Structures of the RNA polymerase-σ⁵⁴ reveal new and conserved regulatory strategies

Abstract

Transcription by RNA polymerase in bacteria requires specific promoter recognition by σ factors. The major variant σ factor (σ⁵⁴) initially forms a transcriptionally silent complex requiring specialised ATP-dependent activators for initiation. Our crystal structure of the 450 kDa RNAP-σ⁵⁴ holoenzyme at 3.8 Å reveals molecular details of σ⁵⁴ and its interactions with RNAP. The structure explains how σ⁵⁴ targets different regions in RNAP to exert its inhibitory function. Although σ⁵⁴ and the major σ factor, σ⁷⁰, have similar functional domains and contact similar regions of RNAP, unanticipated differences are observed in their domain arrangement and interactions with RNAP, explaining their distinct properties. Furthermore, we observe evolutionarily conserved regulatory hotspots in RNAPs that can be targeted by a diverse range of mechanisms to fine tune transcription.

Gene transcription is a tightly regulated event. A range of factors are required to maintain RNAP in a signal responsive, but inhibited state and to target it to specific genes (1). Bacterial sigma (σ) factors, eukaryotic TFIIB and other general transcription factors, are the primary promoter recruitment factors. The major variant σ⁵⁴ (also called σ^N) is utilised in transcribing genes for numerous stress responses (2). Unlike the major σ⁷⁰ class, which recognises the -35 and -10 promoter DNA elements, the σ⁵⁴ class directs RNAP to promoter sites through the -24 and -12 DNA elements and forms a stable closed promoter complex that is unable to spontaneously melt DNA and initiate transcripts (3). Instead, the initiation process requires ATP-dependent activator proteins bound to upstream enhancer sites to actively remodel the RNAP-σ⁵⁴-DNA complex (4).

σ⁵⁴ contains four structural domains connected by long coils and loops that span a large area of the RNAP core enzyme, which consists of two α, β, β’ and ω subunits (5, 6) (Fig. 1A-C, Fig. S1A-B, Table S1). σ⁵⁴ Region I (RI, residues 1-56) forms a hook comprised of two α-helices. Region II (RII, residues 57-120) also contains two α-helices (residues 57-85) in addition to loops that are buried inside the RNAP (Fig. 1A, 1C). The RNAP core-binding domain (CBD) of σ⁵⁴ extends as a structural fold to residue 250 and consists of two α-helical subdomains. Following on from the CBD, the backbone extends back to connect to a loop region, before an extra long α-helix (residues 315-353, hereafter called ELH, spanning 50 Å) followed by the HTH domain (residues 365 – 385), involved in interaction with the -12 promoter elements (3). The domain containing the RpoN box (RpoN domain), responsible for recognizing the -24 promoter elements, consists of a three-helical bundle (residues 415 – 477) (Fig. 1A) (3). The σ⁵⁴ polypeptide chain snakes back and forth through its loop regions embedded in the RNAP. We carried out crosslinking experiments between σ⁵⁴ and RNAP in solution using a p-benzoyl-L-phenylalanine-incorporated RNAP library. The amino acids that are crosslinked to σ⁵⁴ agree well with the RNAP-σ⁵⁴ crystal structure (Fig. 1D yellow, Fig. S1C).

Figure 1. Structure of complete RNAP-σ⁵⁴.

(A) Structure and schematic diagram of σ⁵⁴. (B) Structure of σ⁷⁰ (PDB 4IGC) with similar functional domains colored accordingly. (C) Holoenzyme in different orientations, RNAP core in cylinders. β-cyan, β’-salmon pink, α-grey, ω-pale blue, σ⁵⁴ in ribbon and colored as in (A), DS represents downstream, US represents upstream. (D) Residues of β and β’ subunits that are crosslinked to σ⁵⁴ are mapped onto the structure and colored in yellow (E) Electron density map and structural model of RNAP-σ⁵⁴-DNA.

We crystallized RNAP-σ⁵⁴ in complex with a 28 bp DNA containing the -24 and -12 promoter elements. The crystals diffracted to 8 Å resolution, with the electron density for DNA being unambiguous (Fig. 1E), and the molecular envelope for the holoenzyme clear. We thus constructed a structural model of RNAP-σ⁵⁴-DNA (Fig. 1E, Fig. S1D).

All σ factors contain a major RNAP core-binding domain (CBD) and a major DNA binding domain, recognising either the -35 or -24 elements. The CBD of σ⁵⁴ (Fig. 1A) binds towards the upstream face of the RNAP (Fig. 1C, based on promoter DNA binding orientation) making extensive interactions with many functional modules within the RNAP including the β-flap (residues 835-937), the C-terminus of the β subunit (residues 1267 – 1320), the β’ zipper/Zn-binding domain (residues 35-100), the β’ dock domain (residues 370-420) as well as the α subunit carboxyl terminal domain (α-CTD) (Fig. 2A). Using FeBABE cleavage assays, residue 198 of σ⁵⁴ (within the CBD) can indeed be mapped to a region within the β-flap (7).

