The Expanding Landscape of ADP-Ribosylation: Protein, DNA, RNA, and Mitochondrial Regulation

Abstract

ADP-ribosylation is an essential post-translational modification that contributes to key cellular processes, such as DNA damage repair, cell-cycle progression, chromatin remodeling, mitochondrial function, and immune responses in mammalian cells. This modification derives from NAD⁺ and is regulated by dedicated writer, eraser, and reader proteins that govern its installation, removal, and recognition. Traditionally viewed as a protein-centered modification, ADP-ribosylation has recently been extended to nucleic acids, with ADP-ribosylated DNA and RNA now identified in both mammalian and bacterial systems. These discoveries reveal previously underappreciated layers of nucleic acid-based regulation and suggest that NAD⁺-dependent chemistry integrates genome maintenance, RNA metabolism, and cellular stress responses. In this review, we first outline the major mammalian ADP-ribosylation machineries, including the families of writer, eraser, and reader proteins, and discuss how their activities are coordinated. We then examine emerging roles of ADP-ribosylation in mitochondria, with a focus on mitochondrial DNA repair and metabolic control. Finally, we highlight recent advances in understanding NAD⁺-dependent modifications of DNA and RNA in mammalian and bacterial cells, including terminal and nucleobase-linked ADP-ribosylation and NAD capping, and discuss outstanding questions regarding their physiological functions and interplay with protein post-translational modification and other nucleic acid modifications.

Graphical Abstract

1. Introduction

ADP-ribosylation is a reversible modification that regulates key mammalian signaling pathways, including DNA damage response, chromatin structure, cell cycle progression, and mitochondrial function.^1-5 It involves the transfer of ADP-ribose from nicotinamide adenine dinucleotide (NAD⁺) to target molecules, most prominently proteins (Figure 1).⁶ ADP-ribosylation occurs either as mono-ADP-ribose (MAR) or poly-ADP-ribose (PAR), with PAR chains being linear or branched.^7-9 Protein ADP-ribosylation has been mapped to multiple amino acids, including His, Ser, Tyr, Asp, Glu, Arg, Lys, and Cys residues.^2,10-17 These modifications are dynamically installed by ADP-ribosyltransferases (ARTs), recognized by reader proteins, and removed by eraser enzymes (Figure 1B). ^15,18 For example, PARP10 MARylates GSK3β and controls its kinase activity,¹⁹ and this mark can be reversed by macrodomain hydrolase MacroD2.²⁰ The central role of ADP-ribosylation in genome maintenance and cell survival has motivated the development of PARP inhibitors, such as talazoparib and Olaparib, for cancer therapy.¹⁵

Figure 1. Regulation of mono- and poly-ADP-ribosylation.

(A) Chemical structure of nicotinamide adenine dinucleotide (NAD⁺), the donor substrate for ADP-ribosylation reactions. (B) Schematic overview of the dynamic installation, recognition, and removal of ADP-ribose on substrates. Writer enzymes catalyze mono-ADP-ribosylation (MARylation) or poly-ADP-ribosylation (PARylation), reader proteins recognize ADP-ribose modifications, and eraser enzymes remove ADP-ribose.

Beyond proteins, recent studies have demonstrated that NAD⁺-dependent ADP-ribosylation extends to DNA and RNA in both mammalian and bacterial systems, forming ADP-ribosylated nucleic acids.¹⁸ In parallel, NAD⁺ can serve as a noncanonical RNA cap, giving rise to NAD-capped transcription products that influence RNA stability and processing.²¹ These nucleic acid-linked structures are regulated by dedicated writers and erasers and are increasingly recognized as components of gene-regulatory and host–pathogen networks. Free PAR chains, generated independently of covalent substrates, have likewise emerged as signaling entities in DNA repair and other pathways, as summarized in a recent review.²² A variety of chemical biology tools now enable more selective detection and enrichment of ADP-ribosylated species and have been reviewed elsewhere.²³

In this review, we first outline the major mammalian writers, erasers, and readers of ADP-ribosylation, with a particular focus on mitochondrial regulation. We then concentrate on emerging evidence for DNA and RNA ADP-ribosylation, NAD-capped RNA, and protein RNAylation in mammalian or bacterial systems. Throughout, we highlight areas of agreement and controversy in the current literature and identify key technical and conceptual challenges that must be addressed to define the physiological functions of nucleic-acid ADP-ribosylation.

2. Regulation of ADP-ribosylation in the mammalian system

2.1. ADP-ribosylation writer enzymes

ARTs fall into three major families: diphtheria-toxin-like enzymes (ARTDs, also known as PARPs), cholera-toxin-like ARTs (ARTCs), and NAD⁺-dependent sirtuins.¹⁸ Together, these enzymes generate mono-ADP-ribosylation (MARylation) and poly-ADP-ribosylation (PARylation) on diverse substrates and thereby regulate DNA damage signaling, transcription, chromatin structure, metabolism, and immune responses.

Diphtheria Toxin-like ARTs (ARTDs/PARPs)

Diphtheria toxin-like ARTs are commonly referred to as poly(ADP-ribose) polymerases (PARPs).²⁴ There are 17 mammalian PARP members known to date.^18,24 PARP3, PARP4, PARP6–8, PARP10–12, and PARP14–16 primarily act as MAR writers,^25,26 whereas PARP1, PARP2, Tankyrase 1 (TNKS1), and Tankyrase 2 (TNKS2) are PAR writers.^25,26 PARP9 and PARP13 lack detectable transferase activity.²⁵ Collectively, PARPs control processes such as gene expression, DNA damage response, intracellular signaling, chromatin organization, cell proliferation and apoptosis, and innate and adaptive immunity.^27-31 Comprehensive discussions of PARP family protein structure, signaling pathways, and therapeutic targeting have been provided elsewhere.^{15,24,27,32-34}

Among PARP proteins, PARP1 is the best-characterized DNA damage sensor and accounts for the majority of DNA damage-induced ADP-ribosylation.^2,35 Upon recognizing DNA strand breaks, PARP1 is recruited to DNA damage sites,³⁶ where it PARylates targets such as histones, transcription factors, and other nucleic-acid-processing enzymes, including FEN1.^37,38 PARP1 also undergoes extensive automodification; auto-PARylation drives a transition from a chromatin architectural factor to a histone chaperone.³⁹ Beyond genome maintenance, PARP1 also contributes to CD8+ T cell maturation, underscoring its broad roles in immunity.⁴⁰

Recent work has clarified how DNA binding allosterically regulates PARP catalytic activity. PARP1 and PARP2 are modular enzymes in which N-terminal regulatory (REG) domains sense DNA breaks and transmit an activating signal to a conserved C-terminal catalytic (CAT) domain. Makwana et al. showed that, in vitro, REG and CAT domains from certain PARPs can assemble in trans into catalytically competent complexes and that DNA-bound REG stimulates PARylation even when REG and CAT are on separate polypeptides, indicating that DNA-dependent allosteric activation does not strictly require a single-chain architecture.⁴¹ REG from human PARP2 robustly cross-complemented CAT from Arabidopsis PARP2, underscoring at least partly modular DNA-break recognition and allosteric activation across selected PARP orthologs.

Deeksha et al. demonstrated that PARP1 activity is also regulated by single-stranded DNA (ssDNA), a DNA context relevant to replication intermediates and repair gaps.⁴² The ZnF1–ZnF2 tandem of PARP1 mediates high-affinity ssDNA binding and ssDNA-dependent stimulation of PARP1 activity, establishing ssDNA as a bona fide allosteric activator alongside classical single- and double-strand breaks. PAR and ssDNA are recognized by distinct domain combinations—ZnF3–BRCT–WGR for PAR versus ZnF1–ZnF2 for ssDNA—and PAR binding can displace ssDNA and suppress ssDNA-dependent PARP1 activity, revealing negative feedback by the PAR product on DNA-stimulated catalysis. Together, these studies highlight that PARP engagement with diverse DNA structures may exert allosteric control over its activity and shape when DNA serves primarily as an allosteric activator versus a covalent ADP-ribose substrate in vivo, which will be discussed in the following section. Other detailed summaries of PARP1 structure, catalytic activity, regulation, and drug development are available in recent reviews.^16,43

Importantly, not all PARP family members function solely as protein “writers”. For example, although PARP9 lacks canonical transferase activity, its macrodomain 1 exhibits ADP-ribosyl hydrolase activity.^25,44,45 PARP14 possesses MAR writer, reader, and eraser activities in a single protein.^44,45 Moreover, PARP1, PARP2, and TNKS1 can synthesize free PAR de novo in addition to modifying protein substrates.⁴⁶ These multifaceted activities expand the functional repertoire of the PARP enzyme family and foreshadow the complexity of ADP-ribosylation signaling on both proteins and nucleic acids.

Cholera Toxin-like ARTs (ARTC)

The human ARTC family comprises hARTC1, hARTC3, hARTC4, and hARTC5, which are localized at the cell surface or secreted into the extracellular space and function as ecto-enzymes.⁴⁷ hARTC1 and hARTC5 are active mono-ADP-ribosyltransferases,⁴⁷ whereas hARTC3 and hARTC4 are catalytically inactive.⁴⁷ In humans, ARTC2 is a nonfunctional gene,^48,49 while in mice, two isoforms, ARTC2.1 and ARTC2.2, retain MARylation activity.⁵⁰ hARTC1 can interact with hARTC3, and the interaction enhances the enzymatic activity of hARTC1.⁵¹ In addition, hARTC1 modifies Arg50 of the vesicle-associated membrane protein-associated protein B (VAPB) to regulate calcium homeostasis.⁵¹ Moreover, hARTC1 also ADP-ribosylates chaperone GRP78 to modulate ER stress responses.⁵² Additional aspects of ARTC structure, catalytic activity, cell signaling, and disease relevance have been reviewed previously.⁴⁷

Sirtuins

Sirtuins are NAD⁺-dependent class III lysine.⁵³ The seven mammalian sirtuins (SIRT1–7) participate in inflammatory responses, aging, oxidative stress, cell death, DNA damge and nutrient metabolism.^54-57 They utilize a NAD⁺ cofactor to remove lysine acylation and generate O-acetyl-ADP-ribose (OAADPr).⁵⁸ In addition to the deacetylase activity, sirtuins also exhibit ADP-ribosyltransferase activity in selected cases.

SIRT6 is known as a chromatin-associated histone deacetylase.⁵⁹ Mouse SIRT6 can catalyze intramolecular auto-MARylation in vitro⁶⁰ and MARylation of BAF170 (a subunit of BAF chromatin remodeling complex) to regulate transcription activation in mouse embryonic fibroblasts.⁶¹ The MARylation activity of SIRT6 is activated by binding to double-stranded (ds) DNA.⁶² The identification of a polyhistidine tract, present in several previously identified SIRT6 MARylation substrates, explain the substrate selectivity, although the mechanism by which the motif facilitates MARylation remains to be elucidated. Similarly, SIRT7 undergoes auto-MARylation and is recognized by histone variant macroH2A (mH2A).⁶³ This interaction is further enhanced under glucose starvation, linking ADP-ribosylation signaling to cellular nutritional status.⁶³ Collectively, sirtuin-mediated ADP-ribosylation adds an additional layer of regulation that connects chromatin dynamics, stress responses, metabolism, and NAD⁺ homeostasis.

Taken together, ARTDs/PARPs, ARTCs, and sirtuins distribute ADP-ribosylation activity across intracellular and extracellular compartments and across proteins and, in some cases, nucleic acids. Several family members combine writer, reader, and eraser functions in a single polypeptide, and many are intimately coupled to NAD⁺ metabolism. These features are particularly relevant for understanding how ADP-ribosylation is deployed on mitochondrial substrates and on DNA and RNA, as discussed in subsequent sections.

2.2. ADP-ribosylation eraser enzymes

ADP-ribosylation is a reversible post-translational modification that is critical for cellular processes such as DNA repair and stress response.⁸ Dysregulation of ADP-ribosylation eraser, Poly(ADP-ribose) glycohydrolase (PARG), has been linked to liver, breast, and prostate cancers, and has prompted the development of small-molecule inhibitors.⁶⁴ Also, mutations in the eraser TARG1 are associated with severe neurodegeneration.⁶⁵ These observations underscore the importance of ADP-ribosylation turnover in disease pathology. Comprehensive reviews have summarized the structural features, catalytic mechanisms, and pharmacological inhibition of ADP-ribosylation erasers.^8,15,34 Here, we provide an overview of the major classes of ADP-ribosylation eraser enzymes and their biological functions.

PARG

PARG is an important enzyme responsible for hydrolyzing PAR chains and can also reverse ester-linked MARylation on aspartate and glutamate residues in cells.^66,67 PARG exhibits both endo- and exo-glycosidic activities, generating free ADP-ribose monomers or shorter polymers.^7,8,68 PARG is critically involved in single-strand break repair^69,70 and the double-strand break response.⁷¹ Loss of PARG leads to reduced cell proliferation, PARylation accumulation, and sensitivity to DNA-damaging agents such as N-methyl-N'-nitro-N-nitrosoguanidine and menadione in the absence of PARP inhibition.⁷² Consistent with these cellular phenotypes, PARG^−/− mice are embryonic lethal.⁷² PARG also removes PARylation from PCNA and facilitates the interaction between FEN1 and PCNA, suggesting a role for PARG in DNA replication and Okazaki fragment processing.⁷³ Together, these findings highlight PARG as a central regulator of ADP-ribosylation homeostasis and genome stability.

ARH

The ADP-ribosylhydrolase (ARH) family comprises three members: ARH1, ARH2, and ARH3.⁷⁴ These enzymes exhibit distinct substrate specificities and biological functions. ARH1, but not ARH2 or ARH3, hydrolyzes ADP-ribose from arginine residues modified by cholera toxin.⁷⁵ Mouse embryonic fibroblasts (MEF) lacking ARH1 (ARH1^−/−) or overexpressing an inactive ARH1 variant proliferate faster than wild-type (ARH1^+/+) or ARH1^−/− cells reconstituted with the wild-type ARH1.⁷⁶ In vivo, wild-type ARH1^+/+ mice show a lower incidence of spontaneous tumors than ARH1^−/− and ARH1^+/− mice, indicating a tumor suppressor role for ARH1.⁷⁶ In contrast, ARH2 has no detectable ADP-ribose hydrolytic activity but has been implicated in cardiac function, tumor biology, and inflammation.^75,77 ARH3 hydrolyzes PAR to ADP-ribose and O-acetyl-ADP-ribose (OAADPr) and does not cleave arginine-, cysteine-, diphthamide-, or asparagine-linked ADP-ribose.^74,75 ARH3 also degrades protein PARylation in the mitochondrial matrix⁷⁸ and removes serine and tyrosine ADP-ribosylation.^79,80 ARH3-deficient cells accumulate MAR on histones, leading to altered histone marks and transcriptional profiles.⁸¹ Moreover, treatment of ARH3 knockout cells with PARG inhibitors results in synthetic lethality and altered chromatin modification and gene expression.⁸² Collectively, these findings highlight ARH3 as a critical regulator of ADP-ribosylation and a potential therapeutic target.

