Structures of E. coli σ^S-transcription initiation complexes provide new insights into polymerase mechanism

Significance

As RNA polymerase (RNAP) translocates along the DNA template for repetitive nucleotide additions, its active site opens and closes for NTP association and catalysis, and a pyrophosphate ion (PPi) is generated after each nucleotide incorporation. Understanding the role of PPi release is important for elucidating the polymerase mechanism. The structures of the σ^S-containing transcription initiation complexes (σ^S-TICs) provide insights into the mechanism of σ^S-dependent selective gene expression. In addition, the highly stressed σ^S-TICs trap a PPi at the RNAP active site, a previously unobserved but catalytically relevant functional state. Our study also demonstrates that PPi release is not directly related to either translocation or active site opening but causes extensive conformational changes on the periphery of the RNAP secondary channel.

Abstract

In bacteria, multiple σ factors compete to associate with the RNA polymerase (RNAP) core enzyme to form a holoenzyme that is required for promoter recognition. During transcription initiation RNAP remains associated with the upstream promoter DNA via sequence-specific interactions between the σ factor and the promoter DNA while moving downstream for RNA synthesis. As RNA polymerase repetitively adds nucleotides to the 3′-end of the RNA, a pyrophosphate ion is generated after each nucleotide incorporation. It is currently unknown how the release of pyrophosphate affects transcription. Here we report the crystal structures of E. coli transcription initiation complexes (TICs) containing the stress-responsive σ^S factor, a de novo synthesized RNA oligonucleotide, and a complete transcription bubble (σ^S-TIC) at about 3.9-Å resolution. The structures show the 3D topology of the σ^S factor and how it recognizes the promoter DNA, including likely specific interactions with the template-strand residues of the −10 element. In addition, σ^S-TIC structures display a highly stressed pretranslocated initiation complex that traps a pyrophosphate at the active site that remains closed. The position of the pyrophosphate and the unusual phosphodiester linkage between the two terminal RNA residues suggest an unfinished nucleotide-addition reaction that is likely at equilibrium between nucleotide addition and pyrophosphorolysis. Although these σ^S-TIC crystals are enzymatically active, they are slow in nucleotide addition, as suggested by an NTP soaking experiment. Pyrophosphate release completes the nucleotide addition reaction and is associated with extensive conformational changes around the secondary channel but causes neither active site opening nor transcript translocation.

Cellular organisms transfer genetic information from DNA to RNA using multisubunit RNA polymerases (RNAPs) that are conserved from bacteria to humans (1, 2). In bacteria, a single five-subunit core enzyme of RNA polymerase (α₂ββ′ω) is responsible for all RNA synthesis, whereas multiple σ factors compete to associate with the RNAP core enzyme to form a holoenzyme that is required for initiating the process at DNA promoter sites (3, 4). RNAP remains associated with the upstream promoter DNA during transcription initiation and moves downstream for RNA synthesis, causing DNA scrunching to form a stressed and unstable initiation complex (5–8). Processive RNA synthesis happens only after the initiation complex escapes the promoter as transcription progresses from initiation to elongation (9–11).

RNA synthesis in both transcription initiation and elongation involves repetitive cycles of nucleotide addition comprising translocation, NTP binding, catalysis, and pyrophosphate release steps. During this cycling process, the RNAP active site opens for NTP association and closes to align the incoming NTP with the RNA 3′ hydroxyl group for catalysis. Nucleotide addition extends the RNA by 1 nt and generates a pyrophosphate ion (PPi). Previous structural studies of transcription complexes frequently used DNA fragments that form a partial transcription bubble. Most of these complexes display an open active site in the posttranslocation state, and the active site-closed conformation has been observed only with NTP-bound precatalysis complexes in which the nucleotide addition reaction was prevented by using 3′-deoxy RNA or nonhydrolysable NTP analogs (Table S1). A PPi-associated complex has never been observed previously in cellular RNAP structures. How release of PPi affects the opening of the RNAP active site and transcription translocation remains to be established.

Table S1.

