Introduction
Gliomas, glial cell derived central nervous system malignancies, are a heterogeneous, aggressive tumour type with poor prognosis. The incidence of isocitrate dehydrogenase (IDH) mutations is high in low-grade gliomas. Despite the fact that these malignancies are still incurable, patients with IDH-mutant gliomas have a better prognosis and response to chemo-and radiotherapy than patients with IDH wild-type tumours [
1,
2]. IDH mutations can also be implicated in the formation of other tumour types (acute myeloid leukaemia – AML, chondrosarcomas, intrahepatic cholangiocarcinoma – ICC). In these non-glial malignancies, IDH mutations appear to confer a worse prognosis to the patient; although there is some controversy in case of ICC and AMLs [
3,
4].
Based on highly detailed analyses of the genetic basis for malignant progression (gene amplifications, mutations, loss of chromosome arms, gene expression, DNA methylation status), several altered pathways in IDH-mutated gliomas were characterised, including RTK-PI3K-mTOR, Notch signalling, cell cycle and DNA damage response regulation. These studies concluded that IDH mutation is at the centre of epigenetic alterations in glioma cells [
5,
6]. The prevalence of IDH mutation is high in grade II-III and secondary glioblastomas (70–80%), the most frequently (> 90%) mutated IDH isoform is the cytosolic IDH1 R132H [
7,
8], that has gained a neomorphic activity resulting in conversion of α-ketoglutarate (aKG) to the oncometabolite D-2-hydroxyglutarate (2-HG). In addition, the mutation has a negative impact on the function on the wild-type allele [
9].
2-HG can accumulate to millimolar concentrations in cells and in the extracellular environment [
10,
11]. Moreover, mutated IDH1 not only causes 2-HG production, but also initiates changes in metabolic adaptation of mutated cells [
12]. Mutant IDH1 oxidises NADPH to NADP
+ during 2-HG production, instead of reducing NADP
+ to NADPH as is done by the wild-type IDH1 enzyme during conversion of isocitrate to aKG. Because 2-HG and aKG are structurally very similar metabolites, 2-HG acts as a competitive inhibitor of aKG-dependent dioxygenases including prolyl-hydroxylases, histone- and DNA demethylases [
13‐
15]. Cancer-related metabolic alterations, adaptations and their role in tumourigenesis and progression are extensively studied in many cancers. The metabolic characterisation of human gliomas started with the reinterpretation of Warburg effect suggesting that other substrates than glucose can also be oxidised by glioma cells [
12,
16,
17], such as glutamine and acetate [
18‐
20].
Multiple substrates can fill up the TCA cycle – which is in the centre of metabolic activity – to provide acetyl-CoA; and many studies reported that cells can use glutamine and/or glutamate as catabolic substrates [
21‐
23]. Previous results suggested that glioma cells adaptively run the TCA cycle backwards to power the electron transport chain, especially in case of IDH-mutant cells. Analogous to fumarate hydratase (FH) and succinate dehydrogenase (SDH) mutations, which can also cause oncometabolite production, IDH mutation provides an example of how a single mutation can reprogram cellular metabolism and affect pathobiology [
13,
15,
24].
Recent years have yielded exciting findings in the field of tumour metabolism, and some aspects of these metabolic alterations related to IDH mutations have also been described [
25,
26].
Astrocytes and GABA-ergic and glutamatergic neurons play roles in the maintenance of the neurotransmitter pools of glutamate and gamma-aminobutyric acid (GABA). Astrocytes and neurons are in metabolic symbiosis, astrocytes can catabolise glutamate and GABA for supporting and regulating neuronal functions. GABA can feed the TCA cycle via the activities of GABA transaminase and succinic semialdehyde dehydrogenase (SSADH) to produce succinate [
27]. The expression changes of GABA utilisation enzymes and GABA consumption were only studied in few recent publications in cancer research [
28,
29].
An emerging question is whether glioma cells have the ability to use the GABA shunt to support energy production and proliferation. This indicates a need to understand the effect of produced 2-HG (caused by IDH1 mutations) on the former pathway.
