Divalent and multivalent cations control liquid-like assembly of poly(ADP-ribosyl)ated PARP1 into multimolecular associates in vitro

Abstract

The formation of nuclear biomolecular condensates is often associated with local accumulation of proteins at a site of DNA damage. The key role in the formation of DNA repair foci belongs to PARP1, which is a sensor of DNA damage and catalyzes the synthesis of poly(ADP-ribose) attracting repair factors. We show here that biogenic cations such as Mg²⁺, Ca²⁺, Mn²⁺, spermidine³⁺, or spermine⁴⁺ can induce liquid-like assembly of poly(ADP-ribosyl)ated [PARylated] PARP1 into multimolecular associates (hereafter: self-assembly). The self-assembly of PARylated PARP1 affects the level of its automodification and hydrolysis of poly(ADP-ribose) by poly(ADP-ribose) glycohydrolase (PARG). Furthermore, association of PARylated PARP1 with repair proteins strongly stimulates strand displacement DNA synthesis by DNA polymerase β (Pol β) but has no noticeable effect on DNA ligase III activity. Thus, liquid-like self-assembly of PARylated PARP1 may play a critical part in the regulation of i) its own activity, ii) PARG-dependent hydrolysis of poly(ADP-ribose), and iii) Pol β–mediated DNA synthesis. The latter can be considered an additional factor influencing the choice between long-patch and short-patch DNA synthesis during repair.

Formation of biomolecular condensate via liquid-like assembly of PARylated PARP1 is driven by biogenic cations. This assembly regulates autoPARylation of PARP1, PARG-dependent hydrolysis of poly(ADP-ribose) and Polβ-mediated DNA synthesis.

Introduction

Phase separation of biomolecules is regarded as a basic mechanism behind the formation of membrane-less compartments or so-called biomolecular condensates in live cells¹. Although the nucleus does not contain membrane-bound organelles, many subcompartments in the nucleus seem to be assembled through liquid–liquid phase separation (LLPS) of protein or protein–nucleic acid mixtures^2,3. In particular, space- and time-specific formation of nuclear condensates is often associated with the assembly of DNA repair foci owing to local accumulation of DNA repair proteins^3,4. In the nucleus, nucleic acids and DNA damage sensor proteins represent key factors that contribute to the creation of these subnuclear compartments in the cell through the phase separation mechanism^3–5. Along with damaged DNA and some types of RNA, the key role in DNA repair foci formation belongs to poly(ADP-ribose) [PAR], which attracts repair factors and plays a pivotal role in the regulation of at least base excision repair (BER) and single-strand break (SSB) repair processes^6–9. In the nucleus, PAR is mainly synthesized by PARP1, which is a member of the ADP-ribosyltransferase diphtheria toxin-like family [ARTDs, also known as poly(ADP-ribose) polymerases (PARPs)]⁹. PARP1 is a sensor of DNA damage and catalyzes the transfer of ADP-ribose units from NAD⁺ onto amino acid residues of target proteins, including itself, resulting in their covalent modification: poly(ADP-ribosyl)ation (PARylation)^9–12. Activation and autoPARylation of PARP1 upon binding to damaged DNA is considered a local (at sites of DNA damage) signal attracting repair proteins through noncovalent interaction with PAR^12–16. It has been demonstrated that components of multiprotein BER and SSB repair machinery such as X-ray repair cross-complementing protein 1 (XRCC1), DNA polymerase β (Pol β), and DNA ligase III (Lig III) interact with PAR and could be recruited to a DNA lesion^{8,11,12,17–20}. A substantial amount of evidence suggests that PAR alone or together with other factors, such as FET family proteins (FUS/EWS/TAF15), plays a direct part in the orchestration of the events helping to assemble DNA repair foci or compartments^{6,8,17,21–26}. These data provide new insights into DNA repair regulation via condensation at least for SSB repair and BER because these processes are strongly dependent on PARP1 activity^7,11,27–30.

One of the possible functions of the formation of PAR-dependent condensates during repair is the regulation of activities of the BER/SSB repair enzymes that could affect the overall process^3–5,31. Although inhibition of PARP1 activity typically impairs the efficiency of BER/SSB repair pathways^7,29–31, there is no direct evidence for the role of PAR-dependent condensates in these repair processes⁵. PARP1’s functional interaction with enzymes and DNA intermediates of BER in vitro has been studied extensively^28,30 but generally not in the context of liquid-like phase separation. Recently, the role of biogenic metals in the regulation of phase separation of proteins and nucleic acids was widely discussed for both in vivo and in vitro systems³². Furthermore, the divalent ions of calcium, magnesium, and manganese and polyamines are involved in the main metabolic and biochemical processes within the cell and are responsible for numerous cellular functions, including the regulation of nuclear enzymes’ activity and maintenance of genome stability^33–40. In our earlier study, we observed the supramolecular association of PARP1 during its PARylation in the presence of Mg²⁺ and assembly of a complex of repair proteins such as DNA polymerase β (Pol β) and XRCC1 with PARylated PARP1 in vitro^41,42. These findings suggest that cations may influence the assembly of PARylated PARP1.

In the present study, we analyzed the self-assembly of PARylated PARP1 and protein-free PAR in the presence of physiologically relevant divalent or multivalent cations that act as cofactors for PARP1; some of them, such as Mg²⁺, Mn²⁺, and Ca²⁺ are required for the activity of BER/SSB repair enzymes^43–45. Using a reconstituted in vitro BER system, we also examined the activity of Pol β and Lig III under conditions of Mg²⁺-dependent liquid-like assembly of PARylated PARP1. Our findings indicate that PARP1’s cofactors such as Mg²⁺, Ca²⁺, and Mn²⁺ and natural polyamines (spermidine³⁺ [Spd³⁺] and spermine⁴⁺ [Spn⁴⁺]) are indeed able to drive the liquid-like self-assembly of PARylated PARP1 in vitro. We also observed that the self-assembly of PARylated PARP1 in the presence of Mg²⁺, Ca²⁺, Mn²⁺, Spd³⁺, or Spn⁴⁺ influences its autoPARylation level and that Mg²⁺-dependent assembly of PARylated PARP1 contributes to more effective PAR hydrolysis by poly(ADP-ribose) glycohydrolase (PARG); this process was found to be accompanied by dissociation of the assemblies. In addition, the assembly of a complex of PARylated PARP1 with repair proteins strongly stimulated Pol β-mediated strand-displacement DNA synthesis but had no noticeable effect on the activity of Lig III. In summary, it is likely that liquid-like self-assembly of PARylated PARP1 plays a critical role in the regulation of i) its own activity, ii) the hydrolysis of poly(ADP-ribose) by PARG, and iii) Pol β-mediated DNA synthesis. This mechanism, together with other factors such as the type of DNA lesion or the type of DNA glycosylase⁴⁶, may influence the dynamics of repair complexes’ formation and regulate the choice between DNA synthesis involving several nucleotides (the long-patch pathway) and insertion of one nucleotide (the short-patch pathway) during BER/SSB repair.

Results

Divalent cations (Mg²⁺, Ca²⁺, and Mn²⁺) and polyamines trigger liquid-like self-assembly of PARylated PARP1

PAR synthesis activity of PARP1 in vitro is known to be regulated by cofactors such as divalent and multivalent cations, e.g., Mg²⁺, Ca²⁺, and natural polyamines [putrescine²⁺ (Put²⁺), Spn⁴⁺, or Spd³⁺]^47,48. In addition, divalent cations Mg²⁺ and Ca²⁺ and polyamines (Spd³⁺ and Spn⁴⁺) are also able to induce condensation and/or phase separation of nucleic acids^49,50, in particular PAR in vitro^51,52. Our previous studies indicate unusual assembly of a complex of PARylated PARP1 molecules in vitro, i.e., the formation of multimolecular associates in solution when protein automodification occurs at a micromolar concentration of PARP1 in the presence of 10 mM Mg^2+ 41,42. The addition of a chelating agent, namely ethylenediaminetetraacetic acid (EDTA), to the solution of autoPARylated PARP1 disrupts the formation of the supramolecular assembly⁴¹. Therefore, we hypothesized here that molecules of PARylated PARP1 are a cation-coordinated self-assembly system. In this study, we investigated the effect of the cation type and cation concentration on the supramolecular assembly of PARylated PARP1. First, we focused on determining which cations and at what concentrations will result in the self-assembly of PARP1 after its PARylation. For this purpose, we chose Mg²⁺ and other biologically relevant cations such as Ca²⁺ and Mn²⁺ and polyamines to analyze the possibility of self-assembly of PARylated molecules of PARP1 after the addition of the above cofactors. To assess the self-assembly of PARylated PARP1, dynamic light scattering (DLS) was used to determine the hydrodynamic size (R_h, nm) of modified PARP1 in the absence or presence of different concentrations of Mg²⁺ (5–15 mM), Mn²⁺ (0.5–2.0 mM), Ca²⁺ (1–8 mM), Put²⁺ (1–60 mM), Spn⁴⁺ (0.01–3.50 mM), or Spd³⁺ (1–9 mM) (Supplementary Table 1). By DLS analysis, the size of particles arising in a solution of PARylated PARP1 across the 0.1–10,000.0 nm range was determined at various concentrations of the cations (Supplementary Table 1). In the absence of cations, no assembly was observed (Fig. 1). DLS measurements showed R_h of ∼7.1 nm for PARP1 before the initiation of its PARylation, and ∼16 nm for PARylated PARP1 after its incubation with DNA and NAD⁺ without cations (Fig. 1). Nevertheless, we noticed the formation of large particles with a radius of 600–1000 nm after the addition of cations at the following concentrations: 2 mM Mn²⁺, 15 mM Mg²⁺, 8 mM Ca²⁺, 9 mM Spd³⁺, or 3.5 mM Spn⁴⁺ (Fig. 2a, Supplementary Fig. 1, Supplementary Table 1). Thus, we were able to determine the cation concentrations at which the formation of relatively stable PARylated PARP1 high-order structures or supramolecular assemblies was detectable. Of note, the addition of Put²⁺ did not have any effects on the self-assembly of PARylated PARP1 (Fig. 2a, Supplementary Table 1), whereas Mg²⁺, Ca²⁺, Mn²⁺, Spd³⁺, and Spn⁴⁺ each induced the formation of supramolecular complexes that were stable and could be detected within a certain range of a cation’s concentration (Fig. 2a, Supplementary Table 1). Consequently, the self-assembly of PARylated PARP1 strongly depends on the concentration of Me²⁺ or polyamines that is different for each type of cation. In the case of Me²⁺ cations, EDTA readily disrupted the assemblies, and particles with smaller R_h (∼15–18 nm) were detected (Fig. 2a, Supplementary Table 1), indicating that the self-assembly of PARylated PARP1 is coordinated by Me²⁺ cations. Additionally, we monitored the self-assembly of PARylated PARP1 after the addition of Me²⁺ or polyamines by measuring the turbidity of the solutions at 600 nm (Fig. 2b, Supplementary Fig. 2). The data showed that the solution of PARylated PARP1 became turbid after the introduction of Mn²⁺ up to 5 mM, Mg²⁺ up to 20 mM, Ca²⁺ up to 8 mM, Spn⁴⁺ up to 3.5 mM, or Spd³⁺ up to 9 mM (Fig. 2b and Supplementary Fig. 2). In the case of Me²⁺ cations, the turbidity of the PARylated-PARP1 solution strongly decreased after the addition of EDTA (Fig. 2b and Supplementary Fig. 2). Therefore, the turbidity measurement in conjunction with DLS data indicated that the addition of these Me²⁺ cations to PARylated PARP1 is accompanied by supramolecular assembly formation (Fig. 2). An insignificant change in turbidity was observed after the addition of 1,6-hexanediol: an ethyl-1,6-hexanediol widely used to dissolve LLPS assemblies⁵³ (Fig. 2b and Supplementary Fig. 2); this result suggested that nonionic interactions do not play a major role in this Me²⁺-coordinated self-assembly of PARylated PARP1 molecules. Although a noticeable reduction in turbidity was observed at concentrations of 10% and 20% of 1,6-hexanediol, the reagent was also not able to completely disrupt PARylated PARP1 associates formed in the presence of Spn⁴⁺ or Spd³⁺. It should be noted that EDTA and 1,6-hexanediol showed similar effects on the turbidity of the PARylated-PARP1 solution in the absence of olaparib used to stop PARylation (Supplementary Fig. 3). Thus, cation-mediated electrostatic interactions also contributed to the stability of the supramolecular assemblies in the presence of polyamines (Fig. 2b and Supplementary Figs. 2 and 3).

Fig. 1. Hydrodynamic size of PARP1 before and after its PARylation in the absence of cations.

Typical volume-weighted size distributions for PARP1 (a), a PARP1-and-DNA mixture (b), and PARylated PARP1 (c). The profile was obtained by means of experimental autocorrelation functions in the Zetasizer Nano ZS software. The average hydrodynamic radii (R_h) computed from the distributions are presented as well. R_h is the average calculated from at least three independent experiments. The hydrodynamic size of PARP1 was determined in reaction mixtures consisting of a 2.5 µM PARP1; b 2.5 µM PARP1 and 2.5 µM DNA duplex containing a one-nucleotide gap (DNA-gap); or c 2.5 µM PARP1, 2.5 µM DNA-gap, and 1 mM NAD⁺. The R_h values were measured directly after 1-min incubation of PARP1 with DNA (b) or after 30-min incubation of PARP1 with DNA and NAD⁺ (c).

Fig. 2. PARylated-PARP1 liquid-like self-assembly induced by the addition of cations.

a Histograms present the hydrodynamic radius (R_h, nm) for PARylated PARP1 as determined in the presence of different concentrations of Mg²⁺, Mn²⁺, Ca²⁺, Spn⁴⁺, Put²⁺, or Spd³⁺. R_h is the average (±SD) calculated from at least three independent experiments (Supplementary Table 1). b Histograms present turbidity (optical density [OD] at 600 nm) of PARylated PARP1 in the absence (green bars) or presence of cations, cations and 10% of 1,6-hexanediol (1,6-HD), cations, and 20% of 1,6-HD, or cations and EDTA as indicated (the mean ± SD of three independent experiments) (Supplementary Fig. 2). ***p < 0.001, **p < 0.01, *p < 0.05, ns: not significant, according to the t-test.

To further confirm that the self-assembly of PARylated PARP1 is mainly due to protein-attached negatively charged PAR, we tested whether the variation of cation concentrations was sufficient to induce self-assembly of protein-free PAR (Fig. 3, Supplementary Table 2). We found that PAR possesses an ability to form supramolecular assemblies (R_h ≈ 302–436 nm) within a certain range of cation concentration: 3–10 mM for Mn²⁺, 9–21 mM for Ca²⁺, 11–30 mM for Mg²⁺, or 2–5 mM for Spn⁴⁺ (Fig. 3a, Supplementary Fig. 4, Supplementary Table 2). Just as in the case of PARylated PARP1, the addition of EDTA dissolved the Me²⁺-induced PAR assemblies, and particles with smaller R_h (∼7–46 nm) were registered (Fig. 3a, Supplementary Table 2), confirming that cation-mediated PAR–PAR interactions contributed to the stability of PAR self-assembly in the presence of Me²⁺ ions. In fact, the protein-free PAR self-assembly showed the same tendency as PARylated PARP1 did, namely, the self-assembly was clearly found to occur in the presence of Mn²⁺, Ca²⁺, Mg²⁺, and Spn⁴⁺, and the Me²⁺-dependent assemblies were disrupted by the addition of EDTA (Figs. 2a and 3a, Supplementary Table 2). The solution of PAR supplemented with one of the cations also became turbid, further indicating that PAR alone can form supramolecular assemblies (Fig. 3b, c). These results suggested that the self-assembly of PARylated PARP1 in the presence of a cation is indeed a cation-mediated process driven via PAR–cation interactions.

Fig. 3. PAR liquid-like self-assembly induced by the addition of Me²⁺ cations or polyamines.

a Histograms of the hydrodynamic radius (R_h, nm) for PAR as determined in the presence of different concentrations of Mg²⁺, Mn²⁺, Ca²⁺, Spn⁴⁺, Put²⁺, or Spd³⁺. R_h is the average (±SD) determined from at least three independent experiments (Supplementary Table 2). b Absorbance spectra of PAR in the absence (red line) or presence of a 24 mM Mg²⁺ (blue line), 17 mM Ca²⁺ (green line), 7.5 mM Mn²⁺ (brown line), 25 mM (purple line), 10 mM Spd³⁺ (turquoise line) or 4.5 Spn⁴⁺ (dark blue line). The absorbance of the PAR solution was recorded from 300 to 700 nm. From left to right: two vertical dashed lines indicate the position of 350 and 600 nm wavelengths, respectively. c Histograms of PAR solution turbidity (OD at 350 nm) measured in the absence (green bar) or presence of a Me²⁺ cation or polyamine; the mean ± SD of three independent experiments.

