Introduction
Neural stem/progenitor cells (NSPC) persist in the adult mouse brain in the walls of the forebrain ventricles. This neurogenic niche includes the ventricular-subventricular zones in the walls of the lateral ventricles (V-SVZ), home to a subpopulation of astrocytes (B cells) that function as the NSPCs (Chaker et al., 2016; Doetsch et al., 1997; Ihrie and Alvarez-Buylla, 2011; Lim and Alvarez-Buylla, 2014; Mirzadeh et al., 2008). This neurogenic region has also been referred to as the SVZ or the subependymal zone (Kazanis et al., 2017). B cells generate intermediate progenitors (C cells) that, in turn, give rise to neuroblasts (A cells) that migrate to the olfactory bulb (OB) (Obernier et al., 2018; Ponti et al., 2013). A subpopulation of B cells also generate oligodendrocytes (Figueres-Oñate et al., 2019; Gonzalez-Perez, 2014; Kazanis et al., 2017; Menn et al., 2006; Nait-Oumesmar et al., 1999; Picard-Riera et al., 2002). From the initial interpretation that adult NSPCs are multipotent and able to generate a wide range of neural cell types (Morshead et al., 1994; Reynolds and Weiss, 1992; van der Kooy and Weiss, 2000), more recent work suggests that the adult NSPCs are heterogeneous and specialized, depending on their location, for the generation of specific types of neurons, and possibly glia (Chaker et al., 2016; Delgado et al., 2020; Fiorelli et al., 2015; Merkle et al., 2014, Merkle et al., 2007; Tsai et al., 2012). Previous single-cell sequencing experiments in the V-SVZ have described the many broad classes of cells that reside in the niche. For example, transcriptional analyses after cell sorting have identified stages in the B-C-A cell lineage (Borrett et al., 2020; Codega et al., 2014; Dulken et al., 2017; Xie et al., 2020), as well as populations of NSPCs that appear to activate after injury (Llorens-Bobadilla et al., 2015). Profiling of the entire niche has highlighted differences between quiescent and activated B cells (Mizrak et al., 2020; Zywitza et al., 2018). However, the differences among B cells of equivalent activation state (e.g. quiescent, primed, or activated) or the B cell heterogeneity that leads to the generation of diverse neuronal subtypes remain poorly understood.
NSPC heterogeneity, interestingly, is largely driven by their location within the adult V-SVZ. This concept explains why young neurons in the OB originate over such a wide territory in the walls of the lateral ventricles. Multiple studies have begun to identify regional differences in gene expression among the lateral, septal, and subcallosal walls of the lateral ventricles (Delgado et al., 2021). For example, differences in gene expression of B cells from the septal and lateral walls of the lateral ventricles have been recently observed (Mizrak et al., 2019). Other studies have shown that
Here, we have undertaken single-cell and single-nucleus RNA sequencing of the microdissected V-SVZ to gain insight into these important questions regarding NSPC heterogeneity and their developmental potential. Clustering analysis reveals strong dorso-ventral differences in lateral wall B cells. Validation of these differential gene expression patterns has revealed the anatomical boundary that separates these dorsal and ventral B cell domains. Additionally, our analysis identifies subpopulations of A cells defined by maturation state and dorso-ventral origin. We also find that a subset of dorso-ventral B cell transcriptional differences are retained through the C and A cell stages of the lineage. These new data advance our molecular understanding of how major region-specific neural lineages emerge in the adult V-SVZ and begin to delineate major functional subclasses of adult-born young neurons.
Results
Single-cell RNA sequencing distinguishes B cells from parenchymal astrocytes and reveals B cell heterogeneity
For whole single-cell RNA sequencing (scRNA-Seq), we dissected the lateral wall of the lateral ventricle from hGFAP:GFP mice at postnatal day (P) 29–35 (n=8, Figure 1A; Figure 1—figure supplement 1A). To determine possible sex differences in downstream analyses, two male and two female samples (n=four samples total) were dissociated and multiplexed by labeling cells with sample-specific MULTI-seq barcodes (McGinnis et al., 2019a). Multiplexed samples were then pooled for the remainder of the single-cell isolation protocol. Two technical replicates of pooled samples were loaded in separate lanes of the Chromium Controller chip (10x Genomics) for single-cell barcoding and downstream mRNA library preparation and sequencing (Figure 1A). Cells carrying multiple barcodes or a high number of mRNA reads (4128 out of 35,025 cells, 11.7%) were considered doublets and were eliminated from downstream analysis. Data from each technical replicate were integrated for batch correction (Stuart et al., 2019). We then performed unbiased clustering of cell profiles and calculated UMAP coordinates for data visualization (Figure 1A). The clustering of lateral wall V-SVZ cells was not driven by sample, technical replicate, or sex (Figure 1—figure supplement 1B–J). Cell cluster identities were annotated based on the detection of known cell type markers (Figure 1B–C). We identified 37 clusters, with 14 clusters corresponding to cell types within the neurogenic lineage: NSPCs (B cells), intermediate progenitors (C cells), and neuroblasts (A cells) (Doetsch et al., 1999; Obernier et al., 2018). In addition, our analysis identified cell clusters corresponding to parenchymal astrocytes, ependymal cells, neurons, oligodendroglia, microglia, pericytes, vascular smooth muscle cells, and endothelial cells (Figure 1B–C).
Figure 1.
Whole-cell sequencing captures the cellular diversity and activation cascade of the adult neurogenic niche.
(A) Schematic of the whole-cell single-cell isolation and sequencing protocol (scRNA-Seq). The lateral wall of the lateral ventricles were microdissected from young adult hGFAP:GFP mouse brains (n=four males, n=four females). Four samples were multiplexed with MULTI-seq barcodes and combined together. Two 10x Chromium Controller lanes were loaded as technical replicates, and cells were sequenced and processed for downstream analysis (Figure 1—figure supplement 1). V-SVZ: Ventricular-Subventricular zone. (B) UMAP plot of scRNA-Seq cell types captured after demultiplexing and doublet removal. (C) Dot plot of cell-type-specific marker expression in the clusters from (B) (Figure 1—figure supplement 2).
Figure 1—figure supplement 1.
Biological and technical replicate metadata of the scRNA-Seq dataset.
(A) Diagram representing V-SVZ microdissected areas (red). The V-SVZ was microdissected from 1 mm brain slices from anterior (bregma 1.70) to Posterior bregma (−0.10 mm) regions. Single-cell suspensions from two female samples and two male samples were multiplexed with MULTI-seq barcodes (Bar1-4). (B) UMAP plot of sequenced cells after demultiplexing and quality control filtering, color-coded by MULTI-seq barcode. Cells with no detectable barcode are labeled ‘Negative’. (C) Ratio of cells within each cluster (see Figure 1A, right panel) coming from each Barcode sample. (D) UMAP plots of cells separated by Barcode number and color-coded by cluster number (see Figure 1A, right panel). (E) UMAP plot of sequenced cells color-coded by animal sex: Female cells (green) correspond to Barcodes 3 and 4 combined, male cells (tan) correspond to Barcodes 1 and 2 combined. Cells with no detectable barcode are labeled ‘Unknown’ (gray). (F) Ratio of cells within each cluster (see Figure 1A, right panel) coming from male (tan) or female (green) mice, or of unknown origin (gray). (G) UMAP plots of cells separated by animal sex and color-coded by cluster number (see Figure 1A, right panel). (H) UMAP plot of sequenced cells color-coded by 10x Chromium Controller Chip lane (equivalent to ‘batch’, or technical replicate): Lane 1 (orange) or Lane 2 (navy). (I) Ratio of cells within each cluster (see Figure 1A, right panel) coming from Lane 1 (orange) or Lane 2 (navy). (J) UMAP plots of cells separated by Lane and color-coded by cluster number (see Figure 1A, right panel).
