Introduction
Activated cells of various types are known to produce and shed membrane microvesicles into their surroundings. Some of these are also known as microparticles or ectosomes, whereas others are referred to as exosomes, depending on the properties and specific mechanisms of their generation2. The biological role of these structures is poorly understood, but may include secretory processes, immunomodulation, coagulation and intercellular communication3. Microvesicles may vary in their abundance, size and composition, but they often contain material associated with membrane lipid rafts, including functional transmembrane proteins4. For example, procoagulant receptor tissue factor (TF) can be released in this fashion from inflammatory cells and these microvesicles are subsequently incorporated into membranes of platelets, endothelium and other cells where TF exerts its biological effects4. Similarly, vesicular transfer of the CCR5 receptor contributes to cellular susceptibility to HIV transmission5. The mechanism triggering microvesicle generation by cancer cells is unknown, but loss of the tumour suppressor gene p53 may, in some instances, influence the release of increased amounts of TF-containing6 or secretory7 microvesicles to the blood of tumour-bearing mice, or to the pericellular milieu.
We reasoned that oncogenic receptors often reside within regions of the plasma membrane, from which microvesicles originate in cancer cells (for example, lipid rafts) and therefore they could themselves become included in the microvesicle cargo. This would be of particular interest in malignant brain tumours, where the activation of membrane-associated EGFR (A000823) represents a major transforming event, with nearly 50% of cases of glioblastoma multiforme (GBM) expressing amplified EGFR, whereas a large proportion is positive for a distinct mutant known as EGFRvIII8.
Interestingly, we observed that production of microvesicles by cultured U373 glioma cells lacking the activated EGFR increases markedly with enforced expression of the EGFRvIII mutant (U373vIII cells). Thus, formation of vesicular membrane protrusions was readily detected by scanning electron microscopy of U373vIII cells and the degree of this vesiculation was markedly greater than that of U373 cells. This effect was accompanied by a corresponding increase in the recovery of protein from the microvesicular fraction of the culture medium of the respective cell lines6, 9 (Fig. 1a, b; Supplementary Information, Fig. S1). This material contained flotillin-1, a protein associated with membrane lipid rafts that is often found in raft-related microvesicles4. These observations suggest that the expression of EGFRvIII triggers production of microvesicles derived from membrane lipid rafts.
Figure 1: Production of EGFRvIII (oncogene)-containing microvesicles by human glioma cells.
(a) Generation of multiple microvesicular (mv) structures on the surfaces of U373vIII glioma cells with the EGFRvIII oncogene (white arrowheads; SEM image) is less pronounced in the case of their indolent parental U373 cells. (b) Increase in abundance of the microvesicular fraction of the conditioned medium, as a function of EGFRvIII expression in U373 glioma (measured by total protein content; mean 
s. d.; *P < 0.001, t-test, n = 3). (c) Inclusion of oncogenic forms of EGFR in lipid-raft-derived microvesicles released by EGFR-expressing cancer cells. Mutant/activated EGFRvIII is expressed by U373vIII glioma, whereas the A431 squamous cell carcinoma cells harbour endogenously activated wild-type EGFR oncogene. Both EGFRvIII and EGFR were detected in the corresponding microvesicle preparations, which were also positive for the lipid-raft marker flotillin 1 (bottom panel). Parental U373 glioma cells or non-transformed HUVECs (top panel) did not express EGFR in the plasma membrane and produced small amounts of EGFR-free microvesicles. (d) Dependence of tumorigenic properties of U373vIII cells on the functional EGFRvIII. Tumours did not form in the absence of EGFRvIII (U373), or when this receptor was inhibited by administration of the pan–ErbB inhibitor (CI-1033); data are mean s. d. (see Supplementary Information, Methods). (e) Immunostaining with a monoclonal anti-EGFR antibody that does not recognize truncated EGFRvIII (top panel) or a specific anti-EGFRvIII antibody (bottom panel) shows predominant expression of EGFRvIII, but not EGFR in U373vIII tumours. (f) Release of EGFRvIII-containing, flotillin-1-positive microvesicles to the circulating blood of SCID mice harbouring U373vIII tumours (top panels). Scale bars are 10 
m (a) and 50 m (e). Full scans of all gels are shown in Supplementary Information, Fig. S5.
