The structure of human ADP-ribosylhydrolase 3(ARH3) provides insights into the reversibilityof protein ADP-ribosylationChristoph Mueller-Dieckmann*, Stefan Kernstock†, Michael Lisurek‡, Jens Peter von Kries‡, Friedrich Haag†,Manfred S. Weiss*§, and Friedrich Koch-Nolte†§

*European Molecular Biology Laboratory Hamburg Outstation, c�o Deutsches Elektronen-Synchrotron, Notkestrasse 85, D-22603 Hamburg, Germany;†Institute of Immunology, University Hospital, Martinistrasse 52, D-20246 Hamburg, Germany; and ‡Leibniz-Institut fur Molekulare Pharmakologie, FMP,Robert-Roessle-Strasse 10, Campus Berlin–Buch, D-13125 Berlin, Germany

Communicated by David S. Eisenberg, University of California, Los Angeles, CA, August 12, 2006 (received for review January 27, 2006)

Posttranslational modifications are used by cells from all kingdomsof life to control enzymatic activity and to regulate protein func-tion. For many cellular processes, including DNA repair, spindlefunction, and apoptosis, reversible mono- and polyADP-ribosyla-tion constitutes a very important regulatory mechanism. More-over, many pathogenic bacteria secrete toxins which ADP-ribosy-late human proteins, causing diseases such as whooping cough,cholera, and diphtheria. Whereas the 3D structures of numerousADP-ribosylating toxins and related mammalian enzymes havebeen elucidated, virtually nothing is known about the structure ofprotein de-ADP-ribosylating enzymes. Here, we report the 3Dstructure of human ADP-ribosylhydrolase 3 (hARH3). The moleculararchitecture of hARH3 constitutes the archetype of an all-�-helicalprotein fold and provides insights into the reversibility of proteinADP-ribosylation. Two magnesium ions flanked by highly con-served amino acids pinpoint the active-site crevice. RecombinanthARH3 binds free ADP-ribose with micromolar affinity and effi-ciently de-ADP-ribosylates poly- but not monoADP-ribosylatedproteins. Docking experiments indicate a possible binding modefor ADP-ribose polymers and suggest a reaction mechanism. Ourresults underscore the importance of endogenous ADP-ribosyla-tion cycles and provide a basis for structure-based design ofADP-ribosylhydrolase inhibitors.

protein structure � posttranslational modification �glycohydrolase � docking

Posttranslational modifications (PTMs) are covalent modifi-cations of amino acid side chains in proteins that come in

many different sizes and shapes, ranging from the simple addi-tion of a phosphate group to complex multistep glycosylations.Enzyme-catalyzed PTMs allow rapid responses to environmen-tal stimuli and play crucial roles in signal transduction.NAD-dependent ADP-ribosylation is a reversible PTM in whichmono- and polyADP-ribosyltransferases (ARTs and PARPs)and ADP-ribosylhydrolases (ARHs) and poly(ADP-ribose) gly-cohydrolases (PARGs) catalyze amino acid-specific ADP-ribosylation and de-ADP-ribosylation, respectively (Fig. 1)(1–10). ADP-ribosylation has attracted attention because bac-terial virulence factors, including diphtheria, cholera, and per-tussis toxin, use it as part of their pathogenic mechanism (2, 11).Mono- and polyADP-ribosylation have been recognized also asregulatory mechanisms in many cellular processes, includingDNA-repair, chromatin decondensation, transcription, telomerefunction, mitotic spindle formation, and apoptosis (5–10, 12).

Several enzymes have been cloned that catalyze de-ADP-ribosylation of mono- or polyADP-ribosylated proteins (13–15).Dinitrogenase-activating glycohydrolase (DRAG), an Arg-specificARH from the phototrophic bacterium Rhodospirillum rubrum,regulates a key enzyme of nitrogen fixation (16–18). The humangenome encodes three DRAG-related proteins designated ARH1,ARH2, and ARH3 (19), which are 357, 354, and 363 residues long,

respectively. ARH1, like DRAG, specifically de-ADP-ribosylatesproteins mono-ADP-ribosylated on arginine residues (20, 21).ARH3 de-ADP-ribosylates polyADP-ribosylated proteins, albeit atonly �10% of the activity observed for the PARG (21), with whichit shares little if any sequence similarity (19, 21). ARH3 does not acton ADPR bonds formed with Arg, Asn, or Cys. The function ofARH2 remains elusive because it acts on neither any of theseADP-ribosylated residues nor on polyADPR (21).

The 3D structures of numerous ADP-ribosylating toxins and ofthree toxin-related vertebrate ARTs (chicken PARP-1, mouse

Author contributions: C.M.-D. and S.K. contributed equally to this work; F.H., M.S.W., andF.K.-N. designed research; C.M.-D., S.K., M.L., and J.P.v.K. performed research; C.M.-D., S.K.,M.L., F.H., F.K.-N., M.S.W., and J.P.v.K. analyzed data; and M.S.W. and F.K.-N. wrote thepaper.

