Close this search box.

Mesenchymal glioma stem cells trigger vasectasia—distinct neovascularization process stimulated by extracellular vesicles carrying EGFR – Nature Communications

Glioma stem cell subsets trigger distinct vascular patterns in vivo

Since PN and MES glioma stem cell subtypes markedly differ with regard to profiles of their associated angiogenic mediators27 and EVs26, we chose to examine these differences in greater functional detail. As reported earlier4, we observed that PN-GSC-initiated intracranial xenografts in immune-deficient (NSG) mice progress at a slower rate than their MES-GSC-initiated counterparts (Fig. 1a). Remarkably, this disparity was also associated with stark differences in vascular patterning (Fig. 1b). Thus, while PN tumours exhibited dense networks of capillary sized (angiogenic-type) blood vessels positive for CD31, their more aggressive MES counterparts contained unexpectedly less dense vascular patterns dominated by larger vessels, up to 100–300 μm in diameter (Fig. 1c, d). As these structural features are inconsistent with characteristics of angiogenic capillary sprouting28, we refer to this novel large blood vessel formation process as vasectasia.

Fig. 1: Differential vascular patterns in GSC-driven tumours and vascular activities of soluble and vesicular components of glioma stem cell secretome.
figure 1

a Kaplan-Meier survival curves of mice bearing PN GSC- and MES GSC-derived tumours. (n = 5 independent experiments. Two-tailed paired t test. P = 0.0000965 and 0.000089) b Representative images of immunofluorescence for CD31 reveal phenotypic vascular differences between tumours driven by PN or MES GSCs. (n = 5 independent experiments). c Quantification of vessel size distribution through tracing CD31 positive endothelial cells. Blood vessels in MES tumours present enlarged lumens, up to 90 μm, compared to a mean vessel diameter of 13 μm in the PN tumours (n = 5 mice/group. Two-tailed paired t test P = 5.48−16 and 1.18−12). d Quantification of microvascular density using CD31 staining (n = 5/group. Two-tailed paired t test. P = 5.25−09 and 2.68−09). Microvascular density was expressed as vessel density per high power field (hpf). Scale bars are 50 µm e Schematic diagram illustrating mouse aortic ring endothelial outgrowth assay using conditioned media derived from different glioma stem cells. f Endothelial responses induced by the GSC conditioned media containing soluble fraction of the secretome and EVs. RhVEGF was used as positive control [25 ng/mL]. Cells were imaged with optical microscope (left) to assess the number of endothelial cells growing out of the ring using FIJI software (right) (n = 6 independent experiments. Two-tailed paired t test. MES P = 7.55−05 and 2.65−04). g Schematic diagram illustrating secretome fractions preparation using centrifugation methods. h Endothelial responses induced by the GSC secretome. Mouse aortic rings were seeded under domes of BME and cultured in growth factors-enriched media. After rings began to form endothelial outgrowth (often referred to as ‘sprouts’) they were treated with supernatant fractions (PN P = 0.00037 and 3.31609−06. MES P = 0.0010 and 0.00073) or with (i) 30 μg/mL of EVs  obtained from either PN GSC (157;1079), or MES GSC (83; 1005). RhVEGF was used as positive control [25 ng/mL]. The number of endothelial cell outgrowths from the ring was assessed using FIJI software. (n = 6 wells/3 independent experiments. Two-tailed paired t test. PN P = 0.0229 and 0.0023. MES P = 0.0021 and 0.00037). j VEGF content distribution in EVs and in supernatant fractions of GSC conditioned media. ELISA assay quantification showed that the growth factor is virtually absent in EVs, and preferentially released in soluble form into the culture media supernatant (n = 4 independent experiments. Two-tailed paired t test. PN P = 1.21−05 and 1.30−08. MES P = 8.08−06 and 9.83−07). Results are shown as mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001 treated group versus control group ****P < 0.0001; Detailed data are provided as a Source Data file.

