Small Extracellular Vesicles Released from Glioma Cells Under Hypoxic Conditions Induce Pericyte-phenotype Transition of Glioma Stem Cells.


 Background:The prognosis of malignant glioblastoma (GBM) is dismal despite advances in surgery, radiation and chemotherapy treatments. Thus, alternative therapy strategies are urgently needed. Antiangiogenic therapy for cancer offers the possibility of universal efficacy. However, preclinical and clinical studies suggest that this therapy using anti-VEGF drug Avastin (Bevacizumab) may lead to a pro-migratory phenotype in therapy resistant GBM and thus actively promote tumor invasion and recurrent tumor growth. Methods: An ultracentrifugation strategy was used to isolate glioma-derived sEVs under hypoxic or normoxic conditions. Transmission electron microscopy (TEM), Western blotting, and nanoparticle tracking analysis (NTA) were used to characterize these isolated particles. Cytochalasin D was added to disrupt cellular sEVs uptake. A tube formation assay was used to evaluate angiogenic activity, while ELISAs and Western blotting were used to assess the activated TGF-β signaling pathway. The effects of sEVs on glioma stem cells (GSCs) in vivo were evaluated using subcutaneous xenografts model system in nude mice. Immunofluorescence and immunohistochemical staining were set out to evaluate the pericyte-phenotype transition of GSCs.Results: In this present study, we showed that hypoxia could promote the release of sEVs by glioblastoma cells and hypoxia-induced glioma-derived sEVs could be taken up by GSCs. This internalization of sEVs promoted tumor growth in mouse Xenografts through the pericyte-phenotype transition of GSCs. We also demonstrated hypoxia-derived sEVs can efficiently deliver TGF-β1 to GSCs. The activated TGF-β signaling pathway mediated this kind of phenotype transition. In addition, combination of Ibrutinib and Bevacizumab showed more effective in targeting GBM. Conclusion: This present study provides a new interpretation to the failure of antiangiogenesis therapy in noncurative surgical resection of GBM, and discovers promising brain-specific therapeutic targets for this damaging tumor.


