Poly(ADP-ribose) glycohydrolase

Poly(ADP-ribose) glycohydrolase (PARG), originally identified in a calf thymus nuclear preparation (Ref. Reference Miwa and Sugimura25), is the primary hydrolase involved in PAR chain degradation, being 1–2 orders of magnitude more active than ARH3 in this regard (Ref. Reference Prokhorova26). In mice, a single gene encodes two main PARG isoforms: 60 kDa (localised in the cytoplasm and mitochondria) and 110 kDa (in the nucleus) (Refs Reference Lin27–Reference Meyer29). By contrast, humans have five PARG isoforms: 55 kDa (in the mitochondria), 60 kDa (in the cytoplasm and mitochondria), 99 kDa (in the cytoplasm), 102 kDa (in the cytoplasm) and 111 kDa (in the nucleus) (Refs Reference Meyer29–Reference Whatcott32) (Table 1). The large human PARG isoforms (99, 102 and 111 kDa) are known to be yielded through alternative splicing (Ref. Reference Meyer-Ficca30), whereas it remains unclear whether the short PARGs (55 and 60 kDa) result from degradation of the 111 kDa protein or alternative splicing.

In both humans and mice, the PARG gene consists of 18 exons. The full-length PARG protein comprises a putative regulatory domain (1–426 amino acids, encoded by exons 1–3) at the N-terminus and a conservative catalytic domain (486–838 amino acids, encoded by exons 9–14) at the C-terminus (Refs Reference Meyer29, Reference Min and Wang31) (Fig. 2). One study found that the existence of a mitochondrial targeting sequence (MTS) is the basis to ensure the catalytic activity of PARG, since the mutation of hydrophobic leucine residues in this element led to inactivation of the 59 and 111 kDa isoforms (albeit the specific mechanism behind this is unknown) (Ref. Reference Botta and Jacobson33). In human PARG, two primary α-helical sub-domains flanking a twisted, mixed, 10-stranded β-sheet core are arranged to form a central cleft above the β-sheet in the catalytic domain, whereas in the mouse, the enzyme has one mixed nine-stranded β-sheet (Refs Reference Tucker34, Reference Wang35). The deep cleft is the main ADPr-binding site and catalytic centre. After PARG binds to ADPr, the adjacent β ₁₂–α ₁₀ loop moves in a concerted manner near the binding site, and the Phe-902 side chain rotates to stack against the adenine moiety, which is secured by a network of direct and water molecule-mediated hydrogen bonds (Ref. Reference Tucker34). Glu-755 and Glu-756 (Glu-748 and Glu-749 in mice; Glu-114 and Glu-115 in bacteria) are the key catalytic residues (Refs Reference Tucker34–Reference Slade37). The critical ribose–ribose O-glycosidic linkage at the PAR terminal position is in direct hydrogen-bond contact with Glu-756, which then protonates the ribose' 2′-OH leaving group (Refs Reference Tucker34, Reference Slade37). A tightly bound water molecule, positioned by interactions with Glu-755 and Asp-737, is activated through the concomitant protonation by Glu-756 and attacks the oxocarbenium intermediate, resulting in the release of ADPr and short unbranched PAR chains (Refs Reference Tucker34, Reference Slade37).

Figure 2. Schematic diagrams of human ADP-ribose hydrolases. The hydrolytic domains are Macro (macrodomain) and Ribosyl_crysJ1 (ADP-ribosylation/Crystallin J1 fold). No reports related ARH2 structure has been published. The N-terminal putative regulatory domain of PARG consists of two nuclear localisation signals (NLSs) and two nuclear export signals (NESs). The catalytic C-terminal domain contains one mitochondrial targeting sequence (MTS), one NES and one NLS. The N-terminal region of MacroD1 also contains one MTS.

ADP-ribosyl-acceptor hydrolase

The ADP-ribosyl-acceptor hydrolase (ARH) family is an evolutionarily highly conserved structural module, adopting a predominantly α-orthogonal bundle architecture, typically consisting of 290–360 residues (Ref. Reference Ishiwata-Endo74). The protein family was first discovered in Rhodospirillum rubrum, where the activating factor dinitrogenase reductase-activating glycohydrolase (DraG) was found to reverse dinitrogenase reductase (Fe-protein), an inhibitor of Arg-ADPRylation (Ref. Reference Ludden and Burris75). Subsequently, a similar enzyme with comparable properties, designated ARH1, was identified in animal cells (Ref. Reference Moss, Jacobson and Stanley14).

Although the three members (ARH1, ARH2 and ARH3) of the ARH family exhibit significant amino acid sequence similarity (Refs Reference Rack, Palazzo and Ahel8, Reference Oka, Kato and Moss11, Reference Glowacki76), ARH1 and ARH3 display distinct substrate specificities: ARH1 cleaves MARylated substrates modified on Arg residues (Ref. Reference Moss77), whereas ARH3 hydrolyses PAR and O-acetyl-ADPr (OAADPr) as well as serine-linked MARylation (Ser-ADPr) (Refs Reference Fontana13, Reference Mueller-Dieckmann78–Reference Kasamatsu80) (Fig. 1, Table 1). By contrast, ARH2 has not been shown to hydrolyse MAR or PAR, because it lacks the amino acid residues critical for enzymatic activity, which are conserved in both ARH1 and ARH3 (Refs Reference Oka, Kato and Moss11, Reference Bu, Kato and Moss81, Reference Smith82).

Macrodomain-containing proteins MacroD1 and MacroD2

The macrodomain fold, which is a compact globular-shaped structure of approximately 25 kDa size, is widely distributed among all life forms (including viruses) and evolutionarily conserved (Ref. Reference O'Sullivan112). It is present in MacroD1, MacroD2 and TARG1, which possess the ability to hydrolyse MAR from proteins (Ref. Reference Schuhwerk113). In humans, MacroD1 and MacroD2 are highly similar members of the MacroD-type class. Although both were initially identified as deacetylases of OAADPr (Ref. Reference Chen114), they were later found to hydrolyse MAR, specifically targeting the ester bond formed by ADP-ribosylated Asp/Glu residues (Refs Reference Rosenthal15, Reference Jankevicius115) (Fig. 1). MacroD1 is predominantly localised to the mitochondrial matrix (Refs Reference Neuvonen and Ahola116, Reference Agnew117), whereas MacroD2 is distributed in the cytosol and nucleus (Refs Reference Jankevicius115, Reference Golia118) (Table 1). The mRNA of MacroD1 is expressed in various tissues, with a high level in skeletal muscle (Ref. Reference Žaja119), whereas the mRNA of MacroD2 is expressed in multiple tissues during the embryonic period, including the liver, brain, lung, thymus, heart, kidney, etc., highly expressed in the brain neurons at embryonic and adult stage (Refs Reference Ito120, Reference Nicole121). According to these reports, the distinct localisation of the macrodomain proteins suggests that they have unique roles to play, despite their structural similarities.

The approximately 35 kDa MacroD1 protein is encoded by 11 exons. Analysis of its crystal structure revealed that it comprises an N-terminal region (91–136 amino acids, encoded by exons 1–3), a macrodomain region (151–322 amino acids, encoded by exons 3–9) (Ref. Reference Chen114) and an MTS (1–85 amino acids, encoded by exon 1) (Ref. Reference Jankevicius115) (Fig. 2). MacroD2 is an approximately 47 kDa protein that is encoded by 17 exons. Similar to MacroD1, the crystal structure of MacroD2 constitutes an N-terminal region (7–47 amino acids, encoded by exons 1–2) and a macrodomain region (78–228 amino acids, encoded by exons 3–9) (Ref. Reference Wazir, Maksimainen and Lehtio122). In both MacroD1 and MacroD2, the N-terminal region is arranged as helical segments and a short β-sheet, while the macrodomain region resembles a canonical macrodomain fold, being composed of a three-layered α–β–α sandwich with a central six-stranded β-sheet (Refs Reference Chen114, Reference Wazir, Maksimainen and Lehtio122, Reference Yang123). MacroD1 and MacroD2 have similar catalytic mechanisms for mediating ADPr hydrolysis. Upon either enzyme binding to ADPr, structural re-arrangement occurs to ensure the correct positioning of the substrate (Refs Reference Jankevicius115, Reference Wazir, Maksimainen and Lehtio122, Reference Yang123). Meanwhile, the hydrogen bond network formed among water molecules, ADPr α-phosphates and other elements is responsible for the precise positioning of water molecule in the catalytic pocket (Ref. Reference Yang123). Then, ADPr α-phosphates activate the water molecule, allowing it to make a nucleophilic attack on the distal ribose C1" atom of ADPr, thereby cleaving the glycosidic bond between this atom and the acceptor Asp/Glu residue (Refs Reference Jankevicius115, Reference Yang123).

