rad54-S816D, S817D is defective in stable D-loop formation

To identify the repair intermediate at which the rad54-S816D, S817D strain is defective, we used a previously established D-loop capture assay [14,19,79,80]. This assay also employs an ectopic repair site within the genome and can be used to isolate nascent and extended D-loops. These DNA structures represent the initial strand exchange intermediates. For nascent D-loop capture, after induction of a break using the HO endonuclease, the cells are cross-linked with psoralen. Cross-linking traps the nascent D-loops and prevents dissociation during purification. The efficiency of D-loop capture is then quantified by isolating genomic DNA, oligo hybridization, proximity ligation, and quantifying the amount of isolated DNA by qPCR (Figs 3A and S6AB). We monitored D-loop capture at 4 hours, and as expected, the rad54∆ strains had a signal that was 100-fold lower than the WT. This value was also the same as that obtained from qPCR performed without the oligo to restore restriction enzyme sites, indicating this is the background (S6C Fig). The rad54-S816A, S817A allele was slightly lower than WT. However, this difference was not significant (Fig 3B). The rad54-S816D, S817D strain was 47-fold lower than WT (Figs 3B and S6CDEFG), indicating a severe defect in D-loop capture. To further validate this phenotype, we performed additional D-loop capture experiments using the rad54-D525S, D527S, S816D, S817D allele, which represents one of the intragenic suppressors identified in the complementation analysis. This mutant suppressed the D-loop capture phenotype, increasing capture efficiency approximately 10-fold and making it only 5-fold less efficient than WT (Fig 3B). The observation is consistent with the ectopic repair survival experiments.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Mutations in RAD54 impact recombination intermediates.

(A). Schematic diagram illustrating the D-loop capture assay to trap the formation of nascent D-loops during recombination. (B). Graph representing D-loop capture efficiency at 4 hours post-break induction for RAD54, rad54∆, rad54-S816A, S817A, rad54-S816D, S817D, and rad54-D525S, D527S, S816D, S817D. The bar represents the mean, and the error bars represent the standard deviation for at least three independent experiments. (C). Schematic diagram illustrating the assay to monitor D-loop extension. (D). Graph representing D-loop extension at 6 hours for RAD54, rad54∆, rad54-S816A, S817A, rad54-S816D, S817D, and rad54-D525S, D527S, S816D, S817D. The bar represents the mean, and the error bars represent the standard deviation for at least three independent experiments. (E). Graph representing D-loop capture efficiency at 4 hours post-break induction for sgs1∆, rad54-S816A, S817A sgs1∆, rad54-S816D, S817D sgs1∆. The bar represents the mean, and the error bars represent the standard deviation for at least three independent experiments. (F). Graph representing D-loop extension at 6 hours for hours post-break induction for sgs1∆, rad54-S816A, S817A sgs1∆, rad54-S816D, S817D sgs1∆. The bar represents the mean, and the error bars represent the standard deviation for at least three independent experiments. (G). The percentage of dsDNA among total extension products for RAD54, rad54∆, rad54-S816A, S817A, rad54-S816D, S817D, sgs1∆, rad54-S816A, S817A sgs1∆, rad54-S816D, S817D sgs1∆. The error bars represent the standard error measurement for at least three independent experiments.

https://doi.org/10.1371/journal.pgen.1012136.g003

The extension of D-loops stabilizes the nascent three-strand intermediates. D-loop extension from the newly paired 3’ end can be measured using the same strains, but without psoralen cross-linking (Figs 3C and S7AB). A caveat to this analysis is that the D-loop must be extended by at least 400 nt to receive the second restriction site and be detected in the assay. Therefore, defects in extension could reflect low D-loop capture efficiency or premature disruption of DNA synthesis, resulting in shorter extension products. As with the D-loop capture assay, there was only a modest difference from WT in the ability of rad54-S816A, S817A to extend D-loops, and this was not significant. However, there was a significant loss (20-fold, p = 0.02) of D-loop extension in the rad54-S816D, S817D strain at 6 hours (Figs 3D and S7CDEF). This change is still better than the 100-fold loss in the rad54∆ strain. The rad54-S816D, S817D phenotype was again suppressed in the rad54-D525S, D527S, S816D, S817D strain (Figs 3D and S7CDEF). From this, we conclude that the interaction between S816 and D525 is required for efficient D-loop capture and stabilization.

Sgs1 is a DNA helicase that participates in DNA end-resection, reversal of toxic recombination intermediates, and dissolution of Holliday junctions [81]. Sgs1 is also implicated in the reversal of D-loops [19,79,82]. We reasoned that if rad54-S816D, S817D were subjected to a high level of D-loop reversal, then deletion of SGS1 might suppress most of the observed defects in D-loop capture. However, this was not the case, and only a small increase in D-loop capture was observed in the rad54-S816D, S817D sgs1∆ double mutant strain. Capture efficiency remained 20-fold lower than WT (Figs 3E and S6CDEFG). The increase observed in the rad54-S816D, S817D sgs1∆ double mutants translate into a 3-fold increase in capture efficiency compared with the rad54-S816D, S817D single mutant. Previously, it has been shown that SGS1 has a stronger impact on D-loop capture at earlier time points, and analysis at 4 hours may contribute to the lack of observable phenotype.

There was no observable defect in the amount of D-loop extension for the sgs1∆ or the rad54-S816A, S817A sgs1∆ strains (Fig 3F). Surprisingly, in the rad54-S816D, S817D sgs1∆ strain, there was now a 100-fold loss in D-loop extension. This observation suggests that while D-loop capture was slightly improved, there was insufficient extension to reach the second restriction site (Fig 3F). This observation means that the rad54-S816D, S817D strains are competent for D-loop extension if the primary D-loop is stabilized, because there is a significant (5-fold) reduction in extension between the rad54-S817D, S817D and the rad54-S817D, S817D sgs1∆ strains (Fig 3F) This extension is lost in the rad54-S816D, S817D sgs1∆ mutant, indicating a synthetic genetic interaction between RAD54 and SGS1.

In this assay, one side of the double-strand break lacks homology to the target. The lack of homology prevents repair through SDSA or classical DSBR, making this system a break-induced replication (BIR) system. In BIR based repair, DNA synthesis proceeds through a migrating bubble, using conservative replication [23,83]. The newly extended ssDNA is used as a template for second strand synthesis [83,84]. Therefore, the extended D-loops can also be characterized by their ssDNA-to-dsDNA content (Figs 3G and S7F). The amount of dsDNA is measured by not adding a hybridizing oligo during DNA extraction, and comparing this to the same conditions with an oligo added. The addition of an oligo restores a restriction site that is used to cut the DNA. If DNA is cut in the absence of the oligo, it means that the restriction site has been restored by natural DNA synthesis. The amount of restriction site cutting and re-ligation with and without oligo can then be used to compare the ssDNA to dsDNA content, reflecting the amount of second strand synthesis [19,80]. This value is low in WT but is elevated 3–4-fold in sgs1∆ strains [19]. We observed this same increase in dsDNA in a sgs1∆ strain (Figs 3G and S7F). Surprisingly, this increase was lost in the rad54-S816A, S817A sgs1∆ strain (Fig 3G). This result suggests a genetic interaction between RAD54 and SGS1 during the repair synthesis phase, and a slight defect in the rad54-S816A, S817A mutant. It should be noted that the extension values for rad54∆ and rad54-S816D, S817D sgs1∆ were so low that this type of analysis could not yield reliable results.

