ADP-Ribosylation of Cytidine: A Novel Nucleic Acid Modification Reversed by NADAR Hydrolases

TL;DR

Scientists discovered a new type of nucleic acid modification involving cytidine, which can be reversed by specific enzymes across different species.

Contribution

The study identifies ADP-ribosylation of cytidine as a novel modification and shows its reversal by NADAR hydrolases.

Findings

ADP-ribosylation of cytidine is catalyzed by pierisin family toxins like ScARP and Scabin.

NADAR proteins from diverse species can reverse cytidine ADP-ribosylation.

The conservation of this activity suggests physiological importance of the modification.

Abstract

ADP-ribosylation of nucleic acids is a modification found in both eukaryotes and bacteria, where it contributes to genome maintenance but can also serve as a toxic mechanism used by bacterial toxins to disrupt essential cellular processes. This modification is catalysed by ADP-ribosyltransferases and can be reversed by antagonistic ADP-ribosylgylcohydrolase enzymes. To date, ADP-ribosylation of nucleic acid bases has been described only for adenosine, guanosine, and thymidine. Here we report the ADP-ribosylation of cytidine, catalysed by members of the pierisin family of bacterial toxins—ScARP (SCO5461) and Scabin. We also show that ADP-ribosylation of cytidine is reversible through removal by certain NADAR family proteins, including NADAR proteins from the bacterium Streptomyces coelicolor (SCO5665) and the sponge Amphimedon queenslandica, as well as YbiA-type NADAR proteins. The conservation of cytidine de-ADP-ribosylating activity of NADAR proteins across phylogenetically distant species suggests that this modification may have important physiological significance.

Figures

Tables

Table 1. ADP-ribosylation of nucleic acids.

DNA/RNA-ADPr Linkage	Target	Transferase	Hydrolase	Physiology	Reference
ADPr-5′P-DNA	5′P of blunt-ended dsDNA5′P on ssDNA	PARP1 (poly)PARP3, 14Tpt1/KptA(SCO3953)	PARG, TARG1, MacroDs, ARH3,SARS2 Mac1,PARP14-MD1,PARP9-MD1	DNA damage repair	[7,8,58]
ADPr-5′P-RNA	5′P on ssRNA	PARP10, 11, 14, 15Tpt1/KptA	PARG, TARG1, SCO6735, MacroDs, SCO6450, ARH3,SARS2 Mac1,PARP14-MD1	unknown(RNA capping?)	[9,58,59]
DNA-3′P-ADPr	3′P on ssDNA	PARP1, 10, 14	PARG, TARG1, MacroD2,SARS2 Mac1,PARP14-MD1	unknown (DNA damage repair?)	[7,8,58]
RNA-3′P-ADPr	3′P on ssRNA	PARP10, 14	PARG, TARG1, MacroDs, ARH3,SARS2 Mac1,PARP14-MD1	unknown	[8,58]
DNA-A(N1)-(ADPr)n	A in ssDNA	PARP1	PARG	unknown	[12]
RNA-A(N6)-ADPr	GA in mRNA	CmdT	CmdA	anti-phage defence	[13]
DNA-T(N3)-ADPr	TNTC/TCTN in ssDNA	DarT2	DarG2,TARG1,SCO6735	anti-phage defence,regulation of DNA replication and bacterial growth	[15,16,19,48]
DNA-G(N2)-ADPr	G in ssDNA and dsDNAG nucleosides and mononucleotides	ScARP, Scabin, Pierisin-1,CARP-1,DarT1	NADAR family(SCO5665, DarG1,YbiA NADAR,phage NADAR)	anti-parasite defenceanti-phage defencemorphological differentiation and antibiotic production	[6,14,20,29,31,55], this study
RNA-G-ADPr	G in tRNA and ssRNA	ScARP, Scabin	NADAR family	unknown	[14,31,55]
DNA-C-ADPr	C in ssDNA	ScARP, Scabin	NADAR, YbiA NADAR	unknown	this study
RNA-C-ADPr	C in ssRNA	ScARP, Scabin	NADAR, YbiA NADAR	unknown	this study
RNA-2′O-ADPr	2′OH in dsRNA (5S, 16S, 23S rRNA, tRNA)	RhsP2	PARG,bactPARG	bacterial immunity	[10,31]

Full text

1. Introduction

Nucleic acids undergo a variety of chemical modifications that influence their structure, stability, and function. Among these modifications, ADP-ribosylation has recently emerged as a significant regulatory mechanism affecting DNA and RNA [1,2]. ADP-ribosylation is catalysed by ADP-ribosyltransferases (ARTs), enzymes that transfer adenosine diphosphate ribose (ADPr) from nicotinamide adenine dinucleotide (NAD^+^) to target molecules. ARTs are categorised into two families, the diphtheria toxin (DTX) family and the cholera toxin (CTX) family, based on the characteristic (but not always fully conserved) catalytic triad: H-Y-E for the DTX family and R-S-E for the CTX family [3,4]. The reverse reaction, de-ADP-ribosylation, is catalysed by ADP-ribosylgylcohydrolases, which belong to three protein superfamilies: macrodomain, ARH, and the NADAR family [5,6].

Initially recognised as a post-translational modification of proteins, ADP-ribosylation has now been established as one of the key modifications of nucleic acids [1]. It occurs at various sites, including the nitrogen atoms of nucleobases, phosphate groups at strand termini or breaks, and the ribose hydroxyl group [7,8,9,10,11]. Furthermore, reversible ADP-ribosylation has been identified on adenosine, guanosine, and thymidine residues, suggesting a potential role in epigenetic regulation and DNA repair [6,11,12,13,14,15,16].

