Chapter 154

Environmental Regulation of Virulence Gene Expression in Escherichia,Salmonella, and Shigella spp.

MICHAEL J. MAHAN, JAMES M. SLAUCH, and JOHN J. MEKALANOS

INTRODUCTION

The ability to proliferate in fluids or within cells of a living host is the key attribute of pathogenic bacteria that distinguishes them from commensal species. Like all bacteria, their primary objectives in this interaction are survival and multiplication. The genesis of disease within the host is usually simply a consequence of virulent bacteria successfully achieving these goals by utilizing all the evolutionary adaptations they have at their disposal. In the most general sense, any product that a bacterium synthesizes which enhances the growth or survival of a bacterium during its interaction with the host can be thought of as a virulence factor and its corresponding coding sequence a "virulence gene." Once we decide to eliminate the prerequisite for a virulence factor to be directly involved in host injury, a more expansive definition emerges that allows us to consider many factors whose phenotypes have been defined experimentally in vitro but whose role in vivo has been at best extrapolated based on presumed conditions inside the host.

Included in this broad category of virulence properties are some well-understood adaptive responses to environmental stresses and cues which have been apparently co-opted in the control of bacterial virulence gene expression. Groups of genes responding to given environmental stresses can be thought of as virulence regulons even if their contribution to the infection process and pathogenesis remains relatively obscure. Thus, it follows that some global regulatory genes mediating bacterial responses to certain environmental stresses can be considered sensors of host signals and the controllers of virulence gene expression.

Classical virulence factors are usually not essential components of cell structure or function. The toxins, adherence factors, invasins, capsules, and outer membrane proteins responsible for complex virulence phenotypes (e.g., mucosal surface colonization or intracellular invasion) are accessory products that generally have little known use to bacteria outside of the host. Accordingly, bacteria seldom express virulence genes constitutively, presumably due to the need to exercise economy through the expression of appropriate genes in the appropriate environment. In a natural setting, we expect that the expression of some virulence factors is activated shortly after introduction of the pathogen into the host in order to maximize the chance of establishing a successful infection, while other "nonvirulence" genes whose functions are inappropriate to this new environment are quickly repressed. Not all virulence factors confer a selective advantage to the microbe at the same stage of infection or at the same anatomical site within the host. Consequently, expression of certain virulence factors must be modulated in response not only to signals heralding the transition from external source to host, but also to other environmental signals encountered throughout the infection cycle. Therefore, virulence gene expression is a mosaic of different responses that allow the bacteria to efficiently adapt to their varying environment. It must further be appreciated that the individual responses in this mosaic may all exist in the same bacterial cell or, alternatively, that the overall effect may be the additive results of different responses occurring in different cells.

In the laboratory setting, we can occasionally establish environmental conditions that mimic the microenvironments encountered inside the host. For example, pathogens no doubt encounter changes in temperature, osmolarity, oxygen tension, pH, and nutrient deprivation during infection. Indeed, the expression of certain classical virulence genes in vivo appears to follow the same regulatory response that has been determined in the laboratory. In these cases, presumably the virulence genes often belong to a relatively well-understood global regulatory network, and the changes that cells undergo during the adaptive responses may themselves contribute to virulence directly. It is these categories of virulence gene expression that we understand most clearly and will focus on in this chapter.

REGULATORY MECHANISMS CONTROLLING VIRULENCE GENE EXPRESSION

Mechanistically, virulence gene expression is controlled at the DNA level by cis- or trans-acting regulatory elements. cis-Acting control mechanisms leading to changes in the level and composition of bacterial surface components are found, for example, in Neisseria gonorrhoeae. These examples include DNA rearrangement and slipped-strand DNA synthesis mechanisms leading to phase variation (Pil⁺ to Pil^–) and antigenic variation (PilE variants) (for review see reference 105). However, it appears that virulence genes are most commonly regulated by positive and negative transcriptional regulatory proteins (including PilE of N. gonorrhoeae). In an effort to identify possible connections between virulence properties of pathogenic bacteria and the regulatory components or global responses of Escherichia coli and Salmonella spp., we have compiled a short list of regulatory genes and responses implicated in the environmental control of virulence in diverse bacterial species (Table 1).

Table 1Regulatory proteins and environmental signals implicated in virulence regulation in diverse bacterial speciesa

Studies of global regulatory systems controlling virulence gene expression have uncovered regulatory genes in pathogenic microorganisms that have clear homologs in E. coli and Salmonella that may not have been previously recognized as potential virulence regulators per se. For example, ToxR was originally identified as a transmembrane, DNA-binding protein involved in the transcriptional activation of cholera toxin and other virulence genes in Vibrio cholerae. Recently, a topologically similar membrane protein, CadC, has been described that regulates the expression of genes involved in polyamine biosynthesis (181). Interestingly, both ToxR and CadC respond to both pH and the composition of growth media, but it is not known whether ToxR regulates polyamine biosynthesis in V. cholerae or whether CadC regulates virulence properties of E. coli.

For the majority of examples listed in Table 1, the specific host signals that the regulatory protein detects are not understood, although environmental signals that modulate expression of virulence genes have in many cases been identified empirically in vitro. The homology between regulators noted in references can also be misleading since such homologies may reflect more an underlying molecular mechanism (e.g., protein kinase domain, helix-turn-helix motif, etc.) than a commonality in the regulatory response. For example, there are several examples noted in Table 1 of virulence regulatory systems that are modeled on the now classical "two-component motif" displayed by OmpR-EnvZ of E. coli, including VirA-VirG in Agrobacterium, AlgR1-AlgR2 in Pseudomonas, BvgA-BvgS in Bordetella, and PhoP-PhoQ in Salmonella (87), each of which apparently recognizes largely different environmental signals. Similarly, homologs of the E. coli and Salmonella AraC proteins include VirF in Yersinia (29), FapR in E. coli (100), and ToxT in V. cholerae (86), none of which recognizes arabinose as a signal. In contrast, some regulatory responses are conserved across diverse genera as illustrated by the high degree of conservation of structure and function for the iron-responsive proteins highly homologous to Fur of E. coli, such as in V. cholerae (69) and Pseudomonas aeruginosa (150). Another example is the highly conserved LuxR-LuxI homologs present in Vibrio, Agrobacterium, and Pseudomonas which respond to the presence of autoinducer in the environment (reviewed in reference 48).

