Novel Mycobacterium tuberculosis anti-σ factor antagonists control σF activity by distinct mechanisms



The aetiological agent of tuberculosis, Mycobacterium tuberculosis, encodes 13 σ factors, as well as several putative anti-, and anti-anti- σ factors. Here we show that a σ factor that has been previously shown to be involved in virulence and persistence processes, σF, can be specifically inhibited by the anti-σ factor UsfX. Importantly, the inhibitory activity of UsfX, in turn, can be negatively regulated by two novel anti-anti-σ factors. The first anti-anti-σ factor seems to be regulated by redox potential, and the second may be regulated by phosphorylation as it is rendered non-functional by the introduction of a mutation that is believed to mimic phosphorylation of the anti-anti-σ factor. These results suggest that σF activity might be post-translationally modulated by at least two distinct pathways in response to different possible physiological cues, the outcome being consistent with the bacteria's ability to adapt to diverse host environments during disease progression, latency and reactivation.


Tuberculosis is the leading cause of death world-wide from any known pathogen. It is estimated that 1.86 billion people are infected by Mycobacterium tuberculosis, the causal agent of tuberculosis (Glickman and Jacobs, 2001). Despite the medical importance of this human pathogen, very little is known about the molecular mechanisms governing the control of gene activity, which, in turn, dictate the proper gene expression profiles required for infection and persistence in the human host. Hence, better knowledge of these mechanisms would provide new therapeutic targets and greatly facilitate the design of improved treatments for tuberculosis.

Core bacterial RNA polymerase (RNAP) is composed of β, β′, ω and an α dimer (Zhang et al., 1999; Ghosh et al., 2001). RNA polymerase is rendered competent for transcription initiation when a fifth subunit, the σ factor, associates with the latter to form the holoenzyme. Hence, the σ factor provides most of the promoter's recognition function of the enzyme (Busby and Ebright, 1994; Gross et al., 1998; Kuznedelov et al., 2002). Thus, modulation of σ factors’ expression or regulation of their activity, in combination with their replacement on core RNAP, can give rise to transcription of different sets of genes, named regulons, in response to various physiological and environmental stimuli. Annotation of the complete genomic sequence of M. tuberculosis has revealed the existence of 13 putative σ factors (Cole et al., 1998), where σA has been shown to be the principal σ factor (Gomez et al., 1998). Interestingly, a point mutation in the latter has also been shown to cause loss of virulence in Mycobacterium bovis (Collins et al., 1995).

σF, an alternate σ factor, has been shown to be involved in the resistance to various stresses in M. bovis (DeMaio et al., 1996; Michele et al., 1999). In some cases, σF of M. tuberculosis has been shown to be necessary for the expression of virulence and adaptation gene products of importance for the pathogen's infection and dormancy processes (Chen et al., 2000a; Manganelli et al., 2001). Significantly, deletion of sigF in M. tuberculosis did not affect short-term intracellular survival of the bacteria in mice, but impaired long-term survival in this host (Chen et al., 2000a, b), suggesting that σF is important for survival and proliferation of the pathogen in lung granulomas during infection (Chen et al., 2000b). Importantly, sigF expression has been shown to be induced when M. tuberculosis infected cultured human macrophages (Graham and Clark-Curtiss, 1999).

One common mechanism by which σ factors are regulated is inhibition by proteins referred to as anti-σ factors. These proteins are believed to be specific for a particular σ factor and, in some bacteria such as Bacillus subtilis, they are themselves negatively regulated by specific molecules termed anti-anti-σ factors (or anti-σ factor antagonists). In B. subtilis, two σ factor regulons have been well described as being regulated in such a fashion. σB is a general stress-response transcription factor and σF is the first spore-specific factor to be activated in the sporulation process. Both σB and σF are bound by their cognate anti-σ factors, respectively, named RsbW and SpoIIAB, when their activity is not required. Therefore, the genes under their control are not expressed (reviewed by Helmann, 1999). To release σB and σF and to induce the expression of their regulons, the anti-anti-σ factors RsbV and SpoIIAA must bind to their corresponding anti-σ factor. Moreover, RsbW and SpoIIAB are also serine kinases that can phosphorylate their anti-anti-σ factor, the outcome leading to the inactivation of the latter (Min et al., 1993; Dufour and Haldenwang, 1994; Najafi et al., 1995). These three partner-switching modules are therefore very subtly regulated by phosphorylation, depending on local concentrations of ATP and/or ADP. A cycling process occurs between the phosphorylated and unphosphorylated form of SpoIIAA, depending on the activities of the SpoIIAB kinase and SpoIIE phosphatase (Magnin et al., 1997). It has been shown that ATP greatly favours the binding of SpoIIAB to σF (Alper et al., 1994; Duncan et al., 1996). According to the induced release model (Duncan et al., 1996), a free molecule of SpoIIAA reacts with the SpoIIAB-σF–ATP complex, which results in the release of σF, the phosphorylation of SpoIIAA and the formation of a SpoIIAB–ADP complex. The latter interacts with another molecule of SpoIIAA to form a stable SpoIIAA–SpoIIAB-ADP post-phosphorylation complex. Detailed studies of the phosphorylation step led to the conclusion that the post-phosphorylation complex includes a structurally rearranged form of SpoIIAB (designed as SpoIIAB*) (Lee et al., 2000) and that the very slow relaxation of SpoIIAB* to its usual form accounts for the long half-life of the post-phosphorylation complex (Magnin et al., 1997; Lee et al., 2000). As the intracellular concentrations of SpoIIAB and σF are almost equimolar, sequestration of SpoIIAB in long-lived complexes results in σF activation for transcription.

