Nudix hydrolases degrade protein-conjugated ADP-ribose

ADP-ribosylation refers to the transfer of the ADP-ribose group from NAD+ to target proteins post-translationally, either attached singly as mono(ADP-ribose) (MAR) or in polymeric chains as poly(ADP-ribose) (PAR). Though ADP-ribosylation is therapeutically important, investigation of this protein modification has been limited by a lack of proteomic tools for site identification. Recent work has demonstrated the potential of a tag-based pipeline in which MAR/PAR is hydrolyzed down to phosphoribose, leaving a 212 Dalton tag at the modification site. While the pipeline has been proven effective by multiple groups, a barrier to application has become evident: the enzyme used to transform MAR/PAR into phosphoribose must be purified from the rattlesnake Crotalus adamanteus venom, which is contaminated with proteases detrimental for proteomic applications. Here, we outline the steps necessary to purify snake venom phosphodiesterase I (SVP) and describe two alternatives to SVP—the bacterial Nudix hydrolase EcRppH and human HsNudT16. Importantly, expression and purification schemes for these Nudix enzymes have already been proven, with high-quality yields easily attainable. We demonstrate their utility in identifying ADP-ribosylation sites on Poly(ADP-ribose) Polymerase 1 (PARP1) with mass spectrometry and discuss a structure-based rationale for this Nudix subclass in degrading protein-conjugated ADP-ribose, including both MAR and PAR.

Assessment of contaminating proteolysis activity in SVP prep. For whole cell lysate, 1 mg of proteins from HeLa whole cell lysate was denatured in 8M Urea (Sigma-Aldrich), 50 mM Tris-HCl pH 7.0 for 10 minutes at 37 °C before being reduced in 1 mM Tris(2-carboxyethyl)phosphine (Life Technologies) for 10 minutes and then alkylated in 2 mM 2-chloroacetamide (Sigma-Aldrich) for 10 minutes in the dark. Samples were then diluted to a final concentration of 1M Urea, 50 mM NaCl (Sigma-Aldrich), 15 mM MgCl 2 (Quality Biological) and 0.2M Tris-HCl pH 7.3. 5 μ g of SVP was added to each sample and incubated for 2 hours at 37 °C. Samples were run on an in-house 6-10% SDS-PAGE gel and transferred to a nitrocellulose membrane. Total protein was visualized by ProAct membrane stain (Amresco) per the manufacturer's instructions. For purified recombinant His-PARP1, 1 μ g of His-PARP1 was automodified as previously published 5 and then switched into the same buffer used for the whole cell lysate (described above) with or without 1M urea. 500 ng of SVP was used for each μ g of PARP1 and digestion was carried out at 37 °C for 2 hours. Samples were run on in-house 6-10% SDS-PAGE gels and total protein was visualized by SimplyBlue Safe Stain (Life Technologies) per the manufacturer's instructions. 32 P-labeled PAR was visualized on a phosphor-screen (GE, BAS-III 2040) followed by imaging on a Typhoon FLA7000 (GE Healthcare Life Sciences).

Expression and purification of Nudix hydrolases. The expression and purification have been previously
published for the following hydrolases: E. coli NudF (EcNudF/ADPRase/UniProtKB Q93K97) 18 , E. coli NudE (EcNudE/ORF186/UniProtKB J7QI36) 32 , Bdellovibrio bacteriovorus (Bd3179/BdNDPSase/UniProtKB Q6MIH8), Deinococcus radiodurans Dr1184 (Dr1184/ UniProtKB Q9RV46) 33 , E.coli RppH (EcRppH/ORF176/UniProtKB P0A776 34 , Agrobacterium tumefaciens ORF147 (AtORF147/UniProtKB Q7CX66) and Pseudomonas aeruginosa Pa3470 (Pa3470/UniProtKB Q9HYD6 ) 35 . Homo sapiens NudT16 (HsNudT16/UniProtKB Q96DE0) was expressed and purified as described in the methods from the Structural Genomics Consortium (http://www.thesgc.org/ structures/3cou). For EcRppH, the last step of purification yielded ≥ 90% homogeneity before being loaded onto a HiLoad 26/60 Superdex 200 prep grade gel filtration column (GE Healthcare), equilibrated in gel filtration buffer (50 mM Tris-HCl pH 8.5, 300 mM NaCl); fractions containing ≥ 95% homogeneity, as determined by SDS-PAGE, were pooled. EcYfcD was purified using the method described for EcADPRase 18 . Selected known single domain Nudix enzymes were structurally aligned using SSM 36 and rendered using PyMOL 38 . PAR was modeled in the active site using a 'template guide' , i.e. the mRNA present in the structure of the complex of EcRppH (PDB ID 4S2X) 39 and the IMP present in the structure of HsNudT16 (PDB ID 2XSQ) 40 . Models of the complex structures were built using the selected structures and the PAR. Modeling steps were performed in MOE 38 . Hydrogen atoms and charges were added automatically to the protein. The binding site was defined as all protein atoms within 4.5 angstrom of any atom of bound ligand. A steepest descent energy minimization was used with a MMFF94x forcefield.

