Using Disulfide Bond Engineering To Study Conformational Changes in the β′260-309 Coiled-Coil Region of Escherichia coli RNA Polymerase during σ⁷⁰ Binding

Abstract

RNA polymerase of Escherichia coli is the sole enzyme responsible for mRNA synthesis in the cell. Upon binding of a sigma factor, the holoenzyme can direct transcription from specific promoter sequences. We have previously defined a region of the β′ subunit (β′260-309, amino acids 260 to 309) which adopts a coiled-coil conformation shown to interact with σ⁷⁰ both in vitro and in vivo. However, it was not known if the coiled-coil conformation was maintained upon binding to σ⁷⁰. In this work, we engineered a disulfide bond within β′240-309 that locks the β′ coiled-coil region in the coiled-coil conformation, and we show that this “locked” peptide is able to bind to σ⁷⁰. We also show that the locked coiled-coil is capable of inducing a conformational change within σ⁷⁰ that allows recognition of the −10 nontemplate strand of DNA. This suggests that the coiled-coil does not adopt a new conformation upon binding σ⁷⁰ or upon recognition of the −10 nontemplate strand of DNA.

The DNA-dependent RNA polymerase of Escherichia coli is responsible for the synthesis of messenger, transfer, and ribosomal RNA in the cell. This multisubunit enzyme consists of four different polypeptide chains with subunit architecture α₂ββ′ω (17). While this core enzyme is capable of elongation and termination of transcription, an additional subunit, σ, is required to form holoenzyme, which can initiate transcription selectively at specific promoter sequences (6, 20). Each of the seven E. coli σ subunits directs transcription from a particular set of promoter sequences, thereby allowing E. coli a means of regulating global gene expression (17, 41).

We have previously demonstrated that amino acids (aa) 260 to 309 of the β′ subunit comprise a major interaction domain (β′260-309) for σ⁷⁰ binding (3, 5). From mutational analysis, secondary structure prediction (2), and comparison with the Thermus aquaticus X-ray crystal structure (43), β′260-309 is thought to adopt a coiled-coil conformation in E. coli RNA polymerase. Mutational analysis suggested that the upper face of this coiled-coil contains residues that interact with σ⁷⁰ region 2.2 (2, 18). Region 2 of σ⁷⁰ is composed of tightly packed α-helices (28), which could also undergo a conformational change upon interaction with the coiled-coil of β′260-309.

Previous modeling of these regions involved in σ-core binding has utilized crystal structures of σ⁷⁰ and core enzyme to predict molecular interactions. However, X-ray crystal structures are static pictures of a particular conformational state of a protein. It is known that RNA polymerase core enzyme and σ⁷⁰ undergo several sequential conformational changes as they proceed through transcriptional initiation—from binding of the sigma factor, to binding double-stranded promoter DNA, to melting of the promoter DNA, to transcription initiation, and finally to sigma release (10, 16, 18). Unfortunately, the large size of these multisubunit complexes prohibits nuclear magnetic resonance analysis; therefore, alternative methods must be developed to understand the conformational transitions involved throughout transcription initiation.

Engineering of strategically placed nonnative disulfide bonds within a given protein can be a powerful tool for analyzing potential conformational changes. If a function is unaffected by disulfide locking and constriction of protein movement, it argues strongly that no conformational change in that particular region is required for that function. If the disulfide bond does disrupt function, one can determine at which step conformational changes must occur in a mechanistic pathway. Introduction of disulfide bonds has previously been used to investigate conformational changes in various processes, including heme assembly in cytochrome b5 (37) and binding of the Tet repressor to operator sites on DNA (40). Unfortunately, many previous attempts at disulfide bond engineering traditionally yielded low success rates because there were no methods available for determining the optimal site for introduction of a disulfide bond within a protein.

In this work, we utilize a recently developed disulfide recognition algorithm (12) to engineer a disulfide bond within the β′240-309 coiled-coil to examine whether a major conformational change in the coiled-coil is required for binding to σ⁷⁰. By restricting the conformation of the β′coiled-coil, we show that a conformational change within the β′ coiled-coil is not required for σ⁷⁰ binding. We also present evidence that the “locked” β′240-309 coiled-coil, upon binding to σ⁷⁰, induces σ⁷⁰ to undergo the conformational change which allows binding to a −10 nontemplate strand oligonucleotide.

MATERIALS AND METHODS

Buffers and reagents.

All reagents were purchased from Sigma Biochemical unless otherwise indicated. Spectrophotometric-grade glycerol was purchased from Fisher. Ni-nitrilotriacetic acid (NTA) agarose was purchased from Qiagen.

