INTRODUCTION

Top Previous Next References

Salmonella is one of the most extensively studied bacterial species in terms of its physiology, genetics, cell structure, and development. It is also one of the most extensively characterized bacterial pathogens and is a leading cause of bacterial gastroenteritis. Despite this, we are only just beginning to understand at a molecular level how Salmonella interacts with its mammalian hosts to cause disease. Studies during the past decade on the genetic basis of virulence of Salmonella have significantly advanced our understanding of the molecular basis of the host-pathogen interaction, yet many questions remain. In this review, we focus on the interaction of salmonellae that cause enterocolitis with the intestinal mucosa, because this is the initiating step for most infections caused by these organisms.

Salmonellae are motile, gram-negative, rod-shaped bacteria belonging to the Enterobacteriaceae family; the species is a close relative of Escherichia coli. Salmonella was first described in 1880 by Eberth and cultured in 1884 by Gaffky (40). Strains were differentiated based on their reaction to sera, and for many decades each new serotype was given a new species designation (e.g., S. typhimurium, S. enteritidis, S. pullorum, and S. dublin). It is generally accepted now that there is only a single species of Salmonella (S. enterica) rather than the over 2,000 named serovars (174), but most investigators continue to write, e.g., "S. typhimurium" rather than "S. enterica serovar Typhimurium" out of convenience and for continuity with the previous literature. Thus, in this review, we refer to the Salmonella serotypes by their "species" names. While some serotypes of Salmonella such as S. typhi and S. pullorum have a restricted host range, most serotypes infect a broad range of warm-blooded animals and are capable of causing disease in humans. The majority of serovars that cause disease in humans belong to subgroup 1 (129, 234).

Salmonella is capable of causing a variety of disease syndromes: enteric fever, bacteremia, enterocolitis, and focal infections. Enterocolitis is by far the most common manifestation of disease caused by Salmonella, but bacteremia and focal infections can accompany or follow enterocolitis. Enteric fever (typhoid fever) is caused primarily by S. typhi and S. paratyphi and occasionally by other serotypes. While approximately 2,000 serotypes of Salmonella have been associated with enterocolitis, at a given time it is a smaller set of about 10 serotypes that accounts for the majority of infections; these typically include S. typhimurium, S. enteritidis, and S. heidelberg (46a, 232). The incubation period is typically 6 to 48 h and is followed by headache, abdominal pain, diarrhea, and vomiting. The diarrhea can contain blood, lymphocytes, and mucus. Fever, malaise, and muscle aches are quite common. Symptoms usually resolve within a week but Salmonella can be shed in the feces for up to 20 weeks by children <5 years of age and for 8 weeks by adults (104). Children, especially those <1 year of age, and those over 60 are most susceptible to disease and tend to have more severe infections (reviewed in references ;0104, 129, 234, and 239;0).

There are approximately 40,000 reported cases of salmonellosis in the USA each year (46b, 16, 18, 56, 151, 234); only 1 to 5% of infections with Salmonella are reported (47), thus there are actually 2 to 4 million cases in the US each year with an estimated annual cost of over $2 billion (234). The incidence of salmonellosis in the United States has steadily increased since World War II (46). The reasons for this are complex and include an increase in the proportion of the population older than 60 years (38, 39), changing agricultural and food distribution methods (33, 122, 214), increased consumption of raw or slightly cooked foods (219), an increase in the number of immunocompromised or chronically ill people, and deterioration of the public health infrastructure (12, 30, 44, 45, 133, 232-234, 253). Transmission of Salmonella to humans is usually by consumption of contaminated food, but human-to-human transmission and direct animal-to-human transmission can occur (205, 234, 254). The most common sources of Salmonella are beef, poultry, and eggs (231). The past two decades have seen an increase in the importance of eggs and egg products in transmission (4, 79, 105, 185, 195, 221, 236). Eggs can be contaminated through cracks in the shell or transovarally from an infected ovary or oviduct to the yolk prior to deposition of the shell (221, 223); internally infected eggs left at room temperature can rapidly achieve Salmonella concentrations of 10¹¹ cells per yolk (261). This mode of transmission can be especially difficult to control because the egg-laying hens are usually asymptomatic (223). S. enteritidis is a common cause of internally contaminated eggs, and as the incidence of S. enteritidis infection in chickens has increased, so has the incidence of S. enteritidis infection in humans (185, 209, 223). S. enteritidis now frequently rivals S. typhimurium as the most common cause of salmonellosis in the United States, and there has been a similar increase in many other parts of the world as well (12, 185, 209).

The first step in the disease process is transmission to a susceptible host. As mentioned above, for Salmonella, this is usually by consumption of contaminated food. It is estimated from volunteer studies that 10⁵ to 10¹⁰ bacteria are required to initiate an infection (103), but the exact amount needed varies with the strain (170-172), what is consumed with the bacteria (17), and the physiological state of the host. It is generally believed that a large inoculum is required to overcome the stomach acidity and to compete with the normal flora of the intestinal tract (reviewed in references 234 and 239). There are several lines of evidence to support this. The infectious dose decreases when Salmonella is consumed with food that traverses the stomach rapidly (i.e., liquids) or with food that neutralizes the stomach acidity (i.e., cheese, milk) (234). Individuals with high gastric pH, such as the elderly, are more susceptible to infection (15, 32, 95, 208, 222, 241, 242). It has also been shown that pretreatment of mice with streptomycin, which reduces the amount of the normal flora, decreases the dose of Salmonella required to infect 50% of the mice (35); similar effects of antibiotic treatment have been observed in humans (214).

In an article published in 1966, Kent et al. (142) wrote, "There is very little information on the distribution of the lesions in typical acute human Salmonella gastroenteritis, although textbooks frequently state that salmonellosis is a small intestine disease." These words are as true 33 years later as when they were written. There is very little information about the pathological lesions found in humans during an infection with nontyphoid Salmonella serotypes. The few reports that do exist suggest that the colon rather than the small intestine is the primary site of involvement (37, 68, 119, 161, 173). What was reported in these cases is that there is a wide range of severity ranging from slight to severe edema with infiltration of polymorphonuclear leukocytes (PMNs) and monocytes, focal inflammation of the lamina propria, degeneration of the mucosa, and extravasation of erythrocytes and PMNs. The small intestines appeared normal. It should be noted that all these cases are from patients admitted to the hospital and that in several instances the infections were fatal (37, 119). It is difficult to ascertain how representative these cases are, because most people infected with Salmonella are never hospitalized, nor do they necessarily consult a physician.

