|Home | About | Journals | Submit | Contact Us | Français|
Listeria monocytogenes (Lm) is a foodborne pathogen that crosses the intestinal barrier upon interaction between its surface protein InlA and its species-specific host receptor E-cadherin (Ecad). Ecad, the key constituent of adherens junctions, is typically situated below tight junctions and therefore considered inaccessible from the intestinal lumen. In this study, we investigated how Lm specifically targets its receptor on intestinal villi and crosses the intestinal epithelium to disseminate systemically. We demonstrate that Ecad is luminally accessible around mucus-expelling goblet cells (GCs), around extruding enterocytes at the tip and lateral sides of villi, and in villus epithelial folds. We show that upon preferential adherence to accessible Ecad on GCs, Lm is internalized, rapidly transcytosed across the intestinal epithelium, and released in the lamina propria by exocytosis from where it disseminates systemically. Together, these results show that Lm exploits intrinsic tissue heterogeneity to access its receptor and reveal transcytosis as a novel and unanticipated pathway that is hijacked by Lm to breach the intestinal epithelium and cause systemic infection.
The interaction of the Listeria monocytogenes (Lm) surface protein InlA with its receptor E-cadherin (Ecad), expressed by intestinal epithelial cells (IECs), is critical for the translocation of this human foodborne pathogen across the intestinal barrier and the onset of a systemic infection (Lecuit et al., 2001). InlA interaction with Ecad is species specific: human Ecad (hEcad) is a receptor for InlA, whereas mouse Ecad (mEcad) is not (Lecuit et al., 1999). We have shown that in hosts permissive to InlA–Ecad interaction, such as transgenic mice expressing hEcad and knockin mice expressing humanized mEcad, Lm directly invades enterocytes and crosses the intestinal epithelium (Lecuit et al., 2001; Disson et al., 2008). However, Ecad is an adherens junction protein typically located below tight junctions (TJs) and is therefore thought to be inaccessible to bacteria located in the intestinal lumen (Boller et al., 1985; Sousa et al., 2005). Lm has been reported to invade the tip of intestinal villi at sites of cell extrusion (Pentecost et al., 2006, 2010). An in vitro study in cultured epithelial kidney cells (MDCK cells) has shown that cell extrusion is associated with junction remodeling that transiently exposes Ecad on extruding cells and their immediate neighbors (Pentecost et al., 2006), and in vivo studies have confirmed that this process is accompanied by redistribution of apical junctional complex proteins (Madara, 1990; Marchiando et al., 2011). Nevertheless, apart from intestinal villus tips, the sites of Lm adhesion to and translocation across the intestinal epithelium in vivo have not been investigated systematically, and the molecular mechanisms that lead to Lm actual translocation across the intestinal epithelium and release into the lamina propria remain unknown.
In vitro studies have demonstrated that Lm enters into nonphagocytic cells within 10 min through a zippering process (Mengaud et al., 1996; Cossart and Sansonetti, 2004). Upon internalization, the bacterium is entrapped in a membrane-bound compartment, from which it is able to escape within 10–15 min (Myers et al., 2003; Henry et al., 2006). Vacuolar escape is mediated by the pore-forming activity of the secreted cholesterol-dependent listeriolysin O (LLO; Garcia-del Portillo and Finlay, 1995; Cossart and Sansonetti, 2004). Upon vacuolar lysis, Lm reaches the cytosol, in which it can polymerize host actin to form comet tails that propel bacteria intracellularly and form protrusions that allow Lm transfer into neighboring cells (see schematic representation in Fig. S6 A). This process is strictly dependent on the listerial surface protein ActA, which is necessary and sufficient on the bacterial side to mediate actin polymerization (Kocks et al., 1992). Actin nucleation around bacteria can be observed within 15 min after vacuolar escape, and Lm-containing protrusions can be observed ~2.5 h upon infection (Tilney et al., 1992).
The goal of our study was to determine comprehensively Lm intestinal epithelial target cells and to decipher the mechanisms of Lm crossing of the intestinal epithelial barrier in vivo. To identify Lm intestinal epithelial target cells, we first studied Ecad luminal accessibility by two-photon and confocal imaging of noninfected whole mount intestinal tissue of humanized transgenic mice permissive to InlA and investigated the sites of InlA–Ecad-dependent bacterial association with the intestinal epithelium. We demonstrate that Ecad is luminally accessible at discrete locations, which are (a) junctions between mucus-secreting goblet cells (GCs) and adjacent enterocytes, (b) extruding enterocytes at the tip of villi, as previously proposed by Pentecost et al. (2006, 2010), but also along their lateral sides, and (c) villus epithelial folds. We show that among GCs, those that have expelled their mucus content are those that typically exhibit luminally accessible junctional Ecad. Importantly, we identify GCs as the Lm preferential targets at the intestinal barrier level and establish a positive correlation between the number of GCs and the efficiency of Lm intestinal invasion and systemic dissemination. We further show that Lm translocation across the intestinal epithelium into the lamina propria mostly occurs below intestinal villus tips, is very rapid and efficient, and is strictly InlA dependent, whereas LLO and ActA play no role in this process. Finally, we show that Lm is rapidly transcytosed in a vacuole in a microtubule-dependent manner across enterocytes and egresses from them by exocytosis. Lm rapid InlA–Ecad-dependent translocation across the intestinal barrier results in a similarly rapid bacterial systemic dissemination.
Transcytosis is a so far unsuspected infection strategy for Lm, which is critical for its rapid and efficient crossing of the intestinal epithelium upon targeting of its receptor Ecad. This study provides, to our knowledge, the first in vivo elucidation of how an enteropathogen translocates across the intestinal epithelium into the lamina propria to cause systemic infection. These results illustrate the critical importance of extending cellular microbiology to three-dimensional tissue systems to understand the pathophysiology of infectious diseases.
To study the early stages of Lm intestinal crossing, we inoculated intestinal ligated loops of iFABP-hEcad transgenic mice, which express hEcad under the control of the iFABP promoter (Lecuit et al., 2001), with Lm and imaged whole mount small intestinal tissue by two-photon and confocal microscopy. Noninfected small intestinal tissues immunolabeled for hEcad display a typical epithelial honeycomb-like pattern (Fig. 1 A, Fig. S1 A, and Video 1). In infected intestinal loops, as early as 30 min postinoculation (p.i.), bacteria were associated with intestinal villi (Fig. 1 B): we detected Lm adhering to the apical surface of enterocytes, within enterocytes, and at the basal pole of enterocytes. To our surprise given this early time point, we also detected Lm in the lamina propria (Fig. 1 C and Video 2). Because of the classical intracellular cycle of Lm and its time frame in vitro, we would have anticipated that Lm intestinal translocation in the lamina propria would require at least 45 min to occur. We thus performed a kinetic analysis of Lm invasion of the intestinal tissue and studied the contribution of InlA to this process. Using two-photon and confocal imaging, we compared the association of WT Lm to that of its ΔinlA isogenic mutant with intestinal villi 5, 15, 30, and 45 min p.i. We observed a time- and InlA-dependent Lm association to intestinal villi (Fig. 1 D); as early as 15 min p.i., Lm was significantly associated with intestinal villi in a strict InlA-dependent manner. We further quantified bacteria located (a) on the villus surface, (b) in epithelial cells, and (c) in the lamina propria, along the villus vertical axis 30 min p.i. We observed that most bacteria were 40–80 µm away from the villus tip and were located in the lamina propria (Fig. 1 E). We conclude that Lm translocation across the intestinal barrier into the lamina propria is strictly InlA dependent, occurs all along the villus z axis, and is particularly rapid.
Because Lm early translocation across the intestinal barrier is InlA–Ecad dependent and occurs along the z axis of intestinal villi, we investigated Ecad accessibility from the luminal side of uninfected intestinal epithelium of iFABP-hEcad transgenic mice. We first performed a double staining, in which luminally accessible hEcad was stained before tissue permeabilization and total hEcad was stained after tissue permeabilization with HECD-1, an mAb raised against extracellular domain of hEcad that does not recognize mEcad (Fig. 2 A). Accessible hEcad was observed at the surface of junctional complexes as well as along the lateral sides of the cells (Fig. 2, A and B, insets; Fig. 3 A, B, and D; and Video 3). We next investigated the precise location of accessible hEcad on villi. We observed that hEcad is luminally accessible at the tip of villi (Fig. 2 B, arrow; see other examples in Fig. S4 A, insets; and Video 3), as previously proposed (Pentecost et al., 2006). Importantly, we also detected luminally accessible hEcad at distinct spots along the longitudinal axis of intestinal villi (Fig. 2 B, insets; see other examples in Fig. S4, E and I, insets; and Video 3). Similar results were obtained when total and/or accessible endogenous mEcad was labeled in transgenic and WT C57BL/6 mice with ECCD2, an mAb raised against the extracellular domain of mEcad (Fig. S1, B and C), demonstrating that Ecad accessibility does not result from hEcad transgenic expression.
We next determined the relative number of the IECs that express luminally accessible hEcad and found that ~2% of total IECs fall into this category. To determine the cell subtypes that express accessible hEcad, we costained intestinal villi for luminally accessible hEcad and wheat germ agglutinin (WGA). WGA specifically binds to sialic acid and N-acetylglucosaminyl carbohydrate residues present on the plasma membrane of enterocytes and in mucus of GCs (Fischer et al., 1984; Jang et al., 2004). We observed that among enterocytes, which were WGA positive (WGA+), as expected, some expressed luminally accessible hEcad and were identified as either extruding apoptotic cells or cells located within intestinal epithelial folds (Fig. 2 B, arrow; Fig. S4, A, E, and I; and Videos 3, 8, and 9). Importantly, we also identified isolated WGA–highly positive cells (WGAhi) expressing luminally accessible hEcad (Fig. 2 B, inset; Fig. 3, A, B, and D; Fig. S1 C; Fig. S3 A; and Videos 3 and 4). These WGAhi cells exhibited a weaker cortical actin cytoskeleton compared with WGA+ cells (Fig. S2 A, arrows) and were strongly immunolabeled with antibodies against Mucin 2 (Muc-2) and intestinal trefoil factor (ITF; Fig. S2, B and C), identifying them unambiguously as GCs (Poulsom and Wright, 1993; van Klinken et al., 1999). We also observed that hEcad is enriched along the lateral membranes of GCs compared with neighboring enterocytes (Fig. 3 D), whereas accessible hEcad was not detectable at the apical dome of GCs, where Ecad-negative mucin granules fuse with the apical membrane and exocytose their mucus content (Fig. 3, A, B, and D).