Figure 1—figure supplement 2.
Transcriptomic profile of B cells versus Parenchymal Astrocytes.
(A) UMAP plot of V-SVZ neurogenic lineage cells (blue) and parenchymal astrocytes (purple) isolated from the scRNA-Seq dataset (inset, boxed area). (B) Gene expression of B cell-enriched markers
NSPCs in the V-SVZ correspond to a subpopulation of astrocytes (B cells) derived from radial glia (Doetsch et al., 1999; Laywell et al., 2000; Merkle et al., 2004). B cells have ultrastructure and markers of astrocytes (Borrett et al., 2020; Codega et al., 2014). Therefore, identifying markers that distinguish parenchymal astrocytes from B cells has been a challenge in the field. A fraction of both populations expressed
To better understand the biological differences between parenchymal astrocytes and B cells, we performed gene ontology (GO) enrichment on the differentially expressed genes. Genes associated with synapse regulation (GO:0051965 and 0051963), macroautophagy (GO: 0016241), and dendrite development and morphogenesis (GO: 0016358 and 0048813), among others, were overrepresented in parenchymal astrocytes compared to B cells (Figure 1—figure supplement 2D). In contrast, B cells were enriched in terms associated with RNA regulation (GO:0000463, 0034471, 0000956, and 0000966) and mitochondrial regulation (GO:006626, 0073655, 0090201, and 0010823) (Figure 1—figure supplement 2D). These differentially represented GO terms support the known neuro-regulatory function of parenchymal astrocytes, as well as the increased transcriptional regulation that has been associated with the transition of NSPCs from quiescent to activated states (Dulken et al., 2017; Llorens-Bobadilla et al., 2015). Using this cell type classification, we included B cell clusters, but not striatal astrocytes, in our downstream analysis of the neurogenic lineage.
scRNA-Seq captures neurogenic progression in the V-SVZ
In our scRNA-Seq dataset, we found that the majority of cells were part of the neurogenic lineage, which is composed of primary progenitors, intermediate progenitors, and young neurons (the B-C-A cell lineage). These neurogenic lineage clusters were in the center of the UMAP plot. With this analysis, the data had a ‘hummingbird-like’ shape with B cells (clusters 5, 13, 14, and 22) at the top in the bird’s head and neck, proliferating cells (clusters 8, 10, 16, and 17) and C cells (cluster 12) resembled the bird’s body and wing, and A cells (clusters 0, 1, 4, 6, and 15) formed a tail (Figure 2A). As expected, all B cell clusters expressed
Figure 2.
Characterization of the scRNA-Seq V-SVZ neurogenic lineage.
(A) The neurogenic lineage has a ‘bird-like’ shape, with B cells forming the head (blue), C cells in the body (green), dividing cells in the wing (yellow), and A cells in the tail (red). These are divisible into 14 distinct clusters, including 4 B cell clusters, 1 C cell cluster, 4 clusters of dividing cells, and 5 A cell clusters (right). (B-J). Gene expression captures progression along the lineage, with canonical markers of each stage expressed in its corresponding region of the UMAP plot. (K) Scoring cells by phase of the cell cycle reveals cells in G2M and S phase occupying the wing of the bird. (L-M). Pseudotime calculated by RNA velocity recapitulates the B to C to A trajectory along the neurogenic lineage. (M) Genes associated with B cells (
The unbiased clustering analysis pooled dividing cells into the wing region of the ‘hummingbird’. A closer look at these clusters of mitotic cells in G2 or metaphase (G2M; clusters 10, 16, 17, and 8) showed that their gene expression pattern overlaps with that of the non-dividing neurogenic lineage cell progression: Cluster 10 expresses markers of B cells (
Identified by
Overall, our single-cell dataset recapitulates the known B-C-A cell progression through the neurogenic lineage. Intriguingly, our analysis also reveals heterogeneity among B, C, and A cells, in which each cell type is subdivided into multiple distinct clusters. Among C cells, the heterogeneity was mostly driven by different stages of the cell cycle (Figure 2F–G,I,K). What drives heterogeneity among B and A cell clusters?
Quiescent B cell clusters correspond to regionally organized dorsal and ventral domains
In our scRNA-Seq dataset, we found that quiescent B cells (the head of the ‘bird’; Figure 2A) were subdivided into three clusters: B cell cluster 5 (B(5)), B(14), and B(22) (Figure 3—figure supplement 1A). To understand their molecular differences, we conducted differential expression analysis to identify significantly upregulated genes in each of the three B cell clusters (Figure 3—figure supplement 1B(i)) and candidate cluster-specific marker genes (Figure 3—figure supplement 1B(iI)) (Supplementary file 3). When we examined the top ten candidate markers for each cluster, we found genes corresponding to known markers of dorsal and ventral B cell identity (Figure 3—figure supplement 1C). For example,
Figure 3.
scRNAseq reveals regional heterogeneity among adult neural stem cells.
(A-C) UMAP plots of
Figure 3—figure supplement 1.
Identification of quiescent B cell clusters.
(A) UMAP plot of B cell cluster identities used in the following analysis: B(14), B(5), and B(22). (B) i Venn diagram summarizing differential gene expression analysis between clusters B(14), B(5), and B(22). ii Numbers of candidate marker genes identified after selecting significantly upregulated genes expressed in no more than 40% of cells of the other cluster. (C) Heatmap depicting expression of the top 10 differentially expressed genes between clusters B(14) (left), B(5) (center), and B(22) (right). (D) Dot plot of the average Dorsal or Ventral Dissection predicted identity scores for scRNA-Seq B cell clusters B(14), B(5), and B(22). (E) UMAP plot of
Figure 3—figure supplement 2.
Characterization of the sNucRNA-Seq dataset.
(A) Diagram representing the four V-SVZ microdissected regions: Anterior ventral (AV, pink), anterior dorsal (AD, dark green), posterior ventral (PV, magenta), and posterior dorsal (PD, light green). (B) Single nucleus suspensions from each region were processed as separate samples in parallel. (C) UMAP plot of sequenced nuclei after quality control filtering, color-coded by cluster number. (D) UMAP plot of sNucRNA-Seq cell types. (E) Dot plot of cell-type-specific marker expression in the clusters from (D). (F) UMAP plot of sequenced cells color-coded by region: AD (green), PD (light green), AV (pink), and PV (magenta). (G) UMAP plots of nuclei separated by region and color-coded by cluster number (see panel C). Below each plot is the number of nuclei and the percent of the total dataset originating from each region. (H) Ratio of nuclei within each cluster (see panel C) coming from each region sample. (I) Violin plot of the number of unique molecular identifiers (UMIs) detected per nucleus or cell (left panel), and number of genes detected per cell (right panel) in each sNucRNA-Seq region and in the scRNA-Seq dataset (white). Median UMIs: 9181scRNA-Seq/3,062sNucRNA-Seq, a threefold difference. Median genes: 3549scRNA-Seq/1,679sNucRNA-Seq, a 2.1-fold difference). (J) Pie charts of cell types represented in sNucRNA-Seq (left) and scRNA-Seq datasets (right). The scRNA-Seq dataset contains 60.4% neurogenic lineage cells (shades of orange; 14,660/24,261 total cells), compared to 10.6% in the sNucRNA-Seq dataset (4859/45,820 total nuclei). Conversely, the sNucRNA-Seq dataset contains 61.5% neurons (shades of purple; 28,185 nuclei, 13 subtypes (panel D), while the scRNA-Seq contains only 1.6% (395 cells, three subtypes; Figure 1A–B). The scRNA-Seq and sNucRNA-Seq datasets each contain 37.9% (9206 cells) and 25.1% (11,480 nuclei) glial cells, respectively (shades of plum). Cell types listed in each pie chart legend are plotted in order, clockwise from coordinate 0.
