PARPs and ADP-Ribosylation-Mediated Biomolecular Condensates: Determinants, Dynamics, and Disease Implications

Abstract

Biomolecular condensates are cellular compartments that selectively enrich proteins and other macromolecules despite lacking enveloping membranes. These compartments often form through phase separation triggered by multivalent nucleic acids. Emerging data revealed that poly(ADP-ribose) (PAR), a nucleic acid-based protein modification regulated by the PARP family, plays a critical role in this process. This review focuses on the role of PARPs and ADP-ribosylation, exploring the principles and mechanisms by which PAR regulates condensate formation, dissolution, and dynamics. Future studies with advanced tools to examine PAR-binding sites, substrate interactions, PAR length and structure, and transitions from condensates to aggregates will be key to unraveling the complexity of ADP-ribosylation in health and diseases, including cancer, viral infection, and neurodegeneration.

PARPs and ADP-Ribosylation in Biomolecular Condensates

Biomolecular condensates (see Glossary) are cellular compartments that selectively enrich specific biomolecules, including proteins and nucleic acids, despite lacking enveloping membranes, many of which form through liquid-liquid phase separation (LLPS) [1–3]. These condensates play essential roles in gene regulation, cell signaling, and stress responses [1–3]. Dysregulation of their formation is linked to various diseases, including viral infections, cancer, and neurodegeneration, often due to altered composition and physical properties [3–5]. Understanding the principles of biomolecular condensates could offer new therapeutic strategies.

Although proteins are the major components of biomolecular condensates, nucleic acids like DNA and RNA often trigger condensation due to their multivalency. Poly(ADP-ribose) (PAR), dubbed the “third nucleic acid” for its RNA-like structure (Figure 1A-B), forms linear or branched chains and has emerged as crucial in forming and regulating biomolecular condensates like DNA repair foci, stress granules, mitotic spindles, and nucleoli [6]. PAR is conjugated to proteins through ADP-ribosylation, regulated by 17 human ADP-ribosyltransferases, commonly known as PARPs [7–9]. Four of these enzymes (PARPs 1, 2, 5a, 5b) synthesize PAR chains, while 11 others add single ADP-ribose units [mono(ADP-ribose) or MAR], with two of these (PARPs 9 and 13) being catalytically inactive. Notably, recent data indicate that PAR chains synthesized by these PAR-generating ADP-ribosyltransferases can also exist as protein-free polymers within cells [10]. Unlike other nucleic acids, this enzymatic regulation allows precise control over when and where these condensates form. Furthermore, some condensates undergo changes in formation or composition when cells are treated with PARP inhibitors [6]. Four such inhibitors are FDA-approved for cancer treatment and are being explored for other conditions, including neurological, cardiovascular, and immunological diseases, indicating their translational potential [11]. Yet, how PAR-mediated protein condensation is initiated, sustained, and regulated remains understudied.

Figure 1. PAR Structure, PARP1 Multimerization, and the Functional Implications of PAR in Regulating Protein Condensates.

(A, B) Structures of PAR and polyadenylated RNA. (C) Schematic diagram of PARP1 dimerization and multimerization on dsDNA. The equal signs represent interactions between specific domains, with red indicating interactions required for dimerization and blue indicating those required for multimerization (Adapted from [31]). (D) Functional implications of PAR in regulating protein condensates in the nucleus and cytoplasm.

This review explores the role of PARPs and ADP-ribosylation in modulating protein condensation. We begin by exploring the shared properties of proteins that form condensates with PAR. Next, we discuss how PARPs and ADP-ribosylation drive condensate formation and dissolution, modulate condensation dynamics, and regulate aggregate formation. Finally, we speculate on potential mechanisms by which PAR regulates protein condensation and their physical states, as well as the advancements needed in tools for mechanistic understanding.

Common Properties of Proteins Condensed by PAR

PAR-Induced Protein Condensation

PAR has been shown to induce a dozen proteins to form condensates (Table 1). One of the pioneering discoveries involves FUS (Fused in Sarcoma), a member of the FET family of RNA-binding proteins, which also includes EWSR1 and TAF15, all implicated in cancer and neurodegeneration [12–15]. PAR is a potent inducer of FUS condensation—10 pM of PAR with a length longer than 8-mer can induce condensation of 1 μM FUS, a concentration close to its physiological levels [15]. Furthermore, FUS, once interacted with PAR, can form condensates even after PAR is degraded by poly(ADP-ribose) glycohydrolase (PARG) [15], suggesting that PAR induces a lasting effect on FUS, likely by promoting a state conducive to condensation. Single-molecule studies indicate that PAR transiently interacts with FUS [15], potentially allowing one PAR molecule to interact with multiple FUS molecules.

Table 1.

Experimental Evidence of PAR-Modulated Protein Condensation and Aggregation under Various Conditions in vitro

