Abstract
Magnetic resonance imaging (MRI) has been recognized as a valuable tool for achieving 'reification of clinical diagnosis' of major depressive disorder (MDD). However, the reliability and validity of MRI results are often compromised by genetic, environmental, and clinical heterogeneity within test samples. Here, we combined MRI with other clinical findings using multimodal MRI fusion algorithm to construct a data-driven, bottom-up diagnostic approach. The covariation patterns between the multimodal MRI features and differential expression of exosomal microRNA (miRNA) were identified on a subset of 70 MDD patients and 71 healthy controls (HCs) (served as a training set) as classification features, whereas data from the other 45 MDD patients and 43 HCs served as a test set. Furthermore, longitudinal data from 28 MDD patients undergoing antidepressant treatment for six months were utilized to validate the identified biomarkers, and related signaling pathways were initially explored in depression-like mice. Plasma exosome-derived miR-151a-3p levels were found to be significantly lower in MDD patients compared to HCs and correlated with abnormal changes in functional MRI (fMRI) metrics in the anterior cingulate cortex (ACC), visual cortex, and default mode network, etc. Then, these multimodal MRI features associated with miR-151a-3p expression distinguished MDD patients from HCs with high classification accuracy of 92.05% in support vector machine (SVM) model, outperforming the diagnostic rate when only multimodal MRI features with intergroup differences were entered (70.45%). Furthermore, 10 out of 28 MDD patients exhibited a clinically significant response to the treatment (a reduction of over 50% in Hamilton Rating Scale for Depression (HAMD) score). The significant upregulation of plasma exosomal miR-151a-3p levels and changes of fMRI indicators were also observed in these 10 patients after treatment of six months. Animal experiments have shown that reducing the expression of miR-151-3p in ACC induces depression-like behaviors in mice, while elevating hsa-miR-151a-3p expression in ACC alleviates the depression-like behaviors of mice exposed to chronic unpredictable mild stress. Our study proposed an innovative diagnostic model of MDD by combining the plasma exosome-derived miR-151a-3p expression with its associated multimodal MRI patterns, potentially serving as a novel diagnostic tool.