All experimental procedures were approved by the Malmö–Lund Ethical Committee for the use of laboratory animals and were conducted following the European Union directive on animal rights. Mice expressing a green fluorescent protein (GFP) under the Dcx promoter (Dcx-GFP) (Couillard-Despres et al., 2006), which is highly active in neuroblasts, were bred and housed in the animal facility connected to the Lund University Biomedical Center under sterile conditions in individually ventilated cages (IVCs) at 22°C, with 40%−60% humidity and a 12-h light/dark cycle with ab libitum access to food and water. Adult (3–4 months), middle-aged (12–16 months), and aged (18–24 months) mice were used, with male and female mice equally distributed between groups. Tissues from four to five mice were pooled for each age group.

Heterozygote Dcx-GFP mice were euthanized by cervical dislocation, and their brains were kept in an ice-cold L-15 medium (Invitrogen: Waltham, Massachusetts, United States), as previously described (Ahlenius and Kokaia, 2010). To facilitate the comparison of neurogenic niches, the OBs were isolated and the SVZs and DGs were dissected from 1-mm sections separately. Tissue from 4-5 mice were pooled in ice-cold L-15 medium separately for each region (from the same group of animals) and dissociated into a single -cell suspension using the Adult Brain Dissociation Kit, mouse and rat (Miltenyi Biotech: Bergisch Gladbach, North Rhine-Westphalia, Germany, 130-107-677) following the manufacturer's guidelines. The cells were then purified through a two-step gradient as previously described (Ahlenius and Kokaia, 2010) and resuspended in fluorescence-activated cell sorting (FACS) medium [4% Bovine serum albumin (BSA) (W/V), 2% 2M 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) (V/V), and 0.1% Sodium azide (V/V) in Hanks' balanced salt solution (HBSS) (BSA, HEPES and HBSS: Thermo fisher; Waltham, Massachusetts, United States)].

To enrich for neural progenitors, dissociated cells from Dcx-GFP mice were subjected to fluorescence-activated cell sorting (FACS; Supplementary Figure 1a) for separation based on their endogenous GFP expression, and cell viability was assessed with propidium iodide using FACSARIAII/III (BD Biosciences: Franklin Lakes, New Jersey, United States). The sorted cells were collected into Phosphate-buffered saline (PBS) with 2% Fetal bovine serum (FBS) for single-cell sequencing library preparation on the same day. Alternatively, these cells were collected into QIAzol lysis reagent (Qiagen: Venlo, Netherlands (corporate) Hilden, Germany (operational)) and stored at −80°C for bulk sequencing.

For bulk RNA sequencing, total RNA was isolated from the sorted GFP+ cells using the miRNeasy Micro Kit (Qiagen) according to the manufacturer's instructions. Briefly, the samples frozen in the QIAzol lysis reagent were brought to room temperature and homogenized by pipetting, and the RNA was then separated with a chloroform phase gradient. During RNA purification, on-column DNase digestion (Qiagen) was performed, and the final RNA was eluted into 14 μl of H₂O. Concentrations were measured using a bioanalyzer (Agilent: Santa Clara, California, United States) and loaded using the RNA 6000 Pico Kit (Agilent).

For single-cell RNA sequencing, libraries were generated from FACS-sorted cells expressing endogenous GFP (Dcx-GFP) for each experimental group separately using a Chromium system (10 × Genomics: Pleasanton, California, United States), according to the manufacturer's guidelines [Single Cell 3′ Reagent Kits v2 User Guide (CG00052)], with 12–13 cDNA amplification cycles and 14 sample index cycles. The libraries generated from 2,100 aged DG cells, 3,000 aged OB cells, and 5,000 cells for all other samples were sequenced on a NextSeq 500 System (Illumina: San Diego, California, United States) using NextSeq 500/550 High Output v2 Kits (Illumina). Each sample contained cells from four to five mice. The cells from different regions but of the same age group were obtained from the same group of animals.

Bulk RNA sequencing libraries were prepared using the SMARTer Stranded Total RNA-Seq Kit v2, Pico Input Mammalian (Takara Bio USA, Inc.) according to the manufacturer's user manual (063017). Furthermore, 2 ng of total RNA was used from each sample with 4-min fragmentation time and 14 PCR cycles for amplification of the final library in a Mastercycler X50s (Eppendorf: Hamburg, Germany). The libraries were sequenced on a NextSeq 500 System (Illumina) using NextSeq 500/550 High Output v2.5 Kits (Illumina). The sequencing was performed at the Center for Translational Genomics, Lund University, and Clinical Genomics Lund, SciLifeLab. Each biological sample (n = 5) contained cells from four to five mice for the bulk data analysis. The cells from all different regions were collected from every animal included in the study.

