Crystal Structure of Human ADP-ribose Transferase ARTD15/PARP16 Reveals a Novel Putative Regulatory Domain*

Background: ADP-ribose transferases ARTD1–3/PARP1–3 have an α-helical domain that closes over the NAD+-binding site. Results: Human ARTD15/PARP16 is a mono(ADP-ribose) transferase with a novel α-helical domain that interacts with a catalytic domain loop. Conclusion: The ARTD15 transferase domain is likely regulated by effector binding to the adjacent helical domain. Significance: This provides a basis for understanding the enzymatic mechanism of this previously uncharacterized enzyme. ADP-ribosylation is involved in the regulation of DNA repair, transcription, and other processes. The 18 human ADP-ribose transferases with diphtheria toxin homology include ARTD1/PARP1, a cancer drug target. Knowledge of other family members may guide therapeutics development and help evaluate potential drug side effects. Here, we present the crystal structure of human ARTD15/PARP16, a previously uncharacterized enzyme. ARTD15 features an α-helical domain that packs against its transferase domain without making direct contact with the NAD+-binding crevice or the donor loop. Thus, this novel domain does not resemble the regulatory domain of ARTD1. ARTD15 displays auto-mono(ADP-ribosylation) activity and is affected by canonical poly(ADP-ribose) polymerase inhibitors. These results add to a framework that will facilitate research on a medically important family of enzymes.

Protein ADP-ribosylation, in which ADP-ribose units derived from NAD ϩ are covalently linked to protein side chains (1), was first reported a half-century ago (2). Among the enzymes that modulate their substrate protein activities by mono(ADP-ribosylation) are toxins from various pathogenic bacteria (3) and the eukaryotic ADP-ribose transferase ectoenzymes (4). Poly(ADP-ribosylation), the formation of branched or unbranched chains of ADP-ribose units on protein side chains, is catalyzed by ADP-ribose transferases with diphtheria toxin homology (ARTD), 3 also referred to as poly(ADP-ribose) polymerases (PARPs) (5,6). ARTD1/PARP1 (7,8), a multidomain protein with functions in DNA repair, is a target for development of cancer therapeutics (9).
The human genome encodes at least 18 different ARTD proteins. ARTD1 and ARTD2, as well as ARTD5 and ARTD6 (the tankyrases), have proven poly(ADP-ribose) transferase activity. Recently, ARTD10/PARP10 was identified as a mono(ADP-ribose) transferase in vitro, and it was concluded that most of the remaining family members may also catalyze mono(ADP-ribosylation) rather than ADP-ribose chain synthesis (10). It is not known whether the catalytic activities are an intrinsic property of the transferase domains or a consequence of the presence and identity of other functional domains, which vary widely within the family (5,11).
On the N-terminal side of their transferase domains, ARTD1-3 (and, by prediction, ARTD4) contain an ␣-helical domain that likely has regulatory functions (11,12). This domain interacts with an elongated loop of the transferase domain (the donor or D-loop) and partially blocks the NAD ϩ co-substrate-binding crevice (13)(14)(15)(16). We present the crystal structure of ARTD15, a previously uncharacterized animalspecific member of the family (11,17). ARTD15 has a novel ␣-helical domain that makes interactions with the transferase domain but without sharing sequence or structural homology with the helical domain of ARTD1-4. ARTD15 has automono(ADP-ribosylation) activity and binds to canonical PARP inhibitors.

EXPERIMENTAL PROCEDURES
Recombinant Protein Production-Human ARTD15 cDNA was obtained from the Mammalian Gene Collection (accession number BC031074), provided by the National Institutes of Health. Expression vectors were constructed by inserting cDNA fragments into pNIC28-Bsa4 by ligation-independent cloning (18) Gold(DE3) cells (Stratagene) transformed with the pRARE2 plasmid (Novagen) and protein purification by immobilized metal ion chromatography, size exclusion chromatography, and affinity tag cleavage were performed using standard methods as described under supplemental "Experimental Procedures." Proteins solved in 20 mM HEPES (pH 7.5), 300 mM NaCl, 10% glycerol, and 2 mM tris(2-carboxyethyl)phosphine were stored at Ϫ80°C. For crystallization experiments, the buffer was exchanged to 10 mM sodium acetate (pH 5.5), and the protein was concentrated to 37 mg/ml by ultrafiltration. Crystallization and Structure Determination-Crystallization of ARTD15 Gly-5-Ala-279; derivatization using heavy atoms; native and multiple anomalous diffraction data collection; phasing by multiple isomorphous replacement with anomalous scattering; and model building and refinement using autoSHARP (19), Refmac5 (20), BUSTER (21), and Buccaneer (22) are described in detail under supplemental "Experimental Procedures." ARTD15 crystallized in space group C222 1 , and the structure was refined to a resolution of 2.7 Å. A complete overview of data collection and refinement statistics is given in supplemental Table S1. The final model contained the complete ␣-helical domain, whereas a number of loop residues in the transferase domain were not resolved (supplemental Table S2).

