The Bacillus subtilis extracytoplasmic function (ECF) σM factor is activated by cell envelope stress elicited by antibiotics, and by acid, heat, ethanol and superoxide stresses. Here, we have used several complementary approaches to identify genes controlled by σM. In many cases, expression is only partially dependent on σM because of both overlapping promoter recognition with other ECF σ factors and the presence of additional promoter elements. Genes regulated by σM have a characteristic pattern of induction in response to cell envelope-acting antibiotics as evidenced by hierarchical clustering analysis. σM also contributes to the expression of the Spx transcription factor and thereby indirectly regulates genes of the Spx regulon. Cell envelope stress responses also include regulons controlled by σW, σB and several two-component regulatory systems (e.g. LiaRS, YycFG, BceRS). Activation of the σM regulon increases expression of proteins functioning in transcriptional control, cell wall synthesis and shape determination, cell division, DNA damage monitoring, recombinational repair and detoxification.
The adaptation of bacteria to extracytoplasmic stresses requires mechanisms of trans-membrane signalling. Among the most common signalling mechanisms are the ubiquitous two-component regulatory systems (TCS) in which a trans-membrane sensor kinase controls the activity of a cytosolic response regulator (Mascher et al., 2006), and extracytoplasmic function (ECF) σ factors in which σ factor activity is controlled by one or more trans-membrane regulators (anti-σ factors and accessory proteins) (Raivio and Silhavy, 2001; Helmann, 2002).
Knowledge of the remaining ECF σ factors (σY, σV, σZ and σYlaC) is also sparse. The σY regulon includes an autoregulated heptacistronic operon, postulated to encode a toxic peptide, and one additional verified gene (ybgB) encoding a putative immunity protein for an antimicrobial peptide (Cao et al., 2003; Tojo et al., 2003). Analysis of genes induced by overexpression of σV revealed extensive overlap with target operons that are also activated, under other conditions, dependent upon σX and/or σW (Zellmeier et al., 2005). This reflects the challenge of defining the regulons controlled by ECF σ factors when there is partial, or even extensive, overlap in promoter specificity (Qiu and Helmann, 2001; 2002; Minnig et al., 2003; Mascher et al., 2007). It is also possible that overexpression of σV activates either synthesis or activity of other ECF σ factors. Similarly, a transcriptome analysis of gene expression as measured 2 h after overexpression of each of the seven ECF σ factors in B. subtilis identified numerous operons that were upregulated by multiple σ factors (Asai et al., 2003). However, it is unclear whether these proposed targets result from direct or indirect effects of σ factor activity over this long period of induction and no further analysis or promoter identification studies were reported.
As a class, ECF σ factors are often control cell envelope stress responses (Raivio and Silhavy, 2001; Helmann, 2002; Alba and Gross, 2004). Cell envelope stress can be broadly defined as resulting from chemical or genetic impairment of the assembly, maintenance, or function of the cell membranes and wall. Many antibiotics target enzymes needed for peptidoglycan biosynthesis and exposure of cells to these compounds elicits a cell envelope stress response. Other compounds may interfere with the synthesis or integrity of the cell membranes and induce a related stress response (Darwin, 2005; Raivio, 2005; Rowley et al., 2006). For example, many cationic antimicrobial peptides, bacteriocins, and some organic solvents and detergents interfere with the integrity of the cytoplasmic membrane (Bauer and Dicks, 2005; Breukink and de Kruijff, 2006; Duque et al., 2007; Martinez et al., 2007). Misfolding or overexpression of secreted proteins can also trigger envelope stress and activate the synthesis of appropriate degradation enzymes or protein chaperones (Hyyrylainen et al., 2005). While the regulatory pathways and the precise regulons differ between organisms, it is clear that envelope stresses elicit a complex set of overlapping responses to help maintain the integrity of the cell membrane and wall.
Here, we identify genes controlled by σM. In addition to confirming most of the previously assigned regulon members (Jervis et al., 2007), we provide evidence that σM contributes to the transcription of many additional operons, including genes important for cell wall biosynthesis, shape determination and cell division, DNA damage responses and detoxification enzymes. In addition, a subset of genes previously assigned to the σX and/or σW regulons are here shown to respond to cell envelope-active compounds in a pattern characteristic of σM-dependent control.
Results and discussion
Strategy for defining the σM regulon
In previous work, 13 promoters (controlling 18 genes) were defined as σM-dependent based, in large part, on their induction in a strain engineered to overproduce σM and the confirmation of the corresponding transcription start points by 5′-RACE (Jervis et al., 2007) (Table 1A; and references therein). In a separate study, the mRNA levels for ∼50 genes were found to be upregulated at least threefold 2 h after induction of σM and were therefore assigned as candidate members of the σM regulon (Asai et al., 2003). However, this list of candidates has never been tested and appears to contain a number of false positives. Moreover, the similar promoter recognition properties of different ECF σ factors leads to overlapping recognition (Huang et al., 1998; Qiu and Helmann, 2001; Mascher et al., 2007). While in many cases these regulatory overlaps are physiologically relevant, overexpression of individual ECF σ factors may also lead to non-physiological activation of promoters normally regulated by related σ factors.
Operon structure is indicated based on distance between adjacent genes and coregulation in transcriptome studies (Sierro et al., 2007). Initial genes in parentheses indicate an internal promoter element. In the specific case of (ydbO-ydbP(as))-ddlmurF (as) indicates the antisense orientation of the ydbP gene (see Fig. 5).
Recognition elements for σM (−35 and −10 regions) and mapped start points for transcription are in bold. Note that yceC also has a σA-dependent start site (−10 region in bold).
Distance from the indicated +1 site to the start codon of the first codirectional gene.
