Inoculum Composition and Salmonella Pathogenicity Island 1 Regulate M-Cell Invasion and Epithelial Destruction by Salmonella typhimurium

Abstract

In the mouse model of Salmonella typhimurium infection, the specialized antigen-sampling intestinal M cells are the primary route of Salmonella invasion during the early stages of infection. Under certain experimental conditions, M-cell invasion is accompanied by M-cell destruction and loss of adjacent regions of the follicle-associated epithelium (FAE), although the conditions responsible for expression of the cytotoxic phenotype in a proportion of previous studies have not been defined. In the present study, we have demonstrated that the cytotoxic effect exerted by wild-type S. typhimurium on mouse Peyer’s patch FAE is dependent on the inoculum composition. We have also demonstrated that the extent of FAE destruction correlates with the extent of M-cell invasion. Bacteria inoculated in Luria-Bertani (LB) broth induce extensive FAE loss and exhibit efficient M-cell invasion, whereas bacteria inoculated in phosphate-buffered saline fail to induce significant FAE disruption and invade M cells at significantly lower levels. Similarly, inoculation in LB significantly enhances invasion of Madin-Darby canine kidney cells by wild-type S. typhimurium. Mutants defective for expression of invA, a component of Salmonella pathogenicity island 1 which is vital for efficient invasion of cultured cells, fail to induce FAE destruction and, when inoculated in LB, are attenuated for M-cell invasion. Variation in inv gene expression is, therefore, one possible mechanism by which inoculate composition may regulate the virulence of wild-type S. typhimurium. Our findings suggest that the composition of the gut luminal contents may be critical in determining the outcome of naturally acquired Salmonella infections and that both vaccine formulation and dietary status of vaccine recipients may significantly affect the efficacy and safety of live Salmonella oral vaccine delivery systems.

Salmonella species are an important group of pathogens which infect a wide range of hosts to cause a variety of disease syndromes. One feature common to all of these disease syndromes is that following oral ingestion, the bacterium must penetrate the intestinal epithelial barrier prior to the initiation of disease. The mechanisms by which Salmonella invade the intestinal epithelium are unclear, although in vitro studies have identified several genes which are required to optimize S. typhimurium invasion of epithelial cells. Many of these genes are located in Salmonella pathogenicity island 1 (SPI1), located at centisome 63 on the Salmonella typhimurium chromosome. These genes encode a type III protein secretion system together with several of its target proteins (for a review, see reference 16). In vitro studies have demonstrated that invasion of cultured cells by Salmonella species is modulated by a variety of environmental and growth conditions, including oxygen tension, osmolarity, carbohydrate availability, and bacterial growth phase (13, 24, 27, 35–37). Environmental modulation of in vitro invasion is achieved by the regulation of SPI1-encoded invasion gene expression via a complex array of transcription factors (1, 2, 21, 25, 32) which are thought to ensure that invasion gene expression and consequently epithelial invasion by S. typhimurium are maximal under conditions present in the gut lumen.

The primary sites of Salmonella invasion in the host intestine are the ileal Peyer’s patches and possibly also the cecal lymphoid patches (4). The follicle-associated epithelium (FAE) which overlies these gut-associated lymphoid tissues includes the specialized antigen-sampling M cells which are a major site of invasion by a diverse range of pathogens (14, 19, 28). Recent studies suggest that M cells play a pivotal role in the pathogenesis of S. typhimurium since, at least during the early stages of infection, these cells are the primary site of S. typhimurium invasion in the mouse intestine (7, 22). The role of the SPI1-encoded genes in M-cell invasion is unclear, although recently we have demonstrated that strains carrying mutations in the inv genes of SPI1 are severely attenuated for invasion of cultured cells but retain the capacity to invade mouse M cells (8).

The interaction of S. typhimurium with intestinal M cells induces the formation of prominent membrane protrusions termed membrane ruffles (7, 8, 22). In addition, under certain experimental conditions, wild-type strains of S. typhimurium induce M-cell destruction and extensive sloughing of adjacent areas of FAE (3, 10, 22, 31, 33). The factors regulating the cytotoxicity of S. typhimurium have not been identified, although it is notable that previous studies have used markedly different experimental protocols (3, 7, 8, 10, 22, 31, 33).

