Glioma Cell Secretion: A Driver of Tumor Progression and a Potential Therapeutic Target

Glial cells in the central nervous system (CNS) provide trophic support for neurons (1). In glial tumors, this trophic support is dysregulated creating a pro-oncogenic microenvironment mediated by a heterogeneous array of molecules secreted into the extracellular space (2–15). The glioma secretome includes proteins, nucleic acids, and metabolites that are often overexpressed in malignant tissue and contribute to virtually every aspect of cancer pathology (Table 1; Fig. 1; refs. 2–15), providing a strong rationale to target the cancer cell–secretory mechanisms.

Although the specific mechanisms regulating secretion in malignant cells remain to be fully characterized, there is significant evidence that the secretory mechanisms themselves are altered during oncogenesis (8, 16–36). Well-known mediators of secretion, like the ADP-ribosylation factors (ARF) and the small Rab GTPase proteins (RAB) have been reported to be dysregulated in glioma and several other tumors (17, 20, 22, 28, 31, 32, 37–41). These proteins facilitate secretion of pro-oncogenic molecules (28, 42) and their inhibition diminishes multiple aspects of cancer pathology including cellular proliferation, survival, and invasion (20, 22, 28, 32, 37–39, 41, 42), while showing no signs of obvious toxicities in animal models (18, 19, 21, 26, 36, 43–46). This reliance of cancer cells on secretory pathways is exemplified by the unfolded protein response (UPR; ref. 18). UPR activation is thought to be crucial for oncogenic progression (18), and agents inhibiting the UPR have shown potent antitumorigenic effects in models of glioma, multiple myeloma, and pancreatic cancer (18). Tumor cell secretory “addiction” describes the dependence of tumor cells on secretory pathways like the UPR (18), and suggest a potential therapeutic window to target these pathways. The functional impact of secreted molecules and secretory pathways on glioma biology underscores the potential therapeutic implications of targeting the tumor cell secretion (Table 1).

Glioma cells modify their microenvironment by introducing diverse molecules into the extracellular space (Table 1). To exemplify the pro-oncogenic role that secreted molecules can have on glioma pathology, we review the functional impact of specific cytokines, metabolites, and nucleic acids on glioma biology. By describing some of the potent antitumorigenic effects observed in preclinical therapeutic studies targeting tumor cell secretion, we also highlight how blocking secreted molecules might be of therapeutic impact in gliomas, as well as other tumors.

Cytokines are essential mediators of cellular signaling (2, 13, 15). In glioma, secreted cytokines, including IL1β, IL6, and IL8, create a state of chronic inflammation that promotes the malignant phenotype (15). These cytokines are associated with poor prognosis for patients with high-grade gliomas (HGG; refs. 2, 13, 15). Both in vitro and in vivo studies demonstrate that targeting these mediators of inflammation inhibits important aspects of glioma pathology including angiogenesis, proliferation, and invasion (2, 13, 15). It has also been shown that cytokines, like IL6, IL8, EGF, and TGFβ, promote resistance to antineoplastic therapy in glioma (15, 24, 47, 48), breast (49), and prostate cancer (50). In melanoma, secretion has been identified as an important mechanism facilitating the emergence of drug resistance via the activation of the AKT pathway (51). Importantly, these cytokines also facilitate the maintenance of cancer stem cells (Table 1; ref. 15), which are largely refractory to therapy, and play an important role in cancer progression (52).

Platelet-derived growth factor (PDGF), one of the best characterized cytokines in HGGs and other cancers (2, 13, 53–55), is the ideal example to illustrate the functional impact of cytokines on cancer biology. Autocrine PDGF signaling was determined to play an important role in malignant transformation (2), and mouse models of HGG demonstrate that PDGF signaling is sufficient for tumor initiation and progression (13, 54). Dysregulated PDGF signaling activates MAPK-ERK and PI3K-AKT, two nodal points critical for cell proliferation, resistance to apoptosis, and invasion (10, 12, 56). PDGF-mediated PI3K/AKT activation has also been shown to regulate glucose metabolism facilitating the Warburg effect in HGGs (12). In vitro and in vivo studies confirm PDGF's recognized function enhancing tumor angiogenesis by stimulating endothelial cell migration and promoting endothelial cell proliferation (2). Consistent with these findings, disrupting PDGF signaling markedly reduces angiogenesis, tumor growth, and invasion in multiple mouse models of glioma (2, 10, 12, 13).

