Deciphering the dark side of histone ADP-ribosylation: what structural features of damaged nucleosome regulate the activities of PARP1 and PARP2

Abstract

Poly(ADP-ribose) polymerases are critical enzymes contributing to regulation of numerous cellular processes, including DNA repair and chromatin remodelling. Within the PARP family, PARP1 and PARP2 primarily facilitate PARylation in the nucleus, particularly responding to genotoxic stress. The activity of PARPs is influenced by the nature of DNA damage and multiple protein partners, with HPF1 being the important one. Forming a joint active site with PARP1/PARP2, HPF1 contributes to histone PARylation and subsequent chromatin relaxation during genotoxic stress events. This study elucidates interrelation between the presence and location of a one-nucleotide gap within the nucleosome core particle (NCP) and PARP activities in automodification and heteromodification of histones. Utilizing a combination of classical biochemical methods with fluorescence-based technique and a single-molecule mass photometry approach, we have shown that the NCP architecture impacts the efficiency and pattern of histone ADP-ribosylation and binding to the histones-associated damaged DNA more significantly for PARP2 than for PARP1. Analysis based on existing studies of HPF1-dependent ADP-ribosylome and NCP structural dynamics allows to suggest that the DNA damage location and the conformational flexibility of histone tails modulated by post-translational modifications are crucial for delineating the distinct roles of PARP1 and PARP2 during genotoxic stress responses.

Graphical Abstract

Graphical Abstract.

Introduction

Poly(ADP-ribosyl)ation (PARylation) is a dynamic post-translational modification of biomolecules such as proteins and nucleic acids, catalysed by members of the PARP family. PARPs possess a conserved catalytic domain that facilitates the binding of nicotinamide adenine dinucleotide (NAD⁺) and the transfer of ADP-ribose moieties from the donor NAD⁺ to various amino acid residues on protein targets. PARPs can modify themselves (automodification) or other target molecules (heteromodification) [1–3].

Among the PARP family members, PARP1 and PARP2 are responsible for poly(ADP-ribose) (PAR) synthesis in the nucleus and play pivotal roles in regulating multiple cellular processes, including DNA repair, replication, and modulation of chromatin architecture [2–7]. Covalent PARylation of proteins can impact their function, localization, and stability. Furthermore, the highly negatively charged PAR polymer can serve as a molecular scaffold, recruiting and concentrating various proteins, such as DNA repair factors, chromatin remodellers, and transcription factors [6, 8–11]. Thus, PARPs can regulate functions of proteins both through the covalent PARylation and facilitation of their co-localization at damaged DNA. Being very similar in catalytic mechanisms but different in DNA binding activities and specificities PARP1 and PARP2 have both overlapping and specific functions in multiple DNA repair pathways [6, 7, 12, 13].

Despite extensive research on PARylation, many aspects of this process remain unclear. The initiation of PARylation involves the coordination of NAD⁺ and the acceptor amino acid of the target protein at the enzyme’s active site. The complexity of this coordination likely leads to hydrolysis of NAD⁺ and synthesis of free PAR chains before the successful initiation of covalent PARylation [14–18]. The efficiency of initiation depends on several factors, including the interaction of the acceptor protein with PARP, the flexibility of the modified region, and the nature of the target amino acid residues and their surrounding environment. Many studies have focused on the role of PAR- and DNA-mediated interactions or direct protein-protein interactions between PARP and its target proteins. The major mechanism of selective PARylation is constituted by PAR-binding domains in target proteins [19–22], but the interaction mediated by a specific DNA structure may also contribute to the substrate targeting [23]. However, the factors influencing specific selection of acceptor amino acids are not fully understood. It is believed that while a consensus sequence for the selection of PARylation sites may not be required, certain preferences exist. For instance, ADP-ribosylation of glutamate follows a proline-directed motif (PXE*, EP, PXXE) [24], while serine ADP-ribosylation, catalysed by PARP1/PARP2 in complex with histone PARylation factor 1 (HPF1), preferentially targets the KS motif [25, 26]. HPF1 complements the active site of PARP1/PARP2 and beyond its contribution to NAD⁺ binding and catalysis seems to recognize the region of substrate protein to be modified [27]. Thus, both the steric accessibility of PARylation sites and the primary structure of the target protein influence heteroPARylation, though the precise mechanisms remain to be clarified.

Core histones H2A, H2B, H3, and H4 are well-established substrates for HPF1-dependent ADP-ribosylation [28–31]. Proteomic studies demonstrate that histones are among the most abundant acceptors of Ser-linked ADP-ribosylation upon oxidative stress [31, 32]. This is likely associated with HPF1-dependent PARylation of histones at sites of DNA damage, which promotes chromatin relaxation [5, 33]. In this context, it is important to highlight the distinct roles of PARP1 and PARP2. In vivo experiments with PARP1 depletion show that PARP1 mainly contributes to stress-induced nuclear ADP-ribosylation, and in its absence histone PARylation and chromatin relaxation are significantly reduced [33, 34]. The deletion of PARP1 in conjunction with HPF1 produced much more significant effects on cell survival following DNA damage, compared to those detected for HPF^{−/ −}/PARP2^{−/ −} cells, and HPF1-deficient cells maintained poly-ADP-ribosylation sufficient for XRCC1-mediated repair of single-strand breaks [35]. Thus, the predominant role of PARP1 in initiating the DNA damage response is clear, while the role of PARP2 appears less significant. On the other hand, PARP1 and its catalytic activities facilitate the enrichment of PARP2 at DNA damage foci [36, 37]. Therefore, the reduction in histone PARylation in the absence of PARP1 may, among other factors, be linked to impaired recruitment of PARP2 to DNA damage sites. In other words, there is a potential gap in our understanding of the specific role of PARP2 in the genotoxic stress response. Interestingly, in vitro studies indicate that PARP2 in the complex with HPF1 modifies histones more efficiently than PARP1 does and is specifically activated by base excision repair (BER) related damage in the nucleosome core particle (NCP) [16, 38]. In the current study, we further compare the roles of PARP1 and PARP2 in HPF1-dependent histone modification.

The structured histone domains serve as a foundation for the assembly of stable NCP, while the flexible histone tails at N- and C-termini play a crucial role in regulating interactions between the nucleosome and various factors [30]. Furthermore, the histone tails are critically involved in the modulation of histone-DNA interactions (i.e. nucleosome dynamics) via multiple post-translational modifications [10, 39–42]. Consequently, the well-established sites of ADP-ribosylation and other post-translational modifications are predominantly found within the histone tails.

This research specifically investigates how the architecture of the PARP-NCP complex influences the HPF1-dependent PARylation of histones. The use of NCP assembled in vitro on the 603 nucleosome positioning sequence characterized by Lowary and Widom [43] provides an advantageous model for studying heteroPARylation. The known dynamics of histone tails within the NCP, coupled with the precise positioning of DNA on the histone core facilitated by the Widom 603 sequence, offer clarity in the experimental setup. Furthermore, the position of a DNA damage (one-nucleotide gap intermediate of BER) shown in our previous research to influence activities of PARP1/PARP2 in the auto- and heteromodification reactions [38] can be precisely determined in the compact NCP structure. The present results highlight dependence of the extent and pattern of histone ADP-ribosylation on the presence of BER-specific DNA damage in distinct superhelical locations [SHLs; number of double-helical turns from the central base-pair at the particle dyad axis [44, 45] or outside the NCP. PARP1 and PARP2 revealed different sensitivity to the structural features of NCP, further suggesting their specific roles in DNA damage response and chromatin remodelling.