Structural survey of transcription complexes

NAC conformation	PDB ID	Comments	Reference
I TL/open	1I6H	Yeast Pol II complex with upstream DNA and 13–14 nt RNA synthesized with GTP+ATP+CTP and stalled because of a lack of UTP	(29)
	3M3Y	Yeast Pol II complex with pyriplatin-dG:C base pair at initiation + 1 (i+1) site; translocation impaired	(49)
	4Q5S	TthRNAP initiation complex; stressed with σ3.2 blocking of 6-nt RNA	(50)
I → II ? TL/open	2VUM	Yeast Pol II complex with α-amanitin; translocation intermediate	(31)
I ↔ II ? TL/open (failed or incomplete translocation?)	3RZD	Yeast Pol II complex; hybrid tilted*	(51)
	3RZO	Yeast Pol II complex; hybrid tilted*
	3S1R	Yeast Pol II complex with 3′-deoxy RNA and E site GTP; hybrid tilted*
	4A3B	Yeast Pol II complex; hybrid tilted*	(52)
	4A3C	Yeast Pol II complex; hybrid tilted*
	4A3D	Yeast Pol II complex; hybrid tilted*
	4A3K	Yeast Pol II complex; hybrid tilted*
	4BY7	Yeast Pol II complex; hybrid tilted*	(53)
II TL/open	1SFO	Yeast Pol II complex	(54)
	1Y1W	Yeast Pol II complex	(55)
	2JA5	Yeast Pol II complex	(56)
	2JA6	Yeast Pol II complex
	2JA7	Yeast Pol II complex with dT-dT dimer at i and i+1 sites
	2JA8	Yeast Pol II complex with dT-dT dimer at i and i+1 sites; unusual downstream DNA shifts (similar to 4A93)
	2O5I	TthRNAP complex	(57)
	2R7Z	Yeast Pol II complex with cisplatin-crosslinked T-strand dGi+2 and dGi+3	(58)
	2R92	Yeast Pol II complex with RNA stem loop	(59)
	2R93	Yeast Pol II complex with RNA duplex
	3H3V	Yeast Pol II complex	(60)
	3HOU	Yeast Pol II complex with dT-U mismatch at i site	(61)
	3HOV	Yeast Pol II complex
	3HOW	Yeast Pol II complex with dT-U mismatch at i site and frayed 3′-U
	3HOX	Yeast Pol II complex
	3HOY	Yeast Pol II complex
	3HOZ	Yeast Pol II complex with dT-U mismatch at i site, frayed 3′-G
	3I4M	Yeast Pol II complex with 8-oxo-dG:C base pair at i site, frayed 3′-U	(62)
	3I4N	Yeast Pol II complex with 8-oxo-dG:A base pair at i site, frayed 3′-U
	3M4O	Yeast Pol II complex with pyriplatin-dG at i+1 site	(49)
	3S14	Yeast Pol II complex	(51)
	3S15	Yeast Pol II complex
	3S16	Yeast Pol II complex
	3S17	Yeast Pol II complex
	3S1M	Yeast Pol II complex
	3S1N	Yeast Pol II complex
	3S2D	Yeast Pol II complex
	3S2H	Yeast Pol II complex
	4A3G	Yeast Pol II complex	(52)
	4A93	Yeast Pol II complex with dT-dT dimer at i+1 and i+2 sites, unusual downstream DNA shifts (similar to 2JA8)	(63)
	4BBS	Yeast Pol II initiation complex with TFIIB	(64)
	4G7O	TthRNAP initiation complex	(18)
	4S20	EcoRNAP complex with RapA	(47)
	4XLN	TaqRNAP initiation complex with a 13-nt bubble	(17)
	4XLR	TaqRNAP initiation complex with a 13-nt bubble and TthCarD	(65)
	5C3E	Yeast Pol II complex with a 15-nt bubble (upstream invisible)	(66)
	5C44	Yeast Pol II complex with a 15-nt bubble (upstream invisible)
	5C4A	Yeast Pol II complex with a 12-nt bubble (upstream invisible)
	5C4J	Yeast Pol II complex with a 15-nt bubble
	5C4X	Yeast Pol II complex with a 15-nt bubble
III TL/open	1R9S	Yeast Pol II complex with 3′-deoxy RNA and A site UTP	(37)
	1R9T	Yeast Pol II complex with 3′-deoxy RNA and E site noncognitive ATP (template dA)
	1Y77	Yeast Pol II complex with pre-A site GMPcPP	(55)
	2E2I	Yeast Pol II complex with 3′-deoxy RNA and A site/E site dGTP	(27)
	2E2J	Yeast Pol II complex with A site GMPcPP
	2NVQ	Yeast Pol II complex with 3′-deoxy RNA and A site/E site dUTP
	2NVT	Yeast Pol II complex with A site GMPcPP
	2NVX	Yeast Pol II complex with 3′-deoxy RNA and A site/E site dUTP
	2YU9	Yeast Pol II complex with 3′-deoxy RNA and A site/E site UTP
	2PPB	TthRNAP complex with A site AMPcPP and streptolydigin	(25)
	4OIO	TthRNAP initiation complex with 2 NTPs (ATP and CMPcPP)	(67)
IV? TH/close†	4A3E	Yeast Pol II complex with A site AMPcPP	(52)
	4A3F	Yeast Pol II complex with A site AMPcPP
	4A3J	Yeast Pol II complex with A site GMPcPP
	4A3L	Yeast Pol II complex with A site AMPcPP
	4A3M	Yeast Pol II complex with A site AMPcPP
	4BY1	Yeast Pol II complex with A site AMPcPP	(53)
	4Q4Z	TthRNAP initiation complex with 2 NTPs (ATP and CMPcPP)	(50)
IV TH/close	2E2H	Yeast Pol II complex with 3′-deoxy RNA and A site GTP	(27)
	2NVZ	Yeast Pol II complex with 3′-deoxy RNA and A site UTP
	2O5J	TthRNAP complex with A site AMPcPP	(25)
	3S1Q	Yeast Pol II complex with 3′-deoxy RNA and A site ATP	(51)
V TH/close		σS-EcoRNAP initiation complex with a 14-nt bubble and Ppi	This work
V → VI TH/close		σS-EcoRNAP initiation complex with a 14-nt bubble	This work
		σS-EcoRNAP initiation complex with a 14-nt bubble
VI TH/close	4YLN	σ70-EcoRNAP initiation complex with a 14-nt bubble	(5)
	4YLO	σ70-EcoRNAP initiation complex with a 14-nt bubble
	4YLP	σ70-EcoRNAP initiation complex with a 15-nt bubble
Transcription complexes of other conformations
Ratcheted TL/blocked‡	3AOH	TthRNAP complex with Gfh1; cleft open	(68)
	3AOI	TthRNAP complex with Gfh1; cleft open
Elemental pause TL/blocked‡	4GZY	TthRNAP complex; cleft open	(69)
	4GZZ	TthRNAP complex; cleft open
Backtracked TL/blocked§	3GTG	Yeast Pol II backtracked complex	(70)
	3GTJ	Yeast Pol II backtracked complex
	3GTK	Yeast Pol II backtracked complex
	3GTL	Yeast Pol II backtracked complex
	3GTM	Yeast Pol II backtracked complex with TFIIS
	3GTO	Yeast Pol II backtracked complex
	3GTP	Yeast Pol II backtracked complex
	3GTQ	Yeast Pol II backtracked complex
	3PO2	Yeast Pol II backtracked complex; hybrid tilted*	(71)
	3PO3	Yeast Pol II backtracked complex with TFIIS; hybrid tilted*

