Crystal Structure of the Productive Ternary Complex ofDihydropyrimidine Dehydrogenase with NADPH and 5-IodouracilIMPLICATIONS FOR MECHANISM OF INHIBITION AND ELECTRON TRANSFER*

Received for publication, December 13, 2001, and in revised form, January 16, 2002Published, JBC Papers in Press, January 16, 2002, DOI 10.1074/jbc.M111877200

Doreen Dobritzsch‡§, Stefano Ricagno‡¶, Gunter Schneider‡, Klaus D. Schnackerz�,and Ylva Lindqvist‡**

From the ‡Division of Molecular Structural Biology, Department of Medical Biochemistry and Biophysics, KarolinskaInstitutet, S-17177 Stockholm, Sweden and �Theodor-Boveri-Institut fur Biowissenschaften, Physiologische Chemie I,Am Hubland, D-97074 Wurzburg, Germany

Dihydroprymidine dehydrogenase catalyzes the firstand rate-limiting step in pyrimidine degradation byconverting pyrimidines to the corresponding 5,6-dihydro compounds. The three-dimensional structuresof a binary complex with the inhibitor 5-iodouracil andtwo ternary complexes with NADPH and the inhibitors5-iodouracil and uracil-4-acetic acid were determinedby x-ray crystallography. In the ternary complexes,NADPH is bound in a catalytically competent fashion,with the nicotinamide ring in a position suitable forhydride transfer to FAD. The structures provide acomplete picture of the electron transfer chain fromNADPH to the substrate, 5-iodouracil, spanning a dis-tance of 56 Å and involving FAD, four [Fe-S] clusters,and FMN as cofactors. The crystallographic analysisfurther reveals that pyrimidine binding triggers aconformational change of a flexible active-site loop inthe �/�-barrel domain, resulting in placement of acatalytically crucial cysteine close to the boundsubstrate. Loop closure requires physiological pH,which is also necessary for correct binding of NADPH.Binding of the voluminous competitive inhibitor uracil-4-acetic acid prevents loop closure due to sterichindrance. The three-dimensional structure of theternary complex enzyme-NADPH-5-iodouracil supportsthe proposal that this compound acts as a mechanism-based inhibitor, covalently modifying the active-siteresidue Cys-671, resulting in S-(hexahydro-2,4-dioxo-5-pyrimidinyl)cysteine.

Dihydropyrimidine dehydrogenase (DPD,1 EC 1.3.1.2) is acytosolic enzyme catalyzing the NADPH-dependent reductionof uracil and thymine to the corresponding 5,6-dihydropyrimi-

dines, the first and rate-limiting reaction in the three-steppathway of pyrimidine degradation (1). In mammals, this is theonly pathway leading to the synthesis of the putative neuro-transmitter �-alanine (2).

The human enzyme has recently evolved as an adjunct targetfor anticancer drug design because it also degrades 5-fluoro-uracil (5FU), one of the most widely prescribed chemothera-peutic agents for treatment of many common malignancies (3).5FU primarily targets the enzyme thymidylate synthase andthereby interferes with DNA synthesis (4). Approximately 85%of the administered dose is rapidly degraded by DPD to ther-apeutically inactive but toxic fluorinated products (5, 6). Fur-thermore, accurate determination of an optimal drug doseproved to be very difficult due to substantial differences in theactivity of DPD among individuals (7, 8). Severe life-threaten-ing toxicities have been reported after treatment of cancerpatients with inherited DPD deficiency with standard doses of5FU (9–11). The catabolism of 5FU via DPD and the other twoenzymes of the pyrimidine degradation pathway thus repre-sents a major determinant of the pharmacokinetics of thisanticancer drug. Its inhibition may result in increased 5FUefficacy, administration of lower doses, and diminished drugside effects. To date, several DPD inhibitors (eniluracil,5-chloro-2,4-dihydroxypyridine, and 3-cyano-2,6-dihydropyri-dine) are utilized or under evaluation as modulators of 5FUtreatment (12).

The high sequence identities (�90%) between human (13)and other mammalian DPD, such as bovine (14) and pig (13),suggest very similar reaction mechanisms and three-dimen-sional structures. Recently, the x-ray structure of the best-characterized mammalian enzyme, recombinant pig liver DPD,has been determined to a resolution of 1.9 Å (15). The enzymeis a homodimer of 2 � 111 kDa. Each subunit of 1025 aminoacids carries one FAD, one FMN, and four [4Fe-4S] clusters(16). According to the non-classical two-site ping-pong kineticmechanism, the enzyme contains separate binding sites for theelectron-donating cosubstrate NADPH and the electron-accept-ing pyrimidines, respectively (17).

The pH dependence of the kinetic parameters suggests thefollowing reaction mechanism for the reduction of uracil/thy-mine by pig liver DPD (17). NADPH binds to site 1 with thepro-S side of the nicotinamide facing toward the FAD-isoallox-azine ring. Hydride is transferred from the C-4 of NADPH toN-5 of FAD, leaving NADP� and FADH � . There is evidencefrom kinetic isotope effects that electron transfer to site 2 viathe [4Fe-4S] clusters does not occur until NADP� is released.At site 2 reduced FMN is formed. The pyrimidine substrate isbound with the si-face at the C-6 atom directed toward the N-5

* This work was supported by grants from the Swedish Cancer Foun-dation and the Swedish Research Council. The costs of publication ofthis article were defrayed in part by the payment of page charges. Thisarticle must therefore be hereby marked “advertisement” in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.

The atomic coordinates and structure factors (code 1gte (DPD�5IU), 1gth(DPD�5IU�NADPH), and 1gt8 (DPD�UAA�NADPH)) have been deposited inthe Protein Data Bank, Research Collaboratory for Structural Bioinfor-matics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).

§ This author gratefully acknowledges a fellowship from the Wenner-Gren Foundation.

¶ This author gratefully acknowledges a fellowship from the Found-ation Blanceflor Boncompagni-Ludovisi.

** To whom correspondence should be addressed. E-mail: ylva@alfa.mbb.ki.se.

1 The abbreviations used are: DPD, dihydropyrimidine dehydrogen-ase; 5FU, 5-fluorouracil; 5IU, 5-iodouracil; UAA, uracil-4-acetic acid.

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 277, No. 15, Issue of April 12, pp. 13155–13166, 2002© 2002 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www.jbc.org 13155

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

of FMN and with the si-face of C-5 directed toward an enzymegeneral acid, which has been identified as the thiol group ofcysteine 671 (16). The transfer of hydride from reduced FMNand the proton from Cys-671 is proposed to occur in a concertedanti-addition reaction.

The three-dimensional structure of pig liver DPD (15) re-vealed a highly modular organization of the subunit. It consistsof five distinct domains, each carrying a subset of the variousprosthetic groups. The �-helical domain I (residues 27–172)contains two of the iron-sulfur clusters, nFeS1 and nFeS2, ofwhich the latter shows an unusual cluster coordination com-prising one glutamine and three cysteine residues. The Ross-mann-type nucleotide binding folds of domains II (residues173–286, 442–524) and III (residues 287–441) bind FAD andNADPH, respectively. The interface between these domainsforms site 1 of DPD, whereas the pyrimidine binding site 2,containing FMN, is located on top of the �8/�8-barrel domain IV(residues 525–847). The remaining iron-sulfur clusters cFeS1and cFeS2 are bound to the core of the C-terminal domain V(residues 1–26 and 848–1025). The three-dimensional arrange-ment of the distinct domains in the homodimeric enzyme leadsto the formation of two electron-transfer chains, in which theelectrons are transferred from the FAD to nFeS2 and sub-sequently to nFeS1. The remaining gap between nFeS1 and theFMN is closed by the clusters of the C-terminal domain V ofthe other subunit in the dimer. This domain swapping makesthe dimer the minimal catalytic unit of pig liver DPD.