Figure 3—figure supplement 3.
Gene Regulatory networks of dorsal and ventral B cell markers.
(A-B) Networks of predicted regulatory relationships (by GENIE3, visualized with Cytoscape) between (A) the top 10 markers of putative dorsal B cells (B(5+22)) and all transcripts expressed in B(5+22) or (B) the top 10 markers of putative ventral cells (B(14)) and all transcripts expressed in B(14). Color scale represents log2 fold change between B(5+22) and B(14), with genes upregulated in B(5+22) in green and genes upregulated in B(14) in purple. Black outlines demarcate the top 10 differentially expressed genes used to construct each network. Arrow thickness corresponds to the weight of the link between each pair of nodes. (A) The core dorsal network contains highly connected nodes including cell-cycle gene
Figure 3—figure supplement 4.
(A-B) Network of predicted regulatory relationships between (A)
To confirm regional organization as a major source of heterogeneity in our scRNA-seq dataset, we turned to a previously generated single-nucleus RNA sequencing (sNucRNA-Seq) dataset we had derived from regionally microdissected V-SVZ. We performed single-nucleus sequencing (sNucRNA-Seq) from microdissected V-SVZ subregions of P35 CD1 mice (n=eight males, nine females). We isolated single nuclei from four microdissected quadrants of the V-SVZ: the anterior-dorsal (AD), posterior-dorsal (PD), anterior-ventral (AV), and posterior-ventral (PV) regions (Figure 3—figure supplement 2A; Mirzadeh et al., 2008). The four region samples (AD, PD, AV, and PV) were then processed in parallel for sNucRNA-Seq (Figure 3—figure supplement 2B). Our sNucRNA-Seq dataset contains 45,820 nucleus profiles. The four region samples underwent quality control steps of filtering out low-quality cells and putative doublets (see Materials and methods). Data from each sample were combined and integrated (Seurat v3
In the sNucRNA-Seq data, we identified 42 clusters, including those corresponding to cell types within the neurogenic lineage: NSPCs (B cells), mitotic intermediate progenitors (C cells), and neuroblasts (A cells) (Figure 3—figure supplement 2C–E). Based on the B cell- or astrocyte-specific markers identified in the scRNA-Seq data above, we also identify a parenchymal astrocyte cluster, as well as ependymal cells, striatal neurons, oligodendroglia, microglia, pericytes and vascular smooth muscle cells, endothelial cells, and leptomeningeal cells (Figure 3—figure supplement 2D–E). We also found that all four regions contributed to most clusters (Figure 3—figure supplement 2F–H).
To test the hypothesis that B cells are dorso-ventrally organized in the V-SVZ, we took advantage of the region-specific microdissection of the sNucRNA-Seq cells (Figure 3—figure supplement 2) and metadata Label Transfer to predict scRNA-Seq B cell region identity (Stuart et al., 2019). Each B cell was assigned both a dorsal and ventral ‘predicted identity’ score based on their similarity to dorsal and ventral nuclei (Figure 3D). We then calculated the difference between dorsal and ventral scores for each scRNA-Seq B cell. We found that cells within each cluster were strongly dorsal-scoring (green) or ventral-scoring (magenta), with relatively few cells having similar dorsal and ventral prediction scores (gray) (Figure 3E). We found that cluster B(14) scored more highly for ventral identity on average, while clusters B(5) and B(22) scored more highly for dorsal identity (Figure 3—figure supplement 1D).
To investigate the potential dorso-ventral spatial organization of dorsal-scoring clusters B(5) and B(22) in vivo, we performed RNAscope in situ hybridization for the differentially expressed transcripts
We then asked what genes were differentially expressed between the putative dorsal cluster B(5+22) and the putative ventral cluster B(14) (Figure 3G–I). Among the differentially expressed genes were other known markers of dorsal B cells, such as
Cluster B(14) marker
Figure 4.
Whole mounts reveal CRYM expression in a wide ventral domain.
(A-C) Confocal micrograph of a hGFAP:GFP coronal mouse brain section, where the V-SVZ is immunostained for GFP (green) and CRYM (magenta). B. High-magnification image of the dorsal wedge region of the V-SVZ. (C) High-magnification image of the ventral V-SVZ. (D-G) Immunostaining of CRYM (magenta) in a whole-mount preparation of the lateral wall of the V-SVZ, co-stained with β-CATENIN (white), with a summary schematic depicting the extent of the CRYM+ domain. (E–G) Higher magnification images of boxed regions in (D) showing the distribution of CRYM+ (magenta) in V-SVZ cells outlined by β-CATENIN (white). (H-I) High-magnification images of GFP+ (green) B1 cell-containing pinwheels in the dorsal (H) and ventral (I) V-SVZ outlined by β-CATENIN (white), also immunostained for CRYM (magenta). Quantifications showed that 95.11% ± 2.65 (SD) of the GFP+ B1 cells were CRYM+ in the ventral domain of the V-SVZ. In contrast, only 4.71% ± 1.38 (SD) of the GFP+ B1 cells in the dorsal region were CRYM+ (n=3; T-test, p<0.0001). (J-K) Immunogold transmission electron micrographs of CRYM+ B1 cell in the V-SVZ. K. High magnification of B1 cell apical contact with the lateral ventricle. A: anterior, P: posterior, D: dorsal, V: ventral. Scale bars: 150 μm (A), 20 μm (B and C), 200 μm (D), 50 μm (E, F and G), 10 μm (H and I), 2 μm (J), and 500 nm (K).
We then conducted GO analysis on differentially expressed genes between dorsal B(5+22) and ventral B(14) cells and found that differentially expressed genes in ventral B cells had an overrepresentation of genes involved in the response to growth hormone (GO:0060416) and retinal ganglion cell axon guidance (GO:0031290), while dorsal B cells were associated with oligodendrocyte differentiation (GO:0048709), forebrain generation of neurons (GO:0021872), and central nervous system neuron axonogenesis (GO:0021955)(Supplementary file 3).
To understand the relationships between differentially expressed genes within dorsal and ventral B cell populations, we used gene regulatory network (GRN) analysis. We constructed GRNs based on predicted interactions between the top 10 markers in dorsal B(5+22) cells or ventral B(14) cells and all other genes expressed in each cluster. Dorsal marker genes and their interaction partners formed one large network of 146 genes and three smaller networks with 11–20 genes each (Figure 3—figure supplement 3A), while ventral markers formed two large networks (118 and 91 genes) and three smaller ones containing 20–32 genes each (Figure 3—figure supplement 3B). The dorsal network was much more highly interconnected than the ventral network (dorsal clustering coefficient = 0.121, ventral clustering coefficient = 0.005). The central dorsal network was enriched for genes associated with regulation of apoptosis (GO:0042981) and central nervous system axonogenesis (GO:0021955), while ventral networks were enriched for genes associated with cell fate commitment (GO:0045165), regulation of GABAergic synaptic transmission (GO:0032228), and regulation of transcription (GO:0045449). Notably, the dorsal network contained all three genes that make up nuclear receptor subfamily 4A (
A cell cluster heterogeneity is linked to regionally organized dorsal and ventral domains
We found that A cells (Figure 5A) were separated into two main sets of transcriptionally-related clusters: clusters A(15) and A(6) corresponded to A cells with a strong expression of genes associated with mitosis and cell cycle regulation (such as
Figure 5.
