Report

Modeling Patient-Derived Glioblastoma with

Cerebral Organoids
Graphical Abstract Authors
Amanda Linkous,
Demosthenes Balamatsias,
Matija Snuderl, ..., David Pisapia,
Conor Liston, Howard A. Fine

Correspondence
[email protected]

In Brief
To address limitations with current
preclinical glioblastoma (GBM) models,
Linkous et al. establish a ‘‘GLICO’’
(cerebral organoid glioma) model to retro-
engineer patient-specific GBMs using
patient-derived glioma stem cells and
human cerebral organoids. Resulting
tumors closely phenocopy patient GBMs
and are supported by tumor microtubes
that promote invasion into host tissue.

Highlights
d Glioma stem cells home toward, invade, and proliferate in
human cerebral organoids

d GLICO (cerebral organoid glioma) tumors phenocopy patient

glioblastoma

d Tumor microtubes promote the invasion of GLICO tumors

into normal host tissue

d The GLICO model is a tool to study GBM biology in a human

brain environment

Linkous et al., 2019, Cell Reports 26, 3203–3211

March 19, 2019 ª 2019 The Author(s).
https://doi.org/10.1016/j.celrep.2019.02.063
Cell Reports

Report

Modeling Patient-Derived Glioblastoma

with Cerebral Organoids
Amanda Linkous,1 Demosthenes Balamatsias,1 Matija Snuderl,2 Lincoln Edwards,1 Ken Miyaguchi,1 Teresa Milner,3
Batsheva Reich,3 Leona Cohen-Gould,4 Andrew Storaska,1 Yasumi Nakayama,1 Emily Schenkein,1 Richa Singhania,1
Stefano Cirigliano,1 Tarig Magdeldin,1 Ying Lin,1 Gouri Nanjangud,5 Kalyani Chadalavada,5 David Pisapia,6 Conor Liston,3
and Howard A. Fine7,8,*
1Meyer Cancer Center, Weill Cornell Medicine, New York, NY, USA
2Division of Neuropathology, Department of Pathology, NYU Langone Medical Center and Medical School, New York, NY, USA
3Feil Family Brain and Mind Research Institute, Weill Cornell Medicine, New York, NY, USA
4Department of Biochemistry, Weill Cornell Medicine, New York, NY, USA
5Memorial Sloan Kettering Cancer Center Molecular Cytogenetics Core, New York, NY, USA
6Department of Pathology & Laboratory Medicine, NewYork-Presbyterian Hospital/Weill Cornell Medicine, New York, NY, USA
7Meyer Cancer Center, Division of Neuro-Oncology, Department of Neurology, NewYork-Presbyterian Hospital/Weill Cornell Medicine,

New York, NY, USA

8Lead Contact

*Correspondence: [email protected]
https://doi.org/10.1016/j.celrep.2019.02.063

SUMMARY generate emergent properties specific to their host (Lu et al.,

2012; Quail and Joyce, 2017; Rubenstein and Kaufman, 2008).
The prognosis of patients with glioblastoma (GBM) Three-dimensional (3D) models of cancers, including tumor or-
remains dismal, with a median survival of approxi- ganoids (Shroyer, 2016), patient-derived xenografts (PDXs)
mately 15 months. Current preclinical GBM models (Tentler et al., 2012), and genetic mouse models (Holland,
are limited by the lack of a ‘‘normal’’ human microen- 2001), all address some aspects of GBM but are fundamentally
vironment and the inability of many tumor cell lines limited. Thus, the ability to comprehensively understand and
therapeutically manipulate complex cancer phenotypes requires
to accurately reproduce GBM biology. To address
the creation of clinically relevant models that embrace that
these limitations, we have established a model sys-
complexity yet retain amenability to detailed analysis. We
tem whereby we can retro-engineer patient-specific contend that an ex vivo model that is experimentally manipu-
GBMs using patient-derived glioma stem cells lable, biologically and clinically relevant, logistically pragmatic,
(GSCs) and human embryonic stem cell (hESC)- and scientifically rigorous would enhance the study of GBM as
derived cerebral organoids. Our cerebral organoid a human disease and allow patient-specific high-throughput
glioma (GLICO) model shows that GSCs home to- drug and therapeutic screening.
ward the human cerebral organoid and deeply invade Over the years, evidence generated by our laboratory and
and proliferate within the host tissue, forming tumors others has shown that patient-derived glioma stem cells
that closely phenocopy patient GBMs. Furthermore, (GSCs) are the most biologically and phenotypically relevant
cerebral organoid tumors form rapidly and are sup- cells to the parental tumor in patients (Baysan et al., 2012,
2014; Lee et al., 2006, 2008; Son et al., 2009). Defined by the
ported by an interconnected network of tumor micro-
capacity for self-renewal and multi-lineage differentiation, this
tubes that aids in the invasion of normal host tissue.
subpopulation of cells is imperative for tumor initiation, mainte-
Our GLICO model provides a system for modeling nance, and invasion in vivo. Moreover, GSCs exhibit increased
primary human GBM ex vivo and for high-throughput resistance to ionizing radiation and cytotoxic drugs, implicating
drug screening. their role in treatment resistance of GBM (Bao et al., 2006).
Thus, GSCs have become the accepted standard for studying
INTRODUCTION GBM biology. It is important to acknowledge, however, that
in vivo GSCs are not cell autonomous but rather heavily influ-
Glioblastoma multiforme is the most lethal primary brain tumor in enced by tumor host-cell interactions and the ability to grow
adults. Median survival rates have remained largely unchanged within a 3D extracellular matrix (Lu et al., 2012; Quail and Joyce,
for 30 years. With a 5-year survival rate of less than 5% (Stupp 2017; Rubenstein and Kaufman, 2008). Unknown interspecies
et al., 2005), new strategic approaches to the study and treat- differences between human GSCs and murine brain cells
ment of this disease are clearly needed. coupled with variably long tumor latencies, lack of real-time
Glioblastoma multiforme (GBM) research and drug develop- imaging and genetic manipulation, and various ethical issues
ment rely on cell autonomous in vitro or lengthy, expensive in vivo represent limitations of GSC-derived xenografts (Anisimov
models that poorly recapitulate human disease. Tumors are not et al., 2005; Dragunow, 2008; Gould et al., 2015; Huszthy
cell autonomous but rather inherently complex systems that et al., 2012; Hutchinson and Kirk, 2011; John Lin et al., 2017;

