Structure–function analyses reveal the mechanism of the ARH3-dependent hydrolysis of ADP-ribosylation

ADP-ribosylation of proteins plays key roles in multiple biological processes, including DNA damage repair. Recent evidence suggests that serine is an important acceptor for ADP-ribosylation, and that serine ADP-ribosylation is hydrolyzed by ADP-ribosylhydrolase 3 (ARH3 or ADPRHL2). However, the structural details in ARH3-mediated hydrolysis remain elusive. Here, we determined the structure of ARH3 in a complex with ADP-ribose (ADPR). Our analyses revealed a group of acidic residues in ARH3 that keep two Mg2+ ions at the catalytic center for hydrolysis of Ser-linked ADP-ribosyl group. In particular, dynamic conformational changes involving Glu41 were observed in the catalytic center. Our observations suggest that Mg2+ ions together with Glu41 and water351 are likely to mediate the cleavage of the glycosidic bond in the serine–ADPR substrate. Moreover, we found that ADPR is buried in a groove and forms multiple hydrogen bonds with the main chain and side chains of ARH3 residues. On the basis of these structural findings, we used site-directed mutagenesis to examine the functional roles of key residues in the catalytic pocket of ARH3 in mediating the hydrolysis of ADP-ribosyl from serine and DNA damage repair. Moreover, we noted that ADPR recognition is essential for the recruitment of ARH3 to DNA lesions. Taken together, our study provides structural and functional insights into the molecular mechanism by which ARH3 hydrolyzes the ADP-ribosyl group from serine and contributes to DNA damage repair.

The crystal structure of apo-ARH3 was determined by Mueller-Dieckmann et al. (24) who have shown that ARH3 is a monomeric molecule with 19 ␣-helices. A possible substrate-binding mode was proposed using computer modeling, which was further validated by mutagenesis analysis (24). However, the crystal structure of the ARH3 and ligand complex has not yet been reported. Here, using X-ray crystallography, we determined the detailed structure of ARH3 in the complex with ADP-ribose (ADPR). Combining with biochemistry and cell biology assays, our results demonstrate that the enzymatic activity of ARH3 plays an important role in DNA damage repair.

Overall structure of the ARH3-ADPR complex
To understand the molecular mechanism by which ARH3 mediates ADP-ribosylhydrolyzation, the crystal structure of the ARH3-ADPR complex was determined at 1.58-Å resolution by X-ray diffraction. The final model contains one ARH3 molecule in the asymmetric unit (Fig. 1A). The ARH3 monomer binds to one ADPR molecule and two Mg 2ϩ ions according to the electron density map (Fig. 1B).
The overall structure of ARH3 in the complex shows a compact all ␣-helical-fold (Fig. 1A). Notably, earlier studies on the 3D structure of apo-ARH3 show two crystal forms that are con- The crystal structure of the ARH3-ADPR complex formationally different, especially in the loop between 1 and ␣2 (1-␣2 loop, residues Glu 41 -Asp 47 ) around the active pocket (24). Structural comparison shows that the corresponding 1-␣2 loop in the ARH3-ADPR complex is more similar to that of the orthorhombic crystal form (PDB code 2FOZ) (Fig. S1). Structural alignment using the DALI program reveals that ARH3 shares a similar structural fold with several other ADPribosylglycohydrolases (ARH) despite their relatively low sequence identity (Fig. 2). These ARHs include ARH MJ1187 from Methanococcus jannaschii (PDB code 1T5J), ARH1 from Homo sapiens (PDB code 3HFW), putative ARH from Thermus thermophilus (PDB ID code 2YZV), and mono-ADP-ribosylhydrolase DraG from Rhodospirillum rubrum (PDB code 2WOC), giving a root mean square deviation of 2.4 Ϯ 0.2 Å for all equivalent Ca atoms (Fig. 1C). These ARHs share common structural characteristics representing a compact all ␣-helicalfold, which is remarkably distinguished from that of macro domain ADPR hydrolases including PARG (PDB code 4B1H), MacroD2 (PDB code 4IQY), and TARG1 (PDB code 4J5R). The core structure of these macro domain enzymes contains a three-layered ␣-␤-␣ sandwich, exhibiting the typical macro domain-fold (Fig. 1D).

