RNA polymerase gate loop guides the nontemplate DNA strand in transcription complexes

Significance

The nontemplate DNA strand in the transcription bubble interacts with RNA polymerase and accessory factors to control initiation, elongation, transcription-coupled repair, and translation. During initiation, σ subunit interactions with the nontemplate DNA regulate promoter complex formation and lifetime, abortive synthesis, and start site selection. Here, we show that the β subunit gate loop contacts with an adjacent segment of the nontemplate discriminator region play a similar role during initiation. The deletion of the gate loop alters the structure and properties of promoter complexes and has pleiotropic effects on RNA chain elongation and termination. We propose that, acting in concert with accessory factors, the gate loop mediates the clamp closure and guides the nontemplate strand in initiation and elongation complexes.

Abstract

Upon RNA polymerase (RNAP) binding to a promoter, the σ factor initiates DNA strand separation and captures the melted nontemplate DNA, whereas the core enzyme establishes interactions with the duplex DNA in front of the active site that stabilize initiation complexes and persist throughout elongation. Among many core RNAP elements that participate in these interactions, the β′ clamp domain plays the most prominent role. In this work, we investigate the role of the β gate loop, a conserved and essential structural element that lies across the DNA channel from the clamp, in transcription regulation. The gate loop was proposed to control DNA loading during initiation and to interact with NusG-like proteins to lock RNAP in a closed, processive state during elongation. We show that the removal of the gate loop has large effects on promoter complexes, trapping an unstable intermediate in which the RNAP contacts with the nontemplate strand discriminator region and the downstream duplex DNA are not yet fully established. We find that although RNAP lacking the gate loop displays moderate defects in pausing, transcript cleavage, and termination, it is fully responsive to the transcription elongation factor NusG. Together with the structural data, our results support a model in which the gate loop, acting in concert with initiation or elongation factors, guides the nontemplate DNA in transcription complexes, thereby modulating their regulatory properties.

During each round of transcription, RNA polymerase (RNAP) establishes, maintains, and finally releases contacts with the DNA template and the RNA. These interactions mediate highly selective initiation, processive elongation, and precise termination and can be tuned to enable intricate regulation during the transcription cycle. RNAP resembles a crab claw in which the two pincers composed of the β′ and β subunits form an active site cleft that accommodates the nucleic acid chains (Fig. 1). The mobile β′ clamp domain that forms one of the pincers stands out as a central regulatory feature (1, 2). The open clamp likely allows the loading of the promoter DNA during initiation and the release of the template and the nascent RNA during termination. The clamp is thought to close to form an active initiation complex and to remain closed during elongation but may open partially at a hairpin-dependent pause site (3).

Fig. 1.

Bacterial transcription complexes. (A and B) An overview of the open promoter complex (RPo) with a 6-nt discriminator (A) and the TEC with bound NusG (B). The composite models (Datasets S1 and S2) were generated using T. thermophilus RPo (22) and TEC (38) and elements from other structures as described in SI Materials and Methods. Proteins are depicted by simplified differentially colored molecular surfaces; β, σ^A and NusG are rendered semitransparent. The positions of the N-terminal domain (NTD) of the σ1.1 region, α C-terminal domains (CTDs), and NusG CTDs were chosen arbitrarily within the volume permitted by the length of the flexible linkers; the cyan arrow in A indicates that the NTD of σ1.1 is predominantly located near the β lobe domain in E. coli RPo (39). Nucleic acids and βGL are shown as cartoons, two Mg²⁺ ions in the active site are shown as cyan spheres, and an incoming NTP is shown as red sticks. Selected DNA nucleotides are numbered relative to the TSS in A and from the RNA 3′ end in the posttranslocated TEC in B. (C) A zoomed-in view of the GL and the discriminator region of the NT DNA (the rectangular area outlined in A). The β lobe and σA are semitransparent. E. coli σ⁷⁰ Met102 was modeled into a homologous position of T. thermophilus σA and is shown in a balls-and-sticks configuration. (D) A side view of the TEC. Nucleic acids are depicted as surfaces. The view is clipped along the dashed line in B to expose NusG interactions with the GL and the β′ clamp domain. (Inset) NusG–RNAP contacts.

The clamp movements may be linked to those of the β lobe domain, which forms a part of the second pincer. The β gate loop (GL), which lies across the RNAP cleft from the tip of the clamp (Fig. 1), has been identified as an element that restricts the entry of the duplex promoter DNA into the narrow active site cleft (4), allowing only a single strand of DNA to pass through. This model posited that the DNA stands must separate outside the cleft before entry into RNAP, whereas footprinting studies demonstrated that the promoter DNA enters the active site cleft before it is opened (5). This controversy was a subject of an intense debate until a recent study revealed that the width of the RNAP cleft varies in solution and potentially is able to accommodate the duplex DNA (6).

As its gating role faded, a new role for the GL in transcript elongation has emerged. The GL was shown to interact with Escherichia coli RfaH (7), a specialized paralog of the essential transcription elongation factor NusG. Ubiquitous NusG homologs are thought to enhance elongation by bridging the βGL and the β′ clamp to stabilize the latter in a closed, pause-resistant conformation (8–10). Consistent with an important role of the GL, its sequence is relatively conserved in Bacteria (Fig. S1A), and it is essential for viability in E. coli (7). However, our findings that the enzyme lacking the GL did not exhibit defects in Rho-dependent termination (7), the essential function of NusG, suggested a different role for the GL.

Fig. S1.

GL conservation and essentiality. (A) Multiple sequence alignment of GL sequences from divergent bacterial species. Species names are abbreviated as follows: Eco, E. coli; Hpy, Helicobacter pylori; Tth, T. thermophiles; Syn, Synechocystis sp. PCC 6803; Bsu, Bacillus subtilis; Mtu, Mycobacterium tuberculosis; Tma, Thermotoga maritima; Aae, Aquifex aeolicus; Mge, Mycoplasma genitalium. Amino acid residues within the sequence that were replaced with two glycines are shaded as follows: hydrophobic, green; polar, olive; Pro and Gly, yellow; Asp and Glu, red; Arg, Lys, and His, blue. Amino acid residues flanking the replaced region are shaded gray. Cartoons on the right depict the GL from E. coli RNAP (PDB ID code 4IGC) and outline its interactions with the neighboring TEC elements. The presence of the NT DNA near R371 and E374 is hypothesized based on the assumption that the single-stranded region of the NT-DNA follows a similar route in the TEC and the initiation complex. (B) RNAP lacking the GL does not support cell growth. The E. coli DH5α cells carrying plasmids with the rpoB gene (WT, D516V, or D516V+ΔGL) under the control of an isopropyl β-d-1-thiogalactopyranoside (IPTG)-inducible P_trc promoter were grown in LB to OD₆₀₀∼0.3, induced with 0.2 mM IPTG for 1 h, and then challenged with 50 µg/mL rifapentine (Rif). Cell density was measured at the indicated times after the addition of rifapentine; the averages (± SD) were determined from three independent experiments. (C) A decrease in RNA synthesis upon induction of the ΔGL rpoB. RNA samples were collected from strains grown as in B after 20-min and 60-min incubation with rifapentine. Total RNA was extracted using hot phenol chloroform and RNAlater solution (Thermo Fisher Scientific). Approximately 100 ng of total RNA was analyzed using the 6000 Nano Chip Kit on an Agilent 2100 Bioanalyzer. (Left) Representative electropherograms of RNA samples collected after 60-min incubation with rifapentine. The lower marker (a standard run with every sample) was used for the alignment of the rRNA peaks. (Right) Percentage of 16S rRNA and 23S rRNA species relative to the total RNA at 20 min and 60 min after the addition of rifapentine for WT (blue), WT D516V (red), and ΔGL D516V (green) RNAPs.