RESULTS AND DISCUSSION
Bacterial Expression of Catalytically Active ARTD15-We set out to express a number of ARTD15 protein fragments in E. coli. We observed that expression constructs spanning only the predicted ADP-ribose transferase domain were poorly soluble and that extension of expression constructs toward the N terminus strongly increased the yields of soluble protein (supplemental Fig. S1). No protein was recovered when the N-terminal segment was expressed on its own. The TMHMM server (24) predicted a transmembrane helix including Trp-290 -Ala-312 near the C terminus of ARTD15. Little protein was recovered when the putative transmembrane segment was included in the expression constructs (supplemental Fig. S1).
Overall Structure of ARTD15-We conducted crystallization experiments with the ARTD15 protein and obtained well diffracting crystals of ARTD15  in the presence of the small canonical PARP inhibitor 3-aminobenzamide. As no suitable molecular replacement model could be identified and as the wild-type protein lacks methionine residues necessary for selenomethionine single-wavelength anomalous dispersion approaches, we performed heavy atom soaks. Anomalous scattering experiments were performed using gold-and platinumsoaked crystals, and the structure was solved using three com-plete three-wavelength data sets. The crystal structure of ARTD15 shows a canonical transferase domain with one molecule of 3-aminobenzamide bound in the nicotinamide pocket paired with a novel ␣-helical domain on its N-terminal side (Fig. 1A). The N-terminal domain consists of six ␣-helices that form an irregular bundle with a large hydrophobic core. Threedimensional structure searches using different algorithms (25,26) did not identify any domains with significant structural similarity, and BLAST searches did not identify proteins with significant sequence similarity. We conclude that ARTD15 is a two-lobed enzyme with a novel N-terminal ␣-helical domain and a canonical diphtheria toxin homology transferase domain (Fig. 1, B and C).
ARTD15 Catalytic Site Structure-The NAD ϩ -binding crevice of ARTD15 is lined by the conserved side chains His-152 and Tyr-182 of the HYE motif, as well as the nicotinamidestacking Tyr-193. In diphtheria toxin, mutagenesis of the side chains corresponding to His-152 and Tyr-182 resulted in a strongly reduced affinity for the NAD ϩ co-substrate (27,28). All three ARTD15 side chains align well with the corresponding ones in ARTD1 (Fig. 2) and other family members. By contrast, the third residue of the motif, the catalytic glutamate (Glu-988) of ARTD1, is replaced by a tyrosine (Tyr-254) in ARTD15. Although this was anticipated from sequence alignments, the crystal structure shows that the tyrosine side chain has a similar orientation as Glu-988 of ARTD1. This is suggestive of a role of Tyr-254 in catalysis, and we speculate that the side chain might stabilize a reaction intermediate. Notably, ARTD15 is the only family member that contains neither a glutamate (ARTD1-4 and the tankyrases ARTD5 and ARTD6) nor an aliphatic residue (the remainder of the family) in this position.
Structure and Potential Role of the Novel Helical Domain-The N-terminal ␣-helical domain of ARTD15 packs against the catalytic domain, burying a mainly hydrophobic interface of 1260 Å 2 as determined using the PISA server (29). As in ARTD1, this likely explains why the domains cannot be produced in isolation (Ref. 13 and supplemental Fig. S1). The helical domain of ARTD15 does not share any obvious sequence similarity with the regulatory domains of ARTD1-4. A structure-based sequence alignment illustrates that these domains are quite dissimilar (Fig. 1C).
The helical domains of ARTD15 and ARTD1-3 occupy partially overlapping overall positions relative to the transferase domains. The only similarity in architecture is, however, between ARTD15 helix ␣2Ј and ARTD1 ␣-helix A, which align roughly with one another (Fig. 3A). Nonetheless, the domain arrangement in both ARTD15 and ARTD1-3 is suggestive of a regulatory role for their ␣-helical domains, albeit by different mechanisms.
The helical domain of ARTD1-3 closes over the NAD ϩ site like a lid, with ␣F restricting access to the co-substrate-binding pocket (Fig. 3A) (13)(14)(15)(16). In ARTD1, the D-loop (Pro-881-Gly-888) makes extensive hydrophobic contacts with ␣-helix D of the helical domain (Fig. 3A). This interaction has been interpreted to keep the catalytic domain in a meta-stable state in which it can accept the adenine base of NAD ϩ (12). A similar arrangement is observed in the structures of ARTD2 and ARTD3 (14 -16), whereas the structure of ARTD4 has not been determined. In our ARTD15 structure, the donor loop is partially disordered, but its position is not in the vicinity of the helical domain. Also, there is no secondary structural element that corresponds to ARTD1 helix D. Thus, if the ARTD15 helical domain is involved in the regulation of cofactor binding or activity of the catalytic domain, this must be accomplished by a mechanism that is distinct from that in ARTD1-3.
The crystal structure of ARTD15 suggests two ways by which the helical domain might relay a putative effector-binding event to the NAD ϩ -binding site. (i) The extended or X-loop between ␣2 and ␤3 (Ser-193-Ile-208) makes both polar and hydropho-bic contacts with ␣2Ј (Fig. 3B). The most extensive contact is formed by the Trp-200 side chain, which is inserted into a hydrophobic pocket created between ␣2Ј and the loop connecting ␣4Ј and ␣5Ј. Also, the His-202 and Leu-204 side chains pack against hydrophobic patches in that region (Fig. 3B). Effector binding to the helical domain might therefore affect the positioning of side chains at the C-terminal end of ␣2 that interact with the cofactor, including Tyr-193, which stacks on the nicotinamide moiety. (ii) Helix ␣2Ј and the loop-helix-loop motif that follows in sequence also pack onto ␣1 of the catalytic domain. Modulation of this interaction might have direct con- . ␣-hel., ␣-helical domain; cat., catalytic; TM, predicted transmembrane helix. C, structure-based sequence alignment of the transferase domains of ARTD15, ARTD8, and ARTD1. Sequences were aligned using ClustalW and adjusted manually based on the crystal structures of the three proteins (Protein Data Bank codes 4F0D, 3SMJ, and 3L3M, respectively). ARTD15 and ARTD1 sequences of the ␣-helical domains are shown on a blue background; the ARTD8 sequence is shown truncated at the transferase domain. Secondary structural elements were annotated following previous nomenclature for ARTD1 (13) and ARTD10 (10) with an additional prime to mark the ␣-helices of the novel domain of ARTD15. Red stars mark the positions of the HYE motif residues (6,11), and the blue star highlights the conserved nicotinamide-stacking aromate (6).

