|Home | About | Journals | Submit | Contact Us | Français|
A variety of mechanisms are used to signal extracytoplasmic conditions to the cytoplasm. These mechanisms activate extracytoplasmic function (ECF) sigma factors which recruit RNA-polymerase to specific genes in order to express appropriate proteins in response to the changing environment. The two best understood ECF signaling pathways regulate σE-mediated expression of periplasmic stress response genes in Escherichia coli and FecI-mediated expression of iron-citrate transport genes in E. coli. Homologues from other Gram-negative bacteria suggest that these two signaling mechanisms and variations on these mechanisms may be the general schemes by which ECF sigma factors are regulated in Gram-negative bacteria.
Bacteria live in an ever-changing environment and must alter proteins on their cell surface to adapt to these changes and survive. Environmental changes include the following: osmolarity , membrane composition , light , barometric pressure , temperature [4, 1], and the concentration of some nutrients, specifically chelated iron (III) . Bacteria have developed sets of specific response genes that are regulated by a subset of the σ70-like sigma factors in order to respond to a changing environment. Given a specific external stimulus, these sigma factors recruit RNA polymerase to the appropriate response genes. As these sigma factors are involved in regulating the expression of proteins residing in the outer membrane or periplasmic space, they are given the moniker extracytoplasmic function (ECF) sigma factors .
Because the exterior of invasive bacteria comes into contact with the host and must respond to the host environment, ECF sigma factors also regulate genes associated with bacterial virulence. Lethality of Mycobacterium tuberculosis in mice is dependent on the ECF sigma factor σC . The activity of E. coli σE is involved in uropathogenesis . The virulence of Salmonella typhimurium is also dependent on σE . Notably, the ECF σE homologue AlgU regulates the expression of the mucoid envelope that protects Pseudomonas aeruginosa from antibiotics, oxidative stress, and immune attack . This mucoid form of the bacteria is also associated with morbid and mortal infectivity in the lungs of cystic fibrosis patients .
Some bacterial pathogens are known to express virulence genes in low iron environments [12-14]. The virulence of some bacteria in animal models is associated with iron-binding siderophores that facilitate iron uptake . The iron-siderophore pyoverdine and signaling through its associated TonB dependent ferro-pyoverdine receptor are associated with virulence in Pseudomonas aeruginosa [16, 17]. These genes regulated by ECF sigma factor PvdS [16, 17].
In order to communicate the conditions of the periplasm or the outside environment to the genetic machinery of the cell, a signal must pass through one or two bacterial membranes to the cytoplasm. Thus, signaling to the ECF sigma factors must involve integral membrane proteins to achieve this communication. In many bacterial signaling pathways, transmembrane signaling may be achieved with a two-component signaling mechanism, in which a membrane-bound receptor phosphorylates a cytoplasmic effector. Two-component signaling is the subject of another review in this volume. Sequence homology indicates that two-component signaling activation of an ECF sigma factor may regulate the activity of σE from the Gram-positive bacterium Streptomyces coelicolor A3(2) . However, to date, this has not been seen in Gram-negative bacteria and thus two-component signaling in ECF activation may be limited to Gram-positive bacteria. In the basal state, all known ECF sigma factors in Gram-negative bacteria interact with an anti-sigma factor that inhibits the associated ECF sigma factor from interacting with RNA polymerase . The ECF signaling mechanisms presented here illustrate different methods of signaling and different roles for the anti-sigma factor in regulation of sigma factor activity.
Sequence similarity and functional studies suggest that most ECF sigma factors are likely to be regulated by a mechanism similar to either the activation mechanism of the periplasmic stress response sigma factor, σE, or the activation mechanism of the iron-citrate response sigma factor, σFecI (FecI), both from E. coli. Inhibition of σE is relieved by signals from the periplasm that stimulate the complete degradation of its anti-sigma factor. In FecI activation, extracellular signals travel though a protein network across two membranes and the periplasmic space in order alter the usual anti-sigma factor role; the anti-sigma factor of FecI also plays an important role in activating FecI when stimulated. In this review, we discuss the two major groups of signaling mechanisms represented by the σE and the FecI systems.
σE is an essential gene that controls the periplasmic heat shock regulon that is activated by unfolded protein in the periplasm [20, 5]. The accumulation of unfolded proteins can be caused by a number of factors: heat, overproduction of outer membrane protein C (OmpC) or other outer membrane proteins (OMPs) [21, 22], altered lipopolysaccharide in the outer membrane , or mutations in folding enzymes such as the Dsb proteins and prolyl isomerases [2, 24]. About 20 promoters are transcribed in a σE-dependent manner . The expressed proteins include OMP-folding chaperones, lipopolysaccharide biogenesis enzymes, and periplasmic proteases . The genes regulated by σE also include the cytoplasmic heat shock protein σ32, the house keeping sigma factor σ70, as well as σE itself with all of its regulatory components (RseA, RseB, RseC and RseP) .
Recent discoveries indicate that the σE regulon also contains small RNA (sRNA) transcripts that lower the expression of a number of outer membrane proteins in an Hfq-dependent manner. Hfq functions as a molecular match-maker in facilitation of sRNA-mRNA interactions [26-28]. σE-regulated expression of the sRNA RybB results in decreased levels of mRNAs encoding OmpC, OmpA, and OmpW [27, 29]. Similarly, σE-dependent up regulation of MicA negatively affects the expression of OmpA [27, 29]. Hfq regulates σE itself as well as a number of other genes in the σ32 and σS stress response pathways, which suggest that Hfq may play a pivotal role in regulating all of these different stress response pathways. .
