Portal de Periódicos da CAPES

Structural basis for inhibition of homologous recombination by the RecX protein

2008; Springer Nature; Volume: 27; Issue: 16 Linguagem: Inglês

10.1038/emboj.2008.145

1460-2075

Stefania Ragone, Joseph D. Maman, Nicholas Furnham, Luca Pellegrini,

Enzyme Structure and Function

Article24 July 2008Open Access Structural basis for inhibition of homologous recombination by the RecX protein Stefania Ragone Stefania Ragone Department of Biochemistry, University of Cambridge, Cambridge, UK Search for more papers by this author Joseph D Maman Joseph D Maman Department of Biochemistry, University of Cambridge, Cambridge, UK Search for more papers by this author Nicholas Furnham Nicholas Furnham Department of Biochemistry, University of Cambridge, Cambridge, UK Search for more papers by this author Luca Pellegrini Corresponding Author Luca Pellegrini Department of Biochemistry, University of Cambridge, Cambridge, UK Search for more papers by this author Stefania Ragone Stefania Ragone Department of Biochemistry, University of Cambridge, Cambridge, UK Search for more papers by this author Joseph D Maman Joseph D Maman Department of Biochemistry, University of Cambridge, Cambridge, UK Search for more papers by this author Nicholas Furnham Nicholas Furnham Department of Biochemistry, University of Cambridge, Cambridge, UK Search for more papers by this author Luca Pellegrini Corresponding Author Luca Pellegrini Department of Biochemistry, University of Cambridge, Cambridge, UK Search for more papers by this author Author Information Stefania Ragone1,‡, Joseph D Maman1,‡, Nicholas Furnham1 and Luca Pellegrini 1 1Department of Biochemistry, University of Cambridge, Cambridge, UK ‡These authors contributed equally to this work *Corresponding author. Department of Biochemistry, University of Cambridge, 80 Tennis Court Road, Cambridge CB2 1GA, UK. Tel.: +44 122 376 0469; Fax: +44 122 376 6002; E-mail: [email protected] The EMBO Journal (2008)27:2259-2269https://doi.org/10.1038/emboj.2008.145 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info The RecA/RAD51 nucleoprotein filament is central to the reaction of homologous recombination (HR). Filament activity must be tightly regulated in vivo as unrestrained HR can cause genomic instability. Our mechanistic understanding of HR is restricted by lack of structural information about the regulatory proteins that control filament activity. Here, we describe a structural and functional analysis of the HR inhibitor protein RecX and its mode of interaction with the RecA filament. RecX is a modular protein assembled of repeated three-helix motifs. The relative arrangement of the repeats generates an elongated and curved shape that is well suited for binding within the helical groove of the RecA filament. Structure-based mutagenesis confirms that conserved basic residues on the concave side of RecX are important for repression of RecA activity. Analysis of RecA filament dynamics in the presence of RecX shows that RecX actively promotes filament disassembly. Collectively, our data support a model in which RecX binding to the helical groove of the filament causes local dissociation of RecA protomers, leading to filament destabilisation and HR inhibition. Introduction Homologous recombination (HR) is necessary for the accurate repair of double-strand (ds) breaks in the DNA and is crucial for maintaining genome integrity (van Gent et al, 2001; West, 2003). The conserved RecA-like DNA strand-exchange proteins are DNA-dependent ATPases that polymerise on single-stranded DNA (ssDNA) to form a right-handed nucleoprotein filament, the molecular species that performs the homologous pairing and strand-exchange reactions of HR (Brendel et al, 1997; Roca and Cox, 1997; Bianco et al, 1998; Lusetti and Cox, 2002; Schlacher et al, 2006; Cox, 2007a, 2007b). Uncontrolled recombination can lead to genomic instability and as a result the activity of the nucleoprotein filament is finely tuned by a range of regulatory mechanisms that can promote or repress HR (Sung and Klein, 2006). In bacteria, RecA activity is modulated by a growing network of functional interactions that are only partially understood (Cox, 2007b). The single-strand-binding (SSB) protein