Our data suggested that D-loop extension was less effected than D-loop capture in the rad54-S816D, S817D strain. This would imply that rad54-S816D, S817D may be competent to remove Rad51 from dsDNA. To test this genetically, we overexpressed RAD51 under a galactose-inducible promoter. This assay has been used to show that both RAD54 and its paralog, RDH54, can remove Rad51 from dsDNA [37,85]. The phenotype is weak in rad54∆ strains and is significantly stronger in rdh54∆ strains. However, the effects of RAD54 in this assay are pleiotropic with rdh54∆. Therefore, the weak phenotype associated with rad54∆ is the best metric to determine if our rad54 alleles could rescue this phenotype. In these assays, overexpression of Rad51 does not result in cell death and is reversible if the plasmid overexpressing Rad51 is lost. However, persistent overexpression of can lead to chromosome loss [37]. These phenotypes are caused by extensive binding of Rad51 to chromosomes, even in the absence of DNA repair.

We found that both rad54-S816A, S817A and rad54-S816D, S817D behaved like WT in the RAD51 overexpression experiments (S8A Fig). To enhance the phenotype, we also overexpressed a mutant RAD51 allele, rad51-I345T, which binds ds- and ssDNA more tightly [58]. We expected this would be a greater challenge to remove from the dsDNA and may expose any weak phenotypes associated with the rad54 mutants. Overexpression of rad51-I345T did slightly strengthen the overexpression phenotype in the rad54∆. However, both the rad54-S816A, S817A and rad54- S816D, S817D alleles were able to remove Rad51 from dsDNA (S8A Fig). Despite the weak phenotype in this experiment, our data are consistent with an enzyme that can remove Rad51 from dsDNA but is deficient in D-loop formation.

Interhomolog allelic recombination

The rad54-S816D, S817D mutant had diminished D-loop capture and failed to promote survival during ectopic recombination. Ectopic recombination is generally less efficient than recombination between sister chromatids or homologous chromosomes and may be more sensitive to subtle changes in recombination efficiency [75]. Consequently, we investigated the impact of the rad54-S816A, S817A, and rad54-S816D, S817D alleles on recombination between homologous chromosomes. We used a reporter assay based on the ADE2 gene in diploid S. cerevisiae [86–89]. Each copy of chromosome XV has a different inactive allele of ade2. One copy contains an I-Sce1 nuclease site (ade2-I), and the other has an inactivating mutation located within the gene (ade2-n) (Fig 4A). Upon induction of the nuclease, a double strand break forms when one of the sister chromatids is cut. If this is repaired via the uncut sister, then the restriction site will be restored, and re-cutting can occur. Gene conversion and elimination of the restriction site occur when repair uses the homologous chromosome. Types of gene conversion events can be determined by reconstitution of the ADE2 genes. If long tract repair occurs, then the gene conversion event will acquire the deleterious mutation from the homolog, and the yeast colony will appear red. If short tract repair occurs, the mutation will not be acquired, and a functional ADE2 gene will be reconstituted, resulting in a white colony (Fig 4B). Each sister can be repaired via a long tract or short tract gene conversion event, and due to independent assortment, a colony can be red (two long tracts), white (two short tracts), or sectored (one long tract and one short tract).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Impact of RAD54 mutations on allelic recombination between homologous chromosomes.

(A). Schematic diagram illustrating the DNA reporter used to analyze the effect of Rad54 mutation on allelic recombination. (B). Schematic diagram illustrating the potential gene conversion outcomes and HR based repair pathways during allelic recombination. (C). Graph representing the plating efficiency of strains treated with galactose. These were measured by dividing the number of colonies formed after galactose treatment to a no-galactose control. The bar represents the mean of the data, and the error bars the standard deviation of at least 3 independent experiments. (D). Graph representing the fraction of colonies that undergo chromosome loss for All strains tested. The bar represents the mean and the error bars the standard deviation of at least four independent experiments. (E). Graph representing the percentage of solid red, solid white, and sectored colonies for all strains tested. The bar represents the mean of the data, and the error bars represent the standard deviation of at least three independent experiments. (F). Graph representing the fraction colonies that undergo short tract gene conversion for all strains tested. This data was generated by measuring the colonies that grew on YNB (–Ade) + dextrose. The bar represents the mean, and the error bars the standard deviation of at least four independent experiments.

https://doi.org/10.1371/journal.pgen.1012136.g004

We initially evaluated strains for survival after galactose treatment by comparing the plus and minus galactose conditions. We found that the WT, rad54∆/rad54∆, and rad54-S816A, S817A survived with 80–100% efficiency (Fig 4C). In contrast, the rad54-S816D, S817D mutant survived ~50–70% of the time (Fig 4C). However, this analysis was complicated in the rad54∆/rad54∆ and rad54-S816D, S817D strains because, after galactose treatment, the colonies formed a mixture of normal-sized and microcolonies. We hypothesized that the micro-colonies may represent colonies that have undergone chromosome loss. We tested this by replica-plating the colonies onto YNB (-Met) + dextrose and YP + Hygromycin + dextrose. The MET22 and HPHMX markers are located on opposite sides of the centromere on chromosome XV (Fig 4A), and loss of both markers indicates a chromosome loss event. As expected, the microcolonies exhibited loss of both markers. In the rad54∆/rad54∆ strains, ~ 70% of the surviving colonies exhibited loss of the cut chromosome. In WT and the rad54-S816A, S817A mutant <1% of colonies scored for chromosome loss (Fig 4D). In the rad54-S816D, S817D mutant ~30% of the population exhibited loss of the cut chromosomes (Fig 4D). Importantly, this value was reduced to <1% in the rad54-D525S, D527S, S816D, S817D intragenic suppressor mutant. From this, we conclude that the rad54-S816D, S817D results in a significant loss of the cut chromosome. In subsequent analyses of recombination outcomes, the microcolonies and chromosome loss events were excluded.

The normal size colonies observed in the rad54∆/rad54∆ were red, and there were no sectored or white colonies. To determine if the red colonies were recombinants or uncut colonies, we replica plated cells on YNB (–Ade) + raffinose + galactose plates. If colonies are uncut, they will be forced to convert to ADE2 to survive, resulting in white papillae in the red transferred colonies. If cells don’t survive, they have repaired ade2 by long tract gene conversion or non-homologous end joining (NHEJ). In the rad54∆/rad54∆ strains, this represented most of the large colonies. A limitation of this assay is the inability to distinguish between DSBs that may be uncut in the reinduction assay due to low-fidelity repair via non-homologous end joining (NHEJ). In fact, similar assays have shown that ~35% of repair events occur through Rad51-independent pathways in the absence of Rad51 [90]. Given the uncertainty of competition between Rad51-dependent and independent pathways, we have limited our interpretation to reflect that our observed changes only reflect a loss in gene conversion outcomes. Whether this occurs because of an increase in gene conversion tract length or an increase in competing repair pathways is unclear.

In our hands, sectored colonies are the most common outcome in WT, occurring approximately 65–70% of the time. This was also observed in the rad54-S816A, S817A mutant (Fig 4E). However, with the rad54-S816D, S817D allele, sectored colonies accounted for only about 20–25% of the population, with most colonies appearing solid red. The fraction of short tract gene conversion events can be measured by replica plating colonies on YNB (–Ade) + dextrose. Under these conditions, only colonies/sectors that had undergone short tract gene conversion could grow. In WT and rad54-S816A, S817A, this was ~ 35% of the population (Fig 4F). By contrast, this group was reduced to ~10% in rad54-S816D, S817D strains (Fig 4F, p < 0.0001). Importantly, this phenotype was fully suppressed by the rad54-D525S, D527S, S816D, S817D mutant (Fig 4F).

The helicase Sgs1 is known to shorten gene conversion tracts, and sgs1∆ strains result in higher levels of long tract gene conversion [91]. From our previous experiments, we identified a novel genetic interaction between Rad54 and Sgs1. Therefore, we generated SGS1 deletion strains with RAD54, rad-54S816A, S817A, and rad54-S816D, S817D, to further explore this interaction. Deletion of SGS1 resulted in a moderate increase in chromosome loss (Fig 4D). However, these strains retained a significant amount of recombination (Fig 4F). There was no observable increase in chromosome loss or change in short tract gene conversion in the rad54-S816A, S817A sgs1∆ double mutant. In the rad54-S816D, S817D sgs1∆ double mutant, there were comparable levels of chromosome loss to the rad54-S816D, S817D strains (Fig 4D). However, there was a further decrease in short tract gene conversion (Fig 4F). This is further evidence of a genetic interaction between RAD54 and SGS1.

The reporter strains used in these experiments also carry antibiotic markers downstream of the ade2 genes. The segregation of these markers can be used to determine whether the recombination results in a Crossover (CO), Non-Crossover (NCO), or Break-induced Replication (BIR) outcome (Fig 4A and 4B). Outcomes can be scored by replica plating colonies onto YPD + Nourseothricin (clonNat) or YPD + Hygromycin (Hph). Colonies in which a sector grows on one selection, and the other sector grows on the other selection, can be designated as crossovers. This number is multiplied by two because for every observable crossover outcome, there is the potential for an unobservable crossover. BIR is scored for sectored colonies in which one sector grows under both selections, and the other sector only grows under Nat selection. For red colonies, a similar method can be used for scoring, but these can only be termed BIR-like and could result from other types of repair, as described above. Therefore, these outcomes should be interpreted cautiously.

We analyzed the rad54-S816A, S817A, and rad54-S816D, S817D strains for recombination outcomes. There was a small but significant increase in NCO between the WT and rad54-S816A, S817A, and there was no change in BIR outcomes (S9A Fig). However, there was a more significant drop in CO outcomes and an increase in NCO outcomes associated with the rad54-S816D, S817D (S9A Fig, p < 0.0001). We observed a small increase in BIR associated with this mutant (S9A Fig, p = 0.04). We evaluated the rad54-D525S, D527S, S816D, S817D rescue mutant, and this restored CO and NCO to WT levels (S9A Fig). There was no significant difference in outcomes between the WT and sgs1∆ or rad54-S816A, S817A sgs1∆ strains (S9B Fig). However, in the rad54-S816D, S817D sgs1∆ strain, there was a complete loss of CO (S9B Fig), and this was matched by an increase in NCO and a small increase in BIR events (S9B Fig). The complete loss of CO outcomes was the only significant change from the rad54-S816D, S817D strain. We have included this data in the manuscript, but due to our limited ability to differentiate the competition between Rad51-dependent and independent repair pathways, we have not offered an interpretation.

Rad54 S816D/S817D is defective in binding dsDNA

Through our genetic analysis, we determined that substituting Aspartic acid for S816 and S817 disrupts D-loop stabilization in vivo. To better characterize the mechanism behind this defect, we purified Rad54, Rad54 S816A/S817A, Rad54 S816D/S817D, and Rad54 D525S/D527S/S816D/S817D (S10A Fig). A working hypothesis was that disruption of the interaction between S816/S817 and D525/D527 would impair communication between the two RecA lobes and may destabilize a specific conformation of Rad54. A logical consequence of this disruption may be a reduced affinity for dsDNA. Therefore, we initially tested the ability of these proteins to bind dsDNA using an Electrophoretic Mobility shift assay (EMSA) (Fig 5A). We quantified the apparent K_d and found a two-fold decrease in the affinity of Rad54 S816D/S817D mutant for dsDNA (Fig 5B and 5C, 21 nM versus 41 nM, p = 0.008). There was also a slight increase in the K_d for the Rad54 S817A/S817A mutant, but this change was not significant. Importantly, the K_d was restored to ~18 nM in the Rad54 D525S/D527S/S816D/S817D mutant. From this, we conclude there is a loss in the affinity for dsDNA in the Rad54 S816D/S817D mutant protein.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Rad54 S816D/S816D has reduced affinity for dsDNA.

(A). Representative 0.5% TBE gel for EMSA with Rad54 and Rad54 S816D/S817D with a 90 bp Atto647N labelled dsDNA. (B). Graph representing the quantification of the EMSE data for Rad54, Rad54 S816A/S817A, Rad54 S816D/S817D, and Rad54 D525S/D527S/S816D/S817D. The data are fit by a hill equation. The error bars represent the standard error measurement of at least four independent experiments. (C). Graph representing the K_d measurements for individual EMSA experiments for Rad54, Rad54 S816A/S817A, Rad54 S816D/S817D, and Rad54 D525S/D527S/S816D/S817D. The bar represents the mean, and the error bars represent the standard error measurement of the experiment. (D). Graph representing the amount of ADP produced per minute in an ATP hydrolysis assay for Rad54, Rad54 S816A/S817A, Rad54 S816D/S817D, and Rad54 D525S/D527S/S816D/S817D. The bar represents the mean, and the error bars represent the standard deviation of four independent experiments. (E). Graph representing the amount of ADP produced per min for Rad54 and Rad54 S816D/S817D. The bar represents the mean, and the error bar represents the standard deviation of 3 independent experiments. (F). Schematic diagram illustrating DNA curtains experiments to test the activity of Rad54 on dsDNA. This image is reproduced through a CC-BY 4.0 license from [95]. (G). Widefield microscope image for GFP-Rad54, GFP-Rad54 S816A/S817A, and GFP-Rad54 S816D/S817D, and Rad54 D525S/D527S/S816D/S817D. (H). Mean fluorescent decay rate for Rad54 (Blue), Rad54 S816A/S817A (Green), Rad54 S816D/S817D (Orange), Rad54 D525S/D527S/S816D/S817D (Magenta) and Rad54 with ADP (Cyan). The black line gives the photobleaching rate. The shade represents the standard error measurement of the data. (I). Graph representing the k_off for GFP-Rad54 (+ATP), GFP-Rad54 S816A/S817A, and GFP-Rad54 S816D/S817D, Rad54 D525S/ D527S/S816D/S817D and GFP-Rad54 (+ADP) from dsDNA. The bars represent the mean, and the error bars the standard error measurement of the data.

https://doi.org/10.1371/journal.pgen.1012136.g005

Rad54 is a dsDNA dependent ATPase, and a loss in affinity for dsDNA should result in a reduction in ATP hydrolysis. Therefore, we tested ATP hydrolysis for Rad54, Rad54 S816A/S817A, Rad54 S816D/S817D, and Rad54 D525S/D527S/S816D/S817D. As expected, the Rad54 S816D/S817D mutant hydrolyzed ATP with 10-fold lower efficiency than WT and the other two mutant versions of Rad54 (Fig 5D). We were surprised by the significant loss of ATP hydrolysis, but suspected that Rad54 S816D/S817D may still be activated in the presence of Rad51. The ATPase activity of Rad54 is stimulated by the presence of Rad51, leading to a 3–5-fold increase in ATP turnover [29,34,52]. We next tested whether Rad51 could still stimulate Rad54 S816D/S817D. Rad51 stimulated the mutant enzyme’s activity but did not restore it to WT levels (Fig 5E, p = 0.01). This stimulation is not due to an increase in Rad51 activity, whose ATP hydrolysis activity is at least 100-fold lower than Rad54. Previous reports have shown that there is no stimulation when Rad54 K341R, a catalytically inactive version of Rad54, is substituted in this experiment [52].