All available crystal structures of transcription complexes of cellular RNA polymerases as of September 25, 2015 are categorized in regard to their relative conformations in the nucleotide addition cycle (NAC). NAC conformations: I, pretranslocation state with an unfolded TL and an open active site (TL/open); II, posttranslocation state with TL/open; III, posttranslocation state with a bound NTP and an unfolded TL (TL/open); IV, posttranslocation state with a bound NTP and a active site closed by TH formation (TH/close); V, pretranslocation state immediately after catalysis with a PPi-associated and closed active site (TH/close); VI, pretranslocation state after PPi release with the active site remaining closed by THs (TH/close). AMPcPP, α,β-methyleneadenosine 5′-triphosphate; CMPcPP, α,β-methylenecytosine 5′-triphosphate; Eco, Escherichia coli; GMPcPP, α,β-methyleneguanosine 5′-triphosphate; Pol II, RNA polymerase II; Taq, Thermus aquaticus; Tth, Thermus thermophilus.

^*“Hybrid tilted” refers to the hybrid conformation in which the RNA nucleotide at register n base pairs with the DNA template nucleotide at register n+1.

^†These structures display a largely folded N-terminal helix of the TL, and thus the active sites are closed, more like that of complex IV.

^‡Both “ratcheted” and “elemental pause” display a cleft-open conformation with a bent BH that is incompatible with the helical conformation of the TL. This bent BH conformation also might be incompatible with A-site/pre–A-site NTP binding.

^§Backtracked complexes have backtracked RNA residues that prevent NTP binding and are also incompatible with a folded TL conformation.

In this work we determined the crystal structures of the E. coli transcription initiation complexes (TICs) containing the σ^S factor, a de novo synthesized RNA oligonucleotide, and a complete transcription bubble (σ^S-transcription initiation complexes, σ^S-TICs). The σ^S factor controls the expression of many genes in response to general stresses, such as nutrient deprivation upon entry into stationary phase. The structures show the specific interactions of the σ^S factor with the promoter −10 element and provide insights into the mechanism of σ^S-dependent selective gene expression under stress conditions. In addition, the σ^S-TIC crystals display a pretranslocated initiation complex with a PPi associated at the active site that remains closed. The position of the PPi and the unusual phosphodiester linkage between the two terminal RNA residues suggest an unfinished nucleotide addition reaction that likely is at equilibrium between nucleotide addition and pyrophosphorolysis. The slow enzymatic activity of the σ^S-TIC crystals allowed us to observe PPi dissociation without nucleotide addition in an NTP-soaking experiment. PPi release appears to be associated with extensive conformational changes around the secondary channel but causes neither active site opening nor transcript translocation.

Results and Discussion

Overall Structure of the E. coli σ^S-Based TIC.

By using a synthetic DNA scaffold containing the promoter consensus −35 and −10 sequences, we assembled and crystallized the E. coli TIC that contains the general stress response σ^S factor. The σ^S-TIC crystals are enzymatically active for additional nucleotide incorporation. The structures were solved by molecular replacement, and the final models were refined to about 3.6-Å resolution (Table 1). The σ^S-TIC structures display a closed active site and a well-ordered nascent RNA–DNA hybrid in the pretranslocation register.

Table 1.

Data collection and refinement statistics

Parameters	σS-TIC1	σS-TIC2 1-h soaking with CTP only	σS-TIC3 1-h soaking with CTP, UTP, and GTP
Data collection
Space group	P212121	P212121	P212121
Cell dimensions, Å
a	131.71	132.75	132.87
b	152.67	151.97	152.17
c	226.70	228.06	229.12
Resolution, Å	100–3.60 (3.66–3.60)	100–4.20 (4.27–4.20)	100–4.60 (4.68–4.60)
Rsym, %	12.5 (>100)	10.2 (>100)	8.5 (>100)
I/σI*	9.87 (0.37)	8.44 (0.46)	10.06 (0.55)
Completeness, %	99.6 (99.4)	99.9 (100)	99.3 (99.6)
Redundancy	5.1 (4.2)	7.7 (7.9)	6.9 (6.8)
Refinement
Resolution, Å	100–3.60	100–4.20	100–4.60
No. reflections	50,952	32,692	24,772
Rwork/Rfree, %	24.7/28.7	26.7/32.7	24.0/33.4
No. atoms	27,632	29,036	29,036
Protein	26,167	27,572	27,572
DNA/RNA/ions	1,465	1,464	1,464
B factors	129.5	177.4	183.1
Rmsd
Bond length, Å	0.011	0.011	0.008
Bond angle, °	1.448	1.422	1.179

Values in parentheses are for highest-resolution shell. Each dataset was collected from a single crystal.

^*I/σI = 2.0 at 4.2 Å (σ^S-TIC1), 5.6 Å (σ^S-TIC2), 6.2 Å (σ^S-TIC3); I/σI = 1.0 at 3.9 Å (σ^S-TIC1), 4.7 Å (σ^S-TIC2), 5.2 Å (σ^S-TIC3).

The σ^S-TIC that initially crystallized contains a 14-nt bubble and an RNA tetranucleotide synthesized de novo from three NTPs and traps a PPi at the active site (Fig. 1). Continuous electron densities are seen for both strands of the ssDNA in the bubble region (Fig. S1A). Much of the nontemplate (NT) strand of the bubble that is downstream of the −10 hexamer could not be resolved unambiguously in the density map, suggesting a slippery feature of this DNA segment that displays limited interactions with the protein. The ssDNA residues on the template (T) strand are much better resolved, although residues in the middle of this section appear disordered. The upstream promoter DNA runs into the downstream DNA duplex of a symmetry-related molecule, causing strand separation and disordering of a significant portion of the upstream DNA (Fig. S1B).

Fig. 1.