We next quantified the relative proportion of GCs and non-GCs that display luminally accessible Ecad: GCs represent 2.8% of the total IECs (Fig. 2, C and D) but account for half of the cells that exhibit accessible hEcad (Fig. 2 E), demonstrating a significant overrepresentation of GCs among cells expressing accessible hEcad; in fact, up to one third of GCs (32%) express luminally accessible hEcad, whereas only a very small proportion (1%) of non-GCs do (Fig. 2 F). Accessible hEcad immunolabeling in the presence of Lm was similar to uninfected intestinal tissue (Fig. S2 D), indicating that hEcad luminal accessibility is not induced by Lm but rather reflects physiological tissue heterogeneity.
Cell–cell junctions involving GCs have been described as heterogeneous and relatively weaker than those between classical enterocytes (Madara et al., 1980; Madara and Trier, 1982). Moreover, cell shape changes that GCs undergo during mucus secretion have been associated with the loosening of their TJs (Hull and Staehelin, 1976; Porvaznik et al., 1983). To determine whether mucus secretion might provide luminal access to hEcad, we quantified the amount of mucus-expelling GCs that express luminally accessible hEcad. 87% of GCs that express luminally accessible hEcad were secreting or had secreted their mucus content (mucus-expelling GCs; Fig. 3 A). Because TJ disorganization might provide luminal access to hEcad, we investigated their integrity in GCs by studying the localization of two major TJ components, ZO-1 and occludin. In GCs in which hEcad was accessible, both markers exhibited a pattern distinct from classical absorptive epithelial cells, with a thicker occludin signal (Fig. 3 B, Fig. S3 A, and Video 5) and either a thicker ZO-1 signal in cells that had expelled their mucus content (Fig. 3 B) or a fainter signal in cells full of mucus (Fig. S3 B). PKCζ and PAR-3, together with PAR-6, are known to localize at epithelial junctional complexes and have been implicated in the recruitment of junction proteins at the cell–cell interface and in the establishment and maintenance of cell polarity in Drosophila melanogaster and mammalian epithelial cells (Izumi et al., 1998; Ohno, 2001; Hirose et al., 2002). As expected, we observed that PAR-3 and PKCζ were recruited at junctional complexes of absorptive epithelial cells (Fig. 3 C; Fig. S3, C and D; and Videos 6 and 7). In contrast, in GCs, we observed a striking depolarization of PAR-3, which was detectable all along GCs (Fig. 3 C, asterisks; Fig. S3 C; and Video 6), as well as a disappearance of PKCζ accumulation at junctional complexes (Fig. 3 C, Fig. S3 D, and Video 7).
Together, these results indicate that TJs of GCs are disorganized compared with classical TJs on absorptive cells. These structural features of GCs, together with the extensive cell shape changes occurring during mucus secretion, likely account for luminal accessibility of Ecad in mucus-expelling GCs.
As with GCs, non-GCs expressing luminally accessible hEcad exhibited a relative redistribution of TJ proteins and cortical actin network when stained for occludin, ZO-1, and PKCζ (Fig. S4, B–D, F–H, and J–L; and Table S1) and F-actin (Fig. S4 M), respectively. These observations fit with a redistribution of apical junctional complex proteins and an enrichment of Ecad during apoptotic cell shedding at the tip of intestinal villi, as previously shown (Fig. S4, A–H, and N; and Video 8; Marchiando et al., 2011), and with a relative exacerbated tension exerted on cell–cell junctions within villus epithelial folds (Fig. S4, I–L; and Video 9).
Having identified GCs as strongly overrepresented among cells expressing luminally accessible Ecad (Fig. 2 F), we next investigated Lm preferential cell targets at the intestinal level. To this end, we distinguished three cell categories: (a) GCs, (b) cells next to GCs (NGCs; i.e., cells forming junctions with GCs), and (c) cells away from GCs (AGCs; i.e., all other IECs; Fig. 4 A).
The relative proportions of GCs, NGCs, and AGCs to total IECs per villus were 2.8%, 8.5%, and 89.5%, respectively (Fig. 4 B). To assess whether Lm preferentially targets one cell subtype and to investigate the importance of accessible hEcad in Lm association and subsequent translocation, we evaluated the number of WT Lm that were associated with each of these three cell categories, in conjunction with the presence of accessible hEcad on these cells and in comparison with the ΔinlA Lm isogenic mutant. 40 min p.i., we observed that Lm adhered to each of these three cell categories expressing accessible hEcad (Fig. 4 A, top) and that Lm was translocated from the apical to the basal pole of these three cell categories (Fig. 4 A, middle and bottom). We then determined the preference of Lm for each cell type by calculating the number of Lm per cell, relative to the respective abundance of each cell type (see Materials and methods). We show that although GCs represent a very low proportion of intestinal cells (2.8%; Fig. 4 B), WT Lm interacts >20 times more with GCs than with AGCs (Fig. 4 C, left). In addition, the InlA–Ecad interaction provides a >15-fold Lm preference for GCs compared with the ΔinlA mutant, which demonstrated no cellular preference (Fig. 4 C, right). Finally, WT Lm relative associations per NGC were as low as with AGCs. Overall, these results demonstrate that Lm preferentially targets accessible Ecad-expressing GCs on the epithelium of intestinal villi.
To further reinforce the critical role of GCs in Lm crossing of the intestinal barrier, we next investigated how variation in the number of GCs would influence Lm intestinal tissue invasion and systemic dissemination. We treated hEcad transgenic mice with IL-33, which is well characterized as inducing an increase in the number of GCs and mucus secretion (Moro et al., 2010; Neill et al., 2010). As expected, IL-33–treated mice exhibited an increase in the number of GCs per villus, as well as an increase in the amount of mucus produced (Fig. 5, A and C). Strikingly, this increase in GCs correlated with an increase in Lm invasion of intestinal villi (Fig. 5, B and C) and systemic dissemination to the spleen (Fig. 5 D). Increased intestinal invasion was not observed with the ΔinlA mutant (Fig. 5 B). Together, these results highlight a positive correlation between the number of GCs and Lm intestinal tissue invasiveness and systemic dissemination.
Having identified how and where Lm interacts with the intestinal barrier in vivo and shown that its transfer into the lamina propria is particularly rapid, we next investigated the molecular mechanisms underlying Lm transepithelial translocation. We infected intestinal loops of hEcad transgenic mice with 109 Lm or an isogenic ΔinlA mutant, a Δhly mutant (not expressing LLO), a ΔactA mutant, a nonpathogenic and noninvasive Listeria innocua (Li), or Li-expressing InlA (Li(inlA)). We quantified the amount of bacteria located in the intestinal lamina propria 30 min upon infection. We observed no translocation to the intestinal lamina propria with the ΔinlA deletion mutant or, as expected, with the noninvasive Li (Fig. 6 A). In contrast, Li(inlA) translocated rapidly into the lamina propria in equal numbers to Lm (Fig. 6 A). Strikingly, the Δhly and ΔactA deletion mutants were able to invade and cross the intestinal epithelium as efficiently as WT Lm. This demonstrates that the rapid translocation of Lm across the intestinal barrier is strictly dependent on the InlA–Ecad interaction, whereas the major virulence factors LLO and ActA are not required (Fig. 6 A), showing that the classical cell infection cycle of Lm is not at play in this process (Fig. S6 A). Furthermore, as early as 30 min p.i., Lm disseminated into deeper organs, such as the spleen (Fig. 6 B), in an InlA-dependent but LLO- and ActA-independent manner (Fig. 6 B), clearly demonstrating the significance of the InlA-dependent but LLO- and ActA-independent rapid intestinal translocation for the onset of systemic listeriosis.
The demonstration that LLO and ActA are dispensable for Lm translocation across the intestinal epithelium led us to formulate the hypothesis of a rapid apicobasal transcytosis of Lm-containing vacuole and the exocytosis of Lm at the basal pole of enterocytes in the lamina propria. We first investigated the role of a critical component of the transcytosis machinery, the microtubule network, in Lm translocation across the intestinal barrier using nocodazole, which prevents microtubule polymerization, and colchicine, which leads to dynamic instability and rapid microtubule depolymerization (De Brabander et al., 1976; Lee et al., 1980; Ravelli et al., 2004). Intestinal ligated loops were infected with 109 Lm for 10 min. Nocodazole or colchicine was then added and left in contact with the bacteria for 50 min. In nocodazole- as well as in colchicine-treated tissues, Lm resided within IECs (Fig. 6, C and D). Under these experimental conditions, Lm internalization into enterocytes was not affected, and the integrity of adherens junctions and TJs was preserved (Fig. S5 A). Yet microtubule depolymerization inhibited Lm translocation across the intestinal epithelium. In fact, in nocodazole- and colchicine-treated loops, the amount of bacteria associated with enterocytes was four times higher than in untreated loops (Fig. 6 C). In contrast, taxol, which inhibits disassembly of microtubules and stabilizes the microtubule network (Schiff and Horwitz, 1980), did not reduce Lm rapid translocation (unpublished data).
To investigate the role of vesicle fusion machinery in Lm release from its vacuole into the lamina propria, we investigated the contribution of N-ethylmaleimide–sensitive factor (NSF) during infection of intestinal ligated loops. We used a fusion polypeptide (TAT-NSF700), composed of an HIV transactivating regulatory peptide derived from the TAT protein domain fused to a peptide derived from the D2 domain of NSF, which binds ATP and regulates NSF hexamerization. TAT-NSF700 has been shown to cross cellular membranes, inhibit NSF hydrolysis of ATP, decrease NSF disassembly of SNARE molecules, and disrupt vacuolar trafficking including exocytosis (Matsushita et al., 2005; Jahn and Scheller, 2006). We pretreated intestinal ligated loops with TAT-NSF700 or with a control peptide, TAT-NSF700scr, for 30 min. Entry of TAT-NSF700 peptide in intestinal epithelium was verified using a TAT-NSF700 peptide conjugated to cyanin 3 (Fig. S5 B). This treatment led to an intracytosolic accumulation of vesicles, which is consistent with the inhibition of the fusion machinery by TAT-NSF in enterocytes (Fig. 6 G). We then injected into the loops 109 Lm in the presence of each of the synthetic peptides. By two-photon microscopy of whole mount tissue, we determined and quantified Lm association with the epithelial sheet and its translocation into the lamina propria 45 min p.i. In TAT-NSF700–treated tissue, Lm translocation into the lamina propria was strongly inhibited, and bacteria remained within enterocytes, at the level of their basal pole. In untreated loops and in loops treated with TAT-NSF700scr, bacteria were seen both in the lamina propria and within enterocytes (Fig. 6, E and F). Importantly, entry of Lm was not affected by TAT-NSF700 treatment because total Lm number per villus was not significantly different in PBS-, TAT-NSF700–, or TAT-NSF700scr–treated tissue. Finally, by using transmission electron microscopy (TEM), we observed Lm enclosed in a vacuole below the cell nucleus of an IEC and in the vicinity of its basal pole in a TAT-NSF–treated tissue (Fig. 7 C). Together, these results show that disruption of vacuolar trafficking including exocytosis impairs Lm translocation into the lamina propria.