Oncogenic receptor tyrosine kinases, including EGFRvIII, are known to accumulate in membrane lipid rafts of cancer cells. We reasoned that this may cause release of EGFRvIII as cargo of raft-related microvesicles. Indeed, we noticed that EGFRvIII protein was not only readily detectable in lysates of U373vIII cells, but was also present in their derived flotillin-1-containing microvesicles. As expected, U373 cells released microvesicles containing only trace amounts of wild-type EGFR (wtEGFR) and no EGFRvIII. These results were validated against EGFR-negative endothelial cells (HUVECs) and A431 cells expressing only wtEGFR, as well as their respective microvesicle preparations (Fig. 1c).
In contrast to parental U373 cells, U373vIII readily form subcutaneous tumours in immunodeficient (SCID) mice, in a manner susceptible to inhibition by daily administration of an irreversible, small molecule pan–Erb inhibitor CI-1033 (ref. 6; Fig. 1d). These U373vIII tumours stain strongly for EGFRvIII, but not for wtEGFR and emit EGFRvIII-containing microvesicles into the systemic circulation (Fig. 1e, f). Thus, expression of the mutant EGFRvIII gene causes increased aggressiveness of glioma cells, coupled with extracellular and systemic release of microvesicles containing an intact EGFRvIII oncoprotein.
Heterogenous expression of EGFRvIII in human glioma1 suggests that different tumour-cell subsets could shed EGFRvIII-containing microvesicles into the common intercellular space. As microvesicles can readily fuse with cellular membranes through a phosphatidylserine-dependent mechanism4, we asked whether oncogenic EGFRvIII could be transferred in this manner from more aggressive to indolent glioma cells. EGFRvIII-negative U373 cells were incubated with preparations of microvesicles obtained from either their U373vIII counterparts (harbouring EGFRvIII), or from U373vIII–GFP cells, engineered to express a green fluorescent protein (GFP)-tagged EGFRvIII variant (EGFRvIII–GFP). This caused an extensive uptake of the microvesicular content by U373 cells, and their de novo surface expression of the EGFRvIII antigen, or GFP fluorescence, respectively (Fig. 2a–d).
Figure 2: Microvesicular transfer of the oncogenic EGFRvIII between glioma cells.
(a) Fluorescence-activated cell sorting (FACS) analysis showing that U373 cells incubated with microvesicles released by their EGFRvIII-transformed counterparts (U373vIII) acquired the expression of the EGFRvIII antigen on their surfaces. (b) Direct detection of EGFRvIII on the surface of U373 cells incubated with U373vIII-derived microvesicles. Changes in cell shape (transformation) were seen with microvesicle/EGFRvIII uptake. (c) Generation of the U373/EGFRvIII–GFP cell line by expression of the GFP-tagged EGFRvIII (EGFRvIII–GFP) in U373 cells (left panel). Detection of GFP fluorescence (FACS) on the surface of viable U373 cells preincubated with microvesicles derived from U373/EGFRvIII–GFP cell line — evidence for intercellular transfer of EGFRvIII–GFP. (d) Direct GFP-fluorescence of U373 cells incubated with EGFRvIII–GFP containing microvesicles. All procedures involved several washes, during which non-incorporated microvesicles have been completely removed. All scale bars are 20 m.