Evidence for the indicated +1 assignment is either from 5′-RACE (RACE), primer extension (PE), or run-off transcription (RO) studies as indicated (using the holoenzyme form indicated parenthetically). RACE(2) indicates that the start site was determined both in this work and also by (Jervis et al., 2007). ‘no product’ indicates that the indicated promoter site was not detected using 5′-RACE. References for start site determination are in the last column.
ECF σ indicates which, if any, σ factors have been previously implicated in transcription from the indicated promoter. Those σ factors indicated parenthetically were identified as factors which, when overproduced, would lead to at least a > 3-fold increase in one or more gene of the corresponding operon after 2 h of induction (Asai et al., 2003). No further evidence of promoter activity or recognition have been reported prior to the work reported herein.
Fold change refers to the first gene in the operon. Where data were not available (as a result of a bad spot on the array), numbers shown are for the next gene in the operon (shown in parentheses) or from previous measurements of the vancomycin stimulon (Cao et al., 2002b; shown in brackets).
HC refers to the position of the gene(s) in the hierarchical clustering diagram of Fig. 3. A dash indicates that the gene did not cluster within M1-M4 (note that yebC fell between clusters M2 and M3).
ROMA indicates the signal intensity of the gene(s) in the ROMA experiment of Fig. 2. The indicated a σM-dependent increase in signal intensity of: ‘4’, 1000–2500 units (9 genes total), ‘3′, 333–999 units (15 genes), ‘2’, 100–332 units (34 genes), and ‘1’, 40–99 units (63 genes). Some genes were not detected as a result of bad spots in the microarray (yjbC, ypuA, yrhH), but downstream genes in the operon were detected (except for ypuA which is monocistronic).
Inferred σ factor dependence of the cloned promoter region as judged by analysis of lacZ transcriptional fusions in this work (e.g. Fig. 6). Inferred σ factor dependencies from previous studies of reporter fusions are in parentheses. (1) indicates the promoter fusion was active, but not obviously σM dependent. (2) indicates a promoter fusion with little or no activity. A dash indicates not tested.
A. Previously assigned σM promoters (19 genes in B. subtilis 168)
Herein, we define the σM regulon and provide evidence that this ECF σ factor contributes to the antibiotic-inducible expression of ∼57 genes (30 operons) associated with known or candidate σM-activated promoter elements, including several sites previously assigned to either or both the σX and σW regulons. A comprehensive listing of σM-regulated genes, together with a summary of evidence from this and previous studies, is presented in Table 1. In addition, several genes regulated by Spx, a transcription factor previously associated with disulphide stress responses (Zuber, 2004), are also induced by cell envelope-active compounds in a pattern consistent with σM-mediated induction.
To define the σM regulon we looked for genes that satisfy most or all of the following criteria. First, we identified genes that were induced by vancomycin, a known inducer of σM-regulated genes (Cao et al., 2002b; Mascher et al., 2003). In most cases, these genes were induced little if at all in a sigM null mutant strain. Second, we used ROMA (run-off transcription followed by microarray analysis; Cao et al., 2002a) to identify those genes transcribed by reconstituted σM holoenzyme. Third, we sought to identify plausible σM-dependent promoters using computer-based consensus searches followed by 5′-RACE to identify transcriptional start sites. For several promoters, transcriptional fusions were used to determine the relative role of σM, σW and σX in promoter activity. Fourth, we used hierarchical clustering (de Hoon et al., 2004) to identify additional genes that were induced by cell envelope-active antibiotics in a pattern consistent with a primary role of σM in gene regulation. These four approaches converged to define a set of ∼57 genes (30 operons) that are likely to be direct targets for transcriptional activation by σM under conditions of antibiotic stress (Table 1). An additional 10 genes are associated with candidate promoters which can not yet be confidently assigned to the σM regulon (Table 1D).
Identification of σM-dependent genes in the vancomycin stimulon
Vancomycin, an inhibitor of peptidoglycan biosynthesis, is a known inducer of both the autoregulated sigM operon and several previously described σM-regulated genes (Cao et al., 2002b; Thackray and Moir, 2003). We have used DNA microarray-based analysis to identify σM-dependent members of the vancomycin stimulon. The vancomycin stimulon includes > 250 genes induced at least twofold after 10 min of treatment (Fig. 1), consistent with previous studies (Cao et al., 2002b; Mascher et al., 2003). Most of the induced genes are members of known regulons including those controlled by the LiaRS TCS (Jordan et al., 2006), σW (Cao et al., 2002a) and σB (Price et al., 2001). The two most strongly induced genes, liaI and liaH, are the first two genes in the strongly activated liaI operon controlled by the cell envelope stress-inducible LiaRS TCS (Mascher et al., 2004). Vancomycin also induces genes controlled by the σY and YvrI regulatory proteins (S. MacLellan and J.D. Helmann, unpubl. results), and the ytrA and ywoB operons, consistent with previous studies (Cao et al., 2002b). Induction of the σB regulon by vancomycin was also noted previously (Mascher et al., 2003), although this regulon is apparently not induced by cell wall-active antibiotics in Bacillus licheniformis (Wecke et al., 2006). Numerous other genes, as analysed herein, are known or candidate members of the σM regulon.
To determine which of the vancomycin-inducible genes were likely targets for σM, we determined the vancomycin stimulon in sigM mutant cells. As expected, known members of the σM regulon were induced by vancomycin in wild type, but not in the sigM mutant (Table 1A). A number of additional genes and operons not previously assigned to the σM regulon (e.g. the murG, ycgR, recU and ywaC operons) were also induced in a σM-dependent manner and were associated with plausible σM-dependent promoter elements (Table 1B and C).