Protein	Protein Concentration	PAR Concentration	Buffer Condition	Incubation Duration	Phenotype	Method	Without PAR	References
R-DPR	100 μM	10–50 μM	61.5 mM K2H PO4, 38.5 mM KH2 PO4	20 min	Condensate	Fluorescence microscopy	No condensates at 100 μM	[16]
G3BP1	50–100 μM	2.5–50 μM	PBS	20 min	Condensate	Fluorescence microscopy	No condensates up to 200 μM	[16]
TDP-43	1 μM	3–5 μM	100 mM NaCl, 100 mg/mL dextran	10 min	Condensate	Transmitted Light microscopy	Forms condensates at higher concentrations	[18]
hnRNPA1	20 μM	1–7 μM	25 mM NaCl, 10% PEG	3 min	Condensate	Transmitted Light and fluorescence microscopy	Forms condensates at higher concentrations	[17]
hnRNPA1 + TDP-43(1–274)	12.5 μM hnRNP A1 + 12.5 μM TDP-43(1–274)	3–7 μM	50 mM NaCl	3 min	Condensate	Transmitted Light and fluorescence microscopy	Forms condensates at higher concentrations	[17]
PARIS	5 μM-50 μM	10 nM	150 mM NaCl	2 h	Condensate/aggregate depending on the concentration	Transmitted Light microscopy, ThT fluorescence	Forms condensates/aggregates at higher concentrations	[19]
FUS	400 nM	1 μM	500 mM KCl, 10% dextran	10 min	Condensate	Fluorescence microscopy	Forms condensates at higher concentrations	[12]
FUS	1 μM	10 nM-1 μM	100 mM NaCl	4 h	Condensate, aggregate at higher 1 μM PAR32	Fluorescence microscopy	Forms condensates at higher concentrations	[15]
DDX 21 C-terminal IDR	6 μM	3 μM	150 mM NaCl, 2% PEG-8000, pH 7.0	For microscope it is immediate (undefined); for turbidity it is 15–30 min	Condensate	Fluorescence microscopy, turbidity assay	Forms condensates at 10% PEG-8000	[73]
α-synuclein	0.1 mg/ml	5 nM	PBS	4–72 h	Aggregate	Transmission electron microscopy, ThT fluorescence	Forms aggregates at a slower rate	[43]
PARP1 + DNA	2 μM	–	150 mM KCl	30 min	Condensate	Transmitted Light and fluorescence microscopy, sedimentation assay	–	[31]
PARP1 + DNA	0.5 μM − 8 μM	0–0.5 mM NAD+	50 mM NaCl	Unknown	Condensate	Fluorescence microscopy	–	[32]
PARP7	1 μM	100 μM NAD+	150 mM NaCl, 10% PEG	15 min	Condensate	Transmitted Light microscopy	No condensates	[33]
PARP1 + DNA + FUS	3 nM PARP1 + 40 nM FUS	0.3 mM	12.5 mM NaCl, 100 mM urea	5 min	Compartment (undefined)	Atomic force microscopy	No compartments at the same concentration	[14]
FUS, EWSR1, TAF 15	molar ratio FUS:PAR=53:1; EWS:PAR=41:1; TAF15:PAR=40:1		150  mM KCl	24 h	Aggregate	Transmission electron microscopy	Forms smaller aggregates	[13]
(GR) 20 + G3B P1	100 μM (GR)20 + 2.5 μM G3BP1	2.5–25 μM	PBS	20 min	Condensate	Fluorescence microscopy	Forms co-condensates at higher concentrations	[16]
(GR) 20 + TDP 43	5 μM (GR)20 + 0.5 μM TDP43	2.5–25 μM	PBS	20 min	Aggregate	Fluorescence microscopy, precipitation assay, turbidity assay	Forms co-condensates at higher concentrations	[16]

In addition to FUS, PAR induces the condensation of R-DPR (arginine-containing dipeptide repeat from the C9orf72 gene) [16], G3BP1 (Ras-GTPase-activating protein SH3-domain-binding protein 1) [16], hnRNPA1 (heterogeneous nuclear ribonucleoprotein A1) [17], TDP-43 (TAR DNA-binding protein 43) [18], and PARIS (Parkin-Interacting Substrate) [19] in vitro. This induction occurs under varying conditions of protein concentrations, salt, crowding agents, incubation time, and other factors (Table 1). These proteins, primarily involved in regulating gene expression and stress responses, are often detected in cytoplasmic structures associated with neurodegenerative diseases, where elevated PAR levels are frequently observed [16,20]. Therefore, PAR-induced condensation may play a crucial regulatory role in cellular and pathological processes, with potential therapeutic implications if its mechanisms are understood.

PAR-Induced Condensation Involves Proteins Rich in Repetitive Arginine and Glycine Residues

Most proteins that form condensates with PAR possess low-complexity domains (LCDs) or prion-like domains (PrLDs) [1–3]. Specifically, proteins such as FUS, EWSR1, TAF15, and hnRNPA1 contain conserved regions with either arginine-glycine-glycine (RGG) repeats or glycine-arginine-rich (GAR) domains that often bind nucleic acids [21].

RGG repeats and GAR domains are crucial for the interactions of these proteins with PAR and their enrichment in biomolecular condensates. For instance, the RGG repeats are essential for recruiting FUS to DNA repair foci in cells [13,22]. Truncating the C-terminal RGG repeats of FUS impairs its ability to form condensates with DNA and PARP1 in vitro [14]. A study on the molecular grammar of FUS condensation reveals that arginine residues in FUS are critical for determining the threshold concentration required for condensation (i.e., saturation concentration, C_sat), while glycines maintain condensate liquidity [23]. When FUS condenses with RNA, arginine mutations linked to neurodegenerative diseases such as ALS (Amyotrophic Lateral Sclerosis) and FTD (Frontotemporal Dementia) impair RNA binding, while glycine mutations reduce condensate liquidity [24]. When condensed with PAR, mutating all 31 arginines in FUS abolishes its condensation with PAR and significantly reduces its PAR binding [15]. These findings suggest that RGG repeats are critical in FUS condensation and its interactions with nucleic acids, including PAR.

The RGG repeats are essential in cells for forming biomolecular condensates such as stress granules. Their formation and composition are regulated by noncovalent PAR binding and PARylation, in which PAR chains are covalently attached to proteins by PARPs [17,18,25,26]. The RGG repeats are required for PAR binding of G3BP1 [27], a core scaffolding protein of stress granules, and are crucial for localizing TAF15 and FUS to these condensates [28,29].

The interaction between PAR and repetitive arginine and glycine residues is also observed in pathological settings. The C9ORF72 mutation, the most common genetic cause of ALS and FTD, encodes the highly toxic dipeptide repeats poly(GR), which binds to PAR and induces poly(GR) condensation in vitro [16]. Poly(GR) promotes stress granule formation in cultured cells and TDP-43 aggregation in cell lysates, both suppressed by the loss of PARP1 activity [16].

Similarly, inhibiting PARP1 reduces poly(GR) aggregation in fly models of C9ORF72-related ALS-FTD [16]. These examples highlight the importance of systematically studying the interaction between RGG/GR-containing proteins and PAR, as well as the role this interaction plays in regulating condensation.