MacroD1, also known as leukaemia-related protein 16, is primarily localised in the mitochondrial matrix (Ref. Reference Žaja119). Nonetheless, it has been associated with several nuclear functions, including the activation of NF-κB signalling, binding and regulation of oestrogen and androgen receptors, and counteracting of PARP7-mediated MARylation (Refs Reference Wu124–Reference Bindesbøll127). MacroD1 has also been proposed as a negative regulator of the insulin signalling pathway through its down-regulation of insulin receptor substrate protein 1 (Ref. Reference Zang128). Additionally, MacroD1 expression and gene fusions have been implicated in the tumourigenesis of leukaemia and breast, gastric, liver, lung and colorectal cancers (Refs Reference Imagama129–Reference Hertweck132). Crawford et al. (Ref. Reference Crawford133) reported that Macrod1-knockout mice were viable and fertile but exhibited a female-specific motor coordination defect. Loss of MacroD1 in rhabdomyosarcoma cells resulted in mitochondrial fragmentation (Ref. Reference Žaja119). MacroD2 has been shown to undergo phosphorylation by ataxia-telangiectasia-mutated kinase in response to DNA DSBs and to be involved in reversing the ADPRylation of glycogen synthase kinase 3 beta, a key kinase in the WNT-mediated signal transduction pathway (Refs Reference Golia118, Reference Feijs134). However, the role of MacroD1 in DNA damage repair has yet to be fully elucidated.

MacroD2 is highly expressed in neuronal tissues and cells, which supports the observation that its mutation leads to a neurodegenerative phenotype (Refs Reference Lombardo135, Reference Jones136). The Macrod2 gene locus is linked to several neurological syndromes (e.g. autism, attention-deficit hyperactivity disorder and schizophrenia), which correlates with the notable expression of the protein in neurons during brain development (Ref. Reference Ito120). According to Crawford et al. (Ref. Reference Crawford133), Macrod2-knockout mice exhibited age-dependent hyperactivity along with a gait resembling bradykinesia, but neither Macrod1- nor Macrod2-knockout mice showed any defects in short-term working memory or attention span.

Terminal ADP-ribose glycohydrolase

Terminal ADP-ribose glycohydrolase (TARG1), also called O-acyl-ADP-ribose deacylase 1 (OARD1), which has been well characterised as an enzyme that hydrolyses MARylation, is ubiquitously expressed in various tissues and found in both the nucleus and cytoplasm, with high levels in the nucleolus (Ref. Reference Žaja119). It was initially found to be enriched in chronic lymphocytic leukaemia cells and subsequently demonstrated to possess deacylation activity, producing free ADPr from OAADPr, OPADPr (O-propionyl-ADPr) and OBADPr (O-butyryl-ADPr) deacylation (Refs Reference Marina137, Reference Peterson138). Additionally, TARG1 has been reported to remove ADPr units from MARylated PARP1, which is auto-modified at the Asp/Glu residues (Ref. Reference Sharifi16) (Fig. 1, Table 1).

The approximately 17 kDa TARG1 protein, which is encoded by 8 exons, consists of a macrodomain (1–152 amino acids, encoded by exons 4–8) that contains only the core domain and lacks the N- and C-terminal extension structures common in other macrodomains (Ref. Reference Peterson138) (Fig. 2). Similar to other macrodomains, the TARG1 macrodomain consists of a three-layer α–β–α sandwich containing a six-stranded β-sheet flanked by four α-helical elements (Refs Reference Sharifi16, Reference Peterson138). A hydrophobic pocket that can embed an adenosine moiety is formed by Leu-21 and Phe-22 on the β ₁–β ₂ loop, Ile-44 and Leu-47 on the β ₂–α ₁ loop, and Pro-118, Tyr-150 and the C-terminal Leu-152 residue (Ref. Reference Peterson138). The active centre located in the vicinity of the distal ribose is composed of Ser-35, Lys-84 and Asp-125 (Refs Reference Sharifi16, Reference Peterson138). The hydroxyl group of Ser-35 forms a hydrogen bond with the carbonyl oxygen of an ester, polarising the carbonyl bond (Ref. Reference Peterson138). Lys-84 forms a covalent intermediate with ADPr through the Amadori re-arrangement mechanism, following which Asp-125 hydrolysis and ADPr release ensue (Ref. Reference Sharifi16).

