Conformational Flexibility in ς70Region 2 during Transcription Initiation

Prokaryotic RNA polymerase holoenzyme is composed of core subunits (α2ββ′ω) plus a ς factor that confers promoter specificity allowing for regulation of gene expression. Holoenzyme is known to undergo several conformational changes during the multiple steps of transcription initiation. However, the effects of these changes on the functions of specific regions have not been well characterized. In this work, we addressed the role of possible conformational change in region 2 of Escherichia coli ς70 by engineering disulfide bonds that “lock” region 2.1 with region 2.2 and region 2.2 with region 2.3. When these mutant holoenzymes were characterized for gross defects in multiple-round transcription, we found that insertion of either disulfide bond did not result in a fundamental block, indicating that the disulfide-containing holoenzymes are active. However, both disulfide-containing holoenzymes exhibited defects in formation and stability of the open complex. Our results suggest that conformational flexibility within ς70 region 2 facilitates open complex formation and transcription initiation.

RNA polymerase, a multi-subunit enzyme that is made up of five polypeptide chains (␣ 2 ␤␤Ј), is solely responsible for the synthesis of messenger, transfer, and ribosomal RNA in the bacterial cell. This enzyme has two distinct functional forms: core enzyme and holoenzyme. Although core enzyme is catalytically active and capable of transcription elongation, it is unable to initiate transcription at specific promoter sequences. Promoter recognition is accomplished by binding of the specificity factor , which positions holoenzyme on the promoter sequence (1,2). Through the use of seven factors, Escherichia coli is able to direct transcription from multiple sets of promoter sequences, which confers the ability to regulate gene expression (3,4). Amino acid sequence alignment of factors within the 70 family identified four regions of sequence homology (regions 1-4) (5). Region 2, which contains core binding and Ϫ10 promoter recognition determinants, accounts for the most sequence similarity among 70 family members. This suggests that this region plays an important role in transcription. Because 70 region 2 was shown to be a major interaction domain with the ␤Ј coiled-coil (6 -8), this region may be the site of functionally necessary conformational changes upon interaction with core enzyme.
Based upon the sum of prior mechanistic, x-ray diffraction, electron microscopy, and luminescence resonance energy transfer studies, RNA polymerase core enzyme and 70 are thought to undergo multiple conformational changes through the process of transcription initiation: from binding, to formation of the closed complex, to stable open complex formation, to transcription initiation, and finally to factor release (3, 7, 9 -13). Evidence for intermediates during this process have been identified using the well studied lacUV5 promoter (12, 14 -16). The currently accepted pathway for open complex formation based on this promoter is shown below (15).
R is RNA polymerase holoenzyme, and P is the lacUV5 promoter. RP c is the closed complex where contacts with doublestranded promoter DNA are established. RP i is a kinetically significant intermediate on the pathway to the open complex (RP o ), where melting of the promoter DNA and formation of the transcription bubble occurs. Although conformational changes are thought to accompany many if not all of these steps, the nature and importance of these changes at each sequential step have not been characterized.
Identification of conformational changes accompanying formation of protein-DNA complexes necessitates alternative means of analysis, primarily because traditional methods such as NMR are not suitable for larger multi-subunit protein complexes such as RNA polymerase. Both circular dichroism and NMR require high protein concentrations (micromolar to millimolar) at which RNA polymerase has been shown to aggregate (17,18), making these spectroscopic approaches not useful for analyzing this system. Methods previously utilized to address conformational changes in RNA polymerase holoenzyme, such as fluorescence resonance energy transfer (FRET) and thermodynamic measurements (9,13), allowed the observation of gross conformational changes but were not able to examine specific regions of the proteins or changes during the sequential steps of transcription. Information from the recently available crystal structures of Thermus aquaticus RNA polymerase (19 -21) provided static snapshots of different conformational states of the enzyme; however, transcription is likely to be a continuous and dynamic process. Additionally, the flexibility of accessible mobile regions of a large multi-subunit enzyme may be extremely sensitive to crystal packing interactions or the solution environment, preventing small conformational changes from being seen.
To address these issues, we have utilized a freely available disulfide bond recognition algorithm "Disulfide by Design" (22) to strategically place disulfide bonds within a protein, restricting the conformational flexibility of that region. If no functional consequence arises at a particular mechanistic step as a result of the inserted disulfide bond, one can argue that no conformational change is required in that region of the protein at that time. However, if function of the protein is inhibited, then it strongly argues that some conformational flexibility is essential during that step. Although the introduction of non-native disulfide bonds has been used in the past to examine conformational change and protein stability (23)(24)(25)(26), success rates were often low because of the lack of an algorithm to accurately predict the location at which a disulfide bond was most likely to form. The Disulfide by Design algorithm correctly predicted the location of amino acid residues that form disulfide bonds with over 80% accuracy when tested against proteins of known structure (22). This algorithm was successfully used to engineer a disulfide bond within the coiled-coil domain of the ␤Ј subunit of RNA polymerase to examine whether conformational change was required for 70 binding and binding of a nontemplate strand oligonucleotide by 70 (27).
