In Yersinia, YopN and associated proteins are thought to form a plug/gate in the type III secretion system, which opens on sensing contact with a eukaryotic cell. This, in turn, allows vectorial transport of a pore, composed of YopB and perhaps YopD, into the mammalian membrane17. Therefore, contact with the eukaryotic cell appears necessary for triggering secretion of the Yop pore. It is likely that secretion and pore insertion are physically linked in the Yersinia system: contact during secretion may be required for pore insertion into the eukaryotic membrane. L. pneumophila also requires contact with the eukaryotic cell for pore formation12. This contact dependence may imply a plug/gate trigger mechanism for secretion by the icm/dot complex. It is unlikely that an active pore complex is secreted directly into the medium after host contact because of the lack of cytotoxic activity in cell-free supernatants from infected cells12, 15. Efficient pore-forming activity only occurs during direct contact between bacteria and the host cell. It seems likely that L. pneumophila uses a relatively nonspecific receptor–ligand interaction to trigger pore insertion by the dot/icm transport machinery. Pores appear to be inserted in several cell types, including macrophages, HEp-2 cells (a human epithelial cell line) and red blood cells from seven different mammals tested and from chickens12 (JE Kirby, unpublished). This implies recognition of a common component in eukaryotic membranes, such as lipids or particular sugars on a membrane glycoprotein, rather than a receptor with a more-limited host range. The promiscuity of pore formation may help explain why Legionella is successful in parasitizing organisms as diverse as amoebae and humans. The icm/dot secretion machinery used by L. pneumophila contains remnants of a still functional bacterial conjugation system8–10.