Condensation of PARPs in Both Enzymatic Activity-Dependent and -Independent Manners

Some PARPs form condensates crucial for DNA damage responses and transcriptional regulation [30]. Two recent reports indicate that PARP1 is pivotal in initiating DNA damage responses by condensing with DNA and mediating the synapsis of broken DNA ends to facilitate repair [31,32]. Interestingly, PARP1 lacks any LCDs; instead, its condensation is driven by dimerization and multimerization on DNA (Figure 1C). Dimerization occurs at the structured ZnF1-WGR interface of two dsDNA-bound PARP1 molecules, bridging the DNA strands [31]. Multimerization involves the ZnF1-ZnF3 and ZnF3-WGR interfaces, recruiting additional PARP1 molecules to assemble PARP1-DNA co-condensates [31]. Notably, this assembly does not require enzymatic activity or PAR binding [31]. However, the addition of NAD⁺, the crucial co-factor for ADP-ribosyltransferase activity, can modulate condensate properties (see below) [31,32].

PARP7, also known as TiPARP or ARTD14, is a MAR-adding enzyme that forms nuclear condensates in an ADP-ribose-dependent manner. Unlike PARP1, PARP7 condensation relies on both ADP-ribose binding and catalytic activity [33]. Mutations in the catalytic site or ADP-ribose-binding WWE domain abolish condensate formation in cells [33]. Purified PARP7 from cells forms condensates in vitro in an NAD⁺-dependent manner, while the catalytically dead mutant does not, even with NAD⁺ [33]. These nuclear bodies recruit transcription factors, including HIF-1α, c-Myc, and the estrogen receptor, along with the E3 ubiquitin ligase HUWE1. HUWE1 promotes the ubiquitination and degradation of HIF-1α and other oncogenic transcription factors, potentially serving as a negative feedback regulator for transcriptional activities [33] (Figure 1D, upper panel).

Many PARPs are enriched in biomolecular condensates, formed either constitutively or in response to stress (Table 2). For example, nucleoli contain PARPs 1 and 2 [34], β-catenin degradasomes and mitotic spindles contain PARPs 5a and 5b [35]. A newly discovered condensate of PARP5a and the E3 ubiquitin ligase RNF146 mediates PARylation-dependent ubiquitination, functioning as a necroptosis checkpoint [36] (Figure 1D, lower panel). Besides these PAR-adding PARPs, several MAR-adding PARPs, including PARPs 7, 12, 14, 15, and 16, also contribute to condensate formation (Table 2). Exploring whether other PARPs can form condensates, either in an enzymatic activity-dependent or -independent manner, could provide valuable insights into their mechanisms, regulation, and biological functions within condensates.

Table 2.

Presence of PARPs in Biomolecular Condensates in Eukaryotic Cells

Biomolecular Condensates	PAR-Adding PARPs	MAR-Adding PARPs	Inactive PARPs	Cellular Localization	References
DNA repair foci	1, 2	3		Nucleus at sites of DNA damage	[12,13,87]
Stress granules	5a	12, 14, 15	13	Cytoplasm	[17,25,26,83,84]
Mitotic spindles	1, 2, 5a, 5b	3, 4		Cytoplasm	[88]
Nucleoli	1, 2			Nucleus	[34,89–92]
β-catenin degradasomes	5a, 5b			Cytoplasm	[35]
Wnt signalosomes	5a, 5b			Cytoplasm, near the plasma membrane	[35]
IFNγ-induced ADP-ribose-enriched bodies		14	9	Cytoplasm	[79–81]
TiPARP nuclear bodies		7		Nucleus	[33]
Sec bodies		16		Cytoplasm, particularly near the endoplasmic reticulum	[82]
PARP5A/RNF14 6 condensates	5a			Cytoplasm	[36]

Diverse Roles of PAR in Altering Protein Condensation and Physical States

PAR Promotes Co-Condensation and Regulates the Condensate Composition

PAR not only induces the condensation of individual proteins but also promotes the co-condensation of multiple proteins, thereby regulating the compositional specificity of condensates (Figure 2, Key Figure). For example, PAR facilitates the co-condensation of R-DPR and G3BP1, which typically co-condense only at higher concentrations [16]. Additionally, PAR enhances the dose-dependent co-condensation of hnRNPA1 and TDP-43(1–274), which includes a well-folded N-terminal domain and two highly conserved RNA recognition motifs, lowering the concentration threshold required in the absence of PAR [17].

Figure 2. Diverse Roles of PAR in Altering Protein Condensation and Physical States.

Schematic diagram illustrating how PAR promotes protein co-condensation, regulates condensate dynamics, induces condensate dissolution, and modulates protein aggregation.

DNA repair proteins, such as FET family members, PARP2, and HPF1, can co-condense with PARP1-DNA condensates in vitro [31] (Figure 1D, upper panel). Unlike PARP1, PARP2 is preferentially activated by PAR, leading to branched PAR chain synthesis, suggesting its potential role in creating a second wave of PAR formation with defined structures [37]. FET family proteins increase condensate dynamics and stabilize them against dissolution when PARP1 is activated in the presence of NAD⁺ [31]. Recent studies further show that co-condensation with single-strand break repair proteins such as XRCC1, LigIII, Polβ, and FUS form multiphase structures within PARP1-DNA condensates, where XRCC1 and LigIII segregate with DNA [32] (Figure 2, upper panel). This subcompartmentalization, reminiscent of the multiphase nucleolus [38], may enhance specificity and efficiency in DNA ligation reactions.

PARylation, the primary form of PAR in cells, covalently attaches to proteins and noncovalently recruits specific proteins into condensates, thereby regulating their composition. For instance, only auto-PARylated PARP1, not its unmodified or inactive forms, can bind to the transcription elongation factor Cyclin T1 (CycT1) and promote co-condensation with its intrinsically disordered region in vitro [39]. In cells, PARP1 inhibitor Olaparib prevents the CycT1 recruitment to DNA damage sites [39]. In multiphase PARP1-DNA condensates formed with single-strand break repair proteins, NAD⁺-dependent PARylation enhances XRCC1, LigIII, and FUS enrichment, but not Polβ, likely due to differences in PAR binding capabilities [32]. This differential enrichment allows for precise regulation of condensate composition and organization.