A cell cluster heterogeneity is linked to regionally organized dorsal/ventral domains.
(A) A cells are organized in five distinct clusters (1, 0, 4, 6, and 15). (B) Heatmap showing the top five differentially expressed genes for each of the five clusters. (C) The expression patterns of
Figure 5—figure supplement 1.
Characterization of A cell clusters.
(A) UMAP plot of A cell clusters, and UMAP plots of only A cell clusters labeled with AUCell scores of combined gene expression of genes contained in each of the six GO categories shown. (B)
Interestingly, the combined expression of genes in dorsoventral axonal guidance (GO:0033563) and cerebral cortex regionalization (GO:0021796), terms associated with the regional specification of the brain, was high in clusters A(1) and A(0), despite individual GO terms not being statistically enriched in these clusters (Figure 5—figure supplement 1A; Supplementary file 4). Additionally,
In order to confirm that A cells in A(0) and A(1) correspond to dorsal and ventral young neurons, respectively, we looked for markers of A(1) and A(0) that were minimally present in the other cluster.
Specific markers link regionally distinct B cell and A cell lineages
Interestingly, A(1) marker
Figure 6.
Regional transcriptional signatures are maintained along the neurogenic lineage.
(A) UMAP plot highlighting the putative dorsal and ventral B and A cell clusters. (B) Schematic illustrating the approach to identify genes that are differentially enriched in dorsal B and A, and ventral B and A cells, comparing B(14) to A(1) and B(5) and B(22) to A(0). (C) Expression patterns of dorsal (top row) and ventral markers (bottom row) identified as differentially enriched throughout the B-C-A lineage. (D) Cell scores based on the combined expression of genes in the ventral lineage signature (ventral score; purple) and genes in the dorsal lineage signature (dorsal score; green) throughout the neurogenic lineage. Lineage scores, as the combined expression of genes of the dorsal and ventral signatures, remain relatively constant throughout all cells along the neurogenic lineage. (E) High Score (top quartile of each lineage score) dorsal (green) and ventral (purple) cells in the neurogenic lineage. (F) Expression of dorsal signature genes in High Score dorsal (left) and High Score ventral cells (right) along the neurogenic lineage progression, assessed by pseudotime (see also Figure 6—figure supplement 1). (G) Expression of ventral signature genes in High Score dorsal (left) and High Score ventral cells (right) along the neurogenic lineage progression, assessed by pseudotime. (H) Volcano plot of significantly differentially expressed genes between High Score dorsal and High Score ventral neurogenic lineages. (I - J) RNAscope validation of dorsal lineage marker
Figure 6—figure supplement 1.
Characterization of dorsal and ventral neurogenic lineages.
(A) Schematic illustrating the control comparison to identify genes enriched in both ventral B and dorsal A cells, or in dorsal B and ventral A cells. (B) UMAP plots showing expression of the three genes identified in the analysis in (A), including
To understand the molecular differences between the putative dorsal and ventral lineages, we used the regional gene sets we identified above to calculate a composite AUC score (Aibar et al., 2017) for both the dorsal and ventral gene expression signatures (Figure 6D, Figure 6—figure supplement 1C). We found that cells that scored highly for the dorsal genes were largely located on the right side of the ‘bird’, and the highest ventral-scoring cells were on its left side (Figure 6D, Figure 6—figure supplement 1C–E). In order to understand the functional differences between cells in dorsal and ventral lineages, we normalized the dorsal and ventral scores (see Methods) and selected the top-scoring quartile for each lineage (High Score lineages) to compare their gene expression (Figure 6E). We found that genes enriched in dorsal or ventral lineages have dynamic expression patterns along pseudotime. For example,
To validate RNA expression of putative dorsal and ventral lineage markers in vivo, we combined RNAscope labeling with GFAP and DCX immunostaining in coronal sections of the V-SVZ. We found that the putative dorsal lineage genes
Discussion
During brain development, regional allocations of the neuroepithelium give NSPCs different neurogenic properties. The adult V-SVZ neurogenic niche retains regionally specified NSPCs that generate different subtypes of neurons destined for the OB. A molecular understanding of what makes adult NSPCs different between regions is largely lacking. Our scRNA-Seq and sNucRNA-Seq datasets provide new information about the diverse cell types that populate the V-SVZ. Our lineage analysis reveals parallel pathways of neurogenesis initiated by different populations of B cells. Interestingly, these differences in B cell identity correlate with unique regional patterns of gene expression, which we validated using reference-based metadata label transfer from the second dataset of regionally dissected single V-SVZ nuclei. We confirmed the regional expression of marker genes by immunostaining and RNAscope analysis.
Regional differences in NSPCs potential were demonstrated using restricted viral labeling of non-overlapping territories of the V-SVZ (Merkle et al., 2014; Merkle et al., 2007; Ventura and Goldman, 2007). Labeled ventral B cells produced deep layer granule neurons, calbindin-positive periglomerular cells, and type 1–4 cells in the OB, while dorsal B cells produced superficial granule neurons and tyrosine hydroxylase-positive periglomerular cells (Merkle et al., 2014; Merkle et al., 2007). Similarly, genetic lineage tracing from territories expressing regionally restricted transcription factors also indicates that NSPCs in dorsal and ventral territories generate superficial and deep-layer neurons for the OB, respectively (Kohwi et al., 2007; Kohwi et al., 2005; Merkle et al., 2014; Young et al., 2007). It has been suggested that in the adult V-SVZ, a more primitive population of Oct4+/GFAP- NSCs may be present and that these cells may be earlier in the lineage from the ‘definitive’ GFAP+ B cells (Reeve et al., 2017). However, regionally specified NSPCs can be lineage traced to the embryo (Fuentealba et al., 2015; Furutachi et al., 2015), and we could not detect a population of Oct4+ cells in our datasets. We, however, cannot exclude that rare primitive OCT4+ NSPCs were not captured in our analysis for technical reasons. More recently, genetic labeling of the most ventral domain of the V-SVZ showed the specific contribution of Nkx2.1-expressing B cells to deep layer granule cell neurons in the OB (Delgado and Lim, 2015). The ventral gene expression program is maintained by an epigenetic mechanism across cell divisions. In the absence of myeloid/lymphoid or mixed-lineage leukemia protein 1 (MLL1)-dependent epigenetic maintenance, the neurogenic lineage shifts to produce aberrantly ‘dorsalized’ OB neuronal subtypes (Delgado et al., 2020). This underscores the early embryonic regional specification of adult V-SVZ NSPCs and how these primary progenitors maintain a memory of their regions of origin. Regional genes maintained through the neurogenic lineage could help us understand how NSC identities are maintained to ensure the production of the specific subtypes of interneurons in the OB.