The defect in DNA binding was relatively small, and we sought to use a higher-resolution approach to directly measure Rad54’s dissociation rate. DNA curtains are a high-throughput single-molecule approach that can measure Rad54 dissociation rates (Fig 5F) [92]. In these experiments, GFP-Rad54 is injected into a flow cell and allowed to bind to dsDNA (Fig 5F and 5G). The loss of fluorescent signal is then measured over time in the presence of buffer flow. The rate of GFP photobleaching is also measured. If the protein dissociation rate is faster than the photobleaching rate, then a true dissociation rate of the protein can be determined (Fig 5H). We observed that the rate of dissociation for GFP-Rad54, GFP-Rad54 S816D/S817D, GFP-Rad54 S816A/S817A, and GFP-Rad54 D525S/D527S/S816D/S817D were all faster than photobleaching. By subtracting the bleaching rate from the fluorescent decay rate, we calculated the k_off for each protein (Fig 5I). We found that Rad54, Rad54 S816A/S817A, and Rad54 D525S/D527S/S816D/S817D all dissociated at ~1.3x10^-4 sec^-1, in contrast the Rad54 S816D/S817D mutant dissociated with a rate of 3.3x10^-4 sec^-1 (Fig 5I). This is a 2-fold difference and is consistent with the EMSA. Despite the small difference in dsDNA binding, the difference is significant (p < 0.0001), can be rescued by an intragenic suppressor, and is consistent across multiple techniques. From this, we conclude that Rad54 S816D/S817D is defective in binding dsDNA.

Rad54 can bind to DNA in the absence of ATP [93]. However, it has never been determined if there are differences in affinity for dsDNA during the ATP hydrolysis cycle. The AlphaFold model predicted that Rad54 underwent a conformational change that disrupted the D525-S816 interaction when bound to ADP versus ATP. From this, we generated a hypothesis that the ATP bound form of the enzyme has a higher affinity for dsDNA, and this conformation was disrupted in the S816D/S817D mutant. To test this, we measured the off-rate of GFP-Rad54 in the presence of ADP nucleotide. We found that the k_off was around 3-fold higher than the ATP-bound form (Fig 5I). From this, we conclude that the ADP bound form of the enzyme is less stable on dsDNA. This is consistent with a model where a loss of D525-S816 interaction fails to stabilize the ATP bound form of the enzyme in the S816D/S817D mutant.

Our data suggest that disruption of the Rad54 D525-S816 interaction may result in a loss of dsDNA binding and failure to form a D-loop. If this were true, then overexpression of the protein should result in complete rescue of the phenotype, because it would shift the binding equilibrium of Rad54 toward a bound state. Alternatively, failure to fully complement may suggest a more complex mechanism. To test this, we overexpressed RAD54 and rad54-S816D, S817D from a 2-µ plasmid and assessed their ability to complement MMS sensitivity (S11A Fig). Overexpression of rad54-S816D, S817D did result in more efficient complementation than expression from a centromeric plasmid. However, it could not rescue to RAD54 levels. We additionally evaluated survival using the ectopic repair assay. As with the MMS complementation, there was an increase in survival in ectopic recombination when rad54-S816D, S817D was expressed from a 2-µ plasmid (S11BC Fig, 5% versus 16%, p = 0.03). However, this did not complement the levels of RAD54. From this, we conclude that overexpression of the rad54-S816D, S817D cannot fully complement the phenotype, and that the mechanism by which it fails to promote D-loop formation is likely more complex.

Rad54 S816D/S817D exhibits processivity defects in the absence of Rad51

Rad54 is a highly processive enzyme, meaning it can perform multiple ATP hydrolysis cycles before dissociating from or pausing on a dsDNA template. The energy released by ATP hydrolysis drives conformational changes in the Rad54 enzyme, and the final output is the translocation of the DNA. This mechanical activity can manifest as linear movement along the dsDNA or as extrusion of a DNA loop. Both translocation outcomes require Rad54’s processive activity. Recently, we determined that these two outputs can be controlled by the force applied to the DNA [94]. A simple way to measure Rad54’s processive activity is to monitor its distance and direction of movement along dsDNA using single-molecule imaging. In these experiments, we can add a nominal force from buffer flow to bias the translocation activity towards linear movement and away from loop extrusion.

We used DNA curtains to monitor the translocation activity of GFP-Rad54, GFP-Rad54 S816A/S817A, GFP-Rad54 S816D/S817D, and GFP-Rad54 D525S/D527S/S816D/S817D (Fig 6A). We divided the translocation tracks into those moving toward and away from the barriers (Fig 6B). Rad54 molecules moving against the flow will feel a slight opposing force, while those moving away from the barriers and with the flow will feel an assisting force. Under the conditions used here, a majority of Rad54 molecules (75%) moved toward the barriers. This is likely due to the DNA being more tense in that direction. The Rad54 S816A/S817A (63%) and the Rad54 D525S/D527S/S816D/S817D (82%) mutants also moved toward the barriers (Fig 6C). In contrast, only 17% of Rad54 S816D/S817D molecules moved toward the barrier. This is consistent with weaker-binding molecules being pushed down the DNA by the force of buffer flow.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Rad54 S816D/S817D is poorly processive.

(A). Schematic diagram illustrating DNA curtains experiments to test the activity of Rad54 on dsDNA. This image is reproduced through a CC-BY 4.0 license from [95]. (B). Representative kymographs illustrating the movement of GFP-Rad54, GFP-Rad54 S816A/S817A, GFP-Rad54 S816D/S817D, GFP-Rad54 D525S/D527S, S816D/S817D with and against buffer flow. (C). Bar graph representing the percentage of Rad54 (166/216), Rad54 S816A/S817A (285/446), and Rad54 S816D/S817D (48/271) and GFP-Rad54 D525S/D527S/S816D/S817D that move against the buffer flow. (D). Graph representing the velocity of translocation in kbp/s for Rad54 (N = 216), Rad54 S816A/S817A (N = 446), Rad54 S816D/S817D (N = 271), and Rad54 D525S/D527S/S816D/S817D (N = 81). The molecules that moved against the flow are above the X-axis, and the molecules that moved with the flow are below the X-axis. The bar represents the mean, and the error bars represent the standard deviation of the data. The negative value indicates the direction. (E). Graph representing the distance moved for Rad54 (N = 216), Rad54 S816A/S817A (N = 446), Rad54 S816D/S817D (N = 271), and Rad54 D525S/D527S/S816D/S817D (N = 81). The molecules that moved against the flow are above the X-axis (positive values), and the molecules that moved with the flow are below the X-axis (negative values). The bar represents the mean, and the error bars represent the standard deviation of the data.

https://doi.org/10.1371/journal.pgen.1012136.g006

Surprisingly, there was no difference in velocity (bp/sec) in either direction between any of the Rad54 variants tested, with the exception that there was a slight increase in the translocation rate of the Rad54 D525S/D527S/S816D/S817D mutant (Fig 6D, p = 0.02). Rad54 S816A/S817A and Rad54 S816D/S817D moved a significantly shorter distance against the flow than the WT (Fig 6E). This difference was rescued in the Rad54 D525S/D527S/S816D/S817D mutant (Fig 6E). The Rad54 S816D/S817D mutant moved farther along the DNA with the flow, consistent with a protein being pushed along the DNA. These data indicate that disruption of this interaction in Rad54 results in proteins that bind DNA less tightly, likely leading to lower processivity.