Figure 2. Functional domains of σ⁵⁴ and interactions with RNAP core.

(A) Detailed view of the CBD and RNAP interactions. (B) RI and RIII, (C) DNA-RNA channels, DNA-orange, RNA-magenta and (D) Overlay with RII.1, RII.2 and RII.3.

The RpoN domain of σ⁵⁴ is the most conserved domain among σ⁵⁴ from different organisms (Fig. S2). This domain extends from the main body of RNAP and does not contact other parts of σ⁵⁴ nor core enzyme (Fig. 1A, 1C). Instead, it interacts with an adjacent RNAP molecule in the crystal, suggesting that its location is flexible in solution. In the RNAP-σ⁵⁴-DNA complex model, the RpoN domain is indeed moved relative to the RNAP-σ⁵⁴ structure (Fig. S1D). Unlike the flexible RpoN domain, the RIII-HTH domain of σ⁵⁴, which binds to the -12 promoter region, stably associates with the RNAP core through RI-RIII (Fig. 1E, Fig. S1D) (8-10). Our data thus suggest that -12 binding is a major functional determinant for σ⁵⁴ holoenzyme promoter recognition as well as stable closed complex formation.

The σ⁵⁴ RI plays an inhibitory role and contains contact sites for its cognate activator proteins (11, 12). RI interacts with RIII, forming a structural module that lies along the cleft between β and β’ (Fig. 1C and Fig. 2B), where template strand DNA enters the active site cleft (13). Our structure clearly suggests that the entry of promoter DNA into RNAP is blocked by RI and RIII of σ⁵⁴. Fitting of the RNAP-σ⁵⁴ crystal structure into the cryo-EM map of activator bound RNAP-σ⁵⁴ holoenzyme (14) positions the RI helices inside the connecting density leading to the activator protein, indicating that RI in the RNAP-σ⁵⁴ structure is presented favourably to contact the activator (Fig. S3A).

RII penetrates deeply into the DNA binding channel (Fig. 1, Fig. 2C-D). RII can be divided into three sub-regions based on their locations in the holoenzyme. RII.1 (residues 57-85) occupies the downstream DNA-binding cleft of RNAP, just above the bridge helix, thus playing an inhibitory role (Fig. 1 and Fig. 2C-D, Fig. S3A right). RII.2 (residues 86-105) occupies the space of the DNA template strand (Fig. 2C-D), indicating that RII.2 has to relocate to permit template-strand DNA access into the RNAP active site and for transcription initiation. RII.3 (residues 105-120) extends along the path that is occupied by RNA in the transcribing complex (Fig. 2C-D), suggesting that it will also need to relocate upon RNA synthesis.

Comparisons between σ⁵⁴ and σ⁷⁰ holoenzyme structures reveal that overall they contact similar regions of the RNAP core enzyme (Fig. 3A-B), in agreement with the idea that specific factors unrelated by structure and sequence can be functionally similar (15). However, the relative location of functional domains is completely different. For example, the σ⁵⁴ CBD is located upstream, blocking the RNA exit tunnel (Fig. 3C) while σ₂, the major core binding domain of σ⁷⁰, contacts the downstream β’ region (Fig. 3A-B). The structural differences suggest that σ⁵⁴ CBD, and hence σ⁵⁴, will have to dissociate or relocate to allow progression from transcription initiation to elongation while σ⁷⁰ can loosely associate with the RNAP core even during transcript elongation, in agreement with earlier biochemical data (16,17). σ⁷⁰ region 4 (σ₄), which recognises the -35 promoter element, occupies the upstream surface blocking the RNA exit channel instead (Fig. 3B).

Figure 3. Comparisons of σ⁵⁴ and σ⁷⁰ holoenzymes.

(A) RNAP-σ⁵⁴ with RNAP displayed as spheres and σ⁵⁴ as surface. (B) RNAP-σ⁷⁰, same view as in (A). (C) Elongation complex (PDB code 2O5J).

The inhibitory RI of σ⁵⁴ interacts with RIII across the RNAP cleft, blocking the template strand from entering. On the other hand, σ_1.1 of σ⁷⁰, which also has an inhibitory role (18), is located in the downstream DNA channel, overlapping with the region occupied by σ⁵⁴ RII.1 (Fig. 3), suggesting that σ_1.1 and RII.1 might play related roles in transcription regulation. However, σ_1.1 connects with σ₂ by only a single flexible linker, allowing the σ_1.1 to move readily from the RNAP cleft (Fig. 1B, Fig. 3B). σ⁵⁴ RII.1 is connected to the RI and hence the RIII-ELH as well as the CBD and the linker that connects RII.1 with the CBD is embedded in the DNA/RNA channel of RNAP (Fig. 2C-D). Due to its topological restraint, σ⁵⁴ RII.1 cannot relocate easily without affecting the conformation of other parts of σ⁵⁴, RII.1 will thus block downstream DNA entry.

