HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

J. Biol. Chem., Vol. 277, Issue 15, 13155-13166, April 12, 2002

This Article Abstract Full Text (PDF) All Versions of this Article: 277/15/13155    most recent M111877200v1 Submit a Letter to Editor Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Dobritzsch, D. Articles by Lindqvist, Y. Search for Related Content PubMed PubMed Citation Articles by Dobritzsch, D. Articles by Lindqvist, Y. Social Bookmarking                      What's this?

Crystal Structure of the Productive Ternary Complex of Dihydropyrimidine Dehydrogenase with NADPH and 5-Iodouracil

IMPLICATIONS FOR MECHANISM OF INHIBITION AND ELECTRON TRANSFER^*

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, Karolinska Institutet, S-17177 Stockholm, Sweden and  Theodor-Boveri-Institut für Biowissenschaften, Physiologische Chemie I, Am Hubland, D-97074 Würzburg, Germany

Received for publication, December 13, 2001, and in revised form, January 16, 2002

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Dihydroprymidine dehydrogenase catalyzes the first and rate-limiting step in pyrimidine degradation by converting pyrimidines to the corresponding 5,6- dihydro compounds. The three-dimensional structures of a binary complex with the inhibitor 5-iodouracil and two ternary complexes with NADPH and the inhibitors 5-iodouracil and uracil-4-acetic acid were determined by x-ray crystallography. In the ternary complexes, NADPH is bound in a catalytically competent fashion, with the nicotinamide ring in a position suitable for hydride transfer to FAD. The structures provide a complete picture of the electron transfer chain from NADPH to the substrate, 5-iodouracil, spanning a distance of 56 Å and involving FAD, four [Fe-S] clusters, and FMN as cofactors. The crystallographic analysis further reveals that pyrimidine binding triggers a conformational change of a flexible active-site loop in the /-barrel domain, resulting in placement of a catalytically crucial cysteine close to the bound substrate. 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 steric hindrance. The three-dimensional structure of the ternary complex enzyme-NADPH-5-iodouracil supports the proposal that this compound acts as a mechanism-based inhibitor, covalently modifying the active-site residue Cys-671, resulting in S-(hexahydro-2,4-dioxo-5-pyrimidinyl)cysteine.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Dihydropyrimidine dehydrogenase (DPD,¹ EC 1.3.1.2) is a cytosolic enzyme catalyzing the NADPH-dependent reduction of uracil and thymine to the corresponding 5,6-dihydropyrimidines, the first and rate-limiting reaction in the three-step pathway of pyrimidine degradation (1). In mammals, this is the only pathway leading to the synthesis of the putative neurotransmitter -alanine (2).

The human enzyme has recently evolved as an adjunct target for anticancer drug design because it also degrades 5-fluorouracil (5FU), one of the most widely prescribed chemotherapeutic agents for treatment of many common malignancies (3). 5FU primarily targets the enzyme thymidylate synthase and thereby interferes with DNA synthesis (4). Approximately 85% of the administered dose is rapidly degraded by DPD to therapeutically inactive but toxic fluorinated products (5, 6). Furthermore, accurate determination of an optimal drug dose proved to be very difficult due to substantial differences in the activity of DPD among individuals (7, 8). Severe life-threatening toxicities have been reported after treatment of cancer patients with inherited DPD deficiency with standard doses of 5FU (9-11). The catabolism of 5FU via DPD and the other two enzymes of the pyrimidine degradation pathway thus represents a major determinant of the pharmacokinetics of this anticancer drug. Its inhibition may result in increased 5FU efficacy, administration of lower doses, and diminished drug side effects. To date, several DPD inhibitors (eniluracil, 5-chloro-2,4-dihydroxypyridine, and 3-cyano-2,6-dihydropyridine) are utilized or under evaluation as modulators of 5FU treatment (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-dimensional 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 enzyme is a homodimer of 2 × 111 kDa. Each subunit of 1025 amino acids carries one FAD, one FMN, and four [4Fe-4S] clusters (16). According to the non-classical two-site ping-pong kinetic mechanism, the enzyme contains separate binding sites for the electron-donating cosubstrate NADPH and the electron-accepting pyrimidines, respectively (17).

The pH dependence of the kinetic parameters suggests the following reaction mechanism for the reduction of uracil/thymine by pig liver DPD (17). NADPH binds to site 1 with the pro-S side of the nicotinamide facing toward the FAD-isoalloxazine ring. Hydride is transferred from the C-4 of NADPH to N-5 of FAD, leaving NADP⁺ and FADH  . There is evidence from kinetic isotope effects that electron transfer to site 2 via the [4Fe-4S] clusters does not occur until NADP⁺ is released. At site 2 reduced FMN is formed. The pyrimidine substrate is bound with the si-face at the C-6 atom directed toward the N-5 of FMN and with the si-face of C-5 directed toward an enzyme general acid, which has been identified as the thiol group of cysteine 671 (16). The transfer of hydride from reduced FMN and the proton from Cys-671 is proposed to occur in a concerted anti-addition reaction.