To further our knowledge of the gross conformational changes in 70 previously analyzed by luminescence resonance energy transfer and thermodynamic studies, we investigated whether the ␣-helices in region 2 of E. coli 70 undergo conformational changes during transcription initiation. Using the Disulfide by Design algorithm, we have engineered two disulfide bonds within the essential region 2 of 70 and assayed for the functional consequences of restricting conformational change during transcription initiation. Both mutant holoenzymes exhibit defects in open complex formation and in the stability of the open complex, suggesting that some conformational flexibility is necessary to facilitate transcription initiation.

EXPERIMENTAL PROCEDURES
All of the reagents were purchased from Sigma unless otherwise indicated. Spectrophotometric grade glycerol was purchased from Fisher. Ni-NTA-agarose 1 was purchased from Qiagen.
Disulfide Bond Engineering-The amino acid sequence of E. coli 70 region 2 (amino acids 360 -440) was analyzed using a disulfide recognition algorithm (22) designed to recognize cysteine pairs that are properly oriented to favor disulfide bond formation. This algorithm is part of a Windows-based program called Disulfide by Design that is freely available by e-mailing the author of the program at domski@wayne.edu. This analysis revealed two potential amino acid pairs in region 2, Val 387 /Gly 408 and Ala 415 /Trp 434 , that when substituted with cysteines have a strong potential for forming a disulfide bond ( 3 torsion angles of 91.6 and 96.7°, respectively).
Plasmid Construction-Plasmid characteristics are described in Table I. A cysteine-less variant of 70 was made by the addition of C442S through site-directed mutagenesis using plasmid p[S442C] 70 as template (28). The region containing the substituted serines was subcloned by transferring the PstI-HindIII fragment into pHMK-His 6 -70 (29) to produce plasmid pLA55. Cysteine and serine substitutions at positions Val 387 , Gly 408 , Ala 415 , and Trp 434 of 70 were generated by site-directed mutagenesis of pLA55 using standard QuikChange mutagenesis (Stratagene). Oligonucleotide sequences are available upon request. All of the plasmids generated by PCR were sequenced to ensure that spurious secondary mutations had not been incorporated.
Expression and Purification of Mutant Holoenzymes-Plasmids pHMK 70 and pLA55 through pLA81 were transformed into BL21(DE3)pLysS (Novagen) for overexpression. The cells were grown to an A 600 nm of 0.4 -0.6 in 2-liter flasks at 30°C in LB supplemented with 50 g/ml kanamycin and 10 g/ml chloramphenicol. Isopropyl-␤-D-thiogalactoside was used for induction at a final concentration of 0.5 mM. The cells were harvested 3 h after induction by centrifugation at 8,000 ϫ g for 15 min and frozen at Ϫ80°C until use.
All of the RNA polymerase holoenzymes were purified by the protocol below. The cell pellets were lysed in Ni-NTA buffer, sonicated for two 1-min intervals, and centrifuged at 15,000 ϫ g for 20 min. The resulting pellets contained mostly 70 inclusion bodies, which were discarded. The supernatants were diluted to 35 ml with Ni-NTA buffer and incubated with 3 ml of pre-equilibrated Ni-NTA-agarose (Qiagen) for 1 h at 4°C. Supernatant-agarose suspensions were poured into 15-ml Econopak columns (Bio-Rad) and washed extensively with 20 column volumes of Ni-NTA buffer. Low salt conditions (50 mM NaCl) were used to avoid possible dissociation of the holoenzyme subunits. RNA polymerase holoenzymes were eluted with ten 1-ml fractions of Ni-NTA buffer with 300 mM imidazole.
Because the holoenzyme preparations were still heavily contaminated with cell protein, immunoaffinity chromatography was used to further purify the preparation. The pooled Ni-NTA fractions were diluted 3-fold with TE buffer and loaded onto 0.5-ml immunoaffinity columns containing the polyol-responsive anti-␤Ј monoclonal antibody NT73 (30). The columns were washed with 10 column volumes of TE buffer with 0.15 M NaCl, and eluted with 5 ml of TE buffer containing 0.75 M NaCl with 30% propylene glycol. Fractions containing holoenzyme were pooled, concentrated in 4 ml of 100-kDa cut-off Ultrafree-4 concentrations (Millipore), and dialyzed into storage buffer. Final holoenzyme concentrations in storage buffer were in the range of 100 -200 g/ml and assume 100% active protein. Stoichiometry of the purified holoenzymes was determined by densitometry comparing the ratios of and ␣ subunits using the software package Un-Scan-It Gel (Silk Scientific). Each holoenzyme contained a ratio of 1 subunit to 2 ␣ subunits (Ͻ 10% deviation), indicating that all purified holoenzymes are in proper stoichiometry.
IC5-PE-maleimide Labeling-To verify disulfide bond formation within the subunit of the purified holoenzymes, each holoenzyme was labeled with the fluorescent dye IC5-PE-malemide (Dojindo). Each holoenzyme (0.7 g) was added to buffer containing 50 mM Tris-HCl (pH 7.9), 5.0 M urea, and 10 pmol of IC5-malemide. The samples were incubated at 37°C for 30 min and subjected to electrophoresis along with unlabeled control samples on a 4 -12% NuPAGE Bis-Tris gel with MES buffer. The gel was then visualized on a Molecular Dynamics Typhoon system (Cy5 filter) and later stained with Gel-Code Blue reagent (Pierce).
DNA Purification-DNA concentrations in all of the experiments were determined by absorbance at 260 nm where 1 A 260 nm ϭ 50 g/ml of double-stranded DNA. Supercoiled pRLG593 template (31) for the multiple-round transcription experiments was purified by Midi-prep (Promega), followed by two phenol/chloroform extractions, and finally an ethanol precipitation.
Linear templates (from lacUV5 Ϫ60 to ϩ39) used for the closed complex gel shifts, DNase I footprinting, and filter binding experiments were prepared by digestion of pRLG593 DNA with HindIII, labeling of  Closed Complex Formation Assay-To test for defects in closed complex formation, 12.5 nM of each holoenzyme was added to 2 nM 32 P end-labeled linear lacUV5 template (Ϫ60 to ϩ39) in buffer containing 40 mM Tris-HCl (pH 7.9), 150 mM KCl, 0.1 mM EDTA, 5 mM MgCl 2 , 10% glycerol, and 0.1 g/l BSA. All of the reactions were prechilled and assembled on ice. After incubation on ice for 15 min, the samples were loaded onto a 5% native polyacrylamide gel (prechilled to 0 -2°C) for electrophoresis at 250 V for 1.5 h at 2°C. The gels were then dried and visualized by phosphorus imaging.
DNase I Footprinting-Holoenzyme (1-25 nM) was added to 1 nM 32 P end-labeled linear lacUV5 template (Ϫ60 to ϩ39) in buffer containing 50 mM Tris-HCl (pH 7.9), 50 mM NaCl, 10 mM MgCl 2 , and 65 g/ml of BSA. The samples were incubated for 20 min at 37°C and treated with heparin (10 g/ml) for 30 s, followed by digestion with DNase I at 8 g/ml for 30 s. The samples were electrophoresed on 9% polyacrylamide, 8 M urea gels, dried, and visualized by phosphorus imaging.
Filter Binding Assay-The fraction of RNA polymerase molecules able to form open complexes was measured by filter binding assays. The various holoenzymes (20 nM) were incubated with 2 nM 32 P end-labeled linear lacUV5 template (Ϫ60 to ϩ39) for 30 min at 37°C in buffer containing 50 mM Tris-HCl (pH 7.9), 5 mM MgCl 2 , 100 mM NaCl, and 0.1 g/l BSA. The samples were removed after challenge with 50 g/ml heparin for 30 s, vacuum filtered onto nitrocellulose filters (BA85, Schleicher & Schuell), and washed with 1 ml of buffer. The 32 P-labeled DNA retained on the filters was measured by Cerenkov counting and corrected for background DNA retained in the absence of holoenzyme (Ͻ3% of input). The results were normalized to the cysteine-less 70 holoenzyme, which is represented as having 100% activity.

RESULTS
RNA polymerase core enzyme and 70 are known to undergo several conformational changes throughout the process of transcription initiation (3,7,10); however, the large size of these multisubunit protein-DNA complexes precludes traditional ways of probing for conformational changes, such as NMR analysis. In this work, we used a previously successful disulfide recognition algorithm (22,27) to probe for possible conformational changes in 70 region 2 during the steps of transcription initiation.
This algorithm threads a given amino acid sequence onto the C␣␤ backbone of a known crystal structure and ranks each pair of amino acid residues for the likelihood of disulfide bond formation. Two pairs of residues within 70 region 2, Val 387 /Gly 408 (2.1C-2.2C) and Ala 415 /Trp 434 (2.2C-2.3C), were shown to have strong potential for disulfide bond formation when substituted with cysteines (estimated 3 angles of 91.6 and 96.7°, respec- tively, where 90°is optimal). The positions of these residues within the x-ray crystal structure of the protease-resistant domain of E. coli 70 (32) are presented in Fig. 1. An example of modeling the Cys 387 -Cys 408 disulfide bond within 70 region 2 shows the orientation of the cysteine side chains and the 92° 3 torsion angle (Fig. 1B). Secondary structure predictions (PHD, Swiss model) indicated that substitution of all four residues with cysteine was unlikely to perturb the ␣-helical structure of this region.