Raw data from single-cell RNA sequencing were processed through the Cell Ranger pipeline to generate count matrices, which were then loaded into the R computational environment (4.1.2) using read10xCounts (DropletUtils) and merged into a single-cell experiment object (i.e., SingleCellExperiment). The counts were normalized using logNormCounts (Scater) with size factors based on quickCluster and computeSumFactors (Scran). Then, the logcounts were extracted as dgCMatrices and merged into a Seurat object using CreateSeuratObject (Seurat). The low-quality cells with fewer than 75 or more than 4,000 detected genes (features), more than 1.5% mitochondrial genes, or more than 15% ribosomal genes were discarded (Supplementary Figure 1b). A very low number of cells expressing high levels of Cx3cr1 and Iba1, which are markers for microglia, were also discarded.

For bulk sequencing, FASTQ files were organized using bcl2fastq2 with default settings and without warnings from FastQC (Supplementary Figures 2a–c). Reads were aligned to the mouse reference genome (GRCm38) from the Ensemble database using HISAT2 (Supplementary Figures 2d, e). Quality control of the aligned data, including base distribution and insert size checks, was performed using Picard (Supplementary Figure 2a). Assembly of the alignments into full transcripts and quantification of expression levels were performed using StringTie.

The data analysis was performed in the R computational environment (4.1.2). All genes were scaled with ScaleData (Seurat) to accurately compare the expression between cells. The 2,000 most highly variable genes (HVGs; Supplementary Figure 1c) were identified using the variance stabilizing transformation (VST) method via FindVariableFeatures (Seurat), which were then used for the principal component analysis. The first 20 principle components with the highest standard deviation were selected manually from an elbow plot and used for Louvain clustering (Seurat) with a k parameter of 10 and a resolution of 1 (Figure 1A). Clusters were visualized with uniform manifold approximation and projection (UMAP), initially annotated using singleR (1.8.1) based on a reference mouse brain dataset from Celldex (Benayoun et al., 2019) (Figure 1B), and further adjusted manually to commonly used markers (i.e., Glial fibrillary acidic protein (Gfap) for astroglial and Neural stem cells (NSCs), Achaete-scute family bHLH transcription factor 1 (Ascl1) for NSCs and early progenitors, Doublecortin (Dcx) for neuroblasts, Synaptotagmin 1 (Syt1) for neurons, Platelet derived growth factor receptor alpha (Pdgfra) for oligodendrocyte precursors, and Mbp for mature oligodendrocytes; Figures 1C–H). The clusters assigned as oligodendrocytes were discarded, along with the clusters where more than 40% of the population expressed Gfap and < 40% expressed Ascl1, which were considered astroglial populations without activated NSCs. DimPlot (Seurat) was used to visualize the clusters, while FeaturePlot (Seurat) was used to illustrate gene expression.

For single-cell RNA sequencing data, differential gene expression analysis was performed using findMarkers (Seurat), considering p-values based on Bonferroni correction, using all genes in the dataset; the adjusted p-values (p_val_adj) of < 0.05 are considered statistically significant. ComplexUpset was used for upset plots, Ggplot (ggplot2) for boxplots and extreme point plots, and DoHeatmap (Seruat) for heatmaps. Gene ontology (GO) terms were identified for differentially expressed genes (DEGs), with the adjusted p-values refers to how the DEGs were chosen, while other settings refers to the enrichGO function from ClusterProfiler (4.2.2), with parameters set to minGSSize 1, maxGSSize 500, and both qvalue Cutoff and pvalueCutoff were set to 1. GraphPad Prism v7 was used to visualize population size estimation data.

For bulk RNA sequencing data, DESeq2 was used for differential gene expression (DGE) analysis, which compares adult samples to a combined group of middle-aged and aged samples. Heatmaps were generated using pheatmap, color pallets were created using RColorBrewer, and bar graphs were created using GraphPad Prism v7. Inflammatory response genes can considerably vary due to infection or other external stimuli; therefore, to identify outliers, two independent statistical tests were conducted: Cooks' distance (Supplementary Figures 3a, b) and Grubbs' test (Supplementary Table 1). One adult sample was identified as an outlier for inflammation-related genes in both tests and was not included in the DGE analysis for inflammatory response genes. The DGE analysis included four or five biological replicates, each pooled with the RNA of four to five adult mice. Furthermore, 10 middle-aged and aged mice were included in the DGE analysis.