Identification of genes transcribed by σM holoenzyme in vitro using ROMA
We used ROMA (Cao et al., 2002a) to identify genes likely to be under the direct transcriptional control of σM. Total genomic DNA was transcribed in vitro by purified RNA polymerase with and without the addition of saturating levels of σM. The two RNA samples were labelled with fluorophores and hybridized to DNA microarrays as for a conventional transcriptome analysis. In this case, however, the relevant parameter is not fold-change, but rather the difference in signal intensity in the presence versus the absence of σM (Fig. 2). Absolute signal intensities are not particularly meaningful in this assay, because signal intensity depends on promoter efficiency in vitro, which is imperfectly correlated with in vivo activity. Thus, ROMA provides only a qualitative visualization of the transcriptional activity of the σM holoenzyme.
There is a generally good correspondence between the ROMA results and the in vivo transcriptome analyses: many of the genes that were induced by vancomycin and appeared to be σM-dependent also gave positive signals in the ROMA assay. In other cases, genes associated with positive ROMA signals were not strongly induced by vancomycin, but were activated by other cell envelope-active compounds in a manner characteristic of σM (see below). The appearance of a positive ROMA signal suggests that these genes are likely to be direct targets for the σM holoenzyme. Indeed, of the 19 σM-dependent promoters assigned in this work (Table 1A–C), 14 correspond to positive ROMA signals. The genes that were not detectably transcribed in vitro under these conditions may require additional factors for their expression, the promoters may be relatively weak, or the promoters may be dependent on negative supercoiling for activity. Thus, there appear to be relatively few false negatives in this assay.
In contrast, the ROMA technique does yield a significant number of signals that do not correlate with our consensus list of σM-dependent promoters. Five of these may in fact be recognized by σMin vivo, but this cannot be confidently asserted based on the available date (Table 1D). Further analysis revealed that most of the remaining signals (∼30 genes) result from readthrough transcription from upstream σM-dependent genes (Supplementary material; Table S4). The high sensitivity of the ROMA assay can lead to strong signals in this in vitro assay even though the corresponding genes are not significantly affected by σMin vivo. This likely reflects the facts that transcription termination is more efficient in vivo (in the presence of additional factors such as NusA; Borukhov et al., 2005) and the expression of some of these downstream genes is normally at a sufficiently high level that readthrough transcription is not physiologically significant. For these reasons, in previous analyses we have used restriction enzyme digested genomic DNA to help limit transcription to promoter proximal regions (Cao et al., 2002a; 2003; Cao and Helmann, 2004).
Identification of σM-dependent promoters using consensus searches and 5′-RACE
Consensus-directed searches for promoter elements can identify a significant fraction of the genes regulated by alternative σ factors as shown in our previous work on the σW regulon (Huang et al., 1999) and in related studies of the Streptomyces coelicolorσR (Paget et al., 2001) and enterobacterial σE regulons (Rhodius et al., 2006). Here, we searched the B. subtilis genome for sequences similar to previously verified σM-activated promoter sites (Jervis et al., 2007). Because σM, σX and σW all recognize rather similar promoter sequences (Mascher et al., 2007), this approach identified candidate promoters that may be regulated by any one (or more) of these and perhaps other ECF σ factors.
Candidate ECF-type promoters were found preceding most of the genes that were tentatively assigned to the σM regulon based on the transcriptome and ROMA experiments (Table 1). To determine whether transcription initiation events occur at these sites in vivo, we used 5′-RACE to map transcription start points in RNA samples isolated from cells treated with vancomycin. These experiments confirm the activity of 17 predicted promoter elements (Table 1) including four that were also characterized using 5′-RACE by Jervis et al. (2007) (ypbG, ydaH, yfnI and maf). Start sites for several other genes were previously determined in the course of characterizing the σX and σW regulons (Table 1A and C). Furthermore, the divIC and tarA promoters from B. subtilis W23 are also controlled by both σX and σM (Minnig et al., 2003). Note that the tarA operon, required for synthesis of ribitol-based teichoic acids, is present in B. subtilis W23 strains, but not in B. subtilis 168 (which has glycerol-based teichoic acids).
Hierarchical clustering analysis of cell envelope stress regulated genes
We reasoned that those genes primarily dependent on σM for their expression should share a common pattern of transcriptional responses in cells challenged with a variety of antibiotics. To test this hypothesis, we performed a hierarchical clustering analysis using transcriptional profiling studies of B. subtilis treated with 10 different cell envelope-active compounds including bacitracin (Mascher et al., 2003), vancomycin (in both wild-type and sigM mutant cells; this study), and 8 compounds studied by Hutter et al. (2004a). As expected, genes induced by particular sets of cell envelope-active compounds tend to form discrete clusters, often corresponding to regulons. To simplify the data representation, we repeated the clustering analysis using a subset of 293 genes corresponding to the σM, σY, σW, LiaRS and BceRS regulons, and antibiotic-inducible members of the σB, Spx and YycFG regulons (see Experimental procedures). We also included additional genes that clustered with these regulons in our initial genome-wide analysis and all other genes induced more than threefold by vancomycin.
While the precise nature of the cluster diagram varies with different sets of starting genes, most of the genes controlled (at least in part) by σM were consistently found in one or more distinct clusters, each with a similar overall pattern of responsiveness. These clusters are clearly distinguished from the σB and σW regulons (Fig. 3), and from regulons controlled by σY, LiaRS, or other stress responsive pathways. These results imply, at least for this set of stress conditions, that the contribution of σM to gene expression is sufficiently strong as to impart a characteristic response profile (Fig. 4).