The general scheme for regulation of σE is through stimulated and regulated intramembrane proteolysis (RIP) of its anti-sigma factor RseA (Figure 1). The use of RIP is a common mechanism for communication found in both eukaryotes and prokaryotes (for a review, see ). This mechanism can be used to signal across membranes in either direction. Generally the RIP protease activity is regulated through a signal-dependent cleavage of the substrate protein on one side of the membrane by a protease at a specific site (Site 1). After this cleavage, the RIP protease is free to cleave the substrate protein at a specific site within the membrane (Site 2), releasing a protein fragment on the other side of the membrane. This protein itself often serves as a signal in eukaryotic systems. In the case of σE, the substrate of the RIP protease is ultimately completely eliminated in order to free σE.
RseA spans the inner membrane and includes both a cytoplasmic σE-inhibitory domain and a periplasmic regulatory domain. Destruction of RseA and consequent relief of σE inhibition is regulated by both DegS and RseP, both of which are inner membrane proteins and proteases. The protease activity of DegS is activated on recognition of the C-termini of unfolded OMPs, and upon activation it cleaves off a periplasmic portion of RseA [22, 31, 32]. As a consequence, inhibition of the RIP protease RseP (or YaeL) is relieved and it cleaves RseA in the inner membrane proximal to the cytoplasmic side [33-35]. The liberated RseA continues to inhibit σE until ClpXP and other ATP driven cytoplasmic proteases separate the RseA/σE complex and degrade RseA [20, 36, 37].
Like all ECF (also called Group IV) sigma factors, σE is significantly different from the prototypical, Group I σ70 proteins . The ECF sigma factors are evolutionarily the most divergent of the σ70 proteins . Of the four domains or regions observed in Group I sigma factors, the N-terminal subdomain (σ1.1) is missing and much of the third region (σ3) is missing in ECF sigma factors (Figure 2) [6, 38, 39]. Furthermore, in ECF sigma factors, regions 1 and 3 (σ1 and σ3) have lower conservation and are acidic, whereas regions 2 and 4 (σ2 and σ4) are highly conserved and are more basic . Generally σ2 and σ4 recognize the -10 and -35 regions of the promoter, respectively . The σ1 region prevents sigma factors from binding to the promoter without RNA polymerase, and the truncation in ECF sigma factors may allow promoter binding without RNA polymerase [41, 42, 6]. FliA, an ECF sigma factor that lacks the 1.1 region, can bind its promoter independently of the presence of RNA polymerase . However, RNA polymerase is still required for DNA binding by the ECF sigma factor FecI . Region σ3 allows sigma factors to recognize promoters with extended -10 elements, thus ECF sigma factors should be deficient in this ability . Despite the overall differences between Group I sigma factors and the ECF sigma factors, an X-ray crystal structure of σE indicates that σ2 and σ4 are essentially identical to those regions in the Group I sigma factors [45, 44, 46, 47, 39]. The two-domain sigma factor likely represents the minimal structure required for σ factor activity .
The X-ray crystal structure of the complex of a cytosolic fragment of RseA and σE reveals the mechanism of strong inhibition of σE by RseA. The structure shows that the anti-sigma factor is sandwiched between σ2 and σ4 of σE (Figure 3) . In total, the complex buries 3805 Å2 of surface with hydrophobic interactions dominating . Although interactions between σ2 and σ4 dominate, the spacing provided by the flexible linker region is important for binding . Experimentally induced proteolysis of the RseA/σE complex illustrated the overall protection from proteolysis that is afforded by RseA and generated σ2 and σ4 as individual cleavage products, which can bind to RseA individually . This dominance of σ2 and σ4 interactions with RseA is consistent with the effects of point mutations that interfere with RseA-σE complex formation .
RseA inhibits σE activity by interfering with the formation of the σE-RNA polymerase complex. σE can not compete with σ70 for RNA polymerase in the presence of RseA . In order to identify exactly how RseA inhibits the binding of σE to RNA polymerase, Campbell et al used the similarity of regions σ2 and σ4 to σA and σE, to model a σE-RNA polymerase holoenzyme based on the σA-RNA polymerase holoenzyme structure. With this model, the authors predicted steric clashes between elements of RseA and the primary binding determinants of RNA polymerase (Figure 3). The σ2:β′ coiled-coil interaction between σE and RNA polymerase is incompatible with RseA binding because α-helix 4 of RseA would clash with the β′ coiled-coil structure . Similarly, α-helix 3 of RseA sterically clashes with the β flap-tip helix that is involved in interaction of σ4 with σE .
In order to liberate σE from RseA, a single cut by DegS initiates a series of proteolytic modifications of RseA that ends in its complete degradation. This initial proteolytic step is the critical and only known positively regulated proteolytic event in the regulation of σE. In wild-type E. coli, DegS is required for σE activity and is essential to the cell . Without DegS, σE activity can no longer be induced in response to extracytoplasmic stress . DegS recognizes the three terminal amino acids in the denatured OMPs, which is a good indicator of the folded state of OMPs, as the C-terminal residues are normally sequestered in the lipid bilayer [48, 32, 37]. Although DegS prefers YYF or YYM, DegS also recognizes YXF which is found in OmpC and in other known and putative OMPs . Recognition of peptides with these sequences activates DegS to become proteolytically active and cleave the periplasmic domain of RseA between residues 148 and 149 (Figure 1, Site 1) . These residues in RseA are separated from the periplasmic side of the inner membrane by about 30 residues, which would place the cleavage site far enough from the membrane to be cleaved by DegS .
DegS belongs to the HtrA family of oligomeric serine proteases (for reviews, see [49, 50]). One aspect of the HtrA family members is that at least one PDZ domain accompanies each of their defined catalytic domains. PDZ domains are involved in the mediation of specific protein-protein interactions in both eukaryotes and prokaryotes (for a review, see ). In most cases, PDZ domains recognize the very C-terminal end of the partner protein. Generally, only 6-10 amino acids are necessary and sufficient for binding affinity equal to that of the full length protein .
