RNAPs (RNA polymerases) are complex molecular machines that contain a highly conserved catalytic site surrounded by conformationally flexible domains. High-throughput mutagenesis in the archaeal model system Methanocaldococcus jannaschii has demonstrated that the nanomechanical properties of one of these domains, the bridge–helix, exert a key regulatory role on the rate of the NAC (nucleotide-addition cycle). Mutations that increase the probability and/or half-life of kink formation in the BH-HC (bridge–helix C-terminal hinge) cause a substantial increase in specific activity (‘superactivity’). Fully atomistic molecular dynamics simulations show that kinking of the BH-HC appears to be driven by cation–π interactions and involve amino acid side chains that are exceptionally highly conserved in all prokaryotic and eukaryotic species.
RNAPs (RNA polymerases) underpin all gene-expression activities in prokaryotic and eukaryotic cells and are therefore the focus of numerous studies designed to reveal the molecular details of their functions. On the basis of the longstanding traditions of genetics and biochemistry, studies of bacterial RNAPs are currently most advanced and have provided many key insights into the mechanisms of gene expression. It is clear, however, that eukaryotic transcription processes differ in many respects. Owing to the high degree of complexity, involving a myriad of additional factors, the eukaryotic transcriptional machineries pose many challenges that are only partially met by the currently existing technologies. Since the archaeal transcriptional apparatus closely resembles the core eukaryotic RNAPII initiation complex, such model systems provide important insights into structure–function relationships operating at the heart of the eukaryotic machinery .
We have succeeded in reconstituting the entire archaeal transcriptional machinery from recombinant proteins, providing us with an experimental tool to dissect fundamental processes, such as transcription initiation and elongation, at a level of detail unimaginable in any eukaryotic system [2–4]. Recent efforts, focusing on automation of the in vitro assembly and transcription assay procedures, have allowed the systematic analysis of the structure–function relationships in RNAP domains using hundreds of site-directed mutants [5,6]. Such large datasets shed new light on the structure–function relationships operating within RNAPs, as well as the regulatory roles of accessory transcription factors directing events during transcript initiation, elongation and termination.
In the present review, we discuss new insights gained from a large-scale mutagenesis study of the bridge–helix domain found in the catalytic centre of RNAPs from all three evolutionary domains . A novel approach, based on high-resolution modelling of the structural properties of protein domains using computer simulation, complements the experimentally obtained data to reveal new insights into a fundamental enzymatic mechanism.
Structural and functional evidence for bridge–helix kinking
The bridge–helix domain is a 35-amino-acid α-helix that spans the catalytic site and plays a key role during the NAC (nucleotide-addition cycle) (Figure 1A). Mutations of bridge–helix residues alter the specific activity of RNAPs, ranging from total loss of activity to a near-doubling of the catalytic rate (‘superactivity’) . The bridge–helix is structurally flexible and isomerizes between a ‘straight’ and a ‘kinked’ conformation. A model involving repetitive interconversion between these two bridge–helix conformations suggests an attractive mechanism for translocating RNAPs along the template DNA in single base-pair steps during RNA synthesis [7–10].
Bridge–helix structure and conformation
High-resolution insights into different bridge–helix conformations have thus far only been obtained for an RNAP from the bacterium Thermus thermophilus (‘tth’); in one crystal structure, the bridge–helix is in an essentially straight conformation  (Figure 1A), whereas another structure contains a kinked helix  (Figure 1B). The kink is located near the C-terminus of the bridge–helix and appears to be created via electrostatic interactions between two residues, tthβ' Asp1090 and Arg1096  (Figure 1B). The kinked region displays the properties of a reversible molecular hinge and we will therefore refer to this sequence as the BH-HC (bridge-helix C-terminal hinge) [there is also evidence for a second hinge in the N-terminus of the bridge–helix, which necessitates a distinction between the two hinges via the subscript letter (R.O.J. Weinzierl, unpublished work)].