coats exposed ssDNA, removing DNA secondary structure in the process and inhibiting binding of RecA (Kowalczykowski and Krupp, 1987; Lavery and Kowalczykowski, 1990; Hobbs et al, 2007). The RecBCD complex degrades the ends of a dsDNA break and loads RecA on the resulting 3′-end ssDNA tail in preparation for strand exchange (Kowalczykowski, 2000). The RecR, RecO and RecF proteins cooperate to promote RecA filament assembly on gapped DNA coated with SSB, possibly limiting filament extension onto dsDNA (Umezu et al, 1993; Umezu and Kolodner, 1994; Shan et al, 1997; Webb et al, 1997; Bork et al, 2001b; Morimatsu and Kowalczykowski, 2003). RecF is also capable of promoting RecA filament activity by counteracting the effect of the negative regulator RecX (Lusetti et al, 2006). The DinI protein stabilises RecA filaments (Lusetti et al, 2004), whereas the role of PsiB and RdgC in RecA filament activity is still unclear (Bailone et al, 1988; Moore et al, 2003). The UvrD helicase can disassemble RecA filaments in vitro and may be involved in destabilising recombination intermediates (Mendonca et al, 1993; Morel et al, 1993; Veaute et al, 2005). The Escherichia coli recX gene is located immediately downstream of the recA gene. Both genes are expressed from the same cistron, which is under the transcriptional control of the LexA repressor (Pages et al, 2003). In E. coli, a DNA hairpin between the recA and recX genes allows 5–10% transcriptional read-through of the recX gene, resulting in lower levels of the RecX protein relative to RecA (Pages et al, 2003). The RecX protein is required to overcome the effect of overexpressing RecA, suggesting that RecX is a negative regulator of RecA (Sano, 1993; Papavinasasundaram et al, 1997; Vierling et al, 2000; Sukchawalit et al, 2001). Deletion of the recX gene did not result in a clear phenotype in E. coli, but overexpression of the recX gene inhibited induction of the SOS response (Pages et al, 2003; Stohl et al, 2003). In vitro, the RecX proteins from Mycobacterium tuberculosis and E. coli inhibited RecA-mediated strand exchange and ATPase activity, as well as co-protease function, at substoichiometric concentration relative to RecA (Venkatesh et al, 2002; Stohl et al, 2003; Drees et al, 2004b). Taken together, the evidence indicates that RecX represses HR by interfering with RecA functionalities. The mechanism by which RecX regulates recombination is still unclear. Electron microscopy (EM) reconstructions of AMP·PNP-stabilised RecA filaments on dsDNA in the presence of nearly stoichiometric concentration of RecX revealed that RecX can bind within the helical groove of the RecA nucleoprotein filament (VanLoock et al, 2003). As a result, it was suggested that repression of recombination might occur by steric clash of RecX with the strand-exchange reaction, as well as by prevention of the ATP-coupled allosteric changes in RecA that are required for recombination (VanLoock et al, 2003). An alternative model for RecX-dependent inhibition of recombination was based on its proposed ability to cap the growing 3′-end of the RecA filament, resulting in net disassembly of RecA protomers and eventual dissolution of the filament (Drees et al, 2004a). In this paper, we describe a structural and functional analysis of the E. coli RecX protein and its interaction with the RecA nucleoprotein filament. The crystal structure of RecX reveals a modular architecture of tandem helical repeats, which is strongly suggestive of a mechanism of interaction with the RecA filament. We define a conserved, positively charged surface of RecX as important for its inhibitory function and test this assumption by defining RecX mutants with weakened ability to repress RecA function in vitro. The structural and biochemical data are complemented by a functional analysis that demonstrates how RecX actively promotes RecA dissociation from established filaments. We integrate our findings and existing data in a model of RecX function, which provides a novel example of the variety of mechanisms that regulate HR by interference with the activity of the nucleoprotein filament. Results Modular architecture of the RecX structure Initial crystallisation trials with wild-type RecX were unsuccessful. Mutation of