|Home | About | Journals | Submit | Contact Us | Français|
Activators of bacterial σ54-RNA polymerase holoenzyme are mechanochemical proteins that use ATP hydrolysis to activate transcription. We have determined a 20 Å resolution structure of an activator, PspF(1-275), bound to an ATP transition state analog (ADP.AlFx), in complex with its basal factor σ54 by cryo-electron microscopy. By fitting the crystal structure of apo PspF(1-275) at 1.75 Å into the EM map we identify two loops involved in binding σ54. By comparing enhancer-binding structures in different nucleotide states and mutational analysis, we propose nucleotide dependent conformational changes that free the loops for association with σ54.
Gene expression is regulated at the level of RNA polymerase (RNAP) activity. Bacterial RNAP containing the σ54 factor requires specialized activator proteins, referred to as bacterial Enhancer-Binding Proteins (EBPs) that interact with the basal transcription complex from remote DNA sites by DNA looping (1-4). EBPs bind Upstream Activating Sequences (UAS) via their C-terminal DNA-binding domains and form higher order oligomers that use ATP-hydrolysis to activate transcription (5, 6). EBPs’ central σ54-RNAP interacting domain is responsible for ATPase activity and transcription activation (7-9) and belongs to the larger AAA+ (ATPase Associated with various cellular Activities) family of proteins (10-12). Well studied EBPs include Phage Shock protein F (PspF), nitrogen fixation protein A (NifA), nitrogen regulation protein C (NtrC), and C4-dicarboxylic acid transport protein D (DctD) (1-3, 7, 13).
PspF from Escherichia coli forms a stable oligomeric complex with σ54 at the point of ATP hydrolysis (14). PspF-ADP.AlFx alters the interaction between σ54 and promoter DNA similarly to PspF hydrolyzing ATP (15), and was thus deemed a functional hydrolysis intermediate. Activator nucleotide-hydrolysis dependent events couple the chemical energy of hydrolysis to transcriptional activation. The highly conserved and EBP-specific GAFTGA amino acid motif (Fig. S1) is a crucial mechanical determinant for the successful transfer of energy from ATP hydrolysis in EBP to the RNAP holoenzyme via σ54’s small N-terminal EBP-interacting domain (called Region I, ~ 56 residues and sufficient for PspF interaction) (1, 14, 16-18).
The lack of structural information has hindered progress towards understanding the basis of this energy transfer process required for transcriptional activation. We now present a structure-function analysis of one such system using: 1) a cryo-electron microscopy reconstruction of PspF’s AAA+ domain (residues 1-275, PspF(1-275)) in complex with σ54 at the point of ATP hydrolysis (mimicked by in-situ formed ADP.AlFx), 2) the crystal structure of apo PspF(1-275) at 1.75 Å resolution, and 3) mutational analysis.
Nano-electro spray mass spectroscopy of a PspF(1-275)-σ54 complex with ADP.AlFx established that six monomers of PspF(1-275) are in complex with a monomeric σ54, consistent with AAA+ proteins functioning as hexamers (10, 12).
The 3-dimensional reconstruction of the PspF(1-275)-ADP.AlFx-σ54 complex (~240 kDa) obtained by cryo-electron microscopy (EM) of native samples (Figs. S3 and S4) shows a PspF(1-275) hexamer interacting with one σ54. Class averages show well-defined ring-like structures displaying six subunits of PspF(1-275) within a simple ring and clear extra density for σ54 located ~15 Å above the ring (Fig. 1A, B and C). The clear hexagonal ring structure has a diameter of ~125 Å and a central pore of ~20 Å; from the side, the ring is ~40 Å in height (Fig. 1C and Fig. S5B), dimensions consistent with other hexameric AAA+ proteins (19, 20).
Viewed from the side, the hexamer appears concave with a central depression that readily accommodates the σ54 density. A 90° rotation along the 6-fold axis reveals that the σ54 density, which runs along the ring, is elongated, bent and thicker in the middle (Fig. S5A). This ‘horseshoe’ shape extra density resembles the envelope model of σ54 (21). The estimated mass of σ54 is ~30 kDa, significantly higher than the 6 kDa Region I and less than the 54 kDa σ54. Therefore, we postulate that though we see more than just Region I, which is sufficient for PspF binding, there are parts of σ54 that are not visualized in our reconstruction because they are mobile (22).
To confirm the presence and integrity of σ54 in the particles, we marked N and C-termini of σ54 using nanogold beads. Single-cysteine σ54 constructs (46Cσ54 and 474Cσ54, (23)) were covalently linked to nanogold beads before forming the PspF(1-275)-ADP.AlFx-σ54 complex. The negative stained samples were then analyzed using electron microscopy. Detection of nanogold beads in both experiments confirmed the presence and the integrity of σ54 in the complex (Fig. 1D, (1) and (2)).
When displayed at lower contour levels the 3D electron density map shows weak densities connecting the PspF(1-275) ring to σ54 (see arrow in Fig. 1C). Based on earlier biochemical results (14, 17), we postulate that the connecting densities found almost at right angle to the PspF(1-275) ring (Fig. 1C) mark the location of certain GAFTGA motifs within the PspF(1-275) hexamer in stable association with Region I of σ54.
To facilitate localizing PspF, we determined the crystal structure of apo PspF(1-275) in space group P65 at 1.75 Å resolution using MAD phasing ((24), Table S1). The structure displays a typical AAA+ protein organisation, consisting of an α/β Rossmann fold followed by an α-helical domain (Fig. 2). The GAFTGA motif forms the tip of a loop (L1) inserted into helix3 of the α/β domain. Another loop (L2) consisting of residues 130 to 139 is inserted between helix4 and strand4 (Fig. 2). The extremities of both loops (L1 and L2) show high degrees of flexibility with the tip of L1 being the most flexible, as no electron density was observed for this region (residues 82-89).
Initially a monomer of PspF(1-275) was visually fitted into the EM map as a rigid body so that the α-helical domain sat in a ‘claw’ of the ring (Fig. 3A). This fitting positioned L1 and L2 loops in close proximity to one of the weak connecting densities contacting σ54 (Fig. 3A and B). From the fitted model we generated a hexamer and then visually readjusted individual subunits to better fit the EM map. L1 and L2 loops were either adjusted to fit the connecting densities or removed when no EM density was observed to account for them ((24) and Fig. S6A and B). It appears that at least two adjacent PspF(1-275) monomers contact one σ54 at the point of ATP hydrolysis. We infer that at the point of ATP hydrolysis, certain L1 and L2 loops extend upwards to maintain a stable interaction of PspF with σ54. When successfully engaged with σ54, L1 and L2 are more structured.
To investigate the link between nucleotide bound state and location of the GAFTGA-containing L1, we compared the apo PspF structure with the structures of different forms of Aquifex aeolicus NtrC1 (PDB code 1NY5 and 1NY6), that has 47 % sequence similarity to PspF in its AAA+ domain, including an ADP bound state thought to be incompetent for stable σ54 interaction (3). Our apo PspF(1-275) crystal structure and that of ATP-soaked crystals (data not shown) are similar suggesting that the apo-structure presented here is close to the ATP bound form which is σ54-interaction competent (14). Aligning on the conserved P-loops resulted in overall good alignment (Fig. 4A) with differences in the relative orientation of the α-helical domain to the α/β domain as well as the position of helices 3 and 4, suggesting a nucleotide-dependent change in domain relationships (12, 25).
Closer examination reveals important differences in the properties and positions of L1 and L2 (Fig. 4A, B and C). In NtrC1 (1NY6), the branch part of L1 is at a right angle to the stem part and points into the central pore of the ring, with a twisted L2 lying in close proximity (Fig. 4B, (13)). Discrete interactions stabilize the branch part of L1 in NtrC1, notably the interactions between E212 in L1 (equivalent to E81 in PspF) and R262 (R131), K267 (Q136) in L2. Also, L263 (V132) of the tip of L2, F216 (F85) of the L1 tip, and A206 (Cβ of S75) of helix3 form a hydrophobic cluster which locks the GAFTGA motif in a buried and unfavourable conformation for stable σ54 interaction (Fig. 4B; Fig. S1). Based on our EM map, in this conformation the GAFTGA motif cannot contact σ54.
In PspF(1-275), the relative rotation of helices 3 and 4 disrupts the hydrophobic interactions between F85 of L1, V132 of L2, and Cβ of S75 of helix 3 while interactions between the stem part of L1 and helix3, i.e. between H80 and S75, H92 and E76, are strengthened (Fig. 4C). We hypothesize that these changes in interactions are nucleotide-dependent. Indeed, these interactions form part of a larger network that involves residues R95, R98 of helix3 and S62 of the central β sheet which includes the Walker B motif (D107, E108) responsible for nucleotide hydrolysis (Fig. 4C). This interaction network is well suited to relay nucleotide-dependent movements of the central β sheet, which originate in the Walker B motif, to helix 3 and more importantly to L1 with its GAFTGA motif. A similar network in NtrC1 appears to be based mostly on hydrophobic interactions
To investigate the functionality of the interaction network believed to ultimately control the GAFTGA-containing L1, we conducted structure-based single amino acid mutational analyses. Disruption of the proposed network linking the ATP active site to the GAFTGA loop is predicted to cause a decrease in nucleotide-dependent activities of PspF(1-275). To examine this hypothesis we mutated H92, which interacts with E76 of helix3 (Fig. 4C), into either F, to mimic the overall geometry while eliminating the charge effects, or to R, to maintain the overall charge (R is also the most frequent residue in this position in EBPs, Fig. S1). R95 on the other hand, which interacts with both E76 and S62 of the central β sheet, was mutated into either A or to the more related K. PspF(1-275)R95A and PspF(1-275)H92F each fail in all measurable post-ATP-binding activities ((26), Fig. S1 and Table S2). In both cases the mutations severely disrupt the network of interactions and thus block any communication of the conformational signal beyond the mutated residues. In contrast, PspF(1-275)R95K and PspF(1-275)H92R, retained most of the in vitro activities measured, including transcriptional activation (Table S2). The Walker B to L1 network is consistent with properties of many mutant forms of DctD and NtrC (27, 28), supporting the existence of a common EBP nucleotide-dependent communication pathway.
To determine the functionality of the hydrophobic interactions that lock the tip of L1 in the NtrC1 structure, we mutated F85 and V132. PspF(1-275)V132A retained its nucleotide-binding activity but lacked ATPase activity. PspF(1-275)F85A likewise lacked ATPase activity, but PspF(1-275)F85L and PspF(1-275)F85W showed ATP-binding and some ATPase activity with mutation to W leading to an increase in ATPase activity (Table S2). These results suggest that the stability of the L1-L2 hydrophobic interaction has direct consequences on the ATP hydrolysis cycle. PspF(1-275)V132A can engage σ54 in an ADP.AlFx-dependent manner suggesting an effect upon Pi-release (Table S2). Disrupting this interaction would impair this stage of the ATPase cycle whereas strengthening it (e.g. PspF(1-275)F85W) would increase ATPase activity. PspF(1-275)F85W fails to stably bind σ54, demonstrating that the integrity of the GAFTGA motif is crucial for σ54 interaction (14, 17). The proposed communication pathway in PspF links changes in the ATP hydrolysis site to conformational changes in the GAFTGA-containing L1 loop that remodels the σ54-RNAP-holoenzyme to activate transcription.
We would like to thank the members of MB and XZ’s laboratories for their help and useful discussions throughout this work. We are grateful to Ingar Leiros at ESRF and beamline scientists at Daresbury for their help in data collection, we are also grateful for the support provided by the IC-CBEM. This work is supported by BBSRC funding to XZ and MB. Coordinates for the reported structures have been deposited in the Protein Data Bank (accession codes 2BJW and 2BJV). The EM map has been deposited to EMDB (accession code EMD-1109)