A common feature from different subunits of a homomeric AAA+ protein contacts three spatially distinct transcription elements

Abstract

Initiation of σ⁵⁴-dependent transcription requires assistance to melt DNA at the promoter site but is impeded by numerous protein–protein and nucleo–protein interactions. To alleviate these inhibitory interactions, hexameric bacterial enhancer binding proteins (bEBP), a subset of the ATPases associated with various cellular activities (AAA+) protein family, are required to remodel the transcription complex using energy derived from ATP hydrolysis. However, neither the process of energy conversion nor the internal architecture of the closed promoter complex is well understood. Escherichia coli Phage shock protein F (PspF), a well-studied bEBP, contains a surface-exposed loop 1 (L1). L1 is key to the energy coupling process by interacting with Region I of σ⁵⁴ (σ⁵⁴_RI) in a nucleotide dependent manner. Our analyses uncover new levels of complexity in the engagement of a multimeric bEBP with a basal transcription complex via several L1s. The mechanistic implications for these multivalent L1 interactions are elaborated in the light of available structures for the bEBP and its target complexes.

INTRODUCTION

In bacteria, the multi-subunit core RNA polymerase (RNAP or E) catalyses transcription and is directed to the promoter DNA by association with a sigma factor (σ). Bacterial σ factors fall into two classes: σ⁷⁰ binds to the consensus sequences at –35 (TTGACA) and –10 (TATAAT), whereas σ⁵⁴ binds to the consensus sequences at –24 (GG) and –12 (GC). In σ⁷⁰-dependent transcription, RNAP forms a closed complex (RP_C) on the promoter that can spontaneously isomerize to an open complex (RP_O) (1). In σ⁵⁴-dependent transcription, the isomerization from RP_C to RP_O is energetically unfavourable due to the presence of a stably engaged upstream fork junction DNA around the –12 site. Within the stable RP_C, the –12 fork junction is evident (2), although the downstream DNA melting has not occurred and the +1 transcription start site is outside of the RNAP at this stage (3). The stable RP_C is thought to be preceded by an unstable RP_C in which the –12 fork junction has yet to form (4). The isomerization from RP_C to RP_O requires ATP hydrolysis by bacterial enhancer binding proteins (bEBPs), resembling in part the eukaryotic Pol II system that utilizes TFIIH and ATP for DNA melting (5). σ⁵⁴-dependent transcription not only regulates various adaptive responses (6,7), but is also responsible for regulating pathogenesis determinants in disease-causing agents such as Borrelia burgdorferi (the agent of Lyme disease) and Vibrio cholera (the agent of epidemic diarrheal disease) (8,9). Therefore, an understanding of the σ⁵⁴-transcription pathway is valuable for identification of new antibacterial drug targets (10).

Current information on the organization of σ⁵⁴-transcription complexes have been drawn from: the low-resolution Cryo-electron microscopy (Cryo-EM) studies with purified (E)σ⁵⁴–bEBP complexes (11,12), the NMR and SAXS structures of some regions of σ⁵⁴ (13–16) and crystal studies of bEBPs (12,17–20). Three regions have been identified in σ⁵⁴ (Figure 1B). Region I of σ⁵⁴ (σ⁵⁴_RI) interacts with bEBPs, core RNAP and the –12 promoter region (21–23), participating in promoter melting and isomerization processes (24,25). Region II of σ⁵⁴ (σ⁵⁴_RII) is dispensable for interactions with RNAP and DNA. Region III of σ⁵⁴ (σ⁵⁴_RIII) contains several functionally important modules, including the RpoN box required for the recognition of the –24 promoter element (22).

Figure 1.

Sequence organization and functional properties of PspF_1–275 and σ⁵⁴. (A) Domain organization of PspF_1–275. WA stands for Walker A motif, L1 for Loop 1 (containing the ‘GAFTGA’ motif), WB for Walker B motif and SRH for second region of homology. (B) Domain organization of σ⁵⁴. HTH stands helix-turn-helix motif. Xlink stands for DNA cross-linking region. σ⁵⁴ Regions I–III are separated by slashes. Available structures of two σ⁵⁴ fragments are depicted under the corresponding sequences. Six σ⁵⁴ Region I fragments (underlined in red) are generated for the following binding assays. (C) The rationale of the pBpa-based UV cross-linking assay. The ketone group in the pBpa artificial amino acid cross-links to any C–H bond within 3.1 Å under UV irradiation.

bEBPs belong to Clade 6 of the ATPases associated with various cellular activities (AAA+) protein family (26), and their functionality is dependent on an ability to self-associate, typically to form hexamers. The biochemical functions of bEBPs have been widely studied using as examples the Escherichia coli Phage shock protein F (PspF), which is involved in membrane stress responses (27), and the nitrogen control proteins NtrC and NtrC1. PspF contains an AAA+ domain (residues 1–275, PspF_1–275, Figure 1A) essential for oligomerization and ATP hydrolysis and a C-terminal DNA-binding domain. Two surface-exposed loops of PspF (Figure 1A), namely Loop 1 (L1) and Loop 2 (L2), are particularly important for coupling ATP hydrolysis to the RP_C remodelling event. L1 contains a highly conserved ‘GAFTGA’ motif amongst bEBPs. Residues Phe (F85) and Thr (T86) in the PspF ‘GAFTGA’ motif interact with σ⁵⁴_RI during RP_O formation (21,28). Structural modelling indicates that L1 may have additional roles (11). These roles may be accommodated by the potential availability of up to six L1s across a PspF hexamer (11,12,29).

To seek evidence for functional specialization amongst L1s, we incorporated a photoreactive artificial amino acid, p-benzoyl-L-phenylalanine (pBpa), using an orthogonal tRNA/tRNA synthetase pair to each PspF L1 ‘GAFTGA’ position for identification of potential interacting partners (30); pBpa can cross-link to any C–H bond within 3.1 Å upon UV irradiation [Figure 1C, (31,32)]. Here we provide direct evidence that: (i) L1 contacts the DNA non-template strand immediately upstream of the –24 promoter element, and (ii) the DNA immediately upstream of the –24 element is important for the isomerization from RP_C to RP_O as well as formation of one transcription intermediate. Using a fragmentation approach, we were able to identify two previously unknown PspF L1-binding patches within σ⁵⁴_RI (residues 18–25 and 33–39). The above observations provide evidence that L1 is multifunctional, and makes at least three distinct nucleotide-dependent interactions within its target complex in driving RP_O formation.

