Cooperation of Salmonella pathogenicity islands 1 and 4 is required to breach epithelial barriers


*E-mail; Tel. (+49) 9131 8523 640; Fax (+49) 9131 8522 531.


Invasion is an important microbial virulence strategy to overcome the barrier formed by polarized epithelial cells. Salmonella enterica is a food-borne pathogen that deploys a type III secretion system for the manipulation of the actin cytoskeleton and to trigger internalization into epithelial cells. Here we show that this function is not sufficient to enter polarized cells and report that penetration of epithelia from the luminal side requires both the type III secretion system and novel virulence functions conferred by Salmonella pathogenicity island 4. Salmonella pathogenicity island 4 encodes a type I secretion system for the giant non-fimbrial adhesin SiiE that mediates intimate contact of Salmonella to microvilli on the apical membrane. Mutant strains lacking SiiE fail to invade polarized cells, to destroy epithelial barrier functions and to efface the apical brush border. Deletion analyses of repetitive domains in SiiE indicate that the large size of the adhesin is of functional importance. Our observations demonstrate that efficient penetration of epithelial barriers requires the cooperative activity of two Salmonella pathogenicity islands encoding different secretion systems. These findings underline the role of the epithelial brush border and reveal a new mechanism used by bacterial pathogens to overcome this barrier.


The intestinal epithelium forms a unique barrier that separates the rich microbial flora of the intestine from the sterile underlying tissue. This cell layer is also an efficient barrier for the protection against infectious agents (for schematic representation, see Sansonetti, 2004). Epithelial cells within the intestinal mucosa have a polarized organization with an apical side facing towards the intestinal lumen and a basolateral side (Neutra et al., 2001). These functional entities are efficiently separated by tight junctions (TJs). In addition, there is a distinct distribution of receptors, transporters or patterns of glycosylation on the apical and basolateral side of polarized cells. One of the most remarkable characteristics specific for the apical side is the brush border, a highly organized array of membrane protrusions formed by F-actin (Heintzelman and Mooseker, 1992).

Bacterial pathogens have developed sophisticated approaches to overcome this defence, for example by invasion. Salmonella enterica is a food-borne bacterial pathogen that frequently causes diseases ranging from self-limiting gastrointestinal infections to a life-threatening systemic illness known as typhoid fever (Haraga et al., 2008). These bacteria are invasive, facultative intracellular pathogens and many important virulence traits are clustered within Salmonella pathogenicity islands (SPIs) (Gal-Mor and Finlay, 2006; Gerlach and Hensel, 2007). While SPI1 and SPI2 encode type III secretion systems (T3SS) with well-established roles in invasion (Patel and Galan, 2005) and intracellular lifestyle (Kuhle and Hensel, 2004) respectively, the role of many other putative SPIs still has to be unraveled (Gerlach and Hensel, 2007). The events during the invasion of host cells have been studied in great detail on the cellular and molecular level. Effector proteins translocated by the SPI1-encoded T3SS modify signalling events of the host cell, leading to the rearrangement of the actin cytoskeleton (reviewed in Patel and Galan, 2005). As a result of this manipulation, local membrane protrusions are formed that lead to the engulfment and internalization of Salmonella in a process termed macropinocytosis. Recent studies demonstrated that the invasion phenotype is also closely linked to the induction of intestinal inflammation by Salmonella (Hapfelmeier and Hardt, 2005). So far, the molecular events during Salmonella invasion were mainly analysed with non-polarized epithelial cells. However, polarized cell models more closely resemble the architecture of the mucosal cell layer.

The function of SPI4 in host–pathogen interaction is less well understood. A signature-tagged mutagenesis screen revealed a role of SPI4 for the intestinal colonization in a bovine infection model for S. enterica serovar Typhimurium (S. Typhimurium) (Morgan et al., 2004). We and others have recently been able to demonstrate that SPI4 encodes a type I secretion system (T1SS) with SiiE as a secreted substrate (Gerlach et al., 2007a; Morgan et al., 2007). With a molecular weight of 595 kDa, SiiE is the largest protein of the Salmonella proteome and a protein of highly repetitive nature with 53 immunoglobulin (Ig) domain repeats. We demonstrated that SiiE is a non-fimbrial adhesin that is specifically required for the adhesion of Salmonella to polarized epithelial cells. Interestingly, SiiE function is not required for the adhesion to non-polarized cells (Gerlach et al., 2007a). Furthermore, the SPI4-mediated adhesion is co-regulated with invasion genes and under control of the same regulatory systems (Gerlach et al., 2007b; Main-Hester et al., 2008). The observation that SPI1 and SPI4 are expressed and function during the same phase of host–pathogen interaction prompted us to investigate the role of SPI4 during interaction with epithelial cells in more detail. We found that both SPI1 and SPI4 are required for the efficient entry of Salmonella into polarized epithelial cells. These findings provide novel insight into the function of epithelial barriers and how pathogens can overcome this initial host defence.