In sharp contrast, ATP does not seem to stimulate the binding of RsbW to σB, but it causes the inactivation of RsbV by phosphorylation. In this case, ADP inhibits RsbV phosphorylation and thereby enhances the formation of the RsbV–RsbW complex, the outcome enabling the release of σB (Alper et al., 1996).

As the M. tuberculosisσF factor could presumably control the expression of genes important for pathogenesis, we set out to define its regulatory circuit using in vitro transcription and protein–protein interaction assays. We show that σF can be negatively regulated by the UsfX anti-σ factor and that it is able to physically interact with the latter. We also identify and characterize, two novel anti-anti-σ factors that can negatively regulate UsfX activity. Finally, we show that these anti-σ factor antagonists possess novel properties that could allow σF to respond to various physiological conditions and/or stresses.


σA and σF holoenzymes and target promoters: design of a specific mycobacterial in vitro transcription system

To study transcription initiation by a σF RNAP holoenzyme and to assess its promoter specificity, we set up a transcription system using recombinant M. tuberculosisσ factors (σF and the primary σ factor σA) and purified core RNAP from Mycobacterium smegmatis. Because core RNAP subunits are very well conserved in prokaryotes, even more so in mycobacteria (90%, 91% and 91% of sequence identity for α, β, and β′, respectively, between M. tuberculosis H37Rv and M. smegmatis mc2155), and because of the relative ease at which M. smegmatis can be grown, we have chosen this species as a source of core RNAP. Core RNAP from M. smegmatis was purified to apparent homogeneity (Fig. 1A; left panel), as were the recombinant σA and σF factors obtained by overexpression in Escherichia coli (Fig. 1A; right panel). Two promoter templates were used in our experiments, the first one consisted of the P3 promoter of the Bacillus subtilis sin operon (sinP3; see Fig. 1B, upper panel). That promoter was previously shown to be efficiently transcribed by a purified M. smegmatis RNAP σA-holoenzyme (Predich et al., 1995). The second DNA fragment was that of the M. tuberculosis usf (upstream of sigma F) chromosomal region that contains usfY (Rv3288c), usfX (Rv3287c), and the first half of sigF (Rv3286c), the σF-encoding gene (Fig. 1C; upper panel). This panel denotes possible promoter-containing fragments within this gene cluster. The dotted rectangle indicates a possible intergenic region as suggested previously (DeMaio et al., 1997).

Figure 1.

A mycobacterial in vitro transcription system.

A. Left panel: purification of Mycobacterium smegmatis core RNA polymerase (RNAP). The proteins recovered after each chromatographic step are shown on a 10% SDS-PAGE. Each lane shows the peak fraction of the successive columns as described in the Experimental procedures . The expected core RNAP subunits are indicated as α, β and β′. Right panel: purified recombinant σ A and σ F .

B. Specific transcription initiation by a σ A holoenzyme. Each transcription reaction was carried out with 0.6 pmol of core RNAP and 0.4 pmol of the Bacillus subtilis sin template. No recombinant σ factor (lane 1) or increasing amounts of σ A (lane 2–7) or σ F (lane 8–13) were added to core RNAP and DNA template. The sin operon bears two promoters, P1 and P3 and two terminators T1 and T3 downstream of sinR (Gaur et al., 1988). The template used (Predich et al., 1995), has a deletion as pictured above the gel panel. Transcripts were directly visualized by a run-off assay.

C. Specific transcription initiation by a σ F holoenzyme. Each transcription reaction was carried out with 0.6 pmol of core RNAP and 0.4 pmol of the M. tuberculosis usf template. The template contains the usfY , usfX and the first half of the sigF open reading frames. No recombinant σ factor (lane 1) or increasing amounts of σ A (lanes 2–7) or σ F (lanes 8–12) were added to core RNAP and DNA template. Transcripts were visualized by primer extension.

We postulated that non-coding regions within the usf gene cluster might include promoter sequences recognized by σF, considering that many σ factors, as, for instance, σB in B. subtilis, can initiate transcription of their own structural genes (Hecker et al., 1996). Thus, to test the specificity of σA and σF for different promoters, we designed in vitro transcription reactions to obtain specific transcripts that can be detected either by directly incorporating a radiolabelled nucleotide (as in Fig. 1B), or by primer extension analyses (all figures except 1B). Figure 1B illustrates the in vitro transcription potential of σA and σF holoenzymes at the sinP3 template. Little transcription can be observed in the absence of any σ factor (Fig. 1B, lane 1), demonstrating that the purified core RNAP was virtually free of contaminating endogenous M. smegmatisσ factors that could recognize this promoter. However, addition of increasing σA concentrations (0.5–6 pmols) to reactions using the sinP3 template (Fig. 1B, lanes 2–7) resulted in the appearance of increasing amounts of two transcripts (180 and 140 nucleotides). These two RNAs were of the expected size as previously described and resulted from transcriptional termination at sites T1 and T3 of the sinP3 promoter (Predich et al., 1995; see also Fig. 1B, upper panel). These results confirm those previously obtained with E. coli, B. subtilis and M. smegmatis RNAPs, i.e, that sinP3 is transcribed by primary σ-factor containing holoenzymes (Predich et al., 1995; Shafikhani et al., 2001; M. Gomez, G. Nair and I. Smith, unpublished). However, σF was unable to promote specific transcription initiation at sinP3, even at high protein concentrations (Fig. 1B, lanes 8–13). On the other hand, opposite results were obtained when similar transcription reactions were carried out with the usf template. Figure 1(C) shows that titration of σF (0.5–6 pmols) could efficiently induce transcription at that template, in a concentration-dependent manner (lanes 8–13) whereas σA, even at high protein levels, could not (Fig. 1C, lanes 2–7). Interestingly, we noticed that the primer extension product at the usf template, because of its small size, could not result from a transcript initiated in the DNA region upstream of usfY. Instead, the cDNA size corresponded to the 5′-region immediately upstream of the usfX putative translation start site. Hence, we denoted the newly identified σF-dependent promoter in this region as ‘usfXP1’.