In this work we found that, despite dramatic defects in growth (Fig. S1 B and C), the GL deletion conferred only mild defects in elongation, pausing, intrinsic termination, and RNA cleavage and did not abolish the response to the transcription elongation factors NusA and NusG in vitro. In contrast, the ΔGL enzyme exhibited strong defects during initiation: It formed unstable open complexes in which RNAP interactions with the downstream duplex DNA and the nontemplate (NT) DNA strand in the transcription bubble were compromised. Our results suggest that, as originally predicted (4), the GL guides the promoter DNA into its proper path in the RNAP holoenzyme.

Results

Deletion of the GL Alters the Structure of Open Promoter Complexes.

The discriminator region between the −10 hexamer and the transcription start site (TSS; +1) has recently emerged as an important modulator of the open complex properties (11–18). In a stable open complex formed by Thermus thermophilus RNAP, the −6 and −5 bases of the discriminator NT strand (NT^DISC) make crucial contacts with σ1.2, and βArg371 (E. coli numbering) interacts with the adjacent −4 and −3 bases (14). We hypothesized that the deletion of the GL may disrupt these interactions.

We first tested the effect of the GL at the bacteriophage λP_R promoter (Fig. 2A), which has been studied extensively by footprinting and kinetic approaches. Studies by Record and colleagues identified a series of λP_R promoter complexes that differ in RNAP–DNA interactions, including two unstable open intermediates, I₂ and I₃, and a very stable final open complex, RPo (19). During the isomerization from I₃ into RPo, RNAP tightens its interactions with NT^DISC and the downstream duplex DNA regions, massively stabilizing the complex (2, 19).

Fig. 2.

Footprinting of the ΔGL open complexes. (A) A linear λP_R fragment in which the NT DNA was labeled at the 5′ end with [γ³²P]-ATP. (B) Preformed open complexes were treated with 2 mM KMnO₄ for 30 or 60 s. After quenching, ethanol precipitation, and piperidine cleavage, the DNA was analyzed on an 8% denaturing gel. A representative of three independent experiments is shown; the 0 point is an untreated DNA control. Traces of the 60-s reactions were generated with ImageQuant; the band intensities normalized to the −10 signal (taken as 1) are shown in black (WT) and blue (ΔGL). (C) ExoIII was added to preformed promoter complexes. Aliquots were quenched at the indicated times (0 represents an untreated DNA control) and analyzed on a 6% denaturing gel; a representative of three independent experiments is shown. The positions of the modified residues (B) or the protection boundary (C) were identified using sequencing ladders.

To determine whether the GL removal alters the NT DNA–RNAP interactions, we carried out footprinting with potassium permanganate, which modifies single-stranded or unstacked T residues. In the WT λP_R RPo, the NT strand T residues −10, −4, −3, and +2 were accessible to modification (Fig. 2B), consistent with previous studies (5). A different pattern was observed in the complex formed with the ΔGL RNAP: Although the −10 and +2 residues were as sensitive to modification as in the WT complex, the −4 and −3 residues were significantly protected (Fig. 2B). This reduced reactivity may indicate that, upon the loss of contacts with GL, these bases are inserted into a σ2 pocket, as observed with an unfavorable NT^DISC sequence (18), or are allowed to stack, as has been proposed for the I₂ intermediate at λP_R, in which a quantitatively similar pattern of reduced permanganate reactivity of the −4/−3 residues was observed with the WT RNAP (20).

RNAP contacts with the downstream DNA are established during the final steps of λP_R RPo formation (19) and are absent in earlier unstable open complexes such as I₂. We hypothesized that the downstream DNA interactions also may be destabilized in the ΔGL complex. To probe the latter contacts, we used exonuclease III footprinting (Fig. 2C). RPo formed by the WT RNAP protected the DNA from ExoIII digestion to +23, similar to the downstream boundary observed with other probes (19). In ΔGL complexes, the protection was significantly reduced, with Exo III able to digest DNA to +1 in a fraction of complexes. Together, the footprinting results suggest that the ΔGL open complex resembles the WT I₂ intermediate in which the NT DNA and the downstream duplex are not yet loaded into their tracks.

The GL Stabilizes Open Complexes.

Substitutions of the discriminator bases and the σ and β residues that interact with the NT strand lead to decreases in open complex stability (11, 14, 16). To test whether the loss of the GL–NT^DISC contacts observed in the RPo structure (Fig. 1D) destabilizes the RPo, we carried out standard dissociation assays in which a preformed λP_R RPo is challenged with heparin, followed by measuring the remaining RPo by RNA synthesis (Fig. 3A). We found that deletion of the GL caused a 22-fold decrease in the λP_R RPo lifetime, from 65 to 3 min (Fig. 3A), comparable to the effects of substitutions of key NT bases and RNAP residues that interact with these bases (11, 14–16). Consistent with the deduction that T7A1 and rrnB P1 open complexes resemble λP_R I₃ and I₂ (19), the GL deletion had smaller (4.4- and 1.6-fold) effects at these promoters (Fig. 3A).

Fig. 3.

Effects of the GL deletion on promoter complex properties. (A, Upper) Promoters used in this study. The −35 and −10 hexamers are boxed. The discriminator element is underlined, and its positions are numbered. The TSS and the D2 (−5 at λP_R) base that makes key contacts with σ1.2 are indicated. (Lower) The half-lives of open complexes (in minutes) were measured by assaying the fraction of transcriptionally competent complexes following the competitor challenge (SI Materials and Methods) in at least three independent repeats; data are shown as mean ± SD. The GL effect was defined as the ratio of the half-lives of the WT and ΔGL open complexes. (B, Upper) Open complexes assembled on λP_R with 4-thio-dT at the −4 NT position were exposed to 365-nm UV light. The reactions were separated on 4–12% Bis-Tris gels. (Lower) Relative cross-linking to σ and β (normalized to DNA) was calculated from four independent experiments; error bars indicate the SD. (C) Abortive initiation at λP_R. Open complexes were formed in the presence of [γ³²P]-ApU and were incubated with 250 μM NTPs (see Fig. S2). A representative gel and trace analysis of abortive products are shown. (D) TSS at rrnB P1 were mapped by primer extension of in vitro-transcribed RNAs. Extension products were analyzed on a denaturing 12% gel along with the sequencing ladder generated with the same primer. Positions corresponding to a TSS at 6 and 9 are indicated.

