Brain tumor stem cells: will understanding a new paradigm lead to improved therapies?

Solid tumors are a complex network consisting of cells at various stages of differentiation, neovasculature structures, reactive inflammatory cells, recruited cells and infiltrated parenchyma that interact within the tumor mass. However, evidence strongly indicates that cancer stem cells (CSCs) drive tumorigenesis, as these cells possess self-renewal and tumorigenic capacity absent in the majority of tumor cells. Recent studies have revealed the presence of a slowly cycling, but highly tumorigenic, subpopulation of cells derived from surgically resected glioblastoma multiforme (GBM) tissue similar to those in other solid tumors, such as breast, prostate and colon carcinomas. These cells have properties analogous to the relatively quiescent population of adult neural stem cells (NSCs) embedded within the subventricular zone and hippocampus [1–5].

Despite modernization of drug cocktails and radiotherapeutic regimens over the past half century, median survival for patients with newly diagnosed GBM remains approximately 12–15 months [6,7]. Therapeutic resistance is likely multifactorial; however, a growing body of evidence suggests that the cardinal reason may be the resistance of CSCs to cytotoxic treatments, such as chemotherapy and ionizing radiation. Indeed, recurrent GBM tissue obtained from patients contains a significantly higher percentage of putative CSCs, as compared with their respective newly diagnosed tumors, which suggests that CSCs strongly affect tumor resistance to chemotherapy [8,9]. CSCs that survive therapy repopulate the tumor leading to recurrence.

Mechanisms of treatment resistance

The mechanisms of CSC resistance are only beginning to be discovered. The relative quiescence of CSCs, much like that of normal adult NSCs, is one plausible cause since most common anticancer treatments rely heavily on their target cells being in cycle for effectiveness. However, active mechanisms appear to be involved as well. Glioma CSCs are enriched from a small fraction of tumor cells known to efflux Hoechst dye through membrane-bound pumps. These pumps also remove cytotoxic chemotherapeutic agents from the cell [10,11].

Postoperative radiation therapy affords patients short-term survival benefit, but tumor recurrence, progression and patient death are inevitable. This is likely due to CSCs’ resistance to radiotherapy, despite its effectiveness against bulk tumor cells. Irradiation of GBM cell cultures increases the percentage of stem-like cells, suggesting that radiotherapy has a greater effect on differentiated cells and that CSCs are the resistant population [12]. When subjected to radiation, GBM CSCs robustly activate a DNA damage checkpoint response both in vitro and in vivo when xenografted into immunodeficient mice. An inhibitor of C-terminal Src kinase-homologous kinase (CHK)1/CHK2 sensitizes CSCs to killing by ionizing radiation [13]. Additional pathways and signaling molecules likely mediate CSC insensitivity to radiation in both GBM and other solid tumors. For example, Wnt/β-catenin signaling, an important pathway in neural development, has been shown to confer radioresistance to breast CSCs [14]. The effectiveness of using radiosensitizing agents in conjunction with radiation therapy for therapeutic targeting of GBM CSCs needs to be more aggressively studied. This is particularly important for the treatment of gliomas since radiation remains standard of care following surgical tumor resection and is not limited in delivery by the BBB as many pharmaceutical agents are.

Glioma cancer stem cells: can we eradicate them & can we monitor treatment efficacy?

Given that CSCs are the cells putatively driving tumorgenesis in vivo, they are obvious therapeutic targets. However, there are numerous obstacles that confound our current ability to zero-in on these oncogenic cells therapeutically:

•	Current clinical trial end points are not optimized to detect anti-CSC effects
•	Preclinical anticancer therapies have been, and continue to be, assessed in passaged cell lines
•	Anti-CSC response cannot currently be measured in vivo
•	The basic biology of CSCs largely remains unclear
•	Eradicating CSCs specifically while sparing NSCs will probably prove to be difficult

At present, treatments that show promise as antitumor agents at best provide transient benefit when given to patients with GBM and usually impart little to no benefit. Generally, after treatment safety is determined by Phase I testing, effectiveness is defined by an appreciable reduction in tumor size in Phase II clinical trials. Tumor response by this criterion, however, does not translate into overall survival benefit for patients when tested in Phase III trials. The likely reason for this lack of correlation is that such agents do not target GBM CSCs. In fact, agents that selectively kill CSCs would probably not affect the size of the tumor mass at all. Therefore, progression-free survival or time to progression may be more appropriate end points for determining the effectiveness of therapeutic agents against CSCs [15].

Furthermore, GBM cell lines such as U87MG and others have been the historical standard for preclinical screening of potential therapeutic agents. However, unlike GBM CSCs, the phenotypic and genotypic aberrations found within these cells often bear little resemblance to those in the primary GBM [16]. This likely explains why cancer cell line-based preclinical models have been poorly predictive of therapeutic utility against GBM.