TARG1

TARG1, also known as C6orf130, removes MARylation from glutamate and aspartate residues.^8,65,83 It can also hydrolyzes O-acetyl-, O-propionyl-, and O-butyryl-ADP-ribose, releasing ADP-ribose and acetate, propionate, and butyrate.⁸⁴ Loss of TARG1 sensitizes cells to inhibitors of topoisomerase II, ATR, and PARP.⁸³ TARG1-knockout cells combined with PARG inhibition result in synthetic lethality. These findings suggest that TARG1 may influence responses to DNA-damaging agents and PARP-targeted therapies.⁸³

MACROD1 and MACROD2

Macrodomain proteins are a family of evolutionarily conserved proteins involved in chromatin remodeling, transcriptional regulation, host-pathogen interactions, and other cellular processes.⁸⁵ Human macrodomain-containing proteins MACROD1 and MACROD2 bind MARylated proteins and hydrolyze the O-linked glycosidic ester bond ADP-ribose.^20,86,87 Both enzymes remove MAR from glutamate and aspartate residues^20,85,87,88 and hydrolyze O-acetyl-ADP-ribose.⁸⁹ MACROD2 overexpression in subsets of human breast cancers has been found to mediate tamoxifen resistance and regulate estrogen-independent growth.⁹⁰ In addition, MACROD2 plays a critical role in maintaining chromosome stability, and loss of MACROD2 reduces PARP1 transferase activity and attenuates DNA repair signaling.⁹¹ Detailed information on macrodomain protein structure and their potential as drug targets is provided in a recent review article.⁸⁵

NUDT family members

Nudix (nucleoside diphosphate-linked moiety X) motif hydrolases constitute a broad enzyme family that cleaves pyrophosphate bonds in nucleotide-derived metabolites, including ADP-ribose, and oxidative nucleotides.^86,92-94 Here we will focus on NUDT 5, 9 and 16 which enzymes intersect with multiple facets of ADP-ribosylation signaling and ADP-ribose and PAR-derived metabolites. NUDT16 is a Nudix hydrolase toward protein-conjugated ADP-ribose.⁹⁵ It regulates PAR chains from the DNA damage effector 53BP1, thereby regulating 53BP1 stability, recruitment to double-strand breaks, and its role in cell survival under genotoxic stress.⁹⁶ In addition, NUDT16-mediated removal of ADP-ribose from the histidine methyltransferase SETD3 prevents the ubiquitin-proteasome-dependent degradation of SETD3 and promotes tolerance to radiotherapy.⁹³

NUDT5 hydrolyzes free ADP-ribose and PAR-derived ADP-ribose in the nucleus but does not process PARylated proteins.^8,97 In estrogen-stimulated MCF7 breast cancer cells, PARG first converts PAR to ADP-ribose, and NUDT5 then converts ADP-ribose to ATP in the presence of pyrophosphate, providing a local nuclear ATP source that supports chromatin remodeling, transcriptional activation, and DNA repair.⁹⁷ NUDT5 knockdown impairs homologous recombination (HR) but not non-homologous end joining (NHEJ), indicating DNA repair pathway-specific dependence on nuclear ADP-ribose catabolism and local ATP production.⁹⁸ Besides its ADP ribosylation hydrolase function, recent work has also uncovered a non-enzymatic function of NUDT5 in repressing purine nucleotide synthesis. NUDT5 acts as a scaffold that promotes the oligomerization of phosphoribosyl pyrophosphate amidotransferase (PPAT), the rate-limiting enzyme in de novo purine biosynthesis^99,100, thereby restraining purine nucleotide levels independently of its hydrolase activity.

NUDT9 is a mitochondrial ADP-ribose pyrophosphatase that converts ADP-ribose to AMP and ribose-5-phosphate, limiting the accumulation of free ADP-ribose in the mitochondrial matrix.^92,101,102 Through this activity, NUDT9 is proposed to protect against the potential toxicity of excessive ADP-ribose and to help support mitochondrial nucleotide and redox homeostasis. Together, these NUDT family members illustrate how Nudix hydrolases control ADP-ribosylation at several levels: by buffering free ADP-ribose pools in mitochondria, reversing protein ADP-ribosylation on key genome-maintenance and stress-response factors in the nucleus, and recycling PAR-derived metabolites into nuclear ATP to fuel energy-intensive chromatin transactions.

ENPP1

ENPP1 (ectonucleotide pyrophosphatase/phosphodiesterase 1) is a transmembrane enzyme¹⁰³ involved in a wide range of physiological and pathological processes,¹⁰³ including osteoblastic differentiation,¹⁰⁴ skeletal and soft tissue diseases,¹⁰³ cancer, and immune function.^105,106 ENPP1 hydrolyzes extracellular ATP into AMP and pyrophosphate,¹⁰⁷ and degrade STING agonist cyclic GMP-AMP (cGAMP), , thereby dampening cGAS–STING–dependent innate immune signaling.¹⁰⁶ ENPP1 also hydrolyzes ADP-ribosylated proteins to generate protein-conjugated ribose-5’-phosphate in vitro.¹⁰⁸ These activities highlight the functional complexity of ENPP1 at the interface of nucleotide metabolism, extracellular ADP-ribosylation turnover, and underscore its potential as a therapeutic target in multiple disease settings.

Overall, ADP-ribosylation erasers can be broadly divided into enzymes that cleave PAR chains, hydrolases that remove MAR from specific amino acids or nucleic acids, and enzymes that process free or extracellular ADP-ribose–containing species. The functional partitioning shapes the lifetime, length, and linkage of PAR and MAR chains and determines whether ADP-ribosylation acts as a transient signaling cue or a more persistent modification. It is also particularly relevant for nucleic-acid ADP-ribosylation, where the availability and specificity of hydrolases such as TARG1, MACROD1/2, and ARH3 likely dictate whether DNA and RNA ADP-ribosylation function as repair intermediates, regulatory marks, or toxic lesions.

2.3. Reader proteins of ADP-ribosylation

ADP-ribosylation is a reversible post-translational modification that plays a central role in the DNA damage response, where it functions as a signaling scaffold to recruit DNA repair factors to sites of damage. One of the earliest ADP-ribose readers reported was the macrodomain protein Af1521 from a thermophile, which was shown to bind ADP-ribose and protein PAR chains with high affinity.¹⁰⁹ Subsequent studies identified Amplified in Liver Cancer 1 (ALC1), a macrodomain-containing chromatin remodeler, as a PAR-binding protein.^110,111 ALC1 exhibits PAR-dependent chromatin remodeling activity and facilitates DNA repair, and alters cellular sensitivity to DNA-damaging agents when overexpressed or depleted.^110-112

The WWE domain represents another PAR-recognition module. In 2012, the WWE domain of the ubiquitin E3 ligase RNF146 was shown to specifically recognize iso-ADP-ribose, a structural unit of PAR.¹¹³ Overexpression of the RNF146 WWE domain perturbs PAR dynamics in response to DNA damage.¹¹⁴ PARP13/ZAP, a zinc-finger antiviral protein containing a ZnF5–WWE1–WWE2 architecture, is another ADP-ribose reader.^115,116 Structural and biochemical analyses revealed that the WWE2 domain, but not WWE1, mediates interaction with ADP-ribose.¹¹⁵

Additional reader domains involved in the DNA damage response have also been identified. The tandem BRCA1 C-terminal (BRCT) motifs of BARD1 recognize PAR and promote recruitment of the BRCA1–BARD1 complex to DNA damage sites.¹¹⁷ Forkhead-associated (FHA) domains of APTX and PNKP can bind to iso-ADP-ribose, and the BRCA1 C-terminal (BRCT) domains of Ligase 4, XRCC1, and NBS1 interact with ADP-ribose.¹¹⁸ Through these interactions, PAR-binding BRCT and FHA modules facilitate DNA repair and regulate cell cycle progression.¹¹⁸ More comprehensive discussions of BARD1 function in cancer biology¹¹⁹ and BRCT domain structures¹²⁰ are available in previous reviews.