MATERIALS AND METHODS

Plasmids

Plasmid pPB1 [encoding the E. coli pspF_1–275 sequence, (21)] was used as a template for the subsequent site-directed mutagenesis studies. Each ‘GAFTGA’ position was mutagenized in the context of pPB1 to an amber stop codon (TAG) to yield pET28b-pspF_{1–275 variant} plasmids (Supplementary Table S1).

DNA probes and peptide fragments

The linear DNA probes used in this study are summarized in Supplementary Table S2. The σ⁵⁴_RI fragments were purchased with the highest purity level from Insight Biotechnology.

Protein expression and purification

The expression of PspF_1–275 pBpa variants depends on two plasmids: (i) the pET28b-pspF_{1–275 variant} (Supplementary Table S1) and (ii) the pDULE-pBpa [encoding the Methanococcus jannaschii tRNA/tRNA synthetase pair to specifically charge the intrinsic stop codon with pBpa, (30)]. Typically, 0.26 g of pBpa (Bachem) were dissolved under alkaline conditions and added to a L1 culture. The PspF_1–275 pBpa variants were expressed and purified as previously described (28), treated with thrombin to remove the (His)_x6 tag, and stored in TGED buffer 1 (20 mM Tris-HCl pH 8.0, 50 mM NaCl, 1 mM DTT, 0.1 mM EDTA and 5% glycerol) at −80°C.

Klebsiella pneumoniae σ⁵⁴ was purified as previously described and stored in TGED buffer 2 (20 mM Tris-HCl pH 8.0, 200 mM NaCl, 1 mM DTT, 0.1 mM EDTA and 50% glycerol) at −80°C (33). K. pneumoniae heart muscle kinase (HMK) tagged full-length σ⁵⁴ (HMK-σ⁵⁴_FL) and HMK-tagged σ⁵⁴ fragments (HMK-σ⁵⁴_RI and HMK-σ⁵⁴_ΔRI) were purified and radio-labelled as previously described (34). E. coli core RNAP was purchased from Cambio.

ATPase activity assay

Typically in a 10 μl volume, 4 μM PspF_1–275 was pre-incubated with the ATPase buffer (20 mM Tris-HCl pH 8.0, 50 mM NaCl, 15 mM MgCl₂, 0.1 mM EDTA, 10 μM DTT) at 37°C for 5 min. ATP hydrolysis was initiated by addition of 1 mM unlabelled ATP and 0.6 μCi/μl [α-³²P] ATP (3000 Ci/mmol) and incubated for various time spans at 37°C. Reactions were quenched by addition of 5 volumes of 2 M formic acid. The [α-³²P] ADP was separated from the [α-³²P] ATP by thin layer chromatography (Macherey–Nagel) in 0.4 M K₂HPO₄/0.7 M boric acid. Radioactivity was scanned by PhosphoImager (Fuji Bas-1500) and analysed by Aida software. The ATP turnover rate (k_cat) of each PspF_1–275 pBpa variant was expressed as a percentage of PspF_1–275 wild type (WT) activity. All experiments were minimally performed in triplicate.

Native gel mobility shift assay

Reactions were performed in 10 μl volumes and supplemented with 1 μM ³²P-HMK-σ⁵⁴_FL (or its fragments), ± 0.3 μM core RNAP, ± 50 nM radio-labelled DNA, 5 mM NaF and 4 mM nucleotides (ATP, ADP or AMP) in STA buffer [2.5 mM Tris-acetate pH 8.0, 8 mM Mg-acetate, 10 mM KCl, 1 mM DTT, 3.5% (w/v) PEG 8000] at 37°C for 5 min. Ten μM PspF_1–275 and 0.4 mM AlCl₃ were added for a further 15 min incubation to allow ‘trapped’ complex formation at 37°C. Complexes were either analysed on native gels or subject to UV cross-linking and then analysed.

Gel filtration assay

PspF_1-275 WT or the ADP–AlF_x ‘trapped’ complex was pre-incubated at 4°C with gel filtration buffer (20 mM Tri-HCl pH 8.0, 50 mM NaCl, 15 mM MgCl₂) for 5 min. A Superdex 200 column (10/30, 24 ml, GE Healthcare) assembled on the AKTA system (GE Healthcare) was equilibrated with buffer. Chromatography was carried out at a flow rate of 0.5 ml/min at 4°C.

pBpa-based UV cross-linking assay

‘Trapped’ complexes were formed with either ³²P-HMK-σ⁵⁴ (or its fragments) or ³²P-DNA. Reaction mixtures were UV irradiated at 365 nm on ice for 5 min, 15 min and 30 min then analysed on both native and SDS PAGE gels. The cross-linked protein–protein or nucleo–protein species were scanned by a Fuji PhosphoImager and analysed by Aida software.

Proteinase K-ExoIII footprinting assay

The UV cross-linked nucleo–protein species were generated with ³²P-DNA and subject to Proteinase K-ExoIII footprinting assays as previously described (35). The UV-irradiated samples (20 μl) were digested with 1 μl of 20 mg/ml Proteinase K (Sigma) at 37°C for 1 h to remove the protein components. ³²P-DNA containing the pBpa peptide was phenol-extracted and isopropanol-precipitated. Twenty units of exonuclease III (ExoIII, USB) were added to the DNA sample to a 10 μl final volume. The ExoIII digestion proceeded for various time spans before being quenched by 4 μl of 3X formamide stop dye (3 mg xylene cyanol, 3 mg bromophenol blue, 0.8 ml 250 mM EDTA, 10 ml deionised formamide in 10 ml). The reaction mixtures were heated at 97°C for 5 min before separated on a sequencing gel.

In vitro RP_O formation assay

Open complex formation was measured in 10 μl final volumes containing: 4 μM PspF_1–275, 100 nM holoenzyme (1:4 ratio of E: σ⁵⁴), 20 U RNase inhibitor, 5% (v/v) glycerol, 4 mM dATP and 20 nM Sinorhizobium meliloti nifH promoter (Supplementary Table S2) in STA buffer at 37°C. Transcription was activated for various lengths of time before 0.5 mM dinucleotide primer UpG, 0.2 μCi/μl [α-³²P GTP] (3000 Ci/mmol) and 0.2 mg/ml heparin were added. After extension at 37°C for 10 min, the reaction mixtures were quenched by addition of 4 μl of 3X formamide stop dye and run on a sequencing gel. The activator-bypass activities of the σ⁵⁴ variants were examined in a similar experimental procedure without the addition of PspF_1–275 activators and dATP.

In vitro spRNA assay

The ADP–AlF_x ‘trapped’ complexes were initially formed on the late-melted –10–1/WT DNA probe. Without the addition of dATP, the ADP–AlF_x complexes were allowed to synthesize a UpG-primed RNA product UpGpGpG (the spRNA) in the presence of α-³²P GTP in a manner similar to the RP_O formation assay as described above.