To further delineate specific features of the influence of the cations on PARP1 PARylation, we assayed the formation of the supramolecular assembly when PARP1 modification proceeded in the concentration range of Mg²⁺, Ca²⁺, Mn²⁺, Spn⁴⁺, or Spd³⁺ that was sufficient to induce the self-assembly of already PARylated PARP1 molecules (Fig. 2). By a DLS assay and turbidity measurement, we again estimated the hydrodynamic size of PARP1 and changes in the turbidity of its solution during PARP1 PARylation (Fig. 4, Supplementary Figs. 5 and 6, Supplementary Table 3). When PARP1 PARylation was carried out in the presence of Mg²⁺ (3 mM), Ca²⁺ (10 mM), Mn²⁺ (3 mM), Spn⁴⁺ (3 mM), or Spd³⁺ (13 mM), particles with R_h in the 700–1143 nm range were detectable (Fig. 4a, Supplementary Fig. 5). Under our experimental conditions, all the tested cations, except for Put²⁺, induced the self-assembly of PARylated PARP1 during its automodification (Fig. 4a). Simultaneously, we noticed that the solution of PARP1 during its activation by the presence of the above-mentioned cations becomes turbid (Fig. 4b, Supplementary Fig. 6), indicating that PARP1 undergoes liquid-like self-assembly under these conditions, as observed for PARylated PARP1 and PAR supplemented with these cations (Figs. 2 and 3). As in the case of PARylated PARP1 supplemented with cations (Fig. 2b), the addition of EDTA disrupted the Me²⁺-dependent protein self-assembly during its automodification, but concentrations 5% and 10% of 1,6-hexanediol led to a significant turbidity reduction (by 28–40%) only in the case of Mg²⁺-, Spd³⁺-, or Spn⁴⁺-dependent association (Fig. 4b). Altogether these data suggested that although 1,6-hexanediol (5–15%) readily disrupts weak hydrophobic interactions in liquid-like protein condensates in vitro^53,54, the reagent is not capable of disrupting electrostatic interactions that are responsible for cation-induced assembly of PARylated PARP1.

Fig. 4. Cations induce liquid-like self-assembly of PARP1 during its PARylation.

a Histograms of the hydrodynamic radius (Rh, nm) for PARylated PARP1 as determined when its modification proceeds in the presence of 3 mM Mn²⁺, 10 mM Mg²⁺, 10 mM Ca²⁺, 13 mM Spd³⁺, or 3 mM Spn⁴⁺. R_h is the average (±SD) calculated from at least three independent experiments (Supplementary Fig. 5). b Histograms of PARylated-PARP1 solution turbidity (OD at 600 nm) measured in the absence (green bars) or presence of a cation; the mean ± SD of three independent experiments (Supplementary Fig. 6). ***p < 0.001, **p < 0.01, ns: not significant according to the t-test.

Taken together, these experiments clearly indicated cation-mediated formation of supramolecular complexes of PARylated PARP1 in vitro in response to variation of Mg²⁺, Ca²⁺, Mn²⁺, Spd³⁺, and Spn⁴⁺ concentrations.

PARP1 activity is regulated by its liquid-like self-assembly during PARylation

Our data revealed a well-pronounced dependence of liquid-like self-assembly of PARP1 on the concentration of cations that are cofactors of the enzyme and that this process can also influence PARP1 activity under these conditions. Although the impact of Mg²⁺, Ca²⁺, and polyamines on PARP1 activity in vitro has been studied previously^47,48, the investigation of its activity under the conditions of the assembly formation has not attracted attention before^41,42. To determine whether the self-assembly of PARP1 during automodification would be accompanied by the regulation of its activity, we analyzed its PARylation under conditions of its supramolecular assembly or under conditions when the assembly did not form. Given that the absence of cations or their 2 mM concentration did not trigger the self-assembly of PARylated PARP1 (Fig. 3a), the PARP1 activity was evaluated without cations or in the presence of 2 or 15 mM Mg²⁺, 2 or 4 mM Mn²⁺, 2 or 12 mM Ca²⁺, 2 or 9 mM Spd³⁺, 2 or 3.5 Spn⁴⁺, or 2 or 8 mM Put²⁺ (Fig. 5, Supplementary Fig. 7).

Fig. 5. Liquid-like self-assembly of PARP1 during autoPARylation leads to a decrease in the level of its modification.

a PARP1 PARylation detected by SDS-PAGE and subsequent phosphorimaging. The reaction mixtures were composed of 2.5 µM PARP1, 2 µM DNA-gap, 1 mM NAD⁺ ([³²P]NAD⁺, 2.5 μCi), and a Me²⁺ cation or polyamine as indicated. b Histograms of the relative levels of PARP1 PARylation (the mean ± SD of five to three independent experiments) from (a). The relative PARP1 PARylation levels were normalized to the PARylation level on PARP1 (taken as 100%) in the absence of cations. ***p < 0.001, **p < 0.01 according to the t-test.

There was no clear difference in the PARP1 activity between the presence and absence of 2 mM cations because the level of PARP1 automodification almost did not change (Fig. 5). Products of PARP1 autoPARylation looked like smeared bands, >130 kDa, potentially due to variation of the length of the PAR polymer attached, and were mainly concentrated in the concentrating gel and at the border between the concentrating gel and separating gel (Fig. 5a). PARylation under the conditions of liquid-like assembly changed its reaction product: PARylated PARP1 was resolved mainly as a distinct band at the border between the concentrating gel and separating gel, and the level of PARP1 automodification was lower by 20–50% (Fig. 5b). Consequently, the liquid-like self-assembly of PARylated PARP1 in the presence of cations affects its activity by noticeably reducing own PARylation.

Mg²⁺-dependent assembly of PARylated PARP1 stimulates PARG activity

PARP1 autoPARylation is a reversible process mainly due to PARG activity, which catalyzes the cleavage between ADP-ribose structural units at a terminal position and inside the polymer, thereby releasing ADP-ribose or oligo(ADP-ribose), respectively⁵⁵. Thus, PARG-driven degradation of PAR attached to PARP1 could provide the reversibility of the assembly of modified PARP1. Taking into account that PARG’s catalytic action can be performed in the presence of Mg²⁺ (ref. ⁵⁶), we studied the impact of PARG enzymatic activity on the Mg²⁺-dependent assembly of PARylated PARP1 using a combination of DLS and turbidity measurements (Fig. 6 and Supplementary Fig. 8a). Our results showed that PARG-driven degradation of poly(ADP-ribose) attached to PARP1 leads to the disruption of the supramolecular assembly of PARylated PARP1, as evidenced by a particle size decrease from ~500 to 7 nm (Fig. 6a, b) and reduced turbidity (OD₆₀₀) of the solution (Fig. 6c and Supplementary Fig. 8a). Thus, PAR hydrolysis by PARG is sufficient for disruption of the supramolecular assembly of PARylated PARP1, further indicating that the assembly is PAR dependent.

Fig. 6. Mg²⁺-dependent liquid-like self-assembly of PARylated PARP1 can be disrupted due to PAR hydrolysis by PARG.

Typical volume-weighted size distributions for a mixture of PARylated PARP1 and Mg²⁺ (a), or a mixture of PARylated PARP1, Mg²⁺, and PARG (b). The profile was obtained by means of experimental autocorrelation functions in the Zetasizer Nano ZS software. The average hydrodynamic radii (Rh) computed from the distributions are presented as well. Rh is the average calculated from at least three independent experiments. The hydrodynamic size was determined in reaction mixtures consisting of a PARylated PARP1 and 15 mM Mg²⁺; and b after the addition of PARG to the final concentration of 60 nM and the incubation of the sample overnight at 4 °C. c Graphs showing the change in turbidity (optical density [OD] at 600 nm) of PARylated PARP1 in the absence (0 mM Mg²⁺) or presence of 16 mM Mg²⁺ after addition of olaparib to the final concentration of 250 µM and PARG to the final concentration of 150 nM (Supplementary Fig. 8a); and represent the mean values of three independent experiments with error bars (±SD). d Graphs of the time course of PAR hydrolysis by PARG in the absence of Mg²⁺ (O mM) or in the presence of 2, 4, or 16 mM Mg²⁺. Graphs show the quantification of PARP1 PARylation detected after SDS-PAGE with subsequent phosphorimaging (Supplementary Fig. 8b) and represent the mean values of three independent experiments with error bars (±SD). The level of PAR-PARP1 hydrolysis was calculated as the percentage of the radioactivity of the ³²P-containing bands with respect to [³²P]PAR-PARP1 amounts detected in the absence of PARG and Mg²⁺. For the analysis of the hydrolysis of PARP1-bound PAR by PARG, 2.5 µM PARP1 was incubated with 2 µM DNA-gap in a buffer consisting of 140 mM NaCl, 20 mM HEPES-NaOH pH 7.5, 1 mM DTT, and 1 mM NAD⁺ ([³²P]NAD) at 30 °C for 30 min. After that, the reactions were stopped by the addition of olaparib to a final concentration of 200 µM. Next, 2, 4, or 16 mM Mg²⁺ was added to the samples, the samples were equilibrated for 5 min, and then PARG was added to the final concentration of 150 nM followed by incubation for 10–40 min at 30 °C.