Extracellular vesicles from mesenchymal glioma stem cells transfer EGFR transcripts to endothelial cells

To glean more insights as to the nature of differential vascular effects evoked by PN-GSCs and MES-GSCs cells, their conditioned media (CM) were tested for the ability to induce endothelial cell responses. Thus, we measured endothelial cell outgrowths from mouse aortic ring explants (Fig. 1e–g; Supplementary Fig. 1a, b) along with migration of cultured human primary endothelial cells (HUVEC), and human brain microvascular endothelial cells (HBMVEC; Supplementary Fig. 1c–f) in transwell assays. These responses to tumour CM were measured against negative controls and high concentrations of recombinant VEGF (Fig. 1e–g; Supplementary Fig. 1c, d). All GSC lines exhibited endothelial stimulating activity in these assays albeit to somewhat variable degrees. In view of earlier reports suggesting angiogenic roles of both soluble (VEGF)15 and insoluble (EV)24 mediators, we subjected GSC-derived CM to differential centrifugation26 and compared the activities of supernatants and small EV fractions (Fig. 1h–j; Supplementary Fig. 1d–f). Interestingly, in the case of PN-GSCs, endothelial-stimulating activity was largely (though not exclusively) contained in the culture supernatant, whereas MES-GSC released a considerable proportion of endothelial stimulators in association with EVs. As reported15, the culture supernatants of all GSCs contained appreciable amounts of soluble VEGF, which was virtually undetectable in the corresponding EV fractions, including highly active MES-GSC EVs (Fig. 1k).

In the virtual absence of VEGF in MES-GSC EVs, we sought alternative mediators of endothelial stimulating activity. Since vascular responses to external stimuli often involve activation of receptor tyrosine kinases (RTKs), we interrogated their phosphorylation status in human and mouse endothelial cells treated with GSC-EVs (Fig. 2a). Remarkably, PN-GSC EVs elicited relatively weak phosphorylation across the panel of 49 RTKs included in the immobilised antibody array (e.g., traces of pVEGFR1). On the other hand, MES-GSC-derived EVs triggered several RTKs, of which phosphorylated EGFR (pEGFR) overshadowed other responses (pVEGFR1, pALK, pVEGFR2, pAXL) by orders of magnitude. Notably, endothelial cell expression of pEGFR was selectively stimulated by MES-GSC EVs, but not by PN-GSC EVs (Supplementary Fig. 2a), as early as within 24 h of incubation and was sustained for up to 6 days post-treatment (Fig. 2b; Supplementary Fig. 2b). This is of interest as cancer cell EVs are known to carry oncogenic and bioactive EGFR protein20,23,29, which can be internalised by endothelial cells leading to their reprograming22. Indeed, we observed the acquisition of EGFR expression in human large vessel-derived (HUVEC)- and brain microvessel-derived endothelial cells (HBMECs) in the presence of MES-GSC-EVs, but not in the case of  PN-GSC-EVs (Supplementary Fig. 2b–d). These EVs were bioactive, however,  as mouse endothelial cells (EOMA) exposed to PN-GSC-EVs acquired the expression of CD133 (Supplementary Fig. 2d), a known marker of PN-GSC10.

Fig. 2: Protracted expression of activated EGFR in endothelial cells subjected to extracellular vesicle-mediated transfer of EGFR/EGFRvIII transcript from mesenchymal glioma stem cells.
figure 2

