Main

Bacterial pathogens modulate their gene expression profiles in response to cues perceived during infection of their hosts. Coping with a new environment requires both promoting the expression of some genes and silencing the expression of others that, although helpful in one particular locale, might be detrimental in a different one. Not surprisingly, the constitutive activation of bacterial regulatory systems can attenuate virulence owing to the aberrant expression of products that interfere with a successful infection1,2,3,4. Therefore, the ability of a pathogen to prosper in host tissues and cause disease depends not only on structural genes encoding gene products that mediate the acquisition of nutrients and resistance to noxious compounds, but also on genes encoding proteins involved in the signal-transduction pathways which ensure that structural genes are expressed when and where they are needed.

All signal-transduction cascades begin with signal detection by a sensor. The identification of the signal recognized by a sensor is crucial for understanding the role of a signal-transduction cascade in the lifestyle of a microorganism. Moreover, it can provide insights into the type of niche that a microorganism can explore or the molecular events governing a developmental pathway. Bacterial sensors involved in virulence have been shown to respond to chemicals, such as small peptides5, and physical parameters, such as temperature6. Although significant progress has been made in establishing the conditions that activate virulence genes in laboratory settings, the identity of the signals detected by bacterial sensors during infection remains largely unknown.

It is becoming increasingly clear that the major signal-transduction cascades governing virulence often control basic aspects of bacterial physiology, and are also found in related non-pathogenic bacterial species. For example, the regulatory protein CRP (cyclic AMP (cAMP) receptor protein) is essential for virulence7 and for the utilization of certain sugars as sole carbon sources8 in the enteric pathogen Salmonella enterica serovar Typhimurium ( Salmonella typhimurium ), but has been best characterized in the related commensal species Escherichia coli where it controls the ability to break down several carbohydrates, including the milk sugar lactose8. Likewise, the ToxR protein is required for the production of both cholera toxin and outer membrane porins in Vibrio cholerae 9, and has also been implicated in pressure-responsive gene expression in the deep-sea bacterium Photobacterium profundum strain SS9 (Ref. 10). Therefore, any attempt to define the signal(s) controlling such signal-transduction cascades must take into consideration both the virulence-related and non-virulence-related situations in which the signal-transduction cascade participates.

In this Opinion article, we discuss the biological significance of recent findings suggesting that the S. typhimurium PhoP/PhoQ two-component system can recognize antimicrobial peptides (AMPs), and that this recognition results in changes in the pathogen's gene expression programme.

Activation of PhoP/PhoQ by AMPs

It has recently been reported that certain cationic antimicrobial peptides — polymyxin B, C18G, LL-37 and protegrin — can activate the S. typhimurium PhoP/PhoQ two-component system, modifying the expression profile of genes controlled by the transcriptional regulator PhoP11,12. This activation seems to be direct because the polymyxin B nonapeptide (PMBN), a derivative of polymyxin B that is unable to cross the outer membrane13, bound to the purified sensing domain (that is, the extracytoplasmic domain) of the sensor PhoQ, but not to the sensing domain from the unrelated sensor CitA from Klebsiella pneumoniae 12. Moreover, an S. typhimurium strain that expressed a chimeric protein in which the sensing domain of its PhoQ protein was replaced by the equivalent region from the PhoQ protein of the environmental organism and opportunistic pathogen Pseudomonas aeruginosa failed to respond to the investigated antimicrobial peptide, C18G, but retained the ability to be shut off by high concentrations of Mg2+. Furthermore, C18G and LL-37 modified the biochemical activities of the PhoQ protein present in artificial vesicles only when added to the compartment harbouring the PhoQ sensing domain. It was suggested that cationic peptides bind to acidic residues in the sensing domain of the S. typhimurium PhoQ protein, displacing divalent cations such as Mg2+ and Ca2+ (Ref. 14), which have previously been shown to control PhoQ activity by binding to this domain15,16,17. The authors proposed that antimicrobial peptides and low pH are the major signals detected by the PhoQ protein when S. typhimurium is present in mammalian tissues.

