Novel therapies hijack the blood–brain barrier to eradicate glioblastoma cancer stem cells

Abstract

Glioblastoma (GBM) is amongst the most aggressive brain tumors with a dismal prognosis. Despite significant advances in the current multimodality therapy including surgery, postoperative radiotherapy (RT) and temozolomide (TMZ)-based concomitant and adjuvant chemotherapy (CT), tumor recurrence is nearly universal with poor patient outcomes. These limitations are in part due to poor drug penetration through the blood–brain barrier (BBB) and resistance to CT and RT by a small population of cancer cells recognized as tumor-initiating cells or cancer stem cells (CSCs). Though CT and RT kill the bulk of the tumor cells, they fail to affect CSCs, resulting in their enrichment and their development into more refractory tumors. Therefore, identifying the mechanisms of resistance and developing therapies that specifically target CSCs can improve response, prevent the development of refractory tumors and increase overall survival of GBM patients. Small molecule inhibitors that can breach the BBB and selectively target CSCs are emerging. In this review, we have summarized the recent advancements in understanding the GBM CSC-specific signaling pathways, the CSC–tumor microenvironment niche that contributes to CT and RT resistance and the use of novel combination therapies of small molecule inhibitors that may be used in conjunction with TMZ-based chemoradiation for effective management of GBM.

Introduction

Glioblastoma (GBM) is the most common malignant brain tumor in adults (1) with a 5-year survival rate ranging from 4 to 5% (2). The standard treatment options for newly diagnosed GBM include maximal feasible surgical resection, followed by radiotherapy (RT) and temozolomide (TMZ)-based concomitant and adjuvant chemotherapy (CT) (3). Despite this multimodality therapeutic intervention, GBM is universally fatal (4). Several recent studies have demonstrated that GBM is relatively resistant to CT and RT (5–7), in part due to the presence of small subset of malignant cells called cancer initiating cells or cancer stem cells (CSCs) (6,7).

CSCs are known to have indefinite ability for self-renewal, tumor initiation and propagation (8,9). Identified in 2002 by Ignatova et al. (10), in surgically resected GBM, these CSCs differ from neuronal stem cells (NSCs) in the expression of specific mRNA’s, such as Notch-signaling ligands Jagged-2, Delta (10), bone marrow X-linked kinase (11), nitric oxide synthase-2 (12) and genetic or karyotypic alterations (13). Furthermore, these GBM CSCs are more proliferative than NSCs and will form tumors upon orthotopic transplantation (14). In addition, GBM CSCs are resistant to CT and RT (7,15–17) and have greater oncogenic potential than differentiated tumor cells (18). These GBM CSCs are often characterized by the cell surface glycoprotein marker CD133 (Prom1). Interestingly, a recent study showed that intracranial injection of as few as 100 CD133⁺ brain tumor cells formed tumors in non-obese diabetic-severe immunocompromised (NOD-SCID) mice, whereas no tumors were formed after injection of 10⁵ CD133⁻ cells (18). In contrast, another study by Beier et al. (19) showed that both CD133⁺ and CD133⁻ GBM cells form tumors in nude mice, and hence questioned the appropriateness of CD133 as a GBM CSC-specific marker. In concurrence, Kelly et al. (20) also demonstrated that both CD133⁺ and CD133⁻ GBM cells possess CSC characteristics and can grow without mitogenic stimulation. They further demonstrated that as few as 10 CD133⁺ or CD133⁻ putative CSCs are sufficient to initiate tumor growth in vivo in the immunocompromised mice (20).

CSCs have unique cell surface markers that differentiate them from non-CSCs. Although a single marker cannot specifically identify or help to isolate CSCs, a set of markers is employed to distinguish GBM CSCs including CD15 (21), CD44 (22), CD133, L1CAM (23), A2B5 (24), CD36 (25), integrin α6 (26), cell surface nestin (27), CD90/Thy-1 (28), leucine-rich repeat containing G protein coupled receptor 5 (LGR5) (29) and the intracellular marker SOX2 (30). Although all these markers may be used to identify the CSCs in GBM, an in vivo tumorigenicity assay is the standard procedure to identify CSCs for their tumorigenic behavior (18). Those GBM CSC surface markers generally agreed upon in the literature are listed in Table 1.

Table 1.

List of GBM CSC cell surface markers

Cell surface marker	Reference
CD133/PROM1	18
SSEA-1/Lewis X/ CD-15	21
CD44a	22
L1CAM	23
A2B5	24
CD36	25
Integrin α6	26
Nestin	27
CD90/Thy-1	28
LGR5	29

^aHighly expressed on mesenchymal subtype GBM CSCs.

In this review article, we have discussed the importance of GBM CSCs in imparting CT and RT resistance, the importance of tumor microenvironment (TME) in the sustenance of GBM CSCs and the failure of chemotherapeutic drugs to cross the blood–brain barrier (BBB) and eradicate CSCs. We have also summarized the therapeutic potential of recently developed novel small molecule inhibitors that can breach the BBB and along with TMZ-based CRT can most effectively eradicate CSCs in GBM.

