Glioblastoma multiforme (GBM) is a grade IV glioma, characterized by aggressive vascularization and intracranial dissemination (1). Clinically, GBM is the most aggressive primary brain tumor with a median survival rate of 14 months despite maximal safe resection and adjuvant chemotherapy (2). A high level of vascularization is a hallmark of GBM. The standard of care, including treatment with temozolomide (TMZ), an alkylating agent that forms the backbone of GBM treatment, is met with a limitation of 50% chemoresistance rate (3). To overcome this clinical challenge, more recently, anti-angiogenic approaches have been used clinically, including the administration of bevacizumab (an anti-vascular endothelial growth factor (VEGF) antibody), cilengitide (an integrin inhibitor), cediranib (an anti-VEGF receptor tyrosine kinase inhibitor), and enzastaurin (a protein kinase C beta inhibitor). Despite some evidence of extending progression-free survival (PFS) and clinical improvement, these anti-angiogenic therapies by themselves failed to improve overall survival (OS) in GBM (4, 5). Therefore, it is critical to identify newer molecules with broader anti-angiogenic effects.

As mentioned above, GBM is characterized by a robust vascular network (6). In fact, the systemic levels of pro-angiogenic mediators such as VEGF, FGF-2, IL8, IL2, and GM-CSF have been shown to be twofold higher in GBM patients. Moreover, GBM is associated with a threefold increase in IL6, IL1β, and TNF-α, reflecting a rich pro-angiogenic environment (6). Several of these pro-angiogenic mediators drive multiple vascular processes, including sprouting angiogenesis, vasculogenesis involving endothelial precursor cell (EPC) recruitment, vasculogenic mimicry involving tumor cells lining blood vessels, and intussusceptive angiogenesis (7, 8). This pathologic angiogenesis, also known as neoangiogenesis, is characterized by enlarged and unorganized vessels with abnormal basement membrane and low pericyte density. Moreover, cancer cell-secreted pro-inflammatory cytokines and chemokines also increase endothelial permeability, yielding leaky vessels resulting in interstitial edema, pressure and intravascular metastasis and dissemination, and ultimately death (9). Of note, in addition to inhibiting neoangiogenesis, anti-VEGF therapy has also been shown to reduce vasogenic cerebral edema in brain tumors (10).

Rapidly progressing tumors exhibit a hypoxic microenvironment, especially in the central tumor core. In fact, pseudopalisading necrosis, a histological hallmark of GBM, reflects progression of the core to ischemic necrosis. The pseudopalisading cells are bordered by microvascular hyperplasia, largely driven by hypoxia-induced HIF-1α activation (11). Interestingly, independent of microenvironmental hypoxia, GBM is predisposed to HIF-1α activation due to EGFR (epidermal growth factor receptor) gene amplification and EGFR-dependent PI3K/AKT/mTOR signaling (11). In p53-mutant or p53-deleted GBM, lack of p53 promotes MDM2 ubiquitination and HIF-1α degradation, leading to increased HIF-1α expression and angiogenesis (11). Also, inflammatory cytokines promote angiogenesis, a mechanism amenable to treatment with monoclonal antibodies and small-molecule inhibitors. For example, IL1β induces VEGF expression through p38MAPK- and JNK-dependent AP-1 and NF-κB activation (12). IL6 promotes angiogenesis through the JAK2/STAT3 pathway (13). IL8 induces angiogenesis through JAK2/STAT3 and PI3K/Akt pathways. Of note, PI3K has been shown to independently activate STAT3 (14).

IL-17A, a member of the unique IL-17 cytokine family, plays a causal role in tumor biology, including colorectal cancer, lung cancer, pancreatic cancer, and breast cancer (15). In addition to its role in angiogenesis, IL-17 also exerts pleiotropic pro-tumorigenic effects in GBM. It is prominent in the tumor microenvironment (TME) and significantly overexpressed in GBM (16). In GBM cells, IL-17 promotes cell migration and invasion, with concomitant increases in PI3K, AKT, MMP2/9, Twist, and Bmi1 and reduced expression of the tight-junction protein ZO-1 (16). In contrast, IL-17 has been shown to contribute to the formation of an immunosuppressive TME via decreasing CD8+ T cells and increasing myeloid-derived suppressor cells (17). Consistently, inhibiting IL-17 increases the anti-tumorigenic potential of tumor-infiltrating lymphocytes (18). The synergistic anti-tumorigenic effects of immunotherapy and anti-angiogenic therapy are well recognized (19). Developing strategies and targets that exert anti-angiogenic and immunotherapeutic effects are a major clinical goal.