Reversible poly-ADP-ribosylation of adenosine in ssDNA has been reported in mice and humans, albeit at low levels [12]. The modification is catalysed by PARP1 and reversed by PARG. A modification of adenosine has also been detected in RNA. In this case, it is a sequence-specific mono-ADP-ribosylation that targets adenosine in GA dinucleotides in the mRNA [13]. This modification is catalysed by the bacterial toxin CmdT, which is part of the toxin–antitoxin–chaperone system CmdTAC that, by blocking translation, plays a role in anti-phage defence [13,17]. ADP-ribosylation of DNA, catalysed by PARP1, was observed in human telomeres, where it is reversed by TARG1 and partially by PARG, suggesting that some of these modification events may be in the form of chains [18]. The exact modification site on telomeres has not been elucidated.

Guanosine and thymidine can be ADP-ribosylated by the DarT-DarG toxin–antitoxin (TA) system, which facilitates the reversible ADP-ribosylation of DNA in various pathogenic and non-pathogenic bacteria. DarT1 modifies guanosine without sequence specificity [6], whereas DarT2 modifies thymidine in a sequence-specific manner [15,19]. Their corresponding hydrolases, DarG1 and DarG2, belong to the NADAR and macrodomain families, respectively [6,15]. Both DarTG systems are involved in the anti-phage response through ADP-ribosylation of viral DNA, leading to death of the infected cell by the abortive infection mechanism [20], or regulate DNA replication, gene expression and bacterial growth [16,21]. Guanosine ADP-ribosylation (G-ADPr) can also be catalysed by members of the pierisin family, named after the first DNA-targeting ART found in the cabbage butterfly Pieris rapae [22]. Over the years, the pierisin family has expanded to include SCO5461, also known as ScARP (Streptomyces coelicolor ADP-ribosylating protein), Scabin (a toxin from the plant pathogen S. scabies), and CARP-1 (a toxin from the clam Meretrix lamarckii)—all of which are DNA-modifying ARTs that target the N2 amino group of guanosines [14,23,24,25]. Since ADP-ribosylation of nucleic acids is generally recognised as a form of DNA or RNA damage, these ARTs function as potent genotoxins but are also implicated in other important physiological processes. For example, pierisin serves as a defence mechanism against microbes and parasitoids and may also play a role in P. rapae metamorphosis [14,26]. Its strong cytotoxic effect and apoptosis induction have also been demonstrated in various human cell lines [27,28]. Both ScARP and Scabin are secreted toxins. Scabin is a virulence factor that is specific to potato genomic DNA, a species in which Scabin causes common scab disease [23]. ScARP is probably used in interspecies conflicts but is also involved in morphological differentiation and the production and secretion of antibiotics in S. coelicolor [29].

Guanosine ADP-ribosylation can be removed by different members of the NADAR superfamily. This superfamily includes NADARs associated with the DarTG1 TA system, NADARs that are widely distributed in all domains of life except vertebrates, YbiA-type NADARs and phage NADARs [6,30]. While representatives of all NADAR families can remove DNA G-ADPr in vitro, they are not equally efficient in vivo [31]. All can also remove RNA G-ADPr [11]. Furthermore, some NADARs have defined physiological roles beyond reversing G-ADPr, such as riboflavin synthesis in bacteria and plants [32], swarming motility in E. coli K-12 [33] and mRNA splicing in Caenorhabditis elegans [34]. Phages use NADARs for defence against the bacterial DarTG1 anti-phage system and show specificity towards modifications in phage DNA [35,36].

So far, it has been shown that three nucleotides, adenosine, guanosine and thymidine, can be ADP-ribosylated. Here we report the reversible ADP-ribosylation of the fourth nucleotide, cytidine. This modification is catalysed by members of the pierisin subfamily of ARTs, specifically ScARP and Scabin, which were previously characterised as guanosine-specific transferases. We further demonstrate that the S. coelicolor NADAR glycohydrolase SCO5665 can remove both guanosine- and cytidine-linked ADP-ribosylation. Together, ScARP and SCO5665 form a pair that functions in reversible ADP-ribosylation of DNA and RNA in S. coelicolor. The widespread conservation of NADAR homologues, particularly SCO5665, across phylogenetically distant species suggests an important, possibly protective, physiological function.

2. Results

2.1. ScARP Modifies DNA and RNA at Both Guanosines and Cytidines

ScARP has been primarily described as a guanosine-specific ART [14,37], and it also possesses protein ADP-ribosylating activity in vitro [29,38]. While studying substrate specificities of ScARP, we noticed that when using ssDNA oligonucleotide containing only one guanosine within a random sequence as substrate, we reproducibly observed more than the single expected shift (Figure 1A, lane 2). The observed activity was abolished when we used the catalytically inactive mutant ScARP E164Q or omitted NAD^+^ from the reaction (Figure 1A, lanes 3 and 4).