Our study provides an exciting opportunity to advance our knowledge of the effects of IDH1 mutation and 2-HG accumulation to identify several mechanisms which can help to understand the Janus-faced role of IDH1 mutations in the progression of gliomas – IDH mutations could be driver oncogenic alterations, however, IDH-mutant human gliomas have better prognosis than wild-type. In the present study, the alterations of bioenergetic characteristics and different substrate oxidation capacities in relation to the expressions of relevant proteins in glioma cell lines were analysed (Additional file
1: Figure S1 summarises the analysed pathways). Our presented results demonstrate that SSADH expression – an IDH mutation independent in vivo characteristic of human glioma cells – provides a possibility for GABA oxidation. This may have special importance in survival, proliferation and metabolic adaptation of glioma cells.
Methods
All materials were purchased from Merck-Sigma-Aldrich (Darmstadt, Germany), except where it is indicated in the text.
In vitro cell cultures
U251 MG (U251 wt), genetically engineered U251 MG mutant IDH1 R132H overexpressing (U251 IDH1m) [
25], U87 MG (ATCC) and U373 MG (kindly provided by G. Sáfrány) human astrocytoma/glioblastoma cells were cultured in DMEM high glucose medium supplemented with 10% foetal bovine serum, 2 mM L-glutamine and 100 UI/ml penicillin-streptomycin. R132H mutation status was confirmed by immunocytochemistry using anti-IDH1 R132H (Dianova, H09 clone, 1:80) antibody in different cell lines after paraformaldehyde (4%) fixation. For metabolic flux experiments, cells were washed with DMEM D5030 (glucose-, glutamine- and pyruvate-free medium) followed by incubation with DMEM D5030 medium containing U-
13C-glucose, U-
13C-glutamine or 2-
13C-acetate (Cambridge Isotope Laboratories Inc.). T25 or T75 flasks (for Western blot, liquid chromatography-mass spectrometry (LC-MS) and
13C-labelling experiments) and 96-well plates (sulforhodamine-B and Alamar Blue proliferation assays), special Seahorse plates were used with appropriate cell numbers (3–8 × 10
5 cells/flask or 2.5 × 10
3 – 5 × 10
4 cells/well) for different treatments and measurements. To analyse the direct effect of 2-HG accumulation and the role of GABA metabolism in U251 wt cells, 4 mM 2-HG, 5 mM GABA and 0.6 mM vigabatrin were added for various time periods (24 h, 72 h and 14/28 days). We selected concentrations based on previous publications.
Cell proliferation measurements
Beside cell counting the proliferation rate was analysed by sulforhodamine-B (SRB) assay which correlates to the protein content of the analysed samples. For 72-h proliferation assays 2.5 × 103 untreated or pre-treated glioma cells/100 μl well were seeded; followed by incubation under different conditions. Thereafter the cells were fixed with 10% trichloroacetic acid subsequently stained with SRB using 10 mM Tris, and the absorbance was measured at 570 nm. In addition, Alamar Blue (Thermo Fisher Scientific) test was also used for proliferation studies. The incubation period was 4 h, fluorescence was measured by Fluoroskan Ascent FL fluorimeter software (Labsystems International). Percentages of proliferation were given relative to control samples.
Expression analysis of different proteins by Western blot
Protein concentrations of cell extracts were measured with using Bradford assay (Biorad). Equal amounts of proteins were separated by 8–12% SDS-PAGE gels following their transfer to PVDF membrane applying BioRad Trans Blot Cell. Membranes were stained with Ponceau S, before incubation the following primary antibodies were used in expression studies: anti-phosphofructokinase P (PFKP) (Cell Signaling 1:1000 #8164); anti-hexokinase 2 (HK2) (Cell Signaling 1:1000 #2867); anti-β-F1-ATPase (ATPB) (Abcam 1:2000 #14370), anti-glutaminase (Gls) (Abcam 1:1000 #156876); anti-acetyl-CoA synthetase 2 (ACSS2) (Cell Signaling 1:1000 #3658); anti-alanine, serine, cysteine-preferring transporter 2 (ASCT2) (Bethyl 1:2000 #A304-353A); anti-succinic semialdehyde dehydrogenase (SSADH) (Abcam 1:1000 #129017); anti-GABA transporter 1 (GAT1) (Abcam 1:500 #426). Finally, biotinylated secondary antibodies, Vectastain Elite ABC Kit (Vector) and enhanced chemiluminescence technique (Thermo Fisher Scientific ECL Western Blotting Substrate) were applied by using C-Digit System (Li Cor Biosciences) detection equipment. The membranes were stripped and probed by anti-β-actin antibody (Sigma-Aldrich; #A228; 1:5000), to prove equal protein loading as well.
