Promoter Recognition by a Complex of Spx and the C-Terminal Domain of the RNA Polymerase α Subunit

Background Spx, an ArsC (arsenate reductase) family member, is a global transcriptional regulator of the microbial stress response and is highly conserved amongst Gram-positive bacteria. Bacillus subtilis Spx protein exerts positive and negative control of transcription through its interaction with the C-terminal domain of the RNA polymerase (RNAP) α subunit (αCTD). Spx activates trxA (thioredoxin) and trxB (thioredoxin reductase) in response to thiol stress, and bears an N-terminal C10XXC13 redox disulfide center that is oxidized in active Spx. Methodology/Principal Findings The structure of mutant SpxC10S showed a change in the conformation of helix α4. Amino acid substitutions R60E and K62E within and adjacent to helix α4 conferred defects in Spx-activated transcription but not Spx-dependent repression. Electrophoretic mobility-shift assays showed αCTD interaction with trxB promoter DNA, but addition of Spx generated a supershifted complex that was disrupted in the presence of reductant (DTT). Interaction of αCTD/Spx complex with promoter DNA required the cis-acting elements -45AGCA-42 and -34AGCG-31 of the trxB promoter. The SpxG52R mutant, defective in αCTD binding, did not interact with the αCTD-trxB complex. SpxR60E not only failed to complex with αCTD-trxB, but also disrupted αCTD-trxB DNA interaction. Conclusions/Significance The results show that Spx and αCTD form a complex that recognizes the promoter DNA of an Spx-controlled gene. A conformational change during oxidation of Spx to the disulfide form likely alters the structure of Spx α helix α4, which contains residues that function in transcriptional activation and αCTD/Spx-promoter interaction. The results suggest that one of these residues, R60 of the α4 region of oxidized Spx, functions in αCTD/Spx-promoter contact but not in αCTD interaction.


Introduction
Spx is a member of the ArsC (arsenate reductase) family of proteins and is a unique transcriptional regulator of Gram-positive bacteria [1]. In Bacillus subtilis, its primary function is in the global control of the thiol stress response [2]. A role for Spx in pathogenesis has been inferred from studies of Staphylococcus aureus, Listeria monocytogenes, Bacillus anthracis, and Streptococcus mutans, in which it is produced upon host cell infection and serves to control the expression of virulence determinants [3,4,5,6,7]. B. subtilis Spx interacts with the C-terminal domain of the a subunit (aCTD) of RNA polymerase (RNAP), and in doing so, exerts both positive and negative control of gene transcription [2,8]. Genetic and biochemical studies showed that the G52 residue of Spx and the Y263 residue of aCTD form part of the contact interface between Spx and RNAP, which was confirmed by the x-ray structure determination of the aCTD/Spx complex [9]. The Y263 is conserved in Gram-positive bacteria and is not found in the Gram-negatives [1].
In vitro transcription experiments showed that Spx directly activates transcription of genes involved in thiol homeostasis, including trxA (thioredoxin gene) and trxB (thioredoxin reductase) [10,11]. Spx-dependent transcriptional activation is under redox control, requiring a CXXC disulfide center at the protein's Nterminus [10]. The oxidized form of Spx activates transcription of genes under its control.
Upon oxidative stress, Spx concentrations increase due to enhanced transcription and decreased proteolysis. Increased spx transcription is attributed, in part, to inactivation of the PerR and YodR repressors that negatively control spx transcription during non-stress conditions [12,13]. Spx is also under proteolytic control requiring the ATP-dependent ClpXP protease [8,14], but Spx is able to escape from the proteolysis under oxidative conditions because both the ClpX ATPase subunit and YjbH, an Spx-specific adaptor protein involved in ClpXP proteolysis of Spx, are inactivated by oxidants [15,16,17]. YjbH, a Zn-binding protein, interacts with Spx and the interaction accelerates the proteolysis of Spx by ClpXP [15]. Oxidative stress releases Zn from YjbH and ClpX, which is thought to result in loss of YjbH-mediated proteolysis.
The tight transcriptional and post-translational control of intracellular Spx is logical from a physiological viewpoint, given that the Spx-dependent regulation is exerted over a genome-wide scale. When Spx levels increase, as a result of either a null mutation in clpX or clpP, or in response to oxidative stress, transcription of a large set of genes is reduced [2,8]. Most, if not all, of these genes are positively controlled by transcription activators. For example, ComA-dependent srfA transcription is repressed when Spx is overproduced. ComA binds to the regulatory region of srfA and recruits RNAP to the promoter region through the interaction of ComA and aCTD [18,19,20]. ComA-dependent srfA transcription requires residues C265 and K267 located in the a1 helix of aCTD [20]. Y263, the residue of aCTD that interacts with Spx, is also located in the a1 helix, indicating that Spx likely competes for binding of aCTD with ComA [9,20]. The adverse effect of Spx on ResD-dependent transcription of fnr was explained by a similar interference of ResD-aCTD interaction by Spx [21].
The interference model of activator-aCTD complex disruption by Spx is consistent with the observation that free Spx does not bind to DNA [10]. Furthermore, the DNA sequence-independent mechanism of action can explain why Spx adversely affects transcription of a wide range of genes. On the other hand, it remains unresolved whether Spx-dependent transcriptional activation involves binding of Spx to a specific sequence within target promoter DNA. A previous study showed that oxidized Spx alone does not bind to the trxA or trxB promoters [10], although this does not necessarily eliminate the possibility that Spx, when complexed with RNAP, directly binds to a sequence within an Spx-activated promoter. In fact, previous DNase I footprinting analyses showed that neither RNAP nor Spx individually binds to trxA and trxB, while the presence of both RNAP and Spx resulted in footprints covering a region between 250 and 210 within each promoter [10]. This result indicates that Spx, after forming a complex with aCTD, directs RNAP to Spx-activated promoters, either by altering the way in which RNAP engages the promoter, or by forming part of the binding surface which specifically recognizes Spx-activated promoter sequences. In the hope of obtaining direct evidence of a DNA-protein interaction in the transcription initiation complex at the trxA and trxB promoters, we carried out site-specific DNA and protein crosslinking using trxA (and trxB) promoter DNA, RNAP, and Spx [11]. Contact of Spx with promoter DNA was not detected by crosslinking at any of the nucleotide positions examined. The addition of Spx resulted in enhanced s A contact with the 210 region of the trxA and trxB promoter at 37uC. Similarly, Spx stimulated contact of the bb' subunits of RNAP with nucleotide base positions near the transcription start sites and around 221/222 in both promoters. That study also uncovered evidence of a cis-acting element upstream of the core promoter sequences in Spx-controlled genes that is required for Spx-activated transcription [11].
In this work, we identify the cis-sequence in the trxB promoter that is essential for Spx-dependent transcriptional activation. As previously shown, Spx alone is unable to bind the trxB promoter, but Spx is capable of generating a supershifted trxB-aCTD-Spx complex in an electrophoretic mobility shift assay (EMSA). In parallel, structural studies identify a change in the conformation of a helix a4 of Spx when residue C10 of the redox disulfide center at its N-terminus is mutated to serine. When R60, a residue associated with the helix a4 region, is mutated to glutamate, Spx-dependent transcription of trxB in vivo and binding of the aCTD-Spx complex to the trxB promoter in vitro are abolished.