PSCs with Rad54 S816D/S817D slip during translocation

Rad51 is known to activate the ATP hydrolysis cycle of Rad54, thereby increasing translocation activity [29,95]. In vitro, Rad54 can bind to Rad51-ssDNA filaments, which can then bind to dsDNA and perform translocation [29,95]. This complex is referred to as the presynaptic complex (PSC) because it has not yet paired with a homologous DNA sequence (Fig 7A). We have previously used biochemical reconstitution of the PSC, combined with DNA curtains, to monitor Rad54 activity [29,95]. We next used DNA curtains to evaluate the impact of Rad51 on the linear translocation activity of Rad54 S816D/S817D.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Mutations lead to PSC slippage during translocation.

(A). Cartoon diagram illustrating the use of DNA curtains to perform experiments to monitor the activity of the presynaptic complex (PSC) on dsDNA. This image is reproduce through a CC-BY 4.0 license from [95]. (B). Representative kymographs for PSCs, PSCs with Rad54 S816A/S817A, or PSCs with Rad54 S816D/S817D. Shown are merged images of RPA-mCherry, GFP-Rad54, and Atto647–90-mer ssDNA. (C). Graph representing the percentage of PSC with Rad54 (183/215), PSC with Rad54 S816A/S817A (326/378), and PSC with Rad54 S816DS817D (710/835) that moved against the buffer flow. (D). Measured translocation velocities for the PSC with Rad54 (N = 215), PSC with Rad54 S816A/S817A (N = 378), and PSC with Rad54 S816D/S817D (N = 835). The PSCs that moved against the flow are above the X-axis, and the PSC values that moved with the flow are below the X-axis. The bars represent the mean, and the error bars represent the standard deviation of the data. (E). Measured translocation distances for the PSC with Rad54 (N = 215), PSC with Rad54 S816A/S817A (N = 378), and PSC with Rad54 S816D/S817D (N = 835). The PSCs that moved against the flow are above the X-axis, and the PSC values that moved with the flow are below the X-axis. The bars represent the mean, and the error bars represent the standard deviation of the data. (F). Schematic illustrating the loop extrusion activity of Rad54 extruding loops that can bind RPA-mCherry. (G). Graph representing the number of PSCs with RPA mCherry localized after a five-minute incubation without buffer flow for PSCs with Rad54 and Rad54 S816D/S817D. (H). A dot plot representing the intensity of RPA bound to bound to PSCs with Rad54 (N = 39) and Rad54 S816D/S817D (N = 28). The bar represents the mean of the data, and the error bars represent the standard deviation.

https://doi.org/10.1371/journal.pgen.1012136.g007

As with the Rad54-alone experiments, these experiments were measured using flow-on to increase observation of linear movement during translocation. When experiments were performed with the PSC Rad54, Rad54 S816A/S817A, and Rad54 S816D/S817D, all three moved toward the barrier in ~80–85% of translocation events, marking the first difference in activity between the Rad54 S816D/S817D mutant alone and incorporated into the PSC (Fig 7B and 7C). In contrast to Rad54 alone, there was a difference in the translocation velocity between PSCs with Rad54, Rad54 S816A/S817A, and Rad54 S816D/S817D. The velocity of Rad54 S816A/S817A was 75% of WT (285 bp/sec versus 215 bp/sec, p < 0.0001, and Rad54 S816D/S817D was 54% that of WT (285 bp/sec versus 155 bp/sec, p < 0.0001). For the Rad54 S816A/S817A mutant, this difference was observed only during movement toward the barrier, with no difference in translocation events away from the barriers (Fig 7D). We observed no differences in translocation distance between WT and Rad54 S816A/S817A in either direction. There was a small but significant difference in the distance travelled between the WT and the Rad54 S816D/S817D PSC in the against-flow direction only (Fig 7E). Together, these data suggest that disruption of the Rad54 D525-S816 interaction reduces the rate of linear translocation. This likely stems from slippage of the Rad54 motor due to unstable binding to the dsDNA.

Our experiments suggested that Rad54 S816D/S817D PSC is defective in linear translocation. The PSC can also perform loop extrusion, generating stretches of underwound DNA. This DNA is competent to bind RPA, and the degree of RPA binding can report on the formation of an extruded loop. To establish competent loops that bind RPA, we performed experiments in which the PSC bound to DNA in the absence of buffer flow for 5 min. We then measured the percentage of PSCs that co-localized with mCherry-RPA and the intensity of the associated RPA (Fig 7F). Approximately 63% (53/83) of WT PSCs recruited RPA. In contrast, only 34% (47/138) of Rad54 S816D/S817D PSCs had RPA associated (Fig 7G). Additionally, the mean intensity of RPA bound to the PSC was only 35% of WT (Fig 7H). From this, we conclude that Rad54 S816D/S817D exhibits reduced loop extrusion activity and is defective in both translocation mechanisms.

From our previous work the consequences of loop extrusion are likely to be the recruitment of additional PSC molecules. An indicator of additional recruitment would be an increase in the amount of Rad54 present, and an increase in the amount of Rad51-ssDNA. This change would be evident by comparing the amount of Rad54 and ssDNA with and without buffer flow. By using a photobleaching standard, we directly measured the amount of GFP-Rad54 present with and without buffer flow. In WT, the number of GFP-Rad54 molecules increased two-fold when buffer flow was turned off to allow loop extrusion (S12A Fig, p = 0.0007). This observation was not true in the Rad54 S816D/S817D PSC, and there was an observed decrease in the number of Rad54 molecules when buffer flow was turned off (S12A Fig). It is important to note that WT and Rad54 S816D/S817D exhibit the same number of Rad54 molecules when buffer flow is left on (S12A Fig). A similar pattern was observed for fluorescently labelled ssDNA (S12B Fig). From this, we conclude that loop extrusion by Rad54 can recruit more PSC molecules, and this is significantly reduced when disrupting the D525-S816 interaction.

Finally, we measured the ability of the Rad54 S816D/S817D mutant protein to form D-loops in vitro. We used a previously established assay to show that while Rad54 could promote recombination on a supercoiled plasmid, Rad54 S816D/S817D was unable to promote D-loop formation (S12C Fig). From this we conclude that Rad54 S816D/S817D is defective for D-loop formation both in vitro and in vivo.

Phosphorylation at S816 is likely a repressive mark

Phosphorylation of S816 was identified in two independent phosphoproteomic screens [69,70]. SILAC experiments suggested phosphorylation may occur under damaging conditions during the G1 phase of the cell cycle. The specific kinase is unclear. Based on our experiments, S816 phosphorylation is likely a repressive mark that downregulates Rad54 activity. Inactivation would help reduce recombination during G1, when the sister chromatid is absent. The most likely scenario is phosphorylation during G1 to reduce the potential for ectopic recombination during repair of double-strand breaks, increasing the fidelity of the genome throughout the cell cycle. Future work should be designed to identify the specific kinase responsible for phosphorylating Rad54.

Interactions with Sgs1

We identified an unexpected genetic interaction between RAD54 and SGS1. The specific details of this interaction are still unclear. We observed in the rad54-S816A, S817A sgs1∆ strain that there was a suppression of the increased amount of second-strand DNA synthesis observed in the sgs1∆ strain using a BIR-like system. A reasonable explanation is that in the rad54-S81A, S817A sgs1∆, there is a delay in the transition from strand pairing to the next phase in DNA repair synthesis, resembling the WT. We did not observe a reduction in the level of ssDNA extension. However, we did not perform a detailed kinetic analysis of this process, and a delay in timing is possible. The delay caused by the rad54-S816A, S817A allele may compensate for the loss of SGS1 and reduce the transition to a migrating bubble, lowering the efficiency of the second-strand synthesis on the newly synthesized ssDNA. This hypothesis is highly speculative, and further testing will be required to confirm it.