The catalytic site of ARH3
The ADPR-binding cavity of ARH3 unambiguously contains a binuclear Mg 2ϩ site according to 2F o Ϫ F c electron density map, termed MgA and MgB, respectively (Fig. 1B). The distance between MgA and MgB is 3.8 Å. Both MgA and MgB have octahedral coordination. Structural comparison between apo-ARH3 (PDB codes 2FOZ and 2FP0) and the ARH3-ADPR complex reveals that both Mg 2ϩ ions are shifted about 0.6 Å toward the distal ribose of ADPR, enabling the hydroxyl groups of the distal ribose to coordinate the two Mg 2ϩ ions. MgA is coordinated with Asp 314 , Asp 316 , Thr 317 , three hydroxyl groups of the distal ribose and water269. Meanwhile, MgB is coordinated with the 2Ј-hydroxyl group of the distal ribose, Thr 76 , Asp 77 , Asp 78 , Asp 316 , and two water molecules (water269 and water85) (Fig. 3A). The sequence alignment and structural comparison revealed that the Mg 2ϩ coordinating acidic residues Asp 77 , Asp 78 , Asp 314 , and Asp 316 are fully conserved in homologous ARH enzymes (Fig. 2).
In addition to the abovementioned coordinating residues, we notice that Glu 41 is also involved in the Mg 2ϩ coordination in the apo-ARH3 (PDB code 2FOZ) as the hydroxyl groups from the distal ribose of ADPR are absence in the apo-ARH3. However, in the ARH3-ADPR complex, the carboxyl side chain of Glu 41 forms a hydrogen bond with water351 located at the catalytic center, with a close distance of 3.0 Å. Another hydrogen bond forms between water351 and the 1Ј-hydroxyl group of the distal ribose of ADPR, with a close distance of 2.8 Å. Moreover, the distance between MgA and water351 is 2.5 Å, and the distance between MgB and water351 is 3.4 Å. Thus, a hydrogen bond network is formed among Glu 41 , water351, 1Ј-hydroxyl group of the distal ribose, and MgA and MgB (Fig. 3B).
Moreover, water351 lies close to the C1Ј atom of the distal ribose with a distance of 3.7 Å (Fig. 3B). Because the 1Ј-hydroxyl group of the distal ribose of ADPR is covalently linked to the side chain of serine via a glycosidic bond, it is likely that Glu 41 functions as the catalytic residue to activate water351, which in turn carries out a nucleophilic attack on the C1Ј atom of the distal ribose for the cleavage of the scissile glycosidic bond (Fig.  3C). This hydrolyzation mechanism mimics the canonical catalytic mechanism of glycosidases, in which an acidic residue is used as a proton acceptor to activate a water molecule, and then cleave the scissile glycosidic bond via nucleophilic attack (25).
Both Mg 2ϩ ions are involved in the positioning of water351 in the catalytic center through hydrogen-bonding interaction. In contrast, in MacroD2, both the ␣-phosphate and distal ribose of ADPR contribute to hold the catalytic water (26). In PARG, the stabilizing factors for the catalytic water are Glu 755 , Asp 737 , and the diphosphate moiety of ADPR (27). Thus, the mechanism of ARH3-dependent ADPR hydrolyzation is distinct from macro domain enzymes.
Collectively, this structural analysis suggests that water351 is the catalytic water molecule that exists in the center of the catalytic pocket of ARH3, and Glu 41 works as a key catalytic residue in the ADPR hydrolyzation reaction, which is consistent with earlier reports showing that Glu 41 is required for catalysis (14,28). Thus, our analysis of the catalytic center of ARH3 may reveal the detailed molecular mechanism by which ARH3 mediates the ADPR hydrolyzation.
It has been reported that the conformation of Glu 41 in the ARH3 orthorhombic crystal form (PDB code 2FOZ) is obviously different from that of the monoclinic crystal form (PDB code 2FP0) (Fig. 3D). In the orthorhombic crystal form, Glu 41 is involved in the Mg 2ϩ coordination, whereas Glu 41 moves about 1.8 Å toward to the protein surface and does not coordinate with Mg 2ϩ in the monoclinic crystal form (24). Notably, Glu 41 undergoes a significant conformational change in the ADPRbound ARH3, the side chain of Glu 41 is shifted 2.1 Å compared with that in the apo-ARH3 orthorhombic crystal form (PDB code 2FOZ), and is pushed away from the MgA (Fig. 3D). Taken together, three conformations have been observed for Glu 41 , indicating that Glu 41 is in the dynamic status at the acidic catalytic center. This conformational diversity of Glu 41 is in good agreement with its catalytic roles, which was also confirmed by the ADPR hydrolyzation assays (see below).

