The D-loop intermediate is a vital regulatory point in HR, and several factors regulate D-loop formation as well as disruption (Wright et al., 2018). Factors that form a D-loop may affect D-loop properties such as D-loop position and length, which may, in turn, influence the likelihood of a D-loop disruption. The following D-loop properties influence D-loop disassembly. First, the position of a D-loop will influence whether or not the 3′-OH end of invading DNA is incorporated into the hDNA. D-loops containing an annealed 3′-OH end are primed for an extension by a DNA polymerase and are less likely to be reversed (Li and Heyer, 2009). Conversely, D-loops with a 3′-flap may become a loading pad for D-loop disruption enzymes. Helicases and/or topoisomerases such as Sgs1-Top3-Rmi1, Mph1, and Srs2 reverse D-loops with different specificities (Fasching et al., 2015; Liu et al., 2017; Piazza et al., 2019; Prakash et al., 2009; Putnam et al., 2009). Internal D-loops also increase the possibility of a second invasion near the 3′-end, either in the same donor (Wright et al., 2018) or a different donor DNA (multi-invasions) and the formation of potential genomic rearrangements (Piazza and Heyer, 2018). Second, the presence of mismatches within hDNA enhances the D-loop disruption in a process termed heteroduplex rejection (Chakraborty et al., 2016; Honda et al., 2014). Thus, D-loops formed ectopically at heterologous sites would have a higher likelihood of reversal than at fully homologous donors, influencing the choice of the donor (Chakraborty et al., 2016). Third, the length of a D-loop might influence D-loop stability and likely their susceptibility to disruption enzymes. Long extended D-loops are also more likely to have second-end capture, leading to dHJ formation and the possibility of a crossover outcome (Wright et al., 2018). In summary, D-loop properties that influence D-loop reversibility regulate donor choice, the possibility of forming dHJ, and a crossover outcome, as well as multi-invasions and resultant genomic rearrangements (Piazza and Heyer, 2018; Piazza et al., 2019). Thus, these features underline the importance of studying D-loop characteristics.

Furthermore, several interlinked factors influence D-loop characteristics. First is the resection length, where longer resected tracts permit the formation of longer D-loops. However, in vivo hypo-resection is usually rare, as evidenced by the highly processive resection machinery (Mimitou and Symington, 2011). As resection tract length increases, it allows for more possible internal invasions, influencing D-loop position, length, and potentially also donor choice. Internal D-loops could be disrupted to try end-invasion again, the hDNA may be extended or migrated towards the 3′-end, or the 3′-flap may be cleaved or involved in a second invasion at a different site (Wright et al., 2018). Second, Rad51 filament properties such as length, position, composition (inclusion of accessory factors) and stability, may influence Rad54 translocation activity with subsequent effects on D-loop characteristics. Third, the length of homology between the invading and donor DNA decides the longest possible D-loop length. Subsequently, long homology lengths are associated with increased crossovers (Inbar et al., 2000). Fourth, the topology at donor site or chromatin remodelers altering DNA topology may influence the position and length of a D-loop that is highly favored in negatively supercoiled regions (Wright et al., 2018). Lastly, the extent of DNA synthesis may or may not directly reflect the length of the extended D-loop. The nascent D-loop can either be extended past the initial point of invasion with DNA synthesis or be migrated along with DNA synthesis. Thus, the gene conversion tracts (Guo et al., 2017; Neuwirth et al., 2007) that reflect the length of recombination-associated DNA synthesis do not necessarily mirror the length of the D-loop. Yet gene conversion tract lengths positively correlate with a crossover outcome (Guo et al., 2017). In summary, D-loop properties, as well as factors that influence D-loop properties, impact the repair outcome. Thus, it is crucial that an assay that permits the characterization of D-loops at the single-molecule level to be developed.

