During mammalian brain development, how different astrocytes are specified from progenitor cells is not well understood. In particular, whether astrocyte progenitor cells (APCs) start as a relatively homogenous population or whether there is early heterogeneity remains unclear. Here, we have dissected subpopulations of embryonic mouse forebrain progenitors using single-cell transcriptome analyses. Our sequencing data revealed two molecularly distinct APC subgroups at the start of gliogenesis from both dorsal and ventral forebrains. The two APC subgroups were marked, respectively, by specific expression of Sparc and Sparcl1, which are known to function in mature astrocytes with opposing activities for regulating synapse formation. Expression analyses showed that SPARC and SPARCL1 mark APC subgroups that display distinct temporal and spatial patterns, correlating with major waves of astrogliogenesis during development. Our results uncover an early molecular divergence of APCs in the mammalian brain and provide a useful transcriptome resource for the study of glial cell specification.
Astrocytes are one of the most abundant cell types in the mammalian brain and play crucial functions in supporting and protecting brain organization as well as regulating neuronal circuits formation (Clarke and Barres, 2013; Sloan and Barres, 2014; Ben Haim and Rowitch, 2017). These multiple functions are conceivably executed by different types of astrocytes that are initially specified from progenitor cells during embryonic and early postnatal brain development; however, much of the detail of such developmental programs remains elusive. During development of the mammalian brains, neurons and glial cells are generated by neural progenitor cells (NPCs) in an orderly sequence with neurogenesis preceding gliogenesis. The majority of the main subtypes of cortical projection neurons are sequentially generated during neurogenesis by NPCs, which become progressively restricted in their differentiation potentials from producing early-born neurons to late-born neurons (Gotz and Huttner, 2005; Leone et al., 2008; Kriegstein and Alvarez-Buylla, 2009; Greig et al., 2013). As neurogenesis tapers down, generation of astrocytes becomes increasingly more prominent. However, when compared with our understanding of neuronal specification during neurogenesis, the developmental regulation of astrogliogenesis and subtype specification are only beginning to be addressed (Bayraktar et al., 2014; Tabata, 2015; Akdemir et al., 2020).
Recently, diverse subtypes of mature astrocytes associated with distinct neocortical layers or brain regions were identified in the adult mouse brains (John Lin et al., 2017; Lanjakornsiripan et al., 2018; Batiuk et al., 2020; Bayraktar et al., 2020). It is conceivable that specification of these subtypes of astrocytes in the brain may use a similar progressive restriction mechanism of progenitors, involve early lineage-restricted progenitors or rely on local environmental cues, or a combination of these factors (Rowitch and Kriegstein, 2010; Molofsky and Deneen, 2015; Khakh and Deneen, 2019). Previous studies of astrocyte development in the mouse spinal cord identified two morphologically distinct types of early astrocyte precursors that were subjected to temporal and spatial control (Tien et al., 2012), suggesting that astrocytes could be produced from fate-specified progenitors present during early gliogenesis. In the mouse brains, glial progenitor cell diversity was also observed in juvenile and adult mice (Marques et al., 2016; Weng et al., 2019); however, whether astrocyte progenitor cells (APCs) start as a relatively homogeneous pool or whether different subgroups of progenitors can co-initiate astrocyte production is unclear. If heterogeneity of APCs emerges early in the brain, whether such subtypes of progenitors are present as an intermixed population or whether they can display temporally and/or spatially defined patterns also remain to be determined. Answering these questions requires a detailed analysis of neural progenitor cell heterogeneity at the start of gliogenesis.
Single-cell RNA sequencing (scRNA-seq) technology provides a powerful tool for dissecting the heterogeneity of cell populations. Multiple studies have performed single cell analyses with either mouse or human cortical cells, and the results have documented heterogeneity within the NPC or neuron populations (Johnson et al., 2015; Pollen et al., 2015; Shin et al., 2015; Zeisel et al., 2015, 2018; Okamoto et al., 2016; Yuzwa et al., 2017; Mi et al., 2018; Zhong et al., 2018; Telley et al., 2019; Loo et al., 2019; Polioudakis et al., 2019; Li et al., 2020). Because many of these studies sequenced brain tissues of a mixed cell population, the low percentage of NPCs in total brain cells presented a challenge for a detailed characterization of heterogeneity and molecular markers specifically associated with the NPC population if not enough total brain cells were analyzed. To dissect heterogeneity of NPCs more comprehensively during brain development, we employed an established approach for purification of NPCs from the Nes-GFP/Dcx-mRFP double transgenic reporter mice. Bulk sequencing using the system has verified that the transcriptomes of NPCs are faithfully preserved after purification (Wang et al., 2011; Hahn et al., 2013; Liu et al., 2017, 2018). In addition, we profiled single-cell transcriptomes of purified NPCs from two developmental stages at the neurogenesis-to-gliogenesis transition to better reveal the emerging glial progenitor cell population. Here, we report that the scRNA-seq data revealed multiple clusters of NPCs, among which two molecularly distinct subgroups of APCs were identified. The presence of these APC subgroups and their respective developmental expression patterns were further characterized by analyzing specific markers during early gliogenesis.