ENVIRONMENTAL SIGNALS IMPLICATED IN VIRULENCE GENE EXPRESSION

Certain environmental signals are frequently found associated with the global control of virulence gene transcription, although the reasons for this are at times unclear. For example, low iron concentration and elevated temperature have been proposed to be parameters that signal entry of the microbe into mammalian host tissues where the relatively high temperature of 37°C and the presence of iron-binding host proteins contrast with the low ambient temperature and abundant ferric ion that characterize the environment. Similarly, elevated CO₂ concentrations could accurately reflect the gaseous conditions in metabolizing host tissues. However, regulation of virulence gene expression by abundant ions such as Ca²⁺ is much more difficult to rationalize in terms of host entry cues or microenvironments present in host tissues. Accordingly, it is probably best to appreciate these signals as surrogate but useful environmental cues that may not accurately reflect the actual environmental signals being used by pathogenic bacteria when present in host tissues. What follows is a discussion of a subset of the best-understood environmental signals known to affect the global control of virulence in E. coli, Salmonella, and Shigella.

Iron

In the mammalian host, essentially all iron is bound to protein and sequestered. Ferric iron (Fe³⁺) is virtually insoluble, and ferrous iron (Fe²⁺) is readily oxidized to ferric iron under extracellular conditions. Thus, in the body, most iron is bound intracellularly to ferritin and heme, or extracellularly to transferrin and lactoferrin (for review see references 22, 74, and 183). For most pathogens to adapt and grow in the host, they must have the ability to acquire iron from these high-affinity sources, and iron plays a significant role in pathogenesis. Indeed, during an infection, the host actively reduces the availability of iron (22, 183). Several mechanisms have been proposed. In one model, polymorphonuclear leukocytes, in response to interleukin-1, release lactoferrin at the site of infection. Lactoferrin binds iron better than transferrin at low pH. The lactoferrin is then taken up by macrophages, removing iron from the immediate environment (102, 180). It has also been proposed that the synthesis of ferritin in the liver results in the sequestering of the iron pools (104). Whatever the mechanism, hypoferremia is a clear result of infection, and this serves as an efficient nonspecific defense mechanism. Indeed, injection of iron compounds has been shown to enhance the virulence of a number of bacterial pathogens, including E. coli and Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium) (22).

Many pathogens have the ability to acquire iron in the host (183). Salmonella spp. and E. coli produce both low- and high-affinity iron uptake systems. For example, both organisms are capable of producing the siderophore enterobactin (73, 136, 137). This molecule has an extremely high affinity for iron and effectively removes iron from the mammalian iron-binding proteins. Enterobactin is part of a chromosomally encoded system that includes several outer membrane, inner membrane, and cytoplasmic proteins that are responsible for binding the siderophore, transporting it into the cell, and finally cleaving the molecule to release the iron. Virtually all wild strains of E. coli produce enterobactin (153), and the siderophore has been shown to be produced in vivo (75). However, the role of enterobactin in the virulence of E. coli is not clear, although high-affinity iron uptake is correlated with virulence (see below). The role of enterobactin production in Salmonella spp. is also controversial. Reports of the effects on the virulence of S. typhimurium are mixed (10, 184), whereas S. typhi strains defective in enterobactin synthesis showed decreased ability to grow in HeLa cells in culture and were attenuated in a mouse mucin model (60). Shigella flexneri, an intracellular pathogen, grows normally in HeLa cells in the absence of siderophore-dependent iron uptake systems (146). Thus, the apparently different requirements for iron uptake systems may depend on the assay used and/or the subcellular localization of the pathogen. Certain pathogenic strains of E. coli can also produce another siderophore, aerobactin; the relevant genes can be located either chromosomally or on the pColV plasmid (135). The presence of this additional high-affinity system has been associated with virulence (22, 183). Thus, even though enterobactin has a higher affinity for iron than aerobactin does in vitro, the latter may play the more important role in the host (22, 183).

The roles of iron uptake systems in virulence may differ depending on the anatomical site of infection. For example, Stojiljkovic et al. (172) have reported that E. coli mutants defective in ferric uptake systems based on enterochelin and citrate were capable of efficiently colonizing the intestines of mice, whereas mutants defective in ferrous uptake (feo mutants) were highly defective in colonization. These data suggest that, in the anaerobic environment of the intestine, the Fe²⁺ form of iron is a readily available source, whereas in aerated tissues and aerated mucosal surfaces Fe³⁺ may be more abundant. Consistent with this conclusion, Camilli et al. (23) reported that a low-iron-inducible gene fusion in V. cholerae was not expressed during intestinal infection but was highly derepressed during infection of the mouse peritoneal cavity. Thus, the expression of appropriate iron-assimilating systems might be modulated by host environmental factors that alter the form as well as the concentration of available iron.

E. coli α-hemolysin may also have a role in iron acquisition. This hemolysin is in the RTX family and is produced by 30 to 60% of extraintestinal isolates of E. coli. There is a correlation between the presence of the hemolysin and virulence, and in some cases this could be eliminated with the injection of FeSO₄ (for review see reference 17).

In addition to the iron uptake systems, other virulence factors are also induced in response to low iron. Thus the bacterium uses the low-iron environment as a general signal that it is in the host. For example, in E. coli, the Shiga-like toxin, encoded on a bacteriophage, is induced in response to low iron (chapter 153). In E. coli and Salmonella, regulation by iron is mediated by the Fur protein (171). The apoprotein binds Fe²⁺ as cofactor, and the cofactor-bound protein binds to various sites, termed iron boxes. This complex negatively regulates many of the genes involved in iron uptake, as well as the toxin genes, hemolysin, etc. Thus Fur is an important virulence regulator.