Purification and Verification of Disulfide Bond Formation-All of the cysteine substitutions were introduced via site-directed mutagenesis into pLA55, a pET-derived vector containing a His 6 -tagged cysteine-less variant of 70 . Overexpression of both disulfide-containing 70 variants yielded both large inclusion bodies, which were unable to be refolded into active monomeric protein and a small amount of soluble 70 in the cell supernatants. The majority of the soluble 70 was present as RNA polymerase holoenzyme (data not shown). Because of the difficulty in refolding these factors, holoenzyme instead was isolated using two chromatographic steps. First, holoenzyme in the cell supernatants was purified by native Ni-NTA chromatography, which resulted in eluates containing free disulfidecontaining 70 and disulfide-containing 70 holoenzyme (Fig. 2,  lanes 1 and 3). It is important to note that any wild type 70 present in the cell was removed during this step, because wild type 70 did not have a His 6 -tag. The lack of wild type 70 in the isolated holoenzyme preparations was also confirmed by fluorescent labeling, as discussed below. The Ni-NTA eluates had a significant amount of protein contamination, primarily because the column was run at relatively low salt to ensure that the mutant holoenzymes did not dissociate into core enzyme and 70 . To further purify holoenzyme, the Ni-NTA eluates were passed over an immunoaffinity column containing a monoclonal antibody to the C terminus of the ␤Ј subunit of RNA polymerase (30). Because this step is highly specific, the eluates contained pure disulfide-containing 70 RNA polymerase holoenzyme free of contaminants, with the disulfide-containing 70 in proper stoichiometry with the core subunits (Fig. 2, lanes  2 and 4) as determined by densitometry.
Once purified holoenzyme was obtained, a method was needed to verify that disulfide bond formation occurred within the mutant 70 subunit of the holoenzymes. NMR and mass spectrometry, although extremely useful, were not practical because of the large size of the multisubunit holoenzyme complex. Instead, labeling with the fluorescent dye IC5-PE-maleimide was utilized to visualize free cysteines present in the denatured holoenzyme preparations. This method is advantageous because the IC5-maleimide dye is highly fluorescent, so labeled complexes can be subjected to SDS-PAGE and visualized with the use of a Typhoon imaging instrument. Wild type 70 holoenzyme that contains three cysteines in 70 was used as a positive control for labeling. To verify that the engineered cysteines were accessible to the IC5-maleimide dye, a singlecysteine 70 holoenzyme (2.1S-2.2C) was also used as a control. Note that in all holoenzyme preparations, the ␤Ј, ␤, and ␣ subunits of RNA polymerase contain multiple cysteines that label with IC5-maleimide (Fig. 3A). In wild type holoenzyme, the three naturally occurring cysteines in 70 were labeled with the fluorescent dye (Fig. 3A, lane 1). The single cysteine in the 70 (2.1S-2.2C) subunit of holoenzyme is readily detectable (Fig. 3A, lane 5), demonstrating both the sensitivity of this assay and that the engineered cysteine residues are dye-accessible. In addition, no fluorescence was observed from the negative control, a 70 subunit of holoenzyme where the naturally occurring cysteines are replaced with serine (Fig. 3A, lane 2). This confirmed that the labeling procedure was specific for free cysteines and did not nonspecifically label other amino acid residues. No detectable fluorescence was observed for the 70 (2.1C-2.2C) or 70 (2.2C-2.3C) subunits of holoenzyme, which contain cysteine substitutions in optimal positions for disulfide bond formation (Fig. 3A, lanes 3 and 4). Because these cysteines were not labeled with IC5-maleimide, this strongly suggests that these cysteines are not accessible and form a disulfide bond with near 100% efficiency. This also indicates that the mutant holoenzyme preparations are not contaminated with wild type 70 , because no free cysteines in the 70 subunit were detected. Note that this labeling assay could not be performed on reduced holoenzyme samples, because the reductants (dithiothreitol or 2-mercaptoethanol) needed to maintain the cysteine-containing holoenzymes in a reduced state bind the IC5-maleimide dye irreversibly and prevent cysteine labeling (data not shown).
