PARP–nucleic acid interactions: allosteric signaling, PARP inhibitor types, DNA bridges, and viral RNA surveillance

Abstract

PARP enzymes create ADP-ribose modifications to regulate multiple facets of human biology, and some prominent PARP family members are best known for the nucleic acid interactions that regulate their activities and functions. Recent structural studies have highlighted PARP interactions with nucleic acids, in particular for PARP enzymes that detect and respond to DNA strand break damage. These studies build on our understanding of how DNA break detection is linked to the catalysis of ADP-ribose modifications, provide insights into distinct modes of DNA interaction, and shed light on the mechanisms of PARP inhibitor action. PARP enzymes have several connections to RNA biology, including the detection of the genomes of RNA viruses, and recent structural work has highlighted how PARP13/ZAP specifically targets viral genomes enriched in CG dinucleotides.

INTRODUCTION

PARP enzymes are involved in multiple aspects of human biology, including genome maintenance, gene regulation, viral defense, protein homeostasis, and cell signaling and transport [1,2]. The ADP-ribosyltransferase fold typifies PARP enzymes and embodies the catalytic activity that creates ADP-ribose modifications using NAD⁺ [3,4]. The majority of PARP enzymes create mono-ADP-ribose modifications on proteins or DNA, and a subset of PARP enzymes can form poly(ADP-ribose), a polynucleotide with a distinctive structure and cell signaling capacity. Human PARP enzymes exhibit diverse domain organizations that dictate individual cellular tasks, usually through providing regulatory capacity or affording interaction with partner proteins or nucleic acids. Nucleic acids have long been appreciated as potent regulators of several PARP enzymes, particularly for those PARPs involved in DNA strand break detection and the cellular response to DNA damage. The structural basis for these nucleic acid interactions and their regulatory capacity have only fairly recently begun to be understood at a structural level. This review summarizes recent structural insights into human PARP structural biology with a focus on protein nucleic acid interactions.

PARP1 structural biology

PARP1 is the founding member of the PARP family and has prominent roles in gene regulation and genome maintenance [2]. With respect to genome maintenance, PARP1 rapidly detects DNA strand breaks, and PARP1 engagement of DNA breaks leads to a massive, allosteric stimulation of poly(ADP-ribose) production [5]. PARP1 achieves this sensitivity to DNA strand break damage through multiple domains that collectively interact with DNA and thereby allosterically enforce a catalytic domain conformation that is open and accessible to substrate NAD⁺ [6–11] (Figure 1 A–C). The most profound structural change in this activation process occurs in the PARP1 helical domain (HD), which blocks NAD⁺ binding to the catalytic active site when PARP1 is not engaged on a DNA break. A structure of multiple PARP1 domains bound to a DNA double-strand break suggested a mobile HD [7], and hydrogen-deuterium exchange mass spectrometry pinpointed specific HD helices that undergo massive unfolding transitions in the presence of a DNA single-strand break [8]; however, there were no high-resolution views of the PARP1 HD conformation that would permit NAD⁺ binding. A crystallographic study captured this open conformation by using a mutant version of PARP1 that favored the open HD conformation, and an NAD⁺ mimic was observed bound to the active site [12](Figure 1B,C). A surprising feature of this structure was a large-scale rotation of the ADP-ribosyltransferase (ART) fold relative to the rest of the complex, and this rotation appears necessary to fully open the path to NAD⁺ utilization. Importantly, this same study highlighted that the open configuration of the HD allows it to fully contribute to PARP1 affinity for DNA damage by adding to the multidomain assembly on a DNA break (Figure 1B).

Figure 1. Recent insights into PARP1 structural biology.

A. Human PARP1 domains. Zinc fingers Zn1 and Zn2 have similar structures and modes of engaging DNA damage; Zn1 also forms critical interdomain contacts. Zinc finger Zn3 has a distinct structure and is involved in DNA and interdomain contacts. The BRCT (BRCA1 C-terminus) fold is positioned within an extended unstructured region that bears the major sites of automodification. The WGR domain (Trp-Gly-Arg) is crucial for coupling the activities of the N-terminal DNA binding regions and the C-terminal catalytic region. The catalytic region is composed of a helical domain (HD) and the ADP-ribosyltransferase (ART) fold.