Our dataset is restricted to the V-SVZ, a region where A cells (young neurons) have begun their differentiation, but remain migratory and immature. As A cells move into the OB they complete their differentiation and begin expressing mature neuronal markers like tyrosine hydroxylase and calbindin in different subdomains of the OB. Our study reveals previously unknown markers of young immature type A cells. Previous work has already shown that Pax6, which is associated with a subpopulation of superficial granule neurons and of periglomerular cells, is expressed by a subpopulation of young migrating A cells (Coré et al., 2020; Kohwi et al., 2005). Consistent with these findings, we found
Our dataset provides sets of genes that are differentially expressed in dorsal and ventral B cells. Among these genes, we found well-known regionally expressed transcription factors such as
We confirmed that B cells defined as dorsal or ventral in our scRNA-Seq were predicted to correspond to dorsal and ventral microdissections using unsupervised label transfer of cell identity from the sNucRNA-Seq data (Figure 3D–F). We also found that 59 genes in our scRNA-Seq analysis were highly expressed in sNucRNA-Seq B cells from the dorsal microdissection, including
The border between the
Thyroid hormone signaling has been shown to regulate V-SVZ neurogenesis (Lemkine et al., 2005; López-Juárez et al., 2012; Luongo et al., 2021). Our analysis shows that
In summary, we present a large-scale single-cell description of dorso-ventral identity in the lateral wall of the V-SVZ. Not only do we recapitulate known divisions between dorsal and ventral B cells, but also we identify novel regional B cell markers and uncover gene expression programs that appear to persist throughout lineage transitions (Figure 7). These data form a basis for future investigation of NSPCs identity, lineage commitment, and embryonic origin, providing clues to help us understand how molecularly defined stem cell territories are spatially organized, and what distinguishes V-SVZ regions from one another.
Figure 7.
scRNA-Seq reveals dorsal and ventral neurogenic lineage domains in the V-SVZ.
(A) A summary of cell types in the neurogenic lineage identified by scRNA-Seq and their classification into dorsal and ventral transcriptional identities. (B) Schematic depicting the dorsal and ventral domains newly identified by scRNA-Seq and snucRNA-Seq, and confirmed by staining and RNAscope.
Materials and methods
Key resources table
| Reagent type (species) | Designation | Source or reference | Identifiers | Additional information | |
|---|---|---|---|---|---|
| Genetic reagent | hGFAP:GFP | The Jackson Laboratory (Zhuo et al., 1997) | Cat#003257 | Also referred as hGFAP::GFP: FVB/N-Tg(GFAPGFP)14Mes/J | |
| Antibody | Anti-ß-Catenin (Rabbit polyclonal) | Sigma | Cat#C2206 | IF (1:250) | |
| Antibody | Anti-CRYM (Mouse monoclonal) | Santa Cruz | Cat#sc-376687 | also referred as anti-u-crystallin | |
| Antibody | Anti-DCX (Rabbit polyclonal) | Cell signalling | Cat#4604S | IF (1:200) | |
| Antibody | Anti-GFAP (Chicken polyclonal) | Abcam | Cat#ab4674 | IF (1:500) | |
| Antibody | Anti-GFP (Chicken polyclonal) | Aves Labs | Cat#GFP1020 | IF (1:400) | |
| Antibody | Anti-HOPX (Rabbit polyclonal) | Proteintech | Cat#11419–1-AP | IF(1:500) | |
| Antibody | Anti-S100 (Rabbit polyclonal) | Dako | Cat#Z033 | IF(1:100) | |
| Antibody | Anti-Chicken Alexa Fluor 488 (Donkey) | Jackson ImmunoResearch Labs | Cat#703-545-155 | IF(1:500) | |
| Antibody | Anti-Chicken Alexa Fluor 647 (Donkey) | Jackson ImmunoResearch Labs | Cat#703-605-155 | IF(1:500) | |
| Antibody | Anti-mouse Alexa Fluor 647 (Donkey) | Invitrogen | Cat#A31571 | IF(1:500) | |
| Antibody | Anti-rabbit Alexa Fluor 555 (Donkey) | Invitrogen | Cat#A31572 | IF(1:500) | |
| Antibody | Anti-mouse ultrasmall0.8 nm IgG | Aurion, EMS | Cat#800.022 | IF(1:50) | |
| Antibody | Donkey anti-rabbit biotinylated | Jackson ImmunoResearch | Cat#711-065-152 | IF(1:400) | |
| Antibody | Anti-CD106 (Rat) oligonucleotide tag | BioLegend | Cat#105725 | 1 µg/1 million cells | |
| Antibody | Anti-CD24 (Rat) oligonucleotide tag | BioLegend | Cat#101841 | 1 µg/1 million cells | |
| Sequence-based reagent | MULTIseq barcode | PMID:31209384 | https://gartnerlab.ucsf.edu/more.php | TGAGACCT (A3) | |
| Sequence-based reagent | MULTIseq barcode | PMID:31209384 | https://gartnerlab.ucsf.edu/more.php | GCACACGC (A4) | |
| Sequence-based reagent | MULTIseq barcode | PMID:31209384 | https://gartnerlab.ucsf.edu/more.php | AGAGAGAG (A5) | |
| Sequence-based reagent | MULTIseq barcode | PMID:31209384 | https://gartnerlab.ucsf.edu/more.php | TCACAGCA (A6) | |
| Sequence-based reagent | Chromium i7 Multiplex kit | 10x Genomics | PN-120262 | Index numbers B12 and C1 | |
| Chemical compound | Red blood cell lysis buffer | BioLegend | Cat#420301 | ||
| Chemical compound | Myelin removal beads | Miltenyi Biotec | Cat#130-096-733 | ||
| Chemical compound | TSA fluorescein reagent pack | Akoya Biosciences | Cat#SAT701001EA | ||
| Chemical compound | TNB Blocking Buffer | Akoya Biosciences | Cat#FP1012 | ||
| Commercial assay or kit | Papain Dissociation System-EBSS | Worthington | Cat#LK003150 | ||
| Commercial assay or kit | Chromium Single cell 3’ Library and Gel Bead Kit v2 | 10x Genomics | Cat#PN-120267 | ||
| Commercial assay or kit | Chromium Single cell 3’ GEM, Library and Gel Bead Kit v3 | 10x Genomics | Cat#PN-1000092 | ||
| Commercial assay or kit | RNAscope 2.5 HD Red Detection kit | Advanced Cell Diagnostics | Cat#320497 | ||
| Commercial assay or kit | RNAscope 2.5 HD Duplex Detection kit | Advanced Cell Diagnostics | Cat#322430 | ||
| Commercial assay or kit | Mm-Lphn3 (Adgrl3) | Advanced Cell Diagnostics | Cat#317481 | ||
| Commercial assay or kit | Mm-Crym | Advanced Cell Diagnostics | Cat#466131 | ||
| Commercial assay or kit | Mm-Cnatnp2 | Advanced Cell Diagnostics | Cat#449381 | ||
| Commercial assay or kit | DapB | Advanced Cell Diagnostics | Cat#310043 | used as negative control | |
| Commercial assay or kit | Mm-Dio2 | Advanced Cell Diagnostics | Cat#479331 | ||
| Commercial assay or kit | Mm-Hopx | Advanced Cell Diagnostics | Cat#405161 | ||
| Commercial assay or kit | Mm-Ntng1-C2 | Advanced Cell Diagnostics | Cat#488871-C2 | ||
| Commercial assay or kit | Mm-Pax6 | Advanced Cell Diagnostics | Cat#412821 | ||
| Commercial assay or kit | Mm-PPIB | Advanced Cell Diagnostics | Cat#313911 | used as positive control | |
| Commercial assay or kit | Mm-Rlbp1 | Advanced Cell Diagnostics | Cat#468161 | ||
| Commercial assay or kit | Mm-Snhg15 | Advanced Cell Diagnostics | Cat#889191 | ||
| Commercial assay or kit | Mm-Slit2 | Advanced Cell Diagnostics | Cat#449691 | ||
| Commercial assay or kit | Mm-Trhde | Advanced Cell Diagnostics | Cat#450781 | ||
