ENPP1 processes protein ADP‐ribosylation in vitro

ADP‐ribosylation is a conserved post‐translational protein modification that plays a role in all major cellular processes, particularly DNA repair, transcription, translation, stress response and cell death. Hence, dysregulation of ADP‐ribosylation is linked to the physiopathology of several human diseases including cancers, diabetes and neurodegenerative disorders. Protein ADP‐ribosylation can be reversed by the macrodomain‐containing proteins PARG, TARG1, MacroD1 and MacroD2, which hydrolyse the ester bond known to link proteins to ADP‐ribose as well as consecutive ADP‐ribose subunits; targeting this bond can thus result in the complete removal of the protein modification or the conversion of poly(ADP‐ribose) to mono(ADP‐ribose). Recently, proteins containing the NUDIX domain – namely human NUDT16 and bacterial RppH – have been shown to process in vitro protein ADP‐ribosylation through an alternative mechanism, converting it into protein‐conjugated ribose‐5′‐phosphate (R5P, also known as pR). Though this protein modification was recently identified in mammalian tissues, its physiological relevance and the mechanism of generating protein phosphoribosylation are currently unknown. Here, we identified ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1) as the first known mammalian enzyme lacking a NUDIX domain to generate pR from ADP‐ribose on modified proteins in vitro. Thus, our data show that at least two enzyme families – Nudix and ENPP/NPP – are able to metabolize protein‐conjugated ADP‐ribose to pR in vitro, suggesting that pR exists and may be conserved from bacteria to mammals. We also demonstrate the utility of ENPP1 for converting protein‐conjugated mono(ADP‐ribose) and poly(ADP‐ribose) into mass spectrometry‐friendly pR tags, thus facilitating the identification of ADP‐ribosylation sites.

ADP-ribosylation is a conserved post-translational protein modification that plays a role in all major cellular processes, particularly DNA repair, transcription, translation, stress response and cell death. Hence, dysregulation of ADP-ribosylation is linked to the physiopathology of several human diseases including cancers, diabetes and neurodegenerative disorders. Protein ADP-ribosylation can be reversed by the macrodomaincontaining proteins PARG, TARG1, MacroD1 and MacroD2, which hydrolyse the ester bond known to link proteins to ADP-ribose as well as consecutive ADP-ribose subunits; targeting this bond can thus result in the complete removal of the protein modification or the conversion of poly (ADP-ribose) to mono(ADP-ribose). Recently, proteins containing the NUDIX domainnamely human NUDT16 and bacterial RppHhave been shown to process in vitro protein ADP-ribosylation through an alternative mechanism, converting it into protein-conjugated ribose-5 0 -phosphate (R5P, also known as pR). Though this protein modification was recently identified in mammalian tissues, its physiological relevance and the mechanism of generating protein phosphoribosylation are currently unknown. Here, we identified ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1) as the first known mammalian enzyme lacking a NUDIX domain to generate pR from ADP-ribose on modified proteins in vitro. Thus, our data show that at least two enzyme families -Nudix and ENPP/NPPare able to metabolize protein-conjugated ADP-ribose to pR in vitro, suggesting that pR exists and may be conserved from bacteria to mammals. We also demonstrate the utility of ENPP1 for converting protein-conjugated mono(ADP-ribose) and poly(ADP-ribose) into mass spectrometry-friendly pR tags, thus facilitating the identification of ADPribosylation sites.
Recently, we showed that protein ADP-ribosylation can undergo alternative processing [21,26,35]: rather than complete removal of the ADPr group, the modification can be cleaved down to ribose-5 0 -phosphate (also known as phosphoribose/pR or R5P), a potentially toxic modification for which it is unclear whether and how it can be removed (Fig. 1C) [36]. This reaction can be supported by phosphodiesterase I from Crotalus adamanteus (also referred to as snake venom phosphodiesterase, or SVP) [18,37] as well as the nucleoside diphosphate-linked moiety X (Nudix) family members human NUDT16 [26] and RppH from Escherichia coli [35] (Fig. 1A-C).
Here, using PARP1 and PARP10 as in vitro models, we have identified a new enzyme, ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1), belonging to the family of nucleotide pyrophosphatase/phosphodiesterase (NPP) proteins [38], which is able to process protein ADP-ribosylation to generate phosporibosylated proteins. This enzyme thus implicates a new class of proteins in the modulation of mammalian protein ADP-ribosylation and the production of phosphoribosylated protein substrates. Such activity is consistent with the observation of phosphoribosylated proteins in a recent reanalysis of a phosphoproteome dataset [21]. The identification of both acidic and basic protein ADP-ribosylation sites has recently been made possible by the development of a mass spectrometry (MS)-based method [19,37], which relies upon the conversion of protein PAR and MAR into pR by the enzymes SVP [20,37], NUDT16 [26] or RppH [35]. As a new member of this class of enzymes capable of converting protein-conjugated ADP-ribose to pR, we show here that ENPP1 can replace SVP for the identification of PARP1 and PARP10 automodification sites by MS.