B. The DNA damage-dependent activity of PARP1 can be reconstituted with a minimal set of domains – Zn1, Zn3, WGR, and HD-ART. X-ray crystallographic analysis has captured these essential domains in complex with DNA duplexes mimicking DNA double-strand break damage. The domains collectively assemble on the DNA break, and induce a major increase in the conformational dynamics of the HD. The extent of HD structural change was captured in a recent crystal structure of a PARP1 VE mutant that favored the “open” conformation of the HD (PDB 7s6m[58]; shown with domains in non-grey colors). The mutant removes two residues (denoted VE) from a helix that is highly dynamic when PARP1 binds to DNA based on HXMS analysis (helix B, labeled ‘B’). The mutant structure featured a ~15 degree rotation of the ART domain relative to the rest of the complex, when compared to the wild-type protein (PDB 4dqy[7], all domains shown in grey). The rotation is highlighted by the change in position of the N- and C-terminal ends of the catalytic region (labeled ‘n’ and ‘c’). The ART rotation allowed the NAD⁺-mimic EB47 to access the active site (labeled as PARPi). The “open” conformation of the HD makes additional contributions to the multi-domain assembly on DNA (labeled with an asterisk and denoted by an arrow showing the change in position of a key leucine residue), thereby stabilizing the complex and increasing PARP1 affinity for DNA damage.

C. The same alignment of structures in panel B, but rotated 90 degrees to provide a second view. Helix F in the HD (labeled ‘F’) is sheared away from the ART to allow space for the NAD⁺ mimic.

D. PARP inhibitors can be classified into three types based on their impact on PARP1 allostery and DNA binding affinity. The left pointing arrow references the structures in panels B and C, in which a Type I inhibitor biases the HD to favor the “open” conformation that increases DNA binding affinity and leads to enhanced retention on DNA damage. The right pointing arrow references the structures in panels B and C, in which a Type III inhibitor engages the ART in a way that supports the HD in the closed conformation, and thereby decreases DNA binding affinity and favors release of PARP1 from DNA breaks.

E. A cryo-EM structure captured the BRCT domain of PARP1 in complex with the nucleosome particle (PDB 7scy[22]), illustrating the BRCT contacts made with continuous, undamaged DNA.

PARP inhibitors outcompete NAD⁺ binding to prevent ADP-ribose modifications from being formed. For PARP1, this prevents the creation of poly(ADP-ribose) and the roles that it plays in mediating the efficiency and speed of the DNA damage response. This deficiency is particularly toxic to cancer cells with deficiencies in homologous recombination repair of DNA damage [13–15]. PARP1 automodification with poly(ADP-ribose) weakens interaction with DNA breaks, and thereby contributes to the turnover and mobility of PARP1 at sites of DNA damage. Correspondingly, PARP inhibitors have the effect of modulating PARP1 residency time on DNA breaks by preventing poly(ADP-ribose) formation. The propensity of PARP inhibitors to “stall” PARP1 on DNA is proposed to be a major factor underlying cellular toxicity [16,17], and the inhibitors that best quench catalytic activity typically have the strongest effect of increasing PARP1 residence time on DNA breaks [18]. It has recently been appreciated that small molecules binding to the PARP1 active site have the potential to modulate the PARP1 allosteric activation mechanism in ways that influence affinity for DNA breaks [9,19]. For example, a non-hydrolyzable NAD⁺ mimic increased PARP1 affinity for DNA breaks by supporting the open HD conformation that contributes to DNA binding affinity [9]. A systematic analysis of clinical PARP inhibitor effects on PARP1 allostery and DNA break binding affinity indicated that none of the currently used inhibitors have a strong tendency to increase PARP1 affinity for DNA breaks [19]. Surprisingly, several clinical inhibitors indeed modulated PARP1 allostery, but in a manner that decreased PARP1 affinity for DNA breaks. This analysis led to a classification of PARP inhibitors into three types: type I, allosteric and pro-retention on DNA; type II, non-allosteric and no influence on DNA retention; type III, allosteric and pro-release from DNA [19](Figure 1D). The classification of PARP inhibitors toward PARP1 has also been observed with other structural approaches [20,21]. These findings open a new dimension of PARP inhibitor design, wherein the proximity of the NAD⁺ binding site to the HD permits inhibitor structures to be tailored in ways that engage the HD to influence allostery and PARP1 engagement on DNA.