cysteine residues 113 and 118 to alanine generated a version of the RecX protein that yielded crystals suitable for high-resolution X-ray analysis. The crystal structure was solved to a resolution of 1.8 Å using the anomalous signal of selenomethionine-derivatised protein (Supplementary Table 1). In the crystals, two RecX chains related by a non-crystallographic two-fold axis share an unusually large interface that buries 2800 Å2 of solvent-accessible surface. The homodimeric form of RecX appears to be an artefact of crystallisation because we could not detect it in solution using size-exclusion chromatography, homo-bifunctional chemical crosslinking (not shown), or sedimentation velocity in an analytical ultracentrifuge (Supplementary Figure 1). The mutagenesis required to induce crystallisation did not impair RecX function (Supplementary Figure 2). The crystallographic analysis demonstrates that RecX is a modular protein, consisting of three copies of a three-helix bundle of approximately 50 amino acids each that assemble in tandem repeats, resulting in an elongated protein shape of approximate overall dimensions of 70 Å × 20 Å × 20 Å (Figure 1). The relative arrangement of the three repeats imparts a moderate degree of curvature to the long axis of the RecX structure, resulting in a somewhat flat concave surface on one side and a convex surface on the other side of the molecule. Despite the low degree of sequence similarity, the three repeats can be superimposed with an overall r.m.s.d. of 1.7 Å over 37 aligned Cα positions (Figure 1B). Although the modular nature of the RecX structure could not be detected by inspection of the amino-acid sequence, structural superposition of the three repeats identifies several residues that are preferentially occupied by hydrophobic amino acids (Figure 1D). The most likely explanation for their conservation is their important structural role in RecX folding. Amino acids 113 and 118, which were mutated from cysteine to alanine to induce crystallisation, occupy solvent-exposed positions on the convex side of the protein. Figure 1.3D structure of the RecX protein. (A) The RecX structure consists of three tandem repeats of a three-helix motif. The cartoon representation of RecX highlights the modular nature of its structure. The three repeats are coloured pink (1), purple (2) and red (3), from the N- to the C-terminal end of the protein. The helices of the three-helix motif are numbered in roman numerals. (B) Structural superposition of the three repeats of RecX. Colour coding is as in (A). The side chain of a highly conserved hydrophobic residue in helix II of the three repeats (L36, L96 and L147) is shown in ball-and-stick representation; its position in the multiple sequence alignment of (D) is marked with a black triangle. (C) Structural superposition of repeat 2 of RecX (purple) with the homeodomain of the Oct-1 Pou protein (PDB entry 1POG), in yellow. (D) Structure-based multiple sequence alignment of proteobacterial RecX sequences (ECO, Escherichia coli; CPE, Candidatus pelagibacter; SAG, Stappia aggregata; RET, Rhizobium etli; GBE, Geobacter bemidjiensis; PPR, Pelobacter propionicus; NME, Neisseria meningitidis; RSO, Ralstonia solanacearum; XAX, Xanthomonas axonopodis; PAE, Pseudomonas aeruginosa; HDU, Haemophilus ducreyi; LPN, Legionella pneumophila; VVU, Vibrio vulnificus). Similar residues are in cyan, identical residues in yellow and completely conserved residues in green. The position of the three alpha-helices for each of the three repeats is marked above the alignment; the helices are coloured as in (A). The alignment is split into three parts according to repeat boundaries. The sequences of the repeats are arranged so that structurally equivalent residues are vertically aligned over the three repeats. A sequence insertion in the alignment of the N-terminal repeat is shown to the right side of the alignment. The position of C113 and C118 that were mutated to alanine for RecX crystallisation is indicated by an asterisk. Download figure Download PowerPoint Structural comparison of the RecX structure against the Protein Data Bank (PDB) archive in SSM (http://www.ebi.ac.uk/msd-srv/ssm/) did not find any reliable match, suggesting that the RecX fold