Tethering σ⁷⁰ to RNA polymerase reveals high in vivo activity of σ factors and σ⁷⁰-dependent pausing at promoter-distal locations

Abstract

Bacterial σ factors compete for binding to RNA polymerase (RNAP) to control promoter selection, and in some cases interact with RNAP to regulate at least the early stages of transcript elongation. However, the effective concentration of σs in vivo, and the extent to which σ can regulate transcript elongation generally, are unknown. We report that tethering σ⁷⁰ to all RNAP molecules via genetic fusion of rpoD to rpoC (encoding σ⁷⁰ and RNAP's β′ subunit, respectively) yields viable Escherichia coli strains in which alternative σ-factor function is not impaired. β′::σ⁷⁰ RNAP transcribed DNA normally in vitro, but allowed σ⁷⁰-dependent pausing at extended -10-like sequences anywhere in a transcriptional unit. Based on measurement of the effective concentration of tethered σ⁷⁰, we conclude that the effective concentration of σ⁷⁰ in E. coli (i.e., its thermodynamic activity) is close to its bulk concentration. At this level, σ⁷⁰ would be a bona fide elongation factor able to direct transcriptional pausing even after its release from RNAP during promoter escape.

A fundamental paradigm in transcriptional regulation is the cyclical association of RNA polymerase (RNAP) with dissociable initiation, elongation, and termination factors (Burgess et al. 1969; Chamberlin 1976). In bacteria, σ factors bind tightly to free RNAP to program promoter recognition and initiation (Gill et al. 1991; Maeda et al. 2000), but are thought to release stochastically from elongation complexes (ECs) as, or shortly after, RNAP escapes promoters (Shimamoto et al. 1986). The avidity of σ-factor binding decreases dramatically in part because RNA chains >8 nt compete with σ for sites on RNAP (Krummel and Chamberlin 1989; Daube and von Hippel 1999; Murakami et al. 2002). The elongation factor NusA also may assist σ release by competitive binding to the EC (Gill et al. 1991). A similar cycle of initiation factor binding and release operates for eukaryotic RNAPs, although the details are less well understood (Kimura et al. 2002; Pokholok et al. 2002).

During development, differentiation, and changes in environment, binding, and release of initiation factors allow cells to alter patterns of gene expression by reprogramming RNAP. In bacteria, reprogramming is accomplished by switching among σ factors associated with RNAP (e.g., seven σs in Escherichia coli, 18 in Bacillus subtilis, and at least 65 in Streptomyces coelicolor; Ishihama 2000; Kunst et al. 1997; Bentley et al. 2002). Among these, one σ is the major or housekeeping σ and is typically present at the highest level (σ⁷⁰ in E. coli).

Selection among σs for RNAP binding is thought to be mediated by concentration-dependent competition among available σ molecules for available RNAP molecules (Zhou and Gross 1992; Hicks and Grossman 1996; Farewell et al. 1998; Gross et al. 1998; Kolesky et al. 1999; Ishihama 2000; Maeda et al. 2000). The concentrations of available σs are in turn highly regulated by their expression levels, rates of degradation, and sequestration via binding to inhibitory proteins called anti-σs (Ishihama 2000). Although RNAP outnumbers total σ in E. coli (∼1.7:1; Materials and Methods), σs must compete for the smaller pool of available RNAP (∼0.6 per σ), which includes newly synthesized RNAP and RNAP not engaged in transcription or sequestered nonspecifically (Ishihama 2000; see Discussion).

The idea that σs are released from RNAP during transcript elongation has been challenged recently (Bar-Nahum and Nudler 2001; Mukhopadhyay et al. 2001). In one view, permanent association of σ⁷⁰ with some RNAP is proposed to accelerate recycling of this RNAP for new rounds of transcription by circumventing the need to rebind σ⁷⁰, which could be rate-limiting in the transcription cycle (Bar-Nahum and Nudler 2001).

σ⁷⁰ plays at least one regulatory role during transcript elongation. σ⁷⁰ can remain bound to RNAP long enough to stimulate pausing at promoter-proximal sites that resemble -10 promoter elements (Ring et al. 1996). In this case, σ⁷⁰-dependent pausing was lost when the pause site was moved 20 bp downstream and tested in vitro (Ring et al. 1996). This could be explained by stochastic σ⁷⁰ release after promoter escape (Shimamoto et al. 1986) or by resistance of mature ECs to σ⁷⁰-stimulated pausing.

To assess the effective concentration of σ⁷⁰ in vivo and to gain insight into σ⁷⁰'s effect on the EC, we created an rpoC::rpoD gene fusion that tethered σ⁷⁰ to all RNAP in cells via a covalent polypeptide linkage. Tethering proteins by genetic fusion fixes the local concentration of interacting proteins (Raag and Whitlow 1995; Timpe and Peller 1995; Robinson and Sauer 1998). Depending on the length of the tether and the location of binding sites relative to the tether-attachment points, tethering can generate local protein concentrations of 10^-5 to 10 M (Robinson and Sauer 1998). The bulk concentration of σ⁷⁰ in vivo is ∼15 μM (Materials and Methods). However, a variety of factors, including macromolecular crowding, can dramatically alter the effective concentration of σ⁷⁰ in cells (i.e., its thermodynamic activity). Tethering σ⁷⁰ to RNAP makes it possible to examine the effects in vivo of known local concentrations of tethered σ⁷⁰, and thus gain insight into the effective concentration of σ⁷⁰ in cells.

Results

E. coli is viable with σ⁷⁰ fused to RNAPat the β′ C terminus

We initially tested the in vivo function of σ⁷⁰ tethered to the C terminus of β′ or β when corresponding gene fusions were conditionally expressed from plasmids (Table 1). (RNAP tolerates alterations at the ends of β′ or β; σ⁷⁰ tolerates alterations at its N terminus; Severinov et al. 1997; Sharp et al. 1999.) β::σ⁷⁰ was unable to complement for loss of σ⁷⁰ or β function. However, β′::σ⁷⁰ complemented for loss of σ⁷⁰ and, to a limited extent, β′. We considered this result promising and concentrated further study on β′::σ⁷⁰. As an initial control that β′::σ⁷⁰ was incorporated into RNAP, we replaced region 4 of the tethered σ⁷⁰ with the corresponding region of σ³² (Kumar et al. 1995). This eliminated the ability of β′ in the fusion to provide even weak β′ function (β′::σ^70/32; Table 1), suggesting that the weak complementation of rpoC^TS by β′;::σ⁷⁰ reflected incorporation into a functional RNAP. Based on these results, we next asked if the rpoC;::rpoD fusion could replace rpoC in the chromosome (Materials and Methods).

Table 1.