To get information about the progression of infection relevant to the human situation, Kent et al. (142) infected rhesus monkeys and examined the colonization and histopathology of various organs at several time points (1, 2, 4, and 7 days) postinfection. The symptoms that developed in these monkeys were similar to what is seen in humans (142, 213). During the first 4 days, the ileum, cecum, distal colon, and mesenteric lymph nodes (MLN) were colonized in every animal; there was sporadic colonization of the jejunum, liver, and spleen. The MLN was still colonized in all animals on day 7. In the colon (days 1 and 2), infiltration of PMNs into the mucosa, sloughing of the epithelial cells, and a purulent exudate in the lumen were observed. Occasionally, scattered crypt abscesses and microabscesses in the lymphoid patches were observed. By day 4, bacteria were seen in colonic epithelial cells and within macrophages of the lamina propria. By day 7, involvement of the colonic mucosa was more diffuse (less focal) and the lamina propria contained more chronic inflammatory cells. Progression in the ileum was slower than in the colon. On days 1 and 2, only mild lesions were observed, and these were observed only in one of four animals. However, on days 4 and 7, severe mucosal lesions were observed in 5 of 6 monkeys. There was an infiltration of PMNs and macrophages into the lamina propria, elongation of the crypts along with scattered crypt abscesses, and cuboidal changes in the surface and crypt epithelium. These results suggested that a more typical infection with Salmonella involves both the ileum and colon and that the bacteria invade these tissues and colonize (at least transiently) distal sites such as the MLN, liver, and spleen.

The most detailed ultrastructural study of an infection with Salmonella was done by Takeuchi (229) and Takeuchi and Sprinz (230) using a guinea pig model; these studies represent one of the earliest and best demonstrations of Salmonella traversing the epithelium through the epithelial cells. In this model, the guinea pigs are starved for 4 days prior to oral administration of the bacteria along with opium (to slow intestinal movement); only the ileum was examined in these studies. By 12 h, bacteria could be found within epithelial cells and within PMNs, and some of these PMNs were between epithelial cells or in the lumen. Bacteria that were in very close contact with the brush border seemed to cause degeneration of the microvilli and protruding cytoplasm much like the "ruffles" described in later experiments with mice and cultured cells (137, 146, 259) (Fig. 1B shows an example of ruffles). The initial phagocytic compartments containing bacteria were usually quite large, but they contracted as the vesicles moved to the subnuclear region. Villus rather than crypt epithelium was the primary site of involvement. By 24 h, an intense inflammatory exudate was observed and the brush border was reduced in height. At this time, some bacteria were found within desquamated epithelial cells and within PMNs in the lumen. However, many bacteria were in phagocytic cells within the lamina propria and Peyer's patches (PP). At 48 h, the inflammation was more extensive and extended to the submucosa where bacteria could be found within phagocytes (Fig. 1A shows an example of bacteria within a phagocyte). Extrusion of epithelial cells, emptying of goblet cells, and elongation of crypt cells were also observed.

View larger version (243K): [in this window] [in a new window]   FIG. 1.   Interaction of Salmonella with intestinal mucosa. Transmission electron micrographs of non-follicle-associated epithelia from calves infected with either wild-type S. typhimurium ST4/74 (A and B) or with an isogenic invH mutant (C) are shown. (A) Macrophage from the laminae propriae of the absorptive villi with several internalized bacteria 3 h postinfection. (B) Enterocyte with an internalized bacterium. Note the membrane ruffling. (C) Enterocyte with attached bacteria (invH mutant). No ruffling was observed. Bars, 1 µm. Reprinted from reference 248 with permission of the publisher.

Similar interactions of Salmonella with the mucosa were observed in a rabbit ileal-loop model (96). In this model, the bacteria were seen adhering to the brush border as early as 3 h postinfection and the bacteria could be seen on and in epithelial cells and within the lamina propria by 7 h. Epithelium overlying lymphoid tissue was the most heavily involved and showed acute inflammation. By 12 to 18 h postinfection, as with the guinea pig model, PMNs were observed within the epithelium, in the lumen, and in the lamina propria. By this time, the villi were blunted and swollen and the crypts were hyperplastic. Similar results were found in two other studies with the rabbit ileal-loop model (245, 257). This model was also used in conjunction with a HeLa cell assay for invasion to test a variety of different S. typhimurium strains (94); there was a strong correlation between the ability to invade HeLa cells and the ability to invade the rabbit intestinal mucosa. More recently, the infection of calves with Salmonella has been examined as a model because calves exhibit an enteritis similar to that of humans (220). The pathology observed in this model is very similar to that described above (220, 244, 245, 248) (Figure 1). The fact that similar interactions of Salmonella with the intestinal mucosa occur in at least four different mammalian models (monkeys, guinea pigs, calves, and rabbits) suggests that similar interactions occur during human infections with Salmonella.

Mice are used as a primary animal model of infection today because of their ease of use, low cost, and susceptibility to infection with Salmonella. Oral infection of mice leads to colonization initially of the distal ileum and cecum and then of the draining lymph nodes by 48 h (43). In another study, it was found that although the PP represent only a small fraction of the small intestine, nearly 25% of the Salmonella cells within the small intestine were in the PP (127). It was also found that most of the bacteria found in the cecum could be easily washed off the tissue but that most of the bacteria in the small intestine and PP remained associated with the tissue after extensive washing (127).