GL Contacts with the Discriminator.

Permanganate footprinting (Fig. 2B) suggests that the removal of the GL alters RNAP interactions with the −4 and −3 bases at λP_R. To explore the conformational changes upon the loss of the GL further, we carried out cross-linking with a “zero-length” 4-thio-dT incorporated at the λP_R NT-4 position. This residue would be expected to cross-link to β and σ, as was indeed observed (Fig. 3B). With the WT RNAP, cross-linking to β was more efficient than to σ (∼65 vs. 35%). When holoenzymes were assembled with σ carrying Y101A and M102A substitutions in σ1.2, which have been shown to weaken interactions with the adjacent NT^DISC residues (16), cross-linking to σ was diminished (∼20–25%), as expected. With the ΔGL enzyme, cross-linking to the β and σ subunits was equally efficient (within the experimental error) (Fig. 3B), suggesting that the removal of the GL preferentially weakens the NT −4 base interactions with the β subunit. The residual cross-linking is likely caused by the repositioning of the NT^DISC region, e.g., toward β392 that has been shown to cross-link to D4 (21). Although the loss of cross-linking cannot be interpreted as evidence for direct GL–DNA contacts, these results are fully consistent with the GL–DNA contacts observed in crystal structures (14, 22, 23).

Interactions between the NT DNA and RNAP elements exhibit some sequence preference; e.g., C is strongly disfavored at D2, and G is preferred at +2 (14, 15). We wondered if GL–NT interactions are also sequence specific. We first tested the effect of simultaneously substituting the −3 and −4 T residues with C. This double substitution reduced open complex lifetime at λP_R by fourfold with the WT RNAP (Fig. 3A) but only by 1.5-fold with the ΔGL RNAP. Because both the GC content and purine/pyrimidine bases in the discriminator (18) alter open complex properties, we substituted −3 and −4 with A individually. These changes slightly destabilized the WT complexes but had an opposite effect on the ΔGL complexes (Fig. 3A). The effects were similar at both −3A and −4A promoters, reducing the GL contribution to ∼13.5-fold.

The β residue Arg371 is highly conserved in Bacteria (Fig. S1A) and makes direct contacts with the −3 and −4 NT DNA residues in the bacterial RPo (14). To evaluate the contribution of Arg371 to the GL-specific effects on promoter complex properties, we replaced this residue with alanine and measured the stability of the open complexes formed on the WT and −3, −4 CC λP_R promoters. The βR371A substitution reduced the lifetime of the WT λP_R complexes approximately sevenfold (to 9.7 ± 0.9 min), compared with the 22-fold effect of removing the entire GL (Fig. 3A). On the CC promoter, the βR371A complexes were only slightly more stable (2.5 ± 0.2 min) than those formed by ΔGL RNAP (1.9 ± 0.3 min). These results suggest that Arg371 establishes functional, base-specific interactions with the discriminator region that stabilize the open promoter complexes.

Deletion of the GL Reduces Abortive Synthesis.

rrnB P1 has an 8-nt discriminator with a C residue at the −7 (D2) position (Fig. 3A) that, together with a suboptimal spacer region, precludes productive contacts with σ1.2 (15). Despite forming very unstable complexes, rrnB P1 is one of the strongest promoters, in part because it does not produce abortive RNAs. C–7G substitution restores favorable interactions with σ1.2, stabilizing the rrnB P1 promoter complex ∼40-fold (15) and impeding promoter escape (17). A very stable λP_R RPo with a 6-nt discriminator makes favorable contacts with σ1.2 and produces abundant abortive products; the corresponding G–5C substitution reduces its half-life 14-fold (15). As is consistent with a stabilizing role of GL-NT^DISC contacts, we found that the GL deletion reduced abortive synthesis at λP_R (Fig. 3C). The major effect was observed with the 4-mer RNA, which is too short to clash directly with σ3.2 (24, 25), supporting a model in which a scrunched template strand stimulates RNA release (21).

The decrease in abortive synthesis by the ΔGL RNAP could be caused by reduced RPo stability (Fig. 3A) rather than by the loss of NT strand contacts. To evaluate this possibility, we tested the effects of mutations in RNAP that decreased or increased RPo stability (19, 26) on escape from λP_R. We found that neither destabilizing (β′ΔSI3 and β′R339A) nor stabilizing (σ⁷⁰Δ1–55) changes altered the pattern of abortive products (Fig. S2A). We conclude that decreasing the stability of the λP_R complex per se is not sufficient to facilitate promoter escape.

Fig. S2.

Effects of RNAP variants and promoter region swaps on abortive synthesis. (A) Abortive synthesis at the λP_R promoter. ApU, abortive RNAs, terminated (70 nt), and run-off (153 nt) RNAs and an unidentified arrested product are indicated. (B) A schematic representation of RNAP holoenzyme interactions with promoter elements. The sequences of the T7A1 and λP_R promoters that have a 7- and a 6-nt discriminator, respectively, are shown. (C) The pattern of abortive products depends on the discriminator region. Abortive transcription on synthetic hybrid templates that contain promoter regions from T7A1 (open boxes) or λP_R (hatched boxes). Reactions were carried out as in A with the WT RNAP. Variants with the λP_R discriminator (indicated by red dots) produce abundant abortive products, whereas those with the T7A1 discriminator do not.

We next tested the effect of the discriminator–RNAP interactions on abortive synthesis. Increasing the distance between the −10 region and the start site is expected to weaken these contacts, facilitating promoter escape (17). In agreement with recent reports (12, 17), we found that the discriminator, but not other promoter regions, determined the differences in abortive synthesis between the T7A1 and λP_R promoters (Fig. S2C). The Y101A substitution in σ1.2, which compromises σ contacts with the NT^DISC (16), reduced abortive synthesis at λP_R and eliminated the RNAP ability to “sense” the length of the discriminator (Fig. S3).

Fig. S3.

The length but not the sequence of the discriminator influences promoter escape. (Upper) Variants of the λP_R promoter containing a 6-nt (red) or 7-nt (black) discriminator, including those from λP_R and T7A1. (Lower) Abortive synthesis by the WT and σY101A holoenzymes. For the WT RNAP, promoters with 6-nt discriminators give rise to more abundant abortive RNAs than those with a 7-nt discriminator. These differences disappear when RNAP–NT strand interactions are disrupted by the Y101A substitution.

Deletion of the GL Affects TSS Selection.