Crystallization and Structural Analysis of Reduced Spx
''Reduced'' Spx, in which residue Cys10 was replaced by serine (C10S Spx), was overexpressed in E. coli and purified by Ni 2+ -NTA affinity column chromatography as previously described [9]. Purified, cleaved Spx was dialyzed into 25 mM PIPES, pH 6.5, 100 mM KCl for use in subsequent crystallization trials.
aCTD containing residues 245 to 314 of RpoA was cloned into the IPTG-inducible PET28a overexpression vector (Invitrogen), which contains an N-terminal histidine tag and thrombin cleavage site. The aCTD-PET28a vector was transformed into Rosetta 2 competent cells (EMD Biosciences) for overexpression. aCTD was purified by Ni-NTA affinity chromatography (Qiagen). Following purification, the histidine tag was cleaved by incubation with thrombin protease (GE Healthsciences). Purified, cleaved aCTD was dialyzed into 25 mM Pipes, pH 6.5, 100 mM KCl. The Spx-aCTD complex was formed by mixing equimolar amounts of purified C10S Spx and aCTD. The complex was concentrated to 10 mg/ml. Crystals of the reduced C10S Spx-aCTD complex were grown by the hanging-drop vapor diffusion method. Two microliters of 0.5 mM C10S Spx-aCTD was mixed with 2 mL of reservoir solution containing 25%-30% polyethylene glycol 4000, 0.1 M sodium citrate pH 5.3 and 0.1 M MgCl 2 . Crystals appeared within two to three days and reached maximum dimensions of 0.360.0560.05 mm 3 in approximately one week. The crystals were cryoprotected in the mother liquor plus the addition of 15% glycerol and flash frozen in a nitrogen stream at 2180uC. X-ray intensity data were collected on beamline 8.3.1 at the Advanced Light Source and data were processed using MOSFLM [22] as implemented in the CCP4 suite of programs (CCP4, 2004) ( Table 1). The crystals took the spacegroup P1 with cell dimensions a = 29.4 Å , b = 32.2 Å , c = 50.6 Å , a = 106.1u, b = 91.6u, c = 103.8u.
The structure of the C10S Spx-aCTD complex was solved by molecular replacement using the oxidized Spx-aCTD structure as the search model (Table 1). Manual fitting and adjustment of the model resulted in the placement of amino acid residues 1-115 of Spx and residues 245-311 of aCTD into the electron density map using O [23]. The initial protein model was subjected to rigid body refinement, followed by simulated annealing and positional and Bfactor refinement using CNS [24]. Simulated-annealing omit maps were calculated to ensure the correct placement of all residues and to avoid model bias. The final model included residues 1-115 of Spx, residues 245-311 of aCTD, and 40 solvent molecules. The final R work and R free are 23.6% and 27.8%, respectively, to 1.9 Å resolution. The stereochemistry of the final model was assessed with PROCHECK [25], which revealed 89.7% of all w/y angles in the most favored regions of the Ramachandran plot and none in the disallowed regions. The coordinates and structure factors have been deposited in the RCSB with accession number 3IHQ.

