and resident cell types in human inflammatory disease remain poorly understood. Here, we explored drivers of fibroblast-like synoviocyte (FLS) heterogeneity in inflamed joints of patients with rheumatoid arthritis using paired single-cell RNA and ATAC sequencing, multiplexed imaging and spatial transcriptomics along with in vitro modeling of cell-extrinsic factor signaling. These analyses suggest that local exposures to myeloid and T cell-derived cytokines, TNF, IFN-γ, IL-1ß or lack thereof, drive four distinct FLS states some of which closely resemble fibroblast states in other disease-affected tissues including skin and colon. Our results highlight a role for concurrent, spatially distributed cytokine signaling within the inflamed synovium.

Microenvironment-Driven Dynamic Chromatin Changes in Glioblastoma Recapitulate Early Neural Development at Single-Cell Resolution.

The tumor microenvironment is necessary for recapitulating the intratumoral heterogeneity and cell state plasticity found in human primary glioblastoma (GBM). Conventional models do not accurately recapitulate the spectrum of GBM cellular states, hindering elucidation of the underlying transcriptional regulation of these states. Using our glioblastoma cerebral organoid model, we profiled the chromatin accessibility of 28,040 single cells in five patient-derived glioma stem cell lines. Integration of paired epigenomes and transcriptomes within the context of tumor-normal host cell interactions was used to probe the gene-regulatory networks underlying individual GBM cellular states in a way not readily possible in other in vitro models. These analyses identified the epigenetic underpinnings of GBM cellular states and characterized dynamic chromatin changes reminiscent of early neural development that underlie GBM cell state transitions. Despite large differences between tumors, a shared cellular compartment made up of neural progenitor-like cells and outer radial glia-like cells was observed. Together, these results shed light on the transcriptional regulation program in GBM and offer novel therapeutic targets across a broad range of genetically heterogenous GBMs. SIGNIFICANCE: Single-cell analyses elucidate the chromatin landscape and transcriptional regulation of glioblastoma cellular states and identify a radial glia-like population, providing potential targets to disrupt cell states and improve therapeutic efficacy.

Novel extragenic genomic safe harbors for precise therapeutic T-cell engineering.

Cell therapies that rely on engineered immune cells can be enhanced by achieving uniform and controlled transgene expression in order to maximize T-cell function and achieve predictable patient responses. Although they are effective, current genetic engineering strategies that use γ-retroviral, lentiviral, and transposon-based vectors to integrate transgenes, unavoidably produce variegated transgene expression in addition to posing a risk of insertional mutagenesis. In the setting of chimeric antigen receptor (CAR) therapy, inconsistent and random CAR expression may result in tonic signaling, T-cell exhaustion, and variable T-cell persistence. Here, we report and validate an algorithm for the identification of extragenic genomic safe harbors (GSH) that can be efficiently targeted for DNA integration and can support sustained and predictable CAR expression in human peripheral blood T cells. The algorithm is based on 7 criteria established to minimize genotoxicity by directing transgene integration away from functionally important genomic elements, maximize efficient CRISPR/Cas9-mediated targeting, and avert transgene silencing over time. T cells engineered to express a CD19 CAR at GSH6, which meets all 7 criteria, are curative at low cell dose in a mouse model of acute lymphoblastic leukemia, matching the potency of CAR T cells engineered at the TRAC locus and effectively resisting tumor rechallenge 100 days after their infusion. The identification of functional extragenic GSHs thus expands the human genome available for therapeutic precision engineering.

SETD2 regulates chromatin accessibility and transcription to suppress lung tumorigenesis.