We next attempted to colocalize translocating Lm with its surrounding membrane vacuole. We first used an antibody raised against the cytoplasmic domain of Ecad, which is expected to decorate the cytoplasmic side of Lm-containing vacuole. We observed Lm colocalizing with the cytoplasmic domain of Ecad (Fig. 7 A). To directly visualize vacuolar membrane around Lm, we crossed our mice expressing humanized Ecad with a transgenic mouse line expressing a membrane-targeted Tomato protein (mtd Tomato), in which all membranes are constitutively fluorescently labeled in red (Disson et al., 2008; Muzumdar et al., 2007). Infection of hEcad × mtd Tomato mice allowed the visualization of red membranes surrounding Lm during its apicobasal translocation, whereas Lm was no longer associated with a red signal once released in lamina propria (Fig. 7 B). These data confirm that Lm is transferred in the lamina propria in a membrane-bound vacuole. Collectively, these results show that Lm is transferred apicobasally in a microtubule-dependent manner, is enclosed in a membrane-bound vacuole, and egresses the cell by exocytosis. This process can thus be referred to as transcytosis.
Microbial crossing of host barriers is a critical step of invasive infections. Understanding how microbes cross the intestinal barrier is thus key to comprehend the pathophysiology of foodborne infections. Lm is a model microorganism for which a great deal of knowledge has been obtained by combining in vitro and in vivo approaches. Yet how this foodborne pathogen actually crosses the intestinal barrier has remained elusive. Pentecost et al. (2006, 2010) have reported that Lm targets intestinal villus tips to invade the epithelium at sites of cell extrusion and, based on in vitro observation obtained in cultured kidney epithelial cells, proposed that these were sites of accessible Ecad. Nevertheless, the existence of other sites of intestinal translocation as well as the precise molecular mechanism by which Lm is actually transferred across the intestinal barrier monolayer have so far not been addressed. This prompted us to set up an experimental approach based on the comprehensive imaging of uninfected and infected intestinal tissues of iFABP-hEcad transgenic mice. This approach preserves the intrinsic cell heterogeneity in a tissue context, allowing us to perform a detailed study of the initial steps of Lm invasion and of the mechanisms of translocation across the intestinal epithelium with a level of scrutiny usually reserved for cultured cells. We demonstrate that Ecad is luminally accessible at cell–cell junctions between mucus-expelling GCs and adjacent enterocytes all along the vertical axis of intestinal villi. We show that Lm accesses the intestinal lamina propria mostly all along the villus longitudinal axis and mainly targets luminally accessible Ecad of GCs to adhere to and invade the intestinal epithelium in an InlA-dependent manner. We demonstrate that Lm is then rapidly translocated from the apical to basolateral side of IECs enclosed in a vacuole. This translocation is microtubule dependent but independent of the virulence factors LLO and ActA. Furthermore, the exocytic machinery is involved in Lm release in the lamina propria. Finally, Lm rapid transepithelial transcytosis leads to a similarly rapid systemic dissemination. Together, these results demonstrate that Lm is transcytosed across the intestinal epithelium (Fig. S6 B), thus revealing a novel and unsuspected pathway hijacked by Lm that is crucial for its rapid dissemination within the host.
We have identified novel in vivo sites of accessible Ecad around GCs, at cell–cell junctions with their neighbors, and along villus epithelial folds. Moreover, as previously reported in vitro (Pentecost et al., 2006, 2010), we found that Ecad is luminally accessible at cell extrusion sites. In a similar fashion as extruding cells, mucus-producing GCs undergo extensive shape changes such as stretch and compression during mucus secretion. These shape changes are accompanied with junctional reorganization, notably variations in length and strand cross-linking and discontinuities (Hull and Staehelin, 1976; Madara and Trier, 1982; Porvaznik et al., 1983; Madara, 1990). Accordingly, we demonstrate here that TJs are disorganized between cells on which Ecad is luminally exposed and establish a correlation between mucus secretion and luminal access to Ecad. Furthermore, cortical actin at the apex of GCs is greatly reduced compared with IECs (Oliver and Specian, 1990). We indeed observed a reorganized apical actin network at sites with accessible Ecad (Fig. S4 M), a result which is consistent with the well-established importance of the actin cytoskeleton in maintaining cell junction integrity by stabilizing Ecad clusters (Hagen and Trier, 1988; Cavey et al., 2008; Schill and Anderson, 2009). Finally, we also show that Ecad is luminally accessible in villus intestinal folds (Fig. S4 I and Video 9). The forces of tension and constriction generated at these sites induce cell shape changes that may account for the relocation of Ecad to the apical luminal side, as previously described for adherens junction protein remodeling occurring during Drosophila gastrulation and ventral furrow formation (Dawes-Hoang et al., 2005; Kölsch et al., 2007). Altogether, our investigations point to a correlation between physical tensions, localization of adherens junction components and TJ remodeling at the intestinal epithelium level. Lm targeting of accessible Ecad illustrates how a microbial pathogen can take advantage of tissue dynamics and heterogeneity in vivo, a facet of microbial pathogenesis which cannot be investigated in vitro in more static and homogenous systems.