Because research on TARG1 is still in its infancy, the physiological role of this enzyme in organisms remains largely unknown. TARG1 deficiency caused the senescence of U2OS and 293T cells and decreased the proliferation of 293T but not HeLa cells (Refs Reference Sharifi16, Reference Bütepage139). TARG1 is enriched in the nucleolus, where its loss leads to an increase in the number of nucleoli and hyper-active transcription (Ref. Reference Žaja119), implying its importance in nucleolar homeostasis and function. Moreover, it can shuttle rapidly between the nucleoplasmic and nucleolar compartments (Ref. Reference Bütepage139). When DNA damage occurs, TARG1 moves quickly from the nucleolus into the nucleoplasm where it interacts with the PAR units enriched at the DNA damage site. The deletion of TARG1 rendered cells sensitive to DNA damage agents, topoisomerase II and ataxia telangiectasia and Rad3-related (ATR) inhibition and also damaged HR repair (Refs Reference Sharifi16, Reference Bütepage139, Reference Groslambert140), suggesting that the enzyme is involved in DNA repair. TARG1 mutations which result in truncated proteins without catalytic activity have been reported to cause neurodegeneration in humans (Ref. Reference Sharifi16). Whether this is due to the cytotoxicity of the truncated protein or an imbalance of MARylation remains unclear. Given these findings, the molecular mechanism underlying the roles of TARG1 in nucleolar homeostasis and the DDR remains to be studied.

ADPRylation occurs not only in proteins but also in DNA and RNA (Refs Reference Matta141–Reference Talhaoui144). DNA ADP-ribosyl transferase (DarT), a DNA-modifying PARP-like bacterial toxin, can be released by bacteria into human host cells where it targets single-stranded DNA during DNA replication to induce DNA ADPRylation, thereby triggering the DDR (Refs Reference Tromans-Coia145–Reference Cohen and Chang147). One study found that TARG1 could eliminate the toxic DNA ADPRylation induced by ectopic DarT expression in host cells, thereby ensuring normal cell proliferation (Ref. Reference Tromans-Coia145). This finding suggests that TARG1 plays a protective role against bacterial toxins analogous to DarT, similar to the effect of ARH1 against cholera toxin (Ref. Reference Kato89). In ovarian cancer cells, the PARP14-mediated site-specific MARylation of receptor for activated C kinase 1 (RACK1) promoted the assembly of stress granules in response to external stimuli (Ref. Reference Challa148). RACK1 is a scaffold protein of the 40S ribosome subunit and an essential member of stress granules (Refs Reference Buchan and Parker149, Reference Rabl150). TARG1 erases MARylation and dissociates RACK1 from stress granules (Ref. Reference Challa148), implying that it may serve as a potential molecular target for the treatment of ovarian cancer.

MacroD1, MacroD2 and TARG1 are the most extensively characterised macrodomain-containing hydrolases. Although they belong to the same family and hydrolyse MAR, they target different substrates through distinct molecular mechanisms (Ref. Reference Barkauskaite, Jankevicius and Ahel1). The physiological functions of these three enzymes remain elusive, and further research is necessary to comprehend their precise role in MAR hydrolysis. For instance, the nuclear function of MacroD1 and the reason behind its predominantly mitochondrial localisation warrant further study. Additional much-needed investigations include whether these enzymes exhibit substrate specificity, which depends on different MARylation sites and the physiological status of the cells and tissues.

Inhibitors of PAR hydrolases

PARylation and ADPr metabolism play fundamental roles in DNA repair to maintain genome stability, and the targeting of PARylation synthesizing PARP enzymes within this pathway has shown therapeutic potential (Ref. Reference Zong19). Various PARPi have been developed for cancer treatment, with olaparib, niraparib and rucaparib already approved for clinical use against specific tumour types (Refs Reference Slade17, Reference Lord and Ashworth151, Reference Zeng152). Recently, PARGi have also generated immense interest in relation to the pharmacological intervention of different maladies and emerged as promising targets for a variety of cancers and other diseases (Refs Reference Min and Wang31, Reference Kassab, Yu and Yu153–Reference Pillay158).

