PARP enzyme synthesis of protein-free poly(ADP-ribose): Implications for DNA damage signaling and repair

1.1. De novo synthesis of protein-free PAR by PARP enzymes

The structural properties of PAR have typically been studied by chemically releasing intact PAR chains from the site of modification on protein targets. PAR isolated in this way allows attributes like length distribution to be analyzed and has been used to study phenomena such as PAR-seeded biocondensate formation [18], Parthanatos cell death signaling mediated by PAR [25], and interactions with PAR-binding protein modules [30]. The chemical release step is typically preceded by a protein precipitation step that separates the insoluble modified protein from unutilized NAD⁺ that remains in the soluble fraction after protein precipitation. By analyzing both the insoluble protein-containing fraction and the soluble protein-free fraction, we noted that PARP1 reactions unexpectedly produced PAR in the soluble fraction [29]. Several different physical and chemical fractionation techniques were used to verify that the PAR was directly produced in a protein-free form and not inadvertently released from protein during processing of PARP1 reactions. Moreover, the protein-free PAR had a distinct composition. A portion of the protein-free PAR chains had one end bearing an intact NAD⁺ group, a product that could not arise from PAR released from protein (Figs. 1 and 2). As detailed below, it was found that PARP1 creates protein-free PAR by initiating reactions on individual NAD⁺ or ADP-ribose nucleotides [29].

The mechanism of protein-free PAR production is best explained in the context of the structure of the ADP-ribosyltransferase (ART) domain, the defining feature of PARP family enzymes [31]. The ART binds to an NAD⁺ molecule that donates the ADP-ribose unit that is transferred onto a target biomolecule (Figs. 1 and 2). In addition to this “donor site” certain PARP enzymes bear an adjacent “acceptor site” [32]. The acceptor site engages a biomolecule-attached ADP-ribose unit, and then accepts an ADP-ribose from the donor site to create a ribose-ribose linkage and extend the chain. Mutagenesis of the acceptor site has supported this model of PAR chain growth on PARP1 (automodification), whereby the terminal ADP-ribose unit of a PAR chain docks in the acceptor site for addition of an ADP-ribose unit from the donor site [33]. Using X-ray crystallography, biochemistry, and mass spectrometry, a model was established in which PARP1 de novo synthesis of protein-free PAR results from an individual NAD⁺ or ADP-ribose molecule docking in the acceptor site (Fig. 2). These molecules then serve as an initiation site for adding a subsequent ADP-ribose unit from the donor site. PARP1 has a well-documented hydrolysis activity that cleaves NAD⁺ into ADP-ribose and nicotinamide and has been referred to as treadmilling [34]. The local production of ADP-ribose that can bind to the acceptor site and seed the formation of PAR chains can be considered a self-feeding activity for the generation of protein-free PAR. These experiments thus indicated that PARP1 was effectively capable of modifying small metabolites in addition to proteins and nucleic acids.

The initial observation of PARP1 production of free PAR was made using the isolated ART, which is constitutively active [35]. With the full complement of domains of PARP1 (Fig. 2), there is an autoinhibition mechanism in which a helical domain (HD) blocks the donor site and prevents NAD⁺ from binding. Upon detecting DNA strand breaks, a collection of PARP1 domains including the HD assemble in a complex that effectively reverses autoinhibition and allows NAD⁺ access to the donor site. PARP1 then targets itself and the local environment with ADP-ribose modifications. The automodification activity of PARP1 leads to release from DNA and a corresponding decrease in activity [36,37]. PARP1 activity is thus tightly regulated and increases dramatically in the presence of DNA strand break damage. Using full-length PARP1 and activating DNA in biochemical reactions, free PAR was observable and typically represented around 30 % of the output, with protein-linked PAR representing the remaining 70 %. Protein-free and protein-linked PAR were produced over the same time frame, indicating that the two activities operate under similar conditions and timing, rather than one preceding the other.

Two other PARP enzymes with both donor and acceptor sites, PARP2 and Tankyrase1, were also observed to produce free PAR in biochemical reactions [29]. PARP2 is regulated in a similar manner to PARP1 but with a distinct structural element in the ART that modulates HD autoinhibition [38]. TNKS activity is regulated by polymerization which relieves inhibition by an active site loop that likely prevents efficient use of NAD⁺ substrate [39,40]. Despite these distinct mechanisms of ART regulation, the biochemical production of free PAR by PARP enzymes is dependent on the presence of donor and acceptor sites in the ART.

1.2. Free PAR production in cells

An evaluation of PAR isolated from HEK293 cells and U2OS cells tested for evidence of direct free PAR production as observed biochemically [29]. Chemical fractionation of cells was used to separate protein-linked and protein-free PAR (insoluble and soluble fractions, respectively). PAR was evident but of low abundance in both fractions in untreated cells. The level in both fractions was substantially elevated in cells treated with the DNA damaging agent methylmethane sulfonate (MMS) and a PARG inhibitor (PARGi). A PARP1-specific inhibitor greatly reduced the levels of protein-free and protein-linked PAR. These results indicated that PARP1 was the predominant source of PAR in these conditions where MMS introduced activating DNA damage and PARGi prohibited PAR turnover. These conditions that yielded workable amounts of PAR were used to analyze the composition and abundance of protein-linked and protein-free PAR.

In addition to direct free PAR production by PARP1, the free PAR isolated from cells could have been generated by glycohydrolases that acted on protein-linked PAR to release fragments. However, the treatment with PARGi increased free PAR levels substantially and argued against PARG enzyme as a source of free PAR. Knockout cell lines for the other main PAR degrading enzymes (ARH3 and TARG1) did not have a significant impact on the amount of free PAR, further indicating that digestion of protein-linked PAR was not the source of free PAR. As noted above in the biochemical description, PAR chains initiated on an NAD⁺ molecule are unequivocal indications of direct free PAR production. These types of PAR chains were detectable, but not prominent, in the free PAR isolated from cells. To more firmly establish a link between PARP1 de novo PAR synthesis activity and the free PAR isolated from cells, a non-hydrolyzable NAD⁺ analog was used to elevate the NAD⁺-PAR chain signal in cells. Biochemically, PARP1 uses NAD⁺ to initiate PAR chains on carba-NAD⁺ molecules, a non-hydrolyzable analog of NAD⁺. Carba-NAD⁺ is resistant to PARP1 hydrolysis activity and elevates the level of chains initiated on carba-NAD⁺ compared to ADP-ribose. Carba-NAD⁺ was metabolically incorporated into cells using carba-nicotinamide riboside (carba-NR) as a precursor that was converted in cells to carba-NAD⁺. Under conditions of carba-NR treatment, there was a robust signal for free PAR chains starting with carba-NAD⁺. Thus, a distinct biochemical product that can only arise from de novo PAR synthesis was a major component of protein-free PAR isolated from cells. It is anticipated that NAD⁺-PAR chains are either utilized or otherwise processed in cells, and the non-hydrolyzable carba-NAD⁺ group prevented this utilization or processing. Collectively, these results indicated that PARP1 direct synthesis of protein-free PAR is the major contributor to the pool of free PAR molecules in cells experiencing DNA damage.

1.3. Regulation of free PAR synthesis by HPF1

HPF1 (histone PARylation factor 1) regulates PARP1 and PARP2 activity by binding to the ART and complementing the active site to allow modification of Ser residues [41–44]. In the absence of HPF1, PARP1/2 primarily modify Asp and Glu residues. HPF1 can also stimulate the PARP1 NAD⁺ hydrolase activity that converts NAD⁺ to ADP-ribose and nicotinamide [34]. Notably, HPF1 binding to PARP1/2 obstructs the ART acceptor site and thereby limits acceptor site access for PAR elongation reactions [43,44]. Consistent with acceptor site contribution to free PAR production, HPF1 reduced free PAR production in biochemical assays [29], suggesting a mechanism for HPF1 to downregulate free PAR production. In cell-based experiments, HPF1 knockout did not have a dramatic effect on free PAR levels [29]. A mild increase in PAR levels was observed in HPF1 knockout U2OS cells relative to wild-type cells, consistent with loss of an inhibitory effect on free PAR production. However, no significant change in PAR levels was observed for the same comparison made in HEK293 cells. The role of HPF1 in regulating free PAR production could depend on several factors. For example, the ratio of PARP1:HPF1 is typically 20:1 but could vary between cell types [45,46]. The low relative amounts of HPF1 could mitigate its impact on free PAR synthesis as well as protein-linked PAR production. Moreover, HPF1 is recruited to sites of damage slightly after PARP1 and the ratio PARP1:HPF1 ratio varies over time after DNA damage [47]. Thus, the initial analysis of HPF1 impact on free PAR levels might have focused on a window of PARP1 free PAR production that is largely independent of HPF1 (i.e. acute 30-minute response to MMS), but there could be other time frames or conditions where HPF1 has a greater role and could act to dampen free PAR signaling.

1.4. Free PAR and biomolecular condensates

There is growing appreciation of PAR as an organizer of cellular architecture, in particular in relation to PAR interactions that drive the formation of biomolecular condensates [16,48–51] (Fig. 3). For example, the FET family of proteins (FUS, EWSR1, and TAF15), strongly interact with PAR using RGG repeats and are recruited rapidly to sites of damage [51,52]. FUS is mutated in some cases of familial amyotrophic lateral sclerosis (ALS) [53,54], and it is recruited to sites of DNA damage through PARP1 production of PAR, forming condensates that are important for DNA repair [17,55]. Indeed, loss of FUS delays recruitment of the DNA repair factors 53BP1, NBS1, KU80, and SFPQ [55].

