Supplementary Materials for

Comprehensive cell atlas of the first-trimester developing human brain

Emelie Braun et al.

Corresponding author: Sten Linnarsson, sten.linnarsson@ki.se

Science 382, eadf1226 (2023)

DOI: 10.1126/science.adf1226

The PDF file includes:

Figs. S1 to S13

Other Supplementary Material for this manuscript includes the following:

Tables S1 to S8
MDAR Reproducibility Checklist
(previous page) Fig. S1 Specimen metadata and sampling overview. (A) Sampling overview per specimen: dot size
indicating number of cells sampled per region for each specimen. Bar chart (top) showing female specimens (black)
defined by the fraction of cells expressing the XIST gene. (B) Distribution of molecule counts (V2; median: 3 865, V3;
median: 9 551) and gene counts (V2; median: 2 018, V3; median: 3 737) of all cells in the whole dataset. (C) Violin
plots show distribution of UMI (molecule) and gene counts per cell and specimen, colors indicate chemistry version:
V2 (peach) and V3 (violet). (D) Stacked bar plots showing the distribution of cell classes by specimen, ordered by age
as in A and C. Asterisks indicate samples processed more than 48 hours after collection. (E) tSNE embedding colored
by 10X Chromium chemistry version. (F) tSNE embedding of whole dataset colored by specimen, legend displaying all
specimens and age per specimen. (G) Expression of immediate-early and mitochondrial genes by specimen, ordered
by chemistry and age. Asterisks indicate late-processed samples as in E. Samples obtained from Cambridge and
Karolinska Institutet are indicated in pale blue and dark green. (H) Class distribution for paired (BRC..., UK; XDD...,
Sweden) specimens matched by age and chemistry version.
(previous page) Fig. S2 Quality control metadata and discarded cells. (A) Fraction of droplet class per individual
sample/library, sorted in descending order of fraction of good cells. Black dots indicate samples that were processed more
than 48 hours after collection, which were associated with higher quality (P = 10-5, Mann-Whitney’s U test). (B) Example of
quality metrics in a good (top) and bad sample (bottom). Scatter plots (each dot is a cell/classified droplet) show
mitochondrial fraction (left) and droplet class (right) defined by their unspliced UMI ratio (x-axis) and total UMIs (y-axis).
Good sample: female specimen XHU:297, sample ID 10X101_5, forebrain, 10 pcw, v2 chemistry. Bad sample: male specimen
XDD:351, sample ID 10X187_5, 12 pcw, v3 chemistry. (C) tSNE embeddings of the same samples (good: top; bad: bottom)
after preliminary clustering to detect abundance of certain droplet groups. Pie charts showing the fractions of droplet
classes of the cells in each sample. (D) Embedding of clusters of all cells that did not pass QC (discarded droplets) colored
by droplet class. Pie chart showing percentages of droplet classes/cells that got discarded. (E) Clusters of discarded cells,
colored by individual cluster (cell type). (F) Gene expression of typical markers: TOP2A (cycling), HES1 (radial glia), CLDN5
(endothelial), AQP4 (pre-astrocytic), INA (neuronal, NHLH1 (neuroblast), SLC17A7 (excitatory neurons), GAD2 (inhibitory
neurons), SOX10 (oligodendrocyte precursors). (G) Total molecules of bad droplets (top), fraction of unspliced molecules
(bottom). (H) Predicted regions of each sequenced library. Each bar shows fraction of cells predicted as a certain region per
sample. Dots on x-axis indicate dissected region labels. Samples ordered in ascending age order.
Fig. S3 Metadata corresponding to Fig. 1. (A) Cell cycle score — the total UMIs detected from cell cycle genes
relative to the overall total UMIs — shown as vertical histograms, one per cluster. We used 0.4% as the threshold to
define cycling cells, indicated by dashed line. (B) Cell cycle score shown on the tSNE. (C) Fraction of cells with cell
cycle score > 0.4%, per cell class. (D) Region annotation shown on tSNE embedding. (E) Age shown on tSNE
embedding. (F) Regionalization of major classes. Bar plot showing the variance explained by multivariate linear
regression using 50 PCA components as the dependent variable, and subregion labels as independent variables,