Cell Reports 26, 3203–3211, March 19, 2019 ª 2019 The Author(s). 3203
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Jucker, 2010; Perrin, 2014). In contrast, although tumor organo- Cerebral Organoids Exhibit Stage-Specific Neural
ids overcome some of the logistical constraints of animal Development and Demonstrate Myelinated Axons,
models, they fail to address the critical issue of tumor-normal Dendrodendritic Synapses, Neurons, and Glia
host tissue microenvironment interactions. Because molecular markers are not always cell type specific, we
Although it would be theoretically possible to co-culture performed high-resolution transmission electron microscopy
various brain cells with GSCs in two dimension (2D) to address (TEM) to confirm the presence of specific cell types and neuroan-
host-tumor cell interaction, the resulting disorganized mixture atomical features within the cerebral organoids. As shown in Fig-
of normal cells would lack the human extracellular matrix and ure S2B, cerebral organoids contain both glial cells and neurons,
would not remotely resemble the 3D, carefully organized with a number of the neuronal axons displaying myelination and
arrangement of cells in the human brain. This, however, is exactly the presence of dendrodendritic synapses (black arrows). In
what cerebral organoids achieve. Thus, using hESC-derived ce- contrast to the more common axodendritic synapse, a dendro-
rebral organoids and patient-derived GSCs, we demonstrate a dendritic synapse occurs between the dendrites of two neurons
powerful tool for modeling human GBM within a primitive, human and can mediate bi-directional signaling (Masurkar and Chen,
brain microenvironment. 2012). Interestingly, these synapses are among the most well
characterized microcircuits within the brain.
RESULTS
Patient-Derived GSCs Form Infiltrative Tumors in
Morphological and Immunohistochemical Analysis Cerebral Organoids
Reveals Neural Stem Cells and Lineage-Specific To accurately model human GBM in cerebral organoids, we co-
Differentiation in Cerebral Organoids cultured GFP-labeled GSCs with individual, fully formed cerebral
On the basis of recent work by Lancaster and Knoblich (Lancas- organoids for 24 h. Tumor take rate was 100% for all GSC lines,
ter et al., 2013), others and we are now able to create a realistic and tumor-infiltrated organoids were monitored daily by immu-
in vitro model of the developing human brain. By culturing human nofluorescence microscopy for evidence of tumor formation.
embryonic stem cells (hESCs) or induced pluripotent stem cells As shown in Figure 1A, considerable tumor growth was detected
(iPSCs) in conditions that promote 3D expansion of neuroecto- 1 week after co-culture. Subsequent neuropathological evalua-
derm, a cerebral organoid or ‘‘miniature brain’’ forms. Our cere- tion of tumor-bearing organoids revealed a hypercellular bulk tu-
bral organoids display a primitive ventricular system, multiple mor with an infiltrating edge of GSCs that were invading the
neural rosettes with a proliferative zone of neural stem cells normal tissue, thus recapitulating tumor morphology observed
(NSCs), and a differentiated choroid plexus (Figure S1A). Cere- in human patient GBMs. In addition, immunofluorescence stain-
bral organoids are positive for NSC markers including Nestin, ing for GFP revealed extensive infiltration of GFP-positive tumor
Musashi-1, and Sox2 (Figure S1C). In addition, Pax6, a key tran- cells in 923 cerebral organoid gliomas (GLICOs), with more than
scription factor essential for NSC proliferation and neurogenesis 20% of these tumor cells staining positive for the proliferative
(Sansom et al., 2009), is found extensively within the cerebral or- marker Ki67 (Figure 1D). To further confirm that a portion of the
ganoids (Figure S2A). Expression of these proteins is consistent glioma cells within the GLICO were actively proliferating, we
with neuroectoderm cell fate determination, progenitors of the pulsed 923 GLICOs for 48 h with the modified thymidine analog
ventricular zone (VZ) and forebrain identity. Moreover, co-stain- EdU (5-ethynyl-20 -deoxyuridine). We then performed co-immu-
ing with TBR2, a marker of intermediate progenitors within the nofluorescence staining to identify proliferating tumor cells as
subventricular zone (SVZ) of the human brain (Lancaster et al., evident by dual positivity for Edu and GFP (Figure S4A).
2013), shows distinct populations of SVZ-associated (TBR2+) In addition to the positive Ki67 and EdU staining, we also quan-
and VZ-associated (Pax6+) cells (Figure S2A). titatively assessed GSC proliferation using two additional
Providing additional evidence of regulated development, we methods. First, we plated 20,000 or 200,000 GFP-expressing
show that the lumens of the neural rosettes are positive for the 827 GSCs in triplicate and counted the number of live cells at
apical tight junction protein, N-cadherin (Figure S2A). Luminal day 8 and day 18. Both groups exhibited proliferation at days 8
accumulation of N-cadherin-positive cells suggests that these and 18; live cell number from 20,000 = 50,000 at day 8 and
rosettes display the same type of polarity found within the neural 760,000 at day 18, while live cell number from 200,000 =
tube of the embryonic neural plate (Elkabetz et al., 2008). These 1,212,000 at day 8 and 2,080,000 at day 18 (Figure S4B). We
morphological and immunohistochemical findings suggest that then co-cultured 20,000 or 200,000 GFP-expressing 827 GSCs
the level of differentiation and structural development in our ce- with H1-derived cerebral organoids and performed subsequent
rebral organoids approximate the developmental stage of a 20- qRT-PCR (qPCR) for GFP expression at days 8 and 18. Fig-
week-old human fetal brain (Figure S1B). In addition, temporal ure S4C demonstrates increasing GFP expression at days 8
evolution studies of the cerebral organoids reveal decreased and 18 for GLICO tumors originating from both 20,000 and
neurogenesis and increased gliogenesis over time, with 200,000 GSCs, but GFP expression was significantly higher at
enhanced expression of glial fibrillary acidic protein (GFAP) as day 8 between GLICOs formed from 200,000 GSCs compared
the organoids continue to age, much like the normal mammalian with GLICOs from 20,000 GSCs (p = 0.0005). Day 18 GFP expres-
brain (Figures 4A and S2C). Recent data from Sloan et al. (2017) sion in GLICOs from 200,000 GSCs was also significantly greater
corroborates our findings by demonstrating that gene signatures than GFP expression from the same group at day 8 (p = 0.0094),
in cortical cerebral organoids switch from fetal astrocyte genes indicating that GLICO tumors are proliferating over time, as
to mature astrocyte genes as the organoids age over time. demonstrated by the Ki67 expression and EdU incorporation.

3204 Cell Reports 26, 3203–3211, March 19, 2019

 Figure 1. Patient-Derived GSCs Form Infil-
trative Tumors in Cerebral Organoids
(A) Tumor formation from GFP-expressing 827
GSCs (left) 1 week after co-culture; scale bar,
400 mm; n = 6 organoids per cell line.
(B) Historical development of patient-derived
GLICOs from six patient tumor specimens.
(C) Diffuse GSC infiltration (blue arrow indicates
tumor cells) of the organoid as evident by
morphological (H&E) analysis of control organoids
(top) and GFP-expressing 923 GLICO tumors
(bottom) at day 14; scale bars, 1,000 mm (left) and
100 mm (right); six organoids were analyzed per
group. Tumors exhibit necrosis (black arrow).
(D) Representative image and bar graph showing
quantification of Ki67+/GFP+ cells in GFP+ 923
GLICO tumors after 2 weeks of growth; green,
GFP; red, Ki67; blue, DAPI; n = 3 organoids; scale
bar, 50 mm.

in 2D growth and display a consistent and

similar pattern of invasion that is charac-
terized by tightly woven honeycomb-like
networks of tumor cells (Figure 1B). Last,
0517 and 0607 GSCs are defined by
slow growth rates compared with the
other GSCs, and their respective GLICO
tumors are often characterized by smaller
regional nodes of proliferation, sur-
rounded by distinct areas of limited infil-
tration. Such strikingly different patterns
and degrees of tumor cell invasion
and proliferation between the different
GSC lines is consistently reflected in the
growth of these different GSC lines in
the GLICO model and reflects the hetero-
geneity of the invasive phenotypes that
are clinically observed in GBM patients.
3D topographical analyses and optical
sections of GLICO z stacks further
illustrate the various levels of tumor infil-
tration and proliferation throughout the
Two-photon microscopy of GFP-labeled GSC-bearing orga- entire structural landscape of the organoid (Figures S3B and
noids revealed distinct areas of tumor bulk adjacent to normal S3C). These data were supported by light sheet microscopic
cerebral tissue diffusely infiltrated by glioma cells in a pattern studies in which the resulting 3D tumor volume rendering shows
highly similar to that seen in surgical and autopsy specimens a definitive GFP-positive 923 tumor mass at 14 days post-co-
from GBM patients (Figures 1B and S3A). In order to determine culture (Figure S3D; pseudo-color, red).
if this heterogeneity of glioma growth and invasion within the
GLICO model is a consistent phenotype intrinsic to that partic- GLICO Tumors Are Detectable by Luciferase-Based
ular patient-derived GSC line and the GLICO model or just a Imaging
random event, we interrogated the growth patterns of six newly In order to develop a rapid and quantitative method for repeated
generated patient-derived GSC lines in the GLICO (Figure 1B). in vitro imaging and quantitation of tumor mass in real time
Using six to ten GLICO tumors per cell line, we observed remark- without the need to disrupt the cerebral organoid, we generated
ably consistent GSC line-specific biological behavior in the GSCs that stably express either firefly luciferase or a secreted
GLICO model. For example, 0728 and 1206, two of our most Cypridina luciferase and co-cultured these cells with cerebral
aggressive cell lines in vitro, exhibited a highly diffuse pattern organoids to allow tumor formation. The tumor bioluminescence
of invasion that extended to all areas of the cerebral organoid. signal from firefly luciferase-expressing 923 GLICOs was
IN contrast, 0320 and 0810 often form rather large neurospheres easily detected in a multi-well plate within 1 min of exposure

Cell Reports 26, 3203–3211, March 19, 2019 3205

 Figure 2. GLICO Tumor Growth Is Sup-
ported by a Network of Tumor Microtubes
(A) Formation of 923 GSC-derived tumor micro-
tubes in vitro after 12 days in 2D culture (arrows);
scale bar, 200 mm (top) and 100 mm (bottom).
(B) Two-photon microscopy of tumor microtubes
(arrows) in GLICO tumor from GFP-positive 923
GSCs; day 10; 3D rendering of GLICO tumor, top;
individual z stack images of microtubes, bottom;
scale bar, 80 mm; n = 3 organoids.
(C) Electron micrographs of peroxidase-based
immunolabeling of connexin-43-positive gap
junctions (top) and desmosomes (bottom) in GFP-
positive 923 GLICO tumors; scale bars, 500 nm
(top) and 100 nm (bottom); n = 5 organoids.
(D) Electron micrographs showing tube-like con-
nections between 923 GLICO tumor cells; arrows
indicate peroxidase labeling for connexin-43; scale
bars, 2 mm (left) and 500 nm (right).
(E) Cytoplasmic fusion between two 0728 GLICO
tumor cells (left) and a cytoplasmic fusion between
a neuron and 0728 GLICO tumor cell (right); scale
bars, 500 nm (left) and 2 mm (right); m, mitochon-
dria; N, nucleus; N1, neuronal nucleus; N2, 0728
tumor cell nucleus; n = 5 organoids.
(F) Time-lapse image of calcium waves (Rhod-
2AM) traveling along 923 GSC microtubes; scale
bar, 50 mm.
(G) Two-photon microscopy of tumor microtube
(arrows) in GLICO tumor from GFP- and RFP-
positive 827 GSCs; day 10; n = 3 organoids; scale
bar, 80 mm.

trol organoid tissue, invasion of the GSCs

induced abundant necrosis in cerebral or-
ganoids, in part because of the large num-
ber of proliferating tumor cells (Figure 1C).