In addition to the x-ray structure of ligand-free DPD, that ofa catalytically inactive mutant (C671A) in complex withNADPH and the anti-cancer drug 5FU was determined to 2.0-Åresolution (15). Two observations led to the conclusion that thisternary complex does not represent a productive enzyme-sub-strate complex. First, the nicotinamide ring of NADPH was notplaced in a position suitable for hydride transfer from its C4atom to the FAD N-5 atom. Second, the loop comprising resi-dues 670–682 (referred to as “active site loop”), which is pro-posed to close over the active site while catalysis occurs, is seenin an open and partially disordered conformation. Hence, thealanine at position 671, which in the mutant replaces thegeneral acid Cys-671, was located too distant from the C5 atomof 5FU.

Possible explanations for the failure to trigger the putativeactive-site closure upon substrate binding are the mutation ofthe cysteine to alanine or the pH difference between crystalli-zation conditions (4.7) and optimal catalytic activity (�7.3).The assumption that pH might influence the loop conformationis based on the presence of a histidine residue at position 673,which most likely participates in substrate binding. This his-tidine residue should be positively charged at pH 4.7 but neu-tral at the pH optimal for catalytic activity.

To address this question, we prepared crystals of wild-typepig liver DPD with NADPH and two different inhibitors, 5-

iodouracil (5IU) and uracil-4-acetic acid (UAA), respectively.While UAA is a competitive inhibitor of DPD (apparent Ki �78.6 � 20.2 �M) (18), 5IU serves as a substrate and is convertedto 5-iodo-5,6-dihydrouracil, which has been shown to be a po-tent alkylating agent able to covalently modify the Cys-671-thiol group (19). This may lead to trapping of the active-siteloop in its closed conformation. Additionally and for purposes ofcomparison, the structure of the binary DPD�5IU complex wasdetermined. In this report, we describe the three-dimensionalstructures of these inhibitor-enzyme complexes and discusstheir implication for the mechanism of DPD.

MATERIALS AND METHODS

Protein Purification and Crystallization

Pig liver DPD was produced as recombinant protein in Escherichiacoli, and has been purified and crystallized as described previously (20,21). The binary complex of DPD was obtained by crystallizationof wild-type DPD in the presence of 1 mM 5IU, the ternaryDPD�5IU�NADPH complex, by cocrystallization of DPD with 1 mM 5IUand 5 mM NADPH. A change of pH in the crystals was achieved bysoaking crystals of the complex DPD�5IU�NADPH in 100 mM Hepes (pH7.5), 22% polyethylene glycol 6000 (w/v), 0.8 mM NADPH, for 20 minbefore data collection. The complex DPD�UAA�NADPH was prepared bysoaking wild-type DPD crystals for 2.5 h in a solution containing 100mM Hepes (pH 7.5), 22% polyethylene glycol 6000 (w/v), 1 mM UAA, 2.5mM NADPH. The crystals were stable under these conditions and didnot change their appearance. All DPD complexes crystallized in spacegroup P21, with four molecules (two homodimers) in the asymmetricunit.

Data Collection

Before data collection, DPD�5IU crystals were transferred for 5 mininto a cryo solution containing 100 mM sodium citrate (pH 4.7), 22%(w/v) polyethylene glycol 6000, and 20% (v/v) glycerol. For theDPD�5IU�NADPH crystal, the buffer component of this cryo solutionwas replaced by 100 mM Hepes (pH 7.5). The crystal of theDPD�UAA�NADPH complex was soaked in a cryo solution consistentwith that used for ligand soak/pH change, but 20% (v/v) glycerol wereadded. All crystals were flash-frozen in a nitrogen gas stream.

X-ray diffraction data were collected at 100 K at beam line ID-EH3 ofthe European Synchrotron Radiation Facility (European SynchrotronRadiation Facility, Grenoble, France) with a MAR CCD detector for thetwo ternary complexes and at the EMBL beam line BW7B at the DORISstorage ring, Deutsches Elektronensynchrotron (Hamburg/Germany)with a MARResearch Imaging Plate for the binary complex. Data wereindexed and integrated using DENZO (22) and scaled with the CCP4suite of programs (23). Table I gives the details of the data collectionstatistics.

Structure Determination and Refinement

DPD�5IU (pH 4.7)—Initial rigid body refinement using the refinedstructure of the ligand-free enzyme (without water molecules) as amodel and data to 2.5-Å resolution resulted in values of 27.9% for Rcryst

and 28.8% for Rfree, respectively. The �Fo� � �Fc� map computed from thisinitial model showed clear electron density for 5IU bound close to thecofactor FMN. Models for the inhibitor and water molecules wereadded. The high resolution of the data allowed the identification ofseveral alternative conformations of amino acid side chains (residues

TABLE IData collection statistics

DPD�5IU DPD�5IU�NADPH DPD�UAA�NADPH

Cella (Å) 81.7 82.2 82.0b (Å) 158.4 159.7 159.5c (Å) 162.3 167.6 166.0� (°) 95.8 97.1 97.2Resolution range (Å) 25.0–1.65(1.74–1.65) 30.0–2.25(2.37–2.25) 30.0–3.30(3.36–3.30)Unique reflections 481,684 199,249 62,722Multiplicitya 3.2 (2.7) 3.6 (3.4) 3.5 (3.2)Completenessa (%) 98.1 (94.3) 98.3 (97.4) 98.3 (92.9)Rsym

a (%) 4.7 (23.8) 6.4 (25.0) 9.8 (46.3)I/�a 15.0 (4.3) 14.3 (5.2) 9.8 (2.6)

a Values for the last resolution shell are given in parentheses.

Ternary Complex of DPD with NADPH and 5-Iodouracil13156

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

187, 245, 947, 962 in chain A, residues 14, 167, 410, 577, 932 in chainB, residues 167, 245, 281, 368, 410, 577, 779, 858, 932, 947, 1006 inchain C, and residues 14, 410, 775, 779, 858, 947, 962, 1006 in chain D)that were included into the model. Model building and refinementresulted in Rcryst � 18.3% and Rfree � 19.8% (Table II). The final modelcontains residues 2–674, 681–901, and 908–1017 for chain A, residues2–673, 680–901, and 908–1018 for chain B, residues 2–675 and 682–1017 for chain C, and residues 2–902 and 907–1019 for chain D, fourFAD, four FMN, 16 [4Fe-4S] clusters, four 5IU, and 4849 water mole-cules. All missing amino acids belong either to disordered, surface-located loop regions (such as the active-site loop and the “proline-richloop” (15)) or to the mobile C-terminal tail.