shown separately for each major class. Dashed line, baseline implied by the variance explained in erythrocytes. (G)
Key markers used to identify major classes of cells: radial glia (HES1), neuroblasts (NHLH1), neurons (INA), glioblasts
(TNC and BCAN), OPCs and fibroblasts (PDGFRA), oligodendrocyte lineage (OLIG1). Expression is shown on a
perceptually uniform linear scale truncated at the 99th percentile, with zeros shown in grey.
(previous page) Fig. S4. Spatial transcriptomics on 5 pcw. human brain. (A) Manually curated anatomical regions
according to prosomeric model. (B) Expression of selected marker genes measured by EEL FISH. (C) Expression of
genes marking radial layers: ventricular, subventricular and mantle zone. (D) Maximum posterior probability cluster
assignments (top left) and individual cluster probability distributions indicating spatial locations of single-cell clusters.
(E) Spatial transcriptomics (Visium) showing expression of STMN2 (neurons) and SOX2 (radial glia) on four sagittal
sections of the same embryo as in (A).
Fig. S5. Spatial patterning of the neural tube in the anterior-posterior axis. (A) Gene expression quantification in
the ventricular zone of the most medial sample along the anterior-posterior axis. The area is unrolled in the direction
of the arrow and the colored segments correspond to the prosomeric model. *The borders of these areas were not
well defined in this embryo. (B) Example of genes that are expressed in only one segment (SIX6, OLIG3), in expression
level gradients (FOXG1, RSPO3, SIX3), symmetry between the dorsal and ventral part of the neural tube (FEZF2, EMX2,
EN1, EN2, PAX8), or asymmetry (FGF17, PAX5).
(previous page) Fig S6. Excitatory neurons lineage. (A) UMAP projection of the whole data from this study. Cells
used for pallial excitatory neuron lineage analysis are marked in red. (B) UMAP projection from pallial excitatory
neuron lineage colored by selected patterning genes expression. (C) UMAP projections for each indicated post-
conceptional ages of pallial excitatory neuron lineage colored by major cell classes. (D) Marker genes expression for
major cell classes. Dotplot showing mean expression (color) and ratio of cells (circle size) expressing the gene in each
major cell class. Expression values of each cell were normalized by total counts over all genes. Mean normalized
expression values for each gene are scaled between 0 and 1. (E-F) UMAP projections from all collected pallial
excitatory neuron lineage colored by the fraction of cell cycle genes (E) and genes related to progenitor states (F)
out of total UMIs. (G) Distribution of progenitor states by post-conceptional age in radial glia cells (top) and neuronal
IPC (bottom). (H-I) UMAP projection of pallial IPCs and early neuroblasts in 8.5 pcw (H) and 14 pcw (I) colored by
progenitor states inferred by cell cycle phase score (J) Progenitor states for pallial IPCs and early neuroblasts in 14
pcw inferred by DeepCycle transcriptional phase (based on the cell cycle score shown in I). (K-L) Scatter plot showing
pallial IPCs from 8.5 pcw (K) and 14 pcw (L) UMI counts per cell as a function of the DeepCycle transcriptional phase
show growth trend followed by a drop in RNA counts (UMI) that identifies mitosis (marked by M). Cells expressing
SOX2, NEUROD6, and BCL11B are marked in brown (left). Box plots quantifying SOX2, NEUROD6, and BCL11B
expression in each progenitor state (right). One-sided Mann-Whitney-Wilcoxon test with Bonferroni correction is used
to test the significance. (SOX2 for downregulation, BCL11B and NEUROD6 for upregulation) P-values: * <=0.05, **
<=0.01, *** <=0.001, **** <=0.0001. (M) Box plots showing SOX2, NEUROD6, and BCL11B expression in each
progenitor state for all the samples in the study divided by Chromium versions (v2 and v3). One-sided Mann-
Whitney-Wilcoxon test with Bonferroni correction is used to test the significance (SOX2 for downregulation, BCL11B
and NEUROD6 for upregulation). P-values: **** <=0.0001.(N) UMAP projection of all pallial nIPCs after removing cell