FUS interacts with purified free PAR molecules to form condensates in vitro [16–19]. Interestingly, very low concentrations of PAR are sufficient to induce FUS condensation at its physiological concentration [51]. In comparison, RNA-driven condensation of FUS requires a thousand-fold more RNA molecules than PAR molecules. The minimal length of PAR to support condensation was determined to be greater than 4 units. Interestingly, it was found that PAR is only needed to seed FUS condensation, after which it can be removed by PARG digestion without condensate dissolution [18]. In contrast, RNA-seeded condensates dissolve when the RNA is degraded. Furthermore, TNKS1 but not PARP1 was required for FUS granule formation in the cytoplasm [18]. The recent finding that free PAR can be produced directly by PARP1, PARP2, and TNKS1 (and presumably TNKS2) and can exist in the cell raises the possibility that this free PAR could play a role in FUS condensation in the nucleus and/or the cytoplasm without the need for a PAR glycohydrolase. The length of the free PAR produced by PARP1, PARP2, and TNKS1, which ranges from roughly 3–40 units for PARP1 and 3–20 units for PARP2 and TNKS1, would be sufficient to support PAR-induced FUS condensation even at low levels of free PAR production.

1.5. Free PAR and Parthanatos

Free PAR has been suggested to play an important role in Parthanatos, a cellular response that leads to cell death following over-activation of PARP1 in the presence of an excess of DNA damage [56]. In current models of Parthanatos, the PAR produced by PARP1 in the nucleus following DNA damage is exported to the mitochondria (Fig. 3), where it induces translocation of AIF (apoptosis-inducing factor) and MIF (macrophage migration inhibitory factor) to the nucleus and subsequently DNA fragmentation [25,27]. While one study suggested that PARG is detrimental to Parthanatos by degrading PAR [27], another study proposed that PARG is required to generate PAR chains that are exported out of the nucleus [57]. The regulation of Parthanatos is likely to be complex and involve multiple proteins and signals, but PARP1 hyperactivation and PAR production are key initiating elements [58]. The finding that PARP1 can directly produce free PAR without the requirement of a glycohydrolase activity suggests an alternative way to generate free PAR that can signal beyond the nucleus. The signaling capacity of free PAR likely extends beyond Parthanatos. For example, the Golgi associated protein PARP12 re-localizes to stress granules in the cytoplasm through binding to PAR produced in the nucleus by PARP1 [59], indicating a nucleus to cytoplasm transport of PAR (Fig. 3). The prevalent, direct production of free PAR provides a mobile signal that could extend its influence beyond protein-linked PAR modifications.

1.6. Free PAR versus protein-linked PAR

The discovery of PARP enzyme direct synthesis of protein-free PAR alongside protein-linked PAR raises new questions and possibilities for DNA damage signaling. For example, are there DNA damage signaling and repair roles that might be solely performed by the free PAR pool produced by PARP1, rather than protein-linked PAR? For recruiting repair factors to sites of DNA damage, protein-linked PAR seems better suited as a locally anchored signal. For action-at-a-distance roles, like signaling to the cytoplasm or mitochondria, free PAR seems better suited as a more easily diffusible signal (Fig. 3). The coincident production of protein-linked PAR and free PAR could allow versatility in the type of signaling that can be performed. However, free PAR might also impact the local repair environment. For example, a high local concentration of free PAR could seed biocondensates that persist after the initiating PAR signal dissipates. The free PAR might also serve as a diffusible decoy of sorts that could prevent over-recruitment of factors to the sites of DNA damage, in effect modulating the dynamics of the repair foci, perhaps similar to how the PAR-binding activity of PARP13 regulates the dynamics of cytoplasmic stress granules [60]. It will be challenging to delineate the contributions of protein-linked PAR versus free PAR since they both arise from the same active site. Progress could come from “knocking out” the primary sites for protein-linked PAR modification, and interpreting the remaining repair capacity or function to be mediated by free PAR production. The capacity to modulate the composition of the free PAR pool in a detectable way using carba-NAD⁺ will also permit progress in locating the signaling radius of free PAR in different cellular compartments. The NAD⁺ “handle” on one end of free PAR molecules could serve as a distinguishing feature that dictates specific functions for free PAR relative to protein-linked PAR. One can imagine PAR binding modules that are specific for this type of unique PAR composition. An important aspect of PAR signaling capacity is the rate of PAR turnover, and there could be differences in the turnover of free PAR relative to protein-linked PAR. For example, the regulation of PARG through recruitment to different cellular locations could bias PARG activity toward certain pools of PAR production. These ideas and certainly others will be exciting to investigate and further develop our understanding of the signaling capacity of poly(ADP-ribose).