Another advantage of IC5 labeling is that attachment of this dye molecule to a free cysteine adds ϳ500 Da to the total molecular mass of the protein. This gain in molecular mass can be observed directly by a mobility shift in SDS-PAGE. As indicated above, the core subunits of RNA polymerase (␤Ј␤␣) all contain multiple cysteines that bind IC5-maleimide, thereby altering their mobility on the SDS-PAGE gel (Fig. 3B). The wild type 70 subunit of holoenzyme, which contains three naturally occurring cysteines, undergoes a mobility change of ϳ2 kDa when labeled with IC5-maleimide (Fig. 3B, lanes 1 and  2). As expected, no change in mobility is observed for the serine-substituted 70 subunit of holoenzyme (Fig. 3B, lanes 3  and 4). Likewise, no mobility change was observed for either the 70 (2.1C-2.2C) or the 70 (2.2C-2.3C) subunit of holoenzyme (Fig. 3B, lanes 5-8), indicating the presence of a disulfide bond, confirming the results from the fluorescence detection.
Multiple-round in Vitro Transcription Assays-To assay for any major defects caused by insertion of the disulfide bonds in 70 , the various holoenzymes were tested in multiple-round in vitro transcription from a supercoiled template containing the lacUV5 promoter. As a control, the cysteine-less 70 holoenzyme was shown to have a specific activity similar to that of true wild type 70 enzyme under these conditions (Fig. 4) and in previously published single-and multiple-round transcription experiments (28,33). If the engineered disulfide bond restricted a critical conformational change in 70 , one would expect to see a major defect in transcriptional activity. However, no significant difference in transcriptional activity was observed for either the 70 (2.1C-2.2C) or the 70 (2.2C-2.3C) holoenzyme, when compared with the cysteine-less holoenzyme. These results suggest that no major block in transcrip-tion from a supercoiled template occurs because of the presence of the engineered disulfide bonds within 70 and that the disulfide-containing holoenzymes have similar specific activities compared with wild type 70 holoenzyme. To confirm that the transcription phenotype of the cysteine-substituted 70 holoenzymes was due to the presence of the disulfide bond in 70 and not the amino acid substitutions, positions Val 387 (2.1S-2.2C) and Ala 415 (2.2S-2.3C) were substituted with serines. Because the cysteine and serine side chains are similar, the serine substitutions should mimic the engineered cysteine, except for the ability to form a disulfide bond. Both the 70 (2.1S-2.2C) and 70 (2.2S-2.3C) holoenzymes were greatly defective in the multiple-round transcription assay, probably because of an alteration in secondary structure, resulting from a disruption in the hydrophobic packing surface between the ␣-helices of 70 region 2. This result confirms that the wild type level of activity observed for the disulfide-containing holoenzymes is due to the presence of the disulfide bond and that maintenance of ␣-helical structure is essential for transcription.
Closed Complex Formation-Conformational changes in RNA polymerase and 70 may occur during the series of transitions that occur in a multiple-round transcription assay. However, multiple-round assays involve many post-initiation events, such as release (34 -36) and -dependent pausing (37,38), that may obscure defects in the role of 70 during transcription initiation. A small defect in one process might be missed if it is not part of a rate-determining step. To address this concern, a more detailed analysis of the transcription pathway was performed beginning with examining the ability of the holoenzymes to form closed complexes on linear lacUV5 templates (Ϫ60 to ϩ39). At low temperatures, normal holoenzymes are trapped in the closed complex form and are unable to complete the transition to the open complex. Closed complex formation was analyzed by native gel shift assays at 0°C (Fig.  5). The cysteine-less holoenzyme control forms a closed complex readily, shifting all of the labeled DNA (Fig. 5, lane 2). Likewise, the disulfide-containing holoenzymes were able to form stable closed complexes (Fig. 5, lanes 3 and 4). This indicates that the presence of the disulfide bonds within these regions of 70 does not affect closed complex formation under these conditions and that a major conformational change in 70 region 2, such as separation of the ␣-helices, is not required at this step.
Defects in Open Complex Formation-After formation of a closed complex, holoenzymes isomerize in one or more steps to an open complex, where the transcription bubble is formed and additional protein-DNA contacts are made (13,15). To examine open complex formation, the holoenzyme-DNA template complexes were analyzed by DNase I footprinting at varying holoenzyme concentrations. At the lacUV5 promoter, the region of DNase I protection extends from approximately Ϫ50 upstream of the transcription start site to approximately ϩ20, with a DNase I hypersensitivity site at Ϫ24/Ϫ25 (39). The cysteine-less holoenzyme control gave the expected protection pattern with clear DNase I protection extending to around Ϫ45 upstream, with a clear hypersensitivity site at Ϫ24/Ϫ25 (Fig. 6,  lanes 1-3). Unexpectedly, neither the 70 (2.1C-2.2C) nor the 70 (2.2C-2.3C) holoenzyme exhibited protection of the promoter DNA from DNase I cleavage (Fig. 6, lanes 4 -6 and 7-9, respectively). This result was repeated in several footprinting experiments at different temperatures (25 and 30°C) and at holoenzyme concentrations greater than 25 nM (data not shown). The lack of DNase I protection suggests that the disulfide bonds within 70 possibly affect a conformational change necessary for the transition from the closed complex to a stable open complex.