Plating efficiency of tethered σ⁷⁰ expressed from plasmids in wild-type and mutant strains

		rpoD- straina		rpoB- strainb	rpoC- strainb
Plasmid	Subunit	—IPTG	+IPTG	+IPTG	+IPTG
pTrc99c	None	<10-6	<10-6	<10-6d	<10-6d
pRL663	β′	—	—	—	1
pRL706	β	—	—	1	—
pCL391	σ70	1	1	—	—
pRM314	β′::σ70	<10-6	0.5—1	—	Weak, <10-6 scc
pRM328	β::σ70	<10-6	<10-6	<10-6d	—
pRM317	β′::σ70/32	<10-6	<10-6	—	<10-6d

^aStrain CAG20153 (Supplementary Table 1) in which rpoD transcription is conditionally repressed by TrpR. These strains were tested on plates lacking IAA to prevent rpoD transcription

^brpoC^- strain is RL602 (Supplementary Table 1; functionally rpoC^TS). rpoB strain is RL585 (Supplementary Table 1, functionally rpoB^TS). These strains were plated at 42°C, a nonpermissive temperature for rpoC or rpoB expression

^cWeak growth at 10⁶ cells/mL with ITPG, but not without ITPG; no single colonies (sc) at <10⁶ cells/mL. When struck on IPTG plates, this strain gave growth at 42°C in the heavy part of the streak, but not single colonies. No equivalent growth was observed in plates lacking IPTG

^dNo growth at any cell density with or without IPTG

The resulting strain, in which σ⁷⁰ was tethered to all RNAPs, proved viable. We transduced the rpoC;::rpoD allele into a strain in which rpoD expression can be shut off (Lonetto et al. 1998). This strain exhibited no defect in either growth or recovery from stationary phase when forced to rely on β′;::σ⁷⁰ for both β′ and σ⁷⁰ function (Fig. 1A,B). The simplest interpretation of these results is that β′;::σ⁷⁰ RNAP is a fully functional enzyme and that the weak complementation of rpoC^TS by the plasmid-encoded β′;::σ⁷⁰ reflected partial overexpression toxicity (Table 1).

Figure 1.

Strains and growth phenotypes. (A) Early log-phase cultures of strains with the rpoD and rpoC loci illustrated were serially diluted and plated onto LB plates or LB plates containing indole acrylic acid (IAA). Strains (top to bottom) are RL301, RL1374, CAG20153, and RL1094 (Supplementary Table 1). The efficiencies of plating minus IAA versus plus IAA were ∼10^-5 for the wild-type strain (CAG20153) and ∼1 for the β′;::σ⁷⁰ strain (RL1094). (B) Wild-type or β′;::σ⁷⁰ with TrpR-repressed rpoD strains (C600K^- and RL1094) were monitored during growth in LB at 37°C using a Klett colorimeter (data are averages for two independent cultures). After residing in stationary phase for ∼12 h, the strains were diluted back to Klett <5. (C) Model of σ⁷⁰ tethering to β′ based on the Thermus thermophilis holoenzyme structure (Vassylyev et al. 2002; β′ NCD removed). The upstream and downstream faces of RNAP are shown above schematics of the E. coli β′ and σ⁷⁰ subunits. RNAP subunits are shown in spacefill; σ, is shown as a C_α-trace colored by regions shown in the schematic. The N-terminal and C-terminal residues of σ and β′ resolved in the structure are depicted as black spheres (corresponding to β′1389 and σ⁷⁰91 in E. coli). σ region 1.1 (semitransparent white circle) is positioned based on the hydrated volume of an 82-amino acid globular protein attached to σ⁷⁰91.

The structure of bacterial RNAP holoenzyme (Vassylyev et al. 2002) is consistent with this interpretation (Fig. 1C). Although σ region 1.1 is not resolved in the structure, if spherical, it would occupy a hydrated volume ∼30 Å in diameter on the downstream face of RNAP. At least 27 amino acids may flexibly connect region 1.1 to β′: three amino acids from the gene fusion, plus 19 C-terminal amino acids of E. coli β′ that correspond to the disordered C-terminal segment in the RNAP structure, plus the last five visible amino acids of β′, which form a random coil. These 27 amino acids (∼100 Å fully extended) are sufficient to span the 80 Å between the last resolved amino acids in β′ and the likely positions of region 1.1, proposed to be within the DNA-entry channel in the holoenzyme and at the outside edge of RNAP's lobe domain in initiation complexes (Mekler et al. 2002).

To verify that σ⁷⁰ remained tethered to RNAP in the viable β′;::σ⁷⁰ strain, we examined the subunits present in cell extracts by immunoblotting with anti-β′ and anti-σ⁷⁰ antibodies (see Materials and Methods). Only β′;::σ⁷⁰, not β′ or σ⁷⁰, was present in a whole-cell extract from the β′;::σ⁷⁰ strain when rpoD expression was shut off (Fig. 2A, cf. lanes 1 and 2). Therefore, β′;::σ⁷⁰ was not cleaved to separate β′ and σ⁷⁰ subunits in vivo, and this intact β′;::σ⁷⁰ polypeptide could serve as the sole source of both β′ and σ⁷⁰ in viable E. coli cells.

Figure 2.

Purified RNAP and immunoblot analysis. (A) (Left) Cellular extracts of wild-type (wt) or β′;::σ⁷⁰ (F) strains (RL301 and RL1094) were separated by 3%-8% Tris-Acetate (Novex), transferred to nitrocellulose, and probed with a mixture of β′ and σ⁷⁰ antibodies (Materials and Methods). Equal amounts of cellular protein were present in each lane. (Center) Western blot of wild-type (wt) and β′;::σ⁷⁰ (F) RNAPs. Samples were separated by 3%-8% Tris Acetate (Novex) and probed as in the left panel. (Right) β′;::σ⁷⁰ RNAP (F) purified from the strain containing the rpoC;::rpoD fusion and Trp-repressed rpoD (RL1094) separated by 4%-15% PAGE gel (Pharmacia) shown next to wild-type (wt) RNAP for size comparison. (B) β′;::σ⁷⁰ (F) or wild-type (wt) RNAPs were mixed with λP_R template, allowing the formation of open complexes at 37°C for 20 min (heparin was present where indicated for the last 10 min), separated by native gel electrophoresis (4% NuSieve agarose, 0.5× TBE), and stained with EtBr. The absence of heparin masked the slightly faster mobility of β′;::σ⁷⁰ OCs (see legend to Fig. 4C) because β′;::σ⁷⁰ OCs appeared to aggregate more than wild-type OCs.

A possible explanation for the viability of the β′;::σ⁷⁰ strain would be that one β′;::σ⁷⁰ polypeptide provides β′ and a second polypeptide provides σ⁷⁰ in a single RNAP holoenzyme. If this were true, then purified, active β′;::σ⁷⁰ RNAP should contain two β′;::σ⁷⁰ polypeptides. However, highly purified, fully active β′;::σ⁷⁰ RNAP gave a stoichiometry of one β′;::σ⁷⁰ to one β and contained no β′ or σ⁷⁰ (Fig. 2A, lanes 3,5; densitometry not shown). Another possibility would be that σ⁷⁰ from one β′;::σ⁷⁰ RNAP provided σ⁷⁰ function to a second β′;::σ⁷⁰ RNAP. If this were true, then open complexes (OCs) formed by β′;::σ⁷⁰ RNAP would be of significantly greater mass (920 kD vs. 460 kD, not including DNA), and would exhibit significantly slower electrophoretic mobility. To test this, we formed OCs with wild-type and β′;::σ⁷⁰ RNAPs on a λP_R-containing DNA fragment and compared them by native gel electrophoresis (Fig. 2B). The β′;::σ⁷⁰ OCs exhibited similar mobility to wild type (see legend to Fig. 4C, below). We conclude that β′;::σ⁷⁰ RNAP and its OCs contain a single β′;::σ⁷⁰ polypeptide.