Materials and methods

Materials

Natural core histones were isolated from Gallus gallus erythrocytes and purified as described in the published protocol [46]. Recombinant non-tagged human APE1, human PARP1, murine PARP2 and C-terminally His-tagged human HPF1 were expressed and purified as detailed previously [16, 47, 48]. Homogeneity of purified histones, PARP1, PARP2, and HPF1 was verified by SDS-polyacrylamide gel (PAG) electrophoresis (Supplementary Fig. S1). Escherichia coli uracil-DNA glycosylase (UDG) was from Biosan (Novosibirsk, Russia). The recombinant bovine poly(ADP-ribose) glycohydrolase (PARG) generously provided by E. Ilina (ICBFM, Novosibirsk, Russia) was purified as described previously [49]. Recombinant polymerase RB69 and S. pyogenes Cas9 kindly provided by D. Zharkov (ICBFM, Novosibirsk, Russia) were purified as described previously [50, 51]. DNase I was purchased from Thermo Scientific (USA). The pGEM-3z/603 plasmid vector (Addgene plasmid #26 658; http://n2t.net/addgene:26658; RRID: Addgene_26 658) for PCR synthesis of DNAs containing the 147 base pair Widom 603 sequence was a gift from J. Widom. DNA primers were synthesized by Lumiprobe (Moscow, Russia). pBR322 (Addgene plasmid #1979) was from SibEnzyme (Novosibirsk, Russia). The ³²P-labelled NAD⁺ was synthesized enzymatically following a described method [52], using [α-³²P]ATP with specific activity of 3000 Ci/mmol, synthesized in Laboratory of Biotechnology (ICBFM, Novosibirsk, Russia). NAD⁺, reagents for electrophoresis and basic components of buffers were purchased from Sigma–Aldrich (USA).

Methods

Preparation of DNA and nucleosome constructs

DNA constructs of different length containing the 147-bp core of the 603 nucleosome positioning sequence [43] were created by PCR from the pGEM-3z/603 plasmid as shown schematically in Supplementary Fig. S2, using oligonucleotide primers specified for each DNA (Table 1).

Table 1.

Sequences of forward and reverse primers used for construction of DNAs^a

DNA147 and gap12-DNA147	For: 5′-ACCCCAGGGACTTGAAGTAATAAGG-3′
	Rev: 5′-CCCAGTTCGCG[dU]GCCCACCTACCG[T-FAM]GTGAAG-3’
gap35-DNA147	For: 5′-ACCCCAGGGACTTGAAGTAATAAGG-3′
	Rev: 5′-CCCAGTTCGCGCGCCCACCTACCG[T-FAM]GTGAAGTCG[dU]CACTCGG-3′
pDNA147	For: 5′-ACCCCAGGGACTTGAAGTAATAAGG-3′
	Rev: 5′-pCCCAGTTCGCGTGCCCACCTACCG[T-FAM]GTGAAG-3’
DNA167 and gap12-DNA167	For: 5‘-CGAAACGGGTACCCCAGGG-3‘
	Rev: 5′-CTCTCGGGTGCCCAGTTCGCG[dU]GCCCACCTACCG[T-FAM]GTGAAG-3’
gap35-DNA167	For: 5‘-CGAAACGGGTACCCCAGGG-3’
	Rev: 5′-CTCTCGGGTGCCCAGTTCGCGCGCCCACCTACCG[T-FAM]GTGAAGTCG[dU]CACTCGG-3′
DNA177 and gap12-DNA177	For: 5‘-CGAAACGGGTACCCCAGGG-3‘
	Rev: 5′-ATAATCGACACTCTCGGGTGCCCAGTTCGCG[dU]GCCCACCTACCG[T-FAM]GTGAAG-3’
gap35-DNA177	For: 5‘-CGAAACGGGTACCCCAGGG-3’
	Rev: 5′-ATAATCGACACTCTCGGGTGCCCAGTTCGCGCGCCCACCTACCG[T-FAM]GTGAAGTCG
	[dU]CACTCGG-3′
DNA267 and gap-DNA267	For: 5′-FAM-CCTCTAGAGTCGGGAGCTCGG-3’
	Rev: 5′-ACACGAATAGGCGTTTTCCTAG[dU]ACAAATCACCC-3’
pDNA267	For: 5′-FAM-CCTCTAGAGTCGGGAGCTCGG-3’
	Rev: 5′-pACACGAATAGGCGTTTTCCTAGTACAAATCACCCCAGCG-3’

^a5’-Terminal phosphate, dU and FAM-labelled T residues are shown in bold; the linker sequences are underlined.

To introduce a single uracil residue at the predetermined position from the dyad (centre) of the 603 sequence, the respective dU-containing reverse primers were used. To assemble nucleosomes, DNA and core histones were mixed in the high-salted buffer containing 2 M NaCl, dialyzed against the NaCl concentration gradient from 2 M to 250 mM during 6 h at 4°C and then against the buffer with 10 mM NaCl overnight at 4°C with gentle stirring. The concentrations of DNA and histones were optimized in a preliminary quick-time experiment as described in the published protocol [46]. DNA (both naked and nucleosome-associated) containing a one-nucleotide gap (gap-DNA, gap-NCP) was prepared in a two-step procedure (Supplementary Fig. S3A). First, the AP site was generated by incubation of a dU-containing DNA or NCP with UDG (1 activity unit per 0.6 pmol of DNA) for 30 min at 37°C. Second, the produced AP site-containing DNA or NCP (3.5 μM) was incubated with APE1 (10 nM for DNA and 50 nM for NCP) for 15 min at 37°C. The extent of AP site cleavage was controlled by incubation of the reaction probe in the presence of 15 mM EDTA and 20 mM methoxyamine for 30 min on ice followed by probe heating for 5 min at 97°C and subsequent separation by electrophoresis on a 10% polyacrylamide gel under denaturing conditions (Supplementary Fig. S3B). The final preparation contained mainly the one-nucleotide gapped DNA generated from the product of AP site hydrolysis due to spontaneous cleaving off the 5′-deoxyribose phosphate (5′-dRp) residue and 5′-dRp-lyase activities of APE1 and histones [53, 54]; the remaining 5′-dRp residue was removed upon subsequent incubation with PARP1/PARP2 exhibiting the 5′-dRp-lyase activity [55] (Supplementary Fig. S3C). The nucleosome assembly efficiency was analysed by the electrophoretic mobility shift assay on a 5% non-denaturing PAG (Supplementary Fig. S3D).

Testing of PARP activity in the poly(ADP-ribose) synthesis

Catalysed by PARP1 and PARP2 autopoly(ADP-ribosyl)ation (autoPARylation) and covalent labelling of histones were carried out in a standard 10 μL reaction mixture containing 50 mM Tris-HCl, pH 8.0, 50 mM NaCl, 5 mM MgCl₂, 1 μM [³²P]NAD⁺, 250 nM DNA/NCP (or gap-DNA/gap-NCP), 500 nM PARP1 (or PARP2), and 1 μM HPF1. The reaction was initiated by adding [³²P]NAD⁺ to a protein-DNA mixture preassembled on ice. After incubating the mixtures at 37°C for 45 min, the reactions were terminated by the addition of SDS-PAGE sample buffer and heating for 3 min at 95°C. To perform PAR hydrolysis, the ADP-ribosylation reaction was stopped by addition of 1 μM olaparib and 10 mM EDTA, and the mixture was further incubated with 1 μM PARG for 1 h at 37°C before addition of SDS-PAGE sample buffer. The reaction products were separated by 20% SDS-PAGE (a ratio between acrylamide and bis-acrylamide of 99:1); bands of proteins labelled with [³²P]ADP-ribose were analysed by using the Typhoon imaging system (GE Healthcare Life Sciences) and Quantity One Basic Software (Bio-Rad). The radiolabelled signals of modified proteins were quantified as follows: the total (raw) signal of a smeared band of modified protein (indicated for each protein in autoradiograms) was quantified and the same-size background signal of gel in the respective lane was subtracted from the raw signal. To calculate the absolute amount of [³²P]ADP-ribose in the radiolabelled signals, the exact amounts of [³²P]NAD⁺ spotted on the Whatman paper filters were imaged and quantified in parallel with each gel. The quantitative data presented in histograms were obtained in at least three independent experiments. Absence of DNA contamination in PARP1, PARP2 and HPF1 preparations was checked as described previously [16] by measuring the basal (without addition of DNA) activities of the enzymes in the absence and presence of HPF1, with and without DNase treatment (Supplementary Fig. S4).