Precise identification of a σF-dependent promoter upstream of the usfX open reading frame

To localize a possible σF-dependent promoter upstream of usfX, we used a template that had a 279 bp deletion (denoted as ‘D’ in Fig. 2A) immediately after the stop codon of the usfY locus. As already observed, σA was not able to initiate transcription at a wild-type (WT) template (Fig. 2A, lane 2), whereas σF could (lane 3). However, using the D template, σF-dependent transcription was completely abolished (lane 4).

Figure 2.

Identification of the usfX P1 transcription start site.

A. In vitro transcription of the usf template, followed by primer extension as in Fig. 1 , with no recombinant s factor (lane 1), 3 pmol of σ A (lane 2) or σ F (lane 3) added to core RNAP. A template bearing a Nhe I– Hin dIII deletion was also assayed (lane 4) with the σ F holoenzyme.

B. To confirm that the in vitro and in vivo start sites obtained were the same, primer extension was carried on usfXP1 in vitro transcription product (lane 5) and on independent RNA duplicates extracted from M. tuberculosis in log phase (OD540 0.5) (see duplicate lanes 6 and 7) and in stationary phase (OD540 1.0) (see duplicate lanes 8 and 9). The sequence shown (lanes 1–4 and 10–13) is from pYZ99, a plasmid bearing the genomic locus of the M. tuberculosis sigF operon. The panel under lanes 6–9 shows the amounts of rRNA used in the above primer extension reactions.

C. Sequence of the usfX P1 promoter region. The upper sequence line shows the −35 and −10 boxes (bold underlined) and the identified transcriptional start site (+1). The second sequence line denotes the B. subtilis σ B -dependent promoter consensus sequence (W = A or T).

Primer extension analyses were next performed to identify the in vitro transcription start site at the usfXP1 promoter by the σF holoenzyme. Figure 2B shows that the σF holoenzyme specifically initiates transcription at nucleotide C, located 126 basepairs upstream of the usfX open reading frame (ORF) (lane 5; see Fig. 2C for the +1 usfXP1 DNA region). To validate these in vitro transcription results with in vivo data, primer extension analyses were performed with RNA isolated from either logarithmic phase (OD540 = 0.5) or stationary phase (OD540 = 1.0) cultures of M. tuberculosis. These growth conditions were chosen in an attempt to induce σF expression in vivo, as it was previously shown that sigF was expressed at higher levels when M. bovis was grown to stationary phase (DeMaio et al., 1996). Figure 2B (lanes 6–9) shows that there is indeed a promoter that is used in vivo immediately upstream of usfX (Fig. 2B, lanes 6–9). Furthermore, the same transcription initiation start site was observed with in vivo-isolated M. tuberculosis RNA as with RNA transcribed in vitro using a σF holoenzyme with the usf template. However, we found that usfX transcription was not significantly induced during stationary phase (lanes 8 and 9) as compared with the logarithmic phase (lanes 6 and 7). The apparent intensity increase in stationary phase corresponded to the increased RNA amount loaded in the reactions (see rRNA panel under Fig. 2B). Examination of the DNA sequence in the transcriptional start site region of the σF-dependent promoter identified a B. subtilis consensus σB promoter sequence (Petersohn et al., 1999; Chen et al., 2000b) ending 9 basepairs from the usfXP1 +1 nucleotide (Fig. 2C).

UsfX is a M. tuberculosis anti-σF factor

The action of anti-σ factors provide a very efficient way to negatively control the activity of σ factors (Hughes and Mathee, 1998). As discussed above, the derived amino acid sequence of UsfX, as well as the chromosomal location of its structural gene resembled those of RsbW, the anti-σB of B. subtilis (DeMaio et al., 1997). Hence, we wanted to see whether UsfX had a similar function as an anti-σF factor. We first wanted to test if UsfX could inhibit σF-specific transcription. For these studies, we needed to purify UsfX. The DNA sequence of the usfX gene predicts four possible ORFs, which are all in the same translational frame (DeMaio et al., 1997). Thus, four potential proteins consisting of 145, 168, 208 and 242 amino acids have been deduced from the sequence, all with a common C-terminal segment. We cloned, expressed and purified three of these potential proteins to near homogeneity (Fig. 3A). Figure 3B shows that adding an increasing amount of the 145-amino-acid recombinant UsfX derivative to a fixed level of σF inhibits transcription initiated by the σF holoenzyme at the usf template in a dose-dependent manner (Fig. 3B, lanes 4–6). Experiments with the longer UsfX derivative (242 aa; Fig. 3A, lane 2) also gave similar results (data not shown) although usfXP1 is part of the coding region of that derivative. On the other hand, the addition of UsfX had no effect on σA transcription at the sinP3 promoter even at excess molar concentrations of UsfX (Fig. 3C, lanes 4–6).

Figure 3.

UsfX is an anti-σF factor.

A. Shows a 12.5% SDS-PAGE stained with Coomassie brilliant blue. Three forms of UsfX are shown, respectively, of 242 (lane 2), 208 (lane 3) and 145 aa (lane 4). The dot indicates the 145 aa form used in the following experiments.

