Single-cell multi-cohort dissection of the schizophrenia transcriptome.

The complexity and heterogeneity of schizophrenia have hindered mechanistic elucidation and the development of more effective therapies. Here, we performed single-cell dissection of schizophrenia-associated transcriptomic changes in the human prefrontal cortex across 140 individuals in two independent cohorts. Excitatory neurons were the most affected cell group, with transcriptional changes converging on neurodevelopment and synapse-related molecular pathways. Transcriptional alterations included known genetic risk factors, suggesting convergence of rare and common genomic variants on neuronal population-specific alterations in schizophrenia. Based on the magnitude of schizophrenia-associated transcriptional change, we identified two populations of individuals with schizophrenia marked by expression of specific excitatory and inhibitory neuronal cell states. This single-cell atlas links transcriptomic changes to etiological genetic risk factors, contextualizing established knowledge within the human cortical cytoarchitecture and facilitating mechanistic understanding of schizophrenia pathophysiology and heterogeneity.

Systematic investigation of allelic regulatory activity of schizophrenia-associated common variants.

Genome-wide association studies (GWASs) have successfully identified 145 genomic regions that contribute to schizophrenia risk, but linkage disequilibrium makes it challenging to discern causal variants. We performed a massively parallel reporter assay (MPRA) on 5,173 fine-mapped schizophrenia GWAS variants in primary human neural progenitors and identified 439 variants with allelic regulatory effects (MPRA-positive variants). Transcription factor binding had modest predictive power, while fine-map posterior probability, enhancer overlap, and evolutionary conservation failed to predict MPRA-positive variants. Furthermore, 64% of MPRA-positive variants did not exhibit expressive quantitative trait loci signature, suggesting that MPRA could identify yet unexplored variants with regulatory potentials. To predict the combinatorial effect of MPRA-positive variants on gene regulation, we propose an accessibility-by-contact model that combines MPRA-measured allelic activity with neuronal chromatin architecture.

Lateral mammillary body neurons in mouse brain are disproportionately vulnerable in Alzheimer's disease.

The neural circuits governing the induction and progression of neurodegeneration and memory impairment in Alzheimer's disease (AD) are incompletely understood. The mammillary body (MB), a subcortical node of the medial limbic circuit, is one of the first brain regions to exhibit amyloid deposition in the 5xFAD mouse model of AD. Amyloid burden in the MB correlates with pathological diagnosis of AD in human postmortem brain tissue. Whether and how MB neuronal circuitry contributes to neurodegeneration and memory deficits in AD are unknown. Using 5xFAD mice and postmortem MB samples from individuals with varying degrees of AD pathology, we identified two neuronal cell types in the MB harboring distinct electrophysiological properties and long-range projections: lateral neurons and medial neurons. lateral MB neurons harbored aberrant hyperactivity and exhibited early neurodegeneration in 5xFAD mice compared with lateral MB neurons in wild-type littermates. Inducing hyperactivity in lateral MB neurons in wild-type mice impaired performance on memory tasks, whereas attenuating aberrant hyperactivity in lateral MB neurons ameliorated memory deficits in 5xFAD mice. Our findings suggest that neurodegeneration may be a result of genetically distinct, projection-specific cellular dysfunction and that dysregulated lateral MB neurons may be causally linked to memory deficits in AD.

APOE4 impairs myelination via cholesterol dysregulation in oligodendrocytes.

APOE4 is the strongest genetic risk factor for Alzheimer's disease1-3. However, the effects of APOE4 on the human brain are not fully understood, limiting opportunities to develop targeted therapeutics for individuals carrying APOE4 and other risk factors for Alzheimer's disease4-8. Here, to gain more comprehensive insights into the impact of APOE4 on the human brain, we performed single-cell transcriptomics profiling of post-mortem human brains from APOE4 carriers compared with non-carriers. This revealed that APOE4 is associated with widespread gene expression changes across all cell types of the human brain. Consistent with the biological function of APOE2-6, APOE4 significantly altered signalling pathways associated with cholesterol homeostasis and transport. Confirming these findings with histological and lipidomic analysis of the post-mortem human brain, induced pluripotent stem-cell-derived cells and targeted-replacement mice, we show that cholesterol is aberrantly deposited in oligodendrocytes-myelinating cells that are responsible for insulating and promoting the electrical activity of neurons. We show that altered cholesterol localization in the APOE4 brain coincides with reduced myelination. Pharmacologically facilitating cholesterol transport increases axonal myelination and improves learning and memory in APOE4 mice. We provide a single-cell atlas describing the transcriptional effects of APOE4 on the aging human brain and establish a functional link between APOE4, cholesterol, myelination and memory, offering therapeutic opportunities for Alzheimer's disease.

