FUNCTIONS OF The ECF s FACTORS

Strategies to determine the function of ECF s factors.

            While many of the founding members of the ECF s factor sub-family were discovered first as genetic regulators of known function, the advent of genomic sequencing has led to the inevitable discovery of numerous ECF s factor genes for which functions can not easily be predicted. By far, the most dramatic example of this challenge is the recent discovery of ~50 ECF s factors encoded by S. coelicolor genome (Paget et al., 2002). Thus, we are faced with an increasingly familiar problem in functional genomics: How can we determine the function of regulators identified by genome sequencing?

In many ways, this is a familiar problem for those studying s factors in Gram positive bacteria. In the late 1970s studies of B. subtilis RNAP revealed multiple associated s factors with, a priori, no clear indication of function. The role for many of these factors was gradually elucidated by construction of mutant strains (using “reverse genetics”), analyses of in vitro and in vivo transcription selectivity, and the identification of target genes (Haldenwang, 1995). These studies, for example, led to the indentification of s factors controlling flagellar motility (s^D),  the sporulation cascade (s^E, s^F, s^G, s^K), and the general stress response (s^B). Similarly, biochemical fractionation studies in S. coelicolor revealed a large number of alternative s factors with distinct in vitro transcription selectivity (Buttner, 1989; Buttner et al., 1988; Chater et al., 1989). Assigning physiological roles to these various factors represents a formidable challenge.

            Several strategies can be envisioned to determine the physiological roles of newly described ECF s factors. First, mutant strains lacking one or more ECF s may have phenotypes that will provide clues to function. Second, target genes can be identified and, by understanding their functions, we may be able to predict the phenotype conferred by a s factor mutation. Third, physical stimuli or genetic changes that act to induce expression of each s factor regulon can be identified and used to infer possible function. For example, overexpression of a particular ECF s regulon by mutation of the cognate anti-s factor may reveal a more dramatic phenotype than mutation of the s factor gene itself. Examples of these approaches, and their advantages and limitations, are summarized in Table 2.

Escherichia coli

            E. coli contains 7 s factors including 2 members of the ECF sub-family: s^E and s^FecI. The s^E regulon is activated in response to "periplasmic stress" or extreme heat shock and controls the expression of proteases and folding catalysts active in the periplasm. s^FecI controls the expression of the ferric-citrate uptake system in response to the presence in the periplasm of ferric citrate (Angerer et al., 1995; Braun, 1997). For more detailed discussions of the s^E regulon and its role the reader is referred to other recent reviews (Missiakas and Raina, 1998; Ravio and Silhavy, 2001).

  s^E

            The activity of s^E was originally discovered by Erickson and Gross (1989) as a holoenzyme form (Es²⁴) necessary for transcription of the group 3 heat shock s, s³², at very high temperatures (50°C) due to activation of a new promoter site. A similar promoter controls the heat-inducible periplasmic protease DegP(HtrA) (Lipinska et al., 1988). This led to the hypothesis that E. coli contained a second heat shock regulon, activated by a new s factor. This early work predated the isolation of the gene encoding s^E (Raina et al., 1995; Rouviere et al., 1995), so it was not clear that s^E would in fact become a founding member for the ECF sub-family.

            Extensive work on the s^E regulon in E. coli has led to a detailed model for the activation of this system in response to periplasmic stress (Ravio and Silhavy, 2001). Periplasmic stress can be elicited in several ways including the overexpression of outer membrane proteins or by the production of misfolded proteins in the periplasm (Missiakas and Raina, 1998). The latter condition can also be elicited genetically by mutations in folding catalysts. The importance of this stress response is underscored by the finding that s^E is essential in E. coli (De Las Penas et al., 1997a).

            Many of the key regulators of s^E activity (regulators of sigE; rse genes) are cotranscribed with rpoE to form an operon: rpoE rseA rseB rseC (De Las Penas et al., 1997b; Missiakas et al., 1997). The RseA membrane protein functions stoichiometrically as an anti-s factor and its action is enhanced by the periplasmic RseB protein (Figure 2). The role of the RseC protein is not yet clear, but it has been implicated in thiamine biosynthesis (Beck et al., 1997). Activation of the s^E regulon occurs when s^E is released from RseA inhibition: a phenomenon accompanied by the degradation of RseA by Hho(DegS) protease (Ades et al., 1999; Alba et al., 2001).

            The first two defined targets for s^E, the rpoH and htrA genes, have identical promoter consensus elements (Table 3). Indeed, the similarity between these promoters and the S. coelicolor dagA P2 site contributed to the original recognition of the ECF sub-family of regulators. Subsequent work identified another very similar sequence preceeding the rpoE operon that functions as an autoregulatory site.

            Recently, the s^E regulon of E. coli has been defined using a genetic strategy to identify 20 promoter regions that are upregulated in response to s^E overexpression (Table 3). These genes include the known targets of s^E (including rpoH, htrA, and the rpoE operon itself) together with other proteins associated with either the inner or outer membrane and involved in functions such as lipopolysaccharide biogenesis and protein folding (Dartigalongue et al., 2001; Table 4). As seems common in the characterization of regulons controlled by ECF s factors, many of the target genes are unknown function proteins with a predicted location in the cell envelope. Note that in this work the authors have designated these unknown function genes as ecf (extracytoplasmic function). While this is reasonable in E. coli, which has only one other ECF s factor, this practice would clearly lead to confusion in other organisms and a designation such as cse (controlled by sigE) is preferred (see Figure 1). Even this may lead to confusion, as issues of regulon overlap prevent many genes from being uniquely assigned to a single regulon.

            Inspection of the regulatory regions of many genes under s^E control reveals candidate promoters similar to those that have been biochemically characterized (Table 3). Interestingly, many of the proposed -35 and -10 elements do not display strong similarity to the presumed consensus, suggesting that s^E may have a relaxed promoter selectivity compared to other ECF s factors. One can speculate that the presence of only two ECF s factors in E. coli has allowed s^E promoters to tolerate deviations from consensus while retaining recognition by the s^E holoenzyme. In contrast, in organisms with a great many more ECF s factors, deviations from consensus can rapidly switch a target promoter from control by one ECF s to a regulon controlled by a related, but functionally distinct paralog (e.g. Qiu and Helmann, 2001). This can of course be advantageous, and many target promoters do belong to more than one regulon. However, if it is disadvantageous this will act to restrict the sequences of promoters within each regulon and could, in principle, account for the high degree of sequence conservation noted among promoters in some regulons (e.g. B. subtilis s^W and S. coelicolor s^R; see below).

  s^FecI

            The fecIR genes encode regulators affecting transcription of the fecA operon which encodes a specific ferric-citrate uptake system (Braun, 1997). The fecA operon is only transcribed under Fe-limiting growth conditions due to repression by the iron-sensing ferric uptake regulator (Fur) protein (Angerer and Braun, 1998). However, iron limitation alone is not sufficient to induce expression: the fecA operon is activated by the presence of ferric-citrate (Zimmermann et al., 1984). This activation requires the FecI and FecR proteins to signal the presence of the substrate for transport, ferric citrate. The s^FecI regulon appears to consist of this single target operon. Unlike many other ECF s factors, s^FecI does not autoregulate its own synthesis (Braun, 1997).

            The ability of ferric-citrate to activate the transcription of the appropriate uptake genes provides an elegant example of trans-membrane signaling. The signal transduction mechanism likely involves a direct interaction between ferric citrate bound to the outer membrane FecA protein and the periplasmic domain of the FecR regulatory protein (Enz et al., 2000). This interaction leads to the release of s^FecI, bound to the cytoplasmic domain of FecR, and the resulting free s factor then activates transcription of the fecA operon.

            Genetic analyses indicate that FecR plays a positive regulatory role since fecR mutants are not able to efficiently activate fecA transcription (Ochs et al., 1995). Expression of the first 81 amino acids of FecR, encoding just the cytoplasmic N-terminal domain, is sufficient for full activation of a fecA reporter fusion, but this expression no longer requires ferric-citrate (Welz and Braun, 1998). These results support a model in which the cytoplasmic N terminus of the transmembrane FecR protein interacts with s^FecI to convert it from an inactive to an active s factor (Braun, 1997). However, the nature of this activation event has proved elusive. One could imagine that the activation of s^FecI involves, for example, a post-translational modification. In other systems, s factors are synthesized as an inactive pro-protein that is activated by proteolysis (Kroos et al., 1999). The activity of other regulators (although not, to date, s factors) can be controlled by reversible phosphorylation or other types of covalent modification. No evidence has been presented for any such changes in the FecR:s^FecI system.

            An alternative model can be envisioned that reconciles the apparent positive regulatory role of FecR with the role of s^FecI as a s factor for RNA polymerase. The free s factor may be unstable, perhaps due to proteolytic turnover in the cell, and formation of a FecR:s^FecI complex may stabilize the s against degradation (Stiefel et al., 2001). Then, upon release from the complex upon exposure to ferric citrate the s^FecI protein may bind RNAP and catalyze transcription initiation. By this model, FecR would function both as an anti-s factor, and in a positive role to stabilize the otherwise unstable s factor (Figure 3). A similar scenario may pertain to the P. fluorescens ECF s PrtI, which is regulated by a transmembrane "activator" protein, PrtR (Burger et al., 2000).

Bacillus subtilis

            B. subtilis encodes 7 ECF s paralogs that were all initially identified during the international genome sequencing effort (Kunst et al., 1997). None of these seven loci correspond to previously identified genes, making it unlikely that they are essential regulators of any of the most well studied processes in this organism such as endospore formation, genetic competence, or the heat shock and general stress responses.

            To begin to investigate the roles of the various ECF s factors in B. subtilis we and others have sought to determine mutant phenotypes for strains lacking each s, identify target genes for each s, and identify conditions leading to the activation of each s factor regulon. Most studies to date have concentrated on three of these factors: s^X, s^W, and s^M.  The recent discovery of genes encoding 11 ECF s factors in B. halodurans and at least 8 in B. cereus underscores the importance of these regulators in the Bacilli. Remarkably, of the 11 ECF s factors in B. halodurans only one is an obvious ortholog of a B. subtilis factor: s^W (Takami et al., 2000).

 s^X

            The first ECF s factor sequenced in B. subtilis, and the first to attract experimental scrutiny, was s^X. The sigX gene was postulated to encode a s factor based on its similarity to the newly described ECF subfamily in 1994 (Lonetto et al., 1994). This inference was confirmed when the protein product of the sigX gene was overproduced and purified and found to have s factor activity: addition of s^X to RNAP leads to the specific recognition of a distinct autoregulatory site (P_X) not recognized by RNAP containing the major vegetative s, s^A (Huang et al., 1997).

sigX mutants are slightly more sensitive to heat and oxidative stress. The function of this regulator was not immediately apparent as sigX mutants do not display gross phenotypic abnormalities. The only differences relative to wild-type detected in an initial survey were increased sensitivity to heat and oxidative stress (Huang et al., 1997). However, these properties could well be the indirect effect of any number of changes in cell physiology. Since expression of sigX is itself not heat inducible, and heat shock regulation has been carefully investigated in B. subtilis without the identification of any link to sigX (Price, 2000; Hecker and Volker, 2001), it seems unlikely that s^X is a central regulator of the heat shock response. Similarly, s^X does not control transcription of any of the known antioxidant enzymes and oxidative stress responses, again suggesting an indirect effect of the sigX mutation on resistance.

s^X is not an ortholog of s^FecI.  Since s^X is related to the E. coli s^FecI protein, Brutsche and Braun (1997) postulated that perhaps it also controlled iron uptake functions in B. subtilis. However, B. subtilis failed to use ferric-citrate as an iron source and a sigX mutant was not affected in any known ferri-siderophore uptake systems. Surprisingly, expression of sigX in E. coli was found to partially complement a fecI mutation, suggesting that s^X might be able to activate transcription of the fec transport genes (Brutsche and Braun, 1997). It has not been established whether or not this activation involves recognition of the same promoter sequence recognized by s^FecI. In sum, the available evidence suggests that s^X and s^FecI are homologs (evolutionarily related), but not orthologs (they control distinct functions). Indeed, the only known orthologous alternative s factors in E. coli and B. subtilis are the flagellar regulators s^F and s^D, respectively (Chen and Helmann, 1992).

In the course of their studies, Brutsche and Braun demonstrated that the gene immediately downstream of sigX encodes a negative regulator of sigX activity, designated rsiX. After overproduction in E. coli, they found that s^X could direct transcription in vitro from its autoregulatory promoter site, P­­_X, but that when overexpressed with the negative regulator RsiX, the resulting s^X:RsiX complex was inactive. Moreover, s^X protein fractionated with the cell membrane when overexpressed with RsiX, whereas s^X alone is a soluble protein. While their studies failed to provide a link between s^X and iron utilization, they nevertheless demonstrated the anti-s activity of RsiX and confirmed the predicted membrane localization of this regulatory factor (Brutsche and Braun, 1997).

Characterization of the s^X regulon by promoter consensus search. We reasoned that by defining the promoter selectivity of s^X we might be able to identify target genes and thereby assign a function to s^X. To define the sequence determinants for s^X-dependent recognition we took advantage of the fact that sigX, like many ECF s factors, is transcribed (in part) from an autoregulatory promoter (P_X). Using reporter fusions containing only P_X, we performed saturation mutagenesis to define those bases in the -35 and -10 regions critical for promoter function. We then searched the B. subtilis genome for similar sequences proceeding open reading frames. In total, more than a dozen candidate s^X-dependent promoters were identified and tested for activity (Huang and Helmann, 1998). Of these, two were found to be exclusively recognized by s^X in vivo (csbB, lytR) with the others exhibiting a variable level of residual transcription even in a sigX mutant strain. A current compilation of genes transcribed, in whole or in part, by the s^X holoenzyme is presented in Tables 5 and 6.

Analysis of genes associated with s^X-dependent promoters revealed that most have additional promoter sites. For example, csbB can be transcribed from either a s^X- or a s^B-dependent promoter while lytR is preceded by both s^A- and s^X-dependent sites (Huang and Helmann, 1998). Furthermore, even in a sigX mutant transcripts could still be detected emanating from the sites recognized by s^X, suggesting that other holoenzyme forms could also recognize these sequences. Because of these complexities, a sigX mutation may reduce, but is unlikely to eliminate, the expression of these target genes.

s^X controls modifications of the cell envelope. We found s^X-dependent promoters preceding several genes that affect the composition or metabolism of the cell envelope including lytR (a negative regulator of autolysin; Lazarevic et al., 1992), csbB (a membrane-bound glucosyl transferase; Akbar and Price, 1996), pbpX (a pencillin-binding protein), the dlt operon (controlling the D-alanylation of TA; Perego et al., 1995), and the pssA operon controlling phosphatidylethanolamine synthesis. In addition, s^X contributes to the expression of rapD, a response regulator aspartate phosphatase of as yet unknown function (Perego, 1998; Reizer et al., 1997). The s^X regulon overlaps with regulons controlled by (at least) s^D, s^B, and s^W (Tables 5 and 6)

            By defining the s^X regulon, we were led to a model in which s^X modifies the composition and properties of the cell envelope (Figure 4). B. subtilis has a typical Gram positive envelope containing a cytoplasmic membrane surrounded by a thick peptidoglycan (PG) layer. The cell wall is negatively charged, and functions in a manner analogous to the periplasm of Gram negative bacteria, acting to bind and concentrate proteins, small molecules, and ions near the cell (Merchante et al., 1995; Pooley et al., 1996). The activity of s^X regulates the net charge of the cell wall by controlling transcription of the dlt operon and, in a parallel pathway, may modulate the net charge of the membrane by contributing to transcription of the pssA operon.

While it has been known for many years that teichoic acids are essential, their roles in cell physiology are not entirely clear. The s^X-activated dlt operon controls the modification of teichoic acids by esterification with D-alanine (Perego et al., 1995). B. subtilis contains both membrane-associated lipoteichoic acids (LTA) and wall teichoic acids (WTA) (Fischer, 1988). Both LTA and WTA are extensively substituted by esterification with sugars and D-alanine. The latter modification introduces free amino groups (NH₃⁺) into the cell envelope, and thereby reduces the net negative charge (Perego et al., 1995).

Genetic studies indicate that dlt mutants have pleiotropic phenotypes. They often display altered patterns of autolysis (Wecke et al., 1996), perhaps due to alterations in autolysin binding to the cell wall, have alterations in adhesive properties (Clemans et al., 1999), carbohydrate metabolism (Spatafora et al., 1999), sensitivity to acid (Boyd et al., 2000), and may be affected in protein secretion and folding (Hyyrylainen et al., 2000).

Additional insight into the function of D-alanylation comes from the observation that dlt mutants in Staphylococcus aureus have a greatly increased sensitivity to cationic antimicrobial peptides (CAMPs) (Peschel et al., 1999). CAMPs are a broadly distributed family of peptides that kill bacterial cells (Hancock and Diamond, 2000; Hancock and Scott, 2000). Many are thought to act by accumulating within the cytoplasmic membrane to a critical concentration that allows the assembly of structures that permeabilize the cell. However, in other cases CAMPs also have effects on cell wall biosynthesis. For example, nisin and epidermin, two class I lantibiotics, interact with the lipid II PG synthesis intermediate (Breukink et al., 1999; Brotz et al., 1998). Inactivation of dlt also leads to altered methicillin resistance in S. aureus (Nakao et al., 2000).

The cell membrane also contains a net negative charge due to a preponderance of anionic phospholipids. However, phosphatidylethanolamine (PE), a neutral (zwitterionic) lipid, makes up as much as 50% of the membrane (Matsumoto et al., 1998). Since s^X contributes to transcription of the pssA operon (which encodes both phosphatidyl serine synthase and phosphatidylserine decarboxylase; Cao and JDH unpublished results) we predict that s^X also regulates PE levels, and thus membrane net charge. Just as the dlt gene products lead to the incorporation of NH₃⁺ groups into the cell wall, the pss operon products lead to the incorporation of NH₃⁺ groups into the membrane.

These results lead to the hypothesis that modulation of surface charge, coordinated by s^X, may function in resistance to CAMPs (Figure 4). Using disk diffusion and MIC assays, we tested strains mutant in the sigX, dlt, or pssA operons for sensitivity to a variety of CAMPs. The results, while not as dramatic as those reported for S. aureus (Peschel et al., 1999), demonstrate a 2- to 4-fold increase in sensitivity to various positively charged peptides in the sigX mutant, and in the dlt pssA double mutant strain. The dlt and pssA single mutants also had small, but reproducible, changes in sensitivity. These effects can be rationalized as a direct consequence of altered cell surface charge: D-alanylation acts to reduce the initial binding and accumulation of CAMPs near the cell membrane (Peschel et al., 1999). Similarly, reduced surface charge and altered teichoic acids are associated with nisin resistance in the ruminal bacterium Streptococcus bovis (Mantovani and Russell, 2001). By analogy, an increase in PE content in the cytoplasmic membrane might also be predicted to increase CAMP resistance. Indeed, some Listeria monocytogenes strains selected for nisin  resistance have increased PE contents in their membranes, although the genetic changes or mechanisms resistance responsible for this effect are uncharacterized (Crandall and Montville, 1998).

Induction of the s^X regulon by cell wall antibiotics. We have found that several antibiotics that target cell surface processes are strong inducers of the sigX operon and the s^X regulon (Cao and JDH, ms. submitted). The strongest inducers are inhibitors of PG biosynthesis and tunicamycin, a specific inhibitor of WTA synthesis (Pooley and Karamata, 2000). The genes for the biosynthetic enzymes controlling the synthesis of cell surface associated polymers are generally not well characterized. However, recent studies have revealed a possible link between ECF s factors and TA synthesis in the W23 strain of B. subtilis. When compared with the sequenced strain, 168, the W23 strain is found to carry a similar arrangement of TA biosynthesis genes organized into two divergent operons (the tar locus), with the additional presence of genes that specify synthesis of a ribitol-phosphate copolymer. The intergenic regulatory region for the W23 tar locus contains a 100 bp insertion relative to the sequenced B. subtilis 168 strain and this region carries an additional 2 promoter sites that resemble ECF s recognition sequences, leading to the suggestion that ECF s factors control TA biosynthesis in this strain (Minnig et al., 2001).

 s^W

The s^W regulon includes over 50 different genes activated by cell wall stress and is, to date, the most thoroughly studied ECF regulon in B. subtilis (Table 7). The sigW gene is contranscribed with an anti-s factor gene, rsiW, from a single transcription start site (Huang et al., 1998). This operon is positively autoregulated from a s^W-dependent promoter, P_W, which is similar in sequence to P_X. However, there is no crosstalk: P_W is dependent on s^W in vivo and is not well recognized by s^X in vitro and, conversely, P_X is dependent on s^X in vivo and is not well recognized by s^W in in vitro transcription reactions (Huang et al., 1998).

Overlap between the s^X and s^W regulons. As discussed above, we initially sought to define the s^X regulon by searching the B. subtilis genome for sequences similar to the autoregulatory site, P_X. In parallel with these studies we also began an investigation of a second ECF s factor, s^W. Serendipitously, these two lines of investigations were complementary: several of the promoters originally identified as candidate s^X-dependent sites turned out to also be targets for s^W. The first clue to this overlap came from primer extension mapping of transcripts corresponding to putative s^X promoters using RNA preparations from various strains. For example, the transcripts corresponding to the ywbN and yrhH genes were easily detected in the sigX mutant sample, but not in wild-type or in a sigX sigW double mutant (Huang and Helmann, 1998). This suggested that these genes might be transcribed primarily by s^W, and further suggested that the s^W regulon might be expressed at a higher level in sigX mutant cells. Support for this model emerged when we purified the s^W protein and demonstrated that in vitro both s^X and s^W could program RNAP to recognize the ywbN and yrhH promoter elements.

