A few years after the initial report of a new heat-shock sigma factor in E. coli, the gene specifying this σE activity, rpoE, was identified by two independent experimental approaches. The first one involved searching for sequences homologous to the known σ28 sigma factor of Streptomyces coelicolor (Lonetto et al., 1994). The isolation of S. coelicolor sigE gene (encoding σ28), using reverse genetic experiments, resulted from biochemical studies of transcription of the dagA gene. dagA encodes an extracellular agar-degrading enzyme. The dagAP2 promoter was shown to be transcribed by Eσ28in vitro, however, it turned out that this is not the case in vivo as dagAP2 recognition occurs even in sigE mutant strains (Jones et al., 1997). Clearly, S. coelicolor must encode an additional sigE-like gene product closely related to σ28. Nevertheless, searches for sequences homologous to SigE revealed a whole subfamily of sigma factors, including E. coliσE.
The second experimental approach involved isolating mutations leading specifically to decreased transcription from the htrA and rpoHP3 promoters of E. coli (Raina et al., 1995). Such a genetic screen was facilitated by using promoter fusions to the promoterless lacZ gene and scoring for trans-acting point mutations leading to lower β-galactosidase activity. This approach led to the identification of the rpoE gene encoding a polypeptide sharing sequence similarity to the housekeeping σ70 class of sigma factors. Surprisingly, E. coliσE appeared to be more closely related to S. coelicolor SigE compared with E. coliσ70 (Lonetto et al., 1994). Also, RpoE and SigE show sequence conservation with the AlgU polypeptide of P. aeruginosa, which was reported previously (Martin et al., 1993).
Global similarity searches revealed that various bacteria encode closely related sigma factors (sequences updated in Table 2). All these σE homologues seem to function in response to events occurring in the extracytoplasm and Lonetto et al. (1994) designated this group as the extracytoplasmic function sigma factor, or ECF subfamily. The different sequences of the known σE homologues carry distinctive features that distinguish them from the main σ70 class of sigma factors. A particular feature is the presence of a very short region 3 compared with σ70. This is one of the four highly conserved regions as defined by alignments of the σ70 family of proteins. Regions 1 and 3 exhibit lower conservation and are acidic. Regions 2 and 4 tend to be very basic and are highly conserved (Lonetto et al., 1994), which is consistent with their role in DNA binding. Accordingly, loss of function mutations obtained in the E. coli rpoE gene map to region 2.1 and 4.2 respectively (Raina et al., 1995). Region 2.1, also designated the ‘RpoD box’, is involved in binding to the β-subunit of RNA polymerase core enzyme, whereas region 4.2 encodes a predicted helix–turn–helix motif involved in recognition of the promoter −35 box. The σE protein was purified after overexpression of the cloned rpoE gene and, when mixed with core RNA polymerase, was shown to selectively transcribe the htrA and rpoHP3 promoters in vitro (Raina et al., 1995; Rouvière et al., 1995). Similarly, in P. aeruginosa, the algU (algT ) gene is transcribed by the RNA polymerase–AlgU complex (Schurr et al., 1996; Xie et al., 1996). In E. coli, rpoE was found to be the first gene of a four-gene operon designated rpoE rseA rseB rseC (De Las Peñas et al., 1997a; Missiakas et al., 1997). Mutational analyses revealed that the three other genes in this operon play an important role in the signal transduction process coupled to the extracytoplasmic function of the σE controlled regulon. This genetic organization (rpoE rseA rseB rseC ) is conserved in the genome of other bacterial species such as P. aeruginosa, Azotobacter vinelandii, Photobacterium spp. strain SS9 and M. xanthus (Table 2). In some cases, such as Haemophilus influenzae, only the first three genes are found in an operon. The fourth gene, i.e. rseC, is duplicated elsewhere in the genome. Even in the best-studied cases, the exact role of this fourth gene product is not known.