PAR Regulates Condensate Dynamics

Condensates often exhibit liquid-like properties, allowing dynamic exchange of proteins and other biomolecules both internally and with their surroundings, with PAR also influencing these dynamics (Figure 2, lower panel). For example, PAR interaction with apoptosis signal-regulating kinase 3 (ASK3) is crucial for increasing the liquidity of ASK3 condensates and inactivating its enzymatic activity under hyperosmotic stress (Figure 1D, lower panel) [40]. Overexpressing PAR-degrading PARG, but not its inactive mutant, reduces ASK3 recovery dynamics in cells, as shown by fluorescence recovery after photobleaching (FRAP), indicating that PAR maintains the liquid-like state of ASK3 condensates [40]. This effect is abolished when the PAR-binding motif (PBM) in ASK3 is mutated, confirming the reliance on PAR interaction [40].

Similarly, adding NAD⁺ enhances the fluorescence recovery of PARP1 in its condensates with DNA, likely due to newly synthesized PAR [31,32]. For FUS, adding PAR also increases fluorescence recovery in its condensates compared to adding polyU RNA, suggesting that PAR enhances condensate dynamics [15]. However, PAR can also negatively regulate condensate dynamics. For instance, adding PAR to hnRNPA1 decreases fluorescence recovery within its condensates [17]. The exact mechanism by which PAR regulates condensate dynamics remains unclear but may involve how different proteins form multivalent interactions within condensates and how PAR binding and PARylation modify these interactions.

Notably, increasing cellular PARylation levels by knocking down PARG delays the disassembly of stress granules containing hnRNPA1 and FUS [25]. This finding suggests that while PAR may alter the dynamics of individual components in vitro, it can negatively impact overall condensate dynamics in cells. Since stress granules are often considered a crucible for the aggregation of neurological disease proteins [41], reduced dynamics from elevated PAR levels might contribute to converting these condensates into pathogenic aggregates in neurodegeneration. While some studies highlight the functional relevance of PAR-regulated condensate dynamics in cellular stress responses [40], further research is needed to elucidate how PAR affects individual components, overall condensate behavior, and broader biological implications.

PAR Can Dissolve Condensates

Recent data indicate that, depending on the context, PAR can either promote condensate formation or dissolve them (Figure 2, left panel), leading to significant biological consequences (Figure 1D). For example, PAR not only increases the liquidity of ASK3 condensates but also reduces their quantity, which may partially explain ASK3 inactivation under hyperosmotic stress in cells [40]. During DNA damage, PARP1-mediated PARylation of the histidine-rich domain of CycT1, a subunit of the transcriptional elongation complex P-TEFb (composed of the CDK9-CycT1 heterodimer), disrupts its condensation both in vitro and in cells [39] (Figure 1D, upper panel). This prevents CDK9-mediated hyperphosphorylation of the RNA polymerase II C-terminal domain (CTD), leading to a global shutdown of transcription [39].

Similarly, adding NAD⁺ to activate wild-type PARP1 induces dose-dependent disassembly of PARP1-DNA condensates—an effect absent with chemical inhibitors or a catalytically inactive PARP1 mutant [31]. Conversely, adding PARG to remove PAR restores the disrupted condensates [31]. Furthermore, PARP1 PARylation also dissociates PARP1 from DNA ends [31]. These findings suggest that PARylation of PARP1 dissolves PARP1-DNA condensates, potentially facilitating its release from DNA damage sites.

The exact mechanism by which PAR dissolves condensates remains unclear. One possibility is that the highly negative charge of PAR disrupts the electrostatic multivalent interactions that maintain condensate formation. In this case, the excessive negative charges from PAR might induce a reentrant effect, where initial condensate formation is followed by dissolution at higher concentration thresholds, similar to other nucleic acids [42]. Additionally, PARylation can covalently modify proteins, altering their conformations or interactions with themselves or other proteins, which are critical for condensation.

PAR Modulates Protein Aggregation

Liquid-like biomolecular condensates can transition into solid, structurally defined aggregates, particularly in the presence of specific disease mutations [3–5]. These aggregates, often enriched with cross-β structures, are less dynamic and exhibit more irreversible material properties. This transition is regulated by PAR in proteins associated with neurodegenerative diseases (Figure 2, right panel).

PAR chains can induce or accelerate aggregate formation, as seen in Parkinson’s disease. Purified PAR accelerates the fibrillization of monomeric α-synuclein at 37°C, with cross-β structures detected via transmission electron microscopy (TEM) and thioflavin T (ThT) fluorescence [43]. In vivo, PAR enhances the neurotoxicity of preformed α-synuclein fibrils [43]. Beyond α-synuclein, PAR also induces aggregation of other neurodegeneration-related proteins, such as those in the FET family and PARIS (Parkin-Interacting Substrate) [13,19]. Sub-stoichiometric amounts of polydisperse PAR increase the size of spontaneous aggregates formed by FUS, EWS, and TAF15 in vitro [13]. Even under conditions where PARIS forms liquid-like condensates on its own, 10 nM PAR promotes its solid-like aggregation [19] (Table 1). Notably, in aged mice overexpressing wild-type PARIS, but not in those expressing a PAR-binding-deficient PARIS mutant, SDS-resistant PARIS species are detected in the substantia nigra—a brain region rich in dopaminergic neurons that regulate movement and are implicated in Parkinson’s disease [19].

Similar to its role in condensate formation and dissolution, PAR can both promote and inhibit aggregate formation depending on concentration, mutations, and cellular context. For example, PAR induces wild-type TDP-43 to form liquid-like condensates (Figure 1D, lower panel) but promotes the aggregation of the ALS-linked mutant Q331K in the presence of dextran [18]. However, without dextran, PAR inhibits wild-type TDP-43 aggregation [44]. Given that PAR levels are often elevated in neurodegenerative conditions, understanding PAR-regulated protein aggregation could be key to advancing therapeutic interventions.

Potential Mechanisms of PAR Regulation on Protein Condensation and Physical States

The mechanism by which PAR regulates protein condensation and aggregation remains unclear. Here, we speculate on its potential dual role—both as a protein modification and a nucleic acid polymer—during the nucleation stage of condensate formation, its effects on protein conformation and energy state, and how different chemical elements of PAR may integrate into the sticker-spacer model often used to describe condensation behavior.