a Quantification of the human phospho-RTK expression array; analysis performed with EVs from either PN or MES EVs incubated with primary endothelial cells (HUVEC) for 7 days (n = 4 wells/2 independent experiments). Data are presented as mean values ± SD; b Endothelial cells incubated with EVs from PN or MES EVs followed by protein extraction and western blot to analyse the expression of EGFR. The transfer resulted in lasting ectopic activation of EGFR in HUVEC up to 6 days. β-Actin was used as loading control. (n = 3 independent experiments); c Glioma stem cell-derived EGFR+ EVs were incubated with mouse endothelial cells. The transfer of human-specific EGFRvIII was detected for up to 6 days. Mouse β-Tubulin was used as loading control (n = 3 independent experiments); d Glioma stem cell-derived EVs were incubated with mouse endothelial cells enabling transfer of human EGFR. The transfer of human-specific EGFR mRNA was detected only after treatment with MES GSC EVs (n = 3 independent experiments); e Western blot and RT-PCR analysis showed that GW4869 treatment selectively inhibited the shedding of EVs carrying EGFR protein, while EGFR mRNA EVs were not affected (n = 3 independent experiments); f Schematic diagram illustrating the immunoprecipitation approach using magnetic beads crosslinked with anti-EGFR antibody. Western blot and RT-PCR analysis showed a population of EVs enriched for EGFR protein and mostly depleted for EGFR mRNA, while unbound EVs with no EGFR protein were enriched for EGFR mRNA (n = 3 independent experiments); g EGFR status in endothelial cells treated with EVs carrying only EGFR mRNA. Western blot analysis reveals that transfer of EGFR mRNA is sufficient to express EGFR protein in primary endothelial cells for up to 6 days (n = 3 independent experiments); Source data are provided as a Source Data file.

It is possible that rather than transferring EGFR, the cargo of MES-GSC EVs may trigger the expression of endogenous EGFR by recipient endothelium30. To distinguish between these possibilities, we incubated human MES-GSC EVs with mouse endothelial cells (EOMA) and probed them specifically for human EGFR (hEGFR). Both hEGFR protein and transcript were readily detectable in GSC EVs and in EV-treated EOMA cells (Fig. 2c, d) suggesting intercellular transmission of this receptor. Moreover, when human endothelial cells were incubated with MES-GSC EVs, they became positive for GBM-specific mutant EGFRvIII mRNA and protein expression (Supplementary Fig. 2e) further enforcing the notion that endothelial cells express oncogenic EGFR received from cancer cells22.

While these observations are intriguing, they do not specify whether EGFR protein or mRNA is responsible for the lasting expression of EGFR/pEGFR in EV recipient cells. We reasoned that the transfer of EGFR protein alone22 (with reported EGFR half-life under 30 h31) would unlikely account for the 6-day long ectopic expression of this oncogenic receptor in endothelial cells. This may suggest a possible contribution of EGFR mRNA transfer23 by a subset of tumour EVs. In this regard, a considerable heterogeneity of GSC EVs was previously extensively characterised highlighting the preponderance of neutral sphingomyelinases (SMPD2/3/4) in MES-GSCs and of exosomal markers in their related EVs26. To test the role of this EV biogenesis pathway32, MSC-GSCs were treated with the neutral sphigomyelinase inhibitor (GW4869) and tested for the EV output and EGFR content. While the overall EV release was not changed post GW4869 treatment (Supplementary Fig. 3a) and their content of EGFR mRNA remained comparable, the corresponding levels of EGFR protein declined dramatically (Fig. 2e). These observations suggested the existence of several pathways (GW4869-sensitive and -insensitive) in MES-GSCs, which separately traffic EGFR protein or RNA into EVs and the pericellular milieu. To further explore this possibility, MES-GSC EVs were incubated with magnetic beads coated with anti-EGFR antibody to separate/deplete EVs expressing EGFR protein on their surfaces (Fig. 2f). This treatment efficiently removed the EGFR-antigen-positive EVs from the particulate secretome, while the remaining EVs (flow through) were enriched for EGFR mRNA. Remarkably, EVs containing EGFR and EGFRvIII mRNA (but depleted for EGFR/EGFRvIII proteins) efficiently transferred GSC-related EGFR and EGFRvIII to recipient endothelial cells and their effect persisted for several days (Fig. 2g). EGFRvIII mRNA was also directly detected in endothelial cells exposed to unfractionated MES-GSC-EVs (Supplementary Fig. 3b). These observations suggests that the EV-mediated intercellular transfer of EGFR/EGFRvIII mRNA is sufficient to trigger the EGFR/EGFRvIII protein expression in endothelial cells.

Intercellular transfer of EGFR activity leads to endothelial cell responses that parallel vasectasia

It could be expected that disabling the ectopic (EV-related) EGFR activity either before or after internalisation by endothelial cells, would block their responses to MES-GSC EVs. To assess this possibility, cultured endothelial cells were exposed to MES-GSC EVs and subsequently treated with Dacomitinib (DAC), a potent, clinically relevant, and irreversible panErbB (EGFR) inhibitor. Indeed, treatment with optimised concentrations of DAC (Supplementary Fig. 4a) reversed EGFR phosphorylation in endothelial EV recipient cells (Fig. 3a) and dramatically suppressed the MES-GSC EV-stimulated endothelial cell migration and formation of aortic ring outgrowths (Fig. 3b; Supplementary Fig. 4b). In line with these findings, depletion of EGFR in MES-GSC by CRISPR/Cas9-mediated editing of the EGFR/EGFRvIII gene (KO), resulted in a marked diminution of the EV-related ability to stimulate endothelial aortic ring outgrowths and endothelial cell migration (Fig. 3c; Supplementary Fig. 5a–d). Unexpectedly, EGFR-KO did not impact the proliferation of MES-GSC cells in complete growth media in vitro (Fig. 3d) suggesting that oncogenic EGFR is not essential for intrinsic control of growth and survival of these cells, but instead may possess non-cell-autonomous functions in GBM, at least in part, mediated by its EV-mediated export.

Fig. 3: Obliteration of GSC-EV-derived EGFR in endothelial cells suppresses cellular responses to extracellular vesicles and alters vascular patterning in vivo.
figure 3

a Activation of EGFR in endothelial cells following EV transfer from cancer cells is obliterated by the pan-ErbB inhibitor, Dacomitinib. EGFR phosphorylation in primary endothelial cells is completely inhibited by 2.5 μM of Dacomitinib treatment (n = 3 independent experiments); b Dacomitinib inhibition of endothelial cell migration triggered by EGFR-carrying EVs. Endothelial cells were treated with MES EVs or VEGF and 6 h later exposed to 2.5 μM of Dacomitinib. The number of cells migrated was assessed using FIJI software (n = 6 wells/3 independent experiments; two-tailed paired t test P = 0.0022 and 0.00077); c EGFR depletion reduces the ability of GSC EVs to trigger endothelial cell outgrowths. Endothelial cells were treated with 30 μg/ml of EVs obtained from glioma stem cells either deficient (EGFR-KO, clone 19 and 27), or proficient (EGFR-WT) for EGFR. After 3 days of incubation, cells were fixed, stained with crystal violet and imaged (n = 3 independent experiments; two-tailed paired t test; GSC83 (83): P = 0,00024 and 0,00018; GSC1005 (1005): P = 0.00041 and 0.00019); d Proliferation assay reveals similar growth pattern in culture of glioma stem cells deficient (EGFR-KO) or proficient (EGFR-WT) for EGFR (n = 3 independent experiments); e Appearance of freshly removed BME plugs containing indicated agents three weeks after implantation. BME-embedded EVs were obtained from glioma stem cells: EGFR-KO, EGFR-WT, while control plugs contained VEGF, or vehicle. Scale bars are 20 µm (n = 4 independent experiments); f Kaplan-Meier survival curves of mice bearing EGFR-KO and EGFR-WT MES GSC-driven tumours (n = 5 mice per group); g Representative images of immunofluorescence for CD31 reveals differential vascular patterns between tumours driven by EGFR-WT or EGFR-KO MES-GSCs (GSC83). Scale bars are 20 µm; n = 5 independent experiments have been conducted; two-tailed paired t test P = 0,0008; h Quantification of vessel size distribution according to staining for CD31-positive endothelial cells (n = 5 independent experiments; two-tailed paired t test P = 1.45−12 and 3.41−11); i Quantification of microvascular density using CD31 staining (n = 5 independent experiments; two-tailed paired t test P = 1.39−06 and 7.15−06). Microvascular density was expressed as vessel density per high power field (hpf). Data were presented as means ± SD. Significance: **p < 0.01, ***p < 0.001, ****p < 0.0001 of the treated group versus untreated control group; Source data are provided as a Source Data file.