RNAP can start transcription at a range of positions relative to the −10 hexamer (27), and σ1.2–NT^DISC interactions play a key role in TSS selection (17). WT rrnB P1 complexes, in which DNA is scrunched, initiate at 9A (+1), whereas the C–7G substitution or spacer insertions that reduce scrunching shift the TSS to a “standard” 6A (17). By altering the NT^DISC path through RNAP, the GL also may contribute to TSS choice. We found that the ΔGL RNAP used the 6A start site more efficiently than the WT enzyme (Fig. 3D), suggesting that the GL, which contacts the NT^DISC just upstream of the NT DNA segment extruded upon scrunching (18, 21), may stabilize the scrunched state.

The GL Is Largely Dispensable for Transcript Elongation Control.

The location of the GL in the transcript elongation complex (TEC) suggests that it may affect RNAP processivity and response to NusG (Fig. 1 A and C), which inhibits pausing, particularly at sites where RNAP is prone to backtracking (28). We first tested the effect of the GL deletion on transcription through two tandem GGGAUGCGUGCG pause sites that fit the ubiquitous pause consensus (29). The pausing patterns for the WT and ΔGL enzymes were very similar (Fig. 4A). In the absence of NusG, the ΔGL RNAP reached the end of the template slightly sooner than the WT enzyme, whereas in the presence of NusG, the two enzymes elongated the nascent RNA at nearly the same rate (Fig. 4B). The modest defects in pausing suggest that the GL deletion may inhibit backtracking. This conclusion is supported by our observation that TECs formed with ΔGL enzyme were resistant to the nascent RNA cleavage stimulated by the transcript cleavage factor GreB (Fig. S4) and by observations of RNAP backtracking in real time (30).

Fig. 4.

The GL RNAP is responsive to NusG. (A) Single-round elongation assays were carried out on the pVS54 template; positions of pause sites, the hlyT terminator, and the run-off (RO) are shown. Halted [α³²P]-CMP–labeled A38 TECs formed with the WT or ΔGL RNAP were chased in the presence or absence of NusG. Reactions stopped at the indicated times were analyzed on 8% denaturing gels. (B) The average fraction (± SD) of run-off RNA determined from three independent experiments, including that shown in A. (C) Close-up view of NusG contacts with the GL and NT-DNA in the TEC model. The positions of residues replaced by alanine in NusG (7) and Spt5 (31) are shown in magenta.

Fig. S4.

ΔGL TECs are resistant to GreB-facilitated transcript cleavage. Linear pIA226 DNA template (100 nM), holo RNAP (200 nM; WT or ΔGL), ApU (100 µM), and starting NTPs (1 µM GTP, 5 µM ATP and UTP, 10 µCi [α³²P]-GTP, 3,000 Ci/mmol) were mixed in 30 μL of TGA-2 (20 mM Tris-acetate, 20 mM Na-acetate, 2 mM Mg-acetate, 5% glycerol, 1 mM DTT, 0.1 mM EDTA, pH 7.9) and were incubated for 15 min at 37 °C. Halted A26 complexes were purified by gel filtration through a G-50 spin column (GE Healthcare) equilibrated in TGA-2, diluted fourfold, and stored on ice. Reactions were initiated by shifting samples to 37 °C. GreB (20 nM) was added where indicated. Samples were removed at the times shown and were quenched by the addition of an equal volume of STOP buffer (10 M urea, 20 mM EDTA, 45 mM Tris-borate; pH 8.3). (Upper) A schematic of the experimental setup and the sequence of the A26 RNA transcript, with the positions of [α³²P]-GTP indicated in bold. (Lower) RNA products were separated on an 8% denaturing urea-acrylamide gel. The percent of A26 remaining was calculated relative to A26 RNA at time 0 (before GreB addition).

The lack of a strong NusG effect contrasts with our earlier conclusion that the GL is required for RNAP modification by NusG (7). We identified an HTTT motif as a contact point of RfaH to the GL and replaced corresponding SWHL residues in NusG with four alanines. This substitution abolished the NusG anti-pausing activity but not the binding to the TEC because the same variant was fully capable of enhancing Rho-dependent termination (7). We therefore attributed the loss of the anti-pausing activity to the loss of NusG–GL contacts. However, the replaced residues likely also interact with the NT DNA (Fig. 4C). Crickard et al. recently reported that replacing six residues of Spt5, a yeast homolog of NusG, located on the same face as the SWHL NusG motif with alanines abolished Spt5 cross-linking to the NT DNA and its anti-arrest activity (31). Although it is uncertain if the altered surface residues (which are only weakly conserved) or more extensive structural changes (such as local refolding caused by altered core residues) lead to the observed phenotypes, it seems probable that the four-alanine substitution in NusG abolishes its anti-pausing activity by altering interactions with the NT DNA and not with the GL.

We next asked whether the removal of the GL may influence the clamp dynamics, altering RNAP response to hairpin-dependent sites such as hisP, at which interactions between the nascent RNA hairpin and the β flap domain induce opening of the clamp (3). Acting in concert with RfaH, which fills the gap between the GL and the clamp tip, the GL reduces RNAP pausing at hisP (7). However, consistent with our previous findings (7), deletion of the GL reduced pausing at the hisP site (Fig. S5A), arguing that the GL does not favor the clamp closure in the absence of a bridging factor. Mild defects of the ΔGL RNAP observed at intrinsic terminators that use different mechanisms of release (Fig. S5C) are also likely caused by defects in pausing.

Fig. S5.

Effects of GL deletion on hairpin-dependent pausing and termination. (A) Single-round assays on the pIA171 template encoding the hisP signal. Halted [α³²P]-CMP–labeled A29 TECs were formed with the WT or ΔGL RNAP and chased with NTPs (10 μM GTP, 150 μM ATP, CTP, UTP) and rifapentine in the presence or absence of 100 nM NusA. Samples were analyzed on 8% denaturing gels. The hisP half-life calculated as described in ref. 28 is shown below each panel. (B) Termination at a NusA-dependent rsxC terminator. (Upper) Transcript generated from the λP_R promoter on a linear pIA1239 DNA; the TSS (bent arrow), C-less region (residues 1–26), rsxC terminator (release at 107), and transcript end (179) are indicated. (Lower) Halted A26 TECs were formed at 50 nM with WT or ΔGL RNAP. Termination was assayed in single-round A26 RNA extension by the addition of all four NTPs (to 150 µM) and heparin (at 10 µg/mL) in the absence or presence of 100 nM NusA. The reactions were incubated for 10 min at 37 °C and were quenched. Products were analyzed on a 6% denaturing gel. Positions of terminated (Term) and run-off (RO) RNAs are shown on the left. Termination efficiency (terminated transcript as a fraction of total RNA) was determined in three independent experiments. (C) Efficiency of termination at selected terminators. Halted [α³²P]-labeled complexes were chased with 150 μM NTPs and rifapentine. Termination efficiencies for the WT (white bars) and ΔGL (black bars) RNAPs were calculated from three independent experiments. The average values are shown inside the bars; the error bars indicate SD. All intrinsic termination signals are composed of an RNA hairpin followed by a U-rich region, but their mechanisms of RNA release may differ. Three models of termination have been proposed (52). In the forward translocation model, formation of the hairpin pushes RNAP forward without concomitant RNA extension, shortening the RNA:DNA hybrid and thereby destabilizing the TEC. In the hybrid-shearing model, the hairpin pulls the nascent RNA out. In the allosteric model, the hairpin induces a structural change in the RNAP that facilitates melting of the hybrid. Among the terminators tested here, t500 relies on forward translocation, but tR2 and hisT do not (52); the mechanisms of others are not known. The ΔGL RNAP terminated somewhat less efficiently at five of six terminators. The lack of defect at hisT may be explained by its unusual structure: It is composed of an exceptionally stable hairpin followed by a run of nine U residues. At other terminators the hairpin is weaker, and U-tracks are interrupted by G and C residues. It is possible that the mild pausing defect of ΔGL RNAP is sufficient to reduce pausing at an interrupted U-track, thereby reducing termination, whereas the perfect U would overcome this defect.

