Key Role of Conserved Histidines in Recombinant Mouse β-Carotene 15,15′-Monooxygenase-1 Activity*

Alignment of sequences of vertebrate β-carotene 15,15′-monooxygenase-1 (BCMO1) and related oxygenases revealed four perfectly conserved histidines and five acidic residues (His172, His237, His308, His514, Asp52, Glu140, Glu314, Glu405, and Glu457 in mouse BCMO1). Because BCMO1 activity is iron-dependent, we propose that these residues participate in iron coordination and therefore are essential for catalytic activity. To test this hypothesis, we produced mutant forms of mouse BCMO1 by replacing the conserved histidines and acidic residues as well as four histidines and one glutamate non-conserved in the overall family with alanines by site-directed mutagenesis. Our in vitro and in vivo data showed that mutation of any of the four conserved histidines and Glu405 caused total loss of activity. However, mutations of non-conserved histidines or any of the other conserved acidic residues produced impaired although enzymatically active proteins, with a decrease in activity mostly due to changes in Vmax. The iron bound to protein was determined by inductively coupled plasma atomic emission spectrometry. Bound iron was much lower in preparations of inactive mutants than in the wild-type protein. Therefore, the conserved histidines and Glu405 are absolutely required for the catalytic mechanism of BCMO1. Because the mutant proteins are impaired in iron binding, these residues are concluded to coordinate iron required for catalytic activity. These data are discussed in the context of the predicted structure for the related eubacterial apocarotenal oxygenase.

␤-Carotene 15,15Ј-monooxygenase-1 (BCMO1) 1 is the initial enzyme step in the biosynthesis of vitamin A in animals, symmetrically cleaving ␤-carotene to produce two molecules of alltrans-retinal (1). In the last several years, a number of BCMO1 proteins have been cloned and biochemically characterized (2)(3)(4)(5)(6)(7). BCMO1 belongs to a family of oxygenases of diverse activities, including lignostilbene dioxygenase in bacteria (8 -10); an epoxycarotenoid-cleaving enzyme required for abscisic acid biosynthesis in plants (11); and mammalian RPE65, a retinal pigment epithelial protein required for the production of the visual chromophore 11-cis-retinal in the visual cycle (12). The most recent addition to the characterized carotenoid oxygenases is cyanobacterial apocarotenoid 15,15Ј-oxygenase PCC 6803 (ACO) (13). Despite their unique functionalities and the obvious importance of these proteins, structural studies on this family have been hindered by technical difficulties, and only recently has the crystal structure of eubacterial ACO been published (14).
Originally, Drosophila BCMO1 was identified by homology to a previously characterized plant 9-cis-epoxycarotenoid dioxygenase (15) and by its ability to cleave ␤-carotene (4). Subsequently, mouse BCMO1 was identified by our laboratory based on its homology to mammalian RPE65 (3). Because the biochemical function and mechanism of BCMO1 have been studied in detail (3,5,7), it is functionally the best characterized mammalian member of the family and could serve as a model for studying putative vertebrate oxygenases of more elusive function such as RPE65 and BCMO2 (16).
Although there is a weak overall identity (Ͻ10%) among these proteins, there are four histidines and five acidic residues that are absolutely conserved and a well conserved acidic stretch (which we call the "signature" sequence) that is common to the superfamily (see Fig. 1). In the original characterization of its activity, iron was found to be an essential cofactor for intestinal BCMO1 (17,18) as well as for recombinant proteins (4). Iron is also required for the activity of other members of the family such as lignostilbene dioxygenase (8). Therefore, we propose that four conserved histidines and five acidic residues would be the best candidates as putative metal-binding residues in BCMO1.
