Gliomas are the most common form of central nervous system tumor. The most prevalent form, glioblastoma multiforme, is also the most deadly with mean survival times that are less than 15 months. Therapies are severely limited by the ability of these tumors to develop resistance to both radiation and chemotherapy. Thus, new tools are needed to identify and monitor chemoresistance before and after the initiation of therapy and to maximize the initial treatment plan by identifying patterns of chemoresistance prior to the start of therapy. Here we show how magnetic resonance imaging, particularly sodium imaging, metabolomics, and genomics have all emerged as potential approaches toward the identification of biomarkers of chemoresistance. This work also illustrates how use of these tools together represents a particularly promising approach to understanding mechanisms of chemoresistance and the development individualized treatment strategies for patients.
Accounting for more than approximately 80% of all primary central nervous system malignancies, gliomas are the most common form of central nervous system tumor. They include a wide variety of brain tumors including oligodendrogliomas, ependymomas, and astrocytic tumors such as astrocytoma, anaplastic astrocytoma, and glioblastoma, as well as tumors of mixed population[
While there are a number of new therapies currently in clinical trial[
Unfortunately, even when surgery, radiation, and chemotherapy can be used, high grade tumors inevitably return. The remaining cells develop genetic and metabolic adaptations that result in resistance to radiation and chemotherapy, enabling them to evade further treatment[
Given the problems that both intrinsic and acquired resistance pose for treatment and survival, the need for reliable biomarkers of resistance is clear. Current work is actively examining the sensitivity and specificity of genomic, metabolic, and imaging biomarkers to identify and monitor the response to treatment and the development of chemoresistance leading to individualized treatment plans that maximize progression-free survival.
Currently, the three major types of biomarkers under investigation for detection and monitoring of chemoresistance in glioma are imaging biomarkers, acquired by magnetic resonance imaging (MRI), metabolomics to detect changes in cellular metabolism that confer resistance, and genomics to identify changes in gene expression that lead to molecular alterations that permit cancer cells to evade treatment. This article reviews the current progress in using these three approaches towards the goals of detecting and monitoring chemoresistance, understanding the mechanisms associated with chemoresistance, and using this information to design new, individualized therapies for the treatment of glioma.
MRI is an important tool for diagnosis of glioma as well as determination of the site and size of the tumor before and after surgical resection and treatment. Use of T1- and T2-weighted sequences and contrast-enhanced T1-weighted images is currently regarded as the gold standard[
Work is also being done to refine the use of magnetic resonance spectroscopy[
Advances in sodium magnetic resonance imaging (Na-MRI) have also made it an attractive biomarker of treatment response. The rationale for the use of Na-MRI is based on the knowledge that intracellular sodium concentrations are around 15 mM, while extracellular concentrations are approximately 140 mM[
Given the apparent power of Na-MRI for monitoring the response to treatment, it is not surprising that several clinical studies have now tested the possible use of Na-MRI at 3 T in patient populations with glioma. Together these studies have shown elevated total sodium levels in humans with glioma[
Despite these advances in MRI-based biomarkers for the response to therapy, there is still a critical need for non-invasive imaging tools for detecting and monitoring the development of resistance to treatment. Initial reports showed that diffusion MRI was able to detect the loss of therapeutic response to BCNU in rodent 9L glioma cells implanted intracranially, suggesting that this method, which measures water mobility, can detect the development of chemoresistance[
The next advance resulted from the finding that Na-MRI was even more effective than diffusion at detecting the development of resistance following an initial round of therapy[
Towards this goal, preclinical studies using ultra-high field MRI (21.1 T) have now shown that Na-MRI has the ability to detect chemoresistant tumors before the start of treatment (US Patent 8,880,146). An ultra-short echo time of 0.14 ms and high resolution 3D Na-MRI was able to detect differences between intracranial tumors that were generated from rodent glioma cell lines (9L gliosarcoma) with different levels of resistance to BCNU[
The finding that the sodium signal is altered by chemoresistance raises questions about the role of sodium in glioma cell metabolism and mitochondrial function. It has long been recognized that cancer cells have altered metabolic profiles compared to normal cells[
The degree to which the Warburg effect participates in the development of chemoresistance in cancers such as glioma is controversial[
One limitation of the work on chemoresistance in glioma is that much of this work has been conducted in TMZ–resistant cells. The contradictory data about the role of Warburg in chemoresistance, and the paucity of data in glioma cells resistant to chemotherapeutics other than TMZ led us to use metabolomics to evaluate metabolic changes in BCNU-resistant glioma cells and test the hypothesis that acquired BCNU resistance involves modifications to energy metabolism that enhance the Warburg effect. To test this hypothesis, we first generated a line of rat 9L glioma cells that are resistant to BCNU by repetitive dosing with increasing levels of the alkylating agent BCNU (10-100 µM). Maintenance of cultured 9L rat glioma cells with BCNU resulted in the selection of resistant cells.