E. coli culture medium LBG contained 0.5% Bacto Tryptone, 1.0% Bacto Yeast Extract, 0.5% NaCl, and 0.2% glucose. Ni-NTA buffer contained 25 mM Tris-HCl (pH 7.9), 500 mM NaCl, 20 mM 2-mercaptoethanol (2-ME), 6 M guanidine-HCl, 5 mM imidazole, and 5% glycerol. Refolding buffer contained 25 mM Tris-HCl (pH 7.9), 500 mM NaCl, 20 mM 2-ME, and 5% glycerol. Oxidation buffer contained 25 mM Tris-HCl (pH 7.9), 200 mM NaCl, and 5% glycerol. Storage buffer contained 25 mM Tris-HCl (pH 7.9), 200 mM NaCl, 0.1 mM EDTA, and 50% glycerol. pH values for buffers were determined at 22°C.

Disulfide bond prediction and plasmid construction.

Although the coiled-coiled region of interest is located within β′260-309, the β′240-309 peptide was used for these experiments to facilitate purification from inclusion bodies, as the β′260-309 peptide is more prone to precipitation during the refolding process. A pET24a-derived overexpression vector for C-terminal hexahistidine (His₆)-tagged β′ (aa 240 to 309) (pTA545) was constructed by PCR modification (1, 3).

The amino acid sequence of E. coli β′260-309 was analyzed using a disulfide recognition algorithm (12) designed to recognize cysteine pairs that are in the proper orientation to form a disulfide bond. The sequence of E. coli β′260-309 was threaded onto the structure of the T. aquaticus RNA polymerase β′ subunit (43) to identify residues that have strong potential for forming a disulfide bond. This analysis identified two residues, L268 and L306, within the E. coli β′ subunit that when substituted with cysteine yield a predicted disulfide χ₃ torsion angle of 89.1° (where ±90° is optimal for a disulfide bond). An overexpression vector containing β′240-309 with the L268C and L306C mutations was created through QuikChange mutagenesis (Stratagene) using plasmid pTA545 as a template. The L268C mutation was created by the upper primer 5′TCTGACCTGAACGATTGTTATCGTCGCGTCAAT, and the L306C mutation was created by the upper primer 5′GAAGCGGTAGACGCCTGTCTGGATAACCACCAC. After sequencing the region to rule out secondary mutations that may have arisen during PCR, the plasmid pLA28 (β′240-309 L268C/L306C) was transformed into BL21(DE3)pLysS (Novagen) for overexpression.

Cell growth and lysis.

Overnight cultures of BL21(DE3)pLysS containing pTA545 and pLA28 were diluted 1:100 into 1 liter of LBG medium and grown at 37°C with rapid agitation in a 2-liter baffle flask. Cultures were induced with the addition of 1 mM isopropyl-β-d-thiogalactopyranoside when at an optical density at 600 nm (OD₆₀₀) of 0.5 and were harvested by centrifugation at an OD₆₀₀ of 1.3. The cell pellets (≈3 g [wet weight] cells/liter) were resuspended in 10 ml of Ni-NTA buffer and lysed by sonication. Cell debris was pelleted by centrifugation at 12,000 × g for 15 min.

Denaturing Ni-NTA chromatography.

A Ni-NTA agarose slurry (1 ml of packed resin) was added to the supernatants and incubated with mixing for 30 min at 8°C. The mixtures were applied to Polyprep columns (Bio-Rad) for gravity flow chromatography. The resin was drained and washed with 20 volumes of Ni-NTA buffer followed by 20 volumes of Ni-NTA buffer plus 40 mM imidazole. The wild-type and L268C/L306C β′240-309 proteins were eluted with 1.0-ml portions of Ni-NTA buffer plus 200 mM imidazole. Individual fractions were subjected to electrophoresis on a 12% NuPAGE bis-Tris gel with morpholineethanesulfonic acid (MES) buffer (Novex) and stained with GelCode Blue reagent (Pierce).

Refolding and oxidation of the disulfide bond.

For both the purified denatured wild-type and L268C/L306C β′240-309 proteins, Ni-NTA eluate fractions were pooled (4 ml) and refolded by 64-fold dilution into refolding buffer. Because the total volume was almost 300 ml, the proteins were concentrated by batch Ni-NTA chromatography and eluted with refolding buffer plus 200 mM imidazole.

Once refolded, the pooled protein fractions were diluted 10-fold with oxidation buffer to approximately 0.2 mg/ml and placed inside a Pierce Slide-a-Lyzer dialysis cassette (3-kDa cutoff) to remove residual 2-ME. These samples were dialyzed for 54 h at 4°C against oxidation buffer with one buffer change after 24 h. The extent of oxidation was determined by mobility shifts via electrophoresis.