NusA, an essential transcription factor which acts in part through binding to an RNA hairpin as it emerges from the RNA exit channel (1), stimulates pausing and termination. A failure to respond to NusA could explain the growth defect of ΔGL RNAP. However, we found that NusA increased pausing and termination (Fig. S5) by the ΔGL and WT enzymes similarly. In summary, these and prior (7) results demonstrate that deletion of the GL leads to modest defects in elongation, transcript cleavage, and termination and does not abolish the RNAP response to general transcription factors, at least in vitro.

Discussion

In this work, we show that the β subunit GL modulates every step of the transcription cycle but has the most pronounced effects during initiation. The removal of the GL altered the structure and properties of promoter complexes, apparently blocking isomerization into a long-lived open complex that forms at some promoters, such as λP_R. We cannot exclude the possibility that the GL deletion alters the structure of transcription complexes indirectly, but several arguments support a hypothesis that all the effects of the GL are mediated through its interactions with the single-stranded (ss) NT DNA. First, the GL is a surface loop, which we replaced with two glycine residues to preserve the positions of the flanking β regions. Second, the GL residue Arg371 makes direct contacts with the NT DNA in promoter complex structures (14, 22, 23), and our results show that the βR371A substitution destabilizes promoter complexes. Third, the GL does not interact with other core RNAP regions. Fourth, although the GL closely approaches σ1.2, these hypothetical contacts cannot (i) explain the GL effects during elongation or (ii) fully account for its effects on open complexes, because a large part of the GL effect depends on the discriminator sequence, and a “destabilizing” M102A substitution in σ1.2 and a longer discriminator region in T7A1 reduce the GL effect similarly (Fig. 3A).

During initiation, the GL forms part of a relay in the NT strand interactions that commence with σ capturing the −11A base in a pocket, unzipping the −10 region to place the −7 base in another pocket (32), and making contacts with two bases downstream of the −10 hexamer (14, 16). The DNA then bends into the channel, with the GL stabilizing the NT strand and guiding it toward the β CRE pocket, which may capture the +2 base in a stable final open complex (14). Once RNAP escapes from the promoter, the GL may continue to guide the NT strand, sometimes acting in concert with elongation factors that take the place of the released σ.

GL–NT DNA Interactions in Initiation.

Bacterial promoters exhibit astonishing diversity in sequences, strength, and sensitivity to regulation. Although some, such as rrnB P1, are exquisitely tunable by cellular cues, others, such as λP_R, appear to be optimized for steady RNA output. These properties are determined, in part, by the differing structures of open complexes. Record and colleagues argued that open complex intermediates transiently populated at λP_R correspond to final open complexes at other promoters, with λP_R I₂ resembling the rrnB P1 open complex (19). Footprinting analysis shows that in I₂ the contacts with NT DNA are loose, the clamp is not locked, and the downstream DNA is not held tightly (19, 20). Our results support a model in which RNAP interactions with the discriminator mediate the transition from I₂ to RPo and implicate the GL in this isomerization step.

The discriminator region was proposed to direct an elaborate cascade of interactions and conformational changes that determine the structure of open complexes at different promoters (19). At λP_R, a short 6-nt discriminator with G at D2 favors tight interactions with RNAP and enables dramatic, 10⁵-fold stabilization of the final RPo relative to an early, competitor-sensitive I₂. At T7A1, a longer discriminator with a suboptimal A at D2 weakens the contacts, reducing the extent of stabilization to ∼250-fold and mimicking λP_R I₃. Finally, rrnB P1 appears never to progress beyond the initial (I₂-like) unstable open complex because its 8-nt discriminator with C at D2 cannot form productive contacts with RNAP. In this model, RNAP elements that interact with the discriminator are expected to be critically important at promoters that form long-lived open complexes but largely dispensable at promoters forming unstable complexes. Indeed, alanine substitution of σM102, which makes a van der Waals contact with G at D2 (14, 16), dramatically destabilizes very stable complexes formed at λP_R and rrnB P1 C–7G variant (16) but has only a small effect at WT rrnB P1 (16). Similarly, deletion of the GL has a 22-fold effect at λP_R, compared with 4.4- and 1.6-fold effects at T7A1 and rrnB P1, respectively (Fig. 3A).

It is also possible that ternary interactions between GL, NT^DISC, and σ1.2 not only position the bubble and keep it from collapsing but also stabilize the clamp in a closed conformation. Notably, promoter complex intermediates have been proposed to have an open clamp conformation (19).

GL–NT DNA Interactions in Elongation.

In contrast to its well-established role in initiation, the NT DNA effects on elongation are less clear because of the lack of experimental evidence on the NT DNA path in TEC. Analysis of RNAP lacking the GL could provide insights into the role of the NT strand in elongation. The GL is located nearly 60 Å from the RNAP active site and more than 20 Å from the upstream and downstream DNA duplexes, but deletion of the GL reduces RNAP backtracking (Fig. 4A and Fig. S4) and pausing at hisP (Fig. S5A). We speculate that removal of the GL alters the path of the ss NT DNA through the TEC. The NT DNA is absent or unresolved in TEC structures but is well resolved in factor-stabilized initiation and initially transcribing complexes (14, 22, 33, 34). In the latter structures, the NT DNA separates from the template DNA immediately downstream of the β fork loop, passes along the inner side of the GL, exits the cleft between the β lobe and β protrusion domains, loops around σ2, and rejoins the template DNA to form the upstream DNA duplex (Fig. 1). The NT DNA can be modeled to follow a similar path in the TEC (Fig. 1A), looping around the NusG NTD instead of σ2. These contacts likely account for sequence-specific interactions with NusG homologs (29, 35, 36). In this model, the GL restricts the downstream portion of the bubble inside the cleft, biasing the NT DNA to loop out upstream and facilitating DNA reannealing downstream (Fig. S6), thereby promoting backtracking and pausing at hisP, which occurs in a pretranslocated register. In the absence of the GL, the NT DNA could adopt an unconstrained conformation (Fig. S6), eliminating the backward translocation bias and reducing RNAP sensitivity to backtrack-prone and pretranslocated pauses.