Consensus sequences of σE-dependent promoters
With their original discovery of E. coliσE, Erickson and Gross (1989) had already noticed that the S. coelicolor dagA gene contains a promoter, dagAP2, with significant homology to rpoHP3 and htrA promoters. With the identification of multiple ECF, it has become obvious that there is a striking similarity among the cognate promoters they recognize (Table 1). The −35 sequence is very well conserved, as is the spacing between the −35 and −10 regions (Table 1). The absence of real similarity in the −10 sequences can be explained by the low degree of resemblance in region 2.4 in the members of the ECF subfamily. Nevertheless, these alignments of promoter sequences as displayed in Table 1 suggest that ECF factors have rather similar promoter specificity. In fact, it has been shown that the purified AlgT (AlgU) is capable of recognizing the rpoHP3 promoter of E. coli in vitro (Hershberger et al., 1995).
The diversity of the −10 sequences of promoter and the region 2.4 of ECF factors can probably account for the co-existence of multiple members of the ECF subfamily in the same species. For example, both σE and FecI or AlgU and PvdS co-exist in E. coli and P. aeruginosa, respectively, but do not overlap in their biological functions. It seems that Bacillus subtilis carries up to six ECF factors based on sequence comparison. Also, as mentioned above, SigE (σ28) is able to transcribe dagAP2 in vitro (Lonetto et al., 1994), but sigE mutants do not abolish dagAP2 transcription in vivo. Similarly, the Streptomyces antibioticus SigE homologue has been shown to direct transcription of the phsA gene in vitro but is dispensable in vivo (Jones et al., 1997). These findings argue that Streptomyces species must carry at least two related ECF factors with relaxed specificity in vitro. Such relaxed specificity has already been observed for σ70 and σS in E. coli (Tanaka et al., 1993) or σB and σH in B. subtilis (Igo et al., 1987). In E. coli, this relaxed specificity occurs in vivo as well because some promoters may be recognized equally well by both Eσ70 and EσS polymerases, depending on the physiological conditions (Loewen and Hengge-Aronis, 1994). Hence, specific transcription by ECF factors could be modulated in vivo at various levels including promoter stringency, availability of the ECF factor, physiological state of the cells and environmental stresses.
Stimuli inducing σE-dependent regulon expression in E. coli
The first direct evidence that the E. coliσE factor belongs functionally to the ECF family was the finding that transcription of the σE regulon is specifically induced in response to imbalanced synthesis of outer membrane proteins (Mecsas et al., 1993). This view was substantiated with the finding that misfolding of transported proteins, in general, results in a σE-dependent response (Missiakas et al., 1996).
The above observations were made in earlier attempts to identify the E. coli rpoE gene by screening multicopy libraries for increased σE-dependent transcription using the htrA (Mecsas et al., 1993; Raina et al., 1995) and the rpoHP3 transcriptional fusions to lacZ (Table 3). Identification of the cloned sequences provided valuable information indicative of the stimuli that might trigger a σE-dependent response. Two of the multicopy clones encoded two new outer membrane proteins OmpX (Mecsas et al., 1993) and OmpK (C. Dartigalongue and S. Raina, unpublished). Overproduction or imbalanced synthesis of OmpC and OmpF proteins were also shown to induce the σE-dependent response (Mecsas et al., 1993). Three other multicopy clones specified proteins known to be located in the periplasm of E. coli : asparaginase B (AsnB) and penicillin-binding protein 2 (PBP2) (Missiakas et al., 1996) and, in the inner membrane, DsbD. As for AsnB, the isolated clone resulted in the expression of a C-terminally truncated polypeptide that was presumably misfolded in the periplasm. Increased expression of DsbD, a thiol-disulphide reductant, probably resulted in the accumulation of reduced and thus misfolded polypeptides in an environment that is normally oxidizing. Finally, overproduction of PBP2 was thought to cause an alteration of peptidoglycan structure, thereby affecting the bacterial envelope and its folded proteins in general.
Table 3. . Inducers of the σE response.