SETD2, a H3K36 trimethyltransferase, is the most frequently mutated epigenetic modifier in lung adenocarcinoma, with a mutation frequency of approximately 9%. However, how SETD2 loss of function promotes tumorigenesis remains unclear. Using conditional Setd2-KO mice, we demonstrated that Setd2 deficiency accelerated the initiation of KrasG12D-driven lung tumorigenesis, increased tumor burden, and significantly reduced mouse survival. An integrated chromatin accessibility and transcriptome analysis revealed a potentially novel tumor suppressor model of SETD2 in which SETD2 loss activates intronic enhancers to drive oncogenic transcriptional output, including the KRAS transcriptional signature and PRC2-repressed targets, through regulation of chromatin accessibility and histone chaperone recruitment. Importantly, SETD2 loss sensitized KRAS-mutant lung cancer to inhibition of histone chaperones, the FACT complex, or transcriptional elongation both in vitro and in vivo. Overall, our studies not only provide insight into how SETD2 loss shapes the epigenetic and transcriptional landscape to promote tumorigenesis, but they also identify potential therapeutic strategies for SETD2 mutant cancers.

3D Chromatin Dynamics during Innate and Adaptive Immune Memory Acquisition.

Immune cells responding to pathogens undergo molecular changes that are intimately linked to genome organization. Recent work has demonstrated that natural killer (NK) and CD8 + T cells experience substantial transcriptomic and epigenetic rewiring during their differentiation from naïve to effector to memory cells. Whether these molecular adaptations are accompanied by changes in three-dimensional (3D) chromatin architecture is unknown. In this study, we combine histone profiling, ATAC-seq, RNA-seq and high-throughput chromatin capture (HiC) assay to investigate the dynamics of one-dimensional (1D) and 3D chromatin during the differentiation of innate and adaptive lymphocytes. To this end, we discovered a coordinated 1D and 3D epigenetic remodeling during innate immune memory differentiation, and demonstrate that effector CD8 + T cells adopt an NK-like architectural program that is maintained in memory cells. Altogether, our study reveals the dynamic nature of the 1D and 3D genome during the formation of innate and adaptive immunological memory.

TP53 mutations and RNA-binding protein MUSASHI-2 drive resistance to PRMT5-targeted therapy in B-cell lymphoma.

To identify drivers of sensitivity and resistance to Protein Arginine Methyltransferase 5 (PRMT5) inhibition, we perform a genome-wide CRISPR/Cas9 screen. We identify TP53 and RNA-binding protein MUSASHI2 (MSI2) as the top-ranked sensitizer and driver of resistance to specific PRMT5i, GSK-591, respectively. TP53 deletion and TP53R248W mutation are biomarkers of resistance to GSK-591. PRMT5 expression correlates with MSI2 expression in lymphoma patients. MSI2 depletion and pharmacological inhibition using Ro 08-2750 (Ro) both synergize with GSK-591 to reduce cell growth. Ro reduces MSI2 binding to its global targets and dual treatment of Ro and PRMT5 inhibitors result in synergistic gene expression changes including cell cycle, P53 and MYC signatures. Dual MSI2 and PRMT5 inhibition further blocks c-MYC and BCL-2 translation. BCL-2 depletion or inhibition with venetoclax synergizes with a PRMT5 inhibitor by inducing reduced cell growth and apoptosis. Thus, we propose a therapeutic strategy in lymphoma that combines PRMT5 with MSI2 or BCL-2 inhibition.

CRISPR screening uncovers a central requirement for HHEX in pancreatic lineage commitment and plasticity restriction.

The pancreas and liver arise from a common pool of progenitors. However, the underlying mechanisms that drive their lineage diversification from the foregut endoderm are not fully understood. To tackle this question, we undertook a multifactorial approach that integrated human pluripotent-stem-cell-guided differentiation, genome-scale CRISPR-Cas9 screening, single-cell analysis, genomics and proteomics. We discovered that HHEX, a transcription factor (TF) widely recognized as a key regulator of liver development, acts as a gatekeeper of pancreatic lineage specification. HHEX deletion impaired pancreatic commitment and unleashed an unexpected degree of cellular plasticity towards the liver and duodenum fates. Mechanistically, HHEX cooperates with the pioneer TFs FOXA1, FOXA2 and GATA4, shared by both pancreas and liver differentiation programmes, to promote pancreas commitment, and this cooperation restrains the shared TFs from activating alternative lineages. These findings provide a generalizable model for how gatekeeper TFs like HHEX orchestrate lineage commitment and plasticity restriction in broad developmental contexts.

