Massively parallel in vivo Perturb-seq reveals cell-type-specific transcriptional networks in cortical development.

Leveraging AAVs' versatile tropism and labeling capacity, we expanded the scale of in vivo CRISPR screening with single-cell transcriptomic phenotyping across embryonic to adult brains and peripheral nervous systems. Through extensive tests of 86 vectors across AAV serotypes combined with a transposon system, we substantially amplified labeling efficacy and accelerated in vivo gene delivery from weeks to days. Our proof-of-principle in utero screen identified the pleiotropic effects of Foxg1, highlighting its tight regulation of distinct networks essential for cell fate specification of Layer 6 corticothalamic neurons. Notably, our platform can label >6% of cerebral cells, surpassing the current state-of-the-art efficacy at <0.1% by lentivirus, to achieve analysis of over 30,000 cells in one experiment and enable massively parallel in vivo Perturb-seq. Compatible with various phenotypic measurements (single-cell or spatial multi-omics), it presents a flexible approach to interrogate gene function across cell types in vivo, translating gene variants to their causal function.

Lupus autoantibodies initiate neuroinflammation sustained by continuous HMGB1:RAGE signaling and reversed by increased LAIR-1 expression.

Cognitive impairment is a frequent manifestation of neuropsychiatric systemic lupus erythematosus, present in up to 80% of patients and leading to a diminished quality of life. In the present study, we used a model of lupus-like cognitive impairment that is initiated when antibodies that crossreact with excitatory neuronal receptors penetrate the hippocampus, causing immediate, self-limited, excitotoxic death of hippocampal neurons, which is then followed by a significant loss of dendritic complexity in surviving neurons. This injury creates a maladaptive equilibrium that is sustained in mice for at least 1 year. We identified a feedforward loop of microglial activation and microglia-dependent synapse elimination dependent on neuronal secretion of high mobility group box 1 protein (HMGB1) which binds the receptor for advanced glycation end products (RAGE) and leads to microglial secretion of C1q, upregulation of interleukin-10 with consequent downregulation of leukocyte-associated immunoglobulin-like receptor 1 (LAIR-1), an inhibitory receptor for C1q. Treatment with a centrally acting angiotensin-converting enzyme inhibitor or with an angiotensin-receptor blocker restored a healthy equilibrium, microglial quiescence and intact spatial memory.

Autism genes converge on asynchronous development of shared neuron classes.

Genetic risk for autism spectrum disorder (ASD) is associated with hundreds of genes spanning a wide range of biological functions1-6. The alterations in the human brain resulting from mutations in these genes remain unclear. Furthermore, their phenotypic manifestation varies across individuals7,8. Here we used organoid models of the human cerebral cortex to identify cell-type-specific developmental abnormalities that result from haploinsufficiency in three ASD risk genes-SUV420H1 (also known as KMT5B), ARID1B and CHD8-in multiple cell lines from different donors, using single-cell RNA-sequencing (scRNA-seq) analysis of more than 745,000 cells and proteomic analysis of individual organoids, to identify phenotypic convergence. Each of the three mutations confers asynchronous development of two main cortical neuronal lineages-γ-aminobutyric-acid-releasing (GABAergic) neurons and deep-layer excitatory projection neurons-but acts through largely distinct molecular pathways. Although these phenotypes are consistent across cell lines, their expressivity is influenced by the individual genomic context, in a manner that is dependent on both the risk gene and the developmental defect. Calcium imaging in intact organoids shows that these early-stage developmental changes are followed by abnormal circuit activity. This research uncovers cell-type-specific neurodevelopmental abnormalities that are shared across ASD risk genes and are finely modulated by human genomic context, finding convergence in the neurobiological basis of how different risk genes contribute to ASD pathology.

Distinct subnetworks of the thalamic reticular nucleus.