Assay to detect cellular respiration and extracellular pH changes
Real-time measurements of oxygen consumption rate (OCR), reflecting mitochondrial oxidation and extracellular acidification rate (ECAR), indicated as a parameter of glycolytic activity, were performed on a Seahorse XF96 Analyzer (Agilent Technologies, USA) based on previous descriptions [
30‐
32].
Glioma cell lines were plated in 100 μl complete DMEM growth media at 1.5 × 104 cells/well density onto 96-well Seahorse plates (Agilent Technologies, USA) 24 h prior to the assays. The medium was removed and was replaced by glucose-, glutamine- and pyruvate-free DMEM medium (D5030 pH 7.4). The basal OCR and ECAR were calculated via XF96 Analyzer software (Agilent Technologies, USA) after 1.5-h incubation at this condition.
Following this, substrate utilisation was measured by using different substrates in parallel wells.
During the measurements freshly prepared substrates (glucose 10 mM, glutamine 2.5 mM, citrate 5 mM, GABA 5 mM, lactate 5 mM, malate 10 mM, acetate 10 mM and glutamate 5 mM) and/or metabolic inhibitors/modulators (oligomycin 2 μM, 2,4-dinitrophenol - DNP 100 μM and antimycin A + rotenone 1–1 μM) were injected into each well to reach the desired final working concentration.
LC-MS/MS method for quantitative metabolite analysis
Intra- or extracellular metabolites - citrate, aKG, succinate, fumarate, malate, glutamate, 2-HG - were extracted by a modified method based on Szoboszlai et al. [
33,
34]. Cells (minimum 5 × 10
5 cells) were quenched in liquid nitrogen. Metabolites were extracted from cells and in parallel 300–500 μl supernatant by methanol-chloroform-H
2O (9:1:1) and vortexed at 4 °C. After centrifugation (15,000×g, 10 min, 4 °C) the clear supernatants were stored at -80 °C until measurements. Citrate, aKG, succinate, fumarate, malate, glutamate and 2-HG concentrations were assessed by using calibration curves obtained with the dilution of analytical purity standards in the range of 0.5–50 μM. LC-MS grade water, LC-MS grade methanol and LC-MS grade formic acid were purchased from VWR International Ltd. (Debrecen, Hungary).
LC-MS/MS assays were performed on a Perkin-Elmer Flexar FX10 ultra-performance liquid chromatograph coupled with a Sciex 5500 QTRAP mass spectrometer. For chromatographic separation a Phenomenex Luna Omega C18 column, (100 × 2.1 mm, 1.6 μm) was used (GenLab Ltd., Budapest, Hungary). The mobile phase consisted of water containing 0.1% (v/v) formic acid (A) and methanol containing 0.1% (v/v) formic acid (B). The mass spectrometer was operating in negative electrospray ionisation mode with the following settings: source temperature: 300 °C ionisation voltage: − 4000 V, curtain gas: 35 psi, gas1: 35 psi, gas2: 35 psi, entrance potential: -10 V, CAD gas: medium. Quantitative analysis was performed in multiple reaction monitoring (MRM) mode.
Different substrate consumptions were analysed using 13C-labelling and LC-MS measurements
Sub-confluent cells were washed and incubated in DMEM D5030 medium with 10 mM D-U-
13C-glucose or 4 mM L-U-
13C-glutamine or 10 mM 2-
13C-acetate (Cambridge Isotope Laboratories, Andover, MA, USA) addition for 1 h in short-time labelling experiments before the extraction [
32]. In 24-h labelling experiments – avoiding long-term starvation/non-physiological conditions – cells were seeded in DMEM D5030 medium supplemented with FBS, and the combination of labelled and/or unlabelled glucose (10 mM) and glutamine (4 mM) or acetate (10 mM) were added to cell cultures for 24 h.