Other adaptations to low-iron environments include the undermodification of certain tRNA species. In tRNA species that read codons beginning with U residues, the A residue 3' to the anticodon is hypermodified to 2-methylthio-N ⁶-(Δ ²-isopentenyl)adenosine (ms²i⁶A-37; 27, 28, 73). The enzymatic addition of the thio group is iron dependent. Therefore, during iron limitation, these tRNA species are undermodified from ms²i⁶A-37 to i⁶A-37. The phenotypic results of this undermodification include an increase in aromatic amino acid biosynthesis (21) and a significant increase in aromatic amino acid transport (20). These changes in gene expression could be the result of changes in translational efficiency mediated by the tRNA species (21). Enterobactin is synthesized from chorismic acid, an intermediate in aromatic amino acid biosynthesis, and it has been proposed that the undermodification of the tRNA leading to changes in the aromatic amino acid pathways could be important for the adaptation to a low-iron environment (73). It should also be noted that this undermodification of tRNA in response to limiting iron is mutagenic (27, 28). Thus, in addition to the wide variety of genes regulated by Fur, low-iron adaptation involves sophisticated changes in metabolism and increases in the mutation rate.

Temperature

Growth temperature is an environmental cue controlling the expression of many bacterial virulence determinants in a variety of pathogens. For example, Shigella spp. respond to changes in growth temperature (from 30°C to 37°C; 89, 123) by triggering the expression of plasmid-encoded virulence factors essential for invasion of, replication within, and spread to adjacent cells of the colonic epithelium. This environmental regulation occurs at the transcriptional level, and the regulatory genes are located on the plasmid as well as on the chromosome (reviewed in reference 82). Temperature control is mediated through the plasmid-encoded VirF and VirB proteins in a regulatory cascade in which VirF activates the transcription of virB and VirB, in turn, activates the transcription of plasmid genes directly involved in invasion (ipa), intercellular spread (icsB), and secretion of these virulence factors (mxi and spa) (176, 177). There is one known exception: VirF directly activates icsA (a structural gene involved in intercellular spread) (3, 156; for review see reference 149). Thermoregulation appears to be mediated through virB since its transcription level is low at 30°C (compared to 37°C) and derepression of invasion genes can occur in the absence of VirF by constitutive expression of virB via the tac promoter (176).

In addition to the plasmid-encoded regulatory proteins, the invasion genes are also under control of the chromosomally encoded H-NS (VirR), a histonelike protein involved in changes in DNA topology (36, 90, 124, 149). At low temperatures, hns mutations result in derepression of invasion genes (124). Thus, virB appears to be under the positive control of VirF and negative control of H-NS. VirF, a member of the AraC class of DNA-binding proteins (155, 156, 157), can bind to the upstream region of virB and activate its transcription in a DNA topology-dependent fashion. H-NS can bind to virB upstream regions and block transcription in vitro (177). Similar H-NS effects have been reported in two other thermoregulated systems involving expression of the CFA/1 pilus operon (96) and the plasmid-encoded virulence genes of enteroinvasive E. coli (32). Taken together, these data have led to the suggestion that H-NS may play, in addition to its structural role in DNA topology, a specific role in the regulation of gene expression in response to temperature. However, this suggestion is highly controversial, and the molecular mechanism underlying thermoregulation warrants further investigation. Along these lines, a recent report indicates that hns-mediated Shigella virulence gene expression contains an osmotic component as well (149). Transcription of invasion genes was repressed when cultures were grown under conditions of low osmolarity even at the permissive temperature. The repression was relieved by an hns mutation.

E. coli strains are capable of colonizing a variety of different mucosal surfaces, and in many cases colonization is dependent on the expression of specific adhesins that are often physically associated with fibrous appendages termed pili or fimbriae. The production of pili is often regulated in response to temperature as well as other growth conditions (e.g., liquid versus solid medium, medium composition, and aeration). In only a few cases is the molecular basis for this complex regulation understood. For example, uropathogenic strains often express pyelonephritis-associated pili or Pap. The expression of the genes required for pilin production is regulated by a variety of factors, including temperature, such that they are not transcribed at lower temperatures, e.g., 25°C. Selection for mutations that allowed the expression of one of the transcripts at low temperature yielded mutations in hns (52, 71) and rimJ, which encodes the N-terminal acetylase that modifies the ribosomal protein S5 (15, 182). The mechanism by which these proteins affect temperature regulation of pap is not clear.

Expression of the pap operon at higher temperatures is regulated by several factors including phase variation. This is an epigenetic event that is mediated by Lrp (leucine-responsive regulatory protein) in association with PapI (chapters 11 and 150). Transcription of the pap operons is also activated by cyclic AMP (cAMP)-cAMP receptor protein (CRP) (52). Under conditions of high cAMP, transcription of papI is activated by cAMP-CRP. Increasing PapI concentration increases the probability of being in the phase-on state (117).

E. coli can produce a variety of pili, many of which are regulated by environmental factors and phase variation (81, 117; chapters 11 and 150). For example, transcription of the sfa operon, encoding S fimbrial adhesin, is affected by growth rate, catabolite repression, osmolarity, and temperature (133, 159). Production of the aggregative adherence fimbria I from enteroaggregative E. coli is regulated in response to temperature, osmolarity, oxygen, pH, and media composition (134). Lrp has been shown to regulate the daa (F1845 pili; 179), sfa (179), fan (K99 pili; 16), fae (K88 pili; 91), and fim (type 1 fimbriae; 14) operons in E. coli, as well as the pef operon (plasmid-encoded fimbriae; 117) of S. typhimurium. Interestingly, whereas the regulation of daa, sfa, fae, and fan, like that of pap, involves methylation, phase variation of fim is mediated by DNA inversion. Thus, Lrp is capable of regulating phase variation by a variety of seemingly different mechanisms.

Osmolarity

Several aspects of S. typhimurium virulence are affected by the osmolarity of the growth medium. Galan and Curtiss (62) showed that the ability of S. typhimurium to invade cultured epithelial cells was significantly impaired when the bacteria were grown in a low-osmolarity medium. This defect was reversed with the addition of 300 mM NaCl. These authors showed that the transcription of the invA gene, required for invasion, is induced when the bacteria are grown in high-osmolarity medium (62). This induction was not dependent on ompR (see below). Whether the transcriptional induction of the inv operon is sufficient to explain the enhanced invasion in high osmolarity is not clear.

Invasion of cultured epithelial cells by S. typhi is also affected by the osmolarity of the medium in which the bacteria are grown. Tartera and Metcalf (175) showed that both adherence and invasion of epithelial cells by S. typhi were maximal when cells were cultured in medium with an osmolarity of 676 mosmol (LB, 300 mM NaCl). The cultural conditions did not apparently affect the ability of the bacteria to survive or propagate once inside the epithelial cells. The molecular basis of this induction is not understood.