Several additional motifs and structural elements have been characterized as ADP-ribose readers.² The poly(ADP-ribose)-binding zinc finger (PBZ) motif in APLF (aprataxin PNK-like factor)¹²¹ and the mitotic checkpoint protein CHFR can interact with ADP-ribose.¹²² The 20-amino acid motif containing two conserved regions enriched in basic and hydrophobic residues, such as those in p53, DNA ligase III, and XRCC1, plays an important role in binding to PAR.¹²³ The RNA recognition motif 1 (RRM1) of NONO binds PAR, and the recruitment of NONO facilitates NHEJ while suppressing HR.¹²⁴ The PAR-binding regulatory (PbR) motif in Chk1 interacts with ADP-ribose and regulates cell-cycle progression,³ whereas the oligonucleotide/oligosaccharide-binding-fold motif of hSSB1 recognizes PAR and is recruited to DNA damage sites.⁵

Collectively, these studies demonstrate that ADP-ribosylation is decoded by a diverse array of reader proteins and structural motifs, enabling MAR and PAR to function as versatile signaling platforms. Through these interactions, ADP-ribosylation orchestrates DNA damage responses, cell-cycle control, RNA processing, and broader cellular signaling pathways. Reflecting the growing importance of ADP-ribose readers, chemical proteomic strategies have been developed to systematically identify novel ADP-ribosylation-binding proteins.^125-127 The biological functions and mechanistic roles of many newly identified readers, particularly outside the canonical DNA damage response, remain to be elucidated.

3. ADP-ribosylation in mitochondria

ADP-ribosylation in mitochondria remains less well characterized than in the nucleus, but recent work has begun to reveal mitochondrial writers, erasers, and their links to mitochondrial DNA (mtDNA) maintenance and metabolism.

3.1. PARP1 and mitochondrial DNA repair

In 2009, Rossi and colleagues demonstrated that a fraction of PARP1 localizes to mitochondria in human fibroblasts and HeLa cells, where its interaction with an inner membrane protein Mitofilin is required for mitochondrial targeting.¹²⁸ PARP1 was found in complexes with mtDNA and DNA ligase III. Knocking down either PARP1 or Mitofilin disrupts mitochondrial PARP1 location and leads to mtDNA damage accumulation, suggesting a role of PARP1 in mtDNA damage and repair. In contrast, a 2014 study by Bartosz Szczesny et al. showed that knockdown of PARP1 increases mtDNA integrity under basal and oxidative stress conditions in human lung adenocarcinoma A549 cells, in opposition to its positive role in nuclear DNA repair.¹²⁹ In mitochondria, PARP1 interacts with base excision repair (BER) enzyme EXOG and DNA polymerase γ, and its presence impaired coordination among mitochondrial BER factors and limited mtDNA repair and biogenesis. These opposing findings indicate that PARP1 can both support and restrain mitochondrial genome maintenance depending on context and cell types, highlighting the need for future work to reconcile experimental systems, detection methods, and levels of PARP1 perturbation.

To further explore mitochondrial PARylation, Lee and colleagues used subcellular fractionation to verify the mitochondrial localization of PARP1 and detected PARylated products in isolated mitochondria.¹³⁰ PARylation is stimulated by exogenous NAD⁺, and the ADP-ribose chromatin affinity purification (ADPr-ChAP)-mapped PARylation pattern of mitochondrial nucleoids reveals a correlation between PAR signal and TFAM distribution on mtDNA. NAD⁺ treatment increases PARylation, mtDNA transcription, and TFAM occupancy in the D-loop region. On the other hand, Hopp et al. reported that rotenone-induced mitochondrial ADP-ribosylation persisted in PARP1 knockout cells, as measured by anti-ADP-ribosylation immunofluorescence, suggesting that PARP1 is not the predominant mitochondrial ADP-ribosyltransferase under those conditions.¹³¹ These results underscore uncertainties about which enzymes drive mitochondrial ADP-ribosylation under different stress conditions and in vivo.

3.2. Mitochondrial PARylation and NEURL4

Neuralized-like protein 4 (NEURL4) has recently emerged as a mitochondrial ART.¹³² NEURL4 localizes to mitochondria, and NEURL4 knockout markedly reduces mitochondrial PARylation and causes cytosolic mtDNA leakage and dysregulated cellular homeostasis. SILAC-based proteomics identified approximately 170 NEURL4-dependent ADP-ribosylated proteins, such as mtDNA ligase III, the rate-limiting enzyme in mitochondrial base excision repair and carbamoyl phosphate synthetase 1 (CPS1), a key regulator of urea cycle. NEURL4-dependent PARylation of mtLIG3 is required for proper mtDNA maintenance, and NEURL4 haploinsufficiency in mice leads to mtDNA deletions in spermatocytes and impaired sperm function. These findings position NEURL4 as a central mitochondrial ART that coordinates mtDNA integrity and metabolic signaling.

3.3. Mitochondrial erasers: PARG and MACROD1

ADP-ribosylation erasers also localize to mitochondria. A PARG isoform with a mitochondria-targeting sequence has been detected in the organelle.^133-135 Besides, MACROD1 is highly expressed in human and mouse skeletal muscle cells and localizes predominantly to mitochondria.^136,137 Cardiomyocyte-specific MACROD1 depletion in mice protects mitochondrial function during septic insult by increasing MARylation of Ndufb9 in mitochondrial complex I.¹³⁸ On the other hand, MACROD1 knockout in diabetic cardiomyopathy mice aggravates glycemic control, cardiac remodeling, and mitochondrial dysfunction, indicating context-dependent effects.¹³⁹ A recent study revealed that MACROD1 is essential for maintaining metabolic homeostasis and optimal mitochondrial function.¹⁴⁰ Loss of MACROD1 in mice impairs muscle function and exercise capacity.¹⁴⁰ In C2C12 myoblast cells, MacroD1 knockdown increases reactive oxygen species production and mitochondrial fission and and shifts glucose flux away from the TCA cycle toward the pentose-phosphate pathway, consistent with a compensatory increase in antioxidant production.¹⁴⁰ Together, these data identify MACROD1 as a key mitochondrial MAR hydrolase that supports mitochondrial integrity and metabolic homeostasis.

3.4. NAD⁺-mediated nuclear–mitochondrial crosstalk and ADP-ribosylation regulation

Mitochondrial NAD⁺ availability dynamically shapes ADP-ribosylation in both mitochondria and the nucleus, with mitochondria acting as a buffer for subcellular NAD⁺ pools. Hopp et al. showed that mitochondrial ADP-ribosylation is reversible and responds to metabolic perturbation.¹³¹ Respiratory chain inhibition increases mitochondrial ADP-ribosylation, whereas H₂O₂-induced oxidative stress reduces mitochondrial ADP-ribosylation and strongly induces nuclear ADP-ribosylation, consistent with an almost inversely correlated distribution of strong mitochondrial versus nuclear signals controlled by NAD⁺ redistribution. Elevated mitochondrial ADP-ribosylation dampens genotoxin-triggered nuclear ARTD1 (PARP1) PARylation and increases ARTD1 chromatin retention, while release of NAD⁺ from mitochondria by the uncoupler FCCP lowers mitochondrial NAD⁺, modestly decreases mitochondrial ADP-ribosylation, and reduces PARP inhibitor efficacy, necessitating higher inhibitor concentrations to block H₂O₂-induced nuclear PARylation.