The aim of this study was to test the hypothesis that the phenomenon of M-cell cytotoxicity is dependent on the composition of the S. typhimurium inoculate. In addition, we have investigated the role of the inv genetic locus in Salmonella-induced cytotoxicity. Our results demonstrate that inoculate composition determines the extent of both M-cell invasion and the capacity to induce epithelial disruption. While S. typhimurium invades M cells by both inv-dependent and inv-independent mechanisms, mutants defective for invA are unable to cause epithelial disruption and, under conditions which promote invasion and FAE destruction by the wild type, are attenuated for M-cell invasion. These findings suggest that variation in expression of SPI1-encoded invasion genes is one possible mechanism by which inoculate composition may regulate M-cell invasion and FAE destruction by wild-type S. typhimurium.

MATERIALS AND METHODS

Bacterial strains and culture.

Wild-type S. typhimurium strains IR715 and SR11 and mutant strain SB111 (an isogenic nonpolar invA mutant of SR11 [30]) were prepared as previously described (7, 8). Briefly, a single colony grown on Luria-Bertani (LB) agar was inoculated into 2 ml of LB broth and incubated with agitation at 37°C for 7 h. From this starter culture, 10³ bacteria were inoculated into 5 ml of LB broth (in a 6-ml vial) and grown as a static culture overnight (16 h) at 37°C. Alternatively, to obtain stationary-phase bacteria grown under nutrient-limiting conditions, 10³ bacteria were inoculated into 5 ml of LB broth in a 30-ml vial and shaken vigorously overnight at 37°C. To determine the effect of inoculum composition, the bacteria were then pelleted and resuspended three times in either the original LB growth broth, phosphate-buffered saline, pH 7.4 (PBS), or PBS supplemented with 0.14 M mannitol, 5 g of yeast extract per liter, or 10 g of tryptone per liter. In the remaining studies, bacteria were used directly in the LB growth broth, without pelleting and resuspension.

Mouse gut loop studies.

Ligated jejunal/ileal Peyer’s patch-containing gut segments were created in anesthetized adult female BALB/c mice as described previously (7, 8) and infected with the bacterial (3 × 10⁹ bacteria/ml) or control preparations. After an appropriate incubation period, the mice were culled by cervical dislocation and the gut loops were rapidly removed. Harvested tissues were pinned flat, mucosal surface uppermost, on cork boards, rinsed thoroughly in PBS, and fixed in either 2% glutaraldehyde (in 100 mM sodium phosphate buffer, pH 7.3) at 4°C or methanol (−20°C) for at least 2 h. To facilitate subsequent examination of the FAE, villi were then microdissected away from the domes.

Glutaraldehyde-fixed tissues which had been infected, unless otherwise stated, for 60 min were processed for scanning electron microscopy (SEM) as described previously (6) and examined with a Cambridge S240 SEM. All tissues were examined blind, and the extent of FAE destruction on each dome was assigned a score of 0 to 4, where 0 denotes negligible damage (loss of <20 single cells or <4 small cell clusters), 1 denotes low levels of damage (loss of ≥20 single cells or ≥4 but <10 small cell clusters), 2 denotes moderate damage (loss of ≥10 small cell clusters or ≥5 large cell clusters), 3 denotes extensive damage (loss of ≥1/6 FAE), and 4 denotes very extensive damage (loss of ≥1/2 FAE), small and large cell clusters being defined as areas equivalent to >4 but <10 cells and ≥10 cells, respectively.

Methanol-fixed tissues infected for 15 min were processed for immunocytochemical localization of bacteria by staining with anti-Salmonella antibodies and fluorescein isothiocyanate (FITC)-conjugated secondary antibodies (7). M cells were localized by immersion for 60 min in tetramethylrhodamine isothiocyanate (TRITC)-conjugated Ulex europaeus agglutinin 1 (UEA1; 67 μg/ml; Vector Laboratories, Peterborough, United Kingdom), a marker of mouse Peyer’s patch M cells (6). The stained tissues were mounted in Vectashield mounting medium (Vector Laboratories) and examined with a Bio-Rad MRC-600 confocal laser scanning microscope (CLSM) equipped with a krypton-argon mixed-gas laser. S. typhimurium association with the FAE was quantified by analyzing the z series of confocal optical sections of randomly selected areas of FAE from four discrete predetermined peripheral regions of infected Peyer’s patch domes. M-cell-associated bacteria were classified as adhered or invaded, invaded bacteria being defined as those at least 2 μm below the strongly UEA1-stained M-cell surface. Areas where the FAE was obscured (e.g., by mucus) were excluded from analysis.

In vitro studies.

The effect of inoculum composition on invasion of cultured epithelial cells by wild-type S. typhimurium was also examined. S. typhimurium IR715 was grown as described above and then pelleted and resuspended three times in fresh LB or PBS. After dilution in LB or PBS to a concentration of approximately 10⁸/ml, the bacterial preparations were added to Madin-Darby canine kidney (MDCK) strain I cells cultured on permeable supports (Anocell; 0.5 cm²; Nunc, Roskilde, Denmark) as described previously (20). After 15 min of incubation at 37°C, the cell monolayers were washed six times in PBS to remove nonadherent bacteria and transferred to PBS at 0°C to prevent further invasion. S. typhimurium adhesion and invasion was then quantified by differential immunocytochemical staining as described previously (20). Briefly, the monolayers were incubated sequentially (at 0°C) with goat anti-Salmonella antibodies and FITC-conjugated rabbit anti-goat immunoglobulin to label extracellular bacteria. After permeabilization in methanol, the monolayers were incubated with anti-Salmonella antibodies and TRITC-conjugated rabbit anti-goat immunoglobulin at room temperature to label extracellular and intracellular bacteria. Monolayers were then examined with a Leica DM RBE epifluorescence microscope. Counts of adherent (FITC-labeled) and total (TRITC-labeled) bacteria associated with the monolayers were made in 10 randomly selected fields (ca. 5,000 μm² per field) and used to calculate both the numbers of adhered and invaded bacteria per unit area.

Statistical analyses.

The extent of FAE destruction (as defined by dome score values) induced by alternative inocula was initially analyzed by a Kruskal-Wallis test, which revealed that FAE damage was significantly affected by the composition of the gut loop inoculum. Data obtained for pairs of alternative inocula were subsequently analyzed by Dunn’s multiple-comparison tests. Statistical analyses of S. typhimurium association (adhesion and invasion, as defined by CLSM) with FAE cells in vivo and MDCK cells in vitro were performed by Mann-Whitney U tests. For all statistical tests, P values of <0.05 were considered significant; for direct comparisons of alternative inocula, the data analyzed were obtained from experiments performed on the same day and using the same batch of mice.

RESULTS

Effect of inoculum composition. (i) FAE destruction.

SEM examination of tissues infected for 60 min with wild-type S. typhimurium strain IR715 resuspended in PBS revealed the presence of M-cell-associated ruffles but little evidence of M-cell cytotoxicity or FAE loss (Fig. 1a and b). In contrast, S. typhimurium resuspended in the original LB growth broth induced extensive FAE damage (Fig. 1c to f) typically characterized by areas of cell loss which were surrounded by cells denuded of microvilli and associated with the presence of invading bacteria (Fig. 1e and f). The areas of cell loss varied in size from single cells (presumably M cells) and small cell clusters to large areas of FAE which were frequently scattered around the periphery of the domes, a location which corresponds to the sites of maximum M-cell expression (Fig. 1d). Very extensive areas of FAE representing over half of the dome surface were lost from a proportion (15% [16 of 109]) of infected domes (Fig. 1c).

FIG. 1.

SEM images of mouse Peyer’s patch tissues infected for 60 min with wild-type S. typhimurium strain IR715 resuspended in PBS (a and b) or LB (c to f). (a and b) Bacteria in PBS fail to exert a cytotoxic effect. Shown are whole dome (a) and localized area of a dome surface (b), both with an intact FAE. M cells in panel b either display a normal microvillous morphology (M) or possess membrane ruffles (arrow). (c to f) Destruction of FAE by bacteria resuspended in LB. (c and d) Whole domes, with loss of extensive areas of FAE in panel c (arrows) and loss of localized, peripheral regions of FAE in panel d (arrows). (e) Localized region of FAE showing relatively unaffected areas interspersed by damaged regions (arrows). (f) High-power view of FAE, depicting an area of FAE loss (lower left), an intermediate zone of cells denuded of microvilli (arrows), and a relatively unaffected region (upper mid-right). Bacteria are visible in the damaged areas. Bars: a, c, and d, 100 μm; b and f, 10 μm; e, 20 μm.