Fluorescence studies of PARP1/PARP2 interaction with DNAs/NCPs

In direct titration experiments, fluorescence anisotropy measurements of FAM-labelled DNAs (free or nucleosome-associated) were performed in the absence and presence of various concentrations of the potential protein partner [56]. A mixture containing 3 nM 25-FAM-labelled DNA (or NCP) and 1–60 nM PARP1 (or PARP2) in a binding buffer (50 mM NaCl, 50 mM Tris–HCl, pH 8.0, 5 mM MgCl₂ and 5 mM DTT) was prepared on ice in a 384-well plate and incubated at room temperature for 10 min. In competition binding experiments, reaction mixtures contained 50 mM NaCl, 50 mM Tris-HCl (pH 8.0), 5 mM DTT, 5 mM MgCl₂, 10 nM DNA, 5 nM PARP1, and increasing concentrations of plasmid DNA pBR322. The fluorescent probes were excited at 482 nm (482–16 filter plus dichroic filter LP504), and the fluorescence intensities were detected at 530 nm (530–40 filtre) to measure the fluorescence anisotropy of FAM. Each measurement consisted of 50 flashes per well, and the resulting values of fluorescence anisotropy were automatically averaged. The measurements in each well were done three times with intervals of 1 min.

The average values were used for the final plot, and EC₅₀ values were determined with the MARS Data Analysis software (BMG LABTECH). The data were plotted (F vs C) and fitted by four-parameter logistic equation: F = F₀ + (F_∞ - F₀)/[1 + (EC₅₀/C)ⁿ], where F is the measured fluorescence anisotropy of a solution containing the labelled DNA at a given concentration (C) of PARP1 (or PARP2), F₀ is the fluorescence anisotropy of solution of the labelled DNA/NCP alone, F_∞ is the fluorescence anisotropy of the labelled DNA/NCP saturated with the protein, EC₅₀ is the protein concentration at which F - F₀ = (F_∞ - F₀)/2, and n is the Hill coefficient, which denotes the slope of the nonlinear curve. CC₅₀ values were determined by fitting average F values by four-parameter logistic equation: F = F₀ + (F_∞ - F₀)/[1 + (CC₅₀/C)], where F is the measured fluorescence anisotropy of a solution containing the preformed complex of labelled DNA with PARP1/PARP2 at a given concentration (C) of pBR322, F₀ is the fluorescence anisotropy of the labelled DNA•PARP1/PARP2 complex alone, F_∞ is the fluorescence anisotropy of the free labelled DNA, CC₅₀ is the concentration of the competitive plasmid pBR322 at which F - F₀ = (F_∞ - F₀)/2. Binding curves presented below and in Supplementary Data show the best fits of the respective equation, with R² values matching or exceeding 0.97.

Mass photometry measurements of PARP1/PARP2 interaction with NCPs

Formation of PARP1 and PARP2 complexes with NCP147 and gap35-NCP147 was carried out in a 20 μL reaction mixture containing 50 mM Tris-HCl, pH 8.0, 50 mM NaCl, 5 mM MgCl₂, 3 nM nucleosome (NCP147 or gap35-NCP147), 5 − 6 nM PARP1 or PARP2, and 10 nM HPF1 (when indicated). The buffer was preliminary ultrafiltered by using Sartorius vivaspin columns with 1 000 kDa pore size PES membrane filter. The reaction mixture was prepared on ice, centrifuged for 10 min at 10 000 rpm (4°C), and incubated at room temperature for 10 min. MP measurements were performed using a OneMP mass photometer (Refeyn, UK). Following the standard protocol [57], 18 μL buffer was loaded to the sample chamber and the objective was focused by using autofocus function, then 2 μL sample was added to the chamber and mixed by pipetting. Data were collected for one minute using the AcquireMP software (Refeyn, UK). All samples were measured using the expanded detection area. The MP signals were calibrated using BSA (69 kDa) (Sigma–Aldrich, USA), recombinant RB69 protein (107 kDa) and recombinant Cas9 protein (158 kDa). MP data were processed with the DiscoverMP software (Refeyn, UK) to calculate relative molecular species populations from the areas of the Gaussian peaks (representing free NCP and its complexes with one or two PARP molecules). The number of counts for each species was estimated from three independent experiments to obtain the average species concentration fractions. These data were further processed with program “K_d calculation” as described previously [57, 58].

Electrophoretic mobility shift assay

To detect binding of PARP1/PARP2 to nucleosome, an electrophoretic mobility shift assay (EMSA) was used. The protein (400 nM) was incubated with FAM-labelled NCP147/gap12-NCP147/gap35-NCP147 (200 nM) in a 10 μL mixture containing 50 mM Tris-HCl, pH 8.0, 50 mM NaCl, 5 mM MgCl₂ and 1 mM DTT at room temperature for 10 min. After the addition of Ficoll and bromophenol blue (to a final concentration of 5% and 0.1%, respectively), the incubation mixtures were electrophoresed at 4°C on 5% non-denaturing PAG in a 30 mM Tris-Borate-EDTA buffer. Gels were imaged on a Typhoon FLA 9500 in the FAM channel. Free and complexed NCP bands were quantified using Quantity One Basic software. The extent of NCP binding presented in histograms was calculated as a portion of NCP assigned to its complex(es) with protein.

Results

The DNA damage location in the NCP affects HPF1-dependent ADP-ribosylation of histones

The efficiency of PARP1 and PARP2 catalysed histone PARylation was revealed in our previous study to depend on the presence of one-nucleotide gap in one strand of the nucleosomal DNA [38]. This prompted us to further explore possible influence of location of the BER-specific DNA damage relative to the DNA blunt ends and the nucleosome core histones on the activities of these enzymes in auto- and heteromodification reactions. The sequence of DNA clone 603, comprising 147 bp, provides precise positioning of DNA within the nucleosome [43], enabling to study the impact of the DNA damage position. The interaction of PARP1 and PARP2 enzymes with NCP was previously shown to be sensitive to the DNA damage orientation (inwards or outwards) within NCP [59]. To ensure the outwards orientation, we constructed NCPs containing uracil (which can be removed enzymatically to create the one-nucleotide gap as shown in Supplementary Fig. S2A) at positions 12 and 35 relative to the 603 sequence boundary, at superhelical locations 5.5 and 3.5, respectively. Seven NCP variants used in the experiments (Fig. 1A) differ from each other by the presence of 5′-phosphate, one-nucleotide gap at the two different positions, and 50/70 bp linkers without or with one-nucleotide gap.

Figure 1.

HPF1-dependent ADP-ribosylation of histones is influenced by the NCP structure. (A) Schemes of nucleosome structures. (B) Autoradiograms show covalent binding of ³²P-labelled ADPR to PARP1/PARP2 and histones after incubation of each PARP (500 nM) with [³²P]NAD⁺ (1 μM), HPF1 (1 μM) and a defined NCP variant (250 nM). Positions of ADP-ribosylated proteins and their native forms (and molecular weight markers) in 20% SDS-PAG are indicated on the left and right of the autoradiogram. (C and D) Histograms show the amount of ADPR attached to PARP1/PARP2 and histones in the distinct samples (the mean ± SD, n= 3). Statistically significant differences in the levels of histone PARylation in the presence of different NCPs (NCP147 versus pNCP147, gap12-NCP147 versus gap35-NCP147; NCP267 versus gap-NCP267) are marked: P< 0.01 (**), P< 0.001 (***); t-test..