Notable conformational differences exist between the σ⁷⁰ and σ⁵⁴ holoenzymes in many key modules within the RNAP when superposed on the bridge helix, chosen since the catalytic centre must be conserved (Fig. S3B). The β’ coiled-coil, and the whole β’ subunit, is rotated, narrowing the downstream DNA channel by as much as 5 Å (Fig. S3B). However, the clamp, which consists of β and β’ domains forming the walls of the downstream DNA channel, was shown to adopt multiple conformations in solution (19). The β’ coiled-coil forms helical bundles with σ₂ of σ⁷⁰ and helps to stabilise the unpaired non-template DNA strand, and hence the transcription bubble (20) (Fig. S4A). B-linker of TFIIB also interacts with β’ coiled-coil (Fig. S4B). In the σ⁵⁴ holoenzyme structure, the β’ coiled-coil does not form a helical bundle with σ⁵⁴ (Fig. S4C). However, it is possible that en-route to transcription initiation, this interaction is established, in agreement with our biochemical data (Fig. S4D).

σ⁵⁴ occupies multiple locations along the DNA and RNA path within the RNAP core. These locations are also targeted by other regulatory factors and in other RNAPs. For example, RI and RIII sit across the RNAP cleft, blocking the DNA template strand entry (Fig. 4A). At similar locations, domains 2 and 3 (σ₂ and σ₃) of σ⁷⁰ form a V-shaped wedge that acts as a gate to allow the template strand to enter the active cleft (Fig. 4B) (13, 21). TFIIB also uses two sub-domains, the core binding domain and the B-linker, to occupy similar areas in Pol II (Fig. 4C). This area is thus occupied by σ⁵⁴ to inhibit while by σ⁷⁰ and TFIIB to guide the template strand delivery. Furthermore, σ₂ of σ⁷⁰ contains a number of aromatic residues that are shown to interact with non-template strand, therefore facilitating DNA melting and transcription bubble formation (20). In σ⁵⁴, there are few aromatic residues in this region (RI and RIII-ELH) that could facilitate DNA melting, in agreement with data showing that although deletion of σ⁵⁴ RI can bypass the requirement of activator proteins, it fails on double-stranded DNA (11). Activator proteins may therefore be directly required for transcription bubble opening. The downstream DNA channel is occupied by both RII.1 of σ⁵⁴ and σ_1.1 of σ⁷⁰ (Fig. 4D). A negatively charged region of TFIIF is also located in the downstream DNA channel of Pol II (22). RII.2 of σ⁵⁴ is located in the channel that is occupied by the template strand DNA in the transcribing RNAP (Fig. 2D). σ⁷⁰ and TFIIB lack equivalent structural features. This area, however, is occupied in Pol I by part of the expander (Fig. 4E), a unique insertion in Pol I compared to Pol II and shown to stabilize an expanded active centre cleft (23, 24) (Fig. 4E). Similar to RII.3, both region 3.2 in σ⁷⁰ (σ finger) and the B-reader in TFIIB are shown to overlap with the space that would be occupied by RNA (Fig. 4D-F) (25). Indeed, RII.3 shares sequence homology with σ, especially the highly conserved DDE motif (Fig. S2). Acidic residues in these elements are proposed to facilitate template DNA loading and RNA separation from DNA, guiding it towards the exit channel (25-28). It is possible that RII.3 also performs similar roles.

Figure 4. Regulatory elements of σ⁵⁴ and comparison with σ⁷⁰, TFIIB and Pol I.

(A) σ⁵⁴ (B) σ⁷⁰ and (C) TFIIB in relation to template strand (gold) as defined in the complex with a full transcription bubble (4YLN). (D-F) Comparison of σ⁵⁴ with σ⁷⁰, Pol I (4C2M) and TFIIB (4BBS) in the RNAP DNA-RNA channel; structures are superposed on the bridge helix. σ⁵⁴ represented as spheres while σ⁷⁰, TFIIB and Pol I expander as ribbons, RNAPs as cylinders.

Our structure uncovers a diverse range of regulatory strategies utilised by σ⁵⁴ in tightly controlling transcription initiation. These include blocking the template strand from entering the RNAP active cleft by RI and RIII, occupying the downstream DNA channel by RII.1, interfering with the template strand and synthesized RNA by RII.2-RII.3. The position of RI and RIII plays a vital role in these regulatory functions. Activator proteins interact with RI and could relocate RI, RIII-ELH as well as RII, releasing the inhibition posed by some of these structural elements (Fig. S3A) (29, 30). Precisely how activators overcome these multiple modes of inhibition, and the role of ATP in this process, remains to be determined. Furthermore, we show that although σ⁵⁴, σ⁷⁰, TFIIB, TFIIF and Pol I subunits have no sequence or structural similarities, they target the same elements within their respective RNAPs. Our structure and comparisons thus provide clear evidence that regulatory hotspots within RNAPs are functionally conserved. Different transcription factors, irrespective of their structures and sequences, are utilised to target these hotspots in order to exert a fine-tuned transcription regulation strategy (Fig. S5). It will be interesting to see the extent to which additional multi-subunit RNAPs employ these regulatory strategies.