Fig. S6.

The GL alters the NT DNA path in transcription complexes. (A) In elongation complexes, the GL contacts with the NT strand weaken base pairing at the upstream fork junction while stabilizing the downstream base pairs. In the absence of the GL, the upstream junction is strengthened, and backtracking is inhibited. (B) In prescrunched rrnB P1 ICs, the GL binds the NT DNA just upstream from the point of scrunching, and transcription initiates at a distal (9A) site. In the absence of the GL, the NT DNA is more relaxed, allowing utilization of a proximal TSS at 6A.

The NT DNA has been implicated in the action of accessory factors that target the elongating RNAP (29, 35–37). It is highly likely that, as in σ, these regulators establish simultaneous contacts with the GL and NT DNA. Although some factors (e.g., RfaH) critically depend on both sets of contacts, others (e.g., E. coli NusG) do not (Fig. 4A). We stand by our hypothesis that the protein–protein contacts of RfaH and NusG with the GL restrict the clamp (7), but we now suggest that these paralogous factors lock the clamp in different states. Although NusG may stabilize the clamp in a relatively open conformation (10), the smaller RfaH may restrict the clamp in a more closed state. These differences would explain why RfaH reduces pausing at the hisP site that is accompanied by the clamp opening (3), whereas NusG does not.

In conclusion, our results and the available structural data support a model in which the GL influences the regulatory properties of transcription complexes through a combination of direct contacts with the NT strand and with the initiation/elongation factors.

Materials and Methods

Details for all procedures are in SI Materials and Methods. Plasmids are listed in Table S1, and oligonucleotides are listed in Table S2.

Table S1.

Plasmids

Name	Key features	Source
Transcription templates
pIA171	T7A1 promoter–A29–his pause	(46)
pIA226	λPR promoter–A26–his pause	(24)
pIA255	λPR promoter–A26–his terminator	(24)
pIA263	λPR promoter–A26–T7 Te terminator	(24)
pIA264	λPR promoter–A26–T3 Te terminator	(24)
pIA266	λPR promoter–A26–rrnB T1 terminator	(24)
pIA536	rrnB P1 promoter	(47)
pIA1116	T7A1 promoter–C105–λtR2 terminator	This work
pIA1141	λPR promoter–rrnB T1 terminator	This work
pIA1239	λPR promoter–A26–rsxC terminator	(48)
pTS111	ϕ82 late promoter–U25–t500 terminator	(49)
pVS54	T7A1 promoter–A38–P1–P2–hlyT terminator	(50)
Protein expression vectors
pIA160	Ptrc promoter–His6rpoB	(51)
pIA183	Ptrc promoter–His6rpoB[D516V]	(28)
pIA247	PT7 promoter–His6nusG	(28)
pIA370	PT7 promoter–His6nusA	(28)
pIA577	PT7 promoter–greBHis6	(41)
pIA898	Ptrc promoter–His6rpoB[D516V+Δ368–376)	(7)
pIA1039	PT7 promoter–rpoA–His6rpoB[Δ368–376]–rpoC–rpoZ	(7)
pIA1127	PT7 promoter–rpoD	This work
pIA1165	PT7 promoter–rpoD Y101A	This work
pIA1166	PT7 promoter–rpoD M102A	This work
pVS10	PT7 promoter–rpoA–rpoB–rpoCHis6–rpoZ	(35)

Table S2.

Oligonucleotides

Primer no.	Sequence	Application
2	GTAAAACGACGGCCAGT	Top strand/pTS111 termination
3	AACAGCTATGACCATG	Bottom strand/pTS111 termination
17	CGTTAAATCTATCACCGCAAGG	Top strand/pIA226 termination; footprinting
256	CAGTTCCCTACTCTCGCATG	Bottom strand/pIA171, pIA226, pVS54 elongation; termination
310	TCAGGAATTCGAACGCGGTCAGAAAATTATTTT	Top strand/pIA536 open complex stability
338	GGAGAGACAACTTAAAGAGA	Top strand/pIA171, pVS54 pausing; open complex stability
1527	CATGCACCACTGGAAGATC	Bottom strand/pIA171 open complex stability
1532	GACCCAAGCTTCGTGTCAGTG	Bottom strand/pIA536 open complex stability
1742	GCTTGATTCTAGCTGATCGTGGA	Top strand/pIA1116, pIA1141 abortive synthesis; termination
1743	TAATCTAGCTGATCGTGGACCGA	Bottom strand/pIA1116, pIA1141 abortive synthesis; termination
2442	GTTGTGATATCGTCAGGATGATGGTGATG	Bottom strand/pIA226 ExoIII footprinting
2461	CATACAACCTCCTTACTACATGCAACCATT	Bottom strand/pIA226 open complex stability
2462	CATACAACCTCCTTACTACATGCGGCCATT	Bottom strand/pIA226 (−4, −3 CC) open complex stability
2471	AGAAAATTATTTTAAATTTCCTCTTGACAAAAGTGTTAAATTGTGCTATAATGGTTGCATGTAGTAAGGAGGTTGTATG	Consensus promoter, top strand open complex stability
2472	CATACAACCTCCTTACTACATGCAACCATTATAGCACAATTTAACACTTTTGTCAAGAGGAAATTTAAAATAATTTTCT	Consensus promoter, bottom strand open complex stability
2474	CATACAACCTCCTTACTACATGCATCCATTATCA	Bottom strand/pIA226 (−4 A) open complex stability
2475	CATACAACCTCCTTACTACATGCTACCATTATCA	Bottom strand/pIA226 (−3 A) open complex stability
2494	GTTCGGCATGGGGTCAGGTG	Bottom strand/pIA536 TSS mapping
2500	CTAACACCGTGCGTGTTGACTATTTTACCTCTGGCGGTGATAATGG[S4dT]TGCATGTAGTAAGGAGGTTGTAT	Top strand λPR cross-linking
2501	ATACAACCTCCTTACTACATGCAACCATTATCACCGCCAGAGGTAAAATAGTCAACACGCACGGTGTTAG	Bottom strand λPR cross-linking

Elongation Assays.

Halted [α³²P]-NMP–labeled TECs were formed on linear DNA templates with ApU dimer and starting NTP subsets. Following incubation with NusG (100 nM), where indicated, transcription was restarted by the addition of nucleotides at the concentrations indicated in the figures and 50 μg/mL rifapentine at 37 °C to limit the elongation to a single round. Aliquots were withdrawn at selected times, quenched, and analyzed on denaturing urea-acrylamide gels.

Open Complex Stability Assays.

Linear DNA templates were incubated with RNAP holoenzyme at 37 °C for 15 min. At time 0, a competitor was added (heparin at 20 μg/mL for λP_R, heparin at 10 μg/mL for T7A1, or a 200 nM consensus promoter DNA fragment for rrnB P1). Aliquots were withdrawn at selected times and were added to a prewarmed mixture of a dinucleotide primer and NTP substrates specified by the promoter sequence, including an [α³²P]-NTP.