DPD�5IU�NADPH (pH 7.5)—Initial rigid body refinement using thesame model as for DPD�5IU did not result in reasonable R values or ininterpretable electron density maps. Therefore, molecular replacementwas performed with the program AMoRe (24), again with the structureof the ligand-free DPD without water molecules as the search model,resulting in a solution with a correlation coefficient of 78.3. After onecycle of crystallographic refinement (including rigid body refinement,annealing, minimization, and individual B-factor refinement), values of24.1% for Rcryst and 26.6% for Rfree were obtained, and 2�Fo� � �Fc� and�Fo� � �Fc� maps were calculated. The maps showed clear electron den-sity for NADPH, with its nicotinamide ring bound close to the FAD inall four molecules in the asymmetric unit, four 5IU molecules, or inhib-itor derivatives (see “Results”) bound in the pyrimidine binding sites aswell as a change in active-site-loop conformation. Inhibitor models(5IU) and NADPH molecules were fitted into the electron density. Theactive-site-loop residues were rebuilt into the observed density, andminor readjustments of several amino acid side chains were necessary.The final model contained residues 2–1020 for chains A and D, residues2–900 and 908–1020 for chain B, residues 2–902 and 907–1020 forchain C, all cofactors, one 5-iodo-5,6-dihydrouracil molecule (chain B),one 5IU molecule (chain C), one uracil molecule (chain D), four NADPH,and 3073 water molecules. In chain A, Cys-671 is covalently modified toS-(hexahydro-2,4-dioxo-5-pyrimidinyl) cysteine. Because the activesites of molecules B and C do not exclusively contain 5IU and 5-iodo-5,6-dihydrouracil, respectively, but also uracil or 5,6-dihydrouracil (see“Results”), the occupancies for the iodine atoms of both molecule modelswere adjusted correspondingly.

DPD�UAA�NADPH (pH 7.5)—For initial rigid body refinement theAMoRe solution for the DPD�5IU�NADPH ternary complex (after re-moval of inhibitor, cosubstrate, and water molecules) was used, result-ing in R values of 29.6% (Rcryst) and 29.7% (Rfree). Because the initial�Fo� � �Fc� map indicated binding of NADPH and UAA to the active sitesof all four molecules in the asymmetric unit, models of cosubstrate andinhibitor were added to the structure. Due to the low resolution of thedata, grouped instead of individual B-factor refinement was carried out.

The final model contains residues 2–677 and 680–1018 for chain A,residues 2–676, 680–900, and 907–1018 for chain B, residues 2–672and 682–1016 for chain C, and residues 2–676, 680–901, and 906–1018for chain D. R-factors were refined to 21.7% for Rcryst and 27.3% forRfree, respectively.

All model building was carried out in O (25), the refinement andwater-picking routine for the complexes was done using the programCNS (26). NCS restraints were used during refinement, with looserrestraints for side-chain atoms. Excluded from these NCS restraintswere residues 51–54, 264, 323–327, 397–404, 415–417, 671–683, and899–907, which belong to solvent-exposed loop regions, showingslightly different conformations in the four molecules present in theasymmetric unit. All three models have good stereochemistry, as deter-mined by the program PROCHECK (27). Refinement statistics aregiven in Table II.

Structural alignments were achieved with the programs TOP (28) orthe LSQ commands in O (25). All figures were generated usingBOBSCRIPT (29, 30) and RASTER3D (31). Crystal contacts were de-termined with CONTACT of the CCP4 suite of programs (23). Thecrystallographic data have been deposited in the Protein Data Bank,with accession codes 1gte (DPD�5IU), 1gth (DPD�5IU�NADPH), and1gt8 (DPD�UAA�NADPH).

RESULTS

Effects of pH Change on Crystal Packing—The pH changefrom 4.7 to 7.5 did not have any obvious effects on the appear-ance of the DPD crystals. Nevertheless, it caused significantchanges in unit cell dimensions and packing of the moleculeswithin the crystal. Both ternary complexes obtained at pH 7.5showed an elongation of unit cell dimension c by �4 Å. Thiswas accompanied by a change in packing of the two tightlyassociated homodimers within the asymmetric unit, leading tosubstantial alterations in dimer-dimer and crystal contacts.The majority of the contacts observed in unliganded DPD areabsent in DPD�5IU�NADPH (pH 7.5) and DPD�UAA�NADPH(pH 7.5), and new but fewer contacts are formed.

Overall Structure of the Complexes—The binding of the py-rimidine analogs at pH 4.7 did not introduce major changes inthe protein backbone structure, as reflected by a root meansquare deviation of 0.24 Å, measured for all C� atoms of thesubunits of ligand-free DPD and the binary complex DPD�5IU.For the ternary complexes obtained at pH 7.5, the superposi-tion with ligand-free DPD yielded root mean square deviationsof 0.40 Å for all C� atoms.

In complex DPD�5IU�NADPH (pH 7.5), but not in complexDPD�UAA�NADPH (pH 7.5), the active-site loop was observedin its closed conformation. Differences to ligand-free DPD oc-curring in both ternary complexes are located in the NADPH-binding site, where several amino acids adopt different confor-mations to allow proper NADPH binding, as described below.In addition, a slight movement of the NADPH binding domainwith respect to the domain arrangement for unliganded DPDwas observed. Superposition of the �8/�8-barrel domain IV ofthe corresponding structures reveals that the NADPH bindingdomain III changes its position by a modest movement (maxi-mum difference in C�-coordinates is 2.7 Å for residue 415) withrespect to the FAD binding domain II, resulting in a slightwidening of the NADPH binding cleft (Fig. 1). The displace-ment corresponds to a rotation by approximately 2° and atranslation along the rotation axis by 0.7 Å.

NADPH-binding Site—Binding of NADPH in the cleftformed between domains II and III was accompanied by onlylocal changes of amino acid conformations within domain II,whereas it involved both local conformational changes andmore global rigid body movements of residues originating fromdomain III (Fig. 2). A reorientation of the side chains of Phe-438 and Arg-364 resulted in stacking interactions of theseresidues with the NADPH adenine moiety. Residues 485–488,comprising a loop and the first residue of the following �-helixof domain II, moved in all NADPH-containing complexes with

TABLE IIRefinement statistics

Final models DPD�5IU DPD�5IU�NADPH DPD�UAA�NADPH

Rcryst (%) 18.1 17.7 21.7Rfree (%) 19.7 20.8 27.3B-factor from

Wilson plot17.7 29.9 71.7

Average Bvalue (Å2)

Protein atomsa 21.1 (30790) 32.9 (30993) 50.4 (30884)Cofactor atomsa 14.9 (464) 31.7 (656) 39.1 (656)Inhibitor atoms 19.8 (36) 34.6 (40) 124.5 (48)H2O moleculesa 35.1 (4848) 37.4 (3081) � (�)Root mean

squaredeviation

Bond length 0.0086 0.0090 0.0115Bond angles 1.38 1.37 1.48Ramachandran

plot,Residues (%) in

mostfavorableregions

89.6 89.1 86.2

Additionalallowedregions

9.8 10.6 13.4

Disallowedregions

0.0 0.0 0.0

a The total number of atoms is given in parentheses.

