Cell Systems, Volume
14
Supplemental information
A microwell platform for high-throughput
longitudinal phenotyping
and selective retrieval of organoids
Alexandra Sockell, Wing Wong, Scott Longwell, Thy Vu, Kasper Karlsson, Daniel
Mokhtari, Julia Schaepe, Yuan-Hung Lo, Vincent Cornelius, Calvin Kuo, David Van
Valen, Christina Curtis, and Polly M. Fordyce
Supplementary Materials for
A microwell platform for high
-
throughput longitudinal phenotyping and selective
retrieval of organoids
Alexandra Sockell
1,*
%
, Wing Wong
1,2,*
, Scott Longwell
3
, Thy Vu
4
, Kasper Karlsson
1,2
,
Daniel Mokhtari
5
, Julia
Schaepe
3
, Yuan
-
Hung Lo
2
, Vincent Cornelius
3
, Calvin Kuo
2
,
David Van Valen
6
, Christina Curtis
1,2,7
,8,9
,‡
, Polly M. Fordyce
1,3
,9,10
‡
1
Department of Genetics, Stanford University, Stanford, CA 94305
2
Department of Medicine, Stanford University,
Stanford, CA 94305
3
Department of Bioengineering, Stanford University, Stanford, CA 94305
4
Department of Biochemistry, UT Austin, Austin, TX 78712
5
Department of Biochemistry, Stanford University, Stanford, CA 94305
6
Division of Biology and Bioengineering,
California Institute of Technology, Pasadena, CA 91125
7
Department of
Biomedical Data Science
, Stanford University, Stanford, CA 94305
8
Stanford
Cancer Institute, Stanford
University, Stanford, CA 94305
9
Chan Zuckerberg Biohub, San Francisco, CA
94110
10
ChEM
-
H Institute, Stanford University, Stanford, CA 94305
*These authors contributed equally to this work
‡
Correspondence should be addressed to
pfordyce@stanford.edu
and Christina Curtis
cncurtis@stanford.edu
%
Current address: Pacific Biosciences, Menlo Park, CA94025
This
Supplementary Material
file includes:
Supplementary
Figs. S1 to S
13
Supplementary Figure S1
.
Comparison of microwell imaging versus bulk platform imaging
.
Images of organoids taken from bulk culture (Incucyte, Sartorius Inc.) (left) versus our microwell
platform (right). In bulk culture (left), organoids are embedded within the full height of the
3D ECM
scaffolds and single z
-
plane imaging cannot return in
-
focus images for most organoids; red arrows
indicate out
-
of
-
focus organoids. By contrast, the microwell platform ensures that organoids are
seeded within a single z
-
plane for imaging (right).
Supplementary Figure S
2.
Microwell supports the growth of different organoid types.
Brightfield and fluorescence images of breast (top) and colon (bottom) organoids grown in
microwells. Images were taken 7
-
day post
-
seeding. Breast
organoids expressed CK8 which is a
breast luminal epithelial cell marker. Colon organoids expressed Vilin1, which is characteristic of
the intestinal epithelium.
Supplementary Fig
ure
S
3
.
PDMS
microwell array fabrication
.
Microwell arrays of pr
e
-
defined
dimensions (100 x 100 x 80 or 200 x 200 x 80
μm
) were designed in AutoCAD
and
printed on
transparencies
that were used to fabricate silicon wafer
molding masters via standard
photolithography.
Microwell d
evices were
created
by spin
-
coating
PDMS onto molding masters
and cutting out individual microwell arrays. After fabrication, devices were exposed
to oxygen
plasma
to render surfaces hydrophilic
and then inserted into
the
bottom of
individual ‘macrowells’
within
stand
ard 12 well tissue culture plates.
Supplementary Figure S
4
. PDMS spin curve.
Relationship between
final device thickness and
PDMS spin speed
. S
pin coating at 200 rpm yielded 0.5 mm thick devices used in
downstream
experiments
.
Supplementary
Fig
ure
S
5
.
I
mage analysis
pipeline
.
(1)
Devices are imaged via a tiled
acquisition at each timepoint
and then
stitched to
form a single image of the entire device at each
timepoint, (
2
)
stitched
images
are rotated to
align
device
microwells
perpendicular to
the borders
of each image, (3)
rotated images are subdivided by
subarrays,
and
(4)
subarray images are
further subdivid
ed by microwells
. I
ndividual microwell images are extracted from the subarrays,
and images over time are collated for each microwell
. Processing is done for all timepoints in
parallel and downstream image processing allows identification of
microwells containing cells
.
