3.1 Assembly of the NPF

In the presynaptic phase of HR, the formation of the NPF is the first step catalysed by RecA and RAD51 recombinases (San Filippo et al. Reference San Filippo, Sung and Klein2008). A two-step mechanism involving nucleation and growth has been proposed for the formation of a recombinase filament (Chabbert et al. Reference Chabbert, Cazenave and Hélène1987), but a precise quantification of both nucleation and growth rate has not been possible using standard bulk biochemical assays. Furthermore, full understanding of the mechanism of nucleoprotein-filament formation requires detailed knowledge of the oligomeric state of the recombinases in solution their multimerization kinetics and the DNA-binding properties of the different oligomeric states (Takahashi, Reference Takahashi1989). In the following section, we will examine and discuss the major findings on the dynamic assembly of the NPF provided by single-molecule technologies. The assembly of both RecA and RAD51 recombinase filaments on ss and dsDNA was addressed using a diverse set of single-molecule techniques, including magnetic and optical tweezers, FRET spectroscopy and single-molecule fluorescence microscopy.

Using magnetic tweezers, changes in end-to-end length of DNA covered with RecA or RAD51 NPF were measured in real time (Figs 3 and 4) (Van der Heijden et al. Reference van der Heijden, Seidel, Modesti, Kanaar, Wyman and Dekker2007; Van Loenhout et al. Reference Van Loenhout, Van der Heijden, Kanaar, Wyman and Dekker2009). Monte-Carlo simulations were performed to extract kinetic parameters characterizing RAD51 and RecA assembly (Van der Heijden & Dekker, Reference Van der Heijden and Dekker2008). On basis of these simulations, the authors proposed that RAD51-filament formation is characterized by: (i) a low ratio between growth and nucleation rate (cooperativity), approximately 200 for RAD51; (ii) pentamers of RAD51, preassembled in solution, being the units of filament nucleation and growth; (iii) RAD51-filament assembly being independent of the DNA being double- or single stranded and the tension applied to the DNA. The same experimental method and analysis were also applied to E. coli RecA, and its filament formation mechanism on ssDNA was addressed (Van Loenhout et al. Reference Van Loenhout, Van der Heijden, Kanaar, Wyman and Dekker2009). Similarly to RAD51, the data suggested that RecA assembly on ssDNA occurs by nucleation and growth of preformed multimers of RecA (hexamers) that form filaments with low cooperativity (between 10 and 20).

Fig. 3.

RAD51 NPF formation at the single-molecule level. (A) Magnetic tweezers assay for studying the assembly of a recombinase protein on ssDNA and dsDNA. Constant tension is applied across a DNA molecule using a paramagnetic bead and a pair of permanent magnets. Binding of RAD51 induces a change in the end-to-end distance, which can be monitored in real time. (B) DNA length is monitored as a function of time before and after flushing RAD51 in the flow-cell. The increase in DNA length follows a typical profile that can be further analysed in order to estimate the parameters of RAD51 filament formation including nucleation and growth rate. (C) Free rotation magnetic tweezers technique. The magnet configuration produces a magnetic field in the z-direction, allowing the bead to freely rotate. (D) Sequence of video images showing the bead rotational movement produced by the loading of RAD51 on dsDNA. (E) Rotation traces at different RAD51 concentration show a stepping behaviour with a typical step size of 65° (black-10 nM, red-75 nM, blue-50 nM). (F) Single optical trapping combined with flow stretching and microfluidics is used to visualize the nucleation of RAD51 on dsDNA. Fluorescent snapshots are taken at different time points using increasing concentrations of RAD51 (from 100 to 500 nM) reveal the formation of multiple nuclei. (G) The clusters formation can be quantified as a function of time, showing a linear dependence on the time. (H) Nucleation rate can be plotted as a function of free RAD51 concentration, showing a non-linear dependence. The experimental data can be fitted with a using a power law with a 2.4 exponent. Figures reprinted from (Arata et al. Reference Arata, Dupont, Miné-Hattab, Disseau, Renodon-Cornière, Takahashi, Viovy and Cappello2009; Hilario et al. Reference Hilario, Amitani, Baskin and Kowalczykowski2009; Van der Heijden et al. Reference van der Heijden, Seidel, Modesti, Kanaar, Wyman and Dekker2007).

Fig. 4.

RecA assembly studied using fluorescence visualization and FRET spectroscopy. (A) Optical tweezes combined with flow stretching for studying RecA assembly on dsDNA. Snapshots taken at different incubation times show a drastic increase in the fluorescence intensity along the DNA molecule, corresponding to RecA binding to dsDNA. (B) Quantification of RecA nucleation rate (appearance of fluorescent cluster along the DNA molecule) at different RecA concentrations in the presence of ATP (red dots) and ATP-γ-S (black). Fit corresponds to the functions k _nucl=k[RecA]ⁿ. The value found for the parameter n was very similar for both nucleotide cofactors used in the experiment and was 4.5±0.5 for ATP and 4.0±0.4 ATP-γ-S. (C) Schematic representation of FRET-based assay for determining the extension unit of RecA filament assembly. A donor–acceptor FRET pair was positioned 13 nucleotides apart and RecA dynamics at the 5′-end of the filament could be observed by monitoring the FRET fluctuations in real time. (D) Four distinct FRET states were observed. Interestingly fast and reversible transitions occurred between states, indicating that association and dissociation of RecA was taking place. The transitions between FRET states were predominantly occurring between neighbouring states ($M_{i} \rightleftharpoons M_{i \plus \setnum{1}} $), indicating that the fundamental unit regulating assembly and disassembly is a RecA monomer. Figures reprinted from (Galletto et al. Reference Galletto, Amitani, Baskin and Kowalczykowski2006; Joo et al. Reference Joo, McKinney, Nakamura, Rasnik, Myong and Ha2006).