Ternary Complex of DPD with NADPH and 5-Iodouracil 13157

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

respect to their position in ligand-free DPD. The observeddisplacement of C� atoms was maximal for Ala-486 (�1.7 Å).The position of the side chain of Asn-487 as observed in ligand-

free DPD would cause steric clashes with a bound NADPHmolecule. By rotation around the N-C� and C�-C bonds ofAla-486, a shift of the Asn-487 side chain by 5.6 Å measured for

FIG. 1. Stereo view of the superim-posed subunits of ligand-free DPDand its ternary complex with 5IU andNADPH (pH 7.5). The superposition isbased on the �8/�8-barrel domains IV. ForDPD�5IU�NADPH (pH 7.5), the N-termi-nal FeS-cluster domain I is shown ingreen, the FAD binding domain II isshown in yellow, the NADPH binding do-main III is shown in orange, the FMNbinding domain IV is shown in red, andthe C-terminal FeS binding domain V isshown in blue. The subunit structure ofligand-free DPD is black. Cofactor mole-cules, NADPH, and 5IU are shown asball-and-stick models. The arrow at theNADPH binding domains highlights theregion with the largest conformationalchanges in this domain upon formation ofthe ternary complex.

FIG. 2. NADPH binding to DPD atpH 7.5. A, stereo view of the superim-posed NADPH-binding sites of ligand-freeDPD (dark blue) and DPD�5IU�NADPH(pH 7.5) (cyan). Amino acid side chainsdirectly involved in NADPH binding orsignificantly changing position or confor-mation upon NADPH binding are labeledand shown as ball-and-stick models. Forligand-free DPD, the cofactor FAD isshown in dark blue. NADPH and FADbound to DPD�5IU�NADPH (pH 7.5) aregiven in cyan. B, stereo view of NADPH asbound in complex DPD�5IU�NADPH (pH7.5) together with the �Fo� � �Fc� map at 3�-contour level calculated after omittingthe NADPH atoms from the model.

Ternary Complex of DPD with NADPH and 5-Iodouracil13158

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

the carboxamide nitrogen is achieved, which then interactedwith the 3-hydroxyl of the nicotinamide-ribose via a hydrogenbond. The new backbone conformation of stretch 485–488 isstabilized by two new hydrogen bonds formed between themain chain nitrogen atoms of Asn-487 and Thr-488 to thecarboxyl-group of Glu-491. The most significant conformationalchange necessary for the proper positioning of the nicotinamidemoiety of NADPH involves Asp-342. In all complexes obtainedat pH 4.7, its side chain partially occupies the space for thenicotinamide of NADPH when bound in productive fashion. Inligand-free DPD a carboxylate oxygen is at a 3.6-Å distancefrom the FAD N5 at a position very close to the C4 atom ofNADPH in complex DPD�5IU�NADPH (pH 7.5). After NADPHbinding the C� of Asp-342 is displaced by 1.9 Å, the NADPH-nicotinamide moiety is stacked between the FAD isoalloxazinering and the side chain of Asp-342, and one of the carboxyloxygens of Asp-342 is hydrogen-bonded to the backbone amideof Val-373. Additionally, this movement of Asp-342 causes adisplacement of residues 340–344, resulting in disruption ofthe hydrogen bonds between the backbone oxygen of Ala-340and the backbone nitrogen atoms of Arg-371 and Ala-372,which are found in ligand-free DPD, the DPD�5IU (pH 4.7)complex, and also in the previously reported complexDPD�5FU�NADPH (pH 4.7) (15).

In the latter, the mode of NADPH binding was non-produc-tive, because the nicotinamide ring was partially disorderedand flipped away from the FAD isoalloxazine ring onto theprotein surface. In that complex no domain rearrangementcould be observed. Nevertheless, most of the more local alter-ations in the conformation of residues lining the walls of theNADPH binding pocket are seen in DPD�5FU�NADPH (pH 4.7).

Ligands in the Pyrimidine-binding Site—For all complexesof DPD, electron density shows the inhibitor molecules boundalmost parallel to the FMN isoalloxazine ring plane, replacingseveral water molecules found in the active site of the unligan-ded enzyme (Fig. 3A).

In DPD�5IU (pH 4.7), the electron density is well defined forall atoms of 5-iodouracil (Fig. 3C). The binding geometrycorresponds to that observed for 5-fluorouracil in theDPD�5FU�NADPH structure (15) (Fig. 3B), with the C6 atom ofthe inhibitor closest to the N5 of FMN (3.8 Å), the O2 atom ina 3.7-Å distance to both, the N10 and C9A of FMN, and theiodine located 3.7 Å from the FMN O4. As already noted for5FU, there are no hydrogen bonds between amino acid sidechains and the 5-halogen atom.

The ligand-associated electron density seen in DPD�UAA�

NADPH (pH 7.5) is large enough to fit an uracil-4-acetic acidmolecule, but due to the low resolution of the data, the bindinggeometry of the inhibitor cannot unambiguously be determinedand will therefore not be discussed in further detail (Fig. 3D).

The electron density observed for inhibitor molecules boundin the active sites of the DPD�5IU�NADPH (pH 7.5) complex isheterogeneous and differs between each of the four moleculesin the asymmetric unit (Fig. 4, A–D). Furthermore, there areambiguities in assignment of the observed electron density todistinct inhibitor derivatives for these active sites. In the fol-lowing, we will therefore describe only the features of the mostprominent inhibitor derivative species (with the highest occu-pancy) for each active site.

In chain A, no density was observed for the iodine atom, butthere was continuous density between the sulfhydryl group ofCys-671 and the C5 atom of the pyrimidine ring. This electrondensity suggests that in subunit A the expected enzymaticreduction of 5-iodouracil occurred, and the reduction product,5-iodo-5,6-dihydrouracil, subsequently attacked Cys-671, lead-

ing to its covalent modification to S-(hexahydro-2,4-dioxo-5-pyrimidinyl)cysteine.

The active sites of molecules B and C contain mainly uraciland 5,6-dihydrouracil, respectively, but also some 5-iodouracil(chain C) or 5-iodo-5,6-dihydrouracil (chain B). It is possible todistinguish between the substrate (5IU) and the product of theenzymatic reduction (5-iodo-5,6-dihydrouracil) by analyzingthe position of the 5-substituent. In the fully oxidized substrate5-iodouracil, the iodine is placed in-plane with the pyrimidinering. For 5-iodo-5,6-dihydrouracil with its sp3-hybridized C5, itis positioned out of the ring plane, with the iodine surprisinglypointing toward Cys-671. Because the C5 atom receives itsproton from Cys-671 (Scheme 1), one would expect to find theopposite enantiomer. However, the direct neighborhood of theC4 carbonyl group allows an inter-conversion between bothenantiomers via a keto-enol mechanism, although it is not clearfrom the structure whether the reduction product can undergothis conversion within the active site or is first released andthen rebound in the observed enantiomeric form.

There is almost no density for a 5-substituent of the inhibitorderivative bound in the vicinity of the FMN cofactor in mole-cule D (Fig. 4D). Because a distinction between uracil and5,6-dihydrouracil is not possible at the current resolution of2.25 Å for the DPD�5IU�NADPH data, the decision to interpretand refine the ligand as uracil rather then 5,6-dihydrouracilwas arbitrarily made.