The three-dimensional structure of pig liver DPD (15) revealed a highly modular organization of the subunit. It consists of five distinct domains, each carrying a subset of the various prosthetic groups. The -helical domain I (residues 27-172) contains two of the iron-sulfur clusters, nFeS1 and nFeS2, of which the latter shows an unusual cluster coordination comprising one glutamine and three cysteine residues. The Rossmann-type nucleotide binding folds of domains II (residues 173-286, 442-524) and III (residues 287-441) bind FAD and NADPH, respectively. The interface between these domains forms 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 cFeS1 and cFeS2 are bound to the core of the C-terminal domain V (residues 1-26 and 848-1025). The three-dimensional arrangement of the distinct domains in the homodimeric enzyme leads to the formation of two electron-transfer chains, in which the electrons are transferred from the FAD to nFeS2 and subsequently to nFeS1. The remaining gap between nFeS1 and the FMN is closed by the clusters of the C-terminal domain V of the other subunit in the dimer. This domain swapping makes the dimer the minimal catalytic unit of pig liver DPD.

In addition to the x-ray structure of ligand-free DPD, that of a catalytically inactive mutant (C671A) in complex with NADPH and the anti-cancer drug 5FU was determined to 2.0-Å resolution (15). Two observations led to the conclusion that this ternary complex does not represent a productive enzyme-substrate complex. First, the nicotinamide ring of NADPH was not placed in a position suitable for hydride transfer from its C4 atom to the FAD N-5 atom. Second, the loop comprising residues 670-682 (referred to as "active site loop"), which is proposed to close over the active site while catalysis occurs, is seen in an open and partially disordered conformation. Hence, the alanine at position 671, which in the mutant replaces the general acid Cys-671, was located too distant from the C5 atom of 5FU.

Possible explanations for the failure to trigger the putative active-site closure upon substrate binding are the mutation of the cysteine to alanine or the pH difference between crystallization conditions (4.7) and optimal catalytic activity (~7.3). The assumption that pH might influence the loop conformation is based on the presence of a histidine residue at position 673, which most likely participates in substrate binding. This histidine residue should be positively charged at pH 4.7 but neutral at the pH optimal for catalytic activity.

To address this question, we prepared crystals of wild-type pig 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 K_i = 78.6 ± 20.2 µM) (18), 5IU serves as a substrate and is converted to 5-iodo-5,6-dihydrouracil, which has been shown to be a potent alkylating agent able to covalently modify the Cys-671-thiol group (19). This may lead to trapping of the active-site loop in its closed conformation. Additionally and for purposes of comparison, the structure of the binary DPD·5IU complex was determined. In this report, we describe the three-dimensional structures of these inhibitor-enzyme complexes and discuss their implication for the mechanism of DPD.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Protein Purification and Crystallization

Pig liver DPD was produced as recombinant protein in Escherichia coli, and has been purified and crystallized as described previously (20, 21). The binary complex of DPD was obtained by crystallization of wild-type DPD in the presence of 1 mM 5IU, the ternary DPD·5IU·NADPH complex, by cocrystallization of DPD with 1 mM 5IU and 5 mM NADPH. A change of pH in the crystals was achieved by soaking crystals of the complex DPD·5IU·NADPH in 100 mM Hepes (pH 7.5), 22% polyethylene glycol 6000 (w/v), 0.8 mM NADPH, for 20 min before data collection. The complex DPD·UAA·NADPH was prepared by soaking wild-type DPD crystals for 2.5 h in a solution containing 100 mM Hepes (pH 7.5), 22% polyethylene glycol 6000 (w/v), 1 mM UAA, 2.5 mM NADPH. The crystals were stable under these conditions and did not change their appearance. All DPD complexes crystallized in space group P2₁, with four molecules (two homodimers) in the asymmetric unit.

Data Collection

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

X-ray diffraction data were collected at 100 K at beam line ID-EH3 of the European Synchrotron Radiation Facility (European Synchrotron Radiation Facility, Grenoble, France) with a MAR CCD detector for the two ternary complexes and at the EMBL beam line BW7B at the DORIS storage ring, Deutsches Elektronensynchrotron (Hamburg/Germany) with a MARResearch Imaging Plate for the binary complex. Data were indexed and integrated using DENZO (22) and scaled with the CCP4 suite of programs (23). Table I gives the details of the data collection statistics.

Structure Determination and Refinement

DPD·5IU (pH 4.7)-- Initial rigid body refinement using the refined structure of the ligand-free enzyme (without water molecules) as a model and data to 2.5-Å resolution resulted in values of 27.9% for R_cryst and 28.8% for R_free, respectively. The |F_o|  |F_c| map computed from this initial model showed clear electron density for 5IU bound close to the cofactor FMN. Models for the inhibitor and water molecules were added. The high resolution of the data allowed the identification of several alternative conformations of amino acid side chains (residues 187, 245, 947, 962 in chain A, residues 14, 167, 410, 577, 932 in chain B, residues 167, 245, 281, 368, 410, 577, 779, 858, 932, 947, 1006 in chain C, and residues 14, 410, 775, 779, 858, 947, 962, 1006 in chain D) that were included into the model. Model building and refinement resulted in R_cryst = 18.3% and R_free = 19.8% (Table II). The final model contains residues 2-674, 681-901, and 908-1017 for chain A, residues 2-673, 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, four FAD, four FMN, 16 [4Fe-4S] clusters, four 5IU, and 4849 water molecules. All missing amino acids belong either to disordered, surface-located loop regions (such as the active-site loop and the "proline-rich loop" (15)) or to the mobile C-terminal tail.

