Angiocrine extracellular vesicles impose mesenchymal reprogramming upon proneural glioma stem cells

Abstract

Glioblastoma (GBM) is an incurable form of primary astrocytic brain tumor driven by glioma stem cell (GSC) compartment closely associated with the vascular niche. GSC phenotypes are heterogeneous and range from proneural to mesenchymal-like, the latter characterised by greater invasiveness. Here we document the secretory (angiocrine) role of endothelial cells and their derived extracellular vesicles (EVs) as drivers of proneural-to-mesenchymal reprogramming of GSCs. These changes involve activation of matrix metalloproteinases (MMPs) and NFκB, and inactivation of NOTCH, while altering responsiveness to chemotherapy and driving infiltrative growth in the brain. Our findings suggest that EV-mediated angiocrine interactions impact the nature of cellular stemness in GBM with implications for disease biology and therapy.

Fig. 1. Dynamic proximity between glioma stem cells and endothelium.

a–f Human GBM tissue co-stained for the endothelial cell (EC) marker, CD31, and for GSC marker, NES. White arrowheads point to NES⁺ GSCs juxta-positioned along CD31⁺ blood vessels in the GBM tissue. g Schematic of steps taken to obtain live mouse tissues for high resolution confocal imaging of GSCs and ECs in mouse tumor xenografts. Phase tile image (h) and Green fluorescence tile image (i) of a vibratome cut 200 µm thick coronal section of mouse brain injected with GSC157 proneural-GSCs. Lycopersicon lectin-stained blood vessels in the tumor hemisphere (TH; j) and contralateral hemisphere (CH; m) of the mouse tumor xenografts. GFP labeled GSC157 cells in the tumor hemisphere (k) and contralateral hemisphere (n) of the mouse brain. Merged Lycopersicon lectin /GFP with insets showing GSC proximity with ECs and cellular morphologies of GFP⁺ cells in the tumor hemisphere (l) and contralateral hemisphere (o). p Quantification of the number of GFP⁺ proneural glioma stem cells found in close proximity to-, or not near the vicinity of endothelial cells. q Cartoon representation of the use of mouse aortae to generate sprouting endothelial-tumor cell ‘dynamic’ co-cultures. Green cells represent GFP⁺ GSC83 that were found to migrate towards sprouting ECs. q–t Three days after GFP⁺ GSCs were placed with aortic ring, GFP⁺ GSCs migrated towards ECs and adopted more elongated cellular morphologies. Green arrowheads indicate the presence of GFP⁺ GSCs and red arrowheads point towards ECs. GSC glioma stem cells, ECs endothelial cells, TH tumor hemisphere, CH contralateral hemisphere, EC endothelial cells.

Fig. 2. Paracrine disruption of glioma stem cell sphere formation by endothelial cells.

a Schematic of culturing GSCs alone, or in co-culture with HUVECs for 6 days. Red represents GSCs and light green represents HUVECs. b, c Mesenchymal GSC83 stained with PKH26 (red) were cultured alone for 6 days and growth patterns were recorded using Incucyte imaging. d, e Mesenchymal GSC83 stained with PKH26 and co-cultured with unlabeled HUVECs for 6 days and recorded using Incucyte. f, g Proneural GSC157 stained with PKH26 and cultured for 6 days (Incucyte). h, i Proneural GSC157 stained with PKH26 and cultured with unlabeled HUVECs for 6 days (Incucyte). Incucyte quantification of PKH26 cluster area of mesenchymal GSC83 (j) or proneural GSC157 (k) cultured with or without HUVECs (day 6). l Schematic of obtaining conditioned media from the indicated cells lines. Conditioned media effect on proneural GSC1079 cells cultured, as spheres, in either (m) their own conditioned media (CMO) or in (n) endothelial conditioned media (CME) for 7 days. o Schematic of measuring cell sharpness. The cells which float in media are recorded with a smaller cell sharpness index relative to the cells which adhere to the bottom of the plate. The readings were collected using an inverted microscope of the Incucyte system. p Quantification of cell sharpness on day 7 for two proneural (GSC157 and GSC1079) and two mesenchymal (GSC83 and GSC1005) GSCs treated with either their own (CMO) or HBEC5i-CM (CME). Impedance-related cell shape index measured over time using xCelligence system in cultures of either (q) proneural (GSC157 and GSC1079) or (r) mesenchymal (GSC83 and GSC1005) cells treated with CMO (black line) or HBEC5i-CM (CME; blue line) for 12 h. While cell indices are dramatically impacted by CME in proneural GSC cultures, they were similar between CMO and CME conditions for mesenchymal GSCs. GSC glioma stem cells, CMO own conditioned media, CME conditioned media derived from endothelial cells, HBEC5i immortalized human brain endothelial cells, CM conditioned media.

