Reviews

Multidimensional communication in
the microenvirons of glioblastoma
Marike L. Broekman1,2*, Sybren L. N. Maas1,2, Erik R. Abels1, Thorsten R. Mempel3,4,
Anna M. Krichevsky5 and Xandra O. Breakefield1*
Abstract | Glioblastomas are heterogeneous and invariably lethal tumours. They are characterized
by genetic and epigenetic variations among tumour cells, which makes the development of
therapies that eradicate all tumour cells challenging and currently impossible. An important
component of glioblastoma growth is communication with and manipulation of other cells in the
brain environs, which supports tumour progression and resistance to therapy. Glioblastoma cells
recruit innate immune cells and change their phenotype to support tumour growth. Tumour cells
also suppress adaptive immune responses, and our increasing understanding of how T cells
access the brain and how the tumour thwarts the immune response offers new strategies for
mobilizing an antitumour response. Tumours also subvert normal brain cells — including
endothelial cells, neurons and astrocytes — to create a microenviron that favours tumour
success. Overall, after glioblastoma-​induced phenotypic modifications, normal cells cooperate
with tumour cells to promote tumour proliferation, invasion of the brain, immune suppression and
angiogenesis. This glioblastoma takeover of the brain involves multiple modes of communication,
including soluble factors such as chemokines and cytokines, direct cell–cell contact, extracellular
vesicles (including exosomes and microvesicles) and connecting nanotubes and microtubes.
Understanding these multidimensional communications between the tumour and the cells in its
environs could open new avenues for therapy.

Glioblastomas remain one of the most aggressive malig- these tumours to manipulate and exploit normal brain
nancies, with no change in the standard of care for cells. Almost all cell types in the tumour environs are
almost 20 years and a median lifespan from time of diag- affected: the tumour is able to stimulate angiogenesis and
nosis to death of about 15 months1. This bleak outcome co-​opt existing vasculature2, suppress immune cell func-
has stimulated ongoing efforts to reveal new insights tions3, disarm microglia and macrophages that should
into these tumours and the surrounding cells to facilitate recognize and fight foreign elements in the brain 4,
development of new treatment strategies. New studies coerce astrocytes into joining tumour support5 and
and technologies have deepened our understanding of even change the extracellular matrix (ECM) to facilitate
the factors that make these tumours so formidable but invasion6. Conversely, new insights into the presence
have highlighted two major challenges. First, a lack of of adaptive immune cells in the brain and the presence of
models that can authentically reproduce the genetic and a CNS lymphatic system7,8 might give rise to therapeutic
phenotypic properties of human glioblastoma (Box 1), opportunities that manipulate this system to recognize
especially regarding the analysis of glioblastoma tumour neoantigens9 similarly to the strategies currently
microenvironmental communication, is hampering being clinically applied for some melanoma and lung
progress into the development of new therapies for the cancer patients.
condition. Second, as underlined by the 2016 WHO Glioblastoma recruits normal cells in its environs
classification system, evidence increasingly demon- to promote growth, sustenance and encroachment of
strates that glioblastoma is genetically heterogeneous the tumour into the brain (Fig. 1). This glioblastoma
(Box 2) and thus will probably require combinatorial ‘takeover’ of the brain involves multiple types of com-
*e-​mail: approaches for different subtypes of tumour cell even munication and directive exchanged between tumour
M.L.D.Broekman-16@
within a single glioblastoma tumour. cells and surrounding cells. Cell-​secreted soluble factors,
umcutrecht.nl; breakefield@
hms.harvard.edu In addition to this deepening understanding of the including transforming growth factor-​β (TGFβ), IL-6,
https://doi.org/10.1038/ genetic and phenotypic variability within glioblastoma, Notch, platelet-​derived growth factor (PDGF), epidermal
s41582-018-0025-8 the field has gained increasing awareness of the ability of growth factor (EGF), vascular endothelial growth factor

482 | AUGUST 2018 | volume 14 www.nature.com/nrneurol

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Reviews

Key Points
junctions or a cytoplasmic continuum between cells to
enable transport of molecules and organelles21–23. The
• Glioblastomas use numerous forms of communication to hijack many different cell involvement of microtubes has been indicated in the
types in the brain environs to support tumour progression. regrowth of tumours after surgery and in conferring
• Communication routes include secreted proteins and molecules, gap junctions resistance to chemotherapy24, although they are not
between cells, extracellular vesicles, tunnelling nanotubes and microtubes. apparent in some glioma models25.
• Tumour cells co-​opt microglia and infiltrating macrophages for their own benefit These different modes of physical support among
through the release of cytokines and extracellular vesicles. tumour cells, and the two-​way crosstalk between tumour
• Glioblastomas and pericytes generate a state of reduced T cell effector function that cells and normal cells in their vicinity, together with
is commonly referred to as T cell exhaustion or dysfunction. the epigenetic flexibility of cells, enable the tumour
• The interaction of tumour cells with normal brain cells, such as neurons, is not to create a pluripotent environment that can adapt to
unidirectional, and neuronal activity is subverted to promote glioblastoma changes and thus give the tumour many options to sur-
progression. vive therapeutic assault. Given the presence of multiple
• Comprehension and disruption of tumour directives in the glioblastoma routes of information transfer in the tumour microen-
microenvironment could improve therapeutic intervention for these lethal tumours. vironment, it is difficult to determine which route
mediates which functions and how best to disrupt this
(VEGF) and stromal cell-​derived factor 1 (SDF1; also multifaceted tumour-​supportive network.
known as CXCL12), are well known to serve as signalling This Review focuses on new findings that elucidate
molecules by binding to receptors on target cells, but the the complex and dynamic modes of communication
importance of other routes of communication — such between tumour cells and normal cells in the tumour
as gap junctions, extracellular vesicles and nanotubes — microenvironment. We discuss the ways in which
are now being recognized (Fig. 2). A distinguishing tumours recruit and subvert normal cells to change their
feature of glioblastomas is their ability to form a virtual phenotype to support tumour progression, suppress
nuclear and cytoplasmic continuum with neighbouring immune defences and defy therapeutic interventions.
cells, whereby they can introduce not only inorganic
elements but also genetic elements and proteins into Innate immune system and glioblastoma
normal cells to change their phenotype and rescue fellow Interaction between glioblastoma and microglia,
tumour cells that are in trouble, for example, as a result of monocytes or macrophages. The glioblastoma micro­
radiotherapy or chemotherapy. These newly recognized environment contains brain-​resident microglia and
transit routes can transmit non-​secretable molecules, infiltrating monocytes. Once monocytes have infiltrated
including transcription factors, directive RNAs and DNA the tumour, they can differentiate into macrophages26,27.
and even mitochondria and nuclei. Small molecules Although often grouped together under the term
such as Ca2+, ATP, metabolites and microRNAs (miRNAs) tumour-​a ssociated macrophages or myeloid cells
can be transferred between adjacent cells through (TAMs), these cells represent distinctly different pop-
gap junctions10,11. Connexins, which form a structural ulations26. Microglia are derived from immature yolk
component of these junctions, are upregulated in tumour- sac progenitors during early embryonic development
initiating cells12 and are associated with increased inva- and maintain themselves in the brain through self-​
siveness of gliomas11. Non-​secretable proteins (including renewal28,29. In non-​pathological settings, microglia
transcription factors), RNA, DNA, lipids and metabo- are the main innate immune cells in the brain and are
lites can be transferred through extracellular vesicles important in the defence against pathogens and nox-
released from cells via fusion of multivesicular bodies ious stimuli30. Glioblastoma leads to some disruption
with the cell membrane (which yields exosomes), bud- of the blood–brain barrier (BBB), which enables bone
ding from the plasma membrane (giving rise to microve- marrow haematopoietic stem cell-​derived monocytes
sicles and large oncosomes)13–15 or budding off the tips and macrophages to infiltrate the tumour26,27,31. Studies
of nanotubes that extend out from the cells16,17. These have shown that in specific cases up to 50% of the glio-
tumour-​derived extracellular vesicles can change the blastoma mass can consist of TAMs32. Chimeric and
phenotype of normal cells to promote angiogenesis, cell lineage models have shown that the exact com-
immune suppression, tumour cell invasion and metabolic position of TAMs changes over time26,31. One study
regulation 18–20. Tumour cells can also be linked by examined the infiltration of peripheral immune cells
‘tunnelling’ nanotubes and microtubes that form gap in a chimeric GL261 mouse glioma model that received
head-​protected irradiation, in which BBB disruption due
Author addresses to irradiation is avoided. Fluorescently tagged myeloid-​
derived monocytes and macrophages transplanted into
1
Department of Neurology and Center for Molecular Imaging Research, Department of these mice constituted up to 25% of TAMs in the glio-
Radiology, Massachusetts General Hospital and Program in Neuroscience, Harvard blastoma tumour after 21 days, with lower percentages of
Medical School, Boston, MA, USA. myeloid-​derived TAMs observed at earlier time points31.
2
Department of Neurosurgery, Brain Center Rudolf Magnus, Institute of Neurosciences,
The influx of myeloid-​derived monocytes in mouse glio-
University Medical Center, Heidelberglaan, Utrecht, Netherlands.
3
The Center for Immunology and Inflammatory Diseases and Department of Medicine, blastoma tumours was confirmed in a haematopoietic
Massachusetts General Hospital, Charlestown, MA, USA. stem cell lineage tracing model, in which >35% of TAMs
4
Program in Immunology, Harvard Medical School, Boston, MA, USA. were myeloid-​derived26. As such, the population of
5
Department of Neurology, Ann Romney Center for Neurologic Diseases, Initiative for RNA gliob­lastoma TAMs can progress from strictly microglial
Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, USA. in early phases to a mixture of microglia and infiltrating