Construction of Spx Amino Acid Substitution Mutants
All bacterial strains and plasmids are listed in Table 2. The effect of Spx amino acid substitution was examined with the Spx construct carrying amino acid substitutions (AN to DD) at the carboxyl-terminal end, which renders Spx insensitive to ClpXP proteolysis [2]. The previously constructed plasmids pSN56 [2] and pZY14 [20] are pDR111 derivatives that carry spx DD and spx C10A-DD , respectively. pDR111 is an amyE integration vector, and the cloned spx genes are transcribed from the IPTG-inducible Pspankhy promoter [26].
Three additional spx mutations conferring single amino acid substitutions were generated by two-step PCR-based mutagenesis using a pair of complementary mutagenic oligonucleotides 2 oMMN07-351 and oMMN07-352 for R60E, oMMN07-353 and oMMN07-354 for K62E, and oMMN07-355 and oMMN07-356 for K66E. Each oligonucleotide pair was used for the first PCR, together with either the upstream oligonucleotide oMMN01-173 or the downstream oligonucleotide oMMN01-174 and the plasmid pSN56 as template. The two PCR products carrying short complementary ends were annealed, filled-in by ExTaq polymerase (Takara Bio USA), and used as template for the second-round of PCR using oMMN01-173 and oMMN01-174. The PCR product was digested with SalI and HindIII and cloned into pDR111 and digested with the same enzymes to generate pMMN683 (spx DD-R60E ), pMMN684 (spx DD-K62E ), and pMMN685 (spx DD-K66E ).
Plasmid pMMN754 carrying spx DD-G52R was generated as follows. The spx gene carrying the G52R mutation was amplified from chromosomal DNA isolated from ORB4055 using oligonucleotides oMMN01-173 and oMMN01-174. The PCR product was digested with HindIII and BclI and the 59-end of spx containing the G52R mutation was isolated. The 39-end of spx carrying the DD mutation was isolated from pSN56 digested with BclI and SalI. The two fragments were cloned into pUC19 digested with HindIII and SalI by three-fragment ligation to generate pMMN753. The spx fragment was isolated from pMMN753 digested with HindIII and SalI and cloned into the HindIII-SalI sites of pDR111 to generate pMMN754.
The effect of the R60E, K62E, and K66E mutations of Spx on trxB expression was determined by measuring lacZ expression driven by the trxB promoter (2510 to +190 relative to the transcription start site) as previously described [10]. Plasmids pMMN683 (spx DD-R60E ), pMMN684 (spx DD-K62E ), and pMMN685 (spx DD-K66E ) were used to transform ORB4566 carrying spx::neo and trxB-lacZ at thrC, and transformants were selected for spectinomycin resistance (Spc r ) to generate ORB6895, ORB6896, and ORB6897, respectively. The transformants were screened for the amylase-negative phenotype, which is indicative of doublecrossover recombination [27]. As a control, pSN56 carrying the wild-type Spx DD was used to transform ORB4566, and ORB6894 was obtained.
The strains carrying srfA-lacZ and the wild-type or mutant Spx DD were constructed as follows. The strain JH642 was transformed with pMMN683, pMMN684, and pMMN685, and the strains ORB6930, ORB6931 and ORB6932 were constructed as described above. Each strain was transduced with SPb phage carrying pMMN92-borne srfA-lacZ [18] to generate ORB6934, ORB6935, and ORB6936. A control strain, ORB6129, carrying the wild-type Spx DD at the thrC locus and the srfA-lacZ fusion was constructed as previously described [20].

Construction of trxB Promoter Mutations
All mutant trxB promoters are derivatives of pDYR9 [11], which carries the trxB promoter (2115 to +47) fused to lacZ. Base substitution mutations of the trxB promoter were constructed by two-step PCR using complementary mutagenic primer pairs in a procedure similar to that used for the amino acid substitutions of Spx. The sequences of mutagenic oligonucleotides are listed in Table 3, and the outside forward (oDYR07-52) and reverse primers (oDYR07-32) were previously described [11]. DNA fragments resulting from the second PCR were digested with EcoRI and HindIII and cloned into pDG793 [28] that had been digested with the same enzymes. pDG793 is a thrC integration plasmid and double-crossover recombination is selected by Thrphenotype. Each plasmid was used to transform ORB3834 (spx::neo), and erythromycin-resistant (Erm r ) Thrtransformants were then transformed with chromosomal DNA isolated from strains carrying the wild-type and mutant spx DD at the amyE locus. All plasmids are listed in Table 2.

Measurement of b-galactosidase Activity
The effect of the Spx amino acid substitutions on the expression of trxB and srfA was determined by measuring b-galactosidase activity in cells carrying trxB-lacZ (ORB6894 to ORB6897) and srfA-lacZ (ORB6129, ORB6930 to ORB6932) in the presence and absence of IPTG. The strains were grown at 37uC overnight on DS agar plates [29] supplemented with spectinomycin and erythromycin (for trxB-lacZ) or spectinomycin and chloramphenicol (for srfA-lacZ). The overnight cultures were used to inoculate the same liquid medium at a starting optical density of 600 nm (OD 600 ) of 0.02. When the OD 600 of the cultures reached 0.4 to 0.5, the cultures were divided into two flasks and 1 mM IPTG was added to one of the flasks. Samples were taken at 0.5-to 1-hr intervals to assay b-galactosidase activity, which was expressed as Miller units [30].

Western Blot Analysis
The strains ORB6894 (Spx DD ), ORB6895 (Spx DD-R60E ), ORB6896 (Spx DD-K62E ), and ORB6897 (Spx DD-K66E ) were cultured in DS liquid medium supplemented with spectinomycin and erythromycin as described above. Each culture was divided into two tubes at an OD 600 of 0.4 to 0.5, and 1 mM IPTG was added to one of the tubes. Two milliliter samples were harvested after a 1.5-hr incubation and resuspended with 0.5 ml of 20 mM potassium phosphate buffer pH 7.5, 15 mM MgCl 2 , 20% sucrose. Lysozyme (1 mg/ml) was added and the suspension was incubated by gently shaking at 37uC for 30 min. The protoplasts were collected by centrifugation at 7,000 x g for 5 min and washed once with the same buffer. The precipitated protoplasts were lysed by resuspending with 0.5 ml of lysis buffer (30 mM Tris-HCl, pH 8.0, 1 mM EDTA) to obtain crude extract. Protein concentrations in the crude extract were determined using BioRad protein assay solution, and 15 mg of total protein was applied to an  thrC integration vector with promoter-less lacZ [28] SDS-polyacrylamide (15%) gel. The Western blot experiment was done as previously described using an anti-Spx antibody [31].

Protein Purification
RNAP was purified from ORB4028 (spx::neo his 10 -rpoC) as previously described using a Ni-NTA affinity column, a heparin agarose, and a Bio-Rad High Q column [11,32,33]. The selfcleavable intein system (New England Biolabs) was used for overproduction and purification of aCTD and Spx. aCTD (residues 225-314) was overproduced in Escherichia coli BL21/pLysS carrying pSN37 [8] and purified using a chitin column and a BioRad High Q column. s A was overproduced using pSN64 in E. coli ER2566 (New England Biolabs) and purified using a chitin column and a High Q column as previously described [34]. The wild-type Spx protein was overproduced from pMMN470 [31] in ER2566 and purified using a chitin column and a BioRad High S column. To overproduce the Spx R60E protein, pMMN712 was constructed as follows. The fragment, amplified by PCR using oligonucleotides oMMN01-135 and oMMN01-137 together with pMMN683 as template, was digested with NcoI and SmaI and then cloned into pTYB4 (New England Biolabs) digested with the same enzymes. ER2566 carrying pSN21 and pMMN712 were used to overproduce the Spx G52R [8] and Spx R60E mutant proteins, respectively, and the proteins were purified similarly to the wild-type Spx.