Development of BCNU-resistant 9L glioma cells. BCNU sensitive 9L glioma cells (9L-S) were grown in the presence of increasing concentrations of BCNU resulting in a 9L subculture that was BCNU-resistant (9L-R). Cell viability of 9L-S and 9L-R cells was quantified after exposure of both cell types to increasing concentrations of BCNU and expressed as percent survival (mean ± SD,
Treatment with dichloroacetate (DCA) inhibits the enzyme pyruvate dehydrogenase kinase, activates pyruvate dehydrogenase, and increases oxidative phosphorylation. It has been used in combination with other drugs to relieve the Warburg effect in glioma and induce cell death[
Dichloroacetate (DCA) improves the responsiveness of resistant glioma cells to BCNU. Treatment of BCNU-sensitive (9L-S) and BCNU-resistant (9L-R) glioma cells with (A) 75 µM or (B) 100 µM BCNU resulted in significant death of 9L-S, but not 9L-R cells. Addition of 25 mM DCA potentiated the action of BCNU in 9L-R cells at both concentrations of BCNU. Bars (mean ± SD,
We then followed the metabolism of 13C-labeled glucose using gas chromatography-mass spectrometry (GC-MS) of cell extracts to compare the metabolic profile of BCNU-sensitive (9L-S) and BCNU-resistant (9L-R) cells. It has been well-established that the specific labeling pattern of key metabolites (called the isotopomer pattern) can be used to determine the relative activities of metabolic pathways contributing to the synthesis of those metabolites after correction for incorporation of the natural abundance isotope. The goal of this work was to establish a metabolomics approach that would permit the identification of key components of glucose metabolism that are altered in BCNU-resistant cells and provide additional insights into the mechanisms associated with acquired chemoresistance. This goal was supported by comparison of metabolic profiles in the presence or absence of DCA.
In brief, this metabolomic analysis was accomplished by first growing cells in the presence or absence of 75 or 100 µM BCNU and/or 25 mM DCA followed by a 24 h incubation with 13C-labeled glucose that were quenched with degassed, ice-cold 50% aqueous acetonitrile with 0.17 mg/mL norleucine, dried under vaccum overnight, redissolved in anhydrous pyridine, and derivatized by the addition of 20 µL N-methyl-N-(tert-butyldimethylsilyl)trifluoro-acetamide (TBDMS) containing 1% tert-butyldimethylchlorosilane (Thermo Scientific, Rockford, IL). Derivatized samples were injected in splitless mode into an HP Agilent 6890 series gas chromatograph coupled with an HP Agilent 5973 mass selective detector and separated on DB5-MS 30 cm × 250 µm × 0.25 µm column with a 10-cm guard column (J&W Scientific, Folsom, CA). Metabolites were identified by retention time and mass spectrum comparison with standards. The prevalence of particular mass isotopomers (species having the same isotopic mass) for a given compound provides information about the incorporation of labeled media components into metabolites. A mass isotopomer distribution vector (MID) was calculated for the molecular and fragment ions for assigned observed metabolites using in-house scripts written in the programming language[
Comparison of the M3:M2 ratio of TCA cycle intermediates permited the evaluation of cellular metabolism including metabolism of pyruvate via anaplerotic conversion to oxaloacetate or malate versus the pyruvate dehyrogenase/citrate synthase pathway (PDH/CS)[
Effect of chemoresistance on aspartate isotopomer ratios. A: Model depicting the pathways for aspartate M2 and M3 isotopomer production from U-13C glucose; B: Aspartate M3:M2 isotopomer ratios in BCNU-sensitive (9L-S) and BCNU-resistant (9L-R) glioma cells in the absence and presence of 40 mM DCA. Bars (mean ± SD,
The chemoresistance-induced reorganization of pyruvate metabolism and the corresponding effect of DCA treatment was also evident from the isotopomer patterns observed in other TCA cycle intermediates. For instance, the M3:M2 ratio for fumarate was significantly higher in the resistant cells than in the 9L-S cells, and treatment with DCA nearly equalized this ratio
Effect of chemoresistance on fumarate and succinate isotopomer ratios. Isotopomer ratios for (A) fumarate and (B) succinate in BCNU-sensitive (9L-S) and BCNU-resistant (9L-R) glioma cells in the absence and presence of 40 mM DCA. Bars (mean ± SD) with different letters are significantly different from each other at
These data show that pyruvate was primarily metabolized by the TCA cycle through pyruvate dehydrogenase and citrate synthase in 9L-S cells, but that anaplerotic reactions of pyruvate carboxylase and/or malic enzyme provided the primary route in the 9L-R cells. Our method was unable to determine whether the higher M3:M2 ratio observed for aspartate, malate, and fumarate in the 9L-R cells resulted from an actual increase in pyruvate carboxylase activity, from a decrease in PDH activity, or from a combination. Regardless, the increased M3:M2 ratio observed for these compounds in 9L-R cells shows that acquired chemoresistance increases the relative activity of pyruvate metabolized through the anaplerotic reactions and decreases the dependence of the PDH pathway. The reduced activity of pyruvate dehydrogenase is typically seen in cancer cells and our data suggest that acquisition of BCNU resistance exacerbates this effect by creating a more Warburg-like state compared to drug-sensitive glioma cells.
