Interplay Between V-ATPase G1 and Small EV-miRNAs Modulates ERK1/2 Activation in GBM Stem Cells and Nonneoplastic Milieu

Despite a better understanding of the molecular pathogenesis of glioblastoma (GBM), this remains an aggressive and invariably fatal disease. Local tumor recurrence is the main cause of GBM-related death and surgery remains the only potentially curative option. We previously showed that the vacuolar proton pump V-ATPase is key for GBM stem cell (GSC) survival (1) and that GSCs with higher V-ATPase G1 expression exhibit heightened tumorigenicity in vivo (2). This is achieved, at least in part, through GSC secretion of large extracellular vesicles (EV) called large oncosomes in the tumor milieu. This is in line with other studies (3–5) that identified V-ATPase subunits in smaller EVs such as exosomes. Small EVs, including exosomes, are nano-sized (40–150 nm) particles that originate from the late endosomal trafficking machinery (6, 7). These small EVs can exert pleiotropic biological functions both in physiologic and pathologic conditions and can influence the proximal or distant microenvironment via the horizontal transfer of bioactive molecules. Because of these characteristics, small EVs or exosomes have been implicated in every stage of tumorigenesis, from tumor onset (8), maintenance and propagation of the cancer stem cell niche (9, 10), to metastatic dissemination (11, 12). Specifically, both tumor- and stroma-derived exosomes have been also involved in metastatic dissemination (13–15) affecting vascular permeability or by conditioning premetastatic sites in distant organs (11, 16–18). The biological functions exerted by exosomes in recipient cells involve miRNAs (8), a class of the epigenetic regulators qualitatively and quantitatively altered in transformed cells.

In this study, we investigated the interplay between V-ATPase and small EV-miRNAs in glioblastoma taking advantage of primary, patients-derived GSCs and nonneoplastic brain cultures.

Primary cells were isolated directly from patient's sample after surgery at Fondazione IRCCS Ca' Granda-Ospedale Maggiore Policlinico (Milan, Italy). Written informed consent was obtained from all subjects involved in the study and the protocol was approved by a local Ethic Committee (IRB#275/2013).

Samples obtained from tumor cores (n = 51) and/or nonneoplastic tumor margins (MG, n = 20) were disaggregated enzymatically and mechanically using a tumor dissociation kit (Miltenyi Biotec). Then, tumor cells were washed twice with HBSS (Thermo Fisher Scientific) and seeded in Neurocult media supplemented with NS-A Proliferation Kit (Human) and growth factors (bFGF and EGF; all from Stem Cell) to maintain undifferentiated state, or RPMI supplemented with 10% FBS (both from Gibco-Thermo Fisher Scientific) for differentiated condition. MG cultures were seeded in RPMI supplemented with 10% FBS and maintained for up to 1 month in culture.

Small EVs (sEV) were isolated from supernatant of neurosphere (NS) cultures (n = 12) left untreated (controls), or after 24 hours of the indicated treatment. In all cases, cells were seeded at a concentration of 15 NS/mL and media was recovered after 48 hours (controls), or cultures were left untreated for 24 hours followed by incubation with V-ATPase or ERK1/2-targeting drugs for 24 hours and then media were recovered. Cells were pelleted at 250 rcf for 5 minutes and supernatants were collected and concentrated using Amicon Ultra centrifugal filter tubes (Millipore Sigma, part of Merck KGaA). Then, the collected supernatants were centrifuged at 1,000 rcf for 10 minutes at 4°C to remove debris and/or apoptotic bodies (pellet), followed by a second centrifugation at 10.000 rcf for 30 minutes at 4°C to remove large EVs.

Finally, the sEV fraction was isolated from supernatants by qEV size exclusion column (SEC–iZON) as described by Lobb and colleagues (19). Briefly, 500 μL of concentrated supernatant, obtained from 50 mL of starting supernatant, were overlaid on qEV column, eluted with PBS, and collected in 1.5 mL Eppendorf. sEVs are present in phases 7–8–9 (20, 21). The concentration and size distribution of particles was assessed by Nanoparticle Tracking Analysis (NTA) using the Nanosight NS300 instrument (Malvern Instruments Ltd).

