Structure of human biliverdin IX reductase, an early fetal bilirubin IX producing enzyme
Pedro José Barbosa Pereira1, 2, Sandra Macedo-Ribeiro1, 2, Antonio Párraga2, Rosa Pérez-Luque2, Orla Cunningham3, Kevin Darcy3, Timothy J. Mantle3
& Miquel Coll2
1 These authors contributed equally to this work and share first authorship.
2 Instituto de Biologia Molecular de Barcelona, Consejo Superior de Investigaciones Científicas, Jordi Girona, 18-26, 08034 Barcelona, Spain.
3 Department of Biochemistry, Trinity College, Dublin 2, Ireland.
Biliverdin IX reductase (BVR-B) catalyzes the pyridine nucleotide-dependent production of bilirubin-IX, the major heme catabolite during early fetal development. BVR-B displays a preference for biliverdin isomers without propionates straddling the C10 position, in contrast to biliverdin IX reductase (BVR-A), the major form of BVR in adult human liver. In addition to its tetrapyrrole clearance role in the fetus, BVR-B has flavin and ferric reductase activities in the adult. We have solved the structure of human BVR-B in complex with NADP+ at 1.15 Å resolution. Human BVR-B is a monomer displaying an / dinucleotide binding fold. The structures of ternary complexes with mesobiliverdin IV, biliverdin IX, FMN and lumichrome show that human BVR-B has a single substrate binding site, to which substrates and inhibitors bind primarily through hydrophobic interactions, explaining its broad specificity. The reducible atom of both biliverdin and flavin substrates lies above the reactive C4 of the cofactor, an appropriate position for direct hydride transfer. BVR-B discriminates against the biliverdin IX isomer through steric hindrance at the bilatriene side chain binding pockets. The structure also explains the enzyme's preference for NADP(H) and its B-face stereospecificity.
The first identifiable heme catabolite in the human fetus is bilirubin IX. It is detectable at 14−15 weeks1 and remains the major bile pigment at 20 weeks2. Unlike bilirubin IX, the IX isomer of bilirubin does not undergo internal hydrogen bonding, having a much higher solubility. Due to this property, it can be excreted without previous conjugation to glucuronic acid, a reaction catalyzed by UDP-glucuronyltransferase, which is not expressed at significant levels until the first week after birth. Bilirubin IX is transported out of the fetal liver and into bile and the lumen of the fetal intestine1, thereby facilitating fetal clearance of potentially neurotoxic tetrapyrroles. Heme cleavage at the -meso position produces biliverdin IX, which is reduced by biliverdin IX reductase (BVR-B) to form bilirubin IX3. This heme catabolism pathway in utero is distinct from that in adults, in which heme cleavage occurs almost exclusively at the -meso position, producing biliverdin IX and bilirubin IX. The physiological relevance of the apparent switch in heme degradation from a IX pathway in utero to a IX pathway at parturition may be coupled to the switch from embryonic through fetal to adult hemoglobin. It was shown that BVR-B is identical to flavin reductase (FR)4, which is abundant in adult erythrocytes. FR can provide free reduced flavins for the reduction of methemoglobin5. BVR-B is thus a promiscuous enzyme that catalyzes the NAD(P)H-dependent reduction of a range of non- isomers of biliverdin6, several flavins7, pyrroloquinoline quinone (PQQ)8 and ferric ion7. This enzyme can accommodate a wide range of synthetic tetrapyrrole substrates with propionate side chains variously positioned around the tetrapyrrole backbone, although in clear contrast to human BVR-A, it cannot tolerate even one propionate side chain on either side of the C10 position6.
The structures of human BVR-B in complex with NADP+ and with biliverdin and flavin substrates and inhibitors show a number of features that are consistent with its broad substrate specificity.