Ultrastructural analysis of mouse ligated intestinal loops infected with S. typhimurium revealed that the bacteria initially invaded M cells almost exclusively (137); similar results were found in mouse ligated loops infected with S. typhi (146). In both these studies, the M cells frequently were associated with numerous bacteria and were ultimately destroyed by the invading bacteria. For these studies, high concentrations of bacteria were used (10⁸ to 10⁹ cells) in a relatively small ileal segment (4 to 5 cm); thus, it is possible that the destruction of the M cells observed is at least in part a function of the high bacterial load. As with the studies in rabbits, calves, and guinea pigs, host cells in close contact with Salmonella exhibited membrane ruffling during the invasion process (137, 146). It is not clear why exclusive invasion of M cells and destruction of the M cells was observed in the mouse ligated loop model but not in the other models. This difference could be due to the animal model chosen, the method of inoculation, the time point for observation, or differences in the test strains of Salmonella. However, it is clear that regardless of the model chosen, an early manifestation of the host-pathogen interaction is attachment to and invasion of the intestinal epithelium by the bacteria and subsequent inflammation of the lamina propria and lymph nodes. Thus, much of the research on Salmonella virulence has focused on identifying and studying the bacterial products involved in triggering these responses by the host.

Prior to invasion of any cell type, bacteria must encounter and attach to one or more cell types found in intestinal tissue. Such tropism by S. typhimurium may involve several types of fimbriae or pili, four of which have been genetically defined.These include type 1 fimbriae (Fim), plasmid-encoded (PE) fimbriae, long polar (LP) fimbriae, and thin aggregative fimbriae (curli). Some studies suggest that these fimbriae each have a specific tropism for a certain cell type or for cells from particular animal species (25, 26) (Table 1). S. typhimurium induces transepithelial migration of neutrophils in an in vitro assay (168); therefore, it has been hypothesized that fimbriae might play a role in the recruitment of neutrophils to a site of infection (26). Fimbriae might not be directly involved in neutrophil recruitment, but it is possible that they help the bacteria attain close contact with host cells and therefore allow for the interaction of factors that do stimulate transepithelial migration of neutrophils.

The organization of genes encoding the putative fimbriae of S. typhimurium is similar to that of other characterized fimbrial loci of related members of the family Enterobacteriaceae. The functions assigned to many of the Salmonella fimbrial proteins have been assigned based on information from the best-characterized system, that of type 1 fimbriae (55). It is notable that not all fimbriae (type 1 or otherwise) are exclusively associated with virulence, since nonpathogenic strains may contain one or more of the known fimbrial operons. Following is a description of four genetically defined loci in S. typhimurium and their contributions to pathogenesis (Table 1 and Fig. 2 give summaries of these loci).

View larger version (15K): [in this window] [in a new window]   FIG. 2.   Organization of genes within the four fimbrial operons of S. typhimurium. Gene lengths (in base pairs) are indicated below each open reading frame. Genes encoding the major fimbrial subunits are indicated in yellow, genes encoding chaperones are indicated in green, genes encoding ushers are indicated in blue, genes encoding minor subunits are indicated in black, and genes encoding regulatory proteins are indicated in red.

S. typhimurium type 1 fimbriae (Fim) are encoded by the fimAICDHF operon at centisome 15 (63, 215) and are morphologically similar to but antigenically distinct from that of the E. coli type 1 fimbriae (147). In E. coli and Salmonella, type 1 fimbriae are defined as peritrichous, 7 nm wide, and 0.2 to 2.0 µm (145, 147). Type 1 fimbriae specifically bind -D-mannose receptors on various eucaryotic cell types; therefore, binding of bacteria with type 1 fimbriae to eucaryotic cells can be inhibited by the addition of -D-mannose in vitro ("mannose-sensitive fimbriae") (145). As in other members of the Enterobacteriaceae, type 1 fimbriae of S. typhimurium consist of a major subunit, FimA (21 kDa), and several other associated proteins including the FimH adhesin (55). Although the function of FimH as an adhesin appears the same in S. typhimurium and E. coli, the respective proteins are not highly similar (147) and do not show identical binding specificities (82). Perhaps these differences dictate to which tissues or surfaces these bacteria can bind.

E. coli and S. typhimurium undergo phase variation between a fimbriated and a nonfimbriated state (72). Although growth conditions that result in reduced or increased expression of fimA have been studied, it is not clear how environmental signals manifest a change in fimbriae expression (54). Moreover, the mechanism of phase variation of type 1 fimbriae in E. coli and S. typhimurium appears to be unrelated. E. coli fimA expression is regulated by a 314-bp promoter region that can undergo an inversion resulting in an on or off state (3). Unlike E. coli, S. typhimurium fimA has not been shown to be regulated by an invertible DNA sequence, since both fimbriated and nonfimbriated strains are observed to contain promoters in an on orientation (54). Perhaps not surprisingly, S. typhimurium does not appear to contain fimB or fimE, which encode putative recombinases involved in the inversion of the E. coli fimA promoter (144). On the other hand, the E. coli fim locus does not contain fimY or fimW, which may play a role in regulation of S. typhimurium fimA (258). In S. typhimurium, FimZ directly binds the fimA promoter and is required for expression of fimA, whereas no ancillary fim gene plays a direct role in the regulation of E. coli fimA. It should be noted, however, that some E. coli strains appear to have a FimZ homologue (71% identity) in the 5' region of argU (fimU in S. typhimurium) based on sequence comparison analysis (258). However, it is not known if this homologue plays any role in the fimbriation of E. coli. Expression of a S. typhimurium fimA-lacZ reporter in E. coli (turned off) differs from that of the native E. coli fimA, further suggesting that these loci are regulated by different mechanisms (228).

The S. typhimurium type 1 fimbriae mediate adhesion to HeLa cells in vitro; however, they do not appear to play any role in adhesion to various other tissue culture cell types such as the human cell lines HEp-2, T84, and Int-407 or the canine cell line MDCK (25). At least two independent studies have shown that a fim deletion mutant appears to be slightly more virulent than a fim⁺ strain (158, 240). It has been suggested that type 1 fimbriated S. typhimurium strains as well as E. coli strains are removed by the liver more efficiently in the mouse model (155); therefore, it is not clear what role, if any, type 1 fimbriae play in adhesion or virulence of S. typhimurium in the mouse or other animal hosts.

