Single-cell profiling of human subventricular zone progenitors identifies SFRP1 as a target to re-activate progenitors

Following the decline of neurogenesis at birth, progenitors of the subventricular zone (SVZ) remain mostly in a quiescent state in the adult human brain. The mechanisms that regulate this quiescent state are still unclear. Here, we isolate CD271+ progenitors from the aged human SVZ for single-cell RNA sequencing analysis. Our transcriptome data reveal the identity of progenitors of the aged human SVZ as late oligodendrocyte progenitor cells. We identify the Wnt pathway antagonist SFRP1 as a possible signal that promotes quiescence of progenitors from the aged human SVZ. Administration of WAY-316606, a small molecule that inhibits SFRP1 function, stimulates activation of neural stem cells both in vitro and in vivo under homeostatic conditions. Our data unravel a possible mechanism through which progenitors of the adult human SVZ are maintained in a quiescent state and a potential target for stimulating progenitors to re-activate.

isoflurane and oxygen (0.5 mL/min). Pups were fixed on a plate placed in a stereotaxic rig and 2 µL of a mix of the plasmid (final concentration of 3 µg/µL) and Fast Green was injected at the level of the right lateral ventricle at a depth of 1.5 mm from the surface of the skull using a Hamilton syringe with a 34G needle (Nevada, USA). Following injection, mice were immediately subjected to five electrical pulses (85V, 50 ms, separated by 950 ms intervals) using the ECM 830 square wave electroporator and a 7 mm platinum tweezertrode (BTX, Holliston, USA).

Intraperitoneal injection of small molecule
Inhibition of SFRP1 was done by intraperitoneal injections of WAY-316606 (HY-10858 MedChemExpress, NJ, USA), which binds to SFRP1. Intraperitoneal injection was given between P2-P5 for the early postnatal stimulation, twice a day (six injections of 10 µL each), with a final concentration of 0.3 mM (dissolved in sterile PBS with 10% DMSO). Control pups received six injections of PBS with 10% DMSO. An n=3 (control) and n=4 (treated) was used for immunofluorescence analysis. Fourteen pups were terminated (n=7 controls and n=7 treated) and the SVZ was isolated for qRT-PCR analysis. Briefly, the brain was removed and the entire SVZ was dissected and immediately frozen in liquid nitrogen until further use.
Images used for analysis were taken with a Zeiss AxioScope A1 epifluorescence microscope. All images were captured with the same exposure time for both conditions. Six images were quantified per condition (n=6) and analyzed in ImageJ v1.53c. An automated counting method using the Moments threshold for KI67 and SOX2, and the RenyiEntropy threshold for DAPI staining, which is provided by the software, was used to determine the number of cells positive for both markers. High magnification images were taken with a Zeiss AxioScope A1 epifluorescence microscope using 100x/ 1.3NA oil objective, an AxioCam camera (Zeiss), and the software AxioVision v4.8.2.0 or on a Zeiss LSM880 confocal laser microscope using 63x/ 1.4NA oil DIC M27 objectives, an AxioCam MRm camera (Zeiss), and the software Zen black Z.1SP3. Images were taken with a z-step of 1.0 µm and a resolution of 1024x1024.

Quantification and statistical analysis Alignment and processing
Alignment to the human transcriptome was performed using a custom pipeline (https://github.com/anna-alemany/transcriptomics/tree/master/mapandgo/starmap) 8 . Raw reads were trimmed, aligned to the Ensembl release 75 Homo sapiens genome using RNA STAR version 2.53A (Spliced Transcripts Alignment to a Reference) 9 and demultiplexed using cell specific barcodes. Reads that mapped equally well to multiple locations were discarded. Duplicate reads were removed in the case of identical combinations of library, cellular, and molecular barcodes and were mapped to the same gene. Transcript count was adjusted based on read counts and the presence of one of the 4096 possible UMI's.

Filtering and normalization
All analysis were performed on R-studio version 4.0.2. Quality check and filtering was performed on Seurat v3.2.2 10 using the following parameters: Only genes that were detected in at least two cells were taken for downstream analysis. Cells that had less than 100 genes or more than 3000 genes detected and that had more than 6% of their counts mapped to the mitochondrial genome were removed. This resulted in a Seurat Object of 728 cells and 22682 genes.
Normalization was done using Seurat (NormalizeData function with LogNormalize method) where a generalized linear model for each gene is constructed 11 . The effect of differences in sequencing depth, library preparation, and donor was removed from the normalized expression values with the ScaleData function from Seurat.

Single-cell clustering and visualization
Highly variable genes were identified using Seurat v3.2.2 (FindVariableFeatures function). The Wilcoxon rank sum test was used to identify marker genes. P-value adjustment was performed using the Bonferroni correction.

Integration of single-cell datasets
Integration of single-cell datasets published in Zhong et al. 14 and Jäkel et al. 15 with our dataset was done using Seurat v3.2.2 as described in Stuart et al. 16 . This pipeline allows the comparison of multiple datasets by identifying in each dataset the 2000 most variable genes using the FindVariableGenes function. Assuming that there are similarities between datasets and that a subset of cells have a shared biological state, a set of molecular features could be identified. These so-called "anchors" were determined with the FindIntegrationAnchors function with dims set on 35. These anchors are then passed to the IntegratedData function, which creates a Seurat object. This object contains an integrated expression matrix of all the cells, which will then cluster by cell type instead of platform technology or species.
Every gene was fitted in a linear regression model. Multiple hypothesis testing was corrected using the Benjamini and Hochberg test. Genes that had a time-dependent expression were filtered. Differentially expressed genes were considered significant when q-value < 0.01.

Imaging and quantification
The stained sections were analyzed on a Zeiss LSM880 confocal laser microscope using 40x/ 1.3NA oil DICII objectives (EC PlnN), an AxioCam MRm camera (Zeiss), and the software Zen black Z.1SP3. Images were taken with a z-step of 2.5 µm and a resolution of 1024x1024.
For staining of fetal forebrain sections two to four images were quantified. For the quantifications on the human SVZ, around 50% of the progenitors present in one SVZ section was imaged and analyzed. For mouse staining, the entire dorsal, medial, and lateral walls from the SVZ were quantified on at least two brain sections to avoid biases related to regional heterogeneity. The number of GFP positive cells was quantified within the SVZ (bin number 1) and from the border of the SVZ until below cortical layer 6 (bins 2-4). Because both frontal and caudal SVZ sections were quantified, the size of bins 2-4 were determined based on the distance from the SVZ to the subplate layer, which was divided by three to ensure equal bin sizes within frontal and caudal sections. Cells were quantified using the cell counter plugin from Fiji (version 1.52p) 19 .

Supplementary Figures
Supplementary Figure 1