DNA Recombination

Holliday junctions are critical intermediates for homologous, site-specific recombination, DNA repair, and replication. A wealth of structural information is available for immobile four-way junctions, but the controversy on the mechanism of branch migration of Holliday junctions remains unsolved. Two models for the mechanism of branch migration were suggested. According to the early model of Alberts-Meselson-Sigal (Sigal, N., and Alberts, B. (1972) J. Mol. Biol. 71, 789–793 and Meselson, M. (1972) J. Mol. Biol. 71, 795–798), exchanging DNA strands around the junction remain parallel during branch migration. Kinetic studies of branch migration (Panyutin, I. G., and Hsieh, P. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 2021–2025) suggest an alternative model in which the junction adopts an extended conformation. We tested these models using a Holliday junction undergoing branch migration and time-lapse atomic force microscopy, an imaging technique capable of imaging DNA dynamics. The single molecule atomic force microscopy experiments performed in the presence and in the absence of divalent cations show that mobile Holliday junctions adopt an unfolded conformation during branch migration that is retained despite a broad range of motion in the arms of the junction. This conformation of the junction remains unchanged until strand separation. The data obtained support the model for branch migration having the extended conformation of the Holliday junction.

The Holliday junction (HJ) 1 suggested in 1964 by Robin Holliday (6) is a central intermediate in homologous and sitespecific recombination (7). This type of DNA structure is also involved in double-stranded break repair (7,8). Recent data show that HJs are critical intermediates in replication fork stalling leading to subsequent correction of the corrupting template lesion (9 -11). Movement of the crossover along DNA allowing for length extension of the heteroduplex is termed branch migration. If this process is not terminated by resolvases, branch migration leads to complete strand separation as shown in Fig. 1A. Branch migration, whether spontaneous or mediated by proteins, is a key step in various genetic processes involving the Holliday junction. Various models for the Holliday junction intermediate have been utilized to unravel the structural basis for branch migration and resolution of the junction. The immobile four-way junction was a primary model system for the Holliday junction, and a great deal of information on the structure of HJs was obtained from their study. Numerous techniques (12)(13)(14)(15)(16)(17), including very recent x-ray crystallography analysis (18,19), show that in the presence of multivalent cations, the junction adopts an antiparallel orientation in which the four helices stack in pairs to form two double-helical domains. Immobile HJs can adopt two conformational states undergoing transition between them via extended conformation as a transient state (20). However, these findings are in contrast to data obtained for HJs formed by inverted repeats (2-fold sequence symmetry), which allow branch migration to occur. Atomic force microscopy (AFM) imaging of a supercoil-stabilized cruciform (21) shows that the cruciform adopts various conformations and that parallel orientation of the arms is the predominant HJ structure in the presence of Mg cations (22). This finding is in line with results on the structure of the 2-fold symmetry intermediate for the Flp recombination reaction even after the removal of the protein (23). It has been hypothesized that DNA supercoiling and the mobility of the junction are important factors contributing to HJ conformation (21,22). Little is known as to how the structural features of the HJ relate to the mechanism of branch migration. A textbook model suggested 30 years ago (1) utilizes the parallel orientation of exchanging strands (Path I in Fig. 1A). In this model, hybrid DNA molecules are formed by rotatory diffusion of the arms (2). This view has been challenged with an alternative model suggesting that an extended configuration of the junction (Path II in Fig. 1A) is more appropriate for spontaneous movement of the HJ (3)(4)(5). However, there is no direct experimental evidence for either of these models. In this study, we have used single molecule AFM (21, 24 -26) and a mobile HJ capable of spontaneous branch migration to test the models and to observe the conformations and dynamics of this biologically relevant model. We show that the mobile junction adopts an extended conformation during branch migration supporting the Path II model for branch migration (Fig. 1A).
