Glioblastoma tumor cells release microvesicles (exosomes) containing mRNA, miRNA and
angiogenic proteins. These microvesicles are taken up by normal host cells, such as
brain microvascular endothelial cells. By incorporating an mRNA for a reporter protein
into these microvesicles we demonstrate that microvesicle-delivered messages are translated
by recipient cells. These microvesicles are also enriched in angiogenic proteins and
elicit tubule formation by endothelial cells. Tumor-derived microvesicles therefore
serve as a novel means of delivery of genetic information as well as proteins to recipient
cells in the tumor environment. Glioblastoma microvesicles also stimulated proliferation
of a human glioma cell line, indicating a self-promoting aspect. Messenger RNA mutant/variants
and microRNAs characteristic of gliomas can be detected in serum microvesicles of
glioblastoma patients. The tumor-specific EGFRvIII was detected in serum microvesicles
from 7 out of 25 glioblastoma patients. Thus, tumor-derived microvesicles may provide
diagnostic information and aid in therapeutic decisions for cancer patients through
a blood test.
Glioblastomas are highly malignant brain tumors with a poor prognosis despite intensive
research and clinical efforts1. These tumors as well as many others have a remarkable
ability to mold their stromal environment to their own advantage. Tumor cells alter
surrounding normal cells to facilitate tumor cell growth, invasion, chemoresistance,
immune evasion and metastasis 2–4. The tumor cells also hijack the normal vasculature
and stimulate rapid formation of new blood vessels to supply tumor nutrition 5. Although
the immune system can initially suppress tumor growth, it is often progressively blunted
by tumor activation of immunosuppressive pathways 6. Recent studies show the importance
of communication between tumor cells and their environment through shedding of membrane
microvesicles which can fuse to cells in the vicinity 7.
Microvesicles are 30–100 nm in diameter and shed from many different cell types under
both normal and pathological conditions 8. These exosomes can be formed through inward
budding of endosomal membranes giving rise to intracellular multivesicular bodies
(MVB) that later fuse with the plasma membrane, releasing the exosomes to the exterior
8,9. They can also be shed directly by outward budding of the plasma membrane, as
shown for Jurkat T-cells 10.
Microvesicles in Drosophila, termed argosomes, contain morphogens such as Wingless
protein and move throughout the imaginal disc epithelium in the developing embryos
11. Microvesicles found in semen, known as prostasomes, can promote sperm motility,
stabilize the acrosome reaction, facilitate immunosuppression and inhibit angiogenesis
12. On the other hand, prostasomes released by malignant prostate cells promote angiogenesis.
It has been shown that microvesicles can transfer some of their contents to other
cell types 13–16.
The content of microvesicles and their biological function depends on the cell of
origin. Microvesicles derived from B-cells and dendritic cells have potent immuno-stimulatory
and antitumor effects in vivo and have been used as antitumor vaccines 17. Dendritic
cell-derived microvesicles contain co-stimulatory proteins necessary for T-cell activation,
whereas most tumor cell-derived microvesicles do not. Instead they act to suppress
the immune response and accelerate tumor growth and invasiveness 18–21. Breast cancer
microvesicles stimulate angiogenesis, and platelet-derived microvesicles promote tumor
progression and metastasis of lung cancer cells 22,23.
Human glioblastoma tissues were obtained from surgical resections and tumor cells
were dissociated and cultured as monolayers in medium using fetal bovine serum (FBS)
depleted for microvesicles (dFBS). Cultured primary cells obtained from three glioblastoma
tumors were found to produce microvesicles at early and later passages (1–15 passages).
Tumor cells were covered with microvesicles varying in size from about 50 – 500 nm
(Fig. 1a and b). The microvesicles contained RNA and protein in an approximate ratio
of 1:80. To evaluate whether the RNA was contained inside the microvesicles, they
were either exposed to RNase A or left untreated before RNA extraction (Fig. 1c).
