Major Depressive disorder (MDD) is a chronic and recurrent brain disorder characterized by episode and remission phases, and poor therapeutic responses. The molecular correlates of MDD have been investigated in case-control settings, but the biological changes associated with trait (regardless of episode/remission) or state (illness phases) remains largely unknown, hence preventing therapeutic opportunities. To address this gap, we generated transcriptome profiles in the subgenual anterior cingulate cortex of MDD subjects who died during a single or recurrent episode or when in remission. We show that biological changes associated with MDD trait (inflammation, immune activation, reduced bioenergetics) are distinct from those associated with MDD phases or state (neuronal structure and function, neurotransmission). On the cell-type level, gene variability in subsets of GABAergic interneurons positive for corticotropin-releasing hormone, somatostatin or vasoactive-intestinal peptide was associated with MDD phases.Applying a probabilistic Bayesian network approach, we next show that gene modules enriched for immune system activation, cytokine response and oxidative stress, may exert causal roles across MDD phases. Finally, using a database of drug-induced transcriptome perturbations, we show that MDD-induced changes in putative causal pathways are antagonized by families of drugs associated with clinical response, including dopaminergic and monoaminergic ligands, and uncover potential novel therapeutic targets. Collectively, these integrative transcriptome analyses provide novel insight into cellular and molecular pathologies associated with trait and state MDD, and a method of drug discovery focused on disease-causing pathways.