Because DNase I footprinting of the open complex requires holoenzyme to bind to the promoter DNA at high occupancy, the percentage of open complexes formed was determined by filter binding assays (40). Open complexes were allowed to form at 37°C on linear lacUV5 template (Ϫ60 to ϩ39). The reactions were challenged with heparin before filtering to remove any fast dissociating complexes (i.e. nonspecific complexes) from the population. Under these conditions, heparin binds free RNA polymerase, irreversibly preventing rebinding. The 70 (2.1C-2.2C) holoenzyme had a severe defect in promoter occupancy during open complex formation, having only 5% of the control cysteine-less 70 holoenzyme activity (Fig. 7). The 70 (2.2C-2.3C) holoenzyme was also defective for promoter occupancy having only 12% of the cysteine-less 70 holoenzyme activity. As an additional control, the serine-substituted 70 (2.1S-2.2C) holoenzyme, which was previously found to be inactive in the in vitro transcription assay, was tested and showed no occupancy of the lacUV5 template. The low promoter occupancies observed in this experiment explain why neither disulfide-containing holoenzyme was able to protect the lacUV5 template from DNase I digestion. The results from the filter binding experiments were also confirmed by single-round in vitro transcription assays from linear lacUV5 template, with the 70 (2.1C-2.2C) holoenzyme and 70 (2.2C-2.3C) holoenzyme having 8 and 34% of wild type activity, respectively (data not shown). Together, these results indicate that restriction of major conformational changes in 70 region 2, such as separation of the ␣-helices, causes defects in the transcription initiation process.
Defects in Open Complex Stability-Although the disulfidecontaining 70 holoenzymes have no apparent defects in transcriptional activity in the multiple-round transcription assay (Fig. 4), they have significant defects in promoter occupancy during open complex formation (Fig. 7). One possibility is that the restrictions imposed by the disulfide bonds within 70  The reactions were assembled, incubated at 37°C, and exposed to heparin before spotting onto nitrocellulose filter disks. 32 P-Labeled DNA retained on the filter was quantitated by Cerenkov counting and corrected for background in the absence of holoenzyme. The values were normalized to the cysteine-less 70 holoenzyme control, which is represented as 100%. The experiments represent the averages of three experiments with less than 5% deviation. examined. Holoenzyme-promoter complexes were preformed on either supercoiled or linear lacUV5 templates. After addition of the heparin competitor, the fraction of complexes remaining at subsequent time intervals were measured by transcription.
It has been previously shown that open complexes on supercoiled lacUV5 templates are extremely long-lived (40). As expected, the cysteine-less 70 holoenzyme control formed very stable open complexes with half-lives greater than 40 h (Fig.  8A). The 70 (2.1C-2.2C) holoenzyme exhibited a defect in open complex stability with a half-life of ϳ7 h. Likewise, the 70 (2.2C-2.3C) holoenzyme had a lesser defect with a half-life of greater than 16 h. These differences in open complex stability were not dramatic enough to alter multiple-round transcriptional activity from a supercoiled template, as seen in the results from Fig. 4.
In contrast, open complex stability on a linear template was dramatically lower than that observed with supercoiled templates (40). A wild type 70 holoenzyme open complex has been shown to have a half-life of 17 min on linear lacUV5 templates (40). This result agrees with the 16-min half-life observed with the cysteine-less 70 holoenzyme control (Fig. 8B). The 70 (2.1C-2.2C) holoenzyme did show a major defect in open complex stability on the linear template, with a half-life of only 4.5 min. Alternatively, no significant decrease in stability is observed with the 70 (2.2C-2.3C) holoenzyme, whose half-life was 14 min. These results strongly suggest that the defect of 70 (2.1C-2.2C) holoenzyme in open complex formation was caused by the instability of the open complex on linear DNA. Unfortunately, these results do not explain why the 70 (2.2C-2.3C) holoenzyme has a defect in open complex promoter occupancy on linear templates. DISCUSSION Transcription initiation in prokaryotes occurs through the multi-subunit RNA polymerase holoenzyme, consisting of core subunits ␣ 2 ␤␤Ј and the specificity subunit (1). Because core RNA polymerase and the factor are the only two proteins necessary for transcription initiation to occur, conformational changes may be required at various steps along a series of intermediary steps. These steps include: formation of holoenzyme, closed complex formation upon promoter binding, open complex formation and promoter melting, and finally formation of the processive elongation complex (3,10). Previous experiments to identify conformational change within this process have resulted in information about individual steps but have not addressed transitions between intermediates.