In this study, we demonstrate the necessity of the conserved histidines and acidic residues in the catalytic mechanism of BCMO1, supporting our hypothesis that these residues are necessary for iron coordination in BCMO1. Our data agree well with the crystal structure prediction for an iron(II) center coordinated by four conserved histidines and fixed by three glutamates in the related ACO (14) and provide empirical evidence for the validity of this prediction.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis-A panel of mutant BCMO1 proteins was made using a previously described pBAD/BCMO1 construct (3) (3), with small modifications. Briefly, overnight cultures of E. coli cells transformed with constructs were grown in LB broth supplemented with 100 g/ml ampicillin. These cultures were used to inoculate 500-ml cultures. At mid-log phase (A 600 ϭ 0.6 -0.8), protein expression was induced with 0.02% (w/v) L-arabinose. Cultures were allowed to grow for an additional 4.5 h. Cells were harvested by centrifugation at 5000 ϫ g for 30 min at 4°C. The pellet was resuspended in B-PER detergent (Pierce; one-twenty-fifth of the initial culture volume) with one EDTA-free Complete protease inhibitor tablet (Roche Applied Science)/25 ml. After a 10-min incubation on ice, soluble proteins were isolated by centrifugation at 20,000 ϫ g for 15 min at 4°C. Recombinant His-tagged protein was purified using Talon CellThru resin (Clontech). The B-PER extract was applied to the resin and incubated with gentle agitation for 1 h at 4°C. The resin was extensively washed with 50 mM sodium phosphate buffer (pH 7.0) containing 300 mM NaCl and 10 mM MgCl 2 and then transferred to a column from which bound protein was eluted with 300 mM imidazole supplemented with EDTA-free Complete protease inhibitor. The purified enzyme was stored in 40% (v/v) glycerol at Ϫ20°C until used. Prior to use, protein was concentrated on Amicon Ultra-15 filters with a molecular weight cutoff of 30,000.
In Vivo Assay of the Enzymatic Activity of Mouse BCMO1 and Mutants-pBAD/BCMO1, pBAD/mutant BCMO1, and pBAD/LacZ expression constructs were transformed into competent cells prepared as described (19) from a strain of E. coli cells transformed with the pAC-BETA vector, which produces and accumulates ␤-carotene (20). Overnight cultures of pAC-BETA-transformed E. coli cells with constructs were grown in LB broth supplemented with 150 g/ml ampicillin and 30 g/ml chloramphenicol at 30°C. One ml of the overnight culture was used to inoculate 50 ml of LB broth supplemented with the same antibiotics. The culture was allowed to grow at 37°C to mid-log phase (A 600 ϳ 0.6) and was then split; one-half was induced with 0.02% (w/v) L-arabinose, whereas the other was not induced. Both cultures were grown for another 3 h at 30°C; the absorbance was measured; and cells were collected by centrifugation at 5000 ϫ g for 20 min. Pellets were frozen in dry ice and stored at Ϫ70°C. To quantify the activity of the various constructs, ␤-carotene and all-trans-retinal were extracted from frozen bacterial cultures. The pellet was resuspended in lysis buffer (0.9 ml/20 ml of initial culture with A 600 ϳ 1.8) containing 0.1 M Tricine, 0.1 M NaCl, 4% formaldehyde, 0.5% Tween, 1% pyrogallol, and 1.18 M ␣-tocopherol acetate as an internal standard. Resuspended cells were divided into three equal aliquots, and each was centrifuged at 16,060 ϫ g for 1 min. Supernatants were collected, and carotenoids were extracted from each pellet with three sequential aliquots of 300 l of acetonitrile with rigorous vortexing. Supernatants and acetonitrile extracts were combined and filtered, and 100 l was injected onto a reverse phase HPLC column (see below). The amount of ␤-carotene was quantified as the area of ␤-carotene peak normalized to the area of the ␣-tocopherol acetate internal standard. Samples were assayed in triplicate from induced and uninduced cultures, and the activity is expressed as the ratio (%) of ␤-carotene extracted from induced cells to ␤-carotene extracted from uninduced cells, normalized to cell density.
In Vitro Enzymatic Activity Assay-The in vitro assays were performed following the method of Lindqvist and Andersson (5). Typically, reactions were run in glass vials under subdued light in volume of 200 l in 100 mM Tricine-KOH (pH 8.0) supplemented with 10 M FeSO 4 , 5 mM tris(2-carboxyethyl)phosphine hydroxychloride, 1% (w/v) octyl ␤-D-1-thioglucopyranoside, and 10 -20% (v/v) glycerol. The BCMO1 enzyme concentration in preparations varied between 100 and 1600 ng/reaction as determined by enzyme-linked immunosorbent assay (ELISA). The concentration of ␤-carotene in the reaction mixtures ranged from 0.1 to 20 M (from a stock solution of 34 pmol/l). The substrate was solubilized in hexane with 50 l of 4% octyl ␤-D-1-thioglucopyranoside in a glass vial, and the solvent was evaporated under argon. The assay buffer with substrate/detergent was preincubated at 37°C for 5 min; the enzyme in glycerol was added; and the reaction was incubated at 37°C in a water bath for 45 min to 2 h. To analyze reaction products, we used a modified protocol of During et al. (21). Fifty l of 37% (v/v) formaldehyde was added to stop the reaction, and the incubation was continued for 10 min at 37°C. Then, 500 l of acetonitrile was added, and the solution was vortexed and put on ice at 4°C for 5 min. The upper acetonitrile phase was collected, and 100 l was injected and analyzed by reverse phase HPLC.