RESULTS

PspF_1-275 G83pBpa can cross-link to σ⁵⁴ but not to core RNAP

The photo-reactive artificial amino acid pBpa was incorporated at each L1 ‘GAFTGA’ position, generating six PspF_1–275 pBpa variants (G83pBpa, A84pBpa, F85pBpa, T86pBpa, G87pBpa and A88pBpa). Characterization of each pBpa variant for bEBP functions is summarized in Table 1. Position 83 in the PspF L1 ‘GAFTGA’ motif (the position of interest, as the following experiments demonstrate) was substituted with Ala (to remove the side chain) and Phe (to mimic the pBpa cross-linker). The resulting G83A and G83F variants were also characterized (Table 1).

Table 1.

Functional characterization of PspF_1–275 pBpa variants

PspF1–275	σ54 interaction (% of WT)		ATPase activity (% of WT)	RPO formation (% of WT)
	ADP–AIFX	AMP–AIFX
G83pBpa	50	0	10	0
A84pBpa	0	0	116	0
F85pBpa	0	0	96	0
T86pBpa	0	0	146	0
G87pBpa	0	0	138	0
A88pBpa	0	0	100	0
G83A	56	0	23	0
G83F	90	29	10	0.1

The PspF_1–275 pBpa variant and the two G83 variants (substitution with Ala to remove the side chain and substitution with Phe to closely mimic the pBpa cross-linker) were characterized in terms of σ⁵⁴ interaction (in the presence of the ADP–AlF_x ‘trapping’ reagent), ATPase activity and RP_O formation (on a super-coiled S. meliloti nifH promoter).

To investigate whether introducing the bulky hydrophobic pBpa cross-linker could affect the overall L1 exposure and the local σ⁵⁴ interaction, we performed the well-established ‘trapping’ assay. The transient L1 ‘GAFTGA’– σ⁵⁴ interaction is stabilized in the presence of an ATP transition state analogue, ADP–AlF_x. The resulting ‘trapped’ complex, PspF_1–275–(E)σ⁵⁴–ADP–AlF_x, reflects one of the intermediate states (RP_Is) en route to RP_O formation (36,37). G83pBpa was the only variant tested capable of maintaining the σ⁵⁴ interaction in an ADP–AlF_x dependent manner (Table 1 and Figure 2A). By irradiating the ‘trapped’ PspF_1–275 G83pBpa-σ⁵⁴–ADP–AlF_x complex and analysed under denaturing conditions, multiple cross-linked PspF_1–275 G83pBpa × σ⁵⁴ species were observed (Figure 2B). Prolonged irradiation shifted the cross-linked species towards higher molecular forms (Figure 2B); it is likely that higher oligomeric states of G83pBpa (G83pBpa can self-associate and self-crosslink) associated with multiple σ⁵⁴ regions (as demonstrated below). The cross-linked PspF_1–275G83pBpa × σ⁵⁴ species was only be observed in the ‘trapped’ complexes (in the presence of ADP–AlF_x but not in the presence of ATP, ADP or AMP, Supplementary Figure S1), consistent with the proposal that at the point of ATP hydrolysis, the PspF L1s assume a raised conformation to contact and thereby cross-link to σ⁵⁴ (20). The G83pBpa variant was unable to drive RP_O formation (Table 1) possibly due to sub-optimal L1 exposure (50% of WT activity) and low ATPase hydrolysis rate (10% of WT activity). However, this derivative of PspF did support partial functionalities required for forming RP_O, and was able to function for engagement of RP_C and support RP_O formation in a mixed oligomer with WT subunits (see below).

Figure 2.

PspF L1 cross-links to multiple locations within σ⁵⁴ Region I (σ⁵⁴_RI) but not to core RNAP. The HMK-tagged full-length σ⁵⁴ was radio-labelled (³²P-σ⁵⁴_FL) and used in (A–D) but not in (E). The native gel (A) and the SDS PAGE gel (B) depict the binding and cross-linking, respectively, between PspF_1–275 pBpa variants and ³²P-σ⁵⁴_FL in the presence of the ADP–AlF_x ‘trapping’ reagent. A protein band (G83pBpa, Panel A) increases with UV-irradiation time, possibly corresponding to a subpopulation of the ‘trapped’ complex due to cross-linking. The A84pBpa variant was used as a negative control for cross-linking as it cannot bind to σ⁵⁴. The cross-linked species between the G83pBpa variant and ³²P-σ⁵⁴_FL were depicted as G83pBpa × ³²P-σ⁵⁴_FL. From this point, all the following UV cross-linking reactions were performed with 15 min irradiation (the ADP–AlF_x-dependent complexes are most stable within 30 min—in this case, 15 min for complex formation and 15 min for UV irradiation). To assess the interaction and cross-linking between PspF L1 and different regions of σ⁵⁴, the ADP–AlF_x-dependent ‘trapping’ reactions were performed with radio-labelled Region I (³²P-σ⁵⁴_RI), Region I delete (³²P-σ⁵⁴_ΔRI) and mixed RI/ΔRI (a 3:1 ratio). The presence of ‘trapped’ complexes was analysed on a native gel (C); the cross-linking event was analysed on an SDS PAGE gel (D). The band in the G83pBpa × ³²P-σ⁵⁴_RI cross-linking reaction (D) indicated by an asterisk corresponds to an artefact also present in ³²P-σ⁵⁴_RI alone. To assess whether PspF L1 cross-links to core RNAP, core RNAP was added to the ADP–AlF_x-dependent ‘trapping’ reaction (E). The SDS PAGE gel was stained with Invitrogen Sypro Ruby stain and scanned by PhosphoImager. The presence of core RNAP does not seem to alter the cross-linking profile between G83pBpa and σ⁵⁴. Similar results were obtained when ³²P-σ⁵⁴_FL was used.

Based on the Cryo-EM structure, Bose et al. (11) proposed that up to three subunits in a PspF hexamer could potentially contact the RNAP holoenzyme via the PspF surface-exposed L1s. In addition, it has been shown that the AAA+ domain of the S. meliloti DctD (another well-studied hexameric bEBP) can cross-link to the core RNAP β subunit (38). To assess whether PspF L1s can directly contribute to core RNAP binding, we added core RNAP to the cross-linking reactions. As shown in Figure 2E, the cross-linking pattern between PspF_1–275 G83pBpa and σ⁵⁴ was not altered by the presence of core RNAP. The outcomes did not provide evidence to support a direct contact between the L1 ‘GAFTGA’ motif and core RNAP, rather the protein contacts appear to be primarily with σ⁵⁴.