RESULTS AND DISCUSSION
Identification of two distinct astrocyte progenitor subgroups from cortical NPCs by scRNA-seq
Purification of NPCs from the Nes-GFP/Dcx-mRFP double transgenic reporter mice was combined with the 10X Chromium platform to obtain single-cell transcriptomes (Fig. 1A). NPCs (GFP+RFP− cells, Fig. S1) were purified from the cortices at embryonic day 15.5 (E15.5; peak neurogenesis stage) and E18.5 (end neurogenesis/early gliogenesis stage). After quality control and filtering, 878 cells at E15.5 and 1150 cells at E18.5 were used for subsequent analyses. The 1325 most variable genes were selected, and the top 11 principal components (PCs) were used in the plot analysis. The results could clearly delineate ten cell clusters (cluster 0 to cluster 9 inclusive), within which the E18.5 cortex contained more cell clusters than the E15.5 cortex (Fig. 1B-E).
Heatmap results showed that the ten clusters were marked by differential expression patterns of representative genes (Fig. 1D). Cluster 1 was the group of intermediate progenitor cells (IPCs), marked by the expression of Eomes/Tbr2. Clusters 2 and 4 were cycling progenitors enriched with cell-cycle related genes such as Ccnb1, Ccnb2 and Cdc20. Interestingly, the data revealed three clusters of glial cell lineages prominently enriched in the E18.5 cortices. These included two astrocyte progenitor cell (APC) subgroups (clusters 5 and 7) expressing Aldh1l1, Slc1a3 and Fabp7 in both groups (Fig. 1D,F), and an oligodendrocyte progenitor cell (OPC) subgroup (cluster 3) expressing most known OPC markers, such as Olig1, Olig2, Pdgfra and Sox8 (Fig. 1D). Between the two APC subgroups (Fig. 1F,G), Sparc and Sparcl1 transcripts were specifically enriched in cluster 5 and cluster 7 cells, respectively (Fig. 1D). Cluster 5 showed relatively higher expressions of metallothionein (MT) family members Mt1, Mt2 and Mt3, while cluster 7 had higher expression of Olig1, Olig2 and Egfr (Fig. 1D,F, Fig. S2). Cluster 6 is a population expressing interneuron lineage-specific markers, such as Lhx6 and Arx, etc. (Fig. 1D,E), representing the differentiating interneuron progenitors that migrated tangentially into the cortex from the ventral forebrain (Anderson et al., 1997; Wu et al., 2011). The sequencing data also included a small group of projection neurons (cluster 8) and endothelial lineage cells (cluster 9) (Fig. 1E). The former likely resulted from the stringency of gating during FACS purification of NPCs; the latter might be co-isolated due to expression of Nes-GFP in proliferating endothelial cells (Suzuki et al., 2010). Unlike other clusters with distinct markers, cells of cluster 0 express relatively high Eomes, Tubb3 and Fabp7, but not other differentiated markers or cell cycle-related genes, representing a pro-neuronal and quiescent progenitor cell group, which are likely related to the embryonic cells of origin of the adult subventricular zone (SVZ) neural stem cells (Fuentealba et al., 2015).