Structure of ARTD15
JULY 13, 2012 • VOLUME 287 • NUMBER 29 sequences on the positioning of side chains at the C-terminal end of ␤3, including His-152, the histidine that is part of the ARTD signal motif HYE (6).
Catalytic Activity of ARTD15-To investigate whether ARTD15 has ADP-ribose transferase activity, we conducted activity assays in the presence of 2% biotinylated NAD ϩ followed by either Western blotting or capture of hexahistidinetagged ARTD15 on nickel-chelating multiplates and detection of modified products using anti-biotin antibodies and bioluminescence. ARTD15 displayed automodification activity (Fig. 4) but was unable to modify histone proteins (data not shown). We observed no apparent signs of poly(ADP-ribose) polymer formation, although we did not study this aspect in detail. Next, we used the protein capture strategy in multiplate format to determine an apparent K m of 267 Ϯ 52 M with respect to NAD ϩ and automodification. This is ϳ2-3-fold higher than the K m determined for DNA-independent ARTD1 automodification using the same method (23). We recently showed that several compounds, including the widely used unselective PARP inhibitor EB-47, can stabilize ARTD15 in thermal unfolding assays (30). Along with our crystal structure in complex with 3-aminobenzamide, this implies that ARTD15 can be targeted by canonical PARP inhibitors. We conclude that ARTD15 has ADP-ribose transferase activity, is capable of automodification, and can be inhibited by canonical PARP inhibitors.
While this manuscript was under review, Di Paola et al. (31) published a cell biological and biochemical characterization of ARTD15. In agreement with our results, ARTD15 was shown to possess mono(ADP-ribose) transferase activity. Furthermore, the protein was shown to localize to the cytosolic face of endoplasmic reticulum membranes and the nuclear envelope, a localization that depended on the presence of the predicted transmembrane helix (Ser-288 -Ile-308). Finally, Di Paola et al. (31) identified the nuclear transport factor karyopherin-␤1 as a substrate protein for ARTD15.
In summary, we have shown that ARTD15 contains a novel ␣-helical domain that is in contact with the active site of the transferase domain. Important functions of the helical domain are supported by its conservation from insects to humans (11), and we anticipate that this domain functions in ARTD15 regulation. The crystal structure of ARTD15 reported here will guide future research on (i) ADP-ribose transferase mechanisms, (ii) regulation of ARTD15 and the role of the helical domain therein, (iii) evaluation of the specificity of preclinical  (6) and the nicotinamide-stacking tyrosine are indicated by red and blue asterisks, respectively, as in Fig. 1C. Note that the position of the catalytic glutamate of ARTD1 is occupied by a tyrosine that maintains a similar orientation with respect to the active site. The bound ligand of ARTD1 was omitted for clarity. 3-AB, 3-aminobenzamide. and clinical PARP inhibitors, and (iv) development of ARTD15selective inhibitors as research tools.