Neurons burdened by DNA double-strand breaks incite microglia activation through antiviral-like signaling in neurodegeneration.

DNA double-strand breaks (DSBs) are linked to neurodegeneration and senescence. However, it is not clear how DSB-bearing neurons influence neuroinflammation associated with neurodegeneration. Here, we characterize DSB-bearing neurons from the CK-p25 mouse model of neurodegeneration using single-nucleus, bulk, and spatial transcriptomic techniques. DSB-bearing neurons enter a late-stage DNA damage response marked by nuclear factor κB (NFκB)-activated senescent and antiviral immune pathways. In humans, Alzheimer's disease pathology is closely associated with immune activation in excitatory neurons. Spatial transcriptomics reveal that regions of CK-p25 brain tissue dense with DSB-bearing neurons harbor signatures of inflammatory microglia, which is ameliorated by NFκB knockdown in neurons. Inhibition of NFκB in DSB-bearing neurons also reduces microglia activation in organotypic mouse brain slice culture. In conclusion, DSBs activate immune pathways in neurons, which in turn adopt a senescence-associated secretory phenotype to elicit microglia activation. These findings highlight a previously unidentified role for neurons in the mechanism of disease-associated neuroinflammation.

BACE-1 inhibition facilitates the transition from homeostatic microglia to DAM-1.

BACE-1 is required for generating ß-amyloid (Aß) peptides in Alzheimer's disease (AD). Here, we report that microglial BACE-1 regulates the transition of homeostatic to stage 1 disease-associated microglia (DAM-1) signature. BACE-1 deficiency elevated levels of transcription factors including Jun, Jund, Btg2, Erg1, Junb, Fos, and Fosb in the transition signature, which transition from more homeostatic to highly phagocytic DAM-1. Consistently, similar transition-state microglia in human AD brains correlated with lowered levels of BACE-1 expression. Targeted deletion of Bace-1 in adult 5xFAD mice microglia elevated these phagocytic microglia, correlated with significant reduction in amyloid plaques without synaptic toxicity. Silencing or pharmacologically inhibiting BACE-1 in cultured microglia-derived cells shows higher phagocytic function in microglia. Mechanistic exploration suggests that abolished cleavage of IL-1R2 and Toll-like receptors via BACE-1 inhibition contributes to the enhanced signaling via the PI3K and p38 MAPK kinase pathway. Together, targeted inhibition of BACE-1 in microglia may offer AD treatment.

Gene Regulatory Network Dynamical Logical Models for Plant Development.

Mathematical and computational approaches that integrate and model the concerted action of multiple genetic and nongenetic components holding highly nonlinear interactions are fundamental for the study of developmental processes. Among these, gene regulatory network (GRN) dynamical models are very useful to understand how diverse types of regulatory constraints restrict the multigene expression patterns that characterize different cell fates. In this chapter we present a hands-on approach to model GRN dynamics, taking as a working example a well-curated and experimentally grounded GRN developmental module proposed by our group: the flower organ specification gene regulatory network (FOS-GRN). We demonstrate how to build and analyze a GRN model according to the following steps: (1) integration of molecular genetic data and formulation of logical rules specifying the dynamic behavior of each gene; (2) determination of steady states (attractors) corresponding to each cell type; (3) validation of the GRN model; and (4) extension of the deterministic model with the inclusion of stochasticity in order to model cell-state transitions dependent on noise due to fluctuations of the involved gen products. The methodologies explained here in detail can be applied to any other developmental module.

Patient-Derived Bone Marrow Spheroids Reveal Leukemia-Initiating Cells Supported by Mesenchymal Hypoxic Niches in Pediatric B-ALL.