The lpfABCDE (for "long polar fimbriae") locus was identified in a search for S. typhimurium chromosomal loci not present in related members of the Enterobacteriaceae (23). This operon was mapped to centisome 80 of the S. typhimurium LT2 chromosome and is flanked by sequences homologous to those in E. coli K-12, suggesting that lpf may have been acquired by horizontal transfer during the evolution of S. typhimurium. Homologues of lpf have not been found in various other pathogenic Enterobacteriaceae, including enterotoxigenic, enteroinvasive, and enteropathogenic E. coli and Shigella. Moreover, at least two Salmonella strains, S. typhi and S. arizoneae, appear to have lost or never acquired lpf (22).

Expression of the lpf operon in a nonfimbriated E. coli strain results in the appearance of polar filaments in a subpopulation of transformants (23), hence the LP designation. Although the fimbriae encoded by lpf suggest a polar location on the bacterial cell surface, it has not been determined conclusively that LP fimbriae are polar on Salmonella. There is no direct evidence that lpf actually encodes the LP fimbriae; it is possible that lpf induces the expression of cryptic fimbrial genes on the E. coli chromosome. However, in addition to evidence described below, the organization of the lpf genes is similar to that of fim and the deduced Lpf protein sequences are homologous to proteins encoded by fim, strongly suggesting that lpf does encode components of fimbriae (23, 26).

LP fimbriae mediate adhesion to the cells of the Peyer's patches of the small intestine in a mouse model of infection (27). In an in vitro, mouse small intestine model of infection, a mutation in lpfC (which encodes a putative outer membrane usher for fimbrial assembly) resulted in reduced colonization of PP but not of villous enterocytes. Complementation of the lpfC mutation restored the ability of S. typhimurium to associate with the PP and also enhanced its ability to colonize the villous intestine. An lpfC mutation has only a modest effect on virulence in a mouse model (a threefold increase in oral 50% lethal dose [LD₅₀]). A mutation inactivating InvA, a component of the type III secretion system required for invasion, had a moderate effect on virulence in a mouse model (16- to 50-fold increase in oral LD₅₀). However, the lpfC invA double mutant was significantly less virulent than the wild-type strain or either single mutant (150-fold increase in oral LD₅₀) (28). Neither mutation alone or in combination with each other affected virulence by intraperitoneal infection, supporting the hypothesis that lpfC and invA are required for interaction of S. typhimurium with the intestinal mucosa. Perhaps for salmonellae to efficiently invade host tissues by using the type III secretion system, the bacteria must first be associated intimately with target cells. This requirement could be fulfilled by use of LP fimbriae which appear to be specific for cells in the PP, the primary site of infection by salmonellae in the mouse model (137).

A lpf-lacZYA transcriptional fusion reporter strain was tested to characterize the expression of lpf in vitro and in vivo (189). In vitro, phase variation from a Lac⁺ (on) to a Lac (off) colony phenotype and from a Lac to a Lac⁺ colony phenotype occurred at a frequency of 6.8 × 10³ and 2.4 × 10⁴, respectively. However, mice infected with the lpf-lacZYA reporter strain showed an increased selection for bacteria in an LP-on phase of expression. This selection was observed for bacteria isolated from PP but not for bacteria isolated from the MLN or spleen. Mice that were immunized with a glutathione S-transferase-LpfA (major fimbrial subunit) fusion protein prior to infection showed a strong selection against the LP-on expression state in the PP. These results together suggest that LP fimbriae are important at an early step of disease by an oral route of infection. The mechanism of regulation of expression of LP fimbriae has not yet been characterized, but it is thought that inverted repeats flanking lpfA might mediate phase variation of LP fimbriae by causing inversion of the lpfA gene (23).

Several serotypes of Salmonella contain virulence plasmids ranging in size from 50 to 90 kb (108). S. typhimurium contains a 90-kb virulence plasmid (pSLT) which is required for full virulence of the organism after oral infection (110). The role of plasmid-encoded virulence determinants, anything from invasion to serum resistance to survival, is not clear due to conflicting reports on their function (111); however, it has been generally accepted that the virulence plasmid of S. typhimurium is important for causing a systemic infection after oral inoculation of experimental animals (109). Several virulence loci have been cloned and identified from the virulence plasmid, including the spv (for "Salmonella plasmid virulence") (109) and rck (for "resistance to complement killing") (113) genes. The spv region is sufficient for complementation of splenic infectivity of a plasmid-cured strain (109). The rck gene (see below) can confer in vitro serum resistance to serum-sensitive strains of E. coli and plasmid-cured Salmonella (53, 113, 120); however, its specific role in S. typhimurium virulence in vivo has not been determined. Upstream of rck, a 13.9-kb region was sequenced and determined to contain another putative fimbrial operon called pefBACD (for "plasmid-encoded fimbriae") (85). Only four serotypes, S. typhimurium, S. enteritidis, S. choleraesuis, and S. paratyphi C, contain pef sequences, as determined by DNA hybridization to a pefA probe (26). The significance of this finding has yet to be established.

Transmission electron micrographs of nonfimbriated E. coli with pef sequences carried on a cloning vector reveal numerous peritrichous fimbriae (24, 85). E. coli producing PE fimbriae can also adhere to histological sections of murine small intestine more effectively than can nonfimbriated E. coli (24). A mutation in the putative outer membrane usher gene, pefC, does not affect adhesion to HeLa or T84 human cell lines of either S. typhimurium or E. coli carrying the pef locus (25). This is consistent with experiments demonstrating that plasmid-cured strains of Salmonella could still adhere to and invade cultured CHO (Chinese hamster ovary) cells as efficiently as the plasmid-containing strain could (158). The effects of a mutation in pefC were also tested in an intestinal-organ culture (IOC) model. The IOC model in this study involved infection of BALB/c mouse small intestines which had been removed from the mice just prior to infection and maintained under tissue culture conditions. Under these perhaps more native conditions, the S. typhimurium pefC mutant did not attach to the small intestine as well as the wild-type bacteria did when the two strains were used to coinfect the same tissue sample. This result was mirrored by coinfection with nonfimbriated E. coli carrying the pef operon on a plasmid and a plasmidless strain in the IOC model. The authors concluded that pef enhanced attachment to villous small intestine; however, it was not clear in this experiment if "villous intestine" included the M cells. Experiments to distinguish attachment to PP from villous small intestine have been performed with LP-fimbriated strains (27); therefore, it should be possible to test the PE-fimbriated strains in the same way.