To discern the nucleotide specificity of ScARP, we tested five possible nucleotide bases (including deoxyuridine) as potential DNA substrates. We observed that ScARP can also ADP-ribosylate cytidine, in addition to guanosine, but not adenosine, deoxyuridine or thymidine (in the polyT backbone) (Figure 1B). To ensure that we used the preferred position of cytidine within the oligonucleotide, we tested cytidine at the 5′, middle, and 3′ positions within a polyT backbone and observed a slight preference for the middle-positioned cytidine (Figure S1A). We further tested a possible preference for guanosine or cytidine in a double C, double G, or the combination of the two in ssDNA oligonucleotides. ScARP ADP-ribosylated all of the available G and C nucleotides (to a higher or lesser extent), suggesting that ScARP acts without strict sequence or order constraints (Figure S1B). Since ScARP has been reported to be able to modify total RNA [14], we tested its activity on ssRNA oligonucleotides (polyU-A, -C and -G). Here, we also observed ScARP activity on ssRNA targeting both C and G bases (Figure 1D), as observed for DNA, whereas it was inactive on polyU-A ssRNA. Notably, the results suggest higher ScARP catalytic activity for G-ADPr in the RNA context as it demonstrated complete modification of polyU-G at a lower concentration (0.1 µM) than on ssDNA polyT-G (1 µM) (Figure 1C, lane 8 vs. Figure 1D, lane 10).

2.2. ScARP and Scabin Are Nucleotide-Modifying ARTs with Dual Cytidine/Guanosine Specificity

To address the questions of how widespread and evolutionarily conserved the modification of cytidine is, we searched for representatives of nucleotide-modifying ARTs and conducted a phylogenetic analysis. The phylogenetic tree of nucleotide-modifying ARTs distinguished six branches (Figure 2A). ScARP, its closest relative Scabin, a toxin from the plant pathogen S. scabies and their homologues are restricted to Actinomycetes (Gram-positive bacteria with high GC content) and form a separate group within the pierisin family (Figure 2A). In addition to the founder of this family—pierisin—it also includes the mosquitocidal toxin (MTX) from Bacillus sphaericus, the only protein-modifying ART in this family. Pierisin/MTX homologues are found in bacteria and several eukaryotic lineages, including Amoebozoa, Ascomycota and Basidiomycota. The next group/branch comprises exclusively eukaryotic CARP-1 homologues, which are mainly distributed among the Bivalvia, but also occur in representatives of the Cnidaria, Porifera and Bilateria. DarT toxins are only found in bacteria and include the guanosine-modifying DarT1 and the thymidine-modifying DarT2, which form two separate branches. The last branch consists of PARP1 homologues, with human PARP1 as a main representative, which can modify adenosine as well as protein substrates.

All the nucleotide-modifying ARTs mentioned belong to two major ART families: ScARP, Scabin, pierisin, MTX and CARP-1 belong to the CTX family, while DarT1 and DarT2 belong to the DTX family along with well-characterised PARP enzymes [16,39] (Figure 2B). With the exception of PARPs, which have complex domain architectures (39), other nucleotide-modifying ARTs are mostly single-domain proteins; only pierisin and MTX have a C-terminal ricin domain in addition to the ART domain and undergo proteolytic cleavage to become a functional ART-domain protein [40,41]. A common feature of the secreted toxins is the presence of an additional transmembrane sequence at their N-terminus (“ST” box in Figure 2B).

To determine whether the sequence similarity corresponds to a common substrate specificity, we tested the activity of the bacterial ART representatives ScARP, Scabin, MTX, DarT1 and DarT2 on both guanosine and cytidine substrates. Notably, only Scabin also displayed dual specificity for both cytidine and guanosine (Figure 2C,D). On the other hand, the ART domain of the MTX showed no detectable activity on nucleotides, but exhibited auto-ADP-ribosylation, most likely on an arginine residue (Figure 2C and Figure S2; and as shown in [41,42,43]). We also attempted to test the activity of CARP-1; however, all our efforts to produce a reliable amount of the active protein were unsuccessful. The DarTs showed only their previously identified activities—DarT1 was active only on guanosine in the conditions used [6] and DarT2 on the second thymidine in the TCTC motifs [15] (Figure 2E).

2.3. Structural Background for Cytidine/Guanosine Binding and Modification

In the ScARP structure with bound NADH and GDP (PDB: 5ZJ5), the guanosine base of GDP is recognised and positioned by specific interactions with the ARTT loop. The guanosine ring is stacked between the N-ribose of NAD^+^ and Trp159—the aromatic residue in the ARTT loop (X–X–ϕ–X–X–E/Q–X–E), which is responsible for substrate recognition [44]. The N2 and N3 atoms of guanosine form hydrogen bonds with oxygen and nitrogen atoms from the Gln162 side chain. These interactions orient the N2 amino group of guanosine to accept the ADPr moiety from NAD^+^, thereby conferring guanosine specificity and excluding adenosine as a receptor [37].

To understand how nucleotide identity influences productive binding in the ScARP active site, we performed docking studies to compare the binding modes of CDP and GDP, assessing whether one forms a more favourable interaction that better positions the NH_2_ group for modification. The docking results shown in Figure 3A (centre) indicate that CDP could bind in the active site of ScARP in a conformation that positions the NH_2_ group in the same way as in GDP, making it suitable for accepting the ADP-ribose. Based on the best predicted binding modes, it seems that the primary difference in binding between CDP and GDP lies in the orientation of the ribose (Figure 3A). In the crystal structure of the ScARP:NADH:GDP complex, the ribose can form a hydrogen bond with a water molecule (W2), which seems to be important for substrate binding. In addition, the guanosine substrate appears to be more effectively stabilised within the active site due to stronger π–π stacking interactions with Trp159 and a hydrogen bond with the backbone carbonyl of Asn114 (Figure 3A). In contrast, in our ScARP:NAD^+^:CDP docking model, the hydrogen bonding interaction with Asn114 is absent and the π–π stacking interactions are comparatively weaker. Collectively, these differences may contribute to the lower binding affinity of CDP and, consequently, its reduced efficiency in cytidine modification.