To explore these findings further, EVs from EGFR-proficient (WT) MES-GSCs were embedded in subcutaneous implants of the basement membrane extract (BME) in mice in the absence of cancer cells. In this setting, EGFR-carrying EVs elicited formation of vascular networks reminiscent of those observed in corresponding xenografts (Fig. 1b) and comprising unusually large vessels along with smaller endothelial structures (Fig. 3e; Supplementary Fig. 6). In contrast, EVs from the isogenic EGFR-KO MES-GSCs triggered mostly capillary ingrowths (Fig. 3e; Supplementary Fig. 6), similar to those observed in VEGF-containing BME pellets. These observations suggest that EGFR-EVs may be sufficient to trigger the peculiar vascular patterning associated with MES-GSC-driven gliomas.

Moreover, EGFR gene disruption dramatically curtailed the aggressiveness of MES-GSC intracranial xenografts in vivo (Fig. 3f; Supplementary Fig. 7a). In the absence of EGFR, these tumours also failed to develop large vessel patterns and instead contained dense capillary networks (Fig. 3g–i; Supplementary Fig. 7b–d), reminiscent of those observed in PN-GSC-initiated lesions (Fig. 1b). Finally, selective restoration of the oncogenic EGFRvIII in MES-GSC-KO cell lines (KO-EGFR+) resulted in a partial restoration of their tumourigenic phenotype, with increased aggressiveness, and reduced microvascular density coupled with enlarged vessel diameters (Supplementary Fig. 8a–d).

Collectively, these observations point to at least two different blood vessel growth processes induced by GSCs in vivo. In the absence of oncogenic EGFR (PN-GSCs, EGFR-KO-GSCs), GSCs produce mostly soluble angiogenic factors (including VEGF), which appear to drive the formation of dense capillary networks, ostensibly through a process of angiogenesis2. In contrast, MES-GSCs deploy EGFR/EGFRvIII-containing EVs that possess endothelial stimulating activity and trigger responses that may lead to circumferential extension of tumour blood vessels, a process that we termed vasectasia. In the latter case, our morphometric analysis of evolving vascular patterns post-tumour cell inoculation indicated a progressive increase in vessel diameter beginning at the boundary between brain parenchyma and MES-GSC-driven tumour masses, a change that extended toward tumour interior. This response included the enlargement/remodelling of the same vessels at the point of their tumour entry and discernible as early as 5 days post cancer cell inoculation and evident by day 9 (Fig. 4a). Thus, the capillary brain vasculature external to the tumour contained mostly vessels of ~5 μm in diameter. Upon entry into the tumour, the same vessels expanded to reach diameters of 15 μm, or more, and further expanded to 25 μm and beyond by day 9 (Fig. 4a, b) to ultimately reach 50 to >100 μm at the end point (Fig. 1c).