Introduction
Glioblastoma (GBM) is the most common primary malignant brain tumors, with an annual incidence of about 5 per 100 000 population and median survival 14.6 to 16 months with estimated 5 years survival rate of only 3.4% (1,2). Standard rst line management for newly diagnosed GBM involves a multimodality approach, including surgical resection, followed by radiation with concurrent and adjuvant temozolomide (3). Even with optimal surgical resection and chemo-radiation, however, essentially all excised tumors will recur, and after recurrence, the two year survival is less than 30% (4,5). Thus, alternative therapy strategies are urgently needed. One of the most distinct characteristics of GBM is the aggressive angiogenesis (6). Thus, it seems promising to perform antiangiogenic therapy to treat GBM.
As proposed by Judah Folkman in 1971, prevention of new vessel sprouts or "antiangiogenesis" could result in tumor dormancy and thus be a novel cancer therapy (7,8). Antiangiogenic therapy for cancer offers the possibility of universal e cacy, low toxicity, and little possibility of resistance (9). Blockade of the vascular endothelial growth factor (VEGF) pathway has yielded the most promising results both in animal models and in patients (10,11). Bevacizumab (Avastin), an anti-VEGF monoclonal antibody, was grant FDA approval for use in treating recurrent GBM, with high initial radiographic response and disease control rates but most patients will progress after only a few months (12,13). Upon recurrence, tumors may not enhance, secondary to vascular normalization. In addition, other recent prospective comparisons of Bevacizumab with other chemotherapeutic regimens have shown no signi cant increase in overall survival (OS) (14,15). The mechanisms of Bevacizumab failure are likely complex and multifactorial. In this study, we emphasized the importance of GSCs in playing a vital role in the failure of antiangiogenic therapy.
GSCs promote GBM pathogenesis and contribute to the suboptimal e cacy of current treatment options through mediating angiogenesis, resistance to radiotherapy and chemotherapy, invasion into normal brain, and self-renewal (16). GSCs typically reside in speci c intratumoral niches, such as the hypoxic core of a tumor or the perivascular niche (17). Once the blood vessels were disrupted by antiangiogenic therapy in tumor, GSCs then exposed to an arti cial hypoxic microenvironment. There was an excitement generated by reports suggesting that GSCs may trans-differentiate into pericytes(18). This kind of phenotypic transformation promotes angiogenesis, helps repair the damaged endothelial cells and accelerate tumor growth (19). Within this context, we hypothesized that this kind of arti cial hypoxic microenvironment caused by antiangiogenic therapy could induce GSCs to trans-differentiate into pericytes to sustain vascular normalization, making "antiangiogenesis" failure even distant or diffuse recurrence. In addition, this kind of phenotypic transition was endowed by small extracellular vesicles (sEVs),which were secreted by surrounding glioma cells under hypoxic condition.
SEVs have been described to underlie GBM initiation, invasion, angiogenesis, modulation of the immune response and chemoradiotherapy resistance (20). These particles (< 200 nm diameter) are produced when polycystic intracellular structures fuse with the cell membrane (21). SEVs secreted to extracellular microenvironment are involved in the cell-cell neighboring communication and distant cells, favoring proteins, DNA, and mRNAs/miRNAs from donor cells to recipient cells (22). The secretory quantity of sEVs and their contents vary according to their biogenesis, cell of origin and cell's status. Accumulating evidence has demonstrated that the biogenesis of secretory growth factors in sEVs are tightly controlled, and dysregulation of these factors is linked to various cancers' initiation and progression (23,24). Furthermore, sEV delivered growth factors interact with Cancer stem cells (CSCs) to confer tumoradvantageous changes that in uence tumor progression and drug resistance (25,26).
However, the knowledge regarding the interaction between GBM-derived sEVs and GSCs remains unknown. Whether this secretion of sEVs is induced by hypoxic microenvironment is unclear. Exploring the mechanisms behind this will help us not only to understand the biology of interaction between GBM cells and GSCs, which is thought to explain the failure of antiangiogenic therapy in noncurative surgical resection of gliomas, but also to discover promising brain-speci c therapeutic targets for this damaging tumor.
In this study, our ndings revealed that sEVs derived from hypoxic GBM cells induce the pericytephenotype transition of GSCs through TGF-β pathway signaling, which attenuates the therapeutic effect of "antiangiogenesis". Combination of Ibrutinib and Bevacizumab showed a stronger effectiveness in targeting GBM.

GSC Differentiation Analyses
U87 cell-derived GSC-like cells were randomly separated into ve treatment groups, after which growth media was replaced with DMEM containing 10% FBS after 0,2,4,6, or 8 days. Cellular proteins were then isolated following 0,2,4,6, or 8 days of growth, and CD133, SOX2, GFAP, and Nestin expression levels were assessed via Western blotting. SEV Isolatio n Cells were grown until 70% con uent, at which time media was exchanged for DMEM containing 10% sEV-free FBS (prepared via ultracentrifugation at 120,000×g, 4°C for 18h) and 1% penicillin/streptomycin. Cells were then cultured for 48 h under normoxic (37°C 5% CO 2 95% air) or hypoxic (1% O 2 , 94% N 2 ,and 5%CO 2 ) conditions as mentioned above, after which the media was collected and sEV were isolated via differential centrifugation. Brie y, cellular debris was removed by spinning samples for 10 minutes at 300 × g. Media was then spun for 30 minutes at 10,000 ×g at 4°C, following by passage through a 0.22 µm lter. Filtering media were then isolated via ultracentrifugation for 90 minutes at 120,000×g (Beckman Coulter, Optima XPN-100, USA) at 4°C. Precipitated sEV were resuspended in PBS and subjected to a second round of ultracentrifugation under identical conditions, after which they were resuspended in 300uL PBS at 4°C and used for downstream analyses. sEV was stored at -80°C for long term.