Mutations that result in the expression of altered lipopolysaccharide (LPS), for example mutations in the htrM (rfaD ) gene, also cause the constitutive expression of the σE protein (Missiakas et al., 1996). Because it is known that alterations of LPS structure greatly affect the ratio of outer membrane proteins (OMPs), this result provided additional evidence for the hypothesis of Mecsas et al. (1993), who proposed that the accumulation of misfolded OMPs in the extracytoplasmic compartment was responsible for the σE-dependent stress response.
Mutations in genes known to catalyse the folding of proteins in the periplasm were found to have a similar effect. For example, mutating the dsb genes, which encode thiol:disulphide oxido-reductases (DsbA/B/C/D/G) that are required for the proper folding of disulphide-containing proteins in the periplasm (Missiakas and Raina, 1997), led to a two- to fivefold increase of σE activity. Mutations in the surA, fkpA and skp genes were found to confer a constitutively elevated σE-dependent response (Table 3). SurA and FkpA are two periplasmic peptidyl-prolyl isomerases, whereas Skp promotes the assembly of OMPs into the membrane bilayer (Missiakas et al., 1996).
Taken together, these data indicate that the accumulation of misfolded polypeptide in the periplasm leads to a σE-dependent heat-shock response. This induction is specific because it does not affect the cytoplasmic heat-shock genes controlled by the σ32 regulon. Therefore, it can be concluded that the two heat-shock regulons have evolved to respond to protein misfolding in different cellular compartments. The σ32 regulon senses misfolding in the cytoplasm, whereas the σE regulon responds to the accumulation of misfolded proteins in both the periplasm and the outer membrane.
The σE regulon in E. coli
The observation that σE participates in the regulation of heat-shock genes in E. coli strongly suggests that it functions to sustain bacterial growth under extreme stress conditions. The promoters of htrA and rpoH genes were the first found to be heat-shock regulated in a σE-dependent manner. rpoH encodes the major heat-shock σ factor, σ32, which is responsible for the transcription of genes such as groE, dnaK, dnaJ or grpE (reviewed by Gross, 1996). Many of the σ32 transcribed genes are involved in folding or degradation of heat damaged polypeptides in the cytoplasm. htrA encodes a periplasmic protease that has been shown to degrade abnormal proteins. More recently, an additional gene, fkpA, has been shown to belong to the σE regulon (Danese and Silhavy, 1997; C. Dartigalongue and S. Raina, unpublished).
fkpA encodes a peptidyl prolyl isomerase belonging to the FK506 binding protein family. This gene has been isolated in a genetic selection aimed at identifying periplasmic-folding catalysts, which in multicopy can correct many periplasmic-folding defects (Missiakas et al., 1996). In multicopy, it can also dampen the increased σE-dependent response constitutively induced in the absence of various periplasmic folding catalysts (Missiakas et al., 1996). The σE-dependent promoter of fkpA is rather weak.
None of the genes belonging to the σE regulon is essential. However, rpoE itself is known to be essential for E. coli growth at all temperatures (De Las Peñas et al., 1997b). Thus, it is very likely that some other genes are transcribed by EσE. A possible candidate is the newly identified gene ompK (Table 4), which appears to be essential for bacterial growth and whose transcription is EσE dependent (D. Missiakas et al., unpublished).
Table 4. . E. coliσE regulon.
Mechanism of regulation of σE and AlgU (AlgT) activity
In E. coli, genetic experiments have revealed that several genes are responsible for regulating σE activity in vivo. A search for mutations that cause increased LacZ expression in the E. coli reporter strains htrA–lacZ and rpoHP3–lacZ, even in the absence of misfolded polypeptide, led to the identification of the rse genes (regulator of sigma E). The gene designated as rseA was demonstrated both genetically and biochemically, to encode a major, negative regulator of σE activity (De Las Peñas et al., 1997a; Raina et al., 1997). rseA is the second gene of the rpoE four-gene operon (Table 2). In P. aeruginosa, it was known for a long time that inactivation of the mucA gene is responsible for conversion to mucoidy (Martin et al., 1993). mucA was later found to be the second gene of the algU (algT ) four-gene operon (Table 2). MucA is homologous to RseA and its anti-AlgU activity has also been demonstrated in vitro (Xie et al., 1996; Schurr et al., 1996), based on the following observations.