Overall structure Human BVR-B is a 206-residue monomeric enzyme. It shares 91.2% sequence identity with the bovine enzyme and 59.6% and 58.6% with turkey and chicken calcium-binding proteins, respectively (Fig. 1a). The structure of BVR-B has a single-domain architecture consisting of a central parallel -sheet with -helices on either side (Fig. 1b). This characteristic dinucleotide binding fold comprises, in this case, a seven-stranded parallel -sheet further extended by an antiparallel strand. In addition to the seven long strands of the main pleated sheet, a short parallel -sheet (strands 6a and 6c) is formed within the loop joining strand 6 and -helix F. The central -sheet and the two groups of helices are held together mainly through hydrophobic interactions. One group of helices is made up of -helices C, D and E. The second group is composed of -helices A and F and includes a short 310-helix between strands 2 and 3, in contrast to typical dinucleotide binding proteins in which a regular -helix flanks these -strands. The most flexible loop in the structure corresponds to loop 120, between strand 5 and -helix E, which contains residues with the highest main chain B-factors, with the exception of the N-terminal region.
Figure 1. BVR-B displays an / dinucleotide binding fold.
a, Sequence alignment of human (SwissProt entry FLRE_HUMAN28) and bovine (SwissProt entry FLRE_BOVIN29) BVR-B and related enzymes from turkey (SwissProt entry Q91089 (ref. 11)) and chicken (SwissProt entry Q90940 (ref. 11)). Amino acid residues identical to human BVR-B are shaded. Numbers above the alignment refer to the human BVR-B sequence. The secondary structure elements shown correspond to the human BVR-B structure. -Helices are shown in red, 310-helices in orange and -strands in blue. b, Overall structure of human BVR-B in complex with NADP+. The secondary structure elements are displayed in red (-helices) and blue (-sheets). Important loops, and the C-terminus and N-terminus are labeled. The NADP molecule is shown as a ball-and-stick model with carbon atoms in yellow, oxygens in red, nitrogens in blue and phosphorous in orange.
A detailed comparison of the overall structures of the BVR-B−NADP+ binary complex with the ternary complexes with biliverdin IX, mesobiliverdin IV, FMN or lumichrome shows that there are no large structural differences between them. The root mean square (r.m.s.) deviation for main chain atoms is in no case higher than 0.2 Å for all residues.
The overall structure of BVR-B is closely related to that of the N-terminal NAD binding domain of Escherichia coli UDP-galactose 4-epimerase (Protein Data Bank (PDB) entry 1XEL9) as well as to other members of the so-called short chain dehydrogenase/reductase (SDR) protein family. The substrate discriminating C-terminal domain of UDP-galactose 4-epimerase does not have a structural equivalent in the shorter BVR-B, which thus incorporates cofactor and substrate selection and binding into a more compact unit.
Cofactor binding and specificity BVR-B is a NAD(P)H dependent reductase7. The binding mode of oxidized NADP to BVR-B is shown in Fig. 2a. The cofactor is deeply buried inside the protein and only the nicotinamide ring is partially exposed. The first -- unit contains the GxxGxxG (where x represents any amino acid) NAD(P)H binding motif, which is highly conserved among BVR-Bs of known sequence and is also in the proteins described as calcium-binding proteins10,
11. These are most probably avian BVR homologs, as suggested by their high sequence identity to human and bovine BVR-Bs (Fig. 1a) and their conservation of all the functionally important regions (see below).
a, Stereo view of the NADP+ cofactor binding site. Putative hydrogen bonds are shown as dotted green lines, and solvent molecules as red spheres. The atom color code is the same as in Fig. 1b. The 1.15 Å 2Fo - Fc electron density map for the cofactor, contoured at 1.5 , is shown in blue. Residues interacting with NADP+ are numbered. b, Solid surface representation of human BVR-B in complex with NADP+ showing the wide substrate binding site adjacent to the cofactor. Electrostatic surface potentials are contoured from -10 (red) to 10 (blue) kBT e-1. The cofactor carbon atoms are shown in white, phosphorous in yellow, nitrogens in blue and oxygens in red.
