Poor T lymphocyte reconstitution limits the use of haploidentical stem cell transplantation (SCT) because it results in a high mortality from viral infections. One approach to overcome this problem is to infuse donor T cells from which alloreactive lymphocytes have been selectively depleted, but the immunologic benefit of this approach is unknown. We have used an anti-CD25 immunotoxin to deplete alloreactive lymphocytes and have compared immune reconstitution after allodepleted donor T cells were infused at 2 dose levels into recipients of T-cell-depleted haploidentical SCT. Eight patients were treated at 10(4) cells/kg/dose, and 8 patients received 10(5) cells/kg/dose. Patients receiving 10(5) cells/kg/dose showed significantly improved T-cell recovery at 3, 4, and 5 months after SCT compared with those receiving 10(4) cells/kg/dose (P < .05). Accelerated T-cell recovery occurred as a result of expansion of the effector memory (CD45RA(-)CCR-7(-)) population (P < .05), suggesting that protective T-cell responses are likely to be long lived. T-cell-receptor signal joint excision circles (TRECs) were not detected in reconstituting T cells in dose-level 2 patients, indicating they are likely to be derived from the infused allodepleted cells. Spectratyping of the T cells at 4 months demonstrated a polyclonal Vbeta repertoire. Using tetramer and enzyme-linked immunospot (ELISPOT) assays, we have observed cytomegalovirus (CMV)- and Epstein-Barr virus (EBV)-specific responses in 4 of 6 evaluable patients at dose level 2 as early as 2 to 4 months after transplantation, whereas such responses were not observed until 6 to 12 months in dose-level 1 patients. The incidence of significant acute (2 of 16) and chronic graft-versus-host disease (GVHD; 2 of 15) was low. These data demonstrate that allodepleted donor T cells can be safely used to improve T-cell recovery after haploidentical SCT and may broaden the applicability of this approach.