The authors declare no conflict of interest.

Abbreviations: ADPR, ADP-ribose; ARH, ADP-ribosylhydrolase; ART, monoADP-ribosyl-transferase; DRAG, dinitrogenase-activating glycohydrolase; PARG, poly(ADP-ribose)gly-cohydrolase; PARP, polyADP-ribosylpolymerase.

Data deposition: The atomic coordinates and structure factor amplitudes of the crystalforms of hARH3 have been deposited in the Protein Data Bank [PDB ID codes: 2FOZ(orthorhombic) and 2FP0 (monoclinic)].

§To whom correspondence may be addressed. E-mail: msweiss@embl-hamburg.de ornolte@uke-hamburg.de.

© 2006 by The National Academy of Sciences of the USA

Fig. 1. Posttranslational modification of proteins by reversible ADP-ribosylation. ARTs and PARPs transfer the ADP-ribose (ADPR) moiety from�-NAD onto specific amino acid side chains or onto ADPR moieties (X) of targetproteins under the release of nicotinamide. This modification may lead toeither activation or inactivation of the target protein. Protein-ADP-ribosylhydrolases (ARHs and PARGs) hydrolyze the �-glycosidic bond betweenADPR and the side chain, thereby restoring normal protein function. X can beArg, Asp, Cys, diphthamide, Glu, or ADPR. In the case of mono-ADP-ribosylation, R and R� are OH groups. In the case of polyADP-ribosylation,attachment of ADPR can take place at the R site (elongation) or at the R� site(branching). In mammals, two distinct subfamilies of ARTs (ART1–5, PARP1–17) and two distinct subfamilies of ARHs (ARH1–3, PARG) exist.

15026–15031 � PNAS � October 10, 2006 � vol. 103 � no. 41 www.pnas.org�cgi�doi�10.1073�pnas.0606762103

Dow

nloa

ded

by g

uest

on

July

5, 2

020

PARP-2, and rat ART2.2) have provided fascinating insights intothe structure and function of ADP-ribosylating enzymes (22–27). Incontrast, very little is known about the structure and function of theenzymes catalyzing protein de-ADP-ribosylation (18, 21, 28). Here,we report the 3D structure of a de-ADP-ribosylating enzyme. Basedon binding, docking, and mutagenesis studies, we are able topropose the substrate-binding mode and the reaction mechanism.

ResultsOverall Structure. The 3D structure of hARH3 lacking the N-terminal 16 aa (Mr � 37,745 Da) was determined in two differentcrystal forms, one grown in the absence and the other in thepresence of ADP, and refined to high resolution (1.6 Å). Datacollection, phasing, and refinement statistics indicate that thestructure is of excellent quality (Table 1; and see Table 3, which ispublished as supporting information on the PNAS web site). Allthree protomers from the two different crystal forms superimposewell, with an r.m.s. deviation of 0.5 Å between the two crystal formsand of 0.6 Å for the two protomers from the monoclinic crystalform (Fig. 6, which is published as supporting information on thePNAS web site). The main differences are found in four small loopregions. The overall structure of hARH3 is that of a compactall-�-helical monomeric molecule containing 19 �-helices, withapproximate dimensions of 53 � 50 � 62 Å3 (Fig. 2 A and B; andsee Fig. 7, which is published as supporting information on thePNAS web site). Size-exclusion chromatography confirmed that themonomeric state of hARH3 seen in the crystal structures is alsopresent in solution. The complete amino acid chain, except for aninternal loop between Pro-46 and Glu-55, is well defined in theelectron-density map. Both N and C termini are located on thesame side of the molecule, �18 Å apart. Two bound Mg2� ions arelocated directly opposite of the termini in a cleft spanning the entiremolecule. The core structure of hARH3 consists of four quasido-mains, designated A, B, C, and D (Fig. 2 C and D) exhibiting somedegree of internal twofold symmetry. The two central quasidomainsA and C consist of a three-helical bundle, each with all helical axesrunning approximately parallel to each other. B and D are locatedon the sides of A and C; their helices run roughly perpendicular tothose of A and C. The molecular architecture of hARH3 constitutes

the archetype of an all-�-helical protein family, which, based onsymmetry considerations, may have arisen from gene duplication.

Structural and Sequence Similarities. A secondary structure-matching search against all structures in the PDB yielded one hitwith a Z score of 9.3 for the Protein Data Bank entry 1T5J. Thisprotein is annotated as the hypothetical protein Mj1187 fromMethanococcus jannaschii. Despite only 22% sequence identity, thestructural homology to hARH3 is apparent. The structures aresuperimposable, with an r.m.s. deviation of 1.3 Å for 240 alignedC�-positions (Fig. 8A, which is published as supporting informationon the PNAS web site). A structure-based sequence alignment ofhARH3 and the 1T5J structure shows that 14 of the 19 �-helicesalign very well (Fig. 8B). Considerable deviations are observed onlyin loop regions. The structural similarity suggests that the hypo-thetical protein Mj1187 is an ADP-ribosylhydrolase.