Since D-loops are dynamic in nature, it has been challenging to develop an experimental method to study D-loops. Recent advancements in the study of nascent and extended D-loop dynamics includes the development of the D-loop capture and the D-loop extension assays, as well as modifications to a widely-used assay called the D-loop assay (Piazza and Heyer, 2018; Piazza et al., 2019; Wright and Heyer, 2014). Though D-loop assays typically utilize a short,~100 nt ssDNA as the broken strand, Wright and Heyer, 2014 showed that D-loops formed using physiological length ssDNA substrates with a 5′-duplex DNA may better recapitulate D-loop dynamics in vitro. Wright and Heyer, 2014 also developed a restriction-digestion based assay to map the location of D-loops across the region of homology. However, the assay is limited by the presence and distribution of unique restriction sites across the homologous region. Moreover, the method relies on the use of a supercoiled duplex donor to restrict D-loop length. Additionally, the distance between two restriction sites should be greater than the maximum possible D-loop length in a supercoiled donor to distinguish between the D-loops formed at the two sites. Hence, while the restriction-digestion mapping assay broadly reveals D-loop distribution in a population of in vitro D-loops, it cannot define the length of individual D-loops.

In vivo detection of D-loops has been even more challenging, especially somatic D-loops. Meiotic single-end invasions (SEIs) that likely represent a particular form of extended D-loop destined to becoming a crossover are detected using 2D-gel electrophoresis (Hunter and Kleckner, 2001). Recently, using a proximity-ligation-based method, Piazza and colleagues demonstrated the ability to quantitatively measure somatic nascent D-loop formation in vivo (Piazza et al., 2018). The sensitivity of the assay relies on preserving the D-loops through psoralen-crosslinking. However, the crosslinking efficiency limits the assay by prohibiting the detection of D-loops shorter than the crosslinking density of ~500 bp. Thus, the assay falls short in the ability to distinguish a change in D-loop signal caused by a change in D-loop length or an alteration in D-loop quantity among a population of cells. Other assays either detect extended D-loops, a downstream product of nascent D-loop (Hicks et al., 2011; Piazza et al., 2018), or measure the extent of DNA synthesis via gene conversion tracts (Guo et al., 2017; Neuwirth et al., 2007), and thus, do not reflect the dynamic nascent D-loop properties. Thus, there is an unmet need for a method allowing the study of D-loop properties, such as its length, position, and distribution.

Here, we developed a novel high-throughput assay to detect for the first time the length and position of D-loops formed in vitro at the single-molecule level and near base-pair resolution. Subsequent analysis of individual D-loops reveals the distribution of D-loop lengths and position in a population of D-loops. Lack of enrichment or a crosslinking step permits relatively unbiased D-loop analysis. The assay is based on non-denaturing bisulfite sequencing, adapted from the approach used to map R-loops that are formed by RNA invasion in DNA (Malig et al., 2020; Yu et al., 2003). The method allows easy multiplexing of several D-loop samples with high coverage. The method is accompanied by computational analysis and data visualization pipeline for D-loop profiling. The method is widely applicable to study (i) D-loops formed with a wide range of ssDNA substrates and donor DNA, (ii) factors altering D-loop properties (iii) D-loop disruption factors and their preferences for D-loop features. We also discuss the limitations and possible refinements of this method.

To establish and optimize the D-loop Mapping Assay (DMA), we started by using a size-restricted D-loop sample formed in vitro. To do this, we used a supercoiled plasmid as a dsDNA donor to perform an in vitro D-loop reaction (Figure 1A; Wright and Heyer, 2014). For every 10.4 bp of hDNA generated in the D-loop, one negative supercoil from the dsDNA is consumed. Plasmids purified from E. coli have a mean supercoiling density of σ = −0.07 supercoil/turn (Champion and Higgins, 2007), which we verified independently (Solinger et al., 2002). Thus three kbp plasmids such as those used in this study should contain ~20 ± 5 negative supercoils, meaning that they can accommodate a maximum of ~210 ± 50 bp hDNA, even if the length of the invading homologous ssDNA is longer. Longer D-loops would cause the introduction of compensatory positive supercoils, which is energetically unfavorable. By contrast, fully relaxed plasmids carrying D-loops ~ 210 bp long are expected to be in the lowest possible energy state and thus favored at equilibrium (Wright et al., 2018). Thus, such an experimental setup allowed us to establish the DMA by predefining the expected lengths of the D-loops.

