Learn more: PMC Disclaimer | PMC Copyright Notice
A DNA translocating Snf2 molecular motor: S. cerevisiae Rdh54 displays processive translocation and can extrude DNA loops
Associated Data
Abstract
We have used total internal reflection fluorescence microscopy (TIRFM) to investigate the characteristics of the yeast homologous recombination factor Rdh54 on DNA. Our results demonstrate translocation of Rdh54 on DNA and extrusion of DNA loops by Rdh54 in an ATP hydrolysis-dependent manner. The translocating Rdh54 was highly processive and displayed a variety of different behaviors, including variations in translocation rate and distance, pauses, and reversals. We provide evidence that The DNA loops generated encompass an average of six kilobases and Rdh54 often abruptly releases the extruded DNA. Rdh54 forms a multimeric complex which we speculate has at least two functionally distinct DNA-binding sites, one of which enables translocation while the other remains anchored to another DNA locale. Our work, together with other recent studies, suggests that translocation-coupled DNA loop extrusion is a common mechanistic feature among the Snf2-family of chromatin-remodeling proteins.
INTRODUCTION
Rdh54 belongs to the Snf2-family of chromatin-remodeling proteins and is required for mitotic and meiotic DNA recombination 1; 2. The Snf2-family of proteins are structurally related to the Saccharomyces cerevisiae chromatin-remodeling protein Snf2. These proteins possess seven conserved motifs labeled I, Ia, Ib, II, III, IV and V 3; 4. Motifs I and II are the Walker A and B nucleotide-binding motifs commonly found in ATP hydrolyzing enzymes. Snf2 proteins are ubiquitous in nature and are involved in many aspects of DNA metabolism, including chromatin remodeling, DNA replication, transcription, translation, and DNA repair 5. A recent analysis of public databases by Owen-Hughes and colleagues has classified the Snf2 proteins into at least 24 distinct subfamilies 5. Some of the better characterized Snf2 proteins are the ATPase subunits of complexes such as SWI/SNF, ISWI, RSC, NURF, ACF, CHRAC, INO80, Swr1, NURD, and the DNA repair protein Rad54. S. cerevisiae alone has 17 known Snf2 proteins that play important roles in a broad range of biological processes 5. Although originally regarded as helicases, these proteins appear to be ATP hydrolysis-dependent motors that translocate along duplex DNA 6; 7; 8.
Rad54 is the defining member of one Snf2 subgroup (the Rad54-like subfamily) and is among the most well-characterized proteins of the Snf2-family 9; 10. RAD54 was originally identified in S. cerevisiae as a member of the RAD52 epistasis group required for the repair of DNA double-strand breaks (DSBs) via homologous recombination, and mutations in RAD54 lead to increased sensitivity to DNA damaging agents 2. The crystal structure of zebrafish Rad54 revealed that the protein has a pair of tandemly repeated RecA-like folds, which contain the seven conserved helicase motifs 11. Similar domains are found in the SF1 DNA helicases PcrA, UvrD, and Rep, and the SF2 proteins RecG, UvrB, eIF4A, and NS3 4. It is these conserved RecA-like domains that mediate ATP-hydrolysis and DNA translocation. Early work with Rad54 had suggested that the protein was a DNA translocase 8; 12 and this prediction was confirmed in a recent single-molecule study, which demonstrated that Rad54 translocates on double-stranded DNA in an ATP hydrolysis-dependent manner 13. Rad54 appears to utilize its DNA translocase activity to remove Rad51 from duplex DNA 14 and to process various recombination DNA intermediates by branch migration 15; 16.
Rdh54 (Rad homolog 54) is another member of the Rad54-like Snf2 subfamily and was identified based on sequence homology with Rad54, and also independently found as Tid1 in a yeast two-hybrid screen for proteins that interact with the meiosis specific recombinase Dmc1 1; 17; 18. The role of RDH54 in homologous recombination was verified by genetic analyses, which revealed that null mutants are defective in meiotic recombination and crossover interference and show a deficiency in mitotic recombination, DNA repair, and DNA damage checkpoint adaptation, thus placing RDH54 within the RAD52 epistasis group 1; 17; 18; 19; 20. Rdh54 and Rad54 are closely related (37% sequence identity and 55% similarity) and appear to be somewhat functionally redundant. Cells can survive in the absence of one of the two proteins, however, rad54 rdh54 double-mutants exhibit growth defects and are more sensitive to DNA damaging agents than the single mutants. During normal cell growth, Rdh54 is found at kinetochores and may facilitate communication between the DNA damage and spindle checkpoints 21. Exposure of cells to γ-irradiation causes Rdh54 to partially redistribute to DNA repair centers, which appear as foci comprised of many different DNA repair and checkpoint proteins 21. In vitro experiments have revealed that Rdh54 is a robust ATPase that modifies the topology of DNA, suggesting that the protein could translocate on duplex DNA 22. Moreover, Rdh54 promotes Rad51-catalyzed strand invasion of duplex DNA 22, removes Rad51 and Dmc1 from DNA 23; 24, remodels chromatin in vitro, and may help establish the accessibility of chromatinized DNA templates during homologous recombination (Y.H. Kwon and P.S., unpublished observations).