IL-17 signals predominantly via the IL-17R complex that activates inflammatory signaling. For example, binding of IL-17A to the IL17RA/RC complex recruits TRAF3IP2 (also known as CIKS or ACT1) via a homotypic interaction mediated by the SEFIR domain, resulting in activation of multiple downstream signaling cascades. TRAF3IP2 exerts a U-box E3 ubiquitin ligase activity to polyubiquinate TRAF6 at K63, recruitment of TAK1, activation of IKK, IkBa phosphorylation/degradation, and NF-kB activation. TRAF6 also activates MAPK and c/EBP. By associating with the EGFR and IL-17R complex, TRAF3IP2 also activates ERK5 (20, 21). It also plays a role in mRNA stability of some chemokines, including CXCL1. Interestingly, both IKK and TBK1 phosphorylate TRAF3IP2 at serine residue 331, leading to the formation of the TRAF3IP2/TRAF2/TRAF5/ASF complex, which prevents ASF binding to the 3′ UTR of CXCL1 mRNA and CXCL1 mRNA degradation. Interestingly, CXCL1 is a potent driver of angiogenesis (22), and its increased expression predicts poor outcomes in GBM (23). These signaling interactions, downstream of IL-17/IL-17R, also activate stress-activated kinases and pro-angiogenic transcription factors, including p38MAPK and AP-1, which regulate the expression of VEGF and ANG2. We have previously shown that TRAF3IP2 is expressed at significantly higher levels in malignant U87 and U118 glioblastoma cells compared to the non-malignant glial cell line SVG p12 (24). We have also demonstrated that targeting TRAF3IP2 reduces VEGF expression, likely through blockade of NF-κB signaling, in a flank model of GBM (24). Indeed, NF-κB activation is a critical regulator of VEGF expression in glioblastoma (25). Here, we extend our findings to an intracranial model of GBM to accurately recapitulate neoangiogenesis in the brain tumor microenvironment. Because of the comprehensive function and upstream position of TRAF3IP2, we tested the hypothesis that targeting TRAF3IP2 decreases neoangiogenesis.

For the first time, our novel data show that TRAF3IP2 levels are significantly upregulated in highly vascularized areas in the GBM tumor microenvironment, including the tumor-infiltrating region ( Figure 1 ). This strongly implicates TRAF3IP2 as a critical contributor of GBM dissemination in the brain, which depends on angiogenesis-dependent hematogenous routes. Further, data from The Cancer Genome Atlas (TCGA) reveal that human GBM exhibits significantly higher levels of TRAF3IP2 ( Figure 4D ). Importantly, our data demonstrate that silencing TRAF3IP2 significantly reduced the expression of pro-angiogenic markers, including CD31, CD34, and VEGF, in induced intracranial GBM tumors ( Figure 2 ). Our functional in vitro results corroborate our in vivo findings and demonstrate that targeting TRAF3IP2 reduces the ability of GBM to induce brain endothelial cell-mediated vascularization. We showed that this effect is due to a significant reduction in VEGF expression ( Figure 3 ). It has been shown that angiogenic factors, such as VEGF, are released directly or via exosomes into the GBM tumor microenvironment to enhance angiogenesis (32). Therefore, it is plausible that targeting TRAF3IP2 may affect the direct as well as exosome-mediated release of angiogenic mediators. Our transcriptomic analysis identified that targeting TRAF3IP2 perturbed the expression of several genes with pro-angiogenic effects. Specifically, we identified that silencing TRAF3IP2 reduces the levels of proinflammatory cytokines, growth factors and receptors, and cell adhesion molecules that are involved in angiogenesis. Therefore, silencing TRAF3IP2 has the therapeutic potential in GBM by targeting neoangiogenesis.