GBM CSCs are intrinsically chemoresistant

A major challenge with many tumors including GBM is resistance to CT (5) and RT (7), resulting in the development of refractory tumors associated with poor patient prognosis. Although most chemotherapeutic drugs targeting cell cycle progression and affect the proliferative bulk of the tumor, these drugs are mostly ineffective against the more quiescent and non-proliferative CSCs (31,32). Many recent studies have conclusively established the importance of CSCs in imparting CT resistance (CTR) through multiple mechanisms (6,33,34). Among the many possible mechanisms, quiescence, which is a transient cell-cycle arrest, enables CSCs to escape from therapies. This quiescent nature of the CSCs has been demonstrated in mouse models of glioma and in human GBM biopsies by the absence of proliferative marker Ki-67 staining (15,35–38). It is also important to mention that single-cell RNA sequencing on 430 cells obtained from 5 primary GBM patient tumors revealed that quiescence-related genes TSC22D1 and KDM5B expression were increased in non-proliferating tumor cells (39). In addition, CSCs express higher numbers of ATP binding cassette (ABC) transporters, which bestow a broad spectrum of drug resistance (40–42). Among the major ABC transporter genes including breast cancer resistance protein-1 (BCRP1)/ABCG2, ABCB1 and ABCC1 (43), ABCG2 is overexpressed in glioma CSCs (33) and expression of ABCG1 has also been associated with poor overall survival (OS) of GBM patients (44). Though these studies implicated ABC transporters in CTR in CSCs (45,46), others cautioned multiple other factors in addition to these ABC transporters (33) which are summarized in Figure 1. Surprisingly, Eramo et al. (47) recently demonstrated greater resistance of GBM CSCs to several chemotherapeutic drugs over CSCs isolated from small cell lung cancers (SCLC), despite retaining higher intracellular doxorubicin concentrations. These studies suggest that CTR of GBM CSCs is not merely due to compromised drug uptake or active drug efflux (47), but also by some unidentified mechanisms. Second, CSCs express higher levels of O⁶-methylguanine-DNA methyltransferase (MGMT) (48,49), a protein that repairs TMZ-induced DNA damage thereby conferring resistance (50). These studies suggest that traditional cytotoxic drugs—specifically alkylating agents that target highly proliferating cells reliant upon MGMT inactivation—are thus ineffective against quiescent GBM CSCs.

Figure 1.

Targeting GBM CSC-mediated chemo/radio resistance by small molecule inhibitors. GBM CSCs express higher amount of MGMT, which repairs the DNA damage caused by temozolomide. Increased expression of ABC transporters facilitates drug efflux to confer chemoresistance. Radiation kills the cells largely by DNA damage. GBM CSCs overexpress DNA damage repair proteins Rad17, CHK1 and CHK2 to facilitate the repair of radiation-induced DNA damage and imparts RR. In addition, cell surface adhesion molecule L1CAM activates early DDR and confers RR in GBM CSCs by nuclear translocation of intracellular domain of L1CAM (L1-ICD) followed by c-Myc upregulation and increased expression of NBS1, which is one of the core protein in the MRN (MRE11, RAD50 and NBS1) complex. MRN complex is known to activate early DNA damage checkpoint response through activation of ATM kinase. Small molecule inhibitors such as GO eradicate GBM CSCs by selectively targeting CSCs signaling pathways such as Wnt, Notch and Stat3. Cyclopamine, celecoxib and gamma secretase inhibitor RO4929097 target CSCs by affecting SHH, Wnt and Notch pathway, respectively. DSF eliminates CSCs by inhibiting ALDH activity and the NF-κB pathway. Likewise, metformin and sorafenib eradicates GBM CSCs by inhibiting Akt signaling pathways. IR, ionizing radiation; RR, radioresistance; GSI, γ secretase inhibitors; NBS1, Nijmegen breakage syndrome 1; ATM, ataxia telangiectasia mutated; ***small molecule inhibitors; ALDH, aldehyde dehydrogenase; dotted arrow indicates possible effects.

GBM CSCs are radioresistant

Although RT provides the backbone of therapy for GBM, it can also serve to enrich the CSC population and result in refractory tumors (51). Bao et al. observed that RT in human glioma xenograft cultures resulted in a 4-fold enrichment of the CD133⁺ CSC population compared with untreated cultures, and these CSCs were more radioresistant than CD133⁻ cells. These CD133⁺ cells more efficiently repaired RT-induced DNA damage via activation of the DNA damage checkpoint machinery. Furthermore, inhibition of the DDR proteins increased radiosensitivity (RS) (7). Accordingly, recent studies have shown overexpression of DDR proteins including chk1, chk2 and rad17 in the CD133⁺ population, and that inhibition of these proteins sensitizes the CSCs to radiation (7,52). Similarly, CD133⁺ CSC enrichment was also reported in GBM patient tissues after RT (7). In addition, overexpression of a cell surface adhesion molecule L1CAM has also been associated with radioresistance (RR) in GBM CSCs (53). This L1CAM activates early DDR and confers RR in GBM CSCs possibly through nuclear translocation of the intracellular domain of L1CAM (L1-ICD) followed by c-Myc upregulation and increased expression of Nijmegen breakage syndrome 1 (NBS1), which is one of the core proteins in the MRN (MRE11, RAD50 and NBS1) complex (53). The MRN complex is known to activate early DNA damage checkpoint response through activation of ataxia telangiectasia mutated (ATM) kinase and siRNA-mediated silencing of either L1CAM or NBS1 impaired DDR and increased sensitivity to RT in GBM CSCs (53). Because RS varies based on cell cycle distribution with S-phase cells being more resistant than cells in the mitotic phase, the quiescent state of CSCs is one more reason for their RR (54). In addition, RT leads to a disproportionately prolonged G₂/M arrest in GBM CSCs than in differentiated cancer cells, allowing them more time to efficiently repair DNA damage. However, inhibition of ATM using the small molecule inhibitor KU-55933 increased RS of GBM CSCs by abrogating the DNA double stand break repair mechanism irrespective of their cell cycle distribution.

In addition to a hyperactivated DDR, the Wnt/β-catenin signaling pathway also imparts RR to CSCs. Silencing of the Wnt/β-catenin signaling transcription factor, T-cell factor 4 in colorectal cancer cells increased response to CRT (55). Activation of the Wnt/T-cell factor 4 signaling pathway has also been associated with GBM RR (56). Investigators further revealed that inhibition of Wnt signaling by pharmacological and siRNA approaches decreased the population of ABC/Sox2 positive cells (markers for stem cells) thereby increasing RS, suggesting that stem cell associated Wnt/β-catenin signaling imparted RR in GBM. Similar upregulation of the Wnt/β-catenin pathway has also been shown to facilitate RR in mouse progenitor cells (51).