The apparent intercellular microvesicle-mediated transfer of the ostensibly intact EGFRvIII receptor raises the question as to the signalling consequences (if any) of this event for the 'recipient/acceptor' (U373) cells. To address this question, we examined U373 cells 24 h after their exposure to EGFRvIII-containing microvesicles, for activation of the EGFRvIII downstream signalling pathways, such as MAPK and Akt cascades8, 10. Indeed, incorporation of EGFRvIII into the U373 plasma membrane resulted in a consistent increase in Erk1/2 phosphorylation and this event was dependent on the transfer of active (phosphorylated) EGFRvIII, as U373-derived microvesicles (containing no EGFRvIII) were ineffective. Moreover, the irreversible blockade of the microvesicle-associated EGFRvIII by preincubation of this (cell-free) material with pan–ErbB inhibitor (CI-1033) markedly reduced its ability to trigger Erk1/2 phosphorylation (Fig. 3a, b). Phosphorylation of Erk1/2 was also abrogated by preincubation of EGFRvIII-containing microvesicles with annexin V (A000281), which blocks their exposed phosphatidylserine residues and thereby their uptake by U373 cells4. These results suggest the actual (phosphatidylserine-dependent) microvesicle integration and EGFRvIII transfer, and not merely contact between the EGFRvIII-containing microvesicles and the surface of U373 cells, are required for triggering the activation of MAPK pathway in the 'recipient' cell (Fig. 3c). Incorporation of EGFRvIII-containing microvesicles into U373 cells also induced phosphorylation of Akt, in a manner that could be inhibited by annexin V pretreatment (Fig. 3d), and triggered several other events, such as phosphorylation of PDK1 and Raf (data not shown). These events were also related to the transfer of EGFRvIII and not its effectors, such as Erk1/2 and Akt, as the latter were largely undetectable in the microvesicle lysates (Supplementary Information, Fig. S5).
Figure 3: Activation of growth-promoting signalling pathways in cells that have acquired oncogenic EGFRvIII through microvesicle-mediated intercellular transfer.
(a) EGFRvIII-dependent increase in Erk1/2 phosphorylation in U373 cells that had incorporated microvesicles shed by U373vIII cells. Preincubation of these microvesicles with the irreversible pan–ErbB inhibitor (CI-1033) abrogated their ability to trigger Erk1/2 phosphorylation, in a concentration-dependent manner. EGFRvIII-negative microvesicles from U373 cells were inactive in this assay. (b) Phosphorylation of microvesicle-associated EGFRvIII: immunoprecipitates of microvesicles (IP) with indicated antibodies were probed (WB) for EGFR. (c). Decreased Erk1/2 phosphorylation in U373 cells in which the uptake of EGFRvIII-containing microvesicles was blocked with annexin V. U373vIII microvesicles were pretreated with increasing concentrations of annexin V, which blocks phosphatidylserine residues required for fusion with the plasma membrane. (d) Increase in phosphorylation of Akt in U373 cells that had incorporated EGFRvIII-containing microvesicles. Akt phosphorylation could be prevented by blocking the microvesicle uptake with annexin V. Full scans of all relevant gels are shown in the Supplementary Information, Fig. S5.
Full size image (125 KB)The transforming effect of EGFRvIII-dependent pathways is ultimately mediated by deregulation of several genes responsible for tumour growth, survival and angiogenesis8, 10. With regard to the latter we noted that U373 cells exposed to U373vIII-derived microvesicles exhibited a marked (2–3-fold) increase in the production of vascular endothelial growth factor (VEGF), a potent mediator of brain tumour angiogenesis10 and a known EGFR target11. EGFRvIII activity was essential for this effect, as U373-derived microvesicles (devoid of EGFRvIII), or U373vIII-derived microvesicles preincubated with CI-1033 were both unable to induce the increased release of VEGF (Fig. 4a). Under these conditions, EGFRvIII-containing microvesicles also robustly stimulated VEGF promoter activity and this effect was abrogated by their pretreatment with annexin V (Fig. 4b). Collectively, these observations suggest that the incorporation of U373vIII microvesicles by recipient cells triggers EGFRvIII-dependent increase in VEGF gene expression, probably through activation of the MAPK and Akt pathways10.