PARP1 contains a BRCT fold that is not required for interaction with DNA breaks or for catalytic activation by DNA breaks; however, the PARP1 BRCT fold was recently shown to interact with undamaged DNA and was proposed to contribute to how PARP1 navigates through the nucleus [22]. Hydrogen-deuterium exchange mass spectrometry experiments indicated that the BRCT domain of PARP1 interacted with undamaged DNA, and binding studies confirmed this interaction. A cryo-EM structure of the BRCT fold in complex with the nucleosome core particle provided insights into the mode of interaction and the residues likely to mediate the interaction (Figure 1E). The hydrogen-deuterium exchange mass spectrometry experiments also indicated that other domains of PARP1 interact with undamaged DNA, but these collective interactions do not lead to catalytic activation. PARP1 is an abundant nuclear protein and is likely to generally exist in the presence of more undamaged DNA than damaged DNA, so the interaction with undamaged DNA will be an important aspect of PARP1 to continue to decipher.

Insights into PARP1 mechanism have thus far come from individual domains or combinations of domains in complex with DNA, but a full-length structure at high resolution has yet to be captured. Negative stain analysis of full-length PARP1 in complex with a DNA single-strand break indicates that the composite structures are indeed representative of the full-length structure [23], and extending these studies to cryo-EM holds the promise of delivering the first high-resolution views of full-length PARP1.

PARP2 structural biology

PARP2 can be viewed as a more compact version of PARP1, with only the WGR domain and a short N-terminal extension to regulate its catalytic response to DNA strand break interaction (Figure 2A). Although there are conserved features of how PARP1 and PARP2 are allosterically activated by DNA breaks through WGR-HD contacts [24], recent structural analysis has indicated some sharp differences in how PARP1 and PARP2 utilize the WGR domain to engage DNA breaks. In crystal structures of the PARP2 WGR domain in complex with duplex DNA, the ends of DNA duplexes were juxtaposed in the crystal lattice and the WGR domain spanned the two DNA ends, suggesting that the protein might bring together DNA ends [25]. This binding mode is distinct from the WGR of PARP1, which collaborates with zinc fingers to effectively engage DNA breaks. A more recent structure of the WGR-CAT fragment of PARP2 also indicated that the WGR domain bridges DNA ends [26](Figure 2B). A curious feature of this structure was that the ART fold was substantially displaced from the WGR and HD, and the terminus of one DNA strand (not involved in a bridging interaction) was positioned near the active site. Although the exact positioning of the ART fold could be influenced by the crystal packing environment, it clearly highlights the mobility of the ART in the PARP2 activation process.

Figure 2. Recent insights into PARP2 structural biology.

A. Human PARP2 domains. PARP2 bears WGR and HD-ART domains similar to PARP1. The WGR domain interacts with DNA strand breaks and communicates with the HD-ART. The N-terminal region (NTR) also contributes to overall DNA binding affinity.

B. A crystal structure of human PARP2 was determined in complex with duplex DNA (PDB 7aeo[26]). The WGR domain was positioned at the junction of two DNA molecules (DNA1 in light grey, DNA2 in dark grey), effectively forming a bridge across the discontinuity in the backbone of the juxtaposed DNA ends. For clarity, a WGR domain bound to the opposite side of the DNA junction is not shown. The HD forms contacts with the WGR, and the ART is translated away from the HD relative to catalytic domain structures in the absence of DNA. For reference to PARP1 in Figure 1, helix B is labeled.

C. In a recent cryo-EM study, the human PARP2/HFP1 complex was captured bridging the DNA ends extending from two different nucleosomes (PDB 6x0n[27]). In the most complete map/model produced from the study, a complete PARP2/HPF1 complex is positioned on one DNA junction (see panel D), and the WGR alone is positioned on the other DNA junction.

