Glioblastoma (GBM) is the most representative tumor of the central nervous system and is also a refractory tumor in humans. It is also known as glioblastoma multiforme because of its diverse and heterogeniety histomorphology among different tumors and within the same tumor. In the last two decades, the understanding of GBM heterogeneity has grown to refer not only to its histological manifestations, but also to its heterogeneous biological behavior (proliferation and apoptosis, angiogenesis and invasive migration) and, by extension, to its heterogeneous therapeutic response [1,2]. Throughout the 1990s, the heterogeneity of GBM was attributed only to genetic phenotypic changes in tumor cells. The discovery of brain tumor stem cells was an important event in neuro-oncology research and has profoundly influenced our scientific research on the mode of GBM development, tumor heterogeneity, tumor cell-microenvironment relationship, and therapeutic strategies [1-3].
I. Brain tumor stem cell identification and screening
Only about 0.01-1% of tumor cells in acute myeloid leukemia can be induced by experimental transplantation, which was the first evidence of the existence of tumor stem cells [2]. Subsequently, tumor stem cells were also found in solid tumors such as breast, colon, pancreas, and lung [4]. Tumor stem cells have also been found in high-grade primary brain tumors, such as gliomas, medulloblastomas, ventricular meningiomas, and neuroblastomas; no positive findings have been made in low-grade brain tumors [5]. Brain tumor stem cells exhibit similar characteristics to neural stem cells, such as self-renewal ability, formation of cell balls in serum-free medium, expression of stem cell markers and multispectral differentiation potential; brain tumor stem cells differ from neural stem cells in their abnormal expression of differentiation markers, chromosomal abnormalities and ability to form tumors [5]. If there is a need to establish and standardize what is most urgent in the field of brain tumor stem cell research, it is the identification and screening of brain tumor stem cells, which concerns the whole theoretical system of brain tumor stem cells. In fact, the results of studies on brain tumor stem cells are inconsistent with each other or even completely contradictory, and the different methods and procedures for the isolation and enrichment of tumor cell subpopulations are one of the important reasons.
The most frequently used cell surface molecular markers in GBM are CD133, A2B5, CD5, CD171, ITGA6 and EGFR [3]. CD133, A2B5 and CD15 are surrogate markers for which no clear biological function has been found, while CD171, ITGA6 and EGFR are functional markers [3]. Immunomagnetic bead sorting and fluorescenceactivatingcellsorting (FACS) use these molecular markers precisely for the isolation of tumor stem cells.
We use CD133 as an example to illustrate the impact of methodologies using surrogate markers of tumor stem cells on the interpretation of study results.CD133 was first used for the isolation of hematopoietic and neural stem cells and is also widely used for the identification and sorting of brain tumor stem cells [6,7]. In GBM, CD133+ cells were grown intracranially in NOD-SCID mice with tumorigenic capacity compared to CD133- cells [3]. However, different results have also been reported, where CD133- cells isolated from either GBM tissue specimens or in vitro cultured cell lines also had tumorigenic capacity [8,9]. Such inconsistent results may be caused by methodological defects. First, compared to the FACS method, the immunomagnetic bead sorting method lacks specificity, and the CD133+ cells obtained by sorting are mixed with CD33- cells that bind non-specifically to the immunomagnetic beads, and not only CD133+ cells are enriched [10]. Analogously, the conclusion that “CD133-cells are tumor-forming” may also be due to the mixing of CD133-cells with CD133+ cells during sorting [11]. Of course, standardized cell purification determination methods and rigorously designed internal control experiments can help to exclude cell confounding factors. Further, endothelial cells in the glioma stroma express CD133, and CD133+ cells isolated from tumor tissue will contain CD133+ endothelial cells, and the growth advantage of CD133+ cells over CD133- cells in the transplantation model. However, a recent study reported that both FACS-purified CD133+ cells and CD133- cells have the ability to promote tumor growth in nude mice, even when the effect of endothelial cells is excluded [11]. The types of anti-CD133 antibodies used for tumor stem cell sorting include two antibodies, anti-CD133/1 and anti-CD133/2, which recognize two different glycosylated epitopes of the CD133 transmembrane protein, thus also affecting the consistency between the results of the studies [12]. The inconsistency of experimental results due to methodological flaws is also present for surrogate molecular markers such as A2B5 and CD15 [3].