Fig. 4: Single cells sequencing to profile transcriptomes of vascular structures in human xenograft brain tumours.
figure 4

a Enlargement of blood vessels diameter within the GSC83 tumour mass was detected 5 and 9 days after injections. Quantification of vessel size distribution was enabled by CD31 staining of endothelial cells (n = 3 independent experiments; two-tailed paired t test P = 4.22−25 and 2.83−14; Data are presented as mean values ± S.D.); b 3D reconstruction of GSC83 (83) tumour xenograft shows enlargement of blood vessels upon entry into the tumour. c T-distributed stochastic neighbour embedding (tSNE) plot shows clustering of murine cells based on gene expression. In both EGFR-KO and EGFR-WT GSC83 tumours cell colour specifies the assignment of cells to one of 15 different clusters inferred using shared nearest neighbour clustering; d tSNE plot of murine endothelial cells and pericytes shows clustering of cells based on gene expression. Two different populations of pericyte are detected and four different subpopulations of endothelial cells are identified, including: angiogenic, migrating, permeable, and proliferative; e Relative proportion of endothelial cell subpopulations in either EGFR-KO or EGFR-WT GSC83 tumours. f Relative proportion of pericyte cell populations in either EGFR-KO or EGFR-WT GSC83 tumours. g Quantification of immunostaining for vascular markers. Birc5 is selectively elevated in endothelial cells associated with vasectasia in EGFR-WT tumours, while Apln and Vegfr2 were upregulated in angiogenic blood vessels in EGFR-KO tumours (n = 3 independent experiments; two-tailed paired t test P = 6.60−23–4.44−06–1.84−16 and 0.00041); Source data are provided as a Source Data file.

Oncogenic EGFR status defines the molecular landscape of blood vessel-associated tumour microregions

To explore the molecular realm of these vascular changes, MES-GSC xenografts positive or negative for EGFR expression were subjected to digital spatial mRNA profiling with a focus on CD31+ cells associated with large or small vessels in isogenic intracranial tumours. Vasculature in these xenografts was also compared to that in normal mouse brains. Formalin-fixed paraffin-embedded tumour sections (8 lesions) were used to select 42 regions of interest (ROI) (Supplementary Fig. 9a) followed by mRNA sequencing using Illumina NovaSeq platform and processing by GeoMx NGS Pipeline (DND) (NanoString; Supplementary Fig. 9b, c). While ROI-associated transcriptomes of EGFR-positive-, EGFR-negative tumour bearing and control brains revealed several differentially expressed genes (DEGs) between large and small blood vessels, their expression was only partially restricted to endothelial cells and may reveal more global changes in the perivascular microenvironment in EGFR-driven tumours (Supplementary Fig. S9d–f). A quantitative comparison of transcripts associated with large vessels from EGFR-WT tumours (vasectasia) versus small vessels from EGFR-KO tumours (angiogenesis), as depicted in the volcano plot (Supplementary Fig. 9e), demonstrated several significant differences (>2 fold; p < 0.05) with large vessels microregions being enriched for Cd99, Pcbd2 and Ripply2 relative to small vessel microregions. The functional aspects of this enrichment were also apparent from GSEA analysis of the respective molecular pathways, which pointed to upregulation of interferon response, reduced angiogenesis signal and expression of several vascular markers in ROIs associated with large vessels of EGFR-proficient tumours (Supplementary Fig. 10). We also compared the spatial gene expression profiles of GeoMX captured microregions surrounding both large and small vessels in GSC-MES xenografts and in the adjacent normal brain vasculature of comparable calibre. Those patterns were analysed against the recently described ATLAS of vascular markers associated with the brain arteriovenous axis33. Strikingly, while genes expressed in normal brain vessels in our study (large and small) mapped to some extent with the corresponding calibre vessels described in the ATLAS, tumour-associated regions containing blood vessels of comparable sizes were markedly different (Supplementary Fig. 11a) suggesting a disease-specific imposition of abnormal regulatory influences, including both angiogenesis and vasectasia.

