Overall, we found that plasma metabolomics is appropriate for estimating metabolic processes in the marrow or CSF of children with B-ALL. Specifically, the majority of compounds detected in marrow or CSF were also detected in plasma, and, especially for marrow, the number and type of compounds was similar. Substantial proportions (62% in marrow and 41% in CSF) demonstrated moderate or strong correlations with plasma, and many (35% in marrow, 16% in CSF) remained statistically significant at q < 0.05 after multiple testing correction. Of note, we reported strong correlations for compounds previously associated with cancer-related fatigue18 and MRD16, supporting that blood may be useful for biomarker studies of these compounds and endpoints. These findings have practical implications for future metabolomics studies, since blood can be collected more frequently and less invasively.
When evaluating plasma and bone marrow, we observed a robust correlation for asparagine. Asparaginase has long been utilized in ALL chemotherapy, following the discovery of its anti-leukemic effect, mediated by serum asparagine and glutamine depletion, and clinical trials demonstrating improved survival for asparaginase-treated patients31,32. In contrast to a previous magnetic resonance spectroscopy (MRS) and gas chromatography-mass spectrometry (GC-MS)-based untargeted metabolomics study, which reported total depletion of asparagine at D29 in patients who received PEG-asparaginase between induction D4 and D617, we detected asparagine in plasma and marrow samples from all children. Patients on both induction protocols (AALL0932 and AALL1131) received PEG-asparaginase on induction D412,33; our observation is consistent that of Angiolillo et al., who demonstrated that asparagine levels begin to recover approximately 20–25 days after PEG-asparginase administration34. Given the importance of asparagine depletion in ALL chemotherapy, it may be noteworthy that it was readily detected in plasma using this approach, and that plasma and marrow asparagine abundances correlated strongly.
We also observed a strong positive correlation between marrow and plasma pyruvate abundances. In a study of children with newly diagnosed ALL, we reported that bone marrow pyruvate abundance at diagnosis was associated with subsequent MRD (1.9-fold increase among MRD-positive patients, p = 0.02)16. Pyruvate is a key intermediate in glycolysis and gluconeogenesis. Altered glucose metabolism has been described in ALL cells35,36, and we and others have demonstrated that inhibitors of glycolysis exert anti-leukemic effects in vitro16,35. Whether plasma pyruvate abundances similarly associate with treatment response is unclear, but is a promising area for future research that aims to leverage metabolomics to understand ALL outcomes.
In our analysis of plasma and CSF, we observed a strong, albeit inverse, correlation between plasma and CSF abundances of gamma-glutamylglutamine. We previously reported an inverse (cross-sectional) association of CSF gamma-glutamylglutamine abundance and fatigue scores during post-induction chemotherapy, and reported that its abundance at the time of diagnosis was inversely correlated with fatigue severity at the start of delayed intensification18. Gamma-glutamylglutamine is an intermediate in the gamma-glutamyl cycle, in which gamma-glutamyl amino acids are formed by the transfer of glutamyl moieties (e.g., from glutathione to glutamate), then subsequently cleaved to the free amino acid and 5-oxoproline. This pathway may play a role in the regulation of amino acid transport across the blood-brain barrier37, which could explain the observation that plasma and CSF gamma-glutamylglutamine abundances were inversely correlated.
Saito et al. investigated changes in the plasma metabolome of patients with ALL pre- and post-induction, and observed that > 20% of compounds, most notably lipids, were altered38. The authors hypothesized that these alterations may affect the risk of adverse events in children with ALL, as previous studies suggest effects of docosahexaenoic acid (DHA)39–41 and phosphatidylethanolamines42 on asparaginase-associated pancreatitis and relapse. We found that lipids demonstrated somewhat stronger plasma-marrow correlations than other compounds.
Tiziani et al. compared plasma and bone marrow samples from N = 10 children with ALL at diagnosis and reported differences, primarily among several amino acids and ketones17. Collectively, these findings suggest that the plasma and marrow lipids are well correlated, but that special care may be required when comparing diagnostic samples, due to the high titer of leukemic cells. These findings may have particular implications for investigators wishing to perform longitudinal assessments or utilize lipidomics approaches.
Our study should be interpreted in light of certain strengths and limitations. As it was exploratory, the sample size was small, which limited our ability to detect statistically significant correlations. It was also cross-sectional, with all samples collected at the end of induction chemotherapy. On the other hand, we performed metabolomic profiling using a well-described untargeted platform, with broad coverage and an extensive reference panel, which identified > 1,000 unique features. Because this platform is semi-quantitative, we present data on relative abundances rather than absolute concentrations. In future studies, targeted approaches may allow for improved quantitation of metabolite or drug concentrations. Finally, the study sample was relatively homogenous, consisting entirely of pediatric patients with newly diagnosed B-ALL, who were treated on standard protocols. This likely reduced heterogeneity in our analysis, but we caution that it is unknown whether our findings may be applicable to T-ALL.