Two astrocyte progenitor subgroups identified from the ventral forebrain
Previous studies (Hochstim et al., 2008; Tsai et al., 2012) have shown that astrocytes are generated and allocated to a region associated with their original birth place. To examine whether the two APC subgroups identified in the cortex are present in other brain regions, we purified and sequenced 3602 NPCs (1119 at E15.5 and 2483 at E18.5) from the ganglionic eminence regions of the ventral forebrain (Fig. S3A,B). Interestingly, two similar APC subgroups (cluster 5 and 7) could also be detected in the ganglionic eminence (Fig. S3C,D). The two subgroups of APCs were similarly marked by Sparc and Sparcl1, respectively (Fig. S3C,E). In addition, they were also differentially enriched with some other specific genes: Clu, Mt3, Aldoc and Apoe were associated with Sparc-specific lineage, while Slc1a3, Id3, Ptn and Hes5 were enriched in Sparcl1-specific lineage (Fig. S3E). Interestingly, further cell clustering analysis between the cortex and ganglionic eminence showed that the astrocyte lineage progenitors expressing Sparc or Sparcl1 were clustered together, irrespective of their forebrain regions (Fig. S3F). These results collectively indicated that distinct subgroups of APCs were present in both the dorsal and ventral forebrains, and that this molecular divergence started from the early stage of gliogenesis.
Differentially expressed genes between the two APC subgroups in the cortex
To identify potential molecules that can specifically mark the two APC subgroups, we analyzed differential gene expression patterns between the APC subgroups from the cortex. As expected, Sparc and Sparcl1 were shown to specifically identify with APC1 and APC2 subgroups, respectively (Fig. S4A,B). The APC1 subgroup also saw enrichment of Clu (cluster 5, Fig. 1F) and Ptx3 (Fig. S4A,C). The APC2 subgroup was associated with higher expression of the metabotropic glutamate receptor (mGluR) members Grm3 and Grm5 (Fig. S4B). In addition, higher expression of Id (inhibitor of DNA binding) family members Id1, Id2 and Id3 (Fig. S4B,C) was also detected in APC2 subgroup.
SPARC or SPARCL1 expression patterns during brain development
To further examine the two subgroups of APCs, we performed immunostaining of SPARC and SPARCL1 proteins in the mouse brain. Antibody staining showed that SPARC protein was expressed lining the lateral ventricle as early as E16.5 (Fig. 2A). SPARC antibody detected radial processes apparently associated with the ventricular APCs. This was initially seen in the ganglionic eminence from E16.5 (Fig. S5A), but later also became evident in the cortex (Fig. 2B). At later neurogenesis stages, some SPARC+ cells were detected outside the ventricular zone (VZ) in the cortex (Fig. S5B). These cells were presumably the differentiating astrocytes generated by the SPARC+ APCs of the VZ. Between Sparc-expressing (cluster 5) and Sparcl1-expressing (cluster 7) subgroups, the Sparc-expressing subgroup showed higher levels of expression of Pax6, suggesting that SPARC+ APCs might be derived from radial glial cells. On the other hand, both cluster 5 and 7 subgroups showed similar expression levels of other progenitor cell markers: Sox2 and Nes (Fig. 2C, Fig. S5C). Sox2 was also prominent in interneuron-lineage cells (Fig. S5C, cluster 6 versus cluster 8), consistent with previously reported results based on bulk RNA-sequencing (Liu et al., 2016). The Sparc-marked APC1 subgroup also showed enriched expression of Clu (Fig. 1F; Fig. S4A), and we found that CLU protein was detected lining the ventricle at E18.5 (Fig. S5D), similar to the SPARC staining pattern.
Unlike SPARC, SPARCL1 was mainly detected in cells around two discrete locations: the pallium-subpallium boundary (PSB) in the dorsal forebrain; and the septum-subpallium junction in the ventral forebrain (Fig. 2D; E16.5). In the dorsal brains, SPARCL1+ APCs emerged as early as E16.5, with the staining signals originating at the PSB site and gradually spreading medially to the cortical ventricular zone (VZ) at later stages (e.g. P1) (Fig. 2D; P1), concomitant with SPARCL1+ astrocytes (over 96% SPARCL1+ cells were SOX9+) accumulating in the cortex (Fig. S6). In the ventral brains, SPARCL1+ astrocytes were mostly absent in the ganglionic eminence and septum (areas surrounding the lateral ventricle) but were distributed outside (lateral to) the ganglionic eminence, apparently along the ventral extension of the pallium and in the area ventral to the septum-subpallium junction (Fig. 2D; E18.5 and P1).