Autophagy sustains growth of tumor cells during stressful conditions such as CT and RT, which further contributes to cell survival and resistance, and tumor recurrence. Autophagy regulator ATG4B, a substrate for mammalian sterile 20-like kinase 26/MST4 promotes autophagosome formation through reversible modification of ATG8 (57). MST4/sterile 20-like kinase-26 is overexpressed in GBM compared with lower grade glioma and its overexpression is associated with aggressive tumors and poor patient survival (57). Recent study by Huang et al. (57) demonstrated that RT-induced MST4 expression both in vitro and in vivo results in the development of RR through autophagy induction. Their mechanistic studies further showed that MST4-mediated RR was through phosphorylation and activation of cysteine protease ATG4B (ser-383) (57).

The CSC–TME niche promotes chemoradiation resistance in GBM

Stem cell niches offer specific in vivo TME, including extracellular matrix, stromal cells, neural inputs and blood vessels, to stem cells and regulate their tissue homeostasis (58,59). Stem cells are closely regulated by the surrounding microenvironment or niche, and are mainly localized in the subgranular zone and subventricular zone of the adult brain (60,61). Shen et al. (62) identified endothelial cells as the major players in the niche of the central nervous system (CNS). Soluble factors from endothelial cells induce NSCs, neuronal generation, stemness and prevent their differentiation. As in other cancers, TME interactions are thought to play an important role in GBM pathogenesis (63). Calabrese et al. (63) showed that GBM CSCs as well as CSCs from different brain tumors are maintained in the vascular niche. They demonstrated for the first time that Nestin⁺/CD133⁺ cells in GBM were present adjacent to the tumor capillaries (63). Endothelial cells physically interacted with CD133⁺/Nestin⁺ cells in a 3D matrigel and maintained proliferation and self-renewal of CSCs through secretory factors, suggesting a specific interaction between CSCs and endothelial cells (63,64). Further, the addition of human primary endothelial cells to CD133⁺ cells increased brain tumor formation by 25-fold in immunocompromised mice (63). It has also been shown that niche-derived bone morphogenetic protein (BMPs) regulate stem cell proliferation (58). Treatment of GBM cells with a particularly effective BMP, BMP4, resulted in a highly differentiated phenotype with increased glial fibrillary acidic protein (GFAP) expression and simultaneously decreased the CSC population and in vivo tumorigenicity (65).

In addition to regulating proliferation, recent studies have shown that CSC-niche interactions promote CR and RR. To this end, Borovski et al. (66) showed that tumor microvascular endothelial cells (tMVECs) are very resistant to CT and RT. Though they observed RT-induced senescence in the tMVECs, coculturing of these senescent tMVECs with GBM CSCs still stimulated the growth and stemness of CSCs (66). More interestingly, GBM CSCs were recently shown to differentiate into tMVEC-like cells, form their own niche (66) and protect remaining GBM cells from CRT and impart resistance. Accordingly, other researchers also observed differentiation of GBM CSCs into cells like tumor-associated vascular endothelial cells (67,68).

Toll-like receptors (TLRs) are also important players in the TME of GBM, favoring tumorigenesis by exerting both anti- and pro-inflammatory effects. Typically, exogenous factors, such as microbes, activate the innate immune system via TLRs. However, in GBM, tumor progression leads to central necrosis and edema at the tumor border, which results in release of damage-associated molecular patterns to stimulate TLR. TLR9 facilitates CSC self-renewal in GBM (69), whereas TLR4 decreases GBM growth by downregulating the expression of retinoblastoma binding protein 5, a transcription factor upregulated in CSCs with an important role in stemness (70). Recently, Alvarado et al. (70) found that GBM CSCs escaped immune suppression by decreasing TLR4 expression. Interestingly, a transient overexpression of human TLR4 in CSCs was accompanied by decreased expression of stem cell markers, namely, NANOG, SOX2 and OCT4 (70). Moreover, TLR4 expression was inversely associated with glioma grade, i.e. TLR4 expression is lower in GBM than in low-grade astrocytoma and is directly correlated with survival (69).

Aside from TLRs, other components of the immune system are active in the TME. Macrophages and microglia, similar to peripheral macrophages, offer the first line of defense in the CNS (71,72). They can be classified into two types: classically activated M1 and alternatively activated M2. M1 macrophages are involved in antigen presentation and phagocytosis and are cytotoxic (73,74). The M2 macrophages, however, are immunosuppressive and oncogenic, stimulated by either tumor-derived monocyte chemoattractant protein-1 (CCL2) or colony-stimulating factor (75,76). Incubation of macrophages with glioma CSCs results in macrophage polarization into the M2 phenotype and suppressed T-cell proliferation and phagocytic activity while significantly increasing the secretion of immunosuppressive cytokines IL-10 and TGF-β1 (77). PDCD4, a pro-inflammatory protein, downregulates IL-10 secretion through the stimulation of the NF-κB pathway (78). In addition to the innate defenses, the adaptive defenses affect the GBM microenvironment. T regulatory cells and myeloid derived suppressor cells have a critical role in immune suppression in cancer patients. Otvos et al. (79) found that GBM CSCs secrete large amounts of macrophage migration inhibitory factor (MIF) that induce myeloid-derived suppressor-cell-mediated immune downregulation. MIF mRNA levels are significantly greater in GBM than normal tissue and are further increased in recurrent versus newly diagnosed GBM. Levels of MIF mRNA are inversely associated with GBM patient survival. Furthermore, immunosuppressive enzyme arginase-1 expression is induced by adding CSCs to conditioned media which is effectively is nullified by adding the MIF scavenging antibody (79).