The crystal structures of the soluble portion of DegS with and without a peptide that consists of the C-terminal residues of OmpC illustrate how the PDZ domain of DegS functions to specifically activate DegS protease activity . The PDZ domain links the recognition of specific unfolded proteins to allosteric activation of the proteolytic activity towards RseA. Like other HtrA family proteins, DegS is a single pass inner membrane protein that extends into the periplasm of the cell where the soluble portions of the protein interact to form a trimer [49, 50]. The periplasmic portion of DegS forms the shape of a pinwheel consisting of a core of the three proteolytic domains, each accompanied by a PDZ domain (Figure 4) . Similar to other known crystal structures of PDZ domains with peptide ligands, the last four residues of the peptide bind by β-augmentation [53, 54, 52]. These residues form an additional β-strand at the end of the β-sheet in the PDZ domain [53, 54, 52].
A comparison of the peptide-bound and peptide-free structures of DegS reveals a series of conformational changes in loops L1, L2, L3, and LD (Figure 5). Loop 2 is important for substrate recognition, and L1 and LD are loops important to the DegS protease activity . The allosterically driven changes in loop L2 remove residue L218 from the S1 substrate recognition site thus allowing it to recognize RseA (Figure 6) . The LD and L1 loop movements are critical to the backbone conformational change observed in residue H198 (Figure 6). This residue reorients the main chain carbonyl and its associated partial negative charge from a position pointing into the positively charged “anion hole” of the active site to a position pointing away from the active site. The conformational change of this carbonyl removes inhibition on the protease activity of DegS . Furthermore, the geometry between H96 and S201 of the catalytic triad (S201, D126, and H96) improves on peptide binding. Thus, the PDZ domain-mediated binding of the C-terminal tails of unfolded outer membrane proteins stimulates the proteolytic activity of DegS.
On DegS cleavage of the periplasmic domain from RseA, RseP can access the second site of proteolysis (Figure 1, Site 2), which lies in the inner membrane spanning portion of RseA [33, 34]. RseP is an intramembrane protease that belongs to the family of RIP proteases founded by the S2P protease gamma-secretase [55, 56]. These proteases have multiple membrane-spanning regions and include the typical metalloprotease motif HEXXH and a C-terminal LDG motif (Figure 7) [30, 57, 58]. RseP appears to consist of four transmembrane domains with a PDZ protein interaction domain in a periplasmic loop connecting the second and third transmembrane helices . Mutations of the HEXXH and LDG regions of RseP eliminate protease activity . The labeling and functional assays indicate that the RseP protease active site appears to be in a highly structured part of the protein at the interface of the cytosolic side of the inner membrane .
In order for DegS proteolysis to be the rate limiting step in the proteolytic cascade of RseA degradation, substrate accessibility to the active site of RseP must be highly regulated. After the initial cut by DegS, the subsequent proteolysis is so rapid that the DegS cleavage product of RseA is not normally detected [33, 34, 37]. Before cleavage by DegS occurs, the periplasmic domain of RseA inhibits RseP from proteolysis of RseA. DegS proteolysis removes the inhibition of RseP proteolysis of RseA [33, 34].
Direct or indirect interactions between the PDZ domain of RseP and the glutamine rich domain of RseA inhibit the activity of RseP . RseA has two periplasmic glutamine rich domains (residues 162 – 169 and 190 – 200) that are both necessary to confer wild-type resistance of RseA to RseP proteolysis . The PDZ domain of RseP also confers inhibition of the proteolysis of RseP [35, 59]. As PDZ domains can recognize internal sequences as well as terminal sequences, the glutamine rich domains of RseA may be recognized by the PDZ domain of RseP in order to create an inhibitory protein-protein interaction. The mode of inhibition through this interaction is not known but one may speculate that this interaction could position the RseA and RseP in an orientation that prevents substrate accessibility. This protein-protein interaction would keep the protease proximal for quick activity once the first protease site in RseA is cleaved. This interaction may also prevent more indiscriminant activity of RseP, as it has an unexpectedly wide substrate specificity .
Grigorova et al illustrate how negative regulation of RseP enforces the sequential degradation of RseA and enforces the two design features apparent in the σE activation pathway . σE activity is sensitive to a wide range of OMP signals , yet insensitive to variable concentrations of DegS and RseP . RseB binds to RseA as an allosteric inhibition of cleavage by RseP . Without RseB, RseP can cleave full length RseA and allow activation of σE in the absence of DegS . Furthermore, without RseB this cleavage is now dependent on the concentration of RseP . However, in wild-type E. coli, the response to unfolded OMPs is completely dependent on DegS. Neither the affinity of RseB for RseA, nor the activity of RseP is modulated by unfolded OMPs . However, there is the possibility that RseB may be regulated by some as yet unknown signal. Further negative regulation of RseP was discovered by studies with inactivated DegS, which indicated that DegS itself also inhibits DegS independent proteolysis by RseP . These regulatory interactions remove noise from the σE activation system such that small changes in OMPs are detectable, and accidental variations in concentrations of the components of the system do not affect the signaling [21, 33, 60].
RseA and σE interact with picomolar affinity, which ensures that degradation of RseA associated with liberation of σE is almost entirely dependent on ATP-driven proteases . These energized proteases have the ability to unfold proteins, which enables them to disassemble the RseA/σE complex and degrade RseA. RseA is the most quickly degraded substrate of its protease, ClpXP, thus substrate competition does not greatly affect the rate of RseA degradation by ClpXP [36, 37]. Although ClpXP is the most active protease in RseA proteolysis, most of the ATP-dependent proteases can degrade RseA in vivo . Thus, in the unlikely event that ClpXP is competed away by alternative substrates, other ATP-dependent proteases will also degrade RseA . Given that RseA is the substrate of so many proteases, it is unlikely that any type of regulation of this proteolysis is possible. With all of the regulatory elements in place, the only known input that activates the series of proteolytic events is the recognition of OMP C-termini by DegS.