Taken together, formation of the ADP–AlF_x ‘trapped’ complexes with the G83pBpa variant suggests it is a potentially useful reagent, as demonstrated in the following experiments, to elucidate the L1-interacting partners.

Two novel PspF L1-binding patches within σ⁵⁴_RI were identified

The Cryo-EM structure of the ADP–AlF_x ‘trapped’ complex indicated that the PspF hexamer contacted two opposing sites in σ⁵⁴ (12). If one contact site is the PspF L1 target site–σ⁵⁴_RI (36), the other contact site might be outside σ⁵⁴_RI. To explore this proposal, different radio-labelled σ⁵⁴ fragments (σ⁵⁴_RI, σ⁵⁴_ΔRI and mixed σ⁵⁴_RI/ΔRI) were used to form the cross-linked ‘trapped’ complexes with G83pBpa. Both PspF_1–275 WT and G83pBpa can form stable complexes with σ⁵⁴_FL and σ⁵⁴_RI but not with σ⁵⁴_ΔRI (Figure 2C). With the G83pBpa variant, a major cross-linked species of ∼39 kDa was observed (Figure 2D), corresponding to one PspF_1–275 G83pBpa (33 kDa) cross-linked to one σ⁵⁴_RI (6 kDa). Other faint cross-linked G83pBpa × σ⁵⁴_RI species with higher molecular weights were also observed (Figure 2D), suggesting that more than one L1 could contact σ⁵⁴_RI.

From the above observations, we set to explore the PspF L1-binding patches within σ⁵⁴_RI by generating six σ⁵⁴_RI peptide fragments (Figure 1B). As we screened the σ⁵⁴_RI fragments for their ability to stably bind PspF_1–275 hexamers under ADP–AlF_x ‘trapping’ conditions, a slower migrating complex was detected with either σ⁵⁴_Frag 18–25 or 33–39 but not the four other peptides tested (Figure 3A). Formation of the ‘trapped’ complexes in the presence of σ⁵⁴_Frag 18–25 and 33–39 was confirmed by gel filtration, providing evidence that the two peptides can associate with PspF in its ADP–AlF_x-bound state (Figure 3D). Given their relatively small size, these two σ⁵⁴_RI fragments are unlikely to assume ‘complete’ secondary structures in solution; thus their interaction with the PspF_1–275 hexamer may be largely dependent on their primary sequence and independent of their forming a well-ordered structure prior to binding to PspF. From here on, the σ⁵⁴ _RI residues 18–25 will be named Patch 1 and residues 33–39 Patch 2.

Figure 3.

Two σ⁵⁴_RI peptide fragments bind to PspF_1-275 with different affinities. (A) PspF_1–275 WT hexamers bind to two σ⁵⁴_RI fragments (residues 18–25 and 33–39) in the presence of the ADP–AlF_x ‘trapping’ reagent. From here on, fragments 18–25 is depicted as Patch1 and fragments 33–39 as Patch2. (B) Mutations in the PspF ‘GAFTGA’ motif result in sensitivity to σ⁵⁴_RI patch binding. (C) σ⁵⁴_RI Patch1 fragments bind to PspF_1–275 WT hexamers with a markedly higher affinity than Patch2 fragments. A titration experiment was performed with a constant PspF_1–275 WT concentration while the concentration of each Patch fragment was gradually increased. (D) Gel filtration profiles of the ‘trapped’ complexes at 4°C using a Superdex 200 column. The black trace corresponds to 2 µM σ⁵⁴_FL. The brown trace corresponds to 20 µM PspF_1-275 WT in the presence of ADP. The green trace corresponds to 20 µM PspF_1–275 WT in the presence of ADP–AlF_x. The red trace corresponds to 20 µM PspF_1-275 WT binding to σ⁵⁴_RI Patch1 fragments in the presence of ADP–AlF_x. The blue trace corresponds to 20 µM PspF_1-275 WT binding to σ⁵⁴_RI Patch2 fragments in the presence of ADP–AlF_x. The purple trace corresponds to 20 µM PspF_1–275 WT binding to σ⁵⁴_FL in the presence of ADP–AlF_x. The gel elution volumes corresponding to apparent hexamers, tetramers/trimers and dimers are marked by dotted lines.

To demonstrate that binding of the patch fragments to PspF_1–275 is dependent on established determinants for the binding of PspF to σ⁵⁴, the L1 ‘GAFTGA’ variants (F85Y and T86S) were used in the ADP–AlF_x reactions (Figure 3B). The F85Y variant appears to interact with the two patch fragments (the PspF band intensity increases dramatically in the presence of these fragments). However, their binding is possibly different from that of PspF_1–275 WT, as the complex band does not shift much nor into a compact band. The T86S variant can shift the complexes to the same level as WT with Patch 2 fragment but not with Patch 1 fragment. The above observations suggest that the L1 ‘GAFTGA’ motif is indeed the direct target of the σ⁵⁴ _RI patches. Based on the formation of stable ‘trapped’ complexes in a titration experiment (Figure 3C), the Patch 1 fragment exhibited a much higher affinity (at least 10-fold) towards PspF_1–275 than did the Patch 2 fragment.

Sequence alignment of the two σ⁵⁴_RI patches from different organisms (Figure 4A) revealed three highly conserved residues in Patch 1 and two highly conserved residues in Patch 2. Substitutions of these residues with Ala (QAQAAARL and AQQEAQQ, as underlined) produced variant fragments unable to detectably bind the hexameric PspF_1–275 (Figure 4B), suggesting that these residues are key to PspF L1 interactions with σ⁵⁴.

Figure 4.

Scrambled σ⁵⁴_RI Patches fail to bind PspF_1–275 WT in the context of fragments but can bind PspF_1–275 WT in the context of full-length proteins. (A) Sequence alignment of σ⁵⁴_RI Patches 1 and 2 from different bacteria using NCBI BLAST. Highly conserved residues are highlighted in red and subsequently replaced with Ala (‘Scrambled’ or ‘scm’). (B) The scrambled patches failed to bind PspF_1–275 WT hexamers in the context of fragments in the presence of ADP–AlF_x ‘trapping’ reagents. (C) The scrambled patches were able to bind PspF_1–275 WT hexamers in the context of full-length σ⁵⁴ in the presence of ADP–AlF_x ‘trapping’ reagents. σ⁵⁴_scm Patches 1 and 2 corresponds to the full-length σ⁵⁴ harbouring both the scrambled patches.

Taken together, we have identified three highly conserved residues in σ⁵⁴_RI Patch 1 contributing to the high affinity PspF L1 binding, and two highly conserved residues in Patch 2 to a lower affinity PspF L1 binding.

