Studies on the Specificity of the Tetrapyrrole Substrate for Human Biliverdin-IX a Reductase and Biliverdin-IX b Reductase STRUCTURE-ACTIVITY RELATIONSHIPS DEFINE MODELS FOR BOTH ACTIVE SITES*

A comparison of the initial rate kinetics for human biliverdin-IX a reductase and biliverdin-IX b reductase with a series of synthetic biliverdins with propionate side chains “moving” from a bridging position across the central methene bridge ( a isomers) to a “ g -configura-tion” reveals characteristic behavior that allows us to propose distinct models for the two active sites. For human biliverdin-IX a reductase, as previously discussed for the rat and ox enzymes, it appears that at least one “bridging propionate” is necessary for optimal binding and catalytic activity, whereas two are preferred. All other configurations studied were substrates for human biliverdin-IX a reductase, albeit poor ones. In the case of mesobiliverdin-XIII a , extending the propionate side chains to hexanoate resulted in a significant loss of activity, whereas the butyrate derivative re-tained high activity. For human biliverdin-IX a reductase, we suggest that a pair of positively charged side chains play a key role in optimally binding the IX a isomers. In the case of human biliverdin-IX b reductase, the enzyme cannot tolerate even one propionate in the bridging position, suggesting that two negatively charged residues on the enzyme surface may preclude productive binding in this case. The flavin reductase activity of biliverdin-IX b reductase is potently inhibited Stock solutions of both proteins were diluted in 200 m l of distilled water to give a final concentration of 0.6 mg/ml, and buffer was ex-changed into 50 m M ammonium acetate, pH 6.0, through a pre-equili- brated PD-10 column. Samples containing BVR-A and BVR-B were pooled and adjusted to 0.4 and 0.6 mg/ml, respectively. The buffer composition was finally adjusted to 50% (v/v) acetonitrile and 1% (v/v) formic acid. Mass analysis was carried out using a Micromass triple quadrupole electrospray mass spectrometer. A 10- m l sample was intro-duced into the electrospray mass spectrometer via a Rheodyne injector valve fitted with a 10- m l injection loop. Analysis was carried out in positive ion mode. Raw data were collected between a mass-to-charge ratio of 600–1700 m / z , at a cone voltage of 30 V, HV lens of 0.22 kV, and a capillary voltage of 3.60 kV. The raw data were subjected to maximum entropy analysis according to the Micromass schedule.

The formation of linear tetrapyrroles by heme catabolism in mammals has, until recently, been discussed in terms of the IX␣ isomers of biliverdin and bilirubin as both heme oxygenases I and II (HO-1 and HO-2), 1 and biliverdin-IX␣ reductase (BVR-A) are reported to exhibit such specificity (1,2). However, 87% of the bilirubin in human fetal bile has been reported to be the IX␤ isomer (3). Yamaguchi et al. (4) purified a novel enzyme from human liver that catalyzes the reduction of the IX␤, IX␥, and IX␦ isomers of biliverdin to the corresponding rubin (4). We have shown that this enzyme, biliverdin-IX␤ reductase (BVR-B), is identical to NAD(P)H-linked flavin reductase (5). The source of fetal biliverdin-IX␤ has not yet been determined; however, this appears to be a pathway that is only operative at any significant level in the fetus. The physiological relevance of the apparent switch in heme degradation from a IX␤ pathway in utero to a IX␣ pathway at birth is unclear, although it may be coupled to the switch from fetal to adult hemoglobin. O'Carra and Colleran (6) have shown that nonenzymic ascorbate-mediated coupled oxidation of "free" heme (pyridine-heme complexes) produces all four isomers of biliverdin-IX (Scheme 1), in approximately equimolar amounts, whereas coupled oxidation of adult hemoglobin produces a mixture of IX␣ (65%) and IX␤ (35%) isomers of biliverdin. The nature of the protein binding the heme is important because ascorbate-mediated coupled oxidation of myoglobin produces 95% biliverdin-IX␣ (6). It is not known which isomers of biliverdin-IX are produced by ascorbate-mediated coupled oxidation of fetal hemoglobin, although, in preliminary experiments, we cannot support BVR-B-dependent NADPH oxidation using the products of this reaction. 2 Whereas nonenzymic coupled oxidation is demonstrable in vitro, it is clear that at least two forms of heme oxygenase (HO-1 and HO-2) function in vivo. Both of these enzymes produce the IX␣ isomer of biliverdin exclusively; however, the nature of the isomer produced by the recently described HO-3 (7) is not known. Quantitative flux through these three HO pathways in mammals has not been studied, although there is considerable interest that there is a requirement for a functioning HO-1 for effective reutilisation of iron in mammals (8).