The idea that a sensor protein from a bacterial pathogen would respond to a mammalian immune effector is very enticing. After all, there is the related precedent of the bacterial pathogen Agrobacterium tumefaciens responding to plant-derived flavonoids by expressing virulence genes that promote invasion of plant cells and the eventual development of a crown gall tumour18. In addition, it seems fitting that the PhoQ protein would recognize and respond to antimicrobial peptides given that: first, PhoP controls the expression of both gene products that modify lipopolysaccharide (LPS) in ways that bolsters the organism's resistance to antimicrobial peptides19 and gene products that cleave and inactivate antimicrobial peptides20; and second, the inability of an S. typhimurium phoP null mutant to survive within murine macrophages21 could be partially corrected by inactivation of the mouse gene encoding the cathelicidin-related antimicrobial peptide (CRAMP)22. In this context, the activation of the PhoP/PhoQ system by antimicrobial peptides is reminiscent of the activation of the mammalian Toll-like receptor 4 (TLR4) and Nod proteins by bacterial-derived LPS and peptidoglycan23, respectively, except that a bacterial sensor seems to be sensing a host molecule rather than the other way around. When physiological considerations are taken into account, however, in our view, the suggestion that antimicrobial peptides would be natural ligands of the sensor PhoQ is not fully supported, for reasons we outline below.

Failure to activate the whole PhoP regulon

The PhoP/PhoQ system governs the expression of >100 genes in S. typhimurium, which is accomplished either directly by the PhoP protein binding to the promoter of its target genes, or indirectly by other regulatory proteins that are activated by PhoP or its regulated gene products24,25 (Fig. 1). In contrast to low Mg2+, which is a condition that modulates expression of the whole PhoP regulon, polymyxin B activated only a small subset of the PhoP-regulated genes11. Curiously, polymyxin B did not induce some of the PhoP-activated genes encoding gene products that are required for proliferation inside macrophages and for virulence in mice.

Figure 1: Model illustrating the PhoP/PhoQ regulatory network of Salmonella enterica serovar Typhimurium (S. typhimurium).
figure 1

The intramacrophage expression of the spiC gene requires the PhoP protein to promote transcription of the response regulator SsrB, which controls expression of the spiC and spiR genes. PhoP is a transcriptional repressor of the hilA gene, which codes for the major regulator of S. typhimurium invasion, and is required for entry into non-phagocytic cells. The PhoP protein controls the levels of the alternative sigma factor RpoS post-transcriptionally by an as-yet-undescribed mechanism. The PhoP/PhoQ system directly regulates its own transcription, and that of the Mg2+ transporters encoded by mgtA and mgtCB. The PhoP protein is a transcriptional activator of the regulatory protein SlyA, and both proteins bind to the ugtL promoter to activate transcription in a feedforward loop design. The PmrA/PmrB system is activated post-translationally in low Mg2+ by the PhoP-dependent PmrD protein. The ugtL and pbgP genes mediate different modification of the lipid A moiety of the lipopolysaccharide. The katE gene encodes a cytoplasmic catalase. The regulatory proteins SsrB, HilA, RpoS, SlyA and PmrA, as well as PhoP, are required for S. typhimurium to cause a lethal infection in mice.

Full size image

The selective activation of particular PhoP-regulated genes by antimicrobial peptides is evocative of the expression of a limited number of PhoP-regulated genes that takes place when S. typhimurium is exposed to a mild acidic pH26. In fact, an acidic pH was once believed to be the signal sensed by the PhoP/PhoQ system27 before it was demonstrated that acidic pH activation could take place even in a strain lacking the sensor PhoQ28 or the regulator PhoP26. We now know that mild acidic pH is a condition that activates a different signal-transduction system (PmrA/PmrB)26, which can also be activated post-translationally by a PhoP-dependent protein in response to low Mg2+ sensed by the PhoQ protein29. This raises the possibility that antimicrobial peptides act on a regulatory system that promotes the expression of a subset of PhoP-regulated genes independently of the PhoP/PhoQ system. Indeed, the transcriptional activation of the PhoP-activated gene mig-14 that was promoted by polymyxin B was reduced but not eliminated in a phoP null mutant30. Furthermore, genes regulated by the RcsC/YojN/RcsB system were induced to higher levels by polymyxin B than the genes that are regulated by the PhoP protein11. Paradoxically, activation of the RcsC/YojN/RcsB system has been shown to downregulate the PhoP regulon31. In addition, antimicrobial peptides might activate a regulatory protein that functions together with PhoP. In fact, under particular growth conditions, transcription of the S. typhimurium ugd gene requires both the PhoP and RcsB proteins31,32.