After dialysis was complete, the protein samples were removed from the dialysis cassette and placed into Amicon Centriplus-3 ultraconcentrators (3-kDa cutoff) for concentration. The wild-type and L268C/L306C β′240-309 protein samples were then dialyzed against storage buffer in a Slide-a-Lyzer dialysis cassette (3-kDa cutoff) and stored at −80°C.

Purification of σ⁷⁰ and RNA polymerase core enzyme.

Purification of wild-type E. coli σ⁷⁰ from inclusion bodies by solubilization in guanidine-HCl and refolding by gradual 64-fold dilution was performed as previously described (32). RNA polymerase core enzyme was purified by immunoaffinity chromatography and passage over a Bio-Rex 70 cation-exchange column (Bio-Rad) as previously described (38).

Protein-protein native gel shift assay.

The purified oxidized wild-type and L268C/L306C β′240-309 peptides were added at a concentration of either 1 or 5 μM to a constant concentration of σ⁷⁰ (1 μM). The proteins were incubated for 10 min at 25°C in buffer containing 25 mM Tris-HCl (pH 7.9), 100 mM NaCl, and 10% glycerol. To create reduced wild-type and mutant samples, 20 mM 2-ME was added to both the wild-type β′240-309 and oxidized β′240-309 L268C/L306C samples. Both oxidized and reduced complexes were electrophoresed on a native 12% Tris-glycine polyacrylamide gel at 120 V for 3.5 h at 4°C in the cold room and stained with GelCode Blue reagent (Pierce) according to the manufacturer's instructions. To confirm the identity of the binding partners with the shifted complexes, the stained gel was washed for 20 min in water and the complex band was excised from the gel. The gel slice was placed directly in the well of a 12% bis-Tris polyacrylamide gel, subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) with MES buffer (Novex), and stained with GelCode Blue reagent.

Protein-protein-DNA native gel shift assay.

Oligonucleotides labeled with 5′-Cy5 dye containing the consensus −10 nontemplate strand hexamer (5′ATTGGGTATAATTGACTCA) and nonconsensus hexamer (5′ATTGGGATCTCGTGACTCA) (29) were ordered from Sigma-Genosys. For the assay, the following concentrations were used in buffer containing 25 mM Tris-HCl (pH 7.9), 100 mM NaCl, and 10% glycerol: 1.3 μM oligo, 0.7 μM core enzyme, 5.0 μM β′240-309, 5.0 μM β′240-309 L268C/L306C, and 0.5 μM σ⁷⁰. Purified RNA polymerase core enzyme, wild-type and oxidized β′240-309 peptides, and σ⁷⁰ were preincubated separately for 20 min at 37°C to ensure maximal activity. Complexes were allowed to form for 20 min at 37°C and then run on a native 12% Tris-glycine polyacrylamide gel at 120 V for 2.5 h at 4°C. The gel was imaged and fluorescence was measured on a Molecular Dynamics Storm system under red fluorescence mode. The intensity of the entire gel field (see Fig. 4, lanes 7 to 10) was increased to ensure better photographic reproducibility.

FIG. 4.

Native gel shift assay used to determine −10 nontemplate strand oligo binding activity of the oxidized β′240-309 L268C/L306C-σ⁷⁰ complex. Proteins were preincubated at 37°C and then added to the reaction tubes containing either Cy5-labeled −10 nontemplate strand oligo or Cy5-labeled nonspecific control oligonucleotides. Samples were incubated for 30 min at 37°C and analyzed by native gel electrophoresis. Numbers indicate concentrations of components in micromolar. Shifted complexes were visualized on a Storm system under red fluorescence mode. The intensity of the gel region comprising lanes 7 to 10 was increased to improve visualization of the β′240-309 complexes. Positions of the complexes formed and free oligo are indicated on the right.