The interaction observed between rad54-S816D, S817D, and SGS1 is significantly more complicated to interpret. We observed a decrease in short tract gene conversion, and a decrease in extension from nascent D-loops. One possibility is that the D-loops formed in the rad54-S816D, S817D strain are short and structurally distinct from those formed in WT. This difference may alter Sgs1’s role or prevent its proper function. Alternatively, SGS1 is involved in DNA end-resection and the dissolution of dHJ intermediates [75,99,100]. Deletion of SGS1 delays the kinetics of end resection [6,101], potentially disrupting the overall timing of the recombination reaction. Similarly, Rad54’s failure to efficiently stimulate strand exchange may impose similar kinetic limitations on the recombination reaction. Therefore, the combined disruptive effects of RAD54 and SGS1 on the recombination reaction manifest as reduced DNA extension, altering gene conversion, and chromosome loss as the probability of a successful strand exchange is reduced.

Limitations of this study

A major limitation of this study is that we do not use physical methods to directly detect an interaction that may occur between the two RecA lobes of Rad54. Future work should focus on resolving the physical relationship between the two RecA lobes of Rad54. Ideally, this will come on the form of a high-resolution structure.

Yeast strains construction

Strains for the initial spot assay were BY4741 and generated by centromeric plasmid expression in the rad54∆ strain or by gene replacement, as indicated in the context. Strains for the ectopic recombination assay were TGI354 [75] and were generated by gene replacement or centromeric plasmid expression in the rad54∆ strain, as indicated in the context. Strains for the in vivo D-loop capture (DLC) and extension (DLE) assay were generated in the WDHY5511 [19,80] background by gene replacement. Strains for red/white allelic recombination were generated by gene replacement from parent strains LSY2205 and LSY2202 [87]. Strains in W303, TGI354, and WDHY5511 backgrounds were also used in the spot assay as indicated. All yeast strains used in this study are listed in S1 Table and all plasmids are listed in S2 Table.

Protein purification

Rad54, Rad54 S816A/S817A, Rad54 S816D/S817D, Rad51, and RPA-mCherry were purified as previously described [29]. In brief, a protease-deficient yeast strain was transformed with GFP-GST-Rad54, GFP-GST-Rad54 S816A/S817A, or GFP-GST-Rad54 S816D/S817D on a 2-µ plasmid under the control of the Gal1/10 promoter. Cells were grown in YNB (–Ura) plus 3% glycerol and 2% lactic acid. When the cells reached an OD₆₀₀ of 1.5, expression was induced by adding 2% galactose for 6 hours. Cells were harvested by centrifugation and stored at –80 °C.

Cell pellets were resuspended in Rad54 resuspension buffer (30 mM Tris–Cl [pH 7.5], 1 M NaCl, 1 mM EDTA, 10% glycerol, 10 mM BME (β-mercaptoethanol), protease inhibitor cocktail and 2 mM PMSF). Cells were disrupted by manual bead beating, and the lysate was clarified by centrifugation at 26,500 x g for 1 hour. The lysate was fractionated by ammonium sulfate (AS) precipitation. AS was gradually added with mixing to a final concentration of 20% followed by centrifugation at 10,000 x g for 10 minutes. The supernatant was discarded, and the AS concentration was raised to 50% followed by centrifugation at 10,000 x g for 10 min. The protein pellet was resuspended in PBS (phosphate buffered saline) plus 1M NaCl and 10 mM BME. The resulting re-suspended protein was then bound to pre-equilibrated GST resin in batch for 1 hour at 4°C. The GST resin was washed twice with PBS plus 1000 mM NaCl, and twice with PBS plus 500 mM NaCl. The protein was eluted in 20 mM glutathione in PBS plus 500 mM NaCl. The peak fractions were pooled and then applied to a Sephacryl S–300 High Resolution gel filtration column pre-equilibrated with Rad54 SEC buffer (30 mM Tris–Cl [pH 7.5], 500 mM NaCl, 1 mM EDTA, 10% glycerol, and 10 mM BME). The peak was pooled and dialyzed against Rad54 SEC buffer plus 50% glycerol and stored at –80 °C in single-use aliquots.

6xHis-SUMO-Rad51 was transformed into Escherichia coli BL21 (DE3) Rosetta2 cells and grown to an OD₆₀₀ of 0.4 to 0.6 at 37 °C. Expression was induced by addition of 0.5 mM IPTG for 3 hours at 37°C. Cells were harvested and stored at –80 °C. Cells were lysed by freeze–thaw in Cell Lysis Buffer (CLB: 30 mM Tris–Cl [pH 8.0], 1 M NaCl, 10% glycerol, 10 mM imidazole, 5 mM BME, and protease inhibitor cocktail). Crude lysates were sonicated for 6 pulses of 30 seconds on and 2 minutes off, then clarified by centrifugation at 26,500 x g. The extract was precipitated with 50% AS and centrifuged at 26,500 x g for 10 minutes. The pellet was resuspended in CLB and bound to 1 mL of pre–equilibrated Ni–NTA resin for 1 hour with rotation at 4 °C. The resin was washed three times with CLB and eluted in CLB plus 200 mM imidazole. The protein was mixed with 400 units of the SUMO protease Ulp1 and dialyzed overnight at 4 °C into Rad51 buffer (30 mM Tris–Cl [pH 8.0], 150 mM NaCl, 1 mM EDTA, 10% Glycerol, 10 mM imidazole). The 6xHis–SUMO tag and SUMO protease were removed by passing the dialyzed proteins over a second 1 mL Ni–NTA column. The purified Rad51 was then stored at –80 °C in single-use aliquots.

Multiple sequence alignment

Rad54-like eukaryotic sequences were retrieved by BLASTP (NCBI, nr database; organism filter: Eukaryota) using S.c.Rad54 as the query. Redundant entries and obvious fragments were removed. Sequences were aligned with Clustal Omega, the alignment was trimmed with trimAl, and sequence logos were generated from the trimmed alignment using the Python package Logomaker.

Yeast spot growth assay

RAD54 and its designed mutants were expressed using a centromere vector, pRS415, and transformed into the BY4741 rad54 strain or other background strains as indicated, followed by selection on YNB (–Leu) + 2% dextrose plates. Transformed cells were grown overnight in YNB (–Leu) + 2% dextrose medium. The following day, the overnight cultures were diluted to an OD₆₀₀ of 0.3 and grown to an OD₆₀₀ of 1.0. Cells were then serially diluted and spotted on YNB (–Leu) + 2% dextrose plates containing either no drug, 0.01% methyl methanesulfonate (MMS), or 0.02% MMS. Other drugs were also used, including 20 mM Hydroxyurea, 0.02 µg/mL 4-NQO, 0.04 µg/mL 4-NQO, 10 µM camptothecin (CPT), and 20 µM CPT as indicated. Plates were incubated at 30 °C for 3 days and imaged at 72 hours. BY4741 strains with RAD54 mutants were also generated via gene replacement. The procedures for the spot assay for these strains were the same as for the centromeric-expressed strains, except that the overnight culture was grown and diluted in Yeast extract + Peptone (YP) + 2% dextrose medium. The serially diluted cells were spotted on YP + 2% dextrose containing either no drug or the drugs mentioned above. Plates were incubated at 30 °C for 2 days and imaged at 48 hours. For RAD51 or rad51I345T on a pYES plasmid, cells were transformed. Cultures were grown overnight in YNB (–Ura) + 2% dextrose and diluted the next day. After cells reached an OD₆₀₀ of 1.0, they were serially diluted onto YNB (–Ura) + 2% dextrose or YNB (–Ura) + 2% galactose and incubated at 30 °C for 48 hours. They were then images.