Invasion of polarized epithelial cells by S. enterica requires the function of two SPIs

Our previous observation of the polarization-specific adhesion mediated by SPI4 prompted us to re-evaluate the role of various known or predicted SPI in host cell invasion by S. Typhimurium. Mutant strains harbouring deletions of individual genes (invC for SPI1, fimD for Fim fimbriae) or entire virulence loci (SPI2, SPI3, SPI4, SPI5, SPI6, SPI9, CS54) were generated. We performed a systematic analysis of the contribution of these loci to invasion and used HeLa as a common non-polarized epithelial cell line and MDCK as epithelial cells that differentiate into polarized monolayers in a robust and highly reproducible fashion. According to standard protocols for invasion of epithelial cells, bacterial strains were subcultured to late log phase and gentamicin-protected intracellular bacteria were quantified (Fig. 1).

Figure 1.

Role of SPIs for invasion of non-polarized and polarized epithelial cells. S. Typhimurium strains were generated with deletions of various SPI, the CS54 island or the Fim locus. The invasion of the mutant strains was compared with S. Typhimurium WT.
A. The non-polarized cell line HeLa was used for infection for 25 min with bacteria subcultured in LB to late log phase. Non-internalized bacteria were killed by addition of gentamicin. Host cells were lysed 60 min after infection by addition of PBS containing 0.1% Triton X-100 and internalized bacteria were quantified by plating serial dilutions onto agar plates.
B and C. Invasion assays were performed as in (A) but polarized monolayers of MDCK cells were used as host cells.
C. Invasion of the WT strain (WT) was compared with SPI1- and SPI4-deficient strains and a double mutant defective in SPI1 and SPI4. The SPI4 mutation was also complemented by a cosmid harbouring the entire SPI4 locus (pSPI4) but not by the parental cloning vector (pSuperCos1). In the right panel, the invasion of mutants deficient in further SPI or fimbrial adhesins was analysed. Mean values of invaded bacteria and standard deviations of three assays are shown. Statistical analyses by Student's t-test were done by comparison with the WT: **P < 0.01; n.d. below detection limit.

In accordance with previous observations, the entry of a SPI1-deficient strain into non-polarized HeLa was about 200-fold diminished (Fig. 1A). Mutant strains deficient in other SPIs or fimbrial adhesins were not significantly reduced in invasion compared with the wild type (WT). SPI5 encodes SopB, an effector protein of the SPI1-T3SS. However, the lack of this effector caused only minor reduction (fivefold) of entry into HeLa cells, a finding in line with the previously reported partial redundancy of SPI1 effectors (Raffatellu et al., 2005). The epithelial cells encountered by Salmonella during infection of mammalian hosts are most likely part of a polarized epithelial layer. Thus, invasion studies were also performed with polarized cells (Fig. 1B). Mutant strains deficient in SPI2, SPI3, SPI5, SPI6 or SPI9 as well as in fimD or CS54 were not altered in their entry into polarized MDCK cells, while the SPI1-deficient strain was highly reduced in invasion. Interestingly, we found that also a SPI4-deficient strain was highly reduced in entering MDCK cells, although this strain invaded HeLa cells as efficiently as the WT strain. The invasion defect of the SPI4 strain was similar to that of the SPI1 mutant strain and could be fully restored by a cosmid harbouring SPI4 (pSPI4), but not by the empty vector (pSuperCos1) (Fig. 1C). The combination of mutations of SPI1 and SPI4 did not indicate additive effects of both loci in invasion of MDCK cell.

Polarized cells have a distinct distribution of membrane proteins on basolateral and apical sides. In addition, a dense brush border might be present on the apical side. To investigate the role of SPI4 in respect to the polarized organization of host cells, the invasion of MDCK cells from the apical or basolateral side was quantified. MDCK cells were grown in a trans-well system using filters with 3 μm pore size to allow the bacterial access to the basolateral side. Invasion was performed from either the apical or basolateral reservoir. A SPI1 strain was highly reduced in invasion from the apical as well as from the basolateral side, while the SPI4 strain was only reduced in apical invasion (Fig. 2).

Figure 2.

SPI4 and SiiE functions are required for apical invasion.
MDCK cells were grown to polarized monolayers on 3 μm trans-well filters and infected with the indicated strains either from the apical or basolateral reservoir as indicated in the scheme in the left panel. The percentages of invaded bacteria were quantified as described for Fig. 1. Mean values of invaded bacteria as percentage of the inoculum and standard deviations are shown and are representative for at least three independent experiments. Statistical analyses by Student's t-test were done by comparison with the WT: **P < 0.01.

To investigate if the contribution of SPI4 to invasion is restricted to polarized cells of specific hosts or a general phenomenon, various polarized epithelial cell lines were used. The human colonic cell lines CaCo-2 and T84 and the murine renal cell line M-1 (Bertog et al., 1999) were cultured until polarized monolayers were formed and infection with Salmonella was performed from the apical reservoir. A pronounced invasion defect of both the SPI1 and the SPI4 strain was observed for each of these polarized cell lines, indicating the general requirement of both SPI1 and SPI4 for the entry into polarized epithelial cells (Fig. S1).

In conclusion, these data demonstrate that Salmonella can efficiently enter non-polarized epithelial cells or the basolateral side of polarized cells by function of the SPI1-encoded T3SS, while invasion of polarized cells from the apical side in addition requires the function of SPI4.