Biomolecular condensate formation may influence the enzymatic activity of proteins^57,58. To determine whether the PARylated-PARP1 assembly can directly alter PARG’s catalytic function, we performed an assay of PARG activity during the assembly of PARylated PARP1 in the presence of 16 mM Mg²⁺ or its nonassembly at 0, 2, or 4 mM Mg²⁺ (Fig. 6d and Supplementary Fig. 8b, c). PARylated PARP1 and Mg²⁺ of different concentrations were mixed and incubated for 5 min to provide enough time for assembly formation; PARG was next added to start the reaction, and the relative amounts of [³²P]PARylated PARP1 after PAR hydrolysis were determined (Fig. 6d and Supplementary Fig. 8b, c). Compared with the hydrolysis of modified PARP1 in the absence or presence of 2 or 4 mM Mg²⁺, its assembly at 16 mM Mg²⁺ led to stimulation of PARG activity up to 2.0–2.5-fold in a time-dependent manner. Consequently, the supramolecular assembly of PARylated PARP1 promoted more effective PAR hydrolysis by PARG, whose activity was strongly increased by such conditions (Fig. 6d and Supplementary Fig. 8d, c).

Mg²⁺-dependent liquid-like self-assembly of PARylated PARP1 regulates Pol β-dependent DNA synthesis but has no effect on the activity of Lig III

It has been reported that PARP1-dependent synthesis of PAR at sites of nonbulky DNA lesions is a way in which DNA repair factors are recruited to sites of DNA damage^8,17,19. In particular, modified PARP1 directly recruits XRCC1 (which is an SSB and BER scaffolding protein) to sites of DNA damage^{8,17,19,24,59–62}. In turn, XRCC1 forms stable complexes with BER/SSB repair factors such as Lig III and Pol β and may coordinate their activities^60,63–67. Pol β catalyzes one-nucleotide gap-filling or strand-displacement DNA synthesis in BER and SSB repair; in particular, after the gap filling, Pol β generates a nicked DNA intermediate which is sealed by Lig III^68,69. Both ex vivo and in vitro, Pol β and Lig III have been shown to work in close cooperation with XRCC1, which modulates their activities and stabilizes Pol β–DNA and Lig III–DNA complexes^{60,63,66,67,69}. Pol β and Lig III are Mg²⁺-dependent enzymes, and this metal ion plays a critical role in the DNA synthesis and ligation catalyzed by these enzymes^44,45,70. Thus, we analyzed the influence of the liquid-like self-assembly of PARP1 during the autoPARylation triggered by Mg²⁺ cations on the DNA synthesis activity of Pol β and nick-sealing activity of Lig III.

First, we tested whether Mg²⁺-induced assembly of PARylated PARP1 occurs in the presence of the repair enzymes. For this purpose, we monitored the assembly formation by DLS, which enabled direct measurement of particle size when PARP1 was incubated with DNA and NAD⁺ in the presence of complexes Pol β–XRCC1 and/or Lig III–XRCC1 and 5 or 15 mM Mg²⁺. According to the DLS data, when incubation of PARP1 with complexes Pol β–XRCC1 and/or Lig III–XRCC1 was performed in the presence of 15 mM Mg²⁺, particles of size in the range 300–700 nm were found to form after the 30-min reaction (Fig. 7 and Supplementary Fig. 9). Therefore, PARylated PARP1 retained its ability to form large supramolecular assemblies in the presence of 15 mM Mg²⁺ and complexes Pol β–XRCC1 and/or Lig III–XRCC1 (Fig. 7).

Fig. 7. Mg²⁺ can induce a liquid-like assembly of PARP1 during its PARylation in the presence of complexes Pol β–XRCC1 and/or Lig III–XRCC1.

The time dependence of hydrodynamic radii (Rh) of PARP1, XRCC1, and Pol β (a), PARP1, XRCC1, and Lig III (b) and PARP1, XRCC1, Pol β and Lig III (c) mixtures under conditions of PARP1 activation, i.e., in the presence of DNA, NAD⁺ and 5 mM (red line) or 15 mM (blue line) Mg²⁺. The hydrodynamic size of PARP1 was determined in reaction mixtures consisting of (a) 2.5 µM PARP1, 0.1 µM XRCC1, 0.1 µM Pol β, 1 µM DNA, 1 mM NAD+, and 5 or 15 mM Mg²⁺; (b) 2.5 µM PARP1, 0.1 µM XRCC1, 0.1 µM Lig III, 1 µM DNA, 1 mM NAD⁺, and 5 or 15 mM Mg²⁺; (c) 2.5 µM PARP1, 0.1 µM XRCC1, 0.1 µM Pol β, 0.1 µM Lig III, 1 µM DNA, 1 mM NAD⁺, and 5 or 15 mM Mg²⁺. The Rh values were measured directly after 125–1100 s incubation of PARP1 with BER proteins in the presence of DNA and NAD⁺. Rh is the average calculated from at least three independent experiments.

To this end, DNA synthesis and ligation repair steps were reconstituted using purified proteins (Pol β + XRCC1, Lig III + XRCC1, or Pol β + XRCC1+Lig III), and a DNA duplex containing a one-nucleotide gap or nick that mimics DNA intermediates arising during BER or SSB repair.

Under the conditions where PARylated PARP1 self-assembled or not (Fig. 7), we first tested the DNA synthesis catalyzed by Pol β on the DNA duplex (DNA-gap) containing a one-nucleotide gap with phosphate at the 5′ end of a downstream primer in the presence of 5 or 15 mM Mg²⁺ (Fig. 8, Supplementary Fig. 10). DNA-gap mimics the substrate of Pol β arising during BER or SSB repair⁷¹. The effect of PARylation of PARP1 on the DNA synthesis was assessed via a comparison of Pol β’s primer extension activity in the presence of DNA-gap, four dNTPs, NAD⁺, 5 or 15 mM Mg²⁺, with or without PARP1 (Fig. 8). The results showed that Pol β without PARP1 and NAD⁺ or in the presence of 5 mM Mg²⁺, PARP1, and NAD⁺ synthesized primarily a 14-nucleotide product due to single-nucleotide addition or one gap filling; modest strand-displacement synthesis corresponding to primer extension products longer than 14 nucleotides was noted (Fig. 8a). These data are in agreement with studies in which Pol β has been shown to efficiently fill a one-nucleotide gap and to have a limited strand-displacement capacity^72,73. According to our data, when PARylation of PARP1 occurred in the presence of 5 mM Mg²⁺, there was no effect on the DNA synthesis product (Fig. 8). In contrast, when PARylation of PARP1 and its self-assembly took place in the presence of 15 mM Mg²⁺, Pol β primer extension products were mostly longer than 14 nucleotides, thereby representing strand-displacement DNA synthesis (Fig. 8b). Therefore, the Pol β-dependent DNA synthesis under the conditions of liquid-like self-assembly of PARP1 was accompanied by a ∼2-fold increase in the yield of strand-displacement products and a simultaneous decrease in the yield of gap filling (Fig. 8b). Of note, our previous studies have shown that PARP1 mainly suppresses the Pol β-mediated strand-displacement DNA synthesis, and the inhibitory effect of PARP1 is canceled by its PARylation when the level of DNA synthesis is restored to the level observed in the absence of PARP1⁷³. Nonetheless, PARP1 PARylation followed by liquid-like self-assembly led to not only the restoration of the level of DNA synthesis but also stimulation of the strand-displacement synthesis to ∼2-fold of the initial level of the synthesis observed without PARP1 (Fig. 8b). These results allowed us to conclude that DNA synthesis coupled with liquid-like self-assembly of PARylated PARP1 caused a significant increase in strand-displacement activity of Pol β.

Fig. 8. Mg²⁺-dependent liquid-like self-assembly of PARylated PARP1 leads to stimulation of strand-displacement DNA synthesis catalyzed by Pol β.

a The DNA synthesis catalyzed by Pol β was carried out in reaction mixtures consisting of 2.0 µM 5′-³²P-labeled DNA-gap, 0.1 µM Pol β, 0.1 µM XRCC1, 100 µM four dNTPs, 1 mM NAD⁺, and 5 or 15 mM Mg²⁺ in the absence or presence of 2.5 μM PARP1, as indicated, for 5–30 min at 30 °C. Schematic representation of the DNA substrate is shown at the top of the gel. b Histograms of the quantification of the primer extension (%) from (a). The yield of products was calculated as the percentage of the total radioactivity of all the ³²P-containing DNA bands in a lane and is presented as the mean ± SD of three independent experiments. **p < 0.01 according to the t-test.

