OUL232

Reversible ADP-ribosylation of RNA

ABSTRACT
ADP-ribosylation is a reversible chemical modifica- tion catalysed by ADP-ribosyltransferases such as PARPs that utilize nicotinamide adenine dinucleotide (NAD+) as a cofactor to transfer monomer or poly- mers of ADP-ribose nucleotide onto macromolecular targets such as proteins and DNA. ADP-ribosylation plays an important role in several biological pro- cesses such as DNA repair, transcription, chromatin remodelling, host-virus interactions, cellular stress response and many more. Using biochemical meth- ods we identify RNA as a novel target of reversible mono-ADP-ribosylation. We demonstrate that the hu- man PARPs – PARP10, PARP11 and PARP15 as well as a highly diverged PARP homologue TRPT1, ADP- ribosylate phosphorylated ends of RNA. We fur- ther reveal that ADP-ribosylation of RNA mediated by PARP10 and TRPT1 can be efficiently reversed by several cellular ADP-ribosylhydrolases (PARG, TARG1, MACROD1, MACROD2 and ARH3), as well as
by MACROD-like hydrolases from VEEV and SARS viruses. Finally, we show that TRPT1 and MACROD homologues in bacteria possess activities equiva- lent to the human proteins. Our data suggest that RNA ADP-ribosylation may represent a widespread and physiologically relevant form of reversible ADP- ribosylation signalling.

INTRODUCTION
Adenosine diphosphate (ADP)-ribosylation is a covalent modification in which the ADP-ribose (ADPr) group from nicotinamide adenine dinucleotide (NAD+) is transferred to diverse target molecules: proteins, nucleic acids and small molecules such as phosphate or acetate (1). This modifica- tion changes physical and chemical properties or localiza- tion of target molecules and regulates many important cel- lular processes in both prokaryotes and eukaryotes (2).ADP-ribosylation was first described as a mechanism of pathogenicity used by pathogenic bacterial exotoxins that irreversibly modify crucial host cell proteins (3). Two divergent bacterial toxins, diphtheria and cholera toxin are founders of two major ADP-ribosyl transferase (ART) groups (4,5). Poly(ADP-ribose) polymerases (PARPs), the best studied and largest ART subgroup, belong to diphthe- ria toxin-like ADP-ribosyl transferases. PARPs are present in all eukaryotes (except yeast) and sporadically in bac- teria; they regulate important cellular processes such as DNA damage repair, transcription, protein degradation, cell-cycle progression, host-virus interaction, cell division, ageing, cell death and bacterial metabolism (6–9). The human genome encodes for seventeen PARPs with dif- ferent domain architecture and functions (2). tRNA 2r- phosphotransferase 1 (TRPT1/Tpt1/KptA) is sometimes referred to as the eighteenth PARP family member (4). Several PARP family members (PARP1, PARP2 and tankyrases) synthesize long chains of poly-ADPr, while the other PARP family members transfer a single ADPr group on targets (such as PARP3 and PARP16) (7).

ADP-ribosylation is a dynamic chemical modifica- tion that is regulated both at the level of addition and the removal of ADPr groups. PARPs have been shown to target mostly Glu/Asp or Ser residues (8–12). Poly- ADP-ribosylation can be removed by the action of two divergent enzymes, poly(ADP-ribose) glycohydrolase (PARG) and ADP-ribosylhydrolase 3 (ARH3) (13,14). PARG is unable to remove the last ADPr group attached to target proteins (13) while ARH3 is the only hydrolase that can completely remove both poly- and mono-ADPr signal from serine residue (15), a modification catal- ysed by PARP1/HPF1 and PARP2/HPF1 complexes (11,16). Terminal ADPr linked to Glu/Asp is removed by macrodomain containing enzymes namely terminal ADPr glycohydrolase 1 (TARG1/OARD1), MACROD1 and MACROD2 (17–20). Enzymes with phosphodi- esterase activity, NUDT16 (nucleoside diphosphate-linked moiety X-type motif 16) and ENPP1 (ectonucleotide pyrophosphatase/ phosphodiesterase 1), can cleave pyrophosphate from both poly- and mono-ADPr modified targets leaving phosphoribose tags on the proteins (21,22). Although ADP-ribosylation has historically been consid- ered to mostly target proteins, there has been increasing evi- dence that DNA can also be a target for ADP-ribosylation. The first enzymes reported to ADP-ribosylate DNA were pierisins, toxins expressed by cabbage butterfly and re- lated species. Piersins irreversibly ADP-ribosylate DNA on guanines (23,24). More recently, it was discovered that the bacterial toxin-antitoxin system DarT-DarG mediates reversible DNA ADP-ribosylation on thymidine residues in single-stranded DNA (in a sequence specific manner)(25).

Furthermore, it was shown that DNA repair PARPs (PARP1, PARP2 and PARP3) can modify DNA on phos- phates at DNA breaks (26–29). The ADPr groups on phos- phates at DNA breaks are efficiently removed by several cellular hydrolases, most notably by PARG, MACROD1/2, TARG1 and ARH3 (26,27,30).The pool of cellular substrates for ADP-ribosylation has continued to expand and it was recently shown that a PARP-like proteins TRPT1/Tpt1/KptA from bacteria and fungi can ADP-ribosylate RNA and DNA ends (31).In this paper, we reveal that ADP-ribosylation of RNA at the terminal phosphate is more widespread than ini- tially thought. We demonstrate that homologues of TRPT1 in higher organisms as well as human PARP10, PARP11 and PARP15 can ADP-ribosylate phosphorylated ends of RNA. We also show that RNA ADP-ribosylation is a re- versible process that can be accomplished by several hu- man hydrolases as well as by some viral and bacterial macrodomains. Thus, this study provides the first evidence of reversible ADP-ribosylation of RNA.Plasmids expressing full length (FL) PARP3 was cloned into pDEST17 vector with His tag. cDNA encoding the human PARP4 catalytic and BRCT domains (1-572aa) were obtained using gBlock gene fragments and cloned into a pET-His-SUMO-TEV using ligation independent cloning. PARP5bcat (Tankyrase2 catalytic domain) was cloned into pET-His6 vector. PARP10 catalytic domain (818-1025aa) WT and G888W mutant genes were cloned into pGEX-4T1 vector with GST tag. PARP10 FL and PARP16 FL genes were cloned into pET-His6-SUMO-TEV vector. PARP11 catalytic domain (128–338aa) was cloned in pET28a vector. The catalytic domain of human PARP12 (489–684aa) was PCR-amplified from the cDNA library using primers with non-complementary restriction enzyme sites located at the 5r (EcoRI) and 3r (XhoI) ends. The amplified product was cloned into pET-28b+ (Novagen). PARP13 catalytic domain (716–902aa) was PCR amplified and cloned into pET28a vector using BamHI and XhoI cloning sites. PARP14 wwe and catalytic domain (1459– 1801aa) and PARP15 catalytic domain (481-678aa) were cloned into pNic-Bsa4-6xHis vector.

TRPT1 gene (Uniprot Q86TN4) was codon optimized and synthesized with His tag from Invitrogen GeneArt Gene synthesis and then fur- ther cloned into pET28a vector using NcoI and XhoI re- striction enzyme cloning. Point mutants of TRPT1 wereprepared by using Agilent Quick Change Lighting Site Di- rected Mutagenesis kit. Streptomyces coelicolor KptA ho- mologue (SCO3953) was cloned from S. coelicolor genomic DNA into pET15b.PARP3 FL was purified as mentioned earlier (27). PARP4, Tankyrase2 cat (32), PARP12 cat and PARP16 FL plasmids were transformed into Escherichia coli BL21 (DE3) competent cells (Millipore) and grown on LB agar plates with Kanamycin (50 mg/ml) and Chlorampheni- col (34 mg/ml) overnight at 37◦C. A swath of cells were inoculated into a 50 ml starter culture of LB media with Kanamycin and Chloramphenicol at 225 rpm, 37◦C overnight. For each protein of interest 1 L of terrific broth (TB) media (12 g Bacto Tryptone, 24 g yeast extract, 0.4% glycerol, 17 mM KH2PO4, 72 mM K2HPO4, 1% glucose,50 µg/ml Kanamycin, 34 µg/ml Chloramphenicol) was inoculated with the starter culture and grown to 0.8–1.0OD600 at 37◦C, 225 rpm. IPTG (Sigma-Aldrich) was added to 0.4 mM to induce protein expression for 18–24 h at 16◦C, 225 rpm. Cells were harvested by centrifugation, re- suspended in lysis buffer (20 mM HEPES, pH 7.5, 1 mMβ-mercaptoethanol, 1 mM benzamidine, 0.2% NP-40, 0.2% Tween-20, 500 mM NaCl, 1 mM phenylmethylsulfonyl flu-oride (PMSF), 8.3 mg/l DNAse I (Roche)) and lysed by sonication at 0◦C (Branson sonifier 450). Lysates were incu- bated with pre-washed Ni-NTA agarose resin (50% slurry, Qiagen) with end-over-end rotation at 4◦C for 1 h. Follow- ing extensive washing with buffer B1+20 (20 mM HEPES, pH 7.5, 1 mM β-Me, 1 mM PMSF, 1 mM benzamidine,500 mM NaCl, 20 mM imidazole) protein was eluted in fourfractions of B1 containing 100–400 mM imidazole. Frac- tions containing protein were collected and dialysed against 50 mM Tris–HCl, pH 7.5, 0.1 mM EDTA, 1 mM β-Me,0.4 M NaCl at 4◦C.PARP10 catalytic domain (WT and G888W mutant) were purified as mentioned earlier (22). In short, GST- tagged PARP10 was transformed into Rosetta DE3 compe- tent cells and grown in LB media supplemented with Ampi- cillin and Chloramphenicol.

