Methods
No statistical methods were used to predetermine sample size. The
experiments were not randomized and investigators were not blinded
to allocation during experiments and outcome assessment.
Tissue sample collection
The de-identified human tissue collection and research protocols were
approved by the Reproductive Study Ethics Committee of Beijing Anz-
hen Hospital and the institutional review board (ethics committee)
of the Institute of Biophysics. The informed consent was designed as
recommended by the ISSCR guidelines for fetal tissue donation and
fetal tissue samples were collected after the donor patients signing
an informed consent document that was in strict observance of the
legal and institutional ethical regulations for samples from elective
pregnancy terminations at Beijing Anzhen Hospital, Capital Medical
University. All samples used in these studies had not been involved in
any other procedures. All the protocols were in compliance with the
Interim Measures for the Administration of Human Genetic Resources,
administered by the Ministry of Science and Technology of China.
Animals
Timed pregnant female mice at embryonic day 13.5 were used for
in utero electroporation experiments. Embryos for experiments after
in utero electroporation included both male and female mice. Mouse
housing and experimental protocols in this study were in compliance
with the guidelines of the Institutional Animal Care and Use Committee
of the Institute of Biophysics, CAS. All mice had free access to food and
water and were housed in the institutional animal care facility with a
12-h light–dark schedule.
Tissue sample dissection
Gestational age was measured in weeks from the first day of the wom-
an’s last menstrual cycle to the sample collecting date. Fetal brains
were collected in ice-cold artificial cerebrospinal fluid containing
125.0 mM NaCl, 26.0 mM NaHCO 3 , 2.5 mM KCl, 2.0mM CaCl 2 , 1.0 mM
MgCl 2 , 1.25 mM NaH 2 PO 4 at a pH of 7.4 when oxygenated (95% O 2 and 5%
CO 2 ). The hippocampus was dissected and put in hibernate E medium
(Invitrogen, Cat. A1247601). The hippocampus tissue was first digested
in 2 mg/ml collagenase IV (Gibco, Cat. 17104-019) and 10 U/μl DNase I
(NEB, Cat. M0303L) in hibernate E medium and then in 1 mg/ml papain
(Sigma, Cat. P4762) and 10 U/μl DNase I in hibernate E medium. Samples
were vortexed at 300g and 37 °C on a thermocycler for 20 min. Further
pipetting was used to fully digest the tissue into single cells. After that,
the cell suspension was centrifuged at 700g for 5 min to obtain the cell
pellet. The digestion medium was carefully removed and the cell pellet
was resuspended in 300 μl 0.04% BSA in PBS and kept on ice.
RNA library preparation for high-throughput sequencing
Thousands of cells were partitioned into nanolitre-scale Gel Bead-
In-EMulsions (GEMs) using 10x GemCode Technology, where cDNA
produced from the same cell shares a common 10x Barcode. Upon dis-
solution of the single cell 3′ gel bead in a GEM, primers containing an
Illumina R1 sequence (read1 sequencing primer), a 16-bp 10x Barcode, a
10-bp randomer and a poly-dT primer sequence were released and mixed
with cell lysate and Master Mix. After incubation of the GEMs, barcoded,
full-length cDNA from poly-adenylated mRNA was generated. Then
the GEMs were broken and silane magnetic beads were used to remove
leftover biochemical reagents and primers. Prior to library construc-
tion, enzymatic fragmentation and size selection were used to optimize
the cDNA amplicon size. P5, P7, a sample index and R2 (read 2 primer
sequence) were added to each selected cDNA during end repair and
adaptor ligation. P5 and P7 primers were used in Illumina bridge ampli-
fication of the cDNA (http://10xgenomics.com). Finally, the library was
sequenced into 150-bp paired-end reads using the Illumina HiSeq4000.
Data processing of scRNA-seq from Chromium system
Cell ranger 2.0.1 (http://10xgenomics.com) was used to perform quality
control and read counting of Ensemble genes with default parameters
(v2.0.1) by mapping to the hg19 human genome. We excluded poor-
quality cells after the gene-cell data matrix was generated by Cell Ranger
software using the Seurat package (v2.3.4). Only cells that expressed
more than 800 genes and fewer than 7,000 genes were considered, and
only genes expressed in at least 30 single cells (0.1% of the raw data)
were included for further analysis. Cells that expressed haemoglobin
genes (HBM, HBA1, HBA2, HBB, HBD, HBE1, HBG1, HBG2, HBQ1 and HBZ)
were also excluded. Cells with a mitochondrial gene percentage over
15% were discarded. In total, 17,737 genes across 30,416 single cells
remained for subsequent analysis. The data were normalized to a total
of l × 10^4 molecules per cell for the sequencing depth using the Seurat
package. The batch effect was mitigated by using the ScaleData func-
tion of Seurat (v2.3.4).Identification of cell types and subtypes by dimensional
reduction and PAGA analysis
The Seurat package (v2.3.4) was used to perform linear dimensional
reduction. We selected 982 highly variable genes with average expres-
sion between 0.0125 and 8 and dispersion greater than 2 as input for
PCA. Then we identified significant PCs based on the JackStrawPlot
function. Strong PC1–PC10 were used for t-SNE to cluster the cells by
FindClusters function with resolution 1.2. Clusters were identified by
the expression of known cell-type markers and GO analysis. The markers
ASCL1, NEUROD2, GAD1, OLIG2, MBP, AQP4, SPARC and PTPRC were
used to hippocampal cells as progenitor cells, excitatory neurons,
inhibitory neurons, OPCs, oligodendrocytes, astrocytes, endothelial
cells and microglia, respectively.
Three-dimensional t-SNE was applied to cluster all cells in the human
developing hippocampus (dim.embed = 5) with PC1–PC10. Visualiza-
tions were done using rgl package (v0.99.16) implemented in R. We
then applied partition-based graph abstraction (PAGA) to predict a
lineage tree for the hippocampal and the prefrontal cortical cells in an
unbiased way. We produced a consolidated lineage tree that included
all identified cell types rooted to a stem cell group.Identification of DEGs among clusters
The DEGs of each cluster were identified using the FindAllMarkers func-
tion (thresh.use = 0.25, test.use = “wilcox”) with the Seurat R package
(6). We used the Wilcoxon rank-sum test (default), and genes with aver-
age expression difference >0.5 natural log with P < 0.05 were selected
as marker genes. Enriched GO terms of marker genes were identified
using DAVID 6.8^28 ,^29 (https://david.ncifcrf.gov/home.jsp) and Metas-
cape^30 (http://metascape.org).Constructing single cell trajectories in the hippocampus
The Monocle 2 R package (version 2.6.4) and Monocle 3 alpha R package
(version 2.99.2) were applied to construct single cell pseudo-time tra-
jectories to discover developmental transitions^31 –^33. We used highly vari-
able genes identified by Seurat to sort cells into pseudo-time order. The
actual gestational time of each cell informs us which states of cells are at
the beginning of pseudo-time in the first round of “orderCells”. We then
call “orderCells” again, passing this state as the root_state argument.
“DDRTree” and “UMAP” were applied to reduce dimensional space and
the minimum spanning tree on cells was plotted using the visualization
functions “plot_complex_cell_trajectory” or “plot_3d_cell_trajectory”
for Monocle 2 and Monocle 3 alpha, respectively.Cell-cycle analysis
In the cell-cycle analysis, we applied a cell-cycle related gene set
with 43 genes expressed during G1/S and 54 genes expressed during
G2/M^34 ,^35. We defined the G1/S and G2/M states of each cell by comparing