L1–σ⁵⁴_RI sequence-specific interactions play different roles along the activation pathway

After establishing that two σ⁵⁴_RI patches are sequence-specific for PspF L1 interactions, we next examined their impact in the context of σ⁵⁴ and RNAP holoenyzme binding interactions and different steps in the transcription activation pathway. Thus, we generated three full-length σ⁵⁴ variants: σ⁵⁴_scm Patch 1 (harbouring the scrambled Patch 1—‘QAQAAARL’), σ⁵⁴_scm Patch 2 (harbouring the scrambled Patch 2—‘AQQEAQQ’) and σ⁵⁴_scm Patches 1 and 2 (harbouring both the scrambled patches). The initial assessment of the σ⁵⁴_scm patch variants revealed no large defect in forming the ‘trapped’ complexes (Figure 4C), suggesting additional sequences of the σ⁵⁴_RI along with promoter DNA contacts (see below) could compensate for the loss of L1 sequence-specific interactions in forming the ‘trapped’ transcription intermediate (RP_I).

Burrows et al. (39) devised an assay in which the ADP–AlF_x-dependent RP_I could carry out dinucleotide-primed short RNA (spRNA) transcription when the ‘–10 to –1’ transcription bubble was pre-formed. Using this assay, we assessed the impact of the σ⁵⁴_scm patch variants on the amount of transcriptionally ‘active’ RP_Is generated. As shown in Figure 5B, despite starting with a similar amount of ADP–AlF_x-dependent RP_Is (a slightly more pronounced reduction in RP_I was observed with σ⁵⁴_scm Patch 2 and Patches 1 and 2, Figure 5A), all three σ⁵⁴_scm patch variants were able to produce significantly more spRNAs (4–8-fold) than was the σ⁵⁴ WT, indicating the RP_Is were more active than with σ⁵⁴ WT. In contrast, the ATPase-dependent RP_O formation assays replacing ADP–AlF_x with dATP on the same DNA probe revealed significant defects in all three σ⁵⁴_scm patch variants (Figure 5C). Interestingly, all three σ⁵⁴_scm patch variants can generate an RP_O in the absence of the PspF_1–275 activator and hydrolysable nucleotides on the pre-melted DNA probe (Figure 5D), so revealing an activator-bypass phenotype (40). Considering the RP_O generated from the activator-bypass activity may have contributed to the amount of RP_O as observed in Figure 5C, the defect of the σ⁵⁴_scm patch variants in the ATPase-driven isomerization may be more pronounced.

Figure 5.

Functional importance of the PspF L1–σ⁵⁴_RI patch interactions along the activation pathway. A simplified reaction scheme is depicted for each assay. Full-length σ⁵⁴ harbouring the scrambled patches and radio-labelled –10–1/WT DNA probes were used in the following reactions. The linear –10–1/WT DNA probe harbours a mismatch from –10 to –1 on the non-template strand to mimic the DNA conformation in RP_O. (A) The ability of scrambled σ⁵⁴ patch variants (σ⁵⁴_scm patch) to form the ADP–AlF_x-dependent ‘trapped’ complexes, each expressed as a percentage of that of σ⁵⁴ WT. (B) Each σ⁵⁴_scm patch variant was allowed to form the ADP–AlF_x-dependent RP_I complex (DNA–PspF_1–275 WT–Eσ⁵⁴–ADP–AlF_x). The resulting RP_I complexes were tested for their ability to support transcription in the presence of the spRNA mixture (heparin, dinucleotide primer UpG and [α-³²P GTP]). The extent of transcription activity (correlates with the amount of spRNA synthesis) was expressed as a percentage of that of σ⁵⁴ WT. (C) The amount of RP_O generated by σ⁵⁴_scm patch variants in the presence of hydrolysable nucleotide dATP, each expressed as a percentage of that of σ⁵⁴ WT. (D) The amount of activator-bypass RP_O generated by σ⁵⁴_scm patch variants in the absence of PspF_1–275 WT, each expressed as a percentage of that of σ⁵⁴ WT in the activator-dependent assay (C).

Taken together, the above data demonstrate that the PspF L1–σ⁵⁴_RI sequence-specific interactions may play an inhibitory role in the activity of RP_I, possibly to keep the complex in check before moving to RP_O. Once the inhibitory PspF L1–σ⁵⁴_RI interactions were disrupted (by scrambling the Region I patches), the spontaneous transition from RP_I to RP_O is clearly increased as seen in the activator-bypass assays. In contrast, the PspF L1–σ⁵⁴_RI sequence-specific interactions are needed for ATPase-driven RP_O formation, suggesting important roles for the patches in making RP_I from RP_C and in limiting the activity of RP_I in the activator-dependent pathway. Transient interactions between RP_C and PspF in the ATPase-driven reaction may therefore be more dependent upon the integrity of the σ⁵⁴_RI than is the stably engaged RP_I created with ADP–AlF_x.

‘Doped’ WT/G83pBpa heterohexamers can directly cross-link to promoter DNA

After demonstrating the G83pBpa variant can cross-link to σ⁵⁴ in the ADP–AlF_x ‘trapped’ complex, the DNA probe harbouring the early-melted nifH promoter (–12–11/WT, mimicking the –12 fork junction DNA in the RP_C, Figure 6) was added to the reaction mixture. If this fork junction DNA conformation was successfully accommodated in the ‘trapped’ complex, the spatial proximity between G83pBpa and the corresponding promoter region could in principle be determined.

Figure 6.

The ‘doped’ PspF_1–275 WT/G83pBpa heterohexamers cross-link to promoter DNA. The ADP–AlF_x-dependent ‘trapping’ reaction was performed in the presence of radio-labelled –12–11/WT DNA probe. The –12–11/WT DNA probe harbours a mismatch from –12 to –11 on the non-template strand to mimic the DNA conformation in RP_C. PspF_1–275 WT and G83pBpa subunits were mixed with different ratios (6/0, 5/1, 4/2, 3/3, 2/4, 1/5 and 0/6) in the ‘trapping’ reaction. Core RNAP was also added to the reaction mixture to ensure all the transcriptional components were in place. Samples were loaded on a native gel (A) and on an SDS PAGE gel (B). A single cross-linked PspF_1–275 × DNA species was observed.