The discovery that the IX␣ isomer of bilirubin is a ligand for the aryl hydrocarbon (Ah) receptor (9, 10) may explain the transcriptional up-regulation of the rat GST A5 gene in congenital hyperbilirubinaemia (11) and allows us to suggest that a function for the IX␤ isomer (which appears to be a uniquely fetal metabolite) may be related to fetal suppression of the maternal immune system. The Ah receptor is known to be involved in immunosuppression, and, during pregnancy, there is an increased susceptibility to certain types of infection (12)(13)(14). Both biliverdin-IX␣ and bilirubin-IX␣ have been implicated as modulators of the immune system (15,16). Intriguingly, the recent suggestion that indoleamine dioxygenase (EC 1.13.11.42) may be involved in fetal suppression of the maternal immune response (17) could also support Ah receptor involvement as tryptophan metabolites are also known to bind to and activate this transcription factor (18).
High levels of the IX␣ isomer of bilirubin are generally seen at birth (so-called physiological jaundice of the newborn) and are potentially cytotoxic if the protective binding capacity of serum albumin (19,20) is exceeded. However, this is a period when the infant lung experiences a massive increase in the partial pressure of oxygen (21), and the transient increase in serum bilirubin levels seen at this time may represent a temporary boost in the levels of a physiologically significant antioxidant. Indeed, several workers have presented evidence that bilirubin-IX␣ functions as an antioxidant, both in its free form (22) and bound to albumin (23), and low serum bilirubin-IX␣ has been shown to correlate with increased risk of coronary artery disease in two independent studies (24,25). It is conceivable that linear tetrapyrroles may play significant biological roles in mammals, both as antioxidants and as anti-inflammatory agents, and it is therefore important to extend our knowledge of the substrate specificity of the two human enzymes currently known to catalyze the formation of bilirubin isomers. Further study will need to identify the origin of biliverdin-IX␤ in utero and also to define any other linear tetrapyrroles that may occur in the adult and/or the fetus.
The chemical synthesis of the four isomers of biliverdin, cleaved at the ␣, ␤, ␥, and ␦ positions using ascorbate-mediated coupled oxidation of pyridine-heme is not amenable to large scale production. Early work by Colleran and O'Carra (26) demonstrated that replacement of the vinyl side chains by ethyl side chains (defined as mesobiliverdins) had little effect on substrate affinity. We have therefore developed a method of producing symmetrical synthetic mesobiliverdins with propionic acid groups at varying sites along the pigment backbone (Group I verdins, Refs. 27-29; Group II verdins, Refs. 30 -32; Group III verdins, Refs. 33 and 34) and report here the initial rate kinetics of these verdins with recombinant human BVR-A and BVR-B. These structure-activity relationships allow us to propose models for the active site structures of these two enzymes.