Overall structure of the E. coli σ^S-TIC. (A) Schematic representation of the synthetic promoter DNA and a de novo synthesized RNA transcript in the σ^S-TIC crystals. RNA synthesis starts from the −1 position with a GTP residue as observed previously with related promoters in the σ⁷⁰-TICs (5). The disordered upstream DNA residues are shown as gray cycles. (B) Structure of the σ^S-TIC. The E. coli RNAP core enzyme is shown in a tube-and-arrow cartoon representation. The σ^S factor is shown as a Cα trace within a surface representation (wheat). The promoter DNA (NT strand, blue; T strand, green) and the nascent RNA (red) are shown as filled spheres.

Fig. S1.

Nucleic acid in the σ^S-TIC. (A) An overall view of the entire transcription bubble in the σ^S-TIC. The Fo-Fc electron density map (mesh, contoured at 2.0 σ) was calculated using the phase from the protein-only model. (B) Crystal packing involving the dsDNA. The downstream DNA from one complex stacks noncoaxially with the upstream DNA from a symmetry mate. The RNAP core enzyme is omitted for clarity. The DNA from the symmetry mate is colored gray. σ^S is shown in a surface representation. A metal ion (MeI) is shown as a magenta sphere to mark the active center.

E. coli σ^S Factor and σ^S-RNAP Core Interactions.

Like the primary σ factors, the E. coli σ^S factor contains a highly negatively charged N-terminal domain (σ_1.1) that we have not been able to trace in the σ^S-TIC structures. Other than a nonconserved region (NCR) that is commonly present in primary σ factors, σ^S displays very high sequence identity with the primary σ factors (Fig. S2A) (10). Consistent with their sequence similarity, the σ^S factor in the σ^S-TIC forms essentially the same fold as the σ⁷⁰ factor throughout the conserved regions, from region 1.2 to the very C terminus, including the long σ_3.2 loop that inserts deep into the RNAP active site chamber (Fig. 2A).

Fig. S2.

E. coli σ^S factor. (A) A sequence alignment of the E. coli σ^S factor (RPOS_ECOLI, GenBank accession no. gi:1710676) with the Thermus thermophilus σ^A factor (SIGA_THET2, gi:81699349) and the E. coli σ⁷⁰ factor (RPOD_ECOLI, gi:1710680). Sequences are presented with the one-letter amino acid code; numbers at the beginning and end of each line indicate amino acid positions relative to the start of each protein sequence. Secondary structures (α helices) of the E. coli σ^S factor are indicated above the sequences, and evolutionarily conserved regions are labeled below the sequences. NCR marks the nonconserved region of the primary σ factors. (B) A comparison of the interactions of the σ2 domains of different σ factors with the RNAP core enzyme. Structures of the E. coli σ^S-TIC (Left), E. coli σ⁷⁰-RNAP holoenzyme (Ec σ⁷⁰-Holo, PDB ID code 4JKR) (Center), and the RNAP holoenzyme in the T. thermophilus initiation complex (Tth σ^A-Holo, PDB ID code 4G7H) (Right) are superimposed using the clamp helices (E. coli β′ residues 263–308; T. thermophilus β′ residues 538–583) and are shown separately in the same orientation. The σ factors are shown in surface representation; clamp helices are shown in ribbons; and the rest of the core enzymes are shown as Cα traces and are colored as in Fig. 1B. The σ^S residues D87, D135, P136, and E137 that interact with Crl are marked in red. The σ⁷⁰ factor acidic patch (residues 188–209) is modeled and shown in red. A fragment of the upstream DNA in the σ^S-TIC (NT strand, blue; T strand, orange) that interacts with the σ2 and σ3 domains is shown for reference. (C) Stereoview of the upstream fork region of the σ^S-TIC. The promoter DNA is shown as ladders (NT strand, blue; T strand, green); the NT-strand residues −11A and −7T are shown as magenta sticks.

Fig. 2.

E. coli σ^S factor. (A) Overall structure of the E. coli σ^S factor in the σ^S-TIC. The orientation of σ^S here is the same as in Fig. 1. (Left) The Fo-Fc electron density map (mesh, contoured at 2.0 σ) was calculated using the phases from the RNAP core-only model. (Right) The E. coli σ^S factor in the σ^S-TIC (colored) and the σ⁷⁰ factor in the σ⁷⁰-holoenzyme (PDB ID code4JKR) (gray) are superimposed. (B) Recognition of the promoter −10 element in the σ^S-TIC. The N-terminal half of the β subunit (residues 2–667) is omitted for clarity. The nascent RNA (red) and the −10 hexamer residues of the T strand (green) are shown as sticks. The rest of the nucleic acid is shown in cartoon style (T strand, green; NT strand, marine). The RNA polymerase is shown in surface representation: σ^S, wheat; β subunit, light gray; β′ lid, cyan; β′ rudder, forest green; the rest of the β′ subunit is shown in dark gray. A few residues that interact directly with the −10 T-strand residues are labeled in the insets.

A major contribution to the σ-RNAP core association comes from interactions between the σ₂ domain and a helix-turn-helix formation, termed a “clamp helices,” of the clamp domain of the β′ subunit. Compared with σ⁷⁰, which displays strong affinity for the RNAP core enzyme (K_d ∼0.26 nM), σ^S binds relatively weakly to the RNAP core in the absence of nucleic acids (K_d ∼4.26 nM) (12). This weaker affinity of σ^S might be attributed partly to its lack of an NCR domain that provides additional interactions with the clamp helices of the RNAP core enzyme (Fig. S2B).

In E. coli and many γ-proteobacteria, a small protein called “Crl” was found to stimulate σ^S-dependent transcription by promoting the formation of the σ^S–RNAP holoenzyme (13, 14). It was shown previously that the σ^S–Crl interaction involves a general area on the surface of σ^S including residues of Asp87, Asp135, Pro136, and Glu137 (15). This general area corresponds to the attachment face of the NCR to the primary σ factors (Fig. S2B), suggesting that Crl might function like the NCR of a primary σ factor to help encompass the clamp helices and enhance the interactions of σ^S with the core enzyme.