The pefC mutant was also tested in an infant suckling-mouse model of infection (24). According to these studies, PE fimbriae were necessary for attachment to the small intestine and for fluid accumulation in infant mice. As a control, mutations in the downstream rck locus and another fimbrial locus (fim) were also tested in this model and shown to have no effect on fluid accumulation. When E. coli carrying the pef genes on a plasmid was tested in this system and compared to a plasmidless parent strain, no effect on fluid accumulation was seen, demonstrating that PE fimbriae were necessary but not sufficient for fluid accumulation in this model.

Thin aggregative fimbriae (3 to 4 nm wide) (curli) were identified and purified from S. enteritidis (62) and found to be antigenically similar to E. coli curli (210). The locus encoding S. typhimurium curli, agfBAC, was cloned (226) and found to be organized identically to the E. coli csgBAC genes (114). S. typhimurium agfBAC maps to centisome 26 on the LT2 chromosome (63). (Due to the location of agf near a breakpoint of an 815-kb inversion that distinguishes S. typhimurium from S. enteritidis, it should be noted that the S. enteritidis agf genes map between centisomes 40 and 43.3 [63].) As in E. coli, agfC does not appear to be part of the agfBA transcript based on Northern analysis with an agfC-specific probe (60, 210). A divergent operon, agfDEFG, homologous to csgDEFG of E. coli was also found (210, 226). Although agf and csg encode similar proteins, they have low identity at the nucleotide level. However, due to the high level of identity between the csg- and agf-encoded components, csg mutants of E. coli can be complemented by corresponding S. typhimurium agf genes (210).

Expression of agf affects colony morphology, since curli-producing bacteria form a rigid multicellular network within the colony, a phenotype referred to as rdar (211). The rdar phenotype can be observed when bacteria are grown at ambient temperatures (i.e., below 37°C), in rich media and low osmolarity, or at 37°C under iron starvation (210). This inducible phenotype has provided a simple assay for studying the regulation of expression of curli genes. The rdar phenotype is dependent on rpoS, which encodes an alternate sigma factor required for transcription of genes in response to stress and starvation conditions (165), and ompR, a transcriptional activator of the ompC and ompF porin genes (210, 218, 235). Both rpoS and ompR are required for expression of agfD/csgD, which encodes a putative transcriptional activator (211). Agf/CsgD is believed to activate the transcription of agf/csgBA, which encodes the surface-exposed nucleator and major fimbrial subunit, respectively, but it is not known if Agf/CsgD directly interacts with the agf/csgD promoter (210, 211). It is also unknown if RpoS and OmpR directly interact with the agf/csgD promoter.

Curli-producing bacteria tend to autoaggregate, a phenomenon which has been suggested to enhance the survival of salmonellae facing hostile barriers such as stomach acid or other biocides they may encounter (61). Another hypothesis suggests that curli are involved in attachment of bacteria to the host cell epithelium. This idea is supported by experiments showing that curli but not SEF 21 (for "Salmonella enteritidis fimbriae 2") (type 1) or SEF 14 (not found in S. typhimurium) mediate the binding of S. enteritidis to fibronectin, a component of the eucaryotic extracellular matrix (61, 193). An in vitro system has been developed which shows that curli play a significant role in attachment to differentiated murine small intestinal cells (226). In vivo, a preferred site of attachment by curli-producing salmonellae has yet to be determined. The LD₅₀ of an agfB mutant of S. typhimurium was increased only 3.3-fold compared to wild-type bacteria when infection occurred intragastrically (240). Despite this modest effect on LD₅₀, it is clear that agf significantly contributes to pathogenesis. Mouse experiments have revealed that a mutation in agfB, the putative surface-exposed nucleator of AgfA fimbrin subunits (115), dramatically increases the LD₅₀ of a fim/pefC/lpfC mutant (AJB12) from 1.7 × 10⁵ to 1.5 × 10⁷ cells (an 88-fold increase in oral LD₅₀). This was especially notable since AJB12 is slightly more virulent than wild-type bacteria, most probably due to a dominant fim phenotype.

Rck was originally identified as a "cryptic plasmid"-encoded protein that conferred high levels of serum resistance to both E. coli and plasmid-cured, serum-sensitive Salmonella strains (113). Rck is a member of a family of outer membrane proteins (OMPs) including PagC (for "phoP-activated gene") of S. typhimurium (112) and Ail (for "attachment and invasion locus") of Yersinia enterocolitica (180, 181). Unlike PagC, Rck is more similar to Ail in that both proteins confer invasiveness and serum resistance to noninvasive, serum-sensitive E. coli strains (34, 53, 113, 120, 121, 198). Both proteins are predicted to have multiple transmembrane domains and exposed loops based on amino acid sequence analysis (53, 180). Site-directed mutagenesis and "domain-swapping" experiments have been done with Rck and show that loop 3 is required for serum resistance and invasion in E. coli (53). Unfortunately, most probably due to the presence of at least one other invasion system and rck-independent serum resistance mechanisms (227), a specific role for Rck in S. typhimurium pathogenesis has not been reported. Moreover, virulence plasmid-cured strains carrying only the spv genes can cause disease in a mouse model almost as well as wild-type strains can (109), suggesting that rck is not important in the penetration of the intestinal epithelium. However, it is possible that Rck plays a role in the infection of nonmurine hosts of S. typhimurium.