The discovery and optimization of functional markers that can serve both as tumor stem cell enrichment markers and as therapeutic targets is an important research direction. We further illustrate this with the example of the epidermalgrowthfactorreceptor (EGFR), a marker of neural stem cells [13] with a relevant role in regulating the division and stemness maintenance of neural stem cells in the subventricular zone of the brain [14,15]. Molecular neuropathology has long found that EGFR is expressed in more than 60% of primary GBM, but not in secondary GBM, with diagnostic and prognostic implications. The distribution of EGFR+ tumor cells in the same GBM tumor specimen is heterogeneous, suggesting that EGFR can be used as a molecular marker to distinguish cell subpopulations in GBM [16]. GBM cell subpopulations were isolated by FACS from human GBM tissue specimens and tumor stem cell lines, in which EGFR+ cell subpopulations exhibited the most malignant molecular and functional phenotypes regardless of whether they were co-expressed with CD133 and CD15 [3]. Alteration of EGFR expression in cells of tumor stem cell lines by gain or loss of function results in corresponding cell growth promoting or inhibiting effects. Clearly, EGFR expression is essential for glioma formation and can be used not only as a functional marker of glioma cell subpopulations but also as a therapeutic target for GBM. It was also found that EGFR-tumor cell subpopulations can show EGFR re-expression during transplantation tumor formation, suggesting that the regulation of expression of stem cell markers is a dynamic process. Another recent study demonstrated that dynamic maintenance of tumor cell subpopulations does exist in melanoma and GBM [17,18].
II. Theory of tumorigenesis in glioma
The complex tumor cell composition is dynamically changing and there is a constant need for cellular replenishment in tumor cell proliferation, invasion and specialization, and there are two explanatory mechanisms for the tumor inhomogeneity caused by this process, namely: hierarchicalmodel and stochasticmodel.
The hierarchicalmodel refers to the fact that tumorigenesis and maintenance require a small subpopulation of cells with “stemness”, i.e. tumor stem cells. The majority of tumor cells differentiated from tumor stem cells are not capable of self-renewal and have no significant effect on tumor immortalization. Tumor heterogeneity is caused by the coexistence of tumor stem cells and their differentiated progeny cells. The stochastic model is one in which most cells within the tumor have the ability to self-renew, contributing to both tumorigenesis and maintenance. The heterogeneity of tumors is mainly due to molecular genetic and epigenetic differences between tumor cell clones. More importantly, the stochastic model assumes that all tumor cells within a tumor have the capacity for tumor formation, although to some extent, the tumor formation process also requires a team of cells to work together to maintain it; the differences in the phenotypes of the cells replenished within the tumor reflect the cell clones present at different stages of tumor transformation and malignancy [19,20].
Initially, several studies demonstrated that in hematopoietic malignancies, breast and colon cancers, only a small subpopulation of tumor cells has the ability to form new tumors, a strong support for the hierarchical model [21]. However, successive studies have questioned whether the hierarchical model is appropriate for all tumor types. kelly et al. 2007 found that more than 10% of cells isolated from three different mouse models of primary hematopoietic system tumors had the ability to induce tumors in non-irradiated recipient mice [22]. shmelkov et al. 2008 reported that in mouse colon cancer CD133-cells also had tumor formation capacity [23]. Echoing this, single-cell transplantation analysis found more than 25% tumor initiating cells in human melanoma. Similar reports have recently been made in glioblastoma. Indeed, the grade model may be more appropriate for hematopoietic system tumors and several solid tumors, such as breast, colon, and pancreatic cancers, as well as medulloblastoma, which have a strict and well-defined grade of cells in the tissue of origin [11,18]. In contrast, GBM has tissues of origin with mesenchymal-like architecture, such as neural crest and mature brain, which may require a random pattern and a highly flexible tumorigenesis pattern. In GBM, there are multiple groups of active tumor-initiating cells, each of which can be distinguished by specific marker protein expression and different functional phenotypes or gene-molecular characteristics (invasive, pro-angiogenic, or proliferative). Random patterns of tumorigenesis may be more appropriate than hierarchical patterns to explain the intrinsic heterogeneity of GBM, but it cannot be assumed that hierarchical patterns should be excluded altogether. It has been recently suggested in leukemia studies that the hierarchical and random patterns are not mutually exclusive and that tumor cell evolution is governed by both patterns [11].
III. Microenvironment of glioma stem cells
If the tumor is recognized as a microecological system, there exists not only a relationship between different cell clones but also a relationship between the tumor cells and the microenvironment. In this synergistic system, tumor cell clones compete with each other for oxygen, nutrients, and space, and the stronger wins through natural selection; the system creates a local microenvironment between tumor and non-tumor cells to promote tumor growth, invasion, apoptosis and/or treatment resistance, and immune evasion. Components of the microenvironment in brain tumors include microglia, macrophages, astrocytes, oligodendrocytes, neurons, glial and neuronal progenitor cells, extracellular matrix, pericytes, and endothelial cells [1].GBM cells readily invade along myelinated axons, vascular basement membranes, or subventricular membranes, illustrating the influence of the microenvironment on the invasive migration of tumor cells [24]. The interaction between glioma cells and endothelial cells in the microenvironment is important to maintain the “stemness” of tumor stem cells [25]. In xenograft models, endothelial cell or vascular proliferation can expand the self-renewing cell population and accelerate tumor growth. Conversely, targeted inhibition of EGFR downregulation of VEGF by erlotinib or direct neutralization of VEGF by bevacizumab reduces tumor growth and decreases the number of cells with the capacity for self-renewal.