Promoter Recognition by E. coli σ^S Factor.

The E. coli σ^S factor was shown to recognize the same consensus −35 and −10 hexamers as the σ⁷⁰ factor in an in vitro selection experiment (16). The σ^S-TIC structures show that σ^S recognizes the –10 hexamer of the NT strand with specific interactions with the bases of the −7T and −11A residues in the same manner as the primary σ factors (Fig. S2C) (5, 17–19). It was suggested recently that the −10 T strand passes through a cleft between the σ₂ and σ₃ domains in an initiation complex (5, 17), but it remained unclear how the −10 T strand interacts with the RNAP. The σ^S-TIC structures suggest that the single-stranded T-strand residues −7 to −11 stack on each other inside the narrow tunnel formed by the σ₂–σ₃ cleft and the β1 domain (Fig. 2B). The DNA backbone appears to make extensive interactions with positively charged side chains from the β1 domain and the σ₂ and σ₃ domains, and the bases may form several hydrogen-bonding interactions with the σ^S factor (Fig. S3), raising the possibility that the T-strand residues may contribute to the specific recognition of the −10 hexamer by the σ⁷⁰ family factors. Additional evidence is required to determine whether the −10 template region contributes to the promoter recognition.

Fig. S3.

Local contacts at the −10 element of the T strand. (A) The pink Fo-Fc electron density map (mesh, contoured at 2.0 σ) was calculated using the phase from the protein-only model. (B) Stereoview of the local contacts at the −10 element of the T strand. Colors are as in Fig. 2. Most residues involved in the contacts, including R112, I158, R163, N176, R180, and R183, are conserved between σ⁷⁰ and σ^S.

The few differences between σ^S and σ⁷⁰ in their interactions with the extended −10 region include residues Ile169 and Lys173 in σ^S and the corresponding residues Val454 and Glu458 in σ⁷⁰. It was reported that Lys173 is responsible for a preference of cytidine by σ^S at the promoter −13 site (20). Lys173 potentially could interact with the −13 base of the T strand, but a specific recognition could not be established at the current resolution.

In the σ^S-TIC crystals shown here, the helix-turn-helix of the σ^S C-terminal domain (σ₄) that potentially interacts with the promoter −35 hexamer, including helix H42 and the N-terminal half of helix H43, is involved in crystal packing, and thus there is no space to accommodate the −35 region of the promoter DNA duplex. Although σ^S interactions with the promoter −35 element are not observed in the σ^S-TIC crystals, the sequence and structural conservation and the shared recognition of promoter sequences suggest that the σ^S and σ⁷⁰ factors would interact with the DNA promoters in a very similar manner.

Selective Gene Expression Under Stress Conditions.

An intriguing question is how σ^S achieves selective gene expression, given that it recognizes consensus −35 and −10 sequences essentially identical to those recognized by σ⁷⁰. The σ^S-dependent promoters display higher sequence deviations from the consensus −35 hexamer (21); these deviations, again, might be related to the absence of an NCR domain in the σ^S factor. Although the NCR domain in σ⁷⁰ might make it interact better with the RNAP core enzyme, a stretch of acidic residues (18 of the E. coli σ⁷⁰ residues 188–209) (Fig. S2B) is expected to inhibit promoter loading directly through interactions of the σ⁷⁰-holoenzyme with the −10 element (22), thus making interactions with the −35 element important for recruiting the σ⁷⁰-holoenzyme to the promoter. In contrast, the smaller σ₂ domain of the σ^S factor would allow it to interact directly with the promoter −10/extended −10 region, thus making the σ^S-holoenzyme less dependent on the −35 element for promoter loading.

The protein concentration of the σ^S factor in E. coli is tightly regulated at the levels of transcription, translation, and protein stability (23, 24). During exponential growth in rich medium, the σ^S protein level in E. coli is negligible, but when E. coli enters the stationary phase or under certain stress conditions the σ^S protein level might increase by a thousand-fold and reach a level comparable to that of σ⁷⁰. Easier access not only would provide the σ^S-holoenzyme an advantage in competing for promoters in the heavily packed DNA during the stationary phase but also would justify the requirement of tight regulation of the σ^S protein concentration under normal growth conditions.

Stressed TICs.

The σ⁷⁰-TICs that we reported recently contain a complete transcription bubble and display a well-ordered nascent RNA–DNA hybrid lying at the pretranslocation position; we suggested that this pretranslocated hybrid may be a manifestation of the stressed feature of an initiation complex (5). Similar to the σ⁷⁰-TICs, the σ^S-TICs we report here also contain a complete transcription bubble and an RNA oligonucleotide synthesized de novo from NTPs, and the σ^S-TIC structures display a pretranslocated RNA–DNA hybrid as well (Fig. 3). Moreover, the active site of the σ^S-TICs is fully closed by the folding of the trigger loop (TL) into helices (TH). Although the helical conformation of the TL in a pretranslocated transcription complex also might exist in our low-resolution σ⁷⁰-TIC structures (5), it has never been observed previously in any other transcription complexes.

Fig. 3.

σ^S-TIC active site. (A) A close-up view of the DNA–RNA hybrid and PPi at the active site in the structure before NTP soaking. The Fo-Fc electron density map (mesh, contoured at 3 σ) was calculated using the phases calculated from the protein-only model. The inset shows a zoomed-in view of the electron density (contoured at 5 σ) around the catalytic site and the RNA 3′-terminal residue. (B) A close-up view of the active site showing the major interactions involved in stabilizing the PPi. The metal ions (MeI and MeII) are shown as magenta spheres.