| Commercial assay or kit | Mm-Urah-C2 | Advanced Cell Diagnostics | Cat#525331-C2 | ||
| Commercial assay or kit | Silver enhancement kit | Aurion, EMS | Cat#25520 | ||
| Software, algorithm | Cellranger 2.1.0-v2.3.0 and | 10x Genomics | RRID:SCR_017344 | ||
| Software, algorithm | Seurat V3 (Rstudio package) | https://satijalab.org/seurat/ | RRID:SCR_007322 | ||
| Software, algorithm | scVelo | https://scvelo.readthedocs.io | RRID:SCR_018168 | ||
| Software, algorithm | Panther | http://www.pantherdb.org | RRID:SCR_004869 | ||
| Software, algorithm | GENIE3 | https://github.com/vahuynh/GENIE3 | RRID:SCR_000217 | ||
| Software, algorithm | BiNGO | https://apps.cytoscape.org/apps/bingo | RRID:SCR_005736 | ||
| Software, algorithm | Cytoscape | https://cytoscape.org | RRID:SCR_003032 | ||
| Software, algorithm | sp | https://edzer.github.io/sp/ | RRID:SCR_021328 | ||
| Software, algorithm | Imaris v9.6–9.7 | Oxford instruments | RRID:SCR_007370 | ||
| Software, algorithm | Adobe Photoshop | Adobe | RRID:SCR_014199 | ||
| Software, algorithm | Adobe Illustrator | Adobe | RRID:SCR_010279 | ||
| Software, algorithm | AUCell 3.13 | https://bioconductor.org/packages/release/bioc/html/AUCell.html | RRID:SCR_021327 | ||
Mice
Mice were housed on a 12 hr day-night cycle with free access to water and food in a specific pathogen-free facility in social cages (up to five mice/cage) and treated according to the guidelines from the UCSF. Institutional Animal Care and Use Committee (IACUC) and NIH. All mice used in this study were healthy and immuno-competent, and did not undergo previous procedures unrelated to the experiment. CD1-elite mice (Charles River Laboratories) and hGFAP::GFP (FVB/N-Tg(GFAPGFP)14Mes/J, The Jackson Laboratory (003257)) (Zhuo et al., 1997) lines were used. Sample sizes were chosen to generate sufficient numbers of high-quality single cells for RNA sequencing, including variables such as sex, and identifying potential batch effects. Biological and technical replicates for each experiment are described in the relevant subsections below.
Single whole cell sample preparation and multiplexing
Mice received intraperitoneal administration of 2.5% Avertin followed by decapitation. Brains were extracted and 1 mm slices were obtained with an adult mouse brain slicer (Steel Brain Matrix - Coronal 0.5 mm, Alto). Four samples were processed: sample 1: two males P35; sample 2: two males P35; sample 3: two females P29; and sample 4: two females P29. The lateral ventricle walls were microdissected in L-15 medium on ice and the tissue was transferred to Papain-EBSS (LK003150, Worthington). Tissue was digested for 30 mins at 37°C in a thermomixer at 900 RPM. Mechanical dissociation with a P1000 pipette tip (20 s), then fire-polished pasteur pipette was performed for 5 min. Tissue was digested for 10 more min at 37°C, and dissociated with the pasteur pipette for another 2 min. Cells were centrifuged for 5 min, 300 RCF at room temp, and the pellet was resuspended with DNAase/ovomucoid inhibitor according to manufacturer's protocol (Worthington). Cells were incubated in Red blood cell lysis buffer (420301, Biolegend) 3–4 min at 4°C. For MULTI-seq barcoding, cells were suspended with Anchor:Barcode solution (every sample was labeled with a unique barcode: sample 1 Barcode: TGAGACCT (‘A3’); sample two barcode GCACACGC (‘A4’); sample three barcode AGAGAGAG (‘A5’); and sample four barcode TCACAGCA (‘A6’)) for 5 min at 4°C. A Co-Anchor solution was added and incubated for 5 min (McGinnis et al., 2019a). Samples were combined and filtered with a FlowMi 40 µm filter (BAH136800040-50EA, Sigma). To remove myelin, the cell suspension was incubated with Myelin Removal Beads (130-096-733, Miltenyi Biotec) (6 μl/brain) for 15 min at 2–8°C. Cells were washed with 0.5% BSA-PBS and transferred to MACS columns (30-042-401 and QuadroMACS Separator 130-090-976, Miltenyi Biotec). The cell suspension was preincubated with TruStain FcX Plus Antibody (BioLegend, Key resources table) on ice for 10 min, then incubated with oligonucleotide-tagged anti-VCAM1 and anti-CD24 antibodies (BioLegend, Key resources table) on ice for 30 min, then washed twice with 0.5% BSA-PBS by centrifugation (5 min, 4°C, 350 RCF) and filtered with a FlowMi 40 µm filter. The effluent was collected and cell density was counted. Cells were loaded into two wells of a 10x Genomics Chromium Single Cell Controller. We used the 10x Genomics Chromium Single Cell 3’ Library and Gel Bead Kit v3 to generate cDNA libraries for sequencing according to manufacturer’s protocols. GFP expression of isolated cells was observed under an epifluorescence microscope.
MULTI-seq barcode library preparation and cell assignment
MULTI-seq and antibody TotalSeq barcode libraries were assembled as previously described (McGinnis et al., 2019a). Briefly, a MULTI-seq primer is added to the cDNA amplification mix. Afterwards, in the first clean-up step using SPRI beads (0.6x) of the standard 10x library prep workflow, the supernatant is saved, transferred to a new tube and a cleanup step using SPRI (1.6x) is performed to eliminate larger molecules. A library preparation PCR is also performed for the MULTI-seq barcodes. The barcode library is analyzed using a Bioanalyzer High Sensitivity DNA system and then sequenced. The code for demultiplexing samples and detecting doublets can be found at https://github.com/chris-mcginnis-ucsf/MULTI-seq, McGinnis, 2019b.
Whole cell sequencing data alignment and processing
We pooled gene expression and barcode cDNA libraries from each 10x Genomics Single Cell Controller well (technical replicates, ‘Lane’) and sequenced them at the UCSF Center for Advanced Technology on one lane of an Illumina Novaseq 6000 machine. A total of 2,892,555,503 reads were aligned using CellRanger 3.0.2-v3.2.0 (10x Genomics) to a custom version of the mouse reference genome GRCm38 that included the GFP gene (GFP sequence: Supplementary file 7). Reads corresponding to oligonucleotide-tagged TotalSeq antibodies were assigned to cells in CellRanger according to manufacturer instructions.
To identify cell barcodes that most likely corresponded to viable cells, we performed quality control and filtering steps. We excluded cells outside of the following thresholds: UMI count depth: 5th and 95th percentiles; number of genes per cell: below 5th percentile; percentage of mitochondrial gene reads per cell: greater than 10%. We classified cells into sample groups and identified doublets using MULTI-seq barcode abundances (McGinnis et al., 2019a). We used Seurat Integration (Seurat 3) canonical correlation analysis (CCA) to reduce data dimensionality and align the data from technical replicates (Lane 1 and Lane 2) (Stuart et al., 2019).
Single nucleus sample preparation
Brains were extracted and 0.5 mm slices were obtained. We microdissected the anterior ventral, anterior-dorsal, posterior-ventral and dorsal V-SVZ regions of 17 P35 CD1 male (8) and female (9) mice. Briefly, we used a brain matrix to cut one millimeter thick coronal slabs of the mouse forebrain and used histological landmarks to identify each sampling area (e.g. anterior region landmarks: septum; posterior regions: hippocampus). Regions were dissected under a microscope to reduce the amount of underlying striatum in each sample. Each micro-dissected V-SVZ region was processed in parallel as a distinct sample. We processed tissue samples for nucleus isolation and sNucRNA-Seq as previously described (Velmeshev et al., 2019). Briefly, we generated a single nucleus suspension using a tissue douncer (Thomas Scientific, Cat # 3431D76) in nucleus isolation medium (0.32M sucrose, 5 mM CaCl2, 3 mM MgAc2, 0.1 mM EDTA, 10 mM Tris-HCl, 1 mM DTT, 0.1% Triton X-100 in DEPC-treated water). Debris was removed via ultracentrifugation on a sucrose cushion (1.8M sucrose, 3 mM MgAc2, 1 mM DTT, 10 mM Tris-HCl in DEPC-treated water) in a thick-walled ultracentrifuge tube (Beckman Coulter, Cat # 355631) and spun at 107,000 RCF, 4°C for 150 min. The pelleted nuclei were incubated in 250 µL PBS made with DEPC-treated water on ice for 20 min. The resuspended pellet was filtered twice through a 30 µm cell strainer. We counted nuclei with a hemocytometer to determine nucleus density, and loaded approximately 12,000 nuclei from each sample into its own well/lane of a 10x Genomics Chromium Single Cell Controller microfluidics instrument. We used the 10x Genomics Chromium Single Cell 3’ Library and Gel Bead Kit v2 to generate cDNA libraries for sequencing according to manufacturers’ protocols. We measured cDNA library fragment size and concentration with a Bioanalyzer (Agilent Genomics).
Single nucleus sequencing data alignment and processing
We pooled the gene expression cDNA libraries from each single nucleus sample and sequenced them on one lane of an Illumina HiSeq 4000 at the UCSF Center for Advanced Technology. The PV sample was further sequenced to increase sequencing depth. A total of 1,340,031,643 reads were aligned using CellRanger 2.1.0–2.3.0 (AV, AD, PD samples); 3.0.2-v3.2.0 (PV sample) (10x Genomics) to a custom mouse reference genome that includes unspliced ‘pre-mRNA’ (GRCm38), which we expect to be present in cell nuclei (Velmeshev et al., 2019).
To identify cell barcodes that most likely corresponded to viable nuclei, we performed quality control and filtering steps. For each region sample, we excluded nuclei outside of the following thresholds: UMI count depth: 5th to 95th percentiles; number of genes per cell: below 5th percentile; fraction of mitochondrial gene reads per cell (<10%). We used Seurat Integration (Seurat 3) canonical correlation analysis (CCA) to reduce data dimensionality and align the data from each region (Stuart et al., 2019).
Single cell and single nucleus sequencing data normalization and dimensionality reduction
We used Seurat 3 (Stuart et al., 2019) to analyze both the Whole Cell and Single Nucleus datasets: for each dataset, cells or nuclei from each 10x Chromium Controller Lane (scRNA-Seq: Lanes 1 and 2; sNucRNA-Seq: AD, AV, PD, PV lanes) were integrated using IntegrateData and normalized using regularized negative binomial regression (SCTransform) (Hafemeister and Satija, 2019). We calculated 100 principal components (PCs) per dataset, and used 50 (scRNA-Seq) or 100 (sNucRNA-Seq) to calculate cell cluster identities at five distinct resolutions (0.5, 0.8, 1.0, 1.5, and 2.0) and UMAP coordinates. The cell cluster identities presented in this manuscript correspond to resolution 1.5 (scRNA-Seq metadata column integrated_snn_res.1.5) or 2 (sNucRNA-Seq metadata column integrated_snn_res.2), and were chosen based on visual correspondence with the expression of known neurogenic lineage markers. Sequenced antibody tags in the scRNA-Seq dataset were separately normalized using NormalizeData (method: CLR) and ScaleData, and are included in the Seurat object ‘Protein’ assay as ‘VCAM1-TotalA’ and ‘CD24-TotalA’.
Dual fluorescent in situ hybridization-immunofluoresce
Mouse brains (n=3, P30) were serially sectioned using a Leica cryostat (10-µm-thick sections in Superfrost Plus slides). Sections were incubated 10 min with 4% PFA and washed 3x10 min with phosphate-buffered saline (PBS) to remove OCT. Slides were incubated with ACD hydrogen peroxide for 10 min, treated in 1x target retrieval buffer (ACD) for 5 min (at 96–100°C) and rinsed in water and 100% ethanol. Samples were air dried at 60°C during 15 min and kept at room temperature overnight. The day after, samples were treated with Protease Plus for 30 min at 40°C in the RNAscope oven. Hybridization of probes and amplification solutions was performed according to the manufacturer’s instructions. Amplification and detection steps were performed using the RNAscope 2.5 HD Red Detection Kit (ACD, 320497) and RNAscope 2.5 HD Duplex Reagent Kit (ACD, 322430). RNAscope probes used: Mm-Lphn3 (also named Adgrl3) (cat.# 317481), Mm-Rlbp1 (cat.# 468161), Mm-Crym (cat.# 466131), Mm-Pax6 (cat.# 412821), Mm-Slit2 (cat.# 449691), Mm-Cnatnp2 (cat.# 449381), Mm-Urah-C2 (cat.# 525331-C2), Mm-Dio2 (cat.# 479331), Mm-Hopx (cat.# 405161), Mm-Ntng1-C2 (cat.# 488871-C2), Mm-Trhde (cat.# 450781). Mm-Snhg15 was custom made (NPR-0009896, cat.# 889191). DapB mRNA probe (cat.# 310043) was used as negative and Mm-PPIB (cat.# 313911) as positive control. RNAscope assay was directly followed by antibody staining for chicken anti-GFAP (Abcam, ab4674,1:500) and rabbit anti-DCX (Cell signaling, 4604S, 1:200) or rabbit anti-S100 (Dako, Z033, 1:100,
Immunohistochemistry
Coronal sections (n=four mice, P30) and whole mounts (n=three mice, P28) were incubated with Immunosaver (1:200; EMS, Fort Washington, PA) for 20 min at 60°C, and then 15 min at RT. Tissue was then incubated in blocking solution (10% donkey serum and 0.2% Triton X-100 in 0.1 M PBS) for 1 hr. followed by overnight incubation at 4°C with the primary antibodies: mouse anti-CRYM (Santa Cruz, sc-376687,1:100), rabbit anti-BETA-CATENIN (Sigma, C2206, 1:250), Chicken anti-GFP (Aves labs, GFP1020, 1:400), Rabbit anti-HOPX (Proteintech,11419–1-AP,1:500). On the next day, sections were rinsed and incubated with Alexa Fluor secondary antibodies. Samples were mounted with Aqua-poly/mount (Polysciences Inc, 18606–20).
Confocal microscopy
Confocal images were acquired using the Leica Sp8 confocal microscope. Samples processed for RNAscope and immunohistochemistry were imaged at ×20 (low magnification) and ×63 (High magnification). For high-magnification RNAscope images, 10–15 optical sections were acquired sequentially using Leica Application Suite X (LAS X) software.
