Toll-like receptors (TLRs) play an essential role in innate and acquired immune responses. They recognize a wide range of exogenous and endogenous ligands with remarkable specificity. TLR4, in particular, is involved in the recognition of bacterial lipopolysaccharide (LPS) and fibronectin fragments. TLRs owe their name to a highly homologous cytoplasmic domain, the Toll-IL-1 receptor (TIR) domain, that they share with the Drosophila Toll receptor and the IL-1 receptor (IL-1R) family members; this led to the suggestion that TLR/IL-1R family members use an analogous framework of signalling molecules. Upon ligand binding TLRs recruit the adaptor molecule MyD88 through homotypic interactions with a Toll domain found in the C terminus of MyD88. MyD88, in turn, recruits IL-1-receptor-associated kinase (IRAK) and IRAK-2, which induce the activation of MAPK and NF-?B through TNF-receptor-associated factor (TRAF)6. A long-standing problem with this model, however, was the observation that LPS can still induce MAPK and NF-?B activation in peritoneal macrophages from MyD88- and IRAK- but not TRAF6-deficient mice, whereas IL-1 and IL-18 cannot. This was a paradox and suggested that LPS may use, in addition to MyD88/IRAK, an as yet unidentified mediator to activate NF-?B through TRAF6 and MAPKs.