GBM is extremely hypoxic, a state which promotes stemness. CSCs proliferate through hypoxia inducible factor 1α (HIF-1α)-induced Notch activation (80). Soeda et al. (81) found that hypoxia induced self-renewal of CD133⁺ GBM CSCs, and siRNA-mediated silencing of HIF-1α decreased the CD133⁺ population. In addition, HIF-2α expression is also induced by hypoxic conditions in GBM CSCs (82) to regulate Oct-4 expression and maintain stemness (83). Paradoxically, TMZ treatment induces HIF-1α and HIF-2α, and promotes the dedifferentiation of glioma cells to CSCs, whereas silencing of HIF expression prevents this post-CT transformation (84). In addition, silencing of either HIF-1α or HIF-2α in GBM CSCs significantly reduced their tumorigenicity and increased the survival of mice bearing orthotopic xenografts (82). Further, Pistollato et al. also noted that TMZ treatment induced cell death only in the differentiated tumor cells confined to the neo-vascularized periphery of the tumor with no effect on the CD133⁺ GBM CSCs in the hypoxic inner core and middle zone of the tumor (48).

Targeting GBM CSCs

The deregulated cell signaling pathways in CSCs play an important role in RR, CTR and tumor recurrence; selective targeting of these mediators of tumorigenesis would improve therapeutic efficacy and patient prognosis. As discussed earlier, various signaling pathways such as Notch, Sonic hedgehog (SHH), Wnt, BMP, epidermal growth factor receptor (EGFR) and vascular endothelial growth factor (VEGF) are all elevated in CSCs and are therefore attractive targets for therapy. Notch signaling controls NSCs in the CNS, but its activity is significantly upregulated in GBM CSCs (10,85). Stanniocalcin-1 is a secretory glycoprotein with elevated expression in GBM CSCs (86). Recently, Notch 1 signaling was shown to regulate stanniocalcin-1 expression and increase stemness in GBM cells (86). Drug-resistant GBM CSC persisters (cultured in drug for at least 8 weeks) also showed increased Notch activation and maintenance of a quiescent state (87). Fan et al. showed that inhibition of Notch signaling through γ-secretase inhibitors depleted stemness in GBM, whereas expression of active Notch2 in CSCs increased in vivo tumorigenicity (88). Interestingly, adding γ-secretase inhibitors to TMZ significantly decreased the tumor growth and increased the OS of a mouse GBM xenograft model (89). In addition, pretreatment of U87 and U373 neurosphere cultures with a γ-secretase inhibitor and TMZ significantly decreased their tumorigenic potential (89). Moreover, γ-secretase inhibitor treatment or siRNA mediated silencing of Notch1/2, decreased RT-induced Notch activation and increased RS, whereas activation of Notch signaling by the expression of either NICD1 or NICD2 conferred RR (90). Zhu et al. (91) identified CNS endothelial cells as the main source for the Notch ligands that maintain GBM CSC proliferation and self-renewal. Hovinga et al. (92) used saporin toxin conjugated to anti-CD105/endoglin antibody for selective targeting of endothelial cells in GBM explants. They observed that endothelial cell inhibition reduced the Notch effector Hes5 expression (~60%), decreased VEGF secretion and the number of neurosphere formation compared with untreated controls, thus highlighting the importance of Notch signaling in CSCs self-renewal and in angiogenesis (92). Importantly, a phase 0/I clinical trial using a gamma-secretase inhibitor (RO4929097), in conjunction with TMZ and RT decreased CSCs and Notch signaling (93). Unfortunately, they observed tumor recurrence through Notch-independent angiogenesis (93). In addition to the hyperactivation of Notch signaling, 26% of GBM tumors overexpress SHH pathway component Gli1, and concordantly, use of the SHH inhibitor cyclopamine reduced the growth of GBM (94) and is known to damage the BBB integrity (95). Further, induction of differentiation of GBM tumorspheres decreased SHH signaling components and the stem cell population (94, 96). In addition, lentivirus-mediated silencing of SHH signaling significantly reduced the stemness and tumorigenicity in vivo, thus suggesting the role of SHH signaling in the maintenance and tumorigenicity of GBM CSCs. Unfortunately, interim results from the phase II clinical trial (NCT00980343) with a small molecule inhibitor of the SHH pathway (GDC-0449/vismodegib) has not shown improvement in the progression-free survival (PFS) or OS in recurrent GBM patients. This underscores the complexity of GBM and the futility of targeting a solitary pathway when multiple mechanisms for growth and resistance exist.

Wnt signaling plays an important role in embryonic development, adult organ regeneration and several other functions. Uncontrolled activation or mutation of Wnt signaling primes cells to malignant transformation (97, 98). Wnt signaling is highly dysregulated in GBM, and it regulates GBM CSCs self-renewal as well (99,100). Comparative analysis of GBM CSCs and adult human NSCs by Sandberg et al. (101) identified specific activation of 24 signaling pathways in GBM CSCs and among them the Wnt pathway was the second most commonly dysregulated pathway (101). In addition, the naturally present Wnt inhibitor, soluble Frizzled-related protein 1, is downregulated upon upregulation of the Wnt signaling pathway in GBM CSCs (101). Aspirin, a small lipophilic non-steroidal anti-inflammatory drug (NSAID), is known to reduce colon cancer risk (102,103). Lan et al. (104) observed that aspirin also decreased the invasiveness and proliferation in GBM cell lines A172 and U87MG through inhibition of Wnt signaling. Other NSAIDs such as celecoxib and diclofenac by selectively inhibiting the Wnt signaling components decreased proliferation, migration and in vitro tumorigenicity of GBM cells (105,106). Although these NSAIDs inhibited Wnt signaling and decreased the tumorigenic potential of GBM cells, their effect on GBM CSCs still needs to be explored. But despite enticing preclinical efficacy, the phase II clinical trial of celecoxib and TMZ (NCT00112502) did not improve GBM patient survival (107).

