Single-molecule force spectroscopy reveals binding and bridging dynamics of PARP1 and PARP2 at DNA double-strand breaks

To study the mechanics of binding at a double-strand DNA break, we developed a method for making a double-stranded DNA molecule featuring a 20 kDa PEG molecule bridging across a central section (Fig. 1A). The construct was synthesized in a series of steps beginning with a copper-free click reaction between an oligonucleotide containing an internal dibenzylcyclooctyne and a homobifunctional azide–terminated PEG. This reaction creates a stable, covalent linkage between the PEG and DNA. Annealing with complementary oligonucleotides followed by a ligation reaction with PCR amplicons formed the final structure (SI Appendix, Figs. S1–S3). The DNA molecule is 1.8 kbp in length and is flanked by sections labeled with multiple digoxigenins and biotins for attachment to a coverslip surface and superparamagnetic bead, respectively. Light scattering measurements of PEG (27) show that the mean end-to-end distance (M_e-e) obeys a power law dependence on molecular weight (M_w) of M_e-e = 0.047*(M_w)^0.588, giving a value of M_e-e = 16 nm for M_w = 20 kDa (SI Appendix, Fig. S4). We designed the two labeled thymine bases on our synthetic DNA oligonucleotide to have a separation of 46 bp = 16 nm (given 0.34 nm/bp) in order to match the PEG mean end-to-end length. Therefore, upon formation of a DSB in between the two PEG attachment points, the PEG will hold together the two DSB ends in close proximity under conditions of low or zero force. To enable the creation of such a break, we positioned a recognition site for the restriction enzyme NruI which cuts both strands at the same position, leaving a blunt end. Fig. 1B shows a schematic of the DNA before and after restriction enzyme digestion showing how the PEG linker holds together the broken DNA ends. Previous single-molecule studies investigating the behavior of DNA ends under force have used a DNA-based linker, for instance, when analyzing base-stacking interactions or nonhomologous end joining kinetics (28–30). Our synthesis strategy using PEG negates any potential confounding effects due to proteins binding to a DNA linker while allowing controlled manipulation of the DNA break under force.

To test the correct assembly of the DNA construct, we tracked DNA extension as a function of time before and after adding NruI (Fig. 1C). A constant force of 8 pN was applied throughout. Addition of NruI results in a step change in DNA extension of 64 nm in this example trace. Fig. 1D shows a histogram of measurements from 68 magnetic beads, revealing a mean extension change of 69 ± 12 nm (mean ± SD). Our observed extension is consistent with a 20 kDa PEG chain (contour length = 159 nm), giving an expected extension of 56 nm at 8 pN based on a worm-like chain model (SI Appendix, Fig. S5). The DNA cleavage by NruI could be observed on multiple beads across the field of view—the time to cleavage varied from bead to bead due to the stochastic arrival of the enzyme to its recognition site (SI Appendix, Fig. S6). Fig. 1E shows a comparison of the force–extension curve for one bead before and after cutting by NruI. PEG has a much lower persistence length (~0.4 nm) compared to double-stranded DNA (~50 nm) and therefore it requires higher forces to stretch it to a given percentage of its contour length. This is reflected by the shallower gradient of the force–extension curve at forces <15 pN for the curve after NruI cleavage. An overstretching transition is observed after cutting, whereas no such transition is measured before cutting. Previous single-molecule force spectroscopy experiments have shown that torsionally unconstrained DNA forms an overstretched structure (a so-called B–S transition) at approximately 65 pN, whereas torsionally constrained DNA shows this transition at approximately 115 pN (31). Our measured force–extension curves are consistent with this since before cleavage, the DNA is torsionally constrained by the multiple digoxigenin and biotin bonds used for tethering, whereas after cutting, the DNA loses torsional continuity since the single C–C and C–O bonds that make up the PEG chain are able to rotate freely.