In Vitro Transcription
A linear trxB template was generated by PCR with oligonucleotides oDY07-32 and oDY07-52. The template is expected to produce a 66-base transcript. One nM of the template and 25 nM of RNAP together with 25 nM s A were incubated without or with 7.5 nM Spx protein in 62.3 ml of 10 mM Tris-HCl pH 8.0, 50 mM NaCl, 5 mM MgCl 2 , and 50 mg/ml BSA. After 10-min incubation at 37uC, 7.7 ml of nucleotide mixture (700 mM ATP, Table 3. Oligonucleotides used in this study.

Oligonucleotide
Sequence Purpose GTP and CTP, 35 mM UTP, 17.5 mCi a-32 PUTP) was added to each reaction. After incubation at 37uC for 2, 5, and 10 min, 20 ml of the reaction was withdrawn to mix with 10 ml of stop solution (1 M ammonium acetate, 0.1 mg yeast RNA, and 0.03 M EDTA).
The mixture was precipitated with ethanol and resuspended with 5 ml of formamide-dye (0.3% xylene cyanol, 0.3% bromophenol blue, and 12 mM EDTA dissolved in formamide). The samples were heated at 90uC for 2 min and were applied onto an 8% polyacrylamide-urea gel. The gel was dried and autoradiographs were scanned on a Typhoon Trio scanner (GE Healthcare).

Electrophoretic Mobility Shift Assay (EMSA)
The probe used for EMSA was a fragment extending from 256 to 221 of the trxB promoter region, which was generated by annealing complementary oligonucleotides. The 36-mer oMMN08-465 was the template strand and the 35-mer oMMN08-466 was the non-template strand that lacks A (complementary to 256T) at its 39-end. The two oligonucleotides (5 pmoles each) were mixed in 20 ml of 10 mM Tris-HCl pH 7.9, 50 mM NaCl, 10 mM MgCl 2 and heated at 90uC for 5 min, then slowly cooled to room temperature. To radiolabel the nontemplate strand, Klenow fragment and 10 mCi of [a-32 P]dATP (800Ci/mmol) were added to the annealing reaction to fill-in the 39 end. After incubation at room temperature for 15 min, unincorporated [a-32 P]dATP was removed using a nucleotide purification kit (Qiagen). The mutant (G-44A G-33A) trxB and spoVG probes were generated in a similar manner, except using oligonucleotides oMMN08-473 and oMMN08-474, and oMMN09-477 and oMMN09-478, respectively.
Five mM of Spx and 5 mM of aCTD (unless otherwise stated) were incubated at room temperature for 10 min in 20 ml of 20 mM Tris-HCl pH 7.8, 50 mM NaCl, 5 mM MgCl 2 , 10% glycerol. The radiolabeled probe (2,000 cpm/reaction) was added to the preincubated mixture and further incubated at room temperature for 15 min. The reaction mixture was applied onto a pre-run 6% native polyacrylamide gel and run in TGE buffer (50 mM Tris, 0.38 M glycine, 2 mM EDTA) at 180V. The gel was dried and scanned on a Typhoon Trio variable mode imager.