It is very likely that BH-HC-mediated bridge–helix kinking is a fundamental process shared by all multisubunit RNAPs. The most striking evidence comes from a high-throughput mutagenesis screen of the Methanocaldococcus jannaschii (‘mj’) bridge–helix that included a large number of single and multiple amino acid substitutions located within and around the BH-HC. In particular, a proline substitution (mjA′ S824P) in the BH-HC caused a substantial increase in activity, suggesting that increased flexibility of the BH-HC correlates directly with a higher rate of the NAC , as proline destabilizes α-helical structures .
Inspection of BH-HC sequence motifs from a cross-section of bacterial, archaeal and eukaryotic bridge–helices, however, reveals an intriguing puzzle. For example, despite an exceptionally high degree of sequence conservation in the BH-HC region, the Escherichia coli bridge–helix contains an asparagine residue in the position occupied by aspartic acid residue in T. thermophilus, and therefore cannot form the electrostatic bridge found in the kinked tth BH-HC conformation (Figure 1C). A more comprehensive alignment of BH-HC sequences shows that this is not an unusual occurrence (Figure 1D). Although one of the electrostatic partners (orthologous to tthβ′ Arg1096) is absolutely conserved in all species, its potential interaction partner (equivalent to tthβ′ Asp1090) is often either uncharged (e.g. glutamine, asparagine, threonine or isoleucine), or in some cases even positively charged (arginine or lysine) (Figure 1D).
We therefore face an enigma: according to our current understanding, only a subset of species appears to encode in their bridge–helix primary sequence the ‘correct’ configuration of charged residues that allow BH-HC kinking to occur. Considering the strong structural and functional evidence documenting the relevance of BH-HC, we wondered whether a model, focusing on the role of an electrostatic bridge to account for BH-HC kinking, overlooks as yet undefined structural features of the BH-HC structure that play an even more important role in inducing the kinked BH-HC conformation.
BH-HC kinking is stabilized by highly conserved cation–π interactions
In order to address this question in more detail, we carried out MD (molecular dynamics) simulations of the bridge–helix domain. MD studies provide useful insights into the dynamic properties of protein domain structures by simulating motions on an atomic scale with exceptionally high temporal and spatial resolution . The M. jannaschii bridge–helix conforms to the general archaeal/eukaryotic sequence conservation pattern and can therefore be considered as a representative example; in particular, the M. jannaschii BH-HC region is not capable of forming the electrostatic bridge owing to the presence of an uncharged glutamine residue (mjA′ Gln823) in the position orthologous to tthβ′ Asp1090 (Figure 1C). Any kinking movements in the simulated BH-HC would therefore be due to other properties intrinsic to its structure.
We initially modelled the orthologous M. jannaschii sequence on the bridge helix of yeast RNAPII, which was crystallized in the presence of nucleic acid and rNTP substrates and is therefore considered to represent an active elongation complex (PDB code 2E2H) . A series of 27 independent 200 ps atomistic simulations of the modelled M. jannaschii bridge helix was performed using GROMACS  in a fully solvated and ion-populated environment (Figure 2A). Changes in secondary structure were monitored at 5 ps intervals with STRIDE , enabling us to detect any conformational changes in various parts of the bridge–helix in an unbiased and systematic manner.
MD Simulation of the BH-HC kinking mechanism
As expected from a stochastically occurring molecular event, several simulations revealed bridge–helix structures that were distinctly kinked in the BH-HC region. A more detailed investigation of these kinked BH-HC conformations revealed that they were consistently caused by an intramolecular interaction between a tyrosine residue (mjA′ Tyr826) and two arginine residues (mjA′ Arg829 and Arg830). Such interactions are generally known as cation–π interactions and are based on the electrostatic attraction between the negative quadrupole generated by the π electron cloud of aromatic amino acids and positively charged amino acids, such as lysine and arginine [18,19]; interactions between arginine and tyrosine residues, as observed in the bridge–helix, are indeed energetically highly favourable [20,21].