The overlap between the s^X and s^W regulons appears to result from the similar but non-identical sequence recognition properties of the corresponding holoenzymes. As illustrated in Table 8, this difference can ultimately be traced to the -10 element. Based on a comparison of promoter sites exclusively recognized by s^X (including the P_X autoregulatory site), exclusively recognized by s^W (including the P_W autoregulatory site), and those recognized by both holoenzymes, we proposed a simple model for sequence discrimination (Huang et al., 1998). According to this model, s^X recognizes -10 elements with sequence CGaC, s^W recognizes CGTa, and both can recognize CGTC (lower case reflects a non-critical base for recognition). To test this model, we mutated P_X and P_W and tested the effects on in vitro recognition by the s^X and s^W holoenzymes. Thus, we converted the s^X-dependent site P_X (CGAC) into a s^W-dependent promoter by mutation of two bases in the -10 element (CGTA). Conversely, we changed P_W (CGTA) into a promoter recognized by either s^X or s^W with one base change (CGTC), and into an exclusively s^X-dependent promoter with two base changes (CGAC) (Qiu and Helmann, 2001).

Despite the overlap between the s^X and s^W regulons these two systems differ in several respects. First, s^X regulated genes are usually turned on during late-logarithmic phase while s^W-dependent genes are not activated until early stationary phase under laboratory growth conditions (Huang et al., 1997; 1998). Second, these two regulons respond to distinct but overlapping sets of chemical signals. For example, both regulons are induced by antibiotics active on the cell wall, but with differing efficiencies: vancomycin strongly induced the s^W regulon while tunicamycin selectively induces the s^X regulon (M. Cao and JDH, ms. submitted). Other cell active antibiotics induce both regulons. The s^W regulon is also strongly induced by alkali stress (Wiegert et al., 2001), whereas the s^X regulon is not. Finally, in a genetic analysis of transposon mutations that up-regulated either the s^X or the s^W regulon all but one of the identified mutations activated only one of the two regulons (Turner and Helmann, 2000).

Defining the s^W regulon: (a) promoter consensus searches.  As a first approach to define the s^W regulon, we searched the genome for sites similar to the autoregulatory site, P_W. We were startled to find 16 perfect matches in positions suitable to function as promoter elements. Even more remarkable, all 16 promoters are largely, if not exclusively, s^W-dependent in vivo (Huang et al., 1999). Thus, unlike s^X, s^W appears to be required for the expression of its target genes. From this initial study we concluded that s^W controls a large regulon of least 35 genes. By relaxing our search criteria to accommodate a one bp alteration in the spacer length we identified 5 additional s^W promoters including one preceding the fosB fosfomycin resistance gene (Cao et al., 2001). While the consensus search approach was quite successfully in identifying genes controlled by s^W, this approach has several limitations (Table 2). Most obvious, any promoters that differ from the arbitrarily defined consensus sequence will be missed. One example is the pspA(ydjF) gene (Wiegert et al., 2001), which is s^W-dependent but differs in what was presumed to be an invariant base in the -10 region (Table 7).

Defining the s^W regulon: (b) transcriptional profiling. To complement the consensus search approach, we also defined the s^W regulon by transcriptional profiling. A comparison of total RNA from wild-type and sigW mutant cells confirms the s^W-dependence of many known targets and suggests several additional likely targets (M. Cao et al., ms. submitted). One limitation of this approach is that the level of expresion of s^W-dependent genes is quite low in the wild-type cells, so some operons are likely to be missed.

In general, transcriptional profiling experiments provide a much more powerful tool to defining regulons if strong inducing conditions can be identified. The finding that the s^W regulon is induced when cells are shifted to alkaline pH provides one such condition (Wiegert et al., 2001). The alkaline shock stimulon includes 49 genes (of ~80 total) whose up-regulation is dependent on s^W, either directly or indirectly. Significantly, we have observed that s^W-dependent genes are the most strongly induced members of the vancomycin stimulon and this regulation requires the RsiW anti-s factor (M. Cao et al., ms. submitted). Taken together, the consensus search and transcriptional profiling studies identify ~50 genes under s^W control.

Defining the s^W regulon: (c) ROMA: run-off transcription - macroarray analysis. One difficulty with transcriptional profiling studies is that it is difficult to distinguish direct from indirect effects. To address this issue, Min Cao developed a complementary in vitro technique to identify that subset of genes that are directly dependent on s^W for their expression (Cao and JDH, ms. submitted). In this experiment, total genomic DNA is fragmented with restriction enzymes and then used as a template for in vitro transcription using core RNA polymerase with and without a large molar excess of s^W. The resulting radiolabeled RNA populations are then hybridized to cDNA macroarrays (Sigma-GenoSys) to identify those genes proximal to promoters active in vitro. In the case of s^W, 44 strong signals are produced in response to the addition of s^W and at least 22 of these correspond to promoters active in vivo (Cao and JDH, ms. submitted). The rest may result from a relaxed specificity of the holoenzyme under these in vitro conditions.

While the ROMA approach is technically challenging, and requires access to purified s and RNA polymerase as well as cDNA arrays, it provides a very useful complement to conventional transcriptional profiling. The presence of a signal in the ROMA experiment suggests that any effects seen in an in vivo experiment are likely to be direct, rather than indirect effects. Moreover, because of low levels of expression of many ECF regulons in vivo, particularly if inducing conditions are not known, in vivo transcriptional profiling experiments may fail to detect target genes (e.g. Manganelli et al., 2001a). This problem is further compounded by the presence of additional promoter sequences and/or overlapping recognition among ECF s regulons.

s^W controls an "antibiosis" regulon. Our studies suggest that the s^W regulon functions in both the synthesis of, and the defense against, antimicrobial compounds (Cao et al., 2001; Huang et al., 1999). Hence, we refer to the s^W regulon as an "antibiosis regulon." Similarly, the cell envelope modifications orchestrated by s^X may also be an adaptive response to the presence of antimicrobial agents. Historically, the term "antibiosis" was coined to describe the ability of one organism to inhibit the growth of another. Ultimately, this term gave rise to the now much more familiar term antibiotic used to refer to the chemicals mediating the growth inhibition.

The role of s^W in controlling an antibiosis regulon is supported by two key observations: first, this regulon is strongly induced by antibiotics inhibiting cell wall biosynthesis and second, many of the gene products controlled by s^W have known or putative roles in detoxification or antibiotic synthesis. In light of this conclusion, it is interesting to consider the observation of Wiegert et al. (2001) that the s^W regulon accounts for a large portion of the alkali shock stimulon. We suggest that under the stress conditions used (a shift to pH 8.9) the growth-limiting event was an inability of the cell to synthesize cell wall. Since cytoplasmic pH is narrowly regulated over a range of extracellular pH conditions, it makes sense that the first essential enzymes to become inhibited at high pH would be those that function outside the cytoplasmic membrane and these are likely to be those involved in cell wall synthesis. Thus, it is probably not a coinicidence that alkali stress and vancomycin stress are inducing the same target genes. Moreover, it is important to note that resistance to alkali shock itself is not likely to be the defining physiological role for s^W since a sigW mutant strain is no more sensitive to alkali shock than wild-type cells, and none of the target genes of the s^W regulon have an obvious connection with pH homeostasis.

While most genes controlled by sW are of uncertain function, in several cases functional predictions can be made. For example, s^W controls the transcription of PbpE (a low molecular weight penicillin binding protein), the FosB fosfomycin resistance enzyme (Cao et al., 2001), and several enzymes with possible functions in detoxification. These include a bromoperoxidase and an uncharacterized epoxide hydrolase (Huang et al., 1999). In addition, s^W directs the expression of several small, hydrophobic peptides that resemble bacteriocin precursors (Jack et al., 1995) and an ABC transporter with similarity to bacteriocin export systems (Quentin et al., 1999). Indeed, we have shown that sigW mutants display decreased expression of one or more bacteriocins (A. Gaballa and JDH, unpublished results).