The functions of the SPI4-encoded T1SS and of SiiE are required for invasion of polarized cells

We and others have recently demonstrated that SPI4 encodes a T1SS with SiiE as a secreted substrate (Morgan et al., 2007; Gerlach et al., 2007a). To dissect the contribution of individual SPI4 genes, we analysed the invasion of mutant strains with gene-specific deletions. Our previous work demonstrated that siiCDF encode the components of the T1SS and siiE the secreted substrate protein (Gerlach et al., 2007a). The function of siiAB was not required for synthesis or secretion of SiiE (Gerlach et al., 2007a). The invasion of polarized MDCK cells was reduced to the same extent by mutations in siiC, siiD, siiE or siiF. The highly reduced invasion of strains defective in the T1SS could be fully restored by plasmids expressing siiC, siiD or siiF (Fig. 3). The defect of the siiE strain was restored by a cosmid harbouring the entire SPI4 locus (data not shown).

Figure 3.

Role of the SPI4-encoded T1SS in invasion of polarized cells by Salmonella.
Salmonella strains were generated with gene-specific deletions of siiE, encoding the T1SS substrate protein, or siiC, siiD or siiF, encoding the subunits of the T1SS. Mutant strains in siiC, siiD or siiF were also complemented with plasmids for the expression of the deleted gene or the parental vector pWSK29. Various strains as indicated were used for the infection of polarized MDCK cells and the proportion of invaded bacteria was quantified 60 min after infection as described for Fig. 1.

We showed that SiiE is a non-fimbrial adhesin that is specifically required for the adhesion of Salmonella to polarized epithelial cells (Gerlach et al., 2007a). To test further the contribution of SiiE to invasion of polarized cells, we added antiserum against the C-terminal portion of SiiE during invasion of non-polarized cells (HeLa) or polarized cells (MDCK) (Fig. 4). Neither pre-immune nor anti-SiiE immune serum affected the invasion into HeLa cells (Fig. 4A). In contrast, the anti-SiiE serum reduced apical invasion of MDCK cells about 245-fold, while pre-immune serum had no effect (Fig. 4B).

Figure 4.

Antiserum against SiiE inhibits invasion of polarized cells but not of non-polarized cells. Invasion of HeLa cells or apical invasion of MDCK cells was performed without any addition (mock) or in the presence of rabbit pre-immune or αSiiE immune sera in a final concentration of 1%. The amount of invaded bacteria was quantified 60 min after infection and mean and standard deviations are shown for triplicate experiments.

Cooperation of SPI1 and SPI4 is required to destroy epithelial barrier functions

Cultured polarized epithelial cells can express cell contacts that are morphologically and physiologically reminiscent of cells in mucosal epithelia. In cell culture models, this differentiation correlates with the presence of TJ complexes and the development of a trans-epithelial electrical resistance (TER). After invasion of CaCo-2 (Fig. 5A) or T84 cells (Fig. S2) with Salmonella WT, the TER rapidly decreased. The role of SPI1 in reduction of the TER of polarized monolayers has recently been described (Raffatellu et al., 2005) and in line with these findings, infection with a SPI1 strain did not affect the TER of CaCo-2 monolayers. Similar to the phenotype of a SPI1 strain, we observed that infection with a SPI4 strain did not affect the TER of CaCo-2 cells (Fig. 5A). The SPI4 strain caused a highly delayed loss of the TER of T84 monolayers (Fig. S2). The protein ZO-1 is a marker for functional TJ and it has been previously observed that Salmonella invasion leads to the dislocation of ZO-1 from TJ complexes (Jepson et al., 1995; Boyle et al., 2006). The subcellular localization of the TJ protein ZO-1 was analysed in CaCo-2 cells (Fig. 5B). After infection with Salmonella WT, the dissociation of TJ complexes and the appearance of intercellular gaps were evident. In contrast, the integrity of intercellular connections was not affected by infection with either SPI1 or SPI4 strains. These observations indicate that the functions of both SPI1 and SPI4 are required for the destruction of the epithelial barrier by Salmonella invasion.

Figure 5.

SPI1 and SPI4 are both required to destroy epithelial barrier functions. CaCo-2 cells were grown on filter inserts until polarized monolayers were formed. The cells were infected by addition of S. Typhimurium WT or strains deficient in SPI1 or SPI4 to the apical reservoir without subsequent washing and addition of gentamicin.
A. At various time points after infection, the TER was recorded. Mean resistance per cm2 and standard deviations for three assays are shown.
B. The integrity of the intercellular contacts of the epithelial layer was analysed by immuno-staining for the TJ protein ZO-1. CaCo-2 were mock-infected or infected with various strains expressing mCherry (red), fixed 140 min after infection and immuno-stained for ZO-1 (green). Scale bars: upper row, 20 μm; lower row, 5 μm.

SPI1 and SPI4 functions are required for cytoskeletal alterations and brush border effacement by Salmonella

The SPI1-mediated invasion by Salmonella is accompanied by massive reorganization of the actin cytoskeleton and formation of membrane protrusions (Francis et al., 1993). Using the infection model of polarized cells (MDCK), we observed that Salmonella WT frequently induced the formation of ring-like accumulations of F-actin surrounding individual bacteria or bacterial clusters (Fig. 6A). As expected, such membrane ruffles were absent on cells infected with the SPI1 strain (data not shown). If cells were infected with the SPI4 strain, the number of host cell-associated bacteria was very low, but membrane ruffles were present surrounding some of these bacteria (Fig. 6A, arrowhead).

