T he study of interactions between enteric bacterial pathogens and host cells has relied substantially on model systems. This is especially true for human pathogens, such as Salmonella typhi or Shigella species, as experimental studies on human volunteers are restricted and clinical samples are rare. The reductionist approach has been used successfully to dissect the complex phenomenon of disease into single processes, which can be examined in simple model systems. Work on shigellosis, caused by Shigella species’, has provided an instructive example of how using model systems of increasing complexity has furthered the understanding of an enteric disease. The fundamental in vitro model of enteropathogenesis uses semiconfluent monolayers of cultured epithelial cells (such as HeLa or Henle) infected with bacteria. For shigellosis, the entire sequence of key events accompanying the infection process (see Fig. 1) can be reproduced in this system, including bacterial entry, escape from the vacuole, intracellular replication, intraand intercellular movement, and host-cell killing’. The basic in vitro technique of infected monolayers has been refined further to quantitate infection-related events, such as bacterial uptake (by the gentamicin-killing assay2) and cellto-cell spread (by the infectious-foci assay3). However, the two-dimensional cell model system is a poor representation of the intestinal barrier, as it lacks impermeability and polarity. The next step in complexity is an in vitro model consisting of polarized epithelial monolayers4. Cell lines of colonic origin (such as Caco-2 and T-84), grown to confluency on permeable filters, mimic a columnar intestinal barrier. Tight junctions separate an apical (luminal) surface, organized as a brush border, from a basolateral domain. A commonly used experimental system uses a two-chamber device with differential exposure of the apical and basolateral surfaces so that vectorial bacterial infection from either side is possible. The study of invasion by Shigella in this system has revealed an apparent paradoxj. Shigella do not invade a polarized tight epithelium from the apical side, although this is the surface exposed to the gut lumen from which natural infection occurs. Unlike Salmonella species, Shigella need access to the basolateral surface to invade, a property that is shared by some other enteroinvasive bacteria, such as Yersinia and Listeria. A further refinement of this threedimensional, polarized cell model system is the addition of motile immune cells: polymorphonuclear leukocytes (PMNs)~. An experimental system consisting of epithelial cells, bacteria and PMNs enables the examination of the complex interplay between these three cell types, involving direct cell-cell interactions and cell-cell communication via signaling molecules (such as secreted bacterial products and cytokines). Application of this elaborate model to the study of shigellosis has provided one explanation for the paradox of Shigella in-