The structure of the ADPR binding cavity
The electron density map reveals that an ADPR molecule binds to a deep cavity that is mainly encircled by loops (Fig. 4A). A number of residues are responsible for the binding and stabilizing ADPR via main chain and side chain contacts (Fig. 4B). The adenine ring is stacked by Tyr 149 and Phe 143 through the interaction (Fig. 4B). This stacking interaction between the adenine ring and aromatic residues is common in ADPribose hydrolases. For example, Tyr 212 contributes to the stacking interaction with the adenine ring of ADPR in DraG (29), and the corresponding stacking residues are Phe 224 and Phe 902 in MacroD2 and PARG, respectively (26,27) (Fig. 4C). The double stacking interaction in the complex of ARH3-ADPR may clamp the adenine ring tightly and makes the adenine ring in a rigid conformation compared with the single stacking interaction. In addition to theinteraction, the adenine ring hydrogen The crystal structure of the ARH3-ADPR complex  The crystal structure of the ARH3-ADPR complex bonds to the main chain of Gly 147 and Ser 148 , which further contributes to the selectivity of the substrate in the catalytic pocket of ARH3 (Fig. 4B).
Moreover, the adenosine ribose in ADPR forms a hydrogen bond with the main chain amino group of Gly 119 (Fig. 4B). The diphosphate moiety in ADPR sits in the N terminus of two ␣-helices of ARH3, and it hydrogen bonds the main chain amino groups of Ala 118 , Gly 119 , Val 120 , Asn 151 , and Gly 152 , and the side chain of His 182 , generating a network of hydrogen bonds that comprise a large proportion of the interaction between ARH3 and ADPR (Fig. 4B). In addition, several water molecules are involved in this hydrogen bond network formation. Notably, there is no direct interaction between the distal ribose of ADPR and ARH3. Instead, the two catalytic Mg 2ϩ ions interact with the distal ribose via coordination.

Different catalytic mechanisms of ADP-ribose hydrolases
Sequence alignment reveals that these catalytically related residues consisting of Glu 41 , Asp 77 , Asp 78 , Asp 314 , and Asp 316 are not conserved between ARH family enzymes and macro domain ADP-ribose hydrolases, such as PARG, MacroD1/D2, and TARG1. Moreover, unlike ARHs, there is no binuclear metal-binding site in these macro domain enzymes. Thus, the catalytic mechanism of protein ADP-ribose hydrolases can be classified into two different groups, namely metal-dependent and metal-independent catalysis. ARHs, such as ARH3, belong to metal-dependent catalysis, utilizing two Mg 2ϩ ions and acidic residues to complete the catalytic reaction, which might be highly conserved.
In contrast, the catalytic mechanism is not conserved in the macro domain ADP-ribose hydrolases, For example, Glu 756 and a water molecule act together to catalyze the reaction in PARG (27), whereas the key catalytic factor in MacroD2 is an activated water (26). For TARG1, the Lys 84 /Asp 125 catalytic dyad may contribute to the catalytic reaction (30).

Structural comparison of the binding pocket between ARH3 and PARG
The electrostatic potential molecular surface shows that the binding pocket of ARH3 is divided into two subunits by the binuclear Mg 2ϩ sites. The subunit accommodating ADPR is mainly composed of a basic region (positive charged), which has strong affinity for the diphosphate group of ADPR. The other subunit, the putative binding site for the remaining substrate groups, is entirely composed of an acidic region (negative charged) (Fig. 5A). Interestingly, the charge characteristic of the binding pocket in ARH3 is remarkably distinguished from that in PARG. The binding pocket of PARG, accommodating the ADPR dimer, is mostly composed of the basic region (PDB code 5A7R) (Fig. 5B). Moreover, the binding pocket in T. thermophila PARG (PDB code 4L2H) predominantly comprises the basic region (Fig. 5C). The electric charge differences in the substrate-binding pocket may be closely related to their divergent enzyme-degrading activities.