To begin probing the functions of Snf2 proteins in DNA repair, we sought to develop a system for visualizing the interactions between Rdh54 and duplex DNA at the single-molecule level. Here we used TIRFM 25 and microscale engineered DNA curtains 26 to directly observe the behaviors of quantum dot-labeled Rdh54 complexes as they interact with DNA. We show that Rdh54 behaves as a oligomeric complex that exhibits several modes of interaction with DNA, including stationary binding, ATP hydrolysis-driven translocation, changes in velocity, transient pauses, and changes in translocation direction. We have also characterized the collisions between two different complexes of Rdh54 traveling in opposite directions on the same DNA molecule. The colliding Rdh54 complexes are unable to bypass one another and neither of the colliding partners is displaced. Rdh54 also promotes the extrusion of large DNA loops in a reversible reaction that is coupled to DNA translocation. The DNA loops could be released in an abrupt event consistent with the sudden loss of contact between the DNA and one of the protein complexes. Loop release could also occur via a slower process that appeared to arise from backtracking or reversal of the Rdh54. The formation and release of these DNA loops implies a molecular architecture for Rdh54 that must include at least two different DNA-binding sites with distinct biochemical activities to accommodate both stationary DNA binding as well as active translocation. Our study provides evidence that DNA loop extrusion represents a common mechanism by which Rdh54 and other Snf2 chromatin-remodeling proteins alter DNA topology to influence the outcomes of various DNA transactions.
RESULTS
“High-throughput” single-molecule assay for viewing Rdh54
We have developed a new technology that allows us to assemble “DNA curtains” at defined positions on the surface of a fused silica microfluidic sample chamber (Figure 1A) 26. In brief, a fluid lipid bilayer is deposited onto the surface of the sample chamber and DNA molecules are tethered directly to the bilayer via a biotin-neutravidin linkage. The tethered DNA molecules are then organized along the leading edges of microscale diffusion barriers by the application of a hydrodynamic force, which also extends the DNA molecules parallel to the surface of the sample chamber and confines them within the detection volume defined by the penetration depth of the evanescent field (Figure 1A). This approach allows us to simultaneously visualize up to hundreds of physically aligned DNA molecules in real time within a single field-of-view using TIRFM (Figure 1B).
(A) The upper panel shows a schematic illustration of the DNA curtains assembled at the leading edge of a microscale diffusion barrier on the surface of a flow chamber that was coated with a fluid lipid bilayer. The lower panel depicts just one DNA molecule, and illustrates its response to changes in buffer flow and its relative position within the evanescent field. (B) Is an image of an actual DNA curtain stained with YOYO1 shown in the presence and absence of buffer flow. “T” and “F” indicate the tethered and free ends of the DNA molecules, respectively, and these designations are used throughout. These and all other images and graphs are oriented such that the direction of buffer flow is from top to bottom. The observed length of the DNA was ~12.5 μm, yielding a mean extension (〈×〉/L) of 0.8, which corresponds to an applied force of ~0.5 pN 26. (C) Shows an image of Rdh54 bound to the DNA curtain in the presence and absence of buffer flow. The protein was labeled with quantum dots and the DNA was not labeled. There are 264 individual complexes (fluorescent particles) of Rdh54 in the field-of-view and (D) shows a histogram of binding site distributions. The number of fluorescent particles was counted using an algorithm developed for automated analysis of the DNA curtains (See Supplemental Experimental Procedures and Figure S1), and any particles that did not diffuse away from the surface when buffer flow was transiently stopped were discounted. Also refer to Supplemental Video 1.
To visualize the behavior of Rdh54, the protein was labeled with an antibody-coupled fluorescent semi-conducting nanocrystal (quantum dot or Qdot). Quantum dots are an ideal fluorophore for single-molecule imaging because they are extremely bright and they do not photo-bleach on timescales relevant for biological measurements. For labeling, Rdh54 was expressed with a N-terminal thioredoxin tag, and the quantum dots were covalently coupled to anti-thioredoxin antibodies (see Supplemental Experimental Procedures). Thioredoxin-tagged Rdh54 is functional in vitro and in vivo 23, and ATPase assays revealed that Rdh54 was active even in the presence of an excess of either antibody or antibody conjugated quantum dot, indicating that its bulk biochemical properties were not modified by the labeling procedure (Supplemental Experimental Procedures; Figure S3).
