†These authors contributed equally.
Academic Editor: Graham Pawelec
Background: The conserved stem cell signaling network canonical
Wingless (WNT) plays important roles in development and disease. Aberrant
activation of this pathway has been linked to tumor progression and resistance to
therapy. Industry and academia have substantially invested in developing
substances, which can efficiently and specifically block the WNT signaling
pathway. However, a clear clinical proof of the efficacy of this approach is
still missing. Studies on the metabolomics dysregulation of cancer cells have led
to innovations in oncological diagnostics. In addition, modulation of cancer cell
metabolome is at the base of promising clinical oncology trials currently
underway. While onco-protein activation can have profound metabolic outcomes, the
involvement of stem cell signals, such as the WNT pathway, in tumor cell
metabolomics is yet insufficiently characterized. Material and methods:
We determined live cell metabolism and bioenergetics in pathophysiological
relevant, WNT-dependent glioblastoma stem cell (GSC) models. We quantified those
parameters in cells with canonical WNT activity and in isogenic cells where WNT
activity had been inhibited by short hairpin RNA against
Accumulating evidence demonstrates that the Wingless (WNT) network—one of the phylogenetically most conserved developmental signaling pathways—acts as an oncogenic driver in cancers [1, 2]. Various WNT inhibitors are currently being tested in clinical trials [3]. However, not a single WNT inhibitor has been approved for cancer therapy thus far. Previous clinical studies showed that systemic administration of current WNT blockers resulted in intolerable side effects, usually due to targeting the WNT niche in the gastrointestinal tract [1]. The identification of a putative blocker of cancer-specific WNT signaling would be a promising strategy to overcome this hurdle.
The metabolic profile of cancer cells significantly differs from that of non-neoplastic cells. Some cancer cells are abnormally dependent on energy production through anaerobic oxidation and show atypical activation of glutaminolysis as well as fatty acid production [4]. It was suggested that therapeutic resistance is the consequence of these metabolic alterations, which represent a major obstacle in the treatment of cancer [5]. Several studies reported that high activity of glycolysis causes chemotherapy resistance in aggressive cancer cells [6]. Moreover, there is increasing evidence that aggressive cancer cells in various tissues including brain, can re-program their metabolism in response to altered intrinsic and environmental conditions [7, 8, 9, 10].
The impact of WNT signaling on the cellular energy status of tumor cells has
been investigated to some extent. However, limited knowledge is available
regarding its involvement in brain tumor bioenergetics. WNT signaling activates
the potent proto-oncogene MYC in a variety of cancers [11], which subsequently
promotes anaerobic respiration and glutaminolysis [12]. WNT was found to promote
glycolysis in breast [13] and colon [14] cancer. A prominent oncogenic driver,
also present in gliomas, is epidermal growth factor receptor (EGFR) signaling,
which induces translocation of the glycolysis mediator pyruvate kinase isoform M2
(PKM2) to the cell nucleus, where it directly binds to the key WNT transcription
factor
The present study used glioblastoma as a cancer stem cell model since this tumor
type is known to harbor a WNT-driven subpopulation of stem cells [24, 25].
Capitalizing on genetic engineering, gene expression and live cell bioenergetic
analyses, we aimed at identifying the effect of WNT signaling on GSC metabolism.
We found that GSC cell lines manifested variable overall ATP production rates to
which mitochondrial oxidative phosphorylation and glycolysis contributed equally.
Interestingly, suppression of WNT signaling by
The collection of glioma samples was approved by the Beijing Tiantan Hospital institutional review board (IRB) and written informed consent forms were obtained from every patient. The site of tumor sample collection was planned in the neuronavigation software before surgery. Following their surgical removal, tumor tissues were immediately immersed in culture medium and transported to the laboratory within 30 min. Glioma cells, obtained from these primary tumors according to Zhai et al. [27], were cultured in DMEM/F12 serum-free medium (Gibco) supplemented with B27 (Gibco), basic fibroblast growth factor (BFGF, 20 ng/mL), epidermal growth factor (EGF, 20 ng/mL), and heparin (2.5 mg/mL). Cells were regularly passaged according to their growth status. GSC gene expression array data was downloaded from GSE67089 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE67089).
Total RNA was extracted using the mRNA Isolation Kit (Ambion, TX, USA) following
manufacturer instructions. RNA integrity was evaluated by Bioanalyzer and samples
with RNA Integrity Number
The gene list of biological processes was obtained from the GSEA Web portal (http://software.broadinstitute.org/gsea/msigdb/index.jsp). Based on the transcriptome sequencing data, the biological enrichment score of each tumor sample was obtained through the GSVA algorithm. The GSVA algorithm was performed by R under the default parameters.
