A 3D in vitro glioblastoma cell culture system for identification and evaluation of novel radiosensitisers reducing rodent xenograft studies

Glioblastoma is the most common primary brain tumour and is currently incurable. Despite aggressive treatment involving surgery, radiotherapy and chemotherapy, average life expectancy for glioblastoma patients is about one year. Over the past ten years many large, international clinical trials have tested new treatments but none of them has been successful. This is particularly disappointing because many of the new 'targeted' drugs being tested in these trials seemed to be effective when tested in the laboratory. When glioblastoma cells are cultured in the laboratory they are usually grown as single layers of cells in plastic flasks. These 2-dimensional (2D) cell culture conditions cause marked changes in the shape and behaviour of the tumour cells which affect the way they respond to cancer treatments including radiotherapy and targeted drugs.

Because many new drugs appear to be effective in 2D cell cultures, they are then tested in mice and rats implanted with glioblastoma cells. Some drugs show promising results in these animal experiments but are still not effective when tested in patients. So, we need a different method for developing and testing new treatments. First we need to understand why glioblastomas are resistant to radiotherapy, chemotherapy and the new targeted drugs. Then we need to develop new treatments that overcome these mechanisms of resistance. To avoid treating large numbers of animals with ineffective drugs, we also need to be more confident that drugs will work in patients before we start performing animal experiments.

To address these issues we have developed a new, 3D model of glioblastoma that can be grown in the laboratory using tumour cells from patients with glioblastoma. These cells are grown on polystyrene scaffolds coated with special proteins found in glioblastomas, and are nourished with specialised growth factors that are also present in glioblastomas. We have shown that this 3D model contains many of the features that we see in tumours in patients. More importantly, we have shown that drugs which work in patients also work in the 3D model, while drugs which don't work in patients have no effect in the 3D model. Some of these drugs had opposite effects on cells grown in 2D and 3D conditions.

Because we have confidence in the 3D model, we believe it will be valuable to look for genes and proteins that are switched on when 3D cells are treated with radiotherapy, and then test new drugs that can block the effects of these genes and proteins. We have already found several genes that are switched on by radiotherapy in 3D but not 2D cells, and our early experiments suggest that we can overcome resistance to radiotherapy by targeting these genes. in this way we will identify new drugs that are much more likely to be effective in patients.

However we realise that our current 3D model is rather simple and that lots of other cells and structures in glioblastoma might be important. Tumour blood vessels are particularly influential because the cells that line them (endothelial cells) produce chemicals that nourish the tumour cells and make them resistant to radiotherapy. We will therefore develop a more complex 3D model composed of glioblastoma cells and human brain microvascular endothelial cells and see if the new drugs are also effective in this new 'multicellular' model. At the same time we will investigate how the the different cell types interact and how this causes resistance to treatment. Finally, we will convert the new 3D model into a format that allows 'high throughput screening' of new drugs. This will allow us and other researchers around the world to test very large numbers of new drugs as efficiently as possible.

Overall we aim to improve treatments for glioblastoma patients while reducing the number of animal experiments. We will achieve these aims by developing a new 3D model of glioblastoma that accurately predicts which new drugs will be effective in patients.