Structural analysis of Selected
Automodification of WT and E988Q PARP1. Automodification was performed as published previously 5 with the following changes: both WT and E988Q PARP1 were incubated with 0.6 μ M (1.85 kBq/μ L, 37 kBq/sample) Figure 1. Purification of snake venom phosphodiesterase I for the digestion of protein-conjugated PAR. SVP was affinity purified on a blue sepharose column (boxed fractions in panel (a) producing a proteasecontaminated product; (b and c), proteolysis is evident by the loss of protein bands following exposure to SVP). The addition of SVP results in protein degradation within lysates (b) and of purified PARP1 (c). The rightmost section in panel b indicates the size of SVP from the pooled fraction from (a). For (b) 1 mg of whole cell lysate was exposed to 50 pmoles of SVP for 2 hrs at 37 °C, pH 7. For (c) 10 pmoles of PARP1 was exposed to 5 pmoles of SVP for 2 hrs at 37 °C, pH 7. Size exclusion chromatography (d,e) results in a purified SVP fraction (f) that removed the contaminating protease activity (g). Top sections in panels c and g show the autoradiographs of PARylated PARP1 and the bottom sections show the corresponding coomassie staining of the gels. Reaction conditions for (g) were identical to those of (c). ***SDS-PAGE well, **SDS-PAGE interface between stacking and resolving gels, *native size of PARP1; rep = replicate.
Scientific RepoRts | 5:18271 | DOI: 10.1038/srep18271 32 P-NAD + for 10 minutes at room temperature, following which WT PARP1 was incubated with 1 mM NAD + (non-radioactive) for 10 minutes at room temperature to allow for polymer elongation.
Hydrolysis of protein-conjugated ADPr to phosphoribose. For comparative analyses by SDS PAGE: 5 pmoles of 6xHis-PARP1 WT or E988Q mutant were exposed to a hydrolase (various enzymes and amounts) in hydrolysis buffer (50 mM Tris-HCl pH 7.0 (Thermo Scientific), 150 mM NaCl (Sigma-Aldrich), 15 mM MgCl 2 (Quality Biological), 1 mM 3-aminobenzamide (Sigma Aldrich)) for two hours at 37 °C. P-labeled mono-or poly(ADP-ribose) conjugated to PARP1 following exposure to canonical nucleotide sugar hydrolases: EcNudF, EcYfcD and EcNudE and Bd3179. For (a), 5 pmoles of PARylated PARP1 was exposed to 10 pmoles of hydrolase for 2 hrs at 37 °C; for (b), 100 pmoles of hydrolases were utilized, while in (c) 5 pmoles of PARP1 E988Q, a mutant only capable of synthesizing mono(ADP-ribose), was exposed to 100 pmoles of hydrolases for 2 hrs at 37 °C. Ribbon diagrams show the structure of each of the enzymes used: (d) EcNudF/ ADPrase (PDB ID 1KHZ) 18 , (e) EcYfcD (PDB ID 2FKB), (f) EcNudE (PDB ID 1VHG) and (g) Bd3179 (PDB ID 5C7Q) 52 . Panel (h) shows a surface representation of EcNudF with modeled PAR polymer based on the complex with a nonhydrolyzable ADPr 48 . Panel (i) shows a surface representation of EcNudF with the modeled PAR polymer, the inset shows how the terminal ribose is protected and buried within the enzyme. The white arrowhead indicates where protein conjugation would occur on ADPr. ***SDS-PAGE well, **SDS-PAGE interface between stacking and resolving gels, *native size of PARP1, ^co-purified PARP1 protein fragments.