cycle effect by Harmony colored by progenitor state (left) and post conceptional age (right) (O) UMAP projection of
all pallial nIPCs after removing cell cycle effect by Harmony colored by clusters that were defined as neuron-like and
radial glia-like clusters in Fig. 3F and fig. S6P-Q (top) and by the expression of NEUROD6 and SOX2 (bottom) (P) Box
plots showing RNA counts for each progenitor state in neuron-like vs. radial glia-like nIPCs clusters in v2 chemistry
samples. (Q) Box plots showing the number of genes expressed for each progenitor state in neuron-like vs. radial
glia-like nIPCs clusters in v2 (left) and v3 chemistry samples (right). (R) Box plots showing UMI counts in
NEUROD6+/SOX2– vs. SOX2+/NEUROD6– (top) and BCL11B+/SOX2– vs. SOX2+/BCL11B– (bottom) IPCs in v2 (left)
and v3 (right) chemistry samples. One-sided Mann-Whitney-Wilcoxon test with Bonferroni correction is used to test
the significance. ns: not significant; P-values: * <=0.05, ** <=0.01, *** <=0.001, **** <=0.0001 (S-T) Expression
heatmaps of genes associated with radial glia-like (S) and neuron-like (T) IPCs in different progenitors state (left) and
in different classes (right). Mean normalized expression values for each gene were scaled between 0 and 1. Genes
associated with cell cycle were removed. Selected genes are presented, the full gene list in table S4. (U) Volcano plot
illustrating differential expression between radial glia-like and neuron-like IPCs. Significant differential expressed
genes are colored in red. Selected genes are annotated and colored black. LOG2(fold change) values were clipped to
10 and FDR values were clipped to 1e-100.
(previous page) Fig S7. Excitatory neurons lineage – early vs. late states (A) UMAP projection of all collected pallial
excitatory neuron lineage colored by Chromium versions (Left) and by the subsets used for differential gene
expression test (Right). (B) Differential abundance plot of Chromium version 3 samples from 11-14 p.c.w of the pallial
excitatory neuron lineage (top left) and Chromium version 2 samples from 6-10 p.c.w of the pallial excitatory neuron
lineage (bottom left). Each point represents a neighborhood, the size of the point is proportional to the number of
cells in the neighborhood. Neighborhoods are colored by their log-fold change in abundance between early and late
post-conceptional age. Neighborhoods showing significant enrichment (SpatialFDR<10%) are colored.
Neighborhoods with higher abundance of cells from early ages are colored blue and neighborhoods with higher
abundance of cells from later ages are colored red. UMAP projections colored by subset used for differential
expression test and major cell classes for Chromium version 3 samples from 11-14 p.c.w (top panel middle and right,
respectively), Chromium version 2 samples from 6-10 p.c.w (bottom panel middle and right, respectively) (C) Bar plot
showing the number of significant genes associated with early (top), late (middle) and more than one type of change
(“mixed”, bottom) in the major cell classes. The classes in which a significant change was observed are indicated by
asterisks below the bar. For clarity, vertical lines indicate where no significant change was found. (D) Expression
heatmaps of differential expressed genes in radial glia cells, IPCs, neuroblasts and neurons. The normalized mean
expression values for each neighborhood defined by milo were clipped to the 99th percentile to remove extreme
outliers and then scaled between 0 and 1 for each gene. The neighborhoods are ordered according to Log(fold
change) of the differential abundance between early and late states. Selected gene names are indicated. The full
genes list and the data used to create these heatmaps are in Table S5. (E) Gene ontology and pathway enrichment
analysis results. Showing the lowest adjusted p-value results. In case of duplicated terms only the one with the lower
FDR is shown. Axon related gene ontologies are colored purple, cell adhesion related GO are colored green,
transcription related terms are colored orange, extra cellular matrix related GO are colored red, and cell migration