First, loss of function mutations in rseA or mucA lead to constitutive, high-level expression of σE and AlgU (AlgT), respectively, whereas overexpression of RseA/MucA strongly inhibits σE/AlgU activity. Second, RseA and MucA were shown to directly interact with σE or AlgU (AlgT), and as a consequence to inhibit the transcriptional activity of these factors in vitro (Xie et al., 1996; Schurr et al., 1996; De Las Peñas et al., 1997a; Missiakas et al., 1997).
An important clue for the specific role of RseA/MucA in signalling the σE/AlgU-dependent response was provided by studying their subcellular location. Both anti-sigma factors were shown to be inserted in the cytoplasmic membrane in a fashion that positions the N-terminal domain in the cytoplasm, whereas the C-terminal part protrudes into the periplasmic space. It was demonstrated, both in vivo and in vitro, that the N-terminal, cytoplasmic domain of RseA interacts directly with σE and possesses a σE-specific anti-sigma activity (De Las Peñas et al., 1997; Missiakas et al., 1997). Finally, and as speculated in the model (Fig. 1), a possible role for RseA periplasmic domain could be to sense misfolded polypeptides and serve as a ‘signal-transducing’ domain, in a manner reminiscent of histidine kinases of the two component regulatory systems.
Figure 1. . Functional model of ECF factor. The model depicts the sequestration of the ECF σ factor by its cognate anti-sigma factor. The existence of such an interaction has been demonstrated either biochemically or genetically for E. coli (De Las Peñas et al., 1997a; Missiakas et al., 1997), P. aeruginosa (Schurr et al., 1996; Xie et al., 1996), Photobacterium sp. (Chi and Bartlett, 1995) and M. xanthus (Gorham et al., 1996). The stimuli leading to the release of the ECF factor are thought to be protein misfolding in the extracytoplasm (E. coli ) blue light (M. xanthus) and changes in temperatures and pressures in (Photobacterium sp.). The gene encoding for the ECF factor is often the first gene of an operon (see Table 2) and its transcription is autoregulated. The role of the periplasmic component RseB/MucB (E. coli and Pseudomonas sp. respectively) is not very well understood. Evidence for an interaction between RseB/MucB and the anti-ECF factor have been obtained in case of E. coli and P. aeruginosa (Schurr et al., 1996; Xie et al., 1996; De Las Peñas et al., 1997a; Missiakas et al., 1997).
Download figure to PowerPoint
RseB/MucB are encoded by the third gene of the rpoE/algU operon and both seem to function as negative regulators of σE/AlgU. Null mutations in rseB or mucB lead to increased σE/AlgU activity (Table 3). In vivo, RseB interacts with the C-terminal, periplasmic domain of RseA (Raina et al., 1997; De Las Peñas et al., 1997a). Thus, RseB may function to modulate RseA activity or to interact with misfolded proteins itself (Fig. 1). Its role is yet unclear. rseB and mucB must differ in their ability to regulate their cognate ECF factor. Indeed, mutations in both mucA and mucB lead to equal derepression of AlgU (AlgT) activity (Schurr et al., 1996; Xie et al., 1996), whereas mutations in rseB have a relatively milder effect on σE compared with rseA mutants (4-fold derepression versus 12-fold). Finally, rseC/mucC is the fourth gene of the operon in E. coli and P. aeruginosa. RseC is an inner membrane protein with unknown function. Genetic evidence suggests that RseC may also participate in regulating σE, albeit in a positive manner (Missiakas et al., 1997).
At present, it is difficult to propose a general mechanism of regulation of ECF factors. Nevertheless, there seems to be a conserved mechanism of regulation between the ECF factor and a cognate anti-sigma activity. In most cases, for example in P. aeruginosa, it is not even known what stimulus is sensed and leads to the release of the sigma factor from the anti-ECF factor. How such a release is made possible, is not known either. Conformational change, chemical modification or degradation of the anti-sigma factor are possible mechanisms which could be tested.