Luminescence resonance energy transfer experiments have suggested that upon holoenzyme formation, the DNA-binding regions of 70 (regions 2.4 and 4.2) undergo conformational changes that place them in positions compatible with the spacing of the Ϫ10 and Ϫ35 promoter elements (9,28). Additionally, the interaction between region 1.1 and region 4, which prevents free 70 from binding DNA, has been shown to be disrupted upon holoenzyme formation, suggesting that a major conformational change occurs within 70 (41,42). These predicted gross conformational changes in 70 were validated upon inspection of the crystal structure from T. aquaticus A holoenzyme (21), suggesting that 70 is a dynamic protein capable of undergoing conformational changes during holoenzyme formation.
Along with the high resolution crystal structures of T. aquaticus holoenzyme, a complex of this holoenzyme bound to "fork junction" DNA (20,21) has provided new insight into when major conformational changes may occur during transcription. Four large mobile modules, which move as relatively rigid bodies with respect to the main "core" module of holoenzyme, were identified by comparing the two holoenzyme structures to the previously known core enzyme structure (19). This comparison revealed that two of the modules required changes within the DNA-binding regions of (20). The first module, which contains the flap domain of the ␤ subunit of core enzyme bound to region 4, rotated ϳ4°, placing the Ϫ35 recognition helix of in position for binding the Ϫ35 hexamer. The second module, consisting of the RNAP clamp domain and region 2, rotated toward the RNAP channel (20). These conformational changes suggest that flexibility in the holoenzyme structure is required for proper positioning of the DNA-binding determinants of so that sequence specific contacts can be established. In addition, Murakami et al. (20) state that placement of the template strand within the DNA channel must involve opening of the "jaws" of RNA polymerase during open complex formation. Reclosure of the RNA polymerase jaws is then necessary for formation of the transcription tunnel, where the template strand is completely enclosed by protein (20). These hypotheses suggest that holoenzyme is a dynamic structure and may undergo several yet uncharacterized conformational changes as it proceeds through the process of transcription initiation.
Because the large size of the multi-subunit RNA polymerase holoenzyme prohibits NMR analysis or other conventional methods of analyzing conformational changes, we chose to utilize a disulfide prediction algorithm (22) to probe for conformational changes in specific regions of 70 that are required for the proteins to function. We chose to engineer disulfide bonds within 70 region 2 that would prohibit major conformational change involving separation of ␣-helices and then assay for function throughout the multiple steps of transcription initiation. The first disulfide bond, involving residues Cys 387 -Cys 408 , effectively locks region 2.1 to region 2.2 by virtue of the disulfide bond on one end and a short loop connecting the ␣-helices on the other. The presence of the disulfide bond prevents the ␣-helices from separating or rotating with respect to one another by restricting the distance and angle the two ␣-helices can be from one another. The second disulfide bond within 70 , from residues Cys 415 -Cys 434 , locks region 2.2 to region 2.3. Unlike the first disulfide bond locking region 2.1-2.2, this disulfide bond restricts the distance between the ␣-helices but allows for some rotation around the disulfide bond.
Several lines of evidence led us to speculate that the ␣-helical region 2 of 70 might undergo conformational change upon holoenzyme formation or steps thereafter. First, mutational analysis of the coiled-coil of the ␤Ј subunit of RNA polymerase produced several mutants that had increased or decreased affinity for 70 (43). However, these mutations were positioned on the underside of the coiled-coil, away from the residues thought to bind 70 (44). These mutations could be explained if the helices in 70 region 2 separated from each other to form new interactions with the coiled-coil during holoenzyme formation. Second, luminescence resonance energy transfer analysis of 70 indicates that regions 2.4 and 4.2 underwent a significant conformational change upon holoenzyme formation, where these regions of 70 became more solvent exposed upon binding of core enzyme (45). Finally, defects in open complex formation caused by mutations in region 2.2 of 70 suggest that reorganization of the ␣-helices within 70 region 2 may facilitate DNA binding (46).
We sought to examine the effect of restricting helical movement in 70 region 2 during the various steps of transcription initiation. In multiple-round in vitro transcription assays, which test the overall transcriptional activity of the holoenzyme, both the 70 (2.1C-2.2C) and 70 (2.2C-2.3C) holoenzyme displayed wild type levels of activity. This result indicates that the presence of the disulfide bonds do not disrupt a ratelimiting step required for transcriptional activity (Fig. 4). Likewise, the mobility shift assay for closed complex formation on linear template showed no defect from either disulfide-locked 70 holoenzyme (Fig. 5). This result suggests that a major conformational change within the ␣-helices of 70 region 2 is not required for closed complex formation. However, in two assays for open complex formation, DNase I footprinting, and filter binding (Figs. 6 and 7), both disulfide-locked 70 holoenzymes exhibited clear defects. The defects in occupancy of the lacUV5 promoter were severe, suggesting that a conformational change in region 2 might be necessary to facilitate melting of promoter DNA and formation of a stable open complex. Furthermore, heparin stability assays demonstrated that of those open complexes that were able to form, the open complex lifetime for the 70 (2.1C-2.2C) holoenzyme was greatly shortened (Fig. 8).