Figure 4.

Effective local concentration of tethered σ⁷⁰ measured by competitive binding. (A) Equilibrium σ⁷⁰-binding assay (see Materials and Methods). Samples are (left to right) 0.5, 1, 2, and 5 μM [³²P]σ⁷⁰ with 1 μM wild-type RNAP, β′;::σ⁷⁰ RNAP, or no RNAP. The positions of holoenzyme (E σ⁷⁰) and free σ⁷⁰ are indicated; slower σ⁷⁰ bands are σ⁷⁰ dimers. (B) Quantitation of A. The fraction of RNAP binding ³²P-labeled σ⁷⁰ is plotted against the amount of ³²P-labeled σ⁷⁰ present in the reactions. (•) Wild-type RNAP; (○) β′::σ⁷⁰ RNAP. Effective concentrations of σ⁷⁰ were estimated by nonlinear regression (Materials and Methods) from the averages of three experiments. (C) OC mobility assay. β′::σ⁷⁰ or wild-type RNAPs were incubated with the indicated amounts of σ⁷⁰ and then with added promoter DNA (Materials and Methods). OCs and DNA were separated on a native agarose gel and stained with EtBr. The positions of the free DNA, the wild-type OC, the β′::σ⁷⁰ OC, and the supershifted β′::σ⁷⁰ OC formed with a second, untethered σ⁷⁰ are indicated. Heparin was omitted in this experiment to minimize smearing of the supershifted OC band, although this caused variable nonspecific binding of free DNA (evident as aggregates near the top of the gel). Addition of heparin gave constant amounts of free DNA in gels and approximately the same amounts of supershifted OC, allowing use of up to 30 μM σ⁷⁰ (Fig. 4D). The slightly increased mobility of β′::σ⁷⁰ OCs (see also Fig. 2B) and of β′::σ⁷⁰ RNAP (panel A) may reflect a slight structural change caused by tethering that is under study. (β′::σ⁷⁰ RNAP does, however, contain ω.) (D) Quantitation of OC supershifting as a function of σ⁷⁰ concentration. The fraction of RNAP binding a second σ⁷⁰ (the supershifted species in C) is plotted against the concentration of added, untethered σ⁷⁰ (0.1, 0.5, 1, 10, and 30 μM; average of three experiments that included heparin). Predicted binding curves (assuming all σ⁷⁰ species to be equivalent for binding to RNAP) for different effective local concentrations of the tethered σ⁷⁰ are shown for comparison (Materials and Methods).

The results described to this point confirm that RNAP tethered to σ⁷⁰ supports cell growth and functions with the σ⁷⁰ to which it is tethered. This suggests that β′::σ⁷⁰ RNAP can use at least σ³² and σ^E, which are required for viability of E. coli at 37°C (Zhou et al. 1988; De Las Penas et al. 1997), and σ^S, which is necessary for wild-type recovery from the stationary phase (Ishihama 2000). We next examined use of alternative σs more carefully and tested for function of NusA, which also competes with σ⁷⁰ for interaction with RNAP (Gill et al. 1991). In the presence or absence of chromosomally encoded, untethered σ⁷⁰, β′::σ⁷⁰ strains plated a λ phage that requires function of NusA (and other Nus proteins; Friedman et al. 1976) equivalently to wild type (Table 2). The same was true for survival at 45°C (an even more stringent requirement for σ³² and σ^E), for growth on medium requiring σ^N function, for σ^F-dependent swarming motility, and for σ^S-dependent formation of peroxidase in the stationary phase (Table 2). Thus, alternative σs and NusA appear to exhibit high in vivo activity, sufficient to allow normal function in β′::σ⁷⁰ strains despite the presence of σ⁷⁰ tethered to all RNAP molecules. This remains true even in β′::σ⁷⁰ strains also containing chromosomally encoded, untethered σ⁷⁰.

Table 2.

Properties of wild-type and β′::σ⁷⁰ fusion strains in various conditions

Phenotype	Required factor	Wild-type σ70 levela	β′::σ70, no add. σ70a	β′::σ70, plus σ70a
λ growthb	NusA, B, E, G	1	0.8 ± 0.2	1.0 ± 0.4
Growth at 45°Cc	σ32, σE	1.0 ± 0.2	1.0 ± 0.1	1.0 ± 0.1
Growth on Arg/Glnd	σN	1.2 ± 0.3	1.3 ± 0.1	1.0 ± 0.4
Motilitye	σF	1	1.0 ± 0.1	1.0 ± 0.1
Catalase productionf	σS	+	+	+

^aThe strains for the wild-type σ⁷⁰ level were RL301 or C600 K^-. The β′::σ⁷⁰, no additional σ⁷⁰ strain was RL1094 (chromosomally encoded σ⁷⁰ not expressed). The β′::σ⁷⁰, plus σ⁷⁰ strains were RL1374 or RL1390 (both contain additional, chromosomally encoded σ⁷⁰). All measurements are the averages of at least three determinations. The methods for determining phenotypes and plating efficiencies are in the Supplemental Material

^bThe number of plaques formed by λcIc17 (requires E. coli nus functions; Friedman et al. 1976) on β′::σ⁷⁰ strains divided by the number of plaques formed on a wild-type strain

^cEOP on rich medium of β′::σ⁷⁰ strains at 45°C divided by EOP at 37°C

^dEOP on minimal medium containing Arg (requires σ^N-directed transcription of glnA) divided by EOP on Gln-containing medium (allows σ^N-independent growth)

^eThe swarm diameter of β′::σ⁷⁰ strains on motility agar plates divided by the swarm diameter of a wild-type strain. Flagella synthesis requires σ^F function

^fObservation of O₂ evolution upon addition of 3% H₂O₂ to stationary-phase cultures; requires σ^S-dependent expression of catalase from kat

β′::σ⁷⁰ RNAPdisplays wild-type enzymatic properties and activities

We next sought to determine if β′::σ⁷⁰ RNAP also behaved normally in vitro. For this purpose, we tested transcription by wild-type and β′::σ⁷⁰ RNAPs of a linear DNA template encoding the well-characterized his pause site downstream of the T7 phage A1 promoter (Chan and Landick 1989). Upon initiation under conditions that allowed RNAP to transcribe only to A29, no difference was apparent between wild-type and β′::σ⁷⁰ RNAPs in the rate of abortive initiation, the rate of productive initiation, or the ratio of abortive to productive products (as reflected by the AUC abortive RNA and A29 productive RNA; Fig. 3A). When transcription past A29 was allowed after 2 min, both RNAPs escaped the A29 position efficiently and exhibited indistinguishable kinetics of pausing and subsequent transcription to the end of the template. We also tested the ability of NusA protein to increase the duration of pausing at the his pause site and the efficiency of termination by β′::σ⁷⁰ RNAP at several intrinsic terminators; no significant differences from wild-type RNAP were observed (data not shown).

Figure 3.