related GO are colored blue. (F-G) UMAP projection of pallial radial glia (F) and pallial and subpallial cells (G) after
removing cell cycle effect colored by cell cycle phase (left) and post conceptional age (right).
Fig S8. Forebrain GABAergic neurons migration. (A) UMAP projections of DLX2+ GABAergic lineage cells from 12
p.c.w telencephlic samples (left). Cells are colored according to the expression of different ganglionic eminences
marker genes score. Selected interneuron types are marked. Cells from cortical and hippocamapal samples that were
clustered with DLX2+ cells are colored in brown (right). n.d., not determined. (B) UMAP projections of DLX2+
GABAergic lineage cells from 6 p.c.w samples. Cells are colored according to the expression of different ganglionic
eminences marker genes score (left). Selected interneuron types are marked. Cells from diencephalon, thalamic and
hypothalamic samples that were clustered with DLX2+ cells are colored in brown (right). Close-up images show
FOXG1 and FOXD1 expressing cells. (C) UMAP projections of DLX2+ GABAergic lineage from 8 p.c.w samples. Cells
are colored according to the expression of different ganglionic eminences marker genes score (left). Selected
interneuron types are marked. Cells from thalamic and hypothalamic samples that were clustered with DLX2+ lineage
cells are colored in brown (upper right). Cells are colored by FOXG1 expression (lower right). Arrows indicate specific
locations on the manifold with higher concentration of thalamic or hypothalamic cells and low expression of FOXG1.
(D) UMAP projections of DLX2+ GABAergic lineage from 14 p.c.w samples. Cells are colored according to the
expression of different ganglionic eminences marker genes score (left). Selected interneuron types are marked. Cells
from thalamic and hypothalamic samples that were clustered with DLX2+ lineage cells are colored in brown (upper
right). Cells are colored by FOXG1 expression (lower right).
Fig S9. Forebrain GABAergic neurons – cortical migration (A-G) UMAP projections of cells from 12 p.c.w forbrain
GABAergic lineage colored by (A) major cell classes (B) expression of selected patterning genes (C) expression of
radial glia marker genes (D) expression of IPCs marker genes (E) expression of MGE and derived neuronal types
marker genes (F) expression of CGE and derived neuronal types marker genes (G) expression of LGE and derived
neuronal types marker genes; only EMX1– cells are shown (H) UMAP projections of cells from different p.c.w colored
by GE marker genes score. Cells with GE scores below the threshold are colored in grey. (I) UMAP projection of cells
from different p.c.w colored by the indicated Subregion (brown). In case of several cortical dissections for a time
point, showing all the samples together and zoom in figures for each dissection.
Fig S10. DLX2+ cells in the thalamus and hypothalamus (A) UMAP projection of 6 p.c.w forbrain GABAergic lineage
cells colored by expression of selected genes (B) Violin plots showing expression of DLX2, FOXD1, FOXG1 and RGS16
in the cluster that expressed FOXD1 of 6 p.c.w samples in different subregions (right). UMAP showing the cluster cells
colored red (left). (C) Violin plots showing expression of DLX2, FOXD1, FOXG1, LHX6, and SOX6 in the cluster that
probably contain the daughter cells of the FOXD1+ progenitors of 6 p.c.w samples in different subregions (right).
UMAP showing the cluster cells colored red (left). (D) UMAP projection of cells from 8 p.c.w forbrain GABAergic
lineage colored by expression of selected genes (E) Violin plots showing expression of DLX2, FOXD1, FOXG1 and
CALB2 in the cluster that expressed CALB2 of 8 p.c.w lineage in different subregions (right). UMAP showing the
cluster cells colored red (left) (F) UMAP projection of cells from 14 p.c.w forbrain GABAergic lineage colored by major
cell classes. (G) UMAP projection of cells from 14 p.c.w forbrain GABAergic lineage colored by expression of selected
genes . (H) Violin plots showing expression of selected genes in the cluster that expressed CRABP1 and contains
thalamic cells of 14 p.c.w lineage in different subregions (right). UMAP showing the cluster cells coloured red (left). (I)
