To investigate the symptom network structure of major depressive disorder (MDD) with mixed features and implications for treatment. In this post-hoc analysis of a previously reported randomized trial, patients meeting DSM-IV-TR criteria for MDD presenting with two or three manic symptoms (DSM-5 mixed features specifier) were randomized to 6 weeks of double-blind treatment with lurasidone 20-60mg/d (N=109) or placebo (N=100). The network structure of symptoms at baseline and their treatment moderating effects were investigated. Network analyses showed that both ``elevated mood'' (YMRS item 1) and ``increased motor activity-energy'' (YMRS item 2) were associated with ``sleep disturbance'' ("bridge" symptom) and the depressive symptom cluster. Presence of both "elevated mood" and "increased motor activity-energy" at baseline predicted significantly less improvement in MADRS and CGI-S score at week 6 with lurasidone (vs. placebo) compared to patients without these manic symptoms at baseline. The network model also showed "rapid/pressured speech" (YMRS item 6) at baseline predicted improvement in both manic and depressive symptoms with lurasidone vs. placebo treatment. This was a post-hoc analysis where findings need to be confirmed by prospective controlled studies. This post-hoc analysis describes the symptom network structure of MDD with mixed features in a patient sample at study baseline. Specific manic symptoms were found to be linked to sleep disturbance (characterized as a "bridge" symptom), which in turn linked the manic and depressive symptom clusters. The presence (vs. absence) of the specific manic symptoms we identified moderated the antidepressant and antimanic effects of lurasidone in the treatment of MDD with mixed (subthreshold hypomanic) features.