B. Transcription at the sin template. Core RNAP alone (lane 1), or either with 3 pmol of UsfX (lane 2) or 0.6 pmol of recombinant σ A (lane 3) added. Increasing amounts (0.6, 3 and 6 pmol) of UsfX (lanes 4–6) were added to σ A , core RNAP and DNA template. Transcripts were visualized by primer extension.

C. Identical procedure undertaken on the usf DNA template. Core RNAP alone (lane 1), or with again either 3 pmol of UsfX (lane 2) or 3 pmol of recombinant σ F (lane 3) added. Increasing amounts (1.5, 3 and 6 pmol) of UsfX (lanes 4–6) were added to σ F , core RNAP and DNA template.

D. Physical interaction of UsfX with σ F . Interaction assays were carried out with GST fusions as indicated above the panels and 35 S-labelled σ A and σ F factors, and as described in Experimental procedures . L, load (10% total); S, supernatant; P, pellet. A signal in the P lane denotes an interaction.

To further validate the role of UsfX as an anti-σF specific factor, we carried out protein–protein interaction assays to test whether UsfX could physically interact with σF and/or σA. For this purpose, the 145-aa-long UsfX derivative was fused to glutathione-S-transferase (GST) and the resulting fusion protein was used in protein–protein interaction experiments. Figure 3D shows that GST itself cannot interact with radiolabelled σA (lanes 1–3, compare load lane, L, to pellet lane, P), and that GST-UsfX could also not directly interact with σA (lanes 4–6). However, GST-UsfX is able to strongly interact with σF (lanes 10–12). In control experiments, GST was unable to bind to σF (lanes 9–11). These findings indicate that the 145 aa UsfX derivative is sufficient to mediate the interaction with σF, behaving as a functional anti-σF factor. This shorter 145 aa UsfX form will be used in all subsequent experiments. Taken together, these results suggest that UsfX is a σF-specific anti-σ factor, and they further demonstrate that transcription initiation in vitro at the usf template by σF holoenzymes is specific.

A search for UsfX anti-anti-σ factors

Previous sequence homology studies showed that the σF factor of M. tuberculosis is relatively similar, at the primary sequence level, to the σB and σF factors of B. subtilis, and that UsfX is similar to RsbW, the anti-σB factor of B. subtilis (DeMaio et al., 1997). To further extend the network of protein–protein interactions that regulate the activity of σF in M. tuberculosis, and to include putative anti-anti-σF regulators, we considered as possible candidates the orthologues of RsbV and SpoIIAA, the antagonists of RsbW and SpoIIAB in B. subtilis. The occurrence of such orthologues in M. tuberculosis was recently suggested (Kovacs et al., 2001). The COG database (Tatusov et al., 2000) provides a convenient tool for retrieving such orthologues. RsbW and SpoIIAB belong to cluster COG1366, which also includes seven members from M. tuberculosis. Among these, the Rv1365c protein was the most closely related to SpoIIAA and the Rv3687c protein to RsbV. A clustalw alignment of the primary sequences of these four proteins is shown in Fig. 4A, and Fig. 4B displays the pairwise distances between primary sequences, as calculated by clustalw. As both mycobacterial proteins showed a similar relatedness to their bacillar counterparts, we examined them both as possible anti-σF antagonists.

Figure 4.

Search for UsfX antagonists (anti-anti-σ F factors). (A) clustalw alignment of amino acid sequences of putative anti-anti-σ factors Rv1365c and Rv3687c (NCBI loci D70742 and H70791) translated from the complete genome sequence of M. tuberculosis strain H37Rv, with Spo IIAA and RsbV proteins from B. subtilis (NCBI loci NP_390228 and NP_388353). The downward arrows indicate residues highlighted in Fig. 4C . ‘*’, indicates residues conserved in all four proteins.

B. clustalw distances calculated by pairwise alignment of the amino acid sequences of anti-anti-σ factors.

C. Three-dimensional model of Rv1365c protein (residues 14–128) obtained by homology modelling. Residues C73 and C109, which are possibly involved in a disulphide bond, are shown as balls and sticks. Residues V76 and I105, homologous to residues L60 and L89 of Spo IIAA, which are involved in anti-σ factor binding, are shown in spacefill format.

The alignment of Fig. 4A reveals an important difference between Rv1365c and the other three proteins: Rv1365c has a cysteine (C73) in a position corresponding to the serine known to be phosphorylated in SpoIIAA (Najafi et al., 1995; first downward arrow in Fig. 4A). Based on the solution structure of SpoIIAA determined by Kovacs et al. (2001), we obtained, by homology modelling (Bates and Sternberg, 1999), an almost full-length model of the Rv1365c protein (Fig. 4C). This model reveals that C73 and C109 form a cysteine pair well positioned to eventually interact and create a disulphide bond. Such a bond would decrease the accessibility to the hydrophobic amino acids V76 and I105 (second and third downward arrow on Fig. 3A), localized in positions corresponding to residues L61 and L90 of SpoIIAA that were shown to be involved in complex formation with SpoIIAB (Lee et al., 2001). All this suggested that cysteines 73 and 109 could form a redox sensor regulating the activity of the Rv1365c protein.