Cell numbers were normalized to OPCs since it is the population in our data set most stable during aging and present in large enough quantity.

NSCs are fewer in number and less active in the aged brain compared to the adult mammalian brain. This decline leads to the production of fewer IPs and, ultimately, fewer new neurons added to the brain during aging (Katsimpardi and Lledo, 2018). Although recent research has revealed age-related changes in stem cells, little is known about how IPs such as neuroblasts are affected by aging.

To study the effect of aging on IPs, including TAPs and neuroblasts, in the two major neurogenic niches, we dissected the SVZ, OB, and DG from 3-, 14-, and 24-month-old Dcx-GFP mice. The GFP+ cells were isolated using FACS and single-cell RNA sequencing was performed using 10 × and Illumina chemistry. During data processing, we performed unbiased clustering of cells from all ages and regions, which were digitally pooled together (Figure 1A). Clusters were annotated computationally based on previous RNA sequencing studies (Figure 1B) and manually verified and further specified using common markers. Cells in clusters with high expression of both Gfap and Ascl1 were annotated as NCSs, with Ascl1 alone as TAPs, Dcx as neuroblasts, Syt1 as mature neurons, Pdgfra as oligodendrocyte progenitors, and Mbp as mature oligodendrocytes (Figures 1C–H). The expression of cell cycle genes (Figures 1I, J) highlighted highly proliferative cells, such as TAPs.

We identified all major common cell types involved in neurogenesis with a strong enrichment of IPs. Interestingly, some clusters split into substates that were not previously described (Figure 1A), indicating that enrichment, by FACS, for neural progenitors aided in revealing their full heterogeneity.

Clusters with cells from neuronal lineages were selected for further downstream analysis (Figure 1K). We identified two differentiation trajectories from NSCs to mature IPs in the SVZ (Supplementary Figure 1d), one leading to IPs with a potential to generate inhibitory neurons and one to excitatory neurons as previously described (Sequerra et al., 2013). We identified a unique NSC population in the OB, as previously reported (Defterali, 2021), initiating a trajectory leading to inhibitory neurons (Supplementary Figure 1e). In the DG, we identified a single trajectory leading to excitatory neurons (Supplementary Figure 1f).

As expected, from previous histological studies (Jurkowski et al., 2020), the cells from the DG distinctly separated from most cells in the SVZ or OB, and the majority of mature cells could be detected only in the OB or DG. Most cell types at various maturity states identified in the SVZ were also detected in the OB, albeit to a lesser extent, and only a few SVZ and OB cells were clustered with DG cells (Figure 1L).

For each cluster where the population size remained consistent with age and there were sufficient cells in each age group to run the analysis, we performed the DEG analysis between adult and aged cells in the SVZ, OB, and DG separately (Figures 2A, B). We found only a few genes that were upregulated or downregulated in most aged cell types in the DG and SVZ and none in the OB. In the SVZ, none of these DEGs were universal for all cell types; while some were shared, others were unique. The panel of DEGs also differed between neighboring transient cell states, although a larger proportion was shared (Figure 2C).

All non-cycling neuroblast clusters (NB1, NB2, and NB3) exhibited very similar gene expression profiles (Figure 2C) and shared DEGs with age. We, therefore, decided to digitally pool them together to achieve a more powerful DEG analysis, revealing several genes involved in axonogenesis, such as Tubb3, Stmn1, and Nr4a2, to be differentially expressed between age groups (Figure 2D). Tubb3 is a major component of axons (Radwitz et al., 2022) and, along with Nr4a2 (Nurr1) and Stmn1, has been shown to promote neurite growth (Heng et al., 2012; Latremoliere et al., 2018; Kwon et al., 2020).

Next, we performed differential gene expression analysis between adult and middle-aged and between middle-aged and aged SVZ clusters. We detected several genes to have an expression maximum in the mid aged group that either dropped to, or below adult levels in the aged group (Figure 2E). This indicates that transcriptional changes occurring during aging are not always monotonic but dynamic. Many of these extreme-point genes coded for ribosomal proteins in the neuroblast clusters, while Hist1h2ap, a gene encoding a replication-dependent histone, was prominent in the TAP3 cluster of cycling IPs (Supplementary Figure 1b; Figure 2E).