Fig. 3. Mesenchymal reprogramming of proneural glioma cell stemness by endothelial conditioned media.

Time-dependent changes in clonogenicity (ELDA) of GSCs in the presence of endothelial conditioned media (CME) or control media (CMO), compilation of limiting dilution assays across three proneural GSC lines: GSC1079 (a), GSC157 (b), GSC528 (c) and three mesenchymal GSCs lines: GSC1123 (d), GSC83 (e), GSC1005 (f); black lines represent cells treated with fresh media (FM), gray lines are indicative of cells treated with CMO, and blue lines represent cells treated with CME; n = 3 each group and each time point. g TCGA data mining of proneural versus mesenchymal genetic signatures across 51 patients diagnosed with GBM. Proneural hallmarks include, but are not limited to, NES, NOTCH1 and SOX2. Mesenchymal hallmarks include, but are not limited to, VIM, TGM2 and CD44. Green and purple represent mesenchymal (MES) and proneural (PN) genetic signatures, respectively. h Volcano plot showing enriched mesenchymal markers detected in the mass spectrometry (MS) of CME treated GSC157 cells relative to CMO treated GSC157 cells. i Expression of proneural hallmarks (NES and SOX2) and mesenchymal hallmark (TGM2) in proneural cells (GSC157, GSC1079) and mesenchymal cells (GSC83, GSC1005) after 7 day treatment with CMO, or CME. Flow cytometry of mesenchymal-GSC83 (j) and proneural-GSC157 (k) for CD44-APC. Dashed line, IgG control; blue curve, cells treated with CMO; red curve, cells treated with CME. l Kaplan–Meier symptom-free survival curve of tumor bearing mice injected with proneural-GSC157 cells pre-treated with either fresh media, CMO or CME; n = 5 mice per group; mice intracranially injected with cells pretreated with either fresh media (media) - dashed line, with CMO—black line, or with CME—blue line. m Wound healing/cell migration assay of proneural-GSC157 cells grown as monolayer and treated with CMO or CME throughout 48 h. n Quantification of the wound healing assay for CMO or CME treated GSC157 cells. GSC glioma stem cells, CMO own conditioned media, CME conditioned media derived from endothelial cells. Source data are provided as a Source Data file.

Fig. 4. Extracellular vesicles recapitulate mesenchymal reprogramming effects of endothelial secretome against proneural glioma stem cells.

a Schematic of extracellular vesicle (EV) isolation from conditioned media. b Nanoparticle tracking analysis (NTA/Nanosight) of EVs isolated from CME (HUVEC-CM). c Immunoblot for tetraspanins (CD9, CD63, CD81), and EV purity marker, BIP, indicates the absence of cytoplasmic contamination in EV preparations, according to MISEV2018; Ponceau loading control. d Mass spectrometry analysis of the most abundant proteins enriched in HUVEC EVs (EEVs). e EV transfer assay: HUVEC cells were transduced with Cre-mCherry and proneural-GSC157 were transduced with dsRed/LoxP/eGFP lentiviral vectors. EVs from Cre-transduced HUVECs were purified and incubated with dual reporter proneural-GSC-157 for 4 days to observe a red to green fluorescence switch in recipient GSC157 cells. Cre-loxP experiment involving proneural-GSC157 cells treated with OEVs (top panel), Cre-EEVs (middle panel) and culture supernatant (Sup) from Cre-expressing cells (bottom panel). f Morphological differences observed in proneural-GSC157 cells treated with OEVs or HUVEC-EVs (EEVs) after 3 days in culture. g Cell eccentricity ratios measured by the Incucyte software for proneural-GSC157 cells treated with OEV (black line) versus cells treated with EEVs (blue line). h Expression of SOX2, NES, NOTCH1, NICD, and VIM in proneural-GSC157 cells treated with OEVs or EEVs. i Densitometry analysis of NICD expression (activated NOTCH1) in GSC157 cells treated with CMO, HUVEC-conditioned media (CME), CME derived EVs and Sup (n = 3). j Temozolomide (TMZ) dose-response curve for mesenchymal-GSCs and proneural-GSCs. MTS assay quantifying viability of proneural-GSCs after TMZ treatment in cultures pretreated with fresh media (media), OEV or EEV: GSC157 (k), GSC84 (l) and GSC1079 (m). NICD notch intracellular domain, CMO own conditioned media, Sup supernatant, EV extracellular vesicles, OEVs own EVs EEVs endothelial cell derived EVs, TMZ temozolomide, HUVEC human umbilical vein endothelial cells. Source data are provided as a Source Data file.

Fig. 5. Extracellular vesicles impact multiple signaling elements of the glioma stem cell program.

a Mass spectrometry quantification (n = 3) for the presence of MMPs in HUVEC-EVs (EEVs), proneural-GSC157 cells, GSC157 cells treated with CMO and GSC157 cells treated with HUVEC-CM (CME). Comparisons are made relative to untreated GSC157 cells (b, c). MMP activity of GSC157 EVs, HUVEC-EVs and HBEC5i EVs over 1 h (n = 3; b) and cumulatively (n = 3; c). d Cumulative MMP activity for GSC157 cells, GSC157 cells treated with their own EVs (OEVs), HBEC5i EVs and HUVEC-EVs over 1 h (n = 3). e Cumulative MMP activity of GSC157 cells treated with HBEC5i-EVs and either DMSO control or MMP inhibitor, BB94 (n = 3). f Expression of NICD following treatment of proneural-GSC157 cells with OEVs or EEVs in the absence or presence of MMP inhibitors, AG3340 and BB94. g DAVID analysis of the top pathways enriched in proneural-GSC157 cells treated with CME (HUVEC-conditioned media). h Expression of proneural-(NICD) and mesenchymal-(VIM) hallmarks after pharmacological inhibition of: NFκB pathway (Bay11-7082), Wnt pathway (LGK974) and TGFβ pathway (LY2157299). i Immunocytochemistry and quantification of phospho-P65 (p-P65) after treatment of GSC157 cells with EEVs (HUVEC-EVs). Densitometry quantifications of p-P65 (j) and total P65 (k) in GSC157 cells treated with HBEC5i-EVs (EEVs) relative to OEVs (n = 3). l Expression of activated (phospho) and total p65 in the presence of Bay11-7082 in GSC157s. m Expression of activated (phospho) and total p65 in the presence of AG3340 and BB94 in GSC157 cells exposed to EEVs. n Quantification of wound healing assay for OEV, EEV or EEVs+Bay 11-7082 treated proneural-GSC157 cells. o Schematic of the different donor EVs (OEVs and EEVs) competing for being taken up by proneural-GSC157. p Either 5,000 or 50,000 GSC157 cells exposed to a fixed amount of 30μg of HBEC5i-EVs (EEVs). q Expression of NICD in the presence of OEVs, HUVEC-EVs (EEVs), OEV:EEV (1:1) and OEV:EEV (2:1). NICD Notch intracellular domain, CMO own conditioned meida, CME endothelial conditioned media, EV extracellular vesicles, OEVs own EVs, EEVs endothelial cell derived EVs, MMP matrix metaloprotinase; HUVEC human umbilical vein endothelial cells, HBEC5i immortalized human brain brain endothelial cells. Source data are provided as a Source Data file.

Fig. 6. Enhancement of glioma stem cell invasion of the brain following exposure to endothelial cell secretome.

a–h Proneural-GSCs were tracked in the brain of mice 7 days after intracranial injection; tissues were stained for human antigens: hNICD and hVIM. Image is taken at the tumor site. i–p Expression of hNICD and hVIM in the cortex of the contralateral hemisphere in mice intracranially injected with proneural-GSC157 cells pretreated with CMO i–l and CME m–p after 4 months post inoculation. q Quantification of %VIM/DAPI of proneural-GSC157 cells pretreated with CMO or CME in the cortex (n = 5). r Expression of hNICD and hVIM at the tumor site in the brains of mice 7 days after intracranial injections of proneural-GSC157 cells pretreated with OEVs or EEVs. s Quantification of the number of cells positive for hNICD and hVIM in the xenografts of proneural-GSC157 cells pretreated with OEVs or EEVs. t Model summarising the influence of endothelial EVs on mesenchymal reprogramming of proneural GSCs. EEVs carrying MMPs and possibly other bioactive cargo initiate reprograming events leading to downregulation of proneural hallmarks and upregulation of mesenchymal regulators, including NFκB. The model proposes that EEVs compete with homotypic interactions (OEVs and cell-cell contact) for influence over the phenotype of glioma stem cells upon which they impose a mesenchymal-like program. CMO own conditioned media, CME conditioned media derived from endothelial cells, OEVs own extracellular vesicles, EEVs endothelial cell derived extracellular vesicles.