Malignant cells reprogram their metabolism to meet the bioenergetic and biosynthetic demands of tumor growth (57). A consequence of this altered metabolism is the aberrant production of metabolites that can have profound effects on tumor biology (57). Studies suggest that several metabolites, including well-known molecules like glutamate and 2-hydroxyglutarate, are involved in glioma progression (58, 59).

A recently described, secreted pro-oncogenic metabolite in glioma is kynurenine (4, 11). Kynurenine is an intermediate in l-tryptophan catabolism and an endogenous aryl hydrocarbon receptor (AhR) agonist (4, 11). At levels secreted by glioma cells, but not at levels secreted by normal cells, kynurenine promotes genomic instability of HGG-derived cells (4). AhR activation by secreted kynurenine upregulates the trans-lesion synthesis polymerase, hpol κ (4). The dysregulated overexpression of this polymerase, which normally mediates the DNA damage tolerance pathway, increases genomic instability and promotes tumorigenesis (4, 60). Such an autocrine pathway mediating increased genomic instability may contribute to the exceptionally high levels of genomic heterogeneity characterizing HGG (61).

Kynurenine also has been shown to have immunosuppressive effects, to enhance cellular motility and to promote survival of HGG cells (11). Inhibition of kynurenine secretion in glioma increases the lysis of glioma cells by alloreactive blood mononuclear cells, decreases cellular migration in vitro, and markedly decreases clonogenic survival of glioma cells (11). Consistent with these observations, kynurenine-secreting tumors grow significantly more rapidly and have a higher proliferation index than nonsecreting tumors in immunocompetent mice (11).

Secreted nucleic acids are commonly found in the blood of patients with cancer and they have been shown to be indicative of disease progression (62). In glioma and other tumors, nucleic acids can be secreted in extracellular vesicles (EV) and delivered to nearby cells (3, 7, 9, 14). EVs can carry mutated or amplified oncogene sequences, mRNA, transposable elements, or miRNAs (7, 9, 14). EVs from HGG contain an array of pro-oncogenic miRNAs such as miR-19b, miR-20, and miR-21 that promote HGG progression (14). Importantly, genetic material contained in these EVs, such as mRNA and miRNAs, can be delivered to nearby cells where they remain functional (7, 9, 14). Glioma-derived cells incubated in medium supplemented with EVs from patient-derived glioblastoma cells exhibit increased proliferation when compared with glioma cells incubated in normal, untreated growth medium (14). Although further experimentation is required to demonstrate specific effects of each nucleic acid carried within EVs, it is highly likely that they are, at least in part, responsible for these EV-mediated effects in glioma.

Studies have demonstrated that EVs derived from glioma cells can transform fibroblasts and epithelial cells (8). Consistent with these findings, targeting proteins required for EV biogenesis effectively blocks EV-induced oncogenic transformation (8). These findings suggest that DNA transfer in EVs could transform nearby cells contributing to the high degree of heterogeneity and polyclonality observed in human HGGs (61, 63).

Malignant cells utilize the classical secretory pathway. In this pathway, proteins with a signal peptide are recognized by the signal recognition particle and then inserted into the endoplasmic reticulum (ER; ref. 64). In the ER, translation is completed and the newly synthesized proteins are transported through the Golgi apparatus to the plasma membrane in a process mediated by numerous proteins including RABs and ARFs (64).

Accumulating evidence indicates that RABs and ARFs are of significant importance for cancer progression (17, 20, 28, 37, 65, 66). Aberrant expression of RAB proteins, a family of more than 70 members, is observed in glioma and other cancers (22, 28, 31, 32, 40, 67, 68). In HGG RAB3A, RAB34, and RAB27A expression levels correlate with survival and pathologic grade (22, 31, 40). RAB3A, which regulates vesicular transport in neurons (69), is associated with increased proliferation and chemo- and radioresistance of cultured glioma cells and tumor formation and glioma growth in nude mice (22). A recent study has identified RAB35 as a novel activator of PI3K in multiple cells from kidney, colon, prostate, and cervical cancer. Somatic RAB35 mutations found in human tumors were capable of transforming cells in vitro, and this oncogenic RAB35 was sufficient to drive vesicular transport in the absence of stimuli, suggesting that dysregulated vesicular trafficking can contribute to oncogenesis (70).

Like the RAB family of proteins, ARFs and their associated guanine nucleotide exchange factors (GEF) and GTPase-activating proteins (GAP) are emerging as novel mediators of disease progression in gliomas, as well as other tumors (17, 20, 36–39, 65, 66). Dysregulated ARF6 has been reported to be a key driver of tumor cell invasion in several cancer types including glioma (17, 20, 65, 66). A glioma study has shown that ARF6 knockdown by siRNA markedly decreases invasion of intracranial brain xenografts in nude mice (20). ARF6 has also been shown to orchestrate the activity of the oncogenic guanine nucleotide–binding protein G(q) subunit alpha (GNAQ) in uveal melanoma by mediating the vesicular transport of GNAQ and β-catenin (71). Consistent with this finding, targeting ARF6 with the small-molecule inhibitor, NAV-2729, reduces melanoma proliferation and tumorigenesis in vivo (71).