It has been shown that conformational changes of a flexible TL and the bridge helix (BH), which traverses across the active site cleft, remodel the active site of cellular RNA polymerases (1, 2, 25–27). An unfolded TL leaves the active site open to the secondary channel and thus allows NTP binding to the active site, whereas the folding of the TL into two α-helices (THs) closes the active site and helps align the incoming NTP with the RNA 3′-hydroxyl group and with a conserved histidine residue (β′His936 in E. coli RNAP) that serves as a general acid for catalysis (25, 27). The BH stacks against the RNA–DNA hybrid and could bend or bulge toward the template base, which forms the basis of the BH-controlled transcription translocation model (28, 29). The TL folding has a profound effect on the rate of transcription (30) and is affected by the BH conformation (25, 27). It was also suggested that transcription translocation might require an unfolded TL (31).

In the stressed σ^S-TIC, the BH displays a slight bending perpendicular to the translocation direction with a shift of about 3 Å around β′ residues 785–789 (Fig. S4A). This small shift in the BH appears to create enough space to allow the TL to fold into helices. TH formation significantly reduces the dimensions of the secondary channel entrance to the active site, from about 15 × 12 Å in the RNAP apo holoenzyme to about 8 × 6 Å in the σ^S-TIC (Fig. S4B), and thus even the dissociation of PPi might involve a concerted effort by the protein, involving more than amino acid side chain movements. Intriguingly, the phosphate of the RNA 3′-terminal residue appears not to interact directly with the metal ion (MeI) coordinated by the conserved carboxyl triad at the active center (Fig. 3A). The electron density contoured at higher levels displays some discontinuity between the two terminal residues, and the linkage of the terminal phosphate (+3U) to the penultimate RNA residue (+2G) also appears to be significantly distorted at the sugar ring of the penultimate residue. This finding suggests that the σ^S-TIC crystal might contain a mixture of initiation complexes, likely at equilibrium between the forward nucleotide addition reaction and presumably its reverse reaction, pyrophosphorolysis (32–34).

Fig. S4.

Active site conformations. (A) BH bending in the stressed TIC. Structures of the apo E. coli holoenzyme (PDB ID code 4LJZ) (gray), T. thermophilus elemental pause complex (PDB ID code 4GZY) (yellow), and T. thermophilus elongation complex with AMPcPP (APC) and streptolydigin (STD) (PDB ID code 2PPB) (blue) are superimposed using the RNAP β subunit. Streptolydigin and PPi (cyan) are presented as sticks. The nascent RNA and the T-strand DNA in the σ^S-TIC are shown as ladders. The black arrow marks the direction of BH bending related to transcription translocation; the red arrow marks the direction of BH bending observed in the TICs. Both directions of BH bending are based on a comparison with the RNAP apo holoenzyme. (B, Upper) An open active site. A view down to the secondary channel of the RNAP apo holoenzyme (PDB ID code 4LJZ) is presented. Nucleic acids and a PPi were modeled by superimposing the σ^S-TIC structure onto the RNAP apo holoenzyme using the β subunit. (Lower) A closed active site. A view down to the secondary channel of the σ^S-TIC (before NTP soaking) is displayed. The insets on the right are zoomed-in views of the secondary channel.

Unreleased Pyrophosphate in the σ^S-TIC Crystals.

Because RNA polymerase adds nucleotides repetitively to the RNA 3′ end, one pyrophosphate ion is generated after each addition reaction. During this process, it is generally believed that the active site closes after NTP association and opens immediately after or concurrently with the dissociation of the PPi. For T7 RNA polymerase (T7RNAP), a PPi-associated complex has been obtained by supplementing the solution for crystallization with PPi; the PPi-bound and unbound states of T7RNAP were found to associate with the pretranslocation and posttranslocation states, which correspond to the closed and open conformations of the active site, respectively (35). However, a PPi-associated structure has never been observed for the cellular RNA polymerases (Table S1). It was shown previously that translocation occurs shortly after or concurrently with PPi release (36). However, how PPi release affects the opening of the RNAP active site and the translocation of the enzyme along the DNA template remains obscure.

The intrinsic abortive feature of transcription initiation would cause many rounds of oligonucleotide synthesis during the process of complex assembly and crystallization, and these multiple rounds of synthesis could lead to a significant accumulation of PPi in the crystallization drops that reaches a level comparable to or even exceeding the levels of NTPs. On the other hand, because of DNA scrunching, TICs are stressed and tend to rest at the pretranslocation state, a conformation clearly more favorable than others for PPi association. Not surprisingly, a well-ordered PPi remains associated with the σ^S-TIC active site.

The pyrophosphate in the σ^S-TIC interacts via a metal ion (MeII) with the carboxyl group of β′Asp460, one of the conserved carboxyl triad of the active center, with the side chains of the TL residues β′Arg933 and β′His936, and with the side chains of βArg1106 and β′Arg731 that line as the secondary channel (Fig. 3B). The observation that both the PPi and the phosphate of the RNA 3′-terminal residue interact with the side chain of β′His936 is consistent with the proposal that this conserved TL histidine residue is involved in both nucleotide addition and pyrophosphorolysis (25, 27). It is not clear whether the position and the network of PPi interactions we observed here also represent those of the β- and γ-phosphates of an incoming NTP before nucleotide incorporation, although it is likely that they do (Fig. S5).

Fig. S5.

PPi and NTP binding. A comparison of the active site of the σ^S-TIC before NTP soaking with that of the T. thermophilus transcription elongation complex with the NTP analog AMPcPP (APC) (PDB ID code 2O5J) (gray) (A); the T. thermophilus TIC with ATP and CMPcPP (2TM) (PDB ID code 4Q4Z) (light blue) (B); the T. thermophilus transcription elongation complex with APC and streptolydigin (PDB ID code 2PPB) (wheat) (C); and yeast RNAP II elongation complex with 2′-deoxyguanosine-5′-triphosphate (dGTP) (PDB ID code 2E2I) (green) (D). The superimposition was performed using the β subunit of bacterial RNAP or the second-largest chain of the yeast RNA polymerase II.

