Background: There are conflicting reports on the ability of granzyme A (GzmA) to kill cells. Results: Substrate specificity mapping of human and mouse GzmA was used to identify Serpinb6b as a mouse-specific GzmA inhibitor. Conclusion: Mouse but not human GzmA is controlled by an intracellular inhibitor. Significance: The GzmA debate is partly explained by species-specific divergence in cytotoxicity. The granzyme family serine proteases are key effector molecules expressed by cytotoxic lymphocytes. The physiological role of granzyme (Gzm) A is controversial, with significant debate over its ability to induce death in target cells. Here, we investigate the natural inhibitors of GzmA. We employed substrate phage display and positional proteomics to compare substrate specificities of mouse (m) and human (h) GzmA at the peptide and proteome-wide levels and we used the resulting substrate specificity profiles to search for potential inhibitors from the intracellular serpin family. We identified Serpinb6b as a potent inhibitor of mGzmA. Serpinb6b interacts with mGzmA, but not hGzmA, with an association constant of 1.9 +/- 0.8 x 10(5) m(-1) s(-1) and a stoichiometry of inhibition of 1.8. Mouse GzmA is over five times more cytotoxic than hGzmA when delivered into P815 target cells with streptolysin O, whereas transfection of target cells with a Serpinb6b cDNA increases the EC50 value of mGzmA 13-fold, without affecting hGzmA cytotoxicity. Unexpectedly, we also found that Serpinb6b employs an exosite to specifically inhibit dimeric but not monomeric mGzmA. The identification of an intracellular inhibitor specific for mGzmA only indicates that a lineage-specific increase in GzmA cytotoxic potential has driven cognate inhibitor evolution.