SSADH expression analysis by immunohistochemistry (tissue microarray) in clinical samples
47 glioma patients (type: astrocytoma
n = 14, oligodendroglioma
n = 14, glioblastoma
n = 19; sex: female
n = 23, male = 24; WHO grade: II
n = 9, III
n = 19, IV
n = 19; IDH1 R132H mutation: positive
n = 32, negative
n = 15) were included in a tissue microarray (TMA) study. Three normal brain and renal tissues were included as controls. Expression studies by IHC on biopsy materials were approved by the Institutional Ethical Review Board (TUKEB no. 7/2006). TMA sections were deparaffinised, endogenous peroxidase blocking was followed by antigen retrieval (sodium citrate; pH 6). Immunostaining was performed by routine diagnostic antibodies (such as anti-IDH1 R132H), SSADH (Abcam #129017 1:500) antibody and Novolink Polymer Detection System (Novocastra, Wetzlar, Germany), visualised by DAB and counterstained with haematoxylin. H-scores (scale of 0–300) were determined by two independent pathologists with semi quantitative analysis of immunoreactivity using 3DHistech Pannoramic Viewer program according to Krencz et al. [
35]. The H-score was calculated from staining intensity (0, 1+, 2+, or 3+) and positively stained tumour cells (percentage). The final H-score was calculated by averaging the H-scores of all the scores from the same tissue. SSADH expression was categorised as low (H-score 0–149) and high (H-score 150–300).
Statistical analysis
The data are presented as mean with standard deviation and calculated from three independent experiments with minimum three or more parallels depending on the method used. Results were statistically evaluated through one-way ANOVA with post-hoc Tukey’s and Dunnett’s tests for multiple comparisons by IBM SPSS Statistics software, version 22 (SPSS Inc.). Mann-Whitney U test and Kruskal-Wallis test were used to compare SSADH expressions and clinicopathologic parameters in human gliomas using IBM SPSS Statistics software, version 22 (SPSS Inc.). p < 0.05 was considered statistically significant.
Discussion
Our main aim was to characterise the bioenergetic differences related to IDH mutations by using an isogenic cell line pair with and without the IDH1 R132H mutation.
The effects of epigenetic and other (e.g. metabolic) alterations caused by IDH1 mutation have been described previously [
38‐
40]. However, few data are available about the bioenergetic consequences of IDH1 mutation and their role in growth and survival of tumour cells [
25,
41].
Based on our Seahorse measurements (oxygen consumption and extracellular acidification), higher oxygen consumption is strongly correlated to mutant IDH1-produced 2-HG. In addition, after using 2-HG treatment in wild-type cell cultures, the respiration was increased to similar level as in IDH1-mutant cells. Alterations of intra- and extracellular 2-HG besides Krebs cycle metabolite levels were observed. These observations suggest the presence of metabolic compensatory mechanisms in these cells. Our results confirmed, that the Warburg phenotype is dominant in IDH1 wild-type cells, whereas IDH1-mutant cells prefer oxidative phosphorylation using substrates other than glucose. Khurshed et al. published that glycolytic enzyme mRNA expression levels are higher in IDH wild-type tumour tissues and the expression levels of TCA related enzymes are higher in IDH1-mutant human cases [
20]. It was also described that mutant IDH1 enzyme was correlated with high mitochondrial density and increased mitochondrial activity in oligodendroglioma cell line xenografts in vivo [
41].
In our study, the significantly reduced glutamate level in IDH1-mutant cells could correlate to enhanced glutaminolytic pathway (glutamine-glutamate-aKG-2-HG axis). This was not observed in U251 wt cells after 2-HG incubation, suggesting that glutamatolysis in IDH1-mutant cells is related to the necessity to replenish aKG levels, rather than by 2-HG induced effects [
16]. Previous spectrometric analyses have shown and highlighted that the most important source of 2-HG is glutamine, and only low labelling could derive from glucose due to limited glucose turnover [
26]. Glutamine was the main source of 2-HG in our experiments using
13C-labelled substrates in 1-h-and 24-h-labelling periods which is in line with previous report [
23]. Glucose and acetate also contributed to 2-HG-labelling in 24-h
13C-labelling experiments. This potential contribution of acetate substrate in significant 2-HG
13C-labelling after 24 h was detected first in our experiments. Previous TCGA-based analyses suggest that glucose and acetate are preferential substrates in IDH wild-type, rather than IDH-mutant gliomas [
20]; therefore, the relative contribution of glucose and acetate in clinical IDH-mutant gliomas needs further investigations.