Dissociation of PPi and RNAP Conformational Changes.

To test the ability of the σ^S-TIC crystals to incorporate nucleotides, we soaked the crystals in solutions containing CTP that was omitted in the original complex assembly. Unlike the σ⁷⁰-TIC crystals that readily extend the RNA by 1 nt after soaking (5), the σ^S-TIC crystals appear to be much less efficient in RNA synthesis. Soaking σ^S-TIC crystals with CTP only (for a 1-nt extension) or with three NTPs (CTP, UTP, and GTP, for RNA extension of up to 3 nt) for 1 h did not result in noticeable nucleotide addition. RNA extension was observed only after NTP soaking for 2 h or longer. It remains unclear why the σ^S-TIC crystals display such a low reactivity, but it is worth mentioning that the concentration of free Mg²⁺ might be low during the soaking of NTP into the crystals.

Although soaking of the σ^S-TIC crystals in solutions containing NTPs for 1 h did not lead to nucleotide incorporation, the two NTP soaking experiments led to similar, significant conformational changes in both the protein and the nucleic acid in the σ^S-TIC crystals (Fig. 4). Because the soaking solution contains no PPi, soaking leads to the release of the associated PPi as expected, and coordinately, the phosphate group of the RNA 3′-terminal residue shifts toward MeI and makes a more normal phosphodiester linkage with the penultimate RNA residue, suggesting that completion of the nucleotide addition reaction might require PPi release. Interestingly, it appears that there is no noticeable NTP binding to the NTP “entry site” that would overlap with the PPi-binding site (27, 37).

Fig. 4.

Conformational changes after PPi release. (A) The secondary channel of the σ^S-TIC after the crystals were soaked in NTP-containing solutions. RNAP is shown in surface representation. The TL insertion domain (SI3) (purple), F-loop (green), RHs (yellow), BH (blue), and TH1 and TH2 (pink) are also presented in cartoon representations. The insets on the left are close-up views of the flexible structural elements. The nascent RNA (red) and Ppi (cyan) are shown as filled spheres. The PPi was modeled in as a reference. (B) A close-up view of the F-loop and RHs in the σ^S-TICs before (gray) and after (yellow) NTP soaking and PPi release. (C) A close-up view of the active site of the σ^S-TICs before (gray) and after (pink and red) PPi release. The corresponding region of the E. coli RNAP apo-holoenzyme (PDB ID code 4LJZ) (green) is shown for reference. The structures are superimposed on the β subunits.

E. coli RNA polymerase carries a large lineage-specific insertion, termed “sequence insertion 3” (SI3), in the middle of the TL (38). This TL insertion of 188 amino acids could make a large shift around the secondary channel (39) and has been shown to affect E. coli transcription because it disfavors TH formation and stimulates pausing (40). In all of the three σ^S-TIC crystals described here, SI3 is expected to make similar contacts with symmetry-related molecules and contributes negligibly to crystal packing. However, a long hairpin loop in the C-terminal domain of SI3 projects into an intermolecular space and likely prevents SI3 from making large movements. The single most prominent conformational change on the σ^S-TIC after NTP soaking appears to be that the normally disordered SI3 becomes ordered and visible in the electron density map. In this ordered conformation SI3 fills the gap between the lobe and jaw domains of E. coli RNAP and forms part of the wall that separates the primary downstream DNA channel and the secondary channel (Fig. 4A).

In addition to SI3 becoming localized and ordered, the rim helices (β′ residues 650–703) on the edge of the secondary channel are rotated by about 7° toward the SI3–TH linkers (a shift of about 4 Å at the far end). At the same time, the F-loop (β′ residues 742–762) that interacts with both BH and TH makes a similar rotation to press on the SI3–TH linkers (Fig. 4B). The conformational changes of the rim helices and the F-loop, which are likely induced by the release of PPi, appear to function together to tighten the SI3–TH linkers and to lock SI3 against the jaw domain with its cleft between two β-folds (Fig. S6).

Fig. S6.

The E. coli RNAP TL insertion domain SI3. (A) The structure of the free SI3 protein fragment (PDB ID code 2AUK) (gray) is superimposed on the SI3 in the σ^S-TIC after soaking (purple) using the N-terminal β-fold of SI3 (SBHMa). The β' jaw domain in the σ^S-TIC (pink) is shown as a reference. (B) Superimposition of SI3 on this structure (purple), E. coli RNAP model (yellow) (48), and σ⁵⁴-RNAP holoenzyme (PDB ID code 5BYH) (cyan) (39). The superimposition was performed using the β subunit of bacterial RNAP. RNAP and downstream DNA are shown using surface representations. Colors are as in Fig. S4.

Because NTP soaking caused very limited changes to the crystal packing, it is likely that the significant conformational changes of SI3, the rim helices, and the F-loop are directly related to the release of PPi. After PPi dissociation, there appears to be a very minor shift (∼1–2 Å) at the C terminus of the first trigger helix (TH1) (Fig. 4C), and the σ^S-TIC remains at the pretranslocated position with a closed active site.

Concluding Remarks.

The thermodynamic (Brownian ratchet) model of transcription translocation suggests that RNA polymerase oscillates between pre- and posttranslocation states with the forward movement biased by NTP binding. Interestingly, structures of cellular RNA polymerase complexes are overwhelmingly at the posttranslocation state with an open active site (Table S1). Although structural studies could not be used as an evidence for thermodynamic analysis, the structural preference suggests that the posttranslocation state with an open active site might be the thermodynamically more favored state. However, in both the σ⁷⁰-TIC and σ^S-TIC crystals that contain a complete bubble, all the hybrids that we observed lie at the pretranslocation position, and the active site remains closed. It is likely that the structural preference seen in previous structural studies is a biased representation directly related to the substrate designs, which frequently involve DNA fragments lacking the upstream portion of the transcription bubble. Apparently, transcription complexes containing a complete transcription bubble to showcase the thermodynamic barriers for translocation would be of more biological significance, as we show here with the σ^S-TICs.