OmpR and EnvZ make up a two-component system that regulates the major outer membrane porins, OmpF and OmpC, in response to medium osmolarity (for review see reference 163). EnvZ is an inner membrane sensor component that communicates environmental information to OmpR, a transcriptional regulator. These proteins belong to a family of environmental regulators that control a variety of responses from chemotaxis to the regulation of virulence factors in plant and animal pathogens (for review see reference 87). Dorman et al. (35) showed that insertion mutations in ompR dramatically attenuated virulence of S. typhimurium in a mouse model. Insertion mutations in ompF or ompC alone did not affect virulence. However, it was subsequently shown that strains containing mutations in both ompF and ompC were attenuated, although the effect was not as severe as that seen in an ompR mutant (25). In E. coli, loss of both OmpF and OmpC results in a general loss of diffusion across the outer membrane (for review see reference 163). This may explain the virulence defect associated with the loss of porins in S. typhimurium, although this strain also produces the OmpD porin, which is not regulated by OmpR. Mutations in ompD resulted in a 10-fold increase in the oral 50% lethal dose (35). However, the fact that ompR mutants are more attenuated than the ompF ompC double mutants suggests that the regulator has a further role in virulence. In S. typhimurium, OmpR is also known to regulate tppB, the gene encoding tripeptide permease (68), but tppB insertion mutations had no effect on virulence.

Pickard et al. (148) recently showed that synthesis of the Vi capsule in S. typhi was affected by osmolarity and mutations in ompR. Strains containing a deletion of ompR were no longer agglutinated by Vi antiserum, and this defect could be complemented by a plasmid containing the ompR gene. These ompR strains were also defective in the production of OmpF and OmpC. Agglutination by Vi antiserum was also affected by the osmolarity of the culture medium, such that agglutination decreased with increasing osmolarity. The authors provided evidence that it was production and not export of the polysaccharide that was affected.

OmpR and EnvZ also seem to have a significant role in the virulence of Shigella flexneri. Several operons on the virulence plasmid, required for the synthesis and export of products involved in invasion and spread of Shigella flexneri in epithelial tissue, are regulated in response to osmolarity. Using a mxiC-lacZ transcriptional fusion, Bernardini et al. (11) showed that transcription of the mxi operon (membrane expression of invasion plasmid antigens) is induced in high osmolarity. Furthermore, this osmoregulation was not seen in an ompR deletion background. Indeed, expression of the operon was reduced 10-fold in the ompR mutant. However, the ompR deletion strain was reported to have normal expression of Ipa proteins as measured by immunoblot assay using serum from a monkey convalescing from shigellosis. In contrast, a strain containing an uncharacterized spontaneous mutation in either ompR or envZ did show a significant reduction in the expression of the Ipa proteins, and this defect was complemented by a plasmid containing ompR and envZ. The icsB gene, located in an operon that is divergently transcribed from mxi, is also induced in high osmolarity at 37°C (149). This operon also contains the genes encoding the Ipa proteins. Both the mxi operon and the icsB-ipa operon are positively regulated by VirB, and the relationship between VirB regulation and OmpR regulation is not understood. Osmoregulation of these operons is also affected by mutations in the hns gene (149).

Shigella flexneri strains containing a deletion of ompR and envZ were also affected in several in vitro virulence assays. The mutant showed a significant reduction in invasion of HeLa cells and was also negative in a plaque assay, wherein the bacteria form a clearing zone on a confluent monolayer of HeLa cells, indicating a defect in intracellular multiplication and cell-to-cell spreading (11). Interestingly, it was subsequently shown that strains containing insertion mutations in ompC were also defective in invasion and intercellular spread (12). Strains containing mutations in ompF were unaffected. Moreover, the ompR-envZ deletion strain was restored to full virulence by the introduction of a plasmid containing the ompC gene from E. coli. Thus, the main effect of the ompR-envZ deletion appears to be the loss of OmpC. However, the ompR mutant is more affected than the ompC mutant (12). Whether this is due to the lack of both porins and the consequent loss of diffusion or to the ompR effects on plasmid gene expression is not clear.

Oxygen

Oxygen tension may serve as a major environmental cue that signals anatomical (locational) information to the pathogen, triggering alterations in gene expression essential to its pathogenesis. Accordingly, the entry of S. typhimurium into cultured mammalian cells is regulated by oxygen tension; invasiveness is repressed when the organism is grown under conditions of high oxygen and induced under low oxygen (45, 111, 158). The oxrA regulatory circuit involving the anaerobic induction of several proteins was found not to be required for Salmonella invasiveness (111). Such an in vitro induction profile may reflect the in vivo situation in which invasion genes may be expressed in the presumed low-oxygen environment of the intestinal lumen. Another important component to invasiveness is growth phase: salmonellae lose their invasiveness in stationary phase (45, 111).

A regulatory screen designed to search for invasion mutants which invade aerobically (normally repressing conditions) has resulted in the isolation of the hyperinvasion locus (hil) (112), which maps to min 59 of the S. typhimurium chromosome, a 40-kb region that appears dedicated to invasion (6, 77, 110). This region also contains the invasion genes invHFGEABC (6, 39, 61, 64) and spa (77) (surface presentation of antigen), which show similarity to the Shigella spp. virulence plasmid genes mxi and spa which are required for entry of Shigella spp. into mammalian cells and are involved in processing and excretion of surface components that engage in the invasion process. The inv genes are present in all Salmonella serovars tested (63). In addition to these inv genes, S. typhi contains invasion genes that reside in a region close to, but distinct from, the inv region (min 58) (41).

Another essential invasion gene, prgH, also maps to min 59 (see phoP genes in pH section, below). prgH expression is modulated by oxygen and appears to be under the control of the hil negative regulatory locus hilR (9, 110). Curiously, induction of prgH in vitro occurs under high-O₂ conditions (in late log phase), not the low-O₂ conditions expected in the intestine (9). There are at least three possibilities for the oxygen paradox: either (i) a host factor induces the expression of prgH in the presumed low-oxygen environment of the intestine, (ii) PrgH may function as an invasin (or another function) in a high-oxygen environment at another stage in S. typhimurium pathogenesis (e.g., invasion of splenic cells), or (iii) PrgH may be synthesized prior to ingestion by the host. It should be noted that in contrast to invasion mutations such as inv (61), which show a defect only after oral (and not intraperitoneal) delivery (consistent with their proposed functions contributing to invasion of the intestinal epithelium), prgH mutants confer a defect after oral and intraperitoneal administration in a murine typhoid fever model, suggesting that PrgH has an additional function subsequent to invasion of the intestinal epithelium (9).