The presence of at least four fimbrial systems and Rck suggests that attachment to cellular or noncellular surfaces may be a critical step in the survival of S. typhimurium in the environment. Due to the apparent redundancy of fimbrial operons in Salmonella, it has been difficult to assess the role of each system in an in vitro or in vivo assay, because one system may compensate for another. van der Velden et al. have shown by transmission electron microscopy that even an attenuated quadruple fimbrial mutant can still produce what appears to be a fifth fimbrial structure (240). It will be interesting to see how mutations in five fimbrial loci will affect attachment and virulence in vivo.

Several studies have shown that salmonellae preferentially attach to and invade cells in the PP of the small intestine in orally infected mice (127, 137, 197). However, bacteria can also be found in nonphagocytic enterocytes (229). The mechanisms of Salmonella invasion, that is, the stimulation of nonphagocytic cells to internalize bacteria, are clearly complex. Putative invasins that stimulate the uptake of bacteria by epithelial cells have been identified in organisms such as Yersinia by introducing genomic libraries of invasive strains into E. coli "laboratory" strains (134, 181). Little success has been achieved in the search for invasion genes by this technique in S. typhimurium. In S. typhi, one locus, invABCD*, was identified by this technique (77); however, nothing is known about how this locus functions in invasion (the asterisk indicates that a nonhomologous locus with the same name was identified in S. typhimurium). Interestingly, S. typhimurium also has the invABCD* region (182) in addition to the genetically linked invasion system encoded at centisome 63 (updated from the old map position of centisome 59 [215]). This locus, known as Salmonella pathogenicity island 1 (SPI1), is believed to have been acquired by horizontal transfer from another pathogenic bacterial species during its evolution (107) (see Fig. 3 for a map of SPI1).

The first cloned SPI1 genes, invABC(D), were identified by complementing an attenuated, invasion-defective strain of S. typhimurium with a cosmid library of S. typhimurium genomic DNA (89). It was determined subsequently that invD was not actually a SPI1 gene but represented a noncontiguous fragment which had been cloned into the complementing cosmid (182). Since the discovery of invABC, many more genes have been identified in SPI1. The 40-kb SPI1 region encodes at least 33 proteins which include components of a type III secretion apparatus (13, 29, 76, 89, 90, 102, 107, 136, 138, 196), regulatory proteins (19, 135, 138), and secreted effector proteins and their chaperones (59, 87, 131, 139-141, 156). In addition to the SPI1 system, S. typhimurium has two other known type III secretion systems: the SPI2 (Salmonella pathogenicity island 2) system for macrophage survival (centisome 30.5) (123, 124, 191) and the flagellar assembly system for motility (centisomes 27 and 42 to 43) (215) (for a recent review of type III secretion, see reference 130). The hallmark of type III systems is that none of the secreted proteins has a conserved or recognizable signal sequence. Although several type III secretion systems have been identified in many different organisms, the mechanisms by which proteins are secreted has yet to be elucidated.

At least three proteins encoded within SPI1 are known to participate in a supramolecular structure which reaches from the cytoplasmic membrane to the outer membrane. Electron microscopy has visualized this structure, which resembles a syringe (149). This syringe is believed to secrete effector proteins from the bacterium, and these proteins stimulate dramatic cytoskeletal rearrangements in eucaryotic host cells (81, 84, 229). These membrane ruffles facilitate the engulfment of the bacteria by eucaryotic cells. Mutations in genes which encode the apparatus proteins, regulators, and certain secreted proteins eliminate or greatly diminish invasion and the appearance of membrane ruffles (90, 101). In addition to invasion, the SPI1 system appears to function in programmed cell death (apoptosis) of infected cultured macrophages (48, 187) and the recruitment of PMNs to the small intestines in a bovine model of infection (92).

Limited LD₅₀ analysis has been done with nonpolar SPI1 mutants. Transposon insertions in invA (28, 89), invF, invG, hilA, sipC, sipD, spaR, and orgA (197) result in a 16- to 100-fold attenuation of S. typhimurium by oral inoculation. Most of these mutations result in the loss of a functional type III secretion and/or translocation system. Mutations in individual putative effector genes result in subtle in vivo phenotypes, suggesting that no single protein is sufficient for pathogenesis in the mouse model.

The genes encoding the secretion machinery fall into two clusters: inv-spa and prg-org (Fig. 3). The two are separated by several other genes including those which encode secreted proteins (Sip/Ssps, SptP) and an essential regulator (HilA). Mutations in almost any of the apparatus genes are predicted or have been shown to result in the absence of several proteins from culture supernatants.

View larger version (11K): [in this window] [in a new window]   FIG. 3.   Organization of genes in SPI1 at centisome 63. Gene lengths (in base pairs) are indicated below each open reading frame. Gaps longer than 30 bp between genes are indicated above intergenic spaces. Genes encoding regulatory proteins are indicated in red, genes encoding secreted proteins are indicated in mauve, genes encoding chaperones are indicated in green, and genes encoding apparatus proteins or proteins of uncharacterized function are indicated in black.

invH. InvH (16.5 kDa) was originally identified as both an attachment and an invasion factor for S. typhimurium in vitro (13). Stone et al. also identified invH in a screen for invasion mutants of S. enteritidis; however, a mutation in invH did not affect attachment in vitro (224). Recently, Daefler and Russel elegantly showed that InvH is actually an outer membrane protein required for the proper localization of InvG, a PulD-like outer membrane component of the SPI1 type III secretion system (67). Although InvH localizes to the outer membrane, it does not appear to be an integral membrane protein; it is likely to be anchored to the outer membrane by a lipid moiety. InvH was determined to have a consensus lipoprotein sequence which could be labeled by [³H]palmitic acid. Deletion of the consensus sequence abolished palmitic acid labeling. Secretion of one of the SPI1 type III apparatus substrates, SipC (described below), was eliminated in an invH deletion mutant, confirming InvH participation in the secretion apparatus. Secretion could be restored by invH in trans, but an invH clone missing its lipoprotein signal sequence could not complement the invH mutation (67). In light of the above evidence, it seems unlikely that InvH normally acts as an attachment factor; instead, it is likely to be an essential structural component of the type III secretion apparatus.