Cloning and Overexpression of Human Biliverdin-IX␣ Reductase and Human Biliverdin-IX␤ Reductase
Oligonucleotide primers used to amplify the cDNA encoding human BVR-A were designed based on the cDNA sequence reported by Maines et al. (35). The forward (5Ј-GCAGGATCCAAGATGAATGCAGAG-3Ј) and reverse (5Ј-AACCAAATGCTGGTGCCATGGTGGAA-3Ј) primers contained BamHI and NcoI restriction sites, respectively (underlined). A cDNA library derived from the U937 monocyte cell line was used as target in subsequent amplification reactions. The resulting 960-base pair fragment was digested and ligated into the pGEX-KG expression vector to produce the pGEX-BVR-A plasmid. Escherichia coli strain TG1 was transformed according to procedures described in Sambrook et al. (36).
E. coli cells for large scale purification of the recombinant GST fusion protein were cultured as follows. 10-ml cultures were grown in Luria-Bertani medium overnight at 30°C in the presence of 2% glucose and 100 g/ml ampicillin. Two-liter cultures were inoculated with the overnight cultures, grown at 30°C to an A 600 of 0.4 -0.5, and induced with isopropyl-1-thio-␤-D-galactopyranoside at a final concentration of 0.2 mM. After 18 h the cells were harvested by centrifugation (7,700 ϫ g for 10 min) and lysed by sonication in the presence of lysozyme (200 g/ml). Following centrifugation at 12,000 ϫ g for 45 min, the cell supernatant was passed through a glutathione-Sepharose affinity column, and the fusion protein was eluted with 10 mM glutathione in 50 mM Tris-HCl, pH 8.0. Excess glutathione was removed by gel filtration, and the fusion protein was cleaved overnight at 4°C with 20 units of thrombin (1 unit/l). The liberated BVR-A was separated from the GST tag by passage through the glutathione affinity column. The approximate yield of purified BVR-A from a 2-liter culture is 30 mg. Recombinant human biliverdin-IX␤ reductase was prepared as described previously by thrombin cleavage of a GST-BVR-B fusion protein (37).

Mass Spectroscopic Analysis of Recombinant Biliverdin-IX␣ Reductase and Biliverdin-IX␤ Reductase
On SDS-PAGE, recombinant human BVR-A migrates with a mobility corresponding to a molecular mass of 40 kDa. Similar behavior has been reported for the native human enzyme (38,39). This is in contrast to the behavior of the rat, mouse and ox enzymes which exhibit a molecular mass of 34 kDa on electrophoresis (38). Given that the rat enzyme contains 295 amino acids and the human enzyme contains 296 amino acids, mass spectroscopy was carried out to determine the relative molecular mass of human BVR-A. Similar analysis was conducted for BVR-B.  Mesobiliverdin-XII␥ 3 SCHEME 1. Biliverdin isomers resulting from haem cleavage. Four possible isomers of biliverdin can result from the cleavage of haem because of the nonequivalence of the four methene bridge positions ␣, ␤, ␥, and ␦ (P ϭ -CH 2 CH 2 COOH).
Stock solutions of both proteins were diluted in 200 l of distilled water to give a final concentration of 0.6 mg/ml, and buffer was exchanged into 50 mM ammonium acetate, pH 6.0, through a pre-equilibrated PD-10 column. Samples containing BVR-A and BVR-B were pooled and adjusted to 0.4 and 0.6 mg/ml, respectively. The buffer composition was finally adjusted to 50% (v/v) acetonitrile and 1% (v/v) formic acid. Mass analysis was carried out using a Micromass triple quadrupole electrospray mass spectrometer. A 10-l sample was introduced into the electrospray mass spectrometer via a Rheodyne injector valve fitted with a 10-l injection loop. Analysis was carried out in positive ion mode. Raw data were collected between a mass-to-charge ratio of 600 -1700 m/z, at a cone voltage of 30 V, HV lens of 0.22 kV, and a capillary voltage of 3.60 kV. The raw data were subjected to maximum entropy analysis according to the Micromass schedule.

