Structural biology

Snapshots of DNA repair

In recombinational DNA repair, nearly identical sequences in chromosomes are found and swapped. Structures of the RecA–DNA complexes involved provide insight into the mechanism and energetics of this universal process.

Homologous recombination is one of the many processes used by cells to repair damaged DNA and to diversify their genomes. A central step in recombination involves the exchange of DNA strands between identical, or nearly identical, segments of chromosomes. This crucial reaction is catalysed by the RecA family of DNA-strand-exchange proteins, which include the founding member in bacteria, Rad51 in eukaryotes and RadA in archaeans. On page 489 of this issue, Chen et al.1 describe structures of both the substrate (RecA complexed with single-stranded DNA) and the product (RecA complexed with double-stranded DNA) of DNA strand exchange. The structures reveal non-uniform DNA stretching, and suggest a mechanism for strand exchange.

On the face of it, DNA strand exchange is a simple reaction: one strand of double-stranded DNA (dsDNA) is replaced with an identical single strand of DNA (Fig. 1a). This reaction would be simple, were it not for the stability of dsDNA, which resists strand separation, and the need to accurately align DNA sequences. Thus, although the stability is useful for the storage of genetic information, it is an impediment to DNA repair and recombination.

Figure 1: DNA strand exchange promoted by RecA protein.

a, The prototypical DNA-strand-exchange reaction. Double-stranded DNA (dsDNA) pairs with the RecA presynaptic filament, which consists of RecA protein and single-stranded DNA (ssDNA), to produce heteroduplex DNA bound by RecA and the exchanged ssDNA. b, Structures of the participating DNA molecules: B-form dsDNA; ssDNA within the presynaptic filament (as determined by Chen and colleagues1, protein not shown); dsDNA within the filament1 (protein not shown); and randomly coiled ssDNA.

The RecA-like proteins do not deal with this problem by unwinding dsDNA, as helicase enzymes do. Instead, they first assemble on a single strand of DNA, which was generated in the preceding step of recombination, to form a helical nucleoprotein (protein–DNA) filament. When formed with ATP, this filament (termed the presynaptic complex) is the active species in DNA strand exchange that searches for a homologous sequence within the dsDNA. Once found, DNA strand exchange occurs as a concerted swap of DNA strands. The hydrolysis of ATP inactivates the filament, and permits disassembly of the complexes2.

Herein lie the mysteries of DNA strand exchange. How does the RecA nucleoprotein filament recognize DNA sequence identity? And, on finding it, how does the exchange occur? How is the stability of dsDNA overcome? Partial answers to these questions emerged from biochemical studies. The homology search is a 'simple' collisional process because ATP hydrolysis is not essential, only ATP binding. In fact, ATP binding by the RecA nucleoprotein filament is sufficient for DNA strand exchange3. The free energy of ATP hydrolysis is not directly involved in the exchange of DNA strands; rather, the free energy of presynaptic-filament binding to the dsDNA 'activates' it by extending and untwisting it, making the duplex DNA a willing participant in the exchange process4. ATP hydrolysis then allows dissociation of all participants: a classic case of 'credit card' energetics (expend now, pay later ...).

Structural information derived from electron microscopy was particularly revealing with respect to these questions. The ATP-bound form of the RecA nucleoprotein filament is extended by about 50% relative to standard B-form DNA, with around 6.2 RecA monomers and 18 DNA base pairs per turn5; this extended filament is also seen for all RecA homologues6. In contrast, the DNA in the inactive ADP-bound nucleoprotein filament is less extended. Thus, the RecA nucleoprotein filament undergoes ligand-induced structural transitions between an active, extended filament and an inactive, compact filament5. The electron microscopy studies also revealed that the RecA nucleoprotein filaments are structurally polymorphic, varying in pitch, width and extension5, highlighting the challenge facing higher-resolution structural analysis.

Nonetheless, in 1992 the crystal structure of a RecA filament was solved7, but the structure lacked DNA and was of the inactive ADP-bound compact form8. For many years, the absence of a structure of the active form confounded molecular and mechanistic interpretation9.

However, Chen and colleagues1 now elucidate structures of RecA assembled on single-stranded DNA (ssDNA) and on dsDNA, defining both the substrate and product forms of the reaction, respectively. How did they succeed? The authors recognized that the intrinsic conformational flexibility of the RecA nucleoprotein filaments and their capacity to self-assemble indefinitely might hinder crystallization in the active state. Their solution to these problems was ingenious, and is applicable to other self-assembling systems.

First, they created 'pre–polymerized' assemblies of RecA protein by fusing four, five or six monomers of RecA into a single polypeptide chain. To prevent indefinite polymerization of the resulting 'mini-filaments', the sites for monomer–monomer interactions were deleted from the first and last monomers in the chain. Despite the many potential pitfalls, this approach worked splendidly, producing functional proteins. When assembled in the presence of an ATP analogue on DNA that exactly accommodated these fusion proteins (15 and 18 nucleotides, respectively), these mini-filaments formed ordered crystals.

The structures reveal an ordered filament with 6.2 monomeric units per turn and a pitch of 92–95 ångstroms. The DNA is close to the filament axis, is extended relative to B-form DNA, and has global features compatible with the electron microscopy. The ATP is completely buried at an interface between monomers. Each RecA monomer interacts with three nucleotides of the DNA (a triplet) adjacent to itself in the structure, as well as with two more nucleotides, one from each of the preceding and following triplets. As a result, each nucleotide triplet is bound by three monomers.

