Immunotoxin (IT) therapy shows potential for selectively eliminating GVHD-causing T cells in vivo, but the field has been hampered by toxicity. Previously, we showed that a genetically engineered IT consisting of a single-chain protein, including the anti-CD3sFv spliced to a portion of diphtheria-toxin (DT390) has anti-GVHD effects, but pronounced organ toxicity common to this class of agent. A recombinant DT390 anti-CD3sFv protein previously shown to have anti-GVHD activity was modified to reduce its filtration into kidney by genetically inserting a cysteine residue downstream of the sFv moiety at the c-terminus of the protein. This modification produced an intermolecular disulfide bridge, resulting in a bivalent, rather than a monovalent IT, termed SS2, that selectively inhibited T-cell proliferation in vitro. Although monomer and SS2 were similar in in vitro activity, SS2 had a superior therapeutic index in vivo with at least 8- fold more being tolerated with reduced kidney toxicity. Most importantly, in a lethal model of GVHD, 40 μg SS2 given for 1 day, protected 100% of the mice from lethal GVHD for 3 months, whereas the maximum tolerated dose (MTD) of monomer protected only 33%. To our knowledge, this is the first time disulfide bonded ITs have been created in this way and this simple molecular modification may address several problems in the IT field because it (1) markedly increased efficacy curing mice of GVHD after a single daily treatment, (2) markedly decreased organ toxicity, (3) increased the tolerated dosage, and (4) created a therapeutic window where none existed before. (C) 2000 by The American Society of Hematology.