After locating the YOYO1-stained DNA curtains the interchelating dye was removed with a brief rinse of 0.5 M NaCl (Supplemental Information), the labeled Rdh54 (50 μl of 2.5–5 nM) was injected into the sample chamber along with 1 mM ATP, and the unbound proteins and Qdots were rinsed from the sample chamber. As shown in Figure 1C, Rdh54 bound the DNA molecules within the curtain and could be identified as isolated fluorescent signals. An automated particle counting algorithm (Supplemental Experimental Procedures and Figure S1) revealed a total of 264 individual Rdh54 complexes within the field-of-view, and the locations of the proteins on the DNA was random, indicating that there were no preferred binding sequences (Figure 1C and D). Buffer flow was then transiently paused causing the DNA molecules and bound proteins to briefly diffuse out of the excitation volume. This procedure was used as a standard control in all of our TIRFM experiments to verify that the Rdh54 was bound to the DNA and to identify any proteins within the field-of-view that were nonspecifically adsorbed to the surface so that they could be omitted from further analysis (Figure 1B and C). Supplemental Video 1 shows a complete sequence of Rdh54 injection, DNA binding, subsequent removal of the unbound protein, and the transient pause in buffer flow used to verify binding. Importantly, control experiments showed that the quantum dots alone did not bind to the DNA and therefore do not contribute to the measurements (data not shown).
ATP hydrolysis-dependent DNA translocation by Rdh54
Recent studies have shown that Rad54, SWI/SNF, and RSC can translocate on DNA, suggesting that this is a common attribute shared among the Snf2 family members 7; 8; 13; 27; 28. To determine whether Rdh54 could translocate along the DNA, the protein was injected into the sample chamber along with 1 mM ATP and the behavior of the bound proteins was monitored over time by capturing videos at 8.3 frames per second. Approximately 50% of the Rdh54 complexes moved along the DNA during the observation, while the rest appeared to remain stationary (see below). Importantly, we do not believe that the immobile proteins were inactive, because when viewed over longer time intervals we would often see apparently immobile proteins begin moving and vice versa (see below and Figure 6). The kymograms in Figure 2A illustrate the spatial and temporal behavior of Rdh54. These images were generated by excising a region-of-interest (ROI) that corresponded to one DNA molecule from within the curtain and plotting this excised image as a function of time over a 250-second interval. As shown in figure 2A, Rdh54 was able to translocate rapidly along the DNA when ATP was present in the reaction mixture. For those proteins that displayed translocation activity, the movement could occur either against the direction of buffer flow (figure 2A, top panel) or with the flow (Figure 2A, bottom panel), strongly suggesting that it was bona fide DNA translocation.
Examples of Rdh54 translocation in 1 mM ATP are shown as kymograms in (A) with the tethered end of the DNA located at the top of the panel and the free end located at the bottom. Kymograms were generated by excising a 3 × 80 (W × H) pixel ROI and plotting the excised image as a function of time over the indicated interval. The upper panel illustrates Rdh54 movement against buffer flow and the lower panel shows an example that moves with buffer flow. Time-dependent variations in particle intensity are due to “blinking” of the quantum dots and Brownian motion of the DNA that causes the proteins to fluctuate within the evanescent field. (B) Shows the kymogram of a single translocating complex of Rdh54 (upper panel), along with the corresponding particle-tracking data superimposed on the image of the protein (middle panel), or shown independently as a graph of the movement (lower panel). This data was collected using an algorithm that located and tracked the centroid position of each fluorescent particle within the DNA curtain (Supplemental Experimental Procedures for additional details of the analysis procedure). Linear fits to the translocation data are also indicated along with the corresponding translocation rates. (C) Histograms generated from the analysis of 64 different translocating Rdh54 complexes showing the distribution of translocation rates and total distance traveled during the 250-second observation windows. For the unidirectional movement we measured the total distance traveled from the initial to the final position, and for Rdh54 molecules that reversed direction, we measured the maximum distance traveled before turning back. A movie showing the translocation of Rdh54 on DNA is shown in Supplementary Video 2.
Panel (A) shows a kymogram with an example of synchronous movement of different Rdh54 complexes bound to the same molecule of DNA. The upper panel shows the image sequence and the lower panel has superimposed particle-tracking data. The red trace is an immobile Rdh54 complex that serves as a stationary reference point. The green, blue, purple and light blue traces highlight Rdh54 complexes that move synchronously as a consequence of loop extrusion. (B) These graphs highlight details of each looping event (5 total) and the release of each loop is indicated with an arrowhead. At least three distinct Rdh54 complexes are extruding loops in this example. (C) Presents a histogram depicting the lengths of DNA loops generated by Rdh54. This data encompasses 80 total looping events observed on 70 different molecules of DNA. A movie showing transient loop formation by Rdh54 is presented as Supplementary Video 4.