Having established a method to create two opposing double-stranded DNA breaks tethered by a PEG linker, we applied our system to study the mechanical bridging of a DNA DSB by PARP2. Fig. 2A shows the crystal structure (PDB: 7AEO) of PARP2 in complex with DNA (19). A 2:2 stoichiometry of blunt-end, 5′ phosphorylated DNA duplexes with PARP2 was observed in this structure and also in separate literature reports which employed cryo-electron microscopy (18, 20). Each PARP2 protein forms its own bridge via its WGR domain, and there are no significant contacts between the two PARP2 proteins. Previous work suggests that PARP2 activation is highly dependent on the presence of a 5′ phosphate on both SSBs and DSBs (6, 7, 18). After NruI cleavage, the DNA is left with dual 5′ phosphorylated blunt ends. To test how the presence of 5′ phosphate affected binding, we also established methods to prepare the magnetic tweezers construct with either a single or no 5′ phosphate (Fig. 2B) by treating with Antarctic phosphatase and T4 polynucleotide kinase (SI Appendix, Figs. S7 and S8).

Fig. 2 C and D shows a schematic and example time-series data trace of an experiment to measure the rupture force of PARP2 bridging across a DSB. After addition of PARP2 to the flow-cell, the force was initially held close to 0 pN so that the two DNA ends are held in close proximity by the PEG linker, thereby allowing the PARP2 to bind across the DNA DSB interface. The force was then linearly ramped at 2 pN/s. A sudden increase in extension is observed when the PARP2 bridge ruptures as marked by the orange arrows. We found that this force-induced rupture could be repeatedly observed by returning the force to close to 0 pN—allowing the PARP2 bridge to reassemble—before ramping the force again as shown by the five consecutive force ramps displayed in Fig. 2D. Fig. 2E shows comparisons of the force–extension curves at different experimental stages, namely before NruI cleavage, after NruI cleavage, and after addition of 200 nM PARP2. In the presence of 200 nM PARP2, the curve tracks along the fully ligated DNA case before rupture at which point the curve then follows along the broken DNA case. The first force ramp shows a rupture event before the 65 pN overstretching where we observe an increase in the extension of 151 nm corresponding to the 20 kDa PEG contour length. Remarkably, as illustrated by the force–extension curve of the fifth force ramp, we observe that most ramps show rupture at a force greater than the overstretching transition of 65 pN. This indicates that PARP2 bridging restores the torsional continuity of DNA so that the DNA can support forces greater than 65 pN before the rupture of the PARP2 bridge is immediately followed by overstretching of the DNA.

As illustrated in Fig. 2D, we found that after force-induced rupture, the PARP2 bridge could be reformed by reducing the force close to zero so that the PEG linker then brought the two ends in proximity again. This enabled us to perform repeated force ramps to build up statistics on 1) how many cycles showed a characteristic rupture event and 2) the force at rupture. In Fig. 2F, we quantify the percentage of force ramps (calculated from the first 15 ramps after protein addition) that show a rupture event in the presence of 200 nM PARP2. For dual 5′ phosphate (2 × 5′P), we observe that 93 ± 4% (mean ± SEM, N = 6 beads) of force ramps show bridging. In the case of single 5′ phosphate, we observed that 72 ± 13% (mean ± SEM, N = 6 beads) of curves showed a characteristic rupture event. In the case of no 5′ phosphates on the DNA ends, we observed a complete absence of bridging. In Fig. 2G, we show histograms that quantify the rupture force distribution for the dual 5′ phosphate (at 200 nM and 20 nM PARP2 concentration) and single 5′ phosphate constructs (at 200 nM). At 200 nM PARP2 concentration, for the dual 5′ phosphate, we observe a single peak at approximately 85 pN. We fitted these data to the Bell–Evans model which models the rupture of single bonds under force as a lowering of the activation energy barrier by an amount Fδ where F is the force and δ is the distance between the bound state and activation energy peak (32, 33). At constant loading rate, the rupture force probability distribution is given by:

where t_on is the bound lifetime in the absence of force, R is the force ramp rate (in our experiments, 2 pN.s⁻¹), and k_BT is thermal energy. This equation was fitted to the measured rupture force distribution using the Levenberg–Marquardt nonlinear least squares algorithm and yielded fit parameters of t_on = 1.3 × 10⁵ s (i.e., ~1.5 d) and δ = 0.51 nm (see dashed line fit, Fig. 2G).