(A) In vitro transcription assay. β′::σ⁷⁰ or wild-type RNAPs (40 nM) were allowed to initiate transcription on a linear DNA template (25 nM) with ApU, ATP, [³²P]CTP, and GTP. Samples were mixed with 2× STOP buffer at 10, 20, 30, 45, 60, 75, 90, and 120 sec. UTP was then added to allow transcription past A29, and additional samples were taken 10, 20, 30, 45, 60, 120, 240, and 480 sec later (final sample 10 min after reaction started). The samples were separated by 15% denaturing PAGE. (AUC) Trinucleotide abortive product; (A29) A29 EC; (P) his pause RNA; (RO) run-off RNA. (B) Kinetics of promoter association, following the mechanism defined by Saecker et al. (2002). k_f,obs and k_r,obs were obtained for β′::σ⁷⁰ and wild-type RNAPs on the template shown in panel A (≥3 independent experiments; Materials and Methods).

To test whether σ⁷⁰ tethering affects steps on the pathway of OC formation, we measured the overall rates of OC formation at the T7 A1 promoter (k_f,obs) and of preformed OC dissociation (k_r,obs; Fig. 3B; Materials and Methods). Both rates are composites of multiple steps (McClure 1980; Saecker et al. 2002) and would reveal any effects of the tether. Neither k_f,obs nor k_r,obs differed significantly between wild-type and β′::σ⁷⁰ RNAPs (Fig. 3B), consistent with wild-type function of the tethered σ⁷⁰.

Tethering σ⁷⁰ to RNAPincreases σ⁷⁰ local concentration to ∼55 μM

The lack of effect of tethering σ⁷⁰ to RNAP caused us to consider whether tethering actually increased the local concentration of σ⁷⁰ around RNAP. Such increased local concentrations have been measured for other cases of protein tethering (Timpe and Peller 1995; Robinson and Sauer 1998), and are well-grounded in polymer-chain theory (Flory 1969), but must be determined for each case.

To measure the local concentration of tethered σ⁷⁰, we performed competition binding assays using a ³²P-labeled derivative of σ⁷⁰ (Materials and Methods). We compared the ability of [³²P]σ⁷⁰ to displace untethered σ⁷⁰ from wild-type RNAP and to displace tethered σ⁷⁰ from β′::σ⁷⁰ RNAP using conditions in which σ⁷⁰ binding to RNAP equilibrates (Sharp et al. 1999). To detect [³²P]σ⁷⁰ bound to RNAP, we separated the reactions by nondenaturing PAGE and visualized RNAP and σ⁷⁰ by protein staining and [³²P]σ⁷⁰ using a PhosphorImager (Fig. 4A). [³²P]σ⁷⁰ displaced untethered σ⁷⁰ from the wild-type holoenzyme as predicted for equivalent K_ds of the prebound σ⁷⁰ and [³²P]σ⁷⁰. However, only modest [³²P]σ⁷⁰ binding to β′::σ⁷⁰ RNAP could be detected even at high concentrations of [³²P]σ⁷⁰ (Fig. 4A,B). Assuming that all ³²P at or above the position of β′::σ⁷⁰ RNAP in the gels arose from [³²P]σ⁷⁰ binding to β′::σ⁷⁰ RNAP (because binding of a second σ⁷⁰ would slow migration), we calculated a local concentration of tethered σ⁷⁰ of ∼55 μM (Fig. 4B; Materials and Methods). This is >50-fold higher than the concentration of RNAP in the assay (1 μM), but significantly less than measured for optimal cases of tethering (Robinson and Sauer 1998). The extensive topography of σ⁷⁰-RNAP contacts relative to the tether attachment points on σ⁷⁰ and RNAP, interference of the tether with some contacts, or the tether length may limit tether enhancement of local concentration.

This experiment unambiguously demonstrated weaker binding of free σ⁷⁰ to RNAP in the presence of tethered σ⁷⁰. The local concentration estimate for tethered σ⁷⁰ of 55 μM should be a lower limit because it included ³²P that was nonspecifically retarded. However, our estimate conceivably could be inflated if the complex of [³²P]σ⁷⁰ and β′::σ⁷⁰ RNAP dissociated during electrophoresis. This seemed unlikely because release of [³²P]σ⁷⁰ during electrophoresis would form a smear between the positions of holoenzyme and free σ⁷⁰. No such smear was evident; rather, the amount of [³²P]σ⁷⁰ visible at the position of free σ⁷⁰ was the same in the β′::σ⁷⁰ lanes and the σ⁷⁰ alone lanes, but was reduced in the wild-type RNAP lanes by the amount bound to RNAP (Fig. 4A, cf. the PhosphorImager densities at the bottom of the 1 μM and 2.5 μM lanes).

To confirm that tethered σ⁷⁰ produced a local σ⁷⁰ concentration of ∼55 μM, we sought to detect the binding of untethered σ⁷⁰ to β′::σ⁷⁰ RNAP in a manner that would prevent its release during electrophoresis. We reasoned that if the untethered σ⁷⁰ were engaged in OC formation by β′::σ⁷⁰ RNAP, two σ⁷⁰s would be bound to RNAP because untethered σ⁷⁰ would be trapped in the network of RNAP-σ⁷⁰-DNA interactions in the OC (Murakami et al. 2002), and tethered σ⁷⁰ would remain covalently attached to RNAP. To favor trapping untethered σ⁷⁰ bound to β′::σ⁷⁰ RNAP, we used a promoter on which OCs were stable for many hours (P_UPFullcon; Materials and Methods). We equilibrated RNAPs with increasing concentrations of free σ⁷⁰, incubated them with P_UPFullcon

DNA for 10 min, and then separated the reactions on a nondenaturing agarose gel. After visualizing the locations of DNA by ethidium staining, OCs were readily apparent as retarded bands for both wild-type and β′::σ⁷⁰ RNAPs (Fig. 4C). A second, more slowly migrating species appeared when 5-10 μM σ⁷⁰ was added to β′::σ⁷⁰ RNAP (Fig. 4C, thick arrow), but was not present with wild-type RNAP. Nonspecific binding and smearing prevented us from using concentrations of σ⁷⁰ above 30 μM. Because the supershifted species was also not observed with DNA alone or DNA plus 5-10 μM σ⁷⁰ (data not shown), we attribute it to β′::σ⁷⁰ RNAP complexed with a second, untethered σ⁷⁰ that is engaged in promoter contacts. To estimate the local concentration of the tethered σ⁷⁰, we plotted the amounts of the supershifted species versus concentrations of added σ⁷⁰ and compared the data with predicted curves (Materials and Methods; Fig. 4D). They matched reasonably to the prediction for ∼50 μM local concentration of tethered σ⁷⁰ and were more than the prediction for 100 μM. Thus, both assays give an estimate of tethered σ⁷⁰ local concentration consistent with ∼55 μM.