NPP-type ectophosphodiesterases
Phosphoribosylation of cellular proteins has been detected recently [21]. The human Nudix family member NUDT16 has been suggested as an enzyme that produces this modification by acting on ADP-ribosylated proteins [26,35]. In the search for additional human enzymes able to process protein ADP-ribosylation into pR, we looked for potential human homologues of snake venom phosphodiesterase I from Crotalus adamanteus (also known as SVP) [19,20]. SVP belongs to the class of highly conserved nucleotide pyrophosphatase/phosphodiesterase (NPP)-type ectophosphodiesterases/extracellular glycoproteins [38]. These enzymes are known to hydrolyse diesters of phosphoric acid into phosphomonoesters and can be classified, according to the nature of their substrate, into nucleotide and lipid phosphodiesterases [39]. Currently, seven human genes are known to encode NPP protein homologues (ENPP1-7). All human ENPP proteins are unrelated to phospholipases [40], Nudix hydrolases [41] or ectonucleotide triphosphate diphosphohydrolases [42]. Among human NPP enzymes, only ENPP1, ENPP2 and ENPP3 share the same main domains of SVP in addition to the catalytic domain ( Fig. 2A,B). Indeed, the mammalian ecto-enzyme NPP2 (autotaxin) has a secretion motif in the N-terminal domain, similar to SVP, while NPP1 (PC-1) and NPP3 (B10; gp130RB13-6) are characterized by a short N-terminal intracellular domain and a single transmembrane domain (Fig. 2B). Except for these differences, all four enzymes share two somatomedin B-like domains, a catalytic domain and a C-terminal nuclease-like domain. C-terminal to the catalytic domain of NPP1-3 is the nuclease-like domain, structurally similar to the DNA-or RNA-nonspecific endonucleases [43]. However, this domain is probably not catalytically active Poly(ADP-ribosyl)ated protein  because none of the residues that are essential for catalysis by the nonspecific endoribonucleases is conserved in NPP1-3. Furthermore, the nuclease-like domain is likely to harbour isoform-specific determinants of catalysis because NPP2 with the nuclease-like domain of NPP1 is inactive [44]. Two somatomedin Blike domains seem required for protein interaction similar to the somatomedin B domain of vitronectin [39]. Members of the ENPP family hydrolyse various phosphodiester bonds (e.g. in oligonucleotides and artificial substrates like the p-nitrophenyl ester of TMP) and pyrophosphate bonds (e.g. in (d)NTP, (d)NDP, NAD, FAD, diadenosine polyphosphates and UDP sugars) and thereby generate nucleoside 5 0 -monophosphates. In particular, ENPP1 and ENPP3 show specificity to hydrolyse nucleotides [45]. Thus, among the three mammalian ENPP enzymes, ENPP1 and ENPP3 display catalytic properties similar to SVP [46,47]. However, ENPP1 shows much higher hydrolytic activity than ENPP3 in the hydrolysis of various phosphodiester bonds [43]. Activity against ADPr and its use in proteomics of ADP-ribosylated proteins has never been tested for mammalian ENPP proteins [19]. In order to assay activity of mammalian ENPP proteins against protein ADP-ribosylation, we focused on the well-characterized mouse ENPP1 (mENPP1), which has 79% identity with the human ENPP1 protein [48,49]. To purify soluble ENPP1 enzyme, the extracellular and catalytic region of mouse Enpp1 was fused with the secretory signal sequence and the N-terminal nine residues of the somatomedin B-like 1 (SMB1) domain of mouse ENPP2 at the N terminus and, with the addition of a TARGET tag at the C terminus, we generated a recombinant, secreted ENPP2-1 chimera amenable to enrichment via the TARGET tag system (mENPP2-1-T) (top panel Fig. 3A). The protein was then expressed in a clonal human cell line selected for higher activity against p-nitrophenyl ester and purified using a specific antibody capturing the TARGET tag and isolating recombinant protein from cell media [48,49] (left panel Fig. 3B). Of note, mENPP2-1-T was purified from HEK293S GnTI À cells, cells depleted of N-acetyl-glucosaminyltransferase I (GnTI) activity and therefore lack complex N-glycans [48,49]. We also produced an alternative construct for the expression and purification of ENPP2-1 that would enable more accessible and easier purification. For this, we generated and purified an ENPP2-1 chimera with a C terminus eight histidine tag (mENPP2-1-8xHis) from a pool of transiently transfected Expi293 TM cells (Bottom panel Fig. 3A, right panel Fig. 3B). Comparison of mENPP2-1-T and mENPP2-1-8xHis protein preparations on SDS/PAGE (left panel Fig. 3C) showed a main band with a molecular weight of~90 kDa. However, mENPP2-1-8xHis showed an additional high molecular weight contaminant protein. Anti-6xHis western blot confirmed the presence of His-tagged mENPP2-1 protein at~90 kDa (right panel Fig. 3C). To better investigate the glycosylation pattern in mENPP2-1-T and mENPP2-1-8xHis protein preparations, we performed deglycosylation assays treating both proteins with PNGase F (which removes almost all types of N-linked glycosylation: high mannose, hybrid, bi-, tri-and tetra-antennary) and Endo H (which removes only high mannose and some hybrid types of N-linked carbohydrates) (Fig. 3D). As expected, Coomassie staining revealed that mENPP2-1-T presents a homogeneous glycosylation pattern that can be reverted almost completely by PNGase F treatment (black stars, Fig. 3D). By contrast, mENPP2-1-8xHis showed a complex glycosylation pattern as neither PNGase F nor Endo H was able to completely compact the protein to a single band at a lower molecular weight (Fig. 3D). To test the activity of ENPP1 from both constructs against ADP-ribosylated proteins, we incubated serial dilutions of purified mENPP1 proteins and NUDT16a positive controlwith automodified PARP1 protein as a substrate for 3 h in the presence of 15 mM MgCl 2 . PAR was visualized by western blot using an antibody specifically recognizing poly-and oligochains of ADPr but not MAR ( Fig. 4A-C). Our data showed that ENPP1 successfully removes PAR chains from PARP1 protein with activity comparable to that exhibited by NUDT16, though mENPP2-1-T seems to be slightly more active than mENPP2-1-8xHis (Fig. 4A,B). Comparison of mENPP2-1-T and NUDT16 activity showed notably higher activity of NUDT16 against PARylated PARP1 (

