Architecture of a Complex between the ς70 Subunit of Escherichia coli RNA Polymerase and the Nontemplate Strand Oligonucleotide

We used luminescence energy transfer measurements to determine the localization of 5′- and 3′-ends of a 12-nucleotide nontemplate strand oligonucleotide bound to ς70holoenzyme. Five single reactive cysteine mutants of ς70(cysteine residues at positions 1, 59, 366, 442, and 596) were labeled with a europium chelate fluorochrome (donor). The oligonucleotide was modified at the 5′- or at the 3′-end with Cy5 fluorochrome (acceptor). The energy transfer was observed upon complex formation between the donor-labeled ς70 holoenzyme and the acceptor-labeled nontemplate strand oligonucleotide, whereas no interaction was observed with the template strand oligonucleotide. The oligonucleotide was bound in one preferred orientation. This observation together with the sequence specificity of single-stranded oligonucleotide interaction suggests that two mechanisms of discrimination between the template and nontemplate strand are used by ς70: sequence specificity and strand polarity specificity. The bound oligonucleotide was found to be close to residue 442, confirming that the single-stranded DNA binding site of ς70 is located in an α-helix containing residue 442. The 5′-end of the oligonucleotide was oriented toward the COOH terminus of the helix.

Transcription initiation in Escherichia. coli involves two essential steps: (i) initial promoter recognition by RNA polymerase (RNAP 1 ) and (ii) melting of DNA duplex in the vicinity of transcription start site (1)(2)(3)(4)(5). The simplest two-step model describing transcription initiation (Eq. 1) involves a rapid formation of a labile "closed" complex that in a second step isomerizes to a stable "open" complex. In the open complex 10 -15 base pairs of DNA duplex become single-stranded. This DNA duplex melting in the case of E. coli RNAP occurs spontaneously, therefore the energetic cost of duplex melting must be offset by some favorable RNAP-pro-moter interactions.
The E. coli RNA polymerase holoenzyme is a multisubunit enzyme (subunit composition ␣ 2 ␤␤Ј) (1). 70 (the primary subunit) is the RNAP subunit thought to be responsible for the initial recognition of promoter DNA (1)(2)(3)(4)(5)(6). Sequence homology between the conserved region 2.3 of 70 and eukaryotic singlestranded DNA (ssDNA)-binding proteins was used as a basis of the proposal that 70 subunit could also be actively involved in the promoter melting reaction through binding of ssDNA of the open complex (7). This favorable ssDNA-70 interaction could reduce the energetic cost of DNA melting and facilitate open complex formation. The data from several laboratories provided experimental evidence in support of this proposal. The 70 subunit was shown to bind ssDNA (8 -16) or singlestranded bubbles within the DNA duplex (17,18). The binding was specific for the nontemplate DNA strand (9 -16). The ssDNA binding site of 70 is most likely located in the conserved region 2.3 of the 70 subunit (13, 14, 16, 19 -21). The ssDNA binding activity of 70 is regulated allosterically through binding to the core RNAP (16). Free 70 binds ssDNA weakly and does not discriminate between the template and nontemplate sequences. Binding of 70 to the core reduces the affinity of 70 to template ssDNA and increases the affinity to nontemplate ssDNA resulting in a ϳ200-fold difference in the affinity between nontemplate and template ssDNA (16). All of the above properties of 70 are consistent with its active role in the promoter melting reaction.
Region 2.3 of 70 is thought to be a location of the ssDNA binding site of 70 . However, the architecture of the ssDNA-70 complex and the structural determinants of the high selectivity for nontemplate ssDNA binding are not known. Therefore, in this work we used luminescence energy transfer (LRET) (22-25, 31, 32) distance measurements to determine localization of nontemplate strand oligonucleotide bound to 70 with respect to several functional domains of the protein.

EXPERIMENTAL PROCEDURES
Materials-Cy5, monosuccinimidyl ester, was purchased from Amersham Pharmacia Biotech. Succinimidyl ester of 7-amino-4-methylcoumarin-3-acetic acid (AMCA-NHS) was from Boehringer Mannheim. Oligonucleotides were obtained from Midland Certified Reagent Co. (Midland, TX). All other chemicals were of the highest purity commercially available. Core RNAP was purified from E. coli K12 cells (obtained from the University of Alabama fermentation facility) using the method of Burgess and Jendrisak (29).
Preparation of DTPA-AMCA(5)-Maleimide-5 mg of AMCA-NHS was dissolved in 250 l of DMF, and 50 l of 1 M ethylenediamine HCl (pH 7.0) was added. The mixture was incubated for 1 h at room temperature, and another 50 l of 1 M ethylenediamine HCl was added followed by incubation for 1 h at room temperature. The mixture was diluted to ϳ4 ml with buffer A (25 mM triethylammonium acetate buffer (pH 7.0) containing 2% acetonitrile) ϩ 300 l of buffer B (buffer B is buffer A with 95% acetonitrile). A small amount of a white precipitate was removed by a centrifugation, and the sample was loaded on a fast protein liquid chromatography reverse phase column (HR10/10 column (Amersham Pharmacia Biotech)) packed with Resource 15RPC (Amersham Pharmacia Biotech).The column was eluted at 3 ml/min with 100 ml of 0 -50% buffer B gradient. Fractions containing the adduct of 7-amino-4-methyl coumarin-3-acetic acid and ethylenediamine eluting at about 14% B were pooled (7.5 ml) and lyophilized. Dried fractions were dissolved in 500 l of DMF, and 5 mg of succinimidyl ester of maleimidylpropionic acid dissolved in 100 l of DMF was added. The mixture was incubated for 1 h at room temperature, diluted to ϳ 5 ml with 5% buffer B, and loaded onto a Resource 15RPC column. The column was eluted at 3 ml/min with 100 ml of 0 -50% buffer B gradient. Fractions containing the AMCA-maleimide eluting at about 24% buffer B were pooled (7.5 ml) and lyophilized. Dried fractions were dissolved in 500 l of DMF, and 20 mg of DTPA anhydride was added. The mixture was incubated for 1 h at room temperature, diluted to 5 ml with buffer A, and run on a Resource 15RPC column as described above. Fractions containing DTPA-AMCA-maleimide eluting at ϳ15% buffer B were pooled, dispensed to Eppendorf tubes such that each tube contained 0.1 mol of the chelate, and dried. The yield of a purified product was 10 -20%.
Single-cysteine Mutants of 70 -The preparation of single-cysteine mutants of 70 is described elsewhere (27). Briefly, three endogenous cysteine residues of 70 (cysteines 132, 291, and 295) were replaced with Ser residues using site-directed mutagenesis. Single Cys residues were then introduced into the desired locations using single amino acid replacements, with the exception of [1Cys] 70 , in which the cysteine residue was inserted between the initiating Met and the second residue of the protein.  (27).
Fluorochrome-labeled Oligonucleotides-In all experiments a 12-nt oligonucleotide (TCGTATAATGTG) corresponding to positions Ϫ15 to Ϫ4 of the lacUV5 promoter nontemplate strand was used. The Cy5 fluorochrome was attached to the 5Ј end by first adding a 5Ј-aliphatic amino group through a postsynthetic modification of the oligonucleotide with ethylenediamine (28), which results in a two-carbon linker between the 5Ј-phosphate and the reactive amine. The 5Ј-amino-containing oligonucleotide was then reacted with ϳ 1 mM succinimidyl ester of Cy5 for 2-4 h at room temperature in 0.1 M sodium bicarbonate buffer (pH 8.3). The excess of Cy5 was removed on a G-25 spin column (Amersham Pharmacia Biotech), and the labeled oligonucleotide was purified from unlabeled DNA using a reverse phase high performance liquid chromatography column as described previously (26). To attach Cy5 fluorochrome to the 3Ј-end of the oligonucleotide a 3Ј-amino group was introduced during oligonucleotide synthesis using a three-atom linker. The reaction of the 3Ј-amino-containing oligonucleotide with Cy5 and the purification of labeled DNA were performed as described for 5Ј-amino-modified oligonucleotide. The concentration of the oligonucleotides was determined spectrophotometrically using absorbance at 260 nm corrected for the contribution due to Cy5 dye (26).