Tethering σ⁷⁰ to RNAPmodestly increases σ⁷⁰ local concentration in vivo

To compare the local concentration of tethered σ⁷⁰ with the effective concentration of wild-type σ⁷⁰ in vivo, we tested the ability of chromosomally encoded β′::σ⁷⁰ or increased amounts of plasmid-encoded, untethered σ⁷⁰ to protect wild-type cells (also containing chromosomally encoded σ⁷⁰) against the toxic effects of a plasmid-encoded mutant σ⁷⁰ (RC584). σ⁷⁰RC584 binds RNAP and interferes with promoter recognition (Siegele et al. 1989). The presence of tethered σ⁷⁰ significantly increased resistance to σ⁷⁰RC584 (Fig. 5, cf. closed and open circles; wild-type σ⁷⁰ is not toxic when expressed at comparable levels, dashed line). Expression of untethered σ⁷⁰ from a low-copy-number, compatible plasmid to about fivefold above the normal level conferred a level of σ⁷⁰RC584 resistance similar to that conferred by tethered σ⁷⁰ (Fig. 5, triangles).

Figure 5.

β′::σ⁷⁰ inhibits toxicity caused by σ⁷⁰RC584. RL113 (wild type, •), RL113 + pBADσ⁷⁰ (wt + ≥5× σ⁷⁰; ▴), or RL1366 (wt + β′::σ⁷⁰; ○) carrying pRM389 (σ⁷⁰RC584 expressed under LacI control) were plated with or without IPTG as described in Materials and Methods (both strains were deleted for lacY). Data are mean values of three to six independent experiments. The dotted line shows the approximate effect of wild-type σ⁷⁰ expressed from pCL391 for comparison.

These results confirm that β′::σ⁷⁰ RNAP generates a local σ⁷⁰ concentration higher than that of untethered σ⁷⁰ in wild-type cells. A straightforward interpretation would place the effective concentration of wild-type σ⁷⁰ at ∼11 μM because approximately fivefold overexpression gave σ⁷⁰RC584 resistance similar to ∼55 μM local concentration of β′::σ⁷⁰. However, overexpressed wild-type and RC584 σ⁷⁰s could affect each other's levels in ways not measurable in plating experiments. Furthermore, we did not test whether a lower level of σ⁷⁰ overexpression also could protect against RC584 σ⁷⁰ as β′::σ⁷⁰. Thus, we conclude wild-type cells contain σ⁷⁰ at effective concentrations ≥11 μM and significantly less than 55 μM.

Tethering σ⁷⁰ to RNAPonly modestly affects heat shock and σ³² function

We next wanted to examine the effect of tethered σ⁷⁰ on function of the alternate σ factor σ³², which is required for the well-defined heat-shock response in E. coli (Straus et al. 1987). When σ⁷⁰ is overexpressed, σ³² function appears to be compromised by competition for binding to core RNAP (Zhou et al. 1992). However, σ³² is up-regulated in response, resulting in little if any inhibition of synthesis of heat-shock proteins or delay in the heat-shock response. We first confirmed that β′::σ⁷⁰ RNAP could use σ³² in vitro (transcription of groE by β′::σ⁷⁰ RNAP depended on added σ³²; data not shown). We next tested whether β′::σ⁷⁰ RNAP would affect expression of heat-shock proteins and σ³² levels similarly to overexpression of σ⁷⁰.

The heat-shock response in β′::σ⁷⁰ cells was slightly delayed relative to wild-type cells (Fig. 6A). The GroE synthesis rate was slower in β′::σ⁷⁰ cells 1 or 2 min after shifting cells to 42°C, but reached the same level as wild-type by 11 min post-heat shock (Fig. 6A, plot). The characteristic suppression of σ⁷⁰-dependent transcription was similar in β′::σ⁷⁰ and wild-type cells (Fig. 6A, cf. ^* bands).

Figure 6.

Heat-shock response. (A) Cells were grown in minimal media at 30°C and then shifted to 42°C. At the times indicated, samples were incubated with [³⁵S]methionine for 2 min. Samples were separated by 4%-12% PAGE. The positions of DnaK, GroE, and an 89-kD protein likely to be Lon are indicated. (^*) Major proteins whose expression decreases on heat shock; the top band is β′::σ⁷⁰, and the next-to-top band is β, β′. Relative GroE levels are averages of four experiments. (•) Wild-type (RL301); (○) β′::σ⁷⁰ strain with chromosomally encoded σ⁷⁰ (RL1374). (B) Immunoblot analysis using antibodies against σ³² (Materials and Methods). (Lanes 1-3) Wild-type whole-cell lysate (MC1060). (Lanes 4-6) β′::σ⁷⁰ whole-cell lysate (RL1454). (Lanes 7-9) His₆-tagged σ³². Equal amounts of total cellular protein were electrophoresed.

To ask if heat shock was near normal because σ³² levels were up-regulated in the β′::σ⁷⁰ strain, we measured σ³² levels by immunoblotting. The σ³² level in β′::σ⁷⁰ cells was ∼80% of that in wild-type cells, a statistically insignificant difference (Fig. 6B; Materials and Methods). The slight delay in GroE expression in heat-shocked β′::σ⁷⁰ cells is consistent with a local concentration of tethered σ⁷⁰ higher than the wild-type σ⁷⁰ effective concentration, but the nonelevated σ³² levels in β′::σ⁷⁰ cells and the inhibition of σ⁷⁰-directed transcription upon heat shock even when σ⁷⁰ was tethered to RNAP were surprising (see Discussion).

Tethered σ⁷⁰ causes σ⁷⁰-dependent pausing independent of distance from the promoter

An additional consequence of tethering σ⁷⁰ is that its high local concentration will be maintained during transcript elongation. Because untethered σ⁷⁰ is able to stimulate pausing at -10-like sequences near a promoter (Ring et al. 1996), we wondered if tethering would allow σ⁷⁰ to stimulate pausing at -10-like sequences generally. To test this possibility, we examined σ⁷⁰-dependent pausing on templates originally studied by Roberts and colleagues (+16/17 and +37 pauses), and on an additional template encoding a consensus extended -10 sequence near +450 of an artificial transcription unit (+462 pause). β′::σ⁷⁰ RNAP gave slightly increased pausing at the +16/17 site (ctcaAcgAT, -10-like σ⁷⁰-binding sequence; Fig. 7A). This strong pausing also occurred at the +37 site (TGcTATAAT, consensus extended -10-like σ⁷⁰-binding sequence; Fig. 7B), where wild-type RNAP pauses weakly if at all. This result confirms that wild-type RNAP fails to pause at +37 because its σ⁷⁰ is released, rather than because the structure of the EC past +25 precludes pausing (Ring et al. 1996). Tethered σ⁷⁰, in contrast, stimulates pausing because it remains attached to RNAP in a functional state after completion of the transition to an EC.

Figure 7.

The effective concentration of tethered σ⁷⁰ is sufficient to cause promoter-distal, σ⁷⁰-dependent pausing. The effective concentration of tethered σ⁷⁰ is sufficient to cause promoter-distal, σ⁷⁰-dependent pausing. Synchronous in vitro transcription with β′::σ⁷⁰ or wild-type RNAPs (Materials and Methods; time of transcript elongation as indicated in min; C, incubation with additional 200 μM NTP for 8 min after the last indicated time point). (A) +16/17 pause template (Ring et al. 1996). The top panel shows the run-off (RO) product; the lower panel shows the pause. The schematic of the template indicates the σ⁷⁰-binding sequence and the position of the pause. (B) +37 pause template (+20 in Ring et al. 1996). Gel panels and schematic are as in A. The three pause positions on this template have not been mapped precisely and therefore are designated 37a, 37b, and 37c; the third pause band (relative to the +16/17 site) is probably caused by the stronger consensus σ⁷⁰-binding sequence. The time dependence of pause RNA levels (plot) is the average of three independent experiments. (C) +462 pause template. Gel panels and schematic are as in A. σ (1 μM) or NusA (10 μM) were added as indicated.