Figure 1

Download asset Open asset

Once formed, the D-loop reactions were stopped and subjected to bisulfite treatment at room temperature to allow cytosine-to-uracil conversions on the single-stranded displaced strand. Using primers that are non-biased to conversion and outside the region of homology, the DNA was PCR amplified, purified, and subjected to single-molecule real-time sequencing (Figure 1B). The sequencing reads were then processed, mapped to the reference sequence, and analyzed to reveal cytosine conversion regions as D-loop footprints. The cytosine conversion regions were defined as D-loop footprints only when they crossed a pre-specified peak threshold. The peak threshold was implemented to filter out any background cytosine conversions from spontaneous DNA breathing (for details, see Materials and methods).

Negative controls at each step of the analysis further enhanced the strength of this well-defined in vitro design (Figure 1C). First, a > 1,000 bp non-homologous dsDNA region, where no D-loop footprint is expected, flanked the homologous region. This serves as an internal negative control within each sequencing read. Second, the homology window itself can be further altered by using ssDNA substrates of different lengths, varying in this study from 197 nt to 931 nt. Third, D-loop footprints should not exceed 260 nt for D-loops formed using supercoiled dsDNA donors. This control holds true only for topologically restricted supercoiled donors, and not linear dsDNA donors. Fourth, we expect D-loop footprints to exist only on reads originating from the top strand of dsDNA donor plasmids that is displaced upon strand invasion. The bottom-strand, by contrast, should not harbor any footprints. This represents a further internal control within each sample. Lastly, an experimental negative control can be generated by using a D-loop reaction performed in the absence of invading ssDNA or the Rad51 recombinase. Without ssDNA or Rad51, no strand invasion footprint is expected. Thus, these predicted positions and/or length of D-loops will provide confidence that the footprints derived solely from D-loops.

After establishing the assay using a supercoiled donor, a linearized plasmid, free of topological constraints, was used as a donor to map the D-loops in a topologically uninhibited manner. Again, a range of ssDNA substrates was used with varying lengths of homology to the donor (ds98-915, ds98-607, ds98-197-78ss), different sequences (ds98-931 vs. ds98-915) or presence of a non-homologous 3′-flap (ds98-915-78ss) (Figure 4A). Substrates with different homology lengths with respect to the donor were used to determine whether this would result in corresponding changes in D-loop length distribution. While substrates with similar homology length but different homologous sequences to the donor were used to test for any sequence bias contributing to the D-loop distribution. The ds98-931 and ds98-915 substrates differ in sequence by >33%, and the sequences are provided in the Key Resources Table. Lastly, the substrate with a non-homologous 3′-flap was used to test if 3′-heterology affects the enrichment of D-loops at the 3′-end of homology.

Figure 4 with 3 supplements see all

Download asset Open asset

As expected, the D-loop footprints for each of these substrate types were specifically seen only on the top strand, as evident in the cumulative read data (Figure 4B). A total of ~800–3,700 (10–18%) D-loops were observed among ~4,000–40,000 top-strand reads (Figure 4B). Again, while the bottom-strand comprised ~13,000–100,000 reads, it had <0.1% of reads with a footprint. (Here, a 10-fold difference in the total number of reads analyzed was due to upgrade in Pacific Biosciences sequencing system, from Sequel -I to Sequel – II. See Material and methods). Thus, even with a linear, topologically unrestricted donor, the D-loops were mapped with high strand-specificity (>100 fold). Relative D-loop levels quantified from DMA correlated with the D-loop levels from the gel-based assay (Figure 4—figure supplement 1A,B). Although, as expected from previous observations with negatively supercoiled donors, the D-loop levels estimated from DMA were lower than the ones from the gel.