In another, technically innovative magnetic-tweezers study (Arata et al. Reference Arata, Dupont, Miné-Hattab, Disseau, Renodon-Cornière, Takahashi, Viovy and Cappello2009), RAD51 binding to dsDNA was monitored by tracking changes in dsDNA twist instead of length. According to structural studies, the binding of RAD51 results in the elongation and unwinding of duplex DNA: each RAD51 monomer binding to dsDNA increases dsDNA length with 0.17 nm and unwinds it of approximately 45° (Yu et al. Reference Yu, Jacobs, West, Ogawa and Egelman2001). To monitor the rotation of the magnetic bead along the vertical axis in response to RAD51 binding, a magnetic field parallel with the DNA was applied (Fig. 3). A smaller microsphere was attached to the paramagnetic bead connected to the DNA, allowing tracking of bead rotation with an accuracy of 5° (Fig. 3). Using this method, individual rotation steps of 65° were observed (Fig. 3), inconsistent with preassembled RAD51 pentamers being the active species for filament growth (Van der Heijden et al. Reference van der Heijden, Seidel, Modesti, Kanaar, Wyman and Dekker2007). On basis of these observations, the authors proposed a new model, where individual RAD51 monomers are responsible of the filament growth step. A single-molecule approach able to directly visualize the nucleoprotein-filament assembly made use of optical tweezers in combination with DNA flow-stretching and fluorescence microscopy (Galletto et al. Reference Galletto, Amitani, Baskin and Kowalczykowski2006). A DNA molecule attached to a micron-sized bead held in an optical trap was extended using buffer flow. The instrument consisted of a micro-fluidic device that allow introduction of each reaction component separately using laminar flow. In this way, it was possible to expose the DNA molecule to a buffer containing fluorescently labelled RecA for a defined duration. Following this incubation step, the molecule was repositioned in an observation channel, which did not contain fluorescent RecA. Using this experimental assay, it was for the first time possible to directly visualize individual RecA nuclei on dsDNA and to quantify the rate of nucleation (of the order of 10⁻⁶ nuclei s⁻¹ bp⁻¹) (Fig. 4) It was observed that the rate of nucleation, the rate-limiting step of filament formation, depends on RecA concentration with a power law (Fig. 4). Fitting the experimental data with the equation k _nucl=k[RecA]ⁿ indicated that the nucleation step involves approximately 4–5 RecA monomers (n=4.5±0.5). In addition, the RecA-filament nucleation phase could be distinguished from the growth phase by performing multiple, subsequent incubation and visualization cycles. Growth occurred with a rate of ∼2 RecA monomers per second, indicating that RecA–dsDNA filament formation is characterized by a high cooperativity, of the order of 10⁶, which is higher than has been reported from the bulk biochemical assays (Takahashi, Reference Takahashi1989). This difference might arise from the ability to discriminate better the nucleation and growth in single-molecule measurements.

The same experimental protocol was applied to study the assembly mechanism of human RAD51 (Hilario et al. Reference Hilario, Amitani, Baskin and Kowalczykowski2009). Intriguingly, important differences were reported. First, for RAD51 nucleation was shown to involve 2–3 RAD51 monomers. Second, RecA requires only a few nucleation points that grow cooperatively to produce μm-long segments of continuously coated DNA molecule. RAD51, because of its lower cooperativity, reaches complete coverage of the DNA starting from many more nuclei that show limited growth. This observation is a clear indication that the cooperativity of RAD51–dsDNA filament formation is one or two orders of magnitude smaller that of RecA.

The RecA-filament-formation mechanism was addressed at high temporal resolution using spFRET (Joo et al. Reference Joo, McKinney, Nakamura, Rasnik, Myong and Ha2006), with a donor and acceptor fluorophore attached to ssDNA in order to measure ssDNA conformational changes upon filament formation. Without RecA bound, the ssDNA is in a coiled configuration, with donor and acceptor close, resulting in a high FRET efficiency (Fig. 4). RecA binding to ssDNA leads to the formation of a stiff and elongated NPF, resulting in an increased distance between donor and acceptor and consequently a lower FRET efficiency. Attaching the fluorophores to different locations on the DNA, permitted to compare the kinetics of filament formation on the 3′- and the 5′-end of the ssDNA. The authors observed in the first place that stable filament formation only took place only on substrates with a single-stranded overhang exceeding 17 nucleotides. Substrates with shorter ssDNA segments allowed only the transient formation of DNA–protein complexes. The hypothesis connected to this observation was that nucleation clusters composed of smaller structures than RecA pentamers are not stable enough to form a seed for filament formation. Next, RecA binding and unbinding rates were determined for both the 3′- and 5′-ends. These measurements indicated two interesting aspects of RecA-ssDNA interaction: (i) the unbinding rate is equal at the two ends and (ii) RecA filament formation is faster at the 3′-end. The differential assembly speed of RecA at opposite ends, rather than in differences in the disassembly kinetics is therefore the reason of polar growth displayed by RAD51. Finally, the detailed analysis of the FRET traces revealed that filament extension occurs via the addition of RecA monomers.