As expected, many of the genes under the direct control of σM (e.g. cluster M1; Figs 3 and 4) are induced by vancomycin in wild type, but not in the sigM mutant. In contrast, the σW and σY regulons are induced in both genetic backgrounds. The σW regulon was most strongly induced by Triton-X-114 treatment, whereas σY responded strongly to both Triton-X-114 and gramicidin. For reasons that are not yet clear, induction of the σB regulon by vancomycin appeared to depend on σM. One caveat with this analysis is that our studies and those of Hutter et al. (2004a) were carried out in different strains of B. subtilis 168 and with differing times of exposure.
σM-dependent transcripts with unusually long 5′-untranslated regions
Most σM-dependent promoters are located adjacent to the regulated gene(s) with a predicted 5′-untranslated region (5′-UTR) of < 200 nt. However, several exceptions are apparent (Fig. 5). As noted previously, the predicted disA (formerly yacK) and ysxA(radC) transcripts have 5′-UTRs of 890 and 459 nt respectively (Jervis et al., 2007). The promoter upstream of disA (designated Psms) is within the radA/sms gene which encodes a paralogue of the recombination protein RecA and functions in recombination-dependent DNA repair (Lovett, 2006). It is possible that an amino-terminal truncated variant of RadA/Sms could be translated from the resulting transcript, although this speculation has not yet been tested. The downstream disA gene encodes a DNA integrity scanning protein (DisA) important for sensing DNA damage (Bejerano-Sagie et al., 2006). The promoter upstream of ysxA (Pmaf) is within maf which encodes an NTPase that may function in cleansing the cellular NTP pool of xanthine and inosine triphosphates (Zheng et al., 2005).
We identified two additional σM-dependent transcription units with unusually long 5′-UTRs (or possibly encoding amino-terminal truncated proteins). The promoter inside murG was confirmed by 5′-RACE and is predicted to produce a transcript with a 603 nt 5′-UTR upstream of murB. It seems unlikely that an amino-terminally truncated MurG would be functional because the missing protein region includes several active site residues (Crouvoisier et al., 2007). This candidate promoter was originally identified based on a consensus search, but strong supporting evidence emerged from analysis of its unusual pattern of induction: unlike other σM-dependent promoters, murG was apparently induced only by vancomycin and bacitracin and not by any of the antibiotics investigated by Hutter et al. (2004a). Thus, murG does not cluster with other σM-regulated genes, including those presumably encoded by the same transcript. This apparent inconsistency can be explained by the fact that the arrays used by Hutter et al. contain short dsDNA probes corresponding to the 5′-region of each gene and, in this case, the murG probe is upstream of the assigned promoter. In contrast, the arrays we have used contain either polymerase chain reaction (PCR) products corresponding to the entire ORF (bacitracin induction experiments; Mascher et al., 2003) or ssDNA probes (65-mers; this work) from within genes (in this case, downstream of the intragenic promoter). Thus, depending on the design of the arrays, murG is either detected or not as antibiotic-inducible whereas the downstream genes in this operon (murB, divIB, ylxW, ylxX and sbp) are induced in a pattern consistent with σM regulation (cluster M1).
The promoter located within ydbO also generates a transcript with an unusually long 5′-UTR (∼1110 nt) and the first codirectional gene is ddl, encoding D-ala-D-ala ligase. This promoter would also generate an antisense message for the convergent ydbP gene. Consistent with this hypothesis, apparent induction of ydbP was detected with several antibiotics investigated by Hutter et al., which, because of the dsDNA probes used in these arrays (Hutter et al., 2004a), could correspond to antisense transcripts (Fig. 3; cluster M2). Induction of this gene was not detected with the ssDNA probes used in our arrays. The ydbP gene encodes a thioredoxin-like protein thought to be under σB control (Petersohn et al., 1999). It is not yet known whether this antisense transcript is physiologically relevant, but the fact that induction of this RNA is detected in vivo would be consistent with such a role. Both ddl and murF are weakly induced by vancomycin-inducible in wild type (1.6- and 1.3-fold respectively; Table 1), but not in sigM mutant strains (0.7-fold), consistent with their expression from the PydbO promoter element. Weak induction (1.3- to 2.0-fold) of these genes by vancomycin was also observed in previous studies (Cao et al., 2002b; Hutter et al., 2004a) and the overall pattern of response to cell envelope-active antibiotics is consistent with σM activation (Fig. 3).
Regulation of the complex yjbC-spx operon by σM
The spx gene is subject to unusually complex regulation (Fig. 5). This gene is coexpressed with the upstream yjbC gene as a 1.2 kb transcript (Antelmann et al., 2000). Primer extension analysis of transcripts initiating upstream of yjbC revealed multiple start sites, including one corresponding to a promoter recognized in vitro by both σW and σX (Antelmann et al., 2000; Cao et al., 2002a). Both the yjbC and spx genes are induced in a strain engineered to overproduce σM (Jervis et al., 2007). In the arrays used in this study, spx was only weakly induced (3.3-fold), and there was no signal from the yjbC oligonucleotide. However, both genes are induced by vancomycin in previous studies (induction values after 10 min of 9.0- and 5.6-fold, respectively, in Cao et al., 2002b; and 9.0- and 4.0-fold, respectively, in Hutter et al., 2004a). Moreover, spx gives an exceptionally strong signal in the ROMA experiment (Fig. 2). Taken together, these findings lead us to suggest that the induction of spx by antibiotics likely reflects transcription from the upstream ECF-class promoter preceding yjbC (PyjbC).
The yjbC and spx genes are separated by 184 nt and spx is clearly subject to additional regulatory inputs. For example, there is a σA-type promoter (P3) immediately upstream of spx that is negatively regulated by the PerR and YodB repressors (Leelakriangsak et al., 2007; Leelakriangsak and Zuber, 2007). In addition, a σM-dependent promoter was proposed in this intergenic region (Jervis et al., 2007), although it has an atypical −10 element (TGAC) and we failed to detect promoter activity from this region when cloned as a lacZ reporter fusion (data not shown). Together with the observed coinduction of yjbC and spx, we suggest that induction of spx by cell envelope-active compounds originates from PyjbC.
