Immune-checkpoint blockers (ICB) provide limited benefit against T cell-depleted tumours, calling for therapeutic innovation. Here, we aimed at designing a new type of dendritic cell (DC) vaccine by unbiased computational integration of multi-omics data from cancer patients. In a first attempt, a DC vaccine designed to present tumor antigens from cancer cells succumbing to immunogenic cancer cell death (ICD) and to elicit high type I interferon (IFN) responses failed to induce the regression of mouse tumors lacking T cell infiltrates. In lymph nodes (LNs), instead of activating CD4+ and CD8+T cells, DCs stimulated immunosuppressive PD-L1+LN-associated macrophages (LAMs) via a type I IFN response. Moreover, DC vaccines of this type stimulated pre-existing, T cell-suppressive, PD-L1+tumour-associated macrophages (TAMs). This created a T cell-suppressive circuit of PD-L1+macrophages, spanning across LNs and tumours. Accordingly, DC vaccines synergised with PD-L1 blockade to deplete PD-L1+macrophages, suppress myeloid inflammation affecting the tumor bed and draining lymph nodes, and de-inhibit effector/stem-like memory T cells, eventually causing tumour regression. The synergistic interaction between the DC vaccine and PD-L1 blockade was lost when DCs were manipulated to lose Ifnar1or Ccr7 or when macrophages were depleted. Interestingly, clinical DC vaccines also potentiated lymphocyte-suppressive PD-L1+TAMs in patients bearing T cell-depleted tumours. Altogether, our results reveal the existence of a novel PD-L1+LAM/TAM-driven immunosuppressive pathway that can be elicited by DC vaccines, yet can be subverted for improving the outcome of immunotherapy.