B-cell acute lymphoblastic leukemia (B-ALL) results from the expansion of malignant lymphoid precursors within the bone marrow (BM), where hematopoietic niches and microenvironmental signals provide leukemia-initiating cells (LICs) the conditions to survive, proliferate, initiate disease, and relapse. Normal and malignant lymphopoiesis are highly dependent on the BM microenvironment, particularly on CXCL12-abundant Reticular (CAR) cells, which provide a niche for maintenance of primitive cells. During B-ALL, leukemic cells hijack BM niches, creating a proinflammatory milieu incompetent to support normal hematopoiesis but favoring leukemic proliferation. Although the lack of a phenotypic stem cell hierarchy is apparent in B-ALL, LICs are a rare and quiescent population potentially responsible for chemoresistance and relapse. Here, we developed novel patient-derived leukemia spheroids (PDLS), an ex vivo avatar model, from mesenchymal stromal cells (MSCs) and primary B-ALL cells, to mimic specialized niche structures and cell-to-cell intercommunication promoting normal and malignant hematopoiesis in pediatric B-ALL. 3D MSC spheroids can recapitulate CAR niche-like hypoxic structures that produce high levels of CXCL10 and CXCL11. We found that PDLS were preferentially enriched with leukemia cells displaying functional properties of LICs, such as quiescence, low reactive oxygen species, drug resistance, high engraftment in immunodeficient mice, and long-term leukemogenesis. Moreover, the combination of PDLS and patient-derived xenografts confirmed a microenvironment-driven hierarchy in their leukemic potential. Importantly, transcriptional profiles of MSC derived from primary patient samples revealed two unique signatures (1), a CXCL12low inflammatory and leukemia expansion (ILE)-like niche, that likely supports leukemic burden, and (2) a CXCL11hi immune-suppressive and leukemia-initiating cell (SLIC)-like niche, where LICs are likely sustained. Interestingly, the CXCL11+ hypoxic zones were recapitulated within the PDLS that are capable of supporting LIC functions. Taken together, we have implemented a novel PDLS system that enriches and supports leukemia cells with stem cell features driven by CXCL11+ MSCs within hypoxic microenvironments capable of recapitulating key features, such as tumor reemergence after exposure to chemotherapy and tumor initiation. This system represents a unique opportunity for designing ex vivo personalized avatars for B-ALL patients to evaluate their own LIC pathobiology and drug sensitivity in the context of the tumor microenvironment.

MEF2 is a key regulator of cognitive potential and confers resilience to neurodegeneration.

Recent increases in human longevity have been accompanied by a rise in the incidence of dementia, highlighting the need to preserve cognitive function in an aging population. A small percentage of individuals with pathological hallmarks of neurodegenerative disease are able to maintain normal cognition. Although the molecular mechanisms that govern this neuroprotection remain unknown, individuals that exhibit cognitive resilience (CgR) represent a unique source of therapeutic insight. For both humans and animal models, living in an enriched, cognitively stimulating environment is the most effective known inducer of CgR. To understand potential drivers of this phenomenon, we began by profiling the molecular changes that arise from environmental enrichment in mice, which led to the identification of MEF2 transcription factors (TFs). We next turned to repositories of human clinical and brain transcriptomic data, where we found that the MEF2 transcriptional network was overrepresented among genes that are most predictive of end-stage cognition. Through single-nucleus RNA sequencing of cortical tissue from resilient and nonresilient individuals, we further confirmed up-regulation of MEF2C in resilient individuals to a subpopulation of excitatory neurons. Last, to determine the causal impact of MEF2 on cognition in the context of neurodegeneration, we overexpressed Mef2a/c in the PS19 mouse model of tauopathy and found that this was sufficient to improve cognitive flexibility and reduce hyperexcitability. Overall, our findings reveal a previously unappreciated role for MEF2 TFs in promoting CgR, highlighting their potential as biomarkers or therapeutic targets for neurodegeneration and healthy aging.

In vivo and in vitro human gene essentiality estimations capture contrasting functional constraints.

Gene essentiality estimation is a popular empirical approach to link genotypes to phenotypes. In humans, essentiality is estimated based on loss-of-function (LoF) mutation intolerance, either from population exome sequencing (in vivo) data or CRISPR-based in vitro perturbation experiments. Both approaches identify genes presumed to have detrimental consequences on the organism upon mutation. Are these genes constrained by having key cellular/organismal roles? Do in vivo and in vitro estimations equally recover these constraints? Insights into these questions have important implications in generalizing observations from cell models and interpreting disease risk genes. To empirically address these questions, we integrate genome-scale datasets and compare structural, functional and evolutionary features of essential genes versus genes with extremely high mutational tolerance. We found that essentiality estimates do recover functional constraints. However, the organismal or cellular context of estimation leads to functionally contrasting properties underlying the constraint. Our results suggest that depletion of LoF mutations in human populations effectively captures organismal-level functional constraints not experimentally accessible through CRISPR-based screens. Finally, we identify a set of genes (OrgEssential), which are mutationally intolerant in vivo but highly tolerant in vitro. These genes drive observed functional constraint differences and have an unexpected preference for nervous system expression.

NEBULA is a fast negative binomial mixed model for differential or co-expression analysis of large-scale multi-subject single-cell data.

