Cerberal Organoid Glioma
Cerberal Organoid Glioma
Cerberal Organoid Glioma
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
Report
*Correspondence: [email protected]
https://doi.org/10.1016/j.celrep.2019.02.063
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.
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
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).
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.).
METHOD DETAILS
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.
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.