Activation by different peptides

Antimicrobial peptides are widely distributed effectors of innate immunity in animals and plants, and have also been identified in several microbial species33,34. They are typically shorter than 40 amino acids and carry a net positive charge at neutral pH that mediates an electrostatic interaction with the negatively charged bacterial cell surfaces, and often leads to peptide insertion into the membrane and bacterial killing. Other antimicrobial peptides have been described that bear a negative charge at neutral pH or that seem to exert their bactericidal effects by affecting cytosolic targets35. Antimicrobial peptides do not seem to target chiral centres in membranes because the D-enantiomers of certain antimicrobial peptides exhibit identical bactericidal properties to those of the natural L-forms of the peptides36, and because they are typically active against highly divergent species30. Moreover, it is often the case that peptides that differ in their primary amino-acid sequences and secondary structures use the same mechanism to exert their bactericidal activities35. Therefore, it is not immediately obvious how the PhoQ protein can recognize the assortment of peptides derived from both bacteria and mammals that have been shown to promote the expression of particular PhoP-regulated genes.

The sensor PhoQ is located in the bacterial inner membrane. As the cut-off size for molecules that cross the outer membrane through the porins is <600 Da (Ref. 37) and the peptides proposed to activate the PhoP/PhoQ system are three times as large, it is most likely, in our view, that antimicrobial peptides (which typically have membrane-disrupting properties) will compromise the integrity of the S. typhimurium outer membrane before reaching the PhoQ protein. This raises the possibility that, even at the sub-lethal concentrations reported to activate the PhoP/PhoQ system11,12,30, antimicrobial peptides might be promoting non-specific alterations in the bacterial cell envelope that result in non-specific activation of signal-transduction systems. Consistent with this notion, polymyxin B also promoted the expression of S. typhimurium genes regulated by the RcsB protein and by the alternative sigma factor RpoS11. The effect on the levels of RpoS protein could be indirect because RpoS is affected upon activation of the RcsC/YojN/RcsB38,39 and PhoP/PhoQ systems (X. Tu et al., unpublished observations). Polymyxin E was also reported to activate the E. coli RcsC/YojN/RcsB two-component system40, which is known to respond to alterations in the bacterial membrane41. Therefore, antimicrobial peptides might activate other systems in addition to PhoP/PhoQ. Yet the activation of a regulatory system by an antimicrobial peptide does not necessarily mean that such a system has a role in resistance to antimicrobial peptides. For example, inactivation of the S. typhimurium rcsC gene reduced resistance to polymyxin B by ∼2 fold whereas there was >100-fold reduction in resistance when the phoP gene was mutated42. By contrast, growth in low Mg2+ specifically activates the PhoP/PhoQ system but not the RcsC/YojN/RcsB system31,32, in agreement with the requirement for the PhoP/PhoQ system, but not the RcsC/YojN/RcsB system, for growth in low Mg2+ (Ref. 28).

Presence in non-pathogenic species

The phoP and phoQ genes are present in several Gram-negative species43, including commensal organisms such as E. coli K-12 (Ref. 44) that are not normally pathogenic and would not be expected to routinely encounter antimicrobial peptides. The sensing domain of the E. coli PhoQ protein is highly conserved with the S. typhimurium PhoQ sensing domain45, and includes those residues necessary for a normal response to Mg2+ in S. typhimurium46 and E. coli47, and to antimicrobial peptides in S. typhimurium12. Low Mg2+ has been shown to promote PhoP-dependent gene transcription in several bacterial species, including E. coli48, Yersinia pestis 49,50, Shigella flexneri 51, Erwinia chrysanthemi 52 and Photorhabdus luminescens 53. This is in contrast to the idiosyncratic bacterial response to antimicrobial peptides. For example, the peptide CP11CN promoted gene transcription in P. aeruginosa, but this activation was independent of both the PhoP/PhoQ and PmrA/PmrB systems54. This work also showed that the ability of an antimicrobial peptide to induce gene transcription was unrelated to its structure or its mechanism of action.

As discussed above, the S. typhimurium strain harbouring the chimeric PhoQ protein with the sensing domain from the P. aeruginosa PhoQ protein did not respond to the peptide C18G, and this was taken as evidence that the P. aeruginosa PhoQ protein does not respond to antimicrobial peptides. However, this strain was not investigated for its ability to respond to other antimicrobial peptides such as polymyxin B. This is probably relevant because a P. aeruginosa phoP mutant displayed increased susceptibility to polymyxin B but not to C18G (Ref. 55), which is in contrast to the requirement for a functional PhoP/PhoQ system for resistance to both peptides in S. typhimurium12.