Aside from DNA synthesis, we also examined the effect of liquid-like self-assembly of PARylated PARP1 on the nick-sealing activity of Lig III catalyzing the last step of BER/SSB repair. For this purpose, we used a nicked DNA substrate (DNA-nick) that mimics Pol β’s gap-filling product (Supplementary Fig. 11). In the absence of PARP1 and NAD⁺, we observed the formation of a 30-nucleotide ligation product, and the total extent of the ligation was approximately 80% (Supplementary Fig. 11b). Under conditions of PARP1 PARylation in the presence of either 5 or 15 mM Mg²⁺, we observed approximately the same amount of the ligation product at the 30 min time point, and the yield of the ligation product was slightly diminished at time points 5 and 10 min in comparison with the reaction that was carried out in the absence of PARP1 and NAD⁺ (Supplementary Fig. 11b). Unlike our observations with DNA synthesis by Pol β, the assembly of PARylated PARP1 did not affect the nick-sealing activity of Lig III (Supplementary Fig. 11).

Next, we examined the impact of PARP1 PARylation on DNA synthesis coupled with ligation in the reaction mixture that included Pol β, XRCC1, Lig III, and DNA-gap (Fig. 9, Supplementary Fig. 12). The results revealed time courses of product formation for one-nucleotide gap filling (+1 nt), strand-displacement synthesis (+2 to +7 nt), and ligation (30 nt), i.e., nick sealing after gap-filling (Fig. 9). It was observed that at the 20 min time point, the levels of primer extension and ligation products were ~20% and 60%, respectively, and were not visibly affected by the concentration of Mg²⁺ (5 or 15 mM) (Fig. 9b). The yield of ligation products under the conditions of PARP1 PARylation was slightly lower (∼50%) than the yields (∼60%) obtained without PARP1 (Fig. 9b). Nevertheless, we did not observe an obvious difference in the amount of ligation products when PARylated PARP1 self-assembled or not. As in the case of Pol β alone, liquid-like self-assembly of PARylated PARP1 clearly promoted the strand-displacement DNA synthesis even in the coupled reaction system including both Pol β and Lig III (Fig. 9). Under these conditions, an approximately twofold increase in the yield of DNA synthesis was observed, which was accompanied by a decrease in the yield of ligation products (Fig. 9).

Fig. 9. Mg²⁺-dependent liquid-like self-assembly of PARylated PARP1 leads to stimulation of strand-displacement DNA synthesis by Pol β when the DNA synthesis is coupled with ligation but reduces the formation of the ligation product by Lig III.

a) The products of Pol β–catalyzed DNA synthesis coupled with ligation. The DNA synthesis and nick sealing catalyzed by Pol β and Lig III, respectively, were carried out in reaction mixtures composed of 2.0 µM 5′-[³²P]labeled DNA duplex (DNA-gap), 0.1 µM Pol β, 0.1 µM Lig III, 0.1 µM XRCC1, 100 µM four dNTPs, 1 mM ATP, 1 mM NAD⁺, and 5 or 15 mM Mg²⁺ in the absence or presence of 2.5 μM PARP1, as indicated, for 5–20 min at 30 °C. Schematic representation of the DNA substrate is shown at the top of the gel. b Histograms of quantification of the primer extension (%) and ligation product (%) from (a). The yield of the products was calculated as the percentage of the total radioactivity of all the ³²P-containing DNA bands in a lane and represents the mean ± SD of three independent experiments; the SD did not exceed 10% of total counts.

Thus, the upregulation of strand-displacement DNA synthesis by Pol β is directly related to the PARylation of PARP1 and its liquid-like assembly during automodification in the presence of Mg²⁺.

Discussion

Many functions have been attributed to PARP1 during the DNA damage response^74,75. One of the primary roles of the PAR synthesis catalyzed by PARP1 seems to be “a signal molecule for the specific recruitment of DNA repair factors to sites of DNA damage”^{12–14,16,17,74,75}. Negatively charged PAR, which has unique characteristics such as two negatively charged phosphate groups per monomer ADP-ribose, low complexity, branching, lower flexibility, and higher stiffness (as compared to DNA and RNA)⁵², may also trigger LLPS of proteins and formation of biomolecular condensates linked to DNA repair events^{3–6,76–78}. Recent articles indicate that PAR-dependent LLPS in the context of DNA repair is highly dependent on interactions of PAR with RNA-binding proteins, in particular with a member of the FET family, consisting of FUS, TAF15, and EWS^6,26,79,80, suggesting that FET proteins may play a central role in the organization of the DNA repair compartments coupled to LLPS and PARP1 activation^{3–5,31,76–78}.

Previously, it has been shown that divalent cations (Mg²⁺ or Ca²⁺) or polyamines (Spn⁴⁺ or Spd³⁺) cause condensation–compaction of DNA, RNA, or protein-free PAR in vitro^{49–52,81–84}. Results obtained in our study indicate that metal (Mg²⁺, Ca²⁺, or Mn²⁺) or multivalent (Spd³⁺ or Spn⁴⁺) cations promote the intermolecular association of PARylated PARP1 by inducing PARP1 liquid-like self-assembly in vitro both after and during its automodification, thereby serving as cofactors of PARP1 (Figs. 2 and 4). The cations (Mg²⁺, Ca²⁺, Mn²⁺, Spd³⁺, and Spn⁴⁺) are reported to be involved in the regulation of many cell functions and are cofactors in enzymatic reactions^33–40; some of them (Mg²⁺ and Ca²⁺) are considered regulators of LLPS of proteins and nucleic acids³². Although Mn²⁺ concentration in the cell is low (in the submillimolar range: 20–75 μM)⁸⁵, cations Mg²⁺, Ca²⁺, Spd³⁺, and Spn⁴⁺ are abundant in eukaryotic cells and tissues. For example, the range of physiological concentrations of polyamines has been estimated to be 0.88–1.58 mM³⁴; total Mg²⁺ concentration is between 17 and 20 mM in the majority of mammalian cell types^86,87; intracellular concentration Ca²⁺ has been estimated at 0.1 µM to 1 mM^88,89, but some organelles (endoplasmic reticulum) can accumulate Ca²⁺ to 10–50 mM levels⁸⁹. Although there is no direct evidence of condensation of PARylated PARP1 in vivo, our results show that near-physiological concentration of the cations has a strong effect on the phase behavior of PARylated PARP1 in vitro.

Cations Mg²⁺, Mn²⁺, and Ca²⁺ tested in the present study have a coordination number of four to six for phosphate groups^90–93, and one such cation can potentially bind two phosphate groups in PAR simultaneously, by working as a “bridging molecule” that links different PAR molecules. Likewise, intermolecular interactions of PARylated PARP1 can also be stabilized by “polyamine bridges” formed between separate PAR molecules, as in the case of DNA^81,82. This hypothesis is supported by evidence that a chelating agent, EDTA, disrupts the Me²⁺-induced self-assembly of PARylated PARP1, thereby illustrating the cation-dependent nature of such a supramolecular association (Figs. 2 and 4, Supplementary Tables 1 and 3). In contrast, 1,6-hexanediol, i.e., an inhibitor LLPS of biomolecules^53,54, does not disrupt the intermolecular associates of PARylated PARP1 arising in the presence of either Me²⁺ cations or polyamines (Figs. 2b and 4b). Therefore, the intermolecular interactions coordinated by cations are likely to be a major driving force behind the self-assembly of PARylated PARP1 in the presence of cations, which create intermolecular bridges between PAR phosphate oxygen atoms (Fig. 10a).

Fig. 10. The proposed model of the cation-dependent liquid-like assembly of PARylated PARP1 and regulation of BER/SSB repair.

a PARylated PARP1 intermolecular contacts are formed via coordination of cations with PAR chains; the assembly is destroyed by the presence of EDTA owing to disruption of the cation coordination bonds. b Liquid-like assembly of PARylated PARP1 regulates BER/SSB repair. The assembly is composed of PARylated PARP1, DNA repair factors, and damaged DNA; damaged DNA and PAR linked to PARP1 recruit XRCC1, Pol β, and Lig III.

Growing evidence indicates that LLPS—initially shown to be involved in the formation of biomolecular condensates or membrane-less organelles^1,94,95—also plays an active part in the regulation of enzymatic activity^3–5,96. It has been found that divalent or multivalent cations such as Mg²⁺, Ca²⁺, Put²⁺, or Spn⁴⁺ stimulate PARP1 activity, likely by influencing electrostatic repulsion between negatively charged DNA and the synthesized PAR^47,48,97. Nevertheless, our results revealed that the liquid-like self-assembly of PARylated PARP1 in the presence of the cations directly correlates with the diminished level of its modification and hence may be important for the regulation of PARP1 activity (Fig. 5).