Conformational Changes within the Active-site Pocket—Amino acid side chains involved in substrate/inhibitor bindingin all DPD complex structures are asparagines 609, 668, and736 as well as Thr-737 (Scheme 1). These primary bindingpartners interact via hydrogen bonds with both pyrimidine ringnitrogens and carbonyl groups. The only changes occurring inthe pyrimidine binding pocket apply to residues of the so-calledactive-site loop (residues 670–682) as well as to the two adja-cent amino acids Leu-669 and Ala-683 (Fig. 5). In the ligand-free enzyme and in all complexes except DPD�5IU�NADPH (pH7.5), this loop adopts an open conformation, leaving the pyri-midine-binding site solvent-accessible. The hydroxyl group ofSer-670 is then involved in a weak hydrogen bond (3.3 Å) to theO4 oxygen of the pyrimidine, whereas the general acid Cys-671is located �10 Å distant from the ligand C5 atom. Residues675–679 are fully solvent-exposed, and in three of the fourprotein chains in the asymmetric unit disordered, as indicatedby a lack of electron density.

DPD�5IU�NADPH (pH 7.5) is the only complex in which theactive-site loop has been observed in the closed state. Residues674–678 here form a new short 310 helical turn, and residue682 prolongs the 310 helix after the active-site loop. Superpo-sition of the DPD�5IU�NADPH (pH 7.5) structure withDPD�5IU (pH 4.7) or ligand-free DPD (Fig. 5) shows that resi-due Ser-670 changes its position significantly by 3.6 Å, asmeasured for the C� atoms. Its hydroxyl group is no longerinvolved in substrate/inhibitor binding but in two new stronghydrogen bonds to the backbone carbonyl of Lys-709 and to thecarboxamide nitrogen of Asn-736. Hence, it now bridges tworesidues involved in either FMN binding (Lys-709) or sub-strate/inhibitor-binding (Asn-736).

As a direct consequence, Cys-671 is moved closer to theligand, allowing the formation of the covalent bond (in chain A)or van der Waals interactions (chains B, C, D) between its thiolgroup and the C5 atom of the ligand. In the active sites ofmolecules C-D, the distance between Cys-671-S� and the li-gand C5 is �3.3 Å, suitable for proton transfer. His-673 is alsoinvolved in ligand binding, although by van der Waals ratherthan polar interactions. The His-673 C� is, compared with itsposition in the open loop conformation, now located 12.6 Å

Ternary Complex of DPD with NADPH and 5-Iodouracil 13159

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

closer to the active site. Of the remaining active-site loop res-idues none is directly involved in substrate/inhibitor binding.The electron density observed for residues 676–682 is notequally well defined in all 4 molecules in the asymmetric unit.In molecule A containing the covalent modification electrondensity is continuous for these residues but weaker than for thepreceding amino acids 669–675. Considering also the observedhigher temperature factors (average B-value is 69.6 Å2 for

residues 676–683 and 39.4 Å2 for residues 669–675), it can beconcluded that this stretch of amino acids still shows highermobility than surrounding parts of the protein. In chains B-Dthe density is more diffuse, and the temperature factors areeven higher (average B-value for 669–683 is 55.6 Å2 in chain B,61.1 Å2 in chain C, and 58.8 Å2 in chain D), indicative of thatthese residues might be partly disordered. The maintenance ofa certain degree of active-site-loop mobility can be of advantage

FIG. 3. The pyrimidine binding site.Stereo view of the pyrimidine bindingsites of ligand-free DPD (A), DPD�5FU�NADPH (pH 4.7) (B), DPD�5IU (pH 4.7)(C), and DPD�UAA�NADPH (pH 7.5) (D).The cofactor FMN, the inhibitor mole-cules, ligand-replaced water molecules,and residues involved in ligand bindingare given as ball-and-stick models withoxygens in light gray, nitrogens in black,and all other atoms in gray. The 2�Fo� ��Fc� map is contoured at a level of 1 � forA–C. In D, the �Fo� � �Fc� map for theligand is contoured at a level of 3 �; theligand itself is shown at a position bestfitting the electron density.

Ternary Complex of DPD with NADPH and 5-Iodouracil13160

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

taking into consideration that the loop has to switch betweenopen and closed conformation within each catalytic cycle forsubstrate binding/product release. The movements shouldtherefore require only small activation energy.

Nevertheless, the closed loop conformation is stabilized by anumber of new interactions not observed in the open conforma-tion. A ring nitrogen of His-673 is in hydrogen bond distance tothe backbone oxygen of Glu-611. The side chains of Met-675and Met-680 cluster together with the side chains of Met-642,Ile-613, Leu-612, and Val-583, forming a small hydrophobiccore right beside the substrate binding pocket. These residuescreate a hydrophobic environment for the 5-methyl group of thethymine substrate but are located distant enough to accommo-date also larger substituents at the pyrimidine C5. In moleculeA, where electron density for the side chains of both Glu-677and Arg-678 is observed, these residues are involved in alto-gether four new hydrogen bonds. The glutamate carboxyl groupinteracts with the main chain nitrogens of residues Phe-935and Gly-936 of molecule B, thereby participating in the forma-tion of the dimer interface. One of the guanidinyl nitrogens of

Arg-678 interacts with the backbone oxygen atoms of Asp-581and Ile-582.

DISCUSSION

Active-site Loop Closure Is Substrate Binding and pH-de-pendent—With the new DPD-complex structures available,several open questions regarding the reaction mechanism canfinally be addressed. Until now, one obstacle of the interpreta-tion has been that the flexible loop segment, which controlssubstrate entry and product release from the active site, wasonly observed in the open conformation. Loop closure is anabsolute prerequisite for catalytic activity because it not onlyexcludes the surrounding solvent from the active site, but mostimportantly, also places Cys-671 at the location required forproton transfer to the pyrimidine C5.

Our results suggest that substrate/inhibitor binding in theactive-site pocket subsequently triggers closure of the active-site loop. The energy gained by exchanging the pyrimidine O4as a partner of Ser-670 in a rather weak hydrogen bond againstthe backbone carbonyl of Lys-709 and the side chain amide of

FIG. 4. Ligand binding in DPD�5IU�NADPH (pH 7.5). Stereo views of theactive sites of the four polypeptide chains(A–D) in the asymmetric unit of DPD�5IU�NADPH (pH 7.5). Residues 670–673 ofthe active-site loop, the ligand molecules,and the cofactor FMN are shown as ball-and-stick models. Carbon atoms for theligands 5-iodo-5,6-dihydrouracil, 5IU, anduracil as well as for all carbon atoms ofthe covalently modified cysteine 671 inmolecule A (*C671) originating from5-iodouracil are given in cyan. The final2�Fo� � �Fc� map is contoured in blue at 1 �for the active-site-loop residues, and a �Fo�� �Fc� map calculated after omitting theinhibitor derivatives from the model isshown in magenta at a contour level of 4.5�. The �Fo� � �Fc� map, which results aftermodeling of the ligands as uracil, is con-toured in black (contour level 3 �).

Ternary Complex of DPD with NADPH and 5-Iodouracil 13161

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

Asn-736, resulting in two new much stronger hydrogen bondinteractions, may account for part of the energy necessary forthe loop conformational change.