There was always less than a 7% decrease in RNA content following RNase treatment.
Thus, it appears that almost all of the RNA is contained within the vesicles and is
thereby protected from external RNases by the surrounding membrane. Bioanalysis of
RNA from microvesicles and their donor cells revealed that the microvesicles contain
a broad range of RNA sizes consistent with a variety of mRNAs and miRNAs, but lack
the ribosomal RNA peaks characteristic of cellular RNA (Fig. 1d and e).
Microarray analysis of mRNA populations in microvesicles and their donor glioblastoma
cells was performed using the Agilent 44K whole genome microarray. Approximately 22,000
gene transcripts were found in the cells and 27,000 transcripts in the microvesicles
(detected at well above background levels, 99% confidence interval) on both arrays.
Approximately 4,700 different mRNAs were detected exclusively in microvesicles on
both arrays, indicating a selective enrichment process within the microvesicles (Supplementary
Table 1). Consistent with this, there was a poor overall correlation in levels of
mRNAs in the cells as compared to microvesicles from two tumor preparations (Fig.
2a and b), supporting selective enrichment of some cellular mRNAs in microvesicles.
In contrast, a comparison of levels of specific mRNAs in different preparations of
donor cells or of microvesicles showed a strong correlation, indicating a consistent
distribution within these distinct cellular compartments (Fig. 2c and d). We found
3426 transcripts differentially distributed more than 5-fold (p-value <0.01). Of these,
2238 transcripts were enriched (up to 380-fold) and 1188 transcripts were less abundant
than in the cells (up to 90-fold) (Fig. 2e). The intensities and ratios of all gene
transcripts are shown in Supplementary Table 2. Ontologies of mRNA transcripts enriched
or reduced more than 10-fold are listed in Supplementary Table 3.
The mRNA transcripts that are highly enriched in microvesicles compared to cells are
not always the most abundant in the microvesicles. The most abundant transcripts would
be more likely to generate an effect in the recipient cell upon delivery. The 500
most abundant mRNA transcripts in microvesicles were divided into different biological
processes based on their ontology descriptions and displayed in Fig. 2f. Glioblastoma
microvesicle mRNAs belonging to ontologies such as angiogenesis, cell proliferation,
immune response, cell migration and histone modification were plotted to compare their
levels and contribution to the mRNA spectrum (Fig. 2g). These ontologies were selected
as they represent functions that could be involved in remodelling the tumor stroma
and enhancing tumor growth. All five ontologies contained mRNA with very high expression
levels compared to the median signal intensity level of the array.
Mature miRNA in microvesicles and donor cells was detected using quantitative miRNA
reverse transcription PCR. A subset of 11 miRNAs known to be abundant in gliomas (Krichevsky
et al., in preparation) was readily detected in donor cells and microvesicles from
two different primary glioblastomas (GBM 1 and GBM 2) (Fig. 2h). The levels were generally
lower in microvesicles per μg total RNA than in parental cells (10%, corresponding
to approximate 3 Ct-values), but correlated well with the tumor profile.
Glioblastoma microvesicles labelled with the fluorescent dye PKH67 were incubated
with human brain microvascular endothelial cells (HBMVEC) in culture. The PKH67-labelled
microvesicles were internalized into endosome-like structures by brain endothelial
cells (Fig. 3a). Similar results were obtained when adding the fluorescently labelled
microvesicles to primary glioblastoma cells (data not shown).
To determine if the mRNA delivered by glioblastoma-derived microvesicles could be
expressed in recipient cells, glioblastoma cells were first transduced with a lentivirus
vector encoding a secreted luciferase from Gaussia (Gluc) 24, and microvesicles produced
by them were purified from conditioned medium. RT-PCR analysis showed that the mRNA
for Gluc (555 bp product), as well as GAPDH (226 bp product), were present in the
microvesicles (Fig. 3b). Purified microvesicles containing Gluc mRNA were added to
HBMVEC cells and Gluc activity released into the medium by these endothelial cells
was monitored over time (Fig. 3c). Gluc activity produced by recipient cells showed
a continuing increase over 24 hrs, supporting ongoing translation of the Gluc mRNA.