Activation of the MET tyrosine kinase regulates a wide range of cellular signaling pathways that are involved in cancer progression and metastasis (108). Approximately 4% of GBM patients show MET amplification (109) and its overexpression is associated with poor patient survival (110). Kim et al. (111) showed that MET-enriched cells demonstrated greater self-renewal and tumor initiation potential. They further showed that MET by regulating Wnt/β-catenin pathway maintains stemness, as inhibition of MET signaling prevented the nuclear translocation of β-catenin and loss of self-renewal and stemness character (111). These studies established the role of MET in GBM CSC maintenance, and its inhibition decreased the tumorigenicity in vitro and in vivo (112–114).

EGFR family members and CSCs

EGFR is amplified in 40–70% of primary GBM; and 50–60% of GBM patients with EGFR amplification also carry an oncogenic gene rearrangement that results in the expression of constitutively active EGFR variant III (EGFRvIII) (115). Both the EGFR and EGFRvIII signaling pathways play an important role in tumor growth and formation and maintenance of GBM CSCs (26,116–118) at least in part by increased secretion of IL-6 (119,120). In collaboration with these studies, use of either EGFR-specific antibody (20) or siRNA-mediated silencing of EGFRvIII (118) also decreased sphere formation of GBM CSCs by reducing stemness markers nestin and Sox2 and by increasing GFAP levels (118). Overexpression of EGFRvIII in X02 cells (no endogenous EGFRvIII) also increased the expression of nestin and Sox2 and decreased the GFAP levels (118). EGFRvIII is highly co-expressed with CD133 in GBM and targeting these EGFRvIII⁺/CD133⁺ cells with a bispecific antibody significantly reduced their in vivo tumorigenicity (121). In contrast, a phase II clinical trial using the EGFR inhibitor erlotinib yielded only minimal efficacy as a single agent in recurrent GBM patients (122). This might be at least in part through the activation of other EGFR family members as Clark et al. (123) reported compensatory activation of ERBB2/HER2 and ERBB3/HER3 in GBM CSCs upon inhibition. Our group has recently shown that afatinib, a Food and Drug Administration (FDA) approved pan-EGFR inhibitor radiosensitized head and neck squamous cell carcinoma by eradicating CSCs both in vitro and in vivo (124). Though afatinib is known to cross the BBB (125), however, its use with and without TMZ did not show any efficacy in an unselected recurrent GBM population (126). Contrary to all these studies, our preclinical results have shown strong antitumor activities of afatinib along with TMZ in EGFRvIII-expressing GBM cells (unpublished data), suggesting that this novel combination therapy may be effective in GBM patients with EGFRvIII mutation.

VEGF and GBM CSCs

Angiogenesis is a very complex process of the formation of new blood vessels during the development of solid tumors. VEGF is a principal regulator of angiogenesis (127,128), and its overexpression is associated with high-grade gliomas and poor patient prognosis (129). A recent comparative analysis revealed that GBM CSCs secrete greater amounts of VEGF than non-stem-like cells, and treatment of endothelial cells with GBM CSC-conditioned media increased migration and tube formation more effectively than conditioned media from differentiated cells (130). Although use of the anti-VEGF antibody bevacizumab abolished the GBM CSC mediated in vitro angiogenesis and significantly decreased the GBM CSC population in GBM orthotopic xenografts (63), multiple prospective clinical studies using this drug up front and after recurrence have failed to improve OS of GBM patients (131,132). Studies have demonstrated that blocking the VEGF pathway leads to bevacizumab resistance as other regulators of tumor angiogenesis are upregulated (133,134).

Drugs breaching the BBB to eradicate CSCs

The BBB is one of the major obstacles impeding the delivery of therapeutic drugs to brain tumors. The BBB is a complex structural system of endothelial cells with tight junctions, an ensemble of enzymes, multidrug resistance efflux pump proteins, receptors and transporters that by either paracellular or transcellular pathways control and limit the access of molecules into the brain (135). The BBB permits the diffusion of small molecules with a high lipophilicity and a molecular weight <400–500 Da (136,137). GBM vasculature significantly differs from normal brain and is characterized by several layers of endothelial cells, smooth muscle cells and pericytes and forms a thick basement membrane with increased vascular permeability (138). Gadolinium-accumulation-based magnetic resonance imaging in GBM patients revealed both disrupted and intact BBB in tumors (139). Although the BBB integrity is lost at the tumor center, capillaries supplying nutrients to the tumor and adjacent normal tissue are intact and thus minimize drug availability (140). Though novel therapeutic drugs that target the BBB are currently underway, many small molecule inhibitors and chemically modified derivatives of drugs have been developed that can breach the BBB and efficiently eradicate GBM CSCs and are discussed and grouped in Table 2.

Table 2.