Proflavine, ethacridine, ellipticine, daunomycin and tilorone, which are all DNA intercalators, were found to inhibit PARG activity, which may be indirectly caused by their insertion into the DNA molecule, resulting in the release of inhibitory histones (Ref. Reference Tavassoli, Tavassoli and Shall159). However, owing to the high cytotoxicity of these DNA intercalators, a second generation of PARGi based on tilorone (viz. GPI16552 and GPI18214) were developed. GPI16552 significantly reduced the cerebral infarction volume in a rat model of focal cerebral ischaemia and attenuated the inflammatory response and tissue damage caused by spinal cord trauma (Refs Reference Lu160, Reference Cuzzocrea161). Treatment with GPI18214 attenuated zymosan-induced multiple organ failure in mice, reducing peritonitis in the animals as well as their mortality rate (Ref. Reference Genovese162). GPI16552 and GPI18214 can mitigate splanchnic artery occlusion as well as reperfusion- and dinitrobenzene sulphonic acid-induced intestinal injury (Refs Reference Cuzzocrea163, Reference Cuzzocrea164). PARG inhibition significantly reduces the expression of the pro-inflammatory cytokines tumour necrosis factor alpha (TNFα) and interleukin 1 beta (IL1β) as well as neutrophil infiltration (Refs Reference Cuzzocrea161, Reference Genovese162, Reference Cuzzocrea164), which account for the protective effect of PARGi.

Tannins, particularly gallotannins and ellagitannins, are naturally occurring polyphenol compounds with PARG-inhibiting activity, which is mediated through their competitive binding to PAR with PARG (Ref. Reference Tsai165). With regard to the oligomeric forms of ellagitannin, the dimer nobotanin B exhibits stronger PARG inhibitory activity than the trimer nobotanin E and the tetramer nobotanin K (Ref. Reference Tsai165). Gallotannins and nobotanin B significantly reduce the oxidative death of astrocytes and neurons by accumulating PAR and thus slowing its turnover, thereby blocking PARP1-mediated cell death (Ref. Reference Ying166). PARG inhibition by gallotannin decreased ischaemic brain damage and significantly ameliorated infarct formation and neurological deficits in rats (Ref. Reference Wei167), indicating that PARGi have a neuroprotective function. Galloyl-glucose derivatives based on gallotannin have potent PARG inhibitory activity, albeit with low cell permeability, reducing PARP1-dependent cell death to some extent (Ref. Reference Formentini168). ADP-HPD is an amino-ribose analogue of ADPr, in which half maximal inhibitory concentration (IC₅₀) reaches the nanomolar grade (Ref. Reference Slama169). However, because of its low cell permeability, it is mostly used in in vitro studies to elucidate the structure and catalytic activity of PARG. Pargamicin, a cyclic peptide isolated from an Amycolatopsis sp. fermentation product, has weak PARG inhibitory activity (Ref. Reference Masutani170). With the advancement of identification methods, a series of PARGi xanthene compounds (Eosin Y and Phloxin B) (Ref. Reference Okita171), salicylanilide (Ref. Reference Steffen172), RBPIs (rhodanine-based PARGi) (Ref. Reference Finch173) and phenolic hydrazide hydrazones (Ref. Reference Islam174) with IC₅₀ values in the micromolar range have been reported. Recently, newly developed quinazolinedione sulphonamides (PDD00017273 and PDD00017238) and a thio-xanthine/methylxanthine derivative (JA2131) have achieved PARG inhibitory activity comparable to that of ADP-HPD (Refs Reference James175–Reference Houl177). PDD00017273 is effective against pancreatic ductal adenocarcinoma cells carrying the BRCA2 mutation but has no effect on breast cancer cell lines with BRCA1 or BRCA2 mutations (Refs Reference James175, Reference Jain178), implying that its PARG inhibitory effect is potent for only certain types of cancer. JA2131 inhibits PARG activity by binding to the enzyme's adenine-binding pocket, thereby killing cancer cells by promoting PAR accumulation and γH2AX foci in the nucleus (Ref. Reference Houl177). COH34, a small molecule with anti-tumour effects, can bind specifically to PARG to inhibit its activity at sub-nanomolar concentrations, resulting in extended PARylation, which blocks DNA repair and inhibits the growth of cancer cells with DNA repair defects (Ref. Reference Chen and Yu179). Recently, the company IDEAYA Biosciences has conducted a clinical trial of PARGi for patients bearing tumours harbouring HR deficiency (HRD), including ER+, Her2-, HRD breast cancer and HRD ovarian cancer (https://www.ideayabio.com/pipeline/). Deploying PARGi or targeting dePARylation enzymes represents a novel clinical approach to fighting human diseases, including cancer.