DPD·5IU·NADPH (pH 7.5)-- Initial rigid body refinement using the same model as for DPD·5IU did not result in reasonable R values or in interpretable electron density maps. Therefore, molecular replacement was performed with the program AMoRe (24), again with the structure of the ligand-free DPD without water molecules as the search model, resulting in a solution with a correlation coefficient of 78.3. After one cycle of crystallographic refinement (including rigid body refinement, annealing, minimization, and individual B-factor refinement), values of 24.1% for R_cryst and 26.6% for R_free were obtained, and 2|F_o|  |F_c| and |F_o|  |F_c| maps were calculated. The maps showed clear electron density for NADPH, with its nicotinamide ring bound close to the FAD in all four molecules in the asymmetric unit, four 5IU molecules, or inhibitor derivatives (see "Results") bound in the pyrimidine binding sites as well as a change in active-site-loop conformation. Inhibitor models (5IU) and NADPH molecules were fitted into the electron density. The active-site-loop residues were rebuilt into the observed density, and minor readjustments of several amino acid side chains were necessary. The final model contained residues 2-1020 for chains A and D, residues 2-900 and 908-1020 for chain B, residues 2-902 and 907-1020 for chain 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 to S-(hexahydro-2,4-dioxo-5-pyrimidinyl) cysteine. Because the active sites 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 models were adjusted correspondingly.

DPD·UAA·NADPH (pH 7.5)-- For initial rigid body refinement the AMoRe solution for the DPD·5IU·NADPH ternary complex (after removal of inhibitor, cosubstrate, and water molecules) was used, resulting in R values of 29.6% (R_cryst) and 29.7% (R_free). Because the initial |F_o|  |F_c| map indicated binding of NADPH and UAA to the active sites of all four molecules in the asymmetric unit, models of cosubstrate and inhibitor were added to the structure. Due to the low resolution of the data, 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 2-672 and 682-1016 for chain C, and residues 2-676, 680-901, and 906-1018 for chain D. R-factors were refined to 21.7% for R_cryst and 27.3% for R_free, respectively.

All model building was carried out in O (25), the refinement and water-picking routine for the complexes was done using the program CNS (26). NCS restraints were used during refinement, with looser restraints for side-chain atoms. Excluded from these NCS restraints were residues 51-54, 264, 323-327, 397-404, 415-417, 671-683, and 899-907, which belong to solvent-exposed loop regions, showing slightly different conformations in the four molecules present in the asymmetric unit. All three models have good stereochemistry, as determined by the program PROCHECK (27). Refinement statistics are given in Table II.

Structural alignments were achieved with the programs TOP (28) or the LSQ commands in O (25). All figures were generated using BOBSCRIPT (29, 30) and RASTER3D (31). Crystal contacts were determined with CONTACT of the CCP4 suite of programs (23). The crystallographic data have been deposited in the Protein Data Bank, with accession codes 1gte (DPD·5IU), 1gth (DPD·5IU·NADPH), and 1gt8 (DPD·UAA·NADPH).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Effects of pH Change on Crystal Packing-- The pH change from 4.7 to 7.5 did not have any obvious effects on the appearance of the DPD crystals. Nevertheless, it caused significant changes in unit cell dimensions and packing of the molecules within the crystal. Both ternary complexes obtained at pH 7.5 showed an elongation of unit cell dimension c by ~4 Å. This was accompanied by a change in packing of the two tightly associated homodimers within the asymmetric unit, leading to substantial alterations in dimer-dimer and crystal contacts. The majority of the contacts observed in unliganded DPD are absent 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 pyrimidine analogs at pH 4.7 did not introduce major changes in the protein backbone structure, as reflected by a root mean square deviation of 0.24 Å, measured for all C atoms of the subunits of ligand-free DPD and the binary complex DPD·5IU. For the ternary complexes obtained at pH 7.5, the superposition with ligand-free DPD yielded root mean square deviations of 0.40 Å for all C atoms.

In complex DPD·5IU·NADPH (pH 7.5), but not in complex DPD·UAA·NADPH (pH 7.5), the active-site loop was observed in its closed conformation. Differences to ligand-free DPD occurring in both ternary complexes are located in the NADPH-binding site, where several amino acids adopt different conformations to allow proper NADPH binding, as described below. In addition, a slight movement of the NADPH binding domain with respect to the domain arrangement for unliganded DPD was observed. Superposition of the ₈/₈-barrel domain IV of the corresponding structures reveals that the NADPH binding domain III changes its position by a modest movement (maximum difference in C-coordinates is 2.7 Å for residue 415) with respect to the FAD binding domain II, resulting in a slight widening of the NADPH binding cleft (Fig. 1). The displacement corresponds to a rotation by approximately 2° and a translation along the rotation axis by 0.7 Å.

View larger version (48K): [in this window] [in a new window]   Fig. 1.   Stereo view of the superimposed subunits of ligand-free DPD and its ternary complex with 5IU and NADPH (pH 7.5). The superposition is based on the 8/8-barrel domains IV. For DPD·5IU·NADPH (pH 7.5), the N-terminal FeS-cluster domain I is shown in green, the FAD binding domain II is shown in yellow, the NADPH binding domain III is shown in orange, the FMN binding domain IV is shown in red, and the C-terminal FeS binding domain V is shown in blue. The subunit structure of ligand-free DPD is black. Cofactor molecules, NADPH, and 5IU are shown as ball-and-stick models. The arrow at the NADPH binding domains highlights the region with the largest conformational changes in this domain upon formation of the ternary complex.