RESULTS

With the increased availability of protein sequence data and structural information, computational methods have been developed for predicting the structure of a given amino acid sequence by comparison with a library of known protein structures. In particular, a fold recognition algorithm has been developed in order to identify residues that participate in disulfide bond formation within proteins of unknown structure (12). This algorithm uses the following criteria: (i) a cysteine pair must have C_α and C_β atoms positioned so that a disulfide bond of normal length can form; (ii) appropriate C_α—C_β—S bond angles (χ₃ torsion angles of ±90°) (31, 39); and (iii) at least two residues between cysteines are present for optimal formation of a disulfide bond (31). The algorithm threads the amino acid sequence onto the C_β backbone of the known structure and ranks the likelihood of disulfide bond formation between each pair of amino acid residues. This method requires that a crystal structure of the given protein is known; however, in the absence of a structure, secondary structure predictions and threading onto a similar protein structure (i.e., Swiss Model) could potentially be used if the degree of sequence identity was particularly high. When tested against proteins of known structure in the Protein Data Bank, this method predicted the location of residues that form disulfide bonds with over 80% accuracy (12). Although the algorithm was successful in silico, predictions for disulfide bond formation have never been tested in vitro. We utilized this algorithm to engineer a disulfide bond within the β′240-309 region of E. coli RNA polymerase and addressed whether possible conformational changes were required during binding of this region to the sigma factor σ⁷⁰.

Two amino acids within β′240-309, L268 and L306, when substituted with cysteine have a strong potential for disulfide bond formation (estimated χ₃ torsion angle of 89.1°). The position of these residues and a model of the disulfide bond within the X-ray crystal structure of T. aquaticus corresponding to E. coli β′260-309 are presented in Fig. 1. Secondary structure predictions (PHD; Swiss Model) suggest that these cysteine substitutions should not significantly alter secondary structure, and therefore the proper coiled-coil formation is retained (35). In addition, these two residues are not thought to have direct interaction with σ⁷⁰.

FIG. 1.

Engineering of a disulfide bond within the β′260-309 region of RNA polymerase. (A) Position of the disulfide bond within β′260-309. The disulfide bond between L268C and L306C is shown as a stick model on the structure of T. aquaticus RNA polymerase. The numbering system has been converted from the T. aquaticus system to the E. coli numbering system. (B) Modeled representation of the engineered disulfide bond on the structure of T. aquaticus β′260-309.

An overexpression vector, pLA28, containing β′240-309 with the L268C and L306C mutations was created and transformed into BL21(DE3)pLysS for overexpression. Initial experiments using normal growth conditions (Luria-Bertani [LB] medium and standard flasks in an air shaker at 37°C) resulted in poor induction, and dark cell pellets like those obtained under anaerobic growth were observed. We hypothesize that the ideal positioning of the cysteines might cause disulfide bond formation in the cell, resulting in a perturbation of the redox environment of the cell requiring additional oxygen and NADH to compensate. By performing several induction trials, it was determined that optimal expression of β′240-309 L268C/L306C occurred when 0.2% glucose was added to the standard LB culture medium and baffle flasks were used to increase aeration. Both the wild-type and mutant β′240-309 peptides were purified in a denatured form by Ni-NTA chromatography in the presence of 2-ME. The strong denaturant and 2-ME were found necessary to prevent oxidation and to reduce the formation of peptide multimers during purification. Once purified, air oxidation of the purified peptides was performed by dialysis against buffer without reductant under dilute conditions to minimize the possibility of obtaining interpeptide disulfide bonds and potential multimer formation.

To verify that a disulfide bond had formed, samples of the purified proteins were subjected to SDS-PAGE under reducing and nonreducing conditions. SDS-PAGE allowed us to discriminate between the reduced and oxidized species because of their differences in Stokes' radius. If a disulfide bond formed within the oxidized β′240-309 peptide, the protein should migrate differently than the reduced peptide. Both the reduced (Fig. 2, lane 3) and oxidized (Fig. 2, lane 4) β′240-309 L268C/L306C peptides ran at higher apparent molecular masses than the wild-type β′240-309 peptide (Fig. 2, lanes 1 and 2), which is probably due to the addition of two negatively charged cysteine resides. There is approximately a 1-kDa difference in mobility between the reduced and oxidized β′240-309 L268/L306C peptides. As expected, the observed difference in mobility may be due to the oxidized peptide having an irregular Stokes' radius that slightly impedes migration through the polyacrylamide matrix. Another explanation is that the restricted conformation of the oxidized form might prevent full denaturation, and less SDS binding by the peptide would affect its mobility. The difference in mobility also confirms that all of the dialyzed β′240-309 L268C/L306C protein is oxidized. It is worth mentioning that the oxidized β′240-309 L268C/L306C peptide, when reduced, ran at the same apparent molecular mass as a purified reduced β′240-309 L268C/L306C peptide (data not shown). If both reduced and oxidized species were present, then two protein bands would have been visible in the SDS-PAGE.

FIG. 2.