Temporal patterns of Sparc- and Sparcl1-expressing APC subgroups
Because available antibodies were from the same species, we further examined temporal and spatial expression patterns of Sparc and Sparcl1 by co-staining using RNAscope. Like the protein patterns, Sparc was expressed along the ventricular zone, whereas Sparcl1 was detected in dorsolateral SVZ region at E16.5 (Fig. 3A). At E18.5, Sparcl1 expression gradually extended into the dorsal cortex (Fig. 3B). Cells expressing Sparcl1 or Sparc were distributed in an intermixed pattern (only blood vessel cells co-expressed Sparcl1 and Sparc) (Fig. 3C). To assess their temporal origins in the cortex, we performed developmental trajectory analyses and cluster-specific genes pseudotime analyses. The results showed that NPCs differentiated mainly into neuronal lineage at E15.5 but more prominently into glial lineage at E18.5 (Fig. 3D,E). Consistent with the expression patterns of protein and RNA, SPARC-marked APC1 showed up earlier than SPARCL1-marked APC2 in developmental trajectory (Fig. 3F), and APC1-related genes were expressed earlier in pseudotime (Fig. S7).
Distinct APC subgroups marked by SPARC and SPARCL1 at early gliogenesis
To validate that SPARC- or SPARCL1-marked cells are astrocyte progenitors, we analyzed the expression of these two proteins using the Aldh1L1-EGFP reporter mouse line. At early gliogenesis stage E16.5, EGFP signals in the Aldh1L1-EGFP reporter brain were seen lining the entire ventricular zone (VZ), as well as at the SVZ along the PSB. SPARC antibody staining showed that expression of SPARC protein and reporter EGFP were positively correlated at the VZ (Fig. 4A-D). In the ganglionic eminence region, SPARC staining revealed processes extending away from the VZ; however, no long processes could be detected in the cortex at this stage, indicating that SPARC+ astrocyte progenitors emerged earlier in the ganglionic eminence than in the cortex. Although SPARC signal was absent at the region along the PSB, SPARCL1 signal could be prominently detected in this location marked by EGFP reporter (Fig. 4E,F). These results further confirmed that SPARC and SPARCL1 mark two distinct subgroups of astrocyte progenitors at early gliogenesis.
Local proliferation of SPARCL1-marked maturing astrocytes in the ventral forebrain
As the emergence of SPARCL1+ cells in the ventral brain appeared to be earlier than in the dorsal brain, we further examined this lineage in the ventral region of the Aldh1L1-EGFP brain sample. Our results showed that SPARCL1 was expressed by almost all EGFP+ astrocyte-lineage cells (Fig. S8A). Co-staining of Ki67 in the E17.5 samples showed that few SPARCL1+ cells at their apparent origin site (the septum-subpallium junction) were positive for Ki67 (Fig. S8B); however, about 40% SPARCL1+ cells in the area beneath the septum-subpallium junction were positive for Ki67 (Fig. S8C). These SPARCL1+Ki67+ cells were likely astrocytes originating from the septum-subpallium junction and migrating to the ventral regions to undergo local proliferation (Fig. S8D), similar to differentiated astrocytes observed in the early postnatal cortex (Ge et al., 2012).
Accumulating evidence indicated that astrocytes in the mouse brain originate from two main sources in multiple waves during early development (Akdemir et al., 2020): one source involves radial glia cells in the late embryonic VZ, which generate astrocytes first through intermediate glial progenitors and then through transforming radial glia (Noctor et al., 2004); another source involves glial progenitors in the perinatal dorsolateral SVZ, which produce astrocytes that subsequently distribute into the cortex and can continue to proliferate at their destination (Ge et al., 2012). Whereas previous studies investigated glial cell heterogeneity in juvenile or adult mouse brain (Marques et al., 2016; John Lin et al., 2017; Lanjakornsiripan et al., 2018; Weng et al., 2019; Batiuk et al., 2020; Bayraktar et al., 2020), our study focused on examination of early heterogeneity of astrocyte lineage cells during embryonic brain development. Our single-cell transcriptome analysis of progenitor cells isolated from the cortex and ganglionic eminence identified two subgroups of APCs, expressing Sparc and Sparcl1, respectively. Immunohistochemistry results further showed that SPARC+ cells were in the germinal VZ and expressed radial glia progenitor marker Pax6. In addition, SPARC staining revealed processes extending from the VZ to cortical pial surface at later stages examined (E18.5 and P1). These results collectively suggested that SPARC+ APCs were most likely radial glia cells that would start the early waves of astrogliogenesis through either producing intermediate astrocyte progenitors or undergoing transformation from neurogenic to gliogenic. The more-prominent presence of Sparc-expressing cell cluster (Fig. 1, cluster 5) at the early stage (E15.5) was also consistent with the above idea. On the other hand, SPARCL1+ APCs were initially restricted around the PSB region that would become the future dorsolateral SVZ, as well as around the septum-subpallium junction, indicating that SPARCL1+ APCs were likely the source of the later wave of astrogliogenesis that produced astrocytes that underwent symmetric divisions locally (Ge et al., 2012). In addition, our data showed that the SPARCL1+ progenitor subgroup also had prominent expression of Egfr and Olig2 (Fig. S2), both of which were expressed by OPCs, suggesting that these cells were likely bipotent for becoming astrocytes and oligodendrocytes. Interestingly, a recent study identified a cluster of SPARCL1+EGFR+OLIG2+ cells as the main group of astrocyte progenitors in the developing human cortex, and these cells were found to mainly distribute in the outer SVZ (Fu et al., 2021) – an expanded proliferative region in the human brain. The fact that the SPARCL1+EGFR+OLIG2+ cells have been shown to be the major subgroup of APCs in both the mouse and human brains indicates that there is considerable conservation of astrogliogenesis between the two species.