The dome score values obtained by SEM examination of 13 Peyer’s patches from six mice infected with IR715 in LB and 12 Peyer’s patches from six mice infected with IR715 in PBS (mean dome scores = 2.2 and 0.7, respectively [Table 1]) were analyzed by Dunn’s multiple-comparison tests. These analyses revealed that inoculum composition significantly (P < 0.001) affects the capacity of S. typhimurium to exert a cytotoxic effect on mouse Peyer’s patch FAE following 60 min of incubation in gut loops. At this time point, bacteria in PBS induced minor damage to the FAE, as indicated by the mean dome score values (0.7 for bacteria in PBS and 0.1 for control PBS [Table 1]), although this damage was not statistically significant (P > 0.05). In contrast, bacteria in LB induced extensive FAE damage which was significantly greater (P < 0.001) than that associated with the control inoculum LB (mean dome scores = 2.2 and 0.04, respectively [Table 1]). Time course studies revealed that bacteria in LB fail to induce significant FAE destruction after only 15 min of incubation in mouse gut loops (mean dome score = 0.2) but that damage is significant (P < 0.001) after 30 min and becomes increasingly extensive after longer incubation periods (mean dome scores = 1.9, 2.2, and 3.2 after 30, 60, and 120 min of incubation, respectively). After 60 min of infection in LB, bacteria grown to stationary phase under nutrient-limiting conditions induced FAE damage comparable to that induced by bacteria grown under our standard protocol where bacteria were grown in static, closed tubes (mean dome scores = 3.2 and 3.0, respectively).

TABLE 1.

Effects of inoculum composition and the invA gene on FAE destruction by S. typhimurium after 60 min of infection of mouse Peyer’s patch gut loops

Inoculum	Mean dome scorea	No. of domes/no. of Peyer’s patches examined
Control preparation
LB	0.04	67/9
PBS	0.1	66/11
PBS + 5 g of yeast extract/liter	0	10/2
PBS + 10 g of tryptone/liter	0	16/2
Filtered bacterial LB growth broth	0.1	28/4
Effect of inoculum composition
IR715 in LB	2.2	109/13
IR715 in PBS	0.7	89/12
IR715 in PBS + 0.14 M mannitol	0.8	52/7
IR715 in PBS + 5 g of yeast extract/liter	2.8	36/5
IR715 in PBS + 10 g of tryptone/liter	2.7	26/4
Effect of mutation in invA
SR11 in LB (wild type)	2.4	57/9
SB111 in LB (invA mutant)	0.1	108/15

^aDetermined by SEM examination of infected tissues. The extent of FAE destruction on each dome was assigned a score between 0 and 4, where 0 = negligible damage and 4 = very extensive damage (see Materials and Methods). 

The induction of cytotoxicity by bacteria resuspended in the original LB growth broth was dependent on the simultaneous presence of bacteria and LB broth and was independent of inoculum components which had accumulated in the LB broth during prolonged bacterial growth, since FAE destruction was absent from tissues incubated with filtered bacterial growth broth (mean dome score = 0.1 [Table 1]) and bacteria resuspended in fresh LB induced FAE destruction comparable to that induced by bacteria resuspended in the original LB growth broth (mean dome scores = 1.9 and 2.2, respectively). Salmonella-induced cytotoxicity was also independent of the pelleting and resuspension of bacteria during inoculum preparation, since IR715 prepared with or without three phases of pelleting and resuspension in the original LB growth broth induced similar levels of FAE damage (mean dome scores = 2.2 and 2.1, respectively).

To determine whether the relatively high osmolarity of LB (424 mosmol/liter) relative to that of PBS (292 mosmol/liter) might be solely responsible for the observed effect of inoculum composition on Salmonella-induced cytotoxicity, bacteria were resuspended in PBS supplemented with 0.14 M mannitol (440 mosmol/liter) and incubated for 60 min in mouse gut loops. SEM examination of infected tissues revealed that raising the osmolarity of PBS failed to enhance Salmonella-induced FAE cell cytotoxicity (mean dome score = 0.8, compared to 0.7 for bacteria in unsupplemented PBS [Table 1]). In contrast, supplementation of PBS with either yeast extract or tryptone at concentrations equivalent to those in LB was sufficient to increase Salmonella-induced FAE cell cytotoxicity (mean dome scores = 2.8 and 2.7, respectively [Table 1]) to that observed following bacterial inoculation in LB, whereas FAE damage was absent from control tissues incubated with PBS supplemented with either yeast extract or tryptone (mean dome scores = 0, [Table 1]).

