Glioma: experimental models and reality

Abstract

In theory, in vitro and in vivo models for human gliomas have great potential to not only enhance our understanding of glioma biology, but also to facilitate the development of novel treatment strategies for these tumors. For reliable prediction and validation of the effects of different therapeutic modalities, however, glioma models need to comply with specific and more strict demands than other models of cancer, and these demands are directly related to the combination of genetic aberrations and the specific brain micro-environment gliomas grow in. This review starts with a brief introduction on the pathological and molecular characteristics of gliomas, followed by an overview of the models that have been used in the last decades in glioma research. Next, we will discuss how these models may play a role in better understanding glioma development and especially in how they can aid in the design and optimization of novel therapies. The strengths and weaknesses of the different models will be discussed in light of genotypic, phenotypic and metabolic characteristics of human gliomas. The last part of this review provides some examples of how therapy experiments using glioma models can lead to deceptive results when such characteristics are not properly taken into account.

Fig. 1

Overview of oncogenic pathways in glioma and possibilities for pharmacological interventions, relevant for the models described in this article. Growth factors bind to RTKs resulting in phosphorylation and Ras/Raf and PI3 K signaling to Akt and mTOR, thus driving cell growth and survival. In a substantial percentage of glioblastomas this process is amplified by loss of the counteracting phosphatase PTEN. Akt also can phosphorylate MDM2 thereby stabilizing the protein. This stabilization causes ubiquitinylation and degradation of the tumor suppressor P53, unleashing cyclin D1 complex activity and leading to uncontrolled G1-S progression. Loss of control over the cell cycle is also induced by CDKN2A mutations as CDKN2A products, p14^ARF an p16^INK4A are important G1-S checkpoint proteins. The PI3 K pathway (activated by phosphorylated growth factor receptor tyrosine kinases [RTKs] such as EGFR, MET or PDGR) may be inhibited by small molecule RTK inhibitors or antibodies directed against ligand binding domains of RTKs or the ligands themselves, thus prevent ligand-receptor interaction. The cyclin D1 pathway (resulting in cell cycle progression from G1 to S phase) may be inhibited by the CDK4/6 inhibitor palbociclib. Finally, mutations in IDH1 or IDH2 result in mutant proteins that catalyze the conversion of α-KG to D-2-HG, causing the G-CIMP phenotype and a transcriptional profile leading to gliomagenesis. AGI-5198 specifically inhibits IDH1^R132H activity by binding to the catalytic site of the protein

Fig. 2

Schematic overview of in vitro and in vivo glioma models. a Intravenous injection of ENU into pregnant rat leads to offspring with a high chance of spontaneous glioma development. Resulting gliomas often carry mutations in oncogenes or tumor suppressor genes that are frequently encountered in human gliomas. ENU-induced gliomas have been processed to murine glioma cell lines that can be implanted as allografts in syngeneic animals. b GEMMs. Shown are examples of compound flox-ed mouse, in which the floxed modifications can be activated by local transduction of cells with lentiviruses encoding Cre recombinase under control of ubiquitous or cell-type specific promoters, or by crossing with transgenic mice expressing Cre under control of neural- or glial-cell type-specific promoters (e.g., nestin or GFAP promoter). In these cases modifications occur early during embryonal development, unless promoter activity is made inducible. These glioma models are molecularly highly defined, and can be processed to novel murine cell lines that are amenable for allografting. c PDGCs or PDX can be generated by preparing cell cultures of surgically obtained glioma material (nowadays mostly spheroid cultures) that can be implanted heterotopically (generally subcutaneously in the flank) or orthotopically (in the brain) of immunocompromised animals; alternatively, surgically obtained human glioma tissue can be directly implanted; ideally, an orthotopic xenograft of a diffuse high grade glioma/glioblastoma in the murine brain recapitulates not only the genotype, but also the phenotype of this tumor with e.g., florid microvascular proliferation (lower right image) and diffuse infiltrative growth in the white matter (upper right image). Of note, as illustrated by a xenograft derived from U87 cells, not all orthotopic glioma models show diffuse infiltration in brain parenchyma, and such models are less relevant for the study of glioma in the context of tumor-brain microenvironment interactions

Fig. 3

Cell culture conditions impact the phenotype of orthotopic E98 glioma xenografts. E98 glioma cells grown as adherent cells in serum-supplemented culture medium lose the capacity to grow diffusely in the brain (left panel). Glioma cells are visualized via immunohistochemical staining for c-MET. Note the sharp demarcation between tumor (growing in the leptomeninges here) and brain parenchyma in the left panel, whereas E98 tumor cells diffusely infiltrate in the brain parenchyma after passaging as neurospheres (right panel). Of note, photographs are representative examples from experiments in which E98 cells, grown as adherent cultures or spheroid cultures, were injected in groups of 5 mice using exactly the same injection procedure. Bars indicate 200 µm

Fig. 4

Example of promising in vitro effect of CDKN2A-mutation targeting that is difficult to translate into clinical efficacy. Two different patient-derived glioblastoma models (E98 adherent cells, N13-16 spheroid cultures, both characterized by dysfunctional CDKN2A), were incubated with placebo (left panels) or with the CDK4/6 inhibitor palbociclib (right panels). Visualization of DNA synthesis (S-phase of the cell cycle) with BrdU or EdU (as indicated) shows that palbociclib effectively prevents entry into the S-phase. Although such in vitro results are very promising, also because 80% of human glioblastomas have dysfunctional CDKN2A leading to a defective G1-arrest, in vivo diffuse gliomas may well be protected from palbociclib because this drug is substrate for p-glycoproteins of the BBB [36]

Fig. 5

Schematic overview of tumor growth via angiogenesis versus vessel co-option. a Schematic representation of the dogma of angiogenesis-dependent growth of a tumor. When a tumor outgrows the capacity of the vasculature, hypoxic stress (indicated by blue cells) initiates sprouting angiogenesis as a rescue pathway (here presented as yellow vessels. b Especially in tissue with rich pre-existent vasculature (such as brain tissue), tumors may grow through vessel co-option in an angiogenesis-independent fashion

Fig. 6

Schematic representation of the glutamine-glutamate cycle in the brain. In the brain, glutamate (Glu) is released by presynaptic neurons to the synaptic cleft. Glutamate activates glutaminergic receptors that undergo a conformational change to allow influx of extracellular calcium. This triggers membrane depolarization in the postsynaptic neuron and induces signal transduction. The excess of glutamate in the synaptic cleft has to be removed to prevent excitotoxicity. Astrocytes take up surplus glutamate through various glutamate importers and convert it to glutamine (Gln) through glutamine synthetase. Glutamine is exported to the capillaries (not shown), or is transferred back to the neurons. Subsequently, neurons can then convert back the glutamine to glutamate, closing the glutamine-glutamate cycle. Glutamate can be imported by IDH ^mut glioma cells in order to supply cells with αKG as a rescue pathway