Cultures were induced with0.5 mM IPTG at 0.8–1.0 OD600 and grown overnight at 16 ◦C. Following centrifugation, PARP10 cat bacterial cell pellet was resuspended in PBS buffer supplemented with BugBuster protein extraction reagent, Benzonase, 10% glyc- erol, 1 mM DTT and Complete Protease inhibitor cocktail and allowed to lyse by incubation at 4◦C for 1 h. Lysate was further centrifuged and cleared lysate was applied to glutathione sepharose beads for 1 h at 4◦C. GST-tagged PARP10 was eluted using lysis buffer supplemented with 20 mM reduced glutathione. Eluted protein was further dialysed against 25 mM Tris–HCl pH 7.5, 150 mM NaCl, 10% glycerol and 1 mM DTT. PARP10 cat WT and G888W mutant were further purified on Superdex 200 column.PARP10 FL was transformed into Rosetta DE3 compe- tent cells and grown in 6 l 2× YT media supplemented with Kanamycin. Cultures were induced with 0.5 mM IPTG at 0.6–0.8 OD600 and grown overnight at 18◦C. Bacterial pellet was resuspended in lysis buffer (20 mM HEPES pH 8, 500 mM NaCl, 10 mM imidazole and 0.5 mM TCEP). PARP10 FL was purified via three-step purification process involv- ing Nickel column purification, heparin column and gel fil- tration column. Cells were lysed by addition of BugBuster,protease inhibitor cocktail, benzonase and lysozyme and al- lowed to lyse for 1 h at 4◦C. Cleared lysate was incubated with pre-washed Ni-NTA agarose resin (50% slurry, Qia- gen) with end-over-end rotation at 4◦C for 1 h. Beads were then further washed with high salt buffer (20 mM HEPES pH 8, 1 M NaCl, 10 mM imidazole and 0.5 mM TCEP) followed by gradient elution over 10 mM–1 M imidazole. Eluted protein was assessed by SDS-PAGE gel and further dialysed against 25 mM Tris pH 7.5, 100 mM NaCl, 1 mM EDTA and 0.1 mM TCEP. Dialysed protein sample was fur- ther diluted using no salt buffer (25 mM Tris pH 7.5, 1 mM EDTA and 0.1 mM TCEP) to get the salt concentration to 30 mM NaCl and was applied onto Heparin column to remove any nucleic acid contamination.

Small fraction of protein bound onto Heparin column while most ran out as flow through in the condition tested. The Heparin col- umn bound protein was eluted with gradient of 30 mM–1 M NaCl concentration. At this stage, the purity of eluted pro- tein was tested by SDS-PAGE gel. Protein fractions were concentrated and further subjected to size exclusion chro- matography using Superdex 200 column.PARP13 cat (716-902aa) and PARP14 wwe and cat (1459–1801aa) plasmids were transformed in Rosetta DE3 competent cells and grown in 2× YT media supplemented with Kanamycin. Induction was carried out at 0.6–0.8 OD600 using 0.5 mM IPTG and cells were allowed to grow overnight at 18◦C. Bacterial pellet was lysed in lysis buffer (20 mM HEPES pH 8, 500 mM NaCl, 10 mM imidazole and 0.5 mM TCEP) supplemented with BugBuster, pro- tease inhibitor cocktail, benzonase and lysozyme. Cleared lysate was then bound to pre-washed Ni-NTA agarose resin followed by washes with lysis buffer. Proteins were eluted using elution buffer (20 mM HEPES pH 8, 500 mM NaCl and 0.5 mM TCEP) with an incremental gradient of 10–500 mM imidazole. Proteins purity was assessed by SDS-PAGE gel. PARP13 protein was dialysed overnight against 25 mM Tris pH 7.5, 100 mM NaCl, 1 mM EDTA and 0.1 mM TCEP buffer. PARP14 protein was dialysed overnight against 20 mM HEPES pH 7.5, 300 mM NaCl and 0.5 mM TCEP and further subjected to Superdex 75 column for size exclusion chromatography. The catalytic do- main of PARP15 (481–678aa) was purified as mentioned earlier (33).TRPT1 was purified as described earlier (34), in short TRPT1 plasmid was transformed into Rosetta DE3 com- petent cells and grown in LB media supplemented with Kanamycin. Cultures were induced with 0.2–0.5 mM IPTG at 0.8 OD600 and grown overnight at 16 ◦C. Following centrifugation, TRPT1 bacterial pellet (WT or mutants) was resuspended in lysis buffer (50 mM Tris–HCl pH 7.5, 150 mM NaCl, 10% glycerol and 10 mM imidazole) sup- plemented with BugBuster, Benzonase, 0.5 mM TCEP and Complete protease inhibitor cocktail and lysed by mixing for 1 h at 4◦C. Lysate was centrifuged at 17 000 rpm for 50 min and the cleared lysate was incubated with prewashed nickel NTA agarose beads for 1 h at 4◦C. His-tagged TRPT1 protein was eluted using elution buffer (50 mM Tris–HCl pH 8, 0.1 M NaCl, 10% glycerol) with an incremental gra- dient of 10–500 mM imidazole.