AFM Procedures-Mica functionalized with aminopropyl silatrane (APS-mica) was used as an AFM substrate (26). Briefly, DNA samples (3-5 l) were placed onto APS-mica for 2 mins, and the mica was rinsed with deionized water (Continental Water System Co., San Antonio, TX) and dried with argon. Images were acquired in air using a MultiMode SPM NanoScope III system (Veeco/Digital Instruments, Santa Barbara, CA) operating in tapping mode using OTESPA probes (Digital Instruments, Inc.). The length, height, and angle measurements were performed using Femtoscan software (Advanced Technologies Center, Moscow, Russia). For imaging in aqueous solutions, the sample was diluted in TNM buffer right before application to APS-mica to make the final DNA concentration 0.8 ng/l. The sample was placed onto APS-mica mounted on the scanning stage of the MultiMode SPM NanoScope III system. The optical head with mounted liquid cell was placed on the sample, and the tip was brought into contact with the surface. TNM buffer was added to fill up the space between the cell and mica surface as needed. Images were acquired in tapping mode using standard silicon nitride probes (Veeco/Digital Instruments). Buffer was changed by injection of TE buffer (10 mM Tris-HCl, pH 7.9, and 5 mM EDTA) with 1 ml plastic syringes attached to the inlet and outlet holes of the flow cell.

RESULTS AND DISCUSSION
The schematic illustrating the procedure for preparation of the mobile HJ is shown in Fig. 1B. The HJ was obtained by hybridization of left and right hemi-junctions having homologous duplex regions via annealing of the single-stranded ends (3,4). The annealing was performed in Mg 2ϩ -containing TNM buffer to slow down branch migration (3,4). A partial sequence of the center of the junction illustrating the initial one-step migration is shown in Fig. 1C. A complete sequence for synthetic oligonucleotides forming the mobile HJ is given in the "Experimental Procedures" section. Removal of Mg 2ϩ speeds up branch migration, which eventually leads to the formation of two DNA duplexes as shown at the bottom of Fig. 1A. Gel electrophoresis analysis showed that more than 90% of the HJs were converted into linear duplexes after a 1-hour incubation in TE buffer at 50°C (data not shown). The HJs were separated from unreacted hemi-junctions by agarose gel electrophoresis and extracted from the gel. Magnesium cations in a 10 mM concentration were present at all steps of the sample purification to decrease the rate of branch migration. AFM data for the sample deposited onto APS-mica and subsequently argon-dried are shown in Fig. 2.
All of the molecules in Fig. 2 are four-way DNA junctions. The lengths of the arms for the molecules vary. This is the expected result for HJs undergoing branch migration as typified by the one-dimensional random walk model (3,5). The conformations of the HJs were classified in the following way. The HJ conformations in which the long arms form an almost continuous strand and with the short arms situated on opposite sides of the long filaments are referred to as being in the trans conformation (type 1 in Fig. 2). The extended conformation, in which the short arms are perpendicular to a continuous long filament, is a subset of the trans conformation. The conformations of the HJs having like arms positioned adjacent to one another are referred to as being in the cis conformation (type 2 in Fig. 2). The parallel conformation, having an angle of 0 -90 degrees between homologous arms is a subset of the cis conformation. The antiparallel conformation is characterized by angles less than 90 o between a short arm and the closest long arm and can be trans or cis conformations (27).
To look directly at the dynamics of mobile HJs and to test the models of branch migration, we employed the capability of AFM to acquire images of fully hydrated samples, omitting the drying step. In this time-lapse imaging mode, AFM performs continued scanning over a selected area allowing the observation of molecular dynamics at the single molecule level (21,24,26). In addition, imaging in aqueous solutions eliminates potential sample preparation artifacts induced by drying of the sample. This AFM imaging mode has been very effective for the single molecule observation of the dynamics of supercoiled In the time-lapse experiments, the sample was deposited onto APS-mica (28) and imaged with AFM continuously immediately after the AFM tip engaged the surface. The results of such experiments performed in TNM buffer are shown in Fig.  3. Molecule 1 is in the cis conformation in Fig. 3A, but adopts the trans conformation in Fig. 3B. There is another molecule (2) on the same scanning area that undergoes conformational transitions. This molecule is in the trans conformation with variable angles between the short and long arms in plates A and B, but the junction adopts the cis conformation in Fig. 3C. The molecule keeps changing and the conformation captured in Fig. 3D is probably an intermediate state of the HJ changing back to the trans conformation. For clarity, the conformational transitions for these two molecules are shown in Fig 3E. These images are traces of the molecules on a black background. In this figure, 1a and 1b and 2a, 2b, 2c, and 2d are the traces of molecules 1 and 2, respectively.