invG. Amino acid sequence analysis of the SPI1 type III syringes demonstrated that InvG is a major component of these structures (149). InvG belongs to a family of proteins known as secretins (206) and is required for the secretion of proteins by the SPI1 type III apparatus (67, 117, 118, 138, 141). InvG is similar to MxiD of Shigella (8), YscC of Yersinia (148, 178), and PulD of the Klebsiella type II pullulanase secretion system (71). Based on these similarities, InvG was assumed to be targeted to the outer membrane (138). Because the PulD secretin requires a lipoprotein for proper localization, it was believed that a lipoprotein may also be involved in the localization of InvG to the outer membrane. Three putative lipoproteins had been identified in SPI1, i.e., InvH (13, 67), PrgH (29, 196), and PrgK (29, 196). InvH, but not PrgH or PrgK, was required for InvG localization (67). In wild-type S. typhimurium, InvG was found in density gradient fractions containing outer membrane proteins; in an invH mutant, InvG was found mostly in the inner membrane fractions. This was confirmed in another study which showed that InvG could form ring-like structures in the outer membranes of E. coli (66). Similar to the previous study, targeting of these structures to the outer membrane required InvH (66). Unlike PulD, which requires the PulS lipoprotein for stabilization of the PulD polypeptide in addition to proper localization, InvG does not require InvH for stabilization (67). Although InvH was shown to be required for the proper localization of InvG in the outer membrane, it is likely that another putative lipoprotein, PrgK (see below), also interacts with InvG, because it was purified from the secretion syringes (149).

invE. invE was identified upon sequence analysis of the region where invABC was found (102). invE encodes a 43-kDa polypeptide required for invasion in vitro. In addition, membrane ruffling was not observed upon analysis of electron micrographs of cultured cells infected with an invE mutant.

invA. InvA encodes a putative inner membrane protein of 71 kDa homologous to type III secretion proteins from various animal- and plant-pathogenic bacteria (130). In addition, InvA has homologues in systems involved in the biosynthesis of flagella in E. coli, Salmonella, and other bacterial species. InvA and MxiD of Shigella are so similar that mxiD could complement an invA mutant almost as well as invA itself (100). Yersinia LcrD is also similar to InvA. However, unlike mxiD, lcrD of Y. pseudotuberculosis could not complement an invA mutation (100). TnphoA mutagenesis and amino acid sequence analysis predict that LcrD (in Y. pestis) is an integral membrane protein with eight transmembrane domains and a C-terminal cytoplasmic domain (199, 200). Based on this model, in addition to TnphoA mutagenesis of invA, it is believed that InvA assumes a similar structure in the inner membrane of Salmonella (89, 90). However, the N-terminal sequences of InvA and LcrD were found to be more similar to each other than were the C-terminal sequences. Ginocchio and Galán (100) hypothesized that the C-terminal domains were required for the recognition and secretion of species-specific proteins and that the conserved N-terminal sequences were not as important for this specificity. To test this, chimeric proteins between InvA and LcrD were constructed and tested for their ability to complement an invA mutant. When the N-terminal region of LcrD was fused to the C-terminal domain of InvA, the chimeric protein complemented the invA mutant for invasion nearly as well as invA did. However, a chimeric protein consisting of the N-terminal half of InvA and the C-terminal domain of LcrD could not complement an invA mutant. These results supported the idea that the C-terminal cytoplasmic domains of these proteins were essential for the recognition of species-specific substrates intended for secretion or for species-specific interactions with other apparatus components.

invB. invB is located 24 bp downstream of invA and encodes a protein with a predicted molecular mass of 15 kDa (76). Like InvA, InvB is homologous to proteins of other type III secretion systems, but its function is unknown. A kanamycin resistance cassette in invB did not affect invasion in vitro; therefore, its role in the secretion apparatus may not be essential under these conditions. Similarly, in Shigella, a mutation in the invB homologue, spa15, does not affect invasion in vitro (217). InvB has not been analyzed further, probably due to the apparent lack of phenotype associated with invB mutants.

invC. The start codon of invC overlaps the stop codon of invB, suggesting that the two genes are cotranscribed (76). Unlike invB, a mutation in invC was shown to reduce invasion considerably in vitro. Invasion of cultured epithelial cells by the invC mutant was partially restored by providing invC in trans; however, the mutation in invC (a kanamycin resistance cassette disruption) probably had a polar effect on expression of the downstream genes.

invI (spaM). In a continuing effort to sequence SPI1, Collazo et al. identified invI and determined that it was needed for invasion in vitro (59). Groisman and Ochman also identified this and other genes while analyzing sequences found in Salmonella but not E. coli K-12 (107). Unlike many of the type III secretion proteins, InvI/SpaM does not have significant homology to proteins of other organisms. Proteins with low identity to InvI/SpaM have been found in Shigella (SpaM) and Salmonella (FliJ), but their similarity is so low that it is difficult to make any meaningful comparisons between them. It is also not known where these related proteins localize in the cell or how they function.

spaP, spaQ, spaR (surface presentation of antigen). As mentioned above, Groisman and Ochman identified several genes which were not found in E. coli K-12 but were homologous to genes found in Shigella (107). spaP was identified downstream of spaO (see below) and predicted to encode a protein of 24 kDa. A spaP mutant has a dramatically reduced capability for invasion in vitro (58, 107). The Shigella homologue of SpaP, Spa24, was able to fully complement the spaP mutant for invasion in vitro (107). SpaP is believed to be localized to the inner membrane; however, evidence showing this has not been reported.

spaS. The transcriptional organization of the inv-spa-sip genes has not been confirmed, but it appears that spaS might be the last gene of the inv-spa cluster. This is similar to the organization of genes encoding SpaS homologues in other systems including yscU of pYV (Yersinia virulence plasmid) and ssaU of SPI2 in Salmonella (11, 124). SpaS, like InvA, is a highly conserved protein of about 40 kDa and is homologous to proteins from several other organisms including Yersinia, Shigella, and Pseudomonas (see reference 130 for a complete list). The best-characterized homologue is Y. enterocolitica YscU, which is believed to be an inner membrane protein (11). YscU is required for secretion of Yops; therefore, it has been assumed that SpaS is also required for secretion of proteins from S. typhimurium.

