Transcriptomic evidence that von Economo neurons are regionally specialized extratelencephalic-projecting excitatory neurons

Abstract

von Economo neurons (VENs) are bipolar, spindle-shaped neurons restricted to layer 5 of human frontoinsula and anterior cingulate cortex that appear to be selectively vulnerable to neuropsychiatric and neurodegenerative diseases, although little is known about other VEN cellular phenotypes. Single nucleus RNA-sequencing of frontoinsula layer 5 identifies a transcriptomically-defined cell cluster that contained VENs, but also fork cells and a subset of pyramidal neurons. Cross-species alignment of this cell cluster with a well-annotated mouse classification shows strong homology to extratelencephalic (ET) excitatory neurons that project to subcerebral targets. This cluster also shows strong homology to a putative ET cluster in human temporal cortex, but with a strikingly specific regional signature. Together these results suggest that VENs are a regionally distinctive type of ET neuron. Additionally, we describe the first patch clamp recordings of VENs from neurosurgically-resected tissue that show distinctive intrinsic membrane properties relative to neighboring pyramidal neurons.

Fig. 1. Cell type characterization in human frontal agranular insular cortex (FI).

a Schematic diagram illustrating nuclei isolation from postmortem human brain specimens. The FI region was isolated, vibratome sectioned, stained with fluorescent Nissl, and layer 5 was dissected and processed for nuclei isolation, fluorescence-activated cell sorting (FACS), and RNA-sequencing. Examples of cells with morphologies typical of von Economo neurons (VENs) are shown in the images of Nissl-stained tissues (arrowheads). Human brain image © 2010 Allen Institute for Brain Science. Allen Human Brain Atlas. Available from: http://human.brain-map.org/. In total 561 single layer 5 neurons passed quality control. b Hierarchical representation of 18 neuronal (5 inhibitory, 13 excitatory) and 4 non-neuronal transcriptomic cell types based on median cluster expression. Major cell classes are labeled at branch points in the dendrogram. The bar plot and associated numbers below the dendrogram represent the number of nuclei within each cluster. Cluster-specific colors and labels are used in all subsequent figures. c Heatmap showing the expression of cell class marker genes across all clusters. Maximum expression values for each gene are listed on the far right-hand side of the plot. Gene expression values are quantified as counts per million of intronic plus exonic reads and displayed on a log10 scale, using a blue-white-red color scheme with blue = 0 and red = the maximum value in the plot (5 × 10³). d Violin plots showing expression of four marker genes per excitatory cluster. Each row represents a gene, black dots show median gene expression within clusters, and maximum expression values for each gene are shown on the right-hand side of each row. Gene expression values are displayed on a linear scale. Box indicates putative VEN cluster.

Fig. 2. Identifying a transcriptomic cell type that corresponds to von Economo neurons (VENs) in situ.

a Violin plots showing distributions of genes further examined by in situ hybridization (ISH). Each row represents a gene, black dots indicate median gene expression within clusters, and maximum gene expression values are shown on the right-hand side of each row. Gene expression values are displayed on a linear scale. The EXC GABRQ FEZF2 type expresses GABRQ and ADRA1A, previously defined markers of VENs. b Chromogenic single gene ISH from the Allen Human Brain Atlas (http://human.brain-map.org/) for GABRQ and ADRA1A confirms a subset of layer 5b cells expressing these genes have spindle-shaped cell bodies typical of VENs (red arrows). The nearest Nissl-stained section is shown for each ISH image for laminar context. c ISH from the Allen Human Brain Atlas (http://human.brain-map.org/) for genes expressed in other excitatory neuron types revealed by our analyses. Genes are expressed in and around layer 5 of FI but labeled cells lack spindle-shaped cell bodies typical of VENs. Red arrows in the nearest Nissl-stained section for each ISH image show representative examples of cells with VEN morphology in the approximate region highlighted in the neighboring ISH image (red rectangle). Scale bars in b, c low magnification, 150 μm, high magnification 50 μm. d Multiplex fluorescent ISH (top left) and double chromogenic ISH for marker genes of Exc FEZF2 GABRQ. Cells with pyramidal (P), VEN (V), and fork (F) morphologies are indicated by labeled arrows in each image. Scale bars, 10 μm. e Quantification of the proportion of ADRA1A+, POU3F1+ cells with pyramidal vs. VEN morphologies (n = 5 human donors). Cells lacking defining features of these morphological classes were called uncharacterized. Bars show the mean and error bars the standard deviation. Individual data points are overlaid on each bar plot.