Donor Fluorochrome-labeled 70 -Samples of each single-cysteine mutant of 70 (0.5-1.0 mg) were precipitated with 60% ammonium sulfate by the addition of the appropriate volume of saturated ammonium sulfate solution. The protein pellet was collected by centrifugation and dissolved in 75 l of 50 mM Tris (pH 8.0), 1 mM EDTA, 5% glycerol, and 6 M GdHCl buffer. Dithiothreitol was added to a final concentration of 0.5 mM, and mixtures were incubated for 1 h at room temperature. Dithiothreitol was removed by a Microspin G-50 column (Amersham Pharmacia Biotech) equilibrated with the above buffer. DTPA-AMCAmaleimide was added to a final concentration of 0.5-1.0 mM, and the reaction was allowed to proceed for 1 h at room temperature. The excess of unreacted DTPA-AMCA-maleimide was removed by Microspin G-50 column equilibrated with 50 mM Tris (pH 8.0), 1 mM EDTA, 5% glycerol, and 6 M GdHCl buffer. The eluate from the G-50 column was diluted to ϳ0.75 ml with the above buffer, dialyzed first against the same buffer for few hours and next to 50 mM Tris (pH 8.0), 5% glycerol buffer overnight with three changes of 100 ml of the buffer. Refolded modified 70 , after a 30-min incubation with 10 M EDTA and 10 M EuCl 3 , was mixed with purified core RNAP in a 1:1 molar ratio and incubated for 30 min at 4°C. The reconstituted holoenzyme was purified from unbound labeled 70 on a fast protein liquid chromatography Superdex-200 size exclusion column (Amersham Pharmacia Biotech). For experiments with free donor-labeled 70 , the protein after refolding through dialysis was purified further on Superdex-200 column.
LRET Measurements-All LRET experiments were performed in a 120-l cuvette in 50 mM Tris-HCl (pH 8.0), 250 mM NaCl, 5% glycerol buffer at 25°C. The concentration of the holoenzyme used was 25-50 nM. Luminescence decays of donor fluorochrome-labeled holoenzyme were recorded in the presence and absence of acceptor-labeled oligonucleotides (0 -250 nM). Luminescence lifetime measurements were performed on a laboratory-built two-channel spectrofluorometer with a pulsed nitrogen laser (LN300, Laser Photonics, Orlando, FL) as an excitation source (26). Donor emission was observed at 617 nm and acceptor emission at 670 nm. Decays of donors in the absence of acceptor were fitted to a single-exponential equation, where ␣ is the amplitude and is the luminescence lifetime, respectively. Decays of donors in the presence of acceptor and decays of sensitized acceptor emission were fitted to a three-exponential equation (see "Results"). Donor and sensitized acceptor decay curves were fitted simultaneously using global nonlinear regression with Scientist (Micromath Scientific Software, Salt Lake City, UT) to the following set of equations: where I d (t) and I a (t) are luminescence intensity of donor and sensitized acceptor, respectively; . 2 The ability to analyze donor-and sensitized-acceptor decay data by global fitting to Equations 3 and 4 is an important advantage of using europium chelate in LRET experiments. It improves very significantly the precision and the confidence of lifetime determination in the presence of energy transfer. The energy transfer (E) between europium chelate-labeled 70 and Cy5-labeled oligonucleotide was calculated from measurements of luminescence lifetime of a donor in the absence ( d ) and in the presence of acceptor ( da ).
The distances between the donor and the acceptor were calculated according to Förster theory (22) where R is a distance between a donor and an acceptor, and R 0 is a distance at which the energy transfer is 0.5. The R 0 (55 Å) was calculated as described previously using assumptions described by Selvin and Hearst (31). One of these assumptions is that the orientation factor ( 2 ) value of 2/3 (completely random orientation of donors and acceptors (22)(23)(24)(25)) could be used in calculating R 0 . This assumption is justified by two factors: long lifetime of the donor and multiple transition dipole moments of the lanthanide (26,(31)(32)(33). The long donor lifetime increases the probability that a donor and an acceptor will rotate to all possible orientations during the donor excited state lifetime. Multiple transition dipole moments result in depolarization of donor emission even if the donor is completely immobile.