prgH (phoP-repressed gene). prgH was identified in a screen for genes repressed in a pho-24 mutant (29). A pho-24 mutation [phoQ(Con)] results in a constitutively active form of PhoP, the response regulator of the phoP/phoQ two-component regulatory system (80, 179). Unlike many of the type III secretion proteins, PrgH has not been found to be homologous to any other protein. PrgH is predicted to be a 45-kDa lipoprotein and a major component of the SPI1-associated syringe structures (29, 149, 196). Upon amino acid sequence analysis of the syringe structure components, PrgH did not appear to be processed. Syringe structures containing functional PrgH tagged with an epitope were labeled with monoclonal antibodies specific for the epitope tag; the antibodies appeared to preferentially label the base of the syringe structures (149).

prgI and prgJ. prgI and prgJ are predicted to encode proteins of 8.8 and 10.9 kDa, respectively (196). Neither protein contains a recognizable signal sequence, and both are highly similar to secretion proteins from Shigella (MxiH and MxiI for PrgI and PrgJ, respectively) (9) and Yersinia (YscF and YscI for PrgI and PrgJ, respectively) (178). Little is known about the function of prgI and prgJ because mutations in these genes have not been constructed and tested for invasion in vitro. Moreover, it is not known where these proteins or their homologues in other systems localize in the bacterium.

prgK. PrgK (28 kDa), like PrgH, is believed to be a membrane-associated lipoprotein (196) and is a major component of the type III secretion syringe structure (149). Unlike PrgH, PrgK is similar to proteins from other systems including MxiJ of Shigella and YscJ of Yersinia (10, 130, 178). Like PrgI and PrgJ, it is not conclusively known where PrgK localizes or if a mutation in prgK would abolish invasion in vitro. However, due to its participation in the formation of syringe structures, it is predicted that PrgK would be membrane associated and that a prgK mutant would be noninvasive (149).

orgA (oxygen-regulated gene). orgA was identified in a screen for genes which were expressed under oxygen-limiting conditions (136). OrgA mutants do not secrete invasion-associated proteins into culture supernatants; therefore, OrgA is presumed to be a part of the type III secretion apparatus (197). Little is known about the function of OrgA (48 kDa), but it is required for invasion into cultured epithelial cells and for cytotoxicity to macrophages (187). In addition, a transposon insertion in orgA results in attenuation of S. typhimurium in intragastrically infected mice (197).

Ten proteins secreted by the SPI1 secretion apparatus have been genetically characterized, and eight of them are encoded within SPI1 (Table 2). Mutations in the apparatus genes result in the absence of these proteins from culture supernatants; however, it is possible that these proteins are secreted via other type III systems under different conditions (124, 128). Three putative chaperones which themselves are not likely to be secreted have also been identified.

(i) invJ (spaN) and spaO. InvJ is encoded within the inv-spa cluster of genes, which predominantly encodes apparatus proteins (59), and it does not have significant similarity to any other protein. Secretion of InvJ (36 kDa) is stimulated by the presence of 10% BCS or cultured epithelial cells (260). Interestingly, fixed tissue culture cells could not stimulate the secretion of InvJ, the significance of which has not been determined. Immediately downstream of invJ, spaO was identified with the putative spa apparatus genes. Like InvJ, SpaO is a secreted protein dependent upon the SPI1 secretion machinery (107, 156).

(ii) sicA (Salmonella invasion chaperone). sicA encodes a putative chaperone for one or more of the secreted proteins encoded in the sip operon (140). Chaperones have been identified in other type III secretion systems and are required for the secretion, stabilization, and/or translocation of effector proteins (50, 87, 154, 176, 204, 249, 251, 252). Nothing is known about the interaction between SicA and the secreted polypeptide(s), but it is required for invasion in vitro. Several secreted proteins have been identified in Yersinia, some with specific chaperones (249, 251, 252). It is intriguing that the four Sip proteins may share one chaperone; however, it is not known which proteins actually require SicA. SicA is homologous to IpgC of Shigella, a chaperone required for the stabilization and secretion of the Shigella invasins IpaB and IpaC (for "invasion plasmid antigen") (176). Perhaps SicA functions similarly and interacts with the Salmonella IpaB and IpaC homologues, SipB and SipC, respectively.

(iii) sipB/sspB (Salmonella invasion protein or Salmonella secreted protein). Several groups independently identified sipB (the locus containing sipB has been designated sip and ssp, but for simplification, we refer to it as sip) in a cluster of genes encoding secreted proteins required for invasion of S. typhimurium and S. typhi into tissue culture cells (125, 131, 140). sipB is located immediately downstream of the putative chaperone gene sicA and upstream of four additional open reading frames. SipB (63 kDa) is highly similar to Shigella IpaB, a protein required for invasion and cytotoxicity of shigellae in vivo and in vitro (21, 41, 164, 262). Both SipB and IpaB localize within eucaryotic cells (57, 237). For SipB to be translocated into eucaryotic cells, SipC and SipD must also be present (57). Moreover, SipB, SipC, and SipD are required for the translocation of other proteins, including the protein tyrosine phosphatase SptP (88, 141) and SopE (256). It has been proposed that SipB, SipC, and SipD are involved in the translocation of secreted effector proteins into the eucaryotic cytoplasm; however, it is not known how these proteins, which are themselves translocated (except SipD), accomplish this. Unlike InvJ and SpaO, the Sips are not believed to be involved in the secretion of proteins from the bacterium (57).

(iv) sipC/sspC. SipC is a 42-kDa protein homologous to Shigella IpaC and is encoded downstream of sipB in S. typhimurium (131, 140). Like sipB, sipC is required for invasion of Salmonella and for translocation of SipB and SptP into eucaryotic cells (88). SipC is itself translocated into cultured epithelial cells, but its function within the eucaryotic cell is unknown (57). As observed with SipB, salmonellae do not need to be internalized in order to translocate SipC into eucaryotic cells; treatment of the eucaryotic cells with cytochalasin D (which inhibits invasion) prior to infection does not inhibit the appearance of intracellular SipC