ENPP1 converts protein ADP-ribosylation to protein phosphoribosylation
Hydrolysis assays were again performed, this time using a more sensitive assay in which PARP1 is automodified in the presence of NAD + labelled on the alpha phosphate group with 32 P (Fig. 5A). PARylated PARP1 was then incubated with mENPP2-1-T, PARG (an enzyme known for removing PAR but leaving MAR) or the positive controls SVP and NUDT16 (Fig. 5A,B). As expected, ENPP1 was able to completely remove the radiolabelled PAR signal from PARP1, mirroring the phosphodiester hydrolysis activity of NUDT16 and SVP and failing to leave MAR at the attachment site, as PARG does (Fig. 5A,B). The main product of this reaction was phosphoribosyl AMP (PRAMP) as seen for SVP and NUDT16 previously [19,26,37] (Fig. 5A), [thin layer chromatography (TLC)] (bottom panel Fig. 5B).
In order to directly address whether ENPP1 is able to remove MAR from proteins, recombinant PARP1-E988Q, a PARP1 mutant able to add only a single unit of ADPr onto target proteins (in this case, itself), and a mono-ADP-ribosyl transferase GST-PARP10 catalytic domain, were automodified and then incubated with NUDT16, mENPP2-1-T, PARG or SVP (Fig. 5C). mENPP2-1-T removed the 32 P-labelled MAR signal as efficiently as SVP and NUDT16, thus proving that mENPP2-1-T is active against proteinconjugated MAR. The TLC analysis showed that, as expected, the main reaction product is AMP (Fig. 5D). Taken together, we conclude that ENPP removes PARP-dependent protein ADP-ribosylation.