NADPH-binding Site-- Binding of NADPH in the cleft formed between domains II and III was accompanied by only local changes of amino acid conformations within domain II, whereas it involved both local conformational changes and more global rigid body movements of residues originating from domain III (Fig. 2). A reorientation of the side chains of Phe-438 and Arg-364 resulted in stacking interactions of these residues with the NADPH adenine moiety. Residues 485-488, comprising a loop and the first residue of the following -helix of domain II, moved in all NADPH-containing complexes with respect to their position in ligand-free DPD. The observed displacement 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 NADPH molecule. By rotation around the N-C and C-C bonds of Ala-486, a shift of the Asn-487 side chain by 5.6 Å measured for the carboxamide nitrogen is achieved, which then interacted with the 3-hydroxyl of the nicotinamide-ribose via a hydrogen bond. The new backbone conformation of stretch 485-488 is stabilized by two new hydrogen bonds formed between the main chain nitrogen atoms of Asn-487 and Thr-488 to the carboxyl-group of Glu-491. The most significant conformational change necessary for the proper positioning of the nicotinamide moiety of NADPH involves Asp-342. In all complexes obtained at pH 4.7, its side chain partially occupies the space for the nicotinamide of NADPH when bound in productive fashion. In ligand-free DPD a carboxylate oxygen is at a 3.6-Å distance from the FAD N5 at a position very close to the C4 atom of NADPH in complex DPD·5IU·NADPH (pH 7.5). After NADPH binding the C of Asp-342 is displaced by 1.9 Å, the NADPH-nicotinamide moiety is stacked between the FAD isoalloxazine ring and the side chain of Asp-342, and one of the carboxyl oxygens of Asp-342 is hydrogen-bonded to the backbone amide of Val-373. Additionally, this movement of Asp-342 causes a displacement of residues 340-344, resulting in disruption of the hydrogen bonds between the backbone oxygen of Ala-340 and 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 complex DPD·5FU·NADPH (pH 4.7) (15).

View larger version (38K): [in this window] [in a new window]   Fig. 2.   NADPH binding to DPD at pH 7.5. A, stereo view of the superimposed NADPH-binding sites of ligand-free DPD (dark blue) and DPD·5IU·NADPH (pH 7.5) (cyan). Amino acid side chains directly involved in NADPH binding or significantly changing position or conformation upon NADPH binding are labeled and shown as ball-and-stick models. For ligand-free DPD, the cofactor FAD is shown in dark blue. NADPH and FAD bound to DPD·5IU·NADPH (pH 7.5) are given in cyan. B, stereo view of NADPH as bound in complex DPD·5IU·NADPH (pH 7.5) together with the |Fo|  |Fc| map at 3 -contour level calculated after omitting the NADPH atoms from the model.

In the latter, the mode of NADPH binding was non-productive, because the nicotinamide ring was partially disordered and flipped away from the FAD isoalloxazine ring onto the protein surface. In that complex no domain rearrangement could be observed. Nevertheless, most of the more local alterations in the conformation of residues lining the walls of the NADPH binding pocket are seen in DPD·5FU·NADPH (pH 4.7).

Ligands in the Pyrimidine-binding Site-- For all complexes of DPD, electron density shows the inhibitor molecules bound almost parallel to the FMN isoalloxazine ring plane, replacing several water molecules found in the active site of the unliganded enzyme (Fig. 3A).

View larger version (26K): [in this window] [in a new window]   Fig. 3.   The pyrimidine binding site. Stereo view of the pyrimidine binding sites 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 molecules, ligand-replaced water molecules, and residues involved in ligand binding are given as ball-and-stick models with oxygens in light gray, nitrogens in black, and all other atoms in gray. The 2|Fo|  |Fc| map is contoured at a level of 1  for A-C. In D, the |Fo|  |Fc| map for the ligand is contoured at a level of 3 ; the ligand itself is shown at a position best fitting the electron density.

In DPD·5IU (pH 4.7), the electron density is well defined for all atoms of 5-iodouracil (Fig. 3C). The binding geometry corresponds to that observed for 5-fluorouracil in the DPD·5FU·NADPH structure (15) (Fig. 3B), with the C6 atom of the inhibitor closest to the N5 of FMN (3.8 Å), the O₂ atom in a 3.7-Å distance to both, the N10 and C9A of FMN, and the iodine located 3.7 Å from the FMN O4. As already noted for 5FU, there are no hydrogen bonds between amino acid side chains 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 acid molecule, but due to the low resolution of the data, the binding geometry of the inhibitor cannot unambiguously be determined and will therefore not be discussed in further detail (Fig. 3D).

The electron density observed for inhibitor molecules bound in the active sites of the DPD·5IU·NADPH (pH 7.5) complex is heterogeneous and differs between each of the four molecules in the asymmetric unit (Fig. 4, A-D). Furthermore, there are ambiguities in assignment of the observed electron density to distinct inhibitor derivatives for these active sites. In the following, we will therefore describe only the features of the most prominent inhibitor derivative species (with the highest occupancy) for each active site.