In addition to inhibiting the release of pro-angiogenic mediators, our results show that silencing TRAF3IP2 in GBM cells decreases the expression of IL1b, IL6, and IL8, which are proinflammatory and pro-angiogenic ( Figure 4 ). It was recently shown that exposure of GBM cells to IL1β significantly changes the proinflammatory secretome, including IL8 and IL6 (33). IL6 promotes GBM development through enhancement of cell invasion and migration (34). Indeed, IL6 signaling through STAT3 is required for glioma development in a preclinical model (35). Increased STAT3 activation, in the form of phosphorylated STAT3, was confirmed in GBM (grade IV), anaplastic astrocytoma (grade III), and diffuse astrocytoma (grade II) (34, 35). Interestingly, STAT3 upregulates VEGF and VEGFR2, thus increasing angiogenic signaling in GBM (34). In addition to pro-angiogenic mediators, STAT3 also upregulates cyclin D1 and Bcl-2, resulting in rapid tumor growth through increased cell-cycle progression and reduced cell death (34). GBM-derived IL8 has also been shown to induce brain endothelial cell permeability and is found at elevated concentrations at the tumor margin of resection, suggesting its role in tumor invasion and dissemination (36, 37). Exposure of human brain microvascular endothelial cells to the U87-CM (U87-conditioned medium demonstrated to contain elevated IL8) induced endothelial remodeling and permeability through ERK. IL8 has also been shown to promote changes in vascular endothelial cadherin (VE-cadherin) localization away from cell junctions, suggesting structural changes that enhance vascular remodeling (36). Further, an association was shown between GBM expression of IL1β, IL6, and IL8 and overall survival and disease-free survival [(hazard ratios for increased expression range from 1.1 to 1.4 (overall survival) and 1.3 to 1.8 (disease-free survival, respectively)] ( Figures 6A, B ) ( Supplementary Figures S1C, D ; S2G, H ).

Among various growth factors and receptors, we found that targeting TRAF3IP2 decreased EGF, PGF, and PDGFRB expression ( Figure 4 ). In GBM, FGF2 enhances tumor growth, angiogenesis, and cancer stem cell renewal. FGFR1, the receptor for FGF, through activation of PLC-γ, leads to resistance to radiation- and hypoxia-induced angiogenesis (26). While the critical role of FGF2 and the therapeutic potential of targeting its activity are recognized preclinically, there is no clinical grade FGF2 antagonist or FGFR modulator. Importantly, blockade of FGF2 by a monoclonal antibody and FGFR1 by RNA interference significantly, but did not completely, arrest GBM growth, through downregulation of MAPK and AKT. However, an improvement in efficacy was observed when anti-FGF therapy was combined with TMZ (38).

In addition to inflammatory mediators with pro-angiogenic effects, our data also show an increased EGF expression in GBM ( Figure 4 ). In GBM, EGF/EGFR signaling has been shown to drive metastasis via STAT3-mediated NF-κB activation. PI3K/AKT and MAPK pathways are also critical in the induction and activation of matrix metalloproteinase 9 (MMP9) (39). In fact, secreted MMP9 enables the rapid migration and dissemination of GBM both intracranially and extracranially (40). Interestingly, MMP9 expression predicts TMZ response on survival, with increased expression of MMP9 portending decreased survival and resistance to TMZ (41). In addition to cell invasion, EGF/EGFR also drives metastasis through angiogenesis. While the myriad pro-tumorigenic roles of EGF/EGFR are reviewed elsewhere (42), its role in driving angiogenesis makes it an attractive target in blocking GBM dissemination (28).

Our data also show increased PDGFR expression in the brain xenograft model of GBM, and this was confirmed by the clinical GBM tumor sample data from TCGA and GEPIA ( Figure 4 ), and this was confirmed by the clinical GBM tumor sample data from TCGA and GEPIA ( Figure 4K ). The PDGF system and its members are also shown to promote GBM growth and metastasis, as well as neoangiogenesis (43). In fact, a histologic analysis of GBM tumors revealed the highest expression of PDGFRB, compared to other members of the PDGF family, especially in regions of hyperplastic blood vessels (43). Similarly, in regions of microvascular proliferation and the leading edge of the tumor, PDGFRB is highly expressed (43). In contrast, PDGFRB showed a decreased expression in the pseudo-palisading regions (43), indicating the critical role of PDGFRB in angiogenesis and GBM invasion, two critical phenotypes that are the subject of therapeutic intervention.