We continued the time-lapse observations in an attempt to visualize the process of strand separation. To facilitate branch migration, Mg 2ϩ -containing TNM buffer in the AFM cell was replaced with TE buffer without the termination of scanning. The set of 4 images captured after changing of the buffer is shown in Fig. 4, A-D. The molecule indicated with the arrow is in the trans conformation in the initial image (Fig. 4A). The arms subsequently move to an extended conformation (Fig.  4B). One of the short arms is barely seen in Fig. 4C probably due to branch migration resulting in the dissociation of the molecule into two linear strands as imaged in Fig. 4D.
To graphically illustrate the dynamics of this HJ, the images of the molecule were copied onto a blank area from a separate image. The subset containing 18 images of the entire data set (54 frames) is shown in Fig. 4E. This data set includes images illustrating the dynamics of the junction in TNM buffer in which the HJ changes conformation from cis to trans (images 1-4 in Fig. 4E shown in black on a white background). The majority of the images of this data set show the dynamics of the HJ observed after the removal of Mg cations (white molecules on a black background). They show that the mobility of the arms is considerably higher in TE buffer than in TNM buffer. Lower mobility of the arms in the presence of Mg 2ϩ is explained by bridging of the DNA molecules to the mica surface in addition to electrostatic interaction of the negatively charged DNA backbone with protonated amines of immobilized aminopropyl groups (24,26). Surprisingly, despite an elevated mobility in TE buffer, the HJ does not change from the trans conformation into the cis (images 8 -14). Moreover, the transition to the extended conformation prior to strand dissociation was observed in all images acquired (images 6 -15). These observations suggest that the extended conformation, rather than the cis conformation is required for branch migration.
We measured the lengths of the arms of the molecules in the images in Fig. 4E and the data are plotted in Fig. 4F. The data show that the lengths of the arms fluctuate without considerable change during the entire observation period. Branch migration leading to the dissociation of the junction is observed as a one-step process between images 15 and 16 in Fig. 4E. However, this is not a surprising finding. The time interval between the two imaging frames is 1.4 min, and according to Refs. 3 and 4, the characteristic time for strand exchange via spontaneous branch migration modeled by a random walk mechanism is 1-2 s. The interaction of the molecule with the surface should slow down the kinetics, therefore a 100-fold difference for free molecules and those confined to the surface is quite reasonable. In fact, spontaneous branch migration is a one-dimensional random walk; therefore the length of the arms can change in either direction.