Footprinting Analysis.

The linear λP_R promoter fragment was made by PCR amplification with the [γ³²P]-ATP–labeled NT strand primer 17. Open complexes were assembled with the WT or ΔGL RNAP and were probed with KMnO₄ or ExoIII (New England Biolabs). The positions of modified/protected residues were identified using sequencing ladders generated with the same primer.

SI Materials and Methods

Reagents and Proteins.

All general reagents were obtained from Sigma Aldrich and Fisher; NTPs, [γ³²P]-ATP, and [α³²P]-NTPs were obtained from GE Healthcare and Perkin-Elmer; PCR reagents and restriction and modification enzymes were obtained from New England Biolabs, Roche, and Epicentre. Oligonucleotides were obtained from Integrated DNA Technologies and Sigma Aldrich. DNA purification kits were from Qiagen. E. coli RNAP (40), NusG (28), NusA (28), and GreB (41) were purified as described previously.

Elongation and Termination Assays.

Linear DNA template generated by PCR amplification (30 nM), holo RNAP (40 nM), ApU (100 µM), and starting NTPs (1 µM CTP, 5 µM ATP and UTP, 10 µCi [α³²P]-CTP, 3,000 Ci/mmol) were mixed in TGA-10 (20 mM Tris acetate, 20 mM Na acetate, 10 mM Mg acetate, 5% glycerol, 1 mM DTT, 0.1 mM EDTA, pH 7.9). Reactions were incubated for 10 min at 37 °C, followed by a 3-min incubation with NusA (100 nM) or NusG (100 nM) where indicated. Transcription was restarted by the addition of nucleotides (at the concentrations indicated in the figures) and 50 μg/mL rifapentine at 37 °C. Aliquots were withdrawn at selected times and quenched by the addition of an equal volume of STOP buffer (10 M urea, 20 mM EDTA, 45 mM Tris-borate; pH 8.3). Termination assays were carried out as described in ref. 24.

Open Complex Stability Assays.

Linear DNA templates generated by PCR amplification (2 nM for rrnB P1, 20 nM for the T7A1 and λP_R promoters) were incubated with the RNAP holoenzyme in TGA-10 (for rrnB P1 and λP_R) or 40 mM Tris⋅HCl (pH 7.9), 10 mM MgCl₂, 30 mM NaCl, 14 mM 2-mercaptoethanol, and 0.1 mM EDTA (for T7A1) at 37 °C for 15 min. At time 0, a competitor was added (heparin at 20 μg/mL for λP_R, heparin at 10 μg/mL for T7A1, or a 200 nM consensus promoter DNA fragment for rrnB P1). Aliquots were withdrawn at selected times and were added to a prewarmed mixture of a dinucleotide primer and NTP substrates specified by the promoter sequence. For λP_R, 200 µM ApU, 4 µM GTP, and 10 µCi [α³²P]-GTP were used to generate the ApUpG product. For T7A1, 200 µM ApU, 200 µM CTP, 4 µM GTP, and 10 µCi [α³²P]-GTP were used to generate ApUpCpG. For rrnB P1, 200 µM ApC, 200 µM UTP, 4 µM GTP, and 10 µCi [α³²P]-GTP were used to generate ApCpUpG. The reactions were incubated for 5 min at 37 °C and were quenched by an equal volume of STOP buffer.

KMnO₄ Footprinting.

A linear DNA fragment containing the λP_R promoter was made by PCR amplification on pIA226 with primers 17 and 256 (Table S2); the NT strand primer (no. 17) was end-labeled with [γ³²P]-ATP using polynucleotide kinase (PNK) (New England Biolabs) and purified using G-50 spin columns (GE Healthcare). Sequencing reactions were performed with [³²P]-labeled primer no. 17 with the Sequenase version 2.0 DNA sequencing kit (Affymetrix USB). Open complexes were formed with RNAP (40 nM) and the labeled DNA template (30 nM) for 15 min at 37 °C in GBB buffer (20 mM Tris⋅HCl, 20 mM NaCl, 14 mM MgCl₂, 5% glycerol, and 0.1 mM EDTA; pH 7.9). Samples were shifted to room temperature and treated with 2 mM KMnO₄ for 30–60 s. Aliquots were quenched by the addition of 2.5 μL of β-mercaptoethanol and 2.5 μL of 0.5 M EDTA. The volume was adjusted to 100 μL with H₂O. Samples were subjected to phenol-chloroform extraction. Fifty microliters of GEA mix (1 mg/mL glycogen, 25 mM EDTA, 0.3 M Na acetate) were added, followed by 2.5 volumes of ethanol. Pellets were dissolved in 20 µL H₂O and incubated with 100 µL of 0.5 M piperidine at 95 °C for 12 min. After the addition of 70 μL GEA and ethanol precipitation, DNA was dissolved in 96% formamide containing 0.1% xylene cyanol and 0.1% bromophenol blue.

Exonuclease Footprinting.

A linear DNA fragment was made by PCR amplification on pIA226 with primers 17 and 2442; the NT strand primer (17) was end-labeled with [γ³²P]-ATP using PNK (New England Biolabs) and purified using G-50 spin columns (GE Healthcare). The open complexes were formed with 100 nM WT or ∆GL RNAPs and 250 nM end-labeled DNA in 20 mM Tris⋅HCl (pH 7.9), 5% glycerol, 40 mM KCl, and 5 mM MgCl₂ (TB-40) for 20 min at 37 °C. The complexes were captured on 15 µL Ni-NTA agarose beads (Qiagen), washed to remove unbound DNA, eluted from beads with the addition of 90-mM imidazole in a 15-µL volume, spun in a Durapore (PVDF) 0.45-µm Centrifugal Filter Unit (Merck Millipore), and resuspended in 400 µL of TB-40. ExoIII (New England Biolabs) was diluted in TB-40 and added to the complexes at a final concentration of 4 U/µL for the indicated amount of time. Reactions were quenched with Exo stop buffer [8 M urea, 20 mM EDTA, 1× Tris/borate/EDTA (TBE), 0.5% Brilliant Blue R and 0.5% Xylene Cyanol FF], and samples were resolved on 6% polyacrylamide-7M urea gels.

Abortive Initiation Assays.