The correlation between biological function enrichment scores were evaluated by the Pearson correlation analysis. Correlation plots were drawn by the ggplot2 package (RStudio Inc, MA, USA) in R language.
GBM1, 233, 644 and JHH520 cells were cultured in DMEM/F12 (Gibco), supplemented
with B27 (Gibco), basic fibroblast growth factor (Peprotech BFGF, 20 ng/mL),
epidermal growth factor (Peprotech EGF, 20 ng/mL), and heparin sodium salt (Sigma
2.5 mg/mL). Cells were tested for mycoplasma before each experiment. Incubated at
37
HEK 293T cells were cultured in 10 cm dishes with DMEM (Gibco) supplemented with
10% FBS (Sigma-Aldrich) and 1% Penicillin/Streptomycin (Sigma) until they
reached 80% confluence. Subsequently, the medium was replaced with antibiotic
free DMEM/FBS. Cells where transfected with transfection mix (800
For transduction, approx. 5
To validate the model by target activation, we performed Western blot analysis.
Cell suspensions were centrifuged, and pellets where lysed with cell lysis buffer
(50 mM TRIS-HCl (pH 8),150 mM NaCl, 0.5% Triton X-100, 0.5% Deoxycholate) with
a general use protease inhibitor cocktail for 30 min on ice. The lysates were
centrifuged at 13000 rpm at 4
GBM1, 233 and JHH520 wildtype (wt), EV and sh
The box and whisker plots show the median and 5–95% confidence interval.
Significant differences among cell lines were determined by one-way ANOVA
followed by Sidak analysis for multiple comparisons. p-values of
We applied extracellular flux analysis to quantify ATP production rates in living cells. Fig. 1A shows that the JHH520 and 233 GSC lines displayed the highest ATP production rates followed by GBM1 and 644. The lowest rates of ATP production were observed in 407 and SF188 cells. The differences in total ATP production rates could be ascribed to changes in both mitochondrial and glycolytic ATP production (Fig. 1B,C). Yet, the two bioenergetic modes equally contributed to total ATP production, as demonstrated by the similarity of glycolytic to mitochondrial ATP ratios among the different cell lines (Fig. 1D).
Glioma stem cells reflect a wide range of total (A), as well as mitochondrial (B) and glycolytic (C) ATP production rates, as assessed by extracellular flux analyses on living cells. However, the range of ratios of glycolytic to mitochondrial ATP productions rates is relatively narrow. The two bioenergetic modes equally contributed to total ATP production, as demonstrated by the similarity of glycolytic to mitochondrial ATP ratios among the different cell lines (D).
To explore the impact of canonical WNT signaling on mitochondrial and glycolytic
function, we created stable cell lines modeling loss of function of WNT signaling
[30]. The efficiency of
Establishment of genetic WNT pathway inhibition models via
Due to technical issues, we used GAPDH as a loading control for cell line 233. GAPDH is not an ideal loading control as it is a glycolytic enzyme, which could be influenced by WNT signaling. However, the fact that glycolytic ATP production was decreased by CTNNB1 KD in the 233 cells, while GAPDH did not change (Supplementary Fig. 1), suggest that GAPDH was not modified by CTNNB1 KD and thus could serve as a proper loading control.
We next tested whether the functional regulation of ATP synthesis by canonical WNT signaling is also mirrored by correlations at gene expression level. This was done using an in silico analysis of datasets derived from GSCs. We assessed single cell RNA sequencing of GSCs retrieved from different datasets and found a significant positive correlation between the expression of ATP synthesis genes and genes belonging to canonical WNT activation. In contrast, the expression of ATP synthesis genes did not correlate with the non-canonical WNT pathway (Fig. 3).
Correlation analysis of gene expression in GSCs from two independent single cell RNA sequencing datasets (BNI-GSC and GSE67089). There was a strong positive correlation between the expression of ATP synthesis related genes and canonical (CTNNB1-dependent) WNT signaling genes. However, the expression of non-canonical WNT genes (CTNNB1 independent pathway) did not correlate with the expression of ATP synthesis-related genes. We also found a positive correlation between the expression of all WNT branches (WNT signaling pathway) and ATP synthesis genes.