The failure to trigger the active-site-loop closure in com-plex DPD�5FU�NADPH (15) can clearly be attributed to thenon-physiological pH of the crystallization solution (an influ-ence of the C671A exchange in this complex can be ruled outdue to the observation of the open loop conformation inDPD�5IU (pH 4.7)). At pH 4.7, the loop residue His-673 car-ries a positive charge, which either hinders the loop to reachthe closed conformation or which the substrate-binding sitecannot accommodate in the closed state. His-673 may there-fore account for the group with a pK � 6.5, identified bystudies on pH dependence of kinetic parameters, which isrequired to be in an unprotonated state for optimum activitybut that is not essential for catalytic activity. InDPD�UAA�NADPH (pH 7.5), the closure of the active site isprevented due to steric clashes of loop residues with atoms ofthe rather voluminous inhibitor molecule.

Comparison of DPD domain IV with the structurally andfunctionally closely related Lactococcus lactis dihydroorotatedehydrogenase class 1A in complex with the reaction productorotate (32) allowed predictions of the positions of the catalyt-ically crucial Cys-671 (the counterpart in dihydroorotate dehy-

drogenase class 1A is Cys-130) and the succeeding residuesPro-672 and His-673 (Pro-131 and Asn-132) in the closed state.Superposition of dihydroorotate dehydrogenase class 1A withdomain IV of complex DPD�5IU�NADPH (pH 7.5) indeed re-veals that residues 669–673 of DPD share comparable posi-tions and side chain conformations with residues 128–132 ofdihydroorotate dehydrogenase class 1A. However, none of theresidues Gly-674 to Cys-684 (Val-133–Leu-139) are structur-ally equivalent. A major difference between the ligand-boundand ligand-free structures of both enzymes is that alterationsin loop conformation are far more pronounced in DPD.

Reaction and Electron Transfer Mechanism—The structure ofDPD�5IU (pH 4.7) clearly shows that pyrimidine binding to do-main IV is not dependent on the presence of NADPH in itsbinding cleft, which is in agreement with spectroscopic and ki-netic data (16). Although the structure of a NADPH�DPD com-plex has not yet been determined, comparison of the availableNADPH-free and NADPH-bound complex structures is consist-ent with the assumption that NADPH binding is also independ-ent of the occupancy of the substrate binding pocket. Further-more, proper NADPH binding (i.e. correct placement of thenicotinamide moiety close to the FAD isoalloxazine ring) does notseem to be coupled to loop-closure events in the pyrimidine-binding site. In both complexes, DPD�5IU�NADPH (pH 7.5) with

SCHEME 1. Schematic view of the pyrimidine-binding site before and after uracil reduction.

FIG. 5. Conformational change of the active-site loop upon ligand binding and pH shift. The figure shows the pyrimidine-binding sitein chain A of complex DPD�5IU�NADPH (pH 7.5). The active-site loop (comprising residues 670–682) and the adjacent residues Leu-669 andAla-683 in the closed conformation (DPD�5IU�NADPH, pH 7.5) are shown in pink, with the loop residues 670, 671, and 673 given as ball-and-stickmodels (the carbon atoms of the modified Cys-671 (C671*) originating from 5IU are shown in brown; all others in pink). The same loop as observedin ligand-free DPD and DPD�5IU (pH 4.7) after superposition of domain IV with that of complex DPD�5IU�NADPH is shown in cyan. Here sidechains are shown only for residues Ser-670 and Cys-671 (carbon atoms in cyan) to indicate their position in the open conformation. Ball-and-stickmodels of the cofactor FMN, substrate/inhibitor binding residues, and residue Lys-709 are given with carbon atoms in yellow. These residues donot change position or conformation upon ligand binding and pH shift. Hydrogen bond interactions of ligand binding residues Lys-709 and Ser-670(as observed in DPD�5IU�NADPH (pH 7.5) and DPD�5IU (pH 4.7), respectively) are indicated by dotted lines. Labels in black mark residues placedat identical positions in both structures, and labels in cyan indicate the active-site-loop residues in ligand-free DPD. Active-site-loop residues ofcomplex DPD�5IU�NADPH (pH 7.5) are labeled in pink.

Ternary Complex of DPD with NADPH and 5-Iodouracil13162

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

the active-site loop closed and DPD�UAA�NADPH (pH 7.5) withthe loop open, NADPH is bound in an identical, “productive”manner.

Analysis of the structures of the DPD complexes reveals thatno information transfer between NADPH and the substrate-binding site occurs other than via the electron transfer chain.NADPH binding domain III and FMN binding domain IV arelocated at opposite ends of the subunit and entirely separatedfrom each other by domains I, II, and V (Fig. 1). The movementof domain III induced by NADPH binding does not cause sig-nificant conformational changes in the other four domains. Theaccelerating effect of uracil on the reduction of DPD byNADPH, which was revealed by monitoring absorbancechanges during the reaction of DPD with NADPH in the ab-sence and presence of uracil (16), is thus not coupled to confor-mational changes within the protein but is most likely exclu-sively based on the “sink function” of the pyrimidine substrateas final electron acceptor.

The catalytically competent binding of NADPH requires aslight widening of the cleft between domains II and III, whichis achieved by a modest movement of domain III. A similar butmore pronounced domain rearrangement as well as alterationsin crystal packing contacts have been reported for adrenodoxinreductase upon binding of NADP� (33). Because no pH changeis involved for adrenodoxin reductase, it can be argued thatbinding of NADPH in the conformation suitable for hydridetransfer rather than the pH shift induces the domain move-ment observed for DPD.

However, there is an indirect pH effect. At a pH of 7.5,NADPH is bound to DPD in the proper conformation,whereas at pH 4.7 it is bound in a non-productive fashion. Weexamined the immediate surroundings of NADPH for aminoacid residues, which could trigger the domain movement byadopting different conformations at the two pH values. Themost obvious candidate for such a trigger function is Asp-342.Inevitably this residue has to move to make room for thenicotinamide moiety of NADPH. In ligand-free DPD and allcomplexes obtained at pH 4.7, the side chain of Asp-342 is not

involved in hydrogen-bonding interactions, suggesting thatAsp-342 is protonated under these conditions. At pH 7.5, thecarboxyl group of Asp-342 needs to form a hydrogen bondinteraction with the main chain amide of Val-373, the onlycandidate within an appropriate distance, to compensate forits negative charge. The formation of this hydrogen bond mayrequire or induce the observed additional conformationalchanges in stretch 371–373, leading to the disruption of back-bone interactions between this stretch and Ala-340 and sub-sequently to the movement of domain III.

Once NADPH is bound, hydride is transferred from C4 of thenicotinamide of NADPH to FAD N5. It appears that there is nocompensation for the generated negative charge of reduced FAD(FADH�) localized at the N1 atom. Unlike other flavoenzymes(34), DPD provides no positively charged amino acid side chain orpartially charged entity such as the N terminus of an �-helix ora cluster of backbone nitrogens in a reasonable distance to theFAD N1. The closest atoms are the side-chain hydroxyl groupand the backbone amide of Thr-489, both located at a distance�3.4 Å to the N1 locus of FAD. Due to the missing chargecompensation, the anionic form of the reduced flavin is not sta-bilized, and the redox potential of the cofactor is kept low by theprotein environment. This might represent one of the drivingforces for electron transfer to the nearest [4Fe-4S] cluster nFeS2.