(ii) M-cell invasion.

To determine whether S. typhimurium invasion of the FAE was also influenced by inoculate composition we examined tissues incubated with bacteria for only 15 min, since at this earlier time point, even bacteria suspended in LB failed to induce significant FAE cell damage, and quantification at later time points is of little value since invasion is obscured by the onset of Salmonella-induced cytotoxicity. Tissues dual stained to localize bacteria and M cells were examined by CLSM; as previously described (7, 8), in all tissues examined, the vast majority of FAE-associated bacteria were associated with the M cells. Our dual-staining technique permitted discrimination between M-cell-adherent and M-cell-invaded bacteria (Fig. 2), since the M-cell marker UEA1 strongly stains the apical surface of M cells (Fig. 2a), and consequently the position of bacteria relative to the M-cell surface can be readily identified by examination of the z series of confocal optical sections. Localization of M-cell-internalized bacteria was further assisted by very much weaker, diffuse UEA1 staining of the M-cell cytoplasm (Fig. 2c). Cumulated data from four Peyer’s patches infected with IR715 in LB (total area FAE = 2.5 × 10⁵ μm²) and five Peyer’s patches infected with IR715 in PBS (total area FAE = 2.4 × 10⁵ μm²) revealed that S. typhimurium invasion of M cells was significantly greater (P < 0.05) when inoculated in LB (3.45 ± 1.35 bacteria per 1,000 μm²; mean ± standard error of the mean [SE]) than when inoculated in PBS (0.07 ± 0.06 bacteria per 1,000 μm²; mean ± SE), demonstrating that invasion of M cells is enhanced by exposure to LB. Inoculate composition had a less dramatic effect on bacterial adhesion, although S. typhimurium adherence to M cells was still significantly greater (P < 0.05) when inoculated in LB (9.68 ± 1.07 bacteria per 1,000 μm²; mean ± SE) than when inoculated in PBS (1.14 ± 0.12 bacteria per 1,000 μm²; mean ± SE). The observed increase in bacterial adherence (8-fold) associated with inoculation in LB may be a contributory factor to the increase in invasion (49-fold) but cannot be the sole factor involved.

FIG. 2.

CLSM images of mouse Peyer’s patch FAE infected for 15 min with wild-type S. typhimurium strain IR715 suspended in LB and dually stained for M cells (a and c) and bacteria (b and d). (a and b) Surface views. Adhesion of three bacteria to the surface of an M-cell is accompanied by redistribution of M-cell surface UEA1 staining (a, arrow). (c and d) Images at 6-μm depth. A bacterium (d) has invaded an M cell (c, arrow); localization of bacteria within the M cell is facilitated by weak UEA1 staining of the M-cell cytoplasm. Bar = 5 μm.

(iii) Invasion of cultured epithelial cells.

Following a 15-min infection with S. typhimurium IR715, invasion of MDCK cells was significantly greater (P < 0.0001) when the bacteria were inoculated in LB (2.14 ± 0.58 bacteria per 1,000 μm²; mean ± SE; n = 12) than in PBS (0.14 ± 0.05 bacteria per 1,000 μm²; mean ± SE; n = 12), demonstrating that inoculum composition has an effect in vitro similar to that observed in vivo. Inoculum composition had a much smaller effect on bacterial adhesion to MDCK cells, although adhesion was also significantly greater (P < 0.05) when bacteria were inoculated in LB (0.61 ± 0.12 bacteria per 1,000 μm²; mean ± SE; n = 12) than in PBS (0.22 ± 0.06 bacteria per 1,000 μm²; mean ± SE; n = 12). In common with the results obtained in vivo, the observed increase in bacterial adhesion (3-fold) can be only a minor contributory factor in the observed increase in bacterial invasion (15-fold) associated with inoculation in LB.

Effect of mutation in the invA gene. (i) FAE destruction.