Our results revealed distinct temporal and spatial patterns of the two subgroups of APCs (Fig. 4G). At the temporal level, two aspects are worth noting. First, SPARC+ cells appeared to develop earlier (around E15.5) than SPARCL1+ cells (Figs 1E and 3F). Second, the locations of SPARC+ cells were well demarcated from very early stage (E16.5) through perinatal stages examined; however, the locations of SPARCL1+ cells displayed a dynamic change during development. No SPARCL1+ cells could be detected in the main cortical radial regions at early stages; however, between E18.5 and P1, some SPARCL1+ APCs began to appear in the lateral corner of the cortical VZ, and increasingly more SPARCL1+ cells (presumably differentiated progeny astrocytes) became visible in the cortex spanning from the VZ to the pial surface at P1 (Fig. 2D). It is conceivable that SPARCL1-expressing APCs in the PSB area expanded tangentially and medially into the cortical VZ, and produced astrocytes that subsequently populated the cortex; or, alternatively, some of the cortical VZ radial glia cells became SPARCL1-expressing APCs and produced SPARCL1+ astrocytes that were then dispersed into the cortex. At the spatial level, the two subgroups of APCs were present in separate domains in the ventral brain, with SPARC+ cells residing in the ganglionic eminence region, while SPARCL1+ cells were absent in this region but present in the neighboring areas, centering on the places of origin for these cells (the PSB or septum-subpallium junction). On the other hand, in the dorsal brain region, SPARC+ cells were initially the main APCs detected at an earlier stage (e.g. E16.5). As development progressed, SPARCL1+ APCs emerged in the cortical VZ in an apparent lateral-to-medial gradient from the PSB (Fig. 2D, Fig. S9). Concomitantly, SPARCL1+ astrocytes accumulated and became intermixed with SPARC+ astrocytes in the cortex. Notably, the temporal and spatial patterns of SPARC+ and SPARCL1+ APCs, and their respective progeny astrocytes identified in this study (Fig. 4G), were overall reminiscent of the regional astrocyte allocation revealed by previous fate-mapping studies (Tsai et al., 2012). The biological implication of the distribution difference in the ventral versus dorsal brain between SPARC+ and SPARCL1+ APCs is not clear but might be related to regional functions of these APCs and their progeny astrocytes. In addition, although our data are consistent with SPARC+ and SPARCL1+ cells being APCs, it is possible that these cells may produce other types of glial cell, such as ependymal cells. Future cell fate mapping using Sparc- and Sparcl1-specific Cre lines will be necessary to define the lineage specifications.
In mature astrocytes, SPARC and SPARCL1 have been found to act as a negative and a positive regulator of synapse formation, respectively (Kucukdereli et al., 2011). This indicates that astrocytes might control the development and plasticity of synapses through regulation of relative levels of SPARC and SPARCL1. Accordingly, it was intriguing that SPARC- and SPARCL1-marked APCs were generated in a temporal sequence during development. The early emergence (by E16.5) of SPARC+ APCs in the VZ (Figs 2A and 4B-D) preceded the establishment of cortical neuronal layers, whereas the later rise and spread of SPARCL1+ APCs from the PSB (Figs 2D and 4E,F; Fig. S9) coincided with the completion of outward migration by most cortical neurons. We speculate that early-generated SPARC+ astrocytes in the cortex might be needed to help prevent premature synapse formation when cortical neurons are still being generated or are in the process of outward migration. In contrast, late-generated SPARCL1+ astrocytes might then be needed to facilitate synapse formation when most neurons are arriving at their destination and the emerging cortical layer structure is being laid out. Further studies are required to examine whether loss of either of the two APC subgroups would compromise the formation of neuronal synapses and the wiring of neural circuits during development.