The O2 hydroxyl of FAD is in hydrogen bond distance (2.6 Å)to a water molecule, which in turn is positioned at a distance of3.0 Å to S� of Cys-130, a ligand of cluster nFeS2 (Fig. 6). Thiswater molecule is present in all four polypeptide chains in theasymmetric unit and shows strong and well defined electrondensity. An additional hydrogen bond is formed with the back-bone amide of Val-490 (2.8 Å). We propose that this is the mostlikely route for the transfer of two electrons per catalytic cyclebetween FAD and nFeS2. Concomitant with the electron trans-fer, the FAD-N5 proton must be released. An apparent route forproton release proceeds via Arg-235, which is located withinhydrogen-bond distance to the N5 and O4 atoms of FAD. Theproton can move to Arg-235 and from there either directly orvia the carboxyl group of Glu-346 to the contacting water mol-

FIG. 6. The electron transfer path-way between the ligand-bindingsites. The cofactors FAD and FMN (car-bon atoms in cyan) and the four [4Fe-4S]clusters creating the electron transferchain between both flavins (iron atoms inmagenta, sulfur atoms in green) as well asall amino acids located in van der Waalsdistance (3.8 Å) to the [4Fe-4S] clusters toatoms N5, N1, and C7M of FMN or toatoms N5, N1, and O2 of FAD are shownas ball-and-stick models. Residues men-tioned in the text are labeled. Amino acidswith carbon atoms in yellow originatefrom the same subunit as FAD and FMN.Orange carbon atoms mark residues orig-inating from the other subunit in thedimer. Hydrogen bond interactions men-tioned in the text are indicated by dottedlines. For clarity, NADPH and the pyrim-idine substrate are not shown.

Ternary Complex of DPD with NADPH and 5-Iodouracil 13163

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

SCHEME 2. Proposed mechanisms of DPD inhibition by 5IU and inhibitor deactivation. a, 5IU is reduced to 5-iodo-5,6-dihydrouracil ina NADPH-dependent reaction. b, active-site loop opening is followed by fast release of the enzymatically generated enantiomer of 5-iodo-5,6-dihydrouracil, which can be transformed into the other enantiomeric form via keto-enol tautomerism and rebound to the enzyme. c, 5-iodo-5,6-dihydrouracil attacks the thiol group of Cys-671, leading to its covalent modification to S-(hexahydro-2,4-dioxo-5-pyrimidinyl)cysteine. d, 5-iodo-5,6-dihydrouracil can also be deactivated via elimination of HI, resulting in formation of uracil, which is a substrate of DPD. Its reduction generates5,6-dihydrouracil. The boxed capital letters indicate the dominant species observed in the four subunits (A-D) of DPD in the asymmetric unit.

Ternary Complex of DPD with NADPH and 5-Iodouracil13164

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

ecules and eventually be released to the bulk solvent. Theelectron transfer pathway between nFeS2 and cFeS2 (vianFeS1 and cFeS1) is shielded from solvent molecules through anumber of hydrophobic residues, mostly isoleucines andleucines (Fig. 6).

Based on the positioning of cluster cFeS2 with respect to theFMN cofactor in the pyrimidine-binding site, we suggest theC7� methyl group of FMN as the site of electron entry intothe flavin ring system. There is no obvious route for electrontransfer between these two redox cofactors. The distances be-tween S� of Cys-989, a cFeS2 ligand, to the closest amino acidresidue (Ile-590 C�1) and from there to the FMN-C7� are 4.4and 3.6 Å, respectively. However, the simple positioning of theredox centers in close proximity (�14 Å) to each other might beeffective enough to achieve electron tunneling (35).

The negative charge developing at FMN N1 upon full reduc-tion of the cofactor to the hydrochinone form is compensated bythe close neighborhood of the positively charged lysine residueLys-709. A proton is most likely recruited from another lysineresidue (Lys-574), which is in hydrogen-bond distance to theFMN N5 and, via Glu-611, connected to the bulk solvent. FromFMN N5, a hydride is transferred to the uracil/thymine C6, andthe C5 atom receives a proton from Cys-671, located at theopposite ring side (Scheme 1).

Inhibition Mechanism for 5IU (Scheme 2)—The inhibition ofDPD (from bovine liver) by 5IU was thoroughly examined byPorter et al. (19), resulting in the revelation of several mechanis-tic details that can now be interpreted in structural terms. It wasshown that 5IU is both a substrate and potent inactivator ofDPD. The NADPH-dependent reduction of 5IU by DPD gener-ates 5-iodo-5,6-dihydrouracil (Scheme 2a), the actual inhibitoryspecies, which acts as an alkylating agent with properties similarto iodoacetamide. Because excess 5IU and dithiothreitol pro-tected the enzyme against inactivation, Porter et al. (19) suggestthat 5-iodo-5,6-dihydrouracil is released from the enzyme beforeit can attack the active-site residue Cys-671 (Scheme 2b). Theanti-addition of hydride at C6 and proton at C5 of the 5IU bringsthe rather voluminous iodo substituent at C5 in immediateneighborhood to the FMN isoalloxazine ring system. The result-ing repulsion forces might in part account for the opening of theactive-site loop and fast release of the product.

In fact, comparison of the stoichiometries of DPD inactivationby racemic 5-iodo-5,6-dihydrouracil and the enzymatically gen-erated enantiomer indicated that dissociation of the reductionproduct from the active site and its rebinding is required forinactivation, since the non-enzymatically generated enantiomeris a more potent inactivator of the enzyme. The observation ofthis enantiomeric form of 5-iodo-5,6-dihydrouracil in the activesite of molecules B of DPD�5IU�NADPH (pH 7.5) supports theseconclusions. The interconversion between both enantiomers pro-ceeds most likely via a keto-enol mechanism (Scheme 2b).

The further fate of 5-iodo-5,6-dihydrouracil is partitionedbetween inactivation of the enzyme by covalent modificationof Cys-671 to S-(hexahydro-2,4-dioxo-5-pyrimidinyl)cysteine(Scheme 2c) and conversion to a product unable to inactivateDPD (Scheme 2d). The covalent adduct is present in the activesites of molecules A of complex DPD�5IU�NADPH (pH 7.5).Hence, the inhibitory effect of 5IU is mechanism-based anddepends on the alkylation of an amino acid essential for cata-lysis. As possible mechanisms for the deactivation of 5-iodo-5,6-dihydrouracil, reductive dehalogenation to dihydrouracil andelimination of HI under formation of uracil had been proposed(19). Both uracil and dihydrouracil are products of the reactionof DPD with racemic 5-iodo-5,6-dihydrouracil (19). It istherefore not surprising that uracil or 5,6-dihydrouracil is

bound in the active sites of molecules D (and partially alsomolecules B and C) of complex DPD�5IU�NADPH (pH 7.5).