For comparative analyses by liquid chromatography-tandem mass spectrometry (LC-MS/MS): 60 pmoles of 6xHis-PARP1 was exposed to 120 pmoles of SVP or 3 nmoles of Nudix hydrolase (EcRppH or HsNudT16) in hydrolysis buffer for two hours at 37 °C. Phosphoenrichment of phosphoribosylated peptides. The ion metal affinity chromatography (IMAC) method was performed as described previously 5 .

LC-MS/MS analysis of phosphoribosylated peptides. Peptides were separated on a Thermo-Dionex
RSLCNano UHPLC instrument with ~10 cm x 75 micron ID fused silica capillary columns with ~10 micron tip opening made in-house with a laser puller (Sutter) and packed with 3 micron reversed phase C18 beads (Reprosil-C18.aq, 120 Angstroms, Dr. Maisch). We applied a 90 min gradient of 10-35% B at 200 nL/min using 0.1% acetic acid as solvent A and solvent B of 0.1% acetic acid, 99.9% acetonitrile. MS data was collected with a Thermo Orbitrap Elite. Data-dependent analysis was applied using Top 5 selection and fragmentation was induced by CID and HCD. Profile mode data was collected in all scans. MS analysis parameters were as previously described 5 .
Database search of MS/MS spectra for peptide and protein identification. Raw files were analyzed by MaxQuant version 1.5.3.8 using protein, peptide and site FDRs of 0.01 and a score minimum of 40 for modified peptides, 0 for unmodified peptides; delta score minimum of 17 for modified peptides, 0 for unmodified peptides. Sequences were searched against an in-house database containing the proteins of interest as well as UniProt Escherichia coli BL21 DE3 database (definitions updated October 15 th , 2014). MaxQuant search parameters: Variable modifications included Oxidation (M), Acetylation (Protein N-term), carbamidomethyl (C), phosphorylation (STY) and phosphoribosylation (DEKRC). Max labeled amino acids was 3, max missed cleavages was 2, enzyme was Trypsin/P, and max charge was 7.

Purification of Snake Venom Phosphodiesterase I from Crotalus adamanteus involves both affinity purification and size exclusion chromatography. Snake venom phosphodiesterase I (SVP) from
Crotalus adamanteus was shown to degrade PAR nearly 50 years ago 41 and has since proven a valuable tool for the degradation of PAR into its linear, branching and terminal subunits, a technique that yields quantitative information regarding the molecular structure of the intact polymer 42,43 . The utility of this enzyme, however, is greatly determined by the purification scheme employed to isolate it from the large number of proteases as well as phosphatases and nucleotidases present in the C. adamanteus venom 44 . Oka et al. successfully isolated the phosphodiesterase activity of commercially available SVP away from the contaminating phosphatase and 5′ -nucleotidase activity through affinity purification using blue sepharose, a molecule which mimics NAD + and therefore interacts with the active domain of SVP 45 . The results from a simplified version of this method used by our group is shown in Fig. 1a, where 150 mM potassium phosphate pH 7.5 is used as a single step elution off a blue sepharose column. This purification scheme paved the way for development of the quantitative method mentioned above to determine the structure of the intact polymer, but did not address the need to eliminate contaminating protease activity. This protease activity can be problematic when using SVP to hydrolyze protein-conjugated ADPr, either MAR or PAR, for the purpose of creating a phosphoribose 'tag' at the otherwise ADP-ribosylated amino acid residue 5,6,12 . Such protease activity is demonstrated in Fig. 1b wherein a complex mixture of proteins is exposed to blue sepharose purified SVP, resulting in the degradation of the target proteins and the appearance of SVP along with its co-purified proteins. This proteolytic activity is further shown against purified, 32 P-PARylated PARP1 (both native and    Fig. 1c. In order to separate the 115 kD SVP from the major contaminating proteins (< 30 kD), we subjected the blue sepharose purified product to size exclusion chromatography, yielding a simple mixture of what we hypothesize to be the various glycolytic forms of SVP, based on the knowledge that most secreted proteins are glycosylated 46 , including those found in snake venom 47 (Fig. 1d-f). When tested against 32 P-PARylated PARP1 as in Fig. 1c, this highly pure form of SVP displayed phosphodiesterase activity without apparent proteolytic activity (Fig. 1g). Similar results were seen against whole cell lysate, allowing for use of this enzyme for ADP-ribosylation site identification by mass spectrometry 5 .