From the experimental results, we propose that conformational flexibility within 70 region 2 may facilitate proper placement of the ␣-helices during transcription initiation. We suggest that the 2.1C-2.2C disulfide in 70 causes defects in holoenzyme open complex formation and stability because of conformational restriction on 70 region 2. One potential con-formational change that could be affected is movement of region 2.2 away from region 2.1, which might position region 2.3-2.4 for contact with the Ϫ10 hexamer and for promoter melting. Additionally, the engineered disulfide bond in 70 could prevent rotation of the region 2.1 or 2.2 helix, which might be necessary to facilitate transcription initiation. It is important to note that the defects in transcriptional activity for the 70 (2.1C-2.2C) holoenzyme were only apparent on linear DNA template and not the supercoiled template in the multiple-round transcription assay. From this, we hypothesize that supercoiling of the DNA and possibly the addition of NTPs is sufficient to relieve the defect caused by preventing conformational change. Supercoiling of the promoter template, and not NTP addition, is most likely, because both disulfide-locked 70 holoenzymes were defective in single-round transcription experiments from linear lacUV5 templates (data not shown). The thermodynamics of transcription from a supercoiled template drives the equilibrium toward a more stable open complex and results in more productive transcription (47).
The 70 (2.2C-2.3C) holoenzyme also exhibited a defect in open complex formation; however, it had a heparin stability profile similar to that observed with the cysteine-less 70 holoenzyme. We suggest that region 2.2-2.3 of 70 does not require a dramatic conformational change such as separation of the ␣-helices but could involve a more subtle shift in the orientation of the ␣-helices to facilitate DNA binding. According to the filter binding and heparin stability assays, the 70 (2.2C-2.3C) holoenzyme had slightly less of a defect than 70 (2.1C-2.2C) holoenzyme for open complex formation. We speculate based upon protein modeling that the disulfide bond locking region 2.2 to region 2.3 is not as rigid because of the large loop connecting regions 2.2 and 2.3 that may allow for some flexibility around the disulfide bond. The flexibility may confer some level of open complex stability, albeit at a lower promoter occupancy than wild type.
It is interesting to speculate in a more in-depth manner about the nature of the defects seen with the disulfide-locked holoenzymes on the lacUV5 linear templates. For example, both holoenzymes exhibited defects in open complex formation but have wild type activity for closed complex formation. One difference in these assays is the temperatures at which each complex forms: 0°C for closed complex formation and 37°C for open complex formation. Previous DNase I footprinting experiments with the lacUV5 promoter performed at either 0 or 37°C suggested that protection of the promoter region occurs in a stepwise process, where protection of the promoter region at 37°C extends from Ϫ53 to ϩ20 but only extends downstream to Ϫ2 at 0°C (14,39). Therefore, on the lacUV5 promoter at 0°C, the DNA may not yet be within the jaws of holoenzyme where it would be protected from DNase I cleavage. This suggests that on the lacUV5 promoter, promoter contacts involved in initial closed complex formation are with the Ϫ35 region of DNA. In our experiments, we were able to obtain stable closed complexes with both disulfide-locked 70 holoenzymes at 0°C but were unable to obtain stable open complexes with high promoter occupancy at 37°C. One possible explanation is that the conformational restrictions imposed by the disulfide bonds allowed for Ϫ35 recognition to form the closed complex but that proper Ϫ10 interactions were defective because a necessary conformational change within region 2 was blocked. A model of a "closed complex-like" structure was created from the Taq holoenzyme-fork junction DNA structure by simply filling in duplex DNA and extending it in B-form conformation (20). The modeling is consistent with our results, suggesting that conformational flexibility in 70 region 2 may facilitate interactions with the Ϫ10 nontemplate strand during the transition between closed and open complexes. This work has looked at a small region of 70 to determine its role in transcription initiation, but it would be interesting to further investigate: 1) what positioning of region 2.3-2.4 is required to begin strand separation and open complex formation, 2) if the disulfide-locked 70 holoenzymes are defective for positioning the DNA within the jaws of RNA polymerase, or 3) if a conformational change in region 2 helps to open the jaws during open complex formation. The continued availability of high resolution structural information and continued use of alternative methods such as disulfide bond engineering will help us to understand the true role of conformational changes within holoenzyme during transcription initiation.