As mentioned previously, ARH3 is the only known ADPr hydrolase to remove ADPr from serine residues, which is the major type of ADPRylation of histones during DNA damage repair. Thus, ARH3 inhibitors are currently being developed as a chemotherapeutic strategy for tumour suppression (Ref. Reference Liu180). In contrast to the PARP family, which consists of multiple members, the PARG family comprises only a single member. This characteristic lends the advantage of specificity to targeted PARG therapy and helps to overcome the issue of tumour cell resistance to PARPi (Ref. Reference Schuhwerk113). Furthermore, PARP1 is highly abundant in cells, with an estimated 10⁶ molecules per cell (Ref. Reference Ludwig181). By contrast, each cell contains approximately 2000 PARG molecules (Ref. Reference Hatakeyama182), suggesting that PARG may offer enhanced potency and cell-type specificity for cancer treatment. Of note, ADPr homeostasis involves not only dePARylation but also deMARylation. Therefore, the identification or development of inhibitors targeting both these processes could have a profound impact on cancer research. Although many PARGi have been reported, whether they also target other enzymes or proteins relevant to ADPr degradation has not been well studied.

The successful Food and Drug Administration (FDA) and European Medicines Agency (EMA) approvals of the clinical use of PARPi for cancer treatment have gained momentum (Ref. Reference Parkes and Kennedy183). However, owing to the issue of cancer resistance to PARPi in some patients, alternative therapeutic strategies need to be explored. Targeting PARG and other ADPr hydrolases to modulate PAR metabolism may represent a novel approach for cancer chemotherapy. The elucidation of the X-ray crystal structure of PARG (Refs Reference Wang35, Reference Slade37, Reference Barkauskaite184) is considered a significant achievement in ADPr hydrolase research, as the findings have the potential to greatly expedite the development of novel and potent PARGi, which will undoubtedly contribute to the creation of therapeutic drugs that target ADPr-degrading enzymes.

Perspectives

ADPRylation is a prevalent PTM that regulates various cellular pathways and is involved in disease pathogenesis processes, including the DDR, cell death, transcription, chromatin remodelling, neurodegenerative disorders and inflammatory reactions (Ref. Reference Suskiewicz185). Unravelling the molecular and biological functions as well as substrate specificities of the synthetising and degrading enzymes of PARylation and MARylation will aid in the elucidation of their roles in specific signalling pathways and the identification of potential targets for clinical applications. Different ADPr hydrolases have distinct substrates and sub-cellular localisations, working collaboratively to coordinate the removal of ADPRylation modifications. PARG predominantly degrades (long and branched) PAR chains but has limited activity on short PAR polymers, indicating that other erasers, such as ARH3, may act in concert with the hydrolases to erase ADPr efficiently (Ref. Reference Barkauskaite184). Complete reversal of MARylation is performed solely or in concert by MacroD1, MacroD2, TARG1, ARH1 and ARH3. However, despite great efforts, the specific similarities and differences in the hydrolysis process catalysed by these enzymes, as well as the precise nature of their regulation, remain poorly understood. Given the importance of PARylation and MARylation in diverse signalling networks, further investigations into the roles of the erasers of these PTM processes in different cell types and tissues, as well as in cancerous versus non-cancerous cells, are needed. These studies will not only help to elucidate the biological roles of the ADPr hydrolases but also provide valuable reference information for the future development of chemotherapeutic strategies against PAR homeostasis-related diseases, including cancer.