GLICO Tumor Growth Is Supported

by a Network of Tumor Microtubes
We further sought to elucidate whether
GSC growth in organoids resembles
human glioma growth in vivo.
Osswald et al. (2015) recently reported
(Figure S3E). The amount of activity in the GLICO tumor imaged that GBMs in situ possess a network of gap junction mediated-
in real time was directly proportional to the number of 923 GSC interconnecting microtubes. This network of tumor microtubes
cells that were initially co-cultured with the cerebral organoid provides potential routes for invasion and proliferation, thus facil-
(Figure S3E). In 827 GSCs that expressed the secreted form of itating multicellular communication between tumor cells. Using
luciferase, luciferase activity in the medium increased over two-photon microscopy, we observed a similar network of mi-
time and was directly proportional to the number of GSCs that crotubes in 923 and 827 GSC-formed GBMs growing in cerebral
were growing in the GLICO tumors (Figure S3F). organoids (Figures 2A, 2B, and 2G; Videos S1, S2, and S3).
These microtubes penetrate deeply into the normal brain and
GSC-Derived GLICO Tumors Recapitulate Features of provide multicellular connections among various tumor subpop-
the Human Disease ulations in the organoid. Electron micrographs of peroxidase-
To further highlight the aggressiveness of GSC-derived tumors in based immunolabeling of connexin-43 in GLICO tumors
cerebral organoids, we evaluated the histopathology of tumor- revealed not only connexin-43-positive gap junctions (Figure 2C,
bearing organoids whose tumors had been growing for approxi- top) but also elevated cytoplasmic labeling for connexin-43
mately 2 weeks. Figure 1C demonstrates marked destruction of at various regions of connectivity within the tumor micro-
cerebral organoid tissue by 923 GLICOs, closely mirroring the hu- tubes themselves (Figure 2D). Furthermore, time-lapse im-
man disease pathology. In contrast to the completely viable con- aging demonstrated that these tumor microtubes effectively

3206 Cell Reports 26, 3203–3211, March 19, 2019

 Figure 3. GLICO Tumors Exhibit Differential
Response to Therapy and Preserve Key Ge-
netic Changes of Parental Tumor
(A) Representative fluorescent microscopic im-
ages (left) and quantification of secreted luciferase
activity (right) for 827 and 923 GLICOs treated with
temozolomide (TMZ) or bis-chloroethylnitrosourea
(BCNU); n = 4–6 organoids per group; **p < 0.01
and ***p < 0.001; scale bar, 500 mm.
(B) Representative fluorescent microscopic im-
ages (left) and quantification of secreted luciferase
activity (right) for 923 GLICOs treated with a single
dose of 0, 5, or 10 Gy of ionizing radiation; n = 4–6
organoids per group; scale bar, 500 mm.
(C) Quantification of EGFR copy number variation
(left) in 2D or GLICO samples from patients with
EGFR amplification; ****p < 0.0001. Gender-mis-
matched 0607 GLICOs or 0810 GLICOS were
generated; 0607 (from a female patient) was co-
cultured with H1 (from a male embryo)-derived
organoids, and 0810 (from a male patient) was co-
cultured with H9 (from a female embryo)-derived
organoids. Representative images of DNA FISH for
EGFR/Cen7/ChrY (Y chromosome) in GLICOs are
shown (right); red, EGFR; orange, Cen7; green,
ChrY; white arrow indicates tumor cell; n = 5 or-
ganoid sections per patient line.

an effect is not recapitulated in standard

2D growth conditions, an oft-cited reason
for the poor predictive value of in vitro drug
screening for selecting clinically active
drugs. We therefore designed experi-
ments to determine whether GSC sensi-
tivity to chemotherapeutic agents and
radiation is different in the GLICO model
compared with standard 2D growth. For
the chemotherapy screens, we used two
alkylating agents that are the most
frequently clinically prescribed agents in
GBM patients: temozolomide (TMZ) and
bis-chloroethylnitrosourea (BCNU). Cyto-
toxicity assays of 2D cultures of 827 and
propagate calcium signals (Figure 2F; Video S3). Thus, we have 923 GSCs revealed a dose-dependent decrease in cell viability
developed a humanized bioengineered model of GBM that reca- when cells were treated with TMZ or BCNU. At 7 days post-treat-
pitulates microtube formation in vitro. ment, TMZ (1 mM) reduced cell viability by more than 80% in both
Additionally, our electron microscopy studies also revealed 827 and 923 2D GSCs (Figure S5A). Similarly, BCNU (100 mM)
multiple desmosome-mediated connections (Figure 2C, bottom) reduced cell viability by more than 90% in both GSC lines (Fig-
as well as cytoplasmic fusions between not only neighboring tu- ure S5A). Despite this dramatic effect in vitro, TMZ (1 mM) treat-
mor cells (Figure 2E, left) but also between neurons and tumor ment in 827 and 923 GLICOs resulted in only 24% and 43%
cells (Figure 2E, right). Multiple organelles, including mitochon- reductions in tumor growth, respectively (Figure 3A). Moreover,
dria, were observed between the nuclei of the fused cells BCNU (100 mM) treatment attenuated tumor growth by 91% in
(Figure 2E). 827 GLICOs but only 5% in 923 GLICOs (Figure 3A).
In order to address the effects of ionizing radiation on the
GLICOs and 2D Cultures Exhibit Differential Sensitivity growth of GLICO tumors, we first aimed to determine the radia-
to Chemotherapeutic Agents and Ionizing Radiation tion tolerance of normal cerebral organoids. Normal cerebral or-
The in vivo tumor microenvironment can have a profound effect ganoids were treated with a single dose of 0, 10, or 80 Gy (positive
on the ability of a tumor cell to respond to cellular stress, including control) and analyzed for cleaved caspase-3 as a marker of
genotoxic damage from chemotherapy and radiotherapy. Such apoptosis at 72 h post-irradiation (Figure S5D). Minimal cleaved

Cell Reports 26, 3203–3211, March 19, 2019 3207

 Figure 4. GLICOs Exhibit Similar Growth
Rates in iPSC-Derived Organoids and in Or-
ganoids of Different Ages
(A) Immunofluorescence staining for GFAP (green)
and DAPI (blue) in 6- or 20-week-old H1-derived
cerebral organoids; n = 4 organoids per time point;
scale bars, 200 mm.
(B) Representative images of GLICO tumor growth
(top) and quantification of secreted luciferase ac-
tivity (bottom) from 923 GLICO tumors grown in
either 1- or 4-month-old H1-derived cerebral orga-
noids; n = 6 organoids per group; scale bar, 500 mm.
(C) Representative images of GLICO tumor growth
(top) and quantification of secreted luciferase ac-
tivity (bottom) from 923 GLICO tumors grown in
either 1-month-old hESC-derived (H1) or iPSC-
derived (H6) organoids; n = 6 organoids per group;
scale bar, 500 mm.

GLICOs (Figure 3B). Remarkably, 10 Gy

slowed tumor growth by little more than
25% compared with 0 Gy.
Taken together, our data clearly
demonstrate that, as is often seen in vivo,
isogenic GSC lines are highly more
resistant to drug and radiation-induced
genotoxic stress when grown within the
microenvironment of the cerebral orga-
noid (GLICO) than when grown under
traditional 2D conditions.

GLICOs Exhibit Similar Growth

Rates in Organoids of Different
Ages
In an attempt to define how other param-
eters of the GLICO model may influence
tumor growth, we aimed to address
whether the age of the organoid affects
the growth rate of the GSCs. To do this,
we co-cultured Cypridina luciferase-ex-
caspase-3 was detected in non-irradiated organoids and those pressing 923 GSCs with 1- or 4-month-old H1 hESC-derived ce-
irradiated with 10 Gy. Conversely, organoids irradiated with rebral organoids. As evident by the secreted luciferase activity,
80 Gy as well as those organoids treated with 1 mM staurosporine there was no statistically significant difference in the GSCs’
(additional positive control for apoptosis) revealed high levels of growth rate within young versus old organoids (Figure 4B). We
cleaved caspase-3. With this knowledge, we then compared did, however, observe a difference in the growth patterns of
the radiosensitivity of isogenic GSC lines grown in 2D versus in GSCs in the different aged cerebral organoids. GLICO tumors
the GLICO model. 923 GSCs were irradiated with 0, 5, or 10 Gy from 1-month-old organoids exhibited large areas of regional
of ionizing radiation, and 72 h post-irradiation, the number of proliferation, whereas GLICO tumors from older, 4-month-old
live cells was counted. Compared with the non-irradiated control, organoids displayed a more infiltrative pattern of growth (Fig-
5 and 10 Gy significantly reduced cell viability by more than 44% ure 4B). Whether this more highly infiltrative pattern in older
and 74%, respectively (Figure S5C). We next sought to determine cerebral organoids is related to the even more aggressive natural
the effects of ionizing radiation on 923 GSCs in the GLICO model. history of GBM in elderly patients, compared with younger pa-
On the basis of the observation that 10 Gy was well tolerated in tients, is unknown at this time.
the normal cerebral organoids, we treated 923 GLICOs express-
ing secreted luciferase with 0, 5, or 10 Gy and analyzed tumor Both hESC- and iPSC-Derived Cerebral Organoids
growth after 1 week. As evident by secreted luciferase activity Facilitate Growth of GLICO Tumors
as well as fluorescent imaging, GLICO tumors irradiated with Because cerebral organoids can be generated from hESCs or
5 Gy exhibited similar levels of growth to that of the non-irradiated iPSCs, we also wanted to determine whether the organoids