We have demonstrated Lm preferential targeting of GC accessible Ecad and a correlation between GC numbers and Lm intestinal invasion and systemic dissemination. Even though mucus secretion is at the front line of the innate host defense in the gastrointestinal tract, Lm seems to exploit this innate defense mechanism by binding in an InlA-dependent manner to GCs. Moreover, although GCs are specialized in basal-apical transport, they also possess an endocytic and apicobasal transcytosis activity that Lm may hijack (Colony and Specian, 1987). Thus, Lm takes advantage of an intestinal defense mechanism, namely mucus secretion, to target and cross the epithelium by transcytosis, uncovering an unsuspected Achilles’ heel of the innate immune system that Lm has evolved to take advantage of. Nevertheless, we have also observed that Lm interacts with intestinal GCs not expressing accessible Ecad, although this is far less frequent than with GCs that express accessible Ecad (Fig. 4 C). It has been previously reported that other internalins of Lm, such as InlB, InlC, and InlJ, contain mucin-binding domains able to bind secreted Mucin-2 (Lindén et al., 2008). However, they do not bind to the membrane-bound Mucin-1, suggesting that Lm attraction to GCs in vivo is not caused by Inl–Muc interactions, but rather results from InlA–Ecad interactions. However, Inl–Muc interactions may retain Lm within the mucus layer in an environment close to their source, the GCs, thus facilitating InlA–Ecad-mediated bacterial adhesion. This is further supported by the >15-fold increase of relative infection of accessible Ecad-expressing GCs compared with ΔinlA interactions with GCs (Fig. 4 C).
In vitro studies have shown that bacteria can transcytose across cultured cell monolayers, including Campylobacter jejuni (Grant et al., 1993; Russell and Blake, 1994; Bacon et al., 2001; Kopecko et al., 2001; Hu et al., 2008; Watson and Galán, 2008), Neisseria gonorrhoeae and Neisseria meningitidis (Makino et al., 1991; Merz and So, 2000; Kuespert et al., 2006; Wang et al., 2008; Sutherland et al., 2010), Streptococcus pneumoniae (Zhang et al., 2000; Kaetzel, 2001), and Escherichia coli (Burns et al., 2001; Xie et al., 2004). Yet, with the exception of bacterial sampling by M cells (see next paragraph), this is the first time to our knowledge that a bacterium has been shown to translocate across an epithelial layer by transcytosis in vivo, resulting in its systemic dissemination. That Lm follows a transcytosis pathway rather than its classical infection cycle was unexpected, as Lm is considered a prototypic intracytosolic pathogen (Fig. S6 A). Yet, having evolved such a furtive means to cross the intestinal barrier seems perfectly in line with the observed rapidity of Lm transepithelial transfer, the absence of intraepithelial bacterial proliferation and epithelial damage during the intestinal phase of listeriosis, and the fact that actin-based motility directionality does not preferentially route Lm to the basal pole of enterocytes. But why doesn’t Lm escape its vacuole at the intestinal epithelium level to cross the intestinal epithelium? It could be that transcytosis is too rapid to allow LLO-dependent Lm escape from its internalization vacuole. In addition, it has been recently shown in macrophages that LLO activity is activated by GILT (Singh et al., 2008). GILT is constitutively expressed in antigen-presenting cells and is inducible by IFN-γ in other cell types (Arunachalam et al., 2000). Because GILT is not constitutively expressed in enterocytes, it may not activate LLO early on in the infection, thus allowing the Lm-containing vacuole to be transcytosed apicobasally. Later on, as IFN-γ expression is triggered, GILT induction would lead to the activation of LLO, allowing Lm to be diverted from transcytosis and delivered into the cytosol. In line with this hypothesis, intracytosolic Lm can be observed in vivo in enterocytes later on in the infection (Lecuit et al., 2001; Pentecost et al., 2006, 2010).
We also show here that inhibition of NSF blocks the release of Lm in the lamina propria and consequently its exit from epithelial cells. NSF and soluble NSF attachment proteins (SNAPs) are essential components of the intracellular membrane fusion apparatus (Jahn and Scheller, 2006). This suggests that Lm-containing vacuoles fuse with the basolateral membrane and exocytose, releasing Lm subepithelially. There are only two in vitro studies illustrating bacterial exocytosis, by means of electron microscopy pictures of C. jejuni (Hu et al., 2008; Baker and Graham, 2010), but the putative role of the exocytic machinery in bacterial escape has not been addressed in vivo in the context of a bacterial infection. Lm seems to exploit the full range of Ecad recycling, not only by mimicking its natural ligand Ecad, binding to it and inducing its internalization, but also by subverting its basolateral recycling in polarized epithelial cells by transcytosis and exocytosis.
Lm transcytosis is strictly InlA dependent and LLO independent, whereas we have shown that the intestinal host response to Lm is InlA independent and strictly LLO dependent (Lecuit et al., 2007). We can interpret these findings as follows: at the intestinal villus level, in which Lm enters via InlA without inducing tissue damage, the bacterium follows a furtive intravacuolar path that avoids detection by cytosolic sensors and remains undetected in the lamina propria, possibly because of its extracellular location and its ability to escape innate immune system recognition (Boneca et al., 2007; Gouin et al., 2010; Personnic et al., 2010). In contrast, after translocation through M cells in an InlA-independent manner (Pron et al., 1998; Lecuit et al., 2001; Corr et al., 2006; Chiba et al., 2011), Lm is phagocytosed by antigen-presenting cells, where LLO mediates vacuolar rupture, switching on the host response. This scenario illustrates how Lm has evolved mechanisms that allow its dissemination into the host via the intestinal lamina propria without inducing, at least initially, strong innate immune responses. This is in sharp contrast with enteropathogens such as Shigella flexneri that are contained at the intestinal level to the expense of strong and destructive inflammation responses.