Based on docking results and structural analyses, we performed site-directed mutagenesis to investigate potential differences in guanosine and cytidine binding and to gain deeper insight into the catalytic mechanism. We first examined the role of hydrogen bonds that could be formed between the side chain of Gln162 and GDP/CDP, as well as the NAD^+^ ribose (Figure 3A), by substituting Gln162 with serine. This mutation was expected to disrupt hydrogen bonding with the ribose while potentially allowing interaction with the nucleotide amino group. Gln162 is a part of the ARTT loop (X-X-ϕ-X-X-E/Q-X-E) and known to be involved in substrate recognition, binding and specificity [44] (marked in Figure 2B and shown in the active site in Figure 3A,B). In addition, we selected Tyr110 and Ser121 because these residues are positioned near the reaction centre in the ScARP:NADH:GDP crystal structure (Figure 3B) and Tyr110 forms a water-mediated hydrogen bond with Ser121 and the NADH diphosphate (via water molecule W3 in Figure 3B). The distance between the hydroxyl groups of Tyr110 and Ser121 is 3.6 Å, while the distance between the carboxylate oxygen of the catalytically important Glu164 and the hydroxyl group of Ser121 is 2.6 Å. Given their proximity to the C1′ atom of the NADH ribose and to Glu164, we hypothesised that these residues may play important roles in NAD^+^ and substrate binding, as well as in transition-state stabilisation. We generated three single (Q162S, Y110F and S121A) and one double (Y110F/S121A) ScARP mutant. We mutated tyrosine to phenylalanine (Y110F) to remove the hydroxyl group while preserving the aromatic ring and serine to alanine (S121A) to eliminate its hydrogen bonding capacity. We tested the ability of the mutants to modify both cytidine- and guanosine-containing substrates, comparing their enzymatic activity to the wild-type enzyme and to the previously mentioned catalytic E164Q mutant (Figure 1A and Figure 3C).

The Y110F mutation did not result in a significant change in catalytic activity, suggesting that Tyr110 is not directly involved in catalysis but may play a minor role in substrate positioning. In contrast, the S121A mutation led to a decrease in catalytic efficiency (especially on cytidine), suggesting that Ser121 contributes to the reaction, possibly by stabilising the transition state or maintaining proper substrate orientation (Figure 3C, lanes 6 and 7). However, the double mutant Y110F/S121A did not show an additive effect but maintained the activity level of the single mutant S121A (Figure 3C, lanes 8 and 9). On the other hand, the Q162S mutation resulted in a significant loss of activity (Figure 3C, lanes 10 and 11). Notably, this effect was comparable to the loss of activity observed upon mutating the catalytic glutamate (E164Q), emphasising the critical role of this glutamine residue in properly positioning the nucleotide for ADP-ribosylation (Figure 3C, lanes 10–13).

The structure of ScARP was compared with its homologues Scabin (PDB: 5TLB), Pierisin-1 (PDB: 5H6J), MTX (PDB: 2VSA), and CARP-1 (AlphaFold2 prediction, Table S3). As expected, the structural alignment and metrics indicate that ScARP shares the greatest structural and sequence similarity with Scabin, with an RMSD of 0.48 Å, an SDM of 9.43, a Q-score of 0.95, and a sequence identity of 77.78%. Next in structural similarity is Pierisin-1, followed by MTX. The least structural and sequence similarity is shared with CARP-1 (Table S3), which has some additional structural elements, such as two prolonged and defined alpha helices on both sides of the binding-site cleft (Figure 3D,E and Figure S4).

As shown in Figure 3D, AlphaFold2 predicted a long, partially structured loop that blocks the active site of the CARP-1 toxin, consistent with the observation in the MTX crystal structure. However, unlike in MTX, this loop does not sufficiently inhibit the high activity of the toxin, which makes the expression of the wild-type protein in bacteria nearly impossible. Based on structural comparisons with MTX and Pierisin-1, which exhibit accessible active sites in their catalytically active conformations, we generated an in silico CARP-1 model lacking this loop to enable visualisation of the catalytic pocket and calculation of its electrostatic surface using APBS.

Electrostatic surface analysis shows that both Pierisin-1 and CARP-1 possess highly electropositive substrate-binding clefts, especially CARP-1 after the removal of the inhibitory loop (Figure 3E and Figure S4). This feature correlates with their proposed strong DNA-modifying activity. In particular, CARP-1 exhibits pronounced electropositivity, consistent with its high cytotoxicity and the difficulty of expressing it in bacterial cells, likely due to its strong DNA-binding and modification efficiency. In contrast, after removing the inhibitory loop, MTX displays a mostly electroneutral surface within the binding site, with an electronegative patch that may facilitate binding of electropositive arginine residues (Figure 3D) [41,45]. This would be in line with the experimental data [46].

2.4. Modification of Cytidine Is Reversible and Can Be Removed by Certain NADAR Hydrolases

The NADAR superfamily is divided into four families: NADARs, YbiA-type NADARs, phage NADARs, and NADARs from the DarTG1 TA system [6] (Figure 4A). Recent studies have reported that representatives of all four NADAR families efficiently reverse guanosine ADP-ribosylation on DNA and RNA [11,31]. We therefore tested their ability to reverse cytidine ADP-ribosylation in both DNA and RNA contexts. At least one representative member from each NADAR family was tested: S. coelicolor SCO5665, E. coli NADAR and eukaryotic Amphimedon queenslandica (Porifera) NADAR from the NADAR family; YbiA NADARs from E. coli and Chondromyces crocatus; phage NADAR from Escherichia phage T4; and DarG1 NADARs from Trichlorobacter lovleyi and Sinorhizobium fredii (Figure 4A, in bold). As substrates, we used ssDNA oligonucleotides ADP-ribosylated by ScARP on a single guanosine (G-ADPr) or cytidine (C-ADPr). All NADARs tested hydrolysed the G-ADPr modification (EcYibA slightly less efficiently). E. coli, T. lovleyi, S. fredii DarGs and PhT4NADAR were not able to remove the C-ADPr modification, while the members of the NADAR and YbiA families could (Figure 4B). Next, we tested the NADARs that were able to remove both modifications from DNA on ssRNA oligonucleotides modified on single cytidine or guanosine by ScARP, and observed similarly robust activity as with ssDNA (Figure 4C). The only noticeable difference was the case of E. coli YibA, which seemed to be more efficient on cytidine (DNA and RNA) and RNA- over DNA-modified G.