Perhaps the most remarkable feature of the nucleoprotein filament is the DNA structure (Fig. 1b). The 50% extension is not manifest as an isotropic extension at the nucleotide level; instead, the DNA is seen to comprise a three-nucleotide segment with a nearly normal B-form distance between bases (an axial rise of 3.5–4.2 Å for ssDNA and 3.2–3.5 Å for dsDNA), followed by a long untwisted inter-nucleotide stretch (approximately 7.1–7.8 Å in ssDNA and 8.4 Å in dsDNA) before the next three-nucleotide element, and so on. This was a surprising result, because most people assumed that the DNA within the RecA–DNA complexes was uniformly stretched to an average of about 5.2 Å between bases.

The unusual repeat pattern of DNA extension in the RecA nucleoprotein filament offers a structural basis for understanding the dynamics of filament assembly. Assembly occurs by rate-limiting initiation of polymer formation (nucleation) followed by growth2. The structure shows that it would be energetically unfavourable for a single monomer to make the full repertoire of molecular contacts with DNA because of the need to both unstack the bases and extend the DNA to the next nucleotide triplet. Thus, the free energy for binding of the first monomer will be unfavourable relative to the binding of a second protein to an existing monomer, explaining the observed cooperativity of RecA binding to DNA2. Binding of a third monomer provides additional net free energy, because now two of the three monomers benefit from the added free energy of cooperative interactions. As more monomers bind, the energetic cost of extending the DNA is 'amortized' over an increasingly greater number of RecA monomers, until the net free energy of nucleus formation is sufficiently negative to permit stable nucleation. Although more complex scenarios can be envisaged, the structures of Chen and colleagues1 now permit detailed energetic modelling of filament formation.

The results also highlight the physical mismatch between the ssDNA within the filament and the naked duplex DNA target (Fig. 1b). How does RecA align these sequences? The structures1 offer provocative insights into how the transient three-stranded intermediate might look, and how the fidelity of DNA strand exchange might be enforced. It is easy to imagine the pairing between an ssDNA triplet within the filament and the naked dsDNA, as both have approximately B-form dimensions. However, pairing of the next three base pairs of DNA requires extension of the dsDNA to conform to the observed extension of the ssDNA in the filament. This energetically unfavourable base unstacking and chain extension could be compensated both by the now lower entropic cost (because the next triplet is part of the already paired dsDNA) and by the favourable base-pairing interactions that would form if the next triplet were fully homologous. However, if even one of the base pairs was non-complementary, then the nascent paired molecule might not be sufficiently stable, and homologous pairing with a partially homologous sequence would be aborted.

Furthermore, the structures show that the strand complementary to the ssDNA in the presynaptic filament makes few contacts with the protein. Hence, it is largely stabilized by correct Watson–Crick base-pairing, thereby requiring accurate DNA pairing. Successful DNA pairing requires at least 15 base pairs of homology10, and the structures suggest how such fidelity is enforced.

Determination of the three-dimensional structure of the active state of RecA nucleoprotein filaments by Chen and colleagues1 is a watershed in recombination biochemistry and mechanics. Not only do the structures inform us about this central protein, but they also enable the formulation of structural hypotheses that relate to the RecA orthologues and to interacting proteins. Although the eukaryotic and archaeal RecA homologues differ in many functional and mechanistic details, the RecA structures will provide a valuable foundation for understanding them. Also, many proteins interact with the various forms of RecA family members to regulate assembly and disassembly of the filament. Having structures of both the ATP– and ADP–RecA nucleoprotein filaments will help clarify the mechanistic basis of their biological functions. Clearly, more (DNA) partner-swapping experiments will be forthcoming.


  1. 1

    Chen, Z., Yang, H. & Pavletich, N. P. Nature 453, 489–494 (2008).

  2. 2

    Kowalczykowski, S. C. Annu. Rev. Biophys. Biophys. Chem. 20, 539–575 (1991).

  3. 3

    Menetski, J. P., Bear, D. G. & Kowalczykowski, S. C. Proc. Natl Acad. Sci. USA 87, 21–25 (1990).

  4. 4

    Mazin, A. V. & Kowalczykowski, S. C. Genes Dev. 13, 2005–2016 (1999).

  5. 5

    Yu, X. et al. Curr. Protein Pept. Sci. 5, 73–79 (2004).

  6. 6

    Bianco, P. R., Tracy, R. B. & Kowalczykowski, S. C. Front. Biosci. 3, d570–d603 (1998).

  7. 7

    Story, R. M., Weber, I. T. & Steitz, T. A. Nature 355, 318–325 (1992).

  8. 8

    Story, R. M. & Steitz, T. A. Nature 355, 374–376 (1992).

  9. 9

    McGrew, D. A. & Knight, K. L. Crit. Rev. Biochem. Mol. Biol. 38, 385–432 (2003).

  10. 10

    Ferrin, L. J. & Camerini-Otero, R. D. Science 254, 1494–1497 (1991).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Kowalczykowski, S. Snapshots of DNA repair. Nature 453, 463–465 (2008).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.