Single-nucleus transcriptomic atlas of the human hippocampus across different ages and after stroke injury

(A) Summary of the experimental strategy. The pair of hippocampi from postmortem human donors at different ages were collected. The anterior (AN) and middle (MI) parts containing dentate gyrus were used for snRNA-sequencing and immunostaining.

(B) 92,966 hippocampal single nuclei were visualized by UMAP plot and categorized into 16 major populations: adult astrocyte (Adult-AS, 1146 nuclei), astrocyte/quiescent neural stem cell (AS/qNSC, 11071 nuclei), primed NSC (pNSC, 2798 nuclei), active NSC (aNSC, 2140 nuclei), neuroblast (NB, 2607 nuclei), granule cell (GC, 24671 nuclei), interneuron (IN, 8601 nuclei), pyramidal neuron (PN, 676 nuclei), oligodendrocyte progenitor (OPC, 5396 nuclei), oligodendrocyte (OLG, 15796 nuclei), microglia (MG, 11823 nuclei), endothelial cell (EC, 1232 nuclei), pericyte (Per, 981 nuclei), Relin-expressing Cajal–Retzius cell (CR, 218 nuclei), and two unidentified populations (UN1 and UN2, 3810 nuclei).

(C) Dot plots of representative genes specific for the indicated cell subtypes. The size of each dot represents the cell percentage of this population positive for the marker gene. The scale of the dot color represents the average expression level of the marker gene in this population.

(D) UMAP feature plots showing expression distribution of cell type specific genes in all cell populations. Astrocyte (ALDH1L1, GFAP), neural stem cell (PAX6, VIM), neuroblast (STMN2), granule cell (SV2B), oligodendrocyte progenitor (OLIG1), microglia (CSF1R), interneuron (GAD1, RELN), relin-expressing Cajal-Retzius cell (RELN), pyramidal neuron (MAP3K15) and endothelial cell (VWF) are shown. Dots, individual cells; grey, no expression; red, relative expression (log-normalized gene expression).

(E) Quantification of each cell population in the hippocampus at three different age stages and after stroke-induced injury.

Confirmation of neurogenic lineage and dissecting of NSC molecular heterogeneity in the postnatal human hippocampus

(A) Neurogenic lineage identification was confirmed by cross-species comparison of transcriptomic signatures. Our human data were integrated with published snRNA-seq data from mice, pigs and rhesus macaque by UMAP (Hochgerner et al., 2018, Franjic et al., 2022).

(B) Expressions of previously reported RGL, nIPC, NB and immature GC markers in the corresponding populations from our human hippocampal snRNA-seq data. RGL, radial glial cell; nIPC, neural intermediate progenitor cell; NB, neuroblast; and immature GC, immature granule cell.

(C) The AS/qNSC population was subclustered into three clusters, astrocytes, qNSC1 and qNSC2. (D) Heatmap of top 10 genes (p-value < 0.05) specific for astrocytes, qNSC1 and qNSC2 after normalization.

(E and F) Using Gene set scores (average, over genes in the set, of seurat function AddModuleScore) based on previously defined gene sets 5,7,10,23-25 to characterize RGL (E) and astrocytes (F).

(G) UMAP feature plots showing expression distribution of cell type specific genes. Astrocyte markers (S100B and GFAP), radial glial like cell markers (HOPX and LPAR1) and neuron development markers (STMN1, PROX1 and SIRT2) are shown.

(H and I) Representative GO terms of the top 1000 genes specifically expressed in pNSCs (H) and aNSCs (I).

(J) Cell-cycle phases of qNSC1, qNSC2, pNSC, aNSC and NB predicted by Cell Cycle Scoring. Each dot represents an individual cell. Steel blue, red and orange dots represent G1, S and G2/M phase cells, respectively

Discovery of novel markers distinguishing various types of NSCs and NBs in the human hippocampus

(A and B) Representative top genes specific for qNSC1, qNSC2, pNSC, aNSC and NB in the neonatal neurogenic lineage identified by single-cell hierarchical Poisson factorization (scHPF) (A) and FindAllMarkers function of seurat (B).

(C) UMAP visualization of several cell type specific genes of the qNSCs, pNSC and NB predicted by scHPF and FindAllMarkers.

(D) Heatmap showing that neuroblast/immature GC highly expressed genes that are previously reported by other literature were widely expressed in human hippocampal interneurons.

(E) Scatter plot showing that several NB genes predicted by scHPF and findmarker from our snRNA-seq data were also widely expressed in human hippocampal interneurons. The genes without/with low expression in the interneurons were selected as NB specific markers (red circle scope).

(F) NB specific genes selected from our snRNA-seq data were not or very low expressed in AS/qNSCs, pNSC, aNSC, NB, GC and interneurons.

The transcriptional dynamics predicated by RNA velocity and pseudotime reconstruction revealed developmental potentials of NSC in the neonatal human hippocampus

(A) RNA velocity analysis indicating the developmental trajectory of hippocampal neurogenic lineage at postnatal Day 4. Cell types are labeled.

(B) Representative GO terms of the differentially expressed genes compare qNSC1, qNSC2 with pNSC.

(C) Pseudotime reconstruction of the neurogenic lineages in the neonatal human hippocampus. Dots showing individual cells. Different color represents different cell types. The arrows indicate the directions of differentiation trajectories. pNSCs as the development root was successively followed by aNSCs and neuroblasts, and then separated into two branches (1 and 2), generating two types of neuronal cells N1 and N2, respectively.

(D) Expression dynamics of cell type specific genes along with the pseudotime. Each dot represents an individual cell. NSC genes (HOPX, VIM and SOX2), granule neuroblast genes (DCX and STMN2), and mature granule cell gene (SYT1) are shown.

(E) Immunostainings of radial glia (NSC) markers (HOPX and NES), active NSC markers (NES and Ki67) and neuroblast marker (PSA-NCAM). The HOPX+NES+ RGL cells and NES+Ki67+ active NSCs with long apical processes were detected in postnatal Day4 hippocampal dentate gyrus (arrows). The PSA-NCAM+ neuroblasts (green) were located across the GCL. Scale bars of HOPX/NES immunostaining are 200 μm; the magnified and further magnified cell images are 100 μm and 10 μm, respectively; the arrowhead indicates the vessel. Scale bars of KI67/NES immunostaining are 100 μm and 10 μm, respectively. Scale bars of PSA-NCAM immunostaining are 100 μm and 10 μm, respectively; arrows indicate the neuroblasts.

(F) Heatmap showing that differentially expressed genes along the pseudotime were divided into four clusters. Representative genes and enriched GO terms of each cluster are shown (p-value < 0.05).

(G) Representative NSC genes (HOPX, VIM, CHI3L1 and TNC), and neuronal genes (NRGN, STMN2 and SV2B) were ordered by Monocle analysis along with the pseudo-time. Cell types along with the developmental trajectory were labeled by different colors.

Age-dependent molecular alterations of the hippocampal NSCs and NBs

(A and B) Feature plots (A) and quantification (B) of the neurogenic populations during aging. Neonatal (N), adult (Ad), aging (Ag). The neurogenic populations include qNSC1, qNSC2, pNSC, aNSC and neuroblast.

(C) The dynamic expression of some representative genes, including newly identified qNSCs genes (LRRC3B, RHOJ, and SLC4A4), NSC genes (HOPX, SOX2, VIM, NES and CHI3L1), neural progenitor or proliferation genes (ASCL1, EOMES and MKI67), and immature granule cell genes (STMN2 and DCX), in human hippocampus across neonatal (D4), adult (31y, 32y) and aging (50y, 56y, 60y, 64y-1, 64y-2, 68y).

(D) Immunostaining of classical NSC markers (HOPX, VIM and NES) in human hippocampal dentate gyrus across different ages (postnatal day 4, 32y, 50y, 56y). Scale bars, 60 μm.

(E) Violin plot showing differentially expressed genes of qNSC1 and qNSC2 in the aging group compared to the neonatal group.

(F) Representative GO terms of significantly (avg(log2FC) > 0.5, p-value < 0.05) up-and down-regulated genes in qNSC1 and qNSC2 during aging.

The transcriptomic signatures of the activated neurogenic lineage in the adult human injured hippocampus induced by stroke

(A) The neurogenic lineage included qNSC1, qNSC2, reactivated pNSC/aNSC and NB. Cell distribution showing by feature plots.

(B) Quantification of qNSC1, qNSC2, pNSCs, aNSCs and neuroblasts in neonatal (N), adult (Ad), aging (Ag) and stroke-injured (I) hippocampus, respectively.

(C) The dynamic expression of NSC markers (VIM, NES and CHI3L1) and oligodendrocyte progenitor cell markers (SOX10) from neonatal (N), adult (Ad), aging (Ag) and injured (I) hippocampus.

(D) Immunofluorescence images of NES (green)/Ki67 (red), VIM (red), and CHI3L1 (red)/NES (green) showing a few active NSC cells in the 48-year-old injured hippocampal dentate gyrus. The arrows indicate radial morphology NES+/KI67+, VIM+ or CHI3L1+/NES+ active NSC cells, respectively. Scale bars, 100 μm; the magnification, 20 μm.

(E and F) Integrative analysis of pNSC and aNSC from stroke injury and neonatal hippocampus showing that these cells were subclustered into 8 clusters (E), which were further annotated into pNSC, aNSC and reactive astrocytes according to gene set scores (average, over genes in the set, of seurat function AddModuleScore) (F).

(G) Pseudotime reconstruction of the neurogenic lineage in the stroke-injured human hippocampus. Different colors represent different cell types. The arrow indicates the developmental direction.

(H) Heatmap showing the expression profiles of differentially expressed genes (DEGs) in four clusters along the pseudotime. Representative DEGs and enriched GO terms of each cluster are shown.

(I) The significantly up-regulated genes in neurogenic lineage upon injury compared with aging.

(J) The GO term analysis of up-regulated genes in the neurogenic lineage upon injury compared with aging.

Cell atlas of human hippocampus across different ages and post stoke-induced injury.

(A) Visualization of major cell types from human hippocampal snRNA-seq data by using 3D UMAP. (B) Cell atlas of each human hippocampal sample. Different colors indicate different samples. (C) Heatmap of top 50 genes (p-value < 0.05) specific for each major population after normalization. Adult-AS, adult astrocyte; AS/qNSC, astrocyte/quiescent neural stem cell; pNSC, primed NSC; aNSC, active neural stem cell; NB, neuroblast; GC, granule cell; GABA-IN, GABAgeric-interneuron; Pyr, pyramidal neuron; OPC, oligodendrocyte progenitor cell; OLG, oligodendrocyte; MG, microglia; EC, endothelia cell; Per, pericyte; CR, Relin-expressing Cajal-Retzius cell; Unknown1 (UN1); Unknown2 (UN2). Related to Figure 1.

Reported neuroblast genes were widely distributed in the adult human interneurons.

(A) Transcriptional congruence of granule cell lineage and interneuron population between our dataset and published mouse, macaque and human transcriptome datasets. The matrix plot indicates the similarity scores of given human hippocampal cell populations from our dataset (rows) assigned to the corresponding literature-annotated cell types (columns). (B) Neuroblast genes reported by several literatures were widely distributed in the adult human interneurons from 10 individuals. (C) Our identified neuroblast specific genes were absent in the adult human interneurons. Related to Figure 3.

Pseudotime reconstruction of the neurogenic lineage development in the neonatal Day 4 human hippocampus.

(A) Cells during neurogenic lineage development were ordered by Monocle analysis along pseudo-time. (B) Cell types located at developmental trajectory were labeled by different colors. (C and D) Heatmap (C) and UMAP (D) visualization of distinct differentially expressed genes between N1 and N2. (E) The dynamic changes of representative genes along the pseudo-time were shown by Monocle analysis. Cell types along with the developmental trajectory were labeled by different colors. (F) Heatmap showing expression dynamics of transcriptional factors (TFs) along the neurogenesis trajectory. The representative TFs in each cluster were shown on the right. Related to Figure 4.

Alterations of the neurogenic lineage in human hippocampus during aging.