RESULTS
Acceptor-labeled 12-nt Oligonucleotide and Donor-labeled 70 - Fig. 1, A and B, shows the structure of the two acceptor fluorochrome-labeled oligonucleotides used in this work. The 12-nt oligonucleotide sequence used corresponds to position Ϫ15 to Ϫ4 of the lacUV5 nontemplate strand sequence. We have shown previously that 70 in a RNAP holoenzyme is capable of binding this oligonucleotide ϳ200-fold better than the template or random sequence 12-nt oligonucleotide (16). The acceptor fluorochrome (Cy5) was attached to the 5Ј-end (Cy5-5Ј-NT) or to the 3Ј-end (NT-3Ј-Cy5) of the oligonucleotide, respectively. Fig. 1C shows an absorbance spectrum of the purified Cy5-5Ј-NT. The characteristic absorbance peaks caused by DNA (at 260 nm) and Cy5 (at 647 nm) are apparent. Using this spectrum we estimated that labeled oligonucleotide contained ϳ1 mol of Cy5 dye/mol of the oligonucleotide. Essentially identical spectrum and degree of modification were observed for NT-3Ј-Cy5 (not shown).  (26). We used a europium chelate for LRET measurements because, as shown recently, these probes offer several important advantages when used as donors in LRET measurements compared with classical organic dye fluorescence probes (26,(31)(32)(33). The fluorescence donors were incorporated into the specific sites of 70 protein through chemical modification of unique cysteine residues placed in different structural domains of the protein (Fig. 2B). Cys-1 and Cys-59 are in conserved region 1, which was shown to be involved in the autoinhibition of promoter DNA binding in the free 70 (34,35). Cys-366 is located in a nonconserved region of the protein near sequences thought to be important for core RNAP binding (36,37). Cys-442 is in region 2.4, responsible for Ϫ10 promoter DNA sequence recognition, and is adjacent to region 2.3, thought to be involved in nontemplate ssDNA binding (37)(38)(39). Cys-596 is in region 4.2, which was shown to be involved in Ϫ35 promoter DNA sequence recognition (40). Fig. 2C shows an example of the absorption spectrum of the purified donor-labeled 70 ([S366C] 70 ). Characteristic peaks caused by protein (at ϳ280 nm) and DTPA-AMCA (at ϳ328 nm (26)) were observed. Using this spectrum we estimated that the labeled protein contained  Tables II and III. In the absence of acceptor-labeled oligonucleotide, luminescence decays of donor-labeled 70 reconstituted with the core RNAP were monoexponential (Fig. 3A). In the presence of acceptor-labeled oligonucleotide, decays of donor-labeled 70 were no longer single exponential, and the presence of faster decaying component(s) was obvious (Fig. 3A). The decay of the donor-labeled holoenzyme in the presence of acceptor-labeled oligonucleotide could be fitted to a three-exponential decay function. At any given concentration of acceptor-labeled oligonucleotide, we expected in solution an equilibrium mixture of RNAP-oligonucleotide complex (capable of LRET) and free RNAP (incapable of LRET). In accordance with this expectation, the slowest decay time ( 3 ) observed in the presence of acceptor-labeled oligonucleotide was very similar to the decay time observed in the absence of acceptor, and its amplitude decreased with the increase of oligonucleotide concentration ( Table I) absence of any DNA (not shown). Second, the appearance of a fast decaying component(s) in the presence of acceptor-labeled oligonucleotide was correlated with the appearance of a large sensitized emission signal of the acceptor (Fig. 3B). Third, the amplitude of the fast component(s) increased with the increase of acceptor-labeled oligonucleotide concentration, whereas the lifetimes of the fast and slow components remained constant (Table I).