is unique. However, the structure of the isolated RecX repeat represents a convincing 3D match for helix-turn-helix (HTH) DNA-binding domains found in homeodomain transcription factors and site-specific recombination enzymes: RecX repeat 2 can be superimposed to the Oct-1 POU homeodomain of PDB entry 1POG with an r.m.s.d. of 1.9 Å over 41 Cα positions (Figure 1C). A weaker structural similarity can also be detected between a RecX segment spanning repeats 1 and 2 and the two HTH modules of the MarA transcriptional activator (PDB entry 1BL0). Thus, RecX extends our knowledge of tandem repeat proteins as a remarkable example of protein architecture assembled by consecutive copies of an HTH-like domain. The finding that the RecX repeats resemble DNA-binding domains raises the possibility that RecX might bind DNA. Published evidence shows that RecX can bind both ss- and dsDNA, but the functional relevance of these observations is still unclear (Stohl et al, 2003; Drees et al, 2004a; Galvao et al, 2004). Inspection of the RecX structure reveals the presence of considerable repeat–repeat interfaces: inter-repeat contacts bury 779 Å2 and 965 Å2 of accessible surface area at the 1–2 and 2–3 repeat interfaces, respectively. The tight packing of the repeats fixes their relative orientation and confers overall rigidity to the structure. Notable common features of the repeat–repeat interface include the presence of a conserved glycine residue in the tight connection between helices II and III of repeats 2 and 3 (G100 and G151, respectively), which allows for close head-to-tail packing of helix II in the three repeats (Supplementary Figure 3). The interface between repeats 2 and 3 is particularly extensive: W117 at the N terminus of helix I in repeat 3 links the interior of repeats 2 and 3 in an hydrophobic continuum (Supplementary Figure 3). The smaller hydrophobic component of the interface between repeats 1 and 2 is compensated for by additional polar interactions: particularly prominent are the partially buried, bidentate interactions linking the carboxylate of D71 at the N terminus of helix I in repeat 2 to the main chain nitrogen atoms of E29 and Q30 in repeat 1; and the carbonyl moiety and Nζ nitrogen atom of K99 at the C terminus of helix II in repeat 2 with the side chains of S28 and E31, respectively, in helix II of repeat 1 (Supplementary Figure 3). Surface properties of the RecX structure The available evidence indicates that RecX regulates HR by direct interaction with the RecA nucleoprotein filament (VanLoock et al, 2003; Drees et al, 2004a). To gain an insight about the structural epitopes of the molecule that might be responsible for protein–protein and protein–DNA interactions, we analysed the charge distribution of the RecX surface. The outstanding feature of the electrostatic properties of RecX is the opposite polarity of the charge distribution on its concave and convex sides: the concave face of RecX is almost uniformly positively charged, whereas the convex face is predominantly negative (Figure 2A). Thus, the electrostatic analysis of the RecX surface indicates that RecX is a highly polarised molecule and suggests that the peculiar charge distribution of the protein might have an important role in its function. Figure 2.Surface properties of the RecX protein. The two views in (A, B) show the concave and convex surfaces of the protein, obtained by a 180° rotation on the long axis of the molecule. (A) Electrostatic properties of the RecX protein. The electrostatic potential was calculated with APBS (Baker et al, 2001) and mapped onto the solvent-accessible surface of the molecule, at contouring levels of ±5 kT (blue/red). The concave and convex sides of the protein display surface potentials of opposite charge. (B) Evolutionary conservation of surface residues in RecX, calculated with ConSurf (Landau et al, 2005) from a structure-based alignment of 74 RecX sequences. Amino-acid conservation on the molecular surface of the protein is indicated by a transition in colour hues, from magenta (most conserved) to cyan (most variable). Download figure Download PowerPoint Solvent-exposed regions of a protein that are involved in functional interactions with other macromolecules are subject to