(A) Bubble plots showing our identified pNSCs, aNSCs and NBs from 10 individuals still express neural stem cell and neuroblast marker genes during aging despite their rare number. 48y donor was a stroke sample. (B) Immunostainings of classical NSC markers (HOPX, VIM and NES), pNSC gene CHI3L1 identified by us and neuroblast marker PSA-NCAM in human hippocampal dentate gyrus across different ages (postnatal day 4, 32y and 56y). Scale bars: 4D-500 μm, 32y-800 μm, 56y-600 μm. Related to Figure 5.

Differentially expressed genes and enrichment functions in pNSC, aNSC, and NB along aging, respectively.

(A) Heatmap showing differentially expressed genes (DEGs) between aging pNSC and neonatal pNSC (p-value < 0.05). (B and C) Representative GO terms of significantly up-regulated (B) and down-regulated (C) genes during pNSC aging (avg(log2FC) > 0.5, p-value < 0.05). (D) Heatmap showing DEGs between aging aNSC and neonatal aNSC (p-value < 0.05). (E and F) Representative GO terms of significantly up-regulated (E) and down-regulated (F) genes during aNSC aging (avg(log2FC) > 0.5, p-value < 0.05). (G) Heatmap showing DEGs between aging NB and neonatal NB (p-value < 0.05). (H and I) Representative GO terms of significantly up-regulated (H) and down-regulated (I) genes during NB aging (avg(log2FC) > 0.5, p-value < 0.05). Related to Figure 5.

Stroke injury induced hippocampal cell apoptosis, astrocyte reactivation and neuronal damages.

(A) Representative GO terms of upregulated genes in stroke-injured hippocampal GCs and INs, compared with the normal aged hippocampus. (B) Genes relative to apoptosis, DNA damage and autophagy were significantly upregulated in the stroke-injured hippocampus, compared with the normal aged hippocampus. I, injury; Ag, aging. (C) TUNEL assay showing obvious cell apoptosis in the 48y dentate gyrus, but not in other adult samples, which confirmed the stroke caused hippocampal injury. Scale bars, 500 μm. (D) CHI3L1 and VIM co-immunostaining showing some CHI3L1+VIM+ and CHI3L1+VIM-cells exhibited morphologies of reactive astrocyte (arrowhead) and neuron (arrow) in the GCL and hilus, respectively. Scale bars, 500 μm; the magnified images, 100 μm and 20 μm. Related to Figure 6.

Initially defined pNSCs and aNSCs from stroke-injured hippocampus contained reactive astrocytes and reactivated NSCs.

The integrative analysis of single cell data was based on initially defined pNSC and aNSC populations from the neonatal Day 4 and 48y-stroke injury hippocampus. (A) Heatmap of top 10 genes (p-value < 0.05) specific for each major cluster after normalization, relative to Fig. 6E. (B) The fraction of subpopulations in total cells (cluster 0-7). (C) UMAP feature plots showing expression distribution of cell type specific genes in cell subpopulations, including RGL marker genes (VIM, HOPX, LPAR1 and SOX2), neurogenic development genes (STMN1), and reactive astrocytes marker gene (OSMR, TIMP and LGALS3). (D) The dynamic expression of cell type specific genes along the pseudotime. Each dot represents an individual cell. These representative genes included RGL genes PAX6 and HOPX, reactivated NSC genes VIM, CD44, TNC, CHI3L1 and SOX2, and cell cycle genes CKAP5 and RANGAP1, and neuroblast gene STMN2. Related to Figure 6.

Integration of our snRNA-seq dataset with other published data.

(A) ETNPPL as a new NSC marker and STMN1/STMN2 as new immature neuron markers validated in Wang’s study were verified in our study. (B) Integration of Zhou’s snRNA-seq dataset of 14 aged donors (from 60-92 years old) with our snRNA-seq dataset. We did not detect evident pNSC, aNSC or NB populations in their dataset (circle with a dotted line). (C and D) UMAP visualization of pNSC/aNSC markers (TNC and VIM) (C), and neuroblast markers (STMN1 and NRGN) (D) in our and Zhou’s snRNA-seq dataset. Related to discussion.