Supplementary Materials for
Whole-animal multiplexed single-cell RNA-seq reveals transcriptional shifts
across
Clytia
medusa cell types
Tara Chari, Brandon Weissbourd, Jase Gehring, Anna Ferraioli, Lucas Leclère,
Makenna Herl, Fan Gao, Sandra Chevalier, Richard R. Copley*, Evelyn Houliston*,
David J. Anderson*, Lior Pachter*
*Corresponding author. Email: copley@obs-vlfr.fr (R.R.C.); houliston@obs-vlfr.fr (E.H.);
wuwei@caltech.edu (D.J.A.); lpachter@caltech.edu (L.P.)
Published 26 November 2021,
Sci. Adv.
7
, eabh1683 (2021)
DOI: 10.1126/sciadv.abh1683
The PDF file includes:
Figs. S1 to S22
Tables S1 to S5
Supplementary Methods
Legends for tables S3 and S5
Other Supplementary Material for this manuscript includes the following:
Tables S3 and S5
Figures
Fig. S1. Cell
Type Limit of Detection for
Starvation Experiment.
Cumulative distribution
function displaying likelihood of detecting at least 10 cells from rare cell types which make
up 0.017 of the total population of cells, assuming ~50 cell types, over a possible range of
number of cells sampled in the given sequencing experiment.
The average number of cells
per jellyfish used in the starvation experiment is denoted by the circle. Plot generated using
the Satija Lab howmanycells calculator
.
Fig. S2. Fold Change in Cells per
Organism Under
Starvation.
Hemocytometer cell
counts
for jellyfish in control and starved conditions
(n = 4 per condition), with two
measurements per animal (cells for each sample counted in two 16 square grids; see
Methods). Box-and-whisker plot displayed, with mean line shown and whiskers denoting 1.5
IQR.
Fig. S3. Egg Cell Pr
oduction Under
Starvation.
Number
of egg cells released in the daily
spawning cycle from 5 control and 5 starved medusae on day four of starvation.
Animals
used are not the same as the individuals in the sequencing experiment, however the same
protocol as in Methods was used to conduct the starvation. Error bars denote standard
deviation.
Fig. S4. Cell Bar
code Filtering fr
om ClickT
ag Clustering. a)
Knee-plot for ranked cell
barcodes (ranked according to number of UMIs) versus ClickT
ag UMI counts per cell
barcode. Line denotes filtering/selection of the top 50,000 cells.
b)
Heatmap of counts for
ClickT
ag barcodes associated with cell barcodes. Pairs of barcodes correspond to the
ClickT
ags added to each or
ganism's samples (T
able S2).
c)
Louvain clustering of ClickT
ags
based on counts across cells, corresponding to the 10 labeled or
ganisms.
d)
Log counts of
ClickT
ags. High count clusters correspond to cells of the 10 dual-barcoded animals.
e)
ClickT
ag counts per cell in control (fed) versus starved cells, subsetted to oocytes. * denotes
p
-value <0.05 from a one-sided t-test, with alternative
hypothesis Control ClickT
ag Counts >
Starved ClickT
ag Counts.
Fig. S5. Distribution of Cell Counts between Starvation Conditions. a)
UMI counts shown on
log scale across cells on the UMAP
embedding.
b)
One-way
ANOV
A on log cell counts within
each cell type (per each of the 5 individuals).
Whiskers denote 1.5 IQR and diamonds denote
outlier points. Pval obtained with an F test. * =
p
-value < 0.05, adjusted with Benjamini-Hochber
g
for FDR correction. Early nematoblasts (cluster 1
1) show significant dif
ferences in numbers
between individuals in starved vs control conditions.
Fig. S6. Concordance of Cell
Types Between
Two Distinct Multiplexed Experiments. a)
Starvation data (training set) used to select nearest neighbors from left-out starvation data and
all the stimulation data. For each selection, the fraction of nearest neighbors that had the same
cell type label is reported (see Methods).
b)
Top
100 markers from the stimulation
experiment with cluster labels applied from K-Nearest Neighbors (see Methods) compared to
top 100 markers per cell type from starvation clustered data.
c)
UMAP/P
AGA embedding of
stimulation data with applied cell type labels.
d)
Heatmap of markers from Fig. 2b shown for
stimulation data.
e)
Inter- and intra-cluster distances
shown for control (SW) versus perturbed
cells for each perturbation (see Methods).
The cell type with lar
gest internal distance (but
non-overlapping) is 20 (tentacle epidermis).