Purified sEVs were stored in aliquots (20 μL each) at −80°C before they were used for functional experiments or RNA/protein extraction. The sEV dose was selected on the basis of initial results from dose–response analysis (Supplementary Fig. S1A). Specifically, each sEV aliquot contained 4 × 10⁷ sEVs. For controls (mock samples), fresh NC medium was processed as before and stored at −80°C in aliquots of 20 μL each.

sEV ultrastructure and V1G1 protein presence was evaluated by electron microscopy (EM) and immunogold staining following the Lobb protocol (19) and as described previously (2).

We have submitted all relevant data of our experiments to the EV-TRACK knowledgebase (EV-TRACK ID: EV200059; ref. 22).

To investigate sEV biological effects on nonneoplastic brain (MG), or on tumor cell maintained in differentiated condition (Diff-TC) cell growth, cell-cycle progression, Ki67 proliferation index, and cell motility assays were evaluated. Instead, to evaluate their effect on nonsphere-forming tumor recipient cultures (NSF-TCs, maintained in undifferentiated growth condition) a sphere formation assay was used. Recipient cells were seeded in multi-well plates and let reach 60% confluence; specifically, 2 × 10⁴ or 2 × 10⁵ cells were plated in 24-well or 6-well plates, respectively. Then, recipient cultures were supplemented with NS-derived sEVs at the concentration of 20 μL per mL of culture media, with a ratio of 4 × 10⁷ sEVs/2 × 10⁴ cells.

For the analysis of cell growth, cells were stained with CellTrace dye (5 μmol/L for 20 minutes at 37°C) and the number of cell divisions was evaluated using flow cytometry by scoring the reduction of the dye intensity caused by redistribution of probe in daughter cells as described previously (2). Specifically, Diff-TC or MG cells were followed for 24 hours or 17 days, respectively. Cell-cycle progression was investigated by propidium iodide (PI) staining and flow cytometry. Briefly, cells were fixed with cold absolute EtOH in ice overnight, then washed two times with PBS-1% FBS and stained with 500 μL of PI/RNase Staining Buffer (BD Biosciences). The FACSCanto I instrument and FlowJo V.10.1 software were used for flow cytometry experiments (Becton Dickinson). Cell proliferation was also evaluated by Ki67 nuclear staining after 24 or 72 hours of sEV coculture with glioma or MG recipient cells, respectively. Images were acquired using a confocal microscope (Leica TCS SP5; Leica Microsystems) at 40× magnification.

To evaluate cell motility, MG cells were cocultured with the indicated sEV preparations. After 1 hour, cell movement was followed using Nikon time-lapse microscope (Eclipse Ti-E, Nikon Instruments) for 12 hours capturing images every hour. Conversely, glioma cell migration was analyzed with a wound-healing assay. Briefly, after 24 hours of sEV coculture, a wound was created in the monolayer using a 200 μL tip. Gap closure was monitored over 64 hours and images were captured every 4 hours using the time-lapse microscope. For both experiments, cells were maintained in controlled gas-humidified chamber at 37°C and images were captured in bright field with 10× objective. Gap closure was measured using ImageJ manual tracking plugin and then normalized on initial gap dimension. Sphere formation was evaluated in nonsphere-forming glioma culture after supplementation with the indicated sEV preparation. Then, the number of generated tumor spheres was scored after 3 and 6 days.

For the detection of nascent protein synthesis the Click-iT HPG Alexa Fluor 488 Protein Synthesis Assay Kit was used following manufacturer's instructions (Thermo Fisher Scientific). Specifically, MG monolayer cultures were grown on cover glasses and let reach 60% confluence. Then, recipient cultures were supplemented with NS-sEVs at the concentration of 4 × 10⁷ sEVs/2 × 10⁴ cells and, after 2, 4, and 6 hours, 50 μmol/L Click-iT were added to the cells and incubated for 30 minutes. Then, cells were fixed and permeabilized in 3.7% paraformaldehyde (PFA) and 0.5% Triton, respectively, for 20 minutes at room temperature. Click-iT was detected with Click-iT reaction cocktail while nuclei were stained with Hoechst 33342. Images were acquired using a confocal microscope (Leica TCS SP5) at 40× magnification. Nascent protein synthesis was assessed by determining signal intensity (MFI) in the area proximal to the nucleus.