Eluted TRPT1 protein was dialysed overnight against column buffer (50 mM Tris–HCl pH 8, 50 mM NaCl, 2 mM DTT and 10% glycerol). TRPT1WT protein was further purified by size exclusion chro- matography using Superdex 75 column.Streptomyces coelicolor KptA homologue (SCO3953) gene was expressed in Escherichia coli BL21(DE3) cells and were grown for 3 h at 30◦C with 0.8 mM IPTG added at 0.8 OD600. Recombinant protein was purified using TALON affinity resin according to standard procedure. Strepto- myces coelicolor MacroD homologue (SCO6450) was ob- tained as described earlier (30).Proteins listed below were gifts from other members of the lab. Mycobacterium tuberculosis (Mtb) DarG-macro was cloned with 155 N-terminal amino acids as described earlier (25). Catalytic domains of PARP11 (128-338aa) (35), MACROD1 (30,36), MACROD2 (19), TARG1 (17), PARG(37), NUDT16 (21) and ARH 1-3 (15) were purified as de- scribed earlier. Viral macrodomain-containing hydrolases from VEEV (38) and SARS Coronavirus (39) were prepared as described earlier.Single stranded (ss) RNA and DNA oligos used in this study were commercially ordered from Sigma-Aldrich and Invitrogen, respectively, and are listed in Table 1. Oligonu- cleotides were diluted to 100 µM stock solution in 20 mM HEPES–KOH (pH 7.6) and 50 mM KCl buffer. Doublestranded (ds) DNA was prepared by annealing complemen- tary strands of DNA (ssDNA oligo with RexT) at 95◦C for 5 min and then allowed to gradually cool down to room temperature.10 pmol noP ssRNA (with or without Cyanine3 tag at 3r end) was radioactively labelled at 5r end using T4 polynu- cleotide kinase 3r phosphatase minus (NEB) in presence of γ 32P ATP (Perkin Elmer) and heated at 37◦C for 30 min followed by heat inactivation at 65◦C for 20 min. Radiola-belled oligo was further desalted on G25 column to remove any unincorporated ATP.

This radiolabelled oligo was used as size marker as indicated in figures.ADP-ribosylation assays with RNA were performed as de- scribed previously for DNA ADP-ribosylation (27). All buffers were made in DNase/RNase free water and filter sterilized prior to use. In short, 10 µl reaction mix was pre- pared in buffer containing 20 mM HEPES–KOH (pH 7.6),50 mM KCl, 5 mM MgCl2 and 1 mM DTT. RNA substrate (10 µM) was added along with 2 µM protein, 50 µM NAD+ (Trevigen) and 50 kBq 32P labelled NAD+ (PerkinElmer) per reaction. Protein and NAD+ concentrations were usedas mentioned above unless stated otherwise. Reactions were incubated at 37◦C for 30 min and stopped by addition of 50 ng/µl Proteinase K and 0.15% SDS and heating the re- action at 50◦C for 30 min, unless stated otherwise. Reactions that were treated with Benzonase or calf intestinal phos-phatase (CIP) were heated at 37◦C for 30 min. Samples were further heated at 95◦C for 3 min with 2× TBE urea sample buffer (8 M urea, 20 µM EDTA pH 8.0, 2 µM Tris pH 7.5 and bromophenol blue). The samples were loaded on a pre- run denaturing urea PAGE gel made of 20% (w/v) polyacry- lamide, 8 M urea and 1× TBE. The gel was run at 7 W/gelin 0.5× TBE buffer. The gel was dried under vacuum and visualized by autoradiography.Non-radioactive RNA ADP-ribosylation assay was per- formed using Cyanine3 labelled RNA essentially similar to radioactive assay with exception of using 1 µM labelledRNA oligo and 500 µM NAD+. The gel was visualized us-ing Molecular Imager PharosFX systems using laser exci-tation for Cyanine3 fluorophore at 532 nM wavelength. All ADP-ribosylation assays were individually repeated three times.To study the effect of time kinetics on PARP10 cat and TRPT1 mediated RNA ADP-ribosylation, reaction sam- ples were prepared as mentioned earlier. Aliquots were taken out at different time points (3, 10 and 30 min). The 0 min time was done by placing the reaction on ice. Reactions were stopped by addition of 2× TBE urea sample buffer.To study the effect of NAD+ concentration dependence of PARP10 cat and TRPT1 the assay was performed as de- scribed earlier with different concentrations (0–1500 µM) of NAD+ in reaction. NAD+ concentration dependence study was performed in non-radioactive setup using Cya-nine3 labelled RNA.RNA ADP-ribosylation reaction studying the ef- fect of adenosine mono-phosphate (AMP) or 5r- phosphoadenosine 3r phosphate (PAP) on PARP10 mediated RNA modification was performed by supple- menting the reaction with 0, 50 or 250 µM concentration of AMP or PAP.PARP10 catalysed RNA ADP-ribosylation reactions were stopped by addition of PARP inhibitor, 3-aminobenzamine (3ABA) before treating with hydrolases. 2 µM hydrolase enzymes were added per reaction and heated at 30◦C for 30 min. Reactions containing NUDT16 were supplementedwith 15 mM MgCl2 (21).