3208 Cell Reports 26, 3203–3211, March 19, 2019

derived from each group could equally facilitate the growth and Medicine/NewYork-Presbyterian Hospital. On the basis of
invasion of GLICO tumors. To address this important issue, we NGS analysis, we isolated two GSC lines (0607 and 0810)
co-cultured Cypridina luciferase-expressing 923 GSCs with from patients with confirmed EGFR amplification in their
1-month-old hESC (H1)- or iPSC (H6)-derived cerebral organo- respective parental tumors. We then generated gender-mis-
ids and evaluated tumor growth. Interestingly, there was no sta- matched 0607 GLICOs or 0810 GLICOS; 0607 (from a female
tistically significant difference in tumor proliferation as assessed patient) was co-cultured with H1 (from a male embryo)-derived
by secreted luciferase activity between hESC-derived versus organoids, and 0810 (from a male patient) was co-cultured with
iPSC-derived GLICOS, as both types of organoids fostered H9 (from a female embryo)-derived organoids. We then per-
tumor growth (Figure 4C). formed triple DNA fluorescence in situ hybridization (FISH) for
EGFR/Cen7/ChrY. The Y chromosome was used as a method
GSC Survival Is Attenuated in Absence of Cerebral to distinguish tumor cells from normal organoid cells within
Organoid Microenvironment intact 0607 and 0810 GLICOs. As shown in Figure 3C, both
Although GSCs clearly infiltrate and proliferate inside our normal 0607 and 0810 GLICO tumor cells exhibited EGFR amplification
cerebral organoids as demonstrated above, we wanted to verify compared with the normal organoid-derived cells from their
that GSC-derived tumor growth is not merely a result of pro-pro- respective GLICOs (****p < 0.0001). In addition, as it has been
liferative cues from the organoid differentiation medium but well described that EGFR amplification is often rapidly lost
rather from the unique relationship between the tumor cells once GBM cells are cultured in 2D (Humphrey et al., 1988; Pan-
and the organoid’s host microenvironment. Accordingly, we dita et al., 2004), we also assessed EGFR amplification in 2D
cultured GSCs in vitro in either GSC medium (NBE) or organoid samples of 0607 and 0810. Our findings revealed that 0810
differentiation medium and analyzed cell growth after 7 days. maintains high levels of EGFR amplification when cultured in
Strikingly, organoid differentiation medium reduced cell 2D, but EGFR amplification is lost in 0607 GSCs (Figure 3C).
growth by at least 70% in all GSC lines (Figures S6A and S6B, Thus, the ability of the GLICO model to maintain EGFR amplifi-
**p < 0.01), indicating that the GSCs are not simply proliferating cation, even when such amplification is lost in 2D cultures, sug-
within the organoid as a result of growth factors or nutrients in the gests that the GLICO may provide a distinct microenvironment
organoid medium. To evaluate whether a 3D environment of that preserves the genetic landscape of the parental tumor.
extracellular matrix (Matrigel) would allow the GSCs to proliferate To determine whether GLICO tumors reflect the same patterns
in the cerebral organoid differentiation medium in the same way of signaling activation as their respective parental tumors, we
they do in our cerebral organoids, we generated Matrigel- simultaneously interrogated the phosphorylation of 49 different
embedded pure 923 GSC-derived tumor organoids that express receptor tyrosine kinases (RTKs) in patient GBMs and their cor-
secreted luciferase. Representative images and corresponding responding GLICOs. As illustrated in Figure S7, phosphorylation
quantitation of luciferase activity reveals that GSC-derived tumor of EGFR, related to RTK, and ephrin type-B receptor 3 (EphB3)
organoids can survive within the organoid differentiation me- was observed in both the parental tumor and the corresponding
dium, but their growth is still substantially reduced compared GLICO for GSC lines 0517 and 0810. Furthermore, for sample
with the tumor organoids grown in the GSC medium (NBE; Fig- 0517, phosphorylation of the following kinases was also
ure S6C). The morphology of the GSCs in the tumor organoids observed in both the parental tumor and GLICO: tyrosine kinase
cultured in organoid differentiation medium, compared with with immunoglobulin-like and EGF-like domain 1 (Tie-1), TEK
that seen in the GSCs in cerebral organoids (GLICO), demon- RTK (Tie-2), ephrin type-B receptor 2 (EphB2), ephrin receptor
strated many areas of single cells with punctate fluorescence, A 10 (EphA10), tyrosine-protein kinase Dtk (Dtk), Axl RTK (Axl),
an indicator that many cells may be initiating apoptosis in and RTK-like orphan receptor 2 (ROR2) (Figure S7). Thus, the
some areas of the tumor organoid (Figure S6C). There was signaling profiles of the GLICO tumors are highly reflective of
also a marked absence of the tumor microtubes that are routinely the signaling patterns that are found in the parental tumors.
observed when the GSCs are cultured in the normal cerebral
organoid, as described above. These findings indicate that the DISCUSSION
cerebral organoid provides a tumor microenvironment that is
distinct from that of a tumor organoid, and this microenviron- Cancer is not a cell-autonomous disease but rather one in which
ment is important for maintaining and enhancing the viability the biology of the cancer cell and host cell is intimately con-
and growth of the GSCs. nected. This is especially true for GBM, which does not metasta-
size but instead propagates and ultimately kills patients through
GLICOs Preserve Key Genetic and Signaling diffuse invasion and infiltration into surrounding normal cerebral
Components of the Parental Tumor tissue. Thus, 2D cultures and traditional tumor organoids do not
In addition to the ability of the GLICO model to recapitulate the model this critically important cell-cell interaction and tumor
invasive phenotype of human GBM, we also aimed to investi- microenvironment. Although patient-derived mouse xenograft
gate whether GLICO tumors maintain key genetic features models address issues of host-tumor cell interactions, the
and molecular signaling networks of the parental tumors. After marked interspecies differences at both the gross neuroanatom-
informed consent was obtained, CLIA (Clinical Laboratory ical (e.g., underdeveloped murine neocortex) and cellular
Improvement Amendments)-certified clinical next-generation level (e.g., astrocytic dendritic complexity and transmission
sequencing (NGS) was performed on GBM tumor samples speed of calcium transients in murine versus human astrocytes)
from patients undergoing surgical treatment at Weill Cornell (Oberheim et al., 2009) introduce additional variables. Finally,

Cell Reports 26, 3203–3211, March 19, 2019 3209

expensive, labor-intensive mouse models prevent real-time ge- STAR+METHODS
netic and molecular studies and raise numerous ethical issues
limiting their utility for both mechanistic studies and high- Detailed methods are provided in the online version of this paper
throughput drug screening. and include the following:
The GLICO model addresses a number of the limitations of
prior models for it allows one to study patient-specific GBMs d KEY RESOURCES TABLE
ex vivo within a microenvironment similar to that of a primitive d CONTACT FOR REAGENT AND RESOURCE SHARING
human brain. The biological behavior and histopathological d EXPERIMENTAL MODEL AND SUBJECT DETAILS
features of patient-derived GBMs growing within the cerebral or- B Patient-Derived GSCs
B hESCs and Induced Pluripotent Stem Cells
ganoids, in a manner that closely phenocopies surgical and au-
topsy specimens, attest to the clinical relevance of the model. d METHOD DETAILS
This is further substantiated by the GLICO’s maintenance of pa- B Cerebral Organoid Generation
B Co-Culture of GSCs and Cerebral Organoids
tient-specific EGFR amplification and phospho-RTK signaling,
B Two-Photon Microscopy
as well as the spontaneous formation of GLICO microtubes,
microscopic structural features also found in situ. Furthermore, B Bioluminescence Imaging of GLICO Tumors
B Light Sheet Microscopy of GLICO Tumors
because the model is grown ex vivo, it is amenable to experi-
B Proliferation of GSCs in Organoid Differentiation Me-
mental manipulation, drug treatment, and precise control of
physiological and environmental variables. Additionally, the dium
model is scalable, enabling one to generate hundreds of pa- B Immunofluorescence and Histopathological Evalua-

tient-specific GLICOs for high-throughput drug screening in a tion

way not currently possible for any in vivo model. B Fluorescence In Situ Hybridization (FISH) for EGFR

To this end, the ability to successfully screen tumor cells Amplification

B Drug Treatment of GSCs and GLICOs
in vitro for clinically active drugs and other interventions has
B Irradiation of GSCs and GLICOs
historically been extremely limited, particularly for GBM.
Although multiple reasons exist for the poor predicative power B Immunohistochemistry for Cleaved Caspase-3
B Immunofluorescence for Ki67 and GFP
of 2D screens for identifying clinically active agents, one un-
B EdU Labeling and Immunofluorescence
doubtedly important factor has been the failure to model the
profound effects of tumor and normal tissue microenvironment B Quantitative RT-PCR (qPCR)
B Human Phospho-RTK Arrays
and cell-to-cell interactions on tumor cell drug sensitivity. The
B Electron Microscopy
ability of the GLICO to model these variable drug and radiation
responses is encouraging and illustrates the potential of the B Calcium Imaging of Tumor Microtubes