Having established a new reversible ADP-ribosylation system, we tested its activity at the genomic DNA level. We confirmed ScARP activity in bacterial cells using dot blot analysis of gDNA isolated from E. coli heterologously overexpressing ScARP (but not with the ScARP catalytic mutant). ADP-ribosylation on this gDNA was completely removed by the ScNADAR (SCO5665) hydrolase and partially by EcDarG (Figure S5), reflecting the presence of both C and G modifications on the gDNA, as ScNADAR removes both modifications while EcDarG removes only the G modification.

2.5. Structural Background for Cytidine/Guanosine Binding and De-Modification

To better understand how NADARs accommodate modified cytidine and guanosine during docking, we prepared ADP-ribosylated cytidine (C-ADPr) and guanosine (G-ADPr) as substrates for docking calculations. These modified nucleobases were then docked into the structure of SCO5665, predicted by AlphaFold2 (Figure 5A). As can be seen in Figure 5B,C, docking results indicate that the proper binding of cytidine and guanosine is probably facilitated by a hydrogen bond with Trp53 and T-shaped π–π stacking interactions with Trp71. The distal ribose could be positioned via hydrogen bonds with Asp172, Glu167, Lys117, and Glu87, while hydrogen bonding interactions with Trp175 and Lys117 stabilise the pyrophosphate moiety. Notably, Glu167 is structurally equivalent to the catalytically important Glu125 in P. nicotianae var. parasitica NADAR and forms a hydrogen bond with the distal ribose, which is consistent with the binding mode observed in the crystal structure of the P. nicotianae var. parasitica NADAR:ADPr complex (PDB: 8BAU and Figure 5D). As shown in Figure 5A, the adenine ring (from ADPr) exhibits positional variability, suggesting that it may adopt two binding orientations.

Highly conserved amino acids (Ser, Glu and Lys) identified in the structural sequence alignment shown in Figure 5D (red background) are shown as red balls in superimposed structures in Figure 5E. These residues are all located within the conserved core region and are a part of the substrate-binding channel. As shown by docking studies, this highly conserved Glu87 in SCO5665 NADAR is involved in the proper positioning of the distal ribose. Previously performed mutagenesis studies in E. coli C7 NADAR showed that replacing this glutamate with alanine (E88A) had only a minor effect on substrate catalysis in vitro compared to the wild-type NADAR domain [6], suggesting that this glutamate is primarily responsible for correct ribose positioning rather than direct catalytic activity. Similarly, Lys94 is positioned next to the diphosphate and seems to be essential for diphosphate binding. Mutagenesis experiments performed by Schuller et al. demonstrated that mutation of the corresponding lysine (K95) in E. coli C7 NADAR significantly impaired G-ADPr hydrolysis, reinforcing the idea that phosphate coordination is critical for NADAR function. Ser69 is also highly conserved and located within the substrate-binding cleft, suggesting a possible role in substrate positioning. Although the docking results did not indicate its direct participation in substrate binding, Ser69 may be involved in interactions with the nucleic acid backbone, which was not included in the docking calculations (Figure 5B,C).

Furthermore, the structural sequence alignment revealed that in NADARs capable of de-ADP-ribosylating both cytidine and guanosine, either tryptophan (W53 in SCO5665) or tyrosine is conserved (Figure 5D, light-blue background). Moreover, these NADARs also tend to conserve a glutamate residue (Figure 5D, blue in the middle row), which in SCO5665 corresponds to E167. However, in the DarG1 NADARs this glutamate is replaced by aspartate, and in the phage NADARs by histidine. Notably, mutation of the structurally equivalent aspartate residue in E. coli C7 NADAR resulted in a significant loss of guanosine de-ADP-ribosylation activity, highlighting its crucial role in catalysis [6]. Therefore, we decided to generate two single mutants and one double mutant of SCO5665 NADAR. To test whether these two amino acids influence substrate specificity toward cytidine nucleotides, we mutated Trp53 to serine and Glu167 to aspartate, as observed in the DarG1 family, which specifically de-ADP-ribosylates guanosine. To further assess the catalytic role of Glu167, we also substituted it with glutamine. Both Glu167 mutants (E167D and E167Q) showed a complete loss of activity toward C- and G-ADPr, confirming the essential catalytic role of this residue. Even though the aspartate has the same catalytic potential, the shorter side chain seems to be detrimental for the catalytic activity in this active-site composition. In contrast, the Trp53 mutation (W53S) did not affect activity, indicating that this position is not crucial for substrate binding or recognition of cytidine and guanosine (Figure S6).