Lm is a model microorganism that has been particularly instrumental for key discoveries in the fields of microbiology, cell biology, and immunology. By extending the cellular microbiology approach to tissues and by using high-resolution tissue-imaging techniques, we have demonstrated that the well-established dogma of Lm access to the cytosol for host tissue invasion does not apply to its early, rapid, and efficient crossing of the intestinal barrier. These results not only revisit and enrich our understanding of Lm pathogenesis, but also provide new clues as to how microbial pathogens breach host barriers. Future investigations will have to focus on the mechanisms of Lm breaching of the placental and blood-brain barriers and also on how other microbial pathogens cross host barriers.
Bacterial strains used in this study are Lm strain EGD (BUG 600), its isogenic deletion mutants ΔinlA (BUG 947), Δhly (BUG 1954), and ΔactA (BUG 2140), and Li WT (BUG 499) and Li-expressing InlA (BUG 1489). The inocula used are shown in figure legends and are 1–3 × 109 per intestinal ligated loop. Bacteria were cultured as described previously (Disson et al., 2009).
We used transgenic mice expressing hEcad on a C57BL/6 genetic background (Lecuit et al., 2001). To generate mice expressing both a membrane-targeted Tomato protein and the humanized E16P Ecad, mT/mG mice were crossed with E16P KI mice (Charles River; Muzumdar et al., 2007; Disson et al., 2008). Animal experiments were performed according to the Institut Pasteur guidelines for laboratory animals’ husbandry. For intestinal ligated loops, 8–12-wk-old mice were fasted for 16 h before surgery. After mild anesthesia of mice with 2.5% (vol/vol) vaporous isoflurane (Aerrane; Baxter), deep anesthesia was induced with a mix of ketamine (50 mg/kg body weight; Imalgene 1000; Merial) and medetomidine (0.5 mg/kg body weight; Domitor; Orion Corporation). A laparotomy was performed, the small intestine was exposed, and jejunal loops of 3 cm long were prepared. 200 µl of inoculum was injected into each loop. At the planned endpoint, the animal was euthanized, and intestinal loops were harvested, opened longitudinally, washed repeatedly in 37°C Dulbecco’s modified Eagle’s medium (Invitrogen), and fixed flat (luminal side up) in 35 × 10–mm tissue culture dishes (BD) containing 4% agarose (Sigma-Aldrich). For IL-33 administration, 400 ng of recombinant mouse IL-33 (R&D Systems) in PBS were administered daily for 3 d intraperitoneally. Intestinal ligated loops were performed 24 h later after the last injection. Control animals received PBS only.
All the procedures were in agreement with the guidelines of the European Commission for the handling of laboratory animals, directive 86/609/EEC (http://ec.europa.eu/environment/chemicals/lab_animals/home_en.htm) and were approved by the Animal Care and Use Committee of the Institut Pasteur.
For CFU enumeration, the organs were retrieved and grounded, and serial dilutions were plated on BHI (brain heart infusion) agar plates as previously described (Disson et al., 2009).
For microtubule depolymerization and inhibition of polymerization, 10 µg/ml nocodazole and 10 µg/ml colchicine (Sigma-Aldrich) were used per ligated loop (Mostov et al., 1984; Ameen et al., 2003; Wang et al., 2008). Loops were initially infected for 10 min and, subsequently, nocodazole or colchicine was added and kept along the infection (50 min). For exocytosis inhibition, we used 20 µg/ml of the recombinant NSF peptide TAT-NSF700 (AnaSpec). This peptide has been described to be the most potent peptide to enter cells and inhibit NSF-mediated ATPase activity and SNARE complex disassembly. TAT-NSF700scr (AnaSpec) is a control peptide that does not inhibit NSF activity (Matsushita et al., 2003, 2005). The TAT-NSF700 peptide conjugated to cyanin 3 (RD-Biotech) was used to verify the entry of the peptide in cells.
Tissues were fixed in 4% paraformaldehyde (Electron Microscopy Sciences) in PBS for 1 h at room temperature under agitation and either stained directly (whole mount) or embedded in 4% low-melting-point agarose (Sigma-Aldrich) for 100–250-µm-thick vibratome sections (Micro HM 650V; Thermo Fisher Scientific; Appleton et al., 2009). Sections were stained in a similar manner to whole mount tissue but mounted in fluorescent mounting medium (Fluoromount; Interchim). Tissues were permeabilized for 1 h in 0.4% Triton X-100 (Sigma-Aldrich) and 3% PBS-BSA (Sigma-Aldrich) or preincubated for 1 h in the blocking buffer 3% PBS-BSA. Then tissues were labeled with the appropriate primary and secondary antibody overnight at 4°C and 2 h at room temperature, respectively.
To label accessible Ecad of uninfected tissues, tissues were directly recovered without intestinal ligation to avoid ischemic and epithelial damages, fixed in 4% paraformaldehyde as described in the previous paragraph, and labeled. To detect accessible hEcad, we used HECD-1, a specific mouse mAb raised against the extracellular domain of hEcad (Invitrogen). To detect accessible mEcad, we used ECCD-2, a specific rat mAb raised against the extracellular domain of mEcad (Invitrogen). We used an antibody against the cytoplasmic domain of Ecad (36/Ecad mouse mAb; BD) to show the association of Lm with a membrane vacuole because internalized Ecad decorates the vacuolar membrane.
For Sytox staining, 1 µM Sytox green (Invitrogen) was injected into an intestinal ligated loop for 30 min. The tissue was then harvested, fixed, and processed as described above for further staining.
For immunofluorescence of cryosections, retrieved tissues were fixed and embedded in OCT compound at −80°C (Tissue-Tek), cut using a cryostat at −20°C (7-µm-thick sections), postfixed in pure ethanol, and labeled with the appropriate primary and secondary antibody for 1 h at room temperature and 2 h at room temperature, respectively. Sections were mounted in fluorescent mounting medium (Fluoromount).
Samples were imaged with an upright LSM510 two-photon META (Carl Zeiss) or with an LSM700 confocal microscope (Carl Zeiss) with a 40× water immersion objective. Image analysis was performed using ImageJ (National Institutes of Health), LSM 5 Image Browser (Carl Zeiss), and Photoshop software (Adobe). Three-dimensional reconstructions were performed using Imaris 5.5.3 software (Bitplane).
Infected intestinal tissues were washed in PBS and fixed in 2.5% glutaraldehyde–0.1 M cacodylate buffer solution overnight at 4°C. Fixed tissues were washed three times in cacodylate buffer and then fixed for 1 h at room temperature in 1% OsO4 in the same buffer. Fixed tissues were washed once in cacodylate buffer and twice in distilled water. After dehydration in an ethanol-graded series and in propylene oxide, tissues were embedded in epoxy resin. Specimens were cut using an ultramicrotome Reichert Ultracut S. 60-nm sections were picked up on Formvar regular mesh grids, contrasted with uranyl acetate and lead citrate, and observed on a JEM 1010 (JEOL) under standard conditions at 60 kV.
Ligated loops were fixed 40 min p.i. and labeled with WGA and for accessible hEcad and Lm. The relative proportions of GCs, NGCs, and AGCs were calculated (see Results). The amount of Lm per cell for each cell subtype was determined, taking into account the accessibility of hEcad. We finally calculated the number of Lm per cell type relative to the respective abundance of each cell type and normalized to Lm associations with AGCs (Fig. 4 C).
The following pAbs, mAbs, and other fluorescent probes were used for immunolabeling of tissues: anti–hEcad clone HECD-1 mouse mAb (Invitrogen), anti-Ecad clone 36/Ecad mouse mAb (BD), anti–mEcad clone ECCD-2 rat mAb (Invitrogen), anti–ZO-1 rabbit pAb (Invitrogen), anti-occludin rabbit pAb (Invitrogen), anti-PARD3A (H-70) rabbit pAb (Santa Cruz Biotechnology, Inc.), anti–PKCζ (C-20) rabbit pAb (Santa Cruz Biotechnology, Inc.), anti-ITF (M-18) goat pAb (Santa Cruz Biotechnology, Inc.), anti–Mucin 2 (R-12) goat pAb (Santa Cruz Biotechnology, Inc.), WGA conjugated with Alexa Fluor 555, Alexa Fluor 647, or Alexa Fluor 488 (Invitrogen), Alexa Fluor 647 or Alexa Fluor 488 phalloidin (Invitrogen), Hoechst 33342 (Invitrogen), Alexa Fluor 488 or Alexa Fluor 546 goat anti–rabbit, Alexa Fluor 488 or Alexa Fluor 546 goat anti–mouse, Alexa Fluor 488 goat anti–rat, donkey anti–goat-FITC, donkey anti–goat-rhodamine, and TUNEL (in situ cell death detection kit fluorescein; Roche).
Values are expressed as mean ± SD. Statistical comparisons were made using two-way analysis of variance with Tukey adjustment for multiple comparisons and timed data analysis (Olsen, 2003) or using a Student’s t test for single comparisons.
Fig. S1 shows that Ecad accessibility is an intrinsic property of intestinal villus epithelium. Fig. S2 shows that WGAhi cells are GCs, lack apical actin cytoskeleton, and are stained for GC-specific markers ITF and Muc-2. Fig. S3 shows additional pictures of GC disorganized TJs. Fig. S4 shows the characterization of non-GC subpopulations expressing luminally accessible Ecad. Fig. S5 shows that nocodazole treatment does not affect adherens junction and TJ integrity. Fig. S6 represents the Lm cell infectious process in vitro and in vivo. Videos 1 and 2 show three-dimensional reconstructions of a noninfected and an Lm-infected intestinal villus. Videos 3 and 4 show accessible Ecad on an intestinal villus and on a GC. Videos 5–7 show TJ disorganization on GCs. Videos 8 and 9 show accessible Ecad on non-GC subpopulations. Table S1 summarizes TJ characteristics around cells expressing luminally accessible Ecad. Online supplemental material is available at http://www.jem.org/cgi/content/full/jem.20110560/DC1.
We thank the members of the Microbes and Host Barriers Group for their support and Matteo Bonazzi and the personnel of the Imagopole Facility for their help and support. We thank Gerard Eberl for helpful discussion and advice.
Georgios Nikitas received a Marie Curie Early Stage Research Training fellowship from the Intrapath program, EU FP6. We thank Institut Pasteur, Institut National de la Santé et de la Recherche Médicale, Fondation pour la Recherche Médicale, Ville de Paris, Fondation BNP Paribas, and the European Research Council for financial support.
The authors declare that they have no competing financial interests.