Two lifetimes ( 1 and 2 ) were necessary to describe the fast decaying portion of the decay curve adequately. There are several possible interpretations for the two fast decaying components observed in the presence of acceptor. They could be a result of two populations of species capable of LRET; for example, the oligonucleotide could bind to the primary binding site and, to a lesser degree, to a secondary site. The appearance of two fast decaying components in LRET with lanthanide chelates was observed previously, and the very fast component was interpreted to be an instrumental artifact (30 -32). In cases where the donor-acceptor distance was large, the two fast lifetimes could be resolved. For example, in the case of donorlabeled [R596C] 70 and NT-3Ј-Cy5 DNA these lifetimes were: 1 ϭ 66 s (9% of total fast component amplitude) and 2 ϭ 412 s (91% of total fast component amplitude). However, when the distance between the donor and acceptor was smaller, the values of these two lifetimes became correlated and could not be well resolved. Thus, for all LRET calculations presented here we used the weighted average of two fast decaying components, avoiding the arbitrary decision of which lifetime component to use. The impact of the mode of LRET calculations on the results was small because the distances obtained using the average of 1 and 2 were very similar to distances calculated using only 2 (the average difference in distances between these two modes of calculation was only 4 Ϯ 2 Å). LRET between the donor-labeled 70 and acceptor-labeled 12-nt nontemplate strand oligonucleotide was specific only for 70 reconstituted with the core RNAP. Free 70 showed no evidence of significant energy transfer in the presence of 100 nM oligonucleotide (Fig. 3, C and D). Donor decays in the presence or absence of acceptor-labeled DNA were monoexponential (Fig.  3C), and essentially no sensitized emission of the acceptor was observed (Fig. 3D). At the same concentration of labeled oligonucleotide a very efficient LRET was observed in the case of holoenzyme (Fig. 3, A and B). These observations are consistent with and further confirm our previous report that the specificity for binding the nontemplate single-stranded oligonucleotide is induced in 70 by an interaction with the core RNAP (16).
Competition experiments were used to determine the specificity of LRET in the nontemplate strand oligonucleotide-ho-loenzyme complex (Fig. 4). In the presence of 50 nM acceptorlabeled nontemplate strand oligonucleotide an efficient LRET was observed as indicated by a much faster decay of the donor compared with the decay in the absence of DNA (Fig. 4A). This efficient LRET could be eliminated by the addition of excess unlabeled 12-nt nontemplate strand oligonucleotide as indicated by almost overlapping decay curves observed in the absence of any DNA and in the presence of 50 nM acceptor-labeled and 1 M unlabeled nontemplate oligonucleotide (Fig. 4B). In contrast, excess unlabeled nontemplate randomized sequence oligonucleotide had essentially no effect, and the efficient LRET was still observed, as indicated by a much faster decay observed in the presence of 50 nM acceptor-labeled nontemplate oligonucleotide and 1 M unlabeled nontemplate randomized sequence oligonucleotide (Fig. 4C). The nontemplate randomized sequence oligonucleotide (TTGATATCGTAG) had the same base composition but a sequence different from that of the nontemplate strand oligonucleotide. Based on the results presented in Figs. 3 and 4 and Table I we concluded that LRET signals observed for donor-labeled 70 and acceptor-labeled nontemplate strand oligonucleotides are the property of a specific complex between the holoenzyme and nontemplate ssDNA. The distances calculated from these LRET measurements provide information regarding the architecture of this complex.
Architecture of 70 -Nontemplate ssDNA Complex-Results of LRET measurements with all five single-cysteine mutants of 70 and nontemplate 12-nt oligonucleotides with acceptors at the 3Ј-or 5Ј-end are summarized in Tables II and III. A wide range of energy transfer efficiencies (from 0.37 to Ͼ0.99) was observed. The range of distances corresponding to these energy transfer efficiencies was from 60 Å to Ͻ 25 Å. For several b Lifetimes of donors in the presence of acceptors ( da ) are weighted averages of the lifetimes of two fast components ( 1 and 2 ) in the three-exponential decay observed in the presence of acceptor-labeled oligonucleotide. b Lifetimes of donors in the presence of acceptors ( da ) are weighted averages of the lifetimes of two fast components ( 1 and 2 ) in the three-exponential decay observed in the presence of acceptor-labeled oligonucleotide.  0  100  596  50  46  609  54  72  100  36  610  64  78  250  26  562  74  63 a Amplitude (fast) is a sum of amplitudes of two fast components in the three-exponential decay observed in the presence of acceptor-labeled oligonucleotide (see "Results").
