In this study, we investigated the comprehensive untargeted metabolomic, lipidomic, and biogenic amine profiles of GBM tissue and patient plasma specimens at diagnosis and at recurrence. Despite a small overall cohort size, our result showed that many metabolites were altered in GBM tissue and patient plasma at recurrence when compared to diagnosis. Our study demonstrated the feasibility of studying GBM tissue and patient blood specimen longitudinally using metabolomic methodology.
GBM display marked metabolic heterogeneity in their microenvironments.[17] Both glucose and lipid metabolisms are abnormally regulated in GBM tissues.[18, 19] In our study, we observed several metabolites that had changed in abundance at recurrence when compared to diagnosis. Many of these metabolites were also identified in a recently published study on the metabolic hallmarks of gliomas.[20] Specifically, we identified 2-methylbutyryl-L-carnitine and ecgonine that were known to reflect tumor metabolic flexibility in brain tumor tissues at diagnosis and at recurrence. Carnitine serves as a “shuttle-molecule” that allows fatty acid acyl moieties to enter the mitochondrial matrix for oxidization via the beta-oxidation pathway[21]. We found that the 2-methylbutyryl-L-carnitine level was significantly reduced in recurrent tumors compared to initial GBM tissue. Carnitine transporter modulation has been thought to be a potential target for cancer treatment.[21] In addition, we also found many altered levels of lipids in GBM tissue at recurrence when compared to initial diagnosis. Our findings are in line with previously published data suggesting lipid metabolic alterations in GBM.[22] In addition, mannitol was upregulated in recurrent tissue compared to the original tumor, suggesting BBB permeability changes after surgical resection and chemoradiation. This may suggest mannitol as a vehicle to guide targeted treatment.
The list of metabolites with significantly altered abundance and fold changes differ when comparing the unpaired samples (9 tissue samples at diagnosis vs 3 tissue samples at recurrence) and the paired samples (3 tissue samples from the same patients at diagnosis and at recurrence). Also, despite a small sample size, the paired tissues samples at diagnosis and at recurrence demonstrate a clearer trend in the differences in compound abundance. Therefore, paired samples are recommended in future studies given their better capacity as an internal control.
In both the PCA plot from patient plasma and the heatmap generated from the top 50 altered metabolites, we found that the metabolomic profiles differed between specimens at diagnosis and at recurrence, except for one patient with early recurrence. The metabolomic profile of this patient’s plasma at diagnosis was different from the rest of the plasma specimens at diagnosis and was similar to the group pattern at recurrence. This interesting finding needs to be validated in a larger cohort of patients with treatment refractory tumors and early recurrence. This metabolomic signature may indicate a high risk and poor prognosis.
In the plasma cohort, we found several significantly altered metabolites with large fold changes. For example, 2,4-difluorotoluene increased in patient plasma at recurrence; this metabolite is incorporated into DNA and undergoes replication by DNA polymerase enzymes.[23] The change suggests rapid growth of recurrent tumors. Diatrizoic acid also increased at recurrence. However, this is a contrast agent used during imaging. Indole-3-acetate is an indol-3-yl carboxylic acid anion and has a role as a human metabolite. We again found several other compounds involved in glucose and lipid metabolism. The overall pattern changes of these compounds need to be validated in targeted metabolomics for their potential candidacy as biomarkers for treatment response and tumor recurrence. The enrichment of these metabolites in recurrent GBM tissue may suggest that targeting metabolic activity can be a potential adjuvant targeted treatment for GBM patients.
Interestingly, when plotting all four groups in the same PCA plot, the brain tissue and plasma specimens separated distinctly from each other regardless of disease status (Fig. 4). There was no overlap between specimen types and, the intra-specimen comparison at diagnosis and at recurrence became smaller. This suggested that plasma metabolite patterns are not reflective of brain tumor tissue, and therefore, two sets of biomarker panels are necessary for tissue and plasma. The pattern of tumor microenvironment can be further altered through end organ metabolism in the plasma. Also, plasma is affected by systemically administered medication for peri-operative care.
Our study has limitations. First, we must acknowledge that the sample size in this pilot study is small. Second, although we have paired data within the brain tissue and plasma cohorts at diagnosis and at recurrence, we were unable to identify if the tissue and plasma specimens were from the same patients due to the restrictions of the UC Davis biorepository consents for GBM tissue. In addition, we did not have normal brain tissue or plasma for comparison. These limitations prevented us from identifying small changes in metabolite abundance. Therefore, only significant, and large fold changes were analyzed. A bigger cohort with paired specimens will be emphasized in future studies. The current study paved the way for the next targeted metabolomics, lipidomic, and biogenic amine studies validating and further investigating these profiles.