Molecular traits of endothelial cells associated with vasectasia

To further explore the molecular make-up of endothelial cell populations involved in vasectasia we subjected MES-GSC xenografts positive or negative for EGFR expression to single-cell transcriptional profiling (scRNAseq). The resulting subpopulations of cancer, stromal and endothelial cells were resolved computationally and compared to data reported in the literature34. Interestingly, GSC tumour cells, clustered separately, revealing distinct cellular landscapes (Supplementary Fig. 12a, b) dependent on the EGFR status. Similarly, we compared the clusters of murine (stromal) cells within both EGFR-proficient and -deficient tumours, which were biologically annotated based on the relative abundance of top-ranking marker genes (Fig. 4c and Supplementary Fig. 12c, d). In this regard, populations of microglia, monocyte/macrophages, oligodendrocytes with their precursors (OPCs), pericytes and endothelial cells were readily distinguished and comparable, but not identical between tumours with intact or disrupted EGFR gene (Supplementary Fig. 12e, f).

Interestingly, single-cell gene expression suggested the existence of at least four subpopulations of endothelial cells classified as migrating, proliferative, permeable and angiogenic. Two distinct subsets of pericytes were also detected (Fig. 4d–f; Supplementary Fig. 13a, b). Notably, EGFR-expressing tumours were enriched in proliferative endothelial cells, while EGFR-KO tumours contained mostly angiogenic endothelia (Fig. 4e).

Single-cell gene expression data also pointed to human EGFR transcripts being detectable in endothelial cells of both human brain tumours and murine brain xenografts (Supplementary Fig. 13c–e). Thus, in MES-GSC xenografts, subpopulations of CD31+ and CD34+ mouse endothelial cells harboured human EGFR transcript (Supplementary Fig. 13d). Similarly, in silico analysis of scRNAseq human GBM datasets (GSE84465) also revealed the presence of EGFR transcripts in CD34+ endothelial cells (Supplementary Fig. 13c). In addition, the presence of human EGFR protein was detected by flow cytometry in CD31-postive endothelial cells isolated from EGFR-WT xenografts (>15 fold enrichment; p < 0.001). Moreover, using magnetic bead separation (MACS) of EGFR-positive endothelial cells, we were able to detect their population containing human phosphorylated EGFR (Supplementary Fig. 13e, f). While a relatively small fraction of endothelial cells exhibited this phenotype, their sustained presence is consistent with in vitro results suggesting a role of activated endothelial EGFR in vasectasia. Finally, heterogeneous populations of microglia and myeloid cells were also detected and varied between tumours with different EGFR status (Supplementary Fig. 14a–e), while oligodendrocytes and their precursors (OPCs) did not exhibit major differences in this regard (Supplementary Fig. 14f).

Molecular traits associated with glioma stem cell-driven neovascularization patterns

A more in-depth examination of differentially expressed transcripts between EGFR-expressing and -non-expressing MES-GSC driven tumours additionally exposed specific molecular distinctions between vascular patterns dominated by either angiogenesis or vesectasia. As predicted from scRNAseq profiles, tissue immunostaining revealed that protein markers of endothelial tip cells associated with capillary sprouting outgrowth, such as Apln (apelin) and Vegfr2 (VEGF receptor 2), were widely expressed among CD31+ cells in MES-GSC83-KO tumours (with disrupted EGFR gene expression). In contrast, larger vessels from EGFR-expressing tumours were enriched for Birc2 (survivin), Socs2 (suppressor of cytokine signalling 2) and Srsf2 (serine and arginine-rich splicing factor 2). In addition, the presence of Ki67 staining not only in angiogenic (KO-EGFR), but also in vasectasia-related (EGFR-WT) blood vessels implicated the ongoing endothelial proliferation, as a part of both processes (Fig. 4g, Supplementary Fig. 15a). Finally, RNAscope in situ hybridisation (ISH) analysis further confirmed the presence of distinct transcripts (Socs2, Srsf2, Birc2) in endothelial cells associated with MES-GSC-driven and EGFR expressing tumours (Supplementary Fig. 15b).