In common with IR715, infection of mouse gut loops for 60 min with the alternative wild-type strain SR11 suspended in PBS failed to induce FAE destruction (8), whereas infection with SR11 suspended in LB resulted in extensive FAE destruction (Fig. 3a and b) similar to that observed for IR715 (Fig. 1c to f). In contrast, while infection with the invA mutant SB111 induced M-cell ruffles, M-cell destruction and FAE loss were virtually absent (Fig. 3c and d) even after bacterial inoculation in LB. Statistical analyses of the dome score values obtained from 9 Peyer’s patches in three mice infected with SR11 and 15 Peyer’s patches in seven mice infected with SB111 (mean dome scores = 2.4 and 0.1, respectively [Table 1]) confirmed that, when bacteria are inoculated in LB, mutation in invA significantly (P < 0.001) attenuates Salmonella-induced FAE cell cytotoxicity. The invA mutant SB111 failed to induce FAE damage beyond that associated with the control inoculum LB (mean dome scores = 0.1 and 0.04, respectively [Table 1]), in contrast to the wild-type strain SR11, which induced significant FAE damage (P < 0.001).

FIG. 3.

SEM images of mouse Peyer’s patch tissues infected for 60 min with wild-type S. typhimurium strain SR11 (a and b) or its isogenic invA mutant SB111 (c and d), both suspended in LB. (a and b) Destruction of FAE by SR11. (a) Whole dome, showing peripheral regions of destruction (arrows); (b) localized region of FAE, depicting an area of FAE destruction (arrows) characterized by cell loss, microvillous denudation and the presence of associated bacteria, and a relatively unaffected area of FAE (upper region). (c and d) The invA mutant SB111 fails to exert a cytotoxic effect. Shown are whole dome (c) and localized area of a dome surface, both with an intact FAE. In panel d, the FAE is composed of enterocytes, a goblet cell (G), and M cells displaying a normal microvillous morphology (M) or surface membrane ruffles (arrows). Bars: a and c, 100 μm; b and d, 10 μm.

(ii) M-cell invasion.

CLSM analysis of dually stained FAE from seven Peyer’s patches infected for 15 min with SR11 (total area FAE = 8.4 × 10⁵ μm²) and six Peyer’s patches infected with SB111 (total area FAE = 5.5 × 10⁵ μm²) revealed that significantly more (P < 0.05) wild-type SR11 (1.73 ± 5.51 bacteria per 1,000 μm²; mean ± SE) than invA mutant SB111 (0.15 ± 0.07 bacteria per 1,000 μm²; mean ± SE) bacteria invaded M cells demonstrating that under these inoculum conditions, the invA mutation attenuates M-cell invasion. The invA mutation also induced a small, but not statistically significant, attenuation of M-cell adhesion (P = 0.0734), the level of M-cell adhesion by the parental wild-type strain SR11 being 4.10 ± 0.97 bacteria per 1,000 μm² (mean ± SE) compared with 1.37 ± 0.42 bacteria per 1,000 μm² (mean ± SE) for the invA mutant SB111.

DISCUSSION

We have demonstrated that the capacity of wild-type strains of S. typhimurium to induce M-cell damage and FAE disruption in the mouse gut loop model of infection is dependent on the inoculum composition. This observation accounts for the divergent results previously obtained following S. typhimurium infection of mouse gut loops. After a 60-min infection period, extensive FAE destruction is induced by bacteria suspended in bacterial growth broth (22, 31, 33) and is absent if the bacteria are suspended in PBS (7, 8, 10). The S. typhimurium-induced FAE damage observed by us in the present study and by others in previous studies (3, 22, 31, 33) was similar to the Peyer’s patch ulcerations induced by S. typhi in human typhoid fever (29), an observation which further validates the mouse model of S. typhimurium infection as a model of human typhoid. Interestingly, there is some evidence that S. typhi also induces M-cell destruction following incubation in mouse Peyer’s patch gut loops (23), although it is unclear whether this damage equates to the lesions associated with S. typhimurium infection (31).

Our results also demonstrate that in addition to regulating FAE destruction, inoculum composition determines the extent of M (and MDCK) cell invasion by S. typhimurium. Efficient M-cell invasion correlates with the induction of cytotoxicity, which suggests that M-cell invasion and Salmonella-induced cytotoxicity are closely related events. Daniels et al. (10) demonstrated that bacteria in PBS induce extensive FAE destruction only after prolonged incubation periods (180 min) in mouse gut loops. Together with our data, this observation suggests that Salmonella-induced cytotoxicity is delayed by bacterial inoculation in PBS as a consequence of the observed decrease in bacterial invasion or, alternatively, that bacteria in PBS fail to express the destructive phenotype and require growth conditions present within the tissues to stimulate expression of the appropriate genes. Although we found no evidence in the present study that cytotoxicity and invasion are separately regulated, this possibility cannot be excluded, particularly since recent evidence suggests that M-cell destruction, but not invasion, is linked to expression of slyA (10).