To measures the activity of signaling pathways in NS, the Cignal Finder Cancer 10-Pathway Reporter Array (Qiagen) was used. NS at basal condition were transfected with the indicated gene reporter (pathway-specific responsive element fused with firefly luciferase gene) and a constitutive Renilla luciferase construct. Dual-luciferase emissions were then measured using the Tecan Infinite F200 luminometer (Tecan Trading AG) and Luciferase to Renilla signal were computed for each pathway. The experiment was performed four times for either V1G1^LOW or V1G1^HIGH-NS.

V-ATPase block was achieved either by V1G1 siRNA or by incubation of NS with Bafilomycin A1 (BafA1) or Concanamycin A. V1G1-targeting siRNA or mock control molecules were described previously (1, 2). BafA1 was used at the nontoxic dose of 10 nmol/L for 24 hours (1). Similarly, we identified the adequate Concanamycin A concentration, corresponding to 10 nmol/L, as the dose able to decrease V-ATPase activity in NS without inducing extensive cell death after 24 hours of treatment (Supplementary Fig. S1B and S1C).

Differences among two or more samples' groups were evaluated using nonparametric Mann–Whitney U or Kruskal–Wallis (with posttest) test, respectively (GraphPad Prism software). Data are expressed as mean ± SD. All biological experiments were performed at least four times and experiments have a technical duplicate unless otherwise indicated in the legend.

For miRNA analysis, miRNA raw data were normalized using the global mean method and relative quantities were obtained applying the 2^−ΔCt formula. Then miRNA relative quantities were median-normalized, log₂-transformed, and imported in R environment for statistical analysis. The R packages ComplexHeatmap from Bioconductor was used for heatmap visualization. The samR package from CRAN was used to identify miRNAs differentially expressed according to V1G1 levels in NS (q value < 0.05, FC > 1 abs). Targets identification of significant miRNAs was achieved using the online tool miRTargetLink Human (https://ccb-web.cs.uni-saarland.de/mirtargetlink/) and transcripts with “strong” and experimentally validated interaction with at least two miRNAs were then imported in STRING database of protein–protein interactions (23) and in WebGestalt (WEB-based Gene SeT AnaLysis Toolkit) functional enrichment analysis web tool (24).

We started this study characterizing sEV preparations obtained from GBM NS with high (V1G1^HIGH-NS) or low (V1G1^LOW-NS) V-ATPase G1 expression (Supplementary Fig. S2A and S2B) and, therefore, different tumorigenic potential in vivo (2, 25). We verified by Western blot analysis, electron microscopy, NTA, and multiparametric flow cytometry assays that NS-isolated sEVs had the expected size, morphology, and phenotype (Supplementary Fig. S2C–S2F). In agreement with MISEV guidelines (26, 27), sEV purified from either NS types had a diameter of 50–200 nm (Supplementary Fig. S2D and S2F), had an integral membrane (as evidenced by cell trace staining) and contained ssRNA (as evidenced by sytoRNA staining; Supplementary Fig. S2G). As for immunophenotypic characterization, sEVs expressed the exosomal and extracellular vesicles markers CD9, CD63, Tsg101, Clathrin, whereas the plasma membrane protein Calnexin was not present (Supplementary Fig. S2C). Moreover, sEVs displayed CD9 and CD81 antigens on their surface (Supplementary Fig. S2E). Finally, NTA showed that V1G1^LOW and V1G1^HIGH-NS produced similar amounts of sEVs (Supplementary Fig. S2F).