The NADP pyrophosphate group binds in a crevice formed by loops 10 and 80 at the central region of the nucleotide binding domain with the nucleosides on each side bound to the carboxy edges of -strands 1−2 and 4−5 (Fig. 2). The pyrophosphate moiety interacts with the amide nitrogens of residues Gln 14 and Thr 15 at the N-terminus of helix A, N-capping it — that is, compensating for the partial positive charge of the helix macrodipole.
NADPH binds more tightly than NADH and is probably the cofactor that BVR-B uses in vivo7. Stabilization of the extra 2'-phosphate group of NADP+ is achieved by strong salt bridges to the side chains of Arg 35 and Arg 78 and a hydrogen bond to the side chain hydroxyl of Thr 12. The guanidinium group of Arg 35 stacks against the adenine moiety of the cofactor. The presence of at least one Arg side chain in the vicinity of the 2'-phosphate of NADP is a common feature of NADP dinucleotide binding fold complexes12,
13,
14,
15. The presence of a hydroxyl containing side chain has also been observed in other NADP specific proteins16, as well as the stacking interactions between the Arg residue and the adenine ring13,
17.
The average temperature factor of the nicotinamide moiety is consistently higher than that of the more tightly bound adenine region. This is in agreement with the higher flexibility of this end of the cofactor observed in the crystal derivatized by the binding of a mercury atom to Cys 109, which is located just below the nicotinamide ring. In these crystals, diffuse electron density for the nicotinamide moiety indicated the displacement of this group upon binding the mercury atom.
Active site and stereospecificity of substrate reduction Inhibition studies suggest that flavins and tetrapyrroles compete for a common binding site on BVR-B7. In the BVR-B−NADP+ complex there is a wide exposed hydrophobic wall adjacent to the nicotinamide moiety (Fig. 2b). The substrate binding site is located between the flexible loops 80 and 120 and includes the N-terminal half of helix E.
We obtained ternary complexes with both flavins (the substrate flavin mononucleotide (FMN) and the inhibitor lumichrome; Fig. 3a) and verdins (the substrate mesobiliverdin IV and the inhibitor biliverdin IX; Fig. 3b). All substrates and inhibitors were found docked at the single binding site, stabilized partially by stacking interactions against the nicotinamide ring and by additional van der Waals contacts with residues composing the active site (Fig. 4). Because the other side of the cofactor is fully buried in the protein, only one substrate at a time can bind near the NADPH. This explains the observed competitive kinetics between biliverdins and FMN, as well as between FMN and lumichrome7. It is easy to conceive that related compounds (such as PQQ) could bind in this wide pocket, in an orientation suitable for processing, thus explaining the relative promiscuity of the enzyme.
Figure 3. Structural formulas of some human BVR-B substrates and inhibitors.
a, FMN is a BVR-B substrate, lumichrome an inhibitor. b, Tetrapyrrole biliverdin IX, mesobiliverdin IV and 12-ethyl-13-methyl-mesobiliverdin IV are substrates; biliverdin IX is an inhibitor. The formation of a bilirubin isomer results from the reduction of C10 of the corresponding biliverdin (numbering as indicated for biliverdin IX, which is conserved for the bilatriene skeleton in all other isomers). The ring nomenclature for all -isomers is that shown for biliverdin IX.
Figure 4. Binding mode of substrates and inhibitors to human BVR-B.
On the upper half of each panel, a transparent Connolly surface representation of human BVR-B is superimposed on a stick diagram of the molecule. The color code is as in Fig. 1b, except for carbon atoms of NADP (light blue) and of substrates/inhibitors (mauve). The lower half of the panel shows a detailed view of the protein−ligand interactions. Omit electron density maps for the ligands, calculated with coefficients Fo - Fc and contoured at 2 are shown in blue, superimposed with each of the compounds. The color code is the same as in Fig. 1b. Important residues are numbered. a, FMN; b, lumichrome; c, mesobiliverdin IV; and d, biliverdin IX are shown as stick models bound to human BVR-B. Substrate selectivity pockets (see text) of BVR-B are labeled A to C in (c and d).
