F or decades, immunologists occupied themselves with trying to understand the initiation of an immune response. This led to a highly detailed molecular portrait of antigen recognition and activation. Partly spurred by the enigma ofautoimmune conditions, recent attention has now focused on how immune responses are turned off. Central to this issue is the maintenance of immune cell homeostasis. Because immune responses are potentially dangerous alterations in normal physiology, they must be carefully controlled or extinguished if the antigenic stimulus either becomes too great or is successfully eliminated. One important mechanism of lymphocyte control is programmed death, which may occur in all immune responses. T cells have at least two apoptotic pathways: first, active death, which is antigen driven, and second, passive death, which occurs at the conclusion of an immune response and may be due to lymphokine withdrawal or other mechanisms (reviewed in 1-3)(Fig. 1). These forms of death are molecularly distinct, since Fas and TNF are major participants in active but not passive death, and Bcl-2 or Bcl-x can abrogate passive but not antigen-driven death (2--4). Therefore, the cellular response to antigen can be viewed as a cycle of birth, propagation, and death. Memory may comprise lymphocytes that escape this circuit and persist (1, 5). The article by Simon et al.(6) illustrates the dangers of defective homeostasis by showing that somatic alterations in the regulation of the death-inducing molecule Fas may be associated with hypereosinophilia and its attendant pathological complications. Fas (also called Apo-1, CD95, and Apt) is a member of the TNF receptor (TNFR) superfamily, whose eldest siblings, TNFR type I (p55) and TNFR type II (p75), have long been known to induce programmed cell death (reviewed in 2, 7). The other transmembrane proteins belonging to the TNFR superfamily include the low affanity nerve growth factor receptor (NGFR), CD27, CD30, CD40, OX40, and 4-1BB. Many of these molecules have well-defined roles in immune regulation and homeostasis. These proteins share up to 20-25% amino acid identity, predominantly in cysteine-rich domains (CRDs) within the extracellular domain. CRDs define membership in the family and are necessary for ligand binding. The intracytoplasmic portions of the TNFR-related proteins are unique except for Fas p55 TNFR and the NGFR. These receptors share homology within a 70-aa" death domain" that is required for transducing death signals (8). The death domain acts as a docking site for a family of soluble cytoplasmic signaling proteins that also contain death domains (9). The death domain is conserved far back in evolution, since it is present in a Drosophila protein, Reaper, which genetic evidence has implicated in cell death (10). The intracellular domains of other TNF family receptors are known to interact with cytosolic proteins that fall into a distinct family of homolognes called TRAF (for TNFR-associated factor)(11, 12).

The ligands for the TNFR-like receptors also constitute a gene superfamily and share structural similarity complementary to the receptor ensemble (13). Both the receptors and the ligands assemble and function as homotrimers in which the threefold axis of symmetry lies perpendicular to the cell membrane. The ligands are clasped on three sides by the CRDs of the receptors, which adopt an extended conformation that projects from the membrane (14). Each ligand appears to bind a single member of the receptor superfamily, with the exception of TNF and lymphotoxin-o~, which bind and signal through both the p55 and p75 TN-FRs. Interestingly, in addition to …