To further analyze the movement of Rdh54 on DNA, we used an automated single-particle tracking algorithm to locate the centroid position of each complex (Supplemental Experimental Procedures), and a detailed example of this analysis is presented in Figure 2B (Supplemental Video 2) 29. This example illustrates the movement of a single molecule of Rdh54 against the direction of buffer flow. The center panel shows the data generated from the particle-tracking algorithm superimposed on the image of the translocating protein. The bottom panel shows the graph of the protein’s movement; the translocation rates were determined from the slopes of linear fits to the tracking data. As illustrated in this figure, the movement of the proteins was heterogeneous and the same Rdh54 complex could display a variety of translocation rates during the course of a single observation. Based on analysis of experiments performed in 1 mM ATP, the average Rdh54 molecule translocated at a rate of 80 bp/sec (0.021 μm/sec; Figure 2C), and the translocation rate decreased as a function of ATP concentration (Figure S2). Rdh54 could translocate an average distance of 13 kilobases during the 250-second observations, indicative of a high degree of processivity (Figure 2C). Neither the translocation rate nor the processivity were appreciably influenced by which direction the protein was moving with respect to the buffer flow (Figure S2). There was a subset of protein molecules that appeared to be pushed by buffer flow, but this movement could be readily distinguished because it was much faster than the mean translocation velocities and only occurred in the direction of flow force (Figure S2). The majority of the proteins did not dissociate from the DNA during the course of the 3–6 minute observations and many of the molecules continued moving even after extended periods of time (see below). Interestingly, although we observed several instances where Rdh54 translocated to the free ends of the DNA, we have never observed the protein dissociate as a consequence of reaching the end of the molecules.
As expected, translocation did not occur in the absence of ATP (Figure 3A), or in the presence of ADP or ATPγS (Figure S6; data not shown), indicating that nucleotide hydrolysis was required for movement. In the absence of nucleotide cofactor, most of the complexes remained stationary or moved only in the direction of buffer flow (Figure 3A). We attribute this type of movement to one-dimensional sliding of the protein along the DNA, which was biased in the direction of flow due to the hydrodynamic force exerted by the buffer. We next tested whether the stationary Rdh54 bound to DNA in the absence of nucleotide cofactor could resume its motor function upon addition of ATP. As shown in figure 3A, when ATP was injected into the sample chamber the stationary Rdh54 complexes began rapidly moving along the DNA. This translocation activity was indistinguishable from that observed when ATP was present throughout the reaction indicating that the stationary complexes were bound in a stalled state. This observation was supported by bulk experiments, which demonstrated that Rdh54 retained ATP hydrolysis activity in solution when DNA was present, even though the protein was rapidly inactivated in the absence of DNA (Figure S3).
In (A) wild-type Rdh54 was bound to the DNA in the absence of ATP. 1 mM ATP was then injected into the sample chamber at approximately 100 seconds after initiating data collection (as indicated by an arrow) at which point the stationary complexes began to rapidly translocate along the DNA. The upper panels show examples of kymograms with proteins moving either with or against flow, and the lower panel shows a graphical representation of the same data. At the outset of the reaction one of the complexes is gradually pushed along the DNA by buffer flow in the absence of ATP (refer to the text for further discussion). (B) Shows the binding of the Rdh54 ATPase mutant K352R to a DNA curtain and a histogram of the binding site distribution. The lower panel shows a kymogram of the Rdh54 K352R mutant bound to DNA in the presence of ATP. Particle-tracking data (red and green traces) are superimposed on two of the Rdh54 complexes.
To confirm that translocation was dependent on nucleotide hydrolysis, we also assayed a mutant of Rdh54 harboring a point mutation in the Walker A nucleotide binding domain (K352R) that renders it defective for ATP hydrolysis (Figure 3B) 23. The Rdh54 K392R protein bound to the DNA molecules (Figure 3B, top panels), however, very few of the proteins were observed moving, even in the presence of ATP (Figure 3B, bottom panel). Those that did move did so slowly and moved in the direction of flow, suggesting that they were pushed along the DNA with the force exerted by the buffer. Occasionally, some of the proteins were observed slowly oscillating over short distances between two stationary proteins (see Figure 3B, lower panel for an example). This type of oscillatory motion was consistent with a one-dimensional diffusion mechanism 29.
Heterogeneous translocation behaviors
Analysis of yeast Rad54 has revealed that the protein displays remarkably uniform translocation kinetics with 80% of protein molecules moving monotonically in one direction 13. In contrast, the majority of the Rdh54 molecules displayed heterogeneous translocation behaviors and variations in kinetics. The most common behaviors included halted translocation, transient pauses for varying durations, forward translocation followed by rapid reversals, and forward translocation followed by more gradual reversals (Figure 4A). The motor proteins could also undergo repetitive cycles of forward and reverse translocation events and during these cycles they often appeared to return to their original locations (Figure 4A, upper right panel, and see below). Interestingly, many of the Rdh54 complexes paused for long periods of time, but then resumed translocation later during the course of the experiment (Figure 4B). This provided further support for the hypothesis that the protein could reversibly enter a stalled state that was inactive for translocation yet remained stably bound to the DNA and capable of resuming movement.