Identification of antibiotic-inducible genes regulated by Spx
In the course of this work, we noted that one previously identified σM-dependent gene (yraA) was assigned an unusual promoter with a non-canonical −35 region (it lacks the ‘AAC’ motif) and −10 region (missing the otherwise invariant GT dinucleotide) (Jervis et al., 2007). Upon closer inspection, we realized that the mapped transcription start site is preceded by plausible −35 (TTGAag) and extended −10 (TGtTATtcT) elements for σA Moreover, a search of the literature revealed that yraA was among the genes most strongly induced by Spx (Nakano et al., 2003a). Because transcription of spx can be activated by σM (Jervis et al., 2007), we hypothesized that Spx, in turn, induced yraA.
In addition to yraA, several other genes positively activated by Spx (Nakano et al., 2003a) are also antibiotic-inducible (Figs 3 and 4; cluster M4). The results of hierarchical clustering suggest that the induction of Spx by σM is physiologically significant. Specifically, the overall response pattern of known Spx-activated genes was more similar to the σM than to the σW regulon (Fig. 4; compare M4 with M1 and W). However, the yjbCspx operon can be induced by other ECF σ factors including σW. Indeed, Spx-regulated genes were weakly induced by vancomycin under our conditions independent of σM, presumably reflecting activation of PyjbC by σW.
Spx has been characterized as a positive regulator of the disulphide stress response (activated by diamide) and a negative regulator of several operons controlled by TCSs (including the ComA-dependent srfABCD operon) (reviewed in Zuber, 2004). However, Spx had not previously been implicated as a regulator of antibiotic stress-induced genes.
Determining the relative contributions of σM, σX and σW using reporter fusions
The collective evidence of transcriptional profiling, ROMA, promoter consensus search and 5′-RACE, and hierarchical clustering analysis defines a large list of 57 genes that are likely to be direct targets for σM (Table 1). However, in some cases the observed vancomycin induction was only partially dependent on σM. This could be resulting from the presence of other antibiotic-inducible promoters or to overlapping recognition among ECF σ factors (Mascher et al., 2007). Indeed, several of the promoters herein assigned as σM-dependent were previously characterized as candidate members of the σX and/or σW regulons (Table 1C).
To further define the role of σM in directing transcription from these promoter elements, we generated a set of ectopically integrated transcriptional reporter fusions. For comparison, we included the autoregulatory site (PM) preceding sigM (Horsburgh and Moir, 1999), a site known to be completely dependent on σM for activity (Fig. 6a). A subset of the newly identified σM-dependent promoters were also largely dependent on σM for activity including ywaC, yrhH and murG (Fig. 6a). In other cases, there was a significant amount of basal promoter activity that was independent of σM (Fig. 6b). Nevertheless, σM contributes, at least partially, to the induction observed upon exposure to vancomycin. For at least some of this latter set of promoters, the residual promoter activity is apparently resulting from overlapping recognition by σX and/or σW. In several cases (maf, ypuD, ycgR, metA, yjbC, secDF) promoter activity was reduced to background levels in the sigM sigX sigW triple mutant (data not shown). In three other cases (yebC, rodA and ywtF) there was full activity even in the triple mutant. This could be because of recognition of this same site by another ECF σ factor, or possibly because of the presence of an additional promoter on the cloned DNA fragment.
Our hierarchical clustering results (Fig. 3) suggest that several sites previously assigned as belonging to the σX and/or σW regulons respond in a pattern consistent with a dominant role of σM in their expression. Whereas most σW-controlled genes cluster tightly together (Fig. 3; cluster W), several promoters recognized by σW (yceC, ywaC; Cao et al., 2002a) or by both σW and σX (e.g. ywbN, yjbC and abh; Helmann, 2002) can now be seen to respond to cell envelope stress in a manner consistent with σM-regulation. Similarly, the σX-activated rapD and dltABCD operons (Cao and Helmann, 2004) respond, at least under these stress conditions, as σM-regulated operons. These findings reinforce the notion that most σW-dependent genes are exclusively expressed by σW, whereas many genes dependent on σX and/or σM have promoter sites recognized by multiple ECF σ factors.
Functional characterization of the σM regulon
The work reported here indicates that σM controls a much larger regulon than previously envisioned. Altogether, we estimate that at least 57 genes (30 operons) are associated with promoters directly responsive to σM (Table 1A–C) and another 20–30 genes are activated indirectly by Spx. Functions under the direct control of σM (Table 2) include (i) gene regulation, (ii) cell wall synthesis, shape determination and cell division, (iii) DNA monitoring and repair and (iv) detoxification. Presumably, upregulation of these genes in cells exposed to inhibitors of cell wall synthesis is adaptive. However, there are clearly redundant pathways for protection against many antibiotic stresses and a sigM mutant strain does not display a significantly increased sensitivity to most cell envelope active compounds (Mascher et al., 2007). The exceptions include bacitracin (Cao and Helmann, 2002), moenomycin, SDS and some beta-lactams (Mascher et al., 2007). This presumably reflects the redundancy in antibiotic-inducible ECF σ factor regulons because a sigM sigX sigW triple mutant is significantly increased in antibiotic sensitivity relative to a sigX sigW double mutant (Mascher et al., 2007).