This novel method shows that mRNA incorporated into the tumor microvesicles can be
delivered into recipient normal cells and generate a functional protein.
Nucleic acids are of high value as biomarkers because of highly sensitive PCR detection.
We evaluated whether the RNA in microvesicles could be used as biomarkers for glioblastoma
tumors. The epidermal growth factor receptor (EGFR) mRNA is particularly interesting
since expression of the EGFRvIII mutant/variant is specific to some tumors and defines
a clinical subtype of glioma 25. We used a nested RT-PCR to determine if EGFRvIII
mRNA was found in resected glioma tissue and compared the result with microvesicles
purified from a frozen serum sample from the same patient. The samples were coded
and the PCRs were performed in a blind fashion. Fourteen of the 30 tumor samples (47%)
contained the EGFRvIII transcript, which is consistent with the percentage of glioblastomas
found to contain this mutant message in other studies 26. EGFRvIII could be amplified
from microvesicles in seven of the 25 patients (28%) from whom serum was drawn around
the time of surgery (Table 1; Supplementary Fig. 1). Interestingly, two patients with
an EGFRvIII-negative tumor sample turned out to be positive in the serum microvesicles,
supporting heterogeneous foci of EGFRvIII expression in the glioma tumor. EGFRvIII
message was not detected in five serum samples drawn two weeks after extensive resection
of the tumor, with four corresponding to EGFRvIIIpositive tumors, consistent with
this tumor being the source of microvesicles. Furthermore, EGFRvIII was not found
in serum exosomes from 30 normal control individuals (Supplementary Fig. 2). We also
found that miRNA-21, known to be over-expressed in glioblastoma tumors27, was elevated
in serum microvesicles from these patients as compared to controls (Supplementary
Fig. 3). The identification of tumor-specific RNAs in serum microvesicles thus provides
a window into somatic mutations and changes in gene expression in the tumor cells.
To address whether glioblastoma microvesicles could contribute to angiogenesis, we
used an in vitro angiogenesis assay. HBMVECs were cultured in matrigel-coated plates
in endothelial basal medium (EBM), EBM supplemented purified glioblastoma microvesicles,
or EBM plus angiogenic growth factors (EGM). In the presence of microvesicles there
was a doubling of tubule length by the HBMVECs within 16 hrs, comparable to when exposed
to angiogenic factors (Fig. 4a). This finding supports a role for glioblastoma-derived
microvesicles in initiating angiogenesis in brain endothelial cells.
To further characterise the angiogenic capability of microvesicles we analyzed levels
of angiogenic proteins in microvesicles and compared them with levels in glioblastoma
donor cells using a human angiogenesis antibody array (Fig. 4b). Seven of the 19 angiogenic
proteins were readily detected in the microvesicles, with 6 of them (angiogenin, IL-6,
IL-8, TIMP-1, VEGF and TIMP-2) being at higher levels on a total protein basis than
in glioblastoma cells (Fig. 4c). The three most enriched angiogenic proteins were
angiogenin, IL-6 and IL-8, all of which have been implicated in glioma angiogenesis
and increased malignancy 28–30. This indicates that the angiogenic effect of microvesicles
is mediated at least in part by angiogenic proteins.
Human U87 glioma cells were incubated in normal growth medium or medium supplemented
with microvesicles isolated from primary glioblastoma cells. After three days, untreated
U87 cells had increased 5-fold in number whereas the microvesicle supplemented cells
had increased 8-fold (Supplementary Fig. 4). Thus, glioblastoma microvesicles appear
to stimulate proliferation of other glioma cells.