Human BVR-B catalyzes the incorporation of radioactivity from B-face labeled [3H]-NADH into bilirubin XIIor FMN, but not when using the A-face labeled cofactor (data not shown). The enzyme is clearly B-face stereospecific, in accordance with the crystal structures, which show in all cases that the nicotinamide and the substrate are in the proper orientation for B-face specific hydride transfer from the C4 of the dinucleotide to the corresponding substrate reaction center. Free rotation of the nicotinamide ring is prevented primarily by a hydrogen bond between the carboxamide oxygen and the amide nitrogen of Ile 154 (Fig. 2a). The side chain of His 153 could also hydrogen bond to this carboxamide oxygen, but it is positioned in all the determined structures in such a way that this bond would be relatively weak.
Specificity of BVR-B for flavin substrates Human BVR-B was shown to catalyze the reduction of FMN, FAD and riboflavin, being competitively inhibited by lumichrome7 (Fig. 3a). Recognition of the flavin by the polypeptide chain occurs through the isoalloxazine ring, mostly by hydrophobic interactions (Fig. 4a). There is a single hydrogen bond between the C4 carbonyl oxygen of FMN and one of the hydroxyls of the NADP nicotinamide ribose. The electron density is clear for the isoalloxazine ring, with the phosphoribityl moiety being disordered, since it makes no significant contact with BVR-B. The central ring of FMN stacks over the nicotinamide moiety of NADP+, placing the acceptor N5 of FMN 3.5 Å above the reactive C4 of the cofactor. There is no protein side chain nor ordered solvent molecule in the proximity of the FMN N1 that could act as a proton donor.
Lumichrome (Fig. 3a) binds to BVR-B in a similar way to that described for FMN (Fig. 4b). The isoalloxazine ring of this compound reaches a little deeper into the active site, resulting in a shortening of the single ligand−cofactor hydrogen bond. This small positional shift of lumichrome results in an increased distance (3.7 Å) between its N5 and the nicotinamide C4. A reduced steric hindrance in comparison with FMN could explain such movement since the phosphoribityl moiety is absent in FMN.
Specificity of BVR-B for the tetrapyrrole substrates Human BVR-B can accommodate a wide range of synthetic tetrapyrrole substrates (Fig. 3b). The substrate mesobiliverdin IV was found bound at the active site with the two central pyrrole rings (C and D) buried in the protein (Fig. 4c). Ring D stacks directly on top of the nicotinamide moiety of NADP+. The propionate side chain at position 13 fits in a cleft, or pocket A, between Val 128 and Thr 77 (Fig. 4c). The methyl group at position 12 fits into pocket B (Fig. 4c), in which some additional space could also accommodate an ethyl or a vinyl group. The other pyrrole stacks between the side chains of Phe 113 and Pro 152. The methyl at position 8 inserts into a very narrow cavity (pocket C) in which a larger group could not fit (Fig. 4c). The propionate at position 7 is largely outside the major docking site, not interacting with the protein, and is thus disordered.
In this orientation (Fig. 4c), the reactive C10 of the substrate lies 3.5 Å away from the C4 of the cofactor, well positioned for direct hydride transfer. No Cys side chain was found to be near enough to this reactive center such that the C10 could adduct to form an intermediate rubinoid structure, as proposed for BVR-A18. Although Cys 109 is located at the cofactor binding site, its side chain is more than 9 Å away from the verdin substrate. The side chain of Ser 111, which is structurally homologous to the catalytic Ser 124 of UDP-galactose epimerase9, is oriented away from the ligand (4.9 Å from the hydride acceptor) in all the ternary complexes. Since rat BVR-A has been crystallized19, it would be very interesting to find out whether the BVR-A and BVR-B reactions proceed through a similar direct reduction pathway. No protein side chain nor ordered solvent molecule was found within 3.5 Å of the pyrrole nitrogens. As for flavin-like substrates, this suggests bulk solvent as a probable proton source for the substrate reduction. However, we can conceive that the side chain of His 153, 4.2 Å away from the nearest pyrrole nitrogen in the mesobiliverdin IV complex, could rotate to within 3.0 Å of the proton acceptor. This could happen as a result of conformational changes in the enzyme upon binding of the reduced cofactor. Conformational changes induced by cofactor reduction have been described for the structurally homologous UDP-galactose-4-epimerase9.