The Wild-Type and Spx(C10S) Structure
Formation of the disulfide bond between C10 and C13 is essential for the positive regulatory role of Spx, but not for its negative role [10,20,35]. We hypothesized that a conformational change caused by formation of the disulfide bond could provide a mechanism for how Spx is involved in the transcriptional activation of genes such as trxA and trxB. Therefore, the crystal structure of Spx (C10S), which mimics the reduced form, in complex with the aCTD was determined to 1.9 Å resolution ( Figure 1A).
The structure of the reduced aCTD-Spx complex is quite similar to that of the oxidized aCTD-Spx complex [9,36] and an overlay of 156 corresponding Ca atoms of both complexes, excluding residues on helix a4 of Spx, results in a root mean square deviation of 0.6 A. As seen previously the aCTD contains four core a helices (a1-a4) and a somewhat extended N-terminal helix designated a19 ( Figure 1A). The reduced C10S Spx protein is a mixed a/b protein with its secondary structural elements arranged: b1a1b2a2a3(3 10 )a5b3b4a6 ( Figure 1A). Thus, Spx retains most of the secondary structure that is found in oxidized Spx. Importantly, the aCTD-Spx interface of the reduced complex is identical to that of the oxidized aCTD-Spx complex indicating that the biological effects of thiol stress readout by these proteins are not a consequence of a radically different structure of this complex.
Interestingly, the loss of the disulfide linkage between residues C10 and C13 does not result in a significant change in the local structure ( Figure 1B). Although free to rotate from their positions in the oxidized state, the S10 and C13 side chains do not move Helix a4, which is observed in oxidized Spx [9] but has unraveled in the reduced form, is colored magenta. The residues mutated in this study, R60 and K62, are labelled and shown as sticks with carbon atoms colored white and nitrogen atoms either blue or magenta. Residues S10 and C13 are labelled and shown as sticks with carbon and sulphur atoms colored yellow and the c-oxygen of S10, red. (B) Close up of the region surrounding helix a4 and residues C10/S10 and C13 after the superposition of the oxidized and reduced aCTD-Spx complex structures. Reduced Spx is shown as a magenta ribbon and oxidized Spx as a teal ribbon. The C10-C13 disulfide bond is shown in orange sticks and S10 and C13 from the reduced structure are shown as yellow sticks. In the reduced form residue R92 has moved 2.8 Å away from its position in the ammonium sulphate-containing oxidized form [9]. The side chain of residue R60 beyond the Cb atom is disordered in the sulphate-containing crystal form of oxidized Spx, which was used in the superposition visualized here [9]. doi:10.1371/journal.pone.0008664.g001 because their positions are buttressed by numerous interactions. Indeed the S10 Oc side chain engages in hydrogen bonds to the backbone amide (NH) group of C13 and Oc atom of residue S12. The C13 Sc sulfhydryl group hydrogen bonds to hydroxyl side chain of residue S10 and makes van der Waals contacts with residues R92 and P93. Also, due to two alternative backbone conformations around residues S7 and P8, the S7 backbone carbonyl oxygen can engage in a weak hydrogen bond to the C13 SHc group. One significant conformational difference is found between oxidized and reduced Spx, however, whereby helix a4 of the reduced mutant Spx structure (residues S61 to N68 in oxidized Spx) unfolds and rotates (Figure 1). As a consequence of the unravelling of helix a4, several basic residues are repositioned. Specifically, the side chain of residues K62, which points into the solvent in oxidized Spx and is disordered, turns inward and now makes hydrogen bonds to the carbonyl oxygen (CO) atom of residue G88 and the Oc atom of residue Ser58. Residue K66, which is also solvent exposed in oxidized Spx, is now pointing into the core of the protein. Hence, the exposed electropositive surface of a4 is decreased significantly in reduced Spx. Other helix a4related conformational changes include the repositioning of residue R60, whereby its Ca carbon has moved 1.3 Å from its oxidized position. The side chain of the R60 residue, which points directly into the solvent and is disordered in one crystal structure of oxidized Spx but not in the second [36], has moved 5.8 Å (Cf-Cf) from its position in the oxidized protein and has shifted towards R92 by 1.3 Å ( Figure 1B). The potential importance of this altered location is tied to residue R92. In reduced mutant Spx, the side chain of residue R92 has rotated outward by ,2.8 Å from its location in oxidized Spx. In the oxidized protein, the guanidinium side chain is engaged in an electrostatic interaction with a bound sulfate ion [9]. In the reduced protein, the R92 side chain rotates inward and makes hydrogen bonds to the peptide backbone carbonyl oxygen atoms of residues G88 and L90 ( Figure 1B). This location of the R92 side chain of reduced Spx is very similar to its location in oxidized Spx crystallized from solutions that do not contain sulfate or phosphate anions [36]. Perhaps, the R92-bound sulfate ion described in the first reported oxidized Spx structure is a surrogate for one of the phosphate groups of (aCTD-Spx)-bound DNA. If so, this also places the solvent exposed guanidinium group of residue R60 near the DNA phosphate backbone and suggests its possible role in DNA binding, either to the backbone or to a guanine. Thus, loss of the Spx C10-C13 disulfide bond results only in small conformational changes that are confined primarily to helix a4. However, the resulting helix-to-coil transition repositions the side chains of several basic residues that could have functional consequences with respect to DNA binding.

Residues in or Near Helix a4 of Spx Are Important for trxB Activation but Not for srfA Repression by ComA Activator Interference
To determine whether the structural change in helix a4 is crucial for the function of Spx in transcription activation, we next introduced single amino acid substitutions around the helix a4 region. Interestingly, there are some basic residues adjacent to and within the helix, namely, R60, K62, and K66. Since basic amino acids are known to interact with DNA through sequence recognition and charge neutralization [37], we decided to substitute each residue with glutamate and to examine the effect of these substitutions on Spx activity. Because under nonstress conditions Spx is degraded by ClpXP protease, we expressed the ClpXP-resistant forms (Spx DD ) of the wild-type and mutant proteins from an IPTG-inducible promoter, as previously described [2] so that we could examine the mutational effect on trxB transcription under nonstress conditions. Western blot analysis showed that the three mutant Spx proteins were produced only in the presence of IPTG at a level similar to the wild-type protein ( Figure 2). trxB-lacZ expression in cells producing Spx K66E was equal to or greater than that observed in cells producing the wild-type Spx ( Figure 3A). In contrast, the K62E mutation reduced transcription and the R60E mutation nearly abolished transcription, indicating that R60, and to a lesser extent K62, are important for transcriptional activation of trxB.
Although Spx R60E and Spx K62E are produced at a level similar to the wild-type Spx in B. subtilis cells, we could not completely eliminate the possibility that the mutant proteins were misfolded, and thus, inactive. As described earlier, Spx plays both a positive and negative role in transcription regulation. To examine whether the mutant proteins retain the ability to exert negative transcriptional control, we determined the effect of the mutations on ComA-dependent srfA transcription. As shown in Figure 3B, srfA  subtilis. B. subtilis cells expressing the wild-type and the mutant Spx DD (the C-terminal two amino acids are substituted with aspartate residues, which renders the Spx protein insensitive to ClpXP protease) were grown in DS medium in the absence (lanes 1, 3, 5, and 7) and the presence of IPTG (lanes 2, 4, 6, and 8) as described in Materials and Methods. The lysate was prepared by the protoplast lysis method as described and 15 mg of total protein was resolved by SDS-polyacrylamide gel electrophoresis. Western blot analysis was carried out to detect Spx as shown previously. Lanes: M, molecular weight marker; 1 and 2, ORB6894 (Spx DD ), ORB6895 (Spx DD-R60E ), ORB6896 (Spx DD-K62E ), and ORB6897 (Spx DD-K66E ). doi:10.1371/journal.pone.0008664.g002 transcription was severely reduced in a strain that produced the wild-type Spx protein and in all of the strains that produced the mutant proteins. These results clearly demonstrated that R60 and K62 of Spx play pivotal roles in positive control but are dispensable for its negative role in transcription.