Since a low-oxygen environment induces the ability of S. typhimurium to invade mammalian cells, a screen for invasion genes was devised based on the isolation of transcriptional fusions that were induced under oxygen-limiting conditions (94). These fusions were subsequently scored for a defect in an in vitro invasion assay, leading to the identification of an invasion gene, orgA (oxygen-regulated gene). orgA shows no sequence similarity to known genes but does map between the prgH and hil invasion loci and is repressed by hilR (94), similar to prgH (9, 110). The orgA mutant was shown to be attenuated for virulence only after oral administration in a mouse model, suggesting that OrgA may function in the process of entry into intestinal cells (similar to inv) and not at later stages of infection.

Oxidative Stress

Neutrophils and macrophages have several mechanisms to kill bacteria. These include the oxygen-dependent respiratory burst that generates superoxide anion (O₂ ^–) and hydrogen peroxide (H₂O₂) (7, 101). Activated murine macrophages also produce nitric oxide (NO) (92). Neutrophils from patients with chronic granulomatous disease are incapable of producing a respiratory burst. These patients have a high mortality rate from bacterial infections including Salmonella spp. (154), suggesting that the respiratory burst is crucial in the defense against bacterial pathogens.

E. coli and S. typhimurium have regulons designed to deal with the cytotoxic products produced during the respiratory burst, with a large number of proteins being induced in response to H₂O₂ or O₂ ^– (for review see references 33 and 34). The OxyR protein activates transcription of approximately nine genes in E. coli in response to H₂O₂. These include katG, encoding HPI catalase, and ahpFC, encoding NADPH-dependent alkyl hydroperoxidase. It was presumed that these gene products would be important for the survival of pathogenic strains. However, it was recently shown that oxyR or katG mutants of S. typhimurium do not show increased sensitivity to killing by human neutrophils in vitro (145). Although this study was done using the laboratory strain of S. typhimurium, LT2, there is also a report that the double catalase mutant (katE katG) of the virulent S. typhimurium strain ATCC 14028 does not show a significant increase in sensitivity to killing by murine macrophages or reduced virulence in a mouse infection model (19). Perhaps even the induced catalases are incapable of inactivating H₂O₂ at the level needed to survive in the phagocyte.

SoxR and SoxS control the expression of approximately 10 genes that can be induced in vitro in response to certain redox-cycling agents such as paraquat (33, 34). SoxR is activated by these compounds to stimulate transcription of soxS (140). The SoxS protein then activates the transcription of a variety of genes including sodA, encoding the Mn-dependent superoxide dismutase, nfo, encoding the DNA repair enzyme endonuclease IV, and micF, which posttranscriptionally decreases production of the major outer membrane protein OmpF (26, 34). Although the agents that induce SoxRS in vitro were thought to do so by increasing the levels of O₂ ^– internal to the cell, externally added O₂ ^– does not induce the regulon (139). However, it was recently shown that the SoxRS regulon is activated in response to NO produced by activated murine macrophages. Moreover, Δ soxRS mutants of E. coli showed decreased survival in mouse peritoneal macrophages compared to the wild-type control (139). Interestingly, strains mutant in nfo or micF also showed reduced survival equal to that of the soxRS mutant. A strain containing a mutation in sodA also showed a survival defect, although this strain survived slightly better than the other mutants in this assay (139). These results suggest that the SoxRS regulon may be important for survival in macrophages but that the regulon’s role in virulence may extend beyond protection against superoxide.

Heat Shock

One of the most hostile environments that a pathogen must overcome during infection is within the host cell macrophage. Here the pathogen is challenged with a variety of antimicrobial mechanisms that have been classified as oxygen dependent (i.e., formation of reactive oxygen species [7]) and oxygen independent (i.e., exposure to low pH of the phagolysosome [50], degradative phagocytic granules [7, 40], and antimicrobial defensins [115, 116]). The ability to survive in this environment is central to Salmonella pathogenesis (and other intracellular pathogens) since mutants isolated as defective in macrophage survival in vitro also confer virulence defects in vivo (49, 50, 130).

Experiments with cultured murine macrophages indicate that S. typhimurium responds to potentially damaging intracellular stimuli by increasing the synthesis of >30 stress-induced proteins. Two of the most abundant were heat shock proteins (hsps) GroEL and DnaJ (2, 18), which function as protein chaperones (reviewed in references 66 and 99), are immunodominant antigens for many infections, and have been implicated in autoimmune diseases (65, 95, 97, 185). htrA (high temperature requirement) is another hsp locus essential for S. typhimurium macrophage survival and full virulence (8, 50, 93). The htrA locus of E. coli encodes a protease, DegP, that degrades abnormally folded proteins located in the periplasm (173) and may serve to prevent their toxic accumulation under stress-inducing conditions such as bacterial growth in macrophages.

The functions of hsps in protecting the bacterial cell from the possible disruptive effects of abnormally folded proteins and their induction in several different stress situations (e.g., heat, oxidative stress, and macrophage survival) suggest that hsps contribute to bacterial survival during infection (reviewed in reference 109); an S. typhimurium 66-kDa hsp that cross-reacts with hsp60 (GroEL) was shown to mediate mucin binding in an in vitro assay (44), suggesting that bacterial hsps may have a direct interaction with host tissues. Their abundance and inherent immunogenicity in antigen processing and presenting phagocytes may contribute both to virulence and to the fact that bacterial hsps are immunodominant antigens (18, 65, 97, 109, 185). Hsps also serve as targets for both antibody and T-cell responses critical to immune protection (42, 43, 138, 185).