Figure 6.

Cytoskeletal rearrangements and brush border effacement by invading Salmonella require SPI1 and SPI4 functions. Polarized monolayers of MDCK cells were infected with WT, SPI1 or SPI4 strains, fixed and processed for immuno-fluorescence (A) or scanning EM (B and D).
A. After infection with Salmonella (green) for 25 min, ring-like condensation of F-actin (red) was frequently observed in close proximity of the bacteria (arrowheads). In contrast, SPI4-deficient bacteria were rarely found associated with the apical side of epithelial cells and F-actin condensation was only found in some fields of view. Scale bars, 20 μm.
B. At 25 min after infection, the WT (pseudo-coloured red) and SPI1 strain (green) were found in close contact with apical surfaces and microvilli (arrowheads). Contact to microvilli was not observed for the SPI4 strain (blue). Scale bar: 1 μm.
C. Salmonella WT, SPI1 and SPI4 mutant strains were scored for the intimate association with microvilli (MV) on the apical surface of MDCK cells as shown in (C). Mean values for at least 10 fields of view per strain are shown.
D. Salmonella-induced remodelling of the apical side of MDCK cells. Cells were analysed 60 min after infection with Salmonella WT. Note the appearance of bacterial clusters on individual host cells, the complete loss of the brush border architecture on infected cells and the formation of invaginations at the site of bacterial attachment. Scale bars: left panel, 10 μm; middle and right panels, 1 μm.

Scanning electron microscopy (EM) revealed that Salmonella WT or the SPI1 strain were closely associated with the apical side of polarized epithelial cells and in intimate contact with microvilli (Fig. 6B, arrowheads). Only few adherent bacteria of the SPI4 strain were detected by scanning EM, and these bacteria were not in contact to microvilli. We scored EM micrographs for the association of microvilli with various strains and observed that about 72% and 66% of WT and SPI1 bacteria respectively, but only 14% of the SPI4 mutant strains were in intimate contact with at least one microvillus (Fig. 6C). Addition of antibody against SiiE had a similar effect as the SPI4 mutation. If 1% α SiiE serum was present during infection with the WT, only 18.2% of the bacteria were in intimate contact with at least one microvillus, while 87.5% of the bacteria bound microvilli in the presence of 1% pre-immune serum. Further adhesion assays to the apical side of MDCK cells were performed at 4°C or in the presence of cytochalasin D in order to inhibit the invasion of the bacteria. Adhesion of WT, SPI1 and SPI4 strains was on a similar, highly reduced level (Fig. S3A). Addition of cytochalasin D resulted in about 15-fold reduced adhesion of the WT strain, but had no significant effect on the adhesion of the SPI4 strain (Fig. S3B).

At 60 min after infection, we observed dramatic change of the brush border architecture of MDCK cells infected with WT Salmonella (Fig. 6D). A complete loss of microvilli was evident, reminiscent of the effacement of the brush border reported for enteropathogenic Escherichia coli (Baldini et al., 1983; DeVinney et al., 1999). Brush border effacement has also been reported for Salmonella (Finlay et al., 1988; Haque et al., 2004), but the phenomenon was not correlated to SPI1 and SPI4 functions. Membrane ruffles were not detected on cells that lost the microvilli architecture and usually several bacteria were associated with these host cells and often formed clusters. Due to the short infection time of 30–60 min, these clusters did not result from replication but rather from collective adhesion. In contrast, on neighbouring cells without bacteria attached to the apical side, entirely normal brush border architecture was found. Interestingly, some of the infecting bacteria were located within invaginations on the apical membrane (Fig. 6D). Brush border effacement and formation of bacterial clusters were not observed if cells were infected with the SPI4-deficient strain (data not shown).

We speculated that the brush border acts as barrier for invasion and that SiiE is required to mediate specific binding to receptors on microvilli. To test this hypothesis, we performed experiments with two rounds of infection (Fig. S4). If MDCK cells were first mock-infected and subsequently infected with a SPI4 strain, only very small numbers of adherent bacteria were observed. However, if the first infection with the WT strain and a second infection with the SPI4 strain were performed, a larger number of adherent SPI4 bacteria were found on the apical side. These observations suggest that the cytoskeletal alterations induced by SPI1-mediated invasion alleviate the requirement for the SPI4-dependent adhesion of Salmonella.

We next investigated the adhesion of Salmonella to the basolateral side of MDCK cells grown on trans-well filters with 3 μm pore size (Fig. S5). As expected from the experiments shown in Fig. 2, WT and SPI4 strains were found more frequently associated with membrane ruffles than SPI1 or SPI1/SPI4 strains. The MDCK cells formed protrusions into the membrane pores and bacterial adhesion was predominantly observed at the position of these pores. Invading SPI4 bacteria were mainly associated with these protrusions, while WT bacteria were associated with protrusions at pores but also with membrane ruffles at other parts of the basolateral membrane. As cell protrusions already showed membrane ruffles without the presence of invading bacteria, a clear identification of Salmonella-induced ruffles at the basolateral side was not possible.