Can PAR Affect the Nucleation Stage?

Condensate and aggregate formation can begin with nanoscale clusters that ultimately influence pathological aggregation [45,46]. FET family proteins and α-synuclein, for instance, form nanoscale clusters at sub-saturated, physiologically relevant concentrations, providing insights into nucleation-stage species [47,48]. Atomic Force Microscopy (AFM) shows that low concentrations of FUS form circular nanoclusters within the first few minutes [48]. After 1 hour, a fraction of these assemblies transforms into fibrillar structures with low circularity (nanofibrils), measuring up to 200 nm in length and 3 nm in height [48]. This transition from nanoclusters to nanofibrils depends on incubation time and protein concentration. Intriguingly, a truncated FUS variant lacking the N-terminal SYQG-rich domain and C-terminal RGG repeats fails to form nanofibrils, underscoring the importance of these regions [48].

Moreover, adding mRNA promotes the formation of more circular, granular structures rather than nanofibrils, which reappear after RNase treatment [48]. Phosphomimetic FUS, generated by mutating 12 specific serine and threonine residues to glutamic acids, similarly fails to trigger nanofibril formation [48]. These findings suggest that nucleic acid binding or post-translational modifications can interfere with nanofibril formation and modulate FUS assembly, potentially reducing its capacity to form structured fibers in favor of granular condensates.

PAR, as a nucleic acid-based protein modification, may play dual roles. Although PAR shares the same building blocks as RNA, it arranges them differently, resulting in an additional phosphate per unit and greater structural rigidity [49] (Figure 1A-B). It is often covalently attached in cells to specific amino acid residues (e.g., E, D, K, S, C) on target proteins as a post-translational modification. These distinct structural features may uniquely enhance PAR’s potency in regulating early submicron-scale cluster formation during the nucleation stage of condensates and aggregates.

In the case of α-synuclein, nanoclusters initiate and facilitate the aging-related development of amyloid fibrils [50]. These nanoclusters rapidly form but require days to transition into larger, macroscopic condensates, indicating a kinetic barrier [50]. Given that PAR can accelerate α-synuclein fibrillization and induce condensation and aggregation in other proteins, it would be intriguing to explore whether PAR can help these proteins overcome the kinetic barrier to form large assemblies from nanoclusters (Figure 3A).

Figure 3. Potential Mechanisms of PAR Regulation on Protein Condensation and Physical States.

(A) Energy landscape when protein physical states are altered from monomers to aggregates, with potential kinetic barriers indicated. (B) Schematic diagram indicating how bending the rigid PAR might alter protein conformation. (C) A sticker-spacer model of PAR-protein interactions. (D) A specific example illustrating PAR interacting with an RGG-containing peptide, with the interactions between the guanidinium head groups of arginine and the diphosphate groups of PAR highlighted.

Can PAR Alter Protein Conformation and Energy State?

Transitioning from a protein’s native state to an aggregated form requires overcoming a significant kinetic energy barrier [51]. Factors such as mutations, interactions with biomolecules, or post-translational modifications can destabilize proteins and promote partially folded states that are prone to aggregation [52–54]. Studies suggest that interactions with nucleic acids can alter the hydration and solvent accessibility of a protein, leading to changes in its folding state. For instance, the sequence-specific binding of p53 to DNA stabilizes the protein and prevents misfolding [55]. In contrast, DNA binding can restructure prion proteins, reducing their hydration and potentially converting them into pathological forms [56]. Moreover, research on G3BP1–RNA condensation reveals that RNA binds to G3BP1, shifting the G3BP1 homodimer from a “closed” conformation into an expanded, “open” conformation, which initiates condensation [57,58]. These precedents raise the possibility that PAR, like nucleic acids, may destabilize the native fold of proteins or induce partially unfolded conformations, which are often metastable and lower the kinetic barrier to aggregation (Figure 3A).

Moreover, PAR is more rigid than RNA at physiological salt concentrations, and while FUS can bend PAR, RNF146 WWE, another PAR binder with similar affinity, cannot [49]. These findings suggest that protein interactions with PAR may induce changes in conformation and energy state, with bending the rigid PAR likely requiring energy and potentially triggering a shift in FUS that promotes condensation (Figure 3B).

In addition to non-covalent interactions, post-translational modifications can significantly alter protein stability and interaction networks [54,59–62]. For example, ubiquitination affects the conformational dynamics of the model protein barstar, inducing the formation of partially unfolded, high-energy conformations primed for proteasomal processing [63]. Similarly, PARylation can induce conformational changes, as seen with the allosteric activation of RNF146, an E3 ubiquitin ligase responsible for PARylation-dependent ubiquitination [64]. Understanding how PAR regulates protein condensation and physical states requires exploring how site-specific PARylation affects protein structural stability, conformational dynamics, and interaction networks.

Should PAR Be Considered as Part of the ‘Sticker-Spacer’ Model of Protein Condensation?

Biomolecular condensates form through specific molecular interactions that organize proteins and other biomolecules into dynamic, phase-separated structures. The “sticker-spacer” model helps explain these interactions: “stickers” are molecular modules that drive condensation, while “spacers” link stickers, providing flexibility [65].

This model can also apply to nucleic acids. For RNA, structured regions, short motifs, or individual nucleotides may act as stickers, with regions between them serving as spacers [66]. Similarly, with its uniquely spaced diphosphate groups, PAR may act as stickers by forming electrostatic interactions with positively charged residues in proteins, such as arginines in RGG domains (Figure 3C-D) [23]. The adenosine and ribose groups between diphosphates could function as spacers, influencing the material properties of condensates. Additionally, the recently discovered cation-induced globular structures formed by PAR may serve as another type of sticker, contributing specificity and unique interactive properties to PAR-protein condensates [67].

The sticker-spacer model offers a promising framework for understanding PAR’s role in protein condensation. Further investigations, such as computational modeling and synthesizing PAR variants with defined lengths and branching for structural studies, may help test this hypothesis and expand our understanding of PAR’s contribution to biomolecular condensates.