The increasing availability of single-cell data revolutionizes the understanding of biological mechanisms at cellular resolution. For differential expression analysis in multi-subject single-cell data, negative binomial mixed models account for both subject-level and cell-level overdispersions, but are computationally demanding. Here, we propose an efficient NEgative Binomial mixed model Using a Large-sample Approximation (NEBULA). The speed gain is achieved by analytically solving high-dimensional integrals instead of using the Laplace approximation. We demonstrate that NEBULA is orders of magnitude faster than existing tools and controls false-positive errors in marker gene identification and co-expression analysis. Using NEBULA in Alzheimer's disease cohort data sets, we found that the cell-level expression of APOE correlated with that of other genetic risk factors (including CLU, CST3, TREM2, C1q, and ITM2B) in a cell-type-specific pattern and an isoform-dependent manner in microglia. NEBULA opens up a new avenue for the broad application of mixed models to large-scale multi-subject single-cell data.

Mapping the epigenomic and transcriptomic interplay during memory formation and recall in the hippocampal engram ensemble.

The epigenome and three-dimensional (3D) genomic architecture are emerging as key factors in the dynamic regulation of different transcriptional programs required for neuronal functions. In this study, we used an activity-dependent tagging system in mice to determine the epigenetic state, 3D genome architecture and transcriptional landscape of engram cells over the lifespan of memory formation and recall. Our findings reveal that memory encoding leads to an epigenetic priming event, marked by increased accessibility of enhancers without the corresponding transcriptional changes. Memory consolidation subsequently results in spatial reorganization of large chromatin segments and promoter-enhancer interactions. Finally, with reactivation, engram neurons use a subset of de novo long-range interactions, where primed enhancers are brought in contact with their respective promoters to upregulate genes involved in local protein translation in synaptic compartments. Collectively, our work elucidates the comprehensive transcriptional and epigenomic landscape across the lifespan of memory formation and recall in the hippocampal engram ensemble.

A multiresolution framework to characterize single-cell state landscapes.

Dissecting the cellular heterogeneity embedded in single-cell transcriptomic data is challenging. Although many methods and approaches exist, identifying cell states and their underlying topology is still a major challenge. Here, we introduce the concept of multiresolution cell-state decomposition as a practical approach to simultaneously capture both fine- and coarse-grain patterns of variability. We implement this concept in ACTIONet, a comprehensive framework that combines archetypal analysis and manifold learning to provide a ready-to-use analytical approach for multiresolution single-cell state characterization. ACTIONet provides a robust, reproducible, and highly interpretable single-cell analysis platform that couples dominant pattern discovery with a corresponding structural representation of the cell state landscape. Using multiple synthetic and real data sets, we demonstrate ACTIONet's superior performance relative to existing alternatives. We use ACTIONet to integrate and annotate cells across three human cortex data sets. Through integrative comparative analysis, we define a consensus vocabulary and a consistent set of gene signatures discriminating against the transcriptomic cell types and subtypes of the human prefrontal cortex.

Cell Type-Specific Transcriptomics Reveals that Mutant Huntingtin Leads to Mitochondrial RNA Release and Neuronal Innate Immune Activation.

The mechanisms by which mutant huntingtin (mHTT) leads to neuronal cell death in Huntington's disease (HD) are not fully understood. To gain new molecular insights, we used single nuclear RNA sequencing (snRNA-seq) and translating ribosome affinity purification (TRAP) to conduct transcriptomic analyses of caudate/putamen (striatal) cell type-specific gene expression changes in human HD and mouse models of HD. In striatal spiny projection neurons, the most vulnerable cell type in HD, we observe a release of mitochondrial RNA (mtRNA) (a potent mitochondrial-derived innate immunogen) and a concomitant upregulation of innate immune signaling in spiny projection neurons. Further, we observe that the released mtRNAs can directly bind to the innate immune sensor protein kinase R (PKR). We highlight the importance of studying cell type-specific gene expression dysregulation in HD pathogenesis and reveal that the activation of innate immune signaling in the most vulnerable HD neurons provides a novel framework to understand the basis of mHTT toxicity and raises new therapeutic opportunities.

Reconstruction of the human blood-brain barrier in vitro reveals a pathogenic mechanism of APOE4 in pericytes.

In Alzheimer's disease, amyloid deposits along the brain vasculature lead to a condition known as cerebral amyloid angiopathy (CAA), which impairs blood-brain barrier (BBB) function and accelerates cognitive degeneration. Apolipoprotein (APOE4) is the strongest risk factor for CAA, yet the mechanisms underlying this genetic susceptibility are unknown. Here we developed an induced pluripotent stem cell-based three-dimensional model that recapitulates anatomical and physiological properties of the human BBB in vitro. Similarly to CAA, our in vitro BBB displayed significantly more amyloid accumulation in APOE4 compared to APOE3. Combinatorial experiments revealed that dysregulation of calcineurin-nuclear factor of activated T cells (NFAT) signaling and APOE in pericyte-like mural cells induces APOE4-associated CAA pathology. In the human brain, APOE and NFAT are selectively dysregulated in pericytes of APOE4 carriers, and inhibition of calcineurin-NFAT signaling reduces APOE4-associated CAA pathology in vitro and in vivo. Our study reveals the role of pericytes in APOE4-mediated CAA and highlights calcineurin-NFAT signaling as a therapeutic target in CAA and Alzheimer's disease.

Inferring multimodal latent topics from electronic health records.

Electronic health records (EHR) are rich heterogeneous collections of patient health information, whose broad adoption provides clinicians and researchers unprecedented opportunities for health informatics, disease-risk prediction, actionable clinical recommendations, and precision medicine. However, EHRs present several modeling challenges, including highly sparse data matrices, noisy irregular clinical notes, arbitrary biases in billing code assignment, diagnosis-driven lab tests, and heterogeneous data types. To address these challenges, we present MixEHR, a multi-view Bayesian topic model. We demonstrate MixEHR on MIMIC-III, Mayo Clinic Bipolar Disorder, and Quebec Congenital Heart Disease EHR datasets. Qualitatively, MixEHR disease topics reveal meaningful combinations of clinical features across heterogeneous data types. Quantitatively, we observe superior prediction accuracy of diagnostic codes and lab test imputations compared to the state-of-art methods. We leverage the inferred patient topic mixtures to classify target diseases and predict mortality of patients in critical conditions. In all comparison, MixEHR confers competitive performance and reveals meaningful disease-related topics.

Reconstruction of Cell-type-Specific Interactomes at Single-Cell Resolution.

The human interactome is instrumental in the systems-level study of the cell and the contextualization of disease-associated gene perturbations. However, reference organismal interactomes do not capture the cell-type-specific context in which proteins and modules preferentially act. Here, we introduce SCINET, a computational framework that reconstructs an ensemble of cell-type-specific interactomes by integrating a global, context-independent reference interactome with a single-cell gene-expression profile. SCINET addresses technical challenges of single-cell data by robustly imputing, transforming, and normalizing the initially noisy and sparse expression of data. Inferred cell-level gene interaction probabilities and group-level interaction strengths define cell-type-specific interactomes. We use SCINET to reconstruct and analyze interactomes of the major human brain and immune cell types, revealing specificity and modularity of perturbations associated with neurodegenerative, neuropsychiatric, and autoimmune disorders. We report cell-type interactomes for brain and immune cell types, together with the SCINET package.

Single-cell transcriptomic atlas of the human retina identifies cell types associated with age-related macular degeneration.

Genome-wide association studies (GWAS) have identified genetic variants associated with age-related macular degeneration (AMD), one of the leading causes of blindness in the elderly. However, it has been challenging to identify the cell types associated with AMD given the genetic complexity of the disease. Here we perform massively parallel single-cell RNA sequencing (scRNA-seq) of human retinas using two independent platforms, and report the first single-cell transcriptomic atlas of the human retina. Using a multi-resolution network-based analysis, we identify all major retinal cell types, and their corresponding gene expression signatures. Heterogeneity is observed within macroglia, suggesting that human retinal glia are more diverse than previously thought. Finally, GWAS-based enrichment analysis identifies glia, vascular cells, and cone photoreceptors to be associated with the risk of AMD. These data provide a detailed analysis of the human retina, and show how scRNA-seq can provide insight into cell types involved in complex, inflammatory genetic diseases.

Integrative construction of regulatory region networks in 127 human reference epigenomes by matrix factorization.

Despite large experimental and computational efforts aiming to dissect the mechanisms underlying disease risk, mapping cis-regulatory elements to target genes remains a challenge. Here, we introduce a matrix factorization framework to integrate physical and functional interaction data of genomic segments. The framework was used to predict a regulatory network of chromatin interaction edges linking more than 20 000 promoters and 1.8 million enhancers across 127 human reference epigenomes, including edges that are present in any of the input datasets. Our network integrates functional evidence of correlated activity patterns from epigenomic data and physical evidence of chromatin interactions. An important contribution of this work is the representation of heterogeneous data with different qualities as networks. We show that the unbiased integration of independent data sources suggestive of regulatory interactions produces meaningful associations supported by existing functional and physical evidence, correlating with expected independent biological features.