Tumor necrosis factor (TNF) seems always to have enjoyed a rather conspicuous visibility in biomedical research. With historical roots in the century-old phenomenon of bacterial-induced hemorrhagic necrosis of tumors, TNF-or, rather, its two homologous forms, TNFa and LTa (lymphotoxin, TNFj3)-were finally molecularly cloned in 1984, among the very first cytokines to bs so unambiguously defined. Although TNFa and LTa, classically the respective products of activated macrophages and T cells, can indeed kill many transformed lines, these functionally similar and extraordinarily pleiotropic cytokines are today viewed as primary mediators of immune regulation and the inflammatory response, closely linked to the develop ment of disease. The crucial involvment of TNF, for example, in septic shock, some autoimmune disorders, and graft-host disease is well established (see Beutler, 1992). Since the cloning of two distinct but structurally homologous receptors for TNF,~ 75 and~ 55 (each of which binds both ligands), the past 3 years have witnessed the rapid emergence of two superfamilies, of which the TNFs and their receptors are only representatives (Farrah and Smith, 1992; Suda et al., 1993; Smith et al., 1993). To date, 12 receptors have been identified (Figure 1) with which we can associate some eight TNF-related cytokines (Figure 2). The distinctive but overlapping cellular responses their interactions produce clearly define developmental and regulatory networks involving cells of the lymphoid, hematopoetic, andother lineages. In thisminireviewwe make no attempt to discuss individual members comprehensively and instead highlight emerging global characteristics that distinguish them from other cytokine families: structure, biological networks, and the intriguing ability of some members to induce cell death. A new face to the TNF system seems at hand.

The TruF Receptor Fam//y/lIkracls w/M a Par@//el Fern//y of Ligands The receptors, with two exceptions, are all type I membrane proteins with sequence homology (almost entirely) confined to the extracellular region. The exceptions, T2 and A53R, are poxvirus gene products that map to different genetic loci and have been shown to encode soluble, secreted forms of TNF receptors (Smith et al., 1991). These function to complex (and thereby inactivate) hostproduced TNF. T2 is clearly an acquired form of the~ 75 cellular receptor, while A53R, since it binds only TNFa and shows much lower sequence homology, may represent a third TNF receptor. The extraordinary virulence of wildtype myxoma poxvirus, uniformly fatal to its host (rabbits), is reduced nearly 50% in recombinants differing only by an inactivated 72 gene (Upton et al., 1991). Interestingly,