Very recently, a substantial advancement towards the understanding of RecA filament assembly has been made. Using a novel assay, Bell et al. (Bell et al. Reference Bell, Plank, Dombrowski and Kowalczykowski2012) monitored RecA nucleation and growth on single-stranded binding protein (SSB)-coated ssDNA. First, a biotinylated dsDNA molecule was attached to a streptavidin-coated glass coverslip, it was subsequently denatured using an alkaline solution and then saturated with fluorescent SSB. Fluorescent RecA was introduced in the flow-cell and after a rinsing step fluorescence visualization was used to inspect the formation of a RecA–ssDNA complex. Successive cycles of incubation, buffer exchange and fluorescence visualization highlighted the complex mechanism of RecA filament formation on SSB-coated ssDNA. A first striking observation was that RecA nucleation increased quadratically with respect to RecA concentration. This suggested that RecA nucleation on SSB-coated ssDNA requires a critical nucleus composed of a dimer, rather than of a pentamer, as reported by the same group in an earlier study on dsDNA (Galletto et al. Reference Galletto, Amitani, Baskin and Kowalczykowski2006). Also, the effects of recombinase mediators such as RecO, RecF and RecR (RecFOR complex) were tested at the single-molecule level using a dual-optical tweezers assay combined with fluorescence microscopy(Galletto et al. Reference Galletto, Amitani, Baskin and Kowalczykowski2006). The analysis of these interactions highlighted the interplay between SSB dynamics on ssDNA, RecA nucleation and filament assembly. The authors proposed that RecA is in fact unable to form filaments on SSB-coated ssDNA and its assembly relies uniquely on the transient local unwrapping of SSB tetramers from the ssDNA. The role of RecOR and RecFOR complex is therefore to destabilize the ssDNA–SSB complex, giving the opportunity to RecA to form the NPF and allow the sequential steps of HR to take place.

As described, several controversies have emerged from the various single-molecule studies of recombinase filament formation and a definitive molecular model describing the kinetics of filament formation is still lacking.

Currently, several experiments have provided evidence supporting a nucleation step of RAD51 and RecA that requires three or more proteins bound to ssDNA and dsDNA. A consensus over the actual number has not been reached yet, and in addition, a physical and structural explanation for this minimal size requirement is missing. Whether RAD51 and RecA accomplish nucleation starting from preassembled species in solution or through the coincident binding of multiple monomers also remain controversial. In fact, neither RecA nor RAD51 are present in solution in a well-defined oligomeric state. The kinetics of recombinase multimerization in solution has been addressed using different approaches (Nomme et al. Reference Nomme, Takizawa, Martinez, Renodon-Cornière, Fleury, Weigel, Yamamoto, Kurumizaka and Takahashi2008; Takahashi, Reference Takahashi1989). It was found that the oligomeric state of RAD51 and RecA is highly sensitive to buffer conditions and protein concentration, which are not always well known.

In addition, using single-molecule approaches such as magnetic and optical tweezers, it is possible to disentangle the filament growth phase from the nucleation step only if the two kinetic rates are highly separated (high cooperativity) (Pierobon et al. Reference Pierobon, Miné-Hattab, Cappello, Viovy and Lagomarsino2010). In magnetic tweezers experiment, for example, the binding of protein to the DNA is monitored by measuring changes in DNA length (or twist), averaging over multiple, simultaneously growing filaments. Techniques employing fluorescence visualization allow discriminating the dynamics of distinct filaments. Fluorescence methods, however, suffer from the high background signal of fluorescently labelled proteins in solution, which makes it very hard to visualize the progress of filament formation in real time. To overcome this problem, DNA molecules are subjected to cycles of incubation (in a buffer containing fluorescently labelled recombinase) and visualization (in a buffer not containing fluorescent proteins), resulting in only snapshots. In the near future, we believe that improved visualization methods, capable of limiting background-fluorescence, will allow visualization in real time the formation of RAD51 and RecA recombinase filament with single-monomer resolution and finally properly address these unresolved questions.

3.2 Conformational transitions of the NPF

The interplay between structural and functional properties of recombinase filaments has been intensely studied, in particular by electron microscopy. The general idea that has emerged is that both RecA and RAD51 filaments can adopt two conformations, an extended, active one with a pitch of ∼9–10 nm and a more compressed one, that is, at least 20% shorter (DiCapua et al. Reference DiCapua, Schnarr, Ruigrok, Lindner and Timmins1990; Yu et al. Reference Yu, Jacobs, West, Ogawa and Egelman2001). The structural heterogeneity present in recombinase filaments is commonly attributed to a conformational transition induced by ATP hydrolysis, which involves the rotation of the C-terminal subunit in RecA or N-terminal one in RAD51 (Yu et al. Reference Yu, Jacobs, West, Ogawa and Egelman2001). Interestingly, the two filament conformations have been argued to have different biochemical functions: the extended one is active in performing stand exchange, while the compressed one appears inactive and might represent an intermediate before filament disassembly (Bugreev & Mazin, Reference Bugreev and Mazin2004). Since electron microscopy can only provide static snapshots of these conformational changes, single-molecule techniques have been applied to address these dynamics transitions in real time. The mechanical properties of recombinase–DNA complexes were explored for the first time using optical tweezers in a set of pioneering experiments at the end of the 1990s in the laboratory of Bustamante. A remarkable behaviour was observed for RecA–ssDNA filaments assembled in the presence of ADP (Hegner et al. Reference Hegner, Smith and Bustamante1999). Upon successive stretching and relaxing of these filaments, the force–extension curves measured in the optical tweezers displayed notable hysteresis (Fig. 5). Extending the molecule to its contour length required forces up to 15 pN. Reaching this force-threshold resulted in a plateau-like lengthening, followed by a further force increase. The relaxation curve showed a substantial lengthening but a more regular behaviour (Fig. 5). Two alternative interpretations were put forward: (i) stretching the ssDNA induces a transition of the NPF from the compressed to the extended state or, (ii) free-RecA in solution can associate with gaps opening along the filament in a force-dependent and reversible way. The latter explanation could not be excluded since the experiments were performed in the presence of protein in solution. Later, similar results were obtained in a study using magnetic tweezers (Van Loenhout et al. Reference Van Loenhout, Van der Heijden, Kanaar, Wyman and Dekker2009). RecA NPFs were assembled on ssDNA in the presence of ATP and Mg²⁺, while the DNA's end-to-end length was monitored at a constant stretching force of 3 pN. Using micro-fluidics, buffers containing or lacking ATP could be flushed in reversibly, resulting in elongation or contraction of the DNA by at least 20% (Fig. 5). These experiments were conducted without RecA in solution, confirming that nucleotide exchange modulates the conformation of the RecA filaments.

Fig. 5.

Conformational transitions within RAD51 and RecA NPFs on dsDNA. (A) Consecutive stretching and releasing cycles of a single ssDNA molecule in complex with RecA–ADP using optical tweezers. The first stretching curve shows the approaching of the contour length approximately at the extension of 3 μm. At 10 pN a plateau is observed, followed by the sudden stiffening of the molecule. The release curve shows hysteresis, indicating the occurrence of a force-induced conformational transition. During the second force-extension cycle the contour length is reached at the extension of 2.5 μm but the elastic behaviour is qualitatively similar to the first extension-release cycle. (B) A RecA NPF is formed on ssDNA held at the force of 3 pN using magnetic tweezers in the presence of ATP. Multiple buffer exchanges containing or lacking ATP induce reversible changes in DNA-lengths. (C) TIRF-microscopy is combined with DNA flow-stretching for studying conformational changes within NPFs. (D) Example of experimental result showing the lengthening and successive shortening arising from NPF assembly and subsequent compression. (E) Cycles of buffer exchange show reversible transition of the NPF from the extended to compressed state and vice-versa resulting from the presence of free ATP in solution. (F) If after RAD51 assembly, ATP is depleted from solution (red trace) the nucleoprotein converts into its compressed form (final length ∼7 μm). When high ATP concentration is supplied to the reaction, the filament compression step is blocked. Figures reprinted from (Hegner et al. Reference Hegner, Smith and Bustamante1999; Robertson et al. Reference Robertson, Moses, Kwon, Chan, Chi, Klein, Sung and Greene2009; Van Loenhout et al. Reference Van Loenhout, Van der Heijden, Kanaar, Wyman and Dekker2009).

Single recombinase filaments on dsDNA also appear to have different conformations, as has been demonstrated in flow-stretch experiments combined with fluorescence microscopy (Robertson et al. Reference Robertson, Moses, Kwon, Chan, Chi, Klein, Sung and Greene2009). In these experiments, a fluorescently labelled dsDNA molecule was anchored to a glass surface and RAD51 was flowed into the micro-fluidics chamber. NPF formation was monitored by visualizing the change in length of the dsDNA (Fig. 5). Upon filament formation, the dsDNA length increased, corresponding with the formation of RAD51 filaments in the extended conformation. Removing RAD51 from solution induced a dramatic shortening of the DNA, until a final length was reached corresponding to a compressed conformation. These results indicated that RAD51 could remain DNA-bound in the compressed form for a significant amount of time. At high ATP concentrations (2 mM), filament compression as well as disassembly were inhibited (Fig. 5). In addition, applying cycles of rapid buffer exchange between ATP-containing and ATP-free solutions resulted in reversible transitions between compressed and elongated filaments, without RAD51 detaching from the DNA (Fig. 5). This study clearly indicates that RAD51 filaments can also adopt distinct conformations, tightly coupled to the nature of the nucleotide bound.

More recently, magnetic tweezers have been used to study whether force can induce conformational transitions in RecA–dsDNA filaments (Conover et al. Reference Conover, Danilowicz, Gunaratne, Coljee, Kleckner and Prentiss2011). In this study, single dsDNA molecules were first coated by RecA and then subjected to step-wise force increases and decreases, while monitoring their extension using video microscopy. In the absence of free RecA in solution, rapid (<1 s) force increases (from 40 to 55 pN) were followed by slow (4 nm s⁻¹) and continuous elongation of the DNA, lasting more than 100 s. This observation indicated that applying force to the filament can induce a conformational transition from the compressed to the elongated conformation, resulting in longer DNA tethers.

Single-molecule approaches have thus provided hints of the dynamic nature of recombinase filaments. Several studies have confirmed that both RecA and RAD51 form filaments on DNA differing in length and mechanical properties that can reversibly interchange. On the one hand, filaments in the compressed conformation (which appears to be inactive in catalysing strand exchange) can be converted to the extended configuration if a high-enough ATP concentration is present in solution. On the other hand, tension on the filament also affects filament conformation, indicating the tight coupling between the DNA substrate and the filament structure. Several of the molecular mechanisms underlying the transition between the extended, ATP-bound filament to the compressed, ADP-bound one are unclear. First of all, what is the temporal coordination between ATP hydrolysis, ADP release and the structural compression step? In addition, what structures are adopted when both ATP and ADP are present within the same filament? Is it possible that heterogeneous structures are present within the same continuous NPF? Furthermore, the quantitative thermodynamic model underlying the transition between these two conformations is not known. Future single-molecule experiments, in conjunction with non-equilibrium thermodynamic modelling will be required to gain further insights in this intriguing aspect of recombinase NPFs and their role in the strand exchange reaction.