We have demonstrated that M-cell invasion can proceed via both inv-dependent (this study) and inv-independent (reference 8 and this study) routes. In our previous study using bacteria suspended in PBS (8), the extents of invasion by wild-type and invA mutant strains appeared to be similar, although the possibility that the invA mutation can cause a small attenuation in invasion could not be excluded since quantification was impractical due to variable and relatively low numbers of bacteria which invaded under these inoculum conditions. In the present study, bacterial suspension in LB increased invasion by wild-type bacteria, thereby facilitating quantification, and we were able to demonstrate that under these alternative inoculum conditions, inv-dependent invasion is more efficient than inv-independent invasion. Since it is now clear that when inoculated in LB, an intact invA gene is essential for efficient M-cell invasion, we must now qualify our previous observation that mutation in SPI1 has no apparent effect on M-cell invasion (8) to include only those bacteria inoculated in PBS.

In the present study, when inoculated in LB, the invA mutant was attenuated for M-cell invasion by ca. 12-fold, a figure which is less than that observed for this and other invA mutants in cultured cells (100- to 500-fold [8, 17]) but is consistent with the observations that invA mutants exhibit increased 50% lethal dose values and decreased Peyer’s patch colonization following oral inoculation of mice but still readily kill infected animals (17). Assuming that M-cell invasion is the major route of intestinal invasion by invA mutants (as indicated by our gut loop studies), the latter observation clearly demonstrates that the inv-independent route of M-cell invasion is of clinical significance. Since wild-type bacteria in PBS exhibited levels of M-cell invasion similar to those exhibited by the invA mutant, we propose that the low levels of M-cell invasion associated with bacterial inoculation in PBS may similarly be sufficient to cause disease. This observation may be relevant to the real-life situation, since similarly low levels of bacterial invasion may occur in unhealthy individuals whose diet is largely restricted to fluids. Our observation that invasion by the inv-independent route is unaccompanied by FAE destruction is consistent with the failure of other groups to detect FAE destruction by inv mutants (22, 33), although at this stage it is unclear whether failure of inv mutants to induce FAE destruction is solely a result of insufficient bacterial invasion, or whether SPI1-encoded proteins have an additional role in the destructive process. Some previous studies have failed to demonstrate M-cell invasion by SPI1 mutants (22, 33), a result which is inconsistent either with our observations (reference 8 and this study) or the relatively modest increase in oral 50% lethal dose values associated with these mutations (17, 33). We suggest that given the uneven distribution of FAE-associated bacteria (7, 19), the low levels of M-cell invasion exhibited by SPI1 mutants which we have identified by CLSM examination of large areas of FAE may be overlooked in studies which have used transmission electron microscopy to examine relatively small samples of M cells.

The mechanisms by which bacterial inoculation in LB enhances invasion and FAE destruction by wild-type S. typhimurium are unclear. It is unlikely that the observed phenomena are a consequence of variations in media pH or oxygen tension, since LB and PBS have similar pH values and the bacterial inocula were prepared in identical fashions to eliminate variations in oxygen tension. Our experiments with high-osmolarity PBS also demonstrate that the effect of LB on S. typhimurium virulence is not solely a consequence of medium osmolarity. In combination with the observation that addition of either yeast extract or tryptone to PBS raises the levels of Salmonella-induced FAE cell cytotoxicity to those associated with bacterial suspension in LB, these findings suggest that the observed increase in S. typhimurium virulence associated with bacterial inoculation in LB may be a consequence of components present in all these media. The possible effect of inoculate composition on Salmonella virulence could be mediated nonspecifically by effects on bacterial growth phase or by more subtle effects on the physiological status of the bacteria, although during our experiments we attempted to minimize the effects of variations in bacterial growth phase by using bacteria derived from overnight cultures at high concentration in the test inocula. In addition, we demonstrated that bacteria grown to stationary phase under nutrient-limiting conditions still readily induce FAE destruction when inoculated in LB. These data suggest that FAE invasion and destruction are regulated by one or more components of LB which are not exhausted by growth to stationary phase. Addition of tryptone or yeast extract to PBS had effects on FAE destruction similar to those of bacterial inoculation in LB, suggesting that amino acid supply may be critical to the destruction process. This proposal is consistent with the observation that Salmonella invasion in vitro is reduced by prolonged inhibition of bacterial protein synthesis, as is likely to occur in bacteria suspended in PBS (26). It is possible that enhanced motility of bacteria suspended in LB relative to those in PBS (unpublished observation) contributed to the increase in bacterial adhesion observed both in vitro and in vivo following inoculation in LB, although other factors may also have affected bacterial adherence. It is clear, however, that variations in bacterial adherence were not alone sufficient to account for the observed effects of inoculum composition on bacterial invasion, and additional factors must have contributed to the effect of inoculum composition on S. typhimurium virulence.