These studies support the ability of microvesicles shed from tumor cells to serve
as a means whereby tumors can manipulate their environment in order to make it more
permissive to tumor growth and invasion. In this study we document the abundant shedding
of microvesicles by primary human glioblastoma cells. We characterized mRNA and miRNA
profiles present in these cells and microvesicles and showed that particular mRNAs
are highly enriched in these microvesicles as compared to donor cells. In fact, more
mRNA transcripts were detected well above background in microvesicles as compared
to cells. This difference could be due in part to the large amount of ribosomal RNA
in the cells compared to the microvesicles, increasing the relative amount of mRNA/μg
total RNA in the microvesicles. Ontology analysis showed that a number of mRNA transcripts
associated with cell migration, angiogenesis, cell proliferation, immune response
and histone modification are present in high levels in the microvesicles. The miRNAs
in microvesicles appeared to parallel their distribution in the glioblastoma cells.
We have shown that glioblastoma microvesicles can enter HMVECs and translate a reporter
mRNA carried by the microvesicles. This suggests that the tumor-derived microvesicles
can modify the surrounding normal cells by changing their translational profile. Further,
we have shown that glioblastoma microvesicles can stimulate an angiogenic phenotype
in normal brain endothelial cells and can stimulate the proliferation of other glioma
cells. In addition to the potential role of mRNAs in these processes, microvesicles
also contain a number of angiogenic proteins, such as angiogenin, FGF., IL-6, IL-8,
TIMP-1, VEGF and TIMP-2. Most of these presumably interact with cognate receptors
on the surface of endothelial cells to promote angiogenesis, and may require extracellular
lysis of the microvesicles with release of proteins contained within them. It has
been proposed that the acidic environment in established tumors can promote lysis
of some of the microvesicles, making intravesicular proteins bioavailable 31. On the
other hand, angiogenic proteins like angiogenin require transportation across the
membrane for biological effect 32, which could be facilitated by the microvesicles.
Tumor microvesicles thus act as a multicomponent delivery vehicle for mRNA, miRNA
and proteins to communicate genetic information as well as signalling proteins to
cells in their environs.
This study presents a thorough analysis of mRNAs that are enriched in the microvesicles
versus donor cells, suggesting that there may be a cellular mechanism for localizing
these messages into microvesicles, possibly via a zip code in the 3′UTR as described
for mRNAs translated in specific cellular locations, e.g. beta actin 33. The conformation
of the mRNAs in the microvesicles is not known, but they may be present as ribonuclear
particles (RNPs) 34. Evidence suggests that retroviruses, like HIV, can utilize the
endogenous microvesicle machinery for budding and generation of new virus particles
10. Interestingly, several endogenous retrovirus RNA sequences were found to be highly
enriched in the microvesicles (Supplementary Table 2).
The RNA found in the microvesicles contains a “snapshot” of a substantial array of
the cellular transcriptome at any given time. Among the mRNAs found in glioblastoma-derived
microvesicles, the EGFR mRNA is of specific interest since the EGFRvIII mutant splice
variant is found specifically in many glioblastomas 26. Using nested RT-PCR we were
able to detect the EGFRvIII message in tumor biopsies and serum microvesicles from
glioblastoma patients, but not in any of 30 normal control sera. We showed that brain
tumors release microvesicles into the bloodstream and that we can genetically type
EGFRvIII status of glioblastoma tumors by nested RT PCR of RNA in microvesicles isolated
from a small amount of patient serum as compared to other methods that require invasive
brain surgery. The sensitivity of this assay may depend on factors as tumor size,
tumor location and serum volume, as well as the method of RNA extraction, cDNA conversion
and PCR used. Information about EGFRvIII status of glioblastoma patients could be
useful in the ongoing EGFRvIII vaccine and other therapeutic clinical trials 35. One
study showed that EGFRvIII-positive gliomas are over 50 times more likely to respond
to treatment with EGFR-inhibitors like erlotinib (Tarceva®) or gefitinib (Iressa®)
36. Thus, we propose a new way of looking at molecular determinants of cancer, including
but not limited to EGFRvIII, by isolating microvesicles from serum and extracting
the RNA for profiling and detection of mutations, splice variants and levels of mRNAs
and miRNAs characteristic of tumor formation, progression and response to therapy.