Identification of trxB Sequences Required for Spx-Dependent Transcription Activation
One possible hypothesis for why R60 (and to a lesser extent K62) is required for trxB transcription and not for inhibition of activator-dependent srfA transcription is that R60 is involved in the interaction of Spx with the trxB promoter DNA for establishing the transcription initiation complex. Although Spx itself did not bind trxB DNA [10], this result does not necessarily eliminate the possibility that Spx, by interacting with aCTD, can contribute part of a DNA-binding surface. The intracellular disulfide bond formation might facilitate the recognition and/or binding of the side-chain of R60 with a specific nucleotide in the trxB promoter region, as suggested above. Alternatively, Spx residue R60 might function indirectly in DNA sequence recognition by changing the conformation of aCTD so that it recognizes specific sequences associated with Spx-activated promoters. We think that this is unlikely as explained in the Discussion.
Our previous work demonstrated that the trxB promoter region between 250 and 236 is required for Spx-dependent transcription activation and that the nucleotides at positions 243 and 244 are essential for transcription [10,11]. The identified region corresponds well with the sequence protected from DNase I digestion in the presence of the wild-type Spx-RNAP complex [10]. The DNase I footprinting analyses also identified a hypersensitive site between 234 and 235 of the trxB template strand in the presence of Spx-RNAP [10]. Interestingly, Spx-RNAP generated a hypersensitive site at a similar position (between 235 and 236) of trxA, a gene that is also activated directly by Spx. Furthermore, the nucleotide sequences (AAAA-TAGCGT) of the trxA (240 to 231) and trxB (239 to 230) regions that include the hypersensitive site are identical. These observations prompted us to carry out further mutational analyses of the 239 to 230 region of trxB.
As in the previous study [11], we used the trxB promoter carrying 2115 to +47 for base substitution experiments, except that we expressed the IPTG-inducible Spx DD in the strain lacking the native spx gene. The results showed that three segments of trxB are important for Spx-dependent transcription activation (Figure 4). The first segment is the AGCA sequence positioned from 245 to 242. Our previous study showed that the G at 244 and C at 243 are indispensable for transcription [10,11]. The second segment is a poly-A stretch between 241 and 236. In our previous study, base substitution of two of the A residues showed a moderate effect on trxB transcription in the spx + background [11], and here, these substitutions showed more adverse effects in cells lacking the native spx gene (Figure 4). It has been known that aCTD binds to AT-rich sequences; for example, the sequence AAAAAARNR at positions 246 to 238 of the E. coli rrnB promoter P1 serves as the proximal aCTD-binding site [38]. We, therefore, propose that the sequence between 241 and 235 is a site for aCTD interaction. The last segment important for trxB transcriptional control by Spx is the GC sequence positioned at 233 and 232. Our previous work showed that Spx-dependent activation of a trxB/srfA hybrid promoter transcription was enhanced further by extending the trxB control region to 231 including the GC at 232/33 [11]. The dinucleotide resides in the center of the AGCG sequence that is similar to the upstream essential AGCA (245 to 242) sequence. The G residue in the second position is the most critical among the tetranucleotides AGCA and AGCG. Because the two tetranucleotides are similar, with the exception that the most 39-end of the upstream sequence is A (242) the corresponding position of the downstream sequence is G (231), we examined the effect of base substitution of the G at position 231. We found that substitution of G with either C or T had no significant effect, whereas the substitution with A led to a more than two-fold increase in trxB expression. One interpretation of this result is that the Spx/aCTD complex binds to trxB and that Spx contacts the AGCA sequence and aCTD contacts the downstream A-rich sequence. If this is true, then the question is which protein, if any, binds to the downstream AGCG sequence. The previous crosslinking study did not show contact of any protein to the 235 region [11], and was, therefore, inconclusive. One could envision three alternative scenarios for a protein/235 Figure 4. Effect of base substitutions of the trxB promoter on trxB expression. Single base pair substitutions were generated in the trxB promoter (2115 to +47). The mutated promoters fused to lacZ were introduced in spx mutant strains expressing spx DD from the IPTG-inducible Pspank-hy promoter. Expression of trxB-lacZ was determined in at least two independent isolates as described in Fig. 3. The effect of each base substitution is shown as a percentage of the peak trxB transcribed from the wild-type promoter, which was used as a control in each experiment. The peak expression was generally seen around 1.5 hr after the addition of IPTG. doi:10.1371/journal.pone.0008664.g004 interaction. One scenario is that s A binds to this sequence, which is facilitated by the interaction between s A and aCTD bound to the site upstream of 235. Another scenario is that the downstream AGCG sequence is a site where Spx binds. A third possibility is that the second aCTD, having undergone a conformational change upon contact with oxidized Spx, now recognizes the sequence centered at 233. For the reasons presented in the Discussion, this possibility seems unlikely. The issue of 233 element recognition will be discussed below (see Discussion).

Activation-Defective Spx Mutants Other Than G52R Are Able to Activate a Mutant trxB Promoter Bearing an A-34T Mutation
The mutational analysis showed that the A-34T substitution resulted in highly elevated trxB transcription, which remained largely dependent on Spx because IPTG was still required for promoter activity (Figure 4). We examined expression of trxB(A-34T)-lacZ in cells expressing spx R60E , spx C10A , and spx G52R mutants, as well as the wild-type spx. As shown in Figure 5A, transcription from the wild-type trxB promoter was severely reduced in cells producing each mutant Spx DD protein as compared with those producing wild-type Spx DD . In contrast, the R60E and C10A Spx mutants were able to activate the A-34T promoter, although the activity of the mutant proteins was approximately 50% of the activity of the wild-type protein ( Figure 5B). No expression was observed in the absence of IPTG, indicating that the observed expression is dependent on mutant Spx protein. Unlike the two Spx activation mutants, the Spx G52R mutant protein was completely impaired in activating transcription from the mutant promoter. This result indicates that the G52 residue, and hence, the Spx-aCTD interaction is absolutely required for Spx-activated trxB transcription, but the requirement of residues R60 and C10 is conditional, given that the spx mutations R60E and C10A are partially suppressed by the A-34T mutation in the trxB promoter.