D. Focused refinement in the same cryo-EM study also yielded a map/model for PARP2 bridging two DNA ends (PDB 6x0l[27]). The WGR contacts across the DNA ends agrees with the X-ray structure in panel B. The ART domain remains in close contact with the HD, in contrast to the X-ray structure in panel B. The second WGR domain is not shown for clarity.

E. The ART domain of PARP2 was crystallized in complex with HPF1 (PDB 6tx3[59]). HPF1 completes the PARP2 (and PARP1) active site by contributing E284 to allow deprotonation of Serine residues, permitting their modification with ADP-ribose.

A cryo-EM study analyzed PARP2 interaction with DNA ends extending from two different nucleosome core particles [27], giving a visually stunning representation of the DNA bridge mediated by the PARP2 WGR domain (Figure 2C,D). In this complex, the ART domain remained in close contact with the HD (Figure 2D), in contrast to the crystal structure described in the last section. The model deposited for the PARP2 complex with DNA ends (Figure 2D) did not exhibit the ART domain rotation described in the previous section for PARP1. It could be that only one dominant conformation of the catalytic region was captured, or that the domain rotation is not conserved in PARP2 activation. The PARP2 structure did not contain substrate or substrate mimics in the active site, so it is unclear whether the modeled conformation represents the active form of PARP2. A different cryo-EM study also captured PARP2 bridging DNA ends extending from nucleosomes, but only the WGR domain was modeled in this case [28].

Several structures have thus captured PARP2 spanning two DNA ends, suggesting a potential role for PARP2-mediated DNA bridging in repair processes. PARP2 DNA bridges were recently studied using a magnetic tweezer-based assay, and the results indicated a remarkably stable bridging interaction [29]. PARP1 did not exhibit the capacity to tether DNA ends in the same experimental setup. Further studies are needed to better understand the potential conditions under which PARP2 might play this end-bridging role in the cell.

Even though PARP1/2/3 are generally considered conserved in the WGR-CAT regions, the recent structures and results described here indicate interesting specialization between PARP1 and PARP2 in terms of DNA break engagement. High-resolution structures of PARP3 in complex with DNA might indicate other possibilities for the WGR domain to engage DNA breaks. It is also notable that PARP inhibitors have different allosteric effects on PARP1 and PARP2. None of the current clinical PARPi exhibit type I behavior toward PARP1 [19]. In contrast, several clinical PARPi exhibit type I behavior toward PARP2, thereby strongly increasing PARP2 affinity for DNA breaks [30]. Thus, we continue to learn more about interesting variations even among highly related members of the PARP family.

HPF1 regulation of PARP1/2

Histone PARylation factor 1 (HPF1) has only recently emerged as a central regulator of PARP1 and PARP2 in the DNA damage response [31–34]. HPF1 permits PARP1 and PARP2 to modify serine residues, notably those located on histone tails, in addition to the glutamate and aspartate residues that are the typical sites of modification [32]. Indeed, the discovery of HPF1 has exposed an underappreciated prevalence of serine-ADP-ribose modifications [35–37]. Several structural studies, including the PARP2/nucleosome complex highlighted above, have now demonstrated that HFP1 “completes” the PARP1/PARP2 ART fold, in particular contributing a glutamic acid residue to a joint PARP/HPF1 active site [27,38,39](Figure 2E). The glutamic acid residue is expected to deprotonate serine residues and make them chemically competent for modification. The discovery of HPF1 opens the exciting possibility of other PARP family members working with core regulatory factors.

PARP interactions with RNA

There is a growing appreciation of how PARP enzymes are connected to RNA biology [40,41], and several PARP enzymes contain known RNA binding modules, including PARP10, PARP12, PARP13, and PARP14 [1]. However, there are limited structural insights into RNA-mediated regulation of PARP enzymes. There are reports of RNA structures that stimulate PARP1 and PARP2 activity [42–44]. The level of activation by RNA is generally much lower than the level of activation seen with DNA breaks, and perhaps this is expected for normal functions in unstressed cells that might require more modest levels of PARP activity, compared to the “firehose” levels of PAR production in response to DNA breaks. Structural biology will clearly play an important role in establishing the basis for these RNA interactions with PARP1 and PARP2 and suggesting mechanisms of regulation.