Limiting Dilution Assay
Limiting dilution assay was performed as previously described (27). In breif, GSC-like spheroid cells were cultured for a minimum of 7 days in serum-free media, then plated in 96-well plates (200 uL/well serumfree media), being diluted to concentrations of 20-240 cells/well. Every three days for 14 days, an additional 25 uL of serum-free media was added to each well. Spheres larger than 50 um were then counted, and a regression approach was used to determine the minimum number of cells necessary to give rise to a minimum of 1 tumorsphere per well.

Assessment of SEV Internalization by GSCs and HUVEC Cells
The uorescent PKH26dyes (MINI26-1KT, Sigma-Aldrich, USA) or PKH67dyes (green, Sigma-Aldrich, USA) were used to label sEV by combining 100 µg of sEV in 200 uL of PBS with 1 mL of an appropriately provided diluent and 1 uL of PKH67 or PKH26 for 3-5 minutes. A 1mL volume of FBS was then added to terminate labeling, and sEV were washed with PBS and reisolated via ultracentrifugation. Staining was conducted at room temperature. In order to assess sEV internalization, GSCs or HUVECs were cultured for 12 h in DMEM containing 10% sEV-free FBS on glass covers lips. The media was exchanged for media containing labeled sEV (20 µg/ml), and cells were cultured for an additional 12 h at 37°C. Cells were then xed using 4% paraformaldehyde (PFA) and stained overnight with α-tubulin (1:200, CST, 2144, USA).
Cells were then washed with PBS, stained using DAPI, blocked with an anti-uorescence quencher, and imaged via uorescence microscope(Leica, DMi8, Germany).

Western Blotting
Western blotting was performed as previously described (28

Migration Assays
GSCs (2 × 10 4 ) were added to the upper chamber of a Transwell insert (BD, 353097, USA) in a 24-well plate in serum-free DMEM/F12. DMEM containing 10% sEV-free FBS was then added to the lower chamber, and cells were incubated for 12, 24,48, or 72 h. Cotton swabs were then used to carefully remove non-migratory cells, while cells that had migrated were xed for 30 minutes with 4% PFA, washed with PBS, stained using crystal violet, and counted via microscopy. Three random elds of view in triplicate samples were analyzed for this assay.
qRT-PCR Trizol (Invitrogen, 15596026, USA) was used to isolate total RNA from GSCs, after which RNase-free DNase I (Biosharp, BL510A, China) was used to treat this RNA. A BcaBest RNA PCR kit (TaKaRa, RR023B, Japan) was then used to prepare cDNA based on provided directions, and qRT-PCR was conducted with an iQ5 Multicolor Real-Time PCR Detection System (Bio-Rad) using a Realtime PCR Master Mix (SYBR Green, RR064B, Japan) and the primers listed below, with actin being used as an endogenous control:

Immuno uorescence and Hematoxylin and Eosin (H&E) Staining
Cells or spheroids on poly-l-lysine-coated (Sigma, P4707, USA) glass coverslips were xed for 30 minutes using 4% PFA, then washed three times in PBS. Tumor tissue samples were also xed with 4% PFA prior to para n embedding and cutting into 6 um-thick sections. Samples were blocked for 30 minutes using with PBS and stained using DAPI. For H&E staining, tissue sections were stained using standard H&E staining methods. Optical or confocal microscopes (Leica, DMi8, Germany) were used to image these samples.

ELISAs
Levels of TGF-β1 in cells, sEV, or culture supernatants were assessed using ELISA kits (Elabscience, E-EL-0162c, China) based on provided directions.

Tube Formation Assay
HUVECs and GSCs (3×10 4 ) were grown in Matrigel-coated 96-well plates (BD, 356234, USA) for 6 h at 37°C. Cells were then cultured with hypoxic or normoxic sEV(50ug/ml), with PBS added as a control and cytochalasin D (5ug/ml) added to appropriate wells as above. Capillary-like tube formation was assessed over time via light microscopy, and ImageJ was used to quantify the degree of angiogenesis based upon the number of branch points formed in these tubes.