PSI-BLAST searches readily identify sequence relationships ofARH3 with ARH1, ARH2, 1T5J, DRAG, and �300 other proteinsfrom many different pro- and eukaryotes but not from plants. Thehits include the JI jellyfish lens crystallins (29) and the zebrafish SelJselenoprotein (30), indicating that the ARH fold may have beenadopted to also serve structural functions. It is interesting to note,however, that neither PSI-BLAST nor secondary-structure predic-tion analyses revealed any sequence or structural resemblance ofPARG to the ARH3�DRAG protein family (28). These findingsindicate that PARG is likely to adopt a fold different from that ofhARH3.

Active Site. The active site of hARH3 is defined by the position oftwo Mg2� ions located in adjacent metal-binding sites, which areonly 3.8 Å apart. Both Mg2� ions are octahedrally coordinated withalmost ideal geometry. Residues contacting the Mg2� ions (Table2) are located on the loop adjacent to �-helix 1, directly before andon �-helix 2, and directly before and on �-helix 17 (Fig. 3A).Asp-300 and one water molecule bridge both metal ions. AlthoughADP cannot be located in the electron-density map of the mono-clinic crystal form, the active sites of the protomers in this crystalform are notably different from the one of the protomer in theorthorhombic crystal form obtained in the absence of ADP (Fig.3B). Asp-300, which lies at the bottom of the metal-binding site, isslightly shifted but still coordinates both Mg2� ions, whereasGlu-25, located on the surface of the molecule, moves by �1.8 Å.All attempts to soak the hARH3 crystals with ADPR, ADP, AMP,ribose-phosphate, ribose, or pyrophosphate resulted in immediatecrystal destruction. Cocrystallization experiments with ADPR,ADP, ribose-phosphate, and ribose did not yield any protein crystalwith ligand bound.

Table 2. Coordination of the two Mg2� ions in the structureof hARH3 (orthorhombic crystal form)

Distance, Å Coordinating atom

Mg-1 2.14 Thr60-OG12.06 Asp61-OD12.01 Asp62-OD21.97 Asp300-OD22.11 Wat3-O2.22 Wat1-O

Mg-2 2.10 Asp300-OD12.12 Glu25-OE22.07 Asp298-OD12.20 Thr301-OG12.12 Wat1-O2.10 Wat2-O

Table 1. Refinement and model statistics for the two hARH3crystal forms

P212121 P21

RefinementResolution limits, Å 40.0–1.60 30.0–2.05No. of reflections

Working set 42,579 42,868Test set 834 852

Rcryst, % 17.5 18.8Rfree, % 20.9 23.1No. of atoms

Protein 2,613 5,186Ions 2 4Water 240 147

StereochemistryR.m.s. deviations

Bonds, Å 0.012 0.023Angles, ° 1.34 1.93

Average B factorsProtein, Å2 24.3 48.6Ions, Å2 11.4 28.2Water, Å2 29.7 45.3

Ramachandran plotMost favored, % 92.9 93.1Allowed, % 6.4 6.1

Mueller-Dieckmann et al. PNAS � October 10, 2006 � vol. 103 � no. 41 � 15027

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

July

5, 2

020

Docking. Docking experiments of hARH3 with ADP-ribosyl–ADPR yielded four significant hits. In all of these, the first ADPRmoiety, the interconnecting ribose, and the first phosphate group ofthe subsequent ADP-ribose moiety are in a virtually identicalposition. The 2�-hydoxyl group of the terminal ribose docked in theactive-site cleft is in close proximity to one of the two Mg2� ions(Fig. 4A), which might act as an electron sink to further increase thesusceptibility of the anomeric carbon atom to a nucleophilic attack.The residues Glu-25, Asp-61, and Asp-298 are located within a 5 Åradius of the anomeric carbon. This ribose moiety and the adjacentpyrophosphate group show most of the interactions with theprotein. Docking experiments with an ADPR monomer alsoyielded significant hits with the terminal ribose moiety slightlyshifted toward the two Mg2� ions and the pyrophosphate group onalmost identical position (data not shown).