To gain functional clues into signals induced by sEV from GBM NS in the surrounding environment, we cocultured sEVs from V1G1^HIGH and V1G1^LOW-NS with three types of recipient cells: nonneoplastic brain cultures obtained from tumor margins (MG), nonsphere-forming primary low-grade glioma cells maintained under differentiating (Diff-TC) or undifferentiating (NSF-TC) conditions (2). sEVs from either NS types were comparably internalized by all recipient cultures tested within 6 hours from addition. However, at later time points, sEVs derived from V1G1^HIGH-NS (sEV^HIGH) were more efficiently uptaken by recipient cells compared with sEV^LOW (Supplementary Fig. S2H and S2I). Under these conditions, coculture with sEV^HIGH increased the survival time of MG cultures from 9 to 17 days, whereas sEV^LOW-supplemented cultures showed a survival time comparable with control (Mock; Fig. 1A and B; Supplementary Fig. S3A). Moreover, sEV^HIGH stimulated MG cell-cycle progression and proliferation, as showed by an increase in the fraction of cell with S-phase DNA content (Fig. 1C; Supplementary Fig. S3B) and greater percentage of Ki67-positive nuclei (Fig. 1D). Similar effects were detected when sEVs were provided to Diff-TC cells (Supplementary Fig. S3C and S3D), with sEV^HIGH inducing the strongest effects on cell proliferation. Consistent with these results, sEV^HIGH stimulated protein synthesis in MG cultures (Fig. 1E).

We next examined the effect of sEVs on the motility of recipient cells. In these experiments, sEV^HIGH stimulated the motility of MG recipient cells, compared with control conditions or sEV^LOW addition (Fig. 1F). Similar effects were observed also when sEV^HIGH were cocultured with Diff-TC cells (Supplementary Fig. S3E). Finally, only sEV^HIGH induced NSF-TC to form NS (Supplementary Fig. S3F), a tumor cell population known to be enriched in cancer stem cells (28).

These set of data suggest that sEV secreted by V1G1^HIGH-NS induce a protumorigenic response in the surrounding environment by promoting greater cell survival, proliferation, motility, and clonogenicity, all events crucial for glioma progression (29).

To verify that the V-ATPase was responsible for sEV^HIGH effects on recipient cells, we impaired the pump activity in NS with BafA1. BafA1 treatment did not alter vesicles secretion or their immunophenotype (Supplementary Fig. S4A–S4C), and significantly increased sEV production by NS (Supplementary Fig. S4B). Also, sEVs produced by V1G1^HIGH treated with BafA1 (sEV^BafA1) were internalized by recipient cells comparably with sEV purified from V1G1^HIGH-NS treated with vehicle (sEV^CTRL; Supplementary Fig. S4D and S4E). Functionally, BafA1 treatment abolished the biological effects of sEV^HIGH in either type of recipient cells, shutting down proliferation of MG and Diff-TC (Fig. 1G and H; Supplementary Fig. S4F), Diff-TC motility (Fig. 1I), and NSF-TC sphere formation (Fig. 1J). Therefore, V-ATPase activity is required for the observed protumorigenic functions of NS with elevated expression of V1G1.

Next, we asked whether NS-derived sEVs carry the V1G1 subunit as a molecular cargo. Accordingly, we demonstrated by immunogold that all sEV types vehiculate the ATP6V1G1 subunit, (Fig. 2A), multiparametric flow cytometry (Fig. 2B and C) and Western blot (Fig. 2D) assays. Moreover, sEV^HIGH carried the highest amount of V1G1 compared with sEV^LOW and sEV^BafA1 (Fig. 2B and C). To assess whether the EV-V1G1 was transferred to recipient cells, we examined its expression in MG cultures after coculture with sEV^HIGH, sEV^LOW or control (Mock). In these experiments, V1G1 protein levels were increased in MG cells upon sEV^HIGH supplementation (Fig. 2F and G). Other pump subunits were equally modulated in recipient cultures, with V1F being more expressed in cells cocultured with sEV^HIGH while V1D was more expressed after sEV^LOW supplementation (Fig. 2F). To exclude that this effect was related to de novo transcription, we next quantified the mRNA expression of V-ATPase subunits in recipient cultures after sEV supplementation. Here, no significant differences in V-ATPase subunit mRNA expression was observed between sEV^HIGH-supplemented and control cells (Fig. 2E). Finally, similar modulation of V-ATPase subunits was observed in Diff-TC after coculture with different sEV (Supplementary Fig. S5A). Inhibition of V-ATPase activity in V1G1^HIGH-NS decreased the amount of V1G1 protein in secreted vesicles (Fig. 2B) and, accordingly, incubation of recipient cells with sEV^BafA1 did not result in increased V1G1 expression (Fig. 2F and G).