Xenografts Model
The

Statistical Analysis
Image J was used to analyze the gray values of Western blot bands and to count the cells of Transwell results. Statistical analysis was performed using GraphPad Prism 6.0. Data were presented as `x ± s, and differences between the two groups were analyzed by independent sample t test. Data were deemed to be signi cantly different at P values less than 0.05.

Result
The Internalization of Hypoxia-Induced Glioma-Derived SEVs by GSCs Promote Tumor Growth in a Mouse Xenograft.
Hypoxia is an important contributor to aggressive behavior and resistance mechanisms in GBM. We asked whether hypoxia could promote the release of sEVs by GBMs and hypoxia-induced glioma-derived sEVs could be taken up by GSCs. We rst isolated sEVs from the culture supernatants of human glioma cell lines U87 and U251 in the absence of oxygen (1% O 2 ) by density gradient centrifugation. SEVs from the two cell lines exhibited a double concave tea-dish-like morphology as determined by TEM, and hypoxia did not affect this morphology (Fig. 1A). In addition, particle size and concentration analysis by NTA approach demonstrated that hypoxia-induced sEVs diameter is ~ 110 nm, and hypoxia did not change the sEV particle diameter distribution (Fig. 1B). We further validated the presence of the known sEV markers: CD63, TSG101, CD81, CD9 and the absence of the negative markers: GRP94, GAPDH, βactin, and α-tubulin by Western blot assay (Fig. 1C). Then, we sorted and identi ed GSCs from U87 and U251 cell lines as previously reported (Supplementary Fig. 1). SEVs were stained with the PKH67(Green) lipophilic dyesin advance. Confocal microscopy was used to measure the internalization of sEVs by GSCs after 24h incubation, and microscopy analysis of the GSCs demonstrated an increasing uptake of labeled sEVs (Fig. 1D). We asked whether this internalization could promote proliferation of GBM cells. Thus, we created U87-GSCs Xenografts model using nude mice, and intraperitoneally injected them with PBS or normoxia/hypoxia-derived sEVs. After 2-week treatment, tumor volumes showed signi cant differences between each groups (Fig. 1E). Taken together, these results show the internalization of hypoxia-induced glioma-derived sEVs by GSCs promotes tumor growth in a mouse Xenograft.

SEVs Derived from Hypoxic Glioma Cells Induce Pericyte-Phenotype Transition of GSCs.
To investigate the exact effects of hypoxia-derived sEVs on promoting tumor growth. We asked whether the hypoxia-induced sEVs could induce pericyte transition of GSCs. Notably, immunohistochemical and uorescence staining of endothelial and pericyte markers, such as CD31 and α-SMA, were signi cantly increased after treated with hypoxic sEVs, indicating the increased pericyte differentiation and angiogenesis in vivo. To our surprise, we also found that hypoxia-derived sEVs increased the expression of perivascular bone marrow and X-linked (BMX), which is preferentially up-regulated in GSCs (29). This indicates that GSCs migrates to the perivascular niche prior to adopting a pericyte-like phenotype ( Fig. 2A,   B). We then set out to evaluate the effects of hypoxia-derived sEVs on pericyte transition of GSCs in vitro. Notably, the pericyte markers, including PDGFR, CD146, Desmin, and α-SMA were signi cantly increased in hypoxia-derived sEVs treated U87 /U251 derived GSCs compared with these treated with normoxiaderived sEVs. These results were consistent with the decrease of GSC markers, such as CD133 and SOX2 (Fig. 2C, D). In addition, the enhanced tube formation abilities, including increased branching, branch length, junction numbers, total branching length, and the total length of these tubules, were observed in GSCs treated with hypoxia-derived sEVs compared with these treated with normoxia-derived sEVs or PBS. In addition, these enhanced tube formation activities can be attenuated by Cyt D ( (Fig. 2E, F). These results suggest that hypoxia-derived sEVs induces the transition of GSCs to pericyte phenotype.
Next, we aimed to understand the underlying mechanism why hypoxia-derived sEVs can induce pericytephenotype transition of GSCs. As has been mentioned in a previous study, GSCs generate vascular pericytes to support vessel function and tumor growth predominantly by TGF-β (30). We rstly downloaded a microarray (GSE101113), which contains mRNA data of 56 polyclonal and monoclonal glioma stem cell lines, from GEO Database. Gene Set Enrichment Analysis (GSEA) was performed with KEGG pathway analysis, the results showed that the stemness marker of GSCs, CD15 (encoded by FUT4), was positively correlated with TGF-β signaling pathway (Fig. 3A). Thus, we determined whether TGF-β signaling pathway mediates this kind of phenotype transition. Notably, when we added hypoxia-derived sEVs to U87/U251 derived GSCs, the protein level of TGF-β1, p-Smad2, and p-Smad3 were upregulated, while the expression of Smad2 and Smad3 remained unchanged, indicating the activation of TGF-β signaling pathway (Fig. 3B). We then questioned whether the upregulated TGF-β1 in GSCs was delivered by hypoxia-derived sEVs. Therefore, we utilized ELISA to evaluate TGF-β1 expression in sEVs. We observed that TGF-β1 were upregulated to a greater level under hypoxic conditions than normoxic conditions (Fig. 3C). The upregulated TGF-β1 was also observed in hypoxia-derived sEVs, which were tested by western blotting (Fig. 3D). In addition, we set out to conduct rescue experiment using anti-TGF-β1. Notably, the pericyte markers, including PDGFR, CD146, Desmin, and α-SMA, were signi cantly decreased in U87/U251 derived GSCs treated with anti-TGF-β1. Taken together, these results suggest that TGF-β1 signaling pathway mediates the pericyte-phenotype transition of GSCs.