On the other hand, the supramolecular assembly of PARylated PARP1 promotes more effective PAR hydrolysis by PARG (Fig. 6). For instance, PARylated PARP1 alone can undergo liquid-like self-assembly, and the assemblies are highly responsive to changes in concentrations of cations and can be rapidly generated or dismantled due to PARG-dependent PAR hydrolysis (Figs. 6a and 10a). PAR-binding repair proteins may mix specifically with these PARylated PARP1 assemblies, thereby forming a condensed phase, and can carry out specialized repair reactions. The regulation of the BER/SSB repair protein activities depending on PARP1 activation has been researched in in vitro settings involving reconstituted systems and cell extracts^28,30 but not in the context of repair condensates. The previous studies on reconstituted systems have shown that PARP1 inhibits activities of BER/SSB repair enzymes such as APE1, Pol β, and FEN1, and that the PARylation of PARP1 cancels its inhibitory action^72,98,99. Here, we observed that PARP1 PARylation under conditions of its self-assembly or nonassembly has different effects on the activities of complexes Pol β–XRCC1 and Ligase III–XRCC1. Pol β was more active in strand-displacement synthesis when automodification of PARP1 was accompanied by its self-assembly (Fig. 8). Although Lig III prevents strand displacement by Pol β under similar conditions in a reconstituted system¹⁰⁰, the self-assembly of PARylated PARP1 here led to a decrease in the ligation and a simultaneous increase in strand-displacement synthesis when Pol β-mediated DNA synthesis was coupled with the ligation (Fig. 9c, d). Thus, the supramolecular assembly of complexes Pol β–XRCC1 and/or Lig III–XRCC1 and PARylated PARP1 (Fig. 7) could play a crucial role in the stimulation of strand-displacement activity of Pol β in BER/SSB repair pathways (Figs. 8 and 9c, d) and in the regulation of the transition from the short-patch to long-path BER/SSB repair pathway and vice versa.

To date, there have been no reports of PAR-dependent organization of biomolecular condensates without RNA-binding proteins in the cell^{3–5,76–78}, and it is possible that repair condensates can be generated with participation of only PARylated PARP1 and of recruitment of such repair factors as XRCC1, Pol β, and Lig III. XRCC1, being a loading platform for BER/SSB repair factors, is also characterized by PAR-dependent accumulation at a DNA lesion^19,24,101 and even can form own supramolecular assemblies in the presence of protein-free PAR in vitro¹⁰². Therefore, in the case of PARP1- and PAR-dependent DNA repair, XRCC1–Pol β–Lig III interactions followed by condensate formation may be directly connected with the formation of transient repairosome compartments, which may implement spatiotemporal regulation of DNA repair (Fig. 10b).

PARP1 PARylation is important for the turnover of DNA SSB and double-strand break repair and repair coupled to DNA replication and transcription as well as to other highly dynamic intracellular processes^{11,21,28,30,103–105}. Further research on how PARylated PARP1 self-assembly can regulate DNA repair reactions is needed to assess the influence of condensation on the DNA repair course and to investigate the function of RNA(PAR)-binding proteins containing low-complexity domains that drive biomolecular condensate formation in a live cell.

Methods

Chemicals, proteins, and DNA substrates

Radioactive [α-³²P]ATP and [γ-³²P]ATP were prepared in the Laboratory of Biotechnology at the ICBFM (SB RAS, Novosibirsk, Russia). NAD⁺ and β-nicotinamide mononucleotide were purchased from Sigma-Aldrich (United States, catalog ## 481911 and N3501, respectively), olaparib (AZD2281, Ku-0059436) was purchased from Apexbio Technology (United States, catalog # A4154), whereas reagents for buffer and electrophoresis components from Sigma-Aldrich, United States (Tris, catalog # T6791; EDTA, catalog # E5134; HEPES, catalog # H3375), PanReacAppliChem, Germany (acrylamide/bis-acrylamide, catalog # A1089/A3636; Urea, catalog # A1049), Molecular Group (DTT, catalog # 19733320), and Merk (NaCl, catalog # 106404). T4 polynucleotide kinase was acquired from Biosan (Russia), and protein molecular weight markers from ThermoScientific (Lithuania).

Recombinant human PARP1 was expressed in insect Trichoplusia ni High Five™ (Hi5) cells and purified as previously described¹⁰⁶. Recombinant bovine PARG was expressed in BL21(DE3) Escherichia coli cells and purified as previously described⁵⁶.

Plasmid pET16BXH-XRCC1 containing full-length cDNA of the human C-His₁₀-tagged XRCC1 gene was a generous gift from Keith W. Caldecott (University of Sussex, Brighton, UK). Recombinant XRCC1 was overexpressed in the Rosetta 2(DE3)pLysS strain of E. coli (Novagen, Germany) and purified as previously described¹⁰⁷ with modifications. Briefly, XRCC1 was purified by Co²⁺-IMAC affinity chromatography (HP column, 5 ml; GE Healthcare, Sweden) in a linear gradient of 5 mM–1 M imidazole, by heparin affinity chromatography (HiTrap HP column, 5 ml, GE Healthcare, Sweden) in a linear gradient of 90–750 mM NaCl, and mono Q anion exchange chromatography (GL 5/50 column, 1 ml, GE Healthcare, Sweden) in a linear gradient of 90–750 mM NaCl. Pure protein-containing fractions were concentrated and kept in a buffer (50 mM sodium phosphate pH 7.2, 300 mM NaCl, 10 mM β-mercaptoethanol, and 50% of glycerol) at −25 °C.

Plasmid pGEX4T-hLIG3 containing full-length cDNA of the human N-GST-tagged Lig3 gene was purchased from Addgene (UK, catalog # 81055). Recombinant Lig III was overexpressed in Rosetta 2 strain of E. coli (Novagen, Germany) as previously described¹⁰⁸. The pellet of E. coli cells was resuspended in lysis buffer (50 mM sodium phosphate pH 7.5, 600 mM NaCl, 7 mM β-mercaptoethanol, 1% of NP-40, 10% of glycerol, 1 mM PMSF, 1 mM benzamidine, and a complete EDTA-free protease inhibitor cocktail [Roche, Switzerland]), sonicated on ice, and clarified by centrifugation (48,000 × g, 60 min, 4 °C). GST-tagged Lig III was purified in batch mode using Glutathione Sepharose (4B resin, 1 ml, GE Healthcare, Sweden). Briefly, the supernatant was incubated with the resin with slow rotation at 4 °C for 1 h and centrifuged for 5–10 min (1000 × g), and the supernatant was carefully removed by pipetting. The subsequent steps were carried out similarly to the above procedure, the resin was incubated with lysis buffer containing 300 mM NaCl for 10 min; with ATP-buffer (see description below) for 10 min; with lysis buffer containing 300 mM NaCl for 5 min; with A-buffer (50 mM sodium phosphate pH 7.5, 7 mM β-mercaptoethanol, 0.1% of Tween 20, and 10% of glycerol) with 1 mM PMSF and 1 mM benzamidine overnight. After that, the protein was eluted with 40 mM reduced glutathione in A-buffer with 1 mM PMSF and 1 mM benzamidine. The ATP-buffer consisted from a denatured protein lysate from E. coli; we used the Glutathione Sepharose flowthrough fraction, which was heated for 10 min at 65 °C, centrifuged (10 min, 12,000 × g, room temperature), and supplemented with 5 mM ATP and 10 mM MgSO₄. This procedure is recommended to avoid undesirable consequences of contamination of GST-fused proteins with bacterial “chaperone” DnaK¹⁰⁹. GST-Lig III–containing fractions were concentrated on Vivaspin centrifugal concentrators (Vivaspin Turbo 15, 10 kDa, PES; Sartorius, Germany) and the GST tag was cleaved off using thrombin. Briefly, the protein was incubated with thrombin (10 U per mg of protein) in A-buffer containing 150 mM NaCl and 50% of glycerol at −20 °C for 3 days. SDS-PAGE was performed to monitor this proteolysis step. Next, Lig III was purified by affinity chromatography on a single-strand-cellulose column (8 ml, USB, USA) in a linear gradient of 50–1000 mM NaCl in A-buffer containing 10% of glycerol and by gel-filtration chromatography (HiLoad Superdex 200 pg column, 120 ml, GE Healthcare, Sweden) in A-buffer containing 50 mM NaCl. The resultant pure protein-containing fractions were concentrated and kept in a storage buffer (50 mM sodium phosphate pH 7.5, 50 mM NaCl, 7 mM β-mercaptoethanol, 2 mM EDTA, and 45% of glycerol) at −25 °C.