b fast is the weighted average of the lifetimes of two fast components ( 1 and 2 ) in the three-exponential decay observed in the presence of acceptor-labeled oligonucleotide.
residues of 70 very significant differences between the distance to 5Ј and 3Ј of the oligonucleotide were found. Region 2.4 (Cys-442) was found to be the closest to the oligonucleotide bound to 70 . It appears also that the 5Ј-end of the bound oligonucleotide was much closer (Ͻ25 Å) to residue 442 than the 3Ј-end (35 Å). Residue 596 was the farthest from the bound oligonucleotide, and this residue seems to be located almost at the same distance from the 5Ј-and 3Ј-ends of the oligonucleotide. Also, the NH 2 -terminal cysteine was found at an approximately equal distance from the 5Ј-and 3Ј-ends of the oligonucleotide. Residue 59 was found closer to the 5Ј-end of the oligonucleotide, and residue 366 was found closer to the 3Ј-end.
Using distances determined by LRET, three-dimensional models of relative localization of different domains of 70 could be built. An example of such a model superimposed on the crystal structure of 70 fragment is shown in Fig. 5. Building these models allowed an indirect determination of several additional distances between the sites in the complex (Table IV). Because there were not enough distance constraints to determine all possible distances uniquely, an analysis of a relative precision of these indirect distance determinations was performed. A set of 25 independent models fulfilling distance constraints from LRET experiments was built, starting each from randomly "scrambled" initial distances between the sites in the complex. In each of these models all possible distances were then measured, and the mean and standard deviation were calculated (Table IV). The standard deviation can thus be used as a convenient measure of a relative precision of these indirect distance estimations. Inspection of the data presented in Table  IV shows that two sets of distances can be identified easily: those very poorly defined (standard deviation Ͼ 20 Å) and those whose precision is good enough (standard deviation Ͻ 10 Å) for use in discussing the architecture of the complex. Distances with relatively high precision of estimation were 5Ј of the oligonucleotide to 3Ј of the oligonucleotide, residues 1-442, 9 -442, 366 -442, and 442-596. We have recently measured several interdomain distances in the holoenzyme, 3 and these distances appear to be in general agreement with the distances calculated by model building (Table IV). Additionally, using these directly determined interdomain distances we attempted to obtain a better estimate of 5Ј N 442 distance by building models that included this distance as a variable. This distance could not be determined by LRET measurements because the donor and acceptor were too close (Table II). A distance of ϳ14 Å was obtained from model building, consistent with LRET results (Table II). DISCUSSION We have determined the distances between several sites in the 70 and the 5Ј-or the 3Ј-end of 12-nt nontemplate strand oligonucleotide in complex with RNAP holoenzyme. The measured distances allowed us to build a model describing a threedimensional architecture of the oligonucleotide-RNAP com-3 S. Callaci, E. Heyduk, and T. Heyduk, unpublished data.  (Tables II-IV) and is shown only for visual illustration of the data presented in these tables. The size of the spheres is proportional to a distance to the viewer. The model was built using distance constraints routine of ChemSite (Pyramid Learning, Stanford, CA), and the figure was produced using Ribbons (45). plex. Several conclusions regarding the architecture of the complex can be made.
The oligonucleotide binds to RNAP apparently in a preferred orientation. In principle, the oligonucleotide could bind to its binding site either in 5Ј 3 3Ј or 3Ј 3 5Ј orientation. If the binding could occur equally well in either orientation, the apparent distances measured between sites in 70 and the 5Ј-or the 3Ј-end of the oligonucleotide should be the same. We observed, however, that for several sites in 70 the distances between the 5Ј-end and the 3Ј-end were very significantly different (Tables II and III), showing that the oligonucleotide was bound in one preferred orientation. Such preference for a specific orientation may be an additional mechanism by which RNAP in the open complex could discriminate between template and nontemplate strands in the transcription bubble. Previous binding experiments with oligonucleotides corresponding to the Ϫ10 region of the nontemplate and the template strand showed that RNAP holoenzyme could bind nontemplate sequence oligonucleotides ϳ200-fold better then the template sequence oligonucleotides (16). Oligonucleotides can freely assume any orientation when they bind to the ssDNA binding site of 70 . The situation will be different in the open complex when ssDNA strands in the Ϫ10 region have restricted mobility. Thus, if the nontemplate strand in the open complex is in a correct orientation for binding to the ssDNA binding site of 70 , the template strand will be forced to be in the opposite, unfavorable for the binding orientation. Thus, the two different mechanisms for discrimination between single-stranded nontemplate and single-stranded template strands in the Ϫ10 region of promoter DNA are apparently being used by 70 . One mechanism is the sequence specificity of the binding, the other mechanism is the strand polarity specificity. This dual mode of discrimination employed by 70 seems to be well suited for the tasks that the ssDNA binding site of 70 needs to perform: selective binding of the Ϫ10 sequences and selective binding of the nontemplate strand.