(A) These panels present traces of individual Rdh54 complexes that are representative of the various behaviors observed as they translocated on the DNA. (B) Shows a histogram of pause time distributions. For this analysis, only the events where translocation was resumed were scored as pauses. (C) Shows examples of kymograms depicting collisions between different complexes of Rdh54 bound to the same molecule of DNA.
We also observed collisions between fluorescent Rdh54 complexes traveling in opposite directions along the same DNA molecule (Figure 4C). Some of these collision events appeared to result in the merger of two independent complexes, which could then travel together along the DNA (Figure 4C, upper panel), whereas other colliding proteins either reversed direction or stopped moving (Figure 4C, lower panel). Although these Rdh54 complexes were all labeled with the same color fluorophore it never appeared as though two colliding entities could bypass one another while translocating on the DNA (see below), nor did either of the colliding partners dissociate from the DNA, suggesting that they were tightly associated with the duplex. We verified that approximately 50% of the quantum dots in our samples were “dark” (Supplemental Information), therefore it is possible that the pausing and reversal events may have resulted from collisions between fluorescent and “dark” proteins.
Oligomeric state of Rdh54 complexes bound to DNA
The oligomeric state of Rdh54 in solution was examined by gel filtration experiments and revealed an apparent molecular mass of 249 kDa (Figure S4), which is twice that of a thioredoxin tagged Rdh54 monomer (predicted molecular mass of 125 kDa). Based on these results we concluded that Rdh54 behaved as a dimer under solution conditions. We next analyzed the conjugated Rdh54-Qdot complexes by analytical ultracentrifugation (Supplemental Experimental Procedures). These studies revealed that there was no more than one Qdot per dimer of Rdh54, that the population of labeled molecules was monodisperse, and that there was no evidence for aggregation in solution (Figure S5).
To investigate the oligomeric state of Rdh54 bound to the DNA we used TIRFM to analyze the fluorescence signal from complexes that were labeled with either green (λem = 565 nm) or red (λem = 705 nm) quantum dots, and then mixed together immediately prior to injection into the sample chamber. If Rdh54 behaved as a dimer labeled with one quantum dot, then we predict that there should only be green or red proteins bound to the DNA, but no yellow complexes should be observed. Figure 5A shows sections of a DNA curtain bound by Rdh54 that was labeled with an equimolar mixture of green and red quantum dots. The differing emission spectra were separated by a dichroic mirror and simultaneously imaged on separate halves of the EMCCD chip. The left panel shows the signal from the green quantum dots, the center panel shows the red quantum dots, and the superimposed images are presented at the right (Figure 5A). As shown in these images, we could detect green, red, and yellow Rdh54 complexes. Based on the analysis of 251 individual complexes we found 59 red, 73 green, and 119 yellow Rdh54 complexes. The slightly greater number of green quantum dots is likely due to minor error in measuring the stock concentrations of the purified quantum dot-antibody conjugates. While the presence of “dark” quantum dots prevents us from accurately calculating the stoichiometry of Rdh54 molecules in these complexes, we however can predict that as the oligomeric state of the protein increases in complexity (e.g. tetramer, hexamer, dodecamer, etc.), so to does the probability that individual complexes will appear yellow. Taken together, these findings suggest that the minimal functional unit of Rdh54 appears to be a dimer, and that the protein also forms larger oligomers when bound to DNA.
(A) Shows a representative example of a DNA curtain bound by Rdh54 complexes that were labeled with a mixture of green and red quantum dots. The design of the TIRFM system allows us to collect both colors simultaneously, and the green image is shown in the left panel, the red image is shown in the center, and the overlaid image is shown on the right. (B) Shows a series of kymograms generated from Rdh54 complexes that were labeled with the two different colored quantum dots. Supplementary Video 3 shows the behavior of differentially labeled complexes of Rdh54 bound to the same molecule of DNA.
Collisions between different motor proteins on the same DNA
In many instances we observed what appeared to be collisions between different molecules of Rdh54 bound to the same strand of DNA (Figure 4C). To verify that their relative positions to each other along the DNA did not change over time we performed a dual-color labeling experiment where Rdh54 was tagged with either red quantum dots or green quantum dots. These experiments were performed at slightly higher concentrations of protein (determined empirically) to ensure that multiple Rdh54 complexes would be bound to the DNA.