To ask if the tethered σ⁷⁰ could act at even greater distances from a promoter, we tested the +462 pause template (TGcTATAAT σ⁷⁰-binding sequence; Fig. 7C). As expected, wild-type RNAP was unable to recognize the +462 pause; however, β′::σ⁷⁰ RNAP exhibited strong pausing. If pausing at +462 occurred simply because the tethered σ⁷⁰ was present at high local concentration, then elevated concentrations of untethered σ⁷⁰ should cause wild-type RNAP to pause. Consistent with this interpretation, wild-type RNAP recognized the +462 pause when additional σ⁷⁰ (1 μM) was added to ECs (Fig. 7C). We concluded that this pause depended on the consensus extended -10-like sequence because a mutant TGcTgTAAg site nearly eliminated pausing (data not shown). This effect of nonconsensus substitutions may explain why Marr et al. (2001) found that additional σ⁷⁰ (0.8 μM) did not direct pausing at the nonconsensus +16/17 pause. Pause escape by β′::σ⁷⁰ RNAP was barely detectable at consensus extended -10-like sequences (e.g., plot in Fig. 7B). However, β′::σ⁷⁰ RNAP was paused, not terminated, because addition of GreA or GreB protein reduced pausing dramatically (data not shown), as shown for the +16/17 pause by Marr and Roberts (2000).

Because NusA is thought to displace σ⁷⁰ from the EC (Gill et al. 1991), we next asked how NusA affected recognition of the +462 pause. Addition of NusA to 10 μM slowed elongation overall, generated a new pause at 419 nt, and reduced, but did not eliminate, σ⁷⁰-dependent pausing caused by untethered σ⁷⁰ (Fig. 7C, wt RNAP + σ + NusA). However, NusA had little effect on +462 pausing by β′::σ⁷⁰ RNAP (Fig. 7C) or by wild-type RNAP when untethered σ⁷⁰ was added to 10 μM (1:1 stoichiometry with NusA; data not shown). These results establish that tethered σ⁷⁰, and even untethered σ⁷⁰ if added at higher concentration (but still well below that predicted to occur in vivo), can direct σ⁷⁰-dependent pausing irrespective of pause site location in a transcriptional unit. The findings are consistent with the paradigm that NusA competes with σ⁷⁰ for interaction with an EC (Gill et al. 1991), but suggest that NusA may not eliminate σ⁷⁰-dependent pausing in vivo (see Discussion).

Discussion

Our study of tethering σ⁷⁰ to RNAP leads to four main conclusions. First, a local σ⁷⁰ concentration of ∼55 μM (the measured value for β′::σ⁷⁰ RNAP) does not significantly perturb the physiology of E. coli, consistent with the effective concentration of untethered σ⁷⁰ in wild-type cells being only slightly lower. Second, tethered σ⁷⁰ at this modestly increased local concentration only slightly delays heat-shock gene expression and is still inactivated upon heat shock (like untethered σ⁷⁰), even though σ³² levels are not elevated. Third, tethered σ⁷⁰ or even 1 μM untethered σ⁷⁰ can stimulate pausing in vitro anywhere in a transcriptional unit. Fourth, the effective concentration of free σ⁷⁰ present in cells appears sufficient to allow it to interact transiently with an EC even after σ⁷⁰ release, and to stimulate pausing at extended -10-like sequences in vivo.

Effective concentration of σ⁷⁰ in vivo

An accurate understanding of σ-factor binding by RNAP requires knowing the effective concentrations of σs and RNAP in vivo. However, many factors can cause the effective concentrations of molecules in cells to differ from their bulk concentrations. First, the cytoplasm of E. coli is a highly concentrated mixture of macromolecules, small molecules, and water more akin to a gel than to a dilute solution. Macromolecular crowding will increase the effective concentrations of σ⁷⁰ and RNAP significantly, quite likely by a factor of 10 or more (Record et al. 1998; Ellis 2001). The magnitude of crowding effects may depend on growth conditions, as the water content of cells changes significantly in response to osmolytes in the growth medium (Cayley et al. 1991).

Second, the effective concentrations of σ⁷⁰ and RNAP will be reduced by interactions that sequester them from participation in the competitive binding equilibria among σ factors and RNAP. ECs sequester about two-thirds of RNAP (Ishihama 2000). Additional RNAP binds nonspecifically to DNA and possibly to RNA; these contributions depend on the binding constants, target sizes, and amounts of DNA or RNA available for nonspecific interaction. Most σs in E. coli also specifically bind anti-σs (FlgM for σ^F, RseA for σ^E, DnaK for σ³², and possibly Rsd for σ⁷⁰; Ishihama 2000 and references therein). 6S RNA specifically sequesters holoenzyme (Wassarman and Storz 2000). Some free σ or RNAP-bound σ could be sequestered nonspecifically. Promoter complexes also will sequester some σs, but it is difficult to estimate how many because some σ may release slowly after initiation (Shimamoto et al. 1986; Bar-Nahum and Nudler 2001; Mukhopadhyay et al. 2001) and because some OCs appear unable to initiate transcription (Susa et al. 2002). The extent to which these various interactions reduce σ⁷⁰ and RNAP availability depends on their avidity, but together they likely reduce substantially the effective concentrations of σ⁷⁰ and RNAP molecules.

Finally, the association of σs with RNAP is unlikely to be in equilibrium in vivo. RNAP constantly enters the pool available for σ binding both by release from DNA at terminators and by new synthesis. σs bind RNAP tightly, which means that the off-rates may be slower than the time it takes RNAP to bind and initiate at a promoter.

Given this complexity, the prospects for calculating the effective concentration of σ⁷⁰ in vivo are poor. However, the effects of tethering σ⁷⁰ to RNAP at known local concentration provides some insight. Both enhanced competition against σ⁷⁰RC584 (Fig. 5) and the slight delay in the heat-shock response (Fig. 6) of tethered σ⁷⁰ suggest that the normal effective concentration of untethered, wild-type σ⁷⁰ is modestly less than the local concentration of tethered σ⁷⁰ (55 μM). Comparison to the effects of overproducing σ⁷⁰ suggest it is ≥11 μM, close to σ⁷⁰'s bulk concentration (∼15 μM). This is reminiscent of a similar conclusion reached for a cellular protein ordinarily present at very different bulk concentration, lac repressor (Law et al. 1993). For lac repressor, the balance of crowding and sequestration effects also yields an effective concentration near its bulk concentration (∼1 nM in wild-type cells).