Steric hindrance prevents biliverdin IX from binding productively at the enzyme's active site with its C10 positioned for direct hydride transfer from the NADPH cofactor (Fig. 4d). The small pockets B and C cannot accommodate the large propionates straddling the C10 methene carbon. Thus, a nonproductive binding of this biliverdin inhibitor was observed in which it is rotated 90° around a vertical axis compared to the orientation of mesobiliverdin IV. In this nonproductive orientation, both the methyl and vinyl groups at positions 7 and 3, respectively, fit perfectly into pockets B and C. Unambiguous electron density for these groups confirms the orientation adopted by the biliverdin inhibitor.
These results support the idea that BVR-B binds substrates, such as biliverdin IX (Fig. 3b), 12-ethyl-13-methyl-mesobiliverdin IV (Fig. 3b) and other biliverdin isomers devoid of propionates straddling the C10 position6, in the same orientation as mesobiliverdin IV. This would explain why BVR-B is able to process these substrates.
Conclusions BVR-B, a crucial enzyme in fetal heme catabolism and abundant in the adult erythrocytes, uses NADPH as an electron donor and a great variety of flavin and biliverdin analogs as electron acceptors. The three-dimensional structure of this enzyme and those of its various complexes allow us to explain its preference for NADP over NAD3, its B-face stereospecificity and its selectivity with respect to the steric and functional patterns present in biliverdins. Flavins and biliverdins share a common binding site, interacting with the enzyme almost exclusively through hydrophobic and stacking interactions. Substrate discrimination occurs essentially by steric constraints for the chemical groups adjacent to the electrophile C10 methene carbon. The structure around the active site explains clearly the published specificity data; Cunningham and co-workers6 found that biliverdins with one or two propionates straddling the reducible methene are functional inhibitors of BVR-B, while isomers without this characteristic are substrates of the enzyme (Fig. 3b). In all our structures, the substrates are positioned such that direct hydride transfer from the C4 of the nicotinamide is conceivable. No thiol residues nor other electron-rich donors are close enough to the C10 methene carbon to form rubinoid adducts with it. The most likely proton source for this reaction is either bulk solvent or His 153, whose side chain could approach the proton acceptor in the tetrapyrrole substrates. For flavin-like substrates, however, to which His 153 could not donate a proton, bulk solvent is the only conceivable proton source.
Methods Crystallization and data collection. Recombinant human biliverdin-IX reductase was prepared as described20. Crystals (space group P212121, one molecule per asymmetric unit) were obtained at 293 K from drops consisting of 4 l of protein solution (10 mg ml-1), 1 l of 15 mM NADP and 1 l of 30% (w/v) PEG 8K, 0.2 M ammonium sulfate, and 0.1 M sodium cacodylate pH 6.5. A mercury derivative was prepared by soaking native crystals in mother liquor containing 1 mM mercury acetate. To prepare the inhibitor and substrate ternary complexes, BVR-NADP+ crystals were soaked in mother liquor containing various concentrations of the target compounds (Table 1). Preliminary attempts to obtain BVR-B−NADP+−biliverdin IX complexes did not provide diffraction quality crystals. Further attempts are underway but have been impaired by the scarcity of the compound. Diffraction data from cryo-cooled crystals (100 K) were collected at EMBL beamlines BW7A and BW7B (DESY, Hamburg) and ID14-3 (ESRF, Grenoble). Data for the native and mercury derivative were collected in-house. Data were evaluated with MOSFLM v6.0 (ref. 21), scaled with SCALA22 and reduced with programs from the CCP4 suite22 (Table 1).