Figure 4: Induction of cellular transformation by the uptake of EGFRvIII-containing microvesicles (a) EGFRvIII-dependent increase in VEGF secretion by U373 cells that have incorporated U373vIII microvesicles.
microvesicles pretreated with CI-1033, or originating from EGFRvIII-negative cells (U373) were unable to stimulate VEGF production (*P < 0.05; **P < 0.001; NS, P > 0.05, compared with untreated; #P < 0.05; ##P < 0.01, compared with U373vIII microvesicles). (b) Stimulation of VEGF promoter activity in U373 cells by incorporation of EGFRvIII-containing microvesicles can be blocked by pretreatment with annexin V. (P < 0.0001); (c) Increased expression of the pro-survival gene Bcl-xL and reduced expression of the cell cycle inhibitor p27 in U373 cells exposed to EGFRvIII-containing microvesicles. (d) Increase in soft agar colony-forming capacity of U373 cells after pretreatment with EGFRvIII containing microvesicles (*P < 0.001, compared with untreated U373vIII cells; **P < 0.0001, compared with untreated U373; #P < 0.001, compared with the respective microvesicles from U373 cells; t-test). All numbers are means s. d.; n = 3/experiment (see Supplementary Information, Methods). Full scans of all gels are shown in Supplementary Information, Fig. S5.
Although VEGF upregulation often heralds the activation of oncogenic pathways, cellular transformation downstream of EGFRvIII is mediated by changes in expression of genes directly involved in cellular proliferation and survival8. In this regard our analysis of U373 cells treated with EGFRvIII-containing microvesicles revealed an increase in the expression of the anti-apoptotic protein Bcl-xL. and a decrease in levels of p27/Kip1 cyclin-dependent kinase inhibitor, both known EGFR targets8 (Fig. 4c). Again, these effects were inhibited by annexin V-mediated blockade of the microvesicle uptake. We have also observed similar EGFRvIII-dependent changes in expression of other EGFRvIII target genes, for example p21/Cip (data not shown).
We reasoned that in cells that have incorporated EGFRvIII-containing microvesicles, the functional consequences of the aforementioned repertoire of molecular responses could amount to a higher degree of cellular transformation. This is suggested by more spindle-like morphology of U373 cells exposed to this material (Fig. 2b). To explore this possibility further, U373 cells were preincubated with EGFRvIII-containing microvesicles and tested for growth in semisolid medium, an assay for malignant transformation. Incorporation of microvesicle-associated EGFRvIII caused a 2-fold increase in anchorage-independent soft-agar colony formation of U373 cells, whereas exposure to the equivalent amount of microvesicles devoid of EGFRvIII was inconsequential (Fig. 4d; Supplementary Information, Fig. S2).
It is well recognized that in human GBMs only a small subpopulation of tumour cells harbour the primary genetic alteration leading to EGFRvIII expression, though this oncogene is thought to contribute to progression of the entire tumour. Our results demonstrate that EGFRvIII expression provokes formation of cellular microvesicles, to which this transmembrane protein becomes incorporated and subsequently shed into the pericellular micromilieu (Supplementary Information, Figs S3, S4) and blood (Fig. 1f). Our experiments suggest that microvesicles containing an active oncogene (oncosomes) may serve as vehicles for rapid intercellular transfer of the transforming activity between cells populating brain tumours. This could lead to a 'horizontal' propagation of an increased proliferative, survival, motogenic and angiogenic capacity, even without enrichment in cells harbouring the respective mutation. This hitherto unappreciated form of oncosome-mediated intercellular interaction is fundamentally different from the previously postulated transfer of DNA fragments containing oncogenic sequences from apoptotic cancer cells to their non-transformed (phagocytic) counterparts12. Microvesicle exchange is also different from paracrine effects induced by secretion of tumour-stimulating soluble ligands13, but it could amplify or modulate the latter's effects by intercellular sharing of membrane-associated (and thereby insoluble) active growth factor receptors. Although the present study focuses on EGFRvIII and human glioma cells, a similar microvesicular transfer mechanism could also involve other mutant, upregulated or otherwise activated membrane-associated oncogenic tyrosine kinases (for example, HER-2, wtEGFR, cKit or MET)14; such a mechanism may be operative in a variety of human tumours and propagate with blood to distant sites. It also remains to be established whether host cells (for example, endothelial cells) may also be targets of oncogenes or growth factor receptors that contain microvesicles, and what might be the consequence of such events for tumour progression, metastasis, angiogenesis and responsiveness to therapy. It is interesting to speculate that agents capable of blocking the exchange of microvesicles (for example, annexin V derivatives) may possess anticancer activities by virtue of their interference with the oncosome transfer.