Characterized σE homologues in Gram-negative bacteria and some homologues from Gram-positive bacteria appear to be regulated by mechanisms similar to that of σE . There is almost always expression of an anti-sigma factor that inhibits the activity of the sigma factor. Activity of the sigma factor is often associated with the degradation of this protein. However, not all of the components of the regulatory systems appear to be homologous and the biochemical characterization of these systems is still incomplete.
AlgU is an ECF sigma factor homologous to σE that regulates the expression of the genes essential to the protective mucoid phenotype in P. aeruginosa. The molecular signals that activate AlgU are still unknown, although there are at least two methods by which AlgU activity is induced. Under some conditions, the AlgU pathway responds to periplasmic stress signals similar to that of E. coli σE . However, in the lungs of cystic fibrosis patients and similar laboratory conditions, the mucoid expressing P. aeruginosa often inactivate the anti-sigma factor MucA through C-terminal mutations [63, 64].
Although these data indicate that there are multiple mechanisms for AlgU activation, the constituents and at least some parts of AlgU regulation are analogous to σE regulation. The algU mucA mucB mucC gene cluster is homologous to the E. coli rpoE(σE) rseA rseB rseC gene cluster . Similar to RseA, MucA appears to act as an anti-sigma factor [65, 66]. Conversion to the mucoid state is associated with a redistribution of AlgU from the inner membrane to the cytosol, which suggests proteolysis of the anti-sigma factor similar to proteolysis of RseA . Truncations or alterations of the periplasmic C-terminus of MucA that dramatically lower the stability of MucA were found in almost all of the mutants of mucoid isolates of P. aeruginosa in the lungs of cystic fibrosis patients [68, 69, 67]. The RseB homologue, MucB, has a similar but more pronounced negative regulatory role than RseB, as mutations in MucB lead to a stronger depression of AlgU activity [65, 66]. RseP homologues AlgQ and MucD negatively repress AlgU activity rather than play a role in activating it as RseP does in σE activation . Although a RIP protease mechanism similar to σE activation may initiate the degradation of the anti-sigma factor MucA, no known DegS or RseP homologues appear to play theses roles in the activation of AlgU.
The exact mechanism of AlgU activity in wild-type cells remains elusive, but the activity of the mucA mutants is partially understood. The instability of the MucA mutants is associated with the activity of the PDZ domain containing periplasmic protease Prc . This protease has no affect on wild-type MucA, but it will degrade the mutant MucAs . One thought is that the mutations in MucA leave unfolded C-terminal tails that can be recognized by the PDZ domain of Prc . It is attractive to think that if there is a RIP protease mechanism for activation of AlgU in wild-type cells, removal of the periplasmic domain by Prc would allow the site 2/RIP protease to release the cytosolic portion of MucA into the cytosol for degradation.
The sigma factor CarQ of the Gram-negative bacterium Myxococcus xanthus also appears to be regulated in a similar method to σE. However, with the exception of CarQ itself none of the regulatory elements in this system are homologous to members of the σE system. CarQ is involved in a response to very specific envelope stress in the form of light-induced oxidative damage. This ECF sigma factor regulates the expression of carotenoids which protect the cell from photo-damage [72-74]. CarQ is cotranscribed with its anti-sigma factor, CarR, which has no similarity to RseA [72, 74, 3, 75]. The inhibitory activity of CarR in turn is regulated by CarF, which appears to directly or indirectly mediate the light-induced deactivation of the CarR anti-sigma factor activity . Without homology to any known protease, CarF may function as an anti-anti-sigma factor, as observed with the partner switching mechanism of SpoIIAB and SpoIIAA regulation of σF and the RsbW and RsbV proteins of σB in B. subtilis [77, 78, 76]. In these cases, the anti-anti sigma factor interacts with the anti-sigma factor and displaces the sigma factor. However, there is evidence that the anti-sigma factor CarR is degraded when enough light is present for CarQ activity . This would suggest a mechanism similar to that of σE .
Although Gram-positive bacteria have no periplasm with which to detect unfolded proteins, there appears to be some cross over between the σE type of signaling between Gram-positive and Gram-negative bacteria. Recently an envelope stress response regulator in the Gram-positive B. subtilis, σW, was found to be regulated in a very similar way to σE from E. coli. It has an anti-sigma factor, RsiW, which is degraded by the RIP protease YluC (also called RasP). YluC can cut RsiW only after YpdC (also called PrsW) cleaves the periplasmic domain of RsiW [79, 80]. Like RseA, the final cytoplasmic cleavage product of RsiW is also a substrate for the ClpXP protease. However, the relationship between RsiW and ClpXP appears to be more exclusive than with RseA, as the proteolysis of RsiW by ClpXP is titratable by competing substrates . Another difference between the σW and σE systems is the lack of homology between the primary cleavage enzymes, YpdC and DegS. YpdC appears to contain multiple transmembrane helices compared to the single transmembrane helix of DegS. In fact, there are no homologues of YpdC in E. coli and the three DegS homologues in B. subtilis do not appear to be involved in this signaling mechanism .
Although there are significant differences between the σE and σW signaling systems, there is an interesting similarity in regulation. Similar to the DegS dependent inhibition of RseP proteolysis in the absence of signal, YluC appears to inhibit YpdC in the absence of signal in σW regulation . However, in this case the RIP protease inhibits the protease that makes the initial cut rather than the opposite, which is found in the regulation of σE activation in E. coli. Beyond this difference it is remarkable that not only is the mechanism of signaling similar for σE and σW, but there seems to be similar regulation of the proteolysis that propagates the signal in this pathway. With the similarities of these two ECF signaling systems and the conservation of the RIP protease signaling mechanism from prokaryotes to eukaryotes, the RIP protease mechanism likely will be found in regulation of many other ECF sigma factors in both Gram-positive and Gram-negative bacteria.
For comparison of signaling mechanisms, it is worth briefly mentioning the Cpx signaling pathway, a primarily parallel but somewhat overlapping periplasmic stress response pathway to the σE system (for a review, see ). Not only are both of these systems activated by unfolded proteins in the periplasm, these systems also regulate some genes with similar activities. However, from stimulus to gene activation, these systems have completely different mechanisms. In Cpx signaling, the unfolded proteins are recognized by CpxP, a periplasmically located inhibitor of the signaling protein CpxA. On ligand recognition, CpxP dissociates from CpxA, allowing CpxA to autophosphorylate itself and to phosphorylate the cytoplasmic effector CpxR. CpxR is not a sigma-factor. CpxR-P is a transcriptional activator that binds to a specific site on DNA to recruit RNA polymerase through interaction with the C-terminal domain of its alpha subunit. Thus, the methods of gene regulation are completely different for these two very similar regulons that are activated by similar stimuli.
A second type of ECF signaling, distinct from mechanisms discussed for σE, is illustrated by the activation of FecI. For Gram-negative bacteria, iron uptake is a serious problem. With an additional membrane they have limited accessibility to an already scarce resource, as iron is insoluble in the aerobic environment [82-84]. Thus, these organisms have developed complicated and energetically expensive systems for detecting and acquiring precious metal ions. Most often siderophores are used to complex the iron and this complex is transported first into the periplasm where it may be bound by a secondary iron binding protein. Then the iron (or ferric siderophore) is transported across the inner membrane into the cytoplasm.
Although the members of the iron-citrate uptake system are not essential proteins, they must give some bacteria an advantage because iron-citrate transport is a costly process. Not only is energy involved in the synthesis of these proteins, but the transport of the iron across each membrane is energized. The outer membrane iron transport proteins associate with TonB, which uses the proton gradient across the inner membrane of E. coli to power the outer membrane transporter [85, 86]. Iron transport across the inner membrane also requires energy in the form of ATP hydrolysis. Once the cell has acquired enough iron, the production of the iron transport machinery is repressed by the Fur repressor . In iron rich conditions, Fur forms a complex with iron and becomes an active repressor of the expression of iron transport genes .
Regulation of the iron-citrate transport genes is remarkable, because FecA not only transports iron, it signals the extracellular presence of iron-citrate to the genetic machinery in the cytoplasm (Figure 8) . When the iron-citrate complex binds to FecA on the outside of the cell, a signal is transferred through FecA to its periplasmic N-terminal domain [89-93]. This signal passes from the N-terminus of FecA to the C-terminal end of the inner-membrane protein FecR [91-93]. FecR then transmits the information by some unknown mechanism through the inner membrane to its N-terminus which consequently affects FecI. FecI is activated to recruit RNA polymerase and bind to the fec operon, thus increasing the expression of the iron-citrate uptake genes [94, 95]. The expressed proteins include FecB , the periplasmic binding protein, and the FecC, FecD and FecE proteins, which form an ATP-dependent ABC transporter complex in the inner membrane [95, 96]. Sequence analysis indicates that FecA is part of a subclass of TonB dependent iron transporters that have N-terminal extensions beyond the TonB interaction site [97, 98]. It is believed that the extensions of these other TonB dependent proteins are likely to interact with downstream signaling components to initiate ECF sigma factor signaling mechanisms that are similar to FecARI.
Apo- and iron-citrate bound structures of FecA by Ferguson et al and Yue et al delineate the structural changes that are likely to contribute to the signaling of the presence of iron-citrate (Figure 9) [90, 91]. Like all known TonB dependent iron-siderophore transporters, FecA consists of a large 22 stranded anti-parallel β-barrel with a plug domain N-terminal to the barrel that folds inside the barrel thus blocking small molecule transfer between the periplasm and the extracellular space [90, 91]. On the cell surface, large extracellular loops that are required for both signaling and transport clamp down over the iron-citrate as it binds to FecA [90, 91, 99]. This binding event creates structural changes in FecA that extend from the extracellular space to the periplasm [91, 93]. Conformational changes in extracellular loops 7 and 8 also induce changes in a small loop in the plug domain (NL4), which is found in an altered conformation on iron-citrate binding (Figure 9) . This loop contains the RGX5YGX4GX2N motif found in TonB dependent transporters, suggesting its importance in signaling the presence of substrate to TonB. Another significant conformational change occurs upon iron-citrate binding: residues constituting the TonB box of the transporter, involved in a subsequent protein-protein interaction with TonB protein, become highly mobile (and therefore disordered in the crystal structures) upon binding iron-citrate. Similar changes in the TonB box region have been associated with other TonB dependent transporters on ligand binding [100-104]. EPR measurements on ligand bound BtuB indicate that the TonB box moves away from the plug domain and extends into the periplasm . Disorder and lengthening of this region may be required for recognition of the substrate bound state of the transporter by TonB [106-108].
Many outer membrane transporters require TonB in order to transport substrates. TonB is thought to provide mechanical energy to this set of TonB dependent transporters, which includes FecA. With some sequence homology to the flagellar motor proteins (MotA and MotB), the inner membrane bound TonB-ExbB-ExbD complex harnesses energy from the proton gradient across the bacterial inner membrane in order to create mechanical energy . TonB protein reaches across the periplasm to recognize iron-citrate bound FecA at a specific amino acid sequence known as the TonB box (D80A81L82T83V84) (Figure 9) . The TonB box is at the end of the plug domain and close to the FecR binding site. Complex crystal structures of other TonB dependent transporters with a TonB fragment indicate that TonB and the TonB box interact through the addition of a β-strand of the transporter's TonB box to a β-sheet in TonB [107, 108]. Interestingly, these structures also indicate there are other interactions between TonB and its transporters other than the TonB box ([107, 108]. These other TonB/transporter interactions appears to occur whether or not the TonB box is exposed, as most TonB dependent proteins also have an affinity for TonB with or without the ligand [111, 112, 107]. It appears that TonB may have transient associations with the transporters, but until it recognizes the TonB box that is exposed on ligand binding, TonB can not act on the transporter. It is still not clear how TonB acts on its transporter to stimulate transport.
There is extensive evidence that suggests that TonB is not only required for FecA mediated transport of iron-citrate into the periplasm, but that TonB is also required for FecARI mediated signaling. Removal of the TonB box eliminates signaling, and specific point mutations in the TonB box can dramatically reduce signaling . Dissipation of TonB's power source, the proton motive force (PMF) across the inner membrane, with CCCP and DNP eliminates iron-citrate stimulated signaling by FecA . Deletion or mutations in other components of the TonB system (ExbD or ExbB) also abrogate signaling as well as transport [88, 89]. Interestingly, TonB independent signaling occurs in the FecA4 mutant (L157P/N529D/R611C; located in the plug domain, extracellular loop 7, and TM beta strand 17, respectively) . However, with this mutant protein the transcription induction is only 10% that of wild-type .
It is not clear why TonB is required for FecA mediated signaling of the presence of iron-citrate. This dependence on TonB suggests that the conformational change required for signaling is energy intensive. One can rationalize an evolutionary reason for the TonB requirement for FecA signaling. Because FecA requires TonB, TonB dependent signaling by FecA ensures that TonB is available to assist in iron-citrate transport when FecA expression increases. Although the activity of FecA4 is low, the existence of this constitutively active FecA mutant suggests that signaling by FecA may not necessarily require a continuous application of energy from TonB. During the TonB dependent transport cycle of FecA, FecA may adopt a transient conformation that signals the presence of iron to FecR as part of the process of iron-citrate transport.
Although ligand binding appears to disorder the TonB box and proximal residues, NMR structures of the N-terminus of FecA indicate that about 10 amino acids from the TonB box the N-terminal domain are ordered [92, 93]. The crystal structures of FecA were probably unable to capture the structure of the N-terminal domain due to the flexibility of the region between the N-terminal domain and the plug domain. The NMR structures indicate that the N-terminal domain is highly ordered up to residue 79 and beyond this the structure becomes much more flexible [92, 93]. The N-terminal domain has a novel fold consisting of two α-helices side by side which are sandwiched between two small β-sheets [92, 93]. DNA sequence analysis of N-terminal extensions from other TonB-dependent proteins indicates that this fold appears to be conserved ; recently a structure was determined of the ferric pyoverdine receptor (FpvA), including its N-terminal signaling domain, confirming the structure conservation of the signaling domain . In this structure, the linker region between the TonB box and signaling domain was disordered, suggesting flexibility in vivo.
The striking similarity of the signaling domains observed in the X-ray crystal structure of the ligand bound form of FpvA and the NMR structure of the FecA N-terminus suggests that no structural change occurs within the N-terminus itself for signaling . One caveat is that active TonB is required for the signaling process, and it is not included in these structures for obvious reasons. However, it seems unlikely that TonB would affect the structure of the N-terminus as part of signaling. As TonB is acting mechanically on FecA to power ligand transport, signaling is likely to be a part of the transport process. Computer modeling of BtuB suggests that transport of the plug bound ligand in TonB dependent transporters can be mediated by TonB dependent unraveling of the plug domain through a pulling action on the TonB box . Although the exact mechanism of TonB dependent transport is not clear, any movement of the TonB box mediated by TonB will cause a movement of the N-terminus relative to the rest of FecA and FecR. This movement would be a clear method of mechanical signaling to FecR.
Study of mutants of both FecA and FecR indicate that FecR interacts at a shallow depression that lies at the end of α2 and/or along β2 (Sites 1 and 2 respectively) in the N-terminal domain of FecA (Figure 10) [115, 116, 93]. However, mutations of some surface residues well outside this region also interfere with signaling. There may be more than one surface of contact with FecR, or some of these residues may somehow be involved in signaling without participating in the FecI-FecR interface. Furthermore, the N-terminus of FecA and the C-terminus of FecR are likely to be constitutively associated, as portions of these two proteins interact with out any iron-citrate present in LexA hybrid repression studies .
At the FecR C-terminus (residues 237-317), only the last 80 residues are required for the interaction with FecA . Mutations that abrogate signaling cluster in and around the conserved leucine rich motif of FecR (247-286) . This leucine rich region may be important for creating an interaction surface with FecA . Some mutants of FecR are constitutively active and may mimic the conformation of FecR that is activated by FecA. These mutants include FecR (D138E/V197A) which is not only constitutively active but also binds to FecA more strongly . Another single missense mutant, FecR (S127F), is also constitutively active .
The constitutive activity of the N-terminal domain of FecR in comparison to its inactive full length form raises questions about the activation of FecR. The cytosolic portion of FecR, FecR1-85, is constitutively active. FecR fragments that include the transmembrane domain and even parts of the periplasmic domain are also constitutively active . However, the longer FecR truncation mutants that extend to residues 207 and 273 are much less active than the shorter constructs . The 273 residue construct is particularly less active. Together these data indicate that some aspect of the C-terminus of FecR inhibits the activity of the N-terminus of FecR. Iron-citrate binding to FecA and the resulting structural changes may allow FecR to undergo a conformational change that relieves the inactivation imposed by the C-terminal end of FecR.
The exact role FecR plays in the signal transmission across the inner membrane is not understood. With a single pass transmembrane domain and as a monomer, it is difficult to explain how a conformational change could be propagated from one side of the membrane to the other. It is also difficult to explain how the C-terminal portion(s) of FecR appears to play an inhibitory role.