As shown in figure 5B, the differentially labeled proteins displayed complex patterns of behavior as they interacted with the DNA (Supplemental Video 3). Different molecules of Rdh54 often appeared to collide with one another, or moved apart and traveled in different directions or at different velocities. However, the overall color patterns remained largely unaltered during the course of the observations, indicating that the different Rdh54 complexes did not bypass one another while moving along the DNA. The few changes in color pattern that were observed could be attributed to either the spontaneous dissociation of a protein (see below), the merging of two different colored complexes as they as they approached one another on the DNA, or the dissociation of a larger yellow complex into smaller red and green components (Supplemental Video 3). Surprisingly, under these conditions it often appeared as though the proteins bound to the same DNA molecule traveled in synchronous patterns, even though they were separated from one another by distances that could span thousands of base pairs (Figure 5B and see below). The experiments presented below provide an explanation for this unexpected behavior.
Rdh54 can form large DNA loops during translocation
As indicated above, many of the Rdh54 molecules underwent repetitive cycles of forward and reverse movement (Figure 4A) and it often appeared as though multiple Rdh54 complexes traveled in unison while bound to distal positions on the same molecule of DNA (Figure 5B and Supplemental Videos 3 and 4). To explore these behaviors further we analyzed the movement of Rdh54 complexes bound under conditions where there were multiple proteins per DNA molecule (Figure 6). These results confirmed that many of the proteins traveled in unison, in fact, this type of synchronous movement was detected for at least 80% of the individual translocating Rdh54 complexes examined. In these cases, the protein complexes often abruptly and simultaneously returned to their original positions relative to the tethered ends of the DNA molecules (Figure 6A and B). For example, in Figure 6A, four different Rdh54 complexes begin moving in unison towards the tethered end of the DNA, while the remaining complexes bound nearest the tethered end remained stationary. Just after the 100-second time point the moving complexes abruptly returned to their original locations; several similar events occurred on this same DNA molecule (Figure 6A and B). These abrupt reversals occurred too quickly to be accounted for by an ATP-dependent translocation mechanism (Figure 6B). Rather, the data were more consistent with the disruption of a single protein-DNA contact. We also observed many examples where the synchronous reversal occurred gradually and more closely resembled DNA translocation rather than sudden release (Figure 5B, and Supplementary Video 3 and 4).
The most reasonable explanation for these correlated movements is that one or more of the Rdh54 complexes translocated along the helical axis while extruding a large loop of DNA. This would in turn cause all of the stationary “downstream” proteins to move in concert with the translocating protein as it extruded the DNA loop. The abrupt return of the proteins to their original locations would suggest that the DNA loop was suddenly released by the translocating complex. Importantly, the loop release events never coincided with dissociation of a fluorescent protein from the DNA (Figure 6A), indicating that even though the loops were disrupted the proteins remained stably bound through additional contacts with the DNA and were capable of reiterative catalytic cycles. In cases where the DNA loops were release more gradually the driving mechanism may have been reverse translocation or backtracking of the Rdh54 motor. Analysis of 80 different looping events revealed that the size of the loops averaged ~6 kilobases, and occasional events were observed in which loops larger than 15–20 kilobases were generated (Figure 6C). The hydrodynamic force experienced by the tethered DNA molecules in the sample chamber was on the order of ~0.5 pN 26, indicating that the Rdh54 motor was capable of generating large DNA loops even against this moderate opposing force. Based on these observations we could not determine which of the Rdh54 complexes was responsible for the observed movement. It could in principle be due to either a mobile protein that generated a loop as it translocated towards the tethered end of the DNA or it could also be caused by a stationary complex that pulled the free end of the DNA towards itself (see Discussion and Figure S6). Distinguishing between these different mechanisms will require further investigation. Taken together, these experiments revealed that Rdh54 displays highly dynamic translocation behavior while at the same time remaining tightly bound to the DNA.
Rdh54 forms a highly stable complex with DNA and can translocate for long periods of time
Previous single-molecule studies with chromatin remodeling complexes have relied on either force measurements with unlabeled proteins or have used optical microscopy to visualize proteins tagged with traditional fluorophores that rapidly photobleached. A drawback of both approaches is that it is difficult to definitively follow the activity of a single complex over long periods of time. As a result it remains unresolved whether a single remodeling complex can remain stably associated with a DNA molecule and continuously interact with and travel along the DNA.