Low Mg 2+ as the physiological signal

That Mg2+ would serve as the ligand that specifically binds to the sensor PhoQ makes sense physiologically because the PhoP protein is a direct transcriptional activator of two of the three Mg2+ transporters present in S. typhimurium — termed MgtA and MgtB — promoting their expression when the bacterium experiences low Mg2+ environments56. In addition, activation of the PhoP/PhoQ system results in downregulation of the activity of the third Mg2+ transporter, CorA44. Moreover, mutants defective in the phoP or phoQ genes are unable to grow in low Mg2+ (Refs 28,48–52).

The negatively charged LPS is predicted to contain a large fraction of the total Mg2+ present in the bacterial cell36. This raises the possibility that some of the PhoP-regulated modifications of the negatively charged residues in the LPS might help S. typhimurium cope with Mg2+-limiting environments by making the Mg2+ present in the LPS available for other cellular activities57. For example, pathogenic and non-pathogenic organisms could face low-Mg2+ conditions in aquatic environments, such as a pond, and activation of the PhoP/PhoQ system might promote bacterial survival by enhancing cell envelope integrity and by maintaining the physiological Mg2+ concentrations that are necessary for ATP-mediated reactions and ribosome stability. Therefore, even if some of the PhoP-regulated LPS modifications do confer resistance to antimicrobial peptides in S. typhimurium, the alternative functions just described might account for the presence of the PhoP-regulated determinants mediating LPS modifications in non-pathogenic species that might not encounter antimicrobial peptides. And it might also explain why disruption of PhoP-regulated genes implicated in peptide resistance, such as pagP58 and ugtL59, does not affect the ability of S. typhimurium to cause a lethal infection in mice60.

PhoP/PhoQ as a master regulator

The PhoP/PhoQ system is a master regulator of S. typhimurium virulence functions. It controls pathogenicity determinants directly as well as indirectly, by regulating the expression and/or activity of several transcription factors required for S. typhimurium virulence, such as SsrB61, SlyA62,63, PmrA29, RpoS (X. Tu et al., unpublished observations) and HilA64 (Fig. 1). This allows the PhoP/PhoQ system to regulate several steps during infection of a mammalian host, including the invasion of epithelial cells in the small intestine and the survival in phagocytic cells of the liver and spleen. phoP null mutants are attenuated for virulence in mice following inoculation by the oral and intraperitoneal routes65, and strains with mutations in the phoP locus are defective for invasion of cultured epithelial cells66 and for survival within macrophages21.

The best-characterized PhoP-repressed gene is hilA in S. typhimurium pathogenicity island 1 (SPI-1), which encodes the main regulator of the type III secretion apparatus and effector proteins that mediate epithelial cell invasion. Several PhoP-activated genes contribute to the ability of S. typhimurium to survive or proliferate within phagocytic cells: mig-14, which uses an unknown mechanism to mediate resistance to CRAMP67; the indirectly regulated spiC gene in the SPI-2 pathogenicity island61, which alters normal cellular trafficking in the host cell68; and mgtC, a gene encoding an inner membrane protein, which enables bacterial growth in low Mg2+ conditions69.

Activation in mammalian tissues

Three different cues have now been proposed to activate the PhoP/PhoQ system inside macrophages: a mild acidic pH12,27, antimicrobial peptides12 and low Mg2+ (Ref. 70). This raises a number of questions. Does S. typhimurium respond to three different signals to promote transcription of PhoP-regulated genes required for virulence? Do these signals act concomitantly or in a time-dependent manner during phagosome maturation? What are the physiological concentrations of the proposed signals in the host tissues where the PhoP/PhoQ system is active? Which antimicrobial peptides are actually present in the different niches that S. typhimurium colonizes during infection of an animal host? Why do the antimicrobial peptides activate only a subset of the PhoP-regulated genes instead of the whole regulon?

The significance of an acidic pH and antimicrobial peptides as PhoQ signals is presently unclear, in our view, for two reasons: first, vacuolar acidification is not a prerequisite for S. typhimurium to survive within cultured macrophages71 but a functional PhoP/PhoQ system is21; and second, the PhoP/PhoQ system has been shown to affect trafficking in host cells72 and this might interfere with the delivery of granule-based antimicrobial peptides into the S. typhimurium-containing phagosome. Indeed, inactivation of the gene encoding CRAMP did not affect the growth of wild-type S. typhimurium inside murine macrophages nor did it restore wild-type growth to the phoP null mutant22. This suggests that the intramacrophage survival defect of the phoP null mutant might be caused by factors other than increased susceptibility to CRAMP. Therefore, the phoP null mutant might replicate less well in CRAMP-proficient macrophages than in CRAMP-deficient macrophages owing to its inability to prevent fusion of the S. typhimurium-containing vacuole with vesicles harbouring bactericidal compounds.