Braun et al suggest that the signal could traverse the inner membrane through proteolysis of FecR by RseP, the RIP protease from the σE activation pathway . RseP has wide substrate specificity so it may cleave FecR, and RIP proteolysis would be consistent with the constitutive activity of the FecR1-85 fragment. In a mechanism similar to that of σE activation, one can imagine that the activated TonB/FecA complex may alter or displace the FecR periplasmic domain to expose a cleavage site to an unidentified protease. Parallel to proteolysis of the RseA periplasmic domain, removal of the periplasmic domain of FecR may remove inhibition of RseP. One can also imagine that activated TonB/FecA complex may alter the structure of FecR so that FecR no longer inhibits the RseP intramembrane protease activity. The proteolysis by RseP may then trigger a conformational change in the FecR N-terminus that activates FecI. Alternatively, FecI could be inhibited by the cell membrane and hydrolysis of the tether to the membrane would allow it to be activated.
Contrary to the anti-sigma factor paradigm of ECF sigma factor regulation, the activated anti-sigma factor FecR induces FecI to bind the β′ subunit of RNA polymerase. Prior to activation there is some slight anti–sigma factor activity of FecR as Lex-hybrid inhibition studies show that inclusion of FecR inhibits the interaction of FecI and a β′1-317 N-terminal fragment . However, unlike other anti-sigma factors, FecR is required for full FecI activity . FecR1-85, which has been shown to constitutively signal for the increased transcription of the FecA promoter, increases the affinity of FecI to the β′ fragment of RNA polymerase . This FecI-β′ interaction correlates with FecI activity, as the FecI mutant which does not stimulate transcription can not bind to β′ . FecI can only perform its function and recognize the fecA promoter as part of the holo-RNA polymerase . Using a His10-affinity tag on the β′1-317, FecI be can be copurified, and the FecR85 can be copurified with both FecI and β′1-317 . Therefore, FecR does not interfere with the FecI-RNA polymerase interaction and together FecI, FecR and RNA polymerase can form an activated complex.
FecR binds specifically to region 4 (σ4) of FecI. Sequential deletions and single point mutations in FecI revealed that only σ4 is required for FecR interaction . This region is thought to contain a helix-turn-helix motif that is critical to the interaction with FecR . Substitution of leucines with prolines within this motif eliminates binding to FecR, but replacing these prolines with arginines allows moderate to full recovery of binding. These data suggest that this structure is critical for the interaction between FecI and FecR . In most sigma factors, this region is also important for recognition of the -35 region of the promoter . However, studies of FecI promoter interactions indicate that -35 recognition may not be important for FecI-fecA promoter binding. Although alterations in the -35 region were found in promoter mutants that could not bind FecI-RNA polymerase, these alterations were also accompanied by changes proximal to the +13 position, which is well outside the canonical promoter region . Single substitutions in the fecA promoter near position +13 also affect FecI binding . Thus, the FecI-RNA polymerase complex is likely to have a novel interaction with the fecA promoter DNA, even compared to other ECF sigma factors. Perhaps this new type of promoter interaction of the sigma factor in the context of the holoenzyme complex allows FecR to activate FecI and remain bound even on polymerase binding.
As FecR85, FecI and β′1-317 can all interact simultaneously, the model for activation suggests that FecR is included in the RNA polymerase initiation complex . Past models have suggested that FecR inhibits FecI, and on activation by FecA, FecR frees FecI in order for FecI to associate with the RNA polymerase complex. Given that FecR, FecI, and β′ can form a tertiary complex, and that activated FecR increases both the affinity of FecI for β′1-317 and the activity of FecI, FecR is likely to remain attached to FecI when FecI recruits RNA polymerase. Although FecI would remain attached to the membrane bound FecR, FecR would not keep the polymerase tethered to the membrane for the duration of transcription. Only about 10 bases are polymerized before the polymerase switches from the transcription initiation to elongation state, and sigma factors are released near this transition . However, this does pose an interesting problem for promoter recognition. If the complex is tethered to the membrane it would seem to be restricted in its ability to find the proper promoter.
The ultimate fate of FecI likely depends on whether or not FecR remains bound to FecI when the RNA polymerase begins transcription. If the FecIR-RNA polymerase complex remains intact during transcription initiation, then upon removal of FecI from the polymerase complex, the proteolytically sensitive FecI would likely be degraded. However it is possible that the FecIR complex may re-form and return to its inactive state. This is of course assuming that in the absence of iron-citrate the signaling process is reversible, which is likely, as there is no known covalent modification or proteolysis involved. Furthermore, studies indicate that even on repression of FecI expression by Fur-Fe+3, FecI is a long-lived activated species . This may, in part, be due to the continued activation and protection by FecR. Given that FecR stabilizes FecI from proteolysis, FecI would probably be proteolysed in the absence of the protection by FecR [121, 119]. On conclusion of the RNA polymerase initiation step and the release of FecI, FecI may be proteolysed if FecR does not remain attached through the whole initiation process. Given that FecIR can form a complex with the β′ subunit of RNA polymerase, it is a distinct possibility that the FecIR complex remains intact during transcriptional activation.
The other TonB dependent proteins with N-terminal extensions in various bacterial species are likely to be involved in a similar signaling mechanism to the FecARI signaling mechanism. However, a genome wide search for TonB dependent transporters (by homology to the TonB box) and TonB dependent transducers (by homology to the N-terminal signaling domain) indicated that the TonB-dependent signaling is the exception rather than the rule  Of the 84 genomes that have TonB-dependent receptor/transporters, only 26 of the genomes have TonB-dependent transducers. In most of these 26 organisms, only one to six different TonB-dependent transporters are present . Only in three organisms were many more TonB-dependent transducers found: Nitrosomaonas europaea (24), Pseudomonas aeruginosa (14), and Pseudomonas putida (11) . As other FecARI type signaling systems would depend on inner membrane components similar to FecR, the genomes were also analyzed for FecR-related proteins. There is a tight correlation between the number of TonB dependent transducers and the number of FecR-like proteins. Interestingly, this analysis indicates that although the FecARI like signaling is found in several environmental bacteria and some pathogens, this signaling appears to be missing in several important human and animal pathogenic genera including Brucella, enteropahtogenic E. coli (EPEC), Haemophilus, Helicobacter, Neisseria, Salmonella, Shigella, Vibrio, and Yersinia .