A substantial proportion of cancer secretion is mediated by the nonclassical secretory pathway (29). This pathway is induced by cellular stress (72), a characteristic of transformed cells (18, 73). There are four types of nonclassical pathway secretion, which may play a role in cancer: type I or pore-mediated secretion, type II or ABC transporter-mediated secretion, type III or endosome/autophagosome-mediated secretion, and type IV or Golgi-bypass secretion (72).

Proteins secreted by the type I pathway rely on the formation of pores in the plasma membrane. The formation of these pores can be mediated by either the cytoplasmic protein to be secreted or by inflammation (72). Secretion of FGF2, a widely studied pro-oncogenic protein of importance in glioma (5), is a well-known example of the former mechanism (72). FGF2 binds to phosphatidylinositol 4, 5-bisphosphate (PI(4,5)P2) on the plasma membrane and oligomerizes promoting the formation of a lipidic pore through which secretion occurs (72). Alternatively, pore formation can be mediated by inflammation (72, 74), which characterizes the glioma microenvironment (75). During inflammation, gasdermin D, gasdermin A, and gasdermin A3 are cleaved by caspase-1 releasing the N-terminal half of these proteins: gasdermin-N, which binds PIP2 on the plasma membrane and forms 16-fold-symmetry in the pore (72, 74). These pores then mediate the secretion of IL1β (74), a cytokine known to promote further inflammation and drive glioma growth (75, 76). While a role for the gasdermin protein family in glioma has not been proposed, members of this family are aberrantly expressed in several tumor types (16) and promote brain metastasis of tumors from outside the CNS (77).

The type II nonclassical pathway relies on the ATP-binding cassette (ABC) transporters (72), which are well-described mediators of chemoresistance in glioma as well as other tumor types (27, 78). These transporters facilitate chemoresistance by mediating the efflux of chemotherapeutic agents from malignant cells (27, 78). Overexpression of the ABC transporter subfamily B member 1, which is also known as the multidrug resistance protein 1 (MDR1), mediates resistance to chemotherapeutic agents in aggressive gliomas (27, 78). Inhibiting MDR1, in vitro or in vivo, increases the sensitivity of HGG cells to vincristine and temozolomide, two antineoplastic agents commonly used in the management of patients with glioblastoma (78). Furthermore, MDR1 inhibitors have been shown to lead to increased concentrations of chemotherapy agents in the brain when used as adjuvants (19, 21). MDR1 is also expressed in endothelial cells of the blood–brain barrier (19), where it pumps metabolic waste products out of the CNS as well as enhancing the efflux of chemotherapy agents and decreasing their availability in brain tumors (19). Thus, clinical therapies that effectively inhibited these transporters could potentially have the added advantage of increasing the CNS availability of chemotherapy agents active in glioma (21), although to date, MDR1 inhibitors have not been useful in the clinic due to the side effects they cause (19). Multiple strategies to prevent such drug toxicities are currently under investigation (26).

The type III pathway of nonclassical secretion relies on autophagosomes (72), double-membraned vesicles that enclose cellular material destined to be degraded or secreted (25). Secretory autophagosomes carry cytokines including IL1β, and IL18 (25, 72), which are well-characterized mediators of glioma progression (15, 75, 76). The process of autophagosome formation is tightly controlled by highly conserved autophagy-related genes (ATG; refs. 25, 72). These genes, as well as the process of autophagy, can have pro-oncogenic or antioncogenic effects in HGG (30, 79, 80). For example, targeting ATG7 or ATG13 has been shown to result in markedly decreased KRAS-driven HGG growth in vivo (80), while inhibiting ATG5 enhances the resistance of cultured HGG cells to temozolomide, a drug routinely used in the treatment of glioblastoma (30). The mechanisms behind this dual role of ATGs in HGG remain to be elucidated (79).