The size of adhesin SiiE is critical for invasion of polarized cells

The adhesin SiiE is an unusual protein due to its large size of 595 kDa and the highly repetitive structure with 53 repeats of an Ig domain (Fig. 7A) (Gerlach et al., 2007a). Despite its highly repetitive nature, the size of SiiE is highly conserved in various S. enterica serovars causing non-typhoidal infections and we did not find length polymorphisms in clinical isolates of one serotype (data not shown). We speculated that the large number of Ig domain repeats might be critically important for proper binding of an adhesive domain at the C-terminal end of SiiE to a host cell ligand. To test this hypothesis, mutant strains were constructed that expressed various alleles of the chromosomal siiE with deletions of defined numbers of Ig domains (for schematic representation, see Fig. 7A). Western blot analysis indicated that the resulting strains synthesized comparable amounts of the truncated variants of SiiE (Fig. 7B). All of the variants were secreted during growth in Luria–Bertani (LB) in amounts comparable with that of WT SiiE, as indicated by ELISA (Fig. S6). Our previous work indicated that SiiE is located in patches on the surface of Salmonella adhering to the apical membrane of polarized cells (Gerlach et al., 2007a). A similar distribution of WT SiiE and the various truncated SiiE forms was observed by confocal microscopy (Fig. 7C). These data show that truncated forms of SiiE are secreted by the SPI4-encoded T1SS and located on the cell surface during bacterial host interaction.

Figure 7.

The size of SiiE is critical for the invasion of polarized cells by Salmonella.
A. Schematic presentation of the domain structure of SiiE and generation of truncated variants of SiiE. SiiE consists of 53 repeats of Ig domains (open boxes). The C-terminal part contains a signal (black rectangles) for the secretion by the SPI4-encoded T1SS. Deletions were generated in order to remove various numbers of Ig domains. The synthesis, secretion and surface localization of the various mutants were confirmed. Deletion of a small, irrelevant region of SiiE (aa198–201) by the same procedure had only little effect on surface localization (C) and secretion of SiiE (Fig. S6) and only led to minor reduction of invasion (Fig. S7).
B. Synthesis of SiiE by Salmonella WT, a siiE deletion strain and various strains harbouring siiE alleles with internal deletion. Western blot analyses of cell lysates were performed using antiserum against recombinant SiiE.
C. Surface location of WT and mutant variants of SiiE. Salmonella strains expressing various truncated alleles of siiE were used for apical infection of MDCK cells. After 15 min of infection, immuno-staining was performed for SiiE (green) and Salmonella LPS (red). Representative bacteria located on the surface of MDCK cells are shown. Note that adherent bacteria expressing siiEΔIg21–30, siiEΔIg21–35 or siiEΔIg21–40 were only very rarely detected. Scale bar, 1 μm.
D. Apical invasion of MDCK cells by Salmonella WT, a siiE strain and various mutant strains harbouring siiE alleles with internal deletion of different extent.
E. Invasion by strains with WT (full LPS) or truncated O-antigen (short LPS), both expressing SiiE ΔIg21–30. The proportion of internalized bacteria was quantified as described for Fig. 1. Data are representative of three independent experiments. Statistical analyses by Student's t-test were done by comparison with WT (D) or the strain with siiEΔIg21–30 and full-length LPS (E), **P < 0.01, ***P < 0.001.

We next investigated the effect of the truncation of SiiE on apical adhesion and invasion of MDCK cells (Fig. 7D). Strains with SiiE deleted for 10 or more Ig domains were only rarely found associated with host cell surfaces (data not shown). We observed a clear correlation between the extent of deletions and the rate of invasion by Salmonella. A mutant allele deleted for two Ig domain repeats was only slightly affected in invasion (threefold reduced invasion), but deletion of 5, 10, 15 or 20 Ig domain repeats gradually decreased entry into polarized cells. As expected, none of these deletions affected the invasion of non-polarized cells (data not shown).

Salmonella possesses an unusually long lipopolysaccharide (LPS) with more than 100 modal repeats of O-antigen whose synthesis requires the function of the regulators WzzST and FepE (Murray et al., 2003). We speculated that the large size of SiiE might be necessary to project an adhesive moiety of the protein beyond the O-antigen layer and tested if truncation of SiiE could be complemented by truncation of the O-antigen length. The presence of truncated LPS in a ΔwzzSTΔfepE strain that expressed WT siiE had no significant effect on the apical invasion of MDCK cells. The apical invasion of MDCK cells by strains expressing siiE with a deletion of 10 Ig domains was highly reduced. Interestingly, the presence of truncated O-antigen resulted in 5.7-fold higher invasion compared with the strain with WT O-antigen (Fig. 7E), indicating a correlation between the size of SiiE and the LPS structure.

In conclusion, the data demonstrate the requirement for the large size of SiiE for proper adhesion to, and invasion of polarized cells, from the apical side.