Moreover, the D-loop footprints were essentially all found within the window of homology for each substrate type, as evident from individual footprint maps (Figure 4C, Figure 4—figure supplement 2A–C), attesting to the homology-specificity of the footprints. In some cases, less than 0.4% of footprints were found outside the region of homology (Figure 4—figure supplement 2B). These footprints were found on both the top strand and bottom-strand reads, attesting their origin to be due to DNA breathing in a C-rich region. Next, the distribution of the D-loop position with a linear donor revealed an enrichment at the 3′-end of homology (Figure 4D, Figure 4—figure supplement 2D,E), similar to the D-loops formed using a supercoiled donor. Moreover, the presence of a non-homologous 3′-flap in ds98-915-78ss sample did not significantly alter the distribution of the D-loop position (Figure 4D) compared to the ds98-915 sample. Despite the flap, the D-loops were enriched at the 3′-end of homology in a similar fashion. Since the non-homologous flap is only 78 nt long, it may not alter the distribution of Rad51 filament significantly, resulting in similar D-loop allocations. It may be interesting to test if a non-homologous flap longer than the average D-loop length of 450 nt has any effect on D-loop distribution.

The distribution of D-loops was also unaltered by changes in the homology sequence between the ds98-931 and ds98-915 substrates (Figure 4D, Figure 4—figure supplement 2D). This suggests that differences in cytosine distribution across the homology do not significantly alter D-loop distribution, as long as the % CG is not very low. In these cases, the % CG were >40%. Thus, these observations further support the 5′-to-3′ directionality of the Rad51 filament growth (Qiu et al., 2013; Špírek et al., 2018).

Interestingly, there was no preferred D-loop length. Instead, a broad range of D-loop lengths was observed, independent of the substrate type (Figure 4E,F). The lengths of the D-loops ranged from ~100 nt to a maximum spanning the entire length of homology. D-loops as long as 900 nt were seen with the long substrates (ds98-915, ds98-931, and ds98-915-78ss) at a frequency of ~5% of the total D-loops (Figure 4E). The minimum D-loop size of 100 nt is limited by the peak threshold (of t40w50) used to define a D-loop footprint. An average D-loop length of ~390 nt was noted with the 900 nt homology series substrates. There was a slight, yet significant difference in the D-loop lengths between the ds98-915 and ds98-931, with a difference in mean of 60 nt. Two factors may contribute to this difference in D-loop lengths. One, the 16 nt longer homology in ds98-931 may contribute to relatively longer average D-loop length. Second, the two substrates differ in their homologous sequences. The ds98-915 substrate includes 192 cytosines, while the ds98-931 has 187 cytosines within the region of homology. Thus, the relatively less frequent cytosine distribution in ds98-931, despite the longer homology length, may consequently result in the observed increase in the mean D-loop length. However, there was no statistical difference in D-loop lengths between the ds98-915 and ds98-915-78ss substrates, suggesting that the 78 nt 3′-flap does not alter D-loop lengths. Overall, >40% of footprints were longer than the average length of 400 nt for each of these three substrates. Thus, the differences in the homology sequence or the presence of a non-homologous 3′-flap did not significantly alter the distribution of D-loop lengths.

As expected, the average D-loop length decreased from ~390 nt to ~320 nt, when the homology length was reduced to 607 nt using the ds98-607 substrate (Figure 4F). The maximum D-loop length for the ds98-607 substrate was 600 nt, restricted only by the homology length. About 6% of the D-loops were longer than 500 nt, traversing across the region of homology. Similarly, the D-loops formed with the ds98-197-78ss substrate had an average length of 176 nt, and ~73% spanned the entire region of homology. The minimum D-loops length observed was ~100 nt limited by the peak threshold of t40w50. In summary, our results demonstrate that homology length defines the maximum observable D-loop length, and in turn, influences the average D-loop length. Note that since the D-loops formed on a linear donor are relatively less stable than those on a supercoiled donor (Wright and Heyer, 2014), it is possible that the footprints may reflect all changes in D-loop length and position that occurred over the course of the bisulfite treatment.

Formation of multi-invasions, where a single substrate forms D-loops simultaneously in multiple donor molecules, was shown both in vitro and in vivo (Wright and Heyer, 2014; Piazza et al., 2017). However, multi-invasions cannot be detected by this assay, since each donor molecule is separately analyzed. Instead, in some cases (~2–5%), multiple D-loop footprints are seen within a single read (Figure 4—figure supplement 2A–C), representing a single donor. This may be due to multiple invasions of one or more substrates into a single donor at two different sites (Figure 4—figure supplement 2F). Multiple invasions from a single substrate at two sites within a donor may arise due to the formation of two Rad51 filaments separated by a gap. However, Rad54 may be able to translocate through such a gap, making these kinds of multiple invasions a rarity. Alternatively, two substrates invading a single donor may also give rise to dual footprints. Conversely, two D-loop footprints on a single read may also arise due to inefficient cytosine conversions across a long D-loop, possibly due to the continued presence of RPA on the displaced strand. Unless the two footprints are separated by >200 nt (~50 cytosines), it is currently hard to distinguish between the two possibilities.