Both in P. aeruginosa and in A. vinelandii, AlgU (AlgT) regulates the biosynthesis of the exopolysaccharide alginate (Hershberger et al., 1995; Yu et al., 1995; Martinez-Salazar et al., 1996). Excretion of alginate is considered to be the major pathogenicity factor in chronic respiratory infection of cystic fibrosis patients. Alginate synthesis is the consequence of a co-ordinate expression of a gene cluster located at 34 min on the chromosome of P. aeruginosa. The key gene in this cluster is algD, the first gene of a 12-gene operon. This gene encodes a GDP mannose dehydrogenase that is specific for alginate synthesis because it converts GDP mannose into GDP mannuronate, a direct precursor for alginate polymerization (Hershberger et al., 1995; Yu et al., 1995). Most P. aeruginosa strains are non-mucoid except when isolated from chronically infected cystic fibrosis patients. It is not clear what the stimuli triggering the mucoidy phenotype are, but the environment of cystic fibrosis lung appears to play a major role in this matter.
Light-induced carotenogenesis and outer membrane protein synthesis
M. xanthus and Photobacterium spp. both encode an ECF factor whose activity is specialized and regulated by a cognate anti-sigma factor. This rpoE homologue in the deep-sea organism Photobacterium sp. strain SS9 has been shown to regulate the synthesis of outer membrane proteins in response to changes in osmolarity, barometric pressure and temperature (Chi and Bartlett, 1995). This homologue is the first gene of a four-gene operon and encodes its own putative anti-sigma factor, quite like in E. coli.
In M. xanthus, carotenoid synthesis has been shown to be regulated by the CarQRS operon. Carotenoids are presumed to be light-protective agents that prevent cell lysis in the light. Their synthesis is light inducible and regulated by CarQ, an ECF factor. Both CarQ and CarR directly control the expression of the carQRS operon as well as the carC gene, which encodes phytoene dehydrogenase. CarS is required for the expression of the carB gene cluster involved in the rest of the structural genes of the carotenogenesis (Gorham et al., 1996). Genetic studies have shown that regulation of carQRS is dependent on the stoichiometry of CarQ and CarR, as well as the translational coupling of the two corresponding genes. CarR is a putative anti-sigma factor that is believed to sequester CarQ in the inner membrane, even although its own topology has not been proven yet. It has been proposed that CarR must be directly destroyed (degraded) by a periplasmic light-induced signal (blue light interacting with the protoporphyrin IX photosensitizer), and thus lead to the release of CarQ from the membrane (Gorham et al., 1996).
FecI is another E. coli ECF factor. It controls the transcription of genes whose products are needed for the periplasmic iron dicitrate transport (Enz et al., 1995). These genes, fecABCDE, form an operon and their synthesis appears to be regulated by both citrate and iron in a FecI- and FecR-dependent manner. The FecR is membrane located and responds to iron dicitrate concentration. FecR plays a critical role in FecI activation, but it has presumably diverged from RseA or MucA functions (Enz et al., 1995). In fact, FecR has not been shown to act as an anti-sigma factor, and its activity remains to be elucidated. A rather similar specialized sigma factor, called PupI, exists in P. putida and is involved in pseudobactin transport (Table 2).
Both the mycobacterial SigE and the B. subtilis SigX proteins are extracytoplasmic function sigma factors which have been recently shown to be involved in survival after stress. The sigE gene has been cloned and characterized in various mycobacterial species. A comparison between wild type Mycobacterium segmatitis and a sigE mutant strain demonstrated decreased survival of the mutant under different stress conditions such as high temperature, acidic pH, exposure to detergents and oxidative stress (Wu et al., 1997).
B. subtilus strains lacking the sigX gene have been shown to be impaired in their ability to survive at high temperature. SigX controls its own gene transcription, which is favoured under high salt, whereas another factor σA predominantly controls the transcription of the sigX region under lower salt conditions (Huang et al., 1997). The exact mechanism leading to such a regulation is not very well understood yet.