Despite being one of the least treatable cancers, GBM is well suited for preclinical, in vitro screening of anti-CSC agents because GBM CSCs can be readily derived using the neurosphere assay. Like normal NSCs, GBM CSCs form non-adherent cell masses known as neurospheres when grown in growth factor supplemented culture medium [3,17]. In fact, the neurosphere assay is already being used to test drug sensitivity of GBM CSCs and preliminary data show that neurospheres collapse and the constituent cells die after exposure to various drugs [18]. Ultimately, treatments that change the stem cell phenotype of glioma-derived neurospheres may yield better outcomes for patients with GBM.

However, the neurosphere assay is not absolute. Tumor-derived neurospheres, once thought to be clonal, or derived from a single stem cell, are likely to be a heterogeneous population of stem cells, intermediate progenitors and differentiated cells [19,20], and experiments with video microscopy of genetically labeled cells show that neurospheres are highly motile and prone to fuse [20]. This does not diminish the importance of examining the effectiveness of agents against tumor spheres, but the neurosphere assay alone will not suffice for preclinical testing. Combinations of in vitro modeling, in vivo xenografting, and mouse models and ex vivo techniques must be employed to thoroughly scrutinize possible therapeutic agents that will yield genuine potential to target GBM CSCs in human patients.

Once developed, can the efficacy of GBM CSC-specific treatments be followed directly? Although enormous strides have been made in the field of molecular imaging, this technology is not suitably developed to monitor the effectiveness of treatment against small groups of cells or even single cells. Although a novel fatty acid metabolite, seemingly specific to human NSCs, has recently been detected using magnetic resonance spectroscopy [21], this method is far from clinical use and a metabolite unique to glioma CSCs remains unfound.

More fundamentally, before CSC-specific treatments can be developed or monitored, a method for identifying the CSC targets needs to be validated. In the past, GBM CSCs have been characterized based on functional criteria, which is not pragmatic when attempting to target a specific cell type in vivo. Recently, in vitro sorting for CSCs by membrane-bound epitopes has received much attention because of the ease with which these techniques could potentially winnow stem cells from the rest of the tumor. However, results have been equivocal [3,22,23]. This is best characterized by the controversy surrounding separation of stem cells by the CD133 epitope.

CD133 is reported to be present on the membrane of stem cells and progenitor cells in various tissues [3,4], and as few as 100 CD133⁺ human GBM CSCs transplanted into the brains of immunodeficient mice initiate the development of gliomas [3]. Importantly, however, CD133^- cells derived from GBM specimens fulfill all stem cell criteria and are equally as tumorigenic as CD133⁺ when xenografted [22].Furthermore, CD133 and other molecular markers used to identify CSCs are not specific and are found on a host of other normal cell types. In summary, molecular fingerprinting of GBM CSC cells for targeting and destruction is inexact at best, given our current knowledge of the GBM CSC phenotype.

For treatment of CSCs to have any significant future impact on the overall survival of patients with GBM, the underlying molecular signaling that drives tumorigenicity in these cells must be elucidated in much greater detail than is currently known. For example, targeting the molecular signaling that CSCs co-opt, such as the Hedgehog pathway, has led to depletion of CSCs in laboratory models. Indeed, cyclopamine, an agent causing Hedgehog pathway blockade, can reduce the GBM CSC population [12]. Adding further complexity, CSCs are believed to reside within a niche that supports their viability. The interaction between the supportive stroma and vasculature of the niche and CSCs themselves needs to be understood so that both the cancer-sustaining cells and their microenvironment can be eradicated.

The fact that CSCs and normal adult NSCs utilize common molecular machinery and have similar protein expression profiles is potentially the most challenging hurdle to overcome in the development of therapies to target CSCs while sparing NSCs. Analyzing toxicity to NSCs is fraught with all the complexities of studying the effectiveness of such treatment on CSCs. Moreover, destruction of NSCs may not be immediately apparent. Although not in the immediate future, the promise of curing GBM by targeting CSCs is real, but unforeseen toxicity to the normal NSCs in the adult subventricular zone and hippocampus could lead to or facilitate neurodegenerative disease or deficiencies of learning and memory.

Conclusion

The mounting evidence that the growth of GBM is driven by biologically distinct CSCs dictates a re-examination of how anticancer therapies are developed and evaluated. There is a clear need for agents that target these cancer-sustaining cells. Emphasis must be placed on understanding events occurring frequently in CSCs and their local microenvironment during cancer progression, and the molecular mechanisms involved in their resistance to current chemotherapeutics. This understanding will undoubtedly have profound implications for the treatment of GBM and may decrease lethality from the disease.

Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this editorial manuscript.