pH

Shigella spp. have a remarkably low infectious dose as compared with other enteric bacteria. Oral inoculation of 10 to 500 organisms is sufficient to cause dysentery in healthy adults (37), contrasting sharply with an oral infectious dose of 10⁵ to 10¹⁰ organisms for Salmonella spp. (13, 88) and 10⁸ organisms for V. cholerae (24). Relative resistance to gastric acid may be the defining factor in the infectious dose; in vitro studies indicate Shigella spp. and E. coli enjoy a >10⁵-fold viability advantage over S. typhimurium when exposed to pH 2.5 for 2 h. This acid tolerance is highly dependent on growth phase (72). Maximal acid resistance (and base resistance) is exhibited at stationary phase and is dependent on rpoS (164), which encodes a stationary-phase sigma factor, σ ^S (108, 125, 161; chapter 106). Additionally, growth of log-phase cultures under moderate pH conditions (pH 5.5 to 6.0) induces extreme acid-resistance mechanisms (pH 2.5), termed habituation (70, 152), similar to the acid tolerance response in S. typhimurium (see below) (54, 56). Acid resistance in E. coli and Shigella flexneri follows a common theme of sharing regulatory circuitry between pH regulation and O₂ tension (reviewed in reference 144). Anaerobic growth restores the acid resistance of stationary-phase-grown rpoS mutant strains, indicating that rpoS is not an absolute requirement under all growth conditions (164). Coregulation of genes responding to acidity and anaerobiosis may reflect changes in respiration and fermentation that occur during growth under low-oxygen conditions (reviewed in reference 144).

Survival in acid may have clinical relevance, as enteric pathogens must pass through the stomach at a pH of <3 for up to 2 h before colonization of the intestine (67). Although the correlation of in vitro acid resistance and low infectious dose is compelling, conclusive evidence requires testing of acid-sensitive mutants in volunteers because the infectious dose of Shigella spp. is low only in humans. However, an acid-susceptible mutant of Shigella flexneri has been shown to confer a defect in the ability to survive passage through the gastrointestinal tracts of mice (165). Although Shigella spp. do not cause dysentery in mice, these data do show a correlation between in vitro acid sensitivity and the ability to survive the required acid challenge of the stomach, an obligatory stage that precedes colonization of the intestinal tract.

Although the infectious dose of S. typhimurium is substantially higher than that of Shigella spp., S. typhimurium nonetheless harbors acid resistance mechanisms that presumably contribute to its survival in the acid environment of the stomach and the phagolysosome. S. typhimurium is capable of growth at pH values down to 5.0, but below pH 4.0 the organisms undergo a swift acid death that is attributed to their inability to maintain an internal pH suitable for viability (53, 56, 57, 72, 85). Logarithmically grown S. typhimurium can mount a two-stage protective response to low pH, termed the log-phase adaptive acid tolerance response (ATR), that promotes survival at extreme low-pH conditions (pH 3.0) (56, 57). S. typhimurium grown in log phase for one generation in mild acid (pH 5.8) induces the expression of ATR pre-shock proteins involved in an inducible pH-homeostasis system that can function at external pH levels where housekeeping homeostasis systems fail (57). The enhanced pH homeostasis allows the synthesis of protective acid shock proteins at low pH levels (pH 3.0) that would compromise protein synthesis in nonadapted cells (54). The second stage, termed post-acid shock, occurs when S. typhimurium is shifted from mild base (pH 7.7) to acid conditions (pH 4.5 or below), altering the expression of >40 acid shock proteins, some of which play a protective role during a subsequent exposure to extreme acid conditions (53). Recently several pH-regulated genes have been identified (55, 59), but their specific role in the ATR response and virulence is not understood. The log-phase ATR response is controlled, at least in part, by fur (ferric uptake regulator), which encodes a regulatory protein involved in iron uptake and utilization (53, 54, 58).

Clearly, Salmonella strains are not always growing in log phase when they encounter low pH either in the environment or in the host. Two other ATR systems have been recently reported that are induced in stationary phase and are distinct from log-phase ATR (114). The first, termed stationary-phase ATR, is induced at low pH (pH 4.3). The second stationary-phase ATR system is not induced at low pH, is associated with the general stationary-phase stress response, and is under the control of rpoS (114). Although the three ATR systems described are generally distinct from each other, all three are required for full protection against extreme acid conditions (114). Characterization of the proteins that are unique and common to these ATR systems will provide a means to understand the molecular mechanism by which acid shock proteins confer protection. However, the pH-signaling and responding processes leading to altered levels of both gene expression and protein activity remain a mystery (reviewed in reference 144).

The S. typhimurium PhoP/Q regulatory system is essential for full virulence (49, 76, 130) and is modulated by multiple environmental factors such as pH, phosphate, carbon, nitrogen, and oxygen (9). PhoP/Q is required for (i) resistance to low pH (56), (ii) invasion of mammalian cells (9), (iii) inhibition of macrophage vacuole acidification (pH 5.0) (5) and spacious phagosome formation (4); (iv) resistance to antimicrobial peptides (49, 78); and (v) macrophage survival (49, 130). The phoP and phoQ genes from a standard two-component regulatory system in which PhoQ acts as the sensor and PhoP acts as a transcriptional regulator of two reciprocally expressed classes of genes, phoP-activated genes (pags) and phoP-repressed genes (prgs; 130). The reciprocal pattern of gene expression suggests that pags and prgs are required at distinctly different stages of the infection process. phoP mutants are attenuated (49, 130), as are mutants that constitutively express the pag genes (131), suggesting that either aberrant expression of pag genes and/or the inability to express prg genes affects the virulence of the bacterium. Mutational analysis of representatives from both pags (pagC; 130) and prgs (prgH; see oxygen section above and reference 9) has yielded mutants that confer a virulence defect in a murine typhoid model. pagC encodes an outer membrane protein that shows homology to a Yersinia enterocolitica invasion protein, ail (151). However, PagC may not function as an invasin, since a pagC mutant shows no defect when assayed for in vitro invasion (9), but it does confer a macrophage survival defect in vitro (130). Elevation of phagosomal pH with the addition of weak bases suppresses pagC expression (5). Thus, low intracellular pH appears to be a signal that induces the expression of pag genes, leading to the production of Pag proteins that may contribute to the delayed and attenuated acidification of the phagolysosome (5). Virulence mechanisms that modify the phagosome to resist acidification appear to be a common feature essential for the survival of many intracellular pathogens (reviewed in reference 166). Further investigation of the PhoP-regulated genes will facilitate our understanding of this important virulence mechanism.