Preparation of Partially Pure Native Biliverdin-IX␣ Reductase and Biliverdin-IX␤ Reductase
Human erythrocytes were centrifuged at 4,000 ϫ g for 10 min at 4°C, and the supernatant was removed. The erythrocytes were gently resuspended in three volumes of ice-cold 0.9% (w/v) sodium chloride. Centrifugation was carried out as described above. This washing step was repeated twice. The packed erythrocytes were lysed by the addition of three volumes of ice-cold distilled water. The pH was adjusted to 6.4 using 5 M HCl. The suspension was centrifuged at 10,000 ϫ g for 1 h at 4°C. The pH was finally adjusted to 7.2 with 5 M KOH. The lysate was dialyzed against 20 liters of 10 mM sodium phosphate, pH 7.2, overnight and then loaded onto a DEAE-cellulose column (5 ϫ 30 cm) equilibrated in 10 mM sodium phosphate, pH 7.2. The column was washed until the A 280 was below 0.1, and the two forms of biliverdin reductase were then eluted with a 10 -350 mM gradient (2 ϫ 250 ml) of sodium phosphate, pH 7.2. The fractions were assayed for biliverdin-IX␣ reductase and flavin reductase activity as described below, and the peak fractions were subjected to immunoblotting with antisera raised against the recombinant BVR-A and BVR-B enzymes.

Enzyme Assays
Biliverdin Reductase Assays-The biliverdin isomers used in this study were synthesized from mono-and dipyrrole components as described previously (27)(28)(29)(30)(31)(32)(33)(34). Dried preparations were dissolved in Me 2 SO to give a final stock concentration of 1 mM. The final concentration of Me 2 SO in the assay was shown to have no effect on enzyme activity.
Preliminary plate assays were conducted at 30°C in 100 mM potassium phosphate buffer, pH 7.5, containing 50 M NADPH and the respective biliverdin isomers at a concentration of 20 M. Initial rate measurements with biliverdin as the variable substrate were made by monitoring biliverdin consumption at 660 nm in the presence and absence of BSA (1 mg/ml) using the extinction coefficients given in Table I. No co-solvents or detergents were added to the assay mixture (Beer Lambert's law was obeyed over the concentration range used). All assays were conducted in 100 mM potassium phosphate buffer, pH 7.5, and contained NADPH at a saturating concentration of 50 M (the K m NADPH for BVR-A and BVR-B is 1 M with biliverdin-IX␣ and FMN, respectively).
Flavin Reductase Assay-The flavin reductase activity of BVR-B was measured using the method described by Yubisui et al. (40). Initial rate studies were carried out under saturating concentrations of FMN (150 M) and NADPH (50 M) in 100 mM potassium phosphate, pH 7.5. Activity was monitored by following the decrease in absorbance of NADPH at 340 nm. Inhibition of flavin reductase (FR) activity was determined under the same conditions using mesobiliverdin-XIII␣ (1 nM to 1 M), lumichrome (5 to 75 M), and protohemin (1 nM to 10 M).

Treatment of Data
The initial rate data were fitted to equations for simple hyperbolic kinetics and total and partial substrate inhibiton (41). Most data sets showed potent substrate inhibition, where few data points were on the upward limb, and fitting (to either total or partial substrate inhibition) produced negative coefficients or large errors. For this reason estimates of the kinetic parameters (Table II) were obtained by manually generating saturation curves for total substrate inhibition (using, where possible, initial estimates from the curve fitting routines) and visually checking the theoretical line against the data set. For several of the substrates with BVR-B, the substrate inhibition is so potent that the only kinetic parameter obtainable was the substrate inhibitory K i value, obtained by plotting the reciprocal of the initial rate against the concentration of the tetrapyrrole. For those substrates exhibiting partial substrate inhibition only the linear part of the v o Ϫ1 versus [biliverdin] curve were used to obtain the substrate inhibitory K i value. All experiments conducted in this study used 50 M NADPH. More extensive kinetic studies, for example varying NADPH concentration, require the development of an assay that will allow initial rates to be determined at low concentrations of the tetrapyrrole substrate to allow curve fitting routines to be used with confidence.