MATERIALS AND METHODS
Generation of transgenic mouse lines Nes-GFP and Dcx-mRFP (The Jackson Laboratory, 024905) have been reported previously (Wang et al., 2011). The transgenic reporter line Aldh1L1-EGFP/Rpl10a was from The Jackson Laboratory (030247) and was maintained as homozygotes without any visible defects. For immunohistochemistry staining experiments, Swiss Webster mice were used if not mentioned specifically. All animals were group housed and maintained in the temperature range and environmental conditions recommended by the American Association for Accreditation of Laboratory Animal Care (AAALAC). Animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the City of Hope and were carried out in accordance with NIH guideline and the Guide for the Care and Use of Laboratory Animals.
Purification of NPCs
For purification of total NPCs, heterozygous Nes-GFP mice were bred with homozygous Dcx-mRFP mice to yield GFP/RFP double-positive embryos. Pregnant mice were sacrificed at corresponding gestation days (E15.5 and E18.5). The dorsal cortices or ventral forebrain GE regions derived from the double-positive embryos were dissociated by trituration in HBSS (Mediatech) as previously described (Liu et al., 2017). In brief, tissues were dissociated in 15 ml tubes containing 10 ml ice-cold HBSS (with Ca2+ and Mg2+) with 5 mM EDTA and 25 μg/ml DNaseI. Tissues were triturated 10 times on ice first, incubated for 5 min and triturated another five times to totally dissociate cells. Cells were centrifuged at 500 g for 10 min, then resuspended in 10 ml ice-cold HBSS (with Ca2+ and Mg2+) containing 12.5 μg/ml DNaseI. After being centrifuged again, cells were finally resuspended in DMEM/F12 medium (Corning, 15-090-CV) containing 12.5 μg/ml DNaseI and 0.5% BSA for sorting. FACS was performed with a four-laser BD FACSAria III System (BD Biosciences). Negative cells, cell debris or cell doublets were excluded with gating of side scatter, forward scatter and pulse width. BD FCASDiva software was used for acquisition and data analysis. Cells were collected into a 1.5 ml microcentrifuge tube containing 500 μl DMEM/F12 with 0.5% BSA.
10X Genomics single-cell RNA-seq
FACS-purified cells were resuspended in PBS containing 0.04% BSA at a concentration of ∼700-1200 cells/µl. Cell number and viability were measured using a TC20 Automated Cell Counter (BioRad). Single-cell RNA libraries were prepared according to the Chromium Single Cell 3′ Reagent Kits v2 User Guide (10X Genomics). Approximately 5000-10,000 cells were loaded on a Chromium Single Cell Controller Instrument (10X Genomics) to generate single-cell gel beads in emulsion (GEMs). The barcoded sequencing libraries were constructed using the Chromium Single-Cell 3′ Library Kit (10X Genomics) for enzymatic fragmentation, end-repair, A-tailing, adaptor ligation, ligation cleanup, sample index PCR and PCR cleanup. Libraries were sequenced with a Hiseq 2500 instrument (Illumina) with a depth of 50,000-100,000 reads per cell.
Analysis of 10X Genomics single-cell RNA-seq data
Raw sequencing data were processed using the 10X Genomics Cell Ranger pipeline (version 1.3.1) to generate FASTQ files and aligned to a mm10 genome to gene expression count. The subsequent data analysis was performed using ‘Seurat’ package v2.3.0 and R scripts. Cells with mitochondrial read rate >10% and <200 detectable genes were considered as low quality and filtered out. The filtered data were normalized to the total reads of 10,000 for each cell, and regressed to remove the effect of UMI and the percentage of mitochondria genes. The residuals were then mean centered to generate the final scaled data. Normalized and scaled data were clustered using the top principal components of highly variable genes with a resolution of 0.6. The t-distributed stochastic neighbor embedding (t-SNE) algorithm was used to visualize the resulting clusters. Cluster-specific markers were identified to generate heatmap and violin plots in the identified cell clusters using ‘Seurat’.