Enzyme Activity and Site-Directed Mutagenesis. The results of iso-thermal titration microcalorimetry experiments (ITC, Fig. 5A)show that ADPR binds to hARH3 with an equilibrium dissociationconstant of KD � 1.6 � 0.1 �M and an expected stoichiometry of1:1. No binding was observed under these conditions with ADP

(data not shown). In accord with a previous study (21), hARH3de-ADP-ribosylated PARP-1 and other polyADP-ribosylated tar-get proteins present in a crude preparation of DNA but not argininemonoADP-ribosylated proteins (Fig. 5B). The possible role of theADPR-binding residues identified in the docking experiments wereprobed by site-directed mutagenesis. Mutants E25A, E25Q, D61N,S132A, Y133A, N135A, H166Q, E259A, D298N, D298E, T301A,and T301S were expressed as soluble proteins with similar efficien-cies as wild-type ARH3 in Escherichia coli (Fig. 5C). MutantsD61N, S132A, H166Q, D298E, and T301A lacked detectableactivity, and mutants E25A, E25Q, Y133A, and D298N showedminimal residual activity (Fig. 5D; and see Fig. 9, which is publishedas supporting information on the PNAS web site). Mutants N135Aand T301S showed partial loss of activity, whereas the replacementof residue E259 had little, if any, effect on enzyme activity (Fig. 5D;and see Fig. 9, which is published as supporting information on thePNAS web site). Fig. 4B shows a surface representation of ARH3in which residues are highlighted in red and green, respectively, that,when mutated, do or do not cause inactivation of enzyme activity(this study and ref. 21) (see also Table 4, which is published assupporting information on the PNAS web site, for a summary ofmutant phenotypes).

Fig. 2. Overall structure of ARH3. (A) Ribbon representationof human ARH3. The coloring scheme is from the N terminus(blue) to the C terminus (red). The two magnesium ions in theactive site are shown as cyan spheres. (B) Surface representa-tion of hARH3 with negative and positive electrostatic poten-tials indicated in red and blue, respectively. The orientation isrotated by �90° relative to that in A. (C) Ribbon representa-tion of the quasidomain arrangement. The quasidomains arecolored as follows: A in red (�-helix 1 representing A� and�-helices 16 and 17 A�); B in dark blue (with helices 3 and 4representing B� and helix 19 representing B�); C in orange andD in light blue. Helices that do not account for the scaffold arein gray; the two magnesium ions are in cyan and marked I andII. (D) Topology plot using the same color code as in C. The twoMg2� ions are indicated. �-Helices approximately perpendic-ular to the viewing plane are represented as circles, thoseoriented roughly horizontally in the viewing plane as rectan-gles. Stars show where �-helices have been omitted for clarity.The ill-defined loop (Pro-46–Glu-55) is shown as a dotted line.A schematic diagram of the sequential arrangement of thequasidomains is indicated below the topology plot. (E) Aminoacid sequence and secondary structure elements of hARH3.Amino acids mutated in this study and in a recent study by Okaet al. (21) are highlighted in yellow and gray, respectively.Mg2�-coordinating residues are underlined. At its N terminus,the amino acid sequence of hARH3 is extended by 16–20 aawhen compared with ARH1, ARH2, or DRAG. The 18 N-terminal residues of ARH3 were replaced by two residues (Mand A) for crystallization. Consequently, the numbering ofamino acid residues here deviates by 16 from that used by Okaet al. (21).

15028 � www.pnas.org�cgi�doi�10.1073�pnas.0606762103 Mueller-Dieckmann et al.

Dow

nloa

ded

by g

uest

on

July

5, 2

020

DiscussionThe structure of hARH3 constitutes an archetype of a new proteinstructure family. The binuclear Mg2� center found in the active site(Fig. 2) is a unique feature among the family of glycohydrolases(31). The ARH fold is completely different from that of arginase(PDB ID code 1RLA) (32), which contains a binuclear Mn centerand which has been used in a previous study as a model for DRAG(18). PSI-BLAST analyses indicate that the new ARH fold is widelyspread in nature and may serve various enzymatic (i.e., hydrolysisof protein-ADP-ribosyl and polyADP-ribosyl bonds) as well asstructural functions (i.e., lens crystallins in jellyfish and zebrafish)(29, 30). Importantly, the results of PSI-BLAST, structure predic-tion, and threading analyses do not show any significant similaritiesof the ARH fold to the catalytic domain of PARG, indicating thatthe ARH fold is not a useful model for PARG.

The 3D structure of hARH3 and the proposed binding mode ofthe docked ADPR (Fig. 4) support our own (Fig. 5D) and publishedbiochemical data (18, 21, 33, 34). In the hARH3 model with thedocked ADP-ribosyl–ADPR, one of the Mg2� ions coordinates the2�-hydroxyl group of the ribose carrying the scissile glycosidiclinkage. Conceivably, this Mg2� ion could increase the propensityof the anomeric carbon atom for a nucleophilic attack by acting asan electron sink (Fig. 10, which is published as supporting infor-mation on the PNAS web site). The pivotal role of the binuclearmagnesium cluster for binding the ribose moiety would account forthe metal dependence of ADP-ribosylhydrolase activities (18). The