NATURe RevIeWS | NeuRoloGy volume 14 | AUGUST 2018 | 483

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

Box 1 | Models for glioblastoma TAM activation state. Interaction between glioblastoma
cells and TAMs is multifactorial and occurs both in close
One strategy to improve models of glioblastoma has been to isolate cells from patient proximity by direct cell–cell contact and distantly by
tumours and maintain them as neurospheres or organoids in serum-​free medium such the release of soluble and membrane-​encased factors.
that they retain their genetic heterogeneity and tumour-​initiating cells (also known
This secretome consists of a multitude of molecules,
as cancer stem cells) when reimplanted into immune-​compromised mice158,159.
including soluble lipids, cytokines and chemokines33,37.
Tumour-​initiating cells represent a population of highly malignant tumour cells that
lurk in different vascular and hypoxic niches within the tumour and are able to Glioblastomas also release extracellular vesicles that
expand to reform malignant tumours after therapeutic intervention 160,161
. Authentic contain a cargo of many types of molecules that have
reproduction of the genetic and phenotypic properties of human glioblastoma can been shown to influence TAM status in a combinatorial
also be achieved in models by implanting and passaging portions of patient tumours way in culture and in vivo25,38. However, no techniques
in immune-​compromised mice, referred to as patient-​derived xenograft models162,163. currently are available to suppress extracellular vesicle
Other currently favoured models of glioblastoma include syngeneic mouse models, in release from glioblastomas specifically; therefore, the
which tumours are initially induced by chemicals or viruses and are then established overall relevance of the interaction between glioblas-
as cell lines that can be transplanted back into the mouse brain164. Spontaneous brain toma extracellular vesicles and TAMs remains to be
tumours can also be induced with known driver mutations in genetically engineered
elucidated. Ultimately, the combination and timing of all
mouse models . However, none of these models are a perfect representation of
165
glioblastoma-​released factors determine the activation
human glioblastoma. Neurospheroid cultures, patient-​derived xenograft models and
cell lines suffer from genetic instability , and cell lines frequently have low
166 state and function of TAMs.
invasiveness. Glioma-​derived cells also display a different genomic methylation The traditional model of the activation states of
pattern and transcriptome in culture and in vivo167. Genetically engineered mouse TAMs describes a binary system of either tumour-​
models represent only a few driver mutations and thus have few neoantigens. Given suppressive (M1) or tumour-​supportive (M2) macro­
the current focus of therapeutic research on alerting the immune system to phages39. This model was based on stimulation of cells
glioblastoma, use of more than one type of mouse model is advisable, and research in culture by IFNγ, lipopolysaccharide (LPS) or IL-4
should include syngeneic mouse glioma lines and genetically engineered mouse-​ and was later extended to include M2 subtypes acti-
derived glioma cells that can be grown in immune-​competent mice. Glioma cell lines vated by other types of stimulation, comprising M2a
that have been passaged extensively are genetically unstable168 and have
(IL-4 and IL-13), M2b (immune complexes, Toll-​like
immunological peculiarities169; hence, they do not represent reliable models of
receptor (TLR) or IL-1R) and M2c (IL-10)40. However,
human glioblastoma.
RNA sequencing extended the number of stimuli to a
combination of 28 known factors and revealed that a
monocytes and macrophages in late phases of tumour wide spectrum of activation states could be induced.
progression. In mice, accurate separation of microglia The findings demonstrated that macrophage differ-
and macrophages can be obtained by fluorescence-​ entiation is much more complex than the binary M1–
activated cell sorting using αM integrin (also known as M2 model41, even when stimulated in culture. This
CD11b) and receptor-​type tyrosine-​protein phosphatase complexity became more apparent when microglia,
C (also known as CD45) markers26. In humans, α4 integ- monocytes and macro­phages were isolated from glio-
rin (also known as CD49D) can accurately separate these blastoma in vivo and analysed by RNA sequencing.
two cell types in tumours26. Here, when studies used The most upregulated genes were found to be shared
these specific markers for separation of microglia and between traditional M1, M2a, M2b and M2c tran-
myeloid-​derived cells we refer to the cellular subpopulation scriptomes, suggesting that the activation state in vivo
studied, otherwise the generic term ‘TAMs’ is used. is very different from that in culture26,42,43. Single-​cell
sequencing confirmed that activation of both M1 and
TAM recruitment. The recruitment of TAMs to glioma M2 signatures can be observed even in individual cells
is mostly mediated by cytokine and chemokine gradients in an in vivo brain trauma model44. Consequently, the
released by glioblastoma cells (Fig. 3). These factors M1 and M2 designations are being replaced by more
have been extensively reviewed elsewhere and include precise situation-​specific models39. Altogether, these
CC-​chemokine ligand 2 (CCL2; also known as MCP1) findings suggest that TAMs express gene sets in vivo
and CCL7 (also known as MCP3), glial cell line-​derived that are associated with stimulation by different factors,
neurotrophic factor (GDNF), hepatocyte growth factor highlighting the variety of information transfer in the
(HGF), SDF1, tumour necrosis factor (TNF), VEGF, ATP, tumour microenvironment.
macrophage colony-​stimulating factor 1 (CSF1) and gran-
ulocyte–macrophage colony-​stimulating factor (GM– TAMs contribute to tumour proliferation. The role of
CSF)32,33. TAMs can also be recruited to a specific subset secreted molecules on TAM function and, subsequently,
of glioblastoma cells, such as oligodendrocyte trans­ on tumour growth has been studied extensively 32.
cription factor 2 (OLIG2)-expressing and transcription This interplay between glioblastoma cells and TAMs
factor SOX2-expressing tumour-​initiating cells, which is especially apparent in tissue remodelling and is nece­
secrete periostin to recruit TAMs34. Medical interven- s­s ­ary for glioblastoma cells to infiltrate the brain
tions can also stimulate TAM recruitment; for example, (Fig. 3). One group of proteins that is crucial in tissue
intracranial biopsies can increase infiltration of circulat- remodelling is matrix metalloproteinases (MMPs)45. In
ing monocytes into the tumour in a CCL2-dependent glioblastoma, MMP2 has an important role in ECM deg-
manner35. Microglia and macrophages themselves also radation, which facilitates glioblastoma cell migration
secrete CCL2 to increase infiltration of CCR2+Ly6C+ and invasion46. MMP2 is released in a precursor form
monocytes, thus creating a positive feedback loop for the (pro-​MMP2) that is cleaved by MMP14 to an active
continued infiltration of myeloid cells36. state32. However, glioblastoma cells secrete pro-​MMP2,

484 | AUGUST 2018 | volume 14 www.nature.com/nrneurol

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Reviews

Box 2 | Genetic and epigenetic heterogeneity of glioblastoma Neutrophils and mast cells. In addition to macrophages
and microglia, other innate immune cells, including
Deep sequencing of the genome and transcriptome together with study of the neutrophils and mast cells, are recruited into the tumour
epigenome of glioblastoma cells has revealed both genetic and epigenetic microenvironment51,52. The release of various chemo­
differences among tumour cells within the same glioblastoma, with many genetic kines and cytokines, such as IL-8, TNF and CCL2, by
drivers represented in almost every glioblastoma170,171. Evidence of the complexity
tumour cells or stromal cells after biopsy or resection can
of this disease can be found in the 2016 WHO classification system as well as in
experimental subclassification studies. The WHO classification system for diffuse recruit neutrophils to the tumour microenvironment35,53.
glioma now defines subtypes of intrinsic brain tumours that have a predictable Tumour-​associated neutrophils secrete cytokines that
prognosis; importantly, IDH1 or IDH2 mutation with chromosome 1p/19q co-​deletion further increase the infiltration of neutrophils and
confers a better outcome than other genetic subtypes do172. In addition, epigenetic monocytes and thus might indirectly support tumour
variation is present among tumour cells. Mutation of IDH1 or IDH2 results in altered growth by increasing TAM density at the glioma site54.
transcriptional regulation of many other genes through interference with Similarly, various chemokines produced by glio-
topologically associated domains, adding another dimension to the genetic blastoma cells, including SDF1, macrophage migration
complexity173. Experimental transcriptome classification has defined three subtypes inhibitory factor and plasminogen activator inhibitor 1,
in glioblastoma with wild-​type IDH1 and IDH2: proneural, classical and recruit mast cells to the tumour microenvironment52,55,56.
mesenchymal171. Glioblastoma driver mutations have been found in up to 45 genes174,
Mast cells are bone-​marrow-derived immune cells that
including common mutations in genes such as TP53, ATRX, TERT, NF1, PTEN and
EGFR172. As such, a strategy of hitting subsets of these genes with combinatorial play a part in innate and adaptive immune responses.
treatments will be difficult to achieve owing to the large number of genes175, although Upon activation by glioblastoma cells, mast cells produce
some pathways are more crucial than others. soluble factors, including IL-6, IL-8, VEGF and TNF,
that can facilitate tumour growth and angiogenesis57.
What role other forms of communication — including
but not MMP14. Conversely, microglia in the tumour extracellular vesicles, nanotubes and microtubes —
microenvironment are a major source of MMP14. Two have in the communication between tumour cells and
different glioblastoma-​derived factors act to increase neutrophils and mast cells remains to be investigated.
microglial MMP14 release38,47. First, the ECM protein An increasing body of evidence, therefore, suggests
versican is released from glioma and induces MMP14 that multiple communication loops in the tumour micro­
release by TAMs through its upstream receptor TLR2 environment are responsible for the recruitment and
(ref. 47) . Second, studies of cell-​c ulture models have activation of various cells of the innate immune system.
shown that glioblastoma-​derived extracellular vesicles An improved understanding of these complex interac-
can also induce microglial expression of MMP14 RNA, tions between glioblastoma and innate immune cells
although the mechanism and in vivo relevance remain could open novel therapeutic avenues for glioblastoma
to be elucidated38. via the blockade of these interactions.
Owing to their rapid growth, glioblastomas are
in constant need of neovascularization and release Adaptive immune response to brain tumours
angiogenic factors, such as EGF and VEGF 33 . Beyond their direct interactions with glioma cells
Additionally, in glioblastoma, microglia and mac- described above, innate immune cells also have
rophages accumulate around blood vessels and also essential roles in the activation of adaptive immune
produce the pro-​angiogenic chemokines VEGF and responses, which in turn can restrict glioma growth.
CXC-​chemokine ligand 2 (CXCL2)48. Furthermore, For a long time, the historic notion that the brain is
glioblastoma cells might promote angiogenesis indi- an immune-​privileged organ that is both invisible to
rectly through microglial cells, as CSF1 secreted by glio- the adaptive immune system and physically excludes
blastoma cells in vitro induces microglia cells to release migratory immune cells through the BBB has discour-
insulin-​like growth factor-​binding protein 1 (IGFBP1), aged the idea that CNS tumours, including glioblastoma,
which can induce angiogenesis49. RAGE (receptor for might be susceptible to spontaneous or therapeutically
advanced glycation end products; also known as AGER) induced adaptive antitumour immune responses. This
is thought to play a part in a number of diseases, includ- dogmatic view has largely eroded over the past few years,
ing tumours. In tumour-bearing mice, RAGE ablation partly owing to recognition of phenomena that conflict
increases survival by reducing the levels of VEGFA with this concept, including the fact that multiple scle-
secreted by infiltrating TAMs, which results in leaky rosis is driven by T cells58, that non-​CNS tumours can
(rather than fully developed) vasculature and disturbed induce paraneoplastic autoimmune reactions that affect
tumour perfusion50. However, these effects were reported the CNS59, that restriction of T cell trafficking can elicit
in syngeneic GL261 mouse tumours (a frequently used opportunistic CNS infections60 and that CNS antigens
cellular model of glioma), which do not represent the are detected by T cells in extracranial lymphoid tissues61.
invasive growth pattern observed in glioblastoma As a result, the traditional concept of passive immune
patients. Thus, TAMs have a crucial role in tumour privilege through physical isolation of the CNS from the
angiogenesis through multiple signalling mechanisms. adaptive immune system has given way to the concept of
Overall, the interaction between glioblastoma and active immune privilege that is dynamically maintained
TAMs is bidirectional and multifactorial. This plethora by mechanisms of immune regulation, which can be
of paracrine loops can determine the ultimate effects of revoked when necessary62. Efforts to understand the
TAMs on tumour growth and can differ depending on role of adaptive immune cells, and particularly T cells,
local variables such as hypoxia, the extent of necrosis, in the glioma tumour environment have consequently
TAM infiltration density and/or TAM activation state. intensified.