RESULTS
In recent years there has been increasing evidence of DNA as a new target for reversible ADP-ribosylation. We wanted to investigate whether RNA could also be similarly ADP- ribosylated by any known ARTs. We decided to initially testPARP3 as it was recently demonstrated that this protein has robust ART activity on DNA (27–29). In addition, we fo- cused on another member of the PARP family, PARP10, which contains a RNA-recognition motif (RRM domain)(40). Purified PARP3 and PARP10 (catalytic domain) were first tested with a 21 nucleotide single-stranded RNA (ss- RNA) oligo with or without a phosphate group at the 5r end in the presence of 32P labelled NAD+ as an ADPr donor. Double-stranded DNA was used as a positive control (Fig- ure 1A, lane 1) for ADP-ribosylation on DNA by PARP3(27). Strikingly, PARP10 substantially modified the 5r phos- phorylated ssRNA oligo (Figure 1A, lane 7) reducing its mobility compared to the phosphorylated oligo labelled on its 5r phosphate using 32P labelled gamma-ATP (Figure 1A, lane 2). PARP10 was not able to ADP-ribosylate ss- RNA without a 5r phosphate group (Figure 1A, lane 7 com- pared to lane 9), suggesting that RNA ADP-ribosylation by PARP10 occurs on the 5r phosphate. Importantly, PARP3 could only ADP-ribosylate DNA ends and did not have any activity on RNA oligos, while PARP10 specifically modi- fied phosphorylated ssRNA oligo in the conditions tested (Figure 1A and B). We further demonstrate RNA ADP- ribosylation activity of PARP10 is time and NAD+ con- centration dependent (Supplementary Figure S1A and B) but independent of MgCl2 presence (Supplementary Fig- ure S1C, lanes 1 and 2). To further ascertain the specificity of ADP-ribosylation by PARP10 on DNA and/or RNA ends we tested both single-stranded oligos with or without a phosphorylated moiety at either 5r or 3r end. PARP10 mod- ified ssRNA oligos phosphorylated at either end but did not show activity on ssDNA oligo irrespective of the terminal phosphorylation state (Figure 1C, lanes 5 and 6). The reac- tion product catalysed by PARP10 in presence of phospho- rylated ssRNA was stable against Proteinase K treatment (Figure 1D, lanes 3 and 6) but not against treatment with Benzonase (Figure 1D, lanes 4 and 7) confirming that the modification was on nucleic acid and not on protein. Next we wanted to assess the specificity of PARP10 for modifica- tion of the phosphate groups on RNA.

For this, we analysed RNA ADP-ribosylation catalysed by PARP10 catalytic do- main in the excess presence of adenosine mono-phosphate (AMP) or 5r-phosphoadenosine 3r phosphate (PAP) as po- tential competitors and we observed no significant change in RNA modification in presence of nucleotide analogue in excess (Supplementary Figure S1D).To further demonstrate that ADP-ribosylation of RNA by PARP10 occurs on terminal phosphates we designed 5r and 3r phosphorylated ssRNAs with the Cyanine3 (Cy3) label at the opposite end to the phosphorylation for more sensitive detection. We also produced non-phosphorylated versions of these oligos as additional controls. The phos- phorylated oligos 5P ssRNA 3Cy3 and 3P ssRNA 5Cy3 treated with PARP10 produced a slower migrating ADP- ribosylated product (Figure 1E and F, lane 2). When the re- action was further treated with calf intestinal phosphatase (CIP, phosphatase), the slower migrating ADP-ribosylated band remains intact while the lower unmodified band shifts upwards due to the removal of a charged phosphate group by CIP phosphatase treatment (Figure 1E and F, lane 3) and now migrates the same as non-phosphorylated oligo control (Figure 1E and F, lane 5). When the phosphory- lated oligos 5P ssRNA 3Cy3 and 3P ssRNA 5Cy3 (Fig- ure 1E and F, lane 1) are treated directly with CIP in ab- sence of PARP10, dephosphorylation of oligo is observed which is confirmed by similar migrating pattern as the non- phosphorylated oligo (Figure 1E and F, lanes 4 and 5). PARP10 modified 5r phosphorylated end more efficiently than 3r phosphorylated RNA end. These results show that ADP-ribosylation of RNA by PARP10 occurs on the phos- phate group thereby protecting the phosphate group from the dephosphorylation activity of CIP. Since the catalytic domain of PARP10 lacks two-thirds of the protein includ- ing the RRM we wanted to assess the activity of the FL PARP10 on RNA substrate. We expressed and purified full length PARP10 and tested it for RNA modification activity. We observed PARP10 full length can also ADP-ribosylate phosphorylated RNA ends (Figure 1G, lane 2), however, in comparison to WT PARP10 catalytic domain the activity was much weaker (Figure 1G, lane 4).