View larger version (45K): [in this window] [in a new window]   Fig. 4.   Ligand binding in DPD·5IU· NADPH (pH 7.5). Stereo views of the active sites of the four polypeptide chains (A-D) in the asymmetric unit of DPD·5IU· NADPH (pH 7.5). Residues 670-673 of the active-site loop, the ligand molecules, and the cofactor FMN are shown as ball-and-stick models. Carbon atoms for the ligands 5-iodo-5,6-dihydrouracil, 5IU, and uracil as well as for all carbon atoms of the covalently modified cysteine 671 in molecule A (*C671) originating from 5-iodouracil are given in cyan. The final 2|Fo|  |Fc| map is contoured in blue at 1  for the active-site-loop residues, and a |Fo|  |Fc| map calculated after omitting the inhibitor derivatives from the model is shown in magenta at a contour level of 4.5 . The |Fo|  |Fc| map, which results after modeling of the ligands as uracil, is contoured in black (contour level 3 ).

In chain A, no density was observed for the iodine atom, but there was continuous density between the sulfhydryl group of Cys-671 and the C5 atom of the pyrimidine ring. This electron density suggests that in subunit A the expected enzymatic reduction of 5-iodouracil occurred, and the reduction product, 5-iodo-5,6-dihydrouracil, subsequently attacked Cys-671, leading to its covalent modification to S-(hexahydro-2,4-dioxo-5-pyrimidinyl)cysteine.

The active sites of molecules B and C contain mainly uracil and 5,6-dihydrouracil, respectively, but also some 5-iodouracil (chain C) or 5-iodo-5,6-dihydrouracil (chain B). It is possible to distinguish between the substrate (5IU) and the product of the enzymatic reduction (5-iodo-5,6-dihydrouracil) by analyzing the position of the 5-substituent. In the fully oxidized substrate 5-iodouracil, the iodine is placed in-plane with the pyrimidine ring. For 5-iodo-5,6-dihydrouracil with its sp³-hybridized C5, it is positioned out of the ring plane, with the iodine surprisingly pointing toward Cys-671. Because the C5 atom receives its proton from Cys-671 (Scheme 1), one would expect to find the opposite enantiomer. However, the direct neighborhood of the C4 carbonyl group allows an inter-conversion between both enantiomers via a keto-enol mechanism, although it is not clear from the structure whether the reduction product can undergo this conversion within the active site or is first released and then rebound in the observed enantiomeric form.

View larger version (11K): [in this window] [in a new window]   Scheme 1.   Schematic view of the pyrimidine-binding site before and after uracil reduction.

There is almost no density for a 5-substituent of the inhibitor derivative bound in the vicinity of the FMN cofactor in molecule D (Fig. 4D). Because a distinction between uracil and 5,6-dihydrouracil is not possible at the current resolution of 2.25 Å for the DPD·5IU·NADPH data, the decision to interpret and refine the ligand as uracil rather then 5,6-dihydrouracil was arbitrarily made.

Conformational Changes within the Active-site Pocket-- Amino acid side chains involved in substrate/inhibitor binding in all DPD complex structures are asparagines 609, 668, and 736 as well as Thr-737 (Scheme 1). These primary binding partners interact via hydrogen bonds with both pyrimidine ring nitrogens and carbonyl groups. The only changes occurring in the pyrimidine binding pocket apply to residues of the so-called active-site loop (residues 670-682) as well as to the two adjacent amino acids Leu-669 and Ala-683 (Fig. 5). In the ligand-free enzyme and in all complexes except DPD·5IU·NADPH (pH 7.5), this loop adopts an open conformation, leaving the pyrimidine-binding site solvent-accessible. The hydroxyl group of Ser-670 is then involved in a weak hydrogen bond (3.3 Å) to the O4 oxygen of the pyrimidine, whereas the general acid Cys-671 is located >10 Å distant from the ligand C5 atom. Residues 675-679 are fully solvent-exposed, and in three of the four protein chains in the asymmetric unit disordered, as indicated by a lack of electron density.

View larger version (48K): [in this window] [in a new window]   Fig. 5.   Conformational change of the active-site loop upon ligand binding and pH shift. The figure shows the pyrimidine-binding site in chain A of complex DPD·5IU·NADPH (pH 7.5). The active-site loop (comprising residues 670-682) and the adjacent residues Leu-669 and Ala-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-stick models (the carbon atoms of the modified Cys-671 (C671*) originating from 5IU are shown in brown; all others in pink). The same loop as observed in 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 side chains are shown only for residues Ser-670 and Cys-671 (carbon atoms in cyan) to indicate their position in the open conformation. Ball-and-stick models of the cofactor FMN, substrate/inhibitor binding residues, and residue Lys-709 are given with carbon atoms in yellow. These residues do not 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 placed at identical positions in both structures, and labels in cyan indicate the active-site-loop residues in ligand-free DPD. Active-site-loop residues of complex DPD·5IU·NADPH (pH 7.5) are labeled in pink.

DPD·5IU·NADPH (pH 7.5) is the only complex in which the active-site loop has been observed in the closed state. Residues 674-678 here form a new short 3₁₀ helical turn, and residue 682 prolongs the 3₁₀ helix after the active-site loop. Superposition of the DPD·5IU·NADPH (pH 7.5) structure with DPD·5IU (pH 4.7) or ligand-free DPD (Fig. 5) shows that residue Ser-670 changes its position significantly by 3.6 Å, as measured for the C atoms. Its hydroxyl group is no longer involved in substrate/inhibitor binding but in two new strong hydrogen bonds to the backbone carbonyl of Lys-709 and to the carboxamide nitrogen of Asn-736. Hence, it now bridges two residues involved in either FMN binding (Lys-709) or substrate/inhibitor-binding (Asn-736).

As a direct consequence, Cys-671 is moved closer to the ligand, allowing the formation of the covalent bond (in chain A) or van der Waals interactions (chains B, C, D) between its thiol group and the C5 atom of the ligand. In the active sites of molecules C-D, the distance between Cys-671-S and the ligand C5 is ~3.3 Å, suitable for proton transfer. His-673 is also involved in ligand binding, although by van der Waals rather than polar interactions. The His-673 C is, compared with its position in the open loop conformation, now located 12.6 Å closer to the active site. Of the remaining active-site loop residues none is directly involved in substrate/inhibitor binding. The electron density observed for residues 676-682 is not equally well defined in all 4 molecules in the asymmetric unit. In molecule A containing the covalent modification electron density is continuous for these residues but weaker than for the preceding amino acids 669-675. Considering also the observed higher temperature factors (average B-value is 69.6 Å² for residues 676-683 and 39.4 Å² for residues 669-675), it can be concluded that this stretch of amino acids still shows higher mobility than surrounding parts of the protein. In chains B-D the density is more diffuse, and the temperature factors are even 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 that these residues might be partly disordered. The maintenance of a certain degree of active-site-loop mobility can be of advantage taking into consideration that the loop has to switch between open and closed conformation within each catalytic cycle for substrate binding/product release. The movements should therefore require only small activation energy.

Nevertheless, the closed loop conformation is stabilized by a number of new interactions not observed in the open conformation. A ring nitrogen of His-673 is in hydrogen bond distance to the backbone oxygen of Glu-611. The side chains of Met-675 and Met-680 cluster together with the side chains of Met-642, Ile-613, Leu-612, and Val-583, forming a small hydrophobic core right beside the substrate binding pocket. These residues create a hydrophobic environment for the 5-methyl group of the thymine substrate but are located distant enough to accommodate also larger substituents at the pyrimidine C5. In molecule A, where electron density for the side chains of both Glu-677 and Arg-678 is observed, these residues are involved in altogether four new hydrogen bonds. The glutamate carboxyl group interacts with the main chain nitrogens of residues Phe-935 and Gly-936 of molecule B, thereby participating in the formation of the dimer interface. One of the guanidinyl nitrogens of Arg-678 interacts with the backbone oxygen atoms of Asp-581 and Ile-582.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Active-site Loop Closure Is Substrate Binding and pH-dependent-- With the new DPD-complex structures available, several open questions regarding the reaction mechanism can finally be addressed. Until now, one obstacle of the interpretation has been that the flexible loop segment, which controls substrate entry and product release from the active site, was only observed in the open conformation. Loop closure is an absolute prerequisite for catalytic activity because it not only excludes the surrounding solvent from the active site, but most importantly, also places Cys-671 at the location required for proton transfer to the pyrimidine C5.

Our results suggest that substrate/inhibitor binding in the active-site pocket subsequently triggers closure of the active-site loop. The energy gained by exchanging the pyrimidine O4 as a partner of Ser-670 in a rather weak hydrogen bond against the backbone carbonyl of Lys-709 and the side chain amide of Asn-736, resulting in two new much stronger hydrogen bond interactions, may account for part of the energy necessary for the loop conformational change.

The failure to trigger the active-site-loop closure in complex DPD·5FU·NADPH (15) can clearly be attributed to the non-physiological pH of the crystallization solution (an influence of the C671A exchange in this complex can be ruled out due to the observation of the open loop conformation in DPD·5IU (pH 4.7)). At pH 4.7, the loop residue His-673 carries a positive charge, which either hinders the loop to reach the closed conformation or which the substrate-binding site cannot accommodate in the closed state. His-673 may therefore account for the group with a pK ~ 6.5, identified by studies on pH dependence of kinetic parameters, which is required to be in an unprotonated state for optimum activity but that is not essential for catalytic activity. In DPD·UAA·NADPH (pH 7.5), the closure of the active site is prevented due to steric clashes of loop residues with atoms of the rather voluminous inhibitor molecule.

Comparison of DPD domain IV with the structurally and functionally closely related Lactococcus lactis dihydroorotate dehydrogenase class 1A in complex with the reaction product orotate (32) allowed predictions of the positions of the catalytically crucial Cys-671 (the counterpart in dihydroorotate dehydrogenase class 1A is Cys-130) and the succeeding residues Pro-672 and His-673 (Pro-131 and Asn-132) in the closed state. Superposition of dihydroorotate dehydrogenase class 1A with domain IV of complex DPD·5IU·NADPH (pH 7.5) indeed reveals that residues 669-673 of DPD share comparable positions and side chain conformations with residues 128-132 of dihydroorotate dehydrogenase class 1A. However, none of the residues Gly-674 to Cys-684 (Val-133-Leu-139) are structurally equivalent. A major difference between the ligand-bound and ligand-free structures of both enzymes is that alterations in loop conformation are far more pronounced in DPD.

Reaction and Electron Transfer Mechanism-- The structure of DPD·5IU (pH 4.7) clearly shows that pyrimidine binding to domain IV is not dependent on the presence of NADPH in its binding cleft, which is in agreement with spectroscopic and kinetic data (16). Although the structure of a NADPH·DPD complex has not yet been determined, comparison of the available NADPH-free and NADPH-bound complex structures is consistent with the assumption that NADPH binding is also independent of the occupancy of the substrate binding pocket. Furthermore, proper NADPH binding (i.e. correct placement of the nicotinamide moiety close to the FAD isoalloxazine ring) does not seem to be coupled to loop-closure events in the pyrimidine-binding site. In both complexes, DPD·5IU·NADPH (pH 7.5) with the active-site loop closed and DPD·UAA·NADPH (pH 7.5) with the loop open, NADPH is bound in an identical, "productive" manner.