FIG. 2. Separation of native human biliverdin-IX␣ reductase and biliverdin-IX␤ reductase on DEAE-cellulose. a, human
BVR-A and BVR-B isolated from erythroyctes were separated using ion exchange chromatography. Fractions were assayed for flavin reductase activity (q), biliverdin-IX␣ reductase activity (q), and conductivity (f). b, Western blot analysis was carried out on fractions over the peak of BVR-A activity (lanes 4-6) and over the peak of flavin reductase activity (lanes 7-9) using antisera raised against BVR-A and BVR-B, respectively. Lane 1, molecular mass markers; lane 2, BVR-A standard (5 g); lane 3, BVR-B standard (5 g).

RESULTS
The cloning of human biliverdin-IX␤ reductase into the pGEX-KG expression vector to produce a GST-BVR-B fusion protein has been reported previously (37). We have constructed a similar vector for human biliverdin-IX␣ reductase that allows the production of 40 mg of GST-BVR-A fusion protein/liter of culture. The recombinant protein has been affinity purified on glutathione-Sepharose, and the GST moiety was removed following cleavage with thrombin, yielding a homogenous preparation of BVR-A of the predicted mobility on SDS-PAGE, as shown in Fig. 1. The partial purification of native BVR-A and BVR-B from human erythrocytes was achieved using DEAE-  Fig.  2a and reveals complete separation of the two activities that was subsequently confirmed by immunoblotting (Fig. 2b).
Both the native and recombinant form of human BVR-A migrate on SDS-PAGE with an apparent molecular mass of 40 kDa. Mass spectroscopic analysis of recombinant human BVR-A gave a molecular mass of 34,330 Da (data not shown), suggesting that the protein runs anomalously on SDS-PAGE. The molecular mass estimated for BVR-B by SDS-PAGE was confirmed using mass spectroscopy as 22,531 Da.
The structures of the verdins used to assess the substrate specificity of BVR-A and BVR-B are shown in Fig. 3. The In these two groups, those isomers that contained bridging propionates (1, 2, and 4), modified bridging propionates (9 and 10), and extended bridging carboxylate side chains (butyrate, 11; hexanoate, 12) were not reduced by BVR-B (either native or recombinant). In our discussion, "bridging propionates" refers to propionate side chains at positions C 8 and C 12 , effectively bridging the central methene bridge. The verdins substituted at C 10 (13)(14)(15)(16) showed no change in the visible spectrum, which is not surprising because they are not reducible at C 10 . Unfortunately, none of these compounds are particularly effective inhibitors of BVR-A or BVR-B (which might have been a starting point for anti-hyperbilirubinaemia therapy). The methyl derivative (16) was the most potent, exhibiting modest inhibition at 25 M.
Although the overnight plate incubations allow a crude definition of whether or not the various compounds behave as substrates for the two enzyme forms, it yields little information about the relative rates of reaction. The initial rate kinetics for compounds 1-6, 8, and 9 in the presence and absence of BSA (1 mg/ml) with recombinant BVR-A are shown in Fig. 5. It is clear that the addition of BSA has a pronounced effect on the activity with mesobiliverdin-IV␣ (3); however, for biliverdin-IX␣ (1) and 12-ethyl mesobiliverdin-XIII␣ (4) the effect is mainly on sequestration of substrate. In previous work with ox kidney BVR-A (42) and rat kidney BVR-A (43), we have attempted to define the effect of BSA as simple sequestration of the verdin substrate; however, detailed work in our laboratory suggests that this is not the only function (41). 3 The extinction coefficient for bilirubin-IX␣ at 460 nm is increased on binding to albumin, and the free concentration of biliverdin IX␣ is also reduced by binding to BSA; however, it is clear that other factor(s) are also operative (44). By monitoring the ⌬A 660, as opposed to ⌬A 460, some of these complications are overcome, albeit at the expense of sensitivity.
As the kinetics of BVR-A involve pronounced substrate inhi-

FIG. 5. Initial rate studies on BVR-A with various isomers of biliverdin.