The s^W regulon also includes two genes encoding signal-peptide peptidase homologs (sppA and yqeZ). The SppA family includes membrane bound peptidases that have been proposed to function in the turnover of signal peptides left in the membrane by the action of leader peptidase (Suzuki et al., 1987). However, sppA mutations do not greatly affect secretion, although some effects were noted (Bolhuis et al., 1999), and the regulation of these two homologs by s^W suggests another possible role. We suggest that SppA and YqeZ may function to cleave bacteriocins and thereby prevent their accumulation within the membrane to toxic levels. This could be either a defense mechanism or an immunity mechanism. In support of this idea, an operon encoding an SppA homolog has been implicated in immunity to enterocin A, a bacteriocin from Enterococcus faecalis (O'Keeffe et al., 1999).

Transition state regulators.  One group of proteins likely to affect the activity of s^X and s^W are the transition state regulators: AbrB, Abh, and Spo0A (Strauch and Hoch, 1993). Bacilli produce numerous antibiotics and Spo0A, the key regulator of sporulation, is required for their synthesis (Marahiel et al., 1993; Schaeffer, 1969). However, this Spo0A effect can be bypassed by mutations in abrB. This is explained by the observation that activation of Spo0A leads to repression of the AbrB repressor and thereby leads to derepression of antibiotic synthesis.

Several lines of evidence suggest that AbrB, and other transition state regulators, also affect the s^W and s^X regulons (Figure 5). For example, AbrB represses the s^W-dependent pbpE gene (Strauch, 1995), and has recently been found to repress both the sigW operon and several other s^W target genes (Qian et al., 2001). In addition, Abh (an AbrB homolog) is transcribed by s^X (Huang and Helmann, 1998). The relationships between these transition state regulators, and the regulons controlled by s^X and s^W are currently under investigation.

 s^M

            The sigM ECF s factor was identified as a result of the B. subtilis genome sequencing project and the corresponding mutant attracted attention when it was found to have an apparent defect in spore outgrowth. Horsburgh and Moir (1999) have demonstrated that sigM is optimally expressed in early logarithmic phase cells from two promoter sites: one under s^A control, and the second an autoregulatory site recognized by the s^M holoenzyme. Expression of s^M is upregulated ~2-fold by growth in high salt, and the sigM mutant strain fails to grow in medium containing high levels of salt. This growth defect, and the consequent formation of swollen and abnormally shaped cells, may be due to defects in cell wall biosynthesis, but biochemical analysis did not reveal gross differences in peptidoglycan structure in the mutant strain (Horsburgh and Moir, 1999).

            The activity of s^M, like many other ECF s factors, is negatively regulated by two downstream genes. Using the pMUTIN plasmid vector, it was possible to regulate the level of expression of the downstream yhdL and yhdK genes and demonstrate that decreased expression of these genes leads to increased activity of s^M. However, these genes cannot be disrupted, suggesting that the resulting up-regulation of the s^M regulon impairs viability (Horsburgh and Moir, 1999).

            In ongoing studies, the Moir and Hecker laboratories have used cDNA macroarray and lac fusion analyses to identify genes that are up-regulated in response to the induction of s^M synthesis (Thackray et al., 2001). A preliminary assessement of the s^M regulon reveals induction of the yacK and yacL genes (which may also be transcribed as part of the heat inducible ctsR operon), radC(ysxA), ydcF, ypbG, yjbC, yjbD, and ywoA. Interestingly, some of these genes are annotated as having a likely role in DNA repair. This is intriguing since it has been noted that dessication, which is related to osmotic stress, may lead to DNA damage. Indeed, the extreme resilience of Deinococcus radiodurans against DNA damage is hypothesized to have evolved as a defense against dessication stress (Mattimore and Battista, 1996).

            Many of the newly identified members of the s^M regulon are associated with candidate promoter elements resembling the well characterized sigM autoregulatory site. However, the sigM autoregulatory site has a distinctive -10 sequence, CGTG, not shared by most of these target operons. Most s^M target genes have candidate -10 elements with sequence CGTC, similar to that noted above as being potentially recognized by either s^X or s^W. Indeed, two of the identified s^M targets (yjbC and ywoA) are also known to be recognized by s^W (Table 6).

Other ECF s factors

            The B. subtilis genome encodes four other ECF s factors: s^V, s^Y, s^Z, and s^ylaC.  Although mutant strains have been generated for each of these factors, few clues have yet emerged to their functions. At least for three of these s factors, candidate autoregulatory promoters can be identified upstream of the s factor operon. These sites have the characteristic ECF s factor -35 element, with the conserved "AAC" motif, followed by a candidate -10 element similar to that noted above for other ECF s factors in B. subtilis. Remarkably, all of these sequences are very similar to each other. This raises several important questions: (1) are these in fact autoregulatory sites?, (2) what is the promoter specificity for each s and how are they distinct?, (3) to what extent do the ECF s factor regulons overlap? and (4) do any of the ECF s factors regulate each other? Results from cDNA microarray studies support the idea that these s factors are all autoregulated and have led to lists of candidate target genes (Fujita, personal communication). As noted above for the s^M regulon, at least some of these target genes have been previously identified as targets of other ECF s factors. If, as this suggests, these s factors overlap in their promoter selectivity, why do they not contribute to gene expression of the promoters we have studied to date? The answer may be that most ECF s factors are synthesized as part of a “two component” regulatory system: the ECF s and the cognate anti-s. Under most conditions, these systems are essentially inactive. Only upon receiving the proper stimulus will s be released and become active. Much additional work will be required to understand the sequence differences that determine which genes are subject to control by which s factor and to define the extent of regulon overlap among these seven ECF s factors. 

Read Also

Streptomyces coelicolor

            Unraveling the complexities of the many ECF s factor regulons in S. coelicolor is a truly daunting task. To date, most studies have focused on just three of the ~50 ECF s factors in this organism: s^E, s^R, and s^BldN (Paget et al., 2002). The s^E regulon includes an operon involved in cell wall biosynthesis and mutants display an increased sensitivity to cell wall perturbants. The s^R regulon responds to oxidative stress conditions that lead to the formation of disulfide bonds in the cytoplasm (more accurately referred to as disulfide stress). The s^BldN regulator participates in the sporulation pathway and mutants in this s are defective for the formation of aerial hyphae (bald phenotype). Analysis of these systems has been very productive: s^E is unusual in that its expression is activated by a two-component regulatory system, studies of s^R led to the discovery of a family of zinc-containing anti-s factors (the ZAS family), while the s^BldN system provides us with the first example of an ECF s factor regulated by proteolytic processing from an inactive precursor.

 s^E

            The biochemical activity referred to as s^E was first detected when RNA polymerase fractions were analyzed for the ability to recognize several promoter sites upstream of the dagA agarase gene (Buttner et al., 1988). The fraction that activated the P2 promoter was found to contain a 20 kD protein designated s^E. As noted above, the cloning of the sigE gene was a key event in the original discovery of the ECF family of regulators (Lonetto et al., 1994).

            Isolation of the sigE gene allowed the construction of sigE null mutant strains. Strains lacking s^E activity have several dramatic phenotypes including the overproduction of the blue-pigmented antibiotic actinorhodin, altered colony morphology, and poor sporulation. In addition, the mutants display an increased sensitivity to cell wall hydrolytic enzymes including both muramidases that cleave the glycan backbone (such as lysozyme) and amidases that cleave the peptide cross-links. These phenotypes can all be suppressed by millimolar levels of Mg(II), which is known to stabilize the cell envelope (Paget et al., 1999b).