Quantifications
For quantifications of RNAscope puncta, the V-SVZ of one coronal section hemisphere was tile-scanned per mouse (n=3). Optical sections were taken through the entire thickness of the section at 0.3-micometer intervals using a 63x objective. The tile-scans were imported into Imaris Image Analysis software (v9.7, Bitplane) to detect RNAscope puncta using the Spots tool. Automatic spot detection was manually adjusted, and spots were categorized according to colocalization with DAPI and S100-Beta or DCX immunolabeling. Spots outside of the lateral wall V-SVZ were manually removed from the dataset, as were spots in z-planes that lacked antibody labeling (e.g. the antibody did not fully penetrate the section). Finally, a line was drawn using the Measurement tool from the ventral-most point in the lateral wall V-SVZ, through the V-SVZ to the dorso-lateral most extent of the wedge (raw data available at Figure 3 and Figure 6—source data 1). These data were exported from Imaris and the density of spots along the length of the V-SVZ was quantified using the R sp package (v.1.4–5).
To determine the proportion of B cells (identified as GFP+ cells) expressing CRYM in coronal sections, tile-scans of the entire V-SVZ (n=4) were acquired. Coronal sections (1.42–0.98 mm anterior to bregma) were utilized. Cells were manually counted using Imaris Image Analysis software (v9.7, Bitplane). The length of the V-SVZ, including the wedge and the lateral wall of the lateral ventricle, was divided in 3. The dorsal domain was defined as the most dorsal third (V-SVZ wedge and the adjacent lateral wall); the ventral domain encompassed the ventral two thirds (Figure 3—figure supplement 1S).
For quantifications of B1 cells expressing CRYM, V-SVZ whole mounts from hGFAP:GFP mice (n=3) were immunostained for GFP, ß-Catenin and CRYM. Images of the apical surface of dorsal and ventral regions (8–10 fields/region) of the lateral wall were acquired by confocal microscopy (Leica SP8). B1 cells were identified by their GFP+ small apical endings demarcated by ß-Catenin. B1 cells expressing CRYM were manually counted using Imaris Image Analysis software. Note that within the wedge region, in the most dorsal domain, there is no ventricle and this region is not visible in whole mount preparation; this region was included in the above quantification in sections, but it is not included in the quantifications from the whole mounts. B1 cell whole mount quantifications are expressed as mean ± SD (standard deviation). Student’s test (Excel, Microsoft) was used for pairwise comparison between two groups.
Transmission electron microscopy
For CRYM pre-embedding immunogold staining, mice (n=2) were perfused with 4% paraformaldehyde (PFA)/ 0.5% glutaraldehyde. Brains were cut into 50 μm coronal sections on a vibratome. Floating sections were incubated with 1% Sodium borohydride in phosphate buffer (0.1M) for aldehyde inactivation, cryoprotected with 25% sucrose and permeabilized by freezing and thawing (5x) in methylbutane on dry ice. Sections were blocked with 0.3% BSAc (Aurion) in PB 0.1M for 1 hr at RT and incubated with mouse anti-CRYM (Santa Cruz, sc-376687, 1:100) 72 hr, 4°C. Sections were rinsed and incubated with goat anti-mouse conjugated to colloidal gold (1:50, UltraSmall, Aurion #25120) for 24 hr at 4°C. Silver enhancement and Gold toning were performed as previously described (Sirerol-Piquer et al., 2012). Sections were postfixed with 1% osmium-7% glucose in phosphate buffer 0.1M, dehydrated and embedded in Durcupan (Fluka). Ultrathin sections (70 nm) were cut, stained with lead citrate and examined under TEM (Tecnai Spirit G2, FEI).
Combined gene expression score
We used AUCell to score cells based on the expression of sets of genes (Aibar et al., 2017). For creating the dorsal and ventral scores, we used the dorsal and ventral signature genes (Figure 6 and Figure 6—figure supplement 1). To identify the High Score dorsal and ventral cells, we normalized both scores to values between 0 (min) and 1(max). We subtracted the ventral score from the dorsal score and the cells in the top quartile corresponded to High Score dosal cells and on the bottom quartile corresponded to the High Score ventral cells.
Differential gene expression, GO analysis, and gene regulatory networks
We used Seurat three functions FindMarkers (two groups) or FindAllMarkers (more than two groups) to identify differentially expressed genes among groups of single cells (p_val_adj < 0.05) using a Wilcoxon rank sum test. For single nuclei, Seurat four function FindMarkers was used to identify differentially expressed genes (p_val < 0.05). For detailed parameters see available code (below). Selection of genes from the resulting lists for further analysis are described in the text. Gene Ontology analyses of differentially expressed genes were performed using a binomial test and comparing differentially expressed genes against the whole mouse genome using Panther v.16 (Mi et al., 2021). GO terms with a false discovery rate lower than 5% were considered statistically significant. To construct Gene Regulatory Networks for dorsal and ventral B cells we used gene network inference with ensemble of trees (GENIE3), a tree regression based method that employs Random Forests to rank regulatory links between genes (Huynh-Thu et al., 2010). After calculating link lists from expression matrices in putative dorsal or ventral B cells, we used the top 10 differentially enriched genes in each population to build networks. We selected the top 300 links among genes predicted to be regulators or targets of these markers. To identify regulatory connections of
Nuclei to whole cell regional label transfer
B cell label transfer: First, we subsetted quiescent B cells from both the scRNA-Seq and sNucRNA-Seq datasets (clusters B(5), B(14), B(22), and sNucRNA-Seq cluster 7). We generated the reference sNucRNA-Seq B cell dataset that consisted of equal numbers of B cells per region, randomly selected from the middle 50% of cells by number of genes identified per cell (25th-75th percentile of SCT_snn_nFeature). This prevented the region with the most nuclei from dominating the prediction scores, and filtering cells by nFeature prior to downsampling resulted in reproducible prediction scoring, likely due to exclusion of low-quality B cells and doublets not rejected in the full dataset quality control steps. Anterior dorsal and posterior dorsal regions were combined to create the Dorsal reference cell set, and the anterior ventral and posterior ventral were combined to create the Ventral reference cell set. Subsetted scRNA-Seq B cells and filtered sNucRNA-Seq B cell sets were individually normalized using SCTransform. We then ran FindTransferAnchors with the following settings:
A cell label transfer: The same method as above was applied to scRNA-Seq A cell clusters A(0), A(1), A(4), A(6), and A(15), and sNucRNA-Seq clusters 12 and 29.
RNA velocity
RNA Velocity in the neurogenic lineage was calculated using scvelo (Bergen et al., 2020), using 2000 genes per cell. Moments were calculated using 30 PCs and 30 neighbors. Velocity was estimated using the stochastic model. Pseudotime was plotted using the original UMAP coordinates.
Data and code availability
The RNA sequencing datasets generated for this manuscript are deposited in the following locations: scRNA-Seq and sNucRNA-Seq GEO Data Series: GSE165555.
Processed data (CellRanger output. mtx and. tsv files, and Seurat Object. rds files) are available as supplementary files within the scRNA-Seq (GSE165554) or sNucRNA-Seq (GSE165551) data series or individual sample entries listed within each data series.
Web-based, interactive versions of the scRNA-Seq and sNucRNA-Seq datasets are available from the University of California Santa Cruz Cell Browser: https://svzneurogeniclineage.cells.ucsc.edu.
The code used to analyze the datasets and generate the figures are available at the following location: https://github.com/AlvarezBuyllaLab/SVZSingleCell (copy archived at swh:1:rev:37402d867adce3ff4295f06e8fd6289e1d3ba075), Cebrian-Silla et al., 2021.
© 2021, Cebrian-Silla et al. This work is published under https://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.