σM-dependent regulatory proteins. Activation of σM enhances expression of several known or putative regulatory proteins. These include σM itself, which is cotranscribed with two membrane proteins that negatively regulate σM activity (Horsburgh and Moir, 1999). YwaC is a recently characterized ppGpp synthase, and may function to modulate gene expression (Nanamiya et al., 2007). Recently, ywaC was shown to be induced by depletion of teichoic acids or treatment with cell envelope-active antibiotics, but not antibiotics that target translation, DNA metabolism, or other functions (E. Brown, personal communication). Other candidate regulators controlled, at least in part, by σM include YwtF, a putative transcription factor related to the LytR family, the transition state regulator Abh (Strauch et al., 2007), and RapD, a putative response regulator aspartate phosphatase that negatively regulates ComA-dependent genes (Ogura and Fujita, 2007). It is interesting to note that both RapD and Spx appear to target the ComPA TCS for negative regulation (Nakano et al., 2003b).
We can speculate about the possible significance of Spx as a target of σM control. Spx controls several genes that contribute to the maintenance of a reducing intracellular environment (Nakano et al., 2005; Zuber, 2004). It has recently been suggested that many bactericidal antibiotics kill cells by the generation of reactive oxygen species (Kohanski et al., 2007). Thus, induction of the Spx regulon may help prevent cell death. It is worth noting that in Lactococcus lactis antibiotic stress activates a TCS orthologous to LiaRS (CesRS) which, in turn, activates one of seven Spx paralogues, SpxB. SpxB contributes to peptidoglycan acetylation and thereby contributes to resistance against cell wall hydrolysis (Veiga et al., 2007). The role of B. subtilis Spx in antibiotic resistance and/or cell envelope modification awaits further study.
Cell wall synthesis, shape determination and cell division. The results presented here identify cell wall synthesis and shape determination genes as a major category of σM-regulated functions. Because many of these loci encode essential functions, it is clear that they are not exclusively under σM control. Many of these loci have multiple promoter elements and can be transcribed as complex sets of overlapping mRNAs. Nevertheless, the hierarchical clustering results imply that these loci are induced by a variety of cell envelope-active compounds with a pattern typical for σM-regulated genes.
σM also appears to activate transcription of the ytpAB operon. The YtpA protein functions in processing of membrane lipids by hydrolysis of the 2-sn-acylated fatty acid chain of phosphatidylglycerol to generate a singly acylated derivative referred to as bacilysocin (Tamehiro et al., 2002). This compound is reported to have antibacterial activity, but it is not clear if it is ever released from growing cells. Thus, this protein may actually play a role in some aspect of membrane synthesis, repair, or recycling.
DNA monitoring and repair. Activation of σM may also increase cellular defences against DNA damage. The ysxA locus encodes a putative DNA repair gene related to E. coli radC which functions in recombinational repair. However, the original radC mutation was later found to map to recG, leaving the function of radC (and by implication, ysxA) uncertain (Lombardo and Rosenberg, 2000). As noted above, activation of the σM-dependent promoter within the sms/radA gene may lead to expression of a truncated form of the Sms/RadA protein, a RecA-like DNA binding protein important for the recombinational repair of stalled replication forks (Lovett, 2006). This same promoter activates expression of disA (formerly yacK), encoding a recently described DNA integrity scanning protein that moves rapidly along the chromosome apparently scanning for DNA damage (Bejerano-Sagie et al., 2006).
Detoxification. The σM regulon includes numerous proteins with probable functions in detoxification, although few molecular details are yet known. YqjL is a putative hydrolase that contributes to resistance to paraquat and other superoxide-generating compounds with a 2,2′-dipyridyl ring (Cao et al., 2005). The yceCDEFG operon is homologous to tellurite resistance determinants. Our results indicate that σM can also activate expression of the yrhH gene encoding a putative methyltransferase. Although the function of this gene is unknown, it has recently been shown to be induced in a mutant strain defective for Smc, the structural maintenance of chromosome protein involved in chromosome compaction and partitioning (Britton et al., 2007). Transcriptome results suggest that activation of the yrhH promoter also leads to increased expression of the downstream yrhI and yrhJ genes. Indeed, Moir and coworkers had previously assigned yrhJ to the σM regulon on the basis of studies using an integrational reporter fusion (Thackray and Moir, 2003). However, 5′-RACE studies targeted to yrhJ only identified the known σA-dependent promoter (Lee et al., 2001) in the yrhH-yrhI intergenic region (Jervis et al., 2007). The yrhI and yrhJ genes encode a regulatory protein (designated FatR; Gustafsson et al., 2001) and cytochrome P450 CYP102A3. This inducible P450 monooxygenase oxidizes long and branched-chain unsaturated fatty acids (Gustafsson et al., 2004). It has been proposed that this system might be involved in regulation of membrane fluidity or as a defence against toxic fatty acids.
The σM regulon is induced by chemical and genetic perturbation of envelope synthesis and function
Representative σM-regulated genes have been previously observed as induced by antibiotic-stress and other conditions potentially affecting the cell envelope including acid, ethanol, heat and superoxide stresses (Thackray and Moir, 2003). In addition, genes herein assigned to the σM regulon have been previously proposed as markers for specific classes of cell envelope stress. For example, Hutter et al. (2004b) identified ypuA as gene selectively induced by cell wall synthesis inhibitors, but not by a variety of other antibiotic classes. Similarly, Brown and colleagues have identified ywaC as a gene strongly induced by depletion of teichoic acids and by several antibiotics that target the cell wall (E. Brown, pers. comm.). σM regulon genes are also induced by cationic antimicrobial peptides including LL-37 (Pietiainen et al., 2005). In addition, a subset were found to be significantly induced by secretion stress elicited by depletion of the PrsA foldase (Hyyrylainen et al., 2005). However, in these and related studies, the nature of the regulatory pathways controlling gene induction were generally not defined. Based on the results reported here, it appears that induction of the σM regulon in general is a good reporter for inhibition of cell envelope biosynthesis and function.