Rv1365c encodes an anti-anti-σF factor that is regulated by redox potential

To verify the prediction that Rv1365c could be an anti-anti-σF factor regulated by redox potential, we cloned, expressed and purified this putative anti-anti-σ factor (Fig. 5A, lane 4). We also generated point mutations to directly test if the predicted cysteines (C73 and C109) could be involved in this redox switch. Cysteines were replaced by alanines, creating two mutants: C73A and C109A, which we predicted would constitutively bind to GST-UsfX. Proteins were then expressed in E. coli and affinity purified (see Fig. 5B, lane 4 for C73A; similar results were obtained with C109A, data not shown). Protein–protein interaction assays were next undertaken under various redox potential conditions. As we expect the nature of the interactions to be near stoichiometric, we carried out GST pull-down experiments between the purified proteins without any labelling. First of all, no interaction has been detected between GST and WT Rv1365c or mutant derivatives in the presence or absence of 4 mM DTT, or at 2 mM H2O2 (Fig. 5A lane 7 and data not shown). However, Fig. 5A shows a clear interaction between GST-UsfX and WT Rv1365c in a buffer containing 4 mM DTT (compare lane 7 and 10). Figure 5B and D shows that WT Rv1365c can interact with UsfX only under reducing conditions whereas the C73A mutant retains its ability to interact with UsfX even in the presence of 2 mM H2O2 (compare lanes 7 and 10 of Fig. 5B and D). Only a weak interaction was observed between GST-UsfX and Rv1365c in a buffer containing no DTT (trace amounts of DTT were nevertheless present in our protein preparations) whereas there is no effect on the C73A mutant as compared with different redox potential buffers (Fig. 5C and D). The same effects were observed with the C109A mutant (data not shown). Moreover, cysteine to serine mutants were also prepared and gave identical results to the alanine mutants under all conditions tested (data not shown). Using β-mercaptoethanol instead of DTT in the binding buffer also produced similar effects (data not shown). Taken together, these results support the prediction that there is indeed a cysteine-based motif (most probably a disulphide bridge) in Rv1365c that needs to be reduced to interact with UsfX.

Figure 5.

Rv1365c encodes RsfA, a redox potential-dependent anti-anti-σ F factor. Interaction assays were carried out with GST or GST-UsfX and Aasf1. The panels show 15% SDS-PAGE stained with Coomassie brilliant blue. L, load; S, supernatant; P, pellet. Lane 1 displays the molecular weight markers, whereas lane 2–4 show the different proteins alone as migration control.

A. GST alone (lane 5–7) and GST-UsfX (lane 8–10) along with WT RsfA at 4 mM DTT.

B. Interaction of GST-UsfX and the WT RsfA (lane 5–7) versus a C73A mutant RsfA (lane 8–10) at 4 mM DTT.

C. The same as (B) but with no added DTT in the assay buffer.

D. Again, the same as (B) but with 2 mM H 2O2.

E. The effect of RsfA was assessed in the in vitro transcription assay. Each transcription reaction was carried out at the usf template as described in Fig. 1 , either with 1 mM DTT (lanes 1–7) or 1 mM H 2O2 (lanes 8–14). No recombinant σF (lanes 1 and 8), 3 pmol of σF alone (lanes 2 and 9) or 3 pmol of σF with 3 pmol of UsfX (lanes 3 and 10) were added. Increasing amounts (3 and 12 pmols) of WT RsfA (lanes 4, 5 and 11, 12) or C73A mutant (lanes 6, 7 and 13, 14) were added to reactions.

Next, we assayed the effect of WT or mutant Rv1365c derivatives on in vitro transcription at the usf template. The reactions contained σF either without UsfX or with an equimolar amount of UsfX (relative to σF) and increasing amounts of Rv1365c derivatives. The left panel shows an experiment carried out in the presence of 1 mM DTT. The figure shows that both WT Rv1365c and the C73A mutant can relieve the inhibitory effect of UsfX in the presence of DTT (Fig. 5E, lanes 4–7). The right panel shows that in the absence of DTT but, in the presence of 1 mM H2O2, only the C73A mutant was able to antagonize UsfX (Fig. 5E, lanes 11–14). Similar results were obtained with the C109A mutant (data not shown). We observed that at lower anti-anti-σ concentration (e.g. 3 pmols), the WT molecule was not as efficient as the C73A mutant (Fig. 5E; compare lanes 4 and 6). Possibly, 1 mM DTT may not be sufficient to completely reduce the WT molecule under our buffer conditions, and we also found that higher concentrations of DTT (4 mM) affected in vitro transcription reactions so they could not be used for this assay. Our in vitro transcription experiments thus support those obtained in protein–protein interaction assays. Taken together, our results strongly suggest that Rv1365c is an anti-anti-σ factor that is antagonistic to UsfX anti-σ factor only under reducing conditions. We therefore named the product of the Rv1365c ORF RsfA (regulator of sigma F A).

Rv3687c encodes an anti-anti-σF factor that may be regulated by phosphorylation

To test the possibility that a second anti-σF antagonist, Rv3687c, could reverse the inhibitory effect of UsfX at usfXP1, we cloned, expressed and purified the candidate anti-anti-σ factor (Fig. 6A, lane 4). We first tested the ability of WT Rv3687c to interact with UsfX in protein–protein interaction assays as described above. Figure 6A thus further shows that WT Rv3687c can efficiently interact with GST-UsfX but not with GST alone (compare lanes 7–10). In experiments not shown here, Rv3687c was also used as a control for redox-dependence experiments, as described previously with RsfA. Rv3687c was able to efficiently bind to GST-UsfX, either at 4 mM DTT or at 2 mM H2O2 (data not shown).

Figure 6.

Rv3687c encodes RsfB, an anti-anti-σ F factor.

A. Interaction assays were carried out with GST alone or GST-UsfX and RsfB as in Fig. 5 . Lane 1 displays molecular weight markers, whereas lane 2–4 show the different proteins alone as migration control. GST alone (lane 5–7) and GST-UsfX (lane 8–10) were mixed with WT RsfB.