Key residue mutations in ARH3 abolish the enzymatic activity
To validate the in vitro structural analysis, we generated a series of mutations in the catalytic binding pocket of ARH3, including E41Q, D77N, S148A, Y149A, H182A, D314A, and T317A. Among these mutations, E41Q, D77N, D314A, and T317A occur at the catalytic site, whereas S148A, Y149A, and H182A exist in the ADPR recognition pocket. We generated the recombinant glutathione S-transferase-ARH3 mutants, and examined these proteins in the in vitro ADPR hydrolyzation assays. Compared with the WT ARH3, all these mutations abolished the enzymatic activity of ARH3, suggesting that all these residues are essential for the ARH3-dependent ADPR hydrolyzation (Fig. S2), which is also consistent with earlier studies (14,24). Given Asp 77 , Asp 314 , and Thr 317 are involved in Mg 2ϩ coordination, our results suggest that Mg 2ϩ is important in the catalytic process.

The enzymatic activity of ARH3 plays an important role in DNA damage repair
It has been shown that ARH3 is involved in DNA damage repair (23). Using a laser microirradiation assay, we examined the recruitment of ARH3 to DNA lesions. WT of ARH3 was recruited to the sites of DNA damage within 30 s, and this relocation process was suppressed by the PARP inhibitor olaparib (Fig. 6A), suggesting that the recruitment of ARH3 to DNA lesions is mediated by DNA damage-induced ADPribosylation. Compared with WT ARH3, S148A, Y149A, and

The crystal structure of the ARH3-ADPR complex
H182A mutants failed to relocate to DNA lesions (Fig. 6B). Because S148A, Y149A, and H182A disrupted the ADPR-binding pocket, it suggests that ADPR recognition plays a key role for the recruitment of ARH3. Notably, E41Q, D77N, D314A, and T317A mutants were still recruited to DNA lesions, albeit at low intensity (Fig. 6C), suggesting that the catalytic site is not essential for the recruitment of ARH3. Instead, our results demonstrate that ADPR recognition is required for the recruitment of ARH3 to DNA lesions.
Next, we examined the roles of ARH3 in DNA damage repair. We knocked down the endogenous ARH3 with siRNA, and expressed WT ARH3 and its mutants in U2OS cells (Fig. 7A). The cells were treated with methyl methanesulfonate, a DNA damaging agent. We performed comet assays to examine the DNA damage repair kinetics. Under both alkaline and neutral conditions, the DNA damage repair was impaired when cells were expressing mutant ARH3 (Fig. 7, B and C), suggesting that ARH3-dependent ADPR hydrolyzation plays an important role in DNA damage repair.

Discussion
In 2006, Mueller-Dieckmann et al. (24) determined the 3D structure of apo-ARH3, which are actually in two different crystal forms, including monoclinic and orthorhombic crystal forms. The significant conformational differences occur in 1-␣2 loop around the active pocket, indicating that this region is very flexible. In our structure analysis, we show that the 1-␣2 loop in the ARH3-ADPR complex is more close to that of orthorhombic crystal form (Fig. S1), suggesting that ligand binding stabilizes this region, which plays a key role for catalysis. Moreover, in the ARH3-ADPR complex, we observed that a potential nucleophilic water molecule (water351) is at the catalytic center (Fig. 3). Combining the conformation diversity of Glu 41 and the canonical catalytic mechanism of glycosidases, we proposed a model of catalysis mediated by Glu 41 (Fig. 3C). Our analysis on the ARH3-ADPR complex provides the structural evidence. Moreover, based on this unique conformational change surrounding Glu 41 , we examined the role of Glu 41 and