SDS-PAGE of the wild-type and L268C/L306C β′240-309 peptides under reducing and oxidizing conditions. To create a reducing environment, 20 mM 2-ME was added to the sample buffer. Lane 1, wild-type reduced β′240-309; lane 2, wild-type oxidized β′240-309; lane 3, reduced L268C/L306C; lane 4, oxidized L268C/L306C. Lane M is broad-range protein molecular mass markers (Novagen). Sizes in kilodaltons are indicated on the right.

An unidentified protein band of ≈23 kDa is present in both the reduced and oxidized β′240-309 L268C/L306C preparations (Fig. 2, lanes 3 and 4). This band may be disulfide-linked peptide dimer resulting from sulfhydryl oxidation during migration through the SDS-PAGE stacking gel (11). While this contaminant could not be removed by chromatographic methods, the 23-kDa protein is shown not to be involved in formation of the β′240-309-σ⁷⁰ protein complex (see below).

Native gel shift assay to determine σ⁷⁰ binding activity.

In order to determine whether or not the β′ coiled-coil changes conformation upon binding to σ⁷⁰, the β′240-309 L268C/L306C peptides were assayed for σ⁷⁰ binding by native gel shift assay under both reducing and oxidizing conditions. Wild-type β′240-309 and β′240-309 L268C/L306C peptides were added to wild-type σ⁷⁰ in either equimolar amounts (1 μM and 1 μM) or with a fivefold excess of β′ peptide (5 μM and 1 μM), and complexes were analyzed via native PAGE.

There are four possible outcomes to the gel shift assay. The first outcome, where binding occurs under both oxidizing and reducing conditions, would be expected if the cysteine substitutions and the disulfide bond do not alter σ⁷⁰ binding. Therefore, one could conclude that the cysteines do not disrupt the β′240-309 coiled-coil and that β′240-309 does not undergo a significant conformational change upon σ⁷⁰ binding (i.e., an opening up of the coiled-coil structure). The second outcome, where binding occurs only under reduced conditions, could be explained if the cysteine substitutions do not alter σ⁷⁰ binding but formation of the disulfide bond prevents a necessary conformational change in β′240-309. A second explanation for this outcome could be that the disulfide bond forms, distorting the secondary structure of the region enough to prevent σ⁷⁰ binding. The third outcome, where binding takes place only under oxidizing conditions, may occur if the cysteine substitutions alter the coiled-coil and, thus, σ⁷⁰ binding. However, oxidation and formation of the disulfide bond allows formation of the correct coiled-coil structure, allowing σ⁷⁰ to bind. Finally, if the mutant peptide is unable to bind σ⁷⁰ under any condition, this would indicate that the cysteine substitutions are detrimental to σ⁷⁰ binding, yielding little information about conformational changes in β′240-309.

Results from the native gel shift assay are presented in Fig. 3A. Because of the highly acidic nature of σ⁷⁰ (7), it migrates approximately halfway into the native Tris-glycine gel. When wild-type β′240-309 is added to σ⁷⁰, mobility through the polyacrylamide matrix is retarded, indicating that a stable β′240-309-σ⁷⁰ complex is formed (Fig. 3A, lanes 2 and 3). Under reducing and oxidizing conditions, formation of the wild-type β′240-309-σ⁷⁰ complex occurs, indicating that this assay can be used to determine defects in interactions between these two proteins (data not shown). When the β′240-309 L268C/L306C peptide is added to σ⁷⁰ under reducing conditions, the stable complex is unable to form, indicating that the cysteine substitutions disrupt binding (Fig. 3A, lanes 5 and 6). However, under oxidizing conditions, the β′240-309 L268C/L306C peptide can bind σ⁷⁰, resulting in the formation of a stable complex that is nearly identical to the wild-type complex in terms of binding activity and mobility through the native gel (Fig. 3A, lanes 8 and 9). These results suggest that the coiled-coil conformation is able to bind σ⁷⁰ and that a major conformational change in the β′240-309 coiled-coil is not required for σ⁷⁰ binding. It is important to note that a locked coiled-coil structure, such as the one introduced by disulfide bond formation, is not needed for σ⁷⁰ binding. The wild-type β′240-309 peptide, which is not restricted by a disulfide bond, is able to bind σ⁷⁰ with a flexible coiled-coil.

FIG. 3.

σ⁷⁰ binding activities of the reduced and oxidized β′240-309 L268C/L306C peptides. (A) Native gel shift binding assay results. β′240-309 peptides were added at either a one- or fivefold excess over σ⁷⁰. Numbers indicate concentrations of components, in micromolar. Positions of free σ⁷⁰ and the σ⁷⁰-β′240-309 complex are indicated on the left. (B) SDS-PAGE of σ⁷⁰-β′240-309 L268C/L306C complex. The complex band (panel A, lane 9) was excised from the gel and subjected to SDS-PAGE. Lane M is broad-range protein molecular mass markers (Novagen). Positions of σ⁷⁰ and the β′240-309 L268C/L306C peptide are indicated on the right.