            The sigE gene is part of a four gene operon that also encodes a predicted membrane protein (CseA), a response regulator (CseB), and a membrane-bound histidine protein kinase (CseC) (Paget et al., 1999a) (Figure 6). Expression of this operon does not appear to be regulated by s^E: instead, the major sigE operon promoter shows similarities to the promoter for whiG, which is under the control of an as yet uncharacterized form of holoenzyme. Expression of sigE requires activation by the CseB response regulator. Since this operon is expressed in culture, the signals perceived by the CseC sensor kinase are presumably present in the laboratory growth conditions used. The CseA protein appears to play a negative role in modulating expression of the sigE operon: an in-frame cseA deletion mutant has increased expression of the sigE operon. While the biochemical activity of CseA is not yet known, it could affect the CseC-CseB two component system that is required for activation of sigE operon transcription (Paget et al., 2002).

            The s^E regulon has not yet been extensively characterized. To date, two promoter sites recognized by s^E both in vitro and in vivo have been documented. The first, hrdDp1 contributes to the expression of the group 2 s factor, HrdD. However, since the role of HrdD is as yet unknown (Paget et al., 2002), the significance of this regulation is unclear. The second target operon includes 12 genes thought to specify cell wall glycan synthesis and has been designated cwg. While the precise role of the s^E regulon has yet to be determined, an important clue comes from the recent observation that the cwg genes are induced by vancomycin (Paget et al., 2002). Thus, the s^E regulon may function to coordinate responses to cell wall stress, much as described above for the s^X and s^W regulons in B. subtilis. The relationship between cwg operon expression and the phenotypes of a sigE mutant are currently under investigation (Paget et al., 2002). Unexpectedly, the sigE mutant is unaffected in utilization of the dagA P2 promoter in vivo, despite the fact that it was transcription from this site that led to the original isolation of s^E protein. This suggests that other s factors can also recognize this site in addition to, or even instead of, s^E.

            While a sigE null mutant of S. coelicolor overproduces actinorhodin, in S. antibioticus s^E plays a positive regulatory role in antibiotic biosynthesis (Jones et al., 1997). Sequencing of the gene for phenoxazinone synthase, which catalyzes the penultimate step in actinomycin biosynthesis, revealed a promoter sequence closely resembling known target promoters for s^E. Purified s^E holoenzyme indeed recognizes this site in vitro, but, curiously, is dispensable in vivo. Nevertheless, a sigE mutant fails to produce actinomycin. These results suggest that s^E does play an essential positive regulatory role in actinomycin synthesis, but this role must extend beyond recognition of the phs promoter region. In fact, the sigE null mutant has greatly diminished activity of the key biosynthetic enzyme actinomycin synthase I. Unlike the situation in S. coelicolor, the S. antibioticus sigE null mutant was unaffected in colony size or development. Nevertheless, the identity and arrangement of the neighboring genes suggests that these are in fact orthologous s factors.

s^R

            Biochemical fractionation of purified RNA polymerase allowed Kang et al. (1997) to identify a 31 kDa s that was enriched in stationary phase cells. When these fractions were incubated with various purified DNA templates, this s was found to activate expression from a second promoter preceeding the hrdD gene, hrdDP2.

            To begin to address the physiological role of s^R, Paget et al. (1998) determined the amino-terminal sequence of the purified protein and designed primers to identify the corresponding gene. By constructing a null mutant, they were able to demonstrate that s^R plays a key role in regulation of oxidative stress responses: the mutant strain is sensitive to both superoxide generators and to diamide, a chemical that oxidizes thiols in the cell cytoplasm leading to the formation of disulfides. Disulfide bonds are normally not present in proteins in the reducing environment of the cytoplasm and their formation can inhibit enzyme activity (Aslund and Beckwith, 1999). The resulting type of oxidative stress has been named "disulfide stress."

            In response to oxidizing conditions, s^R activates the transcription of the trxBA operon encoding both the thiol reductant thioredoxin and thioredoxin reductase (Paget et al., 1998). Together, these proteins allow reducing equivalents, in the form of NADPH, to be efficiently used to reduce oxidized thiols (Carmel-Harel and Storz, 2000). In addition to the trxBA operon, s^R also activates its own synthesis from one of two promoters that preceed the sigR operon. These observations lead to a simple model (Figure 7) in which disulfide stress activates s^R-dependent transcription to restore the intracellular redox balance (Paget et al., 1998).

            The sigR operon contains two genes: sigR and rsrA (regulator of sigR). As noted for other ECF s factors, the second gene in the operon encodes a negative regulator functioning as a specific anti-s factor. However, RsrA is unusual in that it is a soluble, rather than a membrane-localized, protein. In a series of elegant collaborative experiments, the Roe and Buttner laboratories have established that RsrA is a small, zinc-containing protein that directly serves as the sensor of disulfide stress (Kang et al., 1999; Paget et al., 2001a). In its reduced form, RsrA forms a 1:1 complex with s^R and prevents transcription initiation. When exposed to a thiol oxidant, such as diamide, at least one disulfide bond is formed in RsrA and the inhibition of s^R is relieved. The bound Zn(II) ion is lost upon RsrA oxidation, suggesting that at least one of the cysteine residues that is oxidized also functions to coordinate Zn(II) ion. Altogether, RsrA contains 7 Cys residues, but only three are required for activity. These appear to function as Zn(II) ligands and/or components of the thiol-disulfide redox switch.

            To further define the s^R regulon, Paget et al. used a promoter consensus search strategy (Paget et al., 2001b). By searching the S. coelicolor genome for similarities to the sequence GGAAT 18 bp GTT they identified 34 candidate promoters located upstream of target genes. Remarkably, 30 of these are functional sites that are induced by diamide in a s^R-dependent manner (Table 9). Approximately one-half of the s^R target genes are also transcribed from an additional, s^R-independent promoter site. In addition, many of the s^R promoters are still active even in a sigR mutant strain suggesting that another s factor, presumably a member of the ECF sub-family, can also recognize these sites. The differences between those promoter sites recognized exclusively by s^R (class A) and those with residual transcription (class B) are not entirely clear, but a correlation with the sequence of the -10 region is apparent: class A sites usually contain the sequence GGTT while most class B sites contain CGTT (Table 9).

            While these genes are unlikely to represent the entire s^R regulon, they nevertheless provide a very informative overview of the types of functions that comprise this regulon. As expected, several of the regulated operons participate in thiol metabolism including genes likely to be involved in either cysteine biosynthesis or synthesis of the low molecular weight thiol mycothiol (functionally analogous to glutathione in many other bacteria; Newton et al., 1996). Other genes in the s^R regulon are likely to be involved in modulating translation during disulfide stress (Paget et al., 2001b). These include rpmE, relA, and ssrA. The rpmE gene encodes ribosomal protein L31 which contains a Cys-x-x-Cys motif. This leads to the speculation that this protein serves to sense redox stress and slow or pause translation until redox balance can be restored. Similarly, induction of RelA will lead to the synthesis of ppGpp which acts as a global modulator of growth rate. Finally, ssrA encodes a stable RNA that functions to rescue ribosomes stalled due to the lack of a termination codon (for example, at the 3'-end of incomplete mRNAs; reviewed in Karzai et al., 2000). In addition to modulating translational capacity, induction of the s^R regulon may also affect transcription: another s^R target encodes a small RNA polymerase associated protein (RapA) of as yet unknown function. As discussed in more detail below, s^R is likely to be the ortholog of Mycobacterium tuberculosis s^H which also controls a large, diamide-inducible oxidative stress response (Manganelli et al., 2001a). It will be interesting to learn how the regulons controlled by these functionally similar systems are related, and how they might differ.

 s^BldN

            The gene encoding s^BldN was discovered in a screen for mutations that block morphological differentiation (Bibb et al., 2000). Originally classified as a white mutant (whiN), these mutations blocked the formation of spores, but not aerial hyphae. In contrast, subsequent studies of null mutations in this locus demonstrated a failure to form aerial hyphae: thus, whiN was reclassified as a bald mutant, bldN.