We demonstrate here that invasion of polarized epithelial cells by Salmonella requires the cooperative function of two different protein secretion systems encoded by SPI1 and SPI4. Without adhesion mediated by SiiE secreted via the SPI4-encoded T1SS, Salmonella was almost unable to trigger the SPI1-mediated remodelling of the host cell cytoskeleton, leading to uptake of the pathogen. Our observations indicate that translocation of effector proteins into and invasion of epithelial cells in their physiological, i.e. polarized, organization require a close cooperation between an adhesin and a protein translocation system. The functional cooperation between SPI1 and SPI4 is in line with the previous data on the co-regulation of the expression of the SPI1-encoded T3SS and its effector proteins with SPI4 (Ahmer et al., 1999; Gerlach et al., 2007b; Main-Hester et al., 2008). Both systems are active in the same growth phase and are expressed under the control of SirA, a regulator that has been proposed as a global regulator of enteropathogenesis (Ahmer et al., 1999). While SirA-regulated functions of SPI1 and SPI5 in invasion have been demonstrated, the role of SirA for control of SPI4 expression remained obscure until recently. Our work shows that the co-ordinate function of SPI1 and SPI4 is a requirement for the manipulation of polarized cells by Salmonella.

These novel findings lead to a revised model for invasion during gastrointestinal infections by Salmonella (see Fig. 8). The unique architecture of brush border on the apical side of polarized epithelial cells might present a barrier to invading pathogens. SiiE mediates adhesion to unknown ligands on the apical membrane and without this proper binding the T3SS-mediated translocation of effectors remains inefficient. In non-polarized cells that lack the dense array of actin-rich microvilli, the translocation works in the absence of a specific adhesin. After adhesion and translocation, the actin cytoskeleton of the host cell is remodelled. We found that Salmonella induced the destruction of the brush border architecture in a way that is reminiscent of the effacement phenotype of enteropathogenic E. coli (Baldini et al., 1983). In addition to the effacement of microvilli effectors of the SPI1-T3SS of Salmonella induce the formation of membrane ruffles and bacterial uptake that have been studied in molecular detail before. Our observations indicate that effacement of microvilli is necessary to destroy a barrier as well as to make sufficient amounts of monomeric actin available for subsequent polymerization during formation of membrane ruffles.

Figure 8.

Model for the cooperation of SPI4 and SPI1 during invasion of polarized cells by S. enterica. The apical side of polarized epithelial cells has a dense array of microvilli (A). The SPI4-encoded T1SS secretes SiiE that mediates adhesion and intimate contact of Salmonella to the apical membrane (B). This close contact is required for the efficient translocation of the effector proteins by the SPI1-encoded T3SS (C). SPI1 effector proteins remodel the host cell actin cytoskeleton, resulting in formation of membrane ruffles and internalization of Salmonella and in the effacement of the brush border (D). In the absence of the SPI4-encoded T1SS and the adhesin SiiE, contacts between bacteria and host cells are highly reduced and translocation of SPI1 effectors is inefficient. Consequently, these mutant strains are severely compromised in host cell entry from the apical side.

The large size of SiiE is required for the proper function as strains secreting truncated forms of SiiE are reduced in invasion. The specific structure of Salmonella LPS with long and very long O-antigen chains could entail the presence of a long linear SiiE portion composed of Ig domains that project the adhesive domain of SiiE beyond the LPS. Alternatively, the size of SiiE might be a requirement for the contact to host cell ligands that are hidden within the highly organized microvilli array. The proposed linear organization of SiiE requires further experimental proof, but so far we have not been able to visualize SiiE by EM. Also, further deletional analyses have to be performed to define the adhesive domain or domains in SiiE.

In addition to invasion, Salmonella can utilize alternative routes for the penetration of the intestinal epithelium, e.g. invasion of M cells (Jones et al., 1994), uptake by CD18-positive phagocytes (Vazquez-Torres et al., 1999) or via sampling by dendritic cells (Rescigno et al., 2001). However, the SPI1-mediated translocation of effector proteins is of central importance for the induction of intestinal responses typical for gastrointestinal diseases by Salmonella (Hapfelmeier and Hardt, 2005). We have previously observed that both SPI1 and SPI4 are required to induce intestinal inflammation in a murine colitis model (Gerlach et al., 2007a).

The presence of SPI4 might explain the difference in the intestinal pathogenesis of Shigella spp. and S. enterica. Shigella cannot invade epithelia from the apical side and the current model proposes that Shigella has to be transcytosed through M cells to be able to invade epithelial cells from the basolateral side by a mechanism that is similar to the SPI1-mediated invasion of Salmonella (Mounier et al., 1992; Nhieu and Sansonetti, 1999). As the invasion of M cells has also been reported for Salmonella (Jones et al., 1994), basolateral invasion of epithelial cells might also be deployed by Salmonella. However, the presence of SPI4 and the adhesin SiiE enables Salmonella to establish intimate contact to apical side of enterocytes and to invade cells from the apical side. SPI4 has been identified to be important for bovine intestinal colonization (Morgan et al., 2004) and our work using a murine model for intestinal inflammation clearly indicated a role of SPI4 in the intestinal pathogenesis by Salmonella (Gerlach et al., 2007a). Invasion is a prerequisite for the intracellular replication but also contributes to the elicitation of an inflammatory host response (Hapfelmeier and Hardt, 2005). This inflammatory response might actually support Salmonella during competition with the intestinal flora of the host (Stecher et al., 2007).

Recent work demonstrated the role of the adhesin CagL for the proper function of the type IV secretion system of Helicobacter pylori during host–pathogen interaction (Kwok et al., 2007). Although the mechanisms of T3SS and type IV secretion systems are different, both systems require the close contact to a host cell membrane for the translocation of effector proteins. As no specific host cell receptors for the needle of T3SS are known, additional adhesin functions are required to bring the T3SS ‘in position’.