Future Directions for Investigating PARPs and ADP-Ribosylation in Condensation

Understanding Molecular Determinants in PAR-Driven Protein Condensation and Aggregation

In this review, we first examined the common properties of proteins that interact with PAR, particularly those with RGG repeats. The planar guanidinium group in arginine forms two hydrogen bonds, which can enable favorable bidentate interactions with PAR’s diphosphates, promoting the multivalent interactions essential for condensate formation (Figure 3D). Recent chemoproteomic analyses identified 12 of the 16 triple RGG motif-containing proteins in the human proteome as PAR binders [68]. Among the nearly 2,000 PAR-binding proteins identified in proteomic studies, many are also known DNA- or RNA-binding proteins [9,68–71]. However, PAR-binding proteins tend to have a lower median isoelectric point than the typically basic proteins that bind canonical nucleic acids [68], suggesting that non-basic residues may play a role in PAR binding, a prerequisite for condensation. Additionally, these PAR-binding proteins are enriched with low-complexity sequences and condensate components [68]. Systematic investigation is thus needed to uncover additional common motifs or molecular grammar required for condensation.

Notably, PAR drives proteins like FUS and PARIS to form liquid condensates at lower protein concentrations but solid aggregates at higher protein concentrations [15,19]. Whether this applies to all proteins, and the underlying principles, remains unclear. Recent data suggest PAR may act as a nucleation seed, a scaffold for protein condensates, or a catalyst for transitioning proteins into a condensation-conducive state [6,72] (Figure 3A-B). Tools are needed to better define PAR-protein interactions at both bulk and single-molecule levels, explore how PAR binding induces conformational changes, and examine the PAR-mediated transition from condensates to aggregates (Table 3).

Table 3.

Techniques to Study PAR-Mediated Protein Condensation

Techniques	Description	Advantage	Drawback	References
PAR Synthesis, Fractionation, and Labeling
PAR Synthesis	PAR is synthesized in vitro by PARPs (e.g., PARP1 or PARP5a) using NAD+ as a substrate to add ADP-ribose units to a protein, forming long PAR chains.	Allows controlled synthesis of PAR for experiments.	Requires optimization to produce high-quality PAR.	[68,93]
Fractionation	High-performance liquid chromatography (HPLC) separates PAR molecules based on size and charge to isolate specific chain lengths.	Provides high-resolution separation critical for studying length-dependent effects.	Time-consuming and requires expertise in HPLC.	[68,93,94]
Labeling	Labels are attached to PAR at specific termini using chemical reactions, EDC or oxime-based coupling or ELTA labeling with fluorescent or biotin tags.	Enables tracking and visualization of PAR in various assays.	Labeling may alter PAR behavior, requiring careful experimental design.	[95–99]
Traditional Biochemical Assays to Study PAR-Protein Interactions
Electrophoretic Mobility Shift Assay (EMSA)	Analyzes the mobility of protein-PAR complexes through a gel matrix, where a shift in mobility indicates binding.	Easy to perform and useful for analyzing binding interactions.	Requires labeled PAR and may not provide detailed binding kinetics.	[100–102]
Fluorescence Polarization (FP) Assay	Detects binding by measuring changes in light polarization of fluorescently labeled PAR when bound to a protein.	High sensitivity and real-time binding information.	Requires fluorescent labeling of PAR.	[103]
Isothermal Titration Calorimetry (ITC)	Measures heat changes during PAR-protein binding to provide information on binding affinity and thermodynamics.	Label-free and provides comprehensive thermodynamic profiles.	Requires large amounts of protein and PAR.	[104]
Reconstitution of Protein-PAR Condensation
In Vitro Reconstitution	Condensation is reconstituted by mixing purified proteins and PAR under controlled conditions, observed via turbidity or microscopy.	-	-	[12,13]
Turbidity Assay	Measures cloudiness in a solution, indicating condensate formation, with turbidity correlating to the extent of condensation.	Simple, rapid, and suitable for high-throughput screening.	Limited in providing detailed information about condensate morphology.	[105,106]
Fluorescence Microscopy	Visualizes fluorescently labeled proteins and PAR within condensates, tracking formation and dynamics.	High spatial resolution and dynamic process tracking.	Requires labeling, which can introduce artifacts.	[12,13]
Transmitted Light Microscopy (DIC)	Offers high-contrast images of condensates without labeling, revealing structural features and boundaries.	Excellent contrast and resolution of structural details.	Limited in revealing molecular dynamics.	[18]
Transmission Electron Microscopy (TEM)	Provides high-resolution imaging to uncover fine structural details at the nanoscale.	Resolves fine morphological features and internal organization.	Requires sample preparation that may alter native states.	[13,107]
Atomic Force Microscopy (AFM)	Uses a cantilever with a sharp tip to scan the sample surface, revealing nanoscale structures like clusters and fibrils.	High-resolution imaging of initial nucleation and small clusters.	Can be time-consuming and may not capture dynamic processes.	[108]
Techniques for Studying Condensate Properties
Fluorescence Recovery After Photobleaching (FRAP)	Measures molecule mobility within condensates by monitoring fluorescence recovery after bleaching a region.	Provides insights into the liquid-like or solid-like properties of condensates.	Requires advanced microscopy and careful interpretation.	[12,13]
Optical Tweezers and Rheology	Measures mechanical properties like viscosity and elasticity of condensates using laser-based trapping and deformation techniques.	Direct measurement of physical properties critical for understanding biological function.	Requires specialized equipment and expertise.	[31,109,110]
C-Trap (Optical Tweezers with Fluorescence Microscopy)	Combines optical tweezers with fluorescence microscopy and microfluidics to manipulate and visualize single molecules, enabling the study of mechanical properties and molecular interactions in real time.	Allows simultaneous measurement of forces and observation of molecular behavior within condensates.	Requires specific fluorescent labeling and complex setup for real-time observations.	[111,112]
Microfluidics	Involves precisely controlling the environment around biomolecular condensates by introducing different reagents or altering conditions (e.g., temperature, buffer).	Provides fine control over experimental conditions, useful for examining phase separation behavior.	Requires specialized microfluidic devices and expertise in handling delicate experimental conditions.	[113,114]
Dynamic Light Scattering (DLS)	Uses scattered light to measure the size distribution of condensates in solution.	Provides information on particle size and stability.	Limited to particles within a specific size range.	[15,115,116]
Mass Photometry	Measures the mass of individual particles by analyzing their light scattering, providing data on size and distribution.	High sensitivity and label-free measurement.	Limited to providing mass information without detailed structural insights.	[50]
Single-Molecule Techniques
Single-Molecule Fluorescence Resonance Energy Transfer (smFRET)	Measures energy transfer between two fluorophores on a single molecule to study interactions and conformational changes.	High-resolution insights into molecular dynamics at the single-molecule level.	Requires precise labeling and advanced microscopy techniques.	[15,49]
Fluorescence Correlation Spectroscopy (FCS)	Analyzes fluctuations in fluorescence intensity to study diffusion dynamics and molecular interactions in real-time at the single-molecule level.	Provides quantitative insights into molecular concentrations, diffusion rates, and interaction dynamics.	Requires fluorescent labeling and advanced optical setups.	[117,118]
Techniques Probing Protein Conformational Changes
ANS (8-Anilinonaphthalene-1-sulfonic Acid)	Fluorescence increases upon binding to hydrophobic regions, indicating protein conformational changes.	Simple and effective for monitoring changes in hydrophobic regions.	May not be specific to all conformational changes.	[119]
Thioflavin T (ThT)	Binds to beta-sheet structures common in aggregates, providing information on aggregation.	Real-time monitoring of amyloid fibril formation.	Limited to studying beta-sheet-rich aggregates.	[43,120]
Tryptophan Fluorescence	Utilizes intrinsic tryptophan fluorescence to study protein conformational changes and interactions.	Label-free and sensitive to environmental changes.	Limited by the number of tryptophan residues in a protein.	[121]
Differential Scanning Fluorimetry (DSF)	Monitors changes in fluorescence as proteins unfold with increasing temperature, assessing stability.	Simple and effective for screening stability changes.	Limited in providing detailed mechanistic insights.	[102]
Hydrogen-Deuterium Exchange Mass Spectrometry (HDX-MS)	Measures hydrogen-deuterium exchange rates to study protein dynamics and structural changes.	Detailed information on specific sites and structural changes.	Requires meticulous sample handling and complex interpretation.	[122,123]
Limited Proteolysis Mass Spectrometry (LiP-MS)	Uses protease accessibility changes identified by mass spectrometry to study structural changes.	Specific information on structural changes within the protein.	Requires careful interpretation and complex sample preparation.	[124]