HPLC Analysis of ␤-Carotene and All-trans-retinal-␤-Carotene, alltrans-retinal, and ␣-tocopherol acetate (internal standard for in vivo assays) were separated on a 4.6 ϫ 250-mm Supelco Supelcosil LC318 C 18 5-m column with a flow rate 1 ml/min and simultaneous UV detection at 451, 383, and 292 nm (Agilent 1100 HPLC series). The initial conditions consisted of acetonitrile and 0.015 M ammonium acetate (60:40) and were held for 5 min (0 min for in vivo method), followed by a linear gradient to 50:50 acetonitrile/isopropyl alcohol over 10 min (5 min for the in vivo method), which was held for an additional 10 min. ␤-Carotene and all-trans-retinal were quantified from their peak area using standard curves obtained with 2-220 and 0.5-20 ng of material, respectively, in 60:40 acetonitrile/lysis buffer.
ELISA for Measurement of Immunoreactive BCMO1-Direct ELISA was used to determine BCMO1 concentrations in the partially purified protein preparations obtained from bacteria by one-step His tag affinity purification. Additional purification steps led to a total loss of enzymatic activity. As the standard, the BCMO1 construct was amplified using a His tag linear template generation set (Roche Applied Science). The resultant PCR product with regulatory elements and a C-terminal His tag was subcloned into the pCR4TOPO vector and used as a template for the in vitro transcription and translation system (Rapid Translation System (RTS) 500 HY kit, Roche Applied Science). Expressed protein was recovered from inclusion bodies by solubilization in 6 M guanidine HCl, refolding by the rapid dilution method in 0.5 M arginine, and subsequent dialysis and concentration (22). Purified protein was stored in 40% (v/v) glycerol at Ϫ20°C, and the same batch of purified protein was used as a standard for all ELISAs. 96-Well OptiPlate-96FHB (PerkinElmer Life Sciences) plates were coated with 2-100 ng of standard and 0.5-4-l samples/well in 100 l of imidazole-buffered saline (IBS) containing 1 mM tris(2-carboxyethyl)phosphine hydroxychloride and one Complete protease inhibitor tablet/50 ml. Plates were incubated overnight at 4°C; rinsed with IBS wash buffer (KPL, Inc., Gaithersburg, MD); and blocked with 300 l of 10% BlokHen II (Aves Labs Inc., Tigard, OR), 10% fetal bovine serum, and 10% SuperBlock solution (Pierce) in IBS for 1 h at room temperature. After a brief rinse with IBS wash buffer, the plates were incubated with chicken anti-BCMO1 antibody (raised against the mouse BCMO1 peptide sequence NYIRKIDPQTLETLEK (Aves Labs, Inc.)) in IBS with 20% fetal bovine serum and SuperBlock for 1 h at room temperature. After washing with IBS, the plates were incubated for 1 h at room temperature in alkaline phosphatase-conjugated goat anti-chicken IgY (1:1200 in IBS with 20% fetal bovine serum and SuperBlock). The plate was then thoroughly washed (3 ϫ 300 l/well) and developed for 2 h with 5 M 4-methylumbelliferyl phosphate (100 l/well) in 100 mM diethylamine buffer (pH 9.6) containing 1 mM MgCl 2 and 3.2 mM Na 3 N. Fluorescence was read with 355/460-nm filters. All samples were run in duplicate.
Iron Determinations-Proteins were purified as described above for in vitro assays. Samples (0.5 ml) were diluted 16-fold with buffer containing 20 mM imidazole, 20 mM Tris-HCl (pH 7.9), 500 mM NaCl, and 100 nM FeSO 4 and then reconstituted to the original volume using Amicon Ultra filters with a molecular weight cut-off of 30,000. Analytical reagent blanks were prepared by running buffers through the whole procedure using the same filters, glassware, plasticware, and pipette tