The intestinal mucosa forms a barrier that protects against invasion of the host by nonpathogenic bacteria residing in the intestinal lumen. Some enteropathogens, such as Salmonella, Shigella, and Yersinia, have specific tissue-invading capabilities and can physically breach the intestinal mucosal barrier (32). Pathogen-induced traversal of the human intestinal mucosa may involve crossing of absorptive enterocytes or passage through the specialized M cells (14-16). In addition to transcellular entry, some bacteria can disrupt the tight junctions between enterocytes and cross via paracellular mechanisms (47). Salmonella enterica serovar Typhimurium may also cross the intestinal mucosa by sequestering itself within luminal CD18-positive mononuclear cells that shuttle the microbe paracellularly across the villus epithelium into the systemic circulation (48). Shigella (45) and possibly Campylobacter (49) may also breach the intestinal barrier by transcytosis across M cells. C. jejuni has previously been reported to undergo transcellular translocation across the mucosa; there have also been reports of paracellular translocation of Campylobacter (7, 9-12, 25, 28, 35, 47, 49).

Cultured mammalian cells are commonly used as a simple model for investigating the host epithelial cell-bacterium interaction that is more easily controlled than whole animals. Caco-2 cells were derived from transformed human colonic carcinoma cells. During 7 to 10 days in culture, these cells form polarized monolayers that have intercellular tight junctions and defined apical and basolateral surfaces. The differentiated apical surface has dense brush border microvilli containing normal intestinal brush border enzymes and antigens (13, 18, 43, 44). A key advantage of differentiated Caco-2 cells is that they provide a substrate monolayer similar to that for which Campylobacter shows a natural tropism in vivo. This cell line allows assessment of the ability of C. jejuni to penetrate and pass through a differentiated intestinal epithelial barrier, a process that appears to be important in causing overall colonic damage and the occasional bacteremia associated with Campylobacter enterocolitis (9, 10, 26, 28).

C. jejuni 81-176 is a well-characterized strain which exhibits relatively high levels of epithelial cell invasion in cultured host cell assays and causes a bloody colitis, as demonstrated in human challenge studies (6; D. Trible, unpublished data). Cell culture invasion assays conducted with inhibitors of cytoskeletal function initially demonstrated that this strain invades epithelial cells apically via a novel mechanism that is strongly microtubule (MT) dependent (38). Subsequent immunofluorescence studies of this 81-176 invasion process confirmed the involvement of MTs, demonstrated that the MT-associated molecular motor dynein is required for this uptake process, and defined the kinetics of INT407 cell invasion by C. jejuni 81-176 (19). Recent work has also shown that Ca²⁺ release from host intracellular stores is essential for C. jejuni 81-176 internalization into host cells (21). In addition, host signal transduction studies have suggested that strain 81-176 interacts at filipin III-sensitive membrane microdomains of the host cell surface with G-protein-coupled receptors, a process which activates phosphatidylinositol 3-kinase and mitogen-activated protein kinases that appear to be intimately involved in events controlling 81-176 internalization (22). Very recently, the rho GTPases Rac1 and Cdc42 have been implicated in the 81-176 invasion process (31). It is also noteworthy that other researchers, using different invasion assay conditions, have observed mixed involvement of both MTs and microfilaments in 81-176 invasion (5, 36). Watson and Galan (51) recently demonstrated that following internalization into host cells C. jejuni 81-176 survives within modified endosomes that avoid fusion with lysosomes.

Based on previous Campylobacter pathogenesis studies with cultured cells, animals, and humans (4-8, 10, 12, 17, 19-22, 23-40, 46, 49-53), we hypothesize that during the first few hours of infection C. jejuni adheres to and crosses differentiated intestinal epithelial cells by a transcellular process which does not disrupt transepithelial electrical resistance (TER) (transcytosis is apical endocytosis followed by basolateral exocytosis). Thus, a kinetic examination of C. jejuni internalization and microscopic analyses of the events involving adherence, invasion, and transcytosis over a 4-h period were conducted with the goal of obtaining an enhanced understanding of Campylobacter-host cell interactions. Importantly, this study provided new information about entry kinetics and revealed new specific bacterium-host cell interactions involved in epithelial cell translocation that have not been observed in previous electron microscope (EM) studies of Campylobacter entry mechanisms (4, 10, 26, 28, 31).

Both scanning EM and transmission EM were used to examine the early Campylobacter-Caco-2 cell interactions in differentiated monolayers grown on Transwell filters. Scanning EM revealed unique torsional contacts between the sides of flagella and microvillus tips, in contrast to previous reports of flagellar tip adherence (26, 28), possibly due to the higher-resolution images obtained in the current study. Both C. jejuni 81-176 and NCTC 11168 were typically tethered at the host surface by each polar flagellum binding laterally to different microvillus tips (Fig. ​(Fig.22 A, B, E, and G), an interaction which suggests that there is specific binding between the sides of flagella and components of the rounded tips of microvilli. Transmission EM revealed additional contacts between the bacterial cell surface and the sides of microvilli (Fig. 2C and D). The latter contacts may involve capsular polysaccharide, Peb1, JlpA, or other outer membrane proteins which have been reported to be involved in adherence (2, 24, 39-41). C. jejuni has a brainlike convoluted surface (Fig. ​(Fig.2G),2G), which may play a role in host cell-bacterium interactions. We speculate that ligands located at the apical surface (instead of deep in the folds) of the convolutions may be more active in host cell interactions. The bacterium-host cell interactions described above were readily observed early and throughout the invasion period. We presume that they represent different types of adherence, some of which may be reversible.