The crystal structure of the ARH3-ADPR complex
confirmed that loss of Glu 41 abolished the enzymatic activity of ARH3 (Fig. S2), which is in agreement with Mueller-Dieckmann et al. (24). Because ARH3 mediates ADP-ribosylhydrolation in response to DNA damage, we have shown the first evidence that the E41Q mutation impaired ARH3-dependent DNA damage repair (Figs. 6 and 7). Taken together, our study demonstrates the catalytic mechanism by which ARH3 hydrolyzes ADP-ribosylation. Based on the computer modeling of ARH3 and ligand binding, Mueller-Dieckmann et al. (24) performed mutagenesis and characterized the catalytically associated amino acid residues. They mutated 9 individual residues, including Glu 41 , Asp 77 , Ser 148 , Tyr 149 , Asn 151 , His 182 , Glu 275 , Asp 314 , and Thr 317 , to examine the roles of these residues in the enzymatic activity of ARH3 (24). In our studies, we examined the structural analysis of the ARH3-ADPR complex and performed the mutagenesis analysis based on the detailed analysis of the ARH3-ADPR complex. We confirmed these specific residues that mediate the catalysis, which is in agreement with Mueller-Dieckmann et al. (24). Both Mueller-Dieckmann et al. (24) and us reveal that Ser 148 , Tyr 149 , and His 182 are involved in ADPR recognition; Glu 41 , Asp 77 , Asp 314 , and Thr 317 are catalytically related. Notably, in addition to the biochemical analysis, we further examined the roles of these residues in the ARH3-mediated DNA damage repair. Recent studies indicate that ARH3 plays a role in DNA damage repair. Using laser microirradiation assays, we have revealed that ARH3 is recruited by DNA lesions, which is mediated by DNA damage-induced poly(ADP-ribosyl)ation (Fig. 6A). We further analyzed ADPR recognition residues including Ser 148 , Tyr 149 , and His 182 , and found that these residues are required for the recruitment of ARH3 to DNA lesions (Fig. 6B). In contrast, catalytically related residues, including Glu 41 , Asp 77 , Asp 314 , and Thr 317 , although very important, are not essential for the recruitment of ARH3. Mutations of these residues only partially impair the recruitment. Furthermore, we performed comet assays to examine the roles of these residues

The crystal structure of the ARH3-ADPR complex
in DNA damage repair. However, due to loss of the enzymatic activity, mutating catalytically related residues still abolished the ARH3-mediated DNA damage repair (Fig. 7). Thus, our detailed molecular biology analysis reveals the biological functions of two different groups of residues in ARH3. One group is for substrate recognition; the other group is for ADPR hydrolysis. Both groups of residues are important for ARH3-dependent DNA damage repair. In summary, our research work provides further insight on the catalytic mechanism and the structure-function analysis of ARH3.

Protein expression and purification
The ARH3 gene was amplified from genomic DNA of 293T cells and constructed into the modified pET-15b vector, containing an N-terminal hexahistidine tag and a cleavage site for human rhinovirus 3C (HRV 3C) protease. The N-terminal 18 residues of ARH3 were removed during the cloning process.
The recombinant plasmid of pET-15b-ARH3 was transformed into Escherichia coli BL21 (DE3) cells. The transforming cells were grown in liquid L broth media at 37°C to A 600 ϭ 0.8 -1.0, and then 0.12 mM isopropyl ␤-D-1-thiogalactopyranoside was added to induce recombinant protein expression at 16°C overnight. Cells were harvested by centrifugation, resuspended in ice-cold lysis buffer (25 mM Tris-HCl, pH 8.0, 100 mM NaCl), and lysed through sonication. The cell lysate was then centrifuged at 35,000 ϫ g for 45 min.
For protein purification, the supernatant containing the soluble ARH3 was loaded onto a nickel-chelating Sepharose (GE Healthcare) column pre-equilibrated with lysis buffer. After extensively washing with lysis buffer, ARH3 was eluted using elution buffer (25 mM Tris-HCl, pH 8.0, 100 mM NaCl, and 250 mM imidazole). The N-terminal His tag was removed by the HRV 3C protease. Then the eluent was further purified using an ion-exchange column Source 15Q (GE Healthcare) and Superdex 200 Increase column (GE Healthcare) successively. The purified protein was concentrated to 11 mg/ml in 10 mM Tris-HCl, pH 8.0, and 100 mM NaCl.

Crystallization
To screen for crystallization of the ARH3-ADPR complex, ARH3 and ADPR were mixed in the molar ratio of 1:3.