3.3 Searching for a needle in the haystack

During the homology search process the recombinase NPF needs to find, align and form a stable complex with the complementary sequence on a duplex template. This process needs to be relatively fast and efficient, considering that replication takes approximately 1–2 h (in Saccharomyces cerevisiae (Aylon et al. Reference Aylon, Liefshitz, Bitan-Banin and Kupiec2003)) and in that time a unique sequence has to be found among the many millions of base pairs available. The strategy and mechanism of homology search is not clear at present. A mechanism purely involving one-dimensional (1D) sliding has been ruled out both experimentally (Adzuma, Reference Adzuma1998) and from theoretical considerations (Barzel & Kupiec, Reference Barzel and Kupiec2008). Recently, the first direct insights in the homology search mechanism have been obtained using fluorescence microscopy in combination with dual-trap optical tweezers and microfluidics (Forget & Kowalczykowski, Reference Forget and Kowalczykowski2012). In this study, a single dsDNA molecule was manipulated using optical tweezers in order to control its end-to-end length. Using micro-fluidics and an automatized microscope stage (Fig. 6), the trapped DNA molecule was transferred into a reservoir with no buffer flow present, where it was allowed to interact with fluorescently labelled ssDNA molecules pre-incubated with RecA–ATP-γ-S . After a defined incubation time, the DNA was inspected using fluorescence microscopy. Stable strand-exchange complexes were observed by monitoring the binding of fluorescent NPFs to the homologous sequence in the optically trapped DNA (Fig. 6). It was found that the efficiency and speed of homology pairing decreased drastically when the dsDNA was held in an extended conformation. Stretching the target dsDNA (end-to-end length more than 80% of the contour length) resulted in no binding of homologous NPFs. On the other hand, when the target DNA molecule was relaxed (end-to-end length 20% of the contour length), strand-exchange complexes formed instantaneously. In addition, homology pairing was found to strongly dependent on the length of the NPF used in the experiment (Fig. 6). This finding appears to be inconsistent with a search mechanism solely relying on 1D sliding of the NPF along the target DNA. Additional modes of search need to be available, such as hopping (Gorman & Greene, Reference Gorman and Greene2008) and inter-segmental transfer (dissociation and re-association at a distal segments on the DNA) to explain the dependence of search kinetics on tension on the DNA and length of the NPF. The authors proposed a mechanism in which a long RecA NPF is simultaneously in contact with many segments of the target DNA (parallel homology sampling) through short, reversible synapsis events. When the DNA is not stretched, all conformations can be explored due to Brownian motion, allowing the NPF to interact with distant DNA segments within short time intervals and therefore accelerating the search process.

Fig. 6.

Homology search mechanism performed by RecA NPFs. (A) Schematic representation of the five-channel microfluidic device used in the experiment. (B) Sequential multi-step assembly of the single-molecule assay. (a) Capturing two beads in the optical trap. (b) DNA tethering. (c) DNA manipulation and visualization. (d) YOYO destaining and set of end-to-end distance. (e) Incubation of the dsDNA with the reservoir containing the fluorescently labelled NPF. (f) Return in the imaging buffer and visualization of the formation of the strand exchange complex in the absence of fluorescence background. (C) Visualization of the end-product of the strand-exchange reaction. Two different fluorescent NPFs were used, of 430 and 1762 nucleotides respectively. Each of them is homologous to a different section of the Lambda DNA molecule used as a template. Fluorescence visualization allows confirming that strand-exchange took place at the correct position. (D) The efficiency of the strand-exchange complex formation shows a strong dependence on the conformation of the DNA molecule. At the distance of only 8 μm, corresponding to approximately 50% of the contour length of the DNA molecule the strand exchange reaction is highly efficient. (E) The speed of homologous recognition is strongly dependent on the DNA conformation. (F) TIRF-based assay for studying the synapsis lifetime. (G) Fluorescent image of synaptic complexes. (H) Fluorescence versus time for synapsis events using heterologous (top) or homologous (bottom) sequences. Figures reprinted from (Bell et al. Reference Bell, Plank, Dombrowski and Kowalczykowski2012; Mani et al. Reference Mani, Braslavsky, Arbel-Goren and Stavans2010).

Another set of single-molecule fluorescence experiments addressed the issue of discriminating between homologous and non-homologous sequences in the search process by quantifying how the lifetime of the synaptic complex depends on the degree of homology (Mani et al. Reference Mani, Braslavsky, Arbel-Goren and Stavans2010). In the assay used, short dsDNA molecules were anchored to a glass surface while fluorescently labelled RecA-coated ssDNA molecules were introduced. Stable complex formation between target DNA and NPFs were identified by stable and bright fluorescence spots appearing on the glass surface (Fig. 6). It was found that completely heterologous sequences did form complexes with lifetimes of tens of seconds. Fully homologous sequences, on the other hand were found to be even more stable, with lifetimes of hundreds of seconds. In addition, single mismatches were shown to significantly decrease the synapsis stability.

An innovative study, employing the simultaneous manipulation of two DNA molecules, aimed at quantifying the thermodynamics aspects of synaptic complex (Fig. 7) (De Vlaminck et al. Reference De Vlaminck, van Loenhout, Zweifel, den Blanken, Hooning, Hage, Kerssemakers and Dekker2012). In that assay, a dsDNA molecule was tethered between two optical traps and subsequently melted by applying a high tension resulting in a tethered ssDNA molecule (Gross et al. Reference Gross, Laurens, Oddershede, Bockelmann, Peterman and Wuite2011; Van Mameren et al. Reference van Mameren, Gross, Farge, Hooijman, Modesti, Falkenberg, Wuite and Peterman2009b). This ssDNA molecule was incubated with RecA in the presence of ATP-γ-S to obtain a stable RecA–ssDNA filament. In addition to optical tweezers, the experimental setup also consisted of magnetic tweezers, used to independently stretch and twist a dsDNA molecule tethered to a glass surface. The ssDNA–RecA complex was brought in contact with the dsDNA and interactions between both DNA molecules were assessed. The first experimental goal was to quantify the interaction between the secondary binding site of the RecA filament and the target dsDNA. Positively twisted and uncoiled dsDNA did not interact with the NPF, while slightly stretched (tension 3.5 pN) and negatively supercoiled dsDNA did interact (Fig. 7). Increasing negative supercoiling resulted in increased strength and probability of interacting with the NPF. It was proposed that the negative supercoiling makes short segments of ssDNA available by the transient formation of ssDNA bubbles on the target dsDNA. These ssDNA segments, due to their high affinity for the secondary binding site of RecA filaments (Mazin & Kowalczykowski, Reference Mazin and Kowalczykowski1996) stabilize the strand-exchange complex. The need for negative supercoiling for observing interactions indicates that the secondary binding site alone is not able to provide the necessary energy for melting the target dsDNA and forming a complex with the ssDNA. Spontaneous DNA breathing and supercoiling, therefore, appear to play important roles in the homology search process. In order to be tested for homology, the target dsDNA needs to be first opened and distorted substantially from its native B-form structure, resulting in an energy penalty. When homology is found the energy penalty can be overcome by the base-pairing energy that is gained (Fig. 7). This model therefore provides a quantitative framework for the detailed energetic balance leading to successful sequence recognition.

Fig. 7.

Dual-DNA manipulation to study RecA-mediated strand exchange. (A) Dual-optical trap is used to manipulate a RecA–DNA (either ssDNA or dsDNA) complex. Magnetic tweezers is used to stretch and twist a second DNA molecule (either ssDNA or dsDNA) tethered between the glass surface and a magnetic bead. The two DNA molecules can be brought into contact and slided next to each other to probe their interaction. (B) Push-probe experiment between a RecA–dsDNA complex and an ssDNA. A force of ∼3 pN is required to disrupt the complex. (C) Push-probe experiment between a RecA–dsDNA and dsDNA. No interaction is observed. (D) Effect of positive and negative supercoils of dsDNA molecule. When positive supercoils are applied no interaction is observed (top panel). When negative supercoils are instead applied, a stable complex is formed that requires a force of ∼2 pN to be disrupted. Figures reprinted from (De Vlaminck et al. Reference De Vlaminck, van Loenhout, Zweifel, den Blanken, Hooning, Hage, Kerssemakers and Dekker2012).

In conclusion, the energetics and kinetics of the reaction intermediates leading to the formation of a stable synapse have been addressed at the single-molecule level. Several aspects of this process, however, are still unclear. In the first place, a quantitative model explaining how DNA conformational dynamic affects the efficiency and the speed of the homology search process is still missing. Secondly, the observation that heterologous DNA molecules can pair with high stability in presence of RecA is in apparent contradiction with the requirement of the homology search process to scan fast through the large amount of non-complementary sequences. In addition the experimental assays employed so far were not yet able to directly visualize the actual process of search and recognition of homology and were rather indirect in detecting the intermediate steps leading to a stable D-loop formation. Furthermore, the relative balance between 1D and 3D diffusive mechanisms during homology search is still not known. Although the first experiments addressing the thermodynamic aspects underlying the formation of a stable DNA–nucleoprotein complex have been performed, a precise quantification of the lifetime of the intermediates states is required to elaborate a predictive target–search model explaining how two complementary sequences find each other in the genomic haystack.

3.4 Strand-exchange at the single-molecule level

The NPF, after finding the homologous sequence, catalyses the defining event of the HR process: the transfer of information between homologous sequences, the strand-exchange. Owing to its transient nature, it has been very difficult to isolate and study this key-step of HR using bulk biochemical or structural methods. Single-molecule approaches have tackled this problem, addressing its molecular mechanism in real time on single DNA molecules. In the first single-molecule study of strand exchange, a dsDNA molecule was attached to a glass surface stretched and negatively supercoiled (F=0.5 pN and 35 negative rotations). The changes in twist and end-to-end length of the tethered DNA molecule were monitored, while it was strand-invaded by homologous RecA–ssDNA filaments of different lengths (Fig. 8) (Van der Heijden et al. Reference van der Heijden, Modesti, Hage, Kanaar, Wyman and Dekker2008). As DNA bound by RecA is elongated and unwound, strand-exchange results in the removal of negative supercoils, which in turn corresponds to a drastic elongation of the torsionally constrained DNA. The experiment was performed preassembling RecA on ssDNA using two different nucleotide cofactors: ATP-γ-S and ATP. When ATP-γ-S was present, a gradual and steady increase in the tether length was observed. The final length was compatible with the expected unwinding and stretching due to the invasion of the homologous DNA segment. In fact, both the final elongation as well as the duration of the length increase was linearly dependent on the size of the invading filaments. Therefore, it was possible to provide an estimate of the strand-exchange rate, which was found to be approximately 2 bp s⁻¹. Substituting ATP-γ-S with ATP resulted in a drastically different behaviour. In this case, after an initial phase in which the DNA length started to increase, a long plateau was observed, followed finally by a decrease in length until the original DNA extension was reached. Two analysis methods were applied: in the first one, the total interaction time was quantified (length increase+plateau+length decrease). This quantity was also found to be proportional to the size of the invading recombinase filament, following the same linear dependence found for the ATP-γ-S. In the second analysis method, the duration and amount of the length increase phase were quantified. It was observed that they show no correlation with the length of the invading RecA filament. A hypothesis was formulated, in which under ATP-hydrolysis conditions, strand exchange progresses throughout the homologous region only through a region (strand-exchange window) composed of 80 bases. In this model, the exchange starts at one the invading end (5′), with a rate that is equal to ATP-triggered RecA unbinding. The invading filament locally elongates and unwinds the tethered molecule, increasing the end-to-end distance. At the lagging end, instead, RecA release into solution after ATP hydrolysis and a new heteroduplex is created. Therefore, while strand exchange progresses in ATP-hydrolysing conditions, the end-to-end distance remains constant (Fig. 8). After the reaction is completed, and all RecA proteins dissociated, a three-stranded joint structure is obtained, whose length is identical to the initial length.

Fig. 8.

Strand exchange dynamics at the single-molecule level. (A) A single dsDNA molecule is held at constant force and negatively supercoiled using magnetic tweezers. Strand-exchange results in the increase of the end-to-end length of tether and a shift in the rotational offset of the rotation–extension curve (see (B)). (C) Real-time detection of the strand exchange reaction in the absence (red curve) and presence of ATP hydrolysis (green curve). Absence of hydrolysis produces an irreversible lengthening of the DNA tether which is compatible with a three-stranded structure having a pitch of 14.6 bp/turn and an elongation of 0.51 nm bp⁻¹. When ATP hydrolysis is permitted the lengthening is only transient and corresponds to a moving strand exchange windows of 80 bases. (B) A partial duplex DNA molecule is attached to a quartz glass slide trough a biotin–neutravidin bond. A fluorescent molecule (donor) is covalently attached at the single–double-stranded DNA junction. A single-stranded tail with a 5′-end of various length is used to assemble the RecA NPF in the presence of ATP-γ-S. (C) A donor-labelled dsDNA FRET time trace showing a sudden increase in donor fluorescence (docking), followed by a gradual transition to a high FRET value (Δt _delay). (D) The number of steps before the completion of the strand exchange reaction is plotted against the homology length. The strand exchange reaction occurs at the speed of 3.2 bp step⁻¹ and requires the initial docking of 13.5 nucleotides. (E) Strand separation assay using a donor–acceptor-labelled duplex DNA reveals fast FRET fluctuations, indicating that the outgoing strand interacts with the secondary binding site of the RecA–ssDNA filament. Figures reprinted from (Ragunathan et al. Reference Ragunathan, Joo and Ha2011; Van der Heijden et al. Reference van der Heijden, Modesti, Hage, Kanaar, Wyman and Dekker2008).

In a second study, spFRET was used to study RecA-catalysed strand-exchange, allowing discrimination between docking of a RecA–ssDNA complex to the target dsDNA and further progression of strand exchange (Ragunathan et al. Reference Ragunathan, Joo and Ha2011). The experimental configuration used in the experiment is schematically represented in (Fig. 8). Acceptor-labelled RecA–ssDNA NPFs were assembled in the presence of ATP-γ-S and immobilized. Donor-labelled dsDNA was injected into the flow-cell and both donor and acceptor fluorescence intensities were monitored upon donor excitation. An example of an experimental result is shown in Fig. 8. After an initial lag phase, the donor fluorescence signal suddenly increased (due to the docking of the donor-labelled dsDNA). This state quickly converted (during an interaction time Δt _delay) to a more stable configuration, characterized by a high FRET efficiency, interpreted to represent the final strand-invaded complex. The design of the DNA substrates revealed that the initial binding of the duplex DNA to the NPF (docking) occurred at different random positions involving just a few bases. The irreversible conversion from this intermediate state to a high FRET value was interpreted as the actual occurrence of the strand exchange reaction. Therefore, the distribution of strand-exchange times (Δt _delay, the delay time between the formation of the first intermediate and the stable highest FRET configuration) was fitted with a Gamma distribution (Δt ^N−1e^−kΔt), under the assumption that strand exchange involves N-hidden rate-limiting steps. Performing this analysis with substrates of different lengths indicated that strand invasion proceeds in steps involving three bases and is characterized by a stepping rate of ∼70 bp s⁻¹ (Fig. 8). The observation that strand invasion occurs in 3 bp steps is in good agreement with high-resolution structural data (Chen et al. Reference Chen, Yang and Pavletich2008), which shows that bases in the RecA–ssDNA complex are grouped in triplets. A remarkable behaviour was observed when monitoring the dynamics of the displaced strand of the dsDNA, which is displaced by the NPF. After an initial docking phase (Δt ₁), a phase lasting ∼3 s involving rapid fluctuations in FRET efficiency was detected (Fig. 8). The authors proposed that these fluctuations are the results of intermittent interactions between displaced DNA strand and the secondary binding site on the RecA filament.

How recombinase filaments mediate the transfer of strands from one DNA molecule to the other remains largely obscure. The two single-molecule experiments presented, which addressed this key biological step, present important differences and few similarities. First of all, the estimated strand-exchange rates are very different (70 bp s⁻¹ in the spFRET experiment and 2 bp s⁻¹ in the magnetic tweezers study). One possible explanation is that strand exchange is highly sensitive to the torsional constrains imposed on the target DNA molecule (the DNA ends are free in the spFRET experiment and engaged by the glass and bead surfaces in the magnetic tweezers experiment) and the buffer conditions used (100 mM sodium acetate and 10 mM magnesium acetate in the spFRET experiment and 10 mM MgCl₂ in the magnetic tweezers experiment). As strand-exchange can proceed in the absence of ATP hydrolysis, it is likely that experimental circumstances affecting the stability of the DNA duplex, such as electrostatic screening and supercoiling, play important roles in determining how fast the two strands exchange. Both studies, in fact, point to a model in which the breathing dynamics of the double helix could play a crucial role in determining the kinetics of the strand-exchange reaction, as anticipated by bulk studies (Bazemore et al. Reference Bazemore, Folta-Stogniew, Takahashi and Radding1997; Gupta et al. Reference Gupta, Folta-Stogniew, O'Malley, Takahashi and Radding1999; Xiao et al. Reference Xiao, Lee and Singleton2006). Finally, what is the driving force behind the directionality and progression of the strand exchange reaction? How can unidirectional motion be obtained in the absence of ATP hydrolysis between two states having an equal amount of paired DNA bases? Fortunately, due to the rapid development of single molecules techniques, it is expected that these questions will be addressed in the near future.

3.5 RAD51 disassembly from DNA

The final step of HR is the disassembly of recombinase proteins from dsDNA. Disassembly of RAD51 filaments has been examined by the previously described optical tweezers/flow-stretching strategies (Fig. 2) (Hilario et al. Reference Hilario, Amitani, Baskin and Kowalczykowski2009). In one paper, the filament disassembly was monitored in different buffer conditions after forming the RAD51 filaments in the presence of ATP and either Ca²⁺ or Mg²⁺, When RAD51 filaments were assembled in Ca²⁺–ATP conditions and successively exposed to Mg²⁺ (in the presence of ATP) they did not disassemble (Fig. 9). It is well known that Ca²⁺ induces and preserves the active and extended conformation of the RAD51 filament, by blocking ATP hydrolysis (Bugreev & Mazin, Reference Bugreev and Mazin2004). To explain this observation the authors implicated the slow exchange of the divalent cation bound to RAD51. In contrast, RAD51 filaments formed and visualized in a buffer containing Mg²⁺–ATP (in the absence of Ca²⁺), allowing ATP hydrolysis, showed clear signs of filament disassembly: shortening of the dsDNA and disappearance of the fluorescent RAD51. Remarkably, not all RAD51 disassembled from the DNA presumably because ADP-bound RAD51 filaments remained bound to dsDNA for surprisingly long times, in the order of tens of minutes. The existence of a long-lived RAD51-bound state subsequent to ATP hydrolysis has important consequences for the progression of the HR in vivo, blocking the following steps of the reaction. The authors proposed that recombinase modulators such as the translocating motor RAD54 are needed to fully disassemble ADP-bound NPFs.

Fig. 9.

RAD51 disassembly from dsDNA. (A) Flow-stretching assay for characterizing RAD51 disassembly. RAD51 filaments (RAD51 on the left and RAD51-fluorescein on the right) are assembled in the presence of CaCl₂ and disassembly is observed in the presence of Mg(OAc)₂. A very slow reduction in the DNA-length is observed. [A.I] DNA–RAD51 filaments assembled in the presence of Mg(OAc)₂ shows a significant reduction in length when exposed to a buffer lacking ATP. Disassembly, does not complete after a period of 25 min as can be seen from the presence of fluorescent RAD51 bound along the DNA molecule. (B) The effect of DNA tension on RAD51 disassembly studied using dual-optical trapping and fluorescence microscopy. RAD51–dsDNA filaments are formed in the presence of CaCl₂ and disassembly is observed in MgCl₂. The trap position is kept constant. Disassembly induces a fluorescence intensity drop (red line) and a correlated force increase (blue line). Disassembly eventually stalls when the force reaches forces of 50 pN (t∼400 s). Releasing the tension to 20 pN (t≈400 s) allows the disassembly reaction to reinitiate, generating tension across the DNA. The process can be repeated further until complete disassembly of RAD51 is achieved. Figures reprinted from (Hilario et al. Reference Hilario, Amitani, Baskin and Kowalczykowski2009; Van Mameren et al. Reference van Mameren, Modesti, Kanaar, Wyman, Peterman and Wuite2009a).

A different approach, involving dual-trap optical tweezers, quantitative single-molecule fluorescence microscopy and microfluidics provided more detailed insight in filament disassembly. In contrast to the flow-stretching approach discussed above, the use of two optical traps allows a well-defined tension to be applied on the tethered DNA, constant over its whole length (Van Mameren et al. Reference van Mameren, Gross, Farge, Hooijman, Modesti, Falkenberg, Wuite and Peterman2009a) (Fig. 9). Using this approach, it was observed that high tensions slow down RAD51-disassembly, effectively stalling the process at forces above 50 pN, even under conditions favouring ATP hydrolysis. Upon a sudden release of the tension, large bursts of RAD51 disassembly were observed. Furthermore, RAD51 disassembly did not occur in a continuous manner, but in bursts, interrupted by pauses, with exponentially distributed duration. A new model for RAD51 disassembly was proposed in which (i) ATP hydrolysis occurs throughout the filament, independently of the applied tension (ii) RAD51 monomers within the filament (and not on one of the ends) remain DNA bound, irrespective of the nucleotide bound (ATP or ADP). (iii) RAD51 monomers at filament ends detach from the DNA after ATP hydrolysis. The actual detachment (and not ATP hydrolysis) depends on tension: at low tension monomers fall off almost instantaneous (until an ATP-bound RAD51 remains on the end), at high tension (50 pN), detachment can be totally counteracted. To explain the force-dependent dissociation rate, it was hypothesized that the complex between dsDNA and an RAD51 monomer on a filament end has to contract, before detachment can take place. This contraction is counteracted by applying tension on the DNA. The elastic energy stored in the filament due to the DNA elongation (dsDNA is lengthened by 50% compared with its B-form when in complex with RAD51 or RecA) is therefore the driving force for the release of the NPF from the dsDNA.

These results provide an alternative explanation for the incomplete filament disassembly in the flow-stretching experiments discussed above (Hilario et al. Reference Hilario, Amitani, Baskin and Kowalczykowski2009). In that experiment, the DNA was extended using flow stretching with a solution containing a high sucrose concentration (10% w/v). This method is probably capable of exerting high-enough tensions to severely influence the disassembly process. Despite this inconsistency, both single-molecule studies propose possible molecular strategies to efficiently remove RAD51 from dsDNA. One option, for example would consists in increasing the rate of ATP hydrolysis only at the filament-end and exploiting the natural hydrolysis rate of RAD51 to catalyse the low-affinity DNA-binding state rather than actively wiping off RAD51 by RAD54 translocation through DNA-bound RAD51 filaments. A second open question concerns the molecular mechanism of RAD51 disassembly. Specifically, the temporal coordination between the ATP-hydrolysis, the structural change and the release into solution is unknown. Also, the structural reason why individual monomers cannot dissociates from internal regions of the filament but only from disassembling ends is unclear.