The thalamic reticular nucleus (TRN), the major source of thalamic inhibition, regulates thalamocortical interactions that are critical for sensory processing, attention and cognition1-5. TRN dysfunction has been linked to sensory abnormality, attention deficit and sleep disturbance across multiple neurodevelopmental disorders6-9. However, little is known about the organizational principles that underlie its divergent functions. Here we performed an integrative study linking single-cell molecular and electrophysiological features of the mouse TRN to connectivity and systems-level function. We found that cellular heterogeneity in the TRN is characterized by a transcriptomic gradient of two negatively correlated gene-expression profiles, each containing hundreds of genes. Neurons in the extremes of this transcriptomic gradient express mutually exclusive markers, exhibit core or shell-like anatomical structure and have distinct electrophysiological properties. The two TRN subpopulations make differential connections with the functionally distinct first-order and higher-order thalamic nuclei to form molecularly defined TRN-thalamus subnetworks. Selective perturbation of the two subnetworks in vivo revealed their differential role in regulating sleep. In sum, our study provides a comprehensive atlas of TRN neurons at single-cell resolution and links molecularly defined subnetworks to the functional organization of thalamocortical circuits.

Individual brain organoids reproducibly form cell diversity of the human cerebral cortex.

Experimental models of the human brain are needed for basic understanding of its development and disease1. Human brain organoids hold unprecedented promise for this purpose; however, they are plagued by high organoid-to-organoid variability2,3. This has raised doubts as to whether developmental processes of the human brain can occur outside the context of embryogenesis with a degree of reproducibility that is comparable to the endogenous tissue. Here we show that an organoid model of the dorsal forebrain can reliably generate a rich diversity of cell types appropriate for the human cerebral cortex. We performed single-cell RNA-sequencing analysis of 166,242 cells isolated from 21 individual organoids, finding that 95% of the organoids generate a virtually indistinguishable compendium of cell types, following similar developmental trajectories and with a degree of organoid-to-organoid variability comparable to that of individual endogenous brains. Furthermore, organoids derived from different stem cell lines show consistent reproducibility in the cell types produced. The data demonstrate that reproducible development of the complex cellular diversity of the central nervous system does not require the context of the embryo, and that establishment of terminal cell identity is a highly constrained process that can emerge from diverse stem cell origins and growth environments.

A single-cell map of antisense oligonucleotide activity in the brain.

Antisense oligonucleotides (ASOs) dosed into cerebrospinal fluid (CSF) distribute broadly throughout the central nervous system (CNS). By modulating RNA, they hold the promise of targeting root molecular causes of disease and hold potential to treat myriad CNS disorders. Realization of this potential requires that ASOs must be active in the disease-relevant cells, and ideally, that monitorable biomarkers also reflect ASO activity in these cells. The biodistribution and activity of such centrally delivered ASOs have been deeply characterized in rodent and non-human primate (NHP) models, but usually only in bulk tissue, limiting our understanding of the distribution of ASO activity across individual cells and across diverse CNS cell types. Moreover, in human clinical trials, target engagement is usually monitorable only in a single compartment, CSF. We sought a deeper understanding of how individual cells and cell types contribute to bulk tissue signal in the CNS, and how these are linked to CSF biomarker outcomes. We employed single nucleus transcriptomics on tissue from mice treated with RNase H1 ASOs against Prnp and Malat1 and NHPs treated with an ASO against PRNP. Pharmacologic activity was observed in every cell type, though sometimes with substantial differences in magnitude. Single cell RNA count distributions implied target RNA suppression in every single sequenced cell, rather than intense knockdown in only some cells. Duration of action up to 12 weeks post-dose differed across cell types, being shorter in microglia than in neurons. Suppression in neurons was generally similar to, or more robust than, the bulk tissue. In macaques, PrP in CSF was lowered 40% in conjunction with PRNP knockdown across all cell types including neurons, arguing that a CSF biomarker readout is likely to reflect ASO pharmacodynamic effect in disease-relevant cells in a neuronal disorder. Our results provide a reference dataset for ASO activity distribution in the CNS and establish single nucleus sequencing as a method for evaluating cell type specificity of oligonucleotide therapeutics and other modalities.