It has been shown in in silico studies that IDH1-mutant gliomas have significantly higher lactate dehydrogenase B expression comparing to wild-type tumours; and higher lactate oxidation was hypothesised in IDH1 mutation bearing cells [
20]. In our observations, we could not detect significant differences in lactate and acetate oxidation between wild-type and IDH1-mutant glioma cells. However, our results show that glutamine, glutamate and GABA energy substrates could significantly increase oxygen consumption rate in IDH1 wild-type U251 glioma cells (approximately 20% increase in OCR), but not in their IDH1-mutant counterpart cells. These suggest that GABA and glutamine can have an important role in energy production through substrate oxidation preferentially in IDH1 wild-type gliomas.
Dependency on glutamine and/or glutamate of IDH-mutant gliomas has been reported previously [
21,
23]. Our current study has demonstrated for the first time, that GABA oxidation (using GABA as an energy substrate) via the activity of SSADH can produce energy, explaining our finding the GABA stimulated proliferation of U251 wild-type cells. However, emerging data about the exact mechanism of GABA receptor mediated pro-proliferative effects are still unclear and contradictory [
42,
43]. The detected pro-proliferative effect was not the result of GABA-receptor stimulation [
44] as U251 cells lack expression of these. Based on our results, 2-HG could reverse the effect of GABA on proliferation. In line with the absence of SSADH expression in U87 MG and U373 MG, these cell lines were not able to oxidise GABA. Recent publications highlight the importance of GABA shunt in primer and metastatic brain tumours. The heterogeneity in cell lines with respect to SSADH expression and GABA oxidation warrants further investigation towards GABA use in human glioma cases and in other tumours which may rely on the effect of GABA use [
28,
29,
45,
46]. Our IHC studies reveal that both IDH-mutant and wild-type gliomas express high levels of SSADH in contrast to normal brain, suggesting certain role for GABA in growth and survival in clinical tumours. This correlates to data of Human Protein Atlas, where 90% of studied human gliomas showed moderate or high SSADH expression which indicates a significantly higher expression than normal brain tissue.
Related to these El-Habr and his co-workers have demonstrated that the accumulation of gamma-hydroxybutyrate (by-product of GABA in central nervous system) and the related SSADH downregulation contribute to a less aggressive phenotype in glioblastoma cases. They found higher SSADH protein expression in glioma cells than in normal brain tissue. They also described that downregulation, lowered expression and inhibited function of SSADH (shRNA silencing and gamma-hydroxybutyrate – structurally shows similarities to aKG and 2-HG – treatment) correlated to lower proliferation capacity of glioma cells in vitro and in vivo [
47]. Regarding to high-grade gliomas which showed lower SSADH expression at mRNA level in TCGA database, in our study (47 cases were analysed), we could not confirm this difference at protein level. Based on our and others’ results, SSADH protein level is higher in almost all glioma cells than in normal tissues. These suggest that SSADH protein overexpression at tissue level could have special tumour cell survival and growth promoting effect in both IDH-mutant and wild-type cases. To clarify the role of GABA metabolism, SSADH expression and functions need further studies using both mRNA and protein expression studies with patient follow-up. In these studies, overall survival data evaluations and if it is possible some SSADH function related analyses should be performed.
Our findings suggest that 2-HG may contribute to a less aggressive phenotype through its inhibitory effect on GABA oxidation. The detected SSADH overexpression in human cases could support GABA metabolism through either GABA oxidation or alternative GABA consumptions. Regarding to our and others’ results, GABA metabolism of gliomas might be a possible novel entry point for therapy, especially in glioma patients with poor prognosis and limited treating opportunities.
Acknowledgements
We gratefully acknowledge to the clinicians (neurosurgeons, pathologists and oncologists) for their helpful work in the diagnosis and patients care and to Attila Patócs for the accessibility of LC-MS in his laboratory. The authors thank Géza Sáfrány who kindly provided U373 MG cell line in some experiments and László Kopper for his careful reading, helpful advice and discussion during the manuscript preparation. We are grateful to Gergely Sváb for his help in final Seahorse measurements and to Éva Mátrainé Balogh, Anna Tamási and Mónika Szilágyiné Paulusz for their technical assistance in histological processing of human biopsies.