The type IV nonclassical pathway mediates the secretion of proteins during ER stress that bypass the Golgi enroute to the plasma membrane following synthesis (72). The IRE1α protein of the UPR activates this pathway in response to ER stress (18, 81), which is provoked by many conditions common to malignant tissues including hypoxia, nutrient deprivation, increased metabolic activity, and high levels of proliferation (18, 73, 82). Human glioma cells expressing a nonfunctional mutant of IRE1α demonstrate markedly decreased intracranial tumor growth in nude mice (82). Golgi bypass in the type IV pathway is mediated by an ER-sorting machinery that includes heat shock proteins (HSP; ref. 72). The role of HSPs in HGGs has been well described (33, 83). For example, HSP70, which is upregulated in HGGs (33, 83), increases tumor formation, survival, and chemoresistance in glioblastoma (33, 83).

Many different terms are used in the literature to describe EVs including microvesicles, exosomes, oncosomes, microparticles, and ectosomes (84, 85). Even although there is considerable debate in the field as to the proper usage of these terms (84, 85), EVs can be broadly classified into two main groups based on their biogenesis: microvesicles and exosomes (36). Microvesicles are derived from the plasma membrane, while exosomes originate from multivesicular endosomal compartments (36). However, experimental procedures designed to isolate each of these vesicle types, including centrifugation protocols, commercial kits, and filters, are based on the size and weight of these vesicles, and not on their intracellular origin (86, 87). Consequently, material isolated by these experimental techniques contain a mixture of different types of EVs (86, 87), and thus it is important to consider investigational findings in the light of these experimental limitations.

In cancer, EVs transport proteins, lipids, and nucleic acid (36) and can promote multiple aspects of cancer progression (3, 9, 14, 32, 36, 66), and even induce malignant transformation (8, 88). The small Ras homolog gene family member A (RhoA) GTPase regulates EV biogenesis in glioma cells (8, 36). Studies demonstrate that RhoA facilitates the activation of Rho kinase (ROCK) and subsequently the Lim kinase (LIMK; refs. 8, 36). LIMK phosphorylates cofilin causing a reorganization of actin fibers and microvesicle shedding (8, 36). Inhibition of these proteins has been shown to markedly reduce microvesicle release in glioma-derived cells (8). Inhibition of the RhoA–ROCK–LIMK pathway reduces cellular proliferation and migration in glioma (8, 35). Pharmacologic inhibition of this pathway by the ROCK inhibitor, Y-27632, suppresses migration and proliferation of glioma-derived cells. Similarly, silencing of LIMK reduced growth of a breast cancer xenograft in nude mice (8). Consistent with these observations, RhoA, which mediates the activation of ROCK and LIMK as discussed previously (8, 36), is highly expressed in HGGs and its expression is correlated with pathologic grade (34). RAB and ARF family members, which were discussed in previous sections, have also been implicated in EV formation in gliomas as well as other types of cancer (32, 36, 66).

The concept of tumor cell–secretory addiction describes the reliance of cancer cells on secreted molecules and their secretory machinery (18). Studies described above demonstrate the significance of this concept in HGG (Table 1). By describing some of the essential functions of glioma-secreted molecules and the mechanisms facilitating their secretion in glioma pathology (Fig. 1), we have sought to underscore the potential therapeutic benefits of targeting secretion in HGGs. Targeting proteins of the secretory mechanism in glioma has been shown to inhibit key aspects of glioma biology including tumor cell proliferation and survival (Table 1; Fig. 1; refs. 8, 17–23, 26–28, 32, 34–36, 38, 39, 44–46, 89, 90). Furthermore, targeting secretion seems to selectively affect transformed cells, while sparing normal cells, as evidenced by the lack of toxicities reported in multiple animal experiments testing agents that inhibit secretion (18, 19, 21, 26, 36, 43–46). This suggests that targeting proteins of the secretory machinery to disturb the secretion of molecules that enhance tumor pathology may be a promising treatment strategy.

Brefeldin A, a known secretion inhibitor (23), and EHT-1864, which inhibits Rac1 (91), a known mediator of secretion (92, 93), dramatically reduce VEGF secretion, decrease GSC self-renewal, and inhibit tumor growth in a mouse model of HGG (54). The antitumorigenic effects of these agents also have been observed in several other cancer types (38, 39, 41). Multiple high-throughput drug screens have identified novel small-molecule inhibitors that effectively target mediators of cellular secretion (94). Two small-molecule inhibitors of secretion that have been studied in cancer are AMF-26 (37) and Eeyarestatin I (EI; ref. 44). Like Brefeldin A, AMF-26 targets the ARF1-guanine nucleotide exchange factor (GEF) interaction, which prevents the assembly of the vesicle coating protein complex I (COPI) assembly and consequently inhibits protein secretion (37). AMF-26 has potent antitumor activity (37). Oral administration of AMF-26 resulted in complete regression of human breast cancer xenografts in nude mice, and was nontoxic to animals (37) providing further evidence of the potential of such drugs to have an acceptable therapeutic index.