GLICO to improve upon the predictive efficiency of in vitro d QUANTIFICATION AND STATISTICAL ANALYSIS
and ex vivo therapeutic screens; however, further studies
SUPPLEMENTAL INFORMATION
comparing drug sensitivity in GLICO with in vivo tumor drug
sensitivities will be required to know for certain. These future Supplemental Information can be found with this article online at https://doi.
studies will also allow us to evaluate how the tumor heteroge- org/10.1016/j.celrep.2019.02.063.
neity varies when the same GSCs are grown in GLICO versus
orthotopic xenograft models. ACKNOWLEDGMENTS
The GLICO model is neither complete nor will likely ever be
complete, as the technology is expanded to make the model We gratefully acknowledge Sushmita Mukherjee for her assistance with multi-
photon microscopic analysis of tumor volume, as well as Arline Faustin and
an ever closer representative of the human clinical disease.
Luis Chiriboga for their assistance with immunohistochemistry. We thank the
The model is amenable to new bioengineering methodologies WCM Imaging Core and WCM Neuroanatomy EM Core in the Feil Family Brain
as evidenced by the recent demonstration that pluripotent and Mind Research Institute. We also thank the MSKCC Molecular Cytoge-
stem cell self-organization around a microfilament scaffold (Lan- netics Core for their assistance with FISH. Work in H.F.’s laboratory is sup-
caster et al., 2017) may reduce intra-organoid neuroanatomical ported by an NIH Director Pioneer Award (1DP1CA228040-01).
variability and induce a component of organoid patterning. We
AUTHOR CONTRIBUTIONS
and our colleagues are also using novel bioengineering ap-
proaches to introduce a perfused vasculature (with blood-brain
A.L. and H.A.F. conceived the project and wrote the paper. A.L., D.B., M.S.,
barrier characteristics) and an immunologic niche given the po- L.E., K.M., T.M., B.R., L.C.-G., A.S., Y.N., E.S., R.S., S.C., T.M., Y.L., G.N.,
tential autologous nature of the model. K.C., D.P. and C.L. designed, performed, and analyzed experiments.
In summary, the human cerebral organoid represents an
excellent opportunity to explore the biological consequences DECLARATION OF INTERESTS
of human CNS tissue on human glioma growth through a model
The authors declare no competing interests.
system that is easily manipulated both genetically and pharma-
cologically. Capable of real-time imaging at the microscopic
Received: October 15, 2017
level, cerebral organoids and GSCs offer a powerful tool for Revised: November 14, 2018
investigating GBM biology in a primitive human brain environ- Accepted: February 15, 2019
ment and for modeling diverse therapeutic interventions. Published: March 19, 2019

3210 Cell Reports 26, 3203–3211, March 19, 2019

REFERENCES the phenotype and genotype of primary tumors than do serum-cultured cell
lines. Cancer Cell 9, 391–403.
Anisimov, V.N., Ukraintseva, S.V., and Yashin, A.I. (2005). Cancer in rodents: Lee, J., Son, M.J., Woolard, K., Donin, N.M., Li, A., Cheng, C.H., Kotliarova, S.,
does it tell us about cancer in humans? Nat. Rev. Cancer 5, 807–819. Kotliarov, Y., Walling, J., Ahn, S., et al. (2008). Epigenetic-mediated dysfunc-
Bao, S., Wu, Q., McLendon, R.E., Hao, Y., Shi, Q., Hjelmeland, A.B., Dewhirst, tion of the bone morphogenetic protein pathway inhibits differentiation of
M.W., Bigner, D.D., and Rich, J.N. (2006). Glioma stem cells promote radiore- glioblastoma-initiating cells. Cancer Cell 13, 69–80.
sistance by preferential activation of the DNA damage response. Nature 444, Lu, P., Weaver, V.M., and Werb, Z. (2012). The extracellular matrix: a dynamic
756–760. niche in cancer progression. J. Cell Biol. 196, 395–406.
Baysan, M., Bozdag, S., Cam, M.C., Kotliarova, S., Ahn, S., Walling, J., Killian, Masurkar, A.V., and Chen, W.R. (2012). The influence of single bursts versus
J.K., Stevenson, H., Meltzer, P., and Fine, H.A. (2012). G-cimp status predic- single spikes at excitatory dendrodendritic synapses. Eur. J. Neurosci. 35,
tion of glioblastoma samples using mRNA expression data. PLoS ONE 7, 389–401.
e47839.
Milner, T.A., Waters, E.M., Robinson, D.C., and Pierce, J.P. (2011). Degenerat-
Baysan, M., Woolard, K., Bozdag, S., Riddick, G., Kotliarova, S., Cam, M.C., ing processes identified by electron microscopic immunocytochemical
Belova, G.I., Ahn, S., Zhang, W., Song, H., et al. (2014). Micro-environment methods. Methods Mol. Biol. 793, 23–59.
causes reversible changes in DNA methylation and mRNA expression profiles
Oberheim, N.A., Takano, T., Han, X., He, W., Lin, J.H., Wang, F., Xu, Q., Wyatt,
in patient-derived glioma stem cells. PLoS ONE 9, e94045.
J.D., Pilcher, W., Ojemann, J.G., et al. (2009). Uniquely hominid features of
Campeau, E., Ruhl, V.E., Rodier, F., Smith, C.L., Rahmberg, B.L., Fuss, J.O., adult human astrocytes. J. Neurosci. 29, 3276–3287.
Campisi, J., Yaswen, P., Cooper, P.K., and Kaufman, P.D. (2009). A versatile
Osswald, M., Jung, E., Sahm, F., Solecki, G., Venkataramani, V., Blaes, J.,
viral system for expression and depletion of proteins in mammalian cells.
Weil, S., Horstmann, H., Wiestler, B., Syed, M., et al. (2015). Brain tumour cells
PLoS ONE 4, e6529.
interconnect to a functional and resistant network. Nature 528, 93–98.
Dragunow, M. (2008). The adult human brain in preclinical drug development. Pandita, A., Aldape, K.D., Zadeh, G., Guha, A., and James, C.D. (2004). Con-
Nat. Rev. Drug Discov. 7, 659–666. trasting in vivo and in vitro fates of glioblastoma cell subpopulations with
Elkabetz, Y., Panagiotakos, G., Al Shamy, G., Socci, N.D., Tabar, V., and amplified EGFR. Genes Chromosomes Cancer 39, 29–36.
Studer, L. (2008). Human ES cell-derived neural rosettes reveal a functionally Perrin, S. (2014). Preclinical research: Make mouse studies work. Nature 507,
distinct early neural stem cell stage. Genes Dev. 22, 152–165. 423–425.
Gould, S.E., Junttila, M.R., and de Sauvage, F.J. (2015). Translational value of Peters, A., Palay, S.L., and Webster, H.d. (1991). The Fine Structure of the Ner-
mouse models in oncology drug development. Nat. Med. 21, 431–439. vous System: Neurons and Their Supporting Cells, Third Edition (Oxford Uni-
Holland, E.C. (2001). Gliomagenesis: genetic alterations and mouse models. versity Press).
Nat. Rev. Genet. 2, 120–129. Quail, D.F., and Joyce, J.A. (2017). The microenvironmental landscape of brain
Humphrey, P.A., Wong, A.J., Vogelstein, B., Friedman, H.S., Werner, M.H., tumors. Cancer Cell 31, 326–341.
Bigner, D.D., and Bigner, S.H. (1988). Amplification and expression of the Rubenstein, B.M., and Kaufman, L.J. (2008). The role of extracellular matrix in
epidermal growth factor receptor gene in human glioma xenografts. Cancer glioma invasion: a cellular Potts model approach. Biophys. J. 95, 5661–5680.
Res. 48, 2231–2238.
Sansom, S.N., Griffiths, D.S., Faedo, A., Kleinjan, D.J., Ruan, Y., Smith, J., van
Huszthy, P.C., Daphu, I., Niclou, S.P., Stieber, D., Nigro, J.M., Sakariassen, Heyningen, V., Rubenstein, J.L., and Livesey, F.J. (2009). The level of the tran-
P.O., Miletic, H., Thorsen, F., and Bjerkvig, R. (2012). In vivo models of primary scription factor Pax6 is essential for controlling the balance between neural
brain tumors: pitfalls and perspectives. Neuro-oncol. 14, 979–993. stem cell self-renewal and neurogenesis. PLoS Genet. 5, e1000511.
Hutchinson, L., and Kirk, R. (2011). High drug attrition rates—where are we Shroyer, N.F. (2016). Tumor Organoids Fill the Niche. Cell Stem Cell 18,
going wrong? Nat. Rev. Clin. Oncol. 8, 189–190. 686–687.
John Lin, C.C., Yu, K., Hatcher, A., Huang, T.W., Lee, H.K., Carlson, J., Sloan, S.A., Darmanis, S., Huber, N., Khan, T.A., Birey, F., Caneda, C., Reimer,
Weston, M.C., Chen, F., Zhang, Y., Zhu, W., et al. (2017). Identification of R., Quake, S.R., Barres, B.A., and Pasca, S.P. (2017). Human astrocyte matu-
diverse astrocyte populations and their malignant analogs. Nat. Neurosci. ration captured in 3D cerebral cortical spheroids derived from pluripotent stem
20, 396–405. cells. Neuron 95, 779–790.e6.
Jucker, M. (2010). The benefits and limitations of animal models for transla- Son, M.J., Woolard, K., Nam, D.H., Lee, J., and Fine, H.A. (2009). SSEA-1 is an
tional research in neurodegenerative diseases. Nat. Med. 16, 1210–1214. enrichment marker for tumor-initiating cells in human glioblastoma. Cell Stem
Lancaster, M.A., Renner, M., Martin, C.A., Wenzel, D., Bicknell, L.S., Hurles, Cell 4, 440–452.
M.E., Homfray, T., Penninger, J.M., Jackson, A.P., and Knoblich, J.A. (2013). Stupp, R., Mason, W.P., van den Bent, M.J., Weller, M., Fisher, B., Taphoorn,
Cerebral organoids model human brain development and microcephaly. Na- M.J., Belanger, K., Brandes, A.A., Marosi, C., Bogdahn, U., et al.; European
ture 501, 373–379. Organisation for Research and Treatment of Cancer Brain Tumor and Radio-
Lancaster, M.A., Corsini, N.S., Wolfinger, S., Gustafson, E.H., Phillips, A.W., therapy Groups; National Cancer Institute of Canada Clinical Trials Group
Burkard, T.R., Otani, T., Livesey, F.J., and Knoblich, J.A. (2017). Guided self- (2005). Radiotherapy plus concomitant and adjuvant temozolomide for glio-
organization and cortical plate formation in human brain organoids. Nat. Bio- blastoma. N. Engl. J. Med. 352, 987–996.
technol. 35, 659–666. Tentler, J.J., Tan, A.C., Weekes, C.D., Jimeno, A., Leong, S., Pitts, T.M., Arcar-
Lee, J., Kotliarova, S., Kotliarov, Y., Li, A., Su, Q., Donin, N.M., Pastorino, S., oli, J.J., Messersmith, W.A., and Eckhardt, S.G. (2012). Patient-derived tumour
Purow, B.W., Christopher, N., Zhang, W., et al. (2006). Tumor stem cells xenografts as models for oncology drug development. Nat. Rev. Clin. Oncol. 9,
derived from glioblastomas cultured in bFGF and EGF more closely mirror 338–350.

