Structural dynamics of DNA strand break sensing by PARP-1 at a single-molecule level

Single-stranded breaks (SSBs) are the most frequent DNA lesions threatening genomic integrity. A highly kinked DNA structure in complex with human PARP-1 domains led to the proposal that SSB sensing in Eukaryotes relies on dynamics of both the broken DNA double helix and PARP-1’s multi-domain organization. Here, we directly probe this process at the single-molecule level. Quantitative smFRET and structural ensemble calculations reveal how PARP-1’s N-terminal zinc fingers convert DNA SSBs from a largely unperturbed conformation, via an intermediate state into the highly kinked DNA conformation. Our data suggest an induced fit mechanism via a multi-domain assembly cascade that drives SSB sensing and stimulates an interplay with the scaffold protein XRCC1 orchestrating subsequent DNA repair events. Interestingly, a clinically used PARP-1 inhibitor Niraparib shifts the equilibrium towards the unkinked DNA conformation, whereas the inhibitor EB47 stabilizes the kinked state.


Gel electrophoresis
The ligation efficiency was assessed via gel electrophoresis. Phosphorylated DNA, ligated DNA and the control lacking the ligase were loaded on a 15% PAA 1xTBE 7M urea gel (16x20cm). The gel was prerun for 1.5hours at 300 V and run for 8.5hours. The gel was imaged on a ChemiDoc MP imaging system (BioRad) Electromobility shift assays.
Band shift assays were carried out using 6% polyacrylamide gel as described previously 1 with minor modifications. 0.5xTB buffer with 5% glycerol was used for gel preparation and as running buffer. Gels were prerun at 55V for 20 min at 4C. 50nm of DNATamra ligand (methods) was mixed with protein in binding buffer (20mM Tris-HCl, pH 7.5, 100 mM NaCl, 3mM MgCl2, 150 µM ZnSO4, 4mM DTT, 10% glycerol) in a total volume of 10µL. Prior to gel electrophoresis, samples were incubated for 30 min at room temperature. Gel electrophoresis was conducted for 50 min at 55V at 4°C. Fluorescent DNA signal was detected using Typhoon FLA 9500 instrument (GE Healthcare).

Determination of the isotropic Förster radius
In order to quantify the isotropic Förster radius, several parameters had to be measured for the dyes in (methods) and from smFRET measurement of DNATamra. For the radiative rate constant of Atto550 the value given by the manufacturer was used. As for 6-Tamra, the radiative rate constant was quantified by measuring the lifetime and quantum yield of unconjugated 6-Tamra. To calculate the spectral overlap J for DNAAtto550 the emission of the donor strand was recorded. For the alternative DNA construct, DNATamra, to measure the emission spectrum of the donor in absence of FRET, a reference sample (double stranded DNA, internally labeled with 6-Tamra and diluted in the buffer used for smFRET measurements) had to be used. Both donor emission spectra were recorded on a SPEX Fluorolog II (Horiba) (spectral bandwidth of 4.25nm for the excitation and 2.13nm for the emission) under magic angle conditions. For measuring the absorption spectrum of the acceptor, a DNA strand comprising only the acceptor labeled 3' stem of the DNAAT550 (see section on synthesis of the DNA ligand) was diluted in the buffer also used for smFRET measurements. The spectrum was taken on a Cary 50 Bio spectrophotometer (Varian). Both spectra are needed to determine the overlap integral. Additionally, the absorption coefficient of Alexa647 at the absorption maximum (651nm determined from the measured spectrum) is needed and given as 270000 M -1 cm -1 by the manufacturer (ThermoFisher Scientific). A value of 1.35 was assumed for the refractive index of the medium between the two dyes. PhotochemCAD 3 2 was used to calculate the isotropic Förster distance Riso=70Å for DNAAtto Riso=68Å for DNATamra 3 from the determined parameters.
The uncertainty for the isotropic Förster radius ΔRiso was determined according to equation 1 by error propagation of errors in the refractive index n, the donor quantum yield φ and the overlap integral J.
Equations for ΔRiso(n), ΔRiso(n) and ΔRiso(n) are given in (Hellenkamp et al. 2018). Due to the position of our labels on the DNA, we expect that the medium between them will be composed almost exclusively of buffer. We thus assume a refractive index of n=1.35±0.02, which leads to ΔRiso(n)=0.01*Riso. We further assumed an error of 4% for the donor quantum yield based on the uncertainties in the donor lifetime and radiative rate constant, leading to ΔRiso(φ)=0.01*Riso. The value of ΔRiso(J)=0.025%*Riso was adopted from (Hellenkamp et al. 2018). These uncertainties lead to ΔRiso=0.03*Riso=2Å for the construct DNATamra.
Data for simulations: lifetime and time-resolved anisotropy Data for the donor in absence of the acceptor was obtained from the donor-only population of the smFRET measurements. All donor photons after donor excitation (531nm) of the donor-only population (uncorrected stoichiometry >0.9 and more than 100 photons) were combined in one TCSPC histogram.
Data for the acceptor can be gained either from the acceptor-only population (uncorrected stoichiometry <0.25 and more than 70 photons) or from the FRET population (as defined above), because the acceptor photons after acceptor excitation (640nm) are not affected by the FRET process. Results for both species were generally in good agreement, so a mean value was taken. were used to record the instrument response function (IRF) of the green and red channels, respectively. (1) In a first step, the fluorescence lifetime was determined from the combined TCSPC histogram D(t), which adds the data from parallel (Dǁ) and perpendicular (D⊥) detection channels according to with G, l1 and l2 being the correction factors described above. For Atto550 a mono-exponential and a biexponential model function were needed to describe the acceptor Alexa647. Iterative re-convolution of the model decay with the recorded IRF was used to fit the TCSPC histogram D(t), additionally considering a constant background contribution. For the simulations, the amplitude weighted average lifetime of the two donor components was used. The value for the donor lifetime given in Supp. Table S5 represents the average of the fit results from DNA, DNA+F2 and DNA+F1F2, because they were in agreement.
In the second step, the anisotropy information was analyzed in a global fit of the TCSPC data from the parallel (Dǁ) and perpendicular (D⊥) detection channels by iterative re-convolution. The model functions with I(t) being the biexponential fluorescence intensity decay and r(t) being the model function for the anisotropy decay, given by one of the following functions, depending on the complexity of the data. data, the anisotropy model with two lifetimes and two rotational correlation times was used.
As in the case of the donor lifetime, the values given in Supp. In the presence of protein, the situation becomes more complicated than our simulation can describe, with two acceptor states due to PIFE 5 . Even so, on a long timescale, the anisotropy is expected to decay to the lower value of the unrestricted state, which is likely to be dominant in the absence of protein.