B. Interaction assays with GST-UsfX and WT RsfB (lane 5–7) as compared with the S61A mutant (lane 8–10) and (C) the S61E mutant (lane 8–10).

D. The effect of RsfB was assessed with the in vitro transcription assay at the usf template. No σ F (lane 1), 3 pmol of σ F alone (lane 2) or 3 pmol of σ F with 3 pmol of UsfX (lane 3) were added. Increasing amounts (3 and 12 pmols) of either WT RsfB (lane 4–6), S61A mutant (lane 7–9) or S61E mutant (lane 10–12) was added to σ F , UsfX, core RNAP and DNA template.

Rv3687c has a conserved serine at residue 61 which is phosphorylated in the B. subtilis orthologous proteins SpoIIAA and RsbV (Najafi et al., 1995). As mentioned previously, phosphorylation of this serine residue drastically reduces binding of the anti-anti-σ factor to its cognate anti-σ factor SpoIIAA (Alper et al., 1994; 1996; Duncan et al., 1996). Thus, to test if S61, and hence perhaps phosphorylation, was important for the activity of Rv3687c, we mutated serine 61 to either an alanine (S61A) or a glutamic acid (S61E). The latter mutation was created to mimic phosphorylation because of the negatively charged side chain. Both mutants were tested for interaction with UsfX without ATP. Figure 6B shows that the Rv3687c S61A mutant is still able to interact with GST-UsfX (compare lanes 7 and 10). Figure 6C shows that binding of the Rv3687c S61E mutant to GST-UsfX is significantly impaired as compared with WT Rv3687c (compare lanes 7–10).

We next tested the ability of WT Rv3687c to inhibit UsfX in σF-specific in vitro transcription at the usf template. The reactions contained σF either without UsfX (Fig. 6D, lane 2) or with an equimolar amount of UsfX (Fig. 6D, lane 3) and increasing amounts of the Rv3687c derivatives. The antagonistic activity of Rv3687c on UsfX-dependent inhibition of σF-directed transcription was evident with WT Rv3687c (Fig. 6D, lanes 4–5) and with the S61A mutant (Fig. 6D, lanes 6–7), but not with the S61E mutant that is believed to mimic phosphorylation (Fig. 6D, lanes 8–9). Taken together, these results strongly suggest that Rv3687c encodes a UsfX antagonist thus named RsfB. Our results further suggest that mimicking phosphorylation can drastically reduce RsfB's potential to function as an anti-anti-σ factor.


Using a highly purified mycobacterial in vitro transcription system, we first identified a M. tuberculosisσF-specific promoter that strongly resembles a B. subtilisσB consensus sequence. That promoter, termed usfXP1, lies upstream of the usfX and sigF genes which are, as previously suggested, most probably co-transcribed (DeMaio et al., 1997). We chose this ‘sigF operon’ as a potential σF target gene as it is common to find that σ factors and other transcription regulatory proteins control the expression of their own genes (Hecker et al., 1996; Fernandes et al., 1999; Raman et al., 2001). Importantly, a σA mycobacterial holoenzyme bearing σA from M. tuberculosis, could not initiate transcription at usfXP1, a result validating the specificity of our transcription system. In addition, −10 and −35 consensus promoter sequences resembling those recognized by σA were not observed in the usf promoter region (see also DeMaio et al., 1997). On the other hand, we observed that σF holoenzymes could not initiate transcription of a promoter template shown to be transcribed by the σA holoenzyme. Our results thus show that σF has a completely different promoter requirement than the σA principal factor.

The organization of the M. tuberculosis sigF cluster revealed three genes, usfY, usfX and sigF. The usfY gene product has no homology to any protein in the database, but because of its arrangement in the cluster, it has been proposed to be an anti-anti-σ factor (DeMaio et al., 1997). UsfX, on the other hand, has significant homology to the B. subtilis anti-σ factor RsbW, and its structural gene is also arranged in the same position in the gene cluster as in the B. subtilis sigB operon (DeMaio et al., 1997). Moreover rsbV, rsbW and sigB are all arrayed in tandem in the same operon, and furthermore, they are all under the control of σB itself. In the case of the M. tuberculosis usf region, we have found a σF-specific promoter between usfY and usfX, but not upstream of usfY (data not shown). It is clearly possible that another promoter, not responsive to σA or σF, might be upstream of usfY to drive the basal expression of at least the latter gene, or maybe of the entire operon. Based on the facts described above, we suggest that UsfY might not be an anti-anti-σF factor, although this last possibility would have to be tested explicitly.

In support of the prediction that UsfX could be an anti-σF factor is the fact that we have first been able to show a strong physical interaction between UsfX and σF. As other σ factors from M. tuberculosis such as σA, σE or σH could not interact with UsfX in our assays (Fig. 3D and data not shown), the interaction between UsfX and σF is indeed a specific one. In agreement with these observations, we have shown that purified UsfX can completely inhibit σF-dependent transcription in vitro, but has no effect on transcription initiation catalysed by the σA holoenzyme. These two findings thus support a role for UsfX as a bona fide anti-σF factor.

To find upstream regulators of UsfX, we set out to search the M. tuberculosis genome for orthologues of RsbV and SpoIIAA, the B. subtilis anti-anti-σB and -σF factors. Two putative anti-anti-σ factors were identified. We have shown that both these putative anti-anti-σ factors could bind and reverse the inhibition effect of UsfX at usfXP1. We suggest that RsfA, encoded by the Rv1365c gene, can be regulated by redox potential as it can only exert its effect under reducing conditions. A Blast-directed search performed on 138 eubacterial genomes (NCBI database, March 2002) for proteins potentially displaying properties similar to RsfA (recognized by their homology to RsfA and the presence of both functionally important cysteine residues) allowed us to detect such protein sequences only in M. tuberculosis (strains H37Rv, 210 and CDC1551), M. bovis, Mycobacterium avium, Mycobacterium avium ssp. paratuberculosis but neither in M. smegmatis nor in Mycobacterium leprae or any other eubacterium. So far, RsfA-like proteins seem to be present only in mycobacteria causing various forms of tuberculosis.