Antisense oligonucleotide (ASO) drugs are a type of chemically modified DNA that can be injected into cerebrospinal fluid in order to enter brain cells and reduce the amount of RNA from a specific gene. The brain is a complex mixture of hundreds of billions of cells. When an ASO lowers a target gene's RNA by 50%, is that a 50% reduction in 100% of cells, or a 100% reduction in 50% of cells? Are the many different cell types of the brain affected equally? This new study uses single cell RNA sequencing to answer these questions, finding that ASOs are broadly active across cell types and individual cells, and linking reduction of target protein in cerebrospinal fluid to disease-relevant cells.

Genetic analysis of isoform usage in the human anti-viral response reveals influenza-specific regulation of ERAP2 transcripts under balancing selection.

While genetic variants are known to be associated with overall gene abundance in stimulated immune cells, less is known about their effects on alternative isoform usage. By analyzing RNA-seq profiles of monocyte-derived dendritic cells from 243 individuals, we uncovered thousands of unannotated isoforms synthesized in response to influenza infection and type 1 interferon stimulation. We identified more than a thousand quantitative trait loci (QTLs) associated with alternate isoform usage (isoQTLs), many of which are independent of expression QTLs (eQTLs) for the same gene. Compared with eQTLs, isoQTLs are enriched for splice sites and untranslated regions, but depleted of sequences upstream of annotated transcription start sites. Both eQTLs and isoQTLs explain a significant proportion of the disease heritability attributed to common genetic variants. At the ERAP2 locus, we shed light on the function of the gene and how two frequent, highly differentiated haplotypes with intermediate frequencies could be maintained by balancing selection. At baseline and following type 1 interferon stimulation, the major haplotype is associated with low ERAP2 expression caused by nonsense-mediated decay, while the minor haplotype, known to increase Crohn's disease risk, is associated with high ERAP2 expression. In response to influenza infection, we found two uncharacterized isoforms expressed from the major haplotype, likely the result of multiple perfectly linked variants affecting the transcription and splicing at the locus. Thus, genetic variants at a single locus could modulate independent gene regulatory processes in innate immune responses and, in the case of ERAP2, may confer a historical fitness advantage in response to virus.

Author Correction: Comprehensive comparative analysis of 5'-end RNA-sequencing methods.

The original version of this paper contained an incorrect primer sequence. In the Methods subsection "Rampage libraries," the text for modification 3 stated that the reverse primer used for library indexing was 5'-CAAGCAGAAGACGGCATACGAGATXXXXXXXXGTGACTGGAGT-3'. The correct sequence of the oligonucleotide used is 5'-CAAGCAGAAGACGGCATACGAGATXXXXXXXXGTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT-3'. This error has been corrected in the PDF and HTML versions of the paper.

Comprehensive comparative analysis of 5'-end RNA-sequencing methods.

Specialized RNA-seq methods are required to identify the 5' ends of transcripts, which are critical for studies of gene regulation, but these methods have not been systematically benchmarked. We directly compared six such methods, including the performance of five methods on a single human cellular RNA sample and a new spike-in RNA assay that helps circumvent challenges resulting from uncertainties in annotation and RNA processing. We found that the 'cap analysis of gene expression' (CAGE) method performed best for mRNA and that most of its unannotated peaks were supported by evidence from other genomic methods. We applied CAGE to eight brain-related samples and determined sample-specific transcription start site (TSS) usage, as well as a transcriptome-wide shift in TSS usage between fetal and adult brain.

Realizing privacy preserving genome-wide association studies.