stronger evolutionary pressure and therefore show a higher degree of amino-acid conservation than otherwise expected for surface residues. We used the ConSurf server (http://consurf.tau.ac.il/) to map the extent of phylogenetic conservation of RecX residues on our crystallographic model (Landau et al, 2005). The structure-based phylogenetic analysis highlights another striking aspect of the RecX protein: the solvent-exposed amino acids on the concave face of the protein present a higher degree of conservation relative to residues of the convex side (Figure 2B). Inspection of the concave surface of RecX reveals the presence of an extensive network of side chain–side chain interactions involving basic and aromatic residues (Figure 3A). The program CaPTURE (Gallivan and Dougherty, 1999) identifies five of these contacts as energetically significant cation–π interactions: F78–R82, R82–Y87, Y87–R150, K128–F146 and R145–Y149. The mesh of cation–π contacts immobilise the side chains of conserved basic residues R82, K128, R145 and R150, thus keeping them poised for potential interaction. The potential for hydrophilic interaction of the concave surface is augmented by side chains of R25, K35, R91, K99, R127 and K142 that line the perimeter of the cation–π interaction network. Figure 3.Structure-based mutational analysis of RecX. (A) Basic residues on the conserved concave face of RecX. Aromatic residues involved in cation–π interactions with basic side chains, as well as glutamic and glutamine residues hydrogen bonded to the basic side chains, are also shown. Cation–π interactions that are energetically significant according to the program CaPTURE (Gallivan and Dougherty, 1999) are enclosed within a dashed-line box. The RecX polypeptide is represented by a ribbon, whereas side chains are drawn in ball-and-stick representation. Basic residues selected for mutagenesis are indicated by arrows. (B) Inhibition of RecA's DNA-dependent ATPase activity by wild-type and mutant RecX proteins. Each graph shows the amount of ATP hydrolysed by RecA in the presence of the various RecX mutants (see Materials and methods for details). The assay was performed at four different RecX concentrations, reported in each graph. Each data point represents the average of three simultaneous measurements. Download figure Download PowerPoint Structure-based mutational analysis of RecX function Examination of the RecX structure suggested that its conserved, positively charged concave face might represent a functional surface. We therefore selected for mutagenesis the conserved lysine and arginine residues R25, K35, R127, K128, R145 and analysed the ability of the mutant proteins to inhibit RecA activity. The functional importance of the selected amino acids was tested by mutation to glutamic acid, which reverses the charge of the residue while preserving its hydrophilic nature. In the case of residues participating in cation–π contacts (K128 and R145), mutation was to methionine, to maintain the hydrophobic component of the interaction. The RecX mutants were tested for their ability to suppress the ATPase and strand-exchange activities of the RecA filament. The experiments were performed with a range of RecX concentrations to explore the variation in potency between the mutant and wild-type proteins. All RecX mutants showed reduced inhibition of ATP hydrolysis by the RecA filament relative to wild-type RecX, albeit with considerable differences in their effectiveness (Figure 3B). Examination of the concentration dependency of RecA inhibition allowed ranking of the RecX mutants as: R25E>R145M>K128M∼K35E>R127E, in order of decreasing efficacy. R25E was the most effective mutation, extinguishing RecX-dependent inhibition almost completely even at the highest concentration tested (0.8 μM; 1:2.5 ratio of RecX to RecA). The results of the strand-exchange inhibition experiments confirmed that R25E is the most effective single-point mutation in compromising RecX function, whereas the R145M mutant displayed only a mild impairment and the K35E, R127E, K128M mutants behaved similar to wild-type RecX (Figure 4). A double mutation of R25E and R127E showed the strongest effect on RecX-dependent inhibition, rendering the protein virtually inactive