Cell Reports 26, 3203–3211, March 19, 2019 3211

STAR+METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE SOURCE IDENTIFIER

Antibodies
Rabbit monoclonal anti-N-Cadherin Cell Signaling Cat#13116S; RRID: AB_2687616
Rabbit monoclonal anti-SOX2 Cell Signaling Cat# 3579S; RRID: AB_2195767
Mouse polyclonal anti-TUJ1 BioLegend Cat# 801201; RRID: AB_2313773
Mouse monoclonal anti-PAX6 Santa Cruz Cat# sc-81649; RRID: AB_1127044
Rabbit polyclonal anti-TBR2 Abcam Cat#ab23345; RRID: AB_778267
Rabbit polyclonal MAP2 Cell Signaling Cat# 4542S; RRID: AB_10693782
Rabbit monoclonal Musashi Cell Signaling Cat# 85652S; N/A
Rabbit polyclonal Nestin BioLegend Cat#841801; RRID: AB_2565467
Rabbit monoclonal anti-GFAP Cell Signaling Cat# 12389S; RRID: AB_2631098
Rabbit polyclonal anti-GFAP Dako Cat# ZO334; N/A
Rabbit polyclonal anti-Connexin-43 Sigma Aldrich Cat# C6219; RRID: AB_476857
Chicken polyclonal anti-GFP Abcam Cat#ab13970; RRID: AB_300798
Rabbit monoclonal anti-Ki67 Cell Signaling Cat#9129; N/A
Rabbit polyclonal Cleaved Caspase-3 Cell Signaling Cat#9661; RRID: AB_2341188
Goat anti-mouse Alexa Fluor 488 Thermo Fisher Scientific Cat# A11001; RRID: AB_2534069
Goat anti-rabbit Alexa Fluor 594 Thermo Fisher Scientific Cat# A11012; RRID: AB_2534079
Goat anti-rabbit Alexa Fluor 568 Thermo Fisher Scientific Cat#A11036; RRID: AB_10563566
Goat anti-chicken Alexa Fluor 488 Thermo Fisher Scientific Cat#A11039; RRID: AB_2534096
Goat polyclonal anti-mouse IgG-HRP Thermo Fisher Scientific Cat# 31430; RRID: AB_228307
Goat polyclonal anti-rabbit IgG-HRP Thermo Fisher Scientific Cat# 31460; RRID: AB_228341
Goat polyclonal anti-rabbit IgG-Biotin Jackson ImmunoResearch Cat# 111-065-003; RRID: AB_2337959
Laboratories
Chemicals, Peptides, and Recombinant Proteins
ROCK inhibitor (Y-27632 2HCl) Selleck Chemicals Cat# S1049; N/A
Rhod-2AM Invitrogen Cat# R-1244; N/A
Staurosporine Sigma-Aldrich Cat#S6942; N/A
Critical Commercial Assays
VECTASTAIN Elite ABC-HRP Kit Vector Laboratories Cat# PK-7200; RRID:AB_2336828
Proteome Profiler Array R&D Systems Cat# ARY001B; N/A
Click-iT EdU Alexa Fluor 594 Imaging Kit Thermo Fisher Scientific Cat#C10339; N/A
QuantiFast SYBR Green PCR Kit QIAGEN Cat#204054; N/A
Cypridina Luciferase Glow Assay Kit Pierce; Thermo Fisher Scientific Cat#16170; N/A
Experimental Models: Cell Lines
Patient-Derived Glioma Stem Cells: 827, 1228, National Institutes of Health N/A
211, 923, 308, 0728, 0320, 0517, 0607, 0810, and Weill Cornell Medicine/NYU
and 1206 Presbyterian Hospital
NIH-registered human H1 (WA01) embryonic WiCell Research Institute Cat# WA01; RRID: CVCL_9771
stem cells
NIH-registered human H9 (WA09) embryonic WiCell Research Institute Cat# WA09; RRID: CVCL_9773
stem cells
Recombinant DNA
pLenti PGK GFP Blast (w510-5) Campeau et al., 2009 Addgene Plasmid# 19069
pGL4.50[luc2/CMV/Hygro] Promega Cat# E1310; N/A
(Continued on next page)

e1 Cell Reports 26, 3203–3211.e1–e5, March 19, 2019

Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
pLenti-C-tRFP OriGene Cat# PS100074
pLenti-C-TurboRFP-P2A-CLuc Constructed in Lab N/A
Software and Algorithms
ImageJ NIH https://imagej.nih.gov/ij/; RRID: nif-0000-30467
Living Image Living Image Cat# 128113; http://www.perkinelmer.com/
product/li-software-for-spectrum-1-seat-
add-on-128113
GraphPad Prism 7 GraphPad RRID: SCR_000306

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for reagents may be directed to and will be fulfilled by the Lead Contact, Howard A. Fine (haf9016@
med.cornell.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Patient-Derived GSCs
Following informed consent, tumor samples classified as glioblastoma, based on the World Health Organization (WHO) criteria, were
obtained from patients undergoing surgical treatment at the National Institutes of Health (NIH) or from Weill Cornell Medicine/New
York Presbyterian Hospital in accordance with the appropriate Institutional Review Boards. Within 1–3 hours after surgical removal,
tumors were washed in PBS and enzymatically dissociated into single cells. Tumor cells were cultured in NBE medium consisting of
neurobasal medium (Thermo Fisher Scientific), N2 and B27 supplements (Thermo Fisher Scientific), and human recombinant bFGF
and EGF (25 ng/mL each; R&D Systems) plus Heparin sodium and L-Glutamine. Regular mycoplasma screening was performed
using the MycoAlert Detection Kit (Lonza Inc.).

hESCs and Induced Pluripotent Stem Cells

NIH-registered human H1 (WA01) or H9 (WA09) embryonic stem cells were purchased from WiCell Research Institute, Inc. and main-
tained in mTeSR1 medium (STEMCELL Technologies). H6 induced pluripotent stem cells were a generous gift from the Laboratory of
Todd Evans at Weill Cornell Medicine.

METHOD DETAILS

Cerebral Organoid Generation

Once hESC colonies reached 80% confluency, cells were plated in low-attachment 96-well plates (9,000 cells per 150 mL of STEMdiff
Neural Induction Medium (STEMCELL Technologies)). Embryoid bodies were fed every other day for 6-7 days; medium contained
50 mM Rho-associated protein kinase (ROCK) inhibitor for the first 4 days. To initiate neural rosette formation, embryoid bodies
were transferred to Matrigel-coated (Corning #354277) 6-well plates where they received daily medium changes (STEMdiff Neural
Induction, STEMCELL Technologies) for 6-7 days. Neural rosettes were then replated a second time to increase purity for an addi-
tional 5-7 days. Following the second round of formation, neural rosette clusters were transferred to droplets of Matrigel (Corning
#356234) by pipetting into cold Matrigel on a sheet of Parafilm. Droplets were solidified at 37 C and were subsequently grown without
agitation for 4 days in differentiation medium containing a 1:1 mixture of DMEM/F12 and Neurobasal, 1:200 N2 supplement (Thermo
Fisher Scientific), 1:100 B27 supplement without vitamin A (Thermo Fisher Scientific), 87.5 mL of 1:100 2-mercaptoethanol, 1:4,000
insulin (Sigma), 1:100 Glutamax (Thermo Fisher Scientific) and 1:200 MEM-NEAA. Following this stationary growth phase, the tissue
droplets were transferred to a spinning bioreactor or orbital shaker containing differentiation medium as described above except a
B27 supplement with vitamin A (Thermo Fisher Scientific) was used.