MOTIVATION: As genomics moves into the clinic, there has been much interest in using this medical data for research. At the same time the use of such data raises many privacy concerns. These circumstances have led to the development of various methods to perform genome-wide association studies (GWAS) on patient records while ensuring privacy. In particular, there has been growing interest in applying differentially private techniques to this challenge. Unfortunately, up until now all methods for finding high scoring SNPs in a differentially private manner have had major drawbacks in terms of either accuracy or computational efficiency. RESULTS: Here we overcome these limitations with a substantially modified version of the neighbor distance method for performing differentially private GWAS, and thus are able to produce a more viable mechanism. Specifically, we use input perturbation and an adaptive boundary method to overcome accuracy issues. We also design and implement a convex analysis based algorithm to calculate the neighbor distance for each SNP in constant time, overcoming the major computational bottleneck in the neighbor distance method. It is our hope that methods such as ours will pave the way for more widespread use of patient data in biomedical research. AVAILABILITY AND IMPLEMENTATION: A python implementation is available at http://groups.csail.mit.edu/cb/DiffPriv/ CONTACT: bab@csail.mit.edu SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Experimental and Computational Methods for Allelic Imbalance Analysis from Single-Nucleus RNA-seq Data.

Single-cell RNA-seq (scRNA-seq) is emerging as a powerful tool for understanding gene function across diverse cells. Recently, this has included the use of allele-specific expression (ASE) analysis to better understand how variation in the human genome affects RNA expression at the single-cell level. We reasoned that because intronic reads are more prevalent in single-nucleus RNA-Seq (snRNA-Seq), and introns are under lower purifying selection and thus enriched for genetic variants, that snRNA-seq should facilitate single-cell analysis of ASE. Here we demonstrate how experimental and computational choices can improve the results of allelic imbalance analysis. We explore how experimental choices, such as RNA source, read length, sequencing depth, genotyping, etc., impact the power of ASE-based methods. We developed a new suite of computational tools to process and analyze scRNA-seq and snRNA-seq for ASE. As hypothesized, we extracted more ASE information from reads in intronic regions than those in exonic regions and show how read length can be set to increase power. Additionally, hybrid selection improved our power to detect allelic imbalance in genes of interest. We also explored methods to recover allele-specific isoform expression levels from both long- and short-read snRNA-seq. To further investigate ASE in the context of human disease, we applied our methods to a Parkinson's disease cohort of 94 individuals and show that ASE analysis had more power than eQTL analysis to identify significant SNP/gene pairs in our direct comparison of the two methods. Overall, we provide an end-to-end experimental and computational approach for future studies.

The effect of background noise and its removal on the analysis of single-cell expression data.

BACKGROUND: In droplet-based single-cell and single-nucleus RNA-seq experiments, not all reads associated with one cell barcode originate from the encapsulated cell. Such background noise is attributed to spillage from cell-free ambient RNA or barcode swapping events. RESULTS: Here, we characterize this background noise exemplified by three scRNA-seq and two snRNA-seq replicates of mouse kidneys. For each experiment, cells from two mouse subspecies are pooled, allowing to identify cross-genotype contaminating molecules and thus profile background noise. Background noise is highly variable across replicates and cells, making up on average 3-35% of the total counts (UMIs) per cell and we find that noise levels are directly proportional to the specificity and detectability of marker genes. In search of the source of background noise, we find multiple lines of evidence that the majority of background molecules originates from ambient RNA. Finally, we use our genotype-based estimates to evaluate the performance of three methods (CellBender, DecontX, SoupX) that are designed to quantify and remove background noise. We find that CellBender provides the most precise estimates of background noise levels and also yields the highest improvement for marker gene detection. By contrast, clustering and classification of cells are fairly robust towards background noise and only small improvements can be achieved by background removal that may come at the cost of distortions in fine structure. CONCLUSIONS: Our findings help to better understand the extent, sources and impact of background noise in single-cell experiments and provide guidance on how to deal with it.

Mostly natural sequencing-by-synthesis for scRNA-seq using Ultima sequencing.

