Although substantial efforts have been made to elucidate the neuronal basis of both Treatment-resistant depression (TRD) and non-TRD (nTRD), the results are inconsistant. We apply the resting-state dynamic functional connectivity (D-RSFC) to explore the divergence of neuronal basis of both depression subtypes. Our results demonstrated that the prefrontal-limbic circuit is the most stable dysfunctional brain network in depression. The D-RSFC method could reveal the altered dynamic functional connectivity in both MDD subtypes and the divergence of brain networks between TRD and nTRD. Additionally, we speculate that the caudate-ACC circuit might be the biomarker for evaluating treatment response in TRD.