We also similarly tested the D-loop mapping assay on D-loops formed using the human recombination proteins RAD51, RAD54, and RPA. The rationale for using human proteins was to test whether the D-loop length and distribution was an inherent property of the yeast Rad51 recombinase. In vitro D-loop reactions with human proteins were performed using the ds98-931 substrate and a supercoiled donor (Figure 4—figure supplement 3A) or a linearized donor (Figure 4—figure supplement 3C). With both donor types, the D-loop footprints were again specifically seen within the homologous region (Figure 4—figure supplement 3B,D) and on the top strand (Supplementary file 1). The distribution of D-loops was similar to the D-loops produced by the yeast proteins. An enrichment of D-loops was seen at the 3′-end of homology, irrespective of the donor type (Figure 4—figure supplement 3E). Average D-loop length seen with a supercoiled donor was 252 ± 78 nt as expected, while with a linear donor, the average length was 320 ± 140 nt (Figure 4—figure supplement 3F). The average D-loop length from a linear donor was slightly lower than the 410 nt seen with yeast proteins. Conformingly, only 21% of the D-loops were longer than 400 nt with human recombinant proteins compared to >40% seen with yeast recombinants. Only one D-loop spanned the entire length of homology. Thus, compared to the yeast proteins, human proteins seemed to be less efficient under the conditions used in forming longer D-loops. As evident from observation of the D-loops on an agarose gel (Figure 4—figure supplement 3G,H), the efficiency of D-loop formation was also much lower, with only 3% using a linear donor. This could be due to the poor stability of the RAD51 filament. It may also suggest a role for RAD51 accessory factors in stabilizing the RAD51 filament and altering its properties to promote more efficient D-loop formation (San Filippo et al., 2008). Concomitantly, the percentage of D-loops observed on a gel correlated well with the frequency of D-loop footprint (Figure 4—figure supplement 3H). Thus, the D-loop Mapping Assay can be broadly used to test the effect of recombinant proteins across various species on D-loop properties.

While optimization of peak threshold helped minimize false positives and false negatives, we wondered if it had any effect on the overall D-loop features. To address the effect of peak thresholds on D-loops called, we examined D-loop lengths and distribution under the various peak threshold conditions with the ds98-931 substrate and a linear donor.

In terms of the D-loop lengths, as one may expect, more short-length footprints were called with the low stringency conditions. With t25w50, 22% of the footprints were shorter than 250 nt, compared to 5% with a t40w50 threshold. Consequently, the average D-loop length with a t25w50 threshold was 357 nt compared to 410 nt with a t40w50 threshold (Figure 5C). Additionally, lowering the window size from w50 to w20 further lowered the average D-loop length. The average length was 220 nt for t40w20, with 59% of the D-loops being shorter than 200 nt. Thus, with lower window size, more of the short D-loop footprints were captured along with some footprints potentially arising from spontaneous DNA breathing, leading to an overall reduction in average D-loop length.

Note that genuine-D-loops shorter than 250 bp would be most affected by the window size. This is because a window size of 20 would result in a minimum D-loop size of 80–100 bp, considering a CG distribution of >40%. While using a window of 50, the minimum D-loop size would be 120–200 bp. Hence, for w = 50, D-loops that were originally 180 bp would be slightly overestimated in size as 200–250 bp, assuming that the cytosine conversion frequency permits it. On the other hand, longer D-loops will be unaffected by the window size. The length of genuine D-loops will also be unaffected by a change in conversion frequency below the average bisulfite conversion frequency.