Cytotoxic innate lymphoid cells sense cancer cell-expressed interleukin-15 to suppress human and murine malignancies.

Malignancy can be suppressed by the immune system. However, the classes of immunosurveillance responses and their mode of tumor sensing remain incompletely understood. Here, we show that although clear cell renal cell carcinoma (ccRCC) was infiltrated by exhaustion-phenotype CD8+ T cells that negatively correlated with patient prognosis, chromophobe RCC (chRCC) had abundant infiltration of granzyme A-expressing intraepithelial type 1 innate lymphoid cells (ILC1s) that positively associated with patient survival. Interleukin-15 (IL-15) promoted ILC1 granzyme A expression and cytotoxicity, and IL-15 expression in chRCC tumor tissue positively tracked with the ILC1 response. An ILC1 gene signature also predicted survival of a subset of breast cancer patients in association with IL-15 expression. Notably, ILC1s directly interacted with cancer cells, and IL-15 produced by cancer cells supported the expansion and anti-tumor function of ILC1s in a murine breast cancer model. Thus, ILC1 sensing of cancer cell IL-15 defines an immunosurveillance mechanism of epithelial malignancies.

Chromatin interaction-aware gene regulatory modeling with graph attention networks.

Linking distal enhancers to genes and modeling their impact on target gene expression are longstanding unresolved problems in regulatory genomics and critical for interpreting noncoding genetic variation. Here, we present a new deep learning approach called GraphReg that exploits 3D interactions from chromosome conformation capture assays to predict gene expression from 1D epigenomic data or genomic DNA sequence. By using graph attention networks to exploit the connectivity of distal elements up to 2 Mb away in the genome, GraphReg more faithfully models gene regulation and more accurately predicts gene expression levels than the state-of-the-art deep learning methods for this task. Feature attribution used with GraphReg accurately identifies functional enhancers of genes, as validated by CRISPRi-FlowFISH and TAP-seq assays, outperforming both convolutional neural networks (CNNs) and the recently proposed activity-by-contact model. Sequence-based GraphReg also accurately predicts direct transcription factor (TF) targets as validated by CRISPRi TF knockout experiments via in silico ablation of TF binding motifs. GraphReg therefore represents an important advance in modeling the regulatory impact of epigenomic and sequence elements.

Cytotoxic granzyme C-expressing ILC1s contribute to antitumor immunity and neonatal autoimmunity.

Innate lymphocytes are integral components of the cellular immune system that can coordinate host defense against a multitude of challenges and trigger immunopathology when dysregulated. Natural killer (NK) cells and innate lymphoid cells (ILCs) are innate immune effectors postulated to functionally mirror conventional cytotoxic T lymphocytes and helper T cells, respectively. Here, we showed that the cytolytic molecule granzyme C was expressed in cells with the phenotype of type 1 ILCs (ILC1s) in mouse liver and salivary gland. Cell fate-mapping and transfer studies revealed that granzyme C-expressing innate lymphocytes could be derived from ILC progenitors and did not interconvert with NK cells, ILC2s, or ILC3s. Granzyme C defined a maturation state of ILC1s. These granzyme C-expressing ILC1s required the transcription factors T-bet and, to a lesser extent, Eomes and support from transforming growth factor-ß (TGF-ß) signaling for their maintenance in the salivary gland. In a transgenic mouse breast cancer model, depleting ILC1s caused accelerated tumor growth. ILC1s gained granzyme C expression following interleukin-15 (IL-15) stimulation, which enabled perforin-mediated cytotoxicity. Constitutive activation of STAT5, a transcription factor regulated by IL-15, in granzyme C-expressing ILC1s triggered lethal perforin-dependent autoimmunity in neonatal mice. Thus, granzyme C marks a cytotoxic effector state of ILC1s, broadening their function beyond "helper-like" lymphocytes.