Most systems homologous to FecI appear to be involved in iron signaling and transport. The transporters for these systems include: PupA, HasR, FiuA, FpvA, PubB, PbuA, BfrZ, and BhuR [123-125, 97, 126, 93]. Although these systems have the components seen in FecARI, they may not necessarily use the signaling components in the same way. For example, rather than having an activating effect on its associated sigma factor, the FecR homologue PupR appears to have a distinctly inhibitory effect on the sigma factor PupI . Similarly the anti-sigma factor HasS appears to have a solely inhibitory effect on the sigma factor HasI in regulation of heme uptake in Serratia marcescens . Similar to the FecARI system, HasI does not regulate its own expression, but unlike the relationship between FecI and FecR, HasI does regulate the expression of HasS .
One known TonB dependent transducer does not transport ligand. The PrhARI system from the pathogen Ralstonia solanacearum binds a portion of a plant cell wall and signals the presence of the plant to the bacterium . This signal induces the hypersensitive response and pathogenicity genes that are essential for its interaction with non-host plants and induction of the type III secretion system required for disease development in host plants .
Some TonB dependent receptors/transporters also contain an extra domain called the Oar domain . When present, the Oar domain occurs between the N-terminal extension and the TonB box . The role of this domain is not yet clear. Although the Oar domain can be found in proteins without the N-terminal signaling domain, it is interesting to consider that the Oar domain may promote protein-protein interactions that may affect regulation of the FecARI type signaling.
Signaling in ECF activation takes an interesting variety of forms. In this review, we have discussed the two major groups of signaling mechanisms represented by the σE and the FecI signaling. Some of the associations with these groups are based on what appear to be somewhat similar mechanisms rather than homology, like the association of CarQ signaling with σE signaling. Although the components of the CarQ signaling pathway have no homology to the components of the σE signaling pathway, the evidence that activation of CarQ is concomitant with anti-sigma factor degradation suggests a mechanism similar to that of the regulation of σE. Most other systems of ECF regulation, especially in Gram-negative bacteria, appear to have components that are homologous to components of either the σE or the FecI signaling systems. However, these components are often not used identically and illustrate the great diversity of the use of the components of the signaling systems.
Comparison of the σE-like and the FecI-like signaling systems suggests that one simple explanation for some of their differences is that these systems communicate conditions from different cellular locations. The FecARI system signals the presence of extracellular iron-citrate through two membranes and the intervening periplasmic space. The FecARI model of signaling is appropriate for signaling the presence of small molecules that only reach the periplasm by transport, such as chelated iron in the case of FecARI. The two-component, Cpx-like and the RIP protease, σE-like systems that recognize periplasmic stimuli could possibly be used for signaling. However, as periplasmic detection of the extracellular molecules would first require transport, this method of detection and signaling would not be as efficient as the FecARI-like system. This is especially the case for the PrhARI signaling system from Ralstonia solanacearum, because this system detects an integral part of an intact plant cell wall which can not enter into the periplasm.
The σE-like activation pathway is appropriate for recognition of periplasmic stimuli in Gram-negative bacteria or external stimuli in Gram-positive bacteria. Because the detection of the periplasmic stimuli of σE could be detected by a two-component signaling system such as Cpx, activation of many of the regulons with the σE-like pathway may alternatively be regulated by a two-component signaling pathway. It is not clear what pressures determine why a regulon would be controlled by one or the other, thus an interchangeability of these signaling systems is suggested. This is partly reflected by a σE homologue in the Gram-positive bacterium Streptomyces coelicolor A3(2), which appears to be regulated by a two-component system . Although Gram-positive bacteria do not have a periplasm, σE-like RIP protease mechanisms appear to play a role in regulation of ECF sigma factors in these organisms as well. The σW signaling system from B. subtilis contains an anti-sigma factor and a RIP protease. Given the universality of the RIP protease mechanism, it should be no surprise that ECF signaling systems in Gram-positive bacteria use σE-like RIP protease mechanisms.
An almost universal commonality to ECF signaling, especially in Gram-negative bacteria, is the use of anti-sigma factors. The two major mechanisms of ECF sigma factor regulation and the variations all use anti-sigma factors. Anti-sigma factors are probably retained for a number of reasons. Two coupled reasons are that the anti-sigma factor protects the sigma factor from degradation and the sigma factor is ready to act as soon as it is activated or released. The use of an anti-sigma factor as a signaling component also lends itself to a variety of regulatory mechanisms which are not limited to those presented in this review .
Another commonality in signaling in these systems is protein recognition through addition of a β-strand of one protein into a β-sheet in the other. In the ECF signaling mechanisms these interactions are primarily mediated by PDZ domains. The σE signaling mechanism contains PDZ domain interactions with both DegS-OMP recognition as well as regulation of RseP. The Prc protease that plays a role in the regulation of AlgU in mutant P. aeruginosa also contains a PDZ domain which may play a role in AlgU activity. In the FecARI signaling system, TonB interacts with the TonB box through β-strand addition.
In summary, ECF sigma factor signaling is extremely diverse, most likely in order to address different types of signals from various locations. The signaling mechanisms often use similar components, but in different signaling systems these components can be used in different ways.
We are grateful to Andrew Perry and Jennifer K. Hill for critically reading this review and helpful discussions. This work is supported by the Intramural Research Program of the NIH, National Institute of Diabetes and Digestive and Kidney Diseases.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.