Despite an enormous body of work on Salmonella invasion, the specific requirement of SPI4 for this virulence function has previously been missed. Our work indicates that understanding of bacterial pathogenesis critically depends on the use of suitable model systems. As the epithelial cell layer is an important host barrier encountered by gastrointestinal pathogens, the specific features of this tissue have to be rebuilt in infection models. The requirement of an adhesion mechanism for pathogenic interactions with the intestinal epithelium also indicates new directions for therapeutic interference with gastrointestinal infections by Salmonella. ‘Virulence blockers’ interfering with bacterial adhesion are currently under investigation (Cegelski et al., 2008) and might also have future potential to interfere with SPI4 function and gastrointestinal infections by Salmonella.

Experimental procedures

Bacterial strains and culture conditions

The S. Typhimurium NCTC 12023 was used as WT strain and mutant strains used in this study were isogenic to NCTC 12023. Characteristics of the strains are listed in Table 1. Salmonella strains were routinely cultured in LB broth or on LB agar containing antibiotics if required for selection of strains or maintaining plasmids. Antibiotics are used at the following concentrations: carbenicillin, 50 μg ml−1; kanamycin, 50 μg ml−1; chloramphenicol, 34 μg ml−1.

Table 1.  Bacterial strains and plasmids used in this study.
DesignationRelevant characteristicsReference
Salmonella strains  
 NCTC 12023WTLab collection
 MvP813invC::aph KmRRed deletion, this study
 MvP818ΔinvC::FRTpCP20 mediated deletion of aph in MvP813
 MvP499SPI2::aph KmRRed deletion, this study
 MvP492SPI3::aph KmRRed deletion, this study
 MvP589ΔSPI4::FRTRed deletion, Gerlach et al. (2007a)
 MvP504SPI5::aph KmRRed deletion, this study
 MvP919SPI6::aph KmRRed deletion, this study
 MvP494SPI9::aph KmRRed deletion, this study
 MvP604ΔSPI4 invC::mTn5This study
 MvP920CS54::aph KmRRed deletion, this study
 MvP921fimD::aph KmRRed deletion, this study
 MvP595ΔsiiC::FRTRed deletion, Gerlach et al. (2007a)
 MvP597ΔsiiD::FRTRed deletion, Gerlach et al. (2007a)
 MvP599ΔsiiE::FRTRed deletion, Gerlach et al. (2007a)
 MvP812ΔsiiF::FRTRed deletion, Gerlach et al. (2007a)
 MvP781ΔsiiE6540−12493::FRTΔIg21–40 of SiiE, this study
 MvP989ΔsiiE6540−11026::FRTΔIg21–35 of SiiE, this study
 MvP990ΔsiiE6540−9541::FRTΔIg21–30 of SiiE, this study
 MvP991ΔsiiE6540−8038::FRTΔIg21–25 of SiiE, this study
 MvP992ΔsiiE6540−7141::FRTΔIg21–22 of SiiE, this study
 MvP1009ΔsiiE591−601::FRTΔaa198–201 of SiiE, this study
 MvP1218ΔsiiE6540−9541 wzz::FRTthis study
fepE::aph KmR
 p3224Complementation of siiCGerlach et al. (2007a)
 p3229Complementation of siiDGerlach et al. (2007a)
 p3223Complementation of siiFGerlach et al. (2007a)
 pSuperCos1Cosmid vectorStratagene
 p7B4SPI4 in pSuperCosGerlach et al. (2007a)
 pFPV25.1PrpsM::gfpmut3Raphael Valdivia, Stanford, CA
 pFPV-mCherry/2PrpsM::mCherryLeigh Knodler, Hamilton, MT

Construction of deletion mutant strains

Mutant strains deleted for the entire SPI4 locus or harbouring gene-specific deletions of individual SPI4 genes were generated by the ‘one-step’ inactivation protocol as previously described (Chakravortty et al., 2002). Oligonucleotides used for the generation of targeting constructs are listed in Table S1. In-frame deletions in siiE were generated by Red-mediated recombination using pKD13 (Datsenko and Wanner, 2000) as a template vector. This procedure inserts a 27-codon scar sequence within the open reading frame of the target gene after FLP-mediated deletion of the aph-resistance cassette. The deletions were confirmed by PCR using control primers listed in Table S1.

Deletions in SPI4 genes were complemented by plasmids expressing WT alleles of the mutant genes. The construction of these plasmids has been described before (Gerlach et al., 2007a).

Cell lines and culture conditions

The non-polarized epithelial cell line HeLa was cultured in DMEM and maintained as previously described (Gerlach et al., 2007a). Polarized cell lines from various organisms were used as follows: The canine kidney epithelial cell line MDCK was seeded in 24-well plates (Cellstar bio-one, Greiner, Frickenhausen, Germany) at a density of 1 × 105 per well or on semi-permeable polycarbonate membranes (Millicell-PCF, 0.4 or 3 μm pore size, Millipore, Schwalbach, Germany) at a density of 5 × 105 per insert. Cells were allowed to differentiate for 6–10 days in MEM medium (PAA, Pasching, Austria) supplemented with 10% FCS (Sigma-Aldrich, Seelze, Germany), nonessential amino acids (PAA), 2 mM Glutamax (Invitrogen, Karlsruhe, Germany), 100 U ml−1 penicillin and 100 μg ml−1 streptomycin (PAA).

