We have shown that a panel of the most widely used GBM model cell lines fail to faithfully recapitulate the metabolic activity of primary GBM tumours. Furthermore, we have specifically identified several metabolic parameters in which primary GBM cells differ from primary healthy mixed neural cell populations, as well as those parameters which are best mirrored in GBM cell lines. Importantly, the commonly used GBM cell lines showed no difference in baseline glycolysis from primary healthy cells, although they had different basal OCRs, reaffirming the need to target the correct cell lines for bespoke research projects.
Metabolism in the brain relies on an interaction between neurons and astrocytes. Glycogen is predominantly found in astrocytes (18) which contain enzymes to activate high levels of glycolytic metabolism, mainly releasing lactate (19, 20). Lactate and pyruvate transfer can occur to localised neurons (21), where this enters the Krebs cycle and can continue through the oxidative phosphorylation pathway. Thus, an interplay between astrocytes and neurons is required for successful metabolic activity, underscoring our choice of a mixed neural cell population for our primary healthy cells.
Some metabolic parameters of cell activity were found to be similar between primary healthy and GBM cells. As summarised in Table 3, primary healthy and GBM cells were not statistically different in their non-mitochondrial respiration rates or maximal mitochondrial respiration rates, nor were they different in their basal ECAR or ECAR after oligomycin. As such, these parameters may be poor therapeutic targets, given that healthy cells display similar attributes to tumour cells. However, many metabolic parameters were statistically different between primary healthy and GBM cells that could illuminate the most promising targets for novel therapeutics (Table 3).
The addition of FCCP to primary GBM cells (Fig. 3B) did not elicit an increase in OCR, indicating a loss of reserve capacity. This was also seen in the majority of GBM cell lines (U118MG, U251MG, U373MG, D645 and T98G). One explanation for this is that tumour cells are known to predominantly utilise ”The Warburg Effect”, reducing their reliance on mitochondrial respiration (4). Thus, administration of an uncoupling agent may result in a lesser increase in oxygen consumption as the activity of the electron transport chain is already reduced. To support this hypothesis, both the baseline OCR (Fig. 3) and the basal mitochondrial rate (Fig. 5B) of primary GBM cells were observed as significantly lower than that of primary healthy cells.
Studies often utilise GBM cell lines for metabolic research, due to their relative accessibility and propagation potential. The issue that GBM cell lines differ in their metabolic activity from primary disease cells has been previously examined, yet no previous study has incorporated this number of commonly used GBM cell lines for more comprehensive metabolic profiling. Separate tumour cell lines have been found to respond differently to metabolism-altering compounds, with some eliciting heightened responses, due to their endogenous levels of hypoxia-induced factor 1-alpha (22). The PTEN status of cell lines has been shown to affect metabolic activity (23) as well as p53 (24). Potentially due to their mutant p53 alleles (25), the primarily glycolytic lines U251 and T98G have been shown to be unable to properly activate AMPK signalling (26) which is critical for survival in a hypoxic environment (27). In our study, we observed no change in response to rotenone/antimycin A after FCCP administration in D645 and T98G cells, potentially due to similar genomic modifications which may be revealed in further study.
The data presented in this study is useful for the recommendation of cell lines to be used as GBM models in metabolic studies. No single cell line appears to reflect the metabolic activity of primary GBM cells completely (Table 3), although it can be deduced that U87MG and T98G cell lines may be poor models for mitochondrial metabolism. D645 is the only line which accurately reflected the ATP-linked respiration of primary GBM cells, although most other metabolic attributes were not correctly portrayed in the cell line. Therefore, cell lines U251MG, U373MG and D54 cells would be most accurate for modelling the mitochondrial metabolism of primary GBM cells, with D54 cells being the preferred choice for studies researching reserve capacity. T98G clearly would be the optimal model for glycolysis-related metabolic research, mirroring all measured attributes of primary GBM cells.
There are limitations to this study which must be noted. While several GBM cell lines were examined in this study, this list was not exhaustive. Also, this study utilised the Mito Stress Kit (Seahorse Bioscience) for assessing metabolic parameters, which provided a broad assessment of cellular metabolic activity. Further study into the specifics of glycolytic metabolism, fatty acid oxidation and pentose phosphate pathway metabolism could be illuminated using blocking drugs such as 2-DG or 6-aminonicotinamide; these were beyond the scope of this study.