While the pipeline presented here is an effective method for isolating SVP from snake venom, we believe the complexity of the purification scheme, along with lot-to-lot variability observed from commercial sources of SVP which serve as the input for this purification (data not shown), stands to be greatly improved by the availability of a recombinant enzyme which could be reliably expressed, purified and scaled to meet the often large material demands of proteomic pipelines. For this reason, we went on to compare the activity of SVP with that of relatively small, stable and well-characterized Nudix hydrolases, which we hypothesized could hydrolyze protein-conjugated ADPr, either MAR or PAR, to phosphoribose with similar specificity. Nudix ADPrases do not hydrolyze protein-conjugated ADPr. Nudix ADPrases are responsible for the breakdown of free ADPr into its phosphoribose and adenosine monophosphate subunits, thus modulating the levels of free ADPr 18,48 . This knowledge lead us to test a group of Nudix ADPrases for hydrolase activity against protein-conjugated MAR and PAR: PARP1, an enzyme known to autoPARylate in the presence of NAD + , was exposed to 32 P-labeled NAD + producing either 32 P-labeled PARylated (on WT PARP1) or MARylated (on the catalytically deficient PARP1 E988Q mutant) proteins to serve as substrates for hydrolysis by candidate Nudix enzymes or the positive control, SVP (Fig. 2a-c). As Fig. 2b shows, the various Nudix enzymes tested did not significantly hydrolyze ADPr from its conjugated proteins when compared to the positive control SVP. From a structural perspective, the lack of activity towards ADPr could be explained by the dimeric structure of ADPrases, where each dimer is formed by monomers of an N-terminal β -sheet domain and a C-terminal Nudix domain ( Fig. 2d-g). The N-terminal domains are swapped, creating two active sites where both monomers contribute to substrate recognition (each active site is composed of the N-terminal β -sheet of one monomer and the C-terminal Nudix domain of the other monomer). As shown in Fig. 2h,i, ADPr is nested in the active site of the ADPRase EcNudF so that the 1′ -hydroxyl of the terminal ribose group is completely buried by the protein dimer (white arrowhead in Fig. 2i), preventing conjugation to another ADPr group (or a protein). This explanation could likely be extended to the other three nucleoside sugar hydrolases tested in this study as they display the same quaternary arrangement and have a high structural homology with a pairwise root mean square deviation ranging from 0.9 to 2.0 Å (Fig. 2d- Single domain Nudix hydrolases are capable of hydrolyzing protein-conjugated ADPr. In order to consider Nudix enzymes with active sites more open to fit the target ADPr group bound to either a PAR polymer or protein, we turned to Nudix enzymes representative of families which are not swapped dimers and lack N-or C-terminal domain insertions, hypothesizing that enzymes with just the Nudix fold would have a more open active site. We chose four Nudix enzymes known to be monomeric by gel filtration 35 (e.g. RppH as shown in Fig. 3). Two of them (Dr1184/CoAse from Deinococcus radiodurans and EcRppH from Escherichia coli) degraded the 32 P-labeled PAR on the model protein PARP1 (Fig. 4a), while only EcRppH showed slight activity against 32 P-labeled MAR at the tested concentration (Fig. 4b). Structural analysis and modeling revealed that the active site within these enzymes could accommodate protein-conjugated ADPr (Figs. 4c,d), as opposed to the dimeric ADPrases (c.f. Fig. 2d-g).
Both EcRppH and HsNudT16 can degrade protein-conjugated PAR and MAR to a phosphoribose tag for mass spectrometry. A recent study by Palazzo et al. 13  Nudix (deoxy)inosine diphosphatase 49 which is also known to decap small nucleolar RNAs 50 as well as cytoplasmic mRNAs 51 , has the ability to degrade protein-conjugated ADPr. As shown in Fig. 5a-f HsNudT16 has a high structural similarity to both EcRppH and Dr1184 (which showed activity against protein-conjugated ADPr, see Fig. 4a,b) and also possesses an open active site which would allow for the target ADPr to be conjugated to a protein or additional ADPr unit(s). Based on these similarities, we postulated that the activity against protein-conjugated ADPr would be comparable for all three Nudix enzymes. To test this, 32 P-PARylated or MARylated PARP1 was exposed to increasing amounts of SVP, EcRppH, Dr1184 or HsNudT16. As shown via autoradiographs in Fig. 5g,h, both HsNudT16 and EcRppH are able to hydrolyze both protein-conjugated PAR and MAR.