NATURe RevIeWS | NeuRoloGy volume 14 | AUGUST 2018 | 485

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

/QPQE[VGUCPF /KETQINKC /CUVEGNNU 6EGNNU

OCETQRJCIGU 5WRRQTVVWOQWT r4GETWKVGFVQVWOQWTD[5&( ±5WDRQRWNCVKQPUYKVJ
r4GETWKVGFVQVWOQWTD[ RTQITGUUKQPD['%/ r#EVKXCVGFD[VWOQWTFGTKXGF DQVJRTQVWOQWTCPF
E[VQMKPGCPFEJGOQMKPG TGOQFGNNKPICPF UVKOWNK CPVKVWOQWTRQVGPE[
ITCFKGPV KPFWEVKQPQH 2TQFWEGITQYVJGPJCPEKPICPF FGRGPFKPIQPVJG
5WRRQTV'%/ CPIKQIGPGUKU CPIKQIGPKEHCEVQTU EQORNGZCTTC[QH
TGOQFGNNKPICPFKPFWEG KOOWPGTGIWNCVQT[
CPIKQIGPGUKU OGEJCPKUOU

1NKIQFGPFTQE[VGU '%/
s+PJKDKVVWOQWTITQYVJ 5VKHHPGUUQH'%/ECP
VJTQWIJRCTCETKPG CHHGEVINKQOCRTQITGUUKQP
UKIPCNNKPIXKC9+( 6GPCUEKP%KP'%/
OQFWNCVGUINKQOC
$NQQFXGUUGNU UGETGVQOG
&GNKXGT[QHQZ[IGPCPF
PWVTKGPVU 0GWTQPU
5WRRQTVVWOQWTEGNN #HHGEVINKQOCITQYVJ
FKUUGOKPCVKQP CPFKPXCUKQPD[PGWTQPCN
2TQXKUKQPQHCDCTTKGT CEVKXKV[KPENWFKPI
VJCVTGUWNVUKP UGETGVKQPQH
EQORCTVOGPVCNK\CVKQP PGWTQVTCPUOKVVGTU
YKVJKPVJGDTCKP
#UVTQE[VGU
0GWVTQRJKNU 6TCPUHQTOKPVQ
5QWTEGQHE[VQMKPGUVQ PGQRNCUVKEINKQOCEGNNU
KPETGCUGVJGKPȰNVTCVKQPQH
OQPQE[VGU

Fig. 1 | Glioblastoma microenvironment. The glioblastoma environ consists of tumour cells, extracellular matrix (ECM),
blood vessels, innate immune cells (monocytes, macrophages, mast cells, microglia and neutrophils), T cells and non-​
tumorous neurons, astrocytes and oligodendrocytes. +, protumour function; –, antitumour function; ±, mixed protumour
and antitumour functions; SDF1, stromal cell-​derived factor 1; WIF1, WNT inhibitory factor 1.

Glioblastoma neoantigens and T cell activation. CNS to deep cervical lymph nodes and present brain
High-​grade glioma ranks in the bottom half of human tumour antigens, which induce expression of a traffick-
tumours in terms of mutational neoantigen load63, which ing receptor pattern by T cells that enables their homing
currently is considered a crucial determinant of an effec- to the CNS; by contrast, the same activity is not demon-
tive adaptive antitumour response. Nevertheless, T cell strated by APCs that originate from skin tumours. Thus,
infiltration of tumours positively correlates with clinical unique features of the brain tumour environment help
outcome in patients with glioblastoma64–68, suggesting — to educate APCs that then travel to deep cervical lymph
although not proving — that these cells can control nodes via lymphatic vessels to imprint T cells with
tumour growth. In contrast to innate immune cells, cells brain tropism61.
of the adaptive immune system (namely, T and B cells)
become effector cells that are capable of exerting anti­ Routes of T cells into the brain. Once primed in deep
tumour effects only through a dramatic expansion of small cervical lymph nodes, T cells enter the bloodstream and
populations of naive antigen-​specific precursor cells, have two principal access routes to the brain (Fig. 4b).
in conjunction with differentiation that equips them In the steady state, CC-​chemokine receptor 6 (CCR6)-
with immunological effector activities. With regards expressing T helper 17 cells can enter the cerebrospi-
to T cells, effector differentiation enables migration to nal fluid via the choroid plexus and percolate around
non-​lymphoid and tumour tissues. The activation pro- the brain surface, where they might re-​encounter their
cess occurs in secondary lymphoid organs, such as the cognate antigen on meningeal APCs. This interaction
lymph nodes and spleen, where resident and lymph-​ can trigger an initially inflammatory reaction that then
borne migratory dendritic cells that originate in periph- licenses local CNS microvessels to support recruitment
eral tissues present antigens to T cells (Fig. 4a). Cellular of additional T cells and other immune cells in a CCR6-
and molecular tracer studies revealed a functional con- independent way70. Whether this immune-​surveillance
nection between the CNS and the deep cervical lymph pathway is relevant for anti-​glioma responses is currently
nodes decades ago69, but a bona fide lymphatic drainage unknown. Alternatively, inflammation induced by the
system for the CNS was demonstrated only in rodents invasive growth of glioblastomas and by the factors these
and humans in 2015 (refs7,8). Mouse studies also sug- tumours produce could directly drive microvascular
gest that antigen-​presenting cells (APCs) traffic from the changes that support recruitment of T cells and innate

486 | AUGUST 2018 | volume 14 www.nature.com/nrneurol

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Reviews

C D Beyond TGFβ2, glioblastomas have been reported to

produce a wide range of membrane-​bound, secreted and
metabolic factors, as described for other tumours, includ-
ing Fas antigen ligand (FasL; also known as FASLG),
0CPQVWDG
)CRLWPEVKQP
programmed cell death 1 ligand 1 (PDL1), VEGFA,
indoleamine oxidase and others74–77. Glioblastoma can
'ZQUQOG
also induce TGFβ and IL-10 expression in pericytes,
‘O
‘O
which routinely interact with extravasating T cells and
/KETQXGUKENG reduce their expression of co-​stimulatory molecules78.
Both glioblastoma and pericyte-​d erived immune-​
0WENGWU 0WENGWU /KETQVWDG regulatory factors might act on antitumour effector
1PEQUQOG T cells through direct mechanisms or indirectly by
promoting the recruitment and function of tolerogenic
APCs and regulatory T (Treg) cells. Either way, these
factors produce a local state of reduced T cell effector
function commonly referred to as T cell exhaustion
or dysfunction79. This cellular differentiation state is
also observed in chronic viral infection and presuma­
bly serves to prevent immunopathological tissue
damage due to ongoing T cell activity against a virus,
Fig. 2 | Routes of communication between tumour cells and cells in their environs.
a | Gap junctions (24 nm in diameter) form across the adjacent membranes of cells that whereas in the tumour environment it restricts T cell
are in physical contact, enabling the passage of small molecules. Cells also release antitumour functions.
exosomes (50–100 μm) from multivesicular bodies that fuse with the plasma membrane.
In addition, microvesicles (100–200 μm) and even large oncosomes (1–10 μm) bud off T cell dysfunction in glioblastoma. One hallmark
from the plasma membrane and can interact with and be taken up by other cells. of T cell dysfunction is the co-​expression of multi-
b | Tunnelling nanotubes (50–200 nm in width and up to 5 μm in length) extend out from ple co-​inhibitory receptors, including programmed
cells and can either bud off vesicles at their tips or form gap junctions with other cells.
cell death protein 1 (PD1), T cell membrane protein 3
Microtubes extend out from tumour cells (1–2 μm in width and >500 μm in length) and (TIM3; also known as HAVCR2), lymphocyte activa-
can form gap junctions with other cells. tion gene 3 protein (LAG3) and T cell immunoreceptor
with immunoglobulin and ITIM domains (TIGIT)80.
immune cells via pre-​existing brain microvessels and Improved understanding of the mechanisms by which
neo-​vessels that develop in response to pro-​angiogenic expression of these receptors is induced in tumour-​
factors released by tumours. infiltrating T cells will be necessary, but they probably
include T cell receptor activation in the absence of suf-
Immune regulation in the brain. Upon entry into the ficient co-​stimulation, resulting in binding of nuclear
glioma microenvironment, tumour-​reactive T cells factor of activated T cell (NFAT) to exhaustion-​associated
are faced with a complex array of immune-​regulatory genes instead of effector genes81. Treg cells can downreg-
mechanisms (Fig. 4c). Most of these mechanisms are also ulate co-​stimulatory molecules on tumour-​associated
operative in non-​CNS tumours, but it is possible that APCs and thereby promote T cell exhaustion82. In addi-
the conditions underlying active immune privi­­lege in tion, TGFβ and VEGFA, which are both produced by the
the healthy CNS also support a particularly immune-​ glioma, amplify T cell receptor-​driven induction of PD1
suppressive environment in glioblastomas. TGFβ2, and other co-​inhibitory receptors83,84 and might explain
which was discovered under the name glioma-​derived their high expression in tumour-​infiltrating T cells.
T cell suppressor factor, was one of the earliest immu- Glioma cells and APCs in the glioma environment
nosuppressive factors found and was identified on the express co-​inhibitory receptor ligands, which engage
basis of its ability to suppress IL-2-dependent T cell sur- with T cells and attenuate functions triggered by
vival71. TGFβ2 and TGFβ1, the predominant forms of concurrent T cell receptor engagement, reducing the
TGFβ produced by immune cells, have remained centre antitumour activity of these cells. Expression of PDL1,
stage in glioma microenvironment research, not only one of the known ligands for PD1, is induced on tumour
because of their organization of the interplay between cells by IFNγ and therefore reflects ongoing antitumour
tumour cells and myeloid cells (as described above) immune activity85. In addition, oncogenic mutations,
but also because of their direct effects on T cells. CD8+ such as those causing a loss of PTEN function in glioma,
T cells rapidly induce expression of ITGAE (encoding can activate PDL1 expression in tumour cells86.
αE integrin), a gene for which the expression is strictly
TGFβ-​dependent, upon entry into the brain tumour Immune checkpoint inhibition. Use of blocking anti-
microenvironment in mice or humans, indicating bodies to interrupt binding of the co-​inhibitory recep-
that they are locally exposed to strong TGFβ signals72. tors PD1 and cytotoxic T lymphocyte protein 4 (CTLA4)
Although αE integrin expression improves retention of with their ligands — so-​called checkpoint blockade —
cytotoxic T lymphocytes in the brain, TGFβ also attenu­ has recently established itself as a powerful new modality
ates the ability of these cells to express crucial effector of cancer therapy. In a notable study from 2016, blockade of
molecules, such as granzyme B and IFNγ, which probably the PD1 and/or CTLA4 pathways eliminated estab-
reduces their efficacy against glioma73. lished mouse glioblastomas in the majority of treated

NATURe RevIeWS | NeuRoloGy volume 14 | AUGUST 2018 | 487

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

$TCKPTGUKFGPVOKETQINKC %KTEWNCVKPIOQPQE[VGU

%&DJKIJ%&JKIJ%&Fs %&DJKIJ%&JKIJ%&F

%[VQMKPGCPF
EJGOQMKPG #EVKXCVKQPCPFFKHHGTGPVKCVKQP
ITCFKGPV
%%.
%%.
)&0(
*)(
5&(
60(
4GETWKVOGPVQH6#/U 8')(
#62 /KZVWTGQHCEVKXCVKQPCPF
%5( 5GETGVQOG FKHHGTGPVKCVKQPUVCVGUYKVJKP
)/s%5( '8U VWOQWTGPXKTQPOGPV

'%/TGOQFGNNKPI
%5(
8GTUKECP '8U +PVGTCEVKQPDGVYGGPVWOQWTCPF6#/U
'8U

RTQ//2
8')(
%NGCXCIG ')(
//2
6#/OGFKCVGFCPIKQIGPGUKU
//2

'%/FGITCFCVKQP ')(
8')( #
%:%.
+)($2

Fig. 3 | Interactions between glioma and TAMs. Recruitment of tumour-​associated macrophages or myeloid cells (TAMs),
including blood monocytes and brain-​resident microglia, is based on the gradient of chemokines and cytokines released
by the glioblastoma cells. Once recruited, TAMs can be activated and differentiated under the influence of the secretome
and extracellular vesicles (EVs) released by the tumour. The various recruited and activated TAMs can affect tumour
growth by promoting angiogenesis through secretion of epidermal growth factor (EGF), vascular endothelial growth
factor (VEGF), CXC-​chemokine ligand 2 (CXCL2) and insulin-​like growth factor-​binding protein 1 (IGFBP1). This process is
further promoted by the release of tumour-​derived VEGF and EGF. Invasion and growth of the tumour are accomplished by
remodelling the extracellular matrix (ECM) surrounding the tumour. For example, versican and EVs from the tumour
induce the release of matrix metalloproteinase 14 (MMP14) by microglia. The release will facilitate the cleavage of tumour-​
derived pro-​MMP2 following extracellular degradation by the active enzyme MMP2. CCL2, CC-chemokine ligand 2 (also
known as MCP1); CSF1, macrophage colony-​stimulating factor 1; GDNF, glial cell line-​derived neurotrophic factor;
HGF, hepatocyte growth factor; SDF1, stromal cell-​derived factor 1; TNF, tumour necrosis factor.

animals and produced immune memory protective indicated a transient increase in progression-​free sur-
against tumour rechallenge87. However, patients with vival90, a phase III arm of the same trial failed to show
glioblastoma generally carry a lower neoantigen load an increase in overall survival91. In addition to low
than the mouse glioblastoma model used here; therefore, antigenicity, a low degree of inflammation and the
it remains to be determined whether CTLA4-targeted or overwhelming immune-​regulatory activity intrin-
PD1-targeted therapy will be effective in humans, given sic to the CNS might have contributed to this failure.
the correlation between neoantigen load and response Consequently, measures to enhance the basal immune
to either of these therapies88,89. Indeed, although phase I reactivity of glioblastoma might be needed to render
trial data on PD1 blockade in recurrent glioblastoma PD1-targeted therapy effective. However, irrespective

488 | AUGUST 2018 | volume 14 www.nature.com/nrneurol

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Reviews

C
%GTGDTQURKPCNȱWKF &WTCNN[ORJCVKEXGUUGNU
5MWNN
&WTCOCVGT 


6WOQWTCPVKIGP
)NKQDNCUVQOC

#2%


6EGNN FE.0U
D E
6TGIEGNN %6.#
+.


6)(β

!
&GPFTKVKEEGNN
6)(β
2&.CPF2&.
.QY%&CPF%&
%JQTQKF %6.
RNGZWU  /KETQINKC

6)(β
8')(#
+. 6#/
(CU.
+&1
/&5%
6)(β
+.

)NKQOC

Fig. 4 | T lymphocytes in the glioblastoma environment. a | A hypothetical scenario for the induction of
anti-glioblastoma T cell responses. Tumour antigens travel with interstitial fluid along the perivascular glymphatic
system to the brain surface (step 1), where they enter the cerebrospinal fluid or are taken up by local
antigen-presenting cells (APCs). A fraction of cerebrospinal fluid enters dural lymphatic vessels, potentially along
with migratory APCs (step 2), and by this route tumour antigen reaches the deep cervical lymph nodes (dcLNs). There,
resident or migratory APCs interact with and activate tumour antigen-​specific naive T cells (step 3). b | Potential
pathways by which activated or memory T cells reach the CNS and glioblastoma tumours. T cells expressing
CC-chemokine receptor 6 (CCR6; and possibly other chemokine receptors) enter the cerebrospinal fluid in ventricles
via the choroid plexus (step 4) and percolate along the brain surface. Upon re-​encounter with their cognate tumour
antigen (step 5), they initiate a local inflammatory reaction that activates the local microvasculature to recruit
additional T cells in a CCR6-independent and potentially CXC-​chemokine receptor 3 (CXCR3)-dependent fashion
(step 6), facilitating their accumulation in the glioblastoma environment. c | In and around the glioblastoma, effector
T cells, including cytotoxic T lymphocytes (CTLs), are exposed to a network of immune-​regulatory mechanisms
that promote their differentiation into a dysfunctional state. CTL A4, cytotoxic T lymphocyte protein 4; FasL , Fas
antigen ligand; IDO, indoleamine 2,3-dioxygenase; MDSC, myeloid-​derived suppressor cell; PDL1, programmed cell
death 1 ligand 1; TAM, tumour-​associated macrophages or myeloid cells; TGFβ, transforming growth factor-​β; Treg cell,
regulatory T cell; VEGFA , vascular endothelial growth factor A.

NATURe RevIeWS | NeuRoloGy volume 14 | AUGUST 2018 | 489

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

of the question of tumour antigenicity and despite CNS However, insight into environmental mechanisms
immune privilege, brain tumours are, in principle, sus- that regulate integrity of the vasculature and BBB in
ceptible to the effects of checkpoint blockade antibodies glioblastoma is currently limited and would benefit
and the enhanced antitumour T cell function they elicit. from improved, orthotopic animal models and drugs
This conclusion is also supported by data from humans that are designed to access the brain. BBB penetration
showing that brain metastasis of malignant melanoma is by chemotherapeutic drugs and antibodies currently
tractable to checkpoint inhibitor treatment92. remains a major limitation in treating glioblastomas105.

Endothelial cells and tumour vasculature Neurons

Glioblastoma growth is accompanied by angiogenesis, Although many patients with glioblastoma have epi-
and tumour vasculature has several key functions ena- leptic seizures, substantial cognitive dysfunctions,
bling tumour growth, including the delivery of oxygen memory impairment and personality changes, little is
and nutrients, modulation of the immune response, known about the interactions between glioblastoma and
support of tumour cell dissemination and provision surrounding cortical neurons. The effect of the tumour
of barriers for compartmentalization within the brain. on neuronal networks has been proposed to be due to
Notably, tumour-​initiating cells are usually located close mechanical pressure. However, considering the range of
to endothelial cells, in the so-​called perivascular niche, factors released by glioma and the complexity of inter-
which serves as a hub for generation of multiple cellular cellular signalling mechanisms, the effects are probably
phenotypes and has important functions in the main- much more complicated. The most profound neurotox-
tenance of tumour-​initiating cells, in communication icity and seizures are associated with excessive glutamate
between the tumour and vascular compartments and in release by glioma cells106. Interestingly, expression of glu-
therapeutic resistance (reviewed in depth elsewhere2,93). tamate transporters is increased in peritumoural cells,
which seems to have a neuroprotective function and
Factors promoting angiogenesis in the tumour provides some survival benefit in animal models107. In
microenvironment. Glioma cells release many pro-​ addition, recurrent seizures and the associated excessive
angiogenic factors that shape tumour vasculature, glutamate release might lead to increased vascular per-
including TGFβ, VEGF, proteolytic enzymes, ribonucle- meability through the activation of NMDA (N-​methyl-d-
ases (such as plasminogen activators and angiogenin) aspartate) receptors108. This permeability could result in
and chemokines. These factors are released in the forms increased peritumoural oedema, but it could also have a
of extracellular vesicles, soluble factors and extravesic- positive effect on drug delivery to the tumour.
ular ribonuclear complexes. For example, extracellular However, glioblastomas might have developed addi-
vesicle-​derived VEGFA has a pro-​angiogenic effect, at tional mechanisms of neurotoxicity that are unrelated to
least in culture94. In addition, glioma-​derived factors in glutamate release. Our analysis of the proteins secreted
the ECM, such as tenascin, dynamically modulate the by tumour-​initiating cells and the ribonuclear com-
glioma secretome and might also alter endothelial cell plexes released by glioblastoma cells identified multiple
proliferation and tubulogenesis, contributing to overall oncogenic and cell-​cycle-promoting factors109, includ-
angiogenesis95. An intriguing idea that highly plastic ing a cofactor of the DNA polymerase proliferating cell
tumour-​initiating cells might themselves transdifferen- nuclear antigen (PCNA) and a number of miRNAs that
tiate into endothelial cells and promote angiogenesis in are oncogenic in the brain microenvironment. These
glioblastoma96 has also received considerable support in miRNAs include miR-10b and miR-26, which drive
studies based on live-​cell imaging and patient tumour cell-​c ycle progression via interaction with cell-​c ycle
histopathology 97,98. Integrated transcriptomic and inhibitors (cyclin-​dependent kinase inhibitor 1 and 2)
epigenomic analysis has revealed that activation of the and retinoblastoma-​associated protein, respectively110,111.
WNT5a pathway can drive differentiation of tumour-​ These molecules promote tumour growth, but transfer of
initiating cells to endothelial-​like cells99. In this context, such proliferative signals to neurons can lead to neuro-
understanding how the tumour microenvironment degeneration112. For example, increased levels of miR-26
modulates this process will be important. cause aberrant cell-​cycle entry in postmitotic neurons that
leads to neuronal cell death113. Intercellular transfer of
Vascular integrity. Normalization of the tumour vascu- such molecules from glioblastoma to neurons could affect
lature integrity, which is severely compromised in glio- neuronal health, postmitotic state and overall cell viability.
blastoma, represents one therapeutic strategy to improve Although mitogenic signalling from malignant cells to
drug delivery100. Invading glioblastoma cells cause phys- adjacent normal cells is expected, the glioma-​supportive
ical dislocation of astrocytes from endothelial cells and signalling from neurons to glioma is, perhaps, more
thus disrupt gliovascular coupling of the BBB101. VEGF surprising. Several reports suggest that neurotransmit-
can also increase access of drugs to glioblastoma across ters can affect glioma growth and invasion114–116, and
the BBB102. In addition, extracellular vesicles released seminal work in the past few years from Monje and col-
by glioblastoma can disrupt the endothelial barrier leagues has revealed a novel and unexpected mechanism
via delivery of semaphorin 3A, among other factors103. through which neuronal activity supports glioblastoma
Interestingly, vascular integrity can also be modulated growth117,118. Using an optogenetics-based approach, these
remotely by delivery of miR-132 in extracellular vesi- researchers found that firing activity of cortical projec-
cles released from neurons, via the indirect upregula- tion neurons promotes glioma proliferation and growth
tion of the adherens junction protein, VE-​cadherin104. in vivo via regulated secretion of the synaptic protein

490 | AUGUST 2018 | volume 14 www.nature.com/nrneurol

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Reviews

neuroligin 3. This postsynaptic adhesion molecule is injury can cause transcriptome and secretome alter-
secreted by neurons and oligodendrocyte precursor cells ations in reactive astrocytes that further potentiate
and is found mostly in excitatory synapses. Its actions tumour aggressiveness5,125.
include downstream signalling, activation of the focal Signalling between glioma cells and reactive peritu-
adhesion kinase (FAK; also known as PTK)–phosphoinos- moural astrocytes in vivo has yet to elucidated, largely
itide 3-kinase (PI3K)–mammalian target of rapamycin owing to the lack of highly specific markers discriminating
(mTOR) pathway, growth factor receptors and oncogenic these cell populations, but studies of glioblastoma biology
proteins and feedforward expression of neuroligin 3 and the secretome present some intriguing possibilities.
in glioma cells. Remarkably, growth of glioblastoma For example, malignant transformation of astrocytes
xenografts from adult and paediatric patients, as well requires expression of the oncogenic miR-10b, which
as xeno­grafts from diffuse intrinsic pontine glioma, was is normally silenced in these cells126. miR-10b is highly
strongly impaired in Nlgn3-knockout mice. Monje and abundant in all glioblastoma subtypes and is released in
colleagues propose that blockade of neuroligin 3 release, extracellular vesicles from tumour cells. Transfer of this
mediated largely by the membrane protease ADAM10 and other pro-​oncogenic regulators, such as miR-21 and
(disintegrin and metalloproteinase domain-​containing miR-26, from glioblastoma to the peritumoural astro-
protein 10), could represent a therapeutic approach for cytes could facilitate their transformation. Interestingly,
glioblastoma118. Additional factors regulated by neu- a greater density of astrocytes adjacent to the tumour is
ronal activity, such as brain-​derived neurotrophic factor predictive of a shorter lifespan in glioblastoma patients127.
and endoplasmic reticulum chaperone BiP (GRP78; also This finding supports the concept that the tumour-​
known as HSPA5), have also been identified that have associated astrocytes are coerced into promoting tumour
central roles in normal synaptic functions and can act growth and facilitating tumour invasion128,129.
as glioma mitogens in the tumour microenvironment117, By contrast, endogenous oligodendrocytes seem to
highlighting the importance of neuronal-​to-glioblastoma inhibit glioblastoma growth and proliferation by par-
communication. Generally, dissection of molecular mech- acrine signalling via WNT inhibitory factor 1 (ref.130).
anisms that couple the activity of neurons and synapses Although less is known about interactions between
with the behaviour of neighbouring glioma cells, includ- glioma and oligodendrocytes than astrocytes, the
ing tumour-​initiating cells and invading glioma cells, could presence of crosstalk between oligodendrocytes, micro-
lead to a new class of selective and targeted anti-​glioma glia and astrocytes through multiple cytokines131 suggests
therapies that spare normal neuronal functions. a complex relationship with these endogenous cells and
glioblastoma that can include both tumour-​promoting
Astrocytes and other glial populations and tumour-​suppressing signals. Furthermore, a subtype
Astrocytes are the endogenous native cells of the brain of reactive astrocytes has been identified that is induced
that are phenotypically most similar to the bulk of glio- by activated microglia and promotes the death of neurons
blastoma and represent one of the cell types from which and oligodendrocytes132, which points to the multifaceted
gliomas can originate. Astrocytes can be transformed interactions within the glioma microenvironment.
to neoplastic glioma cells or tumour-​initiating cells
in vitro and in vivo by a variety of oncogenes, such as Perspectives and therapeutic insights
MYC, RAS and EGFR variant III (refs119,120), which raises The many new insights into the communication between
the question of whether the communication between glioblastoma cells and other cells within the tumour
glioblastoma and surrounding reactive astrocytes can microenvironment, as outlined in this Review, reveal
promote the process of transformation and thereby new strategies for potential therapies for glioma. These
feed the tumour with newly recruited oncogenic cells. include therapies that target modes of communication
Although the exact mechanism is unknown, media between glioblastoma and surrounding cells, as well as
conditioned by glioma cells inhibits expression of the therapies that directly or indirectly target molecules with
p53 tumour suppressor in normal astrocytes, which in tumour-​supportive properties, including factors involved
turn enriches laminin and fibronectin in the ECM and in the innate and adaptive immune system, angiogenesis
promotes glioma cell survival121. Studies show that the and interaction with normal cells, such as neurons.
inevitable recurrence of the tumour could arise from One mode of communication that is increasingly
quiescent tumour-​initiating cells, infiltrating tumour recognized as a potential method for glioblastoma cells
cells and/or recruitment of apparently normal cells from to influence the microenvironment is tumour-​derived
the peritumoural brain zone122. In support of the latter extracellular vesicles. The contents of extracellular ves-
origin of cells, genes involved in growth, proliferation icles derived from tumour-​initiating cells, for exam-
and motility are induced in peritumoural white mat- ple, have been shown to include various oncogenic
ter that is free of tumour cells, whereas several tumour and cell-​c ycle-promoting factors, including PCNA
suppressors and genes involved in neurogenesis are and miRNAs109. Future studies could vali­date their
downregulated relative to normal white matter123. functional effects on different cell types in the tumour
Indeed, co-​implantation of tumour-​associated glial microenvironment by interfering with extracellular-​
cells (that is, brain-​derived glial cells conditioned by glio- vesicle-mediated communication, for example, by
blastoma xenografts) and glioblastoma cells promoted influencing extracellular vesicle generation or uptake.
tumour growth compared with implantation of tumour Extracellular vesicle release can be reduced by GW4869
cells alone or tumour cells with unconditioned glial — a neutral sphingomyelinase inhibitor that prevents the
cells124. Furthermore, resection and radiation-​induced ceramide-​mediated inward budding of multivesicular

NATURe RevIeWS | NeuRoloGy volume 14 | AUGUST 2018 | 491

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

bodies and the release of mature exosomes into the showed that CSF1R inhibition alone in the recurrent
extracellular space133. GW4869 has been shown to reduce setting had little to no effect; however, the activation of
the levels of extracellular vesicles in various systems, PI3K signalling in many glioblastomas strongly suggests
including in a mouse model of Alzheimer disease, where the effectiveness of CSF1R and IGF1 antagonists as a
it reduced the number of exosomes in both serum and combination therapy145,146.
brain134–136. On the other hand, uptake of extracellular Treatments aimed at ameliorating the adaptive
vesicles by recipient cells can be reduced by heparin137 immune response have been successful in some tumours,
or by silencing of annexin A1 (ref.138). However, these but their effect in glioblastoma has been very
strategies have been applied only in preclinical studies modest, partly owing to the low neoantigen load
thus far and do not distinguish between tumour-​specific of glioblastoma. Thus far, no clinical benefit has
extracellular vesicle interactions and physiological pro- been demon­strated by large clinical trials on dendritic
cesses related to extracellular vesicles. Much work will cell vacci­nation aiming to elicit anti-​glioblastoma T cell
be needed to explore their potential for specific, effective responses; however, overall survival was improved
and safe routine clinical applications. upon vaccination with dendritic cells loaded with
Further strategies for glioma therapy include target- tumour lysate in a small subset of nine patients
ing other mechanisms of communication within the with glioblastoma who exhibited a dominantly mes-
tumour microenvironment, such as cytokines, gap junc- enchymal gene signature, indicating that some forms
tions and nanotubes or microtubes. Inhibition of certain of glioblastoma might be more susceptible to immu-
cytokines has shown promise for glioblastoma treatment notherapy than others 147. Although the failure of
in mouse models, including CCL2, which reversed the immune checkpoint therapy to improve overall sur-
induction of migration and proliferation of tumour vival in patients with recurrent glioblastoma has been
cells observed after biopsy35. In addition, recruitment disappointing (such as in the CheckMate-143 trial148),
of CCR4+ Treg cells mediated by CCL2 can be blocked multiple avenues remain to explore the potential util-
using an antagonist against CCR4 and results in an ity of this approach, especially in the setting of newly
improvement in the median survival of tumour-​bearing diagnosed disease. Phase III trials on PD1 blockade
animals36. The gap junctions formed in glioblastomas for newly diagnosed glioblastoma patients are con-
between astrocytes and tumour cells via connexin 43 tinuing (CheckMate-498 (ref.149) and CheckMate-548
support growth and invasion23 and can be modulated (ref.150)). Similar to tumours with a mesenchymal gene
using the cyclooxygenase inhibitors meclofenamate signature, hypermutant glioblastoma in paediatric
or tonabersat138. These cyclooxygenase inhibitors have patients with germline biallelic mismatch repair defi-
been shown to reduce established brain metastases ciency might constitute a subset of cases with enhanced
in an experimental model139. Therapies that alter the responsiveness to this form of therapy151. However,
structure of the plasma membrane — for example, focal in most cases, additional interventions might be
hyperthermia (in combination with radiation therapy), needed to sensitize cells to immune checkpoint ther-
which has gained clinical attention, with several trials apy. Interestingly, a preclinical study of mouse glio-
under way — could also suppress nanotubule and blastoma showed that intracranial treatment with an
microtubule formation and even extracellular vesicle IL-12-expressing oncolytic herpes simplex virus gave
release. In addition, the ECM is a crucial site for extra- a highly effective response when combined with anti-​
cellular communication in the tumour environment. PD1 and anti-​CTLA4 therapy152. Effectiveness might
Cell-​culture studies have shown that the stiffness of the also be increased by including inflammatory agents,
ECM can affect glioma progression, and interfering with such as oncolytic adenovirus and herpes simplex virus
this substrate might provide an effective therapy beyond vectors, with immune checkpoint inhibitors153. Overall,
direct targeting of the tumour cells140. One approach PD1 blockage in glioblastoma tumours in humans has
would be to block ECM remodelling using inhibitors of not resulted in increased survival, but strategies to
extracellular MMPs141 and heparanases142. increase efficiency or select favourable patients might
Over the past few years, interactions between glio- yield more encouraging results.
blastoma cells and cells of the immune system have Treatments targeting the tumour vasculature, espe-
gained attention as targets for novel treatment options. cially antiangiogenic therapies such as bevacizumab,
One interesting strategy aimed at modulating the com- have been studied extensively (reviewed in detail
munication between the tumour and cells of the innate elsewhere154). However, these trials did not show the
immune system involves inhibition of the CSF1 receptor much hoped for undisputed improvement in overall
(CSF1R), which is expressed on TAMs. For example, this survival. Consequently, an alternative strategy aiming
inhibition decreased the tumour-​supportive capacity to normalize tumour vasculature for improved drug
of TAMs and increased survival in a mouse proneural delivery has received attention. Dual inhibition of the
glioblastoma model143. Unfortunately, subsequent long­ VEGF receptor and angiogenin 2 normalized tumour
itudinal studies have shown that resistance to CSF1R vasculature and prolonged survival in two glioblastoma
blockage is acquired as a result of macrophage-induced mouse models109. In a different model, angiopoietin 1
secretion of insulin growth factor 1 (IGF1), leading to receptor activation and angiogenin 2 inhibition nor-
PI3K activation in glioblastoma cells, which again pro- malized the vasculature, elicited a favourable tumour
motes tumour growth144. Co-​treatment with CSF1R and microenvironment (including an altered immune cell
IGF1 antagonists in a mouse model of glioma reduced profile) and improved the delivery of chemotherapeu-
the levels of resistance substantially144. A clinical trial tic agents into tumours155. The therapeutic potential of

492 | AUGUST 2018 | volume 14 www.nature.com/nrneurol

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Reviews

antiangiogenic drugs, therefore, has the potential to be Conclusions

reinvigorated; however, current means have yet to Glioblastomas represent a complex set of tumour cells
be successful. that differ greatly in genetic and epigenetic character-
Finally, dissection of the mechanisms linking the istics and are surrounded by endogenous cells that are
activity of neurons and other endogenous cells in their transformed into predominantly tumour-​supporting
environs with the behaviour of glioma cells could pro- stromal cells through various modes of bidirectional
vide new opportunities and might result in a new class communication. This rampant multimodal network-
of selective and targeted anti-​g lioma therapies. For ing in the tumour environment offers targets for new
example, release of neuroligin 3, which is cleaved from therapies yet simultaneously highlights the challenges
neurons and oligodendrocytes via ADAM10, promotes ahead. Microenvironmental compensatory mecha-
glioma proliferation117. Furthermore, tumour growth nisms, such as stromal IGF1 secretion after CSF1R
was markedly reduced in a paediatric glioblastoma blockade, demonstrate the need for extended and
and diffuse pontine glioma model using the ADAM10 repeated interventions to overcome dynamic changes
inhibitor GI254023X 118. The ADAM10 inhibitors in tumour growth incentives. These mechanisms also
INCB7839 and XL-784 have been tested in clinical trials underscore the need to undercut the dependency by glio-
for breast cancer and have been shown to penetrate the blastoma on its environs in the development of much-​
BBB156,157. Consequently, further studies are warran­ needed novel therapies for this inevitably lethal disease.
ted to explore inhibition of this neuronal support of
tumour growth. Published online 9 July 2018