Spx R60E Activates trxB(A-34T) Transcription In Vitro
The in vivo results described above showed that Spx R60E and Spx C10A are unable to activate transcription from the wild-type trxB promoter, but are able to significantly activate trxB(A-34T) transcription. We next carried out in vitro run-off transcription experiments to determine whether the effect of the A-34T substitution can be solely attributed to interactions involving the trxB(A-34T) promoter, Spx, and RNAP. Figure 6 shows that the basal level of trxB transcription was markedly elevated by adding wild-type Spx, but only a slight increase in the transcript was detected in reactions containing Spx R60E . The trxB(A-34T) transcript accumulated slightly more than the wild-type trxB transcript at a longer reaction time (10 min) in the absence of Spx, and the transcript levels were increased further when the wild-type Spx was present. Unlike the wild-type transcript levels, those of the mutant trxB were markedly elevated by Spx R60E . We repeated these experiments almost ten times and in some experiments we did not see any difference between the wild-type and the mutant trxB transcript levels activated by the wild-type Spx as shown in Figure 6; however, in other experiments, we observed that the mutant trxB transcript level was significantly higher than the wildtype transcript level. Although we do not understand the variability in the in vitro transcription results, Spx R60E reproducibly activated transcription from the trxB(A-34T) promoter to a higher level than from the wild-type promoter. Based on the in vitro transcription assays, we conclude that the adverse effect of the R60E mutation on trxB transcription is likely caused by either a weaker interaction of the mutant Spx with the trxB promoter, RNAP, or both.

The Spx-aCTD Complex Binds the trxB Regulatory Region
The mutational studies of the trxB promoter uncovered a region of trxB required for Spx-dependent activation [10,11], and this putative cis-acting component of Spx control was further investigated. Given that the A-rich sequence and the flanking AGCA and AGCG sequences are conserved between the trxA and trxB promoters and that the upstream AGCA sequence is also present in other Spx-controlled genes (see Discussion), one could envisage that the AGCA sequence (and possibly AGCG in trxA and trxB) is the site where Spx binds. To test this possibility, we next carried out EMSA analysis using a DNA fragment carrying the trxB promoter (256 to 221). This fragment covers the putative Spx-and aCTD-binding sites, but lacks the 210 region. We first examined whether either aCTD or Spx alone binds DNA. aCTD at 5 mM bound DNA, but Spx at the same concentration did not ( Figure 7A, lanes 2 and 6). Although Spx itself was unable to bind DNA at the concentration tested, Spx addition resulted in a supershifted complex with promoter DNA and aCTD, indicating that a DNA-aCTD-Spx ternary complex was formed ( Figure 7A, lane 3). In contrast, the supershifted band was not detected with   Figure 7A, lane 11, a reaction containing the G52R mutant protein alone with aCTD). The ternary complex, but not the binary complex, was only formed in the absence of DTT ( Figure 7B), which is in good agreement with the hypothesis that the oxidized form of Spx is required for DNA binding.
We next examined whether the trxB mutations that affect Spxdependent activation had lower binding affinities for the Spx-aCTD complex. The trxB promoter fragment bearing the G-44A and G-33A mutations was used for a probe in EMSA to compare the binding affinity for aCTD and Spx-aCTD ( Figure 8). When Spx was added in equal concentrations in reactions containing trxB(wt)-aCTD or trxB(G-44A G-33A)-aCTD, the ternary complex was more abundant when the wild-type promoter was present than when the mutant promoter was present (Figure 8, lanes 4 and 5, and 8 and 9). Even in the reaction in which more aCTD bound to the mutant promoter than the wild-type promoter, Spx was unable to supershift the mutant promoter complex as efficiently as the wild-type promoter complex (Figure 8, lanes 11 to 16). A mutant trxB-lacZ fusion bearing the two nucleotide substitutions produced 0.7 units of b-galactosidase activity (data not shown). These results support the assumption that G-44A (and G-33A) is important for the binding of Spx to the trxB promoter. We carried out a similar experiment with the trxB promoter carrying either the G-44A or G-33A mutation to determine which residue is critical for Spx binding; however, the trxB promoter carrying the single mutation did not show a significant difference from the wild-type promoter in EMSA analysis (data not shown).
We examined whether the DNA-aCTD-Spx complex is specific to promoters of Spx-activated genes by using the spoVG promoter in a similar EMSA analysis. The spoVG promoter has an AT-rich upstream sequence that was shown to have properties of an UP element, and was required for transcription [39,40]. Spx did not activate transcription of spoVG; conversely, the transcription was shown to be inhibited by overproduction of Spx through an as yet undiscovered mechanism [41,42]. The spoVG promoter exhibited a much higher affinity for aCTD than the trxB promoter, yet addition of Spx did not result in a supershifted DNA-aCTD complex even when an excess of Spx over aCTD was added ( Figure 9A). In addition, when the cold spoVG fragment was added, it was able to disrupt both aCTDand aCTD-Spx-binding to the trxB promoter ( Figure 9B). These results suggest that spoVG has a higher affinity for free aCTD than aCTD complexed with Spx, and once aCTD binds the spoVG promoter, Spx is unable to establish an interaction with aCTD.