Co-Culture of GSCs and Cerebral Organoids

To establish cells that stably express GFP, GSCs were transfected with pLenti PGK GFP Blast (w510-5) (Campeau et al., 2009)
(Addgene) and were cultured under blasticidin selection in NBE medium. For co-culture experiments, individual organoids were
transferred to a 24-well plate (one organoid per well). Excess medium was removed, and 10,000 stable GFP-expressing GSCs
were plated in each organoid-containing well (10,000 GSCs/2 mL of NBE per well). Plates were incubated at 37 C for 24 hr with
no agitation. Each organoid was subsequently washed in PBS and transferred to a clean well with 2 mL of organoid differentiation
medium. Tumor-bearing organoids were maintained on an orbital shaker for up to 14 days at 37 C.

Cell Reports 26, 3203–3211.e1–e5, March 19, 2019 e2

Two-Photon Microscopy
For two-photon imaging of tumor volume and tumor microtubes, stable GFP-expressing 923 or RFP-expressing 827 GSCs were co-
cultured with cerebral organoids for 24 hr. GLICO tumors were allowed to grow for 10 days prior to imaging. All images were acquired
using an Olympus FluoView FVMPE-RS multi-photon microscope and a Mai Tai DeepSee ultrafast laser (Spectra Physics) tuned to
920 nm for GFP excitation. For optimal spatial resolution, image stacks were acquired in dual galvanometer scanning mode using a
high-sensitivity gallium arsenide phosphide (GaAsP) detector with a green-light filter cube and a 25X, 1.05 numerical aperture (NA)
objective with a 2-mm working distance (Olympus). Each high-resolution image stack (640x640 pixels, 4 ms dwell time, 3 mm step
size) spanned 525x525 mm (XY) and up to 600 mm in Z. The images in Figure 3B are max projections over the Z axis through
100–250 mm of tissue. All image analysis was produced by ImageJ/Fiji software. The number of organoids used per experiment is
stated for each individual figure.

Bioluminescence Imaging of GLICO Tumors

To establish cells that stably express firefly luciferase, 923 GSCs were transfected with pGL4.50[luc2/CMV/Hygro] (Promega) and
were cultured under hygromycin selection in NBE medium. Luciferase-expressing 923 GSCs were co-cultured with cerebral organo-
ids for 24 hours. GLICO tumors were allowed to form for 8 days with gentle shaking. GLICO tumors containing parental GSCs that did
not express luciferase were used as negative controls for imaging. Luciferin was added to the organoid medium 1 hour prior to im-
aging (100 mM in 2 mL medium). Using a 24-well black-walled plate, GLICO tumors were imaged on an IVIS SpectrumCT imaging
platform (PerkinElmer). Luminescent images were generated using Living Image software. For studies in which tumor growth was
monitored using secreted luciferase, 827 or 923 GSCs were stably transduced with pLenti-TurboRFP-P2A-CLuc and cultured under
neomycin selection in NBE medium. Luciferase-expressing GSCs were co-cultured with cerebral organoids for 24 hours. GLICO
tumor growth was monitored by assaying the culture medium for secreted luciferase acitivity at Day 8 and Day 18 after co-culture.
Tumor growth of pLenti-C-tRFP-expressing control GSCs was monitored in tandem. The number of organoids used per experiment
is stated for each individual figure.

Light Sheet Microscopy of GLICO Tumors

Organoids were collected with ice cold PBS and transferred to 1X PBS/4% PFA at 4 C overnight with shaking and then left at room
temperature for 1 hr. Organoids were then washed with 1X PBS with shaking 3 times and dehydrated in a series of methanol expo-
sures. Samples were bleached in 5% H2O2, in 20% DMSO/methanol overnight at 4 C. Organoids were transferred to 30% H2O2 /1
volume DMSO/4 volumes methanol overnight at 4 C. Following methanol washes and rehydration, samples were embedded in 1%
agarose prepared in TAE buffer. For immunolabeling, organoids were pretreated and incubated in 1X PBS/0.2%Triton X-10020%/
DMSO/0.3M glycine at 37 C overnight. Samples were then blocked in 1X PBS/0.2% Triton X-100/10%/DMSO/6% Donkey Serum at
37 C for 3 days then washed again in 1X PBS/0.2%Tween-20 with 10ug/ml heparin (PTwH), at room temperature 1hr, 2 times. Or-
ganoids were incubated with primary antibody to target GFP expressing cells to reduce autofluorescence diluted in PTwH/5%
DMSO/3%Donkey Serum at 37 C for 4 days. For sample clearing, organoid samples were incubated overnight in 50% Tetrahdro-
furan/H2O (THF, Sigma) in a glass vial with silicon coated cap (Thermo Scientific). Samples were incubated for 1 hr in 80% THF,
2 times for 1 hr in 100% THF and then incubated in Dichloromethane (DCM, Sigma) for 5 min to 1 hr (until samples sank). Organoids
were then incubated in DiBenzyl Ether (Sigma) until the sample was clear (20 min to 2 hr). All images were acquired with an Ultravision
II La Biotec bidirectional triple light-sheet microscope. Stacks of 130 images with 1.04 mm between planes were acquired for the GFP
signal and pseudocolored red. All image processing and analysis was done using Imaris software and where 3D rendering was
appropriate.

Proliferation of GSCs in Organoid Differentiation Medium

GSCs were seeded in triplicate into 12 well plates (100,000 cells/well) with either GSC medium (NBE) or organoid differentiation me-
dium. Live cells were counted after 7 days. A Student t test was performed, and statistical significance was defined as p < 0.05. Data
are presented as mean values with 95% confidence intervals. For tumor organoid experiments, 100,000 luciferase-expressing GSCs
were plated in 50 mL Matrigel and allowed to grow in either GSC medium (NBE) or organoid differentiation medium. Luciferase levels
were measured after 7 days of tumor cell growth using the Cypridina Luciferase Glow Assay Kit (Pierce). The number of organoids
analyzed is indicated in the figure legend.

Immunofluorescence and Histopathological Evaluation

Cerebral organoids were fixed in 4% paraformaldehyde for 45 minutes at room temperature followed by three PBS washes for
10 minutes each. Samples were incubated in 30% sucrose overnight and embedded in OCT for cryosectioning. Frozen sections
(10 mm thickness) were stained with hematoxylin and eosin or used for immunostaining. Sections were blocked and permeabilized
in 0.3% Triton X-100 and 3% normal goat serum in PBS. The following primary antibodies were used: N-cadherin (Cell Signaling,
13116S, 1:200), SOX2 (Cell Signaling, 3579S, 1:200), TUJ1 (BioLegend, 845502, 1:200), PAX6 (Santa Cruz, sc-81649, 1:200), TBR2
(Abcam, ab23345, 1:200), MAP2 (Cell Signaling, 4542S, 1:200), Musashi-1 (Cell Signaling, 85652S, 1:200), Nestin (BioLegend,
841801, 1:200), and GFAP (IHC: Cell Signaling, 12389S, 1:200; IF: Dako, ZO334, 1:1000). Secondary antibodies used were goat Alexa
Fluor 488 and 594 conjugates (Thermo Fisher Scientific, 1:500). For immunofluorescence, DAPI (4 , 6-diamidino-2 -phenylindole,

e3 Cell Reports 26, 3203–3211.e1–e5, March 19, 2019

dihydrochloride) was used as a counterstain at a concentration of 1.0 mg/mL in PBS. Histopathologic evaluation was performed by
board-certified neuropathologist, Dr. Matija Snuderl (NYU Langone Medical Center).

Fluorescence In Situ Hybridization (FISH) for EGFR Amplification

Following informed consent, tumor samples classified as glioblastoma, based on the World Health Organization (WHO) criteria, were
obtained from patients undergoing surgical treatment at the National Institutes of Health (NIH) or from Weill Cornell Medicine/New
York Presbyterian Hospital in accordance with the appropriate Institutional Review Boards. Samples were then sent for targeted
next generation sequencing (NGS) for > 300 cancer-related genes (FoundationOne, Foundation Medicine Inc., Cambridge, MA).
Once the EGFR amplification status was verified by NGS, GSCs from two EGFR-amplified tumors (0607 and 0810) were used to
generate gender-mismatched 0607 GLICOs or 0810 GLICOS; 0607 (from female patient) was co-cultured with H1 (from male em-
bryo)-derived organoids, and 0810 (from male patient) was co-cultured with H9 (from female embryo)-derived organoids. We then
performed triple DNA FISH for EGFR/Cen7/Y chromosome. The Y chromosome was used as a method to distinguish tumor cells
from normal organoid cells within intact 0607 and 0810 GLICOs. After two weeks of incubation on an orbital shaker, GSCs from
GLICO tumors were harvested and fixed in methanol:acetic acid (3:1) as per standard procedures. FISH analysis was performed
on fixed cells using a home-brew EGFR/Cen7/ChrY probe. The probe mix consisted of PAC/BAC clones containing the full length
EGFR gene (clones RP5-1019E12, and RP11-339F13; labeled with Red dUTP) and a centromeric repeat plasmid for chromosome
7 served as the control (clone p7t1; labeled with Orange dUTP). The ChrY probe detects the Y chromosome; labeled with Green
dUTP. Probe labeling, hybridization, post-hybridization washing, and fluorescence detection were performed according to standard
laboratory procedures. Slides were scanned using a Zeiss Axioplan 2i epifluorescence microscope equipped with a megapixel CCD
camera (CV-M4+CL, JAI) controlled by Isis 5.5.9 imaging software (MetaSystems Group Inc, Waltham, MA). The entire hybridized
area was first scanned through 63X or 100X objective to assess quality of hybridization and signal pattern. For each sample, a
minimum of 100 discrete nuclei and 25 metaphases (when available) were scored. Amplification was defined as EGFR:Cen 7 ratio
of R 2.0, R 6 copies of EGFR (independent of control locus) or at least one small cluster of Gene (R4 signals resulting from tandem
repeat/duplication). In cells with high-level amplification, signals R 20 cannot be accurately counted and therefore given a
score of 20.