Intriguing questions are how fast PPi release happens and how PPi release is related to active site opening and transcription translocation. For T7 RNA polymerase, it was suggested that PPi release is directly coupled with active site opening and translocation (35). For cellular RNA polymerases, PPi release was suggested to be a thermodynamically controlled quick process coupled to the RNAP conformational change that is associated with the binding of the next cognate nucleotide (32). The NTP soaking experiment presented here displays release of PPi by the σ^S-TIC crystals and shows that the active site of the σ^S-TIC remains closed after PPi release. Although PPi release appears to be associated with a small shift at the C terminus of TH1 (Fig. 4C), a movement that likely occurs during the unfolding of TH to TL, it causes neither active site opening nor transcript translocation. However, it is possible that a reversal of the observed conformational changes on the periphery of the secondary channel could be coupled to TL unfolding and active site opening.

Materials and Methods

Preparation and Crystallization of E. coli σ^S-TIC.

To form the σ^S-TIC, we used a synthetic DNA scaffold corresponding to the promoter region between positions −38 and +12 relative to the expected transcription start site (Fig. 1A). The synthetic promoter, which contains the consensus −35 and −10 hexamers and the extended −10 motif, was prepared by annealing the NT strand to an equal molar amount of the T-strand DNA that is complementary to the NT strand except for a 6-nt discriminator region (Fig. 1A). The σ^S-TIC was assembled by directly incubating the σ^S-RNAP holoenzyme with a twofold molar excess of the preformed DNA promoter in buffer A [20 mM Tris (pH 7.5), 50 mM NaCl, 0.1 mM EDTA, 5 mM MgCl₂] at 37 °C for 20 min in the presence of ATP, GTP, and UTP (2 mM each). This mixture then was used for crystallization at room temperature by vapor diffusion with a reservoir containing 18% (wt/vol) PEG 3350, 0.1 M NaCl, and 0.1 M Hepes (pH 7.8). After the crystals grew to full size for (for about 1 wk), they were cryo-protected in the mother liquor containing 15% (wt/vol) ethylene glycol before flash-freezing in liquid nitrogen. The σ^S-TIC crystallizes in the orthorhombic P2₁2₁2₁ space group with one copy of the complex per asymmetric unit different from that of the σ⁷⁰-TICs (5).

NTP Soaking for RNA Synthesis in the Crystal.

The initially obtained σ^S-TIC crystals containing an RNA tetranucleotide were soaked in the reservoir solution supplemented with either CTP only (for a 1-nt extension) or three NTPs (CTP, UTP, and GTP, for an RNA extension of up to 3 nt) (2 mM each) at room temperature for various time periods. The crystals then were cryo-protected and flash-frozen in the same manner as aforementioned.

Data Collection, Processing, and Structure Determination.

X-ray diffraction data were collected at 100 K at the beamlines 24-ID-C and 24-ID-E at Argonne National Laboratory, Chicago, IL. All data were integrated and scaled with HKL2000 (41). The structures were solved by molecular replacement with PHASER (42) using a structure of the E. coli σ⁷⁰-TIC (5) as the starting model. The molecular replacement solution was subjected to rigid body refinement with Refmac5 (43) using multiple rigid groups, and the phases were improved by density modification. The maps were improved further by temperature factor sharpening that allowed building the σ^S factor and the nucleic acid models into the density using COOT (44). After model building in Coot, 10 cycles of TLS (translation libration screw-motion) and restrained refinement were performed using Refmac5 (43) in the CCP4 suite (45). Data collection and structural refinement statistics are summarized in Table 1. All figures were created using PyMOL (46).

SI Materials and Methods

Cloning, Expression, and Purification of E. coli σ^S Factor.

A DNA fragment encoding E. coli σ^S factor was amplified from genomic DNA of E. coli K12 and cloned into a pET21d expression vector between the NcoI and XhoI sites. The primers we used were 5′-AAACCATGGGTCAGAATACGCTGAAAGTTCATG-3′ and 5′-AAACTCGAGGAACAGCGCTTCGATATTCAG-3′. Two point mutations (S2G and R329L) were introduced for cloning purpose, and a 6xHis tag was engineered into the C terminus of the σ^S gene to facilitate protein purification. After the expression vector was transformed into E. coli Rosetta2(DE3)pLysS cells, overexpression of the C-terminal his-tagged σ^S protein (σ^S-his) was induced by the addition of 1 mM isopropyl-β-d-thiogalactopyranoside at the late log phase. Cells were harvested after overnight induction at 30 °C and were lysed using a continuous-flow French press. The σ^S-his protein was first purified by a Ni²⁺-affinity column, followed by monoQ anion exchange and size-exclusion chromatography. The purified σ^S-his protein was used directly for assembling the E. coli σ^S-RNAP holoenzyme.

Preparation of E. coli σ^S-RNAP Holoenzyme.

The E. coli RNA polymerase core enzyme that lacks the C-terminal 94 amino acids of the α subunit (Δα236–329, or ΔαCTD) was prepared as described previously (5, 47). The ΔαCTD RNAP holoenzyme containing the σ^S factor (σ^S-RNAP) was prepared by mixing the purified mutant core enzyme with purified σ^S-his protein (about 1:3 molar ratio) at room temperature for 15 min followed by size-exclusion chromatography to remove the extra σ^S-his protein. The holoenzyme was concentrated to about 30 mg/mL in buffer A [20 mM Tris (pH 7.5), 50 mM NaCl, 0.1 mM EDTA, 5 mM MgCl₂] and stored in small aliquots at −80 °C after flash freezing in liquid nitrogen.