Structure solution and refinement. The structure of the BVR-B−NADP+ binary complex was solved by the SIRAS method using a single mercury derivative. Difference Patterson maps and difference Fourier synthesis located two mercury sites (Table 1). Heavy atom refinement and phasing, and solvent flattening were performed with SHARP23 and SOLOMON22, respectively. The electron density obtained enabled building of the complete protein model, which was refined with the high resolution synchrotron data for the BVR-B−NADP+ complex and was also used for the initial phasing of the ternary complexes. Positional and temperature factor refinement with CNS v1.0 (ref. 24), applying bulk solvent and anisotropic B-factor corrections, was alternated with cycles of manual model building with Turbo-FRODO25.
The isotropic CNS models were further refined with SHELX97-2 (ref. 26). An initial round of 10 conjugate gradient least-square (CGLS) cycles using isotropic thermal parameters was followed by refinement with anisotropic displacement parameters (ADPs) for all atoms. Riding hydrogens (as defined in SHELX)26 were added to the models with resolution better than 1.4 Å and were associated with all protein atoms, except for the hydroxyl groups of Ser, Thr and Tyr residues. Finally, the occupancies for the substrates and inhibitors were refined. The final models were obtained after refinement with all the data (Table 1).
Stereospecificity of NADH oxidation. [4-3H] NAD was reduced using glutamate dehydrogenase and alcohol dehydrogenase to produce 3H at the A- and B-faces, respectively27. The A- and B-face labeled [4-3H] NADH was used with mesobiliverdin XII (or FMN) in a BVR-B reaction as described for rat BVR-A27. The tritium labeled bilirubin XII formed was extracted into chloroform and the radioactivity measured by scintillation counting.
Coordinates. The coordinates for all structures have been deposited with the Protein Data Bank (accession codes: 1HDO
, NADP; 1HE2
, biliverdin IX; 1HE3
, mesobiliverdin IV; 1HE4
, FMN; and 1HE5
, lumichrome).
Acknowledgments This work was supported by grants from the Ministerio de Educación y Cultura and the Generalitat de Catalunya to M.C. P.J.B.P. was supported in part by a FEBS Long Term fellowship. P.J.B.P. and S.M.R. acknowledge postdoctoral fellowships from Programa Praxis XXI (FCT, Portugal). O.C. was supported by the Health Research Board, Ireland. We thank D.A. Lightner for supplying the various verdins used in this work. Synchrotron data collection was supported by EU grants and the ESRF.
reductase, an early fetal bilirubin IX
reductase (BVR-B) catalyzes the pyridine nucleotide-dependent production of bilirubin-IX
reductase (BVR-A), the major form of BVR in adult human liver. In addition to its tetrapyrrole clearance role in the fetus, BVR-B has flavin and ferric reductase activities in the adult. We have solved the structure of human BVR-B in complex with NADP+ at 1.15 Å resolution. Human BVR-B is a monomer displaying an 
. It is detectable at 14−15 weeks
, the IX
, is shown in blue. Residues interacting with NADP+ are numbered. b, Solid surface representation of human BVR-B in complex with NADP+ showing the wide substrate binding site adjacent to the cofactor. Electrostatic surface potentials are contoured from -10 (red) to 10 (blue) kBT e-1. The cofactor carbon atoms are shown in white, phosphorous in yellow, nitrogens in blue and oxygens in red.
or FMN, but not when using the A-face labeled cofactor (data not shown). The enzyme is clearly B-face stereospecific, in accordance with the crystal structures, which show in all cases that the nicotinamide and the substrate are in the proper orientation for B-face specific hydride transfer from the C4 of the dinucleotide to the corresponding substrate reaction center. Free rotation of the nicotinamide ring is prevented primarily by a hydrogen bond between the carboxamide oxygen and the amide nitrogen of Ile 154 (
l of protein solution (
10 mg ml-1), 1 