ENPP1 efficiently removes ARTC2.2-mediated ADP-ribosylation
As mentioned before, ENPP1 is an ecto-enzyme initially described as plasma cell membrane glycoprotein 1 (PC-1, CD203); however, its involvement in lymphocyte biology has not been studied thoroughly [30,38,44,49]. Localization of ENPP1 to the cell surface suggests that ENPP1 may process ADP-ribosylation synthesised by extracellular ADP-ribosyl transferases such as membrane-bound cholera-like ecto-ADP-ribosyltransferases (ARTCs) [7][8][9]. Unlike PARPs, these enzymes usually modify protein arginine residues [6,7]. In particular, we focused on the ARTC2.2 protein expressed as an ecto-enzyme on the plasma membrane of mouse immune cells [50]. ARTC2.2 is known to have a crucial regulatory function on the activity and survival of T lymphocytes and NK lymphocytes, ADP-ribosylating several proteins, such as the purinergic P2X7 receptor, LFA-1 and CD8 [50]. To address the question of whether ENPP1 can remove ARTC2.2-dependent ADP-ribosylation in a cellular context, we expressed and purified the catalytic domain of mouse ARTC2.2 (mARTC2.2) in and from E. coli and then in vitro incubated the ART recombinant protein with K562 (human NK lymphocyte) cell extract in the presence of radiolabelled NAD + (Fig. 6A). Under these conditions, mARTC2.2 was able to reproducibly modify several proteins in the K562 cell extract. Concomitant incubation of mARTC2.2 modified extract with NUDT16 showed significant reduction in radiolabelled ADP-ribosylated proteins; however, incubation of cellular proteins with the same concentrations of mENPP2-1-T gave rise to an even more remarkable hydrolysis of mono-ADPribosylated proteins. Of note, NUDT16 was itself modified by mARTC2.2. Our data show that ENPP1 can better perform conversion of ADP-ribosylated proteins into pR-proteins than NUDT16 in this cellfree system and reaction conditions. Taken together, our data demonstrate that ENPP1 is able to remove ARTC-mediated ADP-ribosylation in vitro, and suggest that this event may occur in vivo due to the extracellular proximity of ARTCs and ENPPs.

ENPP1 as a tool for LC-MS/MS aided identification of protein ADP-ribosylation sites
In order to determine whether ENPP2-1 can serve as a replacement for SVP in the recently established ADPr site identification proteomics pipeline [19,35], we treated 60 pmoles of autoPARylated PARP1 protein with 120 pmoles of SVP, mENPP2-1-T or mENPP2-1-8xHis, or 600 pmoles of mENPP2-1-T or mENPP2-1- 8xHis and used LC-MS/MS to search for the 212.01 Da shift characteristic of pR-modified peptides. Ten PARP1 peptides confidently presented with an MS1 mass shift corresponding to a singly or doubly phosphoribosylated state combined with peptide sequencing by MS2 to determine site localization ( Table 1). As shown in Fig. 7, panels A and B, the ambiguity in site localization as reported by Max-Quant can often be overcome by de novo sequencing of the spectrum of interest, shown here for the first two forms of peptide 3, where pR is carried on E168 (panel A) and E169 (panel B). A direct comparison between the peptide forms identified following SVP, mENPP-2-1-T or mENPP-2-1-8xHis digestion is depicted in the '29' columns of Table 1, wherein the enzymes were incubated with PARylated PARP1 in a 2 : 1 molar ratio (note that this ratio is for enzyme: protein, not enzyme:substrate, where the substrate would be ADPr). Of the 19 peptide forms identified in this analysis, 15 were found following exposure to SVP, 16 following mENPP-2-1-T and 11 following mENPP-2-1-8xHis, demonstrating the comparability of these three enzymes in this application. To determine whether these reactions can be driven to completion by the addition of excess enzyme, the ENPP1 proteins were added in a 10 : 1 enzyme/PARylated protein ratio (see Table 1, columns labelled 109), resulting in 19/19 peptide forms being identified in the mENPP-2-1-T sample and 9/19 peptide forms showing up in the mENPP-2-1-8xHis sample. This result suggests that, at least in the case of the highly active form of ENPP1 (mENPP-2-1-T), the transformation of PAR to pR can be driven to completion by the addition of more recombinant enzyme. In order to confirm that ENPP1 is able to convert protein-conjugated mono(ADP-ribose) to a pR tag, we automodified the catalytic domain of PARP10an enzyme restricted to mono(ADP-ribosyl)ation activity and exposed it to mENPP-2-1-T. This resulted in the confident identification of four pR-containing PARP10 peptides after ENPP1 treatment (Table 2). K916, a residue previously identified as a PARP10 automodification site [18] and also known to be acetylated [51], was among the identified mono(ADP-ribosyl)ated sites.