Further support for this hypothesis was obtained by treating 9L-S and 9L-R cells with DCA. As mentioned above, DCA is known to activate pyruvate dehydrogenase, reversing tumor-specific modifications in mitochondrial physiology, both
The development of BCNU-resistance also had a significant effect on the isotopomer patterns of serine and glycine. The M3 and M2 isotopomers of serine reflect the relative amount synthesized from 3-phosphoglycerate (the M3 isotopomer) versus that synthesized from glycine (the M2 isotopomer) [
Effect of chemoresistance on serine isotopomer ratios. A: Model depicting the pathway for serine M2 and M3 isotopomer production from U-13C glucose; B: Serine M3:M2 isotopomer ratios in BCNU-sensitive (9L-S) and BCNU-resistant (9L-R) glioma cells in the absence and presence of 40 mM DCA. Bars (mean ± SD) with different letters are significantly different from each other at
Effect of chemoresistance on glycine isotopomer ratios. A: Model depicting the pathway for glycine M1 and M2 isotopomer production from U-13C glucose; B: Glycine M1:M2 isotopomer ratios in BCNU-sensitive (9L-S) and BCNU-resistant (9L-R) glioma cells in the absence and presence of 40 mM DCA. Bars (mean ± SD) with different letters are significantly different from each other at
As shown in
In summary, while clearly not all resistance is the result of the Warburg effect, much of the data in the literature support the hypothesis that an enhancement of the Warburg effect is associated with acquired chemoresistance. Furthermore, our work with BCNU-resistant cells line reported here adds to the existing literature on TMZ-resistance by identifying several metabolic changes associated with acquired resistance in glioma cells. More importantly, however, this work illustrates the strength of a metabolomic approach for potentially phenotyping glioma, evaluating resistance, and identifying metabolic pathways that may serve as future targets for overcoming chemoresistance and developing novel treatments for chemoresistant tumors.
A large number of studies have established the association between wide-spread genomic changes and glioma tumorigenesis. Evaluation of gliomas at the genomic level is crucial because histologically identical tumors have been shown to have very different genomic alterations[
Like metabolomic approaches, most of the work using genomic approaches to determine the mechanisms responsible for chemoresistance has been conducted in TMZ-resistant cells by comparing them to TMZ-sensitive cells. These cells are characterized by chromosomal instability, alterations in MDM2, ERK, AKT, STAT3, ABC transporters, and PIK3 pathway members[
Hypoxia-related genes have also been implicated in the metabolic changes associated with chemoresistance and prognosis[
The alkylating agent BCNU has been infused into biodegradable wafers for implantation after surgical resection. The clinical outcomes of this approach have been mixed[
Because of the paucity of genomic data on chemoresistance to BCNU, we used rat 9L cells and microarray analysis to compare the gene expression profiles of BCNU-sensitive (9L-S) and BCNU-resistant cells and identify significant changes in gene expression associated with BCNU-resistance. Resistant cells were produced by continuous culture in increasing concentrations of BCNU as previously described producing two lines of resistant cells, 9L-R1 treated with 150 µM BCNU and 9L-R2 treated with 225 µM BCNU in culture. Additionally, 1 × 105 9L-S cells were implanted to produce an intracranial tumor in adult Sprague Dawley rats. After 11 days of tumor growth and confirmation by MRI, rats were treated with 26.6 mg BCNU/kg body weight. Following tumor regrowth, tumors were excised and cultured to propagate an additional line of resistant cells (9L-R3). Total cellular RNA was collected from all four cell lines and analyzed using a Roche/Nimblegen 12-plex microarray. Data analysis was completed with ArrayStar software (DNASTAR, Madison, WI). Expression patterns were normalized using the quantile normalization method and mRNA abundance were considered differentially regulated when there were changes in abundance of at least 1.5-fold with