These data are in line with our previous description of a differential V-ATPase conformation according to glioma aggressiveness, with more aggressive tumors carrying higher levels of the V1G1 and V1F subunits (25).

To better understand the mechanisms of sEV^HIGH protumorigenic effects, we profiled cancer-related signaling pathways in MG recipient cells upon cocultures with different sEV preparations for 72 hours. Using an antibody array assay, we found that sEV^HIGH induced ERK1/2, Stat3, AMPKα, HSP27, PRAS40, and mTOR phosphorylation in recipient cultures (Fig. 3A; Supplementary Fig. S5B). Conversely, recipient cells treated with sEV^LOW showed reduced mTOR and glycogen synthase kinase-3β (GSK3β) phosphorylation (Fig. 3A), suggesting that sEV^LOW blunt PI3K/ERK signaling in recipient cultures (Supplementary Table S1). In keeping with this finding, V1G1^HIGH treatment resulted in greater activation of ERK1/2 signaling compared with V1G1^LOW-NS at both transcriptional and protein level (Fig. 3B and C) and, accordingly, BafA1-mediated inhibition of V-ATPase activity decreased ERK1/2 phosphorylation particularly in V1G1^HIGH-NS (Fig. 3D). Other cancer-related pathways, such as Wnt, Notch, NFκB, and Myc, were not differentially induced in NS with different V1G1 level (Supplementary Fig. S5C) and were not modulated by sEV^HIGH (Supplementary Fig. S5D–S5F).

We then looked at gene expression level of ERK1/2 pathway members in recipient MG cells upon coculture with sEV^LOW or sEV^HIGH using a cancer pathway array. The majority of these genes was upregulated after sEV^HIGH supplementation (Fig. 3E), with Bcl2 being the most upregulated target (P = 0.009; Fig. 3F). To determine whether MAPK-related transcripts or Bcl2 upregulation in recipient cells was due to EV-mediated transfer or by de novo transcription, we profiled sEV-mRNA content. Bcl2 transcript was not present in sEV and, in general, MAPK-transcripts were not overrepresented in sEV^HIGH (Supplementary Fig. S5G). When MG cells were incubated with sEV^BafA1, protein phosphorylation, and Bcl2 or MAPK transcripts were restored to basal levels (Fig. 4A–C). Finally, we detected nuclear accumulation of phosphorylated ERK1/2 only in MG cells incubated with sEV^HIGH and not in control, sEV^LOW- or sEV^BafA1-treated cultures (Fig. 4D). These results suggest that the protumorigenic reprogramming of V1G1^HIGH-NS–derived extracellular vesicles requires activation of ERK1/2 signaling and Bcl2 upregulation.