Here we provide our best current assessment of the scope of the σM regulon and reveal a major role for this ECF σ factor in controlling essential functions related to cell envelope biogenesis and cell division. Defining the σM regulon has been challenging, in large part because of the extensive overlap with other regulatory systems including those controlled by other ECF σ factors. Nevertheless, the strategy used here, employing a combination of in vivo and in vitro transcriptomics, hierarchical clustering and promoter mapping and characterization has yielded a generally consistent picture and allows us to assign at least 30 operons (∼57 genes) as being directly transcribed by σM. Future studies will be required to more fully characterize the physiological significance of these regulatory changes in adaptation to cell envelope stresses.
Bacterial strains and growth conditions
All B. subtilis and E. coli strains, plasmids and oligonucleotides used in this study are listed in the Supplementary material, Tables S1–3. All B. subtilis strains were derived from wild-type 168 strain CU1065. Bacteria were grown in Luria–Bertani (LB) medium at 37°C with vigorous shaking. For E. coli, 100 μg ml−1 of amplicillin was used to select for Ampr. For B. subtilis, antibiotics used for selection were as followed: 100 μg ml−1 of spectinomycin for Spcr, 10 μg ml−1 of kanamycin for Kanr, 10 μg ml−1 of neomycin for Neor and 1 μg ml−1 of erythromycin and 25 μg ml−1 of lincomycin for macrolide-lincomycin-streptogramin B resistance (MLSr).
Construction of the spx-null mutant
A chromosomal deletion was created by long-flanking homology PCR (LFH-PCR). Flanking fragments were amplified with Pfu DNA polymerase (Stratagene), and the flanking fragments and antibiotic resistance gene were joined with the Expand Long Template PCR system (Roche). Primers 3819 and 3820 were used to amplify the upstream fragment, and primers 3821 and 3822 were used to amplify the downstream fragment. Detailed protocols are available at http://www.micro.cornell.edu/cals/micro/research/labs/helmann-lab/supplements.cfm
RNA preparation and Microarray analysis
A B. subtilis microarray, consisting of 4109 gene-specific antisense oligonucleotides (65-mers; Sigma-Genosys), was printed at the W.M. Keck Foundation Biotechnology Resource Laboratory, Yale University. Each slide contains 8447 features corresponding to duplicates of each ORF-specific oligonucleotide, additional oligonucleotides of control genes and a 50% DMSO blank control. To identify sigM-dependent genes, three separate sets of microarray experiments were performed, including (i) a comparison of transcript levels between CU1065 with and without vancomycin, (ii) a comparison of transcript levels between CU1065 and the HB0031 (sigM::kan) strain with and without vancomycin, and (iii) a comparison of transcript levels between CU1065 and HB4728 (spx::spc) strain with and without vancomycin. The cell cultures were grown to an OD600 of 0.4 and split into two flasks with equal volume, vancomycin was added to one flask to a final concentration of 2 μg ml−1 (10X MIC), and cells were harvested 10 min after treatment. The total RNA was prepared from two different cultures (biological replicates). The RNeasy mini kit (Qiagen) was used to extract total RNA. DNase treatment of RNA was performed by using TURBO DNA-freeTM DNase and removal reagents (Ambion). RNA concentrations were quantified using a NanoDrop spectrophotometer (Nanodrop Tech., Wilmington, DE). cDNA was synthesized and differentially labelled using the SuperScriptTM Plus Indirect cDNA labelling System (Invitrogen). cDNA was generated from 20 μg of RNA using random primers in a reverse transcriptase reaction at 42°C for 2 h. cDNA was purified using a Qiagen PCR purification kit (Qiagen) prior to indirect labelling with Alexa Fluor 555 or Alexa Fluor 647 (at least 3 h at room temperature). Labelled cDNA was purified using a Qiagen PCR purification kit (Qiagen) to remove any unincorporated dye and the labelled cDNA was quantified. Both labelled cDNA populations were applied onto a microarray slide and hybridized overnight at 42°C for 16–18 h. After washing, hybridized microarray slides were analysed using a GenePixTM 4000B array scanner (Axon Instruments). Two microarray replicates were performed for each biological replicate. Images were processed using the GenePix Pro 4.0 software package which produces (R,G) fluorescence intensity pairs for each gene. Fluorescent signal intensities were imported into Microsoft Excel. Each expression value is represented by at least four separate measurements (duplicate spots on each of two slides). Mean values and standard deviations were calculated with Excel. The microarray datasets were filtered to remove those genes that were not expressed at levels significantly above background in either condition (sum of mean fluorescence intensity < 100). In addition, the mean and standard deviation of the fluorescence intensity were computed for each gene and those where the standard deviation was greater than the mean value were ignored. The induction values were calculated by using the signal intensities of vancomycin-treated samples divided by untreated samples. Microarray datasets and related files (Microsoft Excel files and files generated using Cluster 3.0 and Treeview) are available at http://www.micro.cornell.edu/cals/micro/research/labs/helmann/-lab/supplements.cfm.