It is possible that inoculum composition modifies S. typhimurium virulence by regulating the expression of Salmonella-encoded genes which control invasion and bacterium-induced FAE destruction. This hypothesis is consistent with previous studies which have demonstrated that in vitro expression of the SPI1-encoded invasion genes (including inv) and, consequently, S. typhimurium invasion of epithelial cell lines are regulated by a variety of environmental and growth conditions, including growth phase, oxygen tension, osmolarity, and pH, via a complex group of transcription factors (1, 2, 5, 18, 21, 25, 32). The complexity of this system is emphasized by the observation that invasion gene expression is regulated in a coordinate manner and is dramatically repressed if only a single factor is suboptimal (2). In this study, we have demonstrated that the inv genetic locus is essential both for efficient M-cell invasion when bacteria are suspended in LB and for Salmonella-induced FAE destruction. From these results, it is tempting to speculate that the observed increase in M-cell invasion and FAE destruction by wild-type S. typhimurium associated with inoculation in LB may be a consequence of enhanced inv gene expression in this medium. It is clear that further studies are required to identify the medium components responsible for the observed effects of inoculum composition on Salmonella virulence and to characterize the mechanisms involved.

Our finding that S. typhimurium invasion and cytotoxicity are regulated by inoculum composition suggests that experiments designed to identify the genetic basis of Salmonella invasion and epithelial destruction must be interpreted with caution, since both the inoculum composition and dietary intake of the experimental animals may affect the observed bacterial virulence. While our studies have so far been restricted to the mouse gut loop model of infection, intragastric infection of mice with wild-type S. typhimurium similarly results in M cell destruction and FAE loss (3, 10), and it may therefore be hypothesized that inoculum composition and the nutritional state of the animal also influence the outcome of orally acquired Salmonella infections. Since any FAE loss is likely to result in extensive invasion by this and other gut luminal microorganisms, our observations suggest that the composition of the gut luminal contents (which is in turn affected by the foodstuffs ingested) is crucial for determining the outcome of naturally acquired Salmonella infections. To further investigate the possible effects of dietary intake on the outcome of Salmonella infection in the naturally infected host, future experiments should be based on conditions which more closely mimic the natural situation (i.e., oral infection of bacteria in feedstuffs or water) and the effects of inoculum size, feedstuff composition, and gastric and intestinal environments on Salmonella virulence in both healthy and diseased individuals should be determined. Interestingly, in addition to the clinical syndromes which may result from the acute stages of S. typhimurium-induced M-cell destruction and FAE loss, it has been proposed that uncontrolled penetration of microorganisms and nutritional macromolecules through damaged areas of intestinal epithelium may also result in chronic disease syndromes due to the excessive stimulation of antigen-presenting cells and activation of potentially damaging inflammatory reactions (9, 11). These pathogenic mechanisms may account for the observations that lymphoid follicle ulceration is an early feature of Crohn’s disease (15, 34) and that M-cell degeneration is present in a proportion of patients suffering from spondylarthropathy (9). The possible role of M-cell destruction in these disease processes awaits further investigation.

In recent years there has been much interest in the use of attenuated Salmonella strains as carriers of heterologous antigens in oral vaccine delivery systems (reviewed in reference 12). Our observations that inoculum composition significantly affects the virulence of S. typhimurium have important implications for the design of Salmonella-based vaccines, since they suggest that vaccine formulation and intestinal contents of the vaccine recipient may have significant effects on both vaccine efficacy and safety. Live Salmonella-based vaccines should clearly be designed such that intestinal epithelial invasion is sufficient to ensure effective immunization but that bacterial invasion occurs in the absence of significant damage to the FAE. Identification of the molecular mechanisms responsible for intestinal invasion and FAE destruction by Salmonella species must now be priority areas for research.