Glioblastoma (GBM) is the most common primary brain tumor. Tumors exhibit inherent chemo- and radioresistance which has been attributed to a subpopulation of cancer cells termed 'GBM stem-like cells' (GSC), and almost inevitably recur. While preclinical studies have shown promising activity of several molecularly targeted agents against GBM cell lines, these agents have failed to improve clinical outcomes for patients. The failure of drug-radiation combinations with promising preclinical data to translate into effective clinical treatments may relate to the frequent use of established GBM cell lines in simplified two-dimensional (2D) in vitro cultures. We have developed a novel 3D-Alvetex GBM model system that recapitulates key histological features of GBM including high cellularity, sparse extracellular matrix and presence of GSC. Using this model, we have reproduced clinical outcomes including (i) lack of response to EGFR-directed therapies alone and in combination with radiation and (ii) enhancement of radiosensitivity by VEGF targeting, providing evidence for this culture model as a clinically relevant platform for evaluating targeted therapies alone and in combination with radiation. Genomic characterisation (RNA-Seq) of two different primary GBM cell lines grown in this 3D system, has identified 71 transcripts that are upregulated following radiation treatment. Following validation of transcripts for which inhibitors are commercially available, using RT-PCR in the 3D model and IHC in an existing human GBM TMA and in house human orthotopic glioblastoma xenografts, we will evaluate the radiosensitising properties of target inhibitors in the 3D model and in a novel multicellular 'perivascular niche' system comprising 3D GSC and human brain endothelial cells. We believe that our models will provide meaningful preclinical assessment of novel molecular targets, improving accuracy of in vitro drug evaluation while reducing and partially replacing in vivo models.

In 2015, according to a Pubmed search using the keywords glioblastoma (GBM) mouse animal models, 43 articles were published reporting preclinical trials of multiple agents. These experiments consisted of either intracranial or flank injection of established GBM cell lines (e.g. U87) or patient-derived primary cells lines to assess the therapeutic effects of multiple agents, from natural killer cells (Lee et al, BMC Cancer. 2015) to molecular targeted agents such as the tumor necrosis factor-related apoptosis-inducing ligand (Crommentuijn et al, Mol Oncol. 2015). From a sample of 20% of the research articles published in 2015, we calculated 70 mice were used per study. Extrapolating these values to the 43 papers found, we estimate that a total of 3165 mice were used in 2015 alone for experiments assessing novel drugs or treatment combinations for GBM. In our laboratory, we used 180 mice for our GBM orthotopic models in the same year. Most of these animal studies are the endpoint of preclinical assessment of molecularly targeted therapies performed in cell lines cultured in the laboratory. Routinely, cell lines grown in the laboratory are cultured in simplified two-dimensional (2D) in vitro systems, which induce drastic changes both in morphology and phenotype of three-dimensional cells in vivo. One very good example of how the phenotypical changes induced by 2D culture might produce misleading conclusions comes from drug studies with 2D cancer cells, where a big discrepancy between 2D cell responses to cytotoxic treatments compared to those observed in patients is observed. These observations suggest that over-simplified 2D cell culture conditions are not predictive of clinical efficacy, and might explain the failure of the new generation of molecularly targeted agents to improve outcomes for cancer patients. Furthermore, the use of inappropriate in vitro models leads to unnecessary animal studies that either fail to show activity in vivo or show activity that does not translate to clinical effectiveness. In summary, molecular targeted therapies that exhibit activity in preclinical studies that utilise conventional experimental models generally fail to deliver clinical benefit to patients. Development of in vitro models that recapitulate the three-dimensional microenvironment of tumours will provide reliable platforms for in vitro assessment of molecular targeted agents which will translate better into the clinic.

In our laboratory, we have developed a novel, three-dimensional (3D) GBM culture system that replicates effects observed in the clinic, such as the lack of response to EGFR-directed therapies alone and in combination with radiation, or the therapeutic gain produced by bevacizumab. We aim to use this model to reduce the number of rodent models required for preclinical evaluation of novel drug/radiation combinations for GBM. Our preliminary data indicate that the 3D-A model developed in our laboratory will provide meaningful preclinical assessment of these molecular targets. This will result in a reduction in animal studies by (i) improving the accuracy of in vitro evaluation of novel compounds thus reducing the number of compounds proceeding to in vivo evaluation and (ii) partially replacing in vivo models in the preclinical drug development pipeline. We hypothesise that our 3D model will successfully eliminate at least 50% of drugs in development for GBM on the basis of lack of efficacy. This will reduce the number of drugs proceeding to animal testing by 50%, and therefore reduce the number of animals used by at least 1500 per year. In the longer term, we predict larger reductions by replacing animal testing with a fully optimised 3D model.