Lastly, changes in peak threshold did not affect the overall distribution of D-loops within the homology window (Figure 5D). Thus, despite the detection of short non-specific footprints at lower thresholds, there was no bias in the position of these non-specific footprints, meaning that there was no overall change in their distribution.

The D-loop Mapping Assay provides the following advantages:

1. DMA provides a tool to determine, for the first time, the lengths of individual D-loops and their distribution in a population of D-loops formed in vitro with near base-pair resolution.

2. DMA allows mapping the exact position of each D-loop in reference to the donor dsDNA sequence. Distribution of the D-loop position can reveal any biases in the position of a D-loop within a region of homology. For instance, it revealed that D-loops are more frequently formed at the 3′-end of the homology.

3. Relative D-loop levels detected by DMA are comparable across the different D-loop samples.

4. The in vitro DMA is backed by several controls that allow testing the specificity of the assay within each sample and also within each read molecule.

5. There is no bias in the length of the D-loop mapped by the DMA above the 120–200 nt threshold limit. The bias is avoided due to the lack of a D-loop enrichment step or a D-loop crosslinking step. Hence, short and long D-loops are mapped with similar frequencies as evident from the vast and equal distribution of D-loop lengths on a linear donor.

6. DMA is a highly sensitive method to detect and map D-loops. DMA can map D-loops formed with efficiency as low as ~1–2% using Sequel-I or as low as ~0.05% with the Sequel-II system (higher coverage).

7. DMA allows multiplexing of samples with up to 50–200 barcoded samples analyzed simultaneously. This provides speedy and cost-efficient high-throughput analysis.

The D-loop mapping assay is limited in the following ways:

1. The minimum detectable D-loop length is limited by the peak threshold used for the analysis of DMA. The peak threshold is used to ensure confidence in that a conversion patch is derived from a genuine D-loop and not spontaneous breathing of the duplex DNA. D-loops shorter than 120–200 nt are undetectable with a threshold window of 50.

2. There is lower confidence in the precise length of those D-loops whose length approaches or is lower than the threshold window. This is because the length may be slightly overestimated due to the requirement of a minimum threshold window. However, D-loop with a length longer than the threshold window is not affected.

3. As with any D-loop assay, some unstable D-loops may be lost during the bisulfite treatment. Some D-loop molecules may also be lost due to the nicking of ssDNA during bisulfite treatment. Despite this, due to the lack of experimental bias, D-loops can be compared across different samples.

4. Some long D-loops may be misidentified as short D-loops due to insufficient bisulfite conversion within the treatment time period. The insufficient bisulfite conversion might be due suboptimal room temperature treatment condition or due to incomplete RPA removal by proteinase K. However, optimizing the conversion threshold by using a spiked ssDNA control helps minimize the possibility of underestimated D-loop lengths or false negatives.

5. Some short D-loops may be misrepresented as long footprints due to any potential D-loop migration during the course of bisulfite treatment. However, D-loops are presumed to be relatively stable at room temperature, as no D-loop dissolution is observed over time (Wright and Heyer, 2014). Moreover, most of the footprints representing supercoiled D-loops had the expected size, minimizing the possibility of branch migration, especially on supercoiled donors.

Lastly, we suggest some potential technical alterations to the DMA that may improve the conversion efficiency or the D-loop detection levels.

1. The proteinase K digestion and removal of RPA from the displaced strand may be less efficient without the presence of a detergent. To minimize the possibility of RPA bound on the displaced strand during bisulfite conversion, we suggest testing the use of CTAB detergent along with proteinase K for any potential improvement in the D-loop footprint signal. CTAB, unlike SDS, is known to prevent D-loop migration (Allers and Lichten, 2000).

2. Currently 35 cycles are performed during each round of PCR. The high cycle number may be promoting the 2-fold bias currently observed among the top and bottom-strand reads. Reducing the cycle number during PCR-1 may help reduce the strand bias. However, these are minor changes and may not significantly change the overall footprint map or relative comparison across different samples.

ds98-607 and ds98-197-78ss were created as described in Wright and Heyer, 2014. Additionally, ds98-931, ds98-915, and ds98-915-78ss substrates were created similarly. The pBSbase phagemid vector was cloned to comprise either the 931 nt or the 915 nt homologies. The 931 or 915 nt homology inserts were amplified from pBSphiX1200 by using pBS-51-931nt-subs-F and pBS-1013nt-subs-R or pBS-1013nt-subs-F and pBS-915nt-subs-R primers respectively. The substrates were then spliced out from the cloned phagemids as previously described (Wright and Heyer, 2014).