Discussion
ENPP1 is a type II transmembrane glycoprotein with nucleotide phosphodiesterase activity [38,39]. This protein has broad specificity and cleaves a variety of substrates, including phosphodiester bonds of nucleotides and nucleotide sugars, and pyrophosphate bonds of nucleotides and nucleotide sugars. This protein may function to hydrolyse nucleoside 5 0 triphosphates to their corresponding monophosphates and may also hydrolyse diadenosine polyphosphates [38,39,46,47].
Mutations in the ENPP1 gene have been associated with 'idiopathic' infantile arterial calcification [52], ossification of the posterior longitudinal ligament of the spine OPLL [53], hearing loss [54] and insulin resistance [49]. ENPP1 has been described as essential for physiological mineralization because its expression on the outer surfaces of mineralizing cells, such as osteoblasts and chondrocytes, regulates the balance between the extracellular concentrations of inorganic phosphate (Pi), a substrate for mineralization, and inorganic pyrophosphate (PPi), an inhibitor of mineralization [55]. Although ENPP1 is essential for the regulation of physiological mineralization, its substrate specificity for different nucleotides is not completely described.
While aligning the nucleotide sequences of the ectonucleotide pyrophosphatase/phosphodiesterase enzymes with SVP, we hypothesized that ENPP1 could act in a similar manner as SVP to convert ADPribosylated proteins into proteins containing a pR tag. In addition to the wide range of intracellular roles played by PAR, its recently observed presence in the extracellular matrix (ECM) suggests an extracellular function for this polymer [56]. In particular, PAR was identified in the ECM of developing bones, where it was likely released from necrotic osteoblasts and playing an essential role as a scaffold for bio-mineralization [56].
Our biochemical data suggest that ENPP1 may be involved in the metabolism of extracellular PAR through its hydrolysis and release of phosphoribosylated proteins, PRAMP and, contextually, free PARa molecule that has previously been described as an extracellular stimulus driving inflammatory signalling [57]. Moreover, colocalization of ARTCs and ENPP1 on the extracellular membrane [50] may suggest a physiological function for ENPP1 in the regulation of immune cell activity and survival.
The advent of mass spectrometry-based methods for identifying ADP-ribosylation sites has allowed for large-scale assessments of the endogenously ADPribosylated proteome at the protein sequence level in human and murine cells [19,58]. All of the available methods for PAR site identification rely upon the creation of a molecular 'tag' at the PTM's amino acid attachment site [20], in the case of SVP digestion this tag is pR. SVP, however, must be purified from snake venom through a multi-step process, resulting in a high prep-to-prep variation due to both source variability (at the level of the snake venom) and inevitable  inconsistencies between protein purification pipelines (at the level of instrumentation, reagents, etc.) [26,35]. Therefore, in order to broadly implement the pR pipeline for proteomics applications, it is necessary to replace SVP with a phosphodiesterase that can be made in large quantities and with high consistency (as a recombinant protein, expressed and purified in the laboratory). We have presented two options for this purpose: mENPP2-1-T and mENPP2-1-8xHis. Of these two, the former (mENPP2-1-T) has proven to be  a more active form of this enzyme, while the latter (mENPP2-1-8xHis) can more easily be adapted for expression and purification in standard laboratories equipped to purify His-tagged proteins. The availability of these tools will allow for greater adoption and implementation of the pR-based proteomics pipeline for the unbiased identification of protein mono and poly(ADP-ribosyl)ation sites. Table 1. ENPP1 as a tool for LC-MS/MS aided identification of protein ADP-ribosylation sites. PARP1 automodification sites identified using the pipeline described by Daniels et al. [19]. Experimental conditions were described in the main text. Posterior error probabilities (PEPs) are reported for corresponding amino acid residues as well as site localization probabilities within the identified pR-containing peptides. The 29 and 109 columns refer to the reaction ratio of enzyme to PARylated PARP1. Key at the bottom describes the confidence ranges reported by MaxQuant for site identification (with possible modification sites of D, E, K and R); often this ambiguity can be resolved by manual inspection of the peptide fragmentation pattern (see Fig. 7 for annotation of two different forms of peptide N166-K182).