Another example of HJ dynamics is shown in Fig. 5. Three frames on the top (A, B, and C) show the dynamics (A and B) and the strand separation (C), and the set of images below (D) illustrates the dynamics on 13 frames including the first two frames taken in TNM buffer. Similar to the data set in Fig. 4E, the traces of the molecules placed on a black background are shown for simplicity in observing the dynamics. Interestingly, this junction retains the cis conformation even after the buffer change (frames 3-6) and dissociates rather quickly after adopting the trans conformation. Linear molecules separate pretty quickly after strand separation (image 13). The graph in Fig.  5E shows the changes in the lengths of the arms during the entire observation period. There is a substantial change in the lengths of the arms starting in frame 7 after the transition from the cis to the trans conformation occurred. The short arms after image 9 start growing and slightly fluctuate in length before the abrupt dissociation of the junction occurs. The lengths of the long arms change in an opposite fashion, as one would expect for coordinated arms changing lengths during branch migration. The strand dissociation also occurs abruptly (images 12 and 13) as we observed for the set in Fig. 4. The complete sets of images for both experiments assembled as animated  files can be viewed in the supplementary materials attached to the paper (Figs. S1 and S2, respectively). We observed substantial elongation and shortening of the arms prior to strand dissociation. The data in Table I show that the shortening of shorter arms A and B accompanied the elongation of longer arms C and D. Similar analysis was applied to the dynamics of HJs in Mg 2ϩ -containing TNM buffer. The results for two molecules imaged in the same experiment are shown in Fig. 6. The data set in Fig. 6A shows the dynamics of the molecule in the trans conformation. The graph below (B) shows the coordinated arms changing for images 3, 4, and 5. Short arms are shortened and long arms are elongated. The dynamics of the molecule in the cis conformation is analyzed in Fig. 6, C and D. There is no substantial change in the lengths of the arms for this molecule. Similar to previous data sets (Fig. 3), we have never observed strand dissociation of the junctions in TNM buffer, suggesting that Mg 2ϩ cations slow down the kinetics of branch migration. This conclusion is in line with the branch migration results of Hsieh and colleagues (3)(4)(5) who reported a 1,000-fold drop in branch migration rate in the presence of Mg 2ϩ . Our data also show that HJs in the presence of Mg 2ϩ cations are very dynamic adopting various conformations including extended and folded parallel and antiparallel conformations. The dynamics of two molecules can be viewed in animated files S3 and S4 in the supplementary materials to this paper. Such broad dynamics of HJs is in line with recent findings (20) where single molecule fluorescence resonance energy transfer (FRET) was applied to immobile HJs undergoing transitions between two folded crossover conformations via the extended conformations as an intermediate state.
The major conclusion from these data is that the extended conformation is the predominant geometry of the HJ during branch migration. Moreover, the transition of HJs from the cis to the extended geometry is required for branch migration to occur, suggesting that the folding of HJs impedes branch migration. These observations are consistent with the branch migration model of Refs. 3-5, suggesting that the unfolding of HJs from the conformation stabilized by Mg 2ϩ was necessary for rapid branch migration. Our data show that neither folded conformation of HJs is involved in branch migration. Moreover, the unfolding of the junction into an extended conformation is needed for branch migration to occur.
Our data show that the extended conformation is a rather stable state for mobile HJs even in the presence of Mg 2ϩ cations and there is a spontaneous transition between different conformations suggesting that the barrier between various conformations is comparable with the kT value. This conclusion is consistent with energetic analysis of the various models of HJs performed (29). In Escherichia coli, branch migration is the protein-mediated process in which the extended conformation of the junction is stabilized by RuvA protein even in the presence of Mg 2ϩ cations (22,30). Similar conformations of HJs are found in a number of site-specific recombination systems (31). Altogether, this evidence suggests that an extended conformation of mobile Holliday junctions is their inherent conformation during branch migration, and that proteins are needed to select the unfolded conformation of the junction and stabilize this conformation for protein-mediated branch migration.
Single molecule AFM was successfully used in this work to assess the dynamics of HJs under conditions allowing branch migration. The advantage of time-lapse AFM imaging is that this technique enables one to observe the dynamics of an entire molecule with nanometer range resolution. Importantly, several molecules in the scanning area can be imaged and analyzed simultaneously allowing the comparison of the dynamics of different molecules at fully identical environmental conditions. However, the dynamics of molecules confines to the space of a surface-liquid interface and cannot be compared with the unrestricted dynamics of molecules free in solution. Our immobilization approach enables us to decrease the strength of the surface-DNA links allowing the observation of extensive dynamics of HJs including the separation of strands, but the rate of branch migration is several orders of magnitude lower in AFM experiments than for freely diffusing molecules (see the estimates above). Therefore, single molecule AFM is a technique that provides information on the range of conformational changes of the molecules rather than the time scale of these changes, which can be retrieved by other single molecule techniques. Note in this regard that a broad dynamics of the arms of mobile HJs at conditions unfavorable for branch migration (in the presence of magnesium cations) is consistent with the recent findings (20) where the arms dynamics of immobile HJs was analyzed with FRET at the single molecule level.