Recombinant non-tagged human PARP1 and mouse PARP2 were used to catalyse DNA-dependent ADP-ribosylation. The mouse PARP2 is highly homologous to the two isoforms of human PARP2: the critical for function WGR and catalytic domains [60] show about 90% of sequence identity (Supplementary Fig. S5). Furthermore, the mouse enzyme is active in functional interactions with its protein partners of human origin [61, 62]. To assemble nucleosomes, we used natural chicken core histones, whose comparison with human histones shows that known ADP-ribosylation sites belong to strictly conserved amino acid sequences (Supplementary Fig. S6). PARP1 and PARP2 were activated by incubation with different NCP variants in the presence of [³²P]NAD⁺ and HPF1, and the reaction products were separated by SDS-PAGE (Fig. 1B) to compare the levels of automodification and heteromodification of histones with ADP-ribose (ADPR). Concentrations of proteins used in these experiments were shown previously to be optimal for HPF1-induced stimulation of PARP1/PARP2 activities [16]. Furthermore, a significant extent of HPF1-dependent switch of the reaction specificity was detected in these conditions: 80/88% of PARP1/PARP2 automodification was stable under treatment with hydroxylamine. We have to emphasize that low concentrations of NAD⁺ were utilized in order to prevent the formation of long PAR chains, which can significantly change the electrophoretic mobility of modified proteins. The primary target for modification catalysed by PARP1 was the enzyme itself, irrespective of the NCP structure used. The amount of ADPR bound to PARP1 in all samples was 1.3–4-fold higher compared to that bound to histones (Fig. 1C). An increased level of histone modification was detected when the nucleosomal DNA contained 5′-phosphate or one-nucleotide gap (pNCP147, gap12-NCP147, and gap35-NCP147 versus NCP147), with the highest yield of heteromodification reaction being observed in the presence of NCP bearing the gap at position 35. The addition of 50 and 70 bp linkers (without or with 5′-phosphate or gap) to NCP147 (NCP267, pNCP267, gap-NCP267) produced insignificant effects on the total levels of auto- and heteromodification, but noticeably decreased the length of PAR attached to histones as shown by the more uniform mixture of modified histone molecules with the increased electrophoretic mobility (Fig. 1B, lanes 5–7). The data suggest that the presence of linkers modulates the interaction mode of PARP1 with NCP (due to an increased distance between the DNA end and nucleosome core or changes in the entry-exit region of NCP), which in turn influences the enzyme activity in heteromodification reaction at the elongation step.

In contrast to PARP1, PARP2 was more active in modification of histones than in automodification as evidenced by 2.5–10-fold higher yields of the heteromodification reaction (Fig. 1D). The ratio between the levels of auto- and heteromodification reactions as well as the length of PAR attached to histones were highly dependent on the NCP structure used for PARP2 activation. The presence of 5′-phosphate or one-nucleotide gap in the nucleosomal DNA favoured the modification of histones. The histone modification detected when NCP contained the gap at position 35 was the most efficient and uniform in the length of PAR. Thus, PARP2-catalysed modification of histones depends on the position of DNA damage. No significant impact of 50 and 70 bp linkers on the relative levels of automodification and histone modification was detected (compare data for NCP147 vs NCP267 and pNCP147 vs pNCP267). However, the introduction of gap into the DNA linker, 47 bp away from the histone core, significantly reduced the level of histone modification. Such an effect can result from primary binding of PARP2 outside the nucleosome core due its higher affinity for gap than for undamaged DNA and DNA ends [63, 64].

ADP-ribosylation experiments performed in the absence of HPF1 revealed no detectable modification of histones catalysed by PARP1/PARP2 (Supplementary Fig. S7). Unexpectedly, the automodification of PARP2 was found to depend significantly on the NCP structure: both the total level of ADP-ribosylation and a set of products of different molecular weights (resolved as discrete bands) varied, depending on the presence of gap and its position as well as on the presence of 5′-phosphate. Similar data were obtained when HPF1-independent automodification of PARP2 was compared in the presence of structural variants of naked DNA (Supplementary Fig. S8B and D). However, no influence of DNA structure on the PARP2 activity was detected in the presence of HPF1, which produced a stimulatory action on the initiation step of the automodification reaction as evidenced from accumulation of the mono-ADP-ribosylated protein (Supplementary Fig. S8B and D). Testing of PARP1 activity under similar conditions revealed no significant influence of the DNA/NCP structure (Supplementary Figs S7 and S8A and C). The difference between PARP1 and PARP2 revealed in these experiments is interesting for further study.

To examine possible impact of NCP structure on the balance between automodification and heteromodification reactions at the initiation step, we performed ADP-ribosylation experiments followed with PAR hydrolysis catalysed by poly(ADP-ribose) glycohydrolase (PARG) known to produce mono-ADP-ribosylated proteins due to exoglycosidase and endoglycosidase activity [65]. No significant change in distribution of mono-ADP-ribose signal between PARP1/PARP2 and histone octamer induced by the presence of 5′-phoshate or gap at position 12 was detected (Fig. 2).

Figure 2.

Influence of NCP structure on histone modification at the initiation step. Autoradiograms (A and B) show covalent binding of ³²P-labelled ADPR to PARP1/PARP2 and histones after incubation of each PARP (500 nM) with [³²P]NAD⁺ (1 μM), HPF1 (1 μM) and a defined NCP variant (250 nM), and subsequent PARG-catalysed hydrolysis. Positions of ADP-ribosylated proteins in 20% SDS-PAG indicated on the left of the autoradiograms correspond to those of native forms; gel positions of coomassie-stained histones are shown. Bands migrated below the major band of modified PARP1 represent ADP-ribosylated PARP1 fragments (minor admixtures not visualized by coomassie staining). Histograms (C and D) show distribution of covalently bound ADPR between PARP and all histones in the distinct samples (the mean ± SD, n= 3). Statistically significant differences in the relative amounts of ADPR covalently bound to histones in gap12-NCP147 and gap35-NCP147 are marked: P< 0.05 (*), P< 0.01 (**); t-test.

However, the initiation of histone ADP-ribosylation by both PARP1 and PARP2 was favoured by the presence of gap at position 35. The addition of 50 and 70 bp linkers changed the balance in favour of automodification in reactions catalysed by PARP1. The increased PARP1 activity in initiation of the automodification may result from the enzyme activation at DNA blunt ends located far away from the histone core. In the case of PARP2, no significant change in the balance due to the presence of linkers was detected. The combined results of ADP-ribosylation experiments performed without and with PARG treatment indicate that distinct structural features of nucleosome-associated DNA may change the efficiency of histone ADP-ribosylation differently at the initiation and elongation step.

Core histones in close proximity to the DNA damage are main targets of PARP2-catalysed modification

As we have found, the efficiency of PARP1/PARP2-catalysed histone modification depends on the location of one-nucleotide gap in the NCP structure. It was interesting to clarify whether this dependence is related to selective modification of distinct core histones. The HPF1-dependent PARylation reaction catalysed by PARP1/PARP2 was performed in the presence of gap12-NCP147 or gap35-NCP147, which revealed impact of the DNA damage position on the balance between auto- and heteromodification reactions catalysed by PARP1/PARP2. Then ADP-ribose oligomers and polymers were degraded with PARG to produce mono-ADP-ribosylated proteins. Mono-ADP-ribosylation does not change the electrophoretic mobility of proteins in SDS-PAG, enabling their identification. The data obtained indicate that all four core histones are targets of PARP1-catalysed modification and the pattern of histone ADP-ribosylation depends to a little extent on the NCP structure (Fig. 3A). Nevertheless, an increased level of H2A modification accompanied with a decreased level of H3 modification in gap35-NCP147 (vs NCP147 and gap12-NCP147) was revealed to be statistically significant (P < 0.01, n= 3, t-test). Such an effect suggests contribution of PARP1 interaction with DNA gap at position 35 to histone modification.