Plasmid pRSETB-pol containing full-length cDNA of the rat POLB gene was a generous gift from Samuel H. Wilson (National Institute of Environmental Health Sciences, Galveston, USA). Recombinant Pol β was overexpressed in the Rosetta 2(DE3)pLysS strain of E. coli (Novagen, Germany) and purified as previously described¹¹⁰ with modifications. Briefly, Pol β was purified by tandem DEAE ion exchange/heparin-Sepharose affinity chromatography (HiPrep DEAE Sepharose FF 16/10 and HiPrep Heparin-Sepharose FF 16/10 columns, 20 ml; GE Healthcare, Sweden) connected consecutively; the columns were washed with 30 ml of buffer A (50 mM Tris-HCl pH 8.0, 1 mM EDTA, 7 mM β-mercaptoethanol, 0.1% of NP-40, and 1 mM PMSF) with 100 mM NaCl, then the DEAE column was disconnected, and the protein was eluted from the heparin-column in a linear gradient of 0.1–1.0 M NaCl in buffer A. After that, fractions enriched with Pol β were purified by Mono S cation exchange chromatography (GL 5/50 column, 1 ml; GE Healthcare, Sweden) in a linear gradient of 100–800 mM NaCl in buffer A. Fractions enriched with Pol β were pooled and passed through a concentrator (Vivaspin Turbo 15, 50 kDa, PES; Sartorius) to remove high-molecular-weight aggregates of Pol β. The flowthrough fraction was diluted with buffer A and subjected to repeated chromatography on the Mono S-column. The obtained pure protein-containing fractions were concentrated and kept in a buffer (25 mM Tris-HCl pH 8.0, 3.5 mM β-mercaptoethanol, 0.12 mM EDTA, and 45% of glycerol) and stored at −25 °C.

Oligodeoxynucleotides were synthesized by the Laboratory of Biomedical Chemistry at the ICBFM (SB RAS, Novosibirsk, Russia). The sequences of oligodeoxynucleotides were as follows: upstream primers: 5′-OH-gggttggtttgcg-3′ and 5′-OH-gggttggtttgcgc-3′; downstream primer: 5′-phosphate-attcacagttctccgc-3′; template: 3′- cccaaccaaacgc g taagtgtcaagaggcg-5′. DNA duplexes (30-mers with a one-nucleotide gap) were obtained by hybridization of an oligonucleotide (3′-cccaaccaaacgc g taagtgtcaagaggcg-5′) with complementary oligonucleotides (5′-OH-gggttggtttgcg-3′ and 5′-phosphate-attcacagttctccgc-3′) in a 1.0:1.5 ratio. The oligonucleotide mixture was incubated for 3 min at 95 °C and then slowly cooled to room temperature.

To prepare [³²P]labeled DNA-gap or DNA-nick, upstream primers were 5′-end-labeled with ³²P using T4 polynucleotide kinase and [γ-³²P]ATP (5000–6000 Ci/mmol) and were purified on MicroSpin™ G-25 columns (Amersham Pharmacia Biotech, USA) according to the recommended protocol. Complementary oligodeoxynucleotides were annealed by heating a solution of equimolar amounts at 90 °C for 2 min and then were slowly cooled to room temperature to obtain ³²P-labeled DNA duplexes containing either a one-nucleotide gap (DNA-gap) or a single-strand nick (DNA-nick).

In vitro synthesis of protein-free PAR was performed as described before⁵⁶. PAR concentration was estimated by measurement of absorbance at 258 nm (A₂₅₈) and application of an extinction coefficient of 13.5 mM⁻¹cm⁻¹ for ADP-ribose.

A radioactive assay of PARP1 PARylation in vitro

NAD⁺ [³²P]labeled on the adenosine phosphate was synthesized as described earlier¹¹¹. The protein PARylation assay was performed in reaction mixtures (15 µl) consisting of 20 mM Tris-HCl pH 7.5, 140 mM NaCl, 1 mM DTT, 2 µM DNA-gap, 2.5 µM PARP1, 1 mM NAD⁺ ([³²P]NAD⁺, 2.5 μCi), and one of the following: 2 or 15 mM MgCl₂, 2 or 4 mM MnCl₂, 2 or 12 mM CaCl₂, 2 or 9 mM spermidine (Spd³⁺), 2 or 3.5 mM spermine (Spn⁴⁺), or 2 or 8 mM putrescine (Put²⁺). The mixtures were incubated at 30 °C for 30 min. The reaction components were mixed on ice. All the reactions were initiated by adding NAD⁺ ([³²P]NAD⁺, 2.5 μCi/50 µl) to a final concentration of 1 mM. The reactions were stopped by the addition of SDS sample buffer and heating for 2 min at 95 °C. Then, the products were separated by denaturing 10% PAGE (with SDS).

A radioactive assay of PARG activity in vitro

The [³²P]labeled poly(ADP-ribosyl)ated PARP1 was prepared in a reaction mixture (50 µl) consisting of 20 mM Tris-HCl pH 7.5, 100 mM NaCl, 1 mM DTT, 2 µM DNA-gap, 2.5 µM PARP1, and 1 mM NAD⁺ ([³²P]NAD⁺, 7.5 μCi). The reaction components were mixed on ice. The reactions were initiated by the addition of NAD⁺ ([³²P]NAD⁺, 7.5 μCi/150 µl) to a final concentration of 1 mM and were incubated at 30 °C for 30 min. The reactions were stopped by the addition of olaparib to a final concentration of 200 µM. After that, the reactions were supplemented with 2.0, 4.0, or 16 mM Mg²⁺, and the samples were equilibrated for 5 min. The hydrolysis of PAR was started by the addition of PARG to a final concentration of 150 nM and was allowed to proceed for 40 min at 30 °C. Three-microliter aliquots were taken at 10, 20, and 40 min, and the reactions were stopped by the addition of SDS sample buffer and by heating for 2 min at 95 °C. The products were separated by denaturing 10% PAGE (with SDS).

Turbidity measurements

For analysis of the turbidity of PARylated PARP1 solutions after its modification, 2.5 µM PARP1 was incubated with 2 µM DNA-gap in a buffer consisting of 100 mM NaCl, 20 mM HEPES-NaOH pH 7.5, 1 mM DTT, and 1 mM NAD⁺ at 30 °C for 40 min. The reactions were stopped by the addition of olaparib to a final concentration of 250 µM. After that, the reactions were supplemented with 5 mM Mn²⁺, 20 mM Mg²⁺, 8 mM Ca²⁺, 9 mM Spd³⁺, or 3.5 mM Spn⁴⁺, and an absorption spectrum of the solutions was recorded. Then, either 1,6-hexanediol or EDTA was added as indicated in a figure legend, and the absorption spectrum of the solutions was recorded again.

For analysis of the turbidity of PARP1 solutions during its modification, 2.5 µM PARP1 was incubated with 2 µM DNA-gap in a buffer consisting of 140 mM NaCl, 20 mM HEPES-NaOH pH 7.5, 1 mM DTT, and 1 mM NAD⁺ in the presence of 15 mM Mg²⁺, 4 mM Mn²⁺, 12 mM Ca²⁺, 16 mM Spd³⁺, or 4 mM Spn⁴⁺ at 30 °C for 30 min. The reactions were stopped by the addition of olaparib to a final concentration of 250 µM. After that, the absorption spectrum of the solutions was recorded. Then, either 1,6-hexanediol or EDTA was added as indicated in a figure legend, and the absorption spectrum of the solutions was recorded again.

For analysis of the turbidity of PAR-PARP1 solutions after PARG treatment, 2.5 µM PARP1 was incubated with 2 µM DNA-gap in a buffer consisting of 140 mM NaCl, 20 mM HEPES-NaOH pH 7.5, 1 mM DTT, and 1 mM NAD⁺ at 30 °C for 40 min. The reactions were stopped by the addition of olaparib to a final concentration of 200 µM. After that, the absorption spectrum of the solutions was recorded. Then, PARG to the final concentration of 150 nM and Mg²⁺ to the final concentration of 16 mM were added as indicated in a figure legend, and the absorption spectrum of the solutions was recorded again during 0.36–4.28 h.

To assess the turbidity of protein-free PAR solutions, 68 µM ADP-ribose was incubated in a buffer composed of 140 mM NaCl, 20 mM HEPES-NaOH pH 7.5, and 1 mM DTT in the absence or presence of 1.0–7.5 mM Mn²⁺, 2–24 mM Mg²⁺, 2–17 mM Ca²⁺, 2–10 mM Spd³⁺, or 2.0–4.5 mM Spn⁴⁺ at 30 °C for 5 min. After that, the absorption spectrum of the solutions was recorded.