Competition of σ⁷⁰ and σ³²

Despite the modest increase in the effective concentration of σ⁷⁰ caused by tethering, σ³² levels are not elevated in the β′::σ⁷⁰ strain as expected for increased σ⁷⁰ concentration (Zhou and Gross 1992), and the tethered σ⁷⁰ is still inactivated upon heat shock. We cannot absolutely exclude the possibility that a small fraction of proteolytically cleaved β′::σ⁷⁰ RNAP allows the near normal σ³² function. However, we favor σ³² function with intact β′::σ⁷⁰ RNAP for three reasons. First, we did not detect cleavage of β′::σ⁷⁰ RNAP after heat shock (data not shown). Second, even if some cleavage occurred, σ³² must compete in these strains against additional, chromosomally encoded σ⁷⁰ that also could bind any available core RNAP. Third, σ³² functions with uncleaved β′::σ⁷⁰ RNAP in vitro and this cannot be explained by proteolytic fragmentation of β′::σ⁷⁰. Rather, we suggest the explanation for the lack of σ³² overexpression when σ⁷⁰ is tethered to RNAP lies in the difference between increasing σ⁷⁰ concentration locally versus globally. Overexpression of σ⁷⁰ produces inclusion bodies and becomes toxic to E. coli at high levels of σ⁷⁰ (data not shown; see Fig. 5). Aggregates of σ⁷⁰ (the precursor to inclusion bodies) may be bound to the chaperone DnaK and thus could induce E. coli's stress response and elevate the σ³² level by releasing σ³² from DnaK. Increasing σ⁷⁰ concentration locally for RNAP, as occurs in the β′::σ⁷⁰ strain, would not provoke this stress response because it would not elevate the global σ⁷⁰ level. In essence, competition of σ⁷⁰ and σ³² may occur for both RNAP and DnaK upon global σ⁷⁰ overexpression, but should be limited to RNAP for σ⁷⁰ tethering.

The inactivation of tethered σ⁷⁰ upon heat shock suggests something other than simple competition for σ³² shuts off σ⁷⁰-dependent gene expression. Indeed, even in wild-type cells, something in addition to an increase in σ³² level seems necessary to explain the decrease in σ⁷⁰-directed gene expression upon heat shock (Fig. 6A; Straus et al. 1987). Although σ³² levels are elevated fivefold 15 min after shift of E. coli to 42°C, σ⁷⁰ levels themselves increase 2.5-fold (Taylor et al. 1984; Straus et al. 1987). This is equivalent to ∼60 σ³² and ∼900 σ⁷⁰ per 1000 RNAP, suggesting that an unknown factor actively inhibits σ⁷⁰ during heat shock. Whatever the mechanism of σ⁷⁰ inhibition, it must also act on tethered σ⁷⁰ and allow σ³² to dominate during heat shock despite its lower level. Such a requirement for additional factors has been a general conclusion in most studies of σ-factor switching (Fujita and Sadaie 1998; Kolesky et al. 1999; Lord et al. 1999; Maeda et al. 2000; Jishage et al. 2002; Rollenhagen et al. 2003). The fact that this mechanism operates on tethered σ⁷⁰ rules out explanations involving physical segregation of σ⁷⁰ away from sites of RNAP function because σ⁷⁰ is fixed to RNAP by covalent linkage.

σ⁷⁰ may function as an elongation factor in vivo

The textbook view of σ⁷⁰ participation in the transcription cycle may need revision, although our results suggest that when or whether σ is released is not the relevant question. Even 1 μM σ⁷⁰ can interact from solution with ECs in vitro and cause pausing anywhere in a transcriptional unit; this behavior is similar to that of σ⁷⁰ permanently tethered to RNAP. Because σ⁷⁰'s effective concentration in cells is significantly higher than 1 μM, continuous transient interactions must occur with ECs regardless of when or whether it is released after promoter escape (Fig. 8).

Figure 8.

Model of σ⁷⁰-RNAP interactions during transcription in E. coli.

Although NusA competes with σ⁷⁰ for binding to the EC (Gill et al. 1991) and may temper the effects of σ⁷⁰-stimulated pausing, it is unlikely to eliminate this pausing in vivo. In vitro, a 10-fold excess of NusA reduced, but did not eliminate, σ⁷⁰-stimulated pausing (Fig. 7C). Although the in vivo concentration of NusA has not been reported, a 10-fold excess over σ⁷⁰ would correspond to >110 μM or more than 2.6% of total cellular protein. Thus, it is unlikely there is enough NusA in cells to eliminate σ⁷⁰-dependent pausing at extended -10-like sequences.

The view that emerges from this study is that the effective concentration of σ⁷⁰ in vivo is sufficient to allow interaction with the EC and stimulation of pausing at extended -10-like sequences, making σ⁷⁰ a bona fide elongation factor. Arndt and Chamberlin (1988) discovered that even 300 nM σ⁷⁰ stimulates release of RNAP from terminators up to 20-fold during multiround transcription in vitro. They concluded that σ⁷⁰ loads directly onto RNAP during the process of termination. Analysis of the time required for each round of RNA synthesis suggested that the rate-limiting step for recycling of RNAP for new rounds of transcription is an isomerization in this RNAP-σ⁷⁰ complex (previously reported by Wu et al. 1976), rather than RNAP-σ⁷⁰ reassociation. Given these observations and our results, we suggest that σ⁷⁰ could be released from ECs upon promoter escape, but remain able to interact transiently with the EC during transcript elongation. If RNAP encounters an extended -10-like sequence (several of which are present within E. coli's transcriptional units), transiently interacting σ⁷⁰ could rebind to stimulate pausing. When RNAP reaches a terminator, σ⁷⁰ could rebind tightly, making the overall behavior of wild-type σ⁷⁰ similar to the behavior of σ⁷⁰ tethered to RNAP without requiring continuous physical retention of σ⁷⁰ by the EC.

Materials and methods

Strains, strain construction, plasmids, protein purification, plating efficiencies, and strain phenotypes are described in the Supplemental Material and Supplementary Table 1.

Templates for in vitro transcription and mobility shift assays

Templates were generated by PCR amplification (Supplementary Table 1) and purified by phenol extraction and spermine precipitation (Hoopes and McClure 1981). The template for mobility-shift assays (Fig. 4C) was isolated from a 1.5% GTG low-melting agarose gel, melted at 65°C in 5 vol of TE, extracted twice with phenol, extracted with CHCl₃, and EtOH-precipitated.

Cell extracts

Cellular extracts were prepared from early exponential-phase cells by centrifugation for 2 min at 10,000g, suspension in SDS buffer (4% SDS, 10% glycerol, 62.5 mM Tris at pH 6.8, 5% β-mercaptoethanol), and incubation at 100°C for 10 min.

Immunoblot analysis

Immunoblotting and detection were performed using ECL Plus reagents (Pharmacia), following the manufacturer's instructions and using primary antibodies against β′ (NT73; 1:10,000), σ⁷⁰ (2G10; 1:5000), and σ³² (3RH1; 1:3000), which was kindly provided by R. Burgess (University of Wisconsin, Madison, WI, USA), and secondary antibody (horseradish peroxidase-coupled mouse IgG) from Pharmacia. Proteolytic cleavage of β′::σ⁷⁰ to yield wild-type-length β′ or σ⁷⁰ was undetectable in β′::σ⁷⁰ strains, with detection limits of 0.5 ng of β′ and 1.2 ng of σ⁷⁰ per microgram of cell lysate. Levels of σ³² in wild-type and β′::σ⁷⁰ strains (Fig. 6B) were obtained from multiple independent cell lysates and blots (average ratio of σ³² in the β′::σ⁷⁰ strain relative to wild type of 0.8 ± 0.4). The estimated error in protein concentration loaded on gels (determined by Bradford assay) was 20%.

