Objectives: Approximately one-third of major depressive disorder (MDD) patients do not respond to standard antidepressants and develop treatment-resistant depression (TRD). We aimed to reveal metabolic differences and discover promising potential biomarkers in TRD.
Methods: Our study recruited 108 participants including healthy controls (n = 40) and patients with TRD (n = 35) and first-episode drug-naive MDD (DN-MDD) (n = 33). Plasma samples were presented to ultra performance liquid chromatography-tandem mass spectrometry. Then, a machine-learning algorithm was conducted to facilitate the selection of potential biomarkers.
Results: TRD appeared to be a distinct metabolic disorder from DN-MDD and healthy controls (HCs). Compared to HCs, 199 and 176 differentially expressed metabolites were identified in TRD and DN-MDD, respectively. Of all the metabolites that were identified, spermine, spermidine, and carnosine were considered the most promising biomarkers for diagnosing TRD and DN-MDD patients, with the resulting area under the ROC curve of 0.99, 0.99, and 0.93, respectively. Metabolic pathway analysis yielded compelling evidence of marked changes or imbalances in both polyamine metabolism and energy metabolism, which could potentially represent the primary altered pathways associated with MDD. Additionally, L-glutamine, Beta-alanine, and spermine were correlated with HAMD score.
Conclusions: A more disordered metabolism structure is found in TRD than in DN-MDD and HCs. Future investigations should prioritize the comprehensive analysis of potential roles played by these differential metabolites and disturbances in polyamine pathways in the pathophysiology of TRD and depression.
Keywords: Biomarker; Machine learning; Metabolomics; Treatment-resistant depression.
Copyright © 2023 Elsevier Inc. All rights reserved.