Mutations of critical cysteine residues in RsfA render the action of the molecule constitutive, i.e. allow σF-dependent transcriptional activity irrespective of the redox potential. We entertain two possible scenarios, possibly overlapping: in the first, this anti-anti-σ factor could be constitutively active in the cytoplasm if it is under reducing conditions. In this case, oxidative stress would shut down σF-specific transcription, and perhaps give way to transcription initiation by other σ factors that respond to such stresses such as σH in M. tuberculosis (Manganelli et al., 1999; Raman et al., 2001) and σR in Streptomyces coelicolor (Paget et al., 2001). In the second scenario, the anti-anti-σ factor could be important for the bacteria's adaptation to hypoxic conditions. It is well documented that during infection, and especially persistence of M. tuberculosis by dormancy in a human host, the tubercle bacilli has to adapt to a low oxygen environment (reviewed by Wayne and Sohaskey, 2001). The level of the sigF gene transcript itself has shown no significant variation under low aeration conditions (24 h incubation at 37°C without agitation) (Manganelli et al., 1999), but a σFlacZ translational fusion has shown an important β-galactosidase activity increase when the bacteria were cultured by the slow-stirring method of Wayne and Hayes (Wayne and Hayes, 1996; Michele et al., 1999). On the other hand, metronidazole has been shown to be active only against anaerobic dormant tubercle bacilli. As this drug exerts its bactericidal effects only after being activated by a redox potential below −430 mV (Wayne and Hayes, 1996), one can expect such a highly reductive potential in the bacterial cytoplasm during dormancy. This could be in agreement with our model of σF release by the reduction-dependent activation of RsfA. These observations are also compatible with a mechanism of σF-directed transcription regulation in which liberation of existing σF protein from its interaction with an anti-σ factor is more important than de novoσF synthesis through increased transcription of its own gene.

RsfB, the second anti-anti-σ factor, is not affected by redox potential, but rather our results suggest that phosphorylation might regulate the latter molecule, like its closest B. subtilis orthologue, RsbV. Unlike the Bacillus case, our findings may suggest that UsfX is not a RsfB kinase or, alternatively, our buffer conditions does not allow for phosphorylation. We disfavour the last possibility as we have been able to observe autophosphorylation of UsfX under similar buffer conditions (data not shown). Therefore, it seems likely that another kinase would regulate RsfB activity.

A recent report (Betts et al., 2002) has shown that the expression of sigF, usfX and four other genes localized immediately upstream (Rv3288c − Rv3291c) is strongly upregulated in nutrient starvation conditions (post incubation in phosphate-buffered saline (PBS) without shaking of 7-day-old cultures grown in nutrient-rich media). Upregulation varied from 3× for sigF and 9× for usfX to almost 42× for Rv3290c, which was annotated in the Tuberculist database as a lysine-e-aminotransferase.

Our finding that two different anti-anti-σ factors can regulate the activity of a σF-specific anti-σ factor is novel and suggests a certain level of functional redundancy in this type of regulatory circuit in M. tuberculosis. The presence of two anti-σF antagonists could extend and ‘fine tune’ the range of physiological conditions in which σF would be partially or fully available for transcription as compared with a single-antagonist regulatory system (see Table 1).

Table 1. . Summary of proposed σ F regulation by anti-σ F factor antagonists.
by σF
High ATP/ADPInactiveInactiveNull
High redox
High ATP/ADPActiveInactivePartial or full
Low redox
Low ATP/ADPInactiveActivePartial or full
High redox
Low ATP/ADPActiveActiveFully active
Low redox

Work is in progress to assess the physiological role of Rsf molecules in M. tuberculosis.

Experimental procedures

M. smegmatis RNAP extraction and purification

Mycobacterium smegmatis cells were grown in 4 L of Luria–Bertani (LB) broth to late log phase (approximately 48–72 h, OD 600  = 1.5–2.0). The cultures were then centrifuged and the cell pellets, weighing approximately 30 g, were frozen at −80°C. Pellets were thawed and cells were disrupted using a French press at 18 000 lbs/sq. in. The lysate was then centrifuged and subjected to (NH 4 ) 2SO4 precipitation. The resulting cytoplasmic extract was aliquoted in 1 ml fractions, quick frozen in liquid nitrogen and kept at −80°C. Further details can be obtained upon request. For purification using an ÄKTA FPLC (Amersham-Bioscience), a 1 ml aliquot was thawed, diluted 1:10 with buffer C (50 mM Tris-Cl pH 8.0; 10 µM ZnSO4; 1 mM EDTA; 5 mM MgCl2; 5 mM β-mercaptoethanol and 20% glycerol) and loaded on a 5 ml HiTrap Heparin column (Amersham-Bioscience). The elution was carried out over five column volumes following a linear gradient of 0–1 M KCl in buffer C. The 1 ml fractions containing RNA polymerase (RNAP), as determined by SDS-PAGE, were pooled and diluted 1:10 with buffer C. This sample was next loaded on a P11 phosphocellulose (Wathman) column and eluted with four column volumes following the same gradient technique. The fractions showing typical protein bands of approximately 120 kDa on SDS-PAGE were again pooled and diluted 1:10 with Buffer C. RNAP was further purified using a Mono Q HR 5/5 (Amersham-Bioscience) column. Elution was carried out over 10 column volumes. The selected fractions, containing at least 95% of enriched core RNAP, as determined by SDS-PAGE, were then loaded onto a final 15 ml bed volume P11 column to eliminate possible endogenous contaminating σ factors. The peak fractions were then tested in in vitro transcription assays. Fractions showing no basal transcription activity on the sin template and large amounts of RNAP core subunits were diluted 1:1 in 50% glycerol, quick frozen in liquid nitrogen and kept at −80°C.