Ectopic recombination assay

The wildtype strain used for this study was described in [75]. The genotypes for modifying these strains are listed in S1 Table. A single colony was picked and grown in YP + 3% glycerol + 2% lactate medium to log phase. The culture was diluted and plated on YP + 2% dextrose and YP + 2% galactose plates, respectively. Plates were incubated at 30 °C for 2–3 days, and the number of colonies was counted. The viability rate was calculated by dividing the colony number on the YP + 2% galactose plate by that on the YP + 2% dextrose plate. The mean and standard deviation were calculated for multiple independent experiments as indicated. For centromeric expression of RAD54 and its mutants in rad54 strains, a single colony was picked and grown in YNB (–Leu) + 2% dextrose overnight. On the second day, the culture was diluted 10-fold into YNB (–Leu) containing 3% glycerol and 2% lactate and grown at 30 °C for 6 hours. Then the culture was diluted and plated onto YNB (–Leu) + 2% dextrose and YNB (–Leu) + 2% galactose, respectively. After being incubated at 30 °C for 3–4 days, the numbers of colonies were determined, and the viability was calculated by dividing the colony number on the galactose plate by that on the dextrose plate.

Allelic Recombination between homologs assay

The assay was performed by growing the appropriate strain overnight in YP + 2% raffinose. The next day, cells were diluted to an OD₆₀₀ of 0.2 and allowed to reach an OD₆₀₀ of 0.4-0.5, then I-Sce1 expression was induced by adding 2% galactose. Cells were allowed to grow for an additional 1.5 hours, then plated onto YP + 2% dextrose (YPD) plates and grown for 48 hours. After 48 hours, they were placed at 4 °C overnight to enable further development of the red color. The number of white, red, and sectored colonies was then counted, followed by replica plating onto YPD + hygromycin B (200 µg/ml) and YPD + nourseothricin (67 µg/ml, clonNat) for analysis of recombination outcomes. Strains were also replica plated on YNB (–Ura or –Met) + 2% dextrose to ensure proper chromosome segregation and YNB (–adenine sulfate) + 2% raffinose +1% galactose to assay for recombinants versus uncut DNA. Finally, colonies were replica plated on YNB (–adenine sulfate) + 2% dextrose to directly count short tract gene conversion events. The data were analyzed by counting sectored colonies and assessing colony survival across different antibiotic sensitivities. The data for each category was then divided by the total population of sectored colonies.

Yeast cell imaging

To immobilize yeast cells, an agarose pad was used. A 1% agarose solution was prepared by adding agarose in S media (YNB + 0.5% ammonium sulfate, then autoclaved) and microwaving to dissolve. An amino acid mix was added to a final concentration of 1x to ensure yeast cells could grow on the agarose pad. The dissolved agarose was kept warm on a 65 °C metal heater. One clean microscope slide was taped at both ends, and 30 µL of 1% agarose was applied to the center of the slide. Another clean and untapped microscope slide was immediately placed on top, and gentle pressure was applied to ensure the agarose spread evenly. Then the slide sandwich was placed on an ice bag for approximately 30 seconds to allow the agarose to solidify. The top slide was carefully slid out, leaving the agarose pad intact on the bottom slide. The agarose pad was allowed to dry for 3 minutes before 3 µL of properly diluted yeast cells was added. A coverslip was then placed over the cells. A Nikon Eclipse Ti microscope equipped with a 488-nm laser. A Plan APO 60XA/1.20 WI objective and an ANDOR Zyla-5.5-CL3 camera were used to observe and capture yeast cells.

ATPase assay

A commercially available ADP-GLO kit (Cat No. V6930, Promega) was used to measure ATP hydrolysis activity. ATP hydrolysis reaction was performed in HR buffer (20 mM Tris-OAc [pH 7.5], 50 mM NaCl, 10 mM Mg(OAc)₂, 200 ng/μl BSA, 1 mM DTT, and 10% Glycerol) and contained 1 mg/ml sheared salmon sperm DNA, 20 nM Rad54, and 200 nM Rad51 (if added).

Electrophoresis mobility shift assay (EMSA) for Rad54

An Atto647N-labeled 90-mer oligo was annealed with an unlabeled complementary oligo to form a labeled 90-bp dsDNA substrate. The oligo sequences are available in S3 Table. The binding reaction was performed in EMSA buffer (35 mM Tris-Cl [pH 7.5], 3 mM MgCl₂, 50 mM KCl, 1 mM DTT, 10% glycerol). The final DNA concentration was 10 nM, and proteins were titrated to be 0, 6.25, 12.5, 25, 50, 100, 150, and 200 nM as final concentrations for Rad54 wildtype and Rad54 S816D/S817D. Rad54 S816A/S817A was titrated to be 0, 6.25, 12.5, 25, 50, 100, and 139 nM as final concentrations. The DNA and proteins were incubated at 30 °C for 5 min and then resolved by 8% Native-PAGE in 0.5x TBE buffer (44.6 mM Tris, 44.5 mM boric acid, 1 mM EDTA, 8% acrylamide/bis-acrylamide (37.5:1), 0.1% APS, 0.1% TEMED) running in 0.5x TBE buffer (44.6 mM Tris, 44.5 mM boric acid, 1 mM EDTA). The hill curve of Rad54-DNA binding was fitted through GraphPad prism.

In vitro D-loop assay

D-loop formation experiments were performed in HR buffer (30 mM Tris-OAc [pH 7.5], 50 mM NaCl, 10 mM Mg(OAc)₂, 1 mM DTT, 0.2 mg/ml BSA) using an Atto647N-labeled DNA duplex (15 nM) with homology to the pUC19 plasmid. Rad51 (300 nM) was incubated with recipient DNA at 30 °C for 15 minutes. The resulting Rad51 filaments were mixed with indicated concentrations of Rad54, RPA (500 nM), and pUC19 plasmid (0.3 nM). Reactions were quenched at the indicated time points and treated with 1 unit of Proteinase K at 37 °C for 20 minutes. The reactions were then resolved by electrophoresis on a 0.9% agarose gel and imaged for fluorescence using a Typhoon imager.

Flow cell construction

Metallic chrome patterns were deposited on quartz microscope slides with predrilled holes for microfluidic line attachment by electron beam lithography to generate flow cells. After metal deposition, a channel was created by placing a small piece of paper between the drill holes and covering the two-sided tape. The paper was excised to make the flow chamber, and a glass coverslip was fixed to the tape. The chamber was sealed by heating to 165 °C in a vacuum oven at 25 mmHg for 60 min. Flow cells were then completed by hot-gluing IDEX nano ports over the drill holes on the opposite side of the microscope slide from the coverslip.