Region 2.4 (Cys-442) was found to be the closest to the oligonucleotide bound to 70 . This is consistent and confirms the proposals that the ssDNA binding site of 70 is localized in region 2.3 (13, 14, 16, 19 -21) because this region is adjacent to region 2.4 and is located in the same ␣-helix (helix 14 (37)). It appears also that the 5Ј-end of the bound oligonucleotide was much closer (Ͻ25 Å) to residue 442 than the 3Ј-end was (39 Å). Thus, the relative orientation of the bound oligonucleotide with respect to helix 14 appears to be 5Ј 3 COOH terminus of the helix. Such an orientation is consistent with the proposed model of nontemplate ssDNA-70 interaction based on the crystal structure of the 70 fragment (37). It is also consistent with data relating mutations in 70 and mutations in the Ϫ10 region of promoter DNA. Based on these studies it was proposed that residues 437 and 440 are involved in recognition of position Ϫ12, whereas residue 441 is involved in recognition of position Ϫ13 (38,39,41,42). Such an alignment of bases in the Ϫ10 region and amino acids in helix 14 is consistent with a 5Ј 3 COOH terminus alignment of helix 14 and the nontemplate oligonucleotide in a holoenzyme-oligonucleotide complex. However, an opposite orientation of the nontemplate strand with respect to helix 14 was also proposed (44). The reasons for this discrepancy are not clear.
Cys-596 is in region 4.2 of 70 , which was suggested to bind the Ϫ35 region of promoter DNA (40). Assuming a simple linear arrangement of Ϫ10 and Ϫ35 DNA sequences and protein domains involved in binding of these sequences, it could be expected that the 5Ј-end of the bound oligonucleotide should be much closer to residue 596 than the 3Ј-end is. Data in Tables II and III show that this was not observed. In contrast, residue 596 seems to be located almost at the same distance from the 5Ј-and 3Ј-ends of the oligonucleotide. This suggests that the orientation of the bound oligonucleotide with respect to residue 596 is as illustrated in Fig. 5, i.e. it is more or less perpendicular, not parallel, to the line joining regions 2.4 and 4.2. This observation suggests that promoter DNA in the open complex is not straight, and thus formation of the open complex involves significant deformation of DNA.
Based on LRET distance measurements it was possible to estimate indirectly with reasonable precision several other distances in the oligonucleotide-RNAP complex. The predicted distance between the 5Ј-and 3Ј-ends of the oligonucleotide was 32 Ϯ 7 Å, a distance somewhat shorter than expected for a linear 12-nt DNA, consistent with the deformation of DNA in the open complex. The predicted distance between residues 366 and 442 was found to be 38 Ϯ 6 Å. This distance is comparable, within the error of estimation, to the distance between these two residues (35 Å) observed in the crystal structure of 70 fragment (37). The agreement of this predicted distance with the crystal structure of the 70 fragment provides an additional validation of distances measured by LRET. The predicted distance between residues 442 and 596 was 58 Ϯ 6 Å. Residues 442 and 596 are located in 70 domains involved in recognition of Ϫ10 and Ϫ35 DNA sequences, which are separated by ϳ17 base pairs. Thus, the predicted distance of 58 Ϯ 6 Å is compatible with the distance expected from the ϳ17-base pair separation between binding sites for these two structural domains of the protein.