Supplementary Figure S
6
. Microwell platform scale and organoid quality.
Stitched brightfield
image from a single “macrowell” of a 12
-
well culture plate
containing
an array of 100 μm x 100
μm microwells
(
A
)
, a “subarray” from that macrowell
(
B
)
, and a close
-
up of 9 microwells from that
subarray
(
C
)
.
(D)
Organoids seeded in microwell with ECM/Matrigel overlay exhibit morphology
like
those grown in bulk.
Supplementary Figu
re S7.
Single
-
cell transcriptomic comparison of organoids grown in
microwell versus bulk.
(A)
UMAP projection of
single
-
cell
transcriptomes of organoids grown in
bulk ECM and those grown in microwells. Differences in global transcriptomic profiles were
observed
(raw data deposited at dbGAP with accession number phs003315.v1.p1)
.
(B)
Cell type
composition of organoids grown in b
ulk ECM and in microwells. The organoids grown in microwell
recapitulate the composition and relative cell type abundance of organoids grown in bulk ECM.
(C)
Gene ontology analysis using differentially expressed genes of organoids grown in microwell
versus
organoids grown in bulk. Cell cycle genes and cellular proliferation were affected in
organoids grown in microwell. On the other hand, these organoids upregulated genes that
associated with various cellular metabolic processes.
UM
AP1
U
M
A
P
2
0
5
10
6
8
10
12
0
25
50
75
100
P
e
r
c
e
n
t
a
g
e
MUC5A
C
Epithelial C
ells
TFF2
P
it C
ells
PGC
Chief C
ells
GKN2
M
uc
osal C
ells
L
GR5
S
t
em C
ells
Bulk
M
icr
o
w
ell
Bulk
M
icr
o
w
ell
A
B
G
O
T
e
r
m
s
L
o
g2 F
old Enrichmen
t
M
icr
o
w
ell v
ersus Bulk
Upr
egula
t
ed
in M
icr
o
w
ell
Upr
egula
t
ed
in Bulk
C
Supplementary Figure S
8
.
Cell loading statistics.
Number of microwells containing 0, 1, 2, 3,
or 4 cells post
-
loading for
100 μm x 100 μm
(A)
and 200 μm x 200 μm
(B)
microwell array
s; green
line and markers indicates best fit Poisson
with annotated
fit parameter.
Supplementary
Fig
ure
S
9
.
DeepCell training and testing.
(
A
)
Per
-
experiment precision, recall,
and F1 scores for DeepCell
performance across Experiments #1
-
3
;
s
cores were highest for
Experiment
1
, a
s expected given that data from this experiment were
used to train the model
.
(
B
)
Correlations between per
-
microwell DeepCell
-
predicted cell counts
vs.
manually labeled
‘ground
truth’
cell counts across
Experiments #1
-
3;
dashed red line indicates linear regression with
annotated Pearson
correlation coefficients.
Supplementary Fig. S10. Additional DeepCell
-
enabled
phenotypic characterization of organoids grown in
experiments #1
-
#3. (A)
Number of microwells containing 0
-
6 cells after loading for
Δ
P53 (top) and DKO (bottom) cell
lines; green line and markers indicate best fit Poisson with
indicated fit parameter.
(B)
Density plots showing distance
moved by single cells prior to first division for
T
P53
KO
(green) and DKO (purple) cell lines; dashed li
nes indicate
population median.
(C)
Distance moved
vs.
time to first
division for single
T
P53
KO
(green) and DKO (purple) cells
loaded within microwells; red line indicates linear regression
with annotated Pearson correlation coefficient.
(D)
Distance
moved
vs.
calculated growth rate for single
T
P53 (green) and
DKO (purple) cells loaded within microwells; red line
indicates linear regression with annotated Pearson
correlation coefficient.
(E)
Boxplots showing distribution of
distances moved by all single cel
ls at each time point prior to
the first cell division.
(F)
Relationship between growth rate
and physical distance of the organoid relative to the microwell
wall. Red line shows linear regression; annotation shows
Pearson correlation coefficient.