Figure 3.

Dependence of histone ADP-ribosylation pattern on the NCP structure. Autoradiograms show covalent binding of ³²P-labelled ADPR to histones after incubation of 500 nM PARP1 (A) or PARP2 (B) with [³²P]NAD⁺ (1 μM), HPF1 (1 μM) and NCP/gap12-NCP/gap35-NCP (250 nM) with following PARG-catalysed hydrolysis and separation of products in 20% SDS-PAG; gel positions of coomassie-stained histones superposed to their labelled products are indicated on the left of the autoradiograms. Histograms show distribution of total covalently bound ADPR between the four histones (the mean ± SD, n= 3). The bottom schemes show which histones within different NCP structures are main targets of ADP-ribosylation.

In the case of PARP2 the histone modification patterns obtained for NCP147 and gap12-NCP147 were very similar to each other but different from that obtained for gap35-NCP147 (Fig. 3B). In NCP and gap12-NCP, H2B and H4 histones were main acceptors of ADPR, while H2A and H2B histones were primarily modified in gap35-NCP structure. Importantly, known ADP-ribosylation sites on histones H2A and H2B tails (Ser2 and Ser6/Ser14, respectively) are localized close to nucleotide 35 (SHL 3.5), while histones H3 and H4 tails comprising known ADP-ribosylation sites (Ser10/Ser28 and Ser2, respectively) are localized nearer to nucleotide 12 (SHL 5.5) [30, 66]. Combined, our results suggest that selectivity of HPF1-dependent histone ADP-ribosylation catalysed by PARP2 is provided primarily by the enzyme interaction with the one-nucleotide gap and its location relative to the nucleosome core.

To further examine whether the distance between the DNA end and the gap position may impact the efficiency and pattern of histone ADP-ribosylation, we designed additional structural variants of gap12-NCP and gap35-NCP containing 10 or 20 bp linkers. The location of gap relative to the core histones remains fixed in two series of variants (gap12-NCP147, gap12-NCP167, gap12-NCP177 and gap35-NCP147, gap35-NCP167, gap35-NCP177), while the distance between the DNA blunt ends and gap varies. Control experiments with non-gapped variants, NCP147, NCP167 and NCP177, revealed no effects of linkers on the yields of PARP1/PARP2-catalysed auto- and heteromodification reactions (Fig. 4A and B, lanes 1–3). Similar data were obtained when the length variants of gap12-NCP or gap35-NCP were compared: the levels of histone modification catalysed by PARP1/PARP2 in the presence of all three variants of gap35-NCP were nearly equal to each other but higher compared to those in the presence of gap12-NCP variants (Fig. 4C and D).

Figure 4.

The location of BER-specific DNA damage within NCP determines the efficiency of histone ADP-ribosylation. (A and B) Autoradiograms show covalent binding of [³²P]ADPR to PARP1, PARP2 and histones after incubation of PARP1/PARP2 (500 nM) with [³²P]NAD⁺ (1 μM), HPF1 (1 μM) and NCP (250 nM) specified for each sample, and further separation of products in 20% SDS-PAG. Gel positions of PARylated proteins, their native forms and molecular weight markers are indicated on the left and right of the autoradiograms. (C and D) Histograms show the amount of ADPR attached to PARP1/PARP2 and histones in the samples (the mean ± SD, n= 3). Statistically significant differences in the levels of histone PARylation in three variants of gap12-NCP compared to those of gap35-NCP are marked: P< 0.01 (**); t-test.

Further ADP-ribosylation experiments followed with PARG-catalysed hydrolysis revealed the patterns of histone ADP-ribosylation catalysed by PARP1/PARP2 being very similar for the three length-variants of NCP/gap12-NCP/gap35-NCP [Supplementary Fig. S9A and B (lanes 1–3/4–6/7–9), C, and D], indicating no impact of 10/20 bp linkers on the selectivity of heteromodification reaction. Furthermore, the balance between the auto- and heteromodification reactions catalysed by PARP1/PARP2 was unchanged due to the addition of short linkers (Supplementary Fig. S9E and F).

Impact of DNA structure on the interaction of PARP1 and PARP2 with NCP

The dependence of histone ADP-ribosylation on the location of DNA damage raises question whether it is related to different affinities of PARP1/PARP2 for the NCP structural variants. The apparent equilibrium dissociation constants of these complexes were determined using fluorescence anisotropy measurements. The respective DNAs containing one-nucleotide gap at position 12 or 35 and their variants with linkers (Fig. 1A) were labelled with FAM fluorophore at position 25. Taking into account binding of PARP1/PARP2 to undamaged DNA with a lower affinity than to DNA lesions [63, 64, 67], the titration experiments were performed in restricted protein concentration range (up to 50–60 nM). In these conditions, contribution of PARP interaction with undamaged DNA to the apparent dissociation constant (EC₅₀ value) may be ignored. The EC₅₀ values determined for PARP1/PARP2 complexes with different ligands are summarized in Table 2. Their analysis shows that PARP1 binds all gapped DNA variants with very similar affinities and 1.8-fold stronger (P < 0.02) than the non-gapped DNA, suggesting contribution of gap as a binding site to the affinity. However, we revealed no statistically significant effect of gap presence on the strength of PARP1 binding to NCP ligands, that may result from a more compact structure of NCP compared to DNA.

Table 2.

Apparent equilibrium dissociation constants of PARP1/PARP2 complexes with DNA and NCP variants (EC₅₀^a, nM).

Ligand	PARP1	PARP2
DNA147, DNA167, DNA177b	3.6 ± 0.8	ndc
gap12-DNA147/167/177; gap35-DNA147/167/177b	2.0 ± 0.3	3.3 ± 1.1
NCP147, NCP167, NCP177b	3.5 ± 0.9	ndc
gap12-NCP147/167/177; gap35-NCP147/167/177b	3.7 ± 0.7	5.3 ± 1.4

^aValues are the mean (±SD) of independent measurements: n= 4 (or 3) for eacn length-variant of DNA (or NCP); n= 2 for each length-variant of gap12-DNA, gap35-DNA, gap12-NCP and gap35-NCP.

^bStructural variants combined in each of four groups revealed no statistically significant difference in the EC₅₀ values.

^cNot determined due to absence of saturation.

Binding of PARP2 to non-gapped DNA/NCP at the protein concentrations used in these experiments could not be quantified due to the absence of saturation (Supplementary Fig. S10). Similar to PARP1, PARP2 displayed practically identical affinities for all structural variants of gap-NCP, which were about 1.5-fold lower (P < 0.002) than respective values for PARP1. It should be noted that the lowest EC₅₀ value (2 nM) is near the minimal concentration of fluorescently labelled ligand we used to perform reliable measurements and therefore represents the upper limit. It is very likely that this method limitation results in the absence of significant difference between the gapped DNA variants in the EC₅₀ values.

The inability to distinguish the affinity of PARP1 for different gap-DNAs in direct titration experiments prompted us to perform further experiments using a competition binding assay. Plasmid DNA containing no breaks was added at increasing concentrations as the undamaged competitor to the preformed PARP1 complex with a defined DNA structural variant. Release of the fluorescently labelled DNA from the complex due to binding of the competitor is accompanied by a decrease in fluorescence anisotropy (Fig. 5A), enabling to calculate CC₅₀ value (directly proportional to stability of the preformed complex). These experiments revealed that the strength of PARP1-DNA interaction depends on the presence of gap and its position relative to the blunt ends of all length variants (Fig. 5B). Interestingly, very similar CC₅₀ values were determined for PARP1 complexes with gap12-DNA177 and gap35-DNA147, in which position of gap is separated from the DNA end by 32 and 35 nucleotides, respectively.

Figure 5.

