PARP1 and PARP2 are early responders to DNA strand breaks. They belong to the ADP-ribosyl transferase diphtheria toxin-like (ARTD) enzymes family [1]. In mammals, the ARTD family has at least 17 members that share a conserved catalytic domain that transfers ADP-ribose (ADPr) from NAD+ to protein substrates or even nucleic acids 2, 3, 4. Except for the rare inactive homologous (e.g., PARP13), most PARP family members can only add one ADP-ribose moiety to their substrates, including the recently identified activity of PARP9 toward ubiquitin [5], termed mono(ADP-ribosyl)ation (MARylation). PARP1, PARP2, and Tankyrase (PARP5a/5b) are unique and capable of extending the initial MARylation with additional ADP-ribose moieties to form poly(ADP-ribose) (PAR) chain in either linear or branched configurations depending on the -OH group used for chain extension. Specifically, the enzymatic activity of PARP1 and PARP2 is induced upon direct interaction with broken DNA. Thus, they are called DNA damage-induced poly-ADP ribose transferases (PARPs). The initial MARylation is thought to be the rate-limiting step for DNA damage-induced PARylation and can occur on the -OH groups of Aspartate (Asp, D), Glutamate (Glu, G), Serine (Ser, S), and more recently, tyrosine (Tyr, Y) side chains [6]. While purified PARP1 and PARP2 can independently modify Asp and Glu on protein substrate, Ser PARylation in the context of Lysine-Serine (KS) motif and potentially Tyr PARylation requires a co-factor histone poly(ADP-ribosylation) factor (HPF1) [7]. PAR chains are highly negatively charged and can facilitate chromatin relaxation 8, 9. PAR chains also serve as landmarks to recruit other repair proteins containing the PAR binding domains, including the XRCC1-LIG3 complex for single-strand break (SSB) repair. The diphosphate linker and the ADP-ribose moiety of PAR chains share many features with nucleic acids and can also be recognized by many nucleic acid binding motifs, including the zinc finger domain found in transcriptional factors and BRCA C-terminal (BRCT) domain with phosphorylation binding properties [10]. Correspondingly, it was reported that ~70% of randomly selected transcriptional factors were transiently recruited to DNA damage sites in a PARP activity and PAR-dependent manner [11]. In this context, PARP1 forms condensate with DNA in vitro [12] and mediates phase separation around DNA damage sites in vivo [13]. Correspondingly, in addition to the role in DNA repair and replication discussed here, PARP1 and its activity have been implicated in ribosome biogenesis 14, 15 and transcriptional regulation 15, 16, 17. While how the role of PARP1 in phase separation and transcription factor recruitment contributes to DNA repair remains elusive, the results highlight the complexity and potentially confounding nature of PAR-dependent DNA damage-induced protein foci formation. Naturally, different PAR binding motifs and their combination dictate the binding preference of the corresponding PAR structures, including linear vs. branched and terminus vs. intermediate units.