Flagella have long been implicated as Campylobacter adhesins (34, 37). Flagellum-dependent motility and the flagellum itself appear to be essential for C. jejuni invasion of epithelial cells, but the mechanism(s) remains unknown (1, 12, 50, 52, 53). It seems unlikely that the Campylobacter flagellar adherence is casual because the infected monolayers were washed six times with Earle's balanced salt solution prior to fixation for scanning EM, a process that involves an additional 10 washes and buffer changes. Although most EM studies were performed with an MOI of 100, using an MOI of 10 resulted in similar bacterium-host cell interactions, but the frequencies were much reduced. We speculate that C. jejuni flagellum-host cell interactions represent a primary adherence mechanism that augments the contact of “invasion-specific” bacterial ligands with host membrane receptors in lipid rafts, which result in signal transduction events that lead to bacterial internalization. Since ~10% of the invasion ability can be restored to nonflagellated C. jejuni cells by centrifugation onto a monolayer (50), we suggest that “invasion-specific” C. jejuni ligands may be located at the bacterial poles, which are normally adjacent to the polar flagella in wild-type C. jejuni. Strain NCTC 11168 is known to be less invasive than 81-176, possibly due to reduced numbers of “invasion-specific” ligands or expression of surface factors that interfere with these ligands.

C. jejuni cells were observed to interact typically with microvilli at the apical Caco-2 cell surface adjacent to intercellular junctions, as noted previously (28). Very infrequently, Campylobacter bacteria were observed bound perpendicular to the host cell, possibly with one polar flagellum interacting with the host cell (Fig. ​(Fig.3A);3A); a similar EM observation of perpendicular bacterial interaction was recently reported (31). This is a relatively difficult event to observe, possibly because the organism is internalized rapidly, and it is technically difficult to find a thin section containing such a physically space-limited event. Figure ​Figure3A3A shows that the local host membrane surface lacks the typical dense microvilli shown in Fig. ​Fig.2D,2D, and this may reflect a differentiated M-like cell surface. We suggest that intimate bacterium-host cell contact, as shown in Fig. 3A, B, or D, may be “invasion specific” and that binding of bacterial ligands with specific host plasma membrane receptors may have triggered a localized collapse of the terminal F-actin web, leading to shortening and coalescence of microvilli. Previously, we and other workers have reported that pretreatment of monolayers with cytochalasin D actually enhances Campylobacter invasion (19, 38, 51), possibly by eliminating the terminal actin web. Thus, initial actin depolymerization and subsequent polymerization are probably involved in Campylobacter entry. In fact, a recent report showed that Rac1 and Cdc42 are involved and that there may be membrane ruffling prior to C. jejuni invasion (31). We have no evidence of effector secretion into host cells via the bacterium-host cell interaction shown in Fig. ​Fig.3A,3A, but we cannot exclude this interesting possibility. Konkel and coworkers (27) suggested that effector secretion is important in Campylobacter invasion.

An early membrane dynamic event observed during Campylobacter entry is the formation of an activated membrane protrusion, which may represent a coalescence of local microvilli. We speculate that invasion-specific bacterium-host cell contact may promote host signaling events that activate the membrane in a localized region, leading to bacterial internalization through this “activated” site (Fig. ​(Fig.3B).3B). “Activation” may encompass host signaling events, as well as cytoskeletal and membrane alterations. In differentiated Caco-2 cells, this “activated” membrane protrusion is approximately the length of one bacterium wide and 2 to 3 lengths high and can apparently undergo multiple bacterial internalization events. In contrast, the comparable invasion-specific, “activated” membrane extension observed in undifferentiated INT407 cells during C. jejuni invasion is smaller and apparently engulfs a single organism (Fig. ​(Fig.3D).3D). Regardless of the host membrane protrusion size, bacterial internalization apparently results from plasma membrane invagination that begins at the bacterium-host cell contact site (19, 26, 29, 31). As observed previously (28, 38), the bacteria are internalized into an endosome, which is transported over time from the apical host surface to the basolateral host surface (Fig. ​(Fig.4).4). Our previous confocal microscopic analyses suggested that C. jejuni 81-176 cells are transported in endosomes via the molecular motor dynein along MTs (19). Watson and Galan (51) recently reported that Campylobacter cells reside and transit within special endosomes that avoid fusion with lysosomes. At later infection times, an occasional bacterium is observed in the intercellular space below apical junctional adherence (Fig. ​(Fig.3C).3C). This event has not been observed commonly or earlier in infection, leading us to suggest that this bacterium just exited the host cell laterally; however, the possibility of paracellular entry from the apical surface followed by junctional resealing cannot be ruled out (8). Based on the current study, we suggest that C. jejuni bacteria in endosomes are transported basolaterally through the cell and can undergo specific endosome-basolateral membrane fusion (Fig. ​(Fig.4D),4D), an exocytosis process which releases C. jejuni subepithelially. This process does not visually appear to be a simple reversal of the apical endocytic mechanism. Although we consider a basolateral endocytosis event unlikely, this possibility cannot be ruled out by the results of EM. Previous reports indicated that the ability of C. jejuni to invade Caco-2 cells is not linked to translocation ability (7, 17), also suggesting that endocytosis and exocytosis are unique mechanistic events. Based on these observations, we speculate that C. jejuni bacteria proficient in transcytosis across the epithelium may be able to alter the Campylobacter-endosome membrane so that they traffic through the cell and can fuse with the host plasma membrane, resulting in basolateral exocytosis. Since monolayer TER is decreased at later infection times (33) and alterations in tight junction proteins have been observed at 24 to 48 h postinfection (8, 33), it seems reasonable to suggest that intercellular junction damage may occur at later times, following C. jejuni transcytosis across the intestinal epithelium.