Analysis of the structures of the DPD complexes reveals that no 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 are located at opposite ends of the subunit and entirely separated from each other by domains I, II, and V (Fig. 1). The movement of domain III induced by NADPH binding does not cause significant conformational changes in the other four domains. The accelerating effect of uracil on the reduction of DPD by NADPH, which was revealed by monitoring absorbance changes during the reaction of DPD with NADPH in the absence and presence of uracil (16), is thus not coupled to conformational changes within the protein but is most likely exclusively based on the "sink function" of the pyrimidine substrate as final electron acceptor.

The catalytically competent binding of NADPH requires a slight widening of the cleft between domains II and III, which is achieved by a modest movement of domain III. A similar but more pronounced domain rearrangement as well as alterations in crystal packing contacts have been reported for adrenodoxin reductase upon binding of NADP⁺ (33). Because no pH change is involved for adrenodoxin reductase, it can be argued that binding of NADPH in the conformation suitable for hydride transfer rather than the pH shift induces the domain movement 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. We examined the immediate surroundings of NADPH for amino acid residues, which could trigger the domain movement by adopting different conformations at the two pH values. The most obvious candidate for such a trigger function is Asp-342. Inevitably this residue has to move to make room for the nicotinamide moiety of NADPH. In ligand-free DPD and all complexes obtained at pH 4.7, the side chain of Asp-342 is not involved in hydrogen-bonding interactions, suggesting that Asp-342 is protonated under these conditions. At pH 7.5, the carboxyl group of Asp-342 needs to form a hydrogen bond interaction with the main chain amide of Val-373, the only candidate within an appropriate distance, to compensate for its negative charge. The formation of this hydrogen bond may require or induce the observed additional conformational changes in stretch 371-373, leading to the disruption of backbone interactions between this stretch and Ala-340 and subsequently to the movement of domain III.

Once NADPH is bound, hydride is transferred from C4 of the nicotinamide of NADPH to FAD N5. It appears that there is no compensation 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 or partially charged entity such as the N terminus of an -helix or a cluster of backbone nitrogens in a reasonable distance to the FAD N1. The closest atoms are the side-chain hydroxyl group and the backbone amide of Thr-489, both located at a distance >3.4 Å to the N1 locus of FAD. Due to the missing charge compensation, the anionic form of the reduced flavin is not stabilized, and the redox potential of the cofactor is kept low by the protein environment. This might represent one of the driving forces 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 of 3.0 Å to S of Cys-130, a ligand of cluster nFeS2 (Fig. 6). This water molecule is present in all four polypeptide chains in the asymmetric unit and shows strong and well defined electron density. An additional hydrogen bond is formed with the backbone amide of Val-490 (2.8 Å). We propose that this is the most likely route for the transfer of two electrons per catalytic cycle between FAD and nFeS2. Concomitant with the electron transfer, the FAD-N5 proton must be released. An apparent route for proton release proceeds via Arg-235, which is located within hydrogen-bond distance to the N5 and O4 atoms of FAD. The proton can move to Arg-235 and from there either directly or via the carboxyl group of Glu-346 to the contacting water molecules and eventually be released to the bulk solvent. The electron transfer pathway between nFeS2 and cFeS2 (via nFeS1 and cFeS1) is shielded from solvent molecules through a number of hydrophobic residues, mostly isoleucines and leucines (Fig. 6).

View larger version (40K): [in this window] [in a new window]   Fig. 6.   The electron transfer pathway between the ligand-binding sites. The cofactors FAD and FMN (carbon atoms in cyan) and the four [4Fe-4S] clusters creating the electron transfer chain between both flavins (iron atoms in magenta, sulfur atoms in green) as well as all amino acids located in van der Waals distance (3.8 Å) to the [4Fe-4S] clusters to atoms N5, N1, and C7M of FMN or to atoms N5, N1, and O2 of FAD are shown as ball-and-stick models. Residues mentioned in the text are labeled. Amino acids with carbon atoms in yellow originate from the same subunit as FAD and FMN. Orange carbon atoms mark residues originating from the other subunit in the dimer. Hydrogen bond interactions mentioned in the text are indicated by dotted lines. For clarity, NADPH and the pyrimidine substrate are not shown.

Based on the positioning of cluster cFeS2 with respect to the FMN cofactor in the pyrimidine-binding site, we suggest the C7 methyl group of FMN as the site of electron entry into the flavin ring system. There is no obvious route for electron transfer between these two redox cofactors. The distances between S of Cys-989, a cFeS2 ligand, to the closest amino acid residue (Ile-590 C1) and from there to the FMN-C7 are 4.4 and 3.6 Å, respectively. However, the simple positioning of the redox centers in close proximity (<14 Å) to each other might be effective enough to achieve electron tunneling (35).

