Poor immune reconstitution after haploidentical stem cell transplantation results in a high mortality from viral infections and relapse. One approach to overcome this problem is to selectively deplete the graft of alloreactive cells using an immunotoxin directed against the activation marker CD25. However, the degree of depletion of alloreactive cells is variable following stimulation with recipient peripheral blood mononuclear cells (PBMCs), and this can result in graft versus host disease (GVHD). We have refined this approach using recipient Epstein-Barr virus (EBV)-transformed lymphoblastoid cell lines (LCLs) as stimulators to activate donor alloreactive T cells. Our studies demonstrate that allodepletion with an anti-CD25 immunotoxin following stimulation with HLA-mismatched host LCLs more consistently depleted in vitro alloreactivity than stimulation with host PBMCs, as assessed in primary mixed lymphocyte reactions (MLRs). Allodepletion using this approach specifically abrogates cytotoxic T-cell responses against host LCLs. In interferon-gamma (IFN-gamma) enzyme-linked immunospot (ELISPOT) assays, antiviral responses to adenovirus and cytomegalovirus (CMV) were preserved following allodepletion. Likewise, using HLA-A2-pp65 tetramers, we have shown that the frequency of CMV-specific T cells is unaffected by allodepletion. Moreover, the donor anti-EBV response is partially retained by recognition of EBV antigens through the nonshared haplotype. Finally, we studied whether allodepletion affects the response to candidate tumor antigens in myeloid malignancies. Using HLA-A2-PR1 tetramer analysis, we found that the frequency of T cells recognizing the PR1 epitope of proteinase 3 was not significantly different in allodepleted and unmanipulated PBMCs from patients with chronic myeloid leukemia (CML) undergoing transplantation. Based on these data, we have embarked on a phase 1 clinical trial of addback of allo-LCL-depleted donor T cells in the haplo-identical setting.