Here we introduce a mostly natural sequencing-by-synthesis (mnSBS) method for single-cell RNA sequencing (scRNA-seq), adapted to the Ultima genomics platform, and systematically benchmark it against current scRNA-seq technology. mnSBS uses mostly natural, unmodified nucleotides and only a low fraction of fluorescently labeled nucleotides, which allows for high polymerase processivity and lower costs. We demonstrate successful application in four scRNA-seq case studies of different technical and biological types, including 5' and 3' scRNA-seq, human peripheral blood mononuclear cells from a single individual and in multiplex, as well as Perturb-Seq. Benchmarking shows that results from mnSBS-based scRNA-seq are very similar to those using Illumina sequencing, with minor differences in results related to the position of reads relative to annotated gene boundaries, owing to single-end reads of Ultima being closer to gene ends than reads from Illumina. The method is thus compatible with state-of-the-art scRNA-seq libraries independent of the sequencing technology. We expect mnSBS to be of particular utility for cost-effective large-scale scRNA-seq projects.

A single-cell map of antisense oligonucleotide activity in the brain.

Antisense oligonucleotides (ASOs) dosed into cerebrospinal fluid (CSF) distribute broadly throughout the brain and hold the promise of treating myriad brain diseases by modulating RNA. CNS tissue is not routinely biopsied in living individuals, leading to reliance on CSF biomarkers to inform on drug target engagement. Animal models can link CSF biomarkers to brain parenchyma, but our understanding of how individual cells contribute to bulk tissue signal is limited. Here we employed single nucleus transcriptomics on tissue from mice treated with RNase H1 ASOs against Prnp and Malat1 and macaques treated with an ASO against PRNP . Activity was observed in every cell type, though sometimes with substantial differences in magnitude. Single cell RNA count distributions implied target suppression in every single sequenced cell, rather than intense knockdown in only some cells. Duration of action up to 12 weeks post-dose differed across cell types, being shorter in microglia than in neurons. Suppression in neurons was generally similar to, or more robust than, the bulk tissue. In macaques, PrP in CSF was lowered 40% in conjunction with PRNP knockdown across all cell types including neurons, arguing that a CSF biomarker readout is likely to reflect disease-relevant cells in a neuronal disorder.

Massively parallel in vivo Perturb-seq reveals cell type-specific transcriptional networks in cortical development.

Systematic analysis of gene function across diverse cell types in vivo is hindered by two challenges: obtaining sufficient cells from live tissues and accurately identifying each cell's perturbation in high-throughput single-cell assays. Leveraging AAV's versatile cell type tropism and high labeling capacity, we expanded the resolution and scale of in vivo CRISPR screens: allowing phenotypic analysis at single-cell resolution across a multitude of cell types in the embryonic brain, adult brain, and peripheral nervous system. We undertook extensive tests of 86 AAV serotypes, combined with a transposon system, to substantially amplify labeling and accelerate in vivo gene delivery from weeks to days. Using this platform, we performed an in utero genetic screen as proof-of-principle and identified pleiotropic regulatory networks of Foxg1 in cortical development, including Layer 6 corticothalamic neurons where it tightly controls distinct networks essential for cell fate specification. Notably, our platform can label >6% of cerebral cells, surpassing the current state-of-the-art efficacy at <0.1% (mediated by lentivirus), and achieve analysis of over 30,000 cells in one experiment, thus enabling massively parallel in vivo Perturb-seq. Compatible with various perturbation techniques (CRISPRa/i) and phenotypic measurements (single-cell or spatial multi-omics), our platform presents a flexible, modular approach to interrogate gene function across diverse cell types in vivo, connecting gene variants to their causal functions.

Lupus autoantibodies initiate a maladaptive equilibrium sustained by HMGB1:RAGE signaling and reversed by LAIR-1:C1q signaling.