Studies of PARP13 have provided the first insights into PARP-RNA interactions. PARP13 also goes by the name zinc finger antiviral protein or ZAP [45–48]. PARP13/ZAP detects and promotes the degradation of viral RNA. PARP13 preferentially interacts with RNA containing CG dinucleotides, which steers its RNA-binding activity toward viral genomes enriched in CG dinucleotides (and some viral genomes with suppressed CG dinucleotide content are less susceptible to PARP13). RNA interaction is mediated by four N-terminal CCCH-type zinc fingers (Figure 3A). Structures of the N-terminal zinc fingers of PARP13 bound to RNA oligonucleotides have illustrated the specificity of PARP13 for CG dinucleotides in RNA [49,50]. The structure of human PARP13 N-terminal zinc fingers indicated that the specificity comes largely from the second zinc finger (Figure 3B). The mouse PARP13 structure captured the same mode of CG dinucleotide detection, and also indicated how other zinc fingers can contribute to RNA interactions (Figure 3C). It is anticipated that multiple PARP13 molecules will bind to the large segments of viral RNA presenting CG dinucleotides, perhaps forming a scaffolding structure that could initiate RNA degradation in concert with other cellular factors.

Figure 3. Recent insights into PARP13/ZAP structural biology.

A. Human PARP13/ZAP domains. Four CCCH zinc fingers (ZnF) are located at the N-terminus, followed by an extended unstructured region. A central structured region is composed of a fifth zinc finger (ZnF5) and two WWE domains: ZnF5-WWE1-WWE2. The ART fold is located at the C-terminus and lacks the residues required for catalytic activity.

B. A crystal structure of human PARP13 in complex with an RNA oligonucleotide demonstrates how specificity toward CG dinucleotides is achieved primarily within ZnF2 (PDB 6uej[49]).

C. A crystal structure of mouse PARP13 in complex with an RNA oligonucleotide also capture the specific interactions made with a CG dinucleotide in ZnF2, and also specific contacts made with a G nucleotide in ZnF3, and a C nucleotide in ZnF4 (PDB 6l1w[50]). The C nucleotide in ZnF4 comes from the terminus of the RNA chain bound to another PARP13 molecule.

D. Crystal structure of the ZnF5-WWE1-WWE2 region of human PARP13 in complex with ADP-ribose (PDB 7tgq[60]). Only the WWE2 domain has a functional binding site. The three domains collectively assemble into a single module.

E. Structures, mutagenesis, and binding analysis support a model for PARP13 interaction with poly(ADP-ribose) in which binding is anchored at one end of the poly(ADP-ribose) chain, and an electropositive groove formed by the collective ZnF5-WWE1-WWE2 structure supports binding of polymers of a certain length [56]. A model of four units of a poly(ADP-ribose) chain are shown on the left for comparison to the binding model.

Interestingly, the PARP13 ART fold at the C-terminus lacks the residues required for catalytic activity, and some PARP13 isoforms lack the ART fold [51,52]. The central region of PARP13 has a CCCH zinc finger and tandem WWE domains (Figure 3A). This central region interacts with poly(ADP-ribose) and the interaction contributes to the antiviral properties of PARP13 [53–55]. Recent structures have provided first insights into how this region of PARP13 engages poly(ADP-ribose), using the collective fold of the zinc finger and tandem WWEs [55,56](Figure 3D,E). More mechanistic work is needed to understand how the RNA-binding, poly(ADP-ribose) binding, and ART fold collectively operate to mediate PARP13 antiviral functions [57].

Conclusion

The structures highlighted in this review represent some of the first views of PARP nucleic acid interactions, but we can likely expect more structures as our understanding of PARP biology continues to mature. Poly(ADP-ribose) is itself a special type of polynucleotide. Detailed analysis of interactions with poly(ADP-ribose) are generally lacking, thus it will be interesting to see this area of PARP biology further develop, as there are interesting opportunities for DNA/RNA binding surfaces to be co-opted for interaction with poly(ADP-ribose), and perhaps thereby exert regulatory functions.