A human tumor necrosis factor-alpha (TNF-alpha) mutant (M3S) with low systemic toxicity in vivo was designed, and its structures in two different crystal packings were determined crystallographically at 1.8 and 2.15-A resolution, respectively, to explain altered biological activities of the mutant. M3S contains four changes: a hydrophilic substitution of L29S, two hydrophobic substitutions of S52I and Y56F, and a deletion of the N-terminal seven amino acids that is disordered in the structure of wild-type TNF-alpha. Compared with wild-type TNF-alpha, it exhibits 11- and 71-fold lower binding affinities for the human TNF-R55 and TNF-R75 receptors, respectively, and in vitro cytotoxic effect and in vivo systemic toxicity of M3S are 20 and 10 times lower, respectively. However, in a transplanted solid tumor mouse model, M3S suppresses tumor growth more efficiently than wild-type TNF-alpha. M3S is highly resistant to proteolysis by trypsin, and it exhibits increased thermal stability and a prolonged half-life in vivo. The L29S mutation causes substantial restructuring of the loop containing residues 29-36 into a rigid segment as a consequence of induced formation of intra- and intersubunit interactions, explaining the altered receptor binding affinity and thermal stability. A mass spectrometric analysis identified major proteolytic cleavage sites located on this loop, and thus the increased resistance of M3S to the proteolysis is consistent with the increased rigidity of the loop. The S52I and Y56F mutations do not induce a noticeable conformational change. The side chain of Phe56 projects into a hydrophobic cavity, while Ile52 is exposed to the bulk solvent. Ile52 should be involved in hydrophobic interactions with the receptors, since a mutant containing the same mutations as in M3S except for the L29S mutation exhibits an increased receptor binding affinity. The low systemic toxicity of M3S appears to be the effect of the reduced and selective binding affinities for the TNF receptors, and the superior tumor-suppression of M3S appears to be the effect of its weak but longer antitumoral activity in vivo compared with wild-type TNF-alpha. It is also expected that the 1.8-A resolution structure will serve as an accurate model for explaining the structure-function relationship of wild-type TNF-alpha and many TNF-alpha mutants reported previously and for the design of new TNF-alpha mutants.