Forty-five nanomoles of ApU (Sigma) were 5′-end–labeled with [γ³²P]-ATP using T4 PNK (Epicentre) in a 15-µL reaction. To remove unincorporated ATP, 200 µL of GEA mix [1 mg/mL glycogen, 0.3 M Na acetate (pH 7.0), 12 mM EDTA] and 600 µL 100% EtOH were added, followed by overnight incubation at −20 °C. The precipitate was pelleted for 15 min at 20,000 × g at 4 °C, washed twice with 1 mL of cold 70% EtOH, air dried, and resuspended in 25 µL diethyl pyrocarbonate-treated H₂O to yield ∼1 mM ³²P-ApU (∼50% of ApU is lost during precipitation). Open complexes were formed on a linear PCR-generated pIA1141 template, which encodes the λP_R promoter followed by the rrnB T1 terminator in TB10 [20 mM Tris⋅HCl (pH 8.0), 120 mM KCl, 10 mM MgCl₂, 1.4 mM β-ME, 0.1 mM EDTA]. The DNA template (100 nM), RNAP holoenzyme (150 nM), and ³²P-ApU (100 µM) were incubated for 15 min at 37 °C. Substrate NTPs (final) were added to 250 µM, followed by a 10-min incubation at 37 °C. (This assay can be run in a single-round format with 25 µM ApU present during the open complex formation and a 50-fold molar excess of unlabeled ApU added during the chase.) Reactions were stopped by adding two volumes of 3× STOP buffer [20 mM Tris⋅HCl (pH 8), 20 mM EDTA, 0.1% amaranth, 0.1% xylene cyanol, 94% formamide] and were separated on a 12% (19:1) gel in 0.5× TBE. The run was stopped when amaranth (which runs as a dimer) was 2 cm away from the end. The gels were dried, exposed to phosphor screens, and scanned using a TyphoonFLA 9000 (GE Healthcare Life Sciences).

Cross-Linking.

A 70-nt long [S4dT] modified NT strand oligonucleotide (Gene Link) was end-labeled with [γ³²P]-ATP and PNK (New England Biolabs), purified with the QIAquick Nucleotide Removal Kit (Qiagen), and annealed to a complementary template DNA strand oligonucleotide (Sigma Aldrich). Open complexes were formed with 80 nM RNAP holoenzyme and 40 nM template in 20 mM Tris-acetate, 20 mM Na-acetate, 2 mM Mg-acetate, 5% glycerol, 1 mM DTT, 0.1 mM EDTA (pH 7.9) at 37 °C for 30 min and were exposed to 365 nM UV light (8W Model UVLMS-38; UVP, LLC) for 20 min on ice. The reactions were quenched with the LDS loading dye and separated on 4–12% Bis-Tris Novex gels (Invitrogen). The gels were dried and visualized by phosphor imaging.

TSS Mapping by Primer Extension.

Oligo 2494 (6 μM) was end-labeled with [γ³²P]-ATP. Supercoiled pIA536 plasmid template DNA (4 nM) was incubated with 40 nM of WT or ΔGL RNAP in TGA-8 buffer (20 mM Tris acetate, 20 mM Na acetate, 8 mM Mg acetate, 5% glycerol, 1 mM DTT, 0.1 mM EDTA, pH 7.9) at 37 °C for 10 min to form an open complex. RNA synthesis was initiated by the addition of 500 μM ATP, GTP, CTP, and UTP. After 20-min incubation at 37 °C, RNA was extracted with GEA and acid phenol/chloroform (Ambion) and was precipitated with ethanol. The RNA was dissolved in 15 μL of H₂O, and 5 μL was annealed to 1 pmol labeled primer by heating at 90 °C for 2 min followed by immediate chilling on ice. Primer extension was performed by the addition of 100 U SuperScript II Reverse Transcriptase (Invitrogen), 1 mM dNTPs, SuperScript buffer, DTT, and 10 U Murine RNase inhibitor (New England Biolabs) in 10-μL volumes. The mixture was incubated at 42 °C for 50 min and at 70 °C for 15 min and was cooled to 4 °C before being quenched in an equal volume of STOP buffer. The samples were analyzed on 10% urea gel. Sequencing was performed on pIA536 with end-labeled primer 2494 using the Sequenase version 2.0, DNA sequencing kit (Affymetrix USB).

Sample Analysis.

Samples were heated for 2 min at 95 °C and were separated by electrophoresis in denaturing 6–12% acrylamide (19:1) gels (7 M urea, 0.5× TBE). The gels were dried, and the products were visualized and quantified using a Typhoon FLA 9000 Phosphorimaging System (GE Healthcare), ImageQuant Software, and Microsoft Excel.

Transcription Complex Models.

The composite model of TEC–NTP–NusG was generated using the structure of the T. thermophilus TEC with the NTP analog [Protein Data Bank (PDB) ID code 2O5J; lineage-specific domain (β′166–448) omitted], the NusG NTD from the model of the T. thermophilus NusG–RNAP complex (9), the NusG CTD (G187–I248) from the crystal structure of Aquifex aeolicus NusG (PDB ID code 1M1G), and αCTDs from the crystal structure of the E. coli holoenzyme (PDB ID code 4YG2). The downstream DNA outside the TEC was extended with the canonical DNA duplex. The upstream duplex DNA was modeled de novo following the overall direction suggested by other TEC models (25, 42, 43) but avoiding clashes with the NusG NTD and maintaining a canonical B-duplex up to 11 bp and DNA base pairing up to 10 bp upstream of the RNA 3′ end as described in ref. 30. The downstream half of the ss NT-DNA was modeled to extend from the downstream DNA duplex toward NusG along a straight path separated from the protein exterior by the GL, whereas the upstream half of the ss NT-DNA was modeled to loop around the surface of NusG before rejoining the template DNA strand. The positions of the NusG CTD and αCTD were chosen arbitrarily within the volume permitted by the length of the flexible linkers tethering those domains to the TEC. Parts of the linkers were modeled de novo using ModLoop (44). The model geometry was evaluated using MolProbity (45). The atomic coordinates of the TEC–NTP–NusG complex are provided in Dataset S2.

The composite model of RPo was generated using the crystal structure of the T. thermophilus RPo with a complete bubble stabilized by an RNA primer (PDB ID code 4XLN; the RNA primer was omitted), the C terminus of the σ1.1 region from the crystal structure of the initially transcribing E. coli RNAP holoenzyme (PDB ID code 4YLP), the NTD of the σ1.1 region from the solution structure of Thermotoga maritima σA σ^A1.1 (PDB ID code 2K6X), and the initiating NTPs from the crystal structure of the T. thermophilus RPo with a partial transcription bubble (PDB ID code 4Q4Z). The TL domain was remodeled into a helical hairpin conformation observed in the structure of the T. thermophilus TEC with the NTP analog (PDB ID code 2O5J). The position of the NTD of the σ1.1 region was chosen arbitrarily within the volume permitted by the length of the flexible linker tethering it to the C-terminal α-helix of σ1.1. The σ1.1 linker was modeled de novo using ModLoop (44). The downstream and upstream DNA duplexes outside the RPo were extended with the canonical DNA duplexes. The model geometry was evaluated using MolProbity (45). The atomic coordinates are provided as Dataset S2. To generate Fig. 1, the simplified surfaces (Gaussian resolution 6, B-factor 50) were calculated and rendered in PyMOL Molecular Graphics System version 1.8.2.1 (Schrödinger), exported in VRML format, converted to OBJ format using MeshLab, and further simplified using sculpting tools of Meshmixer (Autodesk Inc.). The resulting meshes were imported into and rendered in Rhinoceros 5.0 (Robert McNeel & Associates).