Fig. 3. Evolutionary conservation of cell types between human and mouse predicts that VENs project sub-cortically.

a, b Excitatory neurons in human FI (red), human MTG (green), mouse ALM (cyan), and mouse VISp (purple) were integrated and aligned using Seurat v3 with default parameters, and visualized using UMAP. a Cells from each data-set co-cluster, indicating good matching of types between brain regions and species. b Eight Seurat clusters were identified using the Louvain algorithm and labeled based on expected cortical layer and projection target (as described in c). c Membership of cells from excitatory clusters in each data set in the Seurat clusters. Colors indicate the fraction of total cells per cluster assigned to each Seurat cluster (rows sum to 1). Data set clusters are grouped based on maximal fraction of cells in the cluster. Cortical layer of each cluster are inferred based on predominant cortical layer of cells from mouse and human data sets, with the exception of the layer 4 cluster (L4* IT) which primarily includes cells from layer 5 in structures without a layer 4. Projection targets of clusters are inferred based on known projection targets of clusters in mouse ALM and VISp (IT—intratelencephalic, ET—extratelencephalic, NP—near-projecting, CT—corticothalamic). The box highlights that Exc FEZF2 GABRQ is part of the L5 ET cluster.

Fig. 4. Putative extratelencephalic (ET) cells in human frontoinsula (FI) and middle temporal gyrus (MTG) share many common markers but differ in frequency.

a, b Violin plots in human FI (a) and MTG (b) showing expression of all genes enriched in Exc FEZF2 GABRQ, and a subset of genes enriched in the corresponding cluster in human MTG that show divergent patterning between these regions (see the “Methods” section), including many novel marker genes. Gene expression values are displayed on a log2 scale. c Representative inverted images of DAPI-stained sections of FI and MTG. Red dots depict the locations of cells labeled using multiplex fluorescent in situ hybridization (mFISH) for putative ET marker gene POU3F1 and SLC17A7. Scale bars in c: DAPI images 50 μm, mFISH images 10 μm. d Black, quantification of the proportion of SLC17A7+ cells expressing the ET marker POU3F1 in FI and MTG expressed as a fraction of the total number of excitatory (SLC17A7+) cells in layer 5 of either region. Individual data points are represented by black squares (FI) or black dots (MTG). Green, comparable quantification of the fraction of excitatory neurons dissected from layer 5 that are assigned to the ET clusters Exc L4-5 FEZF2 SCN4B (MTG) or Exc FEZF2 GABRQ (FI). By both mFISH and single nucleus RNA-seq (snRNA-seq), a higher fraction of putative ET cells is found in FI than MTG. Bars show the mean and error bars the standard deviation.

Fig. 5. Distinctive electrophysiological properties of putative L5 VENs in ex vivo insula brain slices from a human neurosurgery patient.

a MRI scan indicating the location of the excised insula tissue specimen for research. b Best matched location in the Allen 2D coronal human brain reference atlas, with crosshairs centered on the short insular gyrus. Scale bar: 1 cm. c Biocytin-filled putative VEN in L5 of an ex vivo insula brain slice. Low magnification brightfield and DAPI image confirms the L5 location of the neuron. The boxed region bounding the biocytin-filled neuron is expanded at right. Inset: image of Alexa dye fill following patch clamp recording in live tissue. Scale bars: 1 mm and 100 µm. d Example traces of action potential firing pattern in response to current injection steps for a representative pyramidal neuron (PN) and VEN. Scale bars: 50 pA, 500 msec. e Summary plot of action potential firing in response to current injection steps. *p < 0.0001, 2-way ANOVA. f Summary plot of coefficient of variation (CV) for VENs vs. PNs. *p < 0.05, Mann–Whitney. g Summary plot of spike frequency adaptation (SFA) for VENs vs. PNs. *p < 0.05, Mann–Whitney. Center lines show mean and error bars show standard error of the mean.