Single molecule experiments

All single molecule experiments were conducted on a custom-built prism-based total internal reflection microscope (Nikon) equipped with a 488-nm laser (Coherent Sapphire, 100 mW), a 561-nm laser (Coherent Sapphire, 100 mW), a 640-nm laser (Coherent Obis, 100 mW) and two Andor iXon EMCCD cameras. DNA substrates for DNA curtains experiments were made by attaching a biotinylated oligo to one end of the 50 kb Lambda phage genome and an oligo with a digoxigenin moiety on the other. This enabled double tethering of the DNA between the chrome barriers and the chrome pedestals, as previously described. As for single tethered DNA, if specified in the context, flow cells were attached to a microfluidic system, and sample delivery was controlled using a syringe pump (KD Scientific). Three-color imaging was achieved by two XION 512 × 512 back-thinned Andor EM-CCD cameras and alternative illumination using a 488 nm laser, a 561 nm laser, and a 640 nm laser at 25% power output. The lasers were shuttered, resulting in a 200-msec delay between each frame. Images were collected with a 200-msec integration time. Translocation velocity and distances were measured in HR Buffer. Channel bleed-through is prevented by shuttering of the laser lines, emission filters, and the use of complementary fluorophores. In this case GFP is not activated by the 561 or 647 laser lines, and the 488 or 561 laser lines do not activate Atto647N. mCherry labelled the 488-laser line can poorly activate fluorophores. However, the emitted light is split by a dichroic mirror and filtered through a band-pass and long-pass filter to block wavelengths of light above a certain cutoff. This is sufficient to prevent mCherry signal bleed through into the 488 channel.

Analysis of dsDNA translocation

The velocity and track length for GFP-Rad54 molecules were measured by importing raw TIFF images as image stacks into ImageJ. Kymographs were generated by pointing around a fluorophore signal and defining the whole individual dsDNA molecules where the fluorophore bound to a region of interest (ROI) using a home-made script. Data analysis was performed from the kymographs. The start of translocation was defined when the GFP-Rad54 molecule moved > 2 pixels. Pauses were defined as momentary stalls in translocation that lasted 2–4 frames. Termination was defined by molecules that did not move for > 10 frames. Velocities were calculated using the following formula [(Y_f–Y_i) ×1031.96 bp/ [|X_f–X_i|]) ×frame rate]; where Y_i and Y_f correspond to the initial and final pixel position and X_i and X_f correspond to the start and stop time (in seconds). Negative values stand for translocations moving from barrier to pedestals, which was assisted by a flow. Positive values stand for translocations against flow directions.

Dissociation rate measurement

The dissociation rate of DNA-GFP-Rad54 was measured by initiating data collection, followed by the injection of 10nM GFP-Rad54 in HR buffer onto single-tethered DNA curtains to allow proteins to bind to DNA. The flow rate was maintained at 0.2 mL/min to allow for initial binding and prevent further binding events. Images were collected with a 100 msec integration time at 200 msec intervals for a duration of 5 minutes for the bleaching experiment, or at 1 sec intervals for a period of 15 minutes for the dissociation experiment. The fluorescent signal decay in both experiments was fit to an exponential function, . The difference between the values obtained from bleaching experiment and dissociation experiment was used to calculate the dissociation rate:.

DLC assay

DLC assay was performed as described before [19,80]. In brief, yeast cells were grown in a 5 mL YP + 2% dextrose + 4% adenine sulfate medium overnight. The second day, the culture was diluted by 10-fold in 5 mL YP + 3% glycerol + 2% lactate + 4% adenine sulfate and grown for around 8 hours. Then the culture was inoculated into 100 mL of YP + 3% glycerol + 2% lactate + 4% adenine sulfate medium with an initial OD₆₀₀ ≈ 0.006 and grown for 16 hours. A 5x psoralen stock solution (0.5 mg/mL trioxsalen in 200-proof ethanol) was made in a 50-mL aluminum foil-covered falcon tube and dissolved on a shaker at room temperature overnight with gentle rocking. Next day, the culture should have an OD₆₀₀ at 0.3-0.8. 7.5 OD₆₀₀ of cells was collected as time 0 control, centrifuged at 2,246 x g, 4 °C for 5 minutes.

The cell pellets were resuspended in 1x psoralen buffer. The 1x psoralen buffer was prepared by diluting 5x psoralen in 200-proof ethanol before collecting cells. The resuspended cells were plated in a 60 mm x 15 mm petri dish, put 2–3 cm below a UV light source with the lip removed atop a pre-chilled metal block. The cell samples were exposed under the UV light for 10 minutes with gentle shaking to crosslink DNA. The cells were transferred to a 15-mL Falcon tube. The petri dish was rinsed with TE1 solution (50 mM Tris-Cl [pH 8.0], 50 mM EDTA), and the TE1 buffer was poured over the cells. The cells were then centrifuged at 2,246 x g, 4 °C for 5 minutes again. The supernatant was properly disposed of, and the pellets were stored at –20 °C. Galactose was added to the culture to a final concentration of 2% to induce DSBs. Cells were collected at the designed time points as described above. For lysis, the cell pellets were thawed on ice, then resuspended in spheroplasting buffer (0.4 M sorbitol, 0.4 M KCl, 40 mM sodium phosphate buffer [pH 7.2], 0.5 mM MgCl₂) and transferred to a 1.7 mL microfuge tube.

The cells were spheroplasted in zymolyase solution (2% glucose, 50 mM Tris-Cl [pH 7.5], 5 mg/mL zymolyase 100T) at 30 °C for 20 minutes. The cells were washed with spheroplasting buffer three times at 2,500 x g and restriction enzyme buffer (RE buffer: 50 mM potassium acetate, 20 mM Tris-OAc, 10 mM Mg(OAc)₂, 1 mg/mL BSA) at 16,000 x g three times. The pellets were resuspended in 1.4x RE buffer, either alone or with a hybridization oligo to restore the EcoRI restriction sites, and stored at –80 °C. The DNA was solubilized by incubating the cells with 0.1% SDS at 65 °C for 13 minutes. 1% Triton X-100 quenched the SDS. The DNA was digested by 20 U EcoRI at 37 °C for 1 hour. The restriction enzyme was deactivated by incubating the DNA with 1.5% SDS at 55 °C for 10 minutes. The cells were returned to ice, and SDS was quenched by the addition of 6% Triton X-100. Ligation buffer (50 mM Tris-Cl [pH 8.0], 10 mM MgCl₂, 10 mM DTT, 2.5 μg/mL BSA, 1 mM ATP, pH 8.0, 8 U T4 DNA ligase) was added to perform the ligation reaction at 16 °C for 1 hour and 30 minutes. 25 μg/mL protease K was added to digest the enzymes at 65 °C for 30 minutes. DNA was extracted by adding phenol:chloroform:isoamyl alcohol and vortexing. The upper water phase was moved and incubated with a tenth volume of sodium acetate and a volume of isopropanol at room temperature for 30 minutes and centrifuged at 21,130 x g, 4 °C for 10 minutes to get DNA precipitation. The DNA pellets were dried at 37 °C and dissolved by incubating with 1xTE buffer (10 mM Tris-Cl [pH 8.0], 1 mM EDTA) at 37 °C for 1 hour. The DNA was used as a qPCR template with the primers listed in S3 Table. DLC chimera content was calculated by [DLC amplification efficiency]^[-Cp_(DLC)], and the intramolecular ligation product content was calculated by [intramolecular ligation amplification efficiency]^[-Cp_(ligation)]. The final DLC signal was calculated as DLC chimera content divided by intramolecular ligation product content.

DLE assay

DLE assay was performed as described before [19,80]. The procedures were similar with DLC assay except for (1) 2.5 OD₆₀₀ of cells were collect at each time point; (2) DNA crosslinking was omitted and cell pellets were washed by TE1 buffer twice at 2,246 x g; (3) hybridization oligos were different and listed in S3 Table; (4) DNA was solubilized by 1% SDS at 65 °C for 15 minutes; (4) DNA digestion was performed by adding 20 U HindIII; (5) qPCR was performed using some different oligos listed in S3 Table.