Høyland et al. further showed that subcellular NAD⁺ pools are interconnected, with mitochondria functioning as a rheostat that buffers NAD⁺ when consumption is elevated in other compartments.¹⁴¹ Mitochondria maintain their NAD⁺ supply by import via SLC25A51 and, when NMNAT3 is present, by reversibly cleaving NAD⁺ to nicotinamide mononucleotide (NMN) and ATP, forming a “virtual” NAD⁺ pool stored as NMN that buffers changes in demand. Together, these findings support a model in which mitochondrial NAD⁺ acts as a central, dynamic reservoir that constrains when and where ADP-ribosylation can occur, thereby coordinating mitochondrial and nuclear ADP-ribosylation and linking DNA damage responses to respiratory activity, NAD⁺ transport, and NAD⁺ biosynthesis.

4. DNA ADP-ribosylation

Accumulating evidence indicates that ADP-ribosylation is not limited to proteins but can also occur on DNA, and that some of the writers and erasers known for protein ADP-ribosylation also act on DNA substrates. In this section, we summarize recent findings on DNA ADP-ribosylation and its regulation in mammalian and bacterial systems, emphasizing mechanistic diversity and remaining controversies.

4.1. Mammalian systems

DNA terminal ADP-ribosylation

In 2016, Talhaoui et al. showed that PARP1 and PARP2 can catalyze PARylation at DNA termini (Figure 2A) in vitro using ³²P-labeled oligodeoxynucleotides and gel-based assays.¹⁴² PARP1 ADP-ribosylates recessed dsDNA and oligodeoxynucleotides bearing 3′- or 5′-terminal phosphates,¹⁴² whereas PARP2 preferentially modifies nicked or gapped dsDNA with 5'-phosphate ends.¹⁴² These DNA adducts were validated by Matrix-Assisted Laser Desorption/Ionization Time-of-Flight (MALDI-TOF) mass spectrometry at the oligodeoxynucleotide level and can be reversed by PARG in vitro.¹⁴² Subsequent work showed that PARP1 preferentially ADP-ribosylates 3'-phosphorylated DSB termini,¹⁴³ while PARP3 adds MAR to DNA termini that can be removed by hydrolases PARG, TARG1, MACROD2, and ARH3.¹⁴⁴ Consistent with these findings, PARP2 was found to PARylate and PARP3 to MARylate 5′- and 3′-terminal phosphate groups at DSB and SSB ends within a DNA substrate.¹⁴⁵ Beyond the PARP family, the highly diverged PARP homolog TRPT1 (tRNA 2′-phosphotransferase 1) can ADP-ribosylate 5′-phosphorylated single-stranded DNA (ssDNA) termini.¹⁴⁶ Together, these studies establish terminal DNA phosphates as bona fide ADP-ribosylation acceptors and that both PARPs and non-PARP ARTs can generate these modifications in vitro.

Figure 2. Potential ADP-ribosylated DNA structures.

(A) Representative terminal ADP-ribosylation of DNA. (B) ADP-ribosylation of deoxyadenosine at the N1 position. (C) ADP-ribosylation of deoxycytidine at the N3 position. (D) ADP-ribosylation of deoxythymidine at the N3 position. (E) ADP-ribosylation of deoxyguanosine at the exocyclic N² position.

Building on these findings, Matta et al. systematically examined DNA substrate specificity of PARP1-catalyzed DNA PARylation using dsDNA substrates harboring defined a SSB and overhangs at varying lengths.¹⁴³ They found that a 3′-terminal phosphate at a blunt double-strand break positioned ~13 or 23 bp from a single-strand gap serves as a major acceptor for PARP1-mediated PARylation and that PARP1 preferentially modifies such 3′-phosphorylated DSB termini over auto-PARylation, both with recombinant enzyme and in HeLa extracts. DNA PARylation efficiency is highly sensitive to the distance and helical phase between the activating gap and the acceptor terminus, supporting an SSB-oriented mechanism in which PARP1 bound at a nick or gap modifies a terminal phosphate located on the same face of the DNA helix within roughly one to two turns. These results reinforce that PARP1-dependent DNA ADP-ribosylation is strongly constrained by local DNA break architecture and suggest that the biological impact of DNA PARylation depends on the configuration of complex strand-break clusters generated during damage and repair.

A recent study extended terminal DNA ADP-ribosylation to telomeric DNA, showing that PARP1 ADP-ribosylates telomeres and that this modification is removed by TARG1.¹⁴⁷ Telomeric DNA ADP-ribosylation was detected on the lagging-strand unligated Okazaki fragments and on the 3′ telomere overhang. Loss of TARG1 leads to the accumulation of telomeric DNA ADP-ribosylation and telomere shortening.¹⁴⁷ These data support a model in which terminal DNA ADP-ribosylation participates in DNA end processing and telomere maintenance, but the extent to which these modifications occur in vivo under physiological conditions and how they influence pathway choice in DNA repair remain open questions.

DNA nucleotide ADP-ribosylation

In addition to terminal modifications, DNA ADP-ribosylation has been detected on nucleobases.^148,149 Musheev et al. reported that the N1 position of adenine residues in DNA (Figure 2B) can be PARylated by PARP1 in ssDNA, but not in dsDNA or RNA.¹⁴⁹ The authors demonstrated that adenine PARylation occurs independent of DNA strand breaks using in vitro assays and characterized N1-Poly(ADP-ribosyl)-dA adducts by liquid chromatograph-tandem mass spectrometry (LC-MS/MS). N1-substituted dA adducts readily deaminate to form N1-ribosyl-deoxyinosine (N1-ribosyl-dI). The authors confirmed the presence of N1-ribosyl-dI in pig liver and mouse kidney DNA using LC-MS/MS after extensive ssDNA and modified nucleoside enrichment, indicative of their low abundance in genomic DNA. A recent preprint from the same group estimated a level of ~6 modifications per genome, corroborating their earlier results.¹⁴⁸ Notably, the authors also reported that deoxycytidine can also be PARylated by PARP1 in ssDNA at the N3 position (Figure 2C), as evidenced by LC-MS/MS analysis with a synthetic standard, although the adduct was below the detection limit in big liver DNA.¹⁴⁸ In vitro, the N3-dC adduct deaminates rapidly, and the PARylation is reversible by TARG1 but not PARG, highlighting distinct eraser specificities for nucleobase-linked versus terminal adducts.

There is a clear discrepancy between the nucleobase-centered model and terminal DNA ADP-ribosylation described above. Musheev et al. tested 5′- and 3′-terminally labeled ssDNA substrates and found that PARylation did not protect a 5′-terminal ³²P label from phosphatase treatment and remained sensitive to phosphatase when the 3′ terminus lacked a hydroxyl group,¹⁴⁹ arguing against 5′- or 3′-terminal PARylation under their experimental conditions. These data support N1-Poly(ADP-ribosyl)-dA as the dominant DNA PARylation product in their system. Differences in substrate design, PARP1 and cofactor concentrations, and enrichment and detection strategies likely contribute to these divergent conclusions between nucleobase- and terminal-focused studies. At present, it remains unresolved whether terminal phosphates or nucleobases are the predominant physiological ADP-ribose acceptors in DNA, whether these chemistries occur in distinct cellular contexts, and how frequently nucleobase PARylation occurs in vivo relative to protein PARylation. Addressing these questions will require highly sensitive and orthogonal methods capable of distinguishing terminal from nucleotide-linked ADP-ribose in biological samples.