Microvesicles may provide a means of detecting evolving genetic changes relative to
tumor progression using serum samples drawn over time, presumably no matter what type
of cancer or where the tumor foci are situated in the individual. Further, knowledge
of tumor genotype and phenotype gained through microvesicle analysis may help in designing
tailored therapies to curtail tumor growth. And lastly, microvesicles may prove useful
as a delivery vehicle for therapeutic RNAs and proteins.
Materials and methods
Collection of tumor samples and serum from glioblastoma patients
For cell culture, brain tumor specimens from patients diagnosed by a neuropathologist
as glioblastoma multiforme were taken directly from surgery and placed in cold sterile
Neurobasal media (Invitrogen, Carlsbad, CA, USA). The specimens were dissociated into
single cells within 1 hr from the time of surgery using a Neural Tissue Dissociation
Kit (Miltenyi Biotech, Bergisch Gladbach, Germany) and plated in DMEM 5% microvesicle-depleted
dFBS (prepared by ultracentrifugation at 110,000 × g for 16 hrs to remove bovine microvesicles)
supplemented with penicillin-streptomycin (10 IU ml−1 and 10 μg ml−1, respectively,
Sigma-Aldrich, St Louis, MO, USA). Matched de-identified frozen tumor and serum samples
from confirmed glioblastoma patients were obtained from the Department of Neurosurgery
(Massachusetts General Hospital, Boston, USA and the Cancer Research Center; VU Medical
Center, Amsterdam, The Netherlands). These samples were kept at −80°C until use.
Human glioblastoma cells were placed on ornithine-coated coverslips, fixed in 0.5
× Karnovskys fixative and then washed 2×5 min with PBS. The cells were dehydrated
in 35% EtOH 10 min, 50% EtOH 2×10 min, 70% EtOH 2×10 min, 95% EtOH 2×10 min, 100%
EtOH 4×10 min and then transferred to a Tousimis SAMDRI-795 semi-automatic Critical
Point Dryer followed by coating with chromium in a GATAN Model 681 High Resolution
Ion Beam Coater.
Glioblastoma cells at passage 1–15 were cultured in microvesicle-free media (DMEM
containing 5% dFBS). The conditioned medium from 40 million cells was harvested after
48 hrs. The microvesicles were purified by differential centrifugation 15. In brief,
glioblastomaconditioned medium was centrifuged for 10 min at 300 × g to eliminate
cell contamination. Supernatants were further centrifuged for 20 min at 16,500 × g
and filtered through a 0.22 μm filter. Microvesicles were pelleted by ultracentrifugation
at 110,000 × g for 70 min. The microvesicle pellets were washed in 13 ml PBS, pelleted
again and resuspended in PBS. Exosomes were measured for their protein content using
DC Protein Assay (Bio-Rad, Hercules, CA, USA). Serum exosomes from healthy controls
and glioblastoma patients were diluted up to 13 ml in PBS and sterile filtered before
To evaluate whether RNA was present inside the microvesicles, RNase A (Fermentas,
Glen Burnie, MD, USA) was added to suspensions of microvesicles at a final concentration
of 100 μg/ml and incubated for 15 min at 37°C. Total RNA was then purified using the
MirVana RNA isolation kit (Ambion, Austin TX, USA) according to the manufacturer’s
protocol. The RNA was quantified using a nanodrop ND-1000 (Thermo Fischer Scientific,
Wilmington, DE, USA). Snap frozen tumor biopsies were thawed on RNAlater ICE (Ambion,
Austin TX, USA) according to manufacturer’s recommendation followed by RNA extraction
using the MirVana RNA isolation kit.