PARP1 affinity for gap-containing DNA depends of the damage position relative to the blunt ends. (A) Typical titration curves reflecting dissociation of fluorescently labelled DNA147 or gap35-DNA167 from the complex with PARP1 upon addition of increasing concentrations of the competitor DNA (plasmid pBR322); (B) CC₅₀ values determined for different DNA structural variants. Statistically significant differences between the CC₅₀ values are shown: P < 0.01 (*), P < 0.001 (**); n= 3, t-test.

Using direct fluorescence titration experiments, we revealed only a slight difference between PARP1 and PARP2 in the affinities for gapped NCPs (Table 2). However, their complexes can differ in composition due to propensity of both enzymes for self-association. To test this assumption, we performed mass photometry (MP) experiments. This novel technique measures the mass of individual molecules in solution and allows to obtain molecular mass distribution of proteins and other biomolecules reflecting the molecular composition of the sample and to measure binding affinity [58, 68].

MP experiments with individual components revealed coexistence of monomeric and dimeric forms of PARP1/PARP2 and coexistence of NCP147 with free DNA147 (Fig. 6A). PARP2 detected as more prone to dimerization compared to PARP1 was capable even of trimerization. PARP1 was revealed to form complexes with NCP147/gap35-NCP147 containing one or two protein molecules, PARP1•NCP (320 ± 23 kDa) and (PARP1)₂•NCP (446 ± 28 kDa), respectively (Fig. 6B). In addition to analogous complexes formed by PARP2, PARP2•NCP (271 ± 20 kDa) and (PARP2)₂•NCP (345 ± 30 kDa), a 130−142 kDa species was detected, which could correspond to PARP2 dimer or PARP2 complex with free DNA (Fig. 6C, the asterisked peak; Supplementary Fig. S11).

Figure 6.

Mass photometry data: (A) control measurements of samples containing individual components specified in each panel, (B and C) molecular mass distribution of species in PARP1/PARP2 (6 nM) mixtures with NCP147/gap35-NCP147 (3 nM), (D) histograms show relative amounts of species in the mixtures; difference between the relative amounts of (PARP1)₂•gap-NCP and (PARP2)₂•gap–NCP complexes was statistically significant (**P< 0.01).

From the data of MP experiments, K_d values of PARP1/PARP2 complexes with NCP147 and gap35-NCP147 were determined, using method described by Wu and Piszczek [57]. These values were calculated for the two different types of complexes formed by PARP1/PARP2 (Table 3), based on relative amounts of NCP in free and protein-bound states. Their comparison shows that the non-gapped NCP is more tightly bound by PARP1 than by PARP2. PARP2 has a higher affinity for gap-NCP: K_d values of PARP2•gap-NCP and (PARP2)₂•gap-NCP complexes are 2–2.5-fold lower compared to the respective values of PARP1 complexes. However, the affinity of PARP2 for gap-NCPs measured by fluorescence titration was 1.5-fold lower compared to that of PARP1 (Table 2). The discrepancy between the data obtained by two approaches may result from different oligomerization capabilities of PARP1 and PARP2, depending on the DNA structure. The PARP1 complex containing two protein molecules was detected in the presence of either NCP or gap35-NCP (Fig. 6B and D), suggesting formation of the complex without involvement of the DNA gap. The analogous complex of PARP2 was detected in the similar conditions only for gap35-NCP (Fig. 6C and D), suggesting dimerization of PARP2 on the DNA gap located nearby the position of FAM (in both gap12-NCP and gap35-NCP) and its possible contribution to fluorescent measurements of the apparent binding affinity. Additional MP experiments with varied NCP and PARP2 concentrations revealed binding of two protein molecules at significantly higher concentrations (Supplementary Fig. S12), indicating that the (PARP2)₂•NCP complex has the weakest affinity compared to the other complexes. Comparison of the affinities of PARP2 complexes allowed to suggest that (PARP2)₂•gap–NCP complex was mainly formed via binding of both PARP2 molecules to the DNA gap. This suggestion is favoured by previous results: PARP2 dimerization on the one-nucleotide gap of long DNA visualized by atomic force microscopy was specific for this type of DNA damage [64].

Table 3.

K_d values of PARP1 and PARP2 complexes with NCP147 and gap35-NCP147 (nM)

Enzyme•NCP	Type of complex
	PARP•NCP	(PARP)2•NCP
PARP1•NCP147	8 ± 1	40 ± 10
PARP1•gap35-NCP147	9 ± 2**	40 ± 5***
PARP2•NCP147	33 ± 4	-
PARP2•gap35-NCP147	5 ± 1**	16 ± 3***

The K_d values are averages (±SD) of three independent measurements. Difference between the K_d values of PARP1 and PARP2 complexes with gap35-NCP147 was statistically significant: P< 0.01 (**) and P< 0.001 (***); t-test.

We tried to use MP technique to detect complexes of PARP1 and PARP2 with HPF1, which are characterized by low affinities but can be stabilized by the interaction with nucleosome [69–71]. Detected in the PARP2 mixture with HPF1 an 82 kDa species, which overlapped partially with the individual components, could correspond to the dynamically formed PARP2•HPF1 complex, while a less stable PARP1•HPF1 complex was undetectable (Supplementary Fig. S13A and B). Profiles of molecular mass distribution obtained for the triple mixtures of PARP1/PARP2 with HPF1 and gap35-NCP (Supplementary Fig. S13C and D) were closely similar to those obtained for the double mixtures of PARP1/PARP2 with gap35-NCP, which makes it difficult to draw a conclusion.

The major limitation of MP technique restricting its application is low concentration of biomolecules to be used for better resolution [57]. To overcome this limitation, we used an electrophoretic mobility shift assay (EMSA) for detection of PARP1/PARP2 complexes with NCP variants at concentrations used in ADP-ribosylation experiments (Fig. 7). Two bands migrated more slowly than NCP, which appeared in PARP1 mixtures with either the non-gapped or gapped NCPs correspond most likely to PARP1•NCP and (PARP1)₂•NCP complexes detected in the MP experiments. Two main types of complexes were also detected for PARP2 with each of three NCP variants, but the extent of gapped-NCPs binding was significantly higher. The other slower migrating species formed with low yields (not exceeding 10%) could correspond to complexes of a higher order formed in conditions of EMSA experiments. In accordance with the MP experiments, the non-gapped NCP was more efficiently bound by PARP1 than by PARP2, while the extent of gap35-NCP binding was higher for PARP2 than for PARP1.

Figure 7.

Comparison of PARP1/PARP2 binding to NCP structural variants by EMSA. (A) – Electrophoregram of non-denaturing 5% PAG (visualized by fluorescence imaging) after separation of PARP1/PARP2 mixtures with NCP147, gap12-NCP147 and gap35-NCP147; concentrations used in the experiments: 400 nM PARP1/PARP2, 200 nM NCP147/gap12-NCP147/gap35-NCP147. (B) – Histograms show extent of binding of each NCP variant to PARP1/PARP2 (the mean ± SD, n= 3). Statistically significant differences in the extent of binding are marked: P< 0.05 (*) and P< 0.01 (**); t-test.

Discussion

The architecture of catalytically active PARP1/PARP2 complex with NCP impacts histone modification

Efficient and faithful genomic restoration is facilitated by multiple chromatin remodelling processes, including those triggered by ADP-ribosylation of nuclear proteins [4, 5, 72]. Among them, HPF1-dependent histone ADP-ribosylation is a major contributor to the transient unfolding of chromatin that promotes DNA accessibility in the vicinity of DNA lesions [30−33]. The N-terminal tails of core histones involved in intra- and intermolecular interactions with nucleosomal DNA are primary acceptors of Ser-ADP-ribosylation catalysed by the PARP1/PARP2 complex with HPF1 [5, 30]. Despite significant progress in this field, the mechanisms underlining the selectivity of histone ADP-ribosylation and specific roles of PARP1 and PARP2 in chromatin remodelling and DNA repair remain incompletely understood. Here, we compared activities of PARP1 and PARP2 in ADP-ribosylation of core histones within several NCP variants with distinct structural features.