The substrate specificity of recombinant human BVR-A was examined in the presence (a) and the absence (b) of BSA. Consumption of verdins 2 (q), 3 (E), 4 (f), 5 (Ⅺ), 8 (OE), 9 (‚), 6 (), and 1 (ƒ) was monitored at 660 nm, and initial rates were calculated using the extinction coefficients given in Table I. bition, because the effect of albumin on the initial rate is not clear and, as the present work shows, these effects vary depending on the substrates used, for the present discussion, we define "good" and "poor" substrates for BVR-A in the following way: good substrates are those that, when assayed in the presence of BSA, exhibit an apparent K m biliverdin of less than 10 M and that exhibit a maximal initial rate of greater than 5 mol/min/mg. We define a poor BVR-A substrate as one with an apparent K m biliverdin greater than 20 M and exhibiting a maximal initial rate no greater that 1 mol/min/mg. It is clear that the compounds with a bridging propionate are all good substrates (compounds 1, 2, and 4) with two propionates being preferred (i.e. biliverdin-IX␣ (1) and mesobiliverdin-XIII␣ (2) exhibit lower apparent K m values than 12-ethyl-mesobiliverdin-XIII␣ (4)). Although all of the other verdins tested were substrates for human BVR-A, the rates were very low compared with those for 1, 2, 4, and 11 (see below).
A quite distinct pattern of substrate specificity is shown by BVR-B. This enzyme cannot tolerate even one bridging propionate side chain in a verdin, so that compounds 1, 2, and 4 do not function as substrates (Fig. 6). The plate shown in Fig. 4 has been allowed to go to completion (i.e. it was left overnight at room temperature) for compounds 1-16. Using the spectrophotometric assay, it was not possible to measure BVR-B activity with compounds 1, 2, and 4. In contrast, several of the compounds that are poor substrates for BVR-A are good substrates for BVR-B (Fig. 6). Propionates in any of the nonbridging positions seem to promote activity, with substitutions at positions 3, 5, 6, 7, and 8 being particularly active. It should be noted that the substrate inhibition with BVR-B is far more potent than that observed with BVR-A, having substrate inhibitory K i values in the submicromolar range. This precluded the determination of any standard kinetic parameters for BVR-B using the spectrophotometric assay.
In a separate experiment, we analyzed the effect of the chain length of the carboxylic acid on activity with both BVR-A and BVR-B by using mesobiliverdin-XIII␣ where the propionate side chains at positions 8 and 12 have been substituted by butyrate (11) or hexanoate (12), respectively. In the case of BVR-B, these behave as bridging substituents, and neither are reduced. However, BVR-A can utilize 11 efficiently, whereas the bulkier hexanoate derivative (12) is an extremely poor substrate (Fig. 7).
We have taken advantage of the ability of BVR-B to function as a FR to show that the nonsubstrate mesobiliverdin-XIII␣, (which has bridging propionates), can function as a potent inhibitor of the FR reaction. The inhibition is competitive with respect to FMN and mixed against NADPH, consistent with the hypothesis that the verdin and flavin may compete for a common site (45). Protohemin, which has been reported previously to bind tightly to bovine liver flavin reductase (46), also acts as a potent inhibitor of the human enzyme giving an IC 50 of 0.8 M under standard assay conditions. This is also consistent with the hypothesis of a common tetrapyrrole/flavin site. Unfortunately, we have been unable to obtain any evidence to show that lumichrome (an inhibitor of the FR reaction; Ref. 45) behaves as an inhibitor of BVR-B activity. Lumichrome is not a particularly potent inhibitor of the FR reaction (K i 73 M), and it may be that the marked substrate inhibition seen with BVR-B masks any relatively modest inhibition by lumichrome by shifting the saturation curve to the right. Such an effect may extend the plateau of maximal activity by alleviating substrate inhibition. Intriguingly, protohemin behaves as an activator when BVR-B activity is measured, and this too may reflect an alleviation of substrate inhibition.

