Tolerance to endotoxin has been defined and analyzed either entirely in vivo or entirely in vitro. In contrast, the hyporeactivity of circulating leukocytes reported in patients with sepsis, and often referred to as the phenomenon of tolerance to endotoxin, is an ex vivo observation. Therefore, our objective was to set up an ex vivo model of tolerance to endotoxin.Mice were injected intravenously with lipopolysaccharide (LPS), and their leukocytes, derived from different compartments, were challenged in vitro with LPS or heat-killed Staphylococcus aureus Cowan I (SAC), for production of tumor necrosis factor (TNF).Production of TNF was observed in in vitro cultures of bronchoalveolar cells, peritoneal cells, splenocytes, and whole-blood samples harvested from mice 1-3 h after injection of LPS. Bone marrow cells, in contrast, did not release TNF. In parallel, cells acquired tolerance to LPS within 1-3 h after in vivo injection of LPS. A dramatic decrease in the production of TNF, in response to LPS, was observed by use of circulating leukocytes, splenocytes, peritoneal cells, and bone marrow cells 24 h after injection of LPS. In contrast, LPS-induced production of TNF by bronchoalveolar cells was far less reduced. The kinetics of acquisition of tolerance and recovery were different for different compartments. Cross-tolerance with SAC did not parallel the phenomenon of tolerance to endotoxin that was observed by use of LPS.These data show that tolerance to endotoxin, as monitored by ex vivo analysis, is compartmentalized and that bronchoalveolar cells are less likely than splenocytes, peritoneal cells, and bone marrow cells to develop tolerance to endotoxin.