Promoter kinetics

To determine k_f,obs, 50 nM RNAP (β′::σ⁷⁰ or wild-type) was incubated with 25 nM T7 A1 template at 22°C in transcription buffer (Artsimovitch and Landick 2002). Samples were removed from 10 sec to 8 min, mixed with 0.1 vol of 2.5 mM ApU; 25 μM each ATP, [³²P]CTP, and GTP; and 1 mg/mL heparin; incubated at 37°C for 10 min; and then mixed with an equal volume of 2× STOP buffer (8 M urea, 98 mM Tris-borate at pH 8.3, 10 mM Na₂ EDTA, 0.2% bromophenol blue, 0.2% xylene cyanol). The RNA products were separated by 15% denaturing PAGE and quantified using a PhosphorImager. The results were fit to a single-exponential of formation by nonlinear regression using Kaleidagraph. At the concentrations of RNAP and DNA used, this assay would detect a decrease in the equilibrium binding of RNAP and DNA to form CC, but likely not an increase. To determine k_r,obs, OCs were first formed for 10 min as described for k_f,obs determination. At time 0, heparin was added to 100 μg/mL, and sampled from 15 sec to 100 min. Samples were mixed with 0.1 vol of 2.5 mM ApU and 25 μM each ATP, [³²P]CTP, and GTP; incubated at 37°C for 10 min; and then mixed with an equal volume of 2× STOP buffer. The RNA products were quantified as described for k_f,obs determination, and the results were fit to a single-exponential decay of OC by nonlinear regression using Kaleidagraph.

Equilibrium binding assay

HMKσ⁷⁰ (250 pmole) was ³²P-labeled as described previously (Artsimovitch and Landick 2002). β′::σ⁷⁰ or wild-type RNAPs (1 μM) were incubated with [³²P]σ⁷⁰ at 0.5, 1, 2.5, and 5 μM in equilibrating conditions using E buffer (0.5 M NaCl, 40 mM Tris-HCl at pH 8.0, 0.1 mM EDTA, 3 mM DTT, 0.2% Tween) plus 20% glycerol at 30°C for 20 min (Severinova et al. 1996; Sharp et al. 1999). Samples were then separated on a 4%-15% native Phast gel (Pharmacia) for 125 Vh at 15°C. Gels were Coomassie blue R-stained, analyzed using a PhosphorImager (Fig. 4A), and quantified (Fig. 4B). Effective σ⁷⁰ concentration was obtained by nonlinear regression of fraction RNAP containing [³²P]σ⁷⁰ versus concentration of [³²P] σ⁷⁰ using the equation y = x/(x + E), where x is the concentration of [³²P]σ⁷⁰ and E is the effective concentration of tethered σ⁷⁰.

Open complex mobility assay

β′::σ⁷⁰ or wild-type RNAPs (100 nM) were mixed with increasing amounts of σ⁷⁰ (amount added σ⁷⁰ in micromolar indicated above gel panel) under equilibrium conditions (E buffer + 5% glycerol) at 37°C for 10 min. Linear DNA containing P_UPFullcon (Supplementary Table 1) was added to 150 nM as the reaction was diluted to 0.25 M NaCl and incubated at 37°C for 10 min. OCs formed on this promoter are exceptionally stable (t_1/2 > 18 h; T. Gaal, pers. comm.). Reactions were electrophoresed on 4% NuSieve (FMC), 0.5× TBE, and stained with EtBr. Effective σ⁷⁰ concentration (Fig. 4D) was modeled using the equation y = x/(x + E), where x is the concentration of untethered σ⁷⁰ and E is the effective concentration of tethered σ⁷⁰ (set arbitrarily to 0.1, 1, 10, 50, and 100 μM).

Heat-shock response assay

Wild-type (RL301) or β′::σ⁷⁰ strain (RL1374) cells were grown in supplemented glucose medium at 30°C and labeled with [³⁵S]methionine (10 mCi/mL, Amersham; 0.2 mCi/mL final concentration) for 2 min at various times after shift to 42°C as described (Zhou et al. 1988). The cells were recovered by centrifugation, washed twice, suspended in SDS buffer, boiled for 10 min, and separated by 3%-8% Tris-Acetate (Novex) beside molecular weight markers. The gel was dried down and visualized using a PhosphorImager.

In vitro σ⁷⁰-dependent pause assay

Transcription on the +16/17 and +37 pause templates was performed as described (Ring et al. 1996). For the +462 pause template, A29 ECs were formed at 37°C by mixing 7.5 nM RNAP, 250 nM template, 150 μM ApU, and 10 μM ATP, GTP, and ³²P-CTP. The high excess of template was used to ensure only one RNAP binding per template. σ⁷⁰ (1 μM), NusA (10 μM), or both were added to ECs as indicated. All 4 NTPs (150 μM) and 100 μg Rif/mL were then added to allow elongation to resume. Samples were removed at the times indicated, mixed with equal volume 2× STOP buffer, separated by denaturing PAGE (5%; 0.5× TBE), and analyzed using a PhosphorImager. The size of the σ⁷⁰-dependent pause band (462 nt, predicted) was measured by comparison to size markers generated by transcription of restriction endonuclease-treated +462 pause templates and found to be 465 ± 0.5 nt (data not shown).

In vivo concentrations

The in vivo concentrations of σ⁷⁰ and RNAP are central to understanding the effects of tethering σ⁷⁰. Based on quantitative immunoblotting, Jishage and Ishihama (1995) report a constant in vivo concentration of σ⁷⁰ of 50-80 fmole/μg cellular protein regardless of growth phase in rich medium at 37°C. At a growth rate of 40 min/generation in LB medium (Fig. 1), an E. coli cell contains ∼7 × 10^-13 mL of water (composing 70% of the cell mass) and ∼2.3 × 10^-13 g of protein (Cayley et al. 1991; Bremer and Dennis 1996). Assuming a partial specific volume for protein of 0.72 (Cantor and Schimmel 1980), this gives a total cell volume of ∼1 × 10^-12 mL. Based on these numbers, we calculate a bulk protein concentration of 230 mg/mL [(2.3 × 10^-13 g of protein)/(1 × 10^-12 mL)] and a bulk concentration of σ⁷⁰ of 12-19 μM (12-19 nmole/mL). The molar ratios of total σs to σ⁷⁰ and of RNAP to σ⁷⁰ are reported to be 1.7 and 2.8 in the same conditions, with 65% of RNAP stably bound to DNA (e.g., as ECs; Ishihama 2000). Thus, the bulk concentration of free RNAP (available to bind σ) is also 12-19 μM and of total σs, 19-31 μM. Based on this, we derive the average bulk concentrations of free RNAP, σ⁷⁰, and total σs of ∼15 μM, ∼15 μM, and ∼25 μM, respectively.