In vivo vaccination with cell line-derived whole tumor lysates: neoantigen quality, not quantity matters

Cancer vaccines provide a complex source of neoantigens. Still, increasing evidence reveals that the neoantigen quality rather than the quantity is predictive for treatment outcome.

Using the preclinical Mlh1^-/- tumor model, we performed a side-by side comparison of two autologous cell-line derived tumor lysates (namely 328 and A7450 T1 M1) harboring different tumor mutational burden (TMB; i.e. ultra-high: 328; moderate-high: A7450 T1 M1). Mice received repetitive prophylactic or therapeutic applications of the vaccine. Tumor incidence, immune responses and tumor microenvironment was examined.

Both tumor cell lysates delayed tumor formation in the prophylactic setting, with the A7450 T1 M1 lysate being more effective in decelerating tumor growth than the 328 lysate (median overall survival: 37 vs. 25 weeks). Comparable results were achieved in therapeutic setting and could be traced back to antigen-driven immune stimulation. Reactive T cells isolated from A7450 T1 M1-treated mice recognized autologous Mlh1^-/- tumor cells in IFNγ ELISpot, but likewise YAC-1 cells, indicative for stimulation of both arms of the immune system. By deciphering local effects, vaccines shaped the tumor microenvironment differently. While A7450 T1 M1 prophylactically vaccinated tumors harbored low numbers of myeloid-derived suppressor cells (MDSC) and elevated CD8-T cell infiltrates, vaccination with the 328 lysate evoked MDSC infiltration. Similar effects were seen in the therapeutic setting with stable disease induction only upon A7450 T1 M1 vaccination. Untangling individual response profiles revealed strong infiltration with LAG3⁺ and PD-L1⁺ immune cells when treatments failed, but almost complete exclusion of checkpoint-expressing lymphocytes in long-term survivors.

By applying two tumor cell lysates we demonstrate that neoantigen quality outranks quantity. This should be considered prior to designing cancer vaccine-based combination approaches.