Violin plots showing expression of selected genes in the cluster that expressed PLXNA4 and LHX6 of 14 p.c.w lineage
in different subregions (right). UMAP showing the cluster cells coloured red (left). All the violin plots were scaled to
have the same width.
(previous page) Fig. S11 Ventral midbrain. (A) Region annotation, (B) Cell class, (C) Age (p.c.w.), and (D) Ventral
midbrain regional/cell annotation are shown on the tSNE embedding of Midbrain neural cells. (E) Cell cycling score,
(F) Cell class, (G) Age (p.c.w.) on the UMAP embedding of FOXA1+/FOXA2+ or TH+ cells and their neighbors. (H)
Accuracy and F1 score were evaluated for the parameter regularization strength (C) during logistic regression training
on cell types defined in La Manno et al., 2016 (right). The value of 0.37 was chosen. Precision, recall and F1 score was
evaluated with regularization strength (C=0.37) on 10% held-out test set (left). Cell types are colored according to the
legend below. (I-O) FOXA1+/FOXA2+ or TH+ cells and their neighbors are colored by their log2+1 gene expression of
(I) FOXA2, (J) LMX1A (K) NEUROG2, (L) NEUROG1, (M) NHLH1, (N) WNT1, (O) HEPACAM2 on the UMAP
embedding. Darker color indicates higher expression. Grey indicates no expression. (P) Random walk on the transition
matrix started from cells, annotated with the asterisk, in the encircled cluster (inset). The result of 100 steps with 15
nearest neighbors of the random walk is shown. (Q) Latent time obtained from scvelo’s velocity estimation with
‘mode = dynamical’. (R) Walks terminated in either encircled clusters (inset) were isolated as representative of the
EN1 (top) or PITX2 (bottom) lineages, respectively. Latent-time bins of cells along both lineages are shown. Heatmaps
of scaled expression of differentially expressed genes along latent-time bins for (S) PITX2 and (T) EN1 are shown.
Transcription factor genes are shown in red. Age (p.c.w.), cell class and cluster for each cell are indicated on the top
bars. (U) Heatmap of scaled expression of genes enriched in dopaminergic clusters is shown. Grey indicates no
expression. The left bar indicates cluster affiliation for each cell, as colored.
Fig. S12. Glioblasts. (A) Clusters shown on tSNE embedding. Numbers indicate cluster IDs (table S2). (B) tSNE colored
by age. (C) tSNE colored by cell cycle gene expression score, the total UMIs detected from known cell cycle genes, as
percentage of all UMIs. As in fig. S3A, we used 0.4% as threshold for cycling cells. (D-F) Expression of the indicated
genes shown on the tSNE embedding.
(previous page) Fig S13. Regionalization in OPCs and species comparison. (A) tSNE embeddings labeled by individual
dissected regions. (B) Embeddings showing age (top) and cycling cells (defined as cells that have >0.5% molecules
from cell cycle genes; bottom). (C) Volcano plots of differentially expressed genes in fore-, mid- and hindbrain of
OPCs (top) and COPs (bottom). Miniature tSNEs showing test groups within each region for the differential gene
expression analysis. Heatmap showing region-specific genes of human OPCs, cycling OPCs and COPs after
hierarchical clustering. (D) Embeddings displaying expression of OPC- and COP-specific genes (top three). (E)
Volcano plots of differentially expressed genes between OPCs and COPs in human and mouse respectively. Miniature
tSNEs show test groups (green: OPCs, plum: COPs). (F) Heatmap showing genes differentially expressed between
brain regions, in OPCs and COPs. (G) Heatmaps showing gene expression of shared genes in mouse and human
(right), human- (middle) and mouse-specific genes (right).

Supplementary Tables

Table S1. Samples metadata

Table S2. Clusters metadata
Table S3. Probe sequences, corresponding to Fig. 2
Table S4. Differential expression analysis results corresponding to fig. S6
Table S5. Differential expression analysis results corresponding to fig. S7
Table S6. Gene ontology analysis corresponding to fig. S7E
Table S7. Differential gene expression analysis between regions, corresponding to Fig. 6
Table S8. Differential gene expression analysis across species, corresponding to Fig. 6