Our results demonstrate that the efficiency of HPF1-dependent ADP-ribosylation of histones catalysed by PARP2 is significantly influenced by the presence and position of BER-specific DNA damage within the NCP. PARP2 is more active in heteromodification of histones in NCP structures containing the one-nucleotide gap at positions 12 and especially 35 compared to the non-gapped NCP, regardless of the presence of 10/20 bp linkers (gap12-NCP147/167/177 and gap35-NCP147/167/177 vs NCP147/167/177, Figs 1 and 4). We observed that the presence of gap at position 35 was most favourable for the initiation of histone ADP-ribosylation by PARP2 (Fig. 2) despite the absence of detectable difference in the enzyme affinity for gap35-NCPs vs gap12-NCPs (Table 2 and Fig. 7). Furthermore, the pattern of histone modification strongly depends on the SHL of DNA gap within NCP: histone tails located in close proximity to the gap position are preferable targets of PARP2-catalysed modification (Fig. 3B). These results suggest that gap serves as the primary binding site for PARP2 activation in histone modification and its location relative to ADP-ribose acceptors in core histones determines the efficiency and selectivity of the heteromodification reaction. This conclusion is further supported by experiments with PARP2 and gap-NCP267 bearing gap in the long linker region: a significant reduction in the histone modification level compared to that in the presence of NCP147/NCP267 (Fig. 1D) can be caused by predominant PARP2 binding far away from the nucleosome core. Therefore, we propose that the DNA gap in the NCP represents the main binding site for PARP2, coinciding with our previous data [16, 38], and the enzyme modifies targets in close proximity to this site on DNA, which may be crucial in generating a local DNA damage signal and preventing excessive chromatin ADP-ribosylation.

Heteromodification of histones catalysed by PARP1 was revealed to depend to small extent on the presence of DNA gap only at position 35: the modification of histones at initiation and subsequent steps was most favoured in gap35-NCP147 and its variants with 10/20 bp linkers (Figs 1C, 2C, 3A, and 4C). All four core histones were modified with near similar efficiencies in NCP147/167/177 and gap12-NCP147/167/177 at the initiation step, but the pattern of their modification was noticeably changed in gap35-NCP147/167/177 (Fig. 3A). The balance between reactions catalysed by PARP1 was shown to change in favour of the automodification at the initiation step due to addition of 50/70 bp linkers (Fig. 2C). All the results combined allow to suggest primary activation of PARP1 via specific binding near the entry-exit site of NCP. This specific site of PARP1 binding to NCP visualized recently by atomic force microscopy [73] was shown in several other in vitro studies detected competition of PARP1 with linker histone H1 and interaction of PARP1 with H3 and H4 histones located near the exit-entry site [74–78]. A higher accessibility of H3 histone for ADP-ribosylation catalysed by PARP1 than by PARP2 revealed in our experiments is in agreement with these data. Additionally, very early studies indicated predominant localization of PARP1 within internucleosomal linker regions of HeLa cell chromatin [for review see [22]]. The DNA gap, at least at position 35 of nucleosomal DNA 603, acts as an alternative site for PARP1 binding responsible for the enhanced histone modification in gap35-NCP variants. The decrease in histone modification detected for NCP267 variants (Fig. 2C) is most likely resulted from PARP1 activation via alternative binding to the blunt ends separated by the long linkers from the histone core. Overall, our data clearly show that PARP1 differs from PARP2 in being more active in the automodification than in the ADP-ribosylation of histones and less selective in the interaction with different types of DNA damage in the nucleosome.

While the discovery of HPF1-directed Ser-ADP-ribosylation has greatly expanded the research in the field of DNA damage-dependent ADP-ribosylation signaling, the exact role of this PTM in DNA repair is still insufficiently explored. ADP-ribosylome analysis of DNA-damage-induced Ser-ADP-ribosylation shows an enrichment of proteins involved in BER and double-strand break (DSB) repair [79]. Histone ADP-ribosylation triggered by the PARP1–HPF1 complex facilitates the recruitment of the histone chaperone APLF and the remodelling factors (CHD4, CHD7) at DSB sites, that is probably required to establish a repair-competent chromatin architecture [33]. Single-strand break (SSB) repair at actively transcribed DNA regions mediated by PARP1, PARP2 and HPF1 is regulated by the CSB chromatin remodeller shown to contribute to the recruitment of XRCC1 and HPF1 to the sites of DNA damage [80]. HPF1-dependent ADP-ribosylation of the nuclear mitotic apparatus protein NuMA has been reported to promote SSB repair via facilitating the recruitment of tyrosyl-DNA phosphodiesterase 1 [81]. HPF1-dependent PARP1 activation promotes recruitment of XRCC1-DNA ligase 3 complex for Okazaki fragment ligation, thus contributing to repair of replication-associated DNA damage [82]. First studies have revealed extreme sensitivity of HPF1 knockdown human cells to the alkylating agent methyl methanesulfonate (MMS) [28, 31], suggesting contribution of HPF1-dependent ADP-ribosylation to processing of DNA intermediates of BER via unknown mechanisms. A recent study has demonstrated that HPF1-deficient cells maintain the relatively normal SSB repair capacity despite reduced serine mono-ADP-ribosylation, highlighting the complexity of mechanisms that maintain genomic stability and chromatin remodelling [35]. On the other hand, the combined inhibition of PARG and knockdown of ARH3, the enzymes responsible for the reversal of Ser-ADP-ribosylation, was shown to induce the pronounced cell sensitivity to MMS [83], further suggesting that ADP-ribosylation, particularly of Ser residues, may be preferentially involved in BER. A relationship between HPF1 gene and BER genes in maintaining DNA repair efficiency within breast cancer cell has been unveiled recently [84]. This limited number of studies demonstrates the diversity of functions of HPF1-dependent Ser-ADP-ribosylation in chromatin remodelling and repair of various DNA lesions. Our data show that PARP1 and PARP2 may play specific roles in BER upon processing distinct types of DNA damage. PARP1 generates PAR mainly attached to itself, which can serve for recruitment of chromatin remodelling factors and BER proteins to DNA damage and their organization within biomolecular condensates [9–11]. PARP2 being more active in histone ADP-ribosylation may contribute to local relaxation of nucleosome structure, thus facilitating DNA accessibility to BER machinery.

Interplay between nucleosome dynamics and its processing upon DNA damage response and repair

N-terminal tails of core histones involved in transient interactions with nucleosomal DNA contribute to intrinsic nucleosome dynamics, including nucleosome sliding and DNA unwrapping/wrapping [39–42]. Tails of different histone types preferably interact with the specific DNA regions; the longest H3 tails are arginine rich, interact with DNA in multiple regions and contribute most to the nucleosome stability [40–42]. Various posttranslational modifications of histone tails including ADP-ribosylation, which are most numerous in H3 histone, further enhance nucleosome dynamics and DNA accessibility [10, 33, 40, 42]. Specific canonical histone H3 marks interfere with Ser-ADPR of neighbouring residues, and acetylation of H3 histone at K9 is mutually exclusive with its ADP-ribosylation at S10, indicating complex interplay of modifications that form the histone code [85].

While some BER enzymes can efficiently process DNA damage within NCP, the initiation and completion of this process is inhibited on nucleosome substrates [86, 87]. The enzymatic activities of DNA-glycosylases UDG and OGG1 in NCP structures do not correlate with the solution accessibility of the DNA lesions but significantly depend on the NCP dynamics, which can be modulated by deletion or acetylation of histone tails, as shown by in vitro experiments [88, 89]. ADP-ribosylation of histone variant H2AX at the C-terminal tail mediates the recruitment of Neil3 glycosylase to the sites of DNA damage and facilitates BER [90]. Ligation of nicked nucleosomal DNA is stimulated by PARylation due to enhanced motion of histone tails [10]. The predominant form of PARP-catalysed modification of histones in response to DNA damage is mono-ADP-ribosylation, which is sufficient for destabilization of NCP structure in the vicinity of the DNA lesions [91, 92]. Previous studies have demonstrated that the balance between PAR and MAR attached to histones is controlled by HPF1: the ratio of HPF1 to PARP1/PARP2 determines the number and length of the ADP-ribosylation signal [15, 16, 27, 70, 92].