1. Stupp, R. et al. Effects of radiotherapy with cell-derived extracellular vesicles. Int. Rev. Cell. Mol. 35. Alieva, M. et al. Preventing inflammation inhibits
concomitant and adjuvant temozolomide versus Biol. 326, 175–199 (2016). biopsy-​mediated changes in tumor cell behavior.
radiotherapy alone on survival in glioblastoma in a 19. D’Asti, E., Chennakrishnaiah, S., Lee, T. H. & Rak, J. Sci. Rep. 7, 7529 (2017).
randomised phase III study: 5-year analysis of the Extracellular vesicles in brain tumor progression. 36. Chang, A. L. et al. CCL2 produced by the
EORTC-​NCIC trial. Lancet Oncol. 10, 459–466 (2009). Cell. Mol. Neurobiol. 36, 383–407 (2016). glioma microenvironment is essential for
2. Jhaveri, N., Chen, T. C. & Hofman, F. M. Tumor 20. Redzic, J., Balaj, L., van der Vos, K. & Breakefield, X. O. the recruitment of regulatory T cells and
vasculature and glioma stem cells: contributions to Extracellular RNA mediates and marks cancer myeloid-​derived suppressor cells. Cancer Res.
glioma progression. Cancer Lett. 380, 545–551 progression. Semin. Cancer Biol. 28, 14–23 (2014). 76, 5671–5682 (2016).
(2016). 21. Wang, X., Veruki, M. L., Bukoreshtliev, N. V., Hartveit, E. 37. Wurdinger, T., Deumelandt, K., van der Vliet, H. J.,
3. See, A. P., Parker, J. J. & Waziri, A. The role of & Gerdes, H. H. Animal cells connected by nanotubes Wesseling, P. & de Gruijl, T. D. Mechanisms
regulatory T cells and microglia in glioblastoma-​ can be electrically coupled through interposed of intimate and long-​distance cross-​talk between
associated immunosuppression. J. Neurooncol. 23, gap-junction channels. Proc. Natl Acad. Sci. USA 107, glioma and myeloid cells: how to break a vicious
405–412 (2015). 17194–17199 (2010). cycle. Biochim. Biophys. Acta 1846, 560–575
4. Roesch, S., Rapp, C., Dettling, S. & Herold-​Mende, C. 22. Vignais, M. L., Caicedo, A., Brondello, J. M. & (2014).
When immune cells turn bad-​tumor-associated Jorgensen, C. Cell connections by tunneling 38. de Vrij, J. et al. Glioblastoma-​derived extracellular
microglia/macrophages in glioma. Int. J. Mol. Sci. 19, nanotubes: effects of mitochondrial trafficking on vesicles modify the phenotype of monocytic cells.
E436 (2018). target cell metabolism, homeostasis, and response to Int. J. Cancer 137, 1630–1642 (2015).
5. Okolie, O. et al. Reactive astrocytes potentiate tumor therapy. Stem Cells Int. 2017, 6917941 (2017). 39. Ransohoff, R. M. A polarizing question: do M1 and
aggressiveness in a murine glioma resection and 23. Osswald, M. et al. Brain tumour cells interconnect to a M2 microglia exist? Nat. Neurosci. 19, 987–991
recurrence model. Neuro Oncol. 18, 1622–1633 functional and resistant network. Nature 528, 93–98 (2016).
(2016). (2015). 40. Mantovani, A., Sozzani, S., Locati, M., Allavena, P. &
6. Pencheva, N. et al. Identification of a druggable 24. Weil, S. et al. Tumor microtubes convey resistance to Sica, A. Macrophage polarization: tumor-​associated
pathway controlling glioblastoma invasiveness. surgical lesions and chemotherapy in gliomas. Neuro macrophages as a paradigm for polarized M2
Cell Rep. 20, 48–60 (2017). Oncol. 19, 1316–1326 (2017). mononuclear phagocytes. Trends Immunol. 23,
7. Aspelund, A. et al. A dural lymphatic vascular system 25. van der Vos, K. E. et al. Directly visualized 549–555 (2002).
that drains brain interstitial fluid and macromolecules. glioblastoma-​derived extracellular vesicles transfer 41. Xue, J. et al. Transcriptome-​based network analysis
J. Exp. Med. 212, 991–999 (2015). RNA to microglia/macrophages in the brain. Neuro reveals a spectrum model of human macrophage
8. Louveau, A. et al. Structural and functional features of Oncol. 18, 58–69 (2016). activation. Immunity 40, 274–288 (2014).
central nervous system lymphatic vessels. Nature 26. Bowman, R. L. et al. Macrophage ontogeny 42. Gabrusiewicz, K. et al. Characteristics of the
523, 337–341 (2015). underlies differences in tumor-​specific education alternative phenotype of microglia/macrophages and
9. Boussiotis, V. A. & Charest, A. Immunotherapies for in brain malignancies. Cell Rep. 17, 2445–2459 its modulation in experimental gliomas. PLoS ONE 6,
malignant glioma. Oncogene 15, 1121–1141 (2017). (2016). e23902 (2011).
10. Thuringer, D. et al. Transfer of functional microRNAs 27. Chen, Z. et al. Cellular and molecular identity of 43. Szulzewsky, F. et al. Glioma-​associated microglia/
between glioblastoma and microvascular endothelial tumor-​associated macrophages in glioblastoma. macrophages display an expression profile different
cells through gap junctions. Oncotarget 7, Cancer Res. 77, 2266–2278 (2017). from M1 and M2 polarization and highly express
73925–73934 (2016). 28. Ajami, B., Bennett, J. L., Krieger, C., Tetzlaff, W. & Gpnmb and Spp1. PLoS ONE 10, e0116644
11. Hong, X., Sin, W. C., Harris, A. L. & Naus, C. C. Rossi, F. M. V. Local self-​renewal can sustain CNS (2015).
Gap junctions modulate glioma invasion by direct microglia maintenance and function throughout adult 44. Kim, C. C., Nakamura, M. C. & Hsieh, C. L. Brain
transfer of microRNA. Oncotarget 6, 15566–15577 life. Nat. Neurosci. 10, 1538–1543 (2007). trauma elicits non-​canonical macrophage activation
(2015). 29. Ginhoux, F. et al. Fate mapping analysis reveals that states. J. Neuroinflamm. 13, 117 (2016).
12. Balça-​Silva, J. et al. The expression of connexins and adult microglia derive from primitive macrophages. 45. Kessenbrock, K., Plaks, V. & Werb, Z. Matrix
SOX2 reflects the plasticity of glioma stem-​like cells. Science 330, 841–845 (2010). metalloproteinases: regulators of the tumor
Transl Oncol. 10, 555–569 (2017). 30. Hickman, S. E. et al. The microglial sensome revealed microenvironment. Cell 141, 52–67 (2010).
13. Tkach, M. & Théry, C. Communication by extracellular by direct RNA sequencing. Nat. Neurosci. 16, 46. Du, R. et al. Matrix metalloproteinase-2 regulates
vesicles: where we are and where we need to go. Cell 1896–1905 (2013). vascular patterning and growth affecting tumor cell
164, 1226–1232 (2016). 31. Müller, A., Brandenburg, S., Turkowski, K., Müller, S. survival and invasion in GB. Neuro Oncol. 10,
14. Maas, S. L., Breakefield, X. O. & Weaver, A. M. & Vajkoczy, P. Resident microglia, and not peripheral 254–264 (2008).
Extracellular vesicles: unique intercellular delivery macrophages, are the main source of brain tumor 47. Hu, F. et al. Glioma-​derived versican promotes tumor
vehicles. Trends Cell Biol. 27, 172–188 (2017). mononuclear cells. Int. J. Cancer 137, 278–288 expansion via glioma-​associated microglial/
15. Minciacchi, V. R. et al. Large oncosomes contain (2015). macrophages Toll-​like receptor 2 signaling. Neuro
distinct protein cargo and represent a separate 32. Hambardzumyan, D., Gutmann, D. H. & Kettenmann, H. Oncol. 17, 200–210 (2015).
functional class of tumor-​derived extracellular vesicles. The role of microglia and macrophages in glioma 48. Brandenburg, S. et al. Resident microglia rather than
Oncotarget 6, 11327–11341 (2015). maintenance and progression. Nat. Neurosci. 19, peripheral macrophages promote vascularization in
16. Rilla, K. et al. Hyaluronan production enhances 20–27 (2016). brain tumors and are source of alternative pro-​
shedding of plasma membrane-​derived microvesicles. 33. Li, W. & Graeber, M. B. The molecular profile of angiogenic factors. Acta Neuropathol. 131, 365–378
Exp. Cell Res. 319, 2006–2018 (2013). microglia under the influence of glioma. Neuro Oncol. (2016).
17. Lai, C. P. et al. Dynamic biodistribution of extracellular 14, 958–978 (2012). 49. Nijaguna, M. B. et al. Glioblastoma-​derived
vesicles in vivo using a multimodal imaging reporter. 34. Zhou, W. et al. Periostin secreted by glioblastoma stem macrophage colony-​stimulating factor (MCSF) induces
ACS Nano 8, 483–494 (2014). cells recruits M2 tumour-​associated macrophages and microglial release of insulin-​like growth factor-​binding
18. Fonseca, P., Vardaki, I., Occhionero, A. & Panaretakis, T. promotes malignant growth. Nat. Cell Biol. 17, protein 1 (IGFBP1) to promote angiogenesis. J. Biol.
Metabolic and signaling functions of cancer 170–182 (2015). Chem. 290, 23401–23415 (2015).

NATURe RevIeWS | NeuRoloGy volume 14 | AUGUST 2018 | 493

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

50. Chen, X. et al. RAGE expression in tumor-​associated 75. Berghoff, A. S. et al. Programmed death ligand 1 101. Watkins, S. et al. Disruption of astrocyte-​vascular
macrophages promotes angiogenesis in glioma. expression and tumor-​infiltrating lymphocytes coupling and the blood-​brain barrier by invading
Cancer Res. 74, 7285–7297 (2014). in glioblastoma. Neuro Oncol. 17, 1064–1075 (2015). glioma cells. Nat. Commun. 5, 4196 (2014).
51. Fossati, G. et al. Neutrophil infiltration into human 76. Wainwright, D. A. et al. IDO expression in brain 102. Wen, L. et al. VEGF-​mediated tight junctions
gliomas. Acta Neuropathol. 98, 349–354 (1999). tumors increases the recruitment of regulatory T cells pathological fenestration enhances doxorubicin-​
52. Põlajeva, J. et al. Mast cell accumulation in and negatively impacts survival. Clin. Cancer Res. 18, loaded glycolipid-​like nanoparticles traversing BBB for
glioblastoma with a potential role for stem cell factor 6110–6121 (2012). glioblastoma-​targeting therapy. Drug Deliv. 24,
and chemokine CXCL12. PLoS ONE 6, e25222 (2011). 77. Bao, S. et al. Glioma stem cells promote 1843–1855 (2017).
53. Kolaczkowska, E. & Kubes, P. Neutrophil recruitment radioresistance by preferential activation of the DNA 103. Treps, L. et al. Extracellular vesicle-​transported
and function in health and inflammation. Nat. Rev. damage response. Nature 444, 756–760 (2006). semaphorin3A promotes vascular permeability in
Immunol. 13, 159–175 (2013). 78. Valdor, R. et al. Glioblastoma progression is assisted glioblastoma. Oncogene 35, 2615–2623 (2016).
54. Sionov, R. V., Fridlender, Z. G. & Granot, Z. The by induction of immunosuppressive function of 104. Xu, B. et al. Neurons secrete miR-132-containing
multifaceted roles neutrophils play in the tumor pericytes through interaction with tumor cells. exosomes to regulate brain vascular integrity.
microenvironment. Cancer Microenviron. 8, 125–158 Oncotarget 8, 68614–68626 (2017). Cell Rep. 27, 882–897 (2017).
(2015). 79. Speiser, D. E., Ho, P. C. & Verdeil, G. Regulatory 105. Miller, J. J. & Wen, P. Y. Emerging targeted therapies
55. Põlajeva, J. et al. Glioma-​derived macrophage circuits of T cell function in cancer. Nat. Rev. Immunol. for glioma. Expert Opin. Emerg. Drugs 21, 441–452
migration inhibitory factor (MIF) promotes mast cell 16, 599–6110 (2016). (2016).
recruitment in a STAT5-dependent manner. 80. Wherry, E. J. & Kurachi, M. Molecular and cellular 106. Buckingham, S. C. et al. Glutamate release by primary
Mol. Oncol. 8, 50–58 (2014). insights into T cell exhaustion. Nat. Rev. Immunol. 15, brain tumors induces epileptic activity. Nat. Med. 17,
56. Roy, A. et al. Glioma-​derived plasminogen activator 486–499 (2015). 1269–1274 (2011).
inhibitor-1 (PAI-1) regulates the recruitment of LRP1 81. Martinez, G. J. et al. The transcription factor NFAT 107. Sattler, R. et al. Increased expression of glutamate
positive mast cells. Oncotarget 6, 23647–23661 promotes exhaustion of activated CD8+ T cells. transporter GLT-1 in peritumoral tissue associated
(2015). Immunity 42, 265–278 (2015). with prolonged survival and decreases in tumor
57. Attarha, S., Roy, A., Westermark, B. & Tchougounova, E. 82. Bauer, C. A. et al. Dynamic Treg interactions with growth in a rat model of experimental malignant
Mast cells modulate proliferation, migration and intratumoral APCs promote local CTL dysfunction. glioma. J. Neurosurg. 119, 878–886 (2013).
stemness of glioma cells through downregulation of J. Clin. Invest. 124, 2425–2450 (2014). 108. Vazana, U. et al. Glutamate-​mediated blood-​brain
GSK3β expression and inhibition of STAT3 activation. 83. Park, B. V. et al. TGFβ1-mediated SMAD3 enhances barrier opening: implications for neuroprotection and
Cell. Signal. 37, 81–92 (2017). PD-1 expression on antigen-​specific T cells in cancer. drug delivery. J. Neurosci. 36, 7727–7739 (2016).
58. Compston, A. & Coles, A. Multiple sclerosis. Lancet Cancer Discov. 6, 1366–1381 (2016). 109. Wei, Z. et al. Full-​coverage landscape of extracellular
372, 1502–1517 (2008). 84. Voron, T. et al. VEGF-​A modulates expression of RNAs, coding and non-​coding, released by human
59. Dalmau, J. & Rosenfeld, M. R. Paraneoplastic inhibitory checkpoints on CD8+ T cells in tumors. glioma stem cells. Nat. Commun. https://doi.org/
syndromes of the CNS. Lancet Neurol. 7, 327–340 J. Exp. Med. 212, 139–148 (2015). 10.1038/s41467-017-01196-x (2017).
(2008). 85. Spranger, S. et al. Up-​regulation of PD-​L1, IDO, and 110. Kim, H. et al. Integrative genome analysis reveals an
60. Berger, J. R. & Koralnik, I. J. Progressive multifocal T(regs) in the melanoma tumor microenvironment is oncomir/oncogene cluster regulating glioblastoma
leukoencephalopathy and natalizumab—unforeseen driven by CD8(+) T cells. Sci. Transl Med. 5, 200ra116 survivorship. Proc. Natl Acad. Sci. USA 107,
consequences. N. Engl. J. Med. 353, 414–416 (2013). 2183–2188 (2010).
(2005). 86. Parsa, A. T. et al. Loss of tumor suppressor PTEN 111. Teplyuk, N. M. et al. MicroRNA-10b inhibition reduces
61. Calzascia, T. et al. Homing phenotypes of tumor- function increases B7-H1 expression and E2F1-mediated transcription and miR-15/16 activity
specific CD8 T cells are predetermined at the tumor immunoresistance in glioma. Nat. Med. 13, 84–88 in glioblastoma. Oncotarget 6, 3770–3783 (2015).
site by crosspresenting APCs. Immunity 22, 175–184 (2007). 112. Herrup, K. & Yang, Y. Cell cycle regulation in the
(2005). 87. Reardon, D. A. et al. Glioblastoma eradication postmitotic neuron: oxymoron or new biology? Nat.
62. Galea, I., Bechmann, I. & Perry, V. H. What is immune following immune checkpoint blockade in an Rev. Neurosci. 8, 368–378 (2007).
privilege (not)? Trends Immunol. 28, 12–18 (2007). orthotopic, immunocompetent model. Cancer 113. Absalon, S., Kochanek, D. M., Raghavan, V. &
63. Alexandrov, L. B. et al. Signatures of mutational Immunol. Res. 4, 124–135 (2016). Krichevsky, A. M. MiR-26b, upregulated in Alzheimer’s
processes in human cancer. Nature 500, 415–421 88. Snyder, A., Wolchok, J. D. & Chan, T. A. Genetic basis disease, activates cell cycle entry, tau-​phosphorylation,
(2013). for clinical response to CTLA-4 blockade. N. Engl. and apoptosis in postmitotic neurons. J. Neurosci. 33,
64. Lohr, J. et al. Effector T cell infiltration positively J. Med. 372, 783 (2015). 14645–14659 (2013).
impacts survival of glioblastoma patients and is 89. Daud, A. I. et al. Tumor immune profiling predicts 114. Takano, T. et al. Glutamate release promotes growth
impaired by tumor-​derived TGF-​β. Clin. Cancer Res. response to anti-​PD-1 therapy in human melanoma. of malignant gliomas. Nat. Med. 7, 1010–1015
17, 4296–4308 (2011). J. Clin. Invest. 126, 3447–3452 (2016). (2001).
65. Kim, Y. H. et al. Tumour-​infiltrating T cell 90. Omuro, A. et al. Nivolumab with or without 115. Ishiuchi, S. et al. Ca2+-permeable AMPA receptors
subpopulations in glioblastomas. Br. J. Neurosurg. 26, ipilimumab in patients with recurrent glioblastoma: regulate growth of human glioblastoma via Akt
21–27 (2012). results from exploratory phase 1 cohorts of activation. J. Neurosci. 27, 7987–8001 (2007).
66. Kmiecik, J. et al. Elevated CD3+ and CD8+ tumor-​ CheckMate 143. Neuro. Oncol. 20, 674–686 (2017). 116. El-​Habr, E. A. et al. A driver role for GABA metabolism
infiltrating immune cells correlate with prolonged 91. Reardon, D. A. et al. OS10.3 randomized phase 3 in controlling stem and proliferative cell state through
survival in glioblastoma patients despite integrated study evaluating the efficacy and safety of nivolumab GHB production in glioma. Acta Neuropathol. 133,
immunosuppressive mechanisms in the tumor versus bevacizumab in patients with recurrent 645–660 (2017).
microenvironment and at the systemic level. glioblastoma: CheckMate 143. Neuro. Oncol. 19 117. Venkatesh, H. S. et al. Neuronal activity promotes
J. Neuroimmunol. 264, 71–83 (2013). (Suppl. 3), iii21 (2017). glioma growth through neuroligin-3 secretion. Cell
67. Han, S. et al. Tumour-​infiltrating CD4(+) and CD8(+) 92. Long, G. V. et al. A randomized phase II study of 161, 803–816 (2015).
lymphocytes as predictors of clinical outcome in nivolumab or nivolumab combined with ipilimumab in 118. Venkatesh, H. S. et al. Targeting neuronal activity-​
glioma. Br. J. Cancer 110, 2560–2568 (2014). patients (pts) with melanoma brain metastases (mets): regulated neuroligin-3 dependency in high-​grade
68. Donson, A. M. et al. Increased immune gene the Anti-​PD1 Brain Collaboration (ABC) (abstract glioma. Nature 549, 533–537 (2017).
expression and immune cell infiltration in 9508). J. Clin. Oncol. 35 (Suppl. 15), 9508 (2017). 119. Li, F., Liu, X., Sampson, J. H., Bigner, D. D. & Li, C. Y.
high-​grade astrocytoma distinguish long-​term from 93. Sharma, A. & Shiras, A. Cancer stem cell-​vascular Rapid reprogramming of primary human astrocytes
short-​term survivors. J. Immunol. 189, 1920–2197 endothelial cell interactions in glioblastoma. Biochem. into potent tumor-​initiating cells with defined genetic
(2012). Biophys. Res. Commun. 473, 688–692 (2016). factors. Cancer Res. 76, 5143–5150 (2016).
69. Cserr, H. F. & Knopf, P. M. Cervical lymphatics, the 94. Treps, L., Perret, R., Edmond, S., Ricard, D. & Gavard, J. 120. Jahani-​Asl, A. et al. Control of glioblastoma
blood-​brain barrier and the immunoreactivity of Glioblastoma stem-​like cells secrete the pro-​ tumorigenesis by feed-​forward cytokine signaling.
the brain: a new view. Immunol. Today 13, 507–512 angiogenic VEGF-​A factor in extracellular vesicles. Nat. Neurosci. 19, 798–806 (2016).
(1992). J. Extracell. Vesicles 6, 1359479 (2017). 121. Biasoli, D. et al. Glioblastoma cells inhibit astrocytic
70. Reboldi, A. et al. C-​C chemokine receptor 6-regulated 95. Rupp, T. et al. Tenascin-​C orchestrates glioblastoma p53-expression favoring cancer malignancy.
entry of TH-17 cells into the CNS through the choroid angiogenesis by modulation of pro- and anti-​angiogenic Oncogenesis 3, e123 (2014).
plexus is required for the initiation of EAE. signaling. Cell Rep. 17, 2607–2619 (2016). 122. Lemée, J. M., Clavreul, A. & Menei, P. Intratumoral
Nat. Immunol. 10, 514–523 (2009). 96. Soda, Y. et al. Transdifferentiation of glioblastoma cells heterogeneity in glioblastoma: don’t forget the
71. Schwyzer, M. & Fontana, A. Partial purification and into vascular endothelial cells. Proc. Natl Acad. Sci. peritumoral brain zone. Neuro Oncol. 17,
biochemical characterization of a T cell suppressor USA 108, 4274–4280 (2011). 1322–1332 (2015).
factor produced by human glioblastoma cells. 97. Guelfi, S., Duffau, H., Bauchet, L., Rothhut, B. & 123. Mangiola, A. et al. Gene expression profile of
J. Immunol. 134, 1003–1009 (1985). Hugnot, J. P. Vascular transdifferentiation in the CNS: glioblastoma peritumoral tissue: an ex vivo study.
72. Masson, F. et al. Brain microenvironment promotes a focus on neural and glioblastoma stem-​like cells. PLoS ONE 8, e57145 (2013).
the final functional maturation of tumor-​specific Stem Cells Int. 2016, 2759403 (2016). 124. Leiss, L. et al. Tumour-​associated glial host cells
effector CD8+ T cells. J. Immunol. 179, 845–853 98. Mei, X., Chen, Y. S., Chen, F. R., Xi, S. Y. & Chen, Z. P. display a stem-​like phenotype with a distinct gene
(2007). Glioblastoma stem cell differentiation into endothelial expression profile and promote growth of GB
73. Thomas, D. A. & Massagué, J. TGF-​beta directly cells evidenced through live-​cell imaging. Neuro Oncol. xenografts. BMC Cancer 17, 108 (2017).
targets cytotoxic T cell functions during tumor evasion 19, 1109–1118 (2017). 125. Iwadate, Y., Fukuda, K., Matsutani, T. & Saeki, N.
of immune surveillance. Cancer Cell 8, 369–380 99. Hu, B. et al. Epigenetic activation of WNT5A drives Intrinsic protective mechanisms of the neuron-​glia
(2005). glioblastoma stem cell differentiation and invasive network against glioma invasion. J. Clin. Neurosci. 26,
74. Weller, M. et al. CD95-dependent T cell killing by growth. Cell 167, 1281–1295.e1218 (2016). 19–25 (2016).
glioma cells expressing CD95 ligand: more on tumor 100. Peterson, T. E. et al. Dual inhibition of Ang-2 and VEGF 126. El Fatimy, R., Subramanian, S., Uhlmann, E. J. &
immune escape, the CD95 counterattack, and the receptors normalizes tumor vasculature and prolongs Krichevsky, A. M. Genome editing reveals glioblastoma
immune privilege of the brain. Cell Physiol. Biochem. survival in glioblastoma by altering macrophages. addiction to microRNA-10b. Mol. Ther. 25, 368–378
7, 282–288 (1997). Proc. Natl Acad. Sci. USA 113, 4470–4475 (2016). (2017).