Small molecule inhibitors of interest targeting GBM CSCs

Molecule	Function	Mode of action	Ability to cross the BBB	Preclinical studies	Clinical trials	References
GO	Selectively targets GBM CSCs	Inhibit STAT, Wnt and Notch signaling	Yesa	Analyzed on chorioallantoic membrane model—reduced tumor growth	NA	141–144
Metformin	↑ the differentiation of GBM CSCs and decreases tumorigenicity of CSCs	↓ AKT-dependent survival in GBM CSCs, ↑ differentiation of CSCs through activation of FoxO3	Yes	TMZ + metformin ↓TMZ-resistant T98G xenograft growth	Three Phase 1 trials ongoing (NCT01430351; NCT02149459; NCT02149459)	151–153
Sorafenib	↓ stemness, clonogenicity and stem cell potential	Inhibit survival factor Mcl-1 through inhibition of Akt and MAPK signaling	Yes	Pretreatment with sorafenib (48 h), significantly ↓ tumorigenic potential of GBM CSCs in vivo	NCT00544817—limited efficacy	146,156,157
Disulfiram	Kills GBM CSCs	Inhibit NF-κB and effective in low-MGMT, high-MGMT and TMZ-resistant cell lines. ↑proteaosomal degradation and inhibit the activity of MGMT	Yes	↓MGMT expression in T98 GBM xenografts	NCT01907165—Just completed	158–163
Dichloroacetate	Kills GBM CSCs as well as differentiated cancer cells	↓ mitochondrial membrane potential and ↑ mitochondrial ROS production, p53 stimulation and apoptosis	Yes	↓ C6 glioma growth in rats and mice xenografts	Open label trial in five GBM patients with standard therapy + dichloroacetate —↓ tumor progression in three patients	172, 173
LDH-A inhibitors (NHI 1 and NHI 2)	Active against GBM CSCs and GBM cells	Cell cycle arrest and induced apoptosis in GBM CSCs and GBM cells	NK	NK	NK	185
Gamma secretase inhibitors (RO4929097)	↓ GBM CSCs	Inhibition of Notch signaling	Yes	Combination with TMZ reduced the tumor growth and increased the OS of a mouse GBM xenograft model	Phase 0/1; NCT01119599; Combination with TMZ and RT decreased Notch and CSCs but tumor recurred through Notch independent angiogenesis	89, 90, 93
NSAIDs—celecoxib	↓ proliferation, migration and in vivo tumorigenicity of GBM cells	Inhibits Wnt signaling	Yes	NK	Phase-2-NCT00112502—minimal effect	105–107
CBL0137	Kills GBM CSCs and ↓ in vivo tumorigenicity of GBM CSCs	Inhibits FACT—histone chaperone activity	Yes	↓ tumor growth and ↑ survival of GBM CSCs tumor orthografts	Phase-1 Ongoing—NCT01905228	187

NK, not known.

^aReduced GO and graphene quantum dots cross the BBB.

Graphene oxide

Graphene oxide (GO) is a non-toxic, fairly inert nanomaterial and forms stable dispersions in a range of solvents. Recent research has shown that by concurrently inhibiting the STAT, Wnt and Notch signaling pathways, GO selectively abolished the stemness of CSCs in six malignancies including lung, breast, prostate, ovarian, pancreatic cancer and GBM (141). Further, GO significantly reduced the tumor burden on a chick embryo chorioallantoic membrane in vivo (142). Although no cytotoxicity was observed following i.p injection of GO into mice, reduced GO and graphene quantum dots with an average diameter of 342 and 8 nm, respectively, were shown to cross the BBB (143,144). Because TMZ kills only the differentiated tumor cells and spares CSCs, combining GO nanotherapy along with TMZ could effectively eradicate both the differentiated tumor cells and CSCs. To the best of our knowledge, no clinical trial is currently underway to test the therapeutic efficacy of GO in GBM patients.

Metformin

Metformin is an oral drug for type II diabetes that has been used for more than 50 years (145). Many studies have shown its antitumor effects against various cancers (146). Recently, metformin was shown to selectively deplete CSCs and along with doxorubicin synergistically kill breast cancer cells (147). Metformin also showed antiproliferative and proapoptotic effects against ovarian CSCs, CD133⁺ pancreatic cancer and CD133⁺ colorectal cancer CSCs (148–150). Interestingly, metformin was also shown to cross the BBB and target GBM CSCs via inhibiting the Akt-dependent survival signaling pathway (151). Furthermore, the CD133⁺ GBM CSCs strongly associated with CR were highly susceptible to metformin treatment (151). In addition, metformin decreased the tumorigenicity of GBM CSCs associated with trans-differentiation via activation of the FOXO3 transcription factor (152). Metformin in combination with TMZ also significantly reduced the TMZ-resistant T98G xenograft growth (153). On the basis of these studies, a phase I clinical trial (NCT01430351) is currently ongoing to identify the maximum tolerable dose of metformin with TMZ, mefloquine and/or memantine in post-RT-treated GBM patients.

Sorafenib

Sorafenib is a FDA-approved small molecule multityrosine kinase inhibitor of fibroblast growth factor receptor 1, platelet-derived growth factor receptor β, vascular endothelial growth factor receptor-2 (VEGFR2), vascular endothelial growth factor receptor-3 (VEGFR3), CD117, rearranged during transfection (RET) and fms-like tyrosine kinase receptor-3 (146). Overactivation of these receptors is involved in neo-angiogenesis to support tumor growth (146,154,155). Carra et al. (156) showed that sorafenib decreases cell survival by downregulating survival factor Mcl-1 through inhibition of Akt and MAPK signaling pathways. More interestingly, they reported that sorafenib destroyed the CSCs by suppressing their stemness, clonogenic survival in vitro and tumorigenic potential in vivo using NOD/SCID mice (156). Despite this preclinical efficacy (156), unfortunately, adjuvant sorafenib with TMZ failed to increase the PFS in GBM patients (157). It is important to mention that 40% of patients were removed from the study prior to treatment due to early tumor progression (157).

Disulfiram

Disulfiram (tetraethylthiuram disulfide, DSF) is an inhibitor of the metabolic enzyme aldehyde dehydrogenase (ALDH) that catalyzes the oxidation of acetaldehyde to acetate in presence of NAD⁺ cofactor and therefore is used to treat chronic alcoholism. Many research studies have conclusively established the importance of ALDH in conferring drug resistance and its use as a standard marker for most of the CSCs including GBM CSCs (158). Interestingly, many recent studies have shown that DSF crosses the BBB and also decreases the survival of ABT011 (high MGMT expressing), ABT015 (low MGMT expressing) and TMZ-resistant T98G GBM cells (159). Interestingly, Triscott et al. (160) reported no cytotoxicity of DSF to the normal human astrocytes, but complete blockage of neurosphere growth. Furthermore, they showed that DSF inhibited the self-renewal properties of BT74 and GBM4 primary GBM cells and inhibited the growth of TMZ-resistant SF188 cells (160). Surprisingly, DSF was shown to induce proteasomal degradation and inhibit the activity of MGMT in GBM cells U87 and T98G (161). Although DSF was shown to cross the BBB and inhibit atypical teratoid rhabdoid tumors (162), subsequent studies showed attenuated MGMT activity in the brain tissues upon DSF treatment in normal mice, but preferential inhibition of tumor MGMT in T98 GBM xenograft-bearing nude mice (161). In addition, DSF was also shown to inhibit the NF-κB pathway (159), which plays an important role in the maintenance of GBM CSCs (163). Therefore, use of DSF seems to be a safe and novel therapeutic approach to eradicate the GBM CSCs and prevent tumor recurrence. Though at high concentrations DSF is hepatotoxic (159), a phase I clinical trial using low-dose DSF with adjuvant TMZ for GBM has recently been completed (results not available; NCT01907165).