Discussion
This study was aimed at elucidating how the oxidized form of Spx activates trxB transcription. The questions to be answered were: 1) how disulfide-bond formation at the redox CXXC center of Spx affects its activity as a transcriptional activator; 2) whether oxidized Spx in the Spx-aCTD complex binds to the trxB promoter region and enhances the binding of aCTD or other  subunits of RNAP to DNA; 3) if Spx binds to DNA, what is the consensus Spx-binding site in trxB and other Spx-activated promoters. A comparison of the crystal structures of the oxidized (wild-type) and reduced (C10S) forms of Spx revealed that in the reduced form helix a4 partially unfolds and rotates, suggesting that helix a4 could be important for the positive role of Spx in transcription. Consistent with this assumption, mutations of Spx residues R60 and K62, which are adjacent to and within helix a4, respectively, reduced Spx-dependent trxB transcription but did not show any effect on its repression of ComA-dependent srfA expression. We found that the R60 and K62 residues are well conserved among Spx orthologs, whereas residue 66 is a glutamate instead of lysine in Spx from some Bacillus species such as Bacillus cereus, Bacillus thuringiensis, Bacillus clausii, Bacillus halodurans, and Bacillus weihenstephanensis. The observation supports the important role of residues R60 and K62 in the function of Spx as a transcriptional activator.
The current mutational analysis of the trxB promoter not only confirmed the results from our previous work, but also further defined the cis-sequences required for Spx-dependent activation. Based on the mutational analyses and the EMSA experiments, we now propose that the AGCA sequence at 245 to 242 is the site with which the complex of aCTD with oxidized Spx directly interacts. This hypothesis is further supported by the following observations. Our previous microarray analysis [2] identified nfrA, coding for nitro/flavin reductase [43,44], as one of the genes activated by Spx. nfrA was also shown to be activated in response to a number of stress condition, including oxidative stress [43,45] and the nfrA promoter contains the AGCA sequence at the same position (245 to 242) as in the trxB promoter. Base substitutions of the first three nucleotides in this sequence, particularly G and C, resulted in a substantial reduction in the promoter activity, indicating the essential role of the AGCA sequence in nfrA transcription [43]. Our studies further showed that Spx activates nfrA-lacZ in vivo and the R60E mutation in Spx severely affects nfrA expression (A. L, and P. Z., unpublished results). In addition, an in vitro transcription assay showed that Spx directly activates nfrA transcription (A.L. and P.Z., unpublished results). The results are in good agreement with the hypothesis that the helix a4 of Spx functions in the interaction of aCTD/Spx complex with the AGCA sequence. We feel that it is unlikely that a conformational change of aCTD caused by Spx is responsible for the recognition of promoter DNA solely by the a polypeptide, as no uncharacteristic change in a conformation is observed when aCTD is bound to Spx [9,36].
In contrast to the AGCA sequence, the downstream AGCG sequence at 234 to 231, which is conserved in trxA and trxB, is absent in the nfrA promoter; hence, its role in transcriptional activation is unclear. Interestingly, the substitution of A at 234 with T resulted in a slightly higher basal level of trxB expression and a five-fold increase in Spx-dependent activation as compared with transcription from the wild-type promoter (Figures 4 and 5). Furthermore, trxB transcription from the mutant promoter was significantly stimulated by Spx R60E and Spx C10A (Figures 5 and 6), whereas the G52 residue was absolutely required for transcription. The A-34T change leads to a 3/6 match to the consensus 235 hexamer recognized by s A RNAP [46]. One possible scenario is that s A binds to the 235 region of the mutant promoter and the interaction between Spx and aCTD at the 244 element, as well as interaction of aCTD with s A , stabilizes the three proteins at the mutant promoter, which partially compensates for the defect in trxB interaction conferred by the R60E substitution. In contrast, the G52R mutation, by disrupting the interaction of Spx with aCTD, destabilizes aCTD binding to trxB, resulting in decreased engagement of s A with the 235 region. Does this possibility suggest that s A also interacts with the 235 region of the wild-type trxB promoter? Another mutation, C at 232 to A, which results in a 3/6 match to the consensus 235 hexamer did not increase trxB transcription, and instead, showed a severe adverse effect, suggesting that s A does not bind to the 235 of region the wildtype promoter and/or that the A-34T mutation increases the affinity of s A binding more than C-32A. The A-34T change leads to the sequence TTGCGT. The TTG in the 235 region seem to be the most important nucleotides of the 235 hexamer as these are the most highly conserved [46], which could explain the opposite phenotypes conferred by the C-32A and A-34T mutations.
Our previous study [11] showed that s A crosslinks to trxB at position 211 but not at 234. Furthermore, Spx did not crosslink to any nucleotide tested. One possible reason for this negative result may be the limited positions tested in the crosslinking experiments. Within the three important regions found in this work, the nucleotide positions tested for crosslinking were 246 and 234 of trxB and 247, 235, and 230 of trxA. Only 234A of trxB and 235A of trxA reside within the AGCG sequence. Another possible reason for the negative result is that the modification of a nucleotide with azidophenacyl bromide might have interfered with the binding of either Spx or s A to DNA. It would be worth revisiting the nucleotide-specific crosslinking study by focusing on the AGCA and AGCG sites, as well as the A-rich sequence, and by confirming that the modified templates are transcriptionally active.
Direct interaction between protein and DNA can also be verified genetically by site-specific suppressor analysis. Earlier studies of the sigma subunit-DNA interaction [47,48] demonstrated that mutations of the third G of the 235 hexamer to either A or C, but not to T, were suppressed by the substitution of Arg588 in region 4 of s 70 with His. Similarly, the defect caused by substitution of the fifth C of the 235 hexamer with either T or G was compensated for by the substitution of Arg584 of s 70 with Cys or His. We have investigated whether the various single base substitutions of G at 244, C at 243, G at233, or C at 232, as well as double mutation of either G at 244/233 or C at 243/ 232, were restored by introducing the R60H or R60C mutations in Spx; however, we could not detect any significant suppressing effect in any of the mutant combinations (M.M.N. and P.Z. unpublished results).
The aCTD of RNAP can make specific contact with the UP element sequence of certain promoters such as those of rRNA operons [49] and the spoVG gene of B. subtilis (Figure 9). Evidently, the aCTD in the Spx-aCTD complex engages the trxB promoter DNA in a manner that is different from its interaction with the UP element, as Spx is unable to contact aCTD on the spoVG promoter fragment. Some of the aCTD residues required for UP element contact might also function in its interaction with Spx, and thus, may not be available for Spx-aCTD complex formation. The a1 helix of aCTD contains part of the ''261 element'' that is required for DNA binding, and this helix also contains the essential Tyr residue for Spx interaction. Given that part of the a1 helix interacts with Spx further suggests that only part of the DNAbinding surface in aCTD might be exposed for DNA recognition, and that interaction with Spx is required to complete the DNA binding surface of the aCTD/Spx promoter recognition complex.
We, and others, have successfully cocrystallized aCTD and oxidized Spx as previously reported [9,36]. The EMSA study presented here now opens a powerful approach for cocrystallization of the aCTD-Spx complex with DNA to study the ternary interaction.