Although cancer metabolism is a global and very active research field, metabolic characterization of physiologically relevant molecular subtypes of tumor stem cells, including GSCs, is insufficiently described [32]. Our approach, based on investigating major embryonic stem cell pathways and living cell populations, can potentially bridge this knowledge gap.
WNT-dependent GSCs manifest variable baseline ATP synthesis rates, reflecting
the ability of our disease modeling technology to recapitulate wide spectra of
the cellular and genetic tumor heterogeneity. Our results demonstrate that
inhibition of canonical WNT signaling by CTNNB1 knock-down can suppress the
mitochondrial and glycolytic activity in these cells. Notably, our previous
studies showed that the potential of pharmacological WNT inhibition (using the
Novartis drug candidate WNT974) is independent of the level of baseline WNT
activity, as assessed by target gene expression profiling (AXIN2) [30]. The
current work suggests a correlation between WNT baseline activity and metabolic
state. We previously used AXIN2 as a readout to score the signaling pathway
activity in our cells and found that GBM1 is a model with moderate WNT activity,
while JHH520 is a model with remarkably high WNT activity [20]. Interestingly,
both models showed similarly large drops in ATP synthesis rate upon inhibition of
canonical WNT signaling despite manifesting different ATP production rates. Our
data suggest that anti-WNT directed interventions may be effective independent of
baseline WNT activity. The analyzed cell models belonged to different
transcriptional subclasses of the tumor, with GBM1 being a classical subtype and
JHH520 a mesenchymal subtype [26]. Thus, our findings suggest that novel
therapeutic strategies directed against WNT members—or their downstream signal
mediators—may be able to overcome therapy resistance due to molecular
heterogeneity of glioblastoma (according to Verhaak et al. [32]). That
being said, the models also manifest some homogeneity as they are all wildtype
for IDH1 and present methylated MGMT promoter, two consensus markers in clinical
neuropathology [33]. From a translational point of view, targeting metabolic
pathways, such as glycolysis, to overcome chemotherapy resistance in
WNT-signaling-dependent tumors is very relevant, as pharmacological inhibitors of
glycolysis are already in an advanced stage of development [28]. Inhibitors of
glycolysis may therefore serve as an alternative, possibly less adverse,
therapeutic strategy to eradicate WNT-driven malignant behavior of glioma.
Moreover, as inhibition of glycolysis affects not only ATP production, but also
the availability of NAD
Surprisingly, our single cell gene expression sequencing data suggest that the WNT-ATP synthesis interaction in GSCs is preferentially mediated through the canonical, CTNNB1-dependent branch of WNT. While we acknowledge that our data is only correlative, thereby less robust as compared to functional data, our observation is in contrast to reports from others in other biological contexts. For instance, WNT5a-mediated activation of non-canonical WNT signaling caused increase in glycolytic activity in melanoma and in cortical neurons [36, 37]. Of note, a pioneering work in the field revealed the biological importance of WNT5a to promote GSC maintenance [38, 39], but no assessment of the effect of WNT5a on the metabolomics of these cells has been conducted yet.
We are aware of certain limitations of the study that have to be considered
carefully when interpreting the data: (1) The gene engineering technology used:
the lentiviral approach used in this study cannot exclude that the observed
metabolic phenotype is secondary to processes upstream to
Taken together, our data indicate that metabolic dependency of tumor stem cells are at least in part a result of the aberrant or constitutive activation of phylogenetically conserved canonical WNT signaling. Thus, targeting metabolic pathways is a promising strategy to overcome the resistance to therapy of molecular heterogeneous tumors, such as glioblastoma, as well as an attractive route to overcome the hurdles of clinical translation of hitherto developed anti-tumor stem cell drugs.
MO, MH and UK conceived and designed the experiments; MO and MH performed the experiments; OK, MO, GL, WZ analyzed the data; SM, HJS, EIS, CAD, RSC and DH contributed reagents and materials; UK was administrating the project, UK, MO, MH and OK write the original draft. All authors reviewed and edited the draft.
The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of the Medical Faculty of the Heinrich-Heine University (Study ID #5206). The collection of glioma samples was approved by the Beijing Tiantan Hospital institutional review board (IRB) and written informed consent forms were obtained from every patient.
The authors thank E. Petropoulou, AlfaLab, Greece for her support in the beginning of the study.
This study is a result of the funded by Brigitte & Dr. Konstanze-Wegener Foundation (Dusseldorf, Germany) granted to UDK (Project #26).
The authors declare no conflict of interest.