Because MAPK-related genes were not vehiculated by sEVs, we next asked whether sEVs miRNAs (sEV-miRNA) could contribute to ERK1/2 activation in recipient cells. In these experiments, we identified approximately 150 miRNAs in NS sEVs and sEV-miRNome did not overlap with NS-miRNome, as outlined by principal component analysis (PCA) analysis (Supplementary Fig. S6A). Specifically, sEVs from V1G1^LOW-NS expressed a higher number of miRNAs that were expressed at higher levels, compared with sEV^HIGH (Supplementary Fig. S6A) and, globally, sEV-miRNAs distinguished between the two sEV types at unsupervised level (Fig. 5A). The diversity in miRNAs quantity observed between V1G1^LOW- and V1G1^HIGH-NS was not attributable to different expression of the miRNA-processing enzymes Dicer and Drosha (Supplementary Fig. S6B). Next, we searched for miRNAs that differed between sEV^LOW and sEV^HIGH and we identified 45 miRNAs upregulated in sEV^LOW compared with sEV^HIGH samples (Supplementary Fig S6C; Supplementary Table S2). Among potentially repressed pathways targeted by sEV^LOW-miRNAs were MAPK signaling, PI3K/Akt signaling, and cell cycle (Fig. 5B, Supplementary Fig. S6D; Supplementary Tables S3 and S4). Therefore, we validated MAPK-associated miRNAs (n = 32) in a second set of sEVs and 10 miRNAs (31.3%) were confirmed to be significantly upregulated in sEV^LOW (Fig. 5C). These sEV-miRNAs were increased in recipient cells only after supplementation with sEV^LOW and not with sEV^HIGH, suggesting that the sEV-miRNAs were functional (Fig. 5D). Moreover, the level of these miRNAs was increased in EVs derived from V1G1^HIGH-NS treated with the V-ATPase inhibitors BafA1 (sEV^BafA1; Fig. 5E, Ctrl, sEVs derived from V1G1^HIGH-NS) or Concanamycin A (Supplementary Fig. S6E), comparably with their expression detected in sEV^LOW. To test a potential role for the V1G1-overexpressing V-ATPase in regulating sEV-miRNA composition, we transiently silenced the V1G1 subunit in V1G1^HIGH-NS (Supplementary Fig. S6F) and analyzed the expression of these 10 miRNAs. The expression of five out of 10 sEV-miRNAs, namely miR-21-5p, -30b-5p, -30c-5p, -195-5p, and miR-15b-5p, was increased after V1G1-siRNA (Supplementary Fig. S6G). Functionally, sEV^siV1G1 had no effect on the motility of recipient cells (Fig. 5F). Therefore, we ectopically overexpressed this pool of 5 miRNAs (miR-21-5p, -30b-5p, -30c-5p, -195-5p, and miR-15b-5p; MAPK-miRNAs) or control molecules (Ctrl-miRNA) into MG recipient cells (Supplementary Fig. S6H) followed by coculture with sEV^HIGH or control (Mock). After transfection, the expression of MAPK-miRNAs in MG cells was comparable to that achieved with sEV^LOW co-culture and it did not decrease after sEV^HIGH supplementation (Supplementary Fig. S6H). In this experimental setting, ectopic miRNAs expression abolished the functional effects of sEV^HIGH in recipient MG cells suppressing cell motility (Fig. 6A) and cell proliferation (Fig. 6B) to levels of control transfecting. In keeping with this, presence of the MAPK-miRNAs and sEV^HIGH prevented ERK signaling activity (Fig. 6C; Supplementary Fig. S6I and S6J), and Bcl2 or DUSP1 mRNA expression was not increased (Fig. 6D). The action of the MAPK-miRNAs was selective for ERK, because the expression of JNK transcripts (MAPK8, MAPK10) was unaffected by their presence as compared with cultures incubated with sEV^HIGH alone (Fig. 6D). Finally, treatment of recipient cells with the ERK inhibitor PD98059 after sEV supplementation prevented the pro-motile effect of sEV^HIGH (Supplementary Fig. S6K)

This set of data proposes that there is a signature of sEV-miRNA associated with the V1G1^LOW phenotype that prevents activation of the ERK1/2 pathway in recipient cultures. On the contrary, V1G1^HIGH-NS do not vehiculate these miRNAs in their sEVs and consequently reprogram the surrounding nonneoplastic microenvironment toward a protumorigenic state through ERK1/2 activation (Fig. 6E).

In this study, we identified a novel mechanism exploited by the more tumorigenic V1G1^HIGH-NS to alter the nonneoplastic brain microenvironment by promoting protumorigenic changes such as cell motility, proliferation, and survival. Mechanistically, this involves activation and nuclear translocation of ERK1/2 and upregulation of its downstream effector Bcl2. Interestingly, Bcl2 or MAPK genes are not transferred to recipient cells via sEVs. Instead, we demonstrated that sEVs derived from V1G1^LOW-NS or from NS treated with the V-ATPase inhibitors carry a unique profile of inhibitory miRNAs, which are responsible for blunting the protumorigenic response mediated by V1G1^HIGH-NS.

Taken together, these results are in line with the recent evidence showing that decreased level of tumor suppressor miRNAs in sEVs is a prometastatic mechanism (8). In our studies, target prediction analysis identified a panel of miRNAs, including miR-21-5p, -30b-5p, -30c-5p, -195-5p, and miR-15b-5p, which were over-represented in sEV^LOW-miRNAs and played a key role in antagonizing MAPK signaling. Accordingly, ectopic expression of these miRNAs in MG recipient cells, or use of sEVs derived from NS in which V-ATPase activity was abrogated, was sufficient to abolish ERK1/2 activation and the induction of protumorigenic responses in recipient cells. Therefore, it is possible that the absence of MAPK-targeting miRNAs in secreted sEV from V1G1^HIGH-NS mediates unregulated ERK1/2 activation in recipient cells and downstream mechanisms of heightened cell proliferation and cell motility.