Data collection and processing
To collect the X-ray data, the crystals were immersed in the cryoprotectant buffer containing 20% glycerol (v/v) plus reservoir buffer and then flash-cooled in a 100 K nitrogen stream. The X-ray diffraction data were collected on the beamline BL17u1 at Shanghai Synchrotron Radiation facility (SSRF). The crystal belongs to space group P2 1 with the unit cell dimensions a ϭ 43.885 Å, b ϭ 76.737 Å, c ϭ 48.533 Å, and ␤ ϭ 103.598°. The data were processed using the HKL-3000 software suite (31).

Structure determination and refinement
The crystal structure of the ARH3-ADPR complex was solved through molecular replacement using Phaser (32). The structure of apo-ARH3 (PDB code 2FOZ) was used as the search model. The structure building was performed using the ARP/WARP software (33). The further refinement was carried out repeatedly using COOT (34) and PHENIX (35). The detailed statistics of data collection and structure refinement are summarized in Table 1. All the molecular graphics figures were prepared using PyMol.

Cell culture
U2OS cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and cultivated at 37°C in 5% CO 2 (v/v). U2OS cells were transfected with plasmids encoding ARH3 or its mutants.

Synthesis and purification of PAR
His-PARP1 was expressed in BL21 and purified by nickelnitrilotriacetic acid affinity resin. PAR was synthesized in a 1-ml incubation buffer containing 100 mM Tris-HCl, pH 7.8, 10 mM MgCl 2 , 1 mM NAD ϩ , 10 mM DTT, 50 g of octameric oligonucleotide GGAATTCC, and 100 g of recombinant PARP1. The mixture was incubated at 37°C for 60 min and stopped by an addition of 20 ml of ice-cold 20% TCA. Oligo DNA was removed by DNase I. Following precipitation the pellet was washed with ice-cold 99.8% ethanol and stored at Ϫ20°C.

The crystal structure of the ARH3-ADPR complex ADPR hydrolyzation assay
Approximately 5 pmol of PAR and 1 pmol of each recombinant protein were incubated together in PBS. After incubation for 1 h at room temperature, heating at 90°C for 8 min, 2-l samples were dotted onto nitrocellulose membranes. After incubation for 20 min at 60°C, dot blotting assays were performed with anti-PAR antibody.

Laser microirradiation and microscope image acquisition
Cells transfected with GFP-tagged corresponding plasmids were grown on 35-mm glass bottom dishes (NEST Biotechnology). Laser microirradiation was performed on OLYMPUS IX71 inverted fluorescence microscope coupled with the MicroPoint Laser Illumination and Ablation System (Photonic Instruments, Inc.). A 337.1-nm laser diode (3.4 milliwatts) transmits through a specific Dye Cell and then yields a 365-nm wavelength laser beam that is focused through X60 UPlanSApo/1.35 oil objective to yield a spot size of 0.5-1 m. The pulse energy is 170 microjoules at 10 Hz. Images were taken by the same microscope with CellSens software (Olympus). The GPF fluorescence strips were recorded at the indicated time points and then analyzed with Image J software. 30 cells were analyzed from three independent experiments. Error bars represent the standard deviation.

Comet assay
Single-cell gel electrophoretic comet assay was performed under neutral conditions to detect DNA double strand breaks and alkaline conditions to detect single strand DNA breaks. U2OS cells with or without the indicated treatments were recovered in normal culture medium for the indicated time. Cells were collected and rinsed twice with ice-cold PBS; 2 ϫ 104/ml cells were combined with 1% LMAgarose at 40°C at a ratio of 1:3 (v/v) and immediately pipetted onto slides. For cellular lysis, the slides were immersed in the neutral lysis solution (2% Sarkosyl, 0.5 M EDTA, 0.5 mg/ml of proteinase K in pH 8.0) overnight at 37°C in the dark, or alkaline lysis solution (1.2 M NaCl, 100 mM EDTA, 0.26 M NaOH, 0.1% Sarkosyl) followed by washing in the rinse buffer (90 mM Tris-HCl, pH 8.5, 90 mM boric acid, 2 mM EDTA) and alkaline wash buffer (0.03 M NaOH, 2 mM EDTA) for 30 min with two repeats. Then, the slides were subjected to electrophoresis at 20 V (0.6 V/cm) for 25 min and stained in 2.5 g/ml of propidium iodide for 20 min. All images were taken with a fluorescence microscope and analyzed by Comet Assay IV software.