Early preclinical studies also suggest that other pharmacologic agents that result in the inhibition of ARF1-mediated COPI assembly, like Exo2 (8), LM11 (95), and Golgicide A (96), have significant antitumorigenic activity in breast and prostate cancer models (97–99). However, these agents, as well as other agents that block ER-to-Golgi vesicular transport, including Exo1 (100), LG186 (100), CI-976 (101), dispergo (102), and FLI-06 (103), have not been tested in glioblastoma and their clinical value for cancer therapeutics remains largely unexplored.

EI, which inhibits Sec61-mediated protein translocation in the classical secretory pathway (44), has potent antioncogenic activity (46, 89). Preclinical studies demonstrated that EI treatment causes a significant reduction of invasion and cellular proliferation of a non–small cell lung carcinoma–derived cell line, while increasing cell death by apoptosis (46, 89). In vivo, EI treatment resulted in reduced xenograft growth in nude mice with no signs of toxicity (46). Other Sec61 inhibitors, like apratoxin (104) and cotransins (105), have also been shown to have significant antitumorigenic activity in preclinical studies (106–108). However, these agents, as well as other Sec61 inhibitors, like decatransin (109) and mycolactone (110), remain to be studied in the context of HGG, and they have yet to be studied in clinical trials for any tumor type. These pharmacologic inhibitors of secretion represent a relatively unexplored therapeutic opportunity for tumors, like HGG, that rely on secretory pathways, as discussed above.

Another potential target to inhibit cancer-associated secretion is the UPR (18). Treatment of HGG cells with epigallocatechin gallate, which inhibits GRP78, significantly increases the cytotoxic effect of temozolomide on HGG (90). This drug is currently being evaluated as an adjuvant for treatment of colorectal cancer in a phase I clinical trial (111). Similarly, inhibition of the glucose-regulated protein 78 (GRP78), a key regulator of the UPR pathway, by the novel fusion protein EGF-SubA results in substantial growth reduction of HGG xenografts in nude mice (45).

Botulinum neurotoxin–based pharmacologic agents represent another promising strategy to inhibit secretion (112). These drugs, which are known as targeted secretion inhibitors (TSI), can target specific cells and cleave SNARE proteins (112). TSIs contain a targeting domain, a transport domain, and a blocking domain. The targeting domain can be engineered to bind classic glioblastoma markers, such as EGFR (113). The transport domain facilitates the entry of this agent into the cytosol, where the enzymatic blocking domain cleaves SNARE proteins to inhibit protein secretion (112). Even although TSIs have yet to be explored in the context of cancer, their characteristics discussed above make them well suited for cancer therapeutics.

Cancer cell secretion is a vital process contributing to glioma progression. Numerous studies have characterized the autocrine and paracrine pathways that facilitate many key aspects of tumor progression described as hallmarks of cancer by Hanahan and Weinberg (Table 1; Fig. 1; ref. 114). The capacity of cancer cells to secrete a wide range of different soluble factors with redundant functions (Table 1) could also explain cancer resistance to antineoplastic therapies. In HGG, studies have shown that when signaling from a specific secreted factor is blocked, an alternative secreted factor can maintain the oncogenic functions of the secreted factor that was blocked (54, 115). These secreted factors have also been observed to promote resistance to radiation and chemotherapy (15, 24, 47–51). Therefore, targeting the secretory mechanisms of cancer cells could potentially reduce simultaneously the levels of multiple pro-oncogenic secreted factors and consequently diminish cancer drug resistance and increase patient survival. However, despite accumulating evidence demonstrating the importance of these secretory mechanisms in glioma biology (8, 16–36), there are currently no clinically available therapies targeting these mechanisms.

As pharmacologic agents targeting secretory mechanisms emerge, it will become important to cautiously consider their safety. Although several drugs targeting secretory mechanisms have been tested in animal models with no observable toxicities (18, 19, 21, 26, 36, 43–46), inhibiting such an essential cellular function as secretion poses obvious risks. Many cell types, and especially neurons, glial cells, and immune cells, require secretion to function properly. Much like current chemotherapeutic drugs, dose levels and dosing schedules are likely to require careful evaluation to target secretion from malignant cells, while sparing secretion from normal cells.

No potential conflicts of interest were disclosed.

Support was generously provided by the Theodora B. Betz Foundation (to M.A. Israel), the Jordan and Kyra Memorial Foundation (to M.A. Israel), and the Andrea Clark Nelson Medical Research Endowment (to M.A. Israel).