Combination of Ibrutinib and Bevacizumab Signi cantly Inhibits Tumor Growth In Vivo.
Since we have found that hypoxic microenvironment, which can be caused by Bevacizumab treatment in clinical practice, could induce the pericyte-phenotype transition of GSCs through hypoxia-derived sEVs. We then asked whether combination of Ibrutinib and Bevacizumab could enhance the therapeutic effect of "antiangiogenesis". Therefore, we established U87-GSCs Xenografts model and treated them with 1% O2 sEVs. We observed that the tumor size was signi cantly reduced by the application of Ibrutinib, or Bevacizumab, or together. Whatsmore, combination of Ibrutinib and Bevacizumab could mostly reduce the tumor growth compared with each alone (Fig. 4A). Notably, we observed that the staining of α-SMA and CD31 were signi cantly downregulated, consistent with the decrease of BMX, indicating angiogenesis and pericyte-phenotype transition of GSCs. These results were in accord with the group using the combination treatment of Ibrutinib and Bevacizumab. In addition, we also observed that Ki67 was signi cantly decreased. This indicates the use of Ibrutinib and Bevacizumab could not only target GSCs and endothelial cells but also glioma cells. ( Fig. 4B, C) Taken together, these ndings demonstrate combination of Ibrutinib and Bevacizumab can signi cantly inhibit tumor growth in vivo.