DNA manipulations and plasmid construction

DNA manipulations were carried out according to standard procedures. The DNA templates used for the in vitro transcription assays were a 798 bp BamHI–HindIII fragment from pIS109 (Predich et al., 1995) for the sin promoter (sinP3), a 1047 bp NcoI–EcoRI fragment from pYZ99 (DeMaio et al., 1996) for the usfXP1 promoter and a 768 bp NcoI–EcoRI fragment from pJPB32 (this study) for the ΔusfXP1 promoter. σA was cloned into pET22b (Novagen), σF and UsfX derivatives were cloned into pET16b (Novagen), Rv1365c and Rv3687 derivatives were cloned into pET30a (Novagen). All GST fusion constructs were cloned into pGEX6P-1 (Amersham-Bioscience). Details of plasmid constructions are available upon request.

Protein overexpression and purification

All His-tagged proteins were overexpressed in Escherichia coli BL21(DE3) pLysS, and purified under denaturing conditions according to the manufacturer's recommendations (Qiagen) using buffer C. σA was purified using a modification of the method previously published for B. subtilisσA (Chang and Doi, 1990). Briefly, the inclusion bodies containing the overexpressed proteins were solubilized in 8 M urea buffer C. The extract was loaded on a DEAE Sepharose Fast Flow column (Amersham Bioscience). The elution was carried out over five column volumes following a linear gradient of 0–1 M KCl in buffer C. Peak fractions, as determined by SDS-PAGE, were then further purified by size exclusion on a Superdex 200 column (Amersham-Bioscience). Glutathione-S-transferase fusion proteins were overexpressed in E. coli MM294 and purified under denaturing conditions.

In vitro transcription assays

The in vitro transcription protocol used in our assays was adapted from Predich et al. (1995). Briefly, reactions were carried out in a final volume of 40 µl containing 40 mM Tris-Cl (pH 8.0), 2 mM MgCl2, 75 mM KCl, 50 mM potassium glutamate, 0.1 mM EDTA, 0.1 mM dithiothreitol, 250 µg ml−1 of bovine serine albumin (BSA) and 10% glycerol. DNA template, σ factor (and/or anti-anti-σ factors) and core RNAP were mixed and incubated 30 min at 37°C in the transcription buffer. Transcription reactions were initiated at the same time by adding 2 µl of NTPs. The final NTP concentration were as follows: 0.15 mM ATP, UTP, GTP and 0.4 µM CTP. Immediately after transcription, 2 µl of a solution containing 6 µg of heparin, along with 0.15 mM of CTP, were added. Reactions were left for another 10 min at 37°C. Transcription reactions were then stopped by adding 2 µl of 7.5 M ammonium acetate, 1 µl of 5 mg ml−1 bacterial tRNA and 100 µl of ice-cold ethanol. RNAs were then precipitated for 15 min at −80°C and washed with 70% ethanol. The pellets were resuspended in 10 µl of water. Template specific oligonucleotides were radiolabelled using γ-AT32P and T4 polynucleotide kinase and used as probes for detection by primer extension (USFX-EXT2: TCTGGTCTTCGAGCTGGTCGGTCATGGTC and SINP3-EXT2: TTTCGATACCCCGGATGTCATCGCATCG) according to standard procedures.

Protein–protein interaction assays

When using radiolabelled σ factors in our experiments, either σA and σF were labelled with 35S-l-methionine, according to the manufacturer's instructions using the TNT in vitro coupled transcription–translation system (Promega). The interaction assays were carried out with 6 µg of GST-UsfX or GST alone and 10 µl of 35S-labelled σA or σF.

The interaction assay buffer consisted of 100 mM potassium acetate, 20 mM Hepes pH 7.5, 20% glycerol, 1 mM DTT, 1 mM EDTA pH 8.0, and 0.2% NP40. Samples were incubated for 3 h at 4°C on a rotator; 10% was removed and kept as input material for the assay. Samples were spun at 1500 r.p.m. The supernatant was removed and 25 µl was kept as supernatant. The samples were then washed four times with interaction assay buffer (using 150 mM potassium acetate) and analysed. Bands were visualized using a Molecular Dynamics Phosphor-Imager. For non-radioactive interaction assays, equimolar quantities of each protein partners ranging from 600 pM to 1.2 mM were used. Proteins were mixed in a final volume of 250 ml of pull-down buffer. Samples were incubated 30 min at 37°C on a rotator and treated as described above. After electrophoresis, gels were stained with Coomassie brilliant blue and directly visualized.

In silico studies

Protein sequences were compared and aligned using clustalw (version 1.8). Homology modelling was performed on the 3D-JIGSAW server (Bates and Sternberg, 1999) and atomic clashes removed from the model using the Whatif Web interface (Rodriguez et al., 1998). The model was visualized with RasMol.


We thank W. R. Bishai for the generous gift of pYZ99. This work was supported by funds from the Université de Sherbrooke to L.G. and R.B., and by NIH research grant AI-44856 awarded to I.S. L.G. is a research scholar from the CIHR/CRS Inc. J.B. and S.R. acknowledge support from NSERC studentships.