At 20 nM PARP2 concentration, we observe a similar peak at ~85 pN together with a tail of lower values and a small peak at ~45 pN. Efforts to measure the rupture force statistics at lower concentrations were hampered by the low probability of observing a rupture event. The crystal structure shown in Fig. 2A reveals no significant interactions between the two PARP2 proteins and so the rupture force for a single PARP2 would be expected to occur at roughly half the value for the dual PARP2 configuration. We therefore interpret the main peak at ~85 pN as due to the formation of the 2:2 stoichiometry of PARP2:blunt ends and the peak at ~45 pN as due to the formation of a single PARP2 bridging from the 5′ phosphate at one end to the opposing 3′ hydroxyl. To further investigate the stoichiometry of PARP2 binding, we investigated the rupture force histogram for a single 5′ phosphate (Fig. 2G). Interestingly, we observed a bimodal distribution with two peaks at ~10 pN and ~40 pN. We ascribe the ~40 pN peak to a single PARP2 bridging from a 5′ phosphate to 3′ hydroxyl as observed at 20 nM for 2x5′P. The peak at ~10 pN indicates that there is a second bridging conformation. This conformation has approximately equal likelihood as the 5′ phosphate to 3′ hydroxyl peak at ~40pN given the observed histogram distribution. Since the 0 × 5′P ends showed a complete absence of bridging, we hypothesize that the lower stability state may be due to binding of PARP2 to the single 5′ phosphate but then bridging across to the 5′ hydroxyl rather than the 3′ hydroxyl on the opposing blunt end.

Having characterized the force-induced rupture of PARP2 bridging across DNA, we next studied the influence of bridging on DNA supercoiling. Magnetic tweezers enables measurement of DNA supercoiling by rotating the magnets, which applies a torque due to preferential alignment of the superparamagnetic bead along its easy axis (34, 35). Fig. 3A shows a schematic and example traces of the experimental design we made to test whether PARP2 bridging restores DNA supercoiling. Initially, we measured the fully ligated construct which is torsionally constrained and therefore shows a characteristic decrease in extension when the magnets are rotated and plectonemes form (Fig. 3B). After addition of NruI to the flow-cell, we wait until the NruI has created a break at its recognition sequence such that there is no longer a change in extension upon magnet rotation. At the force of 0.4 pN, used throughout, there is no measurable difference in extension after NruI cleavage since this low force is not sufficient to extend the PEG to a measurable degree (SI Appendix, Fig. S5). Finally, we add in a test protein, incubate for 1 min, and repeat the magnet rotation. Fig. 3B shows an example where we have added PARP2 into the flow-cell and observe that supercoiling is restored, indicating that PARP2 has formed a bridge across the DNA break.

To initially test the system, we performed positive and negative controls using T4 DNA ligase. A concentration of 10 U/μL was used (a typical concentration for bulk DNA ligations in plasmid assembly) and incubated for 1 min in the flow-cell before testing the percentage of beads in the field of view, which show a restoration of supercoiling. Ligase without ATP showed no restoration of supercoiling as expected (Fig. 3C). For ligase with 1 mM ATP, we measured that 97 ± 3% (mean ± SEM, N = 5) of beads in the field of view showed a restoration in supercoiling. This control confirms that, at the low force of 0.4 pN used, the broken DNA ends are held in close proximity by the PEG linker, enabling rapid religation. We then repeated the experiment testing the addition of 200 nM PARP2. We observed that 89 ± 6% (mean ± SEM, N = 3) of beads in the field of view showed a restoration of DNA supercoiling when PARP2 was added, thereby showing that PARP2 bridging supports the formation of torsionally constrained DNA molecules.

Having tested the binding of PARP2 to blunt DNA double-strand ends, we investigated the influence of different overhang structures. Our DNA construct features two restriction sites for the enzyme DrdI at equidistant positions from the PEG linkages (Fig. 4A). Similar to NruI (described above), cleavage of the DNA could be observed in real time as a sudden increase in DNA extension after addition of DrdI to the flow-cell (SI Appendix, Fig. S9). After cleavage, we used the overhangs formed by DrdI to ligate on different synthetic duplex structures in situ in the magnetic tweezers’ flow-cell. Specifically, we made structures with 5′ and 3′ overhangs of four base lengths and with the same 5′-CGCG-3′ sequence (Fig. 4B). We then used our magnetic tweezers assay to perform force ramps and measure the rupture force as described in Fig. 2. Fig. 4C shows histograms of the rupture force for both 3′ and 5′ overhangs (in the absence of PARP2). Both overhang types produce readily detectable rupture events, due to complementary base pairing, with forces between 5 and 20 pN. Fig. 4D shows an example trace comparing force ramps before and after adding 200 nM PARP2 to the 3′-4 bp design. The arrows mark the increase in extension corresponding to when the bridging of the ends ruptures and the PEG linker is extended.

