Abstract
Targeting metabolism has long been a theory for cancer therapy, but clinical development has been limited by toxicities, compound availability, overall efficacy, and patient specificity. CPI-613, a lipoic acid analogue that interferes with enzymes involved in mitochondrial metabolism, has demonstrated clinical activity in lethal malignancies including relapsed or therapy refractory Acute Myeloid Leukemias (AMLs). Using metabolomics, we investigated blood and bone marrow samples from a cohort of 29 relapsed or refractory AML patients involved in Phase I and II studies undergoing CPI-613 treatment, including 13 that achieved a complete response. We show that CPI-613 treatment in patients induced defined alterations related to the tricarboxylic acid (TCA) cycle and associated redox, anabolic, and catabolic metabolism. These findings are consistent with targeting of several ketoacid dehydrogenase (KADH) enzymes that use lipoic acid as a cofactor. The alterations were observed systemically but were more pronounced within the leukemic bone marrow microenvironment. Machine learning revealed that metabolic status and changes associated with mitochondrial metabolism were predictive of treatment response, indicating that mechanism-based metabolite biomarkers to a targeted metabolic cancer therapy may be feasible. Finally, we confirm using isotope tracing and flux analysis that these effects are due to disruptions to substrate utilization into the mitochondria.
Introduction
For decades, the metabolic reprogramming of cancer cells has been recognized as a hallmark of malignancy, yet translating this knowledge into effective therapies has proven difficult. The primary obstacle is the narrow therapeutic window: because the tricarboxylic acid (TCA) cycle is essential for healthy cellular function, inhibiting it often results in severe systemic toxicity. Furthermore, demonstrating "target engagement"—proof that a drug is successfully disrupting its intended metabolic pathway within the human tumor microenvironment—remains a significant challenge in clinical oncology.
This study evaluates CPI-613 (Devimistat), a lipoic acid analog designed to inhibit mitochondrial metabolism by targeting ketoacid dehydrogenase (KADH) enzymes. By analyzing samples from patients with relapsed or refractory Acute Myeloid Leukemia (AML), researchers sought to determine if CPI-613 effectively engages its targets in vivo and whether specific metabolic signatures can predict which patients will respond to treatment.
Methodology
The researchers utilized a multi-omic approach to bridge the gap between clinical observation and mechanistic understanding:
- Clinical Metabolomics: Plasma and bone marrow samples were collected from 29 AML patients across two clinical trials (NCT01768897, NCT02484391). Using Liquid Chromatography coupled to High-Resolution Mass Spectrometry (LC-HRMS), the team profiled hundreds of metabolites to track changes before and after treatment.
- Target Engagement Analysis: The study specifically monitored the accumulation of substrates for KADH enzymes, such as pyruvate and branched-chain amino acids, which rely on lipoic acid as a cofactor. Accumulation of these substrates serves as a direct indicator of enzyme inhibition by CPI-613.
- Machine Learning (ML) Prediction: To identify biomarkers of efficacy, the team employed a Lasso logistic regression model. This model analyzed baseline metabolite levels to distinguish between "Complete Responders" (CR) and those for whom treatment failed (TF).
- Mechanistic Validation: To confirm the drug's impact on carbon flux, the team performed in vitro isotope tracing using [U-13C] glucose in HCT116 cells, allowing them to map the movement of carbon atoms through the TCA cycle under the influence of CPI-613.
Key Insights
The Significance of the Bone Marrow Microenvironment
While metabolic changes were detectable in systemic blood circulation, the most profound shifts occurred within the bone marrow. This finding is critical for AML, as the bone marrow serves as the primary site of leukemic activity. The study demonstrated that CPI-613 effectively penetrates this microenvironment to disrupt mitochondrial function where it is most needed.
Metabolic Signatures of Response
A clear distinction emerged between responders and non-responders. In patients who achieved complete remission, CPI-613 treatment caused a robust reduction in nearly all TCA cycle intermediates (such as citrate, malate, and fumarate) within the bone marrow. Conversely, non-responders showed minimal metabolic shifts, suggesting that the drug failed to adequately engage its targets in those individuals.
The "Predictive Trio"
Through machine learning, the researchers identified a specific three-metabolite signature—aspartate, dihydroorotate, and succinyl carnitine—that was highly predictive of patient outcomes. When measured at baseline, these metabolites allowed the model to predict treatment response with an impressive AUC of 0.93.
Results and Discussion
The study successfully demonstrated that CPI-613 treatment led to the accumulation of KADH substrates, such as the amino acid valine, in 21 out of 29 patients, providing definitive evidence of target engagement in a clinical setting. The isotope tracing experiments further confirmed that the drug specifically reduces the flux of glucose-derived carbons into the TCA cycle via pyruvate dehydrogenase (PDH) inhibition.
These results suggest that the efficacy of metabolic inhibitors like CPI-613 is not uniform but depends on the baseline metabolic state of the patient. By identifying the "Predictive Trio" of metabolites, this research paves the way for a more personalized approach to metabolic cancer therapy, where patients can be screened for likelihood of response before beginning treatment.
Limitations and Future Directions
While the high predictive accuracy is promising, the study's cohort of 29 patients is relatively small. Validation in larger, randomized Phase III trials is necessary to confirm the robustness of the metabolic signature. Additionally, further research is needed to determine if these findings extend to other malignancies, such as pancreatic cancer, where CPI-613 is also being investigated. Finally, the authors suggest exploring how external factors, such as patient diet, might influence the availability of the metabolic pathways targeted by the drug, potentially offering new avenues for combination therapies.
Summary updated