S. typhimurium is normally acquired following ingestion of contaminated water or food and must invade the epithelium of the small intestine to gain access to deeper tissues. If antimicrobial peptides were the physiological signal for the PhoP/PhoQ system, those peptides that are secreted into the lumen of the small intestine might activate the PhoP/PhoQ system resulting in transcriptional repression of the invasion determinants, the expression of which is a prerequisite for S. typhimurium to enter non-phagocytic cells73 and to cause a lethal infection following oral inoculation of mice74. A potential answer to this conundrum resides in the reported ability of S. typhimurium to inhibit the expression of antimicrobial peptides in the murine small intestine75. However, this inhibition would have to occur fast enough for S. typhimurium to change its gene expression programme and produce the PhoP-repressed invasion determinants. Alternatively, S. typhimurium might still gain access to deeper tissues through CD18-positive cells76 in a process requiring neither the invasion determinants nor entry into the cells of the intestinal epithelium.

The data discussed above leave low Mg2+ as a strong candidate for being the PhoP/PhoQ-inducing signal inside host cells. However, a recent study did not find changes in the phagosomal Mg2+ concentration in the first 30 min following S. typhimurium internalization and in the levels of phoP transcription in the first two hours after an ionophore was added to the host cells77. The significance of these results is presently not clear given the fact that transcription of PhoP-activated genes was still increasing at the time the measurements were carried out77, and that PhoP-promoted transcription reaches a maximum only five to six hours after internalization78. Moreover, a phoP null mutant exhibits wild-type survival within macrophages61 and activation of genes within the SPI-2 pathogenicity island two hours post-internalization79, but is defective when these phenotypes are evaluated at five to six hours after internalization61. Therefore, the reported measurements of phagosomal Mg2+ concentrations would be relevant only if one postulates that the PhoP-mediated expression taking place within the first two hours that S. typhimurium is inside a macrophage determines bacterial survival after five hours. That low Mg2+ could be the activating signal inside macrophages is supported by the following findings: low Mg2+ controls all known PhoP-regulated genes under laboratory conditions56; the PhoP-activated mgtC gene is necessary for intramacrophage survival and for growth in low Mg2+ in defined media69; and addition of Mg2+ to macrophages could partially rescue the growth defect of the mgtC mutant but not of other mutants unable to survive within macrophages69. It is noteworthy that Mycobacterium tuberculosis and Brucella suis , which are facultative intracellular pathogens that reside within a macrophage phagosome like S. typhimurium, harbour mgtC homologues. Inactivation of the mgtC gene in these pathogens recapitulates the phenotypes exhibited by the S. typhimurium mgtC mutant, including the inability to grow in defined media of low Mg2+ concentration and to proliferate within macrophages80,81.

Final remarks

The proposed ability of bacterial sensors to detect host immunomodulatory molecules is not limited to pathogenic S. typhimuium as it has been suggested that P. aeruginosa can sense the mammalian cytokine interferon-γ through an outer membrane porin82. However, the porin amino-acid sequence is highly conserved in non-pathogenic species, such as the soil bacterium Azotobacter vinelnadii 83, that are not in contact with mammalian hosts. As P. aeruginosa is normally found in water and soil (rather than as a colonizer of healthy humans), it is presently unclear how its porin evolved as a cytokine sensor. Because S. typhimurium proliferates in both animal hosts and non-host environments84, one could imagine a scenario in which the PhoQ protein incorporated the ability to detect antimicrobial peptides into a Mg2+-sensing ancestral protein. This might explain the in vitro experiments showing that particular antimicrobial peptides can bind to the sensing domain of the PhoQ protein with high affinity12 or modify the ability of the whole PhoQ protein reconstituted in artificial vesicles to alter the phosphorylation state of its partner PhoP12. However, antimicrobial peptides selectively activate only a subset of PhoP-regulated genes, and are likely to disrupt the outer membrane to reach the PhoQ protein in the inner membrane. Further experiments are required to evaluate the physiological significance of the proposal that bacterial sensor proteins can recognize host defence molecules.