As shown in Figure 6, the radio-labelled –12–11/WT DNA probe was not efficiently covalently bound into the ADP–AlF_x ‘trapped’ complexes with G83pBpa homohexamers (ratio of WT/G83pBpa was 0/6), resulting in a cross-linked species with abundance only slightly above the background. Since the PspF_1–275 WT homohexamers (ratio of WT/G83pBpa was 6/0) were able to efficiently form the ‘trapped’ complexes in the presence of DNA (Figure 6A), we reasoned that by ‘doping’ G83pBpa with different ratios of WT subunits, accommodation of the –12–11/WT DNA may be achieved. Not only did the ‘doping’ experiment successfully restore the ‘trapped’ complex formation, but also generated a single PspF_1–275 × DNA cross-linked species (Figure 6B). The above observation provides clear evidence that subunit mixing indeed occurred (see also Figure 7) and that the reconstituted WT/G83pBpa heterohexamers likely contacted the promoter DNA via position 83 or adjacent residues in the L1 ‘GAFTGA’ motif. Taken together, the above data strongly support a chemical bonding interaction between L1 ‘GAFTGA’ and the promoter DNA in the ‘doped’ WT/G83pBpa heterohexamers, as the pBpa cross-linking chemistry requires a distance of 3.1 Å (recall the H-bonding distance is 2 Å). Spatial organizations reported for RNAP-σ⁵⁴ and the L1-DNA cross-linking event together suggest that L1s from different subunits of the hexamer must be involved in DNA and σ⁵⁴_RI contacts.

Figure 7.

The ‘doped’ PspF_1–275 WT/G83pBpa heterohexamers are transcriptionally active. (A) Gel filtration profiles of PspF_1–275 WT, G83pBpa and WT/G83pBpa mixture at 4°C in the absence of nucleotides. The purple trace corresponds to 30 µM PspF_1–275 WT. The blue trace corresponds to 30 µM PspF_1–275 G83pBpa. The green trace corresponds to the theoretical profile of mixing 30 µM WT subunits with 30 µM G83pBpa subunits. The red trace corresponds to the experimental profile of mixing 30 µM WT subunits with 30 µM G83pBpa subunits. A prominent hexameric peak was observed in the experimental profile of the WT/G83pBpa mixture, corresponding to the presence of true heterohexamers. (B) The amount of RP_O formation in the presence of WT homohexamers, WT/G83pBpa heterohexamers (an equimolar ratio of WT and G83pBpa subunits) and G83pBpa heterohexamers under three different total PspF_1-275 concentrations. The –10–1/WT DNA probe was used in this assay. The average amount of RP_O generated by WT/G83pBpa heterohexamer is 55% of that of WT homohexamers. (C) Theoretical activities of the ‘doped’ heterohexamers [taken from (Werbeck et al. (42)]. One inactive subunit ‘doped’ with five WT subunits (blue), two inactive subunits ‘doped’ with four WT subunits (green), three inactive subunits ‘doped’ with three WT subunits (red), four inactive subunits ‘doped’ with two WT subunits (mint), five inactive subunits ‘doped’ with one WT subunit (purple) and six inactive subunits (brown) are necessary to abolish enzyme activity. The WT/G83pBpa heterohexamers (WT/G83pBpa subunit ratio 3/3) should theoretically generate 25% the RP_O of that of WT homohexamers (dotted lines).

Next we ‘doped’ the rest of the pBpa variants (recall all failed to form the ‘trapped’ complexes, Table 1 and Supplementary Figure S2) with the PspF_1–275 WT subunits for DNA cross-linking complementation. Strikingly, the ‘doped’ WT/T86pBpa heterohexamers showed a comparatively strong DNA cross-linking signal with the –12–11/WT DNA probe (Supplementary Figure S2B). This outcome implies a role for the conserved residue T86 or residues adjacent to it in DNA contact in RP_C. Previously, T86 has only been characterized as being a σ⁵⁴-contacting residue of L1.

To assess whether the ‘doped’ WT/G83pBpa heterohexamers were biologically relevant to the bona fide WT homohexamers, we examined the activity in the context of their self-association and RP_O formation by mixing an equimolar amount of PspF_1–275 WT and G83pBpa subunits (Figure 7). By gel filtration of apo forms, we establish that PspF_1–275 WT exists as a mixture of apparent tetramers/dimers (12 ml/13.3 ml elution volumes, Figure 7A purple trace) at a 30 µM injection concentration. The G83pBpa variant exists predominantly as apparent octamers/tetramers (10.7 ml/12 ml elution volumes, Figure 7A blue trace) at a 30 µM injection concentration. The ‘doped’ WT/G83pBpa mixture generated an apparent hexameric peak (11.2 ml elution volume, Figure 7 red trace), eluting at the same volume as the WT homohexamers [11.18 ml elution volume, (28,41)]. This apparent hexameric peak was absent in the theoretical sum of each individual subunit profile (Figure 7 green trace). We thus conclude that the ‘doped’ WT/G83pBpa heterohexamer is very similar in overall geometry to the WT homohexamer and subunit mixing indeed occurred. We also tested the ability of the WT/G83pBpa heterohexamer to form RP_O under three different total concentrations (Figure 7B). The ‘doped’ WT/G83pBpa heterohexamers generated on average 55% the RP_O of the WT homohexamers (Figure 7B). Based on the statistical model for mixing experiments (42), an equimolar amount of WT and an inactive variant (in this case G83pBpa) should theoretically generate ∼25% the RP_O of the WT homohexamers (Figure 7C dotted lines). The above considerations imply that the reconstituted WT/G83pBpa heterohexamers are indeed active in RP_O formation, and that the G83pBpa subunits contribute to this activity.

L1 cross-links to the non-template ‘–29’ region in RP_C/RP_I

To determine the precise DNA region cross-linked by L1, we employed a Proteinase K-ExoIII footprinting method (35). The rationale of this approach is to remove all the protein components by Proteinase K after UV irradiation; a stably cross-linked pBpa peptide will remain attached to the DNA cross-linking site and physically block the read-through of ExoIII (a 3′–5′ exonuclease).

The G83pBpa homohexamer is able to weakly bind and cross-link to DNA (Figure 6B). However, owing to its inability to activate transcription (Table 1 and Figure 7B), we chose the transcriptionally active ‘doped’ WT/G83pBpa heterohexamer for footprinting. The L1 cross-linking site was mapped initially on the –12–11/WT DNA probe as this gave a very clear cross-linking signal (Supplementary Figure S2B). An ExoIII-resistant site was observed from approximately –30 to –27 on the non-template strand but not observed on the template strand (compare Figure 8A with 8B). The above data clearly demonstrate that in RP_C/RP_I, a L1 contacts the non-template strand of the promoter region between –30 and –27 (abbreviated as the ‘–29 region’), immediately upstream of the consensus ‘–24’ GG element (located at –26–25 in the nifH promoter). Removal of the entire upstream sequence of the consensus GG yielded a near 60% reduction in L1-DNA cross-linking, further confirming the ‘–29’ region is the major target site of L1 ‘GAFTGA’ interaction (Figure 8C).