Targeting EGFR disrupts vasectasia and overcomes resistance to anti-VEGF therapy

Antiangiogenic VEGF-directed therapies have not improved overall survival in unstratified GBM patient populations1 likely due to alternative and unrecognised mechanisms of tumour neovascularization17. Since aggressive MES-GSCs express mediators of both angiogenesis (VEGF) and vasectesia (EGFR-EVs; Fig. 1i–k), we considered the consequences of targeting both of these respective processes in vivo, either separately or simultaneously (Fig. 5). To this end, NSG mice were intracranially inoculated with MES-GSC cells (GSC83) expressing Luciferase. Once the bioluminescent tumour signal became apparent, the animals were randomised to receive anti-mouse VEGFR2 blocking antibody (DC101), EGFR inhibitor (DAC), both, or vehicle controls (IgG, Lactate; Fig. 5a; Supplementary Fig. 16a, b;). Interestingly, the administration of DC101 delayed tumour growth somewhat, but this effect wore off within 2–3 weeks and tumours resumed rapid growth trajectory. DAC therapy inhibited GSC83 tumour progression (in spite of the resistance of these cells to DAC treatment in vitro—Fig. 3d), but prolongation of survival was modest. Strikingly, the symptom-free survival of mice receiving both agents was markedly extended and reached approximately 30 days beyond the baseline (3-fold extension; Fig. 5b; Supplementary Fig. 17a).

Fig. 5: Combined targeting of VEGFR2 and EGFR suppresses tumour growth and vasectasia.
figure 5

a Schematic representation of the treatment with EGFR and VEGFR-2 inhibitors. Dacomitinib was administered by gavage at 15 mg/kg for 5 consecutive days followed by a 2-day break, while DC-101 was injected intra-peritoneally at 20 mg/kg, twice a week. Lactate and IgG at corresponding concentrations were injected in parallel, as placebo controls. The mice were treated with single agents alone, or in combination; b Survival curves of mice harbouring mesenchymal glioma stem cell xenografts (GSC83) subjected to a combination therapy targeting VEGF and EGFR pathways. Kaplan-Meier plot depicts groups of mice treated with placebo (IgG + lactate), DC-101 or Dacomitinib alone or with combination therapy of DC-101+ Dacomitinib (n = 5 independent experiments; two-tailed paired t test; P = 0.0075–0.000051 and 0.0000223); c Representative images of immunofluorescent staining for CD31 reveals the impact of therapy on vascular patterns. d Quantification of vessel size distribution based on CD31 staining. Blood vessels in mice treated with placebo (IgG + lactate) and single treatment with DC-101 present enlarged lumen compared to a mean vessel diameter of GSC83 (83) tumours treated with Dacomitinib and combination therapy of DC-101+ Dacomitinib (n = 5 independent experiments; two-tailed paired t test; P = 4.99−08 and 2.18−10); e Quantification of microvascular density following CD31 staining. Microvascular density was increased in tumours treated with Dacomitinib and combination therapy of DC-101+ Dacomitinib as measured by vessel numbers per high power field (hpf; n = 5 independent experiments; two-tailed paired t test; P = 7.92−06 and 1.45−05). Data were presented as means ± SD. Significance: **p < 0.01, ***p < 0.001, ****p < 0.0001 treated group versus untreated control group; Source data are provided as a Source Data file.

These responses were associated with remarkable rearrangements of vascular patterns (Fig. 5c–e). Thus, vasectasia, a network of mostly larger vessels was observed in control tumours and following DC101 treatment alone. In contrast, tumours exposed to DAC were essentially devoid of blood vessels with diameters greater than 50 µm (hallmark of vasectasia), but exhibited dramatically elevated microvessel density, possibly due to compensatory angiogenesis Fig. 5d, e). The latter response was attenuated when DAC was combined with DC101 (Fig. 5e) suggesting a role for an interplay between EGFR and VEGF/VEGFR2 pathways in microvascular growth processes. Overall, these results are consistent with the notion that MES-GSC-initiated brain tumours mount multiple vascular growth responses, either through vasectasia driven by EGFR (EGFR-EVs) or through angiogenesis, the latter dependent, at least in part, on the VEGF pathway.

Latest Intelligence