Closer inspection of the trajectories of bridge–helix structures undergoing kinking in the BH-HC region reveals a wealth of mechanistic insights. Quantitative analysis of BH-HC unfolding simulations shows that a glycine residue, mjA′ Gly825, plays a central role (Figure 2B). Glycine residues display low helix-forming propensity because their high conformational flexibility is entropically unfavourable within geometrically constrained α-helical structures . The importance of mjA′ Gly825 has already been noted in a previous study which has demonstrated that this residue cannot be substituted for by any other amino acid without substantial loss of activity . The glycine residue is, however, not the only critical component of the BH-HC kinking mechanism. In the straight bridge–helix conformation, a tyrosine residue (mjA′ Tyr826) is ideally placed to contact the arginine residues (mjA′ Arg829 and Arg830); owing to fundamental α-helical geometry, the tyrosine and arginine side chains face in the same direction and are only separated by a single turn (Figure 2C, left-hand panel). The flexibility of the long arginine side chains establishes cation–π interactions between mjA′ Tyr826 and Arg829/Arg830 (Figure 2C, middle panel), which are stabilized further through extensive van der Waals contacts between mjA′ Arg830 and Met827 (Figure 2C, right-hand panel). It therefore appears that the cation–π interactions between the side chains of residues mjA′ Tyr826 and Arg829/Arg830 deprive mjA′ Gly825 of intrahelical interactions at its C-terminus, causing local unfolding and kinking of the bridge–helix in a reversible manner. Unlike the kinking model based on the formation of an electrostatic bridge, which is not possible in many species, the residues required by the model proposed here are either universal (mjA′ Gly825, Tyr826 and Arg829), or are exceptionally highly conserved (mjA′ Met827 and Arg830) (Figure 1D) in all prokaryotic and eukaryotic species. The new model is strongly supported by previously published mutagenesis data, showing that any substitutions with other amino acids for mjA′ Gly825, Tyr826, Arg829 and Arg830 cause a severe fall in the catalytic rate of RNAP . The postulated stabilizing effect of mjA′ Met827 via extensive van der Waals interactions with the kinked BH-HC region is also documented by observations showing that other large non-polar aliphatic side chains, such as valine and leucine, can replace Met827 without loss of activity .
MD simulations show that the evolutionarily highly conserved features of the BH-HC region encode a molecular hinge capable of undergoing kinking through cation–π interactions. The kinking motion almost certainly involves a multistep mechanism, progressing from short-lived early stages (initial contacts between tyrosine and arginine residues) towards a more stable structure involving specific alignments of tyrosine, arginine and at least one further large aliphatic amino acid (such as leucine, isoleucine or methionine). The distortions introduced into the α-helical geometry, in conjunction with a helix-destabilizing glycine residue located immediately N-terminal to the tyrosine residue, create a localized and structurally fully reversible kink in the bridge–helix.
From an energetics point of view, it is likely that the cation–π interaction is the most fundamental driving force of BH-HC kinking. In some species, the kink may then be stabilized or positioned, through additional, but not universal, electrostatic interactions. Experimental support for this hypothesis comes from substitutions of acidic amino acids for mjA′ Gln823 (i.e. mjA′ Q823D or mjA′ Q823E), which increase the catalytic rate to superactivity, suggesting that kinking of the M. jannaschii BH-HC sequence can be enhanced through the formation of such an electrostatic bond .
The results shown in the present paper demonstrate the power of MD simulations to explore the structural basis of protein domain functions. Essentially all conformational changes in molecular machines are based on non-covalent interactions that are often very short-lived and energetically unstable. We have demonstrated that alternative approaches, based on a combination of high-throughput mutagenesis and MD, provide additional tools that complement the snapshots of RNAPs caught in immobilized forms in crystals and thus provide new insights into the molecular mechanisms underlying the transcription process.
Molecular Biology of Archaea II: A Biochemical Society Focused Meeting held at Robinson College, Cambridge, U.K., 16–18 August 2010. Organized and Edited by Stephen Bell (Oxford, U.K.) and Finn Werner (University College London, U.K.).
We thank Simone Wiesler, Patricia Burrows and Martin Buck for a critical reading of the paper. We acknowledge the use of the U.K. National Grid Service in carrying out this work.
Research was funded by project grants from the Biotechnology and Biological Sciences Research Council [grant numbers BB/E000975/1 and BB/D5230001/1], the Medical Research Council [grant numbers G0501703] and the Wellcome Trust [grant numbers 078043/Z/05/Z] to R.O.J.W.