4.2. Bacterial systems

DarT and DarG constitute a bacterial toxin-antitoxin system that regulates DNA ADP-ribosylation in species such as Mycobacterium tuberculosis, Thermus aquaticus, and Vibrio cholerae.^150-152 Two DarTG subfamilies have been described, DarTG1 and DarTG2.¹⁵² In 2016, DarT2 was shown to catalyze sequence-specific thymidine ADP-ribosylation (Figure 2D) on ssDNA, thereby inhibiting DNA replication and bacterial growth;^151,153 the cognate macrodomain antitoxin DarG2 removes this DNA ADP-ribosylation.^151,153 In the DarTG1 system, DarT1 catalyzes guanine ADP-ribosylation (Figure 2E) on ssDNA, and DarG1, a NADAR superfamily hydrolase, reverses this modification.¹⁵² These DarTG modules act as a phage defense system by ADP-ribosylating viral DNA in bacteria, blocking productive infection.¹⁵⁴ On the other hand, some bacteriophages encode an anti-DarT factor NADAR (AdfN), a DarG1-like enzyme that removes DarT-induced ADP-ribosylation from phage DNA, illustrating the evolutionary arms race between bacteria and phages.¹⁵⁵ Comprehensive discussions of the DarTG system structure, regulation, and enzymology are available in a previous review.¹⁵⁶

Beyond bacterial systems, DarT-induced thymidine ADP-ribosylation can be reversed by human TARG1 in vitro.¹⁵⁷ The expression of DarT in TARG1-deficient mammalian cells leads more sensitive to heightened sensitivity and a robust DNA damage response. These findings indicate that mammalian hydrolases can resolve DarT-generated ADP-ribosylation and that unrestrained DNA ADP-ribosylation is highly genotoxic.

Building on the DarT-DarG system, a genome engineering approach termed “append editing” has been developed, in which a catalytically weakened DarT2 is fused to Cas9 nickase to induce site-specific ADP-ribosylation of thymidine residues.¹⁵⁸ This strategy promotes homologous recombination in bacteria and enables precise thymine-to-adenine or adenine/cytosine substitutions with minimal indels in yeast, plant, and human cells, thereby expanding precision genome editing.¹⁵⁸ While these applications highlight the utility of DNA ADP-ribosylation as a programmable handle, they also underscore how potent and disruptive nucleotide-linked ADP-ribose modifications can be when not efficiently reversed.

Overall, work in mammalian and bacterial systems demonstrates that DNA ADP-ribosylation encompasses both terminal and nucleobase-linked chemistries and can act as a regulatory signal, a damage-associated lesion, or a weapon in host–pathogen conflict. A major challenge moving forward is to define the prevalence, sequence specificity, and physiological roles of these modifications in vivo, to determine under which conditions DNA serves primarily as an allosteric activator versus a covalent substrate, and to clarify when they function as part of controlled signaling circuits versus toxic marks that must be promptly repaired.

5. RNA ADP-ribosylation, NAD-capped RNA and Protein RNAylation

RNA carries numerous chemical modifications, including the 5′ m⁷G cap, N¹-methyladenosine, and N⁴-acetylcytidine, which regulate RNA stability and translation.¹⁵⁹ Emerging evidence shows that NAD⁺-dependent RNA ADP-ribosylation and NAD⁺ capping add further regulatory layers in both mammalian and bacterial systems.^18,21 This section summarizes recent discoveries on RNA terminal and nucleotide ADP-ribosylation (Figure 3A and B), NAD-capped-RNA (Figure 3C), adenosine of ADP-ribosylation at the N⁶ position (Figure 3D), and ADP-ribosylates 2′-hydroxyl groups of double-stranded RNA (Figure 3E), and highlight key knowledge gaps.

Figure 3. RNA ADP-ribosylation and NAD-capped RNA.

(A) Structure of the terminal RNA of ADP-ribosylation. (B) ADP-ribosylation of uridine at the N1 position. (C) Structure of NAD⁺-capped RNA. (D) CmdT catalyzes ADP-ribosylation at the N6 position of adenine within GA dinucleotides on ssRNA. (E) ADP-ribosylates 2′-hydroxyl groups of double-stranded RNA.

5.1. Mammalian systems

RNA terminal ADP-ribosylation and nucleotide ADP-ribosylation

In vitro studies have established that several human ARTs can ADP-ribosylate RNA termini (Figure 3A). Using Cy3- and ³²P-labeled substrates, Munnur et al. showed that PARP10 ADP-ribosylates the phosphorylated 5′ and 3′ ends of ssRNA, and that phosphorylated 5′ ssRNA is also ADP-ribosylated by PARP11, PARP15, and human TRPT1.¹⁴⁶ PARP10 and TRPT1-installed RNA terminal ADP-ribosylation can be removed in vitro by PARG, TARG1, MACROD1, MACROD2, and ARH3.¹⁴⁶ Cell-based experiments with combined hydrolase knockdown indicate that PARG is the dominant hydrolase for RNA ADP-ribosylation in human cells.¹⁶⁰

Biochemical and structural data now firmly establish human TRPT1 as a phosphate-dependent nucleic acid ADP-ribosyltransferase. Munnur et al. first demonstrated that human TRPT1 ADP-ribosylates 5′-phosphorylated ssRNA ends in vitro.¹⁴⁶ An earlier study by Munir et al.,¹⁶¹ which focused on the canonical RNA 2′-phosphotransferase reaction and on a subset of archaeal and bacterial Tpt1 enzymes, did not detect RNA ADP-ribosylation by human TRPT1 under their assay conditions. However, the authors did show that some Tpt1 homologs, notably Aeropyrum pernix Tpt1, synthesize a 5′-phospho-ADP-ribosylated RNA/DNA cap. Weixler et al.¹⁶² later demonstrated that human TRPT1 preferentially modifies 5′-G/A-phosphorylated ssRNA, that 5′-ADP-ribosylated mRNA is resistant to XRN1 yet translationally inactive, and that TRPT1 overexpression combined with hydrolase knockdown increases ADP-ribose signals across multiple RNA classes in cells. Most recently, Yang et al.¹⁶³ solved crystal structures of human, mouse, and yeast TRPT1 bound to NAD⁺/ADPRp. Together with mutational and biochemical assays, they showed that eukaryotic TRPT1s share a conserved PARP-like ART fold, bind NAD⁺ with micromolar affinity, and ADP-ribosylate 5′-phosphorylated ssRNA and ssDNA via a flexible SGR-containing donor loop, using partially distinct catalytic and nucleic-acid-binding residues from those required for RNA 2′-phosphotransferase activity. Collectively, these studies indicate that nucleic acid ADP-ribosylation is an evolutionarily conserved, intrinsic activity of TRPT1/Tpt1 family enzymes.

Other PARP family members also contribute to RNA ADP-ribosylation. In cells combined the knockdown of hydrolases TARG1, PARG, and ARH3, overexpression of PARP10, PARP11, PARP12, or PARP15 produces distinct ADP-ribose signals across mRNA, large RNA, and small RNA.¹⁶² Treatment with IFN-α, arsenite, MG132, H₂O₂, or EBSS (Earle's balanced salt solution) differentially modulates these RNA-associated signals, suggesting that individual PARPs are engaged by specific stress pathways. In addition, PARP12 has been shown to ADP-ribosylate viral RNA in mammalian cells, accelerating its decay and inducing antiviral host gene expression, consistent with a direct antiviral role for RNA ADP-ribosylation.¹⁶⁴ Intriguingly, macrodomain modules embedded within PARP proteins can also act as RNA ADP-ribose erasers. PARP14 macro1 and PARP9 macro1 remove ADP-ribose from synthetic ADP-ribosylated RNA in vitro,¹⁶⁰ mirroring their ability to hydrolyze protein ADP-ribosylation.^44,45 Overall, these findings broaden the repertoire of RNA ADP-ribose writers and erasers, but nucleotide-resolution maps of endogenous ADP-ribosylation sites and the physiological roles of phosphate-linked RNA ADPr in mammalian cells remain largely undefined.

In addition to RNA terminal ADP-ribosylation, human PARP10 has been identified as ADP-ribosylating uracil bases activity in RNA (Figure 3B).¹⁶⁵ Human TARG1 and TARG1-like macrodomain proteins can reverse this uracil base ADP-ribosylation in different species.¹⁶⁵ This study suggests that nucleotide ADP-ribosylation can occur not only on DNA but also on RNA.