Figure 8.

L1 cross-links to the non-template ‘–29’ region of the –12–11/WT DNA probe. (A and B) PspF_1–275 WT homohexamers, WT/G83pBpa heterohexamers and G83pBpa homohexamers were allowed to form the ‘trapped’ complexes with Eσ⁵⁴ in the presence of ADP–AlF_x. The ‘trapped’ complexes were then subject to Proteinase K-ExoIII footprinting assays. Proteinase K removes all the protein components. The pBpa cross-linker in theory should remain attached to the promoter site and subsequently blocks the read-through of ExoIII nuclease. The cross-linking site is depicted by a half bracket from –27 to –30 (collectively called the ‘–29’ site). Both the template strand, (B) and the non-template strand (A) were radio-labelled on the –12–11/WT DNA probe (³²P as black dots). (C) Nicking and truncation of the ‘–29’ region reduced the L1 cross-linking efficiency. DNA strand nicking/truncation around the non-template ‘–29’ region was made on the –12–11/WT DNA probe to assess the impact on UV cross-linking and DNA binding. Sequential nicking of the non-template ‘–29’ region reduced the UV cross-linking efficiency by ∼30–40% (top panel). The reduction in UV cross-linking was not due to unfavourable DNA binding (DNA binding was largely unaffected, middle panel), except for the –12–11(Δ–60–27)/WT(Δ–60–27) DNA probe where the entire upstream region of the consensus GG was removed. Nicking/truncation may cause the linear DNA probes to adopt slightly different confirmations to the unmodified probe (e.g. secondary sites of cross-linking), which could explain the absence of complete abolishment in UV cross-linking.

The ‘–29’ region is important for activator-dependent RP_O formation

We next addressed whether the ‘–29’ region was important for isomerization from RP_C to RP_O and forming ‘trapped’ complexes to make RP_I. The RP_O formation assays were performed and the amount of RP_O formed from the long DNA probe was compared with that from the short DNA. Both DNA probes harboured a mismatch from –10 to –1 on the non-template strand since this DNA conformation gave the strongest activation signal amongst the three linear probes used in this assay. As shown in Figure 9, although the RP_C formation with the probe lacking DNA upstream of ‘–29’ was reduced by 2-fold (Figure 9 and Supplementary Figure S3), ‘trapped’ complexes (RP_I) with activators were reduced by 18-fold (Figure 9 and Supplementary Figure S4A) and activator-dependent RP_O formation was reduced by 35-fold (Figure 9 and Supplementary Figure S4B). Since truncation of the ‘–29’ region did not reduce the stability of RP_O by more than 2-fold (Supplementary Figure S6), the large RP_O formation defect cannot be attributed to an unstable RP_O generated from the shortened DNA probe. We conclude that the cross-linking of ‘–29’ region DNA to L1 of the activator is important for forming RP_I and RP_O. Parallel experiments with fully duplexed probes confirmed the importance of the ‘–29’ sequence for trapping and activator-dependent formation of RP_O (Supplementary Figure S5).

Figure 9.

Truncation of the ‘–29’ region affects both activator-dependent and -independent transcription activation. The –10–1/WT and –10–1(Δ–60–27)/WT(Δ–60–27) DNA probes were used in the assays to assess the impact of truncation of the ‘–29’ region on activator-dependent RP_C/RP_I/RP_O formation and on activator-independent RP_O formation. The activator-dependent RP_C formation was assessed by the stability of DNA–Eσ⁵⁴ WT complexes (Supplementary Figure S3). The activator-dependent RP_I formation was assessed by the stability of DNA–Eσ⁵⁴ WT–PspF_1–275 WT–ADP–AlF_x ‘trapped’ complexes (Supplementary Figure S4A). The activator-dependent RP_O formation was assess by the amount of RP_O generated by DNA-Eσ⁵⁴ WT- PspF_1-275 WT (Supplementary Figure S4B and Supplementary Figure S7 left panel). The activator-independent RP_O formation was assessed by the amount of RP_O generated by DNA-Eσ⁵⁴ R336A (Supplementary Figure S7 right panel).

After showing that the activator-dependent RP_O formation (with PspF_1–275 WT-σ⁵⁴ WT) involves the interaction between L1 ‘GAFTGA’ motif and the ‘–29’ region, we addressed whether the L1–DNA interaction was important for activator-independent RP_O formation. Residue R336 of σ⁵⁴ is located in the DNA cross-linking sequence of Region III (Figure 1B). The σ⁵⁴ R336A induces an activator-bypass phenotype (40) where the RP_O formation is independent of PspF hexamers and activating nucleotides on a super-coiled or on a linear –10 to –1 pre-opened nifH promoter.

In contrast to the activator-dependent RP_O formation (with PspF_1–275 WT-σ⁵⁴ WT) which was greatly reduced when the ‘–29’ region was truncated (a 35-fold reduction on average, Figure 9 and Supplementary Figure S7), the activator-independent RP_O formation with σ⁵⁴ R336A remained relatively constant on both DNA probes and was only reduced by 8-fold when the ‘–29’ region was removed (Figure 9 and Supplementary Figure S7). The differences in RP_O formation were not due to different affinities of Eσ⁵⁴ WT and Eσ⁵⁴ R336A holoenzymes towards DNA (Supplementary Figure S3). The above observations demonstrate that the activator-independent pathway for RP_O formation is 4-fold less sensitive to loss of the ‘–29’ region than is the activator-dependent pathway. Clearly, the ‘–29’ region is important for σ⁵⁴ WT-containing RNAP to form RP_O in an activator-dependent manner.

DISCUSSION

By using a fragmentation approach, we were able to identify precisely two amino acid patches within σ⁵⁴_RI responsible for PspF L1 contact (residues 18–25 and 33–39). Both σ⁵⁴_RI patches are located within the Leu heptad/hexad repeats (residues 19–44) where the activator-bypass mutations can be found (43,44). It is not known whether these two σ⁵⁴_RI patches are contacted by two L1s simultaneously or in sequence when RP_C passes to RP_I and then to RP_O. Based on the shared phenotypes of the three full-length σ⁵⁴_scm patch variants (Figure 5), it is possible the two σ⁵⁴_RI patches are contacted by two L1s from adjacent PspF subunits in a synchronized manner. This interaction with σ⁵⁴_RI may be responsible for initially holding the holoenzyme and the PspF hexamer together through interactions at the –12 part of the promoter DNA in RP_C and then organizing the RP_I to accept the melted DNA (Figure 10C). The latter reorganization may involve the LI to upstream DNA contact as described below.