DPD reduces uracil and thymine to the corresponding 5,6-dihydropyrimidines in the pH range from 6.0 to 8.0, whichgives raise to the conclusion that no enzymatic reduction orcovalent modification of Cys-671 occurred before the transfer ofthe DPD�5IU�NADPH crystals to pH 7.5. Because substratesand products of the enzymatic reduction compete for the bind-ing in the active site of DPD (19), it was unlikely to obtaincomplete inactivation of the enzyme in the crystals. We couldnot identify the source of the non-equivalency of the four mol-ecules in the asymmetric unit of DPD crystals, which resultedin the accumulation of distinct inhibitor derivatives within thedifferent active sites (A-D). Superposition of the subunits doesnot reveal other deviations in amino acid coordinates except ina few surface-located loop regions, which are generally allsolvent-exposed. This non-equivalency has, however, been ob-served in all DPD structures in so far as that part of the“proline-rich loop” (residues 899–910) is usually less disor-dered in subunits C, and electron density for the active-siteloop (in its open conformation) is usually best defined in sub-units D. The features of the complex DPD�5IU�NADPH (pH 7.5)support the proposal that 5-iodouracil acts as a mechanism-based inhibitor covalently modifying the active-site residueCys-671 and agree well with previously presented biochemicalobservations concerning catalytic and inhibitory mechanisms.

The complex structures presented here provide for the firsttime a complete picture of the electron transfer chain in theproductive enzyme-substrate complex. It was also shown thatpyrimidine binding triggers a conformational change of a flex-ible active-site loop, resulting in placement of the catalyticallycrucial cysteine 671 close to the bound substrate. This loopclosure as well as the correct binding of the cosubstrateNADPH requires physiological pH.

Acknowledgments—We thank A. Mozzarelli for advice regarding thesoaking experiments. We thankfully acknowledge access to the synchro-tron radiation at the European Synchrotron Radiation Facility and theDeutsches Elektronensynchrotron.

REFERENCES

1. Wasternack, C. (1980) Pharmacol. Ther. 8, 629–6512. Traut, T. W., and Jones, M. E. (1996) Prog. Nucleic Acid Res. Mol. Biol. 53,

1–783. Milano, G., and Etienne, M.-C. (1994) Anticancer Res. 14, 2295–22974. Papamichael, D. (2000) Stem Cells 18, 166–1755. Woodcock, T. M., Martin, D. S., and Damin, L. E. M. (1980) Cancer 45,

1135–11436. Heggie, G. D., Sommadossi, J.-P., Cross, D. S., Huster, W. J., and Diasio, R. B.

(1987) Cancer Res. 47, 2203–22067. Lu, Z. H., Zhang, R., and Diasio, R. B. (1993) Cancer Res. 53, 5433–54388. Tuchman, M., Stoeckeler, J. S., Kiang, D. T., O’Dea, R. F., Ramnaraine, M. L.,

and Mirkin, B. L. (1985) N. Engl. J. Med. 313, 245–2499. Shehata, N., Pater, A., and Tang, S.-C. (1999) Cancer Invest. 17, 201–205

10. Van Kuilenburg, A. B. P., Haasjes, J., Richel, D. J., Zoetekouw, L., Van Lenthe,H., De Abreu, R. A., Maring, J. G., Vreken, P., and Van Gennip, A. H. (2000)Clin. Cancer Res. 6, 4705–4712

11. Diasio, R. B., Beavers, T. L., and Carpenter, J. T. (1988) J. Clin. Invest. 81,47–51

12. de Bono, J. S., and Twelves, C. J. (2001) Invest. New Drugs 19, 41–5913. Yokota, H., Fernandez-Salguero, P., Furuya, H., Lin, K., McBride, O. W.,

Podschun, B., Schnackerz, K. D., and Gonzalez, F. J. (1994) J. Biol. Chem.269, 23192–23196

14. Albin, N., Johnson, M. R., and Diasio, R. B. (1996) DNA Sequence 6, 243–25015. Dobritzsch, D., Schneider, G., Schnackerz, K. D., and Lindqvist, Y. (2001)

EMBO J. 20, 650–66016. Rosenbaum, K., Jahnke, K., Curti, B., Hagen, W. R., Schnackerz, K. D., and

Vanoni, M. A. (1998) Biochemistry 37, 17598–1760917. Podschun, B., Cook, P. F., and Schnackerz, K. D. (1990) J. Biol. Chem. 265,

12966–1297218. Naguib, F. N. M., El Kouni, M. H., and Cha, S. (1989) Biochem. Pharmacol. 38,

1471–148019. Porter, D. J. T., Chestnut, W. G., Taylor, L. C. E., Merril, B. M., and Spector,

T. (1991) J. Biol. Chem. 266, 19988–1999420. Rosenbaum, K., Schaffrath, B., Hagen, W. R., Jahnke, K., Gonzalez, F. J.,

Cook, P. F., and Schnackerz, K. D. (1997) Protein Expression Purif. 10,185–191

Ternary Complex of DPD with NADPH and 5-Iodouracil 13165

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

21. Dobritzsch, D., Persson, K., Schneider, G., and Lindqvist, Y. (2000) ActaCrystallogr. Sec. D 57, 153–155

22. Otwinowski, Z., and Minor, W. (1997) Methods Enzymol. 276, 307–32623. CCP4 (1994) Acta Crystallogr. Sec. D 50, 760–76324. Navaza, J. (1994) Acta Crystallogr. Sec. D 50, 157–16325. Jones, T. A., Zou, J.-Y., Cowan, S. W., and Kjeldgaard, M. (1991) Acta Crys-

tallogr. Sec. A 47, 110–11926. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-

Kunstleve, R. W., Jiang, J.-S., Kuszewski, J., Nilges, M., Pannu, N. S.,Read, R. J., Rice, L. M., Simonson, T., and Warren, G. L. (1998) ActaCrystallogr. Sec. D 54, 905–921

27. Laskowski, R. A., MacArthur, M. W., Moss, D. S., and Thornton, J. M. (1993)J. Appl. Crystallogr. 26, 283–291

28. Lu, G. (1996) Protein Data Bank Quarterly Newsletter 78, 10–1129. Kraulis, P. J. (1991) J. Appl. Crystallogr. 24, 946–95030. Esnouf, R. M. (1997) J. Mol. Graphics Modelling 15, 133–13831. Merritt, E. A., and Bacon, D. J. (1997) Methods Enzymol. 277, 505–52432. Rowland, P., Bjornberg, O., Nielsen, F. S., Jensen, K. F., and Larsen, S. (1998)

Protein Sci. 7, 1269–127933. Ziegler, G. A., and Schulz, G. E. (2000) Biochemistry 39, 10986–1099534. Fraaije, M. W., and Mattevi, A. (2000) Trends Biochem. Sci. 25, 126–13235. Page, C. C., Moser, C. C., Chen, X., and Dutton, P. L. (1999) Nature 402, 47–52

Ternary Complex of DPD with NADPH and 5-Iodouracil13166

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from

LindqvistDoreen Dobritzsch, Stefano Ricagno, Gunter Schneider, Klaus D. Schnackerz and Ylva

MECHANISM OF INHIBITION AND ELECTRON TRANSFERDehydrogenase with NADPH and 5-Iodouracil: IMPLICATIONS FOR

Crystal Structure of the Productive Ternary Complex of Dihydropyrimidine

doi: 10.1074/jbc.M111877200 originally published online January 16, 20022002, 277:13155-13166.J. Biol. Chem. 

  10.1074/jbc.M111877200Access the most updated version of this article at doi:

 Alerts:

  When a correction for this article is posted• 

When this article is cited• 

to choose from all of JBC's e-mail alertsClick here

  http://www.jbc.org/content/277/15/13155.full.html#ref-list-1

This article cites 35 references, 7 of which can be accessed free at

by guest on March 21, 2020

http://ww

w.jbc.org/

Dow

nloaded from