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

Inhibition Mechanism for 5IU (Scheme 2)-- The inhibition of DPD (from bovine liver) by 5IU was thoroughly examined by Porter et al. (19), resulting in the revelation of several mechanistic details that can now be interpreted in structural terms. It was shown that 5IU is both a substrate and potent inactivator of DPD. The NADPH-dependent reduction of 5IU by DPD generates 5-iodo-5,6-dihydrouracil (Scheme 2a), the actual inhibitory species, which acts as an alkylating agent with properties similar to iodoacetamide. Because excess 5IU and dithiothreitol protected the enzyme against inactivation, Porter et al. (19) suggest that 5-iodo-5,6-dihydrouracil is released from the enzyme before it can attack the active-site residue Cys-671 (Scheme 2b). The anti-addition of hydride at C6 and proton at C5 of the 5IU brings the rather voluminous iodo substituent at C5 in immediate neighborhood to the FMN isoalloxazine ring system. The resulting repulsion forces might in part account for the opening of the active-site loop and fast release of the product.

View larger version (19K): [in this window] [in a new window]   Scheme 2.   Proposed mechanisms of DPD inhibition by 5IU and inhibitor deactivation. a, 5IU is reduced to 5-iodo-5,6-dihydrouracil in a 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 generates 5,6-dihydrouracil. The boxed capital letters indicate the dominant species observed in the four subunits (A-D) of DPD in the asymmetric unit.

In fact, comparison of the stoichiometries of DPD inactivation by racemic 5-iodo-5,6-dihydrouracil and the enzymatically generated enantiomer indicated that dissociation of the reduction product from the active site and its rebinding is required for inactivation, since the non-enzymatically generated enantiomer is a more potent inactivator of the enzyme. The observation of this enantiomeric form of 5-iodo-5,6-dihydrouracil in the active site of molecules B of DPD·5IU·NADPH (pH 7.5) supports these conclusions. The interconversion between both enantiomers proceeds most likely via a keto-enol mechanism (Scheme 2b).

The further fate of 5-iodo-5,6-dihydrouracil is partitioned between inactivation of the enzyme by covalent modification of Cys-671 to S-(hexahydro-2,4-dioxo-5-pyrimidinyl)cysteine (Scheme 2c) and conversion to a product unable to inactivate DPD (Scheme 2d). The covalent adduct is present in the active sites of molecules A of complex DPD·5IU·NADPH (pH 7.5). Hence, the inhibitory effect of 5IU is mechanism-based and depends on the alkylation of an amino acid essential for catalysis. As possible mechanisms for the deactivation of 5-iodo-5,6-dihydrouracil, reductive dehalogenation to dihydrouracil and elimination of HI under formation of uracil had been proposed (19). Both uracil and dihydrouracil are products of the reaction of DPD with racemic 5-iodo-5,6-dihydrouracil (19). It is therefore not surprising that uracil or 5,6-dihydrouracil is bound in the active sites of molecules D (and partially also molecules 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, which gives raise to the conclusion that no enzymatic reduction or covalent modification of Cys-671 occurred before the transfer of the DPD·5IU·NADPH crystals to pH 7.5. Because substrates and products of the enzymatic reduction compete for the binding in the active site of DPD (19), it was unlikely to obtain complete inactivation of the enzyme in the crystals. We could not identify the source of the non-equivalency of the four molecules in the asymmetric unit of DPD crystals, which resulted in the accumulation of distinct inhibitor derivatives within the different active sites (A-D). Superposition of the subunits does not reveal other deviations in amino acid coordinates except in a few surface-located loop regions, which are generally all solvent-exposed. This non-equivalency has, however, been observed in all DPD structures in so far as that part of the "proline-rich loop" (residues 899-910) is usually less disordered in subunits C, and electron density for the active-site loop (in its open conformation) is usually best defined in subunits 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 residue Cys-671 and agree well with previously presented biochemical observations concerning catalytic and inhibitory mechanisms.

The complex structures presented here provide for the first time a complete picture of the electron transfer chain in the productive enzyme-substrate complex. It was also shown that pyrimidine binding triggers a conformational change of a flexible active-site loop, resulting in placement of the catalytically crucial cysteine 671 close to the bound substrate. This loop closure as well as the correct binding of the cosubstrate NADPH requires physiological pH.

	ACKNOWLEDGEMENTS

We thank A. Mozzarelli for advice regarding the soaking experiments. We thankfully acknowledge access to the synchrotron radiation at the European Synchrotron Radiation Facility and the Deutsches Elektronensynchrotron.

	FOOTNOTES

^* This work was supported by grants from the Swedish Cancer Foundation and the Swedish Research Council.The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

The atomic coordinates and the structure factors (code 1gte (DPD|b15IU), 1gth (DPD|b15IU|b1NADPH), and 1gt8 (DPD|b1UAA|b1NADPH)) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, 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 Foundation Blanceflor Boncompagni-Ludovisi.

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

Published, JBC Papers in Press, January 16, 2002, DOI 10.1074/jbc.M111877200

	ABBREVIATIONS

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

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

CiteULike

Complore

Connotea

Del.icio.us

Digg

Reddit

Technorati

This article has been cited by other articles:

				A. Osterman A hidden metabolic pathway exposed PNAS, April 11, 2006; 103(15): 5637 - 5638. [Full Text] [PDF]

This Article Abstract Full Text (PDF) All Versions of this Article: 277/15/13155    most recent M111877200v1 Submit a Letter to Editor Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Dobritzsch, D. Articles by Lindqvist, Y. Search for Related Content PubMed PubMed Citation Articles by Dobritzsch, D. Articles by Lindqvist, Y. Social Bookmarking                      What's this?