In Fig. 4E, we compare the percentage of force ramp curves showing a rupture event for both 3′ and 5′ overhangs (at 0 nM and 200 nM PARP2). For the 3′ and 5′ overhangs with 0 nM PARP2, we observe that 90 ± 2% and 87 ± 4% (mean ± SEM, N = 4 and N = 3 beads, respectively) of force ramps show a rupture event (due to base-pair hybridization). Similarly, after addition of 200 nM PARP2 to the 3′ overhang structure, we observed a high percentage of bridging of 98 ± 2% (mean ± SEM, N = 4) and with a significantly enhanced rupture force (Fig. 4 D and F). However, after addition of PARP2 to the 5′ overhang DNA construct, rather than an enhancement in bridging rupture force, we observed that bridging was almost completely blocked at 6 ± 3%, (mean ± SEM, N = 9). This indicates that PARP2 no longer bridges across the broken DNA ends but instead binds in a manner that blocks complementary base-pairing between the two ends (see SI Appendix, Fig. S10 for example force ramp trace).

In Fig. 4F, we show the rupture force histograms after addition of 20 nM or 200 nM PARP2 to the four base 3′ overhang constructs. We observe a clear increase in rupture forces compared to 0 nM PARP2 (Fig. 4C), indicating the formation of a bridging interaction by PARP2. Using overhangs with 5′ hydroxyl groups rather than 5′ phosphate completely negated this effect, again showing that PARP2 bridging is specific to the presence of 5′ phosphate (SI Appendix, Fig. S11). The rupture force histograms in Fig. 4F show bimodal distributions, and the shape of the distribution does not change significantly when increasing the concentration from 20 nM to 200 nM. This indicates that the two peaks correspond to distinct states of bound PARP2 (rather than having one or two bound PARP2 as was observed for 2 × 5′P blunt ends in Fig. 2G). Taking the crystal structure of PDB:7aeo (shown in Fig. 2A) and simulating an overhang in the 3′ direction indicates significant steric clashes if two PARP2 proteins are bound across each nick on the DNA (SI Appendix, Fig. S12). This suggests that it is unlikely that two PARP2 proteins can bind to hybridized DNA overhangs—potentially explaining why the rupture forces are lower than those observed for blunt-ended DNA. However, the exact conformational details of the two states remain to be determined.

After characterizing the binding and bridging of PARP2 across different types of DNA DSBs, we investigated the behavior of PARP1. We repeated the tests described in Figs. 2F and 4E where we perform multiple force ramps after addition of protein and then determined the percentage of these force ramps which show bridging of the two ends. Fig. 5A shows the result of these experiments for three types of overhangs—blunt ends (0 bp), 3′-4 bp, and 5′-4 bp (both phosphorylated and with same sequences as Fig. 4). For the blunt ends, we did not observe PARP1 bridging across the two ends in the presence of 200 nM PARP1. Furthermore, PARP1 blocked bridging due to the presence of complementary base-pair overhangs for both 3′-4 bp and 5′-4 bp—with <10% of force ramps showing a rupture event—similar to the activity of PARP2 on the 5′-4 bp overhang (Fig. 4E).

We also performed competition experiments between PARP2 and PARP1. Fig. 5B shows an example trace where we compare behavior under force with 200 nM PARP2 only before adding 200 nM PARP1 + 200 nM PARP2. With 200 nM PARP2 only, we observe the characteristic high force rupture events of PARP2 bridging as indicated by the arrows (Fig. 2D). We then held the magnetic bead at 9pN while adding solution such that the ends are held apart during this time, preventing the formation of a PARP2 bridge. Having added the new solution of 200 nM PARP1 + 200 nM PARP2, we perform additional force ramps. These force ramps show an absence of end-bridging—the only sudden increase in extension during the ramp is due to the 65 pN B–S transition. This indicates that PARP1 rapidly competes away PARP2 when the ends are held in this open conformation. We similarly observed this effect when using calf-intestinal phosphatase rather than PARP1 as a competitor (SI Appendix, Fig. S13)—where the lack of bridging now signifies that phosphatase can rapidly access the 5′ phosphate which is crucial for PARP2 bridging.