ARH3 fold also provides a rationale for the observed properties ofARH3, ARH1, and DRAG mutants analyzed here (Fig. 5D) andin previous studies (Table 4). Residues implicated in binding of theproximal ADPR by our docking experiments (Fig. 4) were mutatedand resulted in loss of enzyme activity, except for E259, whosepotential interaction with the distal adenine evidently does notcontribute significantly to binding. Substitution of the Mg2�-coordinating residues Asp-61, Asp-62, and Asp-298 by nonacidicamino acids resulted in a dramatic reduction of enzyme activity ofhARH3 (Fig. 5D and ref. 21), as did mutations of the correspondingresidues Asp-60 and Asp-61 in rat ARH1 and Asp-243 in DRAG(18, 33). In contrast, substitution of Asp-61 of hARH1 by anotheracidic residue (Glu) had only little, if any, effect on enzyme activity,consistent with a retained capacity to bind Mg2� (18, 33). Substi-tution of Thr-301 by Ala also abolished enzyme activity, whereasthe conservative substitution with Ser led to only partial loss ofactivity. The predicted role of residues Ser-132 and His-166 inpyrophosphate binding is consistent with the observed loss ofenzyme activity of hARH3 mutants S132A and H166Q (Fig. 5D),as reported also for mutants of the corresponding residues inDRAG (18). Further, the ARH3 structure accounts for the findingthat mutations of residues located far from the active site had little,if any, effects on enzyme activity, i.e., Glu-222, Glu-223, Glu-245,and Glu-246 in hARH3 (21); His-65, Arg-139, and Asp-285 in ratARH1 (33); and His-142 and Glu-279 in DRAG (18). A recentmutagenesis study of DRAG identified a region containing residuesVal-98, Asn-100, and Cys-102 as a potential surface for regulating

Fig. 3. Active site of ARH3. (A) Coordination of Mg2� ions in the orthorhombic crystal form of hARH3. Hydrogen bonds are represented as dashed lines. (B)Superposition of the Mg2�-coordinating residues of the orthorhombic (gray) and monoclinic crystal forms. Residue Asp-300 is slightly shifted but retains its bidentatebinding character, whereas Glu-25 of the monoclinic crystal form is shifted by 1.8 Å with respect to those of the orthorhombic crystal form. The residues from themonoclinic crystal form are shown in color.

Fig. 4. In silico docking of ADP-ribosyl–ADPR into the active-site crevice of ARH3 and surface view of ARH3 mutants. (A) Hydrogen-bond network of the dockedADP-ribosyl–ADPR molecule. The main part of the interaction takes place at the terminal ribose moiety and at the subsequent pyrophosphate, whereas the rest of themolecule hardly interacts at all. Note that the mobile residue Glu-25 (Fig. 3) and the residues Asp-61 and Asp-298 are positioned close to the anomeric carbon atom.(B) Surface view of ARH3. Mutants of residues highlighted in green showed wild-type enzyme activity. Mutants of residues highlighted in orange or red showed mildlyand strongly reduced enzyme activities, respectively. The region highlighted in blue corresponds to a proposed regulatory region in DRAG (see Discussion).

Mueller-Dieckmann et al. PNAS � October 10, 2006 � vol. 103 � no. 41 � 15029

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

July

5, 2

020

DRAG activity, e.g., by affecting the association of DRAG withmembranes (35). In accord with this observation, the correspondingregion of ARH3 (�-helix 5) is located on the protein surface, nearto but outside of the proposed ADPR-binding region.

Glycosidases typically contain an acidic amino acid residue thatacts as a proton acceptor and thereby activates a water molecule toperform the nucleophilic attack on the scissile glycosidic bond (36).The observed flexibility of residue Glu-25 in the two crystal formsof ARH3 (Fig. 3B) is compatible with a role of this residue inactivating the nucleophilic water. Alternatively, this function mightalso be performed by the nearby acidic residues Asp-61 or Asp-298(see Figs. 3 and 10).

The biological functions of ARH3 and other de-ADP-ribosylating enzymes in mammalian cells remain to be clarified.There is a striking imbalance between the number of ADP-ribosylating and the number of de-ADP-ribosylating enzymes iden-tified in mammals to date: Mammalian genomes typically encode16–17 PARP-related and 4–6 ART2-related ADP-ribosyltrans-ferases, but only three ARH-related and a single PARG-relatedADP-ribosylhydrolases (19, 37). The best characterized polyADP-ribosylpolymerases have been localized to the cell nucleus (PARP1and PARP2, the tankyrases) and�or cytosol (vault PARP andPARP-10); whereas the best characterized mono-ADP-ribosyl-transferases are GPI-anchored (ART1 and ART2) or secretedectoenzymes (ART5) (6, 38). Moreover, biochemical data has beenobtained for mono- and polyADP-ribosylation reactions in mito-chondria (12, 39). The amino acid sequence of ARH3 is extendedN-terminally by 16–20 aa compared with ARH1, ARH2, andDRAG (19). TargetP analyses (40) predict a mitochondrial local-ization for ARH3 and, indeed, an hARH3-GFP fusion proteincolocalizes with mitochondrial markers in transfected cells (data

not shown). These findings indicate that hARH3 might function asa mitochondrial enzyme.

