CIPN is a debilitating neurotoxicity of paclitaxel that limits the treatment efficacy and reduces quality of life. The primary objective of this analysis was to validate the inverse association of CIPN occurrence with pre-treatment histidine;[22] however, no association was identified. In a secondary analysis of other amino acids, higher glutamate, phenylalanine, tyrosine, and valine were weakly associated with higher CIPN severity. Metabolic network analyses suggested that associations of amino acids with CIPN might act through a metabolic pathway mediated by glutamate.
Our prior analysis of a smaller patient cohort found an inverse relationship between whole blood levels of histidine, phenylalanine, and threonine and CIPN severity.[12] In this analysis of the much larger S0221 clinical trial cohort using archival serum samples, our previous findings were not reproduced. However, unexpectedly, phenylalanine, tyrosine, glutamate, and valine were associated with CIPN severity. Likely explanations for the different findings between the two studies may be the sample type (whole blood vs. serum)[31–33], the timing of blood collection (before paclitaxel vs. before any chemotherapy), and/or the duration the samples have been stored.
Phenylalanine is an essential aromatic amino acid supplied from dietary protein sources and is the precursor of tyrosine. Tyrosine is rapidly metabolized to 3,4-dihydroxyphenylalanine (levodopa), which is the precursor of many catecholamine neurotransmitters in the brain and is associated with many neurological diseases.[34] Prior research in humans and animals showed that genetic diseases that lead to the accumulation of phenylalanine can cause impairment of cognitive development, which requires the restriction of dietary phenylalanine intake. The resulting tyrosine deficiency has been suggested to be the cause of sensory neuropathy symptoms,[35] but a meta-analysis showed that tyrosine supplementation did not improve neuropsychological performance.[36] Whereas the current analysis found that patients with higher glutamate had more severe CIPN, several clinical trials have tested supplementation of glutamate or glutamine to reduce the severity of CIPN from paclitaxel, oxaliplatin, and vincristine, but the results were inconclusive.[37–43] Higher plasma glutamate has been reported in patients with more severe chronic pain syndromes caused by trauma,[44] but we are not aware of any studies comparing glutamate levels in patients with varying severity of CIPN.
A previous analysis conducted patients receiving vincristine suggested that machine learning approaches could be used to identify metabolomic signatures prior to and during treatment that were indicative of vincristine-induced peripheral neuropathy.[13] The investigators provide tools that could be used to predict a patient’s likelihood of CIPN, presumably to inform treatment decisions. We considered predictive CIPN models in the S0221 cohort using various machine learning techniques, however, these models had poor predictive performance and were therefore not included in this report. Our inability to validate our previously discovered amino acid biomarker candidates, or to identify sufficiently predictive amino acid signatures, demonstrates the importance of conducting statistically robust biomarker research prior to recommending biomarkers be used for clinical decision making. Although we find limited evidence supporting the use of amino acid metabolomics to predict CIPN, the amino acid signature in this analysis may provide intriguing avenues for further exploration of the biological mechanism underlying CIPN.
Although this is the largest CIPN metabolomics analysis that has been conducted to date and used CIPN data prospectively collected within a well-conducted clinical trial, there are several limitations of this analysis that should be considered. S0221 only collected CTCAE grade 3 + adverse events, which could have missed patients whose paclitaxel treatment was modified before they reached this level of severity, especially since dose adjustment data were not collected. Additionally, S0221 did not exclude patients with pre-existing neuropathy, who are particularly sensitive to peripheral neuropathy that may be unrelated to amino acid deficiencies. Information on diabetes and hyperlipidemia was also not collected, so possible confounding effects of relevant disease states or recent diet could not be accounted for in our analysis of non-fasting pre-treatment samples[45, 46]. Finally, our metabolomics panel only included the 20 amino acids that comprise proteins. Further research is warranted using broader metabolomics panels and other omics approaches such as lipidomics, proteomics, and microbiomics to better understand the key features that predict the occurrence and severity of CIPN.
In conclusion, this analysis found minimal evidence that pre-treatment levels of amino acids, including histidine, are associated with CIPN from paclitaxel. The metabolic network suggested that the weak association between amino acids and CIPN might be through the metabolism of glutamate. Additional studies that assess non-amino acid metabolomics and other potential predictive biomarkers are encouraged to identify predictive biomarkers that can be used to inform patient care and/or improve mechanistic understanding of CIPN that can be used to develop effective treatment approaches that could long-term treatment outcomes in patients with cancer.