The maintenance and survival of tumor stem cells is regulated both intrinsically, through proliferation and survival pathways, including c-Myc,Oct4 (POU5F1), Olig2 and Bmi1 [26], and extrinsically, through growth factors and interactions with the cellular matrix, which occur in the microenvironment or niches in which the tumor is located. Studies have demonstrated that the mutual communication between tumor stem cells and the microenvironment in which they reside can influence cell fate in GBM [27,28]. However, the interactions between tumor cells and the microenvironment are still far from being understood. In fast-growing gliomas, it remains unclear whether tumor cells establish their own microenvironment or an existing microenvironment to recruit tumor stem cells? What is a bidirectional interaction relationship between tumor cells and stroma? What is the role of tumor stem cells at the leading edge of tumor invasion and how do they interact with the immune system? An in-depth study of the communication mechanisms between tumor stem cells and the microenvironment will be useful for tumor stem cells in tumorigenesis, growth, maintenance, invasion, and therapeutic resistance, and will lead to more effective therapeutic strategies for malignant gliomas.
Immunodeficient mice are commonly used for in vivo studies of tumor stem cells, but immunodeficient mice are unable to replicate components of the immune system that are present in tumor patients and are important drivers of cell grade formation [5]. The immunodeficient state may allow the growth of certain tumor cells, also distorting the real situation in the patient. Genetically engineered modified murine models can generate homozygous hosts, but the relative lack of cellular heterogeneity and the differential factor signaling between different species of cells remains a challenge for tumor stem cell research. It should be said that there is no model system that can fully represent the wide variety of molecular genetic and epigenetic changes under human conditions and faithfully reflect the intercommunication characteristics between tumor stem cells and the microenvironment in which they are located.
IV. Hints for treatment
Molecular targeted therapies are less toxic and more efficient than traditional cytotoxic drugs and can complement existing therapeutic approaches, but the results of first-generation targeted therapy trials in malignant gliomas are not encouraging. Indeed, single-agent therapy with targeted agents has not been successful in recurrent malignant gliomas. Apparently, the heterogeneity of tumor proliferating cell subpopulations is one of the major reasons for the clinical response to therapy [11,18]. For example, the use of small molecule tyrosine kinase (TK) inhibitors to inhibit the EGFR kinase domain could be a therapeutic strategy for GBM, but the only few clinical studies reported to date have demonstrated that EGFR inhibitors alone are only partially effective in GBM [29], due to the presence of high levels of Akt-dependent signaling in tumors with PTEN mutations and the co-activation of multiple TKs in the same tumor cells [30,31]. Indeed, highly malignant EGFR+ GBM cells do respond to TK inhibitors, and the choice of EGFR as a therapeutic target is reasonable, except for the coexistence of EGFR+ and EGFR- cells in the same tumor, and the unresponsiveness of EGFR- cells to treatment leading to tumor recurrence [11]. The development of a multi-targeted combination inhibition strategy, targeting multiple molecules and cells simultaneously, is expected to overcome the resistance of some cell subpopulations of GBM to treatment [3].
In response to the heterogeneous nature of malignant gliomas, further therapeutic strategies should consider the following aspects.
(1) continue to search for molecular markers that can selectively identify GBM cell subpopulations, especially functional ones, not only as potential therapeutic targets, but also to anticipate treatment response and prognosis [32].
(2) Combining genotypes and cell lineage genesis for the typing of malignant gliomas through genomic and transcriptional profiling surveys of bulk clinical tumor specimens can not only facilitate the understanding of functional differences among different cell subpopulations of tumors, but also identify new targets for therapy [33].
(3) Identification of tumor cell subpopulations in GBM with minimal genetic tags prospectively anticipates whether patients will benefit from specific targeted therapies, enabling individualized and optimal treatment of GBM [13,34-36].
(4) An effective treatment strategy for GBM should take into account all actors on the GBM “stage” and keep in mind that the heterogeneity of GBM should be considered as a dynamic process and that GBM cell subpopulations with specific molecular markers also have a temporal or window period and must be treated according to a balanced treatment strategy and time sequence [3]. therapeutic tools [3].