{
  "title": "An atlas of cortical arealization identifies dynamic molecular signatures",
  "abstract": "\nThe human brain is subdivided into distinct anatomical structures, including\nthe neocortex, which in turn encompasses dozens of distinct specialized\ncortical areas. Early morphogenetic gradients are known to establish early\nbrain regions and cortical areas, but how early patterns result in finer and\nmore discrete spatial differences remains poorly understood1. Here we use\nsingle-cell RNA sequencing to profile ten major brain structures and six\nneocortical areas during peak neurogenesis and early gliogenesis. Within the\nneocortex, we find that early in the second trimester, a large number of genes\nare differentially expressed across distinct cortical areas in all cell types,\nincluding radial glia, the neural progenitors of the cortex. However, the\nabundance of areal transcriptomic signatures increases as radial glia\ndifferentiate into intermediate progenitor cells and ultimately give rise to\nexcitatory neurons. Using an automated, multiplexed single-molecule fluorescent\nin situ hybridization approach, we find that laminar gene-expression patterns\nare highly dynamic across cortical regions. Together, our data suggest that\nearly cortical areal patterning is defined by strong, mutually exclusive\nfrontal and occipital gene-expression signatures, with resulting gradients\ngiving rise to the specification of areas between these two poles throughout\nsuccessive developmental timepoints.\n",
  "methods": "\n<section>Sample acquisition</section>\n<p>\nDe-identified tissue samples were collected with previous consent in strict\nobservance of the legal and institutional ethical regulations. Protocols were\napproved by the Human Gamete, Embryo, and Stem Cell Research Committee\n(institutional review board) at the University of California, San Francisco.\nTwo sets of samples included twins: GW20_31 and GW20_34; GW22 and GW22T.\n\n<section>Single-cell RNA sequencing capture and processing</section>\n<p>\nBrain dissections were performed under a stereoscope with regards to major\nsulci to identify cortical regions. Of note, all dissections were performed by\nthe same individual (T.J.N.) to enable reproducibility and comparison between\nsamples. Tissue was incubated in 4 ml of papain/DNAse solution (Worthington)\nfor 20 min at 37 °C, after which it was carefully triturated with a glass\npipette, filtered through a 40-µm cell strainer and washed with HBSS. The GW22\nand GW25 samples were additionally passed through an ovomucoid gradient\n(Worthington) in order to minimize myelin debris in the captures. The final\nsingle-cell suspension was loaded onto a droplet-based library prep platform\nChromium (10X Genomics) according to the manufacturer’s instructions. Version 2\nwas used for all samples except for GW19_2, GW16, and GW18_2 for which version\n3 chemistry was used. cDNA libraries were quantified using an Agilent 2100\nBioanalyzer and sequenced with an Illumina NovaSeq S4.\n\n<section>Quality control and filtering</section>\n<p>\nWe filtered cells using highly stringent quality control (QC) metrics. In\nbrief, we discarded potential doublets using the R package scrublet29 for each\nindividual capture lane, then required at least 750 genes per cell and removed\ncells with high levels (>10%) of mitochondrial gene content. These strict\nmetrics for quality control preserved no more than 40% of cells for downstream\nanalysis, and re-analysis of the data for specific brain structures or cell\ntypes may benefit from less stringent QC for additional discovery. Our goal was\nto obtain clean populations with a high validation rate for a better\nunderstanding of arealization signatures. The resulting ~700,000 cells passing\nall thresholds were used in downstream analyses.\n\n<section>Clustering strategy</section>\n<p>\nWe used a recursive clustering workflow to understand the cell types present in\nour dataset. In order to minimize potential batch effects and to increase\ndetection sensitivity of potential rare cell populations, we performed\nLouvain–Jaccard clustering on each individual sample first. After initial cell\ntype classification, we sub-clustered all the cells belonging to a cell type to\ngenerate the most granular cell subtypes possible. We then correlated subtypes\nbetween individuals based upon the gene scores in all marker genes to bridge\nany batch effects, and iteratively combined clusters across all individuals and\ncell types. For this study, we combined the clusters within a single cell type\nacross all individuals once, and again with all clusters from all individuals\nand cell types, resulting in two iterative combinations. The annotations at\neach step are preserved in the supplementary tables to enable reconstruction at\nany point in the pipeline.\n",
  "paper_url": "https://www.nature.com/articles/s41586-021-03910-8 Bhaduri et al. Nature. 2021.",
  "other_url": "https://data.nemoarchive.org/biccn/grant/u01_devhu/kriegstein/transcriptome/scell/10x_v2/human/processed/counts/ Raw Count Data in NeMO",
  "pmid": "34616070",
  "submitter": "Aparna Bhaduri",
  "version": 1,
  "submission_date": "2021-10-15",
  "lab": "Arnold Kriegstein Lab",
  "institution": "UCSF",
  "inDir": "/hive/data/inside/cells/datasets/dev-brain-regions"
}