Structural analysis of NADAR families reveals a highly conserved globular fold characterised by an α-helical core that forms the active site. The phage NADAR consists solely of this core domain. Among the families, DarG1 possesses the most extensive additional structural elements, including an α-helix and two antiparallel β-sheets, located on the N-terminal side of the core (Figure 5E). In electrostatic surface representation, these extra elements form a well-defined electropositive channel. YbiA also contains two antiparallel β-sheets on the N-terminal side. Even though its surface is less electropositive compared to other representatives, it still indicates the ability of nucleic acid binding (Figure 5F). SCO5665 and other representatives of the NADAR family (aligned in Figure 5D, in purple) lack these additional β-sheets (Figure 5E, purple); instead, their core is flanked by loops and several short α-helices, which together create a distinct electropositive tunnel likely optimised for nucleic acid binding (Figure 5F).

3. Discussion

In this study, we have described a reversible ADP-ribosylation system of S. coelicolor, which consists of the guanosine- and cytidine-specific ART ScARP (SCO5461) and the NADAR hydrolase SCO5665, which efficiently counteracts ScARP ADP-ribosylation on both DNA and RNA substrates. To our knowledge, this is the first reversible DNA ADP-ribosylation system identified in Streptomyces.

Streptomyces is a bacterial genus that utilises numerous different ADP-ribosylation systems [38,47]. The previously described reversible RNA modification system consists of the Tpt1 homologue SCO3953, which modifies the phosphorylated 5′-end of RNA, and the MacroD homologue SCO6450, which removes this modification [9]. Numerous Streptomyces species possess the DarTG2 TA system, which can coexist with the ScARP/NADAR system. Alternatively, some have the SCO6735 hydrolase that can reverse thymidine-ADP-ribosylation [48].

The distribution of ScARP/Scabin homologues is restricted to Actinomycetes, a group of Gram-positive bacteria characterised by a high GC content (Figure 2A). They are always accompanied by a SCO5665/NADAR homologue. Some Streptomyces species lack the ScARP homologue, but still have a SCO5665 homologue, suggesting that these NADARs may have additional roles beyond reversing endogenous G/C-ADP-ribosylation. They may act as immunity proteins preventing ScARP (or other G-specific ARTs that have not yet been discovered) toxicity before exiting the bacterial cell or likely present defence against toxins from competing bacteria, as suggested for SCO6735 [48].

ScARP appears to be a toxin that has the unique ability to modify two nucleobases—guanosines and cytidines. Unlike in the case of the very specific and precise DarTG systems [6,15,16], this promiscuity seems a clear advantage for a toxin. The dual specificity likely equips Streptomyces with a more versatile molecular weapon, targeting free-living bacterial competitors with high GC content, as well as GC-rich actinophages. Notably, we have identified a NADAR protein with conserved biochemical activity against G/C-ADP-ribosylation in the metazoan representative—sponge Amphimedon queenslandica. NADAR homologues can also be found in corals and sea cucumbers, which, like sponges, are known to maintain symbiotic relationships with Actinomycetes [49,50,51]. This correlation supports the hypothesis that NADAR proteins in these metazoans provide a protective defence against G/C-ADP-ribosylating genotoxins produced by symbiotic or environmental Actinomycetes. A recent study suggests that some species of ladybird beetles acquired two bacterial NADARs by horizontal gene transfer and incorporated them into their immune defence against bacterial infections. Experiments using RNAi silencing (on both NADARs) coupled with bacterial infection show detrimental effects on overall survival and intestinal tissue compared to controls [52]. We also cannot exclude the possibility that endogenous G- and/or C-ADP-ribosylation exists in eukaryotes but has not yet been discovered.

ScARP and Scabin belong to the pierisin family of toxins, previously described to comprise only guanosine-modifying ARTs [25,37,40,53,54,55]. We showed comparable activity of both ScARP and Scabin on guanosine- and cytidine-containing ssDNA and ssRNA oligonucleotides. To get more information on substrate binding and the catalytic mechanism of ScARP, we performed molecular docking, structural analyses and site-directed mutagenesis. According to our ScARP:NAD^+^:CDP docking model, we assume that the acceptor site for ADP-ribosylation is the C4 exocyclic amino group of cytosine. This assumption is supported by experimental evidence showing that uracil, the deaminated form of cytosine lacking this amino group, cannot be modified by ScARP/Scabin. Docking comparisons between GDP and CDP show that CDP can adopt a productive binding orientation but lacks the additional stabilising interactions present in GDP. The guanosine substrate appears to be more effectively stabilised within the active site due to hydrogen bonds with Gln162, the backbone carbonyl of Asn114 and a water molecule, as well as stronger π–π stacking interactions with Trp159. In contrast, CDP-bound state lacks hydrogen bonding interaction with Asn114 and the water molecule. Furthermore, the π–π stacking interactions in the ScARP:CDP complex are comparatively weaker. Collectively, these differences likely contribute to the lower binding affinity of CDP and, consequently, its reduced efficiency in cytosine modification. Q162S mutation led to a severe loss of catalytic activity, comparable to the catalytic E164Q mutant, underscoring the critical role of Gln162 in base and ribose positioning for ADP-ribosylation. The S121A mutant displayed reduced catalytic efficiency, indicating that Ser121 likely contributes to catalysis by stabilising the transition state or maintaining proper substrate orientation.

To further corroborate the scope and utilisation of cytidine ADPr modification and de-modification, we looked into the NADAR superfamily. We have already tested and described a similar arsenal on modified guanosine [31]. Here, we found that representatives of the NADAR and YbiA NADAR families were able to de-ADP-ribosylate cytidine modification, whether in DNA or RNA.