Here were have shown that the length of PAR and the number of initiation events in HPF1-dependent heteromodification reaction catalysed by PARP2 depends on the location of BER-specific DNA lesion relative to the nucleosome core: position 35 in DNA (SHL 3.5) is more preferable than position 12 (SHL 5.5) for the efficient modification of core histones with short polymers (Figs 1 and 2). Since no influence of the DNA gap position on the strength of PARP2 interaction with gap-NCP was revealed (Table 2 and Fig. 7B), we can suggest that the efficiencies of initiation and elongation steps upon the heteromodification are determined by the conformational flexibility of target histone tails. Indeed, the histone types modified by PARP2 in gap12-NCP and gap35-NCP are distinct (Fig. 3), implying their different contribution to the NCP dynamics. The longer negatively charged PAR chains could be more disruptive for electrostatic interactions between the DNA and histone tails than the shorter ones. It seems likely that the length of ADPR polymer attached to histone tails and its interrelationship with other posttranslational modifications may control the extent of nucleosome structure relaxation.

Mass Photometry allows analysing the stoichiometry of PARP-NCP complexes and self-association of PARPs

Mass photometry (MP), a relatively novel technique, enables detecting and accurately measuring molecular masses of proteins, nucleic acids, and their complexes [57, 58, 68]. Here, we applied the MP measurements to analyse formation and composition of PARP-NCP complexes and the oligomeric state of PARPs in solution. Considering the self-association of PARP1 and PARP2 and their affinity for undamaged DNA, we performed MP experiments at PARP/NCP concentrations close to EC₅₀ values of their complexes, as determined by the fluorescence anisotropy measurements. Our results revealed the presence of species corresponding to (PARP)₂•NCP complexes even at low concentrations with only a two-fold excess of PARP over NCP (6 and 3 nM, respectively). The patterns of species distribution observed for PARP1 mixtures with NCP147 and gap35-NCP147 were very similar (Fig. 6), indicating very low contribution of PARP interaction with gap to the complex formation at low concentrations, which are used in MP experiments for better resolution. Both MP and fluorescence anisotropy measurements of binding affinities revealed distinction between PARP1 and PARP2: the strength of interaction with NCP strongly depends on the presence of gap only for PARP2 (Tables 2 and 3). Another difference evidenced from the MP experiments is significantly higher stability of PARP2 complex with gap35-NCP containing two protein molecules compared to the respective complex of PARP1 (Table 3). The interaction of PARP1 and PARP2 with NCP variants at high concentrations used in ADP-ribosylation experiments was further explored by EMSA technique (Fig. 7). These experiments confirm the data of MP experiments: a) both PARP1 and PARP2 form at least two types of complexes of different composition with each of NCP variants; b) PARP1 has low sensitivity to the DNA gap presence and interacts more strongly than PARP2 with the non-gapped NCP; c) compared to PARP1, PARP2 displays a higher affinity for the gapped NCPs, binding them more efficiently than the non-gapped one.

Self-association of PARP1 and PARP2 not induced by DNA binding has been detected by different approaches in previous studies [62, 93–95]. Our MP experiments (Fig. 6A) provide the first evidence of a higher propensity of PARP2, in comparison with PARP1, for the oligomerization via di- and trimerization. Cooperative binding of PARP1 to nucleic acid structures and its functional significance were explored in several studies. PARP1 dimerization at a palindromic structure like restriction site and a 5′-recessed DNA end was found to be a requisite for high enzymatic activity [96]. Co-condensation of PARP1 multimers with DNA double-strand breaks (DSB) prevents the broken DNA from separation, thus facilitating DSB repair [97]. Different extent of PARP1 multimerization shown to depend on the type and structure of nucleic acid (RNA/DNA, single-stranded/duplex DNA) may serve for recognition of diverse structures by the functionally versatile protein [98]. The atomic force microscopy study enabled visualizing dimerization of PARP1 and PARP2 at distinct damage types of long DNA, with dimer formation upon the interaction with gap being unique for PARP2 [64]. Binding of two PARP2 molecules to one NCP molecule with the internal BER-specific DNA damage detected in our MP experiments has never been described. The previous cryo-electron-microscopic structural study of PARP2-HPF1 complex bound to NCP revealed bridging of two NCPs via independent interaction of each PARP2 molecule (i.e. without direct interaction between two PARP2 molecules) with both DNAs; the complex is stabilized by interactions of each WGR domain with the nucleosomal DNA end of one nucleosome and several contacts of PARP2 helical domain and HPF1 with the DNA of the second nucleosome [99].

Conclusion

The details of PARP1 and PARP2 interaction with free DNA and nucleosome have been the subject of numerous studies, with some of them being focused on the relationship between the DNA-dependent enzymatic activity and affinity for the distinct DNA structure. The interrelation between the nucleosome (NCP) architecture and PARP-catalysed modification of histones as the earliest targets of ADP-ribosylation in response to DNA damage has never been studied by others. Here, by using a combination of biophysical and biochemical approaches we succeeded to reveal that the extent of histone ADP-ribosylation and the balance between auto- and heteroPARylation reactions at both the initiation and subsequent steps of PAR synthesis depend to different extents for PARP1 and PARP2 on the presence of a one-nucleotide gap (the BER-specific lesion) and its location in the NCP structure (Fig. 8). A higher dependence of the level and pattern of PARP2-catalysed histone modification on the NCP architecture is easily explained by the gap-binding induced enzyme activation. This specific PARP2–NCP interaction favours the enzyme functioning as a dimer. A significantly lower impact of the NCP structure on selectivity of histone modification in the case of PARP1 results most likely from the predominant enzyme binding at the entry-exit region. The analysis of histone modification data taking into account the known HPF1-dependent ADP-ribosylome of histones and dynamics of the NCP structure suggests interrelation between the pattern and extent of histone modification and conformational flexibility of the target histone tails. The location of a specific DNA damage and dynamics of NCP modulated by various post-translational modifications may determine distinct roles of PARP1 and PARP2 in response to genotoxic stress. PARP1 is a main generator of PAR, which mediates assembling of chromatin remodellers and BER proteins on damaged DNA via formation of biomolecular condensates. PARP2 being more active in histone ADP-ribosylation may facilitate DNA accessibility to BER machinery via local relaxation of nucleosome structure.

Figure 8.

Histone PARylation in nucleosome is controlled by the specificity of PARP-NCP interaction. The presence and position of the one-nucleotide gap in nucleosomal DNA were revealed to produce a significant effect on the extent of histone PARylation despite of the absence of detectable difference in the PARP1/PARP2 affinities for the two gapped NCP structures. We suggest that superhelical location of the BER-specific DNA lesion as a potential binding site for PARP activation and the conformational dynamics of surrounding histone tails determine the efficiency and selectivity of histone ADP-ribosylation.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

We have no conflict of interest to declare.

Funding

The reported study was funded by the Russian state-funded project for ICBFM SB RAS № 125012300658-9 (use of shared equipment for experimental work, MP experiments) and by the Russian Science Foundation № 22-74-10059 (preparation of enzymes, histones and nucleosomes, ADP-ribosylation and EMSA experiments). Funding to pay the Open Access publication charges for this article was provided by Russian state-funded project for ICBFM SB RAS (grant number 125012300658-9) and the Russian Science Foundation (grant number 22-74-10059).

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.