Cloning, expression and purification of σM protein
The sigM gene was PCR amplified from B. subtilis chromosomal DNA with oligonuclotides designed to engineer an NcoI site upstream and a BamHI site downstream of the sigM gene. The PCR product was cloned into pET16b (Novagen) via the NcoI and BamHI sites to generate pWE01. The sequence of sigM in pWE01 was verified by DNA sequencing (Cornell DNA sequencing facility). The resultant plasmid was used to transform BL21/DE3(pLys) cells. Cells were grown to mid-logarithmic phase at 37°C in 1 l of LB medium and 100 mg ml−1 of ampicillin. σM expression was induced with 1 mM IPTG for 3 h at 37°C. Cells were collected by centrifugation, resuspended in 20 ml of disruption buffer [50 mM Tris-HCl (pH 8.0), 2 mM EDTA, 0.1 mM dithiothreitol (DTT), 1 mM β-mercaptoethanol, 233 mM NaCl, 10% (v/v) glycerol] and lysed with a French pressure cell at 10 000 psi followed by brief sonication, and the inclusion bodies were recovered by centrifugation. The inclusion bodies were washed twice with 10 ml TEDG buffer [10 mM Tris-HCl (pH 8.0), 10 mM EDTA, 0.1 mM DTT, 10% glycerol] containing 0.5% (v/v) Triton X-100 and then dissolved in 10 ml of the same buffer plus 1% (v/v) Sarkosyl. After centrifugation to remove the insoluble fraction, the supernatant was gradually diluted to 100 ml with TEDG−0.01% Triton X-100, to allow renaturation of σM, and then loaded onto a 5 ml Q-Sepharose (Sigma) column equilibrated with TEDG−0.01% (v/v) Triton X-100. After washing with 50 ml of TEDG−0.15 M NaCl−0.01% Triton X-100, σM was eluted with TEDG−0.4 M NaCl−0.01% Triton X-100. σM was further purified by chromatography on an FPLC Superdex-75 column (Amersham biosciences) in TEDG buffer containing 0.15 M NaCl followed by dialysis into TEDG−0.1 M NaCl−0.01% (v/v) Triton X-100–50% (v/v) glycerol and storage at −80°C.
Run-off transcription followed by microarray analysis
Run-off transcription followed by microarray analysis experiments were performed as described previously (Cao et al., 2002a; Cao and Helmann, 2004) with modifications. Purified σM was added in 20-fold molar excess relative to the purified RNA polymerase (mostly core with some contaminating σA). With the σM autoregulated promoter as a template, the specificity of σM-dependent transcription was optimal between 50 and 100 mM KCl (data not shown). For ROMA experiments, 100 mM KCl (final concentration) was used. Reactions (50 μl) contained 1 μg B. subtilis strain CU1065 genomic DNA which was sheared by vortexing, 0.1 μM RNAP with and without 2 μM σM, 1.25 μM δ factor mixed in transcription buffer (180 mM Tris-HCl (pH 8.0), 100 mM MgCl2, 100 mM NaCl, 100 mM KCl, 10 mM DTT, 100 μg ml−1 BSA, 50% (v/v) glycerol, 40 units RNasin (Invitrogen) and NTP mixture (800 μM ATP, GTP, CTP and UTP). RNAP, σM and δ were mixed on ice for at least 15 min to form holoenzyme before the addition of template DNA. Incubation at 37°C was continued for 10 min to allow the binding of the holoenzyme to promoter regions. The NTPs were then added to the reaction mixture and transcription was allowed to proceed for 20 min at 37°C.
Reactions were terminated by addition of 200 μl stop solution (2.5 M NH4OAc, 10 mM EDTA, 15 μg ml−1 linear polyacrylamide), extracted with phenol/chloroform, and precipitated with ethanol. The RNA pellets were dissolved in 20 μl DEPC-treated water and treated with TURBO DNA-freeTM DNase and removal reagents (Ambion) to remove DNA templates. For each experiment, a control reaction was performed in which σM which was omitted. The cDNA synthesis, cDNA labelling and hybridization were performed as described for microarray analysis. Two or three replicate experiments were performed.
Consensus search procedures and computer analyses
All promoter consensus search protocols were performed using the SubtiList graphical interface http://genolist.pasteur.fr/SubtiList/) and the pattern search algorithm. This site allows the search to be performed on the whole genome, or limited to regions within a defined distance of annotated open reading frames. In addition, specific invariant or degenerate bases can be designated.
Total RNA was isolated from strain CU1065 which was induced with vancomycin as described above. Two micrograms of total RNA was used as a template for reverse transcription using MultiscribeTM Revese transcriptase (Taqman, Roche) and a gene-specific primer (GSP1). The resulting cDNA was purified using a gel extraction purification kit (Qiagen) and a poly(dC) tail added at the 3′-end with terminal deoxynucleotidyl transferase (New England Biolabs). The resulting cDNA was amplified by PCR using a poly dG primer to anneal at the poly(dC) tail and a second gene-specific primer (GSP2), complementary to a region upstream of the GSP1 primer. PCR products were separated by gel electrophoresis and sequenced.
Putative promoter regions were amplified from B. subtilis chromosomal DNA using a forward primer (∼100 bp upstream of the −35 consensus) with restriction site HindIII and a reverse primer (typically ∼50 bp downstream of the start codon) with restriction site BamHI (Table S2). The resulting fragments were digested with HindIII and BamHI and cloned into pJPM122 (Slack et al., 1991) and verified by DNA sequencing. Promoter fusions were introduced into the SPβ prophage by a double-crossover event, in which each pJPM122 derivative was linearized with ScaI and transformed into B. subtilis strain ZB307A with selection for neomycin resistance. The SPβ lysates were prepared by heat induction and used to transduce CU1065, HB0031 and HB4715.
To test the promoter induction, CU1065, HB0031 and HB4715 strains containing promoter fusions were grown overnight in LB medium containing appropriate antibiotics and diluted 1:100 into 5 ml of LB medium. Each culture was grown until an OD600∼0.4, the cultures were induced by adding of 2 μg ml−1 vancomycin (final concentration). The cultures were incubated for additional 30 min at 37°C then 1 ml of uninduced and induced samples were collected by centrifugation. The β-galactosidase activity of each sample was measured according to Miller (1972).
This work was supported by a grant from the National Institutes of Health (GM-047446).