Methods
Reagents.
Antibodies against EGFR (sheep polyclonal and monoclonal), MAPK, AKT, phospho-c- Raf, phospho-PDK1, flotillin-1 and p27 were purchased from Cell Signaling Technology. Anti-EGFRvIII monoclonal antibody was from Zymed. HRP-conjugated secondary antibodies were from Cell Signaling Technology and Alexa fluor secondary antibodies were from Molecular Probes. CI-1033 was a gift from C. Marsolais and L. Levesque (Pfizer). Additional details are included in the Supplementary Information.
Cell culture and isolation of microvesicles.
U373 (human astrocytoma) cells, their stable variant U373vIII expressing Tet-off regulated EGFRvIII or EGFRvIII fused at C-terminus to a GFP (pEGFPN1) cassette (U373vIII–GFP) and A431 were maintained as described previously6 in medium containing microvesicle-depleted fetal bovine serum (FBS). HUVECs were maintained in EGM-2 (Cambrex Bioscience). Microvesicles were collected from conditioned medium or mouse plasma, as described reviously9. Briefly, medium was subjected to two successive centrifugations at 300g and 12,000g to eliminate cells and debris. Microvesicles were pelleted by ultracentrifugation for 2 h at 100,000g, quantified by protein content and analysed for EGFR or EGFRvIII. For scanning electron microscopy (SEM), the cells were grown on coverslips, fixed with 2.5% gluteraldehyde, stained with 1% OsO4, covered with gold and observed using the JEOL 840A instrument. For in vivo analyses, tumours were generated by injection of 1–10 
106 U373vIII or U373 cells into SCID mice (Charles River). In some cases mice were treated daily with the pan–ErbB inhibitor CI-1033 as indicated. Blood was collected from tumour-bearing, or control mice by cardiac puncture into heparinized syringes. Platelet-free plasma was used to prepare microvesicles.
Microvesicle transfer assays.
U373 (acceptor) cells were treated with microvesicles for 24 h and a single-cell suspension was analysed for expression of EGFRvIII or GFP, as indicated. To detect signalling events, U373 were serum-starved before addition of microvesicles, either intact or preincubated with annexin V or CI-1033 at the concentrations indicated. The expression of microvesicle-associated molecules (EGFRvIII) and expression of total and activated MAPK and Akt, as well as other changes, were assayed by immunoblot (Bcl-xL., p27/Kip1), ELISA (VEGF, R&D Systems) or promoter activity assays (VEGF), as described elsewhere6, 9 (Figs 3, 4; Supplementary Information, Methods and Fig. S5). For soft-agar colony formation assays, single-cell suspensions were prepared in 0.3% agarose from equal numbers of cells pretreated with microvesicles or control medium. Cultures were established in plates precoated with 0.5% agarose and all colonies containing more than 4 cells were counted.
Accession codes.
USCD-Nature Signaling Gateway (http://www.signaling-gateway.org): A000823, A000281
Note: Supplementary Information is available on the Nature Cell Biology website.
Author contributions
K. A. N. provided conceptual input, designed and performed experiments, analysed the data and wrote the manuscript; B. M. contributed to conceptual input, performed experiments and coined the term 'oncosomes'; V. L. and L. M. performed experiments; J. M. and A. G. provided reagents and expertise (A. G. in brain tumours), and analysed data; J. R. designed experiments, provided conceptual input and supervision, and wrote the manuscript.