To address this question we followed the long-term behavior of Rdh54 bound to individual DNA molecules. First, Rdh54 was bound to a DNA curtain, as described above, and then 5-minute videos of the same field-of-view were collected at defined intervals over a total period of 2.3 hours (Figure 7A and B). For all of these experiments the DNA was stained with YOYO1, the stain was removed before injecting Rdh54 (Supplemental Experimental Procedures), and the DNA was re-stained after the end of the final measurement to verify its integrity. This procedure was performed to ensure that broken DNA molecules did not contribute to the apparent off rate, and any Rdh54 complexes bound to a DNA molecule that broke away from the surface during the course of the experiment were discounted from the analysis. The apparent dissociation rate was calculated by counting the number of bound proteins at each time point, plotting these values as a function of time, and then fitting the data to a single exponential decay. This yielded an apparent off rate for Rdh54 of koff=1.50±0.45×10−4 sec−1, demonstrating that the protein remained tightly associated with the DNA. As shown by the kymograms in figure 7B, the Rdh54 complexes continued traveling along the DNA molecules during the course of the measurement as evidenced by movement detected during each observation period as well as by the changed locations of the proteins between successive videos (Figure 7B). We verified that the protein could translocate for long periods by continuously monitoring a single molecule of DNA bound by Rdh54 for 60 minutes without interruption (Figure 7C). This movement was ATP-dependent, and the protein would abruptly stop moving if ATP was removed from the buffer, only to resume translocation when ATP was added back (data not shown). Moreover, replacement of ATP with nonhydrolyzable ATPγS prevented the proteins from traveling along the DNA, but the dissociation rates with ATPγS were similar to those observed with ATP (Figure S6 and data not shown). These experiments confirm that Rdh54 can remain bound to DNA in an active state capable of translocation for long periods of time.
To monitor the long-term behavior of Rdh54, the protein was bound to a DNA curtain in the presence of ATP and visualized over a period of 2.3 hours by capturing 5-minute videos at the indicated intervals. In (A), particle counting was used to determine how quickly Rdh54 dissociated from the DNA under these conditions, yielding a rate of koff=1.5±0.45×10−4 sec−1. These traces represent three different experiments. The kymograms in (B) show the dynamic behavior of Rdh54 over time, indicating that the proteins continued to translocate while they remained bound to the DNA. Each separate kymogram was taken from the same individual DNA molecule viewed at the indicated intervals. Similar experiments performed in the presence of ATPγS verified that the observed movements were ATP-dependent (Figure S6). The kymogram in (C) shows complexes of Rdh54 bound to DNA in the presence of ATP that were monitored continuously by capturing a video at 5 frames per seconds for a period of 60 minutes.
DISCUSSION
Our approach to single-molecule biochemistry combines the use of TIRFM with novel surface engineering procedures that allow us to directly visualize up to hundreds of individual reactions in a single experiment. Here we have applied these techniques to the study of the Snf2-related motor protein Rdh54. These assays allowed us to directly visualize Rdh54 binding to and translocating along double-stranded DNA molecules and revealed a variety of behaviors, including: unidirectional translocation, pauses and reversals, reversible DNA loop formation, and continuous translocation along the same DNA molecule for long periods of time.
Rdh54 is an ATP-dependent DNA translocase
We have demonstrated that Rdh54 can actively translocate along double-stranded DNA by directly visualizing single proteins as they move along the helical axis. The translocation activity was rapid, displaying an average velocity of 80 bp/sec at 1 mM ATP, and the proteins were highly processive. The translocation of Rdh54 is reminiscent of that reported for Rad54 13, yet despite their high degree of sequence and functional homology there do appear to be many important differences in their activities. For instance, the majority of Rad54 molecules (80%) exhibited monotonic translocation in a single direction, and no evidence was shown indicating that Rad54 could generate DNA loops 13. In contrast, with Rdh54 variations in translocation behavior appear to be the rule rather than the exception. In fact, most of the translocating Rdh54 complexes displayed some deviation from monotonic translocation kinetics, and at least 80% of the proteins generated DNA loop structures, 60% of which reversed during the course of the observation. Importantly, our data also show demonstrate that Rdh54 can remain bound to DNA for extended periods of time while continually translocating.
Mechanisms of DNA loop extrusion
The formation of DNA loops can occur if individual Rdh54 complexes contain at least two different DNA binding domains that can simultaneously interact with the same DNA molecule. There are several possible ways that this can be achieved, for example: (1) either the protein itself has two distinct DNA binding sites within a single polypeptide chain, one domain to anchor it in place and a separate motor domain used for translocation (Figure S7A); or (2) the protein has a single DNA binding site, but multiple points of contact can be made with the DNA due to the multimeric nature of the complex (Figure S7B). For example, a dimeric complex could, in principle, make two distinct contacts with the DNA. This configuration would support loop formation if only one of the motors translocated and the other remained at fixed position on the DNA. (3) Similarly, it is also possible that loop extrusion could occur via a mechanism similar to that of the type 1 restriction endonucleases, in which two motors translocate in opposite directions even though they are part of the same protein complex (Figure S7C) 30. However, with our experimental setup such a configuration would cause the proteins to always move towards the tethered end of the DNA and the tethered end of the DNA (or any stationary downstream proteins) would appear to move twice as fast as the translocating motor. We have observed no evidence to suggest that either of these is true. Therefore we prefer a model in which the multimeric translocating complexes operate using a single motor in any given instance, while maintaining multiple points of contact with the DNA, but further investigation will be necessary to delineate the precise details of this interaction.