DISCUSSION
Both BVR-A and BVR-B exhibit a fairly broad specificity in terms of the tetrapyrrole substrate, with human BVR-A able to reduce all of the structures tested. Early work on partially purified preparations of guinea pig BVR-A (that may have been contaminated with BVR-B) also suggested that, although the IX␣ isomer was preferred, the ␤, ␥, and ␦ isomers were also substrates for the enzyme (26), which suggests that significant binding energy may be associated with an interaction between the carboxylate side chains and a residue(s) on the enzyme (presumably lysine or arginine). The observation that those compounds with two bridging propionates (1 and 2) have lower apparent K m biliverdin values than the monopropionate verdin substrates (12-ethyl-mesobiliverdin XIII␣; 4) is consistent with the hypothesis that BVR-A may utilize two basic residues to stabilize tetrapyrrole binding (Fig. 8). BVR-B is most distinct in that the bridging propionate rule for BVR-A is the antithesis in this case. This leads us to suggest that, in contrast to BVR-A, there may be a pair of negatively charged residues in BVR-B that do not permit the IX␣ isomer to bind productively. If we argue that the reduced pyridine nucleotide (H) binds in a similar position (Fig. 9) for both BVR-A and BVR-B, then we suggest that a ring of positively charged residues may surround the tetrapyrrole pocket of BVR-B to facilitate the "non-alpha" isomers in binding productively as substrates. The IX-␣ isomers (those with bridging propionates) may bind in a nonproductive mode by rotating through 90°, as illustrated (Fig. 10). This would be in agreement with the competitive kinetics observed for mesobiliverdin-XIII␣ against FMN. It should be noted that open chain tetrapyrroles are forced to adopt a slightly helical structure (47) so that the simple rotation of 90°m ust be accompanied by some degree of "flexibility." Further work is clearly required to test this hypothesis. We have crys- FIG. 8. Proposed BVR-A binding site for biliverdin IX␣. It is proposed that the propionate side chains bridging the C 10 position on the tetrapyrrole interact with the two positively charged residues on BVR-A to promote binding and subsequent catalysis. NADPH is included in the figure to indicate the methene bridge, which is reduced by BVR-A.
FIG. 9. Proposed BVR-B binding site for biliverdin. It is proposed that two negatively charged residues within the active site of the enzyme prevent the binding of biliverdin isomers with propionate side chains bridging the central C 10 methene bridge. Binding of the nonalpha isomers to BVR-B can be rationalized, if in addition to the two negatively charged residues there are a series of positively charged residues around the tetrapyrrole binding site of the enzyme that interact electrostatically with the propionate side chains of mesobiliverdin-IV␣ (green), 8,12-dimethyl-mesobiliverdin-XIII␥ and mesobiliverdin-XII␥ (magenta), and mesobiliverdin-XIII␥ (black). The asterisk indicates the position of the single carboxyl group in the 12-ethyl-13-methyl-mesobiliverdin-IV␣.

FIG. 10. Mesobiliverdin-XIII␣ inhibition of BVR-B.
Mesobiliverdin-XIII␣ is not a substrate for BVR-B; however, it is a potent inhibitor of the flavin reductase activity of the enzyme. This figure illustrates how mesobiliverdin-XIII␣, with its bridging propionate side chains, may orientate itself in the BVR-B tetrapyrrole binding site to overcome the repulsive forces of the two proposed negatively charged amino acids. Its C 10 carbon is no longer oriented in the correct position for reduction; therefore, the isomer cannot act as a substrate for BVR-B. tals of human BVR-B that diffract to 1.6 Å, 4 and more recently, have obtained crystals of BVR-A 5 that should allow the accuracy of these models to be tested. The number of compounds known to interact with BVR-B/FR now includes a wide range of biliverdin isomers in addition to pyrroloquinoline quinone (48), various hemes, fatty acids, and porphyrins (46).