Figure 10.

The proposed organizsation of L1s in engaged RP_C/RP_I. (A) L1 and σ⁵⁴_RIII might contact the promoter ‘–29’ region from different grooves. The σ⁵⁴_RIII RpoN box (blue) binds to the non-template strand of the nifH promoter (PDB 2O8K). The consensus GG element at –25 and –26 is highlighted in green. The ‘–29’ region is highlighted in red. (B) The PspF_1–275 WT hexameric structure with an open spiral (based on energy minimization of monomeric ATP-bound crystal structures, courtesy of M. Rappas) shows that the distances between the centres of two L1s across the hexameric plane are 40.4 Å [measured from subunit (i) to subunit (iii)] and 48.3 Å [measured from subunit (i) to subunit (iv)], consistent with the distance between the boundaries of the ‘–29’ element and the –12 fork junction (∼40–47 Å). (C) The proposed model based on the cross-linking data. The L1–DNA and L1–σ⁵⁴_RI cross-linking sites are depicted by the cyan and yellow stars, respectively. Alternate subunits of the PspF_1–275 open spiral are highlighted by different depths of green.

Two structural features of the RP_C are thought to impede spontaneous RP_O formation: (i) the σ⁵⁴_RI which interacts with core RNAP and blocks DNA entry, functionally reminiscent to σ⁷⁰_1.1 in RP_C (45–48). (ii) The –12 DNA melting site which is modelled in an upstream position with respect to its place in RP_O and is, therefore, misaligned with the active channel of the holoenzyme (11). We speculate that the contact between L1 and the upstream promoter DNA facilitates RP_I and RP_O formation. Although both L1 and σ⁵⁴_RIII contact the non-template ‘–29’ region (Figure 8 and Figure 10A), they do not appear to contact one another (Figure 2C). L1 and σ⁵⁴_RIII may access the ‘–29’ region from different DNA grooves to hold the promoter DNA firmly in place (Figure 10A). Doucleff et al. (14) proposed that σ⁵⁴_RIII may interact with the β G flap. Thus, nucleotide-dependent conformational changes directed by L1 may facilitate the re-alignment of the holoenzyme with the –12 DNA melting site via the proposed σ⁵⁴_RIII–β G flap interaction. In principle, reorganization of σ⁵⁴_RIII triggered by L1 movement could be transmitted to σ⁵⁴_RI and result in disruption of the previously established inhibitory interactions at the –12 fork junction DNA maintained by σ⁵⁴_RI. An LI-directed torsion generated on σ⁵⁴_RI may facilitate the propagation of DNA melting from –12 to the start site. Thus, disruption of the L1–DNA interaction around the ‘–29’ region could partially contribute to the activation defect as observed when this upstream interaction cannot be established (Figure 9). As PspF progresses through cycles of ATP hydrolysis, L1s contacting both the ‘–29’ region and σ⁵⁴_RI are likely to change from an extended state to a folded down state (19,20,49). These changes may move the DNA downstream to facilitate the re-alignment of the –12 fork junction with the active site of the holoenyzme (11). This is analogous to the DNA threading observed in AAA+ helicases by the ‘staircasing’ ssDNA-binding hairpins (50).

An open spiral hexameric configuration has been observed in many AAA+ proteins (51–56). Joly et al. (29) proposed that the PspF_1-275 hexamer also assumed an open spiral configuration employing at least two functional L1s for σ⁵⁴ binding. Thus we incorporated the PspF_1–275 spiral into a proposed transcription complex organization within RP_C/RP_I (Figure 10C). Structural analyses indicate that the distances between the centres of the base of L1 across the PspF_1–275 spiral are 40.4 Å [measured from subunit (i) to subunit (iii)] and 48.3 Å [measured from subunit (i) to subunit (iv), Figure 10B]. This fits well with a 40–47 Å span between the boundary of the ‘–29’ region (position –27, contacted by a L1) and the boundary of the ‘–12’ element [position –14, contacted by σ⁵⁴_RI (57)].

We believe the underlying mechanism proposed can be extended from PspF to other bEBPs. In this context, mutation of the second Gly in the ‘GAFTGA’ motif of Salmonella typhimurium NtrC results in a ‘super’ DNA binding activity (58). The authors suggested that the ‘GAFTGA’ motif may be close to DNA in such inactive bEBP dimmers, and the binding activity may be non-specific due to the additional charge introduced (Lys in place of Gly). In this study, we provide evidence to show that the L1 ‘GAFTGA’ motif is presented for a specific and direct promoter DNA engagement within active bEBP hexamers. Another bEBP that warrants discussion is the Aquifex aeolicus NtrC1 protein. NtrC1 forms closed oligomers in solution [90% heptamers and 10% hexamers (59)]. The negative-stain EM data suggest that although heptameric NtrC1 can engage σ⁵⁴, a significant portion of the density for σ⁵⁴ is missing in the co-complex (17). Quite how far the NtrC1 heptamer–σ⁵⁴ complex might functionally deviate from the transcriptionally active complexes forming with more usual hexameric assemblies of bEBPs, such as PspF, NtrC, ZraR, DpmR, NorR and HrpR/S (18,41,60–63), is unknown. A heptameric arrangement of NtrC1 as compared to, for example, the hexameric ZraR (62) is anticipated to be distinct in terms of the details of interfacial subunit–subunit contacts, some of which are known to control the nucleotide-dependent remodelling output of PspF (64) and to precisely define sites where the ATPase can become uncoupled from remodelling by simple mutation. Although the triple L1 contacts observed in PspF is dictated by the stoichiometry and arrangement of the ring, it can still apply to NtrC1 if the exchange between heptamers and hexamers occurs frequently to allow a faithful and productive engagement of σ⁵⁴ and promoter DNA.

To conclude, our work provides clear evidence that discrete L1s make interactions with three distinct and well separated elements within RP_C/RP_I, these are the two σ⁵⁴_RI patches and the ‘–29’ promoter region. The triple contacts by a single feature of a bEBP contrast directly with many AAA+ proteins (e.g. unfoldases and helicases) that contact only either protein or DNA and the classic bacterial activators (e.g. CRP-cAMP receptor protein) that contact protein and DNA via two distinct and spatially well separated domains.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online: Supplementary Tables 1 and 2 and Supplementary Figures 1–7.

FUNDING

Biotechnology and Biological Sciences Research Council project (BBSRC) [BB/J002828/1 and BB/G001278/1 to M.B.]. Funding for open access charge: BBSRC [BB/J002828/1 and BB/G001278/1].

Conflict of interest statement. None declared.