Static and dynamic FRET lines
In order to test for conformational dynamics, static and dynamic FRET lines were simulated 6 Figure 7), is the fraction of the blinking and is the offset (Methods). An increased is observed for DNA bound by protein datasets. The error bars show the standard error of the mean the computed molecules contributing to the same bin.
Supplementary Figure 11: Donor anisotropy decays for DNA (blue) and DNA with Niraparib (turquise) The anisotropy decay (Supplementary methods) for DNA in presence of niraparib indicates a strong hindrance of the donor dye attached DNA (residual anisotropy of DNA with niraparib is higher than that of DNA in absence of niraparib), whereas the donor dye without niraparib rotates freely ( approaches 0). This hindrance of the donor dye results in slightly lower FRET efficiencies for DNA with PARP-1 and niraparib data (Fig. 6 B, turquoise, Supplementary Figure 2, Supp. Table S1).  Supplementary Figure 9), talazoparib and rucaparib. While niraparib hinders the rotation of the donor dye (residual anisotropy of DNA with niraparib is substantially higher than that of DNA in absence of niraparib or in presence of rucaparib), rucaparib does not seem to affect the rotational motion of the donor dye and similiar time-resolved fluorescence anisotropy as those PARPi that do not affect the FRET efficiency histogram in a (here: data for talazoparib are shown for comparison). The data for rucaparib can be interpreted as an effect on the extinction coefficient of the donor dye. This change in extinction coeffcient of the donor dye results in slightly lower FRET efficiencies for DNA with rucaparib data (a).  Table S1). The error bars show the standard error of the mean for FRET efficiency of the single molecules contributing to the same bin.  Sawaya et al. 1997) were superimposed by using their respective 3' DNA stem, and are shown here in two different orientations. Panels d and e show a superimposition of the PARP-1 structure with the Fen1 and DNA polymerase  structures, respectively. Substantial steric clashes suggest mutually exclusive binding modes. XRCC1 has been proposed to bind damaged DNA on the opposite side of DNA to that bound by polymerase  16 . Such binding might also be compatible with simultaneous binding by PARP-1 during the process of DNA damage recognition and repair.   Table S6. In c, at 60mM salt, full length PARP-1 binds nicked DNA unspecifically resulting in a broad FRET efficiency distribution (blue). In summary, optimal salt concentration for F2 and F1F2 is 60mM, whereas full length PARP-1 shows specific binding at 200mM salt concentration. The error bars show the standard error of the mean the computed molecules contributing to the same bin.
Supplementary Figure 21: Top view on a spherical cone (green) and an open spherical sector (red). Both are directed along the same axis here, indicated by the black crossed circle (center). This axis is also the mean axis of the cone, which does not need to coincide with the 3'-stem. However, we assume that the open spherical sector is directed along the 3'-axis. Its degenerate mean orientation is here shown by the black ring. The standard deviations along the spherical cone or the open spherical sector are indicated by a black double arrows.
Supplementary Figure 22: Exemplary lifetime and anisotropy fits and fit results for DNAAtto in presence of F1F2. Lifetime and anisotropy analysis (Supplementary methods) for DNA, DNA+F2 and DNA+F1F2 was performed to obtain the parameters (Supp . Table S4) necessary for the structural analysis (Methods). a: Fitted (red) donor fluorescence decay histogram (black), instrument response function (grey); the weighted residuals (wres) are shown in the top panel of the fluorescence decay. The fit returned a value of = 3.52 . b: Fitted donor time-resolved polarized fluorescence decays (red is the parallel channel, blue is the perpendicular channel) and the dotted decays correspond to respective IRFs; the lower panel shows the anisotropy decay curve (black) and its fit function (red); the weighted residuals (wres) for both channels are shown in the top panel of the fluorescence decay. The parameters returned from the fit are = 0.27 , = 0.04 , = 1.31, for which the lifetime was fixed to the values given in a.   Fig.7 and Supp. Fig. 13: N is the average number of molecules in observation volume, is the diffusion time, is a time scale characterizing acceptor blinking, is the fraction of the blinking and is the offset (Methods). An increased is observed for DNA bound by protein datasets.