The microarray experiments were performed by Miltenyi Biotech (Auburn, CA, USA) using
the Agilent Whole Human Genome Microarray, 4×44K, two color array. The array was performed
on two different RNA preparations from primary glioblastoma cells and their microvesicles.
The data was analysed using the GeneSifter software (Vizxlabs, Seattle, WA, USA).
The Intersector software (Vizxlabs) was used to extract the genes readily detected
on both arrays.
Quantitative miRNA RT-PCR
Total RNA was isolated using the mirVana RNA isolation kit. Total RNA (30 ng) was
converted into cDNA using specific miR-primers (Applied Biosystems, Foster City, CA,
USA) and further amplified according to the manufacturer’s protocol.
HBMVEC in vitro angiogenesis assay
HBMVECs (30,000) (Cell Systems, Catalogue #ACBRI-376, Kirkland, WA, USA) were cultured
on Matrigel-coated wells in a 24-well plate in basal medium (Lonza Biologics Inc.,
Portsmouth, NH, USA) only, or supplemented with glioblastoma microvesicles (7 μg/well)
or a cocktail of angiogenic factors (EGM; hydrocortisone, EGF, FGF, VEGF, IGF, ascorbic
acid, FBS, and heparin; Singlequots from Lonza). Tubule formation was measured after
16 hrs and analyzed with the ImageJ software (NIH).
Gluc mRNA translation assay
Primary human glioblastoma cells were infected with a selfinactivating lentivirus
vector expressing secreted Gluc under a cytomegalovirus promoter 37 to achieve an
infection efficiency of >95%. The cells were stably transduced and microvesicles generated
during the subsequent passages (2–10) were isolated and purified as above. Microvesicles
(50 μg) were added to 50,000 HBMVEC and incubated for 24 hrs. The Gluc activity in
the supernatant was measured directly after microvesicle addition (0 hrs), and 15
hrs and 24 hrs later and normalised to the Gluc activity in the microvesicles. The
results are presented as the mean ± SEM (n = 4).
PKH67 labelled microvesicle
Purified glioblastoma microvesicles were labelled with PKH67 Green Fluorescent labelling
kit (Sigma-Aldrich, St Louis, MO, USA) as described 21. The labelled microvesicles
were incubated with HBMVEC in culture (5 μg/50,000 cells). Microvesicles were allowed
to bind for 20 min at 4°C and cells were then washed and incubated at 37°C for 1 hr.
RT PCR and nested PCR
RNA was extracted using the MirVana RNA isolation kit. RNA was converted to cDNA using
the Omniscript RT kit (if starting material was >50 ng) or Sensiscript RT kit (if
starting material was <50 ng) (Qiagen Inc., Valencia, CA, USA) using a mix of oligo
dT and random hexamer primer according to manufacturer’s recommendation. The following
PCR primers were used: GAPDH primers; Forw 5′-GAA GGT GAA GGT CGG AGT C-3′, Reverse
5′-GAA GAT GGT GAT GGG ATT TC-3′. EGFR/EGFRvIII PCR1; Forw 5′-CCAGTATTGATCGGGAGAGC-3′,
Reverse 5′-TCAGAATATCCAGTTCCTGTGG-3′, EGFR/EGFRvIII PCR2; Forw 5′-ATG CGA CCC TCC
GGG ACG-3′, Reverse 5′-GAG TAT GTG TGA AGG AGT-3′. The Gluc primers have been described
previously 24. PCR protocol: 94°C 3 min; 94°C 45 s, 60°C 45 s, 72°C 2 min × 35 cycles;
72°C 7 min.
Angiogenesis antibody array
One mg total protein from either primary glioblastoma cells or purified microvesicles
isolated from the same cells were lysed in Promega lysis buffer (Promega, Madison,
WI, USA) and then added to the human angiogenesis antibody array (Panomics, Fremont,
CA, USA) according to manufacturer’s recommendations. The arrays were scanned and
analysed with the ImageJ software (NIH).
The statistical analyses were performed using Students t-test.