In a second approach, we repeated the competition assay with PARP1 but this time held the ends in close proximity before solution exchange such that the PARP2 bridge could reform (Fig. 5C). Here, we observed that the first force ramp after PARP1 addition showed the characteristic high force rupture of a PARP2 bridge before subsequent ramps had an absence of end-bridging. This was measured even after incubating PARP1 for 10 min before applying subsequent ramps [an extension decrease is observed during the 10 min incubation due to DNA condensation by PARP1 (36)]. Overall, this experiment indicates that when a PARP2 bridge has already formed, it is not readily displaced by PARP1.

In Fig. 5 D and E, we present models for PARP1 and PARP2 at blunt 5′ phosphate DSB ends. If the ends are apart (e.g., being held apart under force), then PARP2 exchanges rapidly at the DSB as evidenced by its fast displacement by PARP1 and the activity of calf-intestinal phosphatase. However, if the two DSB ends are in close proximity, such that the ends are able to interact, PARP2 can form a highly stable bridging interaction which resists mechanical forces up to 90 pN (Fig. 2G). PARP1, in contrast, binds stably at DSB ends but does not form a bridging interaction when the two ends are brought together.

Human PARP1 and PARP2 proteins were made using a baculovirus expression system and purified by Immobilized Metal Affinity Chromatography (IMAC) affinity and size-exclusion chromatography (see SI Appendix, Fig. S14 for further details). The PEG–DNA construct synthesis is described in detail in SI Appendix, Figs. S1–S3. Briefly, a 20 kDa bis-azide PEG was reacted overnight with a Dibenzocyclooctyne (DBCO) modified oligonucleotide before agarose gel purification. The full-length DNA construct was then generated in a golden-gate assembly reaction (44) with PCR amplicons before a final agarose gel purification.

Flow-cell fabrication is detailed in depth in a previous publication (45). Briefly, the flow-cell is assembled using silicone tape, glass coverslips, and polydimethylsiloxane. Solutions are flowed through using a gravity-feed method and flow is automatically arrested by a capillary-action stop valve. The bottom coverslip is coated sequentially with nitrocellulose and antidigoxigenin (11333089001, Sigma Aldrich) before cross-linking with glutaraldehyde and passivating with BSA (A0281, Sigma-Aldrich) and β-casein (C6905, Sigma-Aldrich). Three micrometers-sized latex beads (LB30, Sigma-Aldrich) are fixed to the surface, which act as fiducial markers for drift correction.

Superparamagnetic beads (M-280 streptavidin Dynabeads, ThermoFisher) were incubated with gel-purified DNA constructs before introducing into the flow-cell. The beads were imaged by collimated LED illumination (ThorLabs, M625L4-C4) with a 40× or 60× Nikon oil-immersion objective and a monochrome camera (Imaging Source, DMK 33GX249). We used a pair of 5 mm × 5 mm × 5 mm N50 cube magnets (C0057, SuperMagnetMan) in one of the following two configurations: 1) with a 2-mm gap, mounted on a rotating motor to enable DNA supercoiling, and 2) with a 0.3-mm gap. The magnets used for each experiment and calibration methods are described in SI Appendix. The magnetic beads were tracked in real time at 20 frames per second using a previously described video image analysis algorithm (46) with custom C++ and LabVIEW (National Instruments) software. Traces displayed in figures were filtered with a five-point median filter. All measurements with PARP proteins were in a buffer of 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 2 mM MgCl₂, 0.5 mM Tris (2-carboxyethyl) phosphine (TCEP), 1 mg/mL Bovine Serum Albumin (BSA), and 1 mg/mL β-casein. NruI and DrdI cleavage was performed in 1× CutSmart buffer New England Biolabs (NEB) supplemented with 1 mg/mL BSA, 1 mg/mL β-casein. Tests of T4 DNA ligase (Fig. 3) were performed in 1× NEB CutSmart (−ATP buffer) or 1× NEB T4 ligase buffer (+ATP buffer) with a concentration of 10 U/μL. All experiments were conducted at room temperature (22 ± 1) °C.

N.A.W.B. and J.E.M. designed research; N.A.W.B. performed research; N.A.W.B. analyzed data; and N.A.W.B. and J.E.M. wrote the paper.

The authors declare no competing interest.