Discussion
The prognosis of malignant gliomas is still dismal despite advances in surgery, radiation and chemotherapy treatments, the overall 5-year survival rate for people with this type of tumor remains < 10% (31). Thus, alternative therapy strategies are needed. Malignant gliomas are upon the best vascularized tumors in humans and their proliferation is hallmarked by a distinct proliferative vascular component (32). Hence it seems to be a logical consequence to apply anti-angiogenic treatment strategies to malignant gliomas. These treatment strategies have shown promising effects in animal models and some experimental clinical studies. The antiangiogenesis drug Avastin (Bevacizumab), which targets VEGF, has become one of the most popular cancer drugs in the world (33). Anti-angiogenic therapy may lead to vascular normalization and as such facilitate conventional cytotoxic chemotherapy. However, preclinical and clinical studies suggest that anti-VEGF therapy using bevacizumab may also lead to a pro-migratory phenotype in therapy resistant GBMs and thus actively promote tumor invasion and recurrent tumor growth (34).The mechanisms of Bevacizumab failure are likely complex and multifactorial. Over time, tumors demonstrate adaptive upregulation of alternative angiogenic pathways.
One explaination is that hypoxic microenvironment caused by the treatment of Bevacizumab accelerate the formation of abnormal vasculature. However, knowledge regarding this is limited.
In this study, we found that hypoxia could promote the release of sEVs by GBMs and hypoxia-induced glioma-derived sEVs could be taken up by GSCs. In addition, this internalization of hypoxia-derived sEVs promoted tumor growth in a mouse Xenograft through the pericyte-phenotype transition of GSCs. TGF-β1 signaling pathway mediates this phenotype transition. It is well known that sEVs constitute an outstanding communication system in GBM, they can deliver not only a wide variety of coding and noncoding RNAs, but soluble proteins that can alter gene expression and cellular function in recipient cells (35,36). Evidence from the last few years show that GBMs interact with surrounding non-cancer cells through sEVs to maintain a microenvironment that favors tumor proliferation, invasion of the brain, angiogenesis and immunosuppression (37).  (40). Although the communication between sEVs released by GBM and non-cancer cells, including monocytes, macrophages, microglia, endothelial cells, astrocytes, and mesenchymal stem cells, has been fully investigated (41)(42)(43)(44)(45). The knowledge regarding the interaction between GBM-derived sEVs and GSCs is limited. As we all known, the surrounding GBM cells are the closest and most important microenvironment of GSCs in term of quantity and distance. Thus, there is a need to further understand the interaction between GBMs and GSCs through not just the reported direct or indirect communication, including cell junction, paracellular secretory and endocrine pathway(46-48). Based on our study, GSCs could internalize hypoxia-derived sEVs to accomplish pericyte-phenotype transition. It has been reported that high pericyte coverage of GBM blood vessels is associated with poor response to chemotherapy in human patients (49). Thus, this kind of phenotypic transformation can attenuate the Inhibition of Bevacizumab ,which prunes abnormal blood vessels and "normalizes" vasculature into thin-walled vessels composed of a single layer of endothelial cells in GBM.
Based on these ndings, we expect that inhibition of VEGF combined with GSCs can further improve the e cacy in treating GBM. It has been previously reported that BMX is an essential factor for maintaining GSCs-derived pericytes and inhibiting BMX with ibrutinib selectively targeted neoplastic pericytes and disrupts the blood-tumor barrier (BTB) (50). In this present study, Ibrutinib combined with Bevacizumab was utilized to treat U87-GSCs Xenografts, which were previously injected with hypoxia-derived sEVs. Notably, this combination therapy could mostly reduce the tumor growth compared with each alone. On the other hand, in addition to normalizing vasculature, Bevacizumab is also thought to have direct antitumor activity against GBM cells that express VEGF (51). Our results show the staining of Ki67 in Xenograft was signi cantly decreased after the treatment of Bevacizumab alone. This is consistent with the previous study.
In summary, our ndings revealed that sEVs derived from hypoxic GBM cells induce the pericytephenotype transition of GSCs through TGF-β pathway signaling, thus attenuating the therapeutic effect of "antiangiogenesis". Combination of Ibrutinib and Bevacizumab showed more effective in targeting GBM. This present study gives a new interpretation to the failure of antiangiogenesis therapy, and could provide a promising therapeutic strategy in targeting GBM.

Conclusion
This present study provides a new interpretation to the failure of antiangiogenesis therapy in noncurative surgical resection of gliomas, and discover promising brain-speci c therapeutic targets for this damaging tumor.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.     Representative images of U87 derived GSCs xenografts treated with Ibrutinib, Bevacizumab and combined treatment of these two, respectively. The statistical chart of tumor volume was showed on the right. (B-C). Representative immunohistochemical and immuno uorescence images of CD31, α-SMA BMX and Ki67 in GSC xenografts, whcih were treated with Ibrutinib, Bevacizumab and combined treatment of these two respectively. Data are means ± SD.*, P<0.05.**, P<0.01. ***, P<0.001.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.