A possible explana- tion for this finding may be that the isolated FL PARP10 exists in an autoinhibited state, due to the inhibitory func- tion of another domain within the protein. Such autoin- hibitory property has been already well characterized for PARP1 (41). We observed no RNA modification by pre- viously characterized catalytic mutant of PARP10 G888W (Figure 1G, lane 3).ADP-ribosylation of RNA ends by PARP10 is a reversible processADP-ribosylation of proteins and DNA is reversible. We wanted to investigate whether RNA ADP-ribosylation by PARP10 is also a reversible process. Since PARP10 can ADP-ribosylate both 5r and 3r phosphorylated ends of RNA, we tested both of these modified oligos as substrates for well characterized human ADP-ribosylhydrolases: PARG, TARG1, MACROD1, MACROD2 and ARH1-3. We also tested human NUDT16 which is known to cleave the pyrophosphate bond in ADP-ribosylated pro- teins to generate phospho-ribose modified proteins (21). All above mentioned hydrolases, except for ARH1 and ARH2, were able to remove ADP-ribosylation from ei- ther 5r or 3r phosphates at the end of RNA substrates (Figure 2A and B). Importantly, catalytically inactive mu- tants of MACROD1 (G270E) and ARH3 (D77A) (27,30)did not remove ADPr from phosphorylated-ssRNA (Fig-ure 2C), demonstrating that the enzymatic activity of these ADP-ribosylhydrolases is required for efficient re- moval of ADP-ribose from RNA phosphorylated ends. To- gether, these results demonstrate that ADP-ribosylation of phosphorylated-RNA oligos catalysed by PARP10 is a re- versible process.Previous studies have shown PARP10 to be an inter- feron induced gene that inhibits replication of Venezuelan equine encephalitis virus (VEEV) and other alphaviruses (42), yet the physiological substrates for PARP10 an- tiviral activity remain unknown.

Thus, we tested viral macrodomain-containing hydrolases from VEEV and se- vere acute respiratory syndrome coronavirus (SARS CoV) for their ADP-ribosylhydrolase activity on RNA substrates. These viral hydrolases are known to support the ability of viruses to replicate in host cells (38,39,43–45), but their physiological substrates have yet to be identified. Strik- ingly, 5r and 3r phosphorylated ssRNA ADP-ribosylated by PARP10 could be efficiently reversed by the addition of vi- ral macrodomain proteins (Figure 2D). The ability of vi- ral macrodomains to remove ADPr from PARP10-modified phosphorylated-ssRNA could indicate a potential biolog- ical role for PARP10 in antiviral response acting on viral RNAs and the role of viral macrodomains in suppressing this function. Viral macrodomains could also reverse ADP- ribosylation of both double stranded and single stranded DNA similar to single stranded RNA modification (Figure 2E).NAD+ dependent phosphotransferase TRPT1 reversibly caps RNA endsNext we decided to check several other human PARPs for their ability to modify RNA. We tested full length PARP3 and PARP16 (46); catalytic domains of PARP4, Tankyrase2 (47), PARP10, PARP11, PARP12, PARP13, PARP14 andPARP15 and a highly diverged PARP-like protein some- times annotated as 18th human PARP–TRPT1 (4,48) for RNA ADP-ribosylation activity.We observed that, in addition to PARP10, PARP11, PARP15 and human TRPT1 were also able to ADP- ribosylate 5r phosphorylated ssRNA (Figure 3A, lanes 5, 6, 10 and 12). However, in the conditions tested, the other PARPs were unable to ADP-ribosylate 5rphosphorylated ssRNA (Figure 3A). We focused on the ADP-ribosylation activity of TRPT1. We wanted to test whether the ADP- ribosylation by TRPT1 was phosphate dependent and had specificity towards DNA and/or RNA. For this, we tested ssDNA and ssRNA oligo with a phosphate group at ei- ther 5r or 3r end or without phosphate group (noP). We ob- served TRPT1 can ADP-ribosylate both DNA and RNA but only in the presence of 5r phosphorylated end (Figure 3B). We observe TRPT1 based RNA modification is time and NAD+ concentration dependent (Supplementary Fig- ure S2A and B) however independent of MgCl2 (Supple- mentary Figure S1C, lanes 3 and 4).