Cross-regional clustering analysis
To identify the similarity of individual clusters across cortex and GE samples, the data from the cortex and GE samples were normalized and standardized using Seurat v2.3.0 as described above. For each identified cell cluster, the average expression of each gene for all the cells in the cluster was calculated using the scaled data. Hierarchical clustering was applied to the average expression data using one correlation as a distance matrix and average linkage algorithm.
Differential expression analyses
To identify differentially expressed (DE) genes across different single-cell clusters, raw data were normalized using ‘scran’ and compared using Bioconductor package ‘MAST’. DE genes were identified using the criteria of a fold change of more than 1.3 and FDR of less than 0.05. Differential expression analysis between cluster 5 and cluster 7 was conducted using ‘findMarkers’ function implemented in Seurat v4, with a Wilcoxon Rank Sum test. P values were adjusted based on Bonferroni correction.
A single-cell trajectory was constructed by Monocle2 (v2.14.0) using the variable features identified by Seurat. A heatmap of the cluster-specific genes along the pseudotime was also generated using Monocle2.
Immunohistochemistry and image processing
The immunohistochemistry staining was caried out as described previously (Hahn et al., 2013). In brief, slides with brain sections (14 μm) were washed three times in PBS, and then incubated in room temperature for 1 h in blocking buffer containing 0.1% Triton X-100, 5% donkey serum (Jackson ImmunoResearch, 017-000-121) and 5% AffiniPure Fab Fragment Donkey anti-mouse IgG (Jackson ImmunoResearch, 715-007-003) in PBS. Sections were then incubated in primary antibody solution containing specific antibodies in PBS with 0.01% Triton X-100 and 1% donkey serum overnight at 4°C. Sections were washed three times in PBS, and donkey-sourced secondary antibodies were diluted in PBS and applied on sections for 2 h in room temperature. Sections were then counterstained by Hoechst 33342 and washed three or four times in PBS. Slides were mounted using Fluoromount-G (SouthernBiotech, 0100-01). Primary antibodies included goat anti-SPARC (R&D Systems, AF942; 1:200), goat anti-SPARCL1 (R&D Systems, AF2836; 1:1000), rat anti-PECAM1/CD31 (BD Biosciences, 550274; 1:200), goat anti-Clu (R&D Systems, AF2747; 1:200), rabbit anti-Ki67 (Abcam, ab16667; 1:200) and rabbit anti-SOX9 (Abcam, ab185966; 1:500). Secondary antibodies with minimal cross-reactivity to other species were purchased from Jackson ImmunoResearch Laboratories: Cy3-donkey anti-goat (705-165-147; 1:400), 488-donkey anti-rat (712-545-150; 1:400), 647-donkey anti-rat (712-605-150; 1:400). Each antibody staining was performed on multiple brain sections and showed consistent patterns. Images were taken using a fluorescence microscope (Zeiss Observer II) or a Zeiss LSM 880 confocal microscope. ZEN software was used for image taking, tiling and processing. Quantification results were presented from three or four brains with five or six sections counted for each brain. Graphs were generated using GraphPad Prism 8.3.0.
Sparcl1 and Sparc RNA probes were synthesized (Advanced Cell Diagnostics) in fluorescent channels C1 and C2, respectively. RNAscope experiments were performed according to the manuals of RNAscope Multiplex Fluorescent Reagent Kit v2 Assay (Advanced Cell Diagnostics, 323100) and RNA-Protein Co-Detection Ancillary Kit (Advanced Cell Diagnostics, 323180).
We thank Jeremy LaDou and Johnathan Garcia for assistance with animal breeding and care; Lucy Brown and Ni Feng for helping with cell sorting; Jinhui Wang and Chao Guo for performing single-cell sequencing; and Brian Armstrong for the use of imaging facility. Research reported in this study included work performed in the Analytical Cytometry Core and the Integrated Genomics Core supported by the National Cancer Institute under award number P30CA033572.
Conceptualization: Q.L., J.L.; Investigation: J.L.; Data curation: J.L., X.W.; Writing - original draft: J.L.; Writing - review & editing: Q.L.; Supervision: Q.L.; Project administration: Q.L.; Funding acquisition: Q.L.
This work was supported by the National Institutes of Health (NS096130 to Q.L.). Deposited in PMC for release after 12 months.
10X Genomics sequencing data have been deposited in GEO under accession number GSE128948.
Peer review history
The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199985.
The authors declare no competing or financial interests.