SORTING OUT COMPLEX SIGNALS INVOLVED IN VIRULENCE GENE EXPRESSION

The regulation of virulence factors is often not limited to one easily definable environmental signal. Indeed, complex modes of regulation are probably going to be the norm, not the exception. Regulation in response to multiple signals was touched on in the examples above and is further exemplified by the following complex regulatory networks.

Spv

The non-typhoid Salmonella strains that cause invasive diseases in humans and animals usually have a large virulence plasmid (reviewed in references 79 and 80). The primary function of the plasmid is to facilitate the growth of the bacterium in the systemic sites of infection. Restriction analysis has shown that this function can be completely complemented by the conserved 8-kb spv locus. This locus encodes four genes, termed spvABCD, that are transcribed as an operon from two promoters upstream of spvA. However, multiple mRNA species are made, with the predominant species being spvA, followed by spvAB, spvABC, and spvABCD, each in decreasing abundance. The specific functions of each of the encoded proteins are not known. The transcripts are positively regulated by the spvR gene, located upstream of spvA and transcribed in the same orientation. SpvR apparently regulates itself and the spvABCD operon and activation is proportional to the level of SpvR; i.e., overproduction of SpvR induces both transcripts (79, 80).

Using lacZ transcriptional fusions to spvB, Fang et al. (46) showed that the operon in Salmonella dublin was significantly induced in stationary phase or in response to carbon starvation. This induction was strictly dependent on SpvR and rpoS (47). Spink et al. (168) and Abe et al. (1) have provided evidence that SpvR itself is induced in stationary phase. Induction was maximal under low-iron conditions. Transcription of spvR requires rpoS. Therefore, it is unclear whether the effect of rpoS on spvABCD transcription is direct or indirect through its effect on spvR. However, recent evidence shows that overproduction of SpvR can activate the spvABCD promoter in an rpoS ^– background (106).

Several other factors have been shown to affect spv expression. O’Byrne and Dorman (142) confirmed that transcription of an spvB-lacZ fusion was induced 15-fold as the cells entered stationary phase. They showed that mutations in either crp or cya resulted in an additional 10-fold induction. This increased expression was dependent on SpvR. However, expression of an spvR-lacZ fusion was not affected by an insertion mutation in crp. This implies that cAMP-CRP negatively regulates spvABCD transcription directly. However, growth of crp ⁺ cya ⁺ cells in glycerol versus glucose as a sole carbon source had no significant effect on the stationary-phase induction of the spvB-lacZ fusion. These workers have also implicated the DNA-binding protein H-NS in the regulation of the spv operon (141).

Using antibodies directed against SpvA, SpvB, or SpvC, it was shown that the production of these proteins is affected by a number of environmental factors (178). As expected, the proteins were induced in stationary phase or under glucose starvation, except when the cells were grown anaerobically. Low-iron conditions or lowered pH conditions (pH 5.0) caused induction of these proteins in log phase in the presence of glucose. These conditions did not affect stationary-phase induction in this study. Carbon starvation at 42°C caused the strongest induction of the Spv proteins. Thus a variety of environmental factors may influence the expression of the spv operon.

Fierer et al. (51) showed that the spv operon was induced intracellularly. Using a variety of cell lines, the authors showed that an spvB-lacZ fusion was induced 10- to 20-fold, with induction starting rapidly after invasion. Interestingly, alkalinization of J774 macrophages did not affect induction of spvB-lacZ. This is in contrast to intracellular induction of pagC (5) and suggests that the bacteria are independently responding to several intracellular signals to control various virulence genes.

RpoS

The alternative virulence factor, RpoS, regulates a variety of genes in response to starvation in both E. coli and Salmonella spp. (reviewed in references 84 and 161 and chapter 106). Several of these genes could be considered virulence factors in E. coli, including the HPII catalase, which confers resistance to H₂O₂, and csgA, the main subunit of surface fibers called curli that have fibronectin- and laminin-binding properties (143). Curli expression is induced in response to low temperature and low osmolarity as well as stationary phase (143). This is in contrast to expression of many of the fimbrae discussed above and suggests that curli may play a role outside of the host. In S. typhimurium, rpoS null strains are more sensitive to starvation, H₂O₂, acid stress, and DNA damage. These strains also show a 1,000-fold increase in oral 50% lethal dose compared to the rpoS ⁺ parent strain (47, 106). Interestingly, 5 days after oral infection, there was no significant difference in the number of bacteria in the spleens of mice infected with virulence plasmid-cured strains that were rpoS ⁺ versus rpoS ^–, suggesting that the major virulence defect in the rpoS strain is a result of the decreased expression of the spv operon (106). However, at 21 days postinfection, the number of viable bacteria in the spleen was approximately 2 logs higher for the plasmid-cured rpoS ⁺ strain (106), suggesting that RpoS is important for persistence.

CRP

The global regulator, CRP, regulates a variety of genes in E. coli and Salmonella in response to the levels of cAMP, which is synthesized by the product of the cya gene (chapters 84 and 85). Salmonella strains mutant in crp and cya are attenuated and protective (30). Indeed, these mutations have recently received considerable attention in the construction of live-vaccine candidates in a variety of Salmonella serotypes (e.g., 31, 83, 98, 160, 170, 174). CRP negatively regulates the spv operon on the Salmonella virulence plasmid (142; see above), while positively regulating various fimbrial operons (see above) as well as ompD (113). Thus, the virulence defect evidenced in crp cya mutants is likely a combination of effects that decrease, and perhaps increase, pathogenicity. Elucidation of the dominant virulence defects will require further study. In addition, it has been reported that a gene near crp also confers a virulence defect. This gene is removed in some of the crp deletion strains (98).