Because of the unidentified ≈23-kDa protein band present in the purified β′240-309 protein preparations (Fig. 2, lanes 3 and 4), the oxidized β′240-309 L268C/L306C-σ⁷⁰ complex was excised from the native gel (Fig. 3A, lane 9) and subjected to SDS-PAGE (Fig. 3B). This gel showed that only σ⁷⁰ and β′240-309 L268C/L306C were present within the shifted complex, confirming that the 23-kDa contaminant is not a binding partner.

Under reducing conditions, the β′240-309 L268C/L306C peptide was unable to bind σ⁷⁰. This result also provides confirmation that the gain of function (σ⁷⁰ binding activity) is due to the presence of the disulfide bond. This lack of σ⁷⁰ binding by the β′ peptide when reduced could be due either to steric hindrance of the cysteine side chains or to loss of hydrophobic contacts necessary for formation of the proper coiled-coil conformation. To determine the cause of the defect, we substituted one or both cysteines with serines, both of which resulted in a peptide which was unable to bind σ⁷⁰ (data not shown). This suggests that disruption of the hydrophobic core of the coiled-coil is probably responsible for the lack of binding seen with the β′240-309 L268C/L306C peptide under reducing conditions.

Binding activity of −10 nontemplate strand oligonucleotide.

In order to examine whether a conformational change in the β′240-309 coiled-coil is required for binding the β′240-309-σ⁷⁰ complex to a −10 nontemplate strand oligonucleotide, we performed a native gel shift assay using a 5′-Cy5-conjugated −10 nontemplate oligo and an anticonsensus control oligo. Complexes were visualized using the red fluorescence mode of the Storm imaging system and by staining the gel with GelCode Blue reagent.

The β′240-309 peptides were tested for their ability to induce the conformational change within σ⁷⁰ necessary for specific binding of a −10 nontemplate oligo. As expected, neither σ⁷⁰ nor the β′240-309 peptides (wild type or L268C/L306C) alone shifted the −10 nontemplate strand oligo (Fig. 4, lanes 1, 3, and 4) or the control oligo (data not shown). Wild-type β′240-309, when added to σ⁷⁰, is unable to bind to the control oligo (Fig. 4, lane 7), but upon addition of the −10 nontemplate strand oligo the β′240-309-σ⁷⁰ complex shifts the oligo (Fig. 4, lane 8). This suggests that the β′240-309-σ⁷⁰ complex specifically recognizes the −10 nontemplate strand oligo. The oxidized β′240-309 L268C/L306C peptide, which is locked in the coiled-coil conformation, also forms a complex with σ⁷⁰ that is able to shift the −10 nontemplate oligo (Fig. 4, lane 10) but not the control oligo (Fig. 4, lane 9). Again, the oxidized β′240-309 L268C/L306C-σ⁷⁰ complex activity is nearly that of the wild-type in terms of oligo binding activity and mobility in the native gel. The reduced β′240-309 mutant was not tested in this assay since it was deficient in interacting with σ⁷⁰.

A second band is also visible in the lanes that contain β′240-309-σ⁷⁰-oligo complexes (Fig. 4, lanes 8 and 10). Previous protein-protein native gel shift assays have shown that σ⁷⁰ exhibits some potential for dimerization. Because the σ⁷⁰ dimer is only able to shift the −10 nontemplate oligo in the presence of β′240-309 (Fig. 4, compare to lane 1), this dimer complex must contain two molecules of σ⁷⁰, at least one molecule of β′240-309, and the 5′-Cy5-labeled −10 nontemplate oligo.

To validate the obtained results, we also tested the two forms of RNA polymerase for their ability to bind the −10 nontemplate and control oligos. Core enzyme exhibited nonspecific binding to the −10 nontemplate oligo (Fig. 4, lane 2). Holoenzyme showed some nonspecific binding to the control oligo (Fig. 4, lane 5) but, based on comparison to core enzyme in lane 2, this could be due to nonspecific binding by residual core in the holoenzyme prep. Upon addition of the −10 nontemplate oligo (Fig. 4, lane 6), holoenzyme shifts the oligo, confirming that holoenzyme specifically recognizes the −10 nontemplate strand (29).