To measure the turbidity of PARP1 or PAR solutions, nonbinding black 96-well plates with a transparent bottom (Corning 3881) were used. The absorption spectrum of the solutions was recorded using a POLARstar Optima multidetection microplate reader (BMG Labtech, Offenburg, Germany) in the 350–650 nm wavelength range¹¹².

Hydrodynamic size measurements

To evaluate the hydrodynamic radius (R_h) of proteins and PAR, DLS was performed. DLS measurements were carried out on a Zetasizer Nano ZS device (Malvern Instruments Ltd., Malvern, UK) at 25 °C. All stock solutions of the buffer, proteins, PAR, and DNA were pre-ultrafiltered through a polyethersulfone membrane (0.2 μm pore size) in a Vivaspin centrifugal concentrator (Sartorius). The measurements and data processing were performed as described elsewhere⁴². The measurements were performed in a Low-volume quartz batch cuvette (ZEN 2112). Each measurement was conducted at least three times.

For analysis of PARylated PARP1 hydrodynamic size in the absence of cations, 2.5 µM PARP1 was incubated with 2.5 µM DNA-gap in a buffer consisting of 200 mM NaCl, 300 mM urea, 25 mM HEPES-NaOH pH 7.5, and 1 mM DTT. Samples were equilibrated for 1 min, and then the PARP1 activation was initiated by the addition of NAD⁺ to a final concentration of 1 mM. R_h measurement was performed every 3 min after the PARylation reaction initiation for 30 min at 30 °C. Next, 5–15 mM Mg²⁺, 0.5–2.0 mM Mn²⁺, 1–8 mM Ca²⁺, 1–9 mM Spd³⁺, 0.01–3.50 mM Spn⁴⁺, or 1–60 mM Put²⁺ was added to the samples, the samples were equilibrated for 1 min, and R_h was measured.

To determine PARP1 hydrodynamic size during its modifications in the presence of a cation, 2.5 μM PARP1 was incubated with 2.5 μM DNA-gap in a buffer consisting of 200 mM NaCl, 300 mM urea, 25 mM HEPES-NaOH pH 7.5, and 1 mM DTT in the presence of 3 mM Mn²⁺, 10 mM Ca²⁺, 3 mM Spn⁴⁺, 13 mM Spd³⁺, or 15 mM Mg²⁺. The samples were equilibrated for 1 min, and then the PARP1 activation was initiated by the addition of NAD⁺ to a final concentration of 1 mM. R_h measurement was performed every 3 min after the PARylation reaction initiation for 40 min.

For analysis of PAR hydrodynamic size in the presence or absence of a cation, 50 μM ADP-ribose was incubated in a buffer consisting of 200 mM NaCl, 20 mM HEPES-NaOH pH 7.5, and 1 mM DTT in the presence of 1–10 mM Mn²⁺, 3–21 mM Ca²⁺, 1–9 mM Spn⁴⁺, or 1–30 mM Mg²⁺. The samples were equilibrated for 1 min, and then R_h measurement was performed.

To disrupt PAR or modified PARP1 assemblies stabilized by a cation, EDTA to a final concentration of 10–60 mM was added as indicated in figure legends, and R_h was measured in the EDTA-treated samples.

To disrupt modified PARP1 assemblies stabilized by 15 mM Mg²⁺, PARG to a final concentration of 60 nM was added. After that, the reaction mixture was incubated at 4 °C overnight, and R_h was measured in the PARG-treated samples.

To determine PARP1 hydrodynamic size during its modifications in the presence of repair proteins, 2.5 μM PARP1 was incubated with 1.0 μM DNA-gap in a buffer consisting of 140 mM NaCl, 25 mM HEPES-NaOH pH 7.5, and 1 mM DTT in the presence of 0.1 µM Pol β, 0.1 µM XRCC1, 0.1 µM Lig III, and 5 or 15 mM Mg²⁺ as indicated in figure legends. The samples were equilibrated for 1 min, and then the PARP1 activation was initiated by the addition of NAD⁺ to a final concentration of 1 mM. R_h measurement was performed for 20 min every 2 min after the PARylation reaction initiation.

DNA polymerase β and Lig III activity assays

A 30-bp DNA duplex containing a one-nucleotide gap was used as the substrate in the DNA synthesis assay. Standard reaction mixtures (20 µl) consisted of 125 mM NaCl, 25 mM HEPES-NaOH pH 7.5, 1 mM DTT, 1.0 μM ³²P-labeled DNA-gap, 0.1 µM Pol β, 0.1 µM XRCC1, 1 mM NAD⁺, 100 µM four dNTPs, 5 or 15 mM Mg²⁺, and 2.5 μM PARP1 as indicated in figure legends. The reactions were initiated by the addition of the ³²P-labeled DNA-gap to a final concentration of 2 µM followed by incubation at 30 °C for 5–30 min. Aliquots (5 µl) were taken at 5, 15, and 30 min. The reaction was stopped by the addition of a 2-fold volume of the following solution: 90% of formamide, 10 mM EDTA, 0.1% of bromophenol blue, and 0.1% of xylene cyanol.

A 30-bp DNA duplex containing a one-nucleotide gap or a single nick served as the substrate in the DNA-nick-sealing assay. Standard reaction mixtures (20 µl) consisted of 125 mM NaCl, 25 mM HEPES-NaOH pH 7.5, 1 mM DTT, 1 μM ³²P-labeled DNA-nick, 0.1 µM Pol β, 0.1 µM Lig III, 0.1 µM XRCC1, 100 µM four dNTPs, 1 mM NAD⁺, 1 mM ATP, 5 or 15 mM Mg²⁺, and 2.5 μM PARP1 as indicated in figure legends. The reactions were initiated by the introduction of ³²P-labeled DNA-gap or DNA-nick to a final concentration of 2 µM, followed by incubation at 30 °C for 5–20 min. Aliquots (5 µl) were taken at 5, 10, and 20 min. The reaction was stopped by the addition of a 2-fold volume of the following solution: 90% of formamide, 10 mM EDTA, 0.1% of bromophenol blue, and 0.1% of xylene cyanol.

The mixtures were heated at 95 °C for 3 min, and the products were separated by denaturing electrophoresis in a 20% polyacrylamide gel. The gels were dried and subjected to phosphoimaging for quantification using a Typhoon FLA9500 imager (GE Healthcare, UK) and software (Quantity One, Bio-Rad, USA).

Statistics and reproducibility

DLS analysis: data shown in a histogram are representative of the average hydrodynamic radius (R_h) calculated from at least three experiments. The Zetasizer Nano ZS software was used to analyze the acquired correlation function and to derive the translational diffusion coefficient (D). Assuming a spherical shape of particles, the hydrodynamic radius (R_h) of the particles was calculated via the Stokes–Einstein equation: R_h = kT/6πηD, where k is Boltzmann’s constant, T is absolute temperature, and η is the viscosity of the solvent, set here to the viscosity of water with buffer components at 30 °C (0.8998 cP).

Turbidity analysis: data shown in a histogram are representative of the average OD₆₀₀ or OD₃₅₀ values determined from at least three experiments. The experiments were conducted in triplicate, and the mean ± SD of measured values were calculated.

PAGE analysis: bands of [³²P]PAR-labeled PARP1, a [³²P]labeled extended primer, and a ligation product were quantified by means of the Quantity One Basic software (Bio-Rad). The relative level of PARP1 PARylation in the presence of a cation was normalized to the level of PARP1 automodification lasting for 30 min in the absence of cations. The experiments were conducted three to four times; a histogram presents means ± SD of three to four independent experiments. The relative level of PARylated PARP1 hydrolysis in the presence of PARG was normalized to the level of PARP1 automodification observed after 40 min in the absence of cations and PARG. The experiments were conducted three times; the graphs present means ± SD of three independent experiments. The significance of differences between levels of PARP1 PARylation or hydrolysis was assessed by the two-tailed t-test at p < 0.05. The yield of an extended primer and ligation product was calculated as the percentage of the total radioactivity of all the ³²P-containing DNA bands in a lane. The significance of differences between the yields of products was evaluated by the two-tailed t-test at p < 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

O.I.L. was involved in planning and supervised the work, M.V.S. and R.O.A. designed and conducted the experiments, processed the experimental data, drafted the manuscript, and prepared the figures, E.A.M. and M.M.K. expressed and purified all proteins. All authors discussed the results and contributed to the final manuscript.

Peer review

Peer review information

Communications Biology thanks Judith Mine-Hattab and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Joanna Timmins and Johannes Stortz.

Data availability

All data necessary to reproduce our results are included in this published article (and its Supplementary Information and Supplementary Data 1 and Data 2 files).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-024-06811-4.