In conclusion, we here report the 3D structure of a protein-de-ADP-ribosylating enzyme. Our results lend support to the notionthat reversible ADP-ribosylation constitutes an important regula-tory mechanism in mammalian cells. The archetype ARH foldpresented here provides a useful basis for modeling the arginine-specific protein ADP-ribosylhydrolases DRAG and ARH1. Finally,the knowledge of the hARH3 structure provides the basis forstructure-guided mutagenesis of ARHs and for the design ofARH-inhibitors.

Materials and MethodsProtein Production and Crystallization. Production, purification, andcrystallization of human ARH3 were performed as described (41).Human PARP-1, ARH1, and ARH3 mutants were expressed asC-terminally His6-tagged proteins in E. coli and were purified byaffinity chromatography on Protino Ni-TED columns (Macherey-Nagel, Duren, Germany). Mouse ART2.2 was expressed as aC-terminally FLAG-tagged protein and purified from E. coliperiplasm by affinity chromatography on M2-matrix (Sigma, St.Louis, MO) as described (42).

Mutational Studies. Site-directed mutagenesis (QuikChange, Strat-agene, La Jolla, CA) was performed according to the manufactur-er’s instructions. Protein solubility in E. coli lysates and duringpurification was taken as an indicator of overall structural integrity,i.e., mutants that segregated into inclusion bodies or that precipi-tated during purification were excluded from further analyses (e.g.,mutants N135H and H166A, see Fig. 5D, lanes 3 and 6).

Structure Determination and Refinement. Diffraction data werecollected on the European Molecular Biology Laboratory beam-

Fig. 5. ADPR binding and de-ADP-ribosylating activity of ARH3 and ARH3 mutants. (A) Isothermal titration calorimetry profile of ADPR titrated into a solutioncontaining hARH3. The curve shows the fit of the data to an equilibrium binding isotherm. The fit provides an equilibrium dissociation constant (KD) for thebinding of ADPR to hARH3 of 1.6 �M. The stoichiometry (n) of the interaction was measured to be n � 0.98 � 0.02 and the binding enthalpy H � 24.6 � 0.6kJ�mol1. (B) Target proteins (M2 antibody, PARP-1, and proteins present in a crude DNA preparation) were mono-ADP-ribosylated by recombinant mouse ART2.2(M2) or polyADP-ribosylated by recombinant human PARP-1. Reaction products were incubated for 60 min at 37°C in the absence (lane 1) or presence of 1 �gof recombinant ARH1 (lane 2) or ARH3 (lane 3) and analyzed by SDS�PAGE autoradiography. (C) Coomassie stain of representative purified, soluble hARH3proteins. Each lane was loaded with protein equivalent from 0.2 ml of E. coli culture. Lanes: 1, (S132A); 2, (Y133A); 3, (N135H); 4, (N135A); 5, (H166Q); 6, (H166A);7, (D298N); 8, (T301A); 9, (hARH1); and 10, (hARH3). (D) De-ADP-ribosylation of polyADP-ribosylated PARP by ARH3 and ARH3 mutants. PolyADP-ribosylated PARPwas incubated with the indicated amounts of wild-type or mutant ARH3 for 30 min at 37°C before SDS�PAGE autoradiography and densitometric quantificationof radiolabeled bands. Mutants are color-coded according to their predicted interaction with the terminal ribose (red), phosphates (blue), distal ribose (pink),and adenine base (green). Wild-type ARH3 and BSA are shown in black.

15030 � www.pnas.org�cgi�doi�10.1073�pnas.0606762103 Mueller-Dieckmann et al.

Dow

nloa

ded

by g

uest

on

July

5, 2

020

lines BW7A and X12 (Deutsches Elektronen-Synchrotron, Ham-burg, Germany). The raw data were reduced by using DENZO,SCALEPACK (43), and TRUNCATE (44). One Xe-derivatizedcrystal, measured at a wavelength of 1.5 Å, was used in a single-wavelength anomalous-diffraction experiment for phase determi-nation (Table 1). Substructure determination, phase calculation,solvent flattening, and model building were performed automati-cally by using SHELXD, MLPHARE, DM, and ARP�wARPwithin the software pipeline AutoRickshaw (45–48). Structurerefinement was done by using the program REFMAC5 (44).

Binding and Enzyme Assays. Isothermal titration binding assays werecarried out at 16°C in a buffer containing 100 mM Tris, pH 8.0, 150mM NaCl, and 2 mM MgCl2 by using a VP-ITC instrument(Microcal, Northampton, MA). Ligands in the injection syringewere at a concentration of 1 mM, and the concentration of hARH3was 0.05 mM. Thirty injections were performed with a 5-min delaybetween injections to allow the baseline to stabilize. To determinethe heat of dilution, the same experiment was carried out by usingonly buffer as an injectant. Data were analyzed with Origin software(OriginLab, Northampton, MA). ADP-ribosylation of the M2antibody by mouse ART2.2 and autoADP-ribosylation of PARP-1in the absence or presence of salmon sperm DNA were performedin 50 �l of PBS at 37°C. Reactions were initiated by the addition of1 �M [32P]NAD [200 Ci�mmol (1 Ci � 37 GBq)]. After 40 min,ADP-ribosylated proteins were separated from unincorporatednucleotides by gel-filtration chromatography. Purified humanARH1 or human ARH3 were added at the indicated concentra-tions, and incubations were continued for 30–60 min at 37°C in PBScontaining 10 mM MgCl2 and 5 mM DTT. Reactions were stoppedby the addition of SDS�PAGE sample buffer. Reaction productswere analyzed by SDS�PAGE autoradiography.