To validate that EcRppH and HsNudT16 are degrading protein-conjugated ADPr down to its phosphoribose attachment site, we treated 60 pmoles of automodified WT PARP1 with 120 pmoles of SVP, 3 nmoles of EcRppH or 3 nmoles of HsNudT16 (either 2x or 50x molar excess over PARP1, as per the activity assays shown in Fig. 5) before digesting the proteins to peptides and subjecting them to phosphoenrichment on an IMAC matrix 5 . As shown in Table 1, phosphoribosylation sites were confidently identified by mass spectrometry on PARP1 following treatment by each enzyme, demonstrating that all three enzymes are hydrolyzing the ADPr modification down to its phosphoribose attachment site. Notable differences were observed between the site identifications that were achieved following the use of these three enzymes -only two of the 14 phosphoribosylated peptide species were found in all samples, and all enzymes produced at least one unique peptide species. Taken together, these data suggest that these enzymes have an inherent bias for PTM structure, location, or length and therefore may be used in parallel to achieve greater coverage of the ADP-ribosylated proteome. Representative spectra from these runs have been annotated in Fig. 6, further confirming the evidence of phosphoribose modifications on these peptides.

Discussion
This work, along with that of Palazzo et al. 13 , has revealed new substrates for the RNA decapping Nudix hydrolases EcRppH and HsNudT16, as well as the CoAse Dr1184, though it must be noted that the existence of this activity against protein-conjugated ADPr in vivo has yet to be explored. In an attempt to determine the structural basis for this unique protein-conjugated ADPr hydrolase activity, we aligned these three proteins with respect to their known structures and identified a short, novel 3 10 helix downsteam from the helix of the Nudix signature sequence (Fig. 7a, red box), suggesting a possible significance to these structures for the degradation of protein-conjugated ADPr. We caution, however, that the structure is not known of the other two single domain Nudix hydrolases tested which lack the activity of interest: AtORF147 and Pa3470 (c.f. Fig. 4a,b). Moreover, the sequence homology between EcRppH/HsNudT16/Dr1184 and AtORF147/Pa3470 is not high beyond the Nudix signature sequence, so it is difficult to predict if the presence or absence of such short helices are the determining structural factor to facilitate the hydrolysis of protein-conjugated ADPr. Modeling of a trimeric PAR in the binding site of EcRppH positions the ribose towards the front of the enzyme allowing ample space for a protein to bind (Fig. 7b). Notably, the bond to be cleaved to leave a phosphoribose attached to a protein is in an ideal position to be hydrolyzed by the residues of the catalytic site of the Nudix enzyme (Fig. 7b). This model gives compelling structural arguments to the biochemistry data collected. The red background indicates sequence identity, red letters sequence indicates conservation of charge, and white background indicates nonhomologous residues. The secondary structure of each enzyme is shown in blue. The box indicates a short 3 10 helix not observed in other Nudix enzymes such as the ADPRases. The secondary structure elements are more conserved between the two enzymes with activity towards PAR and MAR (EcRppH and HsNudT16) compared to Dr1184 that can only hydrolyze PAR. (b) Surface representation of the model of EcRppH with PAR modeled and minimized using the software package MOE. The diphosphate bond of the terminal ADP-ribose attached to a protein is in the active site, well positioned to be hydrolyzed and released as a phosphoribosylated protein product, where the scissile bond is marked.
Mass spectrometry-based proteomics represents the gold standard for the study of post-translational modifications, and the field of ADP-ribosylation will surely benefit from increased access to the suite of proteomic tools that have been developed for other PTMs, such as phosphorylation, acetylation and ubiquitylation. The advent of tag-based approaches for identifying ADP-ribosylation sites has begun to provide access to these tools, but adoption has been relatively low due to technical difficulties which accompany the current methods. The work presented here promises to streamline one of the major methods for ADPr site identification: the simplification of PAR or MAR to its phosphoribose attachment. Recombinant HsNudT16 and EcRppH can be synthesized and purified from E. coli, allowing for low-cost, high yield production which can be performed in most proteomic laboratories. Furthermore, since the structure is known for both of these enzymes, it is possible to predict mutations and truncations which could potentially increase the enzyme's activity towards protein-conjugated ADPr; for example, introducing mutations which will further open up the active site to allow larger ADP-ribosylated substrates access.
The work presented here has the potential to greatly aid in the elucidation of the ADP-ribosylated proteome by providing researchers the necessary tools for generating phosphoribose as a mass spectrometry friendly site localization tag of protein ADP-ribosylation sites.