Materials and methods
Plasmids and recombinant proteins PARP1 wild-type protein was purchased from Trevigen Inc., (Gaithersburg, MD, USA) (high specific activity) for biochemical assays or expressed and purified as previously described [19] for analysis by mass spectrometry. PARP1-E988Q was expressed from pET28a(+) and purified as previously described [30]. NUDT16 was expressed from pNIC28-Bsa4 and purified as described in Palazzo et al. [26]. pGEX-4T1 GST-PARP10cd (amino acids 818-1025) plasmid was a gift from Bernhard L€ uscher (RWTH Aachen University) [17]. GST [48,49]. In particular, the extracellular region of mouse Enpp1 (residues 92-905) was fused with the secretory signal sequence (residues 1-50) and the N-terminal nine residues of the somatomedin B-like 1 (SMB1) domain (residues 51-59) of mouse Enpp2 at the N terminus and with the TARGET tag at the C terminus to generate a secreted ENNP2-1 chimera (mENPP2-1-T) [48,49]. Of note, HEK293S GnT1 À were used to express mENPP2-1-T, as described in Kato et al. [48,49]. To produce mENPP2-1-8xHis (ENPP1 -OPPF 17442), the full-length gene for ENPP2-1 containing the native signal sequence (bp 1-2619) was amplified by PCR using Phusion Flash polymerase (Life Technologies, Thermo Fisher Scientific, Hemel Hempstead, UK) and the following primers containing the extensions necessary for ligation-independent cloning (Forward primer: aggagatataccatgATGGCAAGACAAGGCTGTTTCGG and reverse primer:gtgatggtgatgtttGTCTTCTTGGCTGAAGAT TGGCAAATGT). The PCR product was cloned into pOPI-NEneo using the InFusion method of ligation-independent cloning as previously reported [59].  reagent and incubated for 20 min at room temperature before adding to 27 mL of cultured Expi293 TM cells. After 18-20 h, 150 lL of ExpiFectamine TM transfection Enhancer 1 and 1.5 mL of ExpiFectamine Transfection TM Enhancer 2 were added to the transfected cells. The supernatant containing the mENPP2-1-8xHis protein was harvested after 96 h. Secreted protein was purified by automated immobilized metal affinity chromatography followed by gel filtration chromatography using the method of Nettleship et al. [60,61]. Briefly, 200 mL of sample was loaded onto a 5 mL HisTrap FF column (GE Healthcare) before washing with 50 mL of 50 mM Tris, pH 7.5, 500 mM NaCl, 30 mM imidazole. This was then repeated until all the samples were loaded. Elution from the HisTrap FF column was of 50 mM Tris, pH 7.5, 500 mM NaCl, 500 mM imidazole and the eluted sample was injected directly onto a HiLoad 16/600 Superdex 200 column. Size exclusion chromatography was performed using 20 mM Tris pH 7.5, 200 mM NaCl and the fractions analysed by SDS/PAGE. The fractions containing the ENPP1 protein were concentrated to 2.1 mgÁmL À1 , based on an extinction coefficient of 1Au = 1 mgÁmL À1 (0.63 mg final yield) before use. pASK60-OmpA-mARTC2.2 6xHis-Flag tag was a gift from Friedrich Koch-Nolte (Universit€ atsklinikum Hamburg-Eppendorf) and purified as previously described [62].