Docking Studies. The 3D ligand input file for ADP-ribosyl–ADP-ribose was generated with Sybyl (Sybyl 7.0; Tripos, St. Louis, MO)and optimized by energy minimization using the Conjugate Gra-dients method and the Tripos force field until the convergencecriterion of 0.05 kcal�(mol Å) was reached. The FlexX dockingprogram (49) interfaced within Sybyl was used to dock the ligandto the active site of hARH3 to examine possible binding modes ofADPR. FlexX takes account of ligand conformational flexibility byan incremental fragment-placing technique. The active site ofhARH3 was defined as all amino acids within 8-Å proximity of theresidues Ser-22, Glu-25, Pro-46, Glu-55, Tyr-58, Gly-101, Ser-132,Phe-133, His-166, and Glu-255. All resulting docking conforma-tions were ranked according to the CScore, and the best rankeddocking conformations (CScore 5) were checked visually.

Database Searches and Structure Analyses. Position-sensitive itera-tive BLAST searches were performed with PSI-BLAST (50),secondary structure-matching searches with SSM [European Bioin-formatics Institute (EBI), Hinxton, Cambridge, U.K.] andPSIPRED (51, 52). Amino acid sequence alignments were per-formed with ClustalW and T-Coffee (53, 54), and structure-basedsequence alignments were performed with STAMP (55). Themitochondrial targeting signal sequence was predicted with Tar-getP (40). Structure models were superimposed with LSQMAN(56). Ribbon plots and hydrogen-bond presentations were preparedwith MOLSCRIPT and Raster3D (57, 58) and surface presenta-tions with GRASP (33) and PyMOL (44, 59).

This work was supported by grants from the Deutsche Forschungsge-meinschaft (to F.K.-N., F.H., and M.S.W) and the Studienstiftung desDeutschen Volkes (S.K.).

1. Haag F, Koch-Nolte F (1997) ADP-Ribosylation in Animal Tissues: Structure,Function and Biology of Mono(ADP-Ribosyl)transferases and Related Enzymes(Plenum, New York).

2. Aktories K, Just I (2000) Bacterial Protein Toxins (Springer, Berlin).3. Rich T, Allen RL, Wyllie AH (2000) Nature 407:777–783.4. Ziegler M (2000) Eur J Biochem 267:1550–1564.5. Smith S (2001) Trends Biochem Sci 26:174–179.6. Corda D, Di Girolamo M (2003) EMBO J 22:1953–1958.7. Seman M, Adriouch S, Scheuplein F, Krebs C, Freese D, Glowacki G, Deterre

P, Haag F, Koch-Nolte F (2003) Immunity 19:571–582.8. Berger F, Ramirez-Hernandez MH, Ziegler M (2004) Trends Biochem Sci 29:111–118.9. Chang P, Jacobson MK, Mitchison TJ (2004) Nature 432:645–649.

10. Kim MY, Zhang T, Kraus WL (2005) Genes Dev 19:1951–1967.11. Aktories K, Barbieri JT (2005) Nat Rev Microbiol 3:397–410.12. Herrero-Yraola A, Bakhit SM, Franke P, Weise C, Schweiger M, Jorcke D,

Ziegler M (2001) EMBO J 20:2404–2412.13. Takada T, Okazaki IJ, Moss J (1994) Mol Cell Biochem 138:119–122.14. Ludden PW, Roberts GP (1989) Curr Top Cell Regul 30:23–56.15. Lin W, Ame JC, Aboul-Ela N, Jacobson EL, Jacobson MK (1997) J Biol Chem

272:11895–11901.16. Pope MR, Murrell SA, Ludden PW (1985) Proc Natl Acad Sci USA 82:3173–3177.17. Fu H, Burris RH, Roberts GP (1990) Proc Natl Acad Sci USA 87:1720–1724.18. Antharavally BS, Poyner RR, Zhang Y, Roberts GP, Ludden PW (2001) J Bacteriol

183:5743–5746.19. Glowacki G, Braren R, Firner K, Nissen M, Kuhl M, Reche P, Bazan F, Cetkovic-Cvrlje

M, Leiter E, Haag F, Koch-Nolte F (2002) Protein Sci 11:1657–1670.20. Moss J, Stanley SJ, Nightingale MS, Murtagh JJ, Monaco L, Mishima K, Chen