While PARP1 and PARP2 share similar catalytic mechanisms, they are different in DNA lesion-binding specificities and activities. While the activation of PARP1 and PARP2 requires DNA breaks and, specifically, the 5’phosphorylated DNA end, PARP1 and PARP2 bind to a much broader range of nucleic acid structures. PARP1 has five DNA binding domains: three zinc finger domains, one BRCT domain, and the conserved Tryptophan-Glycine-Arginine (WGR) domain. Together, PARP1 could bind diverse nucleic acid structures, including RNA [18], DNA (e.g., single-stranded, double-stranded breaks and G-quadruplexes), and even RNA-DNA hybrid in the context of R-loops 15, 19, 20, 21, 22, 23, 24. In nucleosome assemble assay, purified PARP1 can replace H1 and condense nucleosomes without DNA damage, highlighting the ability for PARP1 to bind intact DNA [25]. Meanwhile, PARP2 has only two nucleic acid binding domains – the poorly structured but highly charged N-terminal region (NTR) and the WGR domain. Nucleic acid binding is necessary but not sufficient to activate PARP1 and PARP2. The allosteric enzymatic activation of PARP1 or PARP2 requires direct interaction between their respective WGR domain with the 5’ phosphorylated DNA end; as such, neither RNA [18] nor intact double-stranded DNA (e.g., supercoiled plasmids) 26, 27 can activate PARP1 or PARP2. This selectivity for the 5’ phosphorylate DNA ends enables PARP1 and PARP2 to be sensors for DNA strand breaks. In addition, PARP1 also has three Zinc Finger domains connected with flexible linkers. They allow PARP1 to be activated by diverse DNA end structures, from single-strand nicks and gaps to DNA double-strand breaks (DSBs) with various overhangs. The BRCT domain of PARP1 binds to intact DNA and is dispensable for break-induced allosteric activation of PARP1 but is required for PARP1 dependent condensation formation in vitro 12, 26. The high abundance of PARP1 and its ability to be activated by various DNA breaks [28] explain why PARP1 is responsible for over 80% of the PARylation induced by external DNA damage in cells [29]. In contrast, PARP2 activation solely depends on the WGR domain and can only be efficiently activated by 5’ phosphorylated nick [30]. Since the single WGR domain straddles over the nick and interacts with DNA at both sides, expanding the nick to a gap dramatically reduces PARP2 activation [30]. In structural studies, two molecules of PARP2 combine two dsDNA substrates to form two nicks and bind on each nick [31]. This data explains why dsDNA can also activate PARP2 in early biochemical studies [30]. Although the N-terminal region of PARP2 is highly charged and can bind nucleic acid in vitro, it seems to be dispensable for nick-induced activation of PARP2 [32]. With a combination of in vitro binding assays and single molecule imaging studies, the functional significance and biological implications of DNA lesion selectivity might be explored.

The dual enzymatic inhibitors for PARP1 and PARP2 (referred to as PARPi thereafter) are promising new genotoxic cancer therapy that targets BRCA1/2-deficient cancers. Four NAD+ competitive PARPi have been approved by the FDA in the past decade to treat various BRCA1/2- or homologous recombination-deficient breast, ovarian, pancreatic, and prostate cancers 33, 34. In addition to inhibiting the enzymatic activity of PARP1/2, current FDA-approved PARPi also extend the accumulation of PARP1 and PARP2 at damaged chromatin and DNA damage foci, termed trapping 35, 36, 37, 38. Trapping of PARP1 is critical for the therapeutic effect of PARP inhibition. PARP1 deletion leads to significant resistance to PARPi [39]. However, the molecular nature of trapping and the physiological implications in normal tissues remain poorly understood. Using deuterium exchanging assays, dynamics analyses of PARP1 in vitro have classified PARP inhibiting agents, including some can only be used in vitro, into three types according to their allosteric impacts on PARP1-DNA complex, namely type I (pro-retention), EB-47 and BAD; type II (neutral), talazoparib and olaparib and type III (pro-release), rucaparib, niraparib and veliparib[40]. Curiously, the behavior of PARP inhibitory agents in vitro does not strictly correlate with their ability to cause persistent PARP1 foci in vivo. While BAD and EB47 cannot be used in cells, talazoparib, rucaparib, niraparib, and to a lesser extent olaparib, all cause persistent PARP1 foci in cells 35, 36, 37, suggesting additional factors contribute to the half-life of PARP1 foci beyond allosteric trapping. This might include DNA repair, the other PARylation substrate (e.g., Histones) of PARP1, and functional interaction between PARP1 and PARP2, just to name a few. Indeed, several PARP1/2 regulatory partners have been found in recent years, their impact on PARP1/2 trapping is yet to be fully explored [41]. Here, we review how quantitative live cell imaging tools that have been employed to dissect the complex temporal and special regulation of DNA damage-induced PARylation by PARP1/2 and related regulatory proteins and their role in DNA damage and normal replication under physiological state. We also discuss some pending questions regarding the type of DNA lesions, the amino acid targets, the shape of PAR chains, and their implications in DNA repair and PARP kinetics.