Drug Treatment of GSCs and GLICOs

For 2D cultures, 827 and 923 GSCs were plated in triplicate at a density of 100,000 cells per well in 6-well plates. Cells were treated
with either TMZ (200 mM, 500 mM, or 1 mM) or BCNU (10 mM, 25 mM, 50 mM, or 100 mM). 7 days post-treatment, live cells were counted
using the trypan blue dye-exclusion assay. For GLICOs, cerebral organoids were co-cultured with 100,000 827 or 923 luciferase-ex-
pressing GSCs and allowed to grow for 5 days. GLICOs were then treated with 1 mM TMZ or 100 mM BCNU. Luciferase levels were
measured at Day 0 and 7 days post-treatment using the
Cypridina Luciferase Glow Assay Kit (Pierce)
The number of organoids used per experiment is stated for each individual figure.

Irradiation of GSCs and GLICOs

For 2D cultures, 923 GSCs were plated in triplicate at a density of 300,000 per well in 6-well plates. 48 hours later, cells were irradiated
with 0 Gy, 5 Gy, or 10 Gy using a Shepherd Mark-1 Cesium-137 Irradiator at a dose rate of 9.75 Gy/min. At 72 hours post-irradiation,
live cells were counted using the trypan blue dye-exclusion assay. For irradiation of 923 GLICOs, cerebral organoids were co-
cultured with 100,000 923 luciferase-expressing GSCs and allowed to grow for 5 days. GLICOs were then irradiated with 0, 10, or
80 Gy using a Shepherd Mark-1 Cesium-137 Irradiator at a dose rate of 9.75 Gy/min. Luciferase levels were measured at Day
0 and 7 days post-irradiation using the Cypridina Luciferase Glow Assay Kit (Pierce). The number of organoids used per experiment
is stated for each individual figure.

Immunohistochemistry for Cleaved Caspase-3

Paraffin sections from cerebral organoids treated with 0 Gy, 10 Gy, 80 Gy or Staurosporine (Sigma, St. Louis, MO; 1 mM for 72 hr) were
rehydrated and washed in phosphate buffered saline (PBS). Then, they were incubated in HistoReveal (Abcam, Cambridge, MA)
for 5 min at room temperature. Sections were then washed in PBS and blocked with 3% goat serum with 0.3% Triton X-100 for
30 minutes at room temperature. Primary anti-cleaved caspase-3 antibody (Cell Signaling Technologies, Inc., Danvers, USA) was
applied to the sections with overnight incubation at 4 C. Sections were washed in PBS and incubated with the Alexa 568 secondary
antibody (Molecular Probes-Life Technologies, Grand Island, USA). Washed in PBS, nuclear counterstaining was done with
NucBlueTMstain solution (Molecular Probes-Life Technologies, Grand Island, USA) before microscopic analysis with an Evos FL
Auto imaging system (Carlsbad, CA); n = 4 organoids per group.

Immunofluorescence for Ki67 and GFP

923 GFP-expressing GLICOs were fixed in 4% paraformaldehyde for 45 minutes at room temperature followed by three PBS washes
for 10 minutes each and then embedded in paraffin. Sections of 5 mm were obtained using microtome on poly-lysine-coated slides.
Following deparaffinization and rehydration, antigen retrieval was performed by submerging the slides in Trilogy solution

Cell Reports 26, 3203–3211.e1–e5, March 19, 2019 e4

(Sigma, 920P) by heat in a pressure cooker for 15 min. Sections were permeabilized for 20 min with PBS/0.5% Triton X-100 at R.T.
and blocked 1 hr with PBS/3% BSA. Each section was incubated O.N. at 4 C with primary antibodies against GFP (Abcam, ab13970,
1:200) and Ki-67 (Cell Signaling, 9129, 1:200), followed by an incubation with the secondary antibodies coupled to Alexa 488 (Invi-
trogen, A11039, 1:200) and Alexa 568 (Invitrogen, A11036, 1:200) for 1h at RT. Nuclei were counterstained with DAPI. Images were
obtained using an epifluorescence microscope, processed and analyzed using the Fiji software. The number of Ki67 and GFP pos-
itive cells was quantified and expressed as a percentage of the total GFP positive cells. At least 500 GFP positive cells were included
from each sample over three different fields at 200x. The result is shown as the mean ± standard error of three independent exper-
iments; n = 3 organoids.

EdU Labeling and Immunofluorescence

Cell proliferation was assessed in 923 GFP-expressing GLICOs using the 9129
Click-iT EdU Alexa Fluor 594 Imaging Kit (Thermo, C10339). Briefly, GLICOS were treated for 48h with 20 mM EdU (5-ethynyl-20 -de-
oxyuridine) and fixed in 4% paraformaldehyde for 45 minutes at room temperature, followed by three PBS washes for 10 minutes
each and then embedded in paraffin. Sections of 5 mm were obtained using microtome on polylysine-coated slides. Following de-
paraffinization and rehydration, antigen retrieval was performed using HistoReveal solution (Abcam, ab103720) during 15 min at
R.T. Sections were permeabilized for 20 min with PBS/0.5% Triton X-100 at R.T. The Click-iT reaction was conducted following man-
ufacturer’s instructions. Next, antibody labeling for GFP was performed as described before. Nuclei were counterstained with DAPI.
Images were obtained using an epifluorescence microscope, processed and analyzed; n = 3 organoids.

Quantitative RT-PCR (qPCR)

Total RNA was extracted from 827 GFP-expressing GLICO samples using RNeasy Mini Kit (QIAGEN) and reverse transcription re-
action was carried out using High-Capacity RNA-to-cDNA Kit (Thermo Fisher Scientific), according to the manufacturer’s protocols.
The resulting cDNA was amplified with gene-specific primers, eGFP F – ACGTAAACGGCCACAAGTTC; eGFP R – AAGTCGTGC
TGCTTCATGTG; hGAPDH F – TGCACCACCAACTGCTTAGC; hGAPDH R – GGCATGGACTGTGGTCATGAG and quantified
using QuantiFast SYBR Green PCR Kit (QIAGEN) on QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems). Comparative
Ct (DDCt) method was used for performing relative quantitation of gene expression.

Human Phospho-RTK Arrays

Proteome Profiler human phospho-RTK antibody arrays (R&D Systems) were used according to the manufacturer’s instructions. A
total of 50 mg of fresh protein lysates from 3-5 GLICO tumors and their corresponding parental tumor tissue were incubated overnight
with nitrocellulose membranes dotted with duplicate spots for 49 anti-RTK and control antibodies. Bound phospho-RTKs were
detected with a pan anti-phospho-tyrosine antibody conjugated to horseradish peroxidase using chemiluminescence.

Electron Microscopy
Cerebral organoids were fixed with 3.75% acrolein and 2% paraformaldehyde in 0.1M phosphate buffer (PB; pH7.4) overnight at
4 C(Milner et al., 2011). The next day, the sections were rinsed in PB. Two experiments were performed. Experiment 1: Normal
Morphology-Following en bloc staining with uranyl acetate and graded ethanol dehydration, samples were embedded in an Epon
analog resin. Ultrathin sections (65 nm) were contrasted with lead citrate for use in electron microscopy. Experiment 2: Connexin-
43 Immunolabeling – Free-floating cerebral organoids were incubated with anti-connexin-43 (Sigma; 1:2000) and a goat-anti-
rabbit IgG-biotinylated secondary antibody (Jackson Immunoresearch Laboratories; 1:400) using the avidin-biotin complex perox-
idase method (Vectastain ABC-HRP Kit; Vector Laboratories) (Milner et al., 2011). Organoids were dehydrated and flat-embedded in
EMBed-812. Organoids were sectioned (70 nm thick) on a Leica ultratome (Ultracut UCT) and collected on 400 mesh copper grids
and then counterstained with uranyl acetate and Reynold’s lead citrate. For both experiments, grids were imaged on an FEI Tecnai
BioTwin Transmission Electron Microscope. Elements were identified using morphological criteria defined in Peters et al. (1991). The
number of organoids used per experiment is stated for each individual figure.

Calcium Imaging of Tumor Microtubes

The following small molecule calcium indicator, 5 mM Rhod-2AM (Life Technologies, R-1244), was applied to GFP-expressing 923
GSC tumor microtubes. Time-lapse microscopy of calcium signaling was recorded for 3 minutes.

QUANTIFICATION AND STATISTICAL ANALYSIS

Results were analyzed using GraphPad Prism 7. For 2D experiments, cells were plated in triplicate; results were verified from three
independent experiments. Statistical data for all experiments was analyzed using two-tailed Student’s t tests. Figure legends define
the sample size and significance. Data were judged to be statistically significant when p < 0.05; standard error bars correspond to
95% confidence intervals.

e5 Cell Reports 26, 3203–3211.e1–e5, March 19, 2019