Recent findings suggest that PARP1 activation acidifies the local environment within the nucleolus by producing one mole of proton for each mole of NAD⁺ consumed [73]. Although the mechanisms maintaining the nucleolus’s proton gradient are not fully understood, continuous ribosomal RNA transcription appears necessary for PARP1 recruitment, facilitating acidification [34,73]. Interestingly, PAR-mediated condensation of the intrinsically disordered region of the ribosomal RNA processing factor DDX21 is favored at pH 7 but decreases at higher pH [73], emphasizing the importance of co-monitoring PARP activity and environmental pH for condensate formation.

Untangling the Complexities of PAR-Driven Multiphase Condensation

The mechanisms underlying PAR-induced co-condensation and compositional regulation remain poorly understood. PAR may mediate co-condensation by modulating protein interactions. For example, the PAR-binding motif of hnRNPA1 is required for its interaction with TDP-43, enabling their co-condensation, whereas hnRNPA1 with a mutated motif cannot [17]. PAR binding is also required for hnRNPA1 to be localized to stress granules (Figure 1D, lower panel). Similarly, PAR chains on PARP1 are critical for its interaction and co-condensation with CycT1 [39]. Many proteins that co-condense with PAR, such as R-DPR/G3BP1 and hnRNPA1/TDP-43, are also capable of co-condensing at higher concentrations without PAR [16,17]. PAR, however, facilitates co-condensation at lower concentrations. Therefore, it is critical to develop tools that can map PAR binding sites globally to understand the underlying mechanisms and reveal the breadth of potential co-condensing partners mediated by PAR in cells.

Multiphase condensates in PARP1-DNA complexes with single-strand break repair proteins may involve intricate mechanisms [32]. Drawing parallels from nucleolar multiphase condensation, the immiscibility and multiphasic nature of these condensates may result from unfavorable mutual interactions among the proteins and differences in condensate surface tensions [38]. These phases may also exhibit distinct rheological properties, such as viscosity and viscoelasticity [74], which maintain the structural integrity of their subcompartments and allow for specialized functions. However, our understanding of PAR-mediated co-condensation and compositional control in multiphase condensates remains limited, highlighting the need for further study using advanced biophysical tools.

Unraveling the Influence of ADP-Ribose Length and Structure on Protein Condensation and Aggregation in Normal and Diseased States

PAR is a polymer whose length and branching are dynamically regulated, allowing it to bind specific proteins. Regulation of PAR length and concentration is also linked to pathological states, with slower PAR degradation observed in cancerous tissue and higher levels of PAR frequently associated with neurodegenerative diseases [16,20,43,75,76]. Understanding the impact of these alterations on protein condensation and aggregation, as well as their underlying mechanisms, may open new therapeutic avenues. The length of PAR plays a crucial role in driving the transition of proteins from a monophase state to condensates and aggregates. Unlike PAR, MAR cannot induce the condensation of R-DPR, G3BP1, or TDP-43 [16,18], nor can it promote the co-condensation of R-DPR/G3BP1 or the R-DPR-induced aggregation of TDP-43 [16]. Additionally, PAR length significantly influences the physical state of FUS at its physiological concentration of 1 μM: a 4-mer PAR does not induce FUS condensation, while 8- and 16-mer PARs trigger liquid-like condensation, and a 32-mer leads to solid-like aggregation [15]. This length dependency may arise from differences in valency, with the number of ADP-ribose units, or from PAR adopting different structures depending on chain length.

Although previous studies have suggested that PAR may form secondary structures depending on cations, crystallographic structures beyond dimeric ADP-ribose remain unidentified, and circular dichroism and nuclear magnetic resonance studies on mixed-chain PAR have not revealed any well-defined structures [49,77,78]. Recent studies using small-angle X-ray scattering and molecular dynamics simulations have shown distinct compaction of two different lengths of PAR upon the addition of small amounts of Mg²⁺ ions [67]. Unlike 15-mer PAR, the 22-mer forms ADP-ribose bundles via local intramolecular coil-to-globule transition [67]. Additionally, PAR can form branching structures by connecting through different ribose units in each subunit (Figure 1A). Investigating how PAR length and branching influence condensation and aggregation—and how these largely unexplored structural parameters are regulated by PARPs and their co-factors—could provide valuable insights into the spatiotemporal regulation of protein physical states. Despite advances in synthesizing and labeling PAR polymers [9], tools for studying PAR branching and reconstituting PARylated proteins with defined structures are still greatly needed to assess their impact.

Lastly, emergent data reveal that not only PAR but also MAR can contribute to condensate formation, influencing processes such as transcription, translation, starvation response, and immune regulation (Table 2) [33,79–83]. However, the mechanisms behind MARylation-induced condensation remain unclear—whether it attracts multivalent partners through conformational changes or additional chemical groups. Since the transition from single to multiple ADP-ribose units is a key regulatory node, it is essential to determine whether MARylation functions at distinct stages of biomolecular condensate formation, such as initiation, and how it differs from PARylation in constitutive states, in response to environmental changes, or in diseased conditions. As a case in point, for PAR-enriched stress granules, the systematic knockdown of each of the 17 PARPs has surprisingly revealed that PARP10 and PARP14, both MAR-adding enzymes, play a critical role in their formation [83,84]. Tools capable of monitoring different forms of ADP-ribosylation (MARylation vs. PARylation) and distinguishing ADP-ribose chain lengths are a major gap in current proteomics. Given that different ADP-ribose lengths can recruit distinct proteins [68], analyzing these variations in a single proteomics experiment may reveal critical network interaction specifics within condensates and how they evolve in changing conditions in health and disease.

Concluding Remarks

PARPs and ADP-ribosylation are pivotal drivers of protein condensation, playing essential roles in various cellular processes. PAR’s ability to shift proteins between liquid condensates and solid aggregates highlights its critical function in cellular physiology and pathology. The diverse forms of ADP-ribosylation—including MAR, PAR of varying chain lengths, and its branching—are key to unraveling its regulatory mechanisms (see Outstanding Questions). Beyond proteins, emerging evidence shows that specific DNA bases and certain RNA classes are also targets of ADP-ribosylation [85,86], introducing another layer of complexity. This interplay among the three nucleic acids may further shape condensate assembly, stability, and function, presenting an intriguing direction for future study.

Outstanding Questions.

• What percentage of human proteome form PAR-triggered condensates, and how much PAR is required?

• Which proteins co-condense with PAR, and how does this affect immiscibility and the formation of multiphase structures?

• What causes some proteins to transition from condensates to aggregates based on PAR levels or length, and how might this be relevant in neurodegeneration?

• How does PAR length confer specificity to protein condensation and aggregation? What role does branching or different compact states play?

• How do the differences in configuration of the same building blocks between PAR and RNA affect protein binding, condensation, and aggregation?

• How do MAR and PAR coordinate in regulating protein condensation in cells under normal, stressed, and diseased conditions?

• How does nucleic acid ADP-ribosylation regulate biomolecular condensation?

Highlights

• PARP family members can form condensates in both enzymatic-dependent and -independent manner.

• PAR regulates condensate composition, dynamics, and protein aggregation, and it also promotes the co-condensation of proteins.

• Understanding how PAR influences protein nucleation, conformation, and energy states can shed mechanistic light on its role in regulating protein condensation and aggregation.

• The length and structure of ADP-ribosylation may contribute to the regulatory complexity of protein condensation in health and disease.

Glossary

ADP-ribosyltransferases (PARPs)

Enzymes that transfer ADP-ribose units from NAD⁺ to target molecules, altering their activity, localization, or interactions. In humans, 17 PARPs share a conserved domain, but variations in key residues determine their ability to synthesize MAR, PAR, or remain catalytically inactive. These proteins regulate critical cellular functions, including DNA repair, stress response, and chromatin remodeling

Biomolecular condensates

Cellular compartments that selectively enrich specific biomolecules, including proteins and nucleic acids, despite lacking enveloping membranes. These condensates enable the dynamic organization of biochemical reactions and signaling pathways

Liquid-liquid phase separation (LLPS)

A process where a liquid mixture separates into distinct phases with different compositions. LLPS drives the formation of biomolecular condensates, facilitating compartmentalization and enhancing biochemical reactions

Low-complexity domains (LCDs)

Protein regions characterized by repetitive or simple amino acid compositions. LCDs mediate weak, multivalent interactions that contribute to the assembly of biomolecular condensates

Mono ADP-ribose (MAR)

A single ADP-ribose unit attached to a protein or nucleic acid by specific ADP-ribosyltransferases. MARylation modulates molecular functions and interactions, playing roles in cellular signaling and stress responses

Poly(ADP-ribose) (PAR)

A negatively charged polymer of ADP-ribose units synthesized by PARPs. PAR can exhibit various lengths and branching

Prion-like domains (PrLDs)

Specific types of low-complexity domains that share characteristics with prion proteins, including aggregation propensity. PrLDs are involved in forming biomolecular condensates and pathological aggregates linked to diseases

Stress granules

Structures formed under cellular stress that lack an enclosing membrane and contain stalled translation complexes, mRNAs, and RNA-binding proteins. These granules are thought to regulate mRNA stability and translation during stress conditions, although their exact functions are still under investigation

“Sticker-spacer” model

A theoretical framework explaining biomolecular condensation. Stickers are interaction-mediating motifs, such as charged or aromatic residues, while spacers are flexible regions that modulate the physical properties of condensates, such as viscosity and elasticity. This model describes how sequence composition and valency influence the formation and properties of condensation