494 | AUGUST 2018 | volume 14 www.nature.com/nrneurol

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Reviews

127. Yuan, J. X., Bafakih, F. F., Mandell, J. W., Horton, B. J. 146. Brennan, C. W. et al. The somatic genomic landscape in mice. Methods Mol. Biol. 1741, 183–190
& Munson, J. M. Quantitative analysis of the cellular of glioblastoma. Cell 155, 462–477 (2013). (2018).
microenvironment of glioblastoma to develop 147. Prins, R. M. et al. Gene expression profile correlates with 163. William, D. et al. Optimized creation of glioblastoma
predictive statistical models of overall survival. T cell infiltration and relative survival in glioblastoma patient derived xenografts for use in preclinical
J. Neuropathol. Exp. Neurol. 75, 1110–1123 (2016). patients vaccinated with dendritic cell immunotherapy. studies. J. Transl Med. 15, 27 (2017).
128. Rath, B. H., Fair, J. M., Jamal, M., Camphausen, K. & Clin. Cancer Res. 17, 1603–1615 (2011). 164. Oh, T. et al. Immunocompetent murine models for the
Tofilon, P. J. Astrocytes enhance the invasion potential 148. Reardon, D. A. et al. Randomized phase 3 study study of glioblastoma immunotherapy. J. Transl Med.
of glioblastoma stem-​like cells. PLoS ONE 8, e54752 evaluating the efficacy and safety of nivolumab versus 12, 107 (2014).
(2013). bevacizumab in patients with recurrent glioblastoma: 165. Hambardzumyan, D., Parada, L. F., Holland, E. C. &
129. Roos, A., Ding, Z., Loftus, J. C. & Tran, N. L. Molecular checkmate 143. Neuro. Oncol. 19 (suppl. 3), iii21 Charest, A. Genetic modeling of gliomas in mice: new
and microenvironmental determinants of glioma stem-​ (2017). tools to tackle old problems. Glia 59, 1155–1168
like cell survival and invasion. Front. Oncol. 7, 120 149. Sampson, J. H. et al. A randomized, phase 3, open-​ (2011).
(2017). label study of nivolumab versus temozolomide (TMZ) 166. Ben-​David, U. et al. Patient-​derived xenografts
130. Asslaber, M. et al. Native oligodendrocytes in in combination with radiotherapy (RT) in adult patients undergo mouse-​specific tumor evolution. Nat. Genet.
astrocytomas might inhibit tumor proliferation by (pts) with newly diagnosed, O-6-methylguanine DNA 49, 1567–1575 (2017).
WIF1 expression. J. Neuropathol. Exp. Neurol. 76, methyltransferase (MGMT)-unmethylated 167. Baysan, M. et al. Micro-​environment causes reversible
16–26 (2017). glioblastoma (GBM): CheckMate-498. J. Clin. Oncol. changes in DNA methylation and mRNA expression
131. Peferoen, L., Kipp, M., van der Valk, P., van Noort, J. M. 34 (Suppl. 15), TPS2079 (2016). profiles in patient-​derived glioma stem cells. PLoS
& Amor, S. Oligodendrocyte-​microglia cross-​talk in the 150. Weller, M. et al. A randomized phase 2, single-​blind ONE 9, e94045 (2014).
central nervous system. Immunology 141, 302–313 study of temozolomide (TMZ) and radiotherapy (RT) 168. Bigner, S. H., Mark, J. & Bigner, D. D. Chromosomal
(2014). combined with nivolumab or placebo (PBO) in newly progression of malignant human gliomas from biopsy
132. Liddelow, S. A. et al. Neurotoxic reactive astrocytes diagnosed adult patients (pts) with tumor to establishment as permanent lines in vitro. Cancer
are induced by activated microglia. Nature 541, O6-methylguanine DNA methyltransferase (MGMT)- Genet. Cytogenet. 24, 163–176 (1987).
481–487 (2017). methylated glioblastoma (GBM)—CheckMate-548. 169. Beutler, A. S., Banck, M. S., Wedekind, D. &
133. Trajkovic, K. et al. Ceramide triggers budding of Ann. Oncol. 27 (Suppl. 6), 356TiP (2016). Hedrich, H. J. Tumor gene therapy made easy:
exosome vesicles into multivesicular endosomes. 151. Bouffet, E. et al. Immune checkpoint inhibition for allogeneic major histocompatibility complex in the C6
Science 319, 1244–1247 (2008). hypermutant glioblastoma multiforme resulting from rat gliomamodel. Hum. Gene Ther. 10, 95–101 (1999).
134. Dinkins, M. B., Dasgupta, S., Wang, G., Zhu, G. & germline biallelic mismatch repair deficiency. J. Clin. 170. Patel, A. P. et al. Single-​cell RNA-​seq highlights
Bieberich, E. Exosome reduction in vivo is associated Oncol. 34, 2206–2211 (2016). intratumoral heterogeneity in primary glioblastoma.
with lower amyloid plaque load in the 5XFAD mouse 152. Saha, D., Martuza, R. L. & Rabkin, S. D. Macrophage Science 344, 1396–1401 (2014).
model of Alzheimer’s disease. Neurobiol. Aging 35, polarization contributes to glioblastoma eradication 171. Wang, Q. et al. Tumor evolution of glioma-​intrinsic
1792–1800 (2014). by combination immunovirotherapy and immune gene expression subtypes associates with
135. Asai, H. et al. Depletion of microglia and inhibition checkpoint blockade. Cancer Cell 32, 253–267.e255 immunological changes in the microenvironment.
of exosome synthesis halt tau propagation. (2017). Cancer Cell 32, 42–56.e46 (2017).
Nat. Neurosci. 18, 1584–1593 (2015). 153. Jiang, H. et al. Oncolytic adenovirus and tumor-​ 172. Louis, D. N. et al. The 2016 World Health
136. Phuyal, S., Hessvik, N. P., Skotland, T., Sandvig, K. & targeting immune modulatory therapy improve Organization classification of tumors of the central
Llorente, A. Regulation of exosome release by autologous cancer vaccination. Cancer Res. 77, nervous system: a summary. Acta Neuropathol. 131,
glycosphingolipids and flotillins. FEBS J. 281, 3894–3907 (2017). 803–820 (2016).
2214–2227 (2014). 154. Khasraw, M., Ameratunga, M. S., Grant, R., Wheeler, H. 173. Flavahan, W. A. et al. Insulator dysfunction and
137. Atai, N. A. et al. Heparin blocks transfer of & Pavlakis, N. Antiangiogenic therapy for high-​grade oncogene activation in IDH mutant gliomas. Nature
extracellular vesicles between donor and recipient glioma. Cochrane Database Syst. Rev. 9, CD008218 529, 110–114 (2016).
cells. J. Neurooncol 115, 343–351 (2013). (2014). 174. Parsons, D. W. et al. An integrated genomic analysis
138. Jansen, F. et al. Endothelial microparticle uptake in 155. Park, J. S. et al. Normalization of tumor vessels by of human glioblastoma multiforme. Science 321,
target cells is annexin I/phosphatidylserine receptor Tie2 activation and Ang2 inhibition enhances drug 1807–1812 (2008).
dependent and prevents apoptosis. Arterioscler delivery and produces a favorable tumor 175. Frattini, V. et al. The integrated landscape of driver
Thromb. Vasc. Biol. 32, 1925–1935 (2012). microenvironment. Cancer Cell 30, 953–967 (2016). genomic alterations in glioblastoma. Nat. Genet. 45,
139. Chen, Q. et al. Carcinoma-​astrocyte gap junctions 156. Infante, J., Burris, H. A. & Lewis, N. A multicenter phase 1141–1149 (2013).
promote brain metastasis by cGAMP transfer. Nature Ib study of the safety, pharmacokinetics, biological
533, 493–498 (2016). activity and clinical efficacy of INCB7839, a potent and Acknowledgements
140. Ulrich, T. A., de Juan Pardo, E. M. & Kumar, S. selective inhibitor of ADAM10 and ADAM17. The authors thank S. McDavitt for her skilled editorial assis-
The mechanical rigidity of the extracellular matrix Breast Cancer Res. Treat. 106, S269 (2007). tance. This work was supported by U19 CA179563 by the US
regulates the structure, motility, and proliferation of 157. Friedman, S. et al. Clinical benefit of INCB7839, NIH Common Fund, through the Office of Strategic
glioma cells. Cancer Res. 69, 4167–4174 (2009). a potent and selective inhibitor of ADAM10 and Coordination/Office of the NIH Director (X.O.B., A.M.K. and
141. Nuti, E. et al. Bifunctional inhibitors as a new tool to ADAM17, in combination with trastuzumab in T.R.M.), and the US NIH National Cancer Institute (P01
reduce cancer cell invasion by impairing MMP-9 metastatic HER2 positive breast cancer patients. CA069246 (X.O.B.), R01 AI123349 (T.R.M.) and R21
homodimerization. ACS Med. Chem. Lett. 8, 293–298 Cancer Res. 69, 5056 (2014). NS098051 (A.M.K.)).
(2017). 158. Kim, S. S., Pirollo, K. F. & Chang, E. H. Isolation and
142. Barker, H. E., Paget, J. T., Khan, A. A. & Harrington, K. J.
Author contributions
culturing of glioma cancer stem cells. Curr. Protoc. Cell
All authors contributed to researching data, discussion of
The tumour microenvironment after radiotherapy: Biol. 67, 10.21–10 (2015).
content, writing and review of the manuscript before
mechanisms of resistance and recurrence. Nat. Rev. 159. Hubert, C. G. et al. A three-​dimensional organoid
submission.
Cancer 15, 409–425 (2015). culture system derived from human glioblastomas
143. Pyonteck, S. M. et al. CSF-1R inhibition alters recapitulates the hypoxic gradients and cancer stem Competing interests
macrophage polarization and blocks glioma cell heterogeneity of tumors found in vivo. Cancer Res. The authors declare no competing interests.
progression. Nat. Med. 19, 1264–1272 (2013). 76, 2465–2477 (2016).
144. Quail, D. F. et al. The tumor microenvironment 160. Hira, V. V. V. et al. Periarteriolar glioblastoma stem cell Publisher’s note
underlies acquired resistance to CSF-1R inhibition in niches express bone marrow hematopoietic stem cell Springer Nature remains neutral with regard to jurisdictional
gliomas. Science 352, aad3018 (2016). niche proteins. J. Histochem. Cytochem. 66, 155–173 claims in published maps and institutional affiliations.
145. Butowski, N. et al. Orally administered colony (2018).
stimulating factor 1 receptor inhibitor PLX3397 in 161. Calabrese, C. et al. A perivascular niche for brain Reviewer information
recurrent glioblastoma: an ivy foundation early phase tumor stem cells. Cancer Cell 11, 69–82 (2007). Nature Reviews Neurology thanks W. Wick and the other,
clinical trials consortium phase II study. Neuro Oncol. 162. Xu, Z., Kader, M., Sen, R. & Placantonakis, D. G. anonymous reviewers for their contribution to the peer review
18, 557–564 (2016). Orthotopic patient-​derived glioblastoma xenografts of this work.

NATURe RevIeWS | NeuRoloGy volume 14 | AUGUST 2018 | 495

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.