DNA loops have also been reported for the chromatin-remodeling complexes RSC and SWI/SNF (Lia et al., 2006; Zhang et al., 2006). RSC and SWI/SNF are large complexes comprised of 11–15 different polypeptides, and sequence analysis of their common ATPase subunits places them within the Snf2-like subfamily of Snf2 proteins (Saha et al., 2002; Flaus et al., 2006). Magnetic tweezers were used to show that RSC and SWI/SNF could travel along DNA and also induced the formation of transient DNA loops, and the size of the loops ranged from 20–1200 bps 27. These assays could only detect translocation if it was coupled to loop formation, so it remains unclear whether this protein was able to move without loop formation or whether they remained stably bound to the DNA after loop release and were capable of repeated cycles of loop formation. It is not apparent why the loops observed for RSC and SWI/SNF were smaller than those detected here with Rdh54, but this may reflect differences in the specific biological functions of the two enzymes. For example, RSC or SWI/SNF may only need to move a single nucleosome for relatively short distance to allow RNA polymerase to gain access to a promoter region. In contrast, Rdh54 may need to efficiently clear proteins from a much larger region of DNA 24; 31. Taken together these results show that Rdh54 and RSC both form DNA loops, despite the fact that they are assigned to different Snf2 subgroups, have different biological functions, and have obvious differences in subunit composition. Thus our results confirm that translocation-coupled DNA loop extrusion is a conserved mechanism used by the different Snf2-family of chromatin-remodeling motor proteins and may play an integral role in their biological functions.
DNA translocation, loop extrusion and the biological function of Rdh54
Rdh54 performs a variety of functions during homologous DNA recombination and is likely to act at several different stages of the reaction 9; 10; 23. The challenge now is to understand how translocation and loop extrusion functions are integrated into the requirements of the DNA repair machinery and used to control chromosome structure and/or modify protein-DNA interactions during homologous recombination. It is possible that different aspects of Rad54 activity play distinct roles at the presynaptic, synaptic, and postsynaptic stages of the recombination reaction 10.
Rdh54 is required for meiotic DNA recombination and interacts with both the Rad51 and Dmc1 recombinases 17; 19; 23; 32; 33. One proposed function of Rdh54 (and Rad54) has been as a molecular “stripase” whose function is to remove or remodel stationary proteins from DNA to allow the repair machinery to have unhindered access to its substrates 14; 23; 34. At the early stages of the reaction this would entail removing nucleosomes from chromatin 34; 35; 36. At later stages the stripase activity would ensure that the DNA was cleared of recombinase and made accessible to additional repair factors necessary for downstream steps in the pathway 14; 23; 37. As shown here, Rdh54 translocates and concomitantly generates DNA loops, but it is not clear which of these activities would be most essential for disrupting stationary proteins. One model posits that translocation generates torsional stress that may disrupt protein-DNA complexes 6. A second model proposes that the translocating protein may function like a locomotive’s “cowcatcher” by colliding with sufficient force that any stationary proteins are simply displaced from the DNA 14. A third possibility is that specific protein-protein interactions are necessary between the translocase and the stationary roadblock to specifically trigger dissociation of the bound protein 14; 23. These models are not mutually exclusive and all three mechanisms may play a role during the disassembly of recombinase filaments or during disruption of nucleosomes.
It is possible that DNA loop extrusion plays a direct role in the strand invasion step of homologous recombination. Rdh54 and Rad54 are also involved in earlier steps of homologous recombination and the proteins greatly enhance the invasion of a homologous double-stranded DNA molecule by the Rad51/Dmc1 recombinase presynaptic filament 15; 22; 33; 38. In this mode, it is thought that the translocase associates with the recombinase filament and together they search the duplex DNA for regions of homology and align the two strands of DNA. The function of the translocase in these reactions may be two-fold: (1) it may serve as a molecular motor enabling the presynaptic filament to rapidly sample the duplex DNA 39, and (2) it may extrude supercoiled loops from the duplex DNA which would in turn serve as a more favorable substrate for strand invasion 8; 12; 33. The extrusion of DNA loops during translocation offers a very simple mechanistic explanation for how Rdh54 and other related Snf2 proteins modify the topology of DNA in bulk assays. DNA supercoiling is a requirement for eukaryotic DNA recombinases in the Rad51-family, therefore the formation of extruded supercoiled loops would be of clear benefit to these reactions.
MATERIALS AND METHODS
All details of experimental procedures, data analysis, and videos are provided as Supplemental Information online.
Acknowledgments
This work was supported in part by a March of Dimes Basil O’Connor Starter Scholar Research Award (to E.C.G.) and National Institutes of Health Grants GM074739 (to E.C.G) and GM57814 (to P.S.). R.B.R. was supported by an NIH Biophysics training grant (5T32GM008281). We thank Goran Ahlsen for assistance with analytical ultracentrifugation. We thank Ruben Gonzalez, Lorraine Symington, Rodney Rothstein and members of our laboratories for discussion and comments on the manuscript.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.