PGF (placental growth factor) is a member of the VEGF family, and its function in normal physiology is poorly understood. In cancer, however, PGF has been shown to drive neoangiogenesis. PGF binds the FLT1 receptor to directly stimulate pro-angiogenic effects. Indirectly, PGF shifts VEGFA binding from FLT1 to VEGFR2, resulting in a significant increase in VEGFA-mediated angiogenesis. In addition to its role in angiogenesis, PGF also plays a role in immunomodulation. PGF has been shown to downregulate type 1 T helper immune responses by modulating dendritic cells (44). The secretion of PGF was found to be significantly increased in the Th17 subset of T helper cells and promoted angiogenesis. PGF also supported the pathogenic differentiation of Th17 cells through STAT3. Similar to IL-6, PGF increased IL-17 production but suppressed the formation of regulatory T cells, leading to autoimmunity in the form of autoimmune encephalomyelitis and collagen-induced arthritis (45). Interestingly, TRAF3IP2 functions as an adaptor protein in canonical IL-17 signaling. Consistently, TRAF3IP2 is recognized to play critical roles in the pathogenesis and progression of various autoimmune diseases (46). We previously demonstrated the role of TRAF3IP2 in GBM for the first time in a flank model (24) and here provide novel evidence of its role in driving pathogenic angiogenesis in an intracranial model of GBM. This is significant as our intracranial model of GBM recapitulates the human brain tumor microenvironment.

In GBM, specifically, tumor-derived PGF induces the generation of TGFβ+ regulatory B cells, which suppress CD8+ T-cell proliferation and release of perforin and granzyme B (47). Consistent with its role in modulating both immunity and angiogenesis (48), our data show that targeting TRAF3IP2 downregulates PGF expression and signaling ( Figure 4 ), suggesting that targeting TRAF3IP2 potentially suppresses both angiogenesis and immune suppression in GBM. Therefore, TRAF3IP2 may be an ideal target to decrease angiogenesis and improve an antitumor immune response in GBM. This is especially relevant since the combination of anti-angiogenic therapy and immune checkpoint blockade, which stimulates antitumor immune response, is suggested to be a more effective in treating malignancy (49).

VEGF is considered the most potent pro-angiogenic cytokine, and drugs, including bevacizumab, a monoclonal antibody against VEGF, have been designed to target its function in GBM (50, 51). However, these drugs have limitations, as evidenced by the eventual progression of the tumor and recurrence (52). In addition to promoting neoangiogenesis in GBM, VEGF has also been shown to promote proliferation of GBM stem-like cells through VEGFR2 (53). Paradoxically, despite a reduction in blood supply and angiogenesis, the anti-VEGF treatment has been shown to increase tumor cell invasion in GBM (29), potentially due to an enhanced hypoxic microenvironment and a feedback activation of HIF-1a and PI3K. Both pathways drive pro-tumorigenic signaling, including metabolic changes favoring glycolysis, which fuels cell division through the pentose phosphate cycle, and increases invasiveness partly through the acidification of the ECM by lactic acid. This critical evidence explains the suboptimal clinical outcomes from anti-VEGF therapy and identifies invasiveness and metabolic remodeling as alternative pathways that GBM activates to resist anti-angiogenesis therapy. Nevertheless, the anti-VEGF-mediated reduction of vascularization is an important element in the treatment of GBM and might be more efficacious as an adjunct to other therapies. For example, VEGF blockade results in a more mature dendritic cell phenotype in the brain and reduces levels of PD-1 (exhaustion marker and suppressor of activity CD8+ T cells) on brain-infiltrating CD8+ T cells, changes that result in improved antitumor immunity (54). This is consistent with the current view that immunotherapy combined with anti-angiogenic therapy might promote better outcomes.