IVET

Environmental and regulatory studies (reviewed in references 127 and 128) that have contributed to the discovery of novel virulence factors include (i) environmental screens, based on coordinate environmental expression with known virulence genes (e.g., antigenic modulation of virulence factors of Bordetella pertussis [103, 107]); (ii) regulatory screens, based on the use of regulatory mutants that constitutively express a known virulence gene (e.g., isolation of V. cholerae tcp [tissue coregulated pili] fusions that are regulated by ToxR [147]); (iii) global screens, based on defining global regulatory systems that respond to a specific environmental challenge (e.g., the starvation stress response SSR of S. typhimurium [e.g., carbon, nitrogen or phosphate limitation; reviewed in references 122, 161, and 167]); and (iv) cultured cell global screens, based on defining alterations in global regulatory systems in response to growth in cultured cells and comparing them to those observed under defined extracellular stress conditions (e.g., macrophage versus nutrient limitation, oxidative stress, pH, and heat shock [2]).

The use of these four approaches has facilitated the identification of a number of virulence factors and has contributed immensely to our understanding of pathogenesis. These studies indicate that the expression of any given virulence gene may be modulated simultaneously by a number of environmental signals which may serve as an "environmental address" indicating precise locational (anatomical) information to the pathogen, i.e., what country (inside host versus outside environment), state (gut versus spleen), city (duodenum versus ileum), street (intracellular versus extracellular), and house (phagosome versus phagolysosome). To ensure reliable information, the regulatory circuits and host signature molecules that govern virulence gene expression in the milieu of host tissues are complex and overlapping, making interpretation of the actual in vivo situation, at times, problematic. To overcome these limitations, an in vivo selection system termed IVET was developed that does not depend on the anticipation or reproduction of environmental signals in the laboratory (120, 121; reviewed in references 118 and 162). IVET uses the animal as a selective medium for the identification of microbial virulence factors that are specifically expressed during infection. The original application of IVET relied on the fact that S. typhimurium purA mutations are highly attenuated in a murine typhoid fever model (126). Complementation of this nutritional deficiency in the animal provided a basis for in vivo-selected bacterial promoters that drive the expression of a promoterless purA gene. ivi (in vivo-induced) genes that have answered the selection included carAB (pyrimidine and arginine synthesis), the pheST himA operon (phe tRNA synthetase and integration host factor synthetase), and rfb (O-antigen synthesis). These results have yielded insight into the possible changes in metabolism, gene regulation, and cell-surface properties that may enhance bacterial growth and infectivity in host tissues; mutations in three of the ivi genes tested conferred defects in virulence, suggesting that they play a role in pathogenesis (120).

The applicability of this approach has been expanded to include other microbial pathogens with the advent of an IVET plasmid vector that was based on antibiotic resistance rather than alleviation of nutritional deficiencies (121). This vector does not require in vivo complementation of a well-characterized mutant, and it offers the opportunity to study cultured cell infection systems lacking a defined medium. One S. typhimurium ivi fusion recovered from antibiotic-based selection in BALB/c mice resides in fadB, a gene involved in fatty acid degradation. Induction of this operon in animal tissues may be in response to phagocytic cells that produce antimicrobial short-chain fatty acids (reviewed in reference 169); alternatively, the composition of host fatty acids may serve as signature molecules that communicate anatomical differences to the pathogen. The IVET approach may also be used to study temporal patterns of microbial virulence gene expression. IVET vectors have been constructed that fuse bacterial promoters to a site-specific recombinase which acts on a recombinational reporter system (23); a rise in the number of site-specific recombinants is an indicator of prior gene expression.

The flexibility of the IVET system offers many exciting research opportunities, since a small change in genetic selection often leads to the recovery of remarkably different classes of genes. Thus, the classes of genes that answer the IVET selection should depend on a wide variety of empirical conditions such as (i) the in vivo selectable marker (e.g., purA, thyA, cat), (ii) the reporter gene (e.g., transcriptional versus translational lacZ fusions), (iii) the infection model (e.g., mouse, ileal loops, cultured macrophages), (iv) the route of delivery (e.g., oral, intravenous, intraperitoneal), (v) the infected tissue (intestine, blood, spleen), and (vi) the indicator medium (MacConkey, Tetrazolium, or X-Gal [129]), all of which should influence the type of gene recovered (reviewed in reference 119). Indeed, such variability can also be problematic. However, when coupled with conventional environmental approaches, the IVET approach can address in vivo questions with a utility that appears to outweigh its inherent empirical and theoretical challenges.

CONCLUSIONS

Although disease is the hallmark of pathogenic bacteria, it is frequently not the normal outcome of most microbial-host interactions. Based on a variety of different indicators (e.g., seroconversion rates during epidemics of enteric disease), most encounters with pathogenic bacteria produce subclinical infections more often than overt disease. Indeed, much of the natural evolution of microbial-host interactions may be driven by microbial properties that optimize transmissions as well as growth yield of the microbe. Since death of the host usually is opposed to optimal transmission, one tends to find extreme levels of virulence displayed by unnatural or accidental pathogens rather than ones highly adapted to their particular host. Thus, ultimately, the optimal microbial-host interaction should result in decreased virulence (132). The regulation of virulence gene expression might be one way that an optimal microbial-host interaction is achieved.

We all have a natural tendency to be more interested in the unusually dramatic (disease) than the mundane (infection). Thus, the genetic and environmental factors that elicit the optimal expression of virulence factors have dominated our interest in understanding the regulation of virulence gene expression. However, it is worth considering that the reduction of virulence gene expression may be as important in the evolutionary success of pathogens as expression of virulence per se. As we learn more about the global regulatory responses that are tied to virulence gene expression, we may recognize that beneficial environmental signals are utilized by pathogenic microorganisms to down-regulate virulence factors as often as stressful signals are utilized to induce virulence gene expression. Thus, highly permissive growth conditions (e.g., the presence of glucose, or low bacterial cell density) might signal the microbe to avoid unnecessary virulence gene expression and in that way preserve the delicate balance that must characterize subclinical infections (127). When disease is the outcome, the shift in the balance may be a direct result of an aggressive response on the part of the host, which limits microbial growth through various inhibitory mechanisms. It is hoped that probing the host environment with tools that define the genes that bacteria express during infection will provide a better definition of the microbial-host interaction at both ends of this spectrum.

ACKNOWLEDGMENTS

We thank J. T. Tobias, P. C. Hanna, C. P. Conner, and D. M. Heithoff for critically reading the manuscript. This work was supported by NIH grant AI126289 (J.J.M.) and by NIH grants AI36373 and ACS JFRA 554 and a Beckman Young Investigator grant (M.J.M.)