            When the bldN gene was sequenced it was found to encode a member of the ECF s family (Bibb et al., 2000). Subsequent work has demonstrated that bldN is expressed during development and this regulation involves, at least in part, regulation by a DNA-binding protein (BldD) that represses transcription during growth (Elliot et al., 2001). Intriguingly, BldD also represses expression of another s factor required for development, s^WhiG.

            The downstream targets of s^BldN action are not yet well characterized. One promoter that depends on s^BldN is the bldMp1 site that contributes to expression of the BldM response regulator. Several upstream regulators have also been defined: the expression of bldN depends on several other genetically defined bld loci, but their roles in regulating bldN are not yet well defined.

            One of the most intriguing features of s^BldN is the presence of a large (86 amino acid) N-terminal extension. It is proposed that s^BldN, like some of the late s factors regulating sporulation in B. subtilis (Kroos et al., 1999), is synthesized as an inactive pro-protein that must be proteolytically processed to become active (Paget et al., 2002). The factors and signals that control the processing of s^BldN have yet to be elucidated.

            The s^BldN ortholog in S. griseus (s^AdsA) is also implicated in sporulation (Yamazaki et al., 2000). However, in this case we have a somewhat clearer picture of some of the upstream regulatory factors. In this organism expression of aerial hyphae and the antibiotic streptomycin is regulated in response to a g-butyrolactone signaling molecule known as A-factor (Horinouchi, 1999). A specific A-factor-dependent transcriptional activator (AdpA) coordinates the responses of the cell to this global regulator. One of the direct targets for A-factor action is the bldN ortholog adsA gene (AdpA-dependent sigma factor), while another target is a gene (strR) that functions as an activator of the streptomycin biosynthesis gene cluster (Yamazaki et al., 2000).

Other ECF sigma factors

            Deconvoluting the myriad ECF-s factor regulons in S. coelicolor is likely to keep many researchers busy for a long time to come. In addition to the three regulons described above, analyses of only two other ECF s factors have been reported.

            The first, s^U, was identified during a screen for mutations affecting development (Gehring et al., 2001). However, unlike bldN, in this case the developmentally affected mutant had a transposon insertion in a putative anti-s gene, rsuA, and the developmental phenotype was a result of up-regulated activity of an ECF s factor, rather than loss of activity. Indeed, a sigU mutant is able to differentiate normally, suggesting that the developmental phenotype in this case is a consequence of its uncontrolled activity, perhaps by competing for core enzyme with other s factors required for development.

            The second, s^T, is associated with a downstream gene that encodes a putative anti-s factor similar to RsrA, but likely to be membrane associated (J-H. Roe, personal communication). Like s^R, the s^T regulon functions in defending the cell against oxidative stress: sigT mutants are also sensitive to diamide. This suggests that both of these regulons include functions that are essential for optimal resistance to diamide and a failure to express either one can lead to sensitivity. Thus, even though there may be regulon overlap, these two regulons are not redundant.

 Mycobacterium tuberculosis

            The reemergence of tuberculosis as a clinically important disease during the last couple of decades has led to a resurgence of interest in both M. tuberculosis and its more rapidly growing relative, M. smegmatis. Genome sequencing reveals 10 ECF s factors encoded in the M. tuberculosis genome (Cole et al., 1998).

            Using molecular beacons and real-time PCR as a tool to quantify mRNA levels, Manganelli et al. (1999) reported a survey of the expression of 10 s factor genes, including 7 members of the ECF sub-family, in response to various environmental stresses. During growth sigC mRNA is even more abundant than that for the primary (group 1) s factor gene, sigA. The mRNAs for sigD, sigE, and sigM are also fairly abundant, while sigF, sigH, and sigI messages are present at only low levels. The sigG mRNA could not be detected under these growth conditions. Both sigE and sigH were found to be induced by heat shock, and sigE was additionally induced by the detergent, sodium dodecyl sulfate (SDS). Both of these s factors are also induced during phagocytosis by macrophages, suggesting that these stress responses may be important during infection (Graham and Clark-Curtiss, 1999; Jensen-Cain and Quinn, 2001). Most studies to date have focused on these two ECF s factors.

 s^E

            The first ECF s factor to attract experimental scrutiny in M. tuberculosis was s^E. This gene was originally identified as part of the M. leprae genome sequencing project and the corresponding gene was amplified from several Mycobacterial species (Wu et al., 1997). The encoded s^E proteins are >90% identical in M. tuberculosis, M. smegmatis, and M. avium. Construction of a sigE null mutant in M. smegmatis revealed an increased sensitivity to oxidative stress, heat shock, low pH, and detergent (SDS) stress (Wu et al., 1997). Moreover, it was found that wild-type cells display an adaptive response to hydrogen peroxide that is lacking in a sigE mutant strain.

            Studies of the role of s^E in M. tuberculosis have established that it is also important in resistance to both heat, oxidative, and detergent stresses, and the mutant strain is more sensitive to killing by macrophages (Manganelli et al., 2001b).  When the expression of various s factors was analyzed in the sigE mutant strain it was found that sigE mRNA was itself slightly elevated, suggesting that this gene may not be autoregulated. In addition, the level of mRNA for the group 2 s factor, s^B, was reduced about 10-fold.

            To further define the s^E regulon, cDNA microarray studies were performed to compare the mRNA populations of wild-type and sigE mutant strains both during exponential growth and after imposition of detergent stress. 38 genes were expressed at levels at least 2-fold lower in the sigE mutant than in the wild-type strain, including sigB. Of 62 genes induced by SDS in wild-type, 23 were not significantly induced in the sigE mutant strain. Of these 23, at least 10 operons (13 genes) are associated with putative promoter sites resembling other sites known to be recognized by ECF s factors (Table 10).

 s^H

            The sigH gene was originally identified as a paralog of sigE by searching the M. tuberculosis genome sequence (Fernandes et al., 1999). This gene is conserved in the faster growing species, M. smegmatis, and encodes a protein with 89% identity to M. tuberculosis s^H. Analysis of sigH transcription by primer extension start site mapping identifies two promoters that are active under heat shock conditions (Fernandes et al., 1999), consistent with the results of RT-PCR measurements of RNA levels in stressed cells of M. tuberculosis (Manganelli et al., 1999). Construction of a sigH mutant in M. smegmatis revealed no significant difference from wild-type in the survival of a variety of stress conditions including heat shock, cold shock, acid stress, and hydrogen peroxide treatment. However, the sigH mutant is significantly more sensitive to organic hydroperoxides. Reasoning that s^H and s^E might be partially redundant in function, Fernandez et al. (1999) created a sigH sigE double mutant. This strain was more sensitive to heat shock and organic peroxide stress than either single mutant. 

            In the closely related species M. tuberculosis a sigH null mutant is more sensitive than wild-type to a variety of stresses including heat shock, hydrogen peroxide, organic peroxide, and diamide, but not to superoxide generators. However, unlike sigE (Manganelli et al., 2001b), disruption of sigH does not adversely affect the ability of M. tuberculosis to survive and multiply inside macrophages (Manganelli et al., 2001a).