in both ATPase and strand-exchange assays at all tested concentrations. Figure 4.Inhibition of RecA's strand-exchange activity by wild-type and mutant RecX proteins. The strand-exchange reactions were incubated at 37°C for 45 min with the concentration of wild-type and mutant RecX indicated below each lane (see Materials and methods for details). css: circular single-stranded DNA; lds: linear double-stranded DNA; jm: joint molecule; nc: nicked circle (product of reaction). The position of the 3 kb band in the dsDNA 1 kb ladder (NEB) is indicated by an arrow for reference. Download figure Download PowerPoint These observations implicate conserved basic residues on the concave surface of RecX in the repression of RecA filament activity. The solvent-exposed R25 in the linker between helices I and II of repeat 1 appears to be particularly important for RecX function. However, the involvement of spatially distant residues in RecX's inhibitory function suggests that the integrity of all or most of the concave surface of RecX is required for optimal function. Interaction of RecX with the RecA filament The structure of RecA–dsDNA–AMP·PNP filaments in the presence of RecX, at a 1:1 stoichiometric ratio of RecX to RecA, was investigated by EM (VanLoock et al, 2003). The EM analysis detected additional density located deep inside the helical groove of the RecA filament, which was attributed to RecX. The selection for 3D reconstruction of filament segments enriched in RecX density and the helical averaging of the filament structure resulted in a continuous tubular density wound around the filament axis. Docking of the crystallographic model of RecX in the EM reconstruction reveals that the volume and shape of the protein fit well into the additional RecX-associated density (Figure 5). Figure 5.Interaction of RecX with the RecA nucleoprotein filament. (A) Solid surface representation of the EM reconstruction of a RecA–RecX–dsDNA–AMP·PNP filament. The extra-filament density ascribed to RecX is coloured in light blue, whereas the rest of the filament is coloured in grey. (B) The figure shows the proposed mode of interaction of RecX with the RecA filament, based on manual docking of the atomic model of RecX in the RecA–RecX–dsDNA–AMP·PNP filament. α-Helices in RecX are drawn as ribbons, coloured orange (helix I), cyan (helix II) and light blue (helix III) in all repeats. Two RecA protomers drawn as yellow ribbons have also been fitted in successive filament rungs, above and below RecX. Download figure Download PowerPoint Although it is impossible to fix the position of RecX in the continuous density of the reconstructed filament, it is clear that the curvature of the RecX structure appears well suited to match the helical groove of the filament. As a result of the docking, the conserved, functionally active concave surface of RecX faces inwards towards the filament axis, in a suitable position to interact with both DNA and RecA. We note that the strongly basic nature of the concave surface of RecX would complement the polyanionic nature of DNA and the overall negative charge of the RecA C terminus, which mediates in part the interaction with RecX (Drees et al, 2004b). Taken together, the functional analysis of RecX mutants and the docking of the RecX structure into the EM reconstruction of the RecA–RecX nucleoprotein filament suggest that RecX inhibits RecA functions by binding within the helical groove of the filament through its positively charged, concave surface. RecX actively removes RecA from nucleoprotein filaments After nucleation on ssDNA, RecA filaments grow by the net addition of RecA protomers to the 3′-end of the filament and shrink by net release of RecA protomers from the 5′-end; a higher rate of RecA binding at the 3′-end relative to the rate of RecA dissociation at the 5′-end results in increased filament size (Lindsley and Cox, 1990; Shan et al, 1997; Bork et al, 2001a; Galletto et al, 2006; Joo et al, 2006). Drees et al (2004a) suggested that RecX performs its inhibitory function by binding to the 3′-end of the RecA filament, in a manner that prevents further addition of RecA protomers to the growing end of the filament. As filament disassembly continues unimpeded from the 5′-end, ‘capping’ of the 3′-end by RecX would lead over time to filament dissolution. We chose to test current models of RecX function by surface plasmon resonance (SPR). Nucleoprotein filaments were generated by binding RecA to a 5′-biotinylated ssDNA (60-mer oligo-dT) immobilised on a streptavidin (SA)-coated sensor chip (Figure 6A). RecA was injected into the flow cells until the binding curve indicated that maximal saturation had been achieved. We confirmed that the dynamic behaviour of RecA filaments reconstituted on the SA chip was as previously reported by examining the dependency of RecA dissociation on ATP hydrolysis (Lindsley and Cox, 1990; Arenson et al, 1999; Galletto et al, 2006; Joo et al, 2006) (Supplementary Figure 4). The RecA filaments were then challenged by flowing RecX over the sensor chip. The set-up of the SPR experiment was designed so that it would be possible to discriminate between ‘passive’ and ‘active’ mechanisms of RecX action. In fact, no RecA binding can take place after the initial injection, while dissociating RecA molecules are removed by the continuous flow in the system. Therefore, capping of the RecA filament at its 3′-end by the injected RecX would not alter the rate of RecA dissociation from the filament. Alternatively, any RecX-dependent mechanism involving active RecA removal from an established filament would alter the kinetic profile of filament disassembly. Figure 6.RecX increases the rate of RecA filament disassembly. (A) Illustrative diagram of the surface plasmon resonance experiment, detailing the linkage of the ssDNA substrate to the sensor chip, and the various stages of RecA filament assembly and dissociation during the experiment. RU is resonance units. (B) The effect of RecX concentration (reported in the panel) on the dissociation rate of RecA filaments. (C) The effect of wild-type and mutant R145M, R25E-R127E RecX proteins on the dissociation rate of RecA filaments. RecX concentration in the experiment was 0.4 μM for wild-type and mutant proteins. (D) Same as in (B), except that the non-hydrolysable analogue AMP·PNP was used instead of ATP. Download figure Download PowerPoint The SPR analysis shows that RecX increases the rate of RecA dissociation from the filaments in a concentration-dependent manner (Figure 6B). The two RecX mutants, R25E–R127E and R145M, are impaired in their abilities to increase the rate of RecA filament disassembly compared with wild-type RecX (Figure 6C), a result that mirrors the effect of these mutants on RecA-dependent ATP hydrolysis. To test whether the effect of RecX depends on the ability of RecA to hydrolyse ATP, we reconstituted RecA filaments in the presence of the non-hydrolysable ATP analogue AMP·PNP. AMP·PNP increases considerably the rate of RecA binding to ssDNA, while causing a decrease in the rate of filament disassembly (Figure 6D). These observations agree with the augmented stability reported for RecA filaments formed in the presence of AMP·PNP (VanLoock et al, 2003) or ATPγS (Chabbert et al, 1987; Arenson et al, 1999; Galletto et al, 2006; Joo et al, 2006; Cisse et al, 2007). RecX increases the rate of RecA dissociation in a concentration-dependent manner, as observed for ATP. However, the effect of RecX on RecA dissociation is less pronounced in the absence of RecA-dependent ATP hydrolysis (Figure 6D). Discussion In this paper, we have investigated the structural basis for the role of the RecX protein in HR. The crystal structure of RecX reveals an unanticipated modular architecture of three tandem repeats of a three-helix domain, bearing a clear resemblance to well-known DNA-binding HTH domains similar to the homeodomain. Thus, the crystallographic analysis of RecX increases our knowledge of the predominantly eukaryotic HTH domain by demonstrating its deployment in a novel structural and functional context. Furthermore, using a sequence profile derived from the structural superposition of the three repeats of E. coli RecX, it is possible to identify two additional repeats in the N- and C-terminal tails of a subset of RecX sequences that exceed the consensus size by about 100 amino acids (data not shown). The existence of polypeptides spanning five repeats confirms the modular nature of the RecX protein. We identi