How the V1G1 subunit of the V-ATPase pump regulates the packaging of selective miRNAs in small EV remains to be elucidated. It is intriguing that this subunit is present on the NS plasma membrane, in sEV, and that its inhibition decreases the proportion of sEVs vehiculating V1G1 while increasing sEV-miRNAs targeting MAPKs. We previously described that V-ATPase composition stratifies glioma according to aggressiveness and independently from clinical or molecular variables (25). Now, we show that one of the V-ATPase subunit associated to tumor aggressiveness plays a dominant role in sEV-mediated rewiring of the glioma surrounding microenvironment via ERK1/2 signaling.

The data presented here fit well with an emerging role of V-ATPase pump in the modulation of an intra- and extracellular milieu via acidification and modulation of protein trafficking in organelles such as endosomes and lysosomes (5, 15, 30, 31). Moreover, a feedback loop exists between V-ATPase and ERK, because it has been described that V-ATPase inhibition blocks ERK/MEK signaling in regenerating epithelia (32) and in cisplatin-resistant ovarian cancer cells (33). In addition, PI3K and ERK signaling are required for V-ATPase assembly upon viral infection (32). In this context, V-ATPase subunits are also involved in exosome biogenesis in physiologic and pathologic conditions (34, 35) and are found in EVs released from different cell types (2, 15).

With respect to gliomagenesis, we know that GBM secreted EVs play a major role in the cross-talk between tumor cells and nonneoplastic parenchyma (16, 31, 36, 37) through the horizontal transfer of bioactive molecules (5, 16, 18, 38). Initial studies on GBM use of EVs to reprogram the surrounding microenvironment showed that the oncogenic truncated EGFRVIII protein was transferred through large microvesicles to surrounding cancer cells that lacked the receptor (39). This caused activation in recipient cells of oncogenic signaling that eventually led to phenotypic transformation and increased anchorage-independent cell growth. Moreover, GBM secreted EVs loaded with oncogenic cargoes have been described in glioma patients' cerebrospinal fluid (40) and in the circulation (2). Therefore, it is becoming evident that aggressive glioma use EVs to favor tumor progression (2, 41).

Altogether, our data identify novel molecular mechanisms of gliomagenesis, specifically affecting the GBM stem cell niche, via differential reprogramming of selected neighboring cell populations. As a key requirement in this pathway, the V-ATPase V1G1 subunit could provide an actionable therapeutic target for disease intervention both to prevent establishment of a stem cell niche, as well to limit local infiltration of nonneoplastic brain parenchyma.

No potential conflicts of interest were disclosed.

I. Bertolini: Conceptualization, formal analysis, investigation, writing-original draft. A.M. Storaci: Formal analysis, investigation, methodology. A. Terrasi: Data curation, software, formal analysis, methodology. A. Di Cristofori: Resources, investigation. M. Locatelli: Resources, formal analysis. M. Caroli: Resources, formal analysis. S. Ferrero: Resources, formal analysis, investigation, methodology. D.C. Altieri: Supervision, writing-review and editing. V. Vaira: Conceptualization, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing.

We are thankful to Dr. Mariacarla Panzeri from ALEMBIC-Advanced Light and Electron Microscopy BioImaging Center (San Raffaele Scientific Institute) for technical help with electron microscopy and to Prof. Valentina Bollati from the Department of Clinical Sciences and Community Health (University of Milan, Milan, Italy) for help with NTA experiments. The authors are thankful to the INGM Imaging Facility for scientific and technical assistance. This work was supported by Fondazione Cariplo (2014–1148; to V. Vaira), by the Ricerca Corrente Program 2017 (to S. Ferrero), and by the NIH (grant P01 CA140043; to D.C. Altieri). A.M. Storaci was supported by a Fellowship from the Doctorate School in Molecular and Translational Medicine of Milan University.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.