            Since s^H is the closest homolog of S. coelicolor s^R it seems likely that these two s factors may be functional orthologs. To test this hypothesis microarray analyses were used to measure global mRNA profiles of wild-type and sigH null mutant strains after exposure to the thiol oxidizing agent diamide (Manganelli et al., 2001a). Of the 48 strongly induced genes in the wild-type, 39 were no longer induced in the sigH null mutant. Therefore, s^H is a key regulator of the diamide (disulfide stress) stimulon in M. tuberculosis, as it is in S. coelicolor. In contrast, comparison of the mRNA profiles in non-stressed cells failed to reveal a significant effect of the sigH mutation, indicating that this regulon is not induced during logarithmic growth of non-stressed cells (Manganelli et al., 2001a).

            The s^H regulon includes at least two heat shock proteins (Hsp and ClpB) and several transcription factors including s^B and s^E as well as s^H itself. Interestingly, the sigB promoter recognized by s^H is the same as that controlled by s^E, demonstrating that these two s factors have overlapping promoter selectivity. In addition, s^H controls an operon encoding thioredoxin and thioredoxin reductase as well as a set of genes involved in cysteine biosynthesis (Manganelli et al., 2001a). Control of these latter functions by s^H is consistent with their control in S. coelicolor by s^R and further emphasizes that these are orthologous systems.

            Inspection of the promoter regions for the 27 putative transcription units induced by diamide stress in a s^H-dependent manner reveals candidate ECF-type promoter elements for 14 operons (Table 11). At least one of these sites (sigB) has been validated experimentally. Thus, the s^H regulon likely includes at these 14 operons as direct targets and a roughly equal number that are either controlled by promoters not easily identified by sequence inspection or which are indirect targets for s^H (Manganelli et al., 2001a). It is intriguing that the s^E- and s^H-dependent promoters described to date (Tables 10 and 11) appear to be very similar in sequence. The basis for promoter discrimination between these two s factors is not yet clear.

Pseudomonas  aeruginosa

            ECF s factors have been extensively studied in Pseudomonas spp. where they participate in various iron uptake pathways, alginate secretion, and the expression of virulence factors. The best characterized systems are the P. aeruginosa s^E (also known as AlgU or AlgT) regulator of alginate biosynthesis (Hughes and Mathee, 1998) and s^PvdS, a regulator of pyoverdine siderophore biosynthesis (Vasil and Ochsner, 1999).

 s^E

            In patients with cystic fibrosis the chronic colonization of the lungs by P. aeruginosa is a major factor in the progression, and ultimate lethality, of the disease. Pathogenic isolates of P. aeruginosa from cystic fibrosis patients are typically mucoid in appearance and produce an abundant exopolysaccharide known as alginate. Genetic investigations of the molecular changes leading to mucoidy identified several genetic loci (muc genes) that, in retrospect, are negative regulators of the s^E factor that activates transcription of the alginate biosynthesis genes. The gene encoding s^E, algU(algT), is cotranscribed with the mucABCD genes. The MucA protein functions directly as an anti-s factor for s^E while the MucB protein is located in the periplasm (Hughes and Mathee, 1998). This system is analogous to the regulation of s^E in E. coli by the RseA and RseB proteins. Indeed, expression of E. coli s^E in P. aeruginosa can complement an algU mutation and restore mucoidy (Yu et al., 1995). Alginate is also synthesized by a variety of other Pseudomonads, including the plant pathogen P. syringae (Keith and Bender, 1999). For a detailed review of this system, see Hughes and Mathee (1998).

s^PvdS and its relatives

            Pseudomonads contain a family of ECF s factors that are all regulated by the Fur protein in response to iron-limitation and activate the expression of siderophore biosynthesis and/or transport genes (Leoni et al., 2000). The best characterized member of this family of regulators is s^PvdS, which is required for the synthesis of the siderophore pyoverdine (reviewed in Vasil and Ochsner, 1999). The s^PvdS regulator is highly similar to s^PbrA (89% identity) and s^PfrI (85% identity) which are required to activate pseudobactin biosynthesis in P. fluorescens and P. putida, respectively (Leoni et al., 2000). A scluster of more distantly related Fur-regulated ECF s factors includes s^PupI and E. coli s^FecI (Leoni et al., 2000), both of which direct expression of siderophore receptors and appear to require activation by their cognate membrane bound regulator (Koster et al., 1993).

            The s^PvdS regulon is likely to include at least six target promoters including genes required for pyoverdine synthesis, a positive activator of toxin production, and an extracellular protease (Ochsner et al., 1996; Wilson et al., 2001). The coordinate regulation of these genes by the s^PvdS holoenzyme accounts for the observation that these genes all have a conserved DNA sequence motif in their regulatory regions (Rombel et al., 1995). The s^PvdS protein has been purified and shown to program recognition of its cognate target sites in vitro (Wilson and Lamont, 2000).

            Although P. aeruginosa encodes at least 8 ECF s factors only two have studied in detail. However, it is worth noting that cycle selection experiments using purified Fur protein identified candidate Fur-binding sequences proximal to three additional genes encoding ECF s factors (Ochsner and Vasil, 1996). It remains to be determined whether these ECF s factors are involved in the regulation of iron uptake functions. However, at least one of these s factors (s^FiuI; Genbank entry AF051691) is located adjacent to a putative siderophore receptor gene consistent with the idea that it may regulate iron transport.

            The P. fluorescens siderophore, pseudobactin, is transcriptionally regulated by a close homolog of s^PvdS designated s^PbrA (Sexton et al., 1995). Like s^PvdS, the expression of s^PbrA is regulated by Fur in response to iron levels (Sexton et al., 1996). Interestingly, the expression of both siderophore and its receptor is reduced in mutants in siderophore biosynthesis, suggesting that pseudobactin (like ferric-citrate in E. coli) may have a role in inducing expression of the corresponding uptake operon. However, this response does not require s^PbrA but is instead mediated an unidentified factor (Callanan et al., 1996). An ECF s factor, s^PrtI, has also been implicated in production of protease in a P. fluorescens (Burger et al., 2000).

ECF s factors in other organisms

            ECF s factors are widely distributed among bacteria and offer a convenient mechanism for coordinating gene expression with extracellular signals. In general, bacteria with greater metabolic or developmental complexity tend to have larger genomes compared to highly specialized organisms that may have undergone reductive genome evolution (Table 12). There is a clear correlation between genome size and the proportion of the genome devoted to regulatory functions: in bacterial genomes of 0.5 to 2 Mb in size <3% of the open reading frames encode likely regulatory proteins while this number increases to >8% in the 6.3 Mb P. aeruginosa genome (Stover et al., 2000). Similarly, Caulobacter crescentus, which has a complex dimorphic lifestyle, encodes an estimated 100 or more two-component signal transduction proteins as well as 13 ECF s factors in its 4 Mb genome (Nierman et al., 2001). The expansion, presumably by gene duplication and divergence, of the ECF s factors in many of the more complex bacterial genomes apparently occurs in preference to other classes of s factors. Whereas many simple bacterial genomes encode few s factors, and may not encode any ECF s factors, in the largest genomes sequenced to date the majority of the s factors are of the ECF class.

            Inspection of the phylogenetic relatedness among various ECF s factors reveals several clusters of proteins with related functions (Figure 8). For example, there is a cluster of proteins that regulate iron transport conserved in various Pseudomonads, there is a close relationship between the disulfide stress s factors in S. coelicolor and M. tuberculosis, and there are many homologs of E. coli s^E that may regulate periplasmic or heat stress responses similar to that controlled by s^E. This correlation of sequence with function is reminiscent of that observed among group 3 s factors (Lonetto et al., 1992) and suggests that alternative s factors of the ECF sub-family arose early in many of these lineages.