Conserved amino acid residues Arg-His-Arg-Arg are essential for 2r phosphotrans- ferase function of yeast Tpt1 (34,49). Based on sequence alignment we mutated the corresponding residues in human TRPT1 into single alanine based point mutations (R40A, H41A, R86A and R150A). These point mutants were un-able to ADP-ribosylate 5r phosphorylated end of RNA (Figure 3C). This suggests ADP-ribosylation at 5r end of RNA is also mediated via the same active site as originally studied for NAD-dependent 2r RNA phosphotransferase activity of yeast Tpt1. To further establish that the ADP- ribosylation signal observed in the presence of TRPT1 was an RNA dependent modification, we further treated the re- action with Proteinase K or Benzonase. The band observed in presence of TRPT1 and 5rP ssRNA was resistant towards Proteinase K treatment but not Benzonase, which validates the band to be nucleic acid related ADP-ribosylation (Fig- ure 3D). We treated 5P ssRNA Cyanine3 tagged oligo (5P ssRNA 3Cy3) with TRPT1, which generated 50% of the ADP-ribosylated product of slower mobility in our condi- tions (Figure 3E, lanes 1 and 2). Treatment of unmodified 5rP oligo with CIP led to a band that migrated slower, simi-lar to the non-phosphorylated oligo (Figure 3E, lane 1 ver- sus lanes 4 and 5). CIP treatment of TRPT1 catalysed RNA substrate leaves the upward shifted ADP-ribosylated RNA oligo intact but the lower unmodified RNA oligo migrates similar to non-phosphorylated oligo (Figure 3E, lanes 3 and 5). These results confirm ADP-ribosylation mediated by TRPT1 is on the RNA oligo and the modification oc- curs on the phosphate group at the 5r end which protects RNA against further dephosphorylation by CIP. We also tested smaller 7mer and 12mer RNA oligos and established that TRPT1 activity is not affected by the length of RNA oligo (Figure 3F).Similar to PARP10 mediated RNA ADP-ribosylation, we wanted to test if the TRPT1 catalysed RNA mod- ification could be reversed by known human ADP- ribosylhydrolases. We observe the removal of ADPr signalby PARG, TARG1, MACROD1, MACROD2, ARH3 andNUDT16 (Figure 3G).

ARH1 and ARH2 were unable to reverse the RNA modification. Catalytically inactive mu- tants of MACROD1 (G270E) and ARH3 (D77A) were in- active compared against the wild-type hydrolase as seen ear- lier for PARP10 (Figure 3H). TRPT1 mediated DNA mod- ification can also be reversed by the above tested human hy- drolases (Supplementary Figure S2C). We also tested hy-drolase function on the ADP-ribosylated 5P ssRNA 3Cy3 oligo. The phosphorylated oligo when treated with TRPT1 produces a slow migrating band and a faster migrating un- modified RNA band (Figure 3I, lane 2). Further treatment with hydrolases: PARG, MACROD1 and MACROD2 re- verses the modification observed by the loss of the slow mi- grating ADP-ribosylated band (Figure 3I, lane 3–5). How- ever, the reversal of modification by the hydrolases doesnot change the migration pattern of the lower unmodified band (Figure 3I, lane 3–5), that still matches the migra- tion pattern of phosphorylated ssRNA (as seen in Figure 3I, lane 1 versus 3–5).

This confirms the hydrolase medi- ated hydrolytic cleavage of ADPr group does not affect the phosphate group on which the modification is covalently at- tached.ADP-ribosylation activity of TRPT1 is conserved in different speciesTRPT1 homologues are distributed across eukaryal, ar- chaeal and bacterial domains of life. In E. coli bacteria, these proteins are usually referred to as KptA. We wanted to test if RNA ADP-ribosylation by KptA homologs is con- served across different species. In addition to TRPT1, we also tested KptA homolog from S. coelicolor (Sco KptA/ SCO3953) with different RNA substrates. As with earlier for TRPT1 experiment, we observe that Sco KptA could also exclusively ADP-ribosylate RNA at 5r phosphorylated end (Figure 4A, lanes 2 and 5). Since S. coelicolor also possesses a MacroD-like protein SCO6450 similar to hu- man MACROD1 and MACROD2 we were interested to investigate if this macrodomain could function as poten- tial hydrolase to remove ADP-ribosylation mediated by Sco KptA. Using MACROD1 as a positive control for re- moval of ADP-ribosylation mediated by both Sco KptA and TRPT1 we tested SCO6450 and another known bac- terial macrodomain fold containing hydrolase DarG (25). We observed that SCO6450 was proficient at reversing RNA ADP-ribosylation mediated by both Sco KptA and TRPT1, however, DarG was inactive against KptA mediated modi- fication (Figure 4B and C).

DISCUSSION
ADP-ribosylation is an important chemical modification which helps cells to adapt and survive while maintaining their genomic integrity when faced with challenging envi- ronmental conditions. Classic macromolecular targets of ADP-ribosylation have been proteins, however, there have been several studies in the past few years that have demon- strated DNA as an important target for ADP-ribosylation (25–29). Here, we set out to uncover if any member of the ADP-ribosyltransferase family could also ADP-ribosylate RNA. We demonstrate, for the first time, that ADP- ribosylation of RNA can be catalysed by a few members of PARP family––PARP10, PARP11, PARP15 and a PARP-like protein –TRPT1 previously characterized as an NAD+ dependent phosphotransferase (50,51).PARP10 was one of the first intracellular mono(ADP- ribosyl)ating ARTs identified (52). In addition to the cat- alytic domain, PARP10 also contains a RNA recognition motif (RRM), two functional ubiquitin interaction motifs (UIM), a sequence that promotes nuclear targeting as well as nuclear export and a motif that mediates interaction with PCNA (PIP) (40,52–55). While several protein targets of PARP10 have been suggested (35), the physiological role of PARP10 is unclear. Our study sheds further light on the po- tential biological function of PARP10 through modificationof RNA. Our results show PARP10 can ADP-ribosylate phosphorylated RNA ends with a modest preference for 5r over 3r ends. PARP10 mediated RNA ADP-ribosylation is resistant to phosphatase treatment which would indicate a novel RNA capping mechanism possibly protecting the RNA against the nuclease attack. While the biological rel- evance for this RNA based modification is currently un- known we postulate a potential role in the innate immune response. PARP10 has previously been shown to be induced by interferon and can inhibit viral replication (42,43,56).PARP10 can also inhibit the activation of NF-nB which is activated during infection (55).

ADP-ribosylation of RNA by PARP10 could act as a signal/marker to initiate an ap-propriate immune response. PARP10 has an inhibitory ef- fect on alphavirus replication and on protein biosynthesis (42,56). These inhibitory effects could potentially be medi- ated via RNA ADP-ribosylation, where the ADPr moiety acts as a RNA cap thereby preventing RNA translation or triggering signal transduction. The presence of RRM do- main in PARP10 could have a role in differentiating foreign RNA of invading pathogens from host RNA to work in tandem with the catalytic domain to ADP-ribosylate RNA and to further initiate the immune response. Similar func- tion has been observed for the apoptotic role of PARP10 whereby the RRM domain contributed to pro-apoptotic ac- tivity together with the catalytic domain (57). While RNA ADP-ribosylation could provide an interesting link towards explaining the anti-viral role of PARP10, equally this activ- ity could function in initiating or inhibiting translation thus effecting a cascade of signal transduction.Several human ADP-ribosyl hydrolases PARG, TARG1, MACROD1/2 and ARH3 can reverse RNA ADP- ribosylation mediated by PARP10. Localization of these hydrolases to nucleus, cytoplasm and mitochondria (17,30,58,59) suggests that RNA ADP-ribosylation is utilized in different cellular compartments. Interestingly, in addition to human hydrolases we also observe that VEEV and SARS viral macrodomain-containing hydro- lases can remove RNA ADP-ribosylation mediated by PARP10. This ability of viral macrodomains could indicate a mechanism of pathogenesis by counteracting antiviral activity of PARPs. This could make viral macrodomains good candidates as a potential drug target to combat pathogenesis.In addition to PARP10, PARP11 and PARP15 we also show ADP-ribosylation of RNA catalysed by TRPT1, an ancestral relative of PARP superfamily which is sometimes referred to as the 18th member of PARP family.

This gene is highly conserved in eukaryotic, archaeal and bacterial do- mains of life. While the human gene is known as TRPT1, the yeast and bacterial version of TRPT1 are referred to as Tpt1 and KptA, respectively. The yeast homologue has been characterized for its role in tRNA splicing acting as a NAD+ dependent 2r-phosphotransferase (50,60). The en- zymatic role of 2r-phosphate removal by Tpt1 occurs in a two chemical steps process – first, the 2r-phosphate reacts with NAD+ to release nicotinamide and form 2r-phospho- ADP-ribose RNA intermediate and second step involves generation of ADP-ribose 1rr-2rr cyclic phosphate and leav-ing behind the RNA with hydroxyl group at 2r end. While 2r-phosphotransferase activity is conserved across all di- verse homologues of TRPT1 (31) there is no evidence of intron containing tRNA (that would need TRPT1 activity for splicing) and/or pathway which would generate RNA with 2r-phosphate in most of the organisms except in plants and fungal species (31,61). Furthermore, TRPT1 knockout cells from mouse exhibit levels of tRNA splicing compara- ble to the wild type cells (61). Although some bacteria pos- sess introns in their tRNAs, they are self-splicing introns with very limited distribution to several representatives of proteobacteria and cyanobacteria (62). Therefore, the func- tional role of these widely conserved TRPT1 genes in other species remains elusive. A recent study has demonstrated that several archaeal species such as Aeropyrum pernix, Py- rococcus horikoshii and Archaeoglobus fulgidus and bac- terial Clostridium thermocellum possess Tpt1 protein that can ADP-ribosylate RNA at 5r-phosphorylated ends (31). In our study, we show that TRPT1 from a higher eukaryote (human) and from a bacterium (Streptomyces species) can also ADP-ribosylate 5r-phosphorylated RNA––revealing that RNA ADP-ribosylation activity is widespread among Trpt1 proteins.

To summarize, our results identify RNA as a novel tar- get of reversible ADP-ribosylation that can be catalysed by both PARP and TRPT1 classes of ARTs in vitro. This mod- ification of RNA occurs on phosphorylated terminal ends of RNA; it can be made by PARP10 and TRPT1 ARTs and reversed by several known ADP-ribosylhydrolases. Ef- ficient in vitro activities on RNA substrates by these en- zymes suggest that RNA ADP-ribosylation reactions could be relevant in vivo. We hypothesize that TRPT1/PARP10 could potentially mediate ADP-ribosylation signalling on RNA substrates as an on/off switch thereby controlling the functional state of RNA, protecting RNA ends or act as a platform for recruiting other proteins. In addition we also demonstrate other PARPs––PARP11 and PARP15 to ADP-ribosylate phosphorylated RNA ends, however fur- ther characterization is required to reveal the functional role of these OUL232 proteins.