In summary, aging leads to only a few differentially expressed genes in IPs, with some unique to specific cell types and some shared, but none universal. Despite the small changes, our data indicate that transcriptional changes during aging can be non-monotonic and that aged neuroblasts differentially express genes involved in axon development.

NSCs in the adult DG primarily form excitatory neurons, while SVZ neurogenesis mainly produces inhibitory interneurons (Lledo et al., 2006). Having established that DG and SVZ IPs have different gene expression profiles and rarely share aging-related transcriptional changes, we next assessed if aging affects neural progenitor heterogeneity.

To investigate IP heterogeneity during aging, we used clusters generated from digitally pooled samples (Figure 1A) and identified neural progenitors fated for different neuronal subtypes based on previously identified genes. We used Gad1 (Gad67) and St18 [identified here, and known to regulate inhibitory neuron fate (Nunnelly et al., 2022)] to highlight IPs set to become inhibitory interneurons and Eomes (Tbr2), Tbr1 and Slc17a6 (Vglut2) for IPs set to become excitatory neurons. Excitatory markers were clearly expressed in DG IPs in contrast to the majority of SVZ IPs that instead expressed Gad1. However, when separating the two neurogenic regions, we found that a small fraction of SVZ IPs clustered with DG IPs and expressed markers for excitatory neurons (Figure 3A). This previously described subpopulation of neural progenitors (Sequerra et al., 2013) has not been well studied in the context of aging.

The number of OPCs has shown to remain fairly constant during aging (Wang et al., 2020), while other large populations in our dataset varied. We, therefore, used OPCs as a reference to analyze how the SVZ and DG IP populations changed in size with aging. Interestingly our single cell data suggested that the population of Eomes (Tbr2) positive IPs, in both neurogenic regions, decreased more rapidly with age compared to IPs positive for Gad1 (Figures 3B–E).

To test this statistically, we sorted GFP-positive cells from adult, middle-aged, and aged SVZ of Dcx-GFP mice, performed bulk mRNA sequencing, and analyzed markers specific for the different IP populations. Our bulk data, indeed, showed that the expression of Eomes (Tbr2), Tbr1, and Slc17a6 (Vglut2) decreased more rapidly with age in the SVZ and DG compared to Gad1 and St18 in the SVZ (Figures 3F–H).

In summary, our RNA expression data suggest that the number of IPs set to become excitatory neurons, regardless of the neurogenic region, decreases faster during aging than that of IPs set to become inhibitory interneurons.

Having shown that most aged neural progenitors retain their expression profiles with minor differences, while the composition of neuronal progenitor subtypes changes, we next investigated whether any IPs acquire new cellular identities during aging. Concomitant with decreased neurogenesis, inflammation in the brain increases (Ferrucci and Fabbri, 2018). However, it is unknown whether aged IPs change their gene expression related to immune responses and whether such changes are global or selective.

Since we did not find any major age-related gene expression differences in clusters where population size was unchanged (Figures 2A, B), we decided to analyze cluster NB4 that clearly increased in size in the SVZ with age (Figures 4A, B) and was not detectable at any age in either OB or DG. Interestingly, compared to Gad1+ IPs (NB1, NB2, and NB3), NB4 exhibited a gene expression pattern similar to neuroblasts but with several upregulated genes involved in innate immune responses and responses to external biotic stimuli (Figure 4C). By exploring the differentially expressed genes (Supplementary Table 1), we found Usp18 (Ubiquitin Specific Peptidase 18) to be highly enriched in cluster NB4 and also detected its presence to a greater extent in other cells from aged samples (Figure 4D). Usp18 is induced by viral infection and type I interferons and inhibits apoptosis (Malakhova et al., 2006; Diao et al., 2020). In addition, we found Lgals9 (Galectin-9), a suppressor of T-cells (Yang et al., 2021), to be specific to cluster NB4 (Figure 4E). Therefore, we analyzed Usp18 and Lgals9, along with other top genes defining cluster NB4 (Supplementary Table 1), in our SVZ bulk data and found an increased expression of most of these immune response genes in the SVZ from the middle-aged and aged samples compared to those in adult samples (Figures 4F, G).

In summary, a large portion of IPs in the aged neurogenic niches exhibit only minor changes in gene expression that can be linked to immune responses, while a small selection of IPs in the SVZ not detected in the OB acquire a distinct expression profile related to immune responses.