These experiments suggest that the β′240-309 coiled-coil is not only sufficient to bind σ⁷⁰ but also that the locked coiled-coil conformation is capable of inducing σ⁷⁰ to specifically bind to the −10 hexamer of the nontemplate strand. Therefore, a conformational change in the coiled-coil is not likely during the initial steps of transcription initiation, at least through open promoter complex formation.

DISCUSSION

The process of prokaryotic transcription proceeds through multiple steps: holoenzyme formation, closed complex formation where promoter DNA is bound, melting of the promoter DNA during open complex formation, and formation of an elongation complex (10, 16). During the steps leading up to elongation complex formation, RNA polymerase core enzyme and σ are the only two proteins involved. This suggests that conformational changes may be necessary to transform RNA polymerase and σ⁷⁰ into the various forms necessary to complete the different stages of transcription initiation. However, detailed information about the specific regions of σ⁷⁰ and core enzyme involved in conformational changes is minimal.

Interactions between σ⁷⁰ and RNA polymerase core enzyme where conformational changes are known to occur can be broken down into three possible distinct steps: docking and protein binding (where amino acid contacts are made); closed complex formation (where secondary amino acid contacts may occur to establish interactions within the double-stranded promoter); and open complex formation (where interactions between polymerase and σ⁷⁰ increase the binding affinity of σ⁷⁰ region 2.3 for the −10 nontemplate strand within the transcription bubble greater than 200-fold) (9, 21). Evidence of the conformational change upon binding of σ⁷⁰ to core enzyme includes luminescence resonance energy transfer experiment results, which suggest that a large conformational change occurs where σ⁷⁰ regions 1 and 4.2 move relative to region 2.2, thereby unmasking the DNA-binding regions of σ⁷⁰ (8, 9). The second step, where conformational change occurs upon −10 nontemplate strand interaction, shows an inconsistency between the theoretical buried surface area upon oligo binding and values measured. A 3,000-Ų area of nonpolar surface appears to be buried upon complex formation, but the theoretical maximal nonpolar interaction of σ⁷⁰ with the oligo could only involve approximately 500 Ų of buried surface area (14). These results suggest that the burial of the remaining nonpolar surface upon −10 nontemplate strand binding (≈2,500 Ų) is due to a major conformational change in RNA polymerase. The final step showing conformational change is during the transition between the closed complex and open complex, where two kinetically significant intermediates have been identified (33, 34). Isomerization between these two intermediates results in a large negative ΔC°_p, which is equated to a major conformational change in RNA polymerase during this step (34). In addition to the kinetic and thermodynamic studies of transcription initiation, modeling of the elongation complex, based upon the structure of T. aquaticus core RNA polymerase, illustrates that a major conformational change involves closure of the β and β′ jaws, which would help to keep the enzyme associated with DNA during elongation (25, 30).

In addition to the information known regarding conformational changes in transcription initiation, molecular interactions between core enzyme and σ⁷⁰ have been identified. Recent work has identified regions of the β subunit, including aa 900 to 909 and 1,060 to 1,240 as making initial contacts with region 1.1 and region 3 of σ⁷⁰. Regions 2.1 and 2.2 of σ⁷⁰ have been identified as a major site involved in the initial binding of core RNA polymerase (18, 27, 36). Ordered-fragment far-Western analysis and Ni-NTA coimmobilization experiments demonstrated that aa 260 to 309 of the β′ subunit comprise a major interaction domain for σ⁷⁰ binding (3). The predicted secondary structure of β′260-309, two amphipathic α-helices joined by a loop, is consistent with that of an antiparallel coiled-coil, a motif common in protein-protein interactions (2). The 3.3-Å X-ray crystal structure of T. aquaticus core RNA polymerase (43) verified that this region does adopt a coiled-coil conformation. Site-directed mutagenesis of the β′260-309 coiled-coil revealed several mutations, mostly clustered on the upper face, that disrupted the σ⁷⁰-core interaction (2). Finally, the use of computer-aided protein docking, in conjunction with the mutagenesis data, has provided a visual model which supports the amino acid interactions between the β′ coiled-coil and σ⁷⁰ region 2.2 (4).