The human colonic cell lines CaCo-2, clone C2BBe1 (ATCC CRL-2102) and T84 (ECACC) were seeded on semi-permeable polycarbonate membranes (Millicell-PCF, 0.4 μm pore size) at a density of 2 × 105 cm−2. Cells were allowed to differentiate for 8–12 days in DMEM medium (PAA) supplemented with 10% FCS, 10 μg ml−1 human holo-transferrin (Sigma T0665), 2 mM Glutamax, 100 U ml−1 penicillin and 100 μg ml−1 streptomycin.

The murine M-1 cells (ATCC CRL-2038) were cultured as described previously (Bertog et al., 1999). Briefly, cells were seeded to confluence on semi-permeable polycarbonate membranes (Millicell-PCF, 0.4 μm pore size) and allowed to differentiate for 8–10 days in PC-1 complete culture media (Lonza, Verviers, Belgium) supplemented with 2 mM glutamine, 100 U ml−1 penicillin and 100 μg ml−1 streptomycin.

Cells seeded on semi-permeable supports were grown to confluence and TER of the cultures was checked routinely using an EVOM epithelial volt-ohm-meter (World Precision Instruments, Berlin, Germany).

Epithelial cell invasion model

At least 4 h before infection, cells were washed once with PBS and medium without penicillin/streptomycin was applied. Bacteria were subcultured (1:31 from O/N culture) in LB with appropriate antibiotics for 3.5 h and adjusted in PBS to OD600 = 0.2 (for viable counts experiments) or OD600 = 0.6 (for immuno-fluorescence). According to the intended multiplicity of infection (MOI), a master mix of the inoculum was prepared in complete cell culture media (without antibiotics) and 500 μl were added to the apical or, if indicated, to the basolateral side of the cells. After 25 min cells were washed thrice with PBS and complete cell culture media containing 100 μg ml−1 gentamicin were applied for 1 h. After gentamicin treatment cells were washed once with PBS and lysed using 500 μl pre-warmed 0.1% Triton X-100 in PBS for 5 min at 37°C. Serial dilutions were made in pre-chilled PBS and plated on Mueller-Hinton agar plates.

For the analyses of the TER and the distribution of TJ protein ZO-1, cells were infected by addition of bacteria to the apical reservoir without subsequent washing and addition of gentamicin.


For immuno-fluorescence, cells were fixed after infection with 3% PFA in PBS for 20 min at room temperature and subsequently washed three times with PBS. Cells were permeabilized with 0.1% Triton X-100 in PBS for 5 min at room temperature. After permeabilization, cells were incubated in blocking solution (2% goat serum, 2% BSA in PBS) for 30 min. For ZO-1 staining, cells were fixed overnight in methanol at −20°C, washed thrice in PBS and re-hydrated in PBS for 15 min at room temperature. Antibodies were diluted in blocking solution and applied for 1–3 h (primary antibodies) or 1 h (secondary antibodies). Samples were embedded using Fluoprep (BioMerieux, Nürtingen, Germany) and Entellan (Merck, Darmstadt, Germany). Imaging was done on a Leica TCS NT confocal microscope with subsequent deconvolution using Huygens Essential v2.9 (SVI BV, Hilversum, the Netherlands). For additional image processing Photoshop CS2 (Adobe, München, Germany) was used.

Electron microscopy

MDCK cells were grown as polarized monolayers on semi-permeable polycarbonate membranes (Millipore Millicell PCF), infected with Salmonella strains at various MOI and fixed with 2% glutaraldehyde, 5% formaldehyde in cold cacodylate buffer (100 mM cacodylate, 90 mM sucrose, 10 mM MgCl2, 10 mM CaCl2, pH 6.9) for at least 4 h at 4°C.

Samples were washed with TE-buffer (20 mM TRIS, 1 mM EDTA, pH 6.9) before dehydrating in a graded series of acetone (10%, 30%, 50%, 70%, 90%, 100%) on ice for 15 min for each step. Samples were then subjected to critical-point-drying with liquid CO2 (CPD30, Balzers, Liechtenstein) and sputter-coated with a thin gold film (SCD040, Bal-Tec, Liechtenstein) before examination in a Zeiss field emission scanning electron microscope DSM982 Gemini using the Everhart Thornley SE detector and the inlens SE-detector in a 50:50 ratio at an acceleration voltage of 5 kV.

For imaging of basolateral infections, critical-point-dried cells were separated from the filter surface. For this purpose, a sticky carbon tap mounted onto a stub was gently pressed onto the cells and then carefully pulled off to detach the cells from the filter surface. The basolateral side was then sputter-coated with a thin gold film as described above.


This work was supported by DFG Grants HE1964 to M.H. We like to thank Marko Bertog for help with the M-1 cell culture, Ina Schleicher for technical assistance in the EM studies and Stefanie Hölzer for generation of wzzST and fepE mutant strains. We are indebted to Mathias Hornef and Wolf-Dietrich Hardt for critical comments on the manuscript.