D-loop reactions were performed as described in Wright and Heyer, 2014 with the following modifications. The reactions used a linearized or supercoiled pBSphix1200 plasmid donors. All D-loop reactions were carried out at 30°C. Homologous ssDNA was added at 3 nM with the 3 kb donor dsDNA also present at 3 nM (molecules). Rad51 was saturating with respect to the invading ssDNA (1 Rad51 to 3 nts ssDNA), and RPA was added at one heterotrimer to 25 nt ssDNA, while Rad54 was added at 18 nM monomers. 1x buffer was used as described in Wright and Heyer, 2014. The order of addition was: Rad51 + ssDNA, 10 min incubation; then RPA, 10 min incubation; and finally, Rad54 + linear dsDNA, 15 min incubation. In case a supercoiled dsDNA was used instead of a linear dsDNA donor, the reaction was carried on for 10 min to achieve maximum D-loop formation and prevent D-loop disruption based on Wright and Heyer, 2014. The reactions had a final volume of 25 µl. Reactions were stopped with 2 mg/ml Proteinase K (2.5 μl of 20 mg/ml Proteinase K) and 10 mM EDTA (0.5 μl of 0.5 M EDTA). The reaction was then split, such that 9 μl of the reaction was added to a tube containing 0.2% SDS (0.2 μl of 10% SDS) and 1x DNA loading dye for gel visualization (2.8 μl of 6x dye). This fraction of the reaction was deproteinized by incubating at room temperature (RT) for 1–2 hr. Subsequently, the samples were separated on a 0.8% TBE agarose gel at 70 V for ~3 hr. The gel was stained with SYBR Gold Stain for 30 min at RT, before visualization. The remainder of the D-loop reaction (19 μl) was incubated at room temperature (RT) for 30 min to allow deproteinization, before proceeding to bisulfite treatment for the D-loop Mapping Assay (DMA). Note that SDS was avoided in this fraction of the D-loop sample to prevent branch migration of D-loops (Allers and Lichten, 2000). The reaction volume ensures that >50 ng of dsDNA was incorporated into the bisulfite reaction.

The Gargamel pipeline (available at https://github.com/srhartono/footLoop_sh) allows users to map reads, assign strands, call single-molecule D-loop footprints as peaks of C-to-Tconversion, perform clustering on peaks, and visualize the data, similar to the pipeline described in Malig et al., 2020.

We spiked in ssDNA (pBlueScript KS- circular DNA) in an in vitro D-loop sample to test bisulfite conversion efficiency and to optimize the peak calling threshold. pBSKS- ssDNA was prepared as described in Wright and Heyer, 2014. The D-loop sample formed from ds98-931 and a linear donor was spiked with 0.03 nM (1/100th of ds98-931/dsDNA) pBSKS- ssDNA molecules. The ssDNA was spiked along with the ds98-931 substrate in the D-loop reaction. UNI+pBSKS-F and UNI+pBSKS-R primers (devoid of cytosines/guanines) were used to amplify the spiked ssDNA during the PCR-1 step of DMA.

The computational pipeline for mapping reads, strand assignment, peak calling, clustering, data visualization, and D-loop analysis is available from GitHub (https://github.com/srhartono/footLoop_sh).

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank members of the Heyer laboratory, especially Aurele Piazza and William Wright, for stimulating discussions. We thank Diedre Reitz for helpful feedback on the manuscript. We also thank members of the Chedin laboratory, in particular, Lionel Sanz, for providing AMPure beads and Maika Malig for technical help. We also thank the DNA core technologies at UC Davis and the UC Berkeley Genomics Facility for providing PacBio sequencing services. This research used core services supported by P30 CA93373 and was supported by NIH grants GM58015 and CA92276 to W.-D.H. and NIH grant GM120607 to FC.

© 2020, Shah et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.