NAD⁺-capped RNA

NAD⁺ can act as a noncanonical 5′ cap on eukaryotic RNAs, including mRNAs and small nucleolar RNAs (snoRNAs), generating NAD⁺-capped RNAs (Figure 3C) that are distinct from both ADP-ribosylated RNA caps and the m⁷G cap.¹⁶⁶ NAD⁺-capped RNAs are decapped by the enzymes DXO/Rai1 family enzymes, which regulate intracellular NAD⁺-capped RNA levels and RNA decay. Additional decapping activities are provided by Nudt12 and Nudt16, which also hydrolyze NAD⁺-RNA caps and link NAD-capping to the Nudix hydrolase network.^21,167,168 Levels of NAD⁺-capped RNA respond to cellular NAD⁺ concentrations, suggesting crosstalk between RNA and NAD⁺ metabolism.¹⁶⁹ Comprehensive reviews of the sequencing and detection methods of NAD⁺-capped RNA can be found in elsewhere.^21,170 Together, these studies establish an alternative mammalian RNA cap structure beyond the canonical m⁷G cap and underscore the complexity of NAD⁺-mediated RNA regulation in different biological contexts.

5.2. Bacterial systems

RNA terminal ADP-ribosylation

In bacterial systems, RNA ADP-ribosylation is also reversible and enzyme-specific. The TRPT1 homologue KptA from Streptomyces coelicolor catalyzes ADP-ribosylation of 5′-phosphorylated RNA termini in vitro, similar to human TRPT1.¹⁴⁶ This terminal RNA ADP-ribosylation is reversed by human MACROD1 and SCO6450, a MACROD-like protein from S. coelicolor, but not by the macrodomain DarG protein, underscoring enzyme and substrate specificity within bacterial ADP-ribose cycles.¹⁴⁶ Although these data demonstrate the reversibility and enzyme-specificity of RNA terminal ADP-ribosylation, the endogenous RNA targets and stimuli that engage KptA in S. coelicolor are still unknown.

RNA nucleotide ADP-ribosylation

Bacteria also deploy internal RNA ADP-ribosylation in anti-phage defense. In the CmdTAC system, bacteriophage infection is sensed when CmdC detects viral capsid proteins, leading to ClpP-mediated degradation of the antitoxin CmdA and activation of the toxin CmdT, an ADP-ribosyltransferase.¹⁷¹ CmdT catalyzes ADP-ribosylation at the N⁶ position of adenine within GA dinucleotides on ssRNA (Figure 3D), thereby modifying host or viral transcripts and blocking phage propagation.¹⁷¹ This mechanism reveals mRNA ADP-ribosylation as a previously unrecognized bacterial antiviral strategy, conceptually analogous to DNA-directed toxin–antitoxin systems. More broadly, NAD⁺ utilization as a signaling currency in bacterial immune systems has been comprehensively reviewed elsewhere.¹⁷² It remains unclear how widespread N⁶-adenosine ADP-ribosylation is beyond CmdTAC and whether related mechanisms exist in other bacteria or RNA viruses.

A complementary strategy is mediated by the type VI secretion system effector RhsP2 from Pseudomonas aeruginosa, which targets structured RNAs in competing bacteria.¹⁷³ RhsP2 is an ART toxin whose catalytic domain resembles classical protein-targeting toxins but instead ADP-ribosylates 2′-hydroxyl groups of double-stranded RNA (Figure 3E), with identified substrates including the tRNA pool and the RNA-processing ribozyme RNase P. This 2′-O-linked RNA ADP-ribosylation inhibits aminoacyl-tRNA synthetase charging and compromises RNase P–mediated tRNA maturation, leading to translational arrest and target cell death. Together with CmdT-mediated N6-adenosine RNA ADPr, RhsP2 illustrates that bacteria exploit chemically distinct RNA ADP-ribosylation strategies to disable essential RNA metabolism in rival cells.

NAD⁺-capped RNA

NAD⁺ also functions as a noncanonical 5’ cap on bacterial mRNAs.¹⁷⁴ In several species, NAD⁺ is incorporated co-transcriptionally by RNA polymerase, generating NAD⁺-capped RNAs that can be removed by specialized decapping enzymes. In Escherichia coli, the Nudix phosphohydrolase NudC cleaves the NAD⁺ cap and promotes mRNA decay through RNase E.¹⁷⁵ In Bacillus subtilis, NAD⁺ is similarly incorporated into RNA de novo via RNA polymerase, which is removed by RNA pyrophosphohydrolase BsRppH.¹⁷⁶ NAD⁺-capped RNAs resist RNase J1 and are stabilized, and BsRppH deletion alters gene expression.¹⁷⁶ Thus, whereas NAD⁺ caps in mammalian cells often promote RNA turnover, NAD⁺-capped RNAs in bacteria enhance transcript stability. The determinants that specify stabilizing versus destabilizing outcomes across species are not yet defined.

Protein RNAylation

Beyond RNA metabolism alone, bacteriophages exploit RNA ADP-ribosylation chemistry to reprogram host translation via protein RNAylation (Figure 4).¹⁷⁷ Bacteriophage T4 encodes three ADP-ribosyltransferases (ModA, ModB and Alt).¹⁷⁸ ModA ADP-ribosylates host RNA polymerase, and ModB targets ribosomal protein S1.¹⁷⁸ Recent work showed that T4 ART ModB can use NAD⁺-capped RNA as a donor to covalently attach entire RNA chains to the arginine residues of acceptor proteins, such as ribosomal proteins rS1 and rL1, directly reprogramming the host translation machinery during infection.¹⁷⁷ The full spectrum of RNA and protein substrates, the reversibility of RNAylation, and whether analogous mechanisms operate in eukaryotic systems remain open questions.

Figure 4. Reaction of protein RNAylation.

T4 ART ModB uses NAD⁺-capped RNA as a donor to covalently attach an RNA chain to the acceptor protein.

6. Future perspectives

ADP-ribosylation is a dynamic and reversible modification regulated by dedicated writer, reader, and eraser proteins. Emerging roles in genome maintenance, stress responses, and disease have spurred interest in defining its mechanisms and therapeutic potential. Several key questions remain open, particularly regarding interaction networks and chromatin architecture, crosstalk with other modifications, nucleic acid–linked ADP-ribosylation, and mitochondrial regulation.

Expanding the functional ADP-Ribosylation landscape through the PARP Interactome and Chromatin Architecture

ADP-ribosylation can be regulated not only by the canonical writers, readers, and erasers machinery, but also by its associated interactome and chromatin architecture. First, the PARP functions are shaped by their interaction partners. For example, HPF1 (histone PARylation Factor 1) interacts with PARP1 and PARP2 at DNA damage sites^179-181 and redirects PARP1 and PARP2 auto-PARylation and histone PARylation.¹⁸¹ HPF1-dependent histone ADP-ribosylation is also sensitive to nucleosome structure and the positioning of DNA damage. Specifically, when a one-nucleotide gap is present in nucleosomal DNA, HPF1 enhances histone PARylation (heteromodification), an effect that is more pronounced for PARP2 than for PARP1.¹⁸² This observation reinforces the role of HPF1 as a chromatin context–dependent regulator of PARylation.¹⁸² Furthermore, HPF1 can also enhance the binding of inhibitors, such as olaparib, to PARP1, linking the PARP interactome directly to drug response.¹⁸⁰ Given the growing interest in the PARP protein family interactome, recent studies have systematically mapped PARP interaction networks using TurboID technology.¹⁸³ However, the functional roles of many PARP interactors remain incompletely understood. Therefore, further mechanistic investigation of the PARP interactome will be important for deepening our understanding of ADP-ribosylation biology and for refining PARP-targeted therapies.

