A human skin allograft injury model in immunodeficient mice, engrafted with human peripheral blood mononuclear cells from a different donor, has been used to test whether reagents that block human T cell CD2 interactions with its principal ligand, LFA-3 (CD58), can inhibit immune reactions in vivo. In this model, human skin grafts show a reproducible pattern of progressive human T-cell infiltration and human graft microvascular injury that resembles human first-set skin graft rejection. Murine Mab to human LFA-3 or human LFA-3-IgG1 fusion protein, but not isotype-matched control antibodies, each markedly protected skin grafts from leukocyte infiltration and injury. These data provide the first evidence that LFA-3 functions in vivo and establish the ability of this new model to test human-specific immune modulators.