Dichloroacetate

Glycolysis is the major source for ATP production in most of the proliferating tumor cells (Warburg effect) and inhibition of glycolysis results in reduced tumorigenicity both in vitro and in vivo (164). Though still controversial, Vlashi et al. (165) reported that GBM CSCs/progenitor cells utilize less glucose with less glycolytic activity and produce less lactate as a byproduct, but maintain higher ATP levels through oxidative phosphorylation (OXPHOS) compared with differentiated tumor cells. Further, rapidly growing tumor cells release large amounts of carbonic and lactic acid to create an acidic TME that enhances stemness through an hypoxia independent mechanism by upregulating HIF-2α (166), suggesting that targeting solely the Warburg effect may not be an effective strategy for GBM CSCs. Accordingly, inhibition of either glycolysis or OXPHOS was not shown to significantly reduce the ATP levels in the GBM CSCs or in progenitor cells, but simultaneous inhibition of both these metabolic processes resulted in significant reduction in ATP levels and stemness (165).

Dichloroacetate (DCA) is a small molecular inhibitor of the enzyme pyruvate dehydrogenase kinase (167,168) and has been used to treat patients with hereditary mitochondrial defects for over three decades (169,170). By inhibiting the pyruvate-dehydrogenase-kinase-dependent activation of pyruvate dehydrogenase, DCA has been shown to increase transport of pyruvate into the mitochondria and thereby favor glucose oxidation over glycolysis (168). It is important to mention that increased glycolysis is associated with hyperpolarization of the mitochondrial membrane and resistance to apoptosis (168,171). Accordingly, DCA depletes GBM CSCs by abolishing the mitochondrial membrane potential, increasing reactive oxygen species production, upregulating p53 and inducing apoptosis (172). In addition, DCA also significantly decreased the tumor weight of a C6 tumor-bearing rat and mice in vivo (173). An open label trial in five recurrent GBM patients with standard therapy plus DCA significantly decreased the tumor progression in three patients (172). Phase I clinical trial results supported that DCA was well tolerated in recurrent GBM patients (174), although at higher doses it has been associated with peripheral neuropathy (175). A single agent DCA phase II clinical trial (NCT01029925) in metastatic breast or non-small cell lung cancer was terminated early due to perceived lack of efficacy (176). Currently there are no active clinical trials using DCA in GBM patients.

NHI-1 and NHI-2

The glycolytic enzyme lactate dehydrogenase (LDH) catalyzes the interconversion of pyruvate to lactate and regulates the shift from mitochondrial OXPHOS to glycolysis. LDHA is a tetrameric protein, consisting of two core subunits A and B, which can form five different isoforms (177). Increased expression of LDHA is observed in several cancers including GBM and is associated with poor patient survival (178–180). Many recent studies have reported increased expression and activity of LDHA in CSCs than differentiating tumor cells (181), thus suggesting it may be a novel therapeutic target to counteract CR and tumor recurrence. LDHA is believed to be a central regulator of the Warburg effect and either its downregulation or inhibition slowed tumor progression and suppressed in vivo tumor xenograft growth of many cancers (177,182,183). Di et al. (184) also observed a higher expression of LDHA in GBM over lower grade gliomas and discovered that silencing LDHA decreased proliferation, increased apoptosis and chemosensitized GBM cells to TMZ. Daniele et al. (185) also showed that LDHA inhibition using two small molecular inhibitors NHI-1 and NHI-2 decreased proliferation, induced cell cycle arrest and cell death in GBM cells. More interestingly, they showed these two LDHA inhibitors had even greater cytotoxicity to GBM CSCs (185). Similar results were also reported in lung cancer CSCs using an inhibitor specific to both LDHA and LDHB (186). To the best of our knowledge, none of the studies address the BBB permeability of either NHI-1 or NHI-2.

CBL0137

CBL0137 is a small molecule inhibitor known to target facilitates chromatin transcription (FACT), a histone chaperone that is highly expressed in poorly differentiated cells (187). FACT facilitates transcriptional elongation by disturbing the nucleosome’s interaction with DNA, thus allowing RNA polymerase II to obtain access along the DNA template. Josephine et al. (187) reported overexpression of FACT in GBM compared with normal brain and associated this with GBM CSC marker expression and poor patient survival. Concordantly, use of CBL0137 preferentially targeted GBM CSCs over non-stem cells and oral administration of CBL0137 (0.5 mg/ml in their drinking water) to mice with orthotopically implanted GBM CSCs significantly increased survival (187). Due to its potential to cross the BBB, one clinical trial (NCT01905228) utilizing CBL0137 is currently ongoing for patients with metastatic solid tumors or recurrent GBM.