Angiopoietin 2 (ANGPT2) has also been identified as a therapeutic target in GBM (55). ANGPT2 binds to its receptor tyrosine kinase, TEK (Tie2). During inflammation, ANGPT2 via TEK leads to increased endothelial permeability (56). In human brain tumors, its expression was found to be markedly elevated, compared to low or undetectable levels in a normal brain (55). ANGPT2 levels also positively correlate with WHO grade of tumors in addition to infiltrating macrophages/monocytes. GBM patients were also found to have increased levels of ANGPT2 preoperatively (55). ANGPT2 is expressed at the site of vascular remodeling, such as tumor vessels in the leading and infiltrating edges, as well as microvascular hyperplastic regions of GBM (57). ANGPT2 has also been postulated to be partly responsible for bevacizumab resistance, suggesting that dual targeting of VEGF and ANGPT2 may be more effective (30). In fact, bevacizumab was shown to upregulate ANGPT2, and blockade of both VEGF and ANGPT2 extended survival, decreased vascular permeability by potentially reducing vasogenic cerebral edema, and favored antitumor immunity by enhancing tumor infiltrating lymphocytes and decreasing tumor-associated macrophages (30).

Our robust mechanistic pathway analysis also demonstrates that silencing TRAF3IP2 regulates the expression of pro-angiogenic mediators ( Figure 5 ). These data are supported by clinical tumor data from TCGA ( Figure 4 ), which reveal that in human GBM tumors, TRAF3IP2 expression as well as VEGF, IL6, ANGPT2, IL8, PGF, IL1b, ITGAV, and VEGFR2 are significantly upregulated compared to normal tissue. Further, generally, hazard ratios for high expression of such pro-angiogenic genes are >1 for both OS and disease-free survival (DFS) ( Figure 6 ; Supplementary Figure S3 ). It is important to note, however, the weakness of looking at bulk survival data, as it does not factor variables which may be confounding, such as age, gender, and, importantly, treatment. While significance was generally not achieved by our analysis (95% confidence intervals), it is important to note the general trend that a higher expression of these genes is associated with both worse OS and DFS. Future analysis should prioritize stratification to avoid confounding, yet also maintain sufficient group sizes to achieve significance.

Using a brain xenograft model, we demonstrate the critical role of TRAF3IP2 in driving GBM angiogenesis. TRAF3IP2 is expressed at higher levels in GBM tissue and correlates with VEGF expression and vascularity. Our studies further demonstrate that targeting TRAF3IP2 inhibits angiogenesis by reducing the secretion of multiple pro-angiogenic mediators, including VEGF, leading to reduced angiogenesis. This affect was restored sub-maximally by the addition of recombinant VEGF, suggesting that TRAF3IP2 also induces non-VEGF factors that drive angiogenesis. Pathway analysis reveals that targeting TRAF3IP2 significantly reduces the expression of multiple pro-angiogenic cytokines and factors and/or their receptors. Further, TRAF3IP2 itself is a potent activator of NF-κB, MAPK, JNK, and AP-1, all of which induce the expression of pro-angiogenic genes. Due to inhibition of multiple pro-angiogenic mediators, targeting TRAF3IP2 as an adjunctive therapy may potentially increase the efficacy of current anti-GBM treatment regimens with anti-angiogenic components. Future studies will assess the efficacy of targeting TRAF3IP2 as a monotherapy or a combination therapy in robust animal models of heterogeneous primary GBM, with reductions in tumor size, angiogenesis, cerebral edema, and improvements in overall and disease-free survival as critical endpoints.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by Animal Care and Use Committee at the Tulane University School of Medicine in New Orleans, LA, and conformed to the Guide for the Care and Use of Laboratory Animals, published by the National Institutes of Health (DRR/National Institutes of Health, 1996).

AI, FD, ZB, and KW executed this study. RM, SB, and AD performed in vitro assays. RI and EA provided the rationale and design for the study. RI, EA, BC, and AI wrote the manuscript. All authors read and approved the final manuscript.

This project is supported through grants from the Alliance of Cardiovascular Researchers (to AI, KW, RI and to EA). Additionally, RI received a grant for this study from the Elsa U. Pardee Foundation. Also, this work was supported in part by a grant from the National Institutes of Health (P30GM145498) to RI. BC is a recipient of the Department of Veterans Affairs Research Career Scientist award (BX004016) and is supported by the U.S. Department of Veterans Affairs, Office of Research and Development-Biomedical Laboratory Research and Development (ORD-BLRD) Service Award VA-I01-BX004220.

The authors would like to specially thank and express their gratitude to Hamid Izadpanah for the graphical works in Figure 5 , and the Department of Pathology and Cancer Center at Tulane for histology and flow cytometry. We also thank Yasmine Rashad for her contribution to this manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.