We focused on S. coelicolor NADAR SCO5665 and hypothesised from docking data that SCO5665 could accommodate both ADP-ribosylated guanosine (G-ADPr) and cytidine (C-ADPr) via key interactions: π–π stacking with Trp71, hydrogen bonds with Trp53, and ribose positioning by Asp172, Glu167, Lys117 and Glu87. To pinpoint the residues responsible for base selectivity and catalytic activity in SCO5665, we mutated Glu167 to aspartate, as aspartate occupies the equivalent position in NADARs that specifically de-modifies guanosine (this position is occupied by aspartate in DarG1 NADARs and by histidine in phage NADARs). Furthermore, we mutated Trp53, which is also involved in guanosine and cytidine binding to serine, as found in NADAR subfamilies that de-modify only guanosine. Mutation of catalytic Glu167 to aspartate produced a complete loss of function, which indicates that the shorter side chain cannot fulfil the structural/catalytic role required in SCO5665, likely because it fails to reach or engage the substrate correctly. Moreover, mutation of Glu167 to glutamine further highlights the importance of the negative charge for catalysis, as this substitution also led to a complete loss of enzymatic activity. The Trp53 mutation did not reduce activity, suggesting that nucleotide base recognition may be mediated by steric or hydrophobic interactions, or by other residues such as Ser169 (Figure S7), which has the potential to form a hydrogen bond with the carbonyl oxygen of guanosine or cytidine but was not shown to participate in substrate binding in the docking analysis. Based on overall results it is more likely that base selectivity arises from subtle and potentially cooperative interactions within the active site that are not easily uncoupled by single-point mutations.

Taken altogether, ADP-ribosylation, and more generally nucleotidylation (e.g., adenylation) of nucleic acids, is emerging as a widespread modification strategy that can target bases, phosphate groups or ribose moieties and affect DNA replication, transcription, translation, telomere metabolism and host–pathogen interactions [1,11,13,18,21,28,56] (Table 1). Unlike protein ADP-ribosylation, which is well-characterised, nucleic acid ADP-ribosylation is well on its way to being understood, with recent discoveries providing deeper insights into its role in bacterial defence, toxin–antitoxin systems, and antiviral responses. These modifications may serve as potent regulatory or cytotoxic mechanisms, offering new insights into microbial competition, innate immunity, and possibly even therapeutic targets in infectious disease and cancer.

A recent preprint reports PARP1-catalysed ADP-ribosylation of cytidine in vitro, indicating the potential physiological relevance of this modification in human DNA metabolism [57]. This finding suggests that cytidine ADP-ribosylation may not be restricted to bacterial toxin systems but rather represents a more general and physiologically relevant modification, the understanding of which has only just begun.

4. Materials and Methods

4.1. Plasmid Constructs

The genes coding for the SCO5461/ScARP and SCO5665/ScNADAR proteins were amplified by PCR from S. coelicolor M145 genomic DNA. The ScARP gene was cloned without the first 102 nucleotides, which encode the 34 amino acids of the N-terminal transmembrane signal sequence. This truncated gene and its mutated forms were cloned into the expression vector pET15b. The SCO5665 gene and its mutants were cloned into pET28b. Mutations were introduced using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA).

Genes for the Scabin protein from S. scabies, the MTX protein from Lysinibacillus sphaericus and the NADAR protein from Amphimedon queenslandica were synthesised (and codon-optimised for expression in E. coli) and cloned into pET15b (Scabin) or pET28b (MTX and AqNADAR). As with ScARP, the Scabin gene was cloned without the first 87 nucleotides, which encode the 29 amino acids of the N-terminal transmembrane signal sequence. For MTX, only the gene segment encoding the truncated protein from amino acids 30 to 308 was cloned.

The DarT1 gene from E. coli [6], DarT2 gene from Thermus aquaticus [15], NADAR genes from E. coli and Phytophthora nicotianae, DarG1 NADAR genes from Trichlorobacter lovleyi and Sinorhizobium fredii [6], YbiA NADAR genes from E. coli and Chondromyces crocatus, and the phage NADAR gene from E. coli phage T4 [31] were cloned as previously described.

All plasmid constructs were verified by sequencing.

4.2. Bacterial Strains and Culture Conditions

E. coli TOP10 strain (Invitrogen, Carlsbad, CA, USA) was used for all plasmid manipulations and BL21(DE3) strain (Stratagene, La Jolla, CA, USA) was used for the expression of the recombinant proteins. Both strains were grown in LB medium with the addition of 100 µg/mL ampicillin for the selection of pET15b and 35 µg/mL kanamycin for the pET28b. Strains were grown at 37 °C unless otherwise indicated.

4.3. Protein Expression and Purification

ScARP was expressed and purified as previously described [38]. The same expression and purification protocol was used for the ScARP mutants and Scabin.

Truncated MTX (with a deletion of N-terminal 29 residues) was expressed with a FLAG-tag at the N-terminus and a His-tag at the C-terminus, which ensured detection of the protein before and after Trypsin digestion to remove the C-terminal residues 263–308 (including the inhibitory region (263–285) and the His-tag). The final “MTX(ARTd)” used in the in vitro assays corresponds to aa 30–262. The purification protocol was essentially the same as for the ScARP.