Conclusions and future perspectives

Despite multimodality treatment including surgery, RT and CT, tumor recurrence is nearly universal resulting in death of GBM patients (188). The main reasons for this dismal prognosis are (i) invasive behavior making surgical resection difficult, (ii) intrinsic and acquired resistance to RT and CT, (iii) failure of the targeted therapeutic drugs to cross the BBB (drug impenetrability) (iv) inability for mainstay therapy to eradicate CSCs (iv) heterogeneity and lack of driving mutations. Most importantly, GBM CSCs have been shown to play a pivotal role in the intrinsic treatment resistance and the development of recurrent tumors with <1 year patient survival (188). Many recent studies have shown that traditional CRT successfully manages the bulk of the differentiated tumor cells, but fails to affect the CSCs, thus making tumors more refractory in nature (189). Therefore, development of efficient treatment regimens with long-lasting disease control should include specific and successful targeting of CSCs. Unfortunately, additional challenges for targeting CSCs are their quiescent nature, rapid drug efflux, DNA repair by MGMT and DDR and the complex processes that maintain stemness and dedifferentiation. In this regard, small molecule inhibitors are of interest in targeting GBM CSCs due to their many beneficial properties including (i) BBB penetration, (ii) selective inhibition of signaling pathways specific to CSCs and the CSC-TME niche, and (iii) reduced cytotoxicity to normal brain cells. Notably, GO selectively abolished in vitro tumorsphere formation (believed to be mediated by CSCs) in six different cancer cell lines (141). This effect was associated with an inhibition of stem cell signaling pathways such as Wnt, Notch and Stat3. Metformin and sorafenib eradicated the GBM CSC population by inhibiting Akt signaling pathways and CBL0317 preferentially eliminated CSCs by targeting FACT chaperone. Moreover, DSF eliminated CSCs, mainly by inhibiting ALDH activity and the NF-κB pathway with even more potent cytotoxicity against TMZ-resistant cells in vitro. Despite their strong preclinical efficacy against GBM CSCs, small molecule inhibitors have thus far failed to improve patient survival in clinical studies. This may in part be due to a limited bioavailability of the drugs to the CSCs mediated by ATP dependent P-glycoprotein (P-gp) present in the BBB or drug efflux by the P-gp present in the CSCs themselves. P-gp is overexpressed in GBM CSCs compared with NSCs (190) and its inhibition augmented the cytotoxicity of daunorubicin in prostate cancer multidrug resistance DU145TXR spheroids (191). These studies suggest that using agents that target P-gp and modulate the BBB permeability combined with aforementioned drugs may prove advantageous clinically to GBM patients. In addition, GBM CSCs are dependent upon other signaling pathways than differentiated GBM cells; hence, combination therapy targeting these disparate pathways is needed for effective treatment (Figure 2). Excitingly, many natural compounds including procyanidine and scillarenin have recently been shown to inhibit P-gp and increase drug availability in preclinical models (192,193) and comprehensively reviewed by us recently (135). We propose that combining these natural compounds with the small molecule inhibitors such as DSF, CBL0137, DCA, sorafenib or metformin together with TMZ will not only eradicate GBM CSCs with minimal patient toxicity, but will also prevent tumor recurrence.

Figure 2.

Proposed model for GBM CSCs targeted therapy. (A) TMZ effectively kills mostly differentiated GBM tumor cells with methylated MGMT but is ineffective on CSCs that result in delayed patient relapse. However, (B) ineffectiveness of TMZ to kill both the differentiated and CSCs with unmethylated MGMT results in quick tumor relapse with poor patient prognosis. (C) DSF inhibits the metabolic enzyme ALDH that is highly expressed in stem-like cells and involved in conferring drug resistance. DSF also inhibits the NF-κB pathway that plays an important role in the maintenance of GBM CSCs. In addition, DSF effectively decreases the survival of both high-MGMT- and low-MGMT-expressing and TMZ-resistant GBM cells, sparing normal human astrocytes. P-glycoprotein (P-gp)—which mediates drug efflux—is present in the BBB and its increased expression in the CSCs limits the bioavailability of the drugs. Strikingly, a natural compound procyanidine has recently been shown to inhibit P-gp present in the BBB and increase drug availability in preclinical models. Combination therapy of TMZ with procyanidine and disulfiram, may eradicate both tumor cells, as well as CSCs irrespective of their MGMT status. TMZ- temozolomide; ALDH- aldehyde dehydrogenase; MGMT- O⁶-methylguanine-DNA methyltransferase dotted arrow indicates possible effects.

Funding

National Institutes of Health (P30 CA036727; RO1 CA183459, UO1185148); Fred and Pamela Buffett Cancer Center, University of Nebraska Medical Center, Omaha, NE, and William Langford Creative Mind Fund (to R.V.).

Conflict of Interest Statement: SKB is one of the co-founder for Sanguine Diagnostic and Therapeutics Inc. No potential conflicts of interest were disclosed by the other authors.

Glossary

Abbreviations

ABC

ATP binding cassette

ALDH

aldehyde dehydrogenase

BBB

blood–brain barrier

BMPs

bone morphogenetic proteins

CNS

central nervous system

CR

chemotherapy resistance

CSCs

cancer stem cells

CT

chemotherapy

DCA

dichloroacetate

DDR

DNA damage response

DSF

disulfiram

EGFR

epidermal growth factor receptor

FACT

facilitates chromatin transcription

GBM

glioblastoma

GFAP

glial fibrillary acidic protein

GO

graphene oxide

HIF-1α

hypoxia inducible factor 1α

LDH

lactate dehydrogenase

MGMT

O6-methylguanine-DNA methyltransferase

MIF

migration inhibitory factor

NBS1

Nijmegen breakage syndrome 1

NSAID

non-steroidal anti-inflammatory drug

NSCs

neuronal stem cells

OS

overall survival

OXPHOS

oxidative phosphorylation

P-gp

P-glycoprotein

RR

radioresistance

RS

radiosensitivity

RT

radiotherapy

SHH

Sonic Hedgehog

TLRs

toll-like receptors

TME

tumor microenvironment

tMVECs

tumor microvascular endothelial cells

TMZ

temozolomide

VEGF

vascular endothelial growth factor