Second, PARylation can weaken histone–DNA interactions, promote nucleosome unwrapping and increase local DNA accessibility.¹⁸⁴ This structural remodeling facilitates the recruitment of downstream DNA repair factors and supports the formation of PAR-dependent biomolecular condensates at sites of DNA damage, which concentrate repair machinery and enhance spatial organization and the DNA damage response.¹⁸⁴ Together, these findings indicate that the regulatory landscape of ADP-ribosylation extends well beyond the canonical writers, readers, and erasers to include their associated interactomes and PAR-driven chromatin reorganization as functionally integral components.

ADP ribosylation crosstalk with other post-translational modifications and DNA repair pathway

ADP-ribosylation is a central regulator of DNA damage signaling, chromatin dynamics, and cancer biology. Increasing evidence highlights its crosstalk with other post-translational modifications (PTMs), including ubiquitination, in coordinating DNA repair processes. For example, in the repair of DNA–protein cross-links (DPCs), PARP1 functions as the primary lesion sensor, depositing poly(ADP-ribose) (PAR) chains directly onto the cross-linked substrate.¹⁸⁵ This PARylation regulates the ubiquitination of the DPC, targeting it for proteasomal degradation.¹⁸⁵ According to a preprint, PARP1-dependent PARylation also serves as a recruitment signal for the metalloprotease SPRTN via its Nudix homology domain, thereby facilitating proteolytic degradation of the cross-linked protein.¹⁸⁶ These findings illustrate how PARP1 and ADP-ribosylation act in concert with ubiquitination to orchestrate DPC recognition and removal.

Beyond indirect crosstalk, ADP-ribosylation can also serve as a direct platform for ubiquitin conjugation. ADP-ribose units are recognized by RNF146 and other RNF family E3 ligases, which couple PAR recognition to ubiquitin-mediated proteolysis.¹¹³ Recent work shows that DELTEX E3 ligases can attach ubiquitin directly to ADP-ribosyl residues on protein substrates¹⁸⁷ and that the ADP-ribosylation on DNA and RNA can be ubiquitylated by DTX3L and DTX2.^188,189 Together, these studies emphasize the importance crosstalk of ADP ribosylation to other PTMs and DNA repair pathway.

In addition to canonical PTMs, emerging metabolic modifications further integrate ADP-ribosylation into epigenetic regulation. Lysine lactylation, a metabolite-driven PTM derived from lactate during the Warburg effect,^190-192 has been shown to be involved in DNA repair pathway through modulation of NBS1 lactylation.¹⁹³ Also, PARP1 itself can be modified by lysine lactylation, suggesting reciprocal regulation between these pathways.^194,195 Histone deacetylases (HDACs), which participate in DNA repair,^196,197 also exhibit both “writer” and “eraser” activities toward lysine lactylation and other lysine acylations,^198-201 raising the possibility that HDACs coordinate with PARP-dependent ADP-ribosylation and lactylation to regulate chromatin accessibility and repair pathway choice.

Sirtuins add another layer of complexity. Sirtuins generate O-acetyl-ADP-ribose as a byproduct of lysine deacetylation²⁰² and have now been shown to undergo auto-ADP-ribosylation.^60,63 These findings suggest that ADP-ribosylation may play a key regulatory role via sirtuin proteins. O-acetyl-ADP-ribose and related metabolites can be hydrolyzed by macrodomain eraser proteins,⁸⁹ suggesting metabolic feedback between sirtuin activity and ADP-ribose turnover. However, how this sirtuin–ADP-ribose axis integrates with DNA damage responses and chromatin regulation remains largely unexplored.

Collectively, these studies underscore the intricate crosstalk among ADP-ribosylation, other PTMs, and metabolic pathways that regulate DNA repair. Elucidating how these multilayered modifications coordinate signaling kinetics, substrate fate, and the choice of repair pathway represents an important frontier for future mechanistic and proteomic investigations.

Emerging roles and sequencing methods for DNA and RNA ADP-ribosylation

Protein ADP-ribosylation has been studied; however, nucleic acid ADP-ribosylation has only been discovered more recently. This raises fundamental questions about how PAR/MAR marks on DNA are written, read, and erased relative to protein ADP-ribosylation. The bulky nature of DNA-linked PAR/MAR hampers polymerase reading and amplification, complicating DNA sequencing procedures. Such technical challenges and their low abundance have hindered the identification of their genomic locations.¹⁴⁹ Recently, ADPr-Seq was developed to analyze the Mycobacterium tuberculosis DNA ADP-ribosylation.²⁰³ Therefore, adapting ADPr-Seq in mammalian system will be crucial to identify genomic hotspots and understand how DNA-linked PAR/MAR interfaces with repair pathway choice.

For RNA, terminal ADP-ribosylation and internal base or 2′-O-linked modifications have been demonstrated in vitro and in specific bacterial and mammalian contexts, but it remains unknown which transcripts are preferentially modified in vivo or at what stoichiometry. There is a need for robust, nucleotide-resolution mapping strategies to define endogenous RNA ADP-ribosylation sites, their dynamics during stress or infection, and their impact on RNA stability and translation.

Regulation of ADP ribosylation in mitochondrial DNA repair and function

Mitochondrial ADP-ribosylation is an emerging field, initially described in the context of PARP1-mediated mitochondrial DNA (mtDNA) repair and later expanded to include other mitochondrial ARTs and hydrolases.^128-130 A major unanswered question is the full spectrum of mitochondrial ADP-ribosylation substrates: whether mitochondrial ADP-ribosylation occurs predominantly on proteins, nucleic acids, or both. Moreover, the identity and specificity of mitochondrial reader proteins remain largely unknown, and only a subset of erasers (e.g., certain PARG and MACROD1 isoforms) have been characterized in this compartment. Clarifying these questions will be essential to link mitochondrial ADP-ribosylation to processes such as oxidative phosphorylation, mtDNA maintenance, and mtDNA-mediated signaling.

Emerging evidence showed that sirtuins have auto-ADP-ribosylation activity and can also regulate mitochondrial functions.^60-62,204 This raises the important question of whether mitochondrial sirtuins, including SIRT3, SIRT4, and SIRT5,²⁰⁵ participate in generating or regulating ADP-ribosylation and, consequently, influence mitochondrial metabolism, mtDNA integrity, or aging. Addressing these questions will require integrating organelle-resolved proteomics, metabolite profiling, and functional genetics to dissect the mechanisms, targets, and consequences of mitochondrial ADP-ribosylation.

Biographies

Linlin Zhao is an Associate Professor in the Department of Chemistry and Environmental Toxicology Graduate Program at the University of California, Riverside. His research program investigates the chemical and molecular mechanisms of mitochondrial DNA damage, repair, and turnover, and exploits chemical biology approaches to modulate mitochondrial DNA-mediated cell signaling. Dr. Zhao has been actively involved with the Division of Chemical Toxicology and has served in multiple roles, formerly as the Division's Secretary and, currently, as chair of the Communication Committee and a member of the nomination committee. He is also an editorial board member for Chemical Research in Toxicology.

Yi-Cheng Sin is a postdoctoral researcher in the Zhao laboratory at the University of California, Riverside. His research spans epigenetic regulation, metabolism, and post-translational modifications, with a focus on mass spectrometry–based proteomics and metabolomics. He received his Ph.D. from the University of Minnesota, Twin Cities, where his doctoral work focused on developing and applying chemical proteomics strategies to characterize lysine modifications and their regulatory mechanisms. His work has been published in high-impact peer-reviewed journals, including Science Advances, Chemical Science, Analytical Chemistry, and Nature Chemical Biology. He also contributes to the scientific community as a peer reviewer for several journals.