Fig. S7. Cell-T
ype Specific Gene Expr
ession Changes Following KCl and DI
Treatment.
a)
Diagram of the ‘Stimulation’
experiment. Four biological
replicates (animals) used for
each condition. SW
denotes seawater (control), DI denotes deionized water addition, and KCl
denotes potassium chloride addition (see Methods).
30 minutes following the last stimulation,
animals were dissociated and fixed in methanol.
b)
Summary table of numbers of up- and
down-regulated genes in each cluster (cell type). *Denotes highest number of DE genes in 5,
terminal dif
ferentiating nematocytes.
c)
Volcano plot
of
p
-value and fold-change from
DeSeq2 analysis (see Methods) for DE gene candidates. Dashed line denotes 0.05 alpha
cutoff applied during selection of DE genes (see Methods). Colors indicate the number of cell
types a gene is found to be dif
ferentially expressed in. Genes appear multiple times for each
cell type (if applicable).
d)
Fold-change per condition
across all cells for global (DE in many
cell types), 'IEG'-like candidates.
e)
qPCR for DE
gene candidates shown in panel d in both
conditions (see Methods).
f)
Volcano plot for selection
of upregulated DE genes in Neural
Cells-9 (cluster 9), which consists of the majority of neural cells, colored by perturbation
condition. Gene names denote selected candidates.
g)
Expression for cells in each condition
of upregulated DE genes found in Neural Cells-9 (cluster 9) using the non-parametric,
Wilcoxon test.
P
-values adjusted for multiple testing
with Benjamini-Hochber
g correction. *
denotes
p
-value < 0.01, ** denotes
p
-value < 0.001.
Fig. S8. Distances between Contr
ol Cells in each Multiplexed Experiment.
Histograms of
average pairwise
L
1
distances between cell types of
control animals within (red and blue) and
across (yellow) experiments for visualization of batch ef
fect (see Methods).
Fig. S9. High Resolution Clustering
Annotation of Cell
Atlas fr
om Starvation Dataset.
Cell Atlas with all 36 clusters/cell types labeled with UMAP/P
AGA embedding from the
starvation experiment data (see Methods).
Fig. S10. Marker
Gene Overlaps for
Cell Types in Cell Ranger
versus kallisto Read
Quantifications.
Pairwise overlap of the top 100 marker
genes from the original 36 Cell
Ranger clusters applied to kallisto-bustools processed starvation data versus the top 100
markers from the original Cell Ranger clustered data.
The coloring reflects the Jaccard Index
of each top marker gene set (number of genes in the intersection divided by the number of
genes in the union).
Fig. S1
1. Examples of Diagnostic Genes for
Assigning Cell
Type Identities.
Overview of cluster identities deduced from markers
including the key diagnostic genes
listed.
These include
Clytia
genes whose expression
in the medusa has been previously
characterised (34-36, 39, 58) and genes with clear homology to genes with clear cell type
specificity in other animals (Scyp1= XLOC_008881, ortholog to
Hydra
Scyp1/
Sc4wPfr_899.g22607 (8); Sterovillus components Harmonin/USH-IC =XLOC_003773,
Whirlin=XLOC_
011922
and Sans/USH-1G=XLOC_039341 (43).
For all other genes cited,
in situ
localization profiles, UMAP
embedding and
XLOC identifiers are shown in Fig. 2C ,
Fig. 3, Fig. S14, Fig. S18, and
Table S3.
a
Clusters 2 and 13 have overlapping profiles including
known oocyte specific genes such as GFP
2 (XLOC_004150) and
ones highly expressed in published
Clytia
“Growing”
and “Fully grown“ oocyte transcriptomes (38) .
b
Nematogenis occurs in two distinct phases with distinct
transcriptional profiles expressing genes associated either with the
nematocyst (Clusters 12, 1
1 and 23) or with nematocil formation (Clusters 10, 5 and 21) . Cluster 17 includes two cell
subpopulations preferentially expressing either nematocyst components such as Minicollagen3/4 (XLOC_004102) (36) or
structural genes associated with the mechanosensory cnidocil. It is thus interpreted as a transition phase between distinct
phases of nematogenesis.
c
Details of neuropeptide precursors in
Table S4.
Fig. S12. Cell Counts per
Individual
Across Cell
Types. a)
Cell count for each of the 36
cell types and each or
ganism.
b)
Linear log-log relationship
between variance in cluster size
(cell count) across individuals and average size (cell count), for all 36 cell types.