HC, Williamson KC, Tsai SC (1992) J Biol Chem 267:10481–10488.21. Oka S, Kato J, Moss J (2006) J Biol Chem 281:705–713.22. Sixma TK, Pronk SE, Kalk KH, Wartna ES, van ZB, Witholt B, Hol WG (1991) Nature

351:371–377.23. Choe S, Bennett MJ, Fujii G, Curmi PM, Kantardjieff KA, Collier RJ, Eisenberg D

(1992) Nature 357:216–222.24. Ruf A, Mennissier de Murcia J, de Murcia G, Schulz GE (1996) Proc Natl Acad Sci USA

93:7481–7485.25. Han S, Craig JA, Putnam CD, Carozzi NB, Tainer JA (1999) Nat Struct Biol 6:932–936.26. Mueller-Dieckmann C, Ritter H, Haag F, Koch-Nolte F, Schulz G (2002) J Mol Biol

322:687–696.27. Jorgensen R, Merrill AR, Yates SP, Marquez VE, Schwan AL, Boesen T,

Andersen GR (2005) Nature 436:979–984.

28. Patel CN, Koh DW, Jacobson MK, Oliveira MA (2005) Biochem J 388:493–500.29. Piatigorsky J, Horwitz J, Norman BL (1993) J Biol Chem 268:11894–11901.30. Castellano S, Lobanov AV, Chapple C, Novoselov SV, Albrecht M, Hua D, Lescure

A, Lengauer T, Krol A, Gladyshev VN, Guigo R (2005) Proc Natl Acad Sci USA102:16188–16193.

31. Davies GJ, Gloster TM, Henrissat B (2005) Curr Opin Struct Biol 15:637–645.32. Kanyo ZF, Scolnick LR, Ash DE, Christianson DW (1996) Nature 383:554–557.33. Konczalik P, Moss J (1999) J Biol Chem 274:16736–16740.34. Takada T, Iida K, Moss J (1993) J Biol Chem 268:17837–17843.35. Kim K, Zhang Y, Roberts GP (2004) FEBS Lett 559:84–88.36. Vasella A, Davies GJ, Bohm M (2002) Curr Opin Chem Biol 6:619–629.37. Otto H, Reche PA, Bazan F, Dittmar K, Haag F, Koch-Nolte F (2005) BMC Genomics

6:139.38. Ame JC, Spenlehauer C, de Murcia G (2004) BioEssays 26:882–893.39. Du L, Zhang X, Han YY, Burke NA, Kochanek PM, Watkins SC, Graham SH,

Carcillo JA, Szabo C, Clark RS (2003) J Biol Chem 278:18426–18433.40. Emanuelsson O, Nielsen H, Brunak S, von Heijne G (2000) J Mol Biol 300:1005–1016.41. Kernstock S, Koch-Nolte F, Mueller-Dieckmann J, Weiss MS, Mueller-

Dieckmann C (2006) Acta Crystallogr F 62:224–227.42. Koch-Nolte F, Petersen D, Balasubramanian S, Haag F, Kahlke D, Willer T,

Kastelein R, Bazan F, Thiele HG (1996) J Biol Chem 271:7686–7693.43. Otwinowski Z, Minor W (1997) Methods Enzymol 276:307–326.44. CCP4 (1994) Acta Crystallogr D 50:760–763.45. Schneider TR, Sheldrick GM (2002) Acta Crystallogr D 58:1772–1779.46. Perrakis A, Morris R, Lamzin VS (1999) Nat Struct Biol 6:458–463.47. Cowtan KD, Zhang KY (1999) Prog Biophys Mol Biol 72:245–270.48. Panjikar S, Parthasarathy V, Lamzin VS, Weiss MS, Tucker PA (2005) Acta Crystallogr

D 61:449–457.49. Rarey M, Kramer B, Lengauer T, Klebe G (1996) J Mol Biol 261:470–489.50. Altschul SF, Koonin EV (1998) Trends Biochem Sci 23:444–447.51. McGuffin LJ, Bryson K, Jones DT (2000) Bioinformatics 16:404–405.52. Krissinel E, Henrick K (2004) Acta Crystallogr D 60:2256–2268.53. Thompson JD, Higgins DG, Gibson TJ (1994) Nucleic Acids Res 22:4673–4680.54. Notredame C, Higgins DG, Heringa J (2000) J Mol Biol 302:205–217.55. Russell RB, Barton GJ (1992) Proteins 14:309–323.56. Kolodny R, Koehl P, Levitt M (2005) J Mol Biol 346:1173–1188.57. Esnouf RM (1997) J Mol Graphics Model 15:132–134, 112–113.58. Merritt EA, Murphy ME (1994) Acta Crystallogr D 50:869–873.59. DeLano WL (2002) The PyMOL User’s Manual (DeLano Scientific, San Carlos, CA).

Mueller-Dieckmann et al. PNAS � October 10, 2006 � vol. 103 � no. 41 � 15031

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

July

5, 2

020