Cognitive impairment is a frequent manifestation of neuropsychiatric systemic lupus erythematosus (NPSLE), present in up to 80% of patients and leading to a diminished quality of life. We have developed a model of lupus-like cognitive impairment which is initiated when anti-DNA, anti-N-methyl D-aspartate receptor (NMDAR) cross- reactive antibodies, which are present in 30% of SLE patients, penetrate the hippocampus1. This leads to immediate, self-limited excitotoxic death of CA1 pyramidal neurons followed by a significant loss of dendritic arborization in the remaining CA1 neurons and impaired spatial memory. Both microglia and C1q are required for dendritic loss1. Here we show that this pattern of hippocampal injury creates a maladaptive equilibrium that is sustained for at least one year. It requires HMGB1 secretion by neurons to bind RAGE, a receptor for HMGB1 expressed on microglia, and leads to decreased expression of microglial LAIR-1, an inhibitory receptor for C1q. The angiotensin converting enzyme (ACE) inhibitor captopril, which can restore a healthy equilibrium, microglial quiescence, and intact spatial memory, leads to upregulation of LAIR-1. This paradigm highlights HMGB1:RAGE and C1q:LAIR-1 interactions as pivotal pathways in the microglial-neuronal interplay that defines a physiologic versus a maladaptive equilibrium.

Brain-region-specific changes in neurons and glia and dysregulation of dopamine signaling in Grin2a mutant mice.

A genetically valid animal model could transform our understanding of schizophrenia (SCZ) disease mechanisms. Rare heterozygous loss-of-function (LoF) mutations in GRIN2A, encoding a subunit of the NMDA receptor, greatly increase the risk of SCZ. By transcriptomic, proteomic, and behavioral analyses, we report that heterozygous Grin2a mutant mice show (1) large-scale gene expression changes across multiple brain regions and in neuronal (excitatory and inhibitory) and non-neuronal cells (astrocytes and oligodendrocytes), (2) evidence of hypoactivity in the prefrontal cortex (PFC) and hyperactivity in the hippocampus and striatum, (3) an elevated dopamine signaling in the striatum and hypersensitivity to amphetamine-induced hyperlocomotion (AIH), (4) altered cholesterol biosynthesis in astrocytes, (5) a reduction in glutamatergic receptor signaling proteins in the synapse, and (6) an aberrant locomotor pattern opposite of that induced by antipsychotic drugs. These findings reveal potential pathophysiologic mechanisms, provide support for both the "hypo-glutamate" and "hyper-dopamine" hypotheses of SCZ, and underscore the utility of Grin2a-deficient mice as a genetic model of SCZ.

Heterochronic parabiosis reprograms the mouse brain transcriptome by shifting aging signatures in multiple cell types.

Aging is a complex process involving transcriptomic changes associated with deterioration across multiple tissues and organs, including the brain. Recent studies using heterochronic parabiosis have shown that various aspects of aging-associated decline are modifiable or even reversible. To better understand how this occurs, we performed single-cell transcriptomic profiling of young and old mouse brains after parabiosis. For each cell type, we cataloged alterations in gene expression, molecular pathways, transcriptional networks, ligand-receptor interactions and senescence status. Our analyses identified gene signatures, demonstrating that heterochronic parabiosis regulates several hallmarks of aging in a cell-type-specific manner. Brain endothelial cells were found to be especially malleable to this intervention, exhibiting dynamic transcriptional changes that affect vascular structure and function. These findings suggest new strategies for slowing deterioration and driving regeneration in the aging brain through approaches that do not rely on disease-specific mechanisms or actions of individual circulating factors.

Establishment of an Engaged and Active Learning Community in the Biology Classroom and Lab with Discord.

Discord is an instant messaging platform that facilitates voice and video calls and media exchanges. Although Discord was initially developed for video gaming, it is currently used by other communities where voice chatting is frequent. Here, Discord is described as a useful tool in the science classroom. We implemented Discord in several biology labs and classes and gained experience using it as a professor (AW) and as both a teaching assistant and a student (SS). In this Discord-enhanced learning space, students seemed to be more engaged with the subject material, their peers, and their instructors in comparison with a more traditional physical classroom and an entirely online classroom. In addition, real-time communication was augmented in a physical lab or classroom by Discord's interconnectivity. Discord facilitated real-time peer and instructor discussion and provided opportunities for group interaction outside conventional science teaching spaces. These increased connections also extended to the lab setting, both in and out of scheduled class time, providing a space to build community within and between groups and between students and instructors. Introducing Discord to a physical classroom or using it for an online class provided us with opportunities for a more engaged and active student, and they found it may lead to greater productivity and learning.