For the expression of the NADAR proteins SCO5665, its mutants and AqNADAR, bacterial cultures were induced with 0.8 mM IPTG when an OD_600_ of 0.8 was reached and incubation was continued at 30 °C for the next 3 h. The bacteria were harvested by centrifugation and resuspended in buffer containing 25 mM Tris-HCl (pH 7.5), 500 mM NaCl and 10 mM imidazole, then lysed by addition of lysozyme at a concentration of 1 mg/mL and sonication. The cell debris was removed by centrifugation at 13,000 g for 30 min and the His-tagged recombinant proteins were purified by TALON metal affinity chromatography (Clontech, Palo Alto, CA, USA). The TALON resin was washed with the same buffer containing increasing concentrations of imidazole (10, 20 and 40 mM), while elution was performed with the buffer containing 200 mM imidazole. The purified protein fractions were pooled and desalted using columns (GE Healthcare, Chicago, IL, USA), and then stored in buffer containing 25 mM Tris-HCl (pH 7.5), 50 mM NaCl, 1 mM EDTA, 1 mM DTT and 10% glycerol (v/v) at -80 °C. The protein concentrations were determined by measuring the absorbance at 280 nm with the NanoDrop (DeNovix, Wilmington, DE, USA). All other NADAR proteins were purified as previously described [6,31].

The purity of the recombinant proteins used in this study can be seen in Figure S3.

4.4. ADP-Ribosylation Activity Assay

All oligonucleotides used in this study were synthesised by Metabion, except Cy3-labelled oligonucleotides, which were purchased from IDT.

ADP-ribosylation reactions were performed in 50 mM Tris-HCl pH 7.5 and 50 mM NaCl buffer at 30 °C for 45 min, unless otherwise indicated. For the reactions, NAD^+^ (3 mM) was added to the mixture containing ARTs and oligonucleotides (at various concentrations, as indicated in the figure legends). Reactions were stopped by the addition of formamide-containing loading buffer (80% formamide, 0.5% EDTA, 0.1% xylene cyanol, 0.1% bromophenol blue, 10% glycerol) and brief denaturation at 95 °C. Samples were resolved on a 16% polyacrylamide gel with 8 M urea, stained with SYBR Gold and visualised under UV light (G:Box, Syngene, Baltimore, MD, USA). Cy3-labelled oligonucleotides were visualised using a Molecular Imager PharosFX system (Bio-Rad, Hercules, CA, USA) with 532 nm laser excitation.

In auto-modification assays, 2 µg of recombinant ART (ScARP, Scabin and MTX ARTd) was mixed and treated as described above. Reactions were stopped by denaturing for 4 min at 95 °C. Samples were subjected to PAGE and Western blotting. After transfer, the membrane was stained with Schwartz black. The primary antibody used was Poly/Mono-ADP Ribose (E6F6A) antibody (Cell Signalling Technology, Danvers, MA, USA), diluted 1:5000 in I-Block solution (Thermo Fisher Scientific, Waltham, MA, USA).

4.5. De-ADP-Ribosylation Activity Assay

The transferase reactions were stopped by denaturation at 95 °C for 2 min, cooled on ice, and then 1 µM NADAR hydrolases were added and incubated at 30 °C for 30 min. Reactions were stopped by addition of formamide-containing loading buffer and denaturation at 95 °C for 2 min. Samples were resolved on 16% polyacrylamide gel with 8 M urea, stained with SYBR Gold and visualised under UV light.

4.6. Detection of ADP-Ribosylation on Genomic DNA

E. coli BL21 (DE3) strains overexpressing ScARP or its Q162S mutant were used to isolate gDNA after protein induction. Genomic DNA from the same E. coli strain carrying the empty vector was used as a negative control. Genomic DNA was isolated using a previously described protocol [16]. For gDNA de-ADP-ribosylation, gDNA was incubated with either buffer (control) or 1 µM of the indicated hydrolase at 30 °C for 40 min. Approximately 100 ng of gDNA was dotted onto a nitrocellulose membrane, and Poly/Mono-ADP Ribose (E6F6A) antibody was used for immunoblotting.

4.7. Phylogenetic Analysis

The protein sequences used for the phylogenetic analysis were extracted using National Center for Biotechnology Information (NCBI) resources. ART domains or NADAR/DUF1768 domains of the extracted proteins were defined using the Pfam database [60]. The evolutionary history was inferred by using the Maximum Likelihood method and the Whelan and Goldman model [61]. The trees with the highest log likelihood are shown. Initial tree(s) for the heuristic search were obtained automatically by applying Neighbour-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the JTT model and then selecting the topology with the superior log likelihood value. A discrete Gamma distribution was used to model evolutionary rate differences among sites. Trees are drawn to scale, with branch lengths measured in the number of substitutions per site. The final datasets contained a total of 234 (ART tree) or 284 (NADAR tree) positions, and the evolutionary analyses involved 34 or 28 amino acid sequences, respectively. The evolutionary analyses were conducted in MEGA X [62].

4.8. Molecular Docking

Molecular docking calculations were performed using the Glide module from the Schrödinger software package (Release 2024-4). Before receptor grid generation, the protein structures of ScARP (PDB: 5ZJ5) and SCO5665 (predicted by AlphaFold2 [63]) were prepared using Schrödinger’s Protein Preparation Wizard.

In the case of CDP docking to the active site of ScARP, the receptor grid was generated with NAD^+^ bound in the active site. The structures of CDP and GDP were prepared using LigPrep (Release 2024-4). To validate the accuracy of the docking setup, GDP was first redocked to ensure that the generated receptor grid could successfully reproduce the GDP binding observed in the crystal structure.

To better understand how NADAR accommodates modified cytidine and guanosine during docking, we prepared ADP-ribosylated guanosine (G-ADPr) and cytidine (C-ADPr). The structures of G-ADPr and C-ADPr were built in Maestro and prepared using LigPrep. The receptor grid was generated without applying any constraints. APBS surface calculations were performed using the Poisson-Boltzmann ESP surface method in Schrödinger [64].

Docking results were analysed in Maestro (Release 2024-4).