Consistent with the literature on genomic changes in rapidly dividing cancer cells, each of the resistant 9L cell lines had its own distinct genomic pattern. Comparison of the resistant line to the parent 9L-S line identified 379 genes that were differentially regulated in all three of the resistant lines
BCNU-Resistance Genes
Gene | Percent of Expression in Drug Naïve 9L-S | ||||
---|---|---|---|---|---|
9L-R1 | 9L-R2 | 9L-R3 | mean ± SD | ||
CRABP1 | 0.97 | 0.97 | 2.25 | 1.40 ± 0.74 | 9.7 × 10-12 |
NGF receptor | 3.99 | 3.99 | 3.56 | 2.43 ± 1.51 | 1 × 10-8 |
NGF | 8.26 | 7.59 | 8.55 | 8.13 ± 0.49 | 2 × 10-9 |
Brain creatine kinase | 9.07 | 8.67 | 9.65 | 8.87 ± 0.28 | 2.7 × 10-10 |
VG Na channel, type 1 | 27.7 | 13.9 | 17.5 | 19.70 ± 7.16 | 1.5 × 10-6 |
Na/Ca exchanger 5 | 35.7 | 44.9 | 39.7 | 40.1 ± 4.61 | 2.9 × 10-6 |
FGF 9 | 423 | 818 | 275 | 505 ± 280 | 3.1 × 10-14 |
TGF-β2 | 508 | 849 | 772 | 710 ± 179 | 4.5 × 10-11 |
MGMT | 850 | 642 | 652 | 715 ± 117 | 1.1 × 10-8 |
Galectin 3 | 3055 | 311 | 3342 | 2236 ± 1673 | 5.1 × 10-10 |
Microarray analysis was used to identify genes regulated in chemoresistant lines (9L-R1, 9L-R2, and 9L-R3) compared to the parent drug naïve line of 9L cells. NGF: nerve growth factor; VG Na channel; voltage-gated sodium channel; Na/Ca exchanger; sodium/calcium exchanger; FGF: fibroblast growth factor; TGF: transforming growth factor; MGMT: O-6-methylguanine-DNA methyltransferase
Effect of BCNU exposure to 9L rat glioma cells. Cells were exposed to 150 µM (9L-R1) or 225 µM (9L-R2) in culture or
Another potentially important mechanism is the finding that cellular retinoic acid binding protein-1 (CRABP1) was the most intensely down-regulated mRNA in the three resistant lines. While this gene has not been previously reported to be regulated in glioma cells, CRABPI has been shown to play a role in the induction of apoptosis in a mouse ovarian epithelial cancer cell line treated with retinoic acid[
In conclusion, each of the three approaches to identifying biomarkers of chemoresistance come with specific strengths. Metabolomics and genomics have the potential to identify specific cellular and molecular mechanisms that are responsible for both intrinsic resistance as well as therapy-induced resistance. These mechanisms have the potential to serve as novel targets for reducing chemoresistance, but require surgical intervention or biopsy to obtain tissue. Imaging biomarkers, particularly those utilizing Na-MRI, have the advantage of being rapid, non-invasive, and easily employed before and after treatment. Ultimately, significant advances in treatment are most likely to come from a combination of these approaches. For example, non-invasive imaging-based biomarkers that can be correlated to changes in biochemical mechanisms would enable MRI to be used to collect cellular and molecular information non-invasively in patients with glioma. Thus, used together these powerful tools could inform treatment decisions, facilitate individualized treatment plans, and improve survival time in patients with gliomas.
The authors thank Shannon Gower-Winter, MS for technical support and expert scientific editing of the manuscript. We also thank Mr. Charles Badland for invaluable support producing figures.
Conception and design of the study: Levenson CW, Morgan TJ, Twigg PD, Logan TM and Schepkin VD
Performed data analysis and interpretation: Levenson CW, Morgan TJ, Twigg PD, Logan TM, and Schepkin VD
Not applicable.
This work was supported by a Multidisciplinary Grant (to CWL) from the Florida State University Committee on Research and Creativity; the National High Magnetic Field Laboratory (Tallahassee) supported by NSF, grant (No. DMR-115490).
The author declares that there are no conflicts of interest.
All animal work has been approved by the Institutional Animal Care and Use Committee in accordance to AALAS and NIH guidelines.
Not applicable.
© The Author(s) 2019.