With the knowledge that conformational change occurs during transcription initiation and with modeling and experimental results implicating the β′ coiled-coil as an interaction domain, it was hypothesized that the β′ coiled-coil might undergo a dynamic conformational change upon interaction with σ⁷⁰. Three lines of evidence caused us to speculate that the β′ coiled-coil could open up to form new interactions with σ⁷⁰. First, coiled-coil motifs are often protein interaction sites, and rearrangement of helices to form an intersubunit coiled-coil could be one mechanism of protein interaction between the β′ coiled-coil and the α-helices in region 2 of σ⁷⁰ (13, 19). Second, prior mutational data revealed several β′ mutants that had increased or decreased affinity for σ⁷⁰ (2, 4). These mutations were located on the underside of the coiled-coil away from the residues implicated in binding σ⁷⁰. This could suggest that the β′ coiled-coil undergoes a conformational change upon binding σ⁷⁰, making secondary amino acid contacts between the underside of the β′ coiled-coil and σ⁷⁰. Finally, defects in open complex formation caused by mutations in region 2.2 have been used to suggest that σ⁷⁰ may undergo reorganization of the helices in region 2 to facilitate DNA binding (24). However, it was not previously known if a conformational change in the β′ coiled-coil is also necessary for this reorganization.

Conformational changes in coiled-coil motifs have been previously found to be important in protein binding. In Salmonella enterica serovar Typhimurium, the TlpA protein contains a coiled-coil motif which opens and undergoes helical exchange to form both homo- and heterodimers in the process of forming filament-like structures (22, 23). Not only do coiled-coils undergo conformational changes upon multimerization, they have also been found to participate in strand exchange with coiled-coil domains in binding partners. The small subunit of smooth muscle myosin light chain phosphatase dimerizes via coiled-coil interactions, and it is also thought to form heterodimers with its targeting subunit by coiled-coil strand exchange (26). In addition, the toxin colicin E1 from E. coli contains a coiled-coil domain which undergoes a dramatic conformational change, completely unfolding the coiled-coil upon interaction with the membrane-bound colicin E1 receptor protein (15).

Because nuclear magnetic resonance analysis was not possible due to the large size of RNA polymerase, the advent of a computer prediction algorithm for disulfide bond engineering (12) provided a means of probing for conformational changes necessary for function of specific regions in large proteins. In this work, we utilized this algorithm to engineer a non-naturally occurring disulfide bond within the β′240-309 coiled-coil to inhibit conformational change, such as separation of the α-helices, within this peptide. The presence of the disulfide bond at one end and the short 5-aa loop at the other end effectively lock the coiled-coil in planar conformation, where the distance and the angle between the two α-helices could not be changed without breaking either disulfide or peptide bonds. Although the disulfide bond restricts the coiled-coil as a single fixed unit, our studies do not address movement of this unit in the context of the full core enzyme. It is possible that the whole β′ coiled-coil moves during the process of holoenzyme formation and transcription initiation. This movement could be similar to a hinge mechanism, where the entire coiled-coil moves as one unit, upward or downward with respect to the rest of RNA polymerase. In addition, our experiments do not rule out the possibility that the loop region connecting the α-helices may undergo slight perturbation during this process. Our experiments provide evidence that a major change in the conformation of the β′ coiled-coil (aa 240 to 309) is not required for binding to σ⁷⁰ or for inducing a conformational change within σ⁷⁰ that allows σ⁷⁰ to recognize a −10 nontemplate strand oligonucleotide. Because the β′ coiled-coil is a primary σ⁷⁰ interaction site involved in holoenzyme formation, our results have important implications for understanding and modeling how these proteins interact during the initiation of transcription.

Prior modeling of the binding of σ⁷⁰ and core enzyme in transcription initiation utilized available crystal structures of each individual protein (4, 42). These structures provide information on only one conformational state of the enzyme, and therefore any predicted interactions are based upon docking of rigid proteins, which cannot account for any conformational changes. Our studies involving the β′ coiled-coil show that the coiled-coil conformation is retained through the initial steps of holoenzyme formation and −10 nontemplate strand binding. We conclude that the β′260-309 coiled-coil can be fixed in modeling and docking calculations, and the direct use of the crystal structure of this region is justified in transcription initiation. In contrast, the proposed major conformational changes in σ⁷⁰ during holoenzyme formation currently prohibit the use of the σ⁷⁰ crystal structure in rigid docking models. Further characterization of possible conformational changes in σ⁷⁰ must be performed.

Introduction of the β′ L268C-L306C disulfide bond into full-length β′ and core enzyme would allow one to examine later steps in the transcription initiation process. Unfortunately, due to the large number of cysteines, some of which are essential, in the full-length β′ and other polymerase subunits, verification of formation of the disulfide bond within core enzyme under oxidizing conditions would be difficult. We are currently in the process of engineering disulfide bonds within σ⁷⁰ to determine at what steps in transcription certain regions of σ⁷⁰ are required to undergo conformational changes. We anticipate that our results, combined with future crystal structures of the holoenzyme, open complex, and elongation complex, will further the understanding of the essential conformational transitions that RNA polymerase undergoes during transcription initiation.