Purification of snake venom phosphodiesterase
Phosphodiesterase I (SVP) from Crotalus adamanteus venom was purified as previously described [19,26]. Briefly, 2.52 mg dried weight of partially purified SVP (Worthington Biochemical Corporation, Lakewood, NJ, USA) was dissolved in 1 mL of loading buffer (10 mM Tris/HCl pH 7.5, 50 mM NaCl, 10% glycerol) and loaded onto a pre-equilibrated 1 mL HiTrap blue HP (GE Healthcare). The column was washed with five column volumes (CV) of loading buffer followed by an increasing gradient of KPO 4 pH 8.0 up to 150 mM. The SVP protein was eluted using 1 M KPO 4 buffer. Desired fractions were pooled and loaded onto analytical size-exclusion chromatography Superdex 200 (GE Healthcare) using € AKTA pure (GE Healthcare) in a buffer composed of 10 mM Tris pH 8.0, 50 mM NaCl, 15 mM MgCl 2 and 1% glycerol. Concentration of desired fractions (~97 kDa molecular weight) was measured using Nanodrop (Thermo Fisher Scientific) and stored at À80°C.

Hydrolytic activity assays on ADP-ribosylated proteins
PARylated and MARylated PARP1 proteins were prepared as described [26] in a reaction buffer containing 50 mM Tris/HCl (pH 8.0), 4 mM MgCl 2 , 50 mM NaCl, 0.2 mM DTT, 200 lM NAD+ (Trevigen) and 130 ng activated DNA (BPS Bioscience, Inc., San Diego, CA, USA). Briefly, for the PAR hydrolysis activity assays, 70 nM PARP1 (PARP1-HSA; Trevigen) was automodified as described in [26]. After 20-min incubation, PARP1 was passed three times through SpinTrap G-25 (GE Healthcare). PARylated PARP1 substrate was used in a 10-lL reaction. For the MARylated PARP1, 1 lM of PARP1-E988Q was used as a substrate. Reactions were stopped by the addition of PARP inhibitor Olaparib (1 lM). The MgCl 2 (Sigma) concentration was adjusted to 15 mM to allow full hydrolase activity. Automodified PARP1 was then incubated for indicated times at 30°C with hydrolytic enzymes in 10-lL reaction. Concentrations of hydrolytic enzymes used are as indicated in figures. Reactions were stopped by addition of Laemmli loading buffer, samples boiled at 90°C for 1.5 min and analysed by NuPAGE Novex Bis-Tris 4-12% gel using MOPS buffer (Invitrogen). Radiolabelled experiments were visualized by autoradiography.

Thin layer chromatography
The TLC was performed as previously described [26]. PARylated PARP1 and MARylated PARP1-E988Q/GST-PARP10cd proteins were automodified in the presence of [ 32 P]-labelled NAD + as described above. The product of this reaction was then cleaned up by G25 desalting columns, MgCl 2 was added to a final concentration of 15 mM and 10-lL reaction samples were processed by NUDT16, ENPP1, PARG and SVP, as described above. About 1 lL of reaction was spotted onto polyethyleneimine (PEI)cellulose plates (Macherey-Nagel, Polygram CEL 300 PEI/ UV254) and developed in 0.15 M LiCl and 0.15 M formic acid. Dried plates were exposed on X-ray film or visualized by UV254 shadowing.

Immunoblotting
Fractionated proteins on gradient gels were transferred onto nitrocellulose membranes using Trans-Blot Turbo Transfer System (Biorad) at 1.3 A/25 V for 20 min. Membranes were blocked in 5% nonfat dry milk (NFDM; Bio-Rad Laboratories Ltd., Hemel Hempstead, UK) diluted in 0.1% Tween 20-PBS and subsequently incubated with rabbit polyclonal anti-PAR (1 : 2000; Trevigen) and mouse monoclonal anti-6xHis (1 : 4000; Clontech). Primary antibody incubation was then followed by incubation with secondary antibody as indicated and developed with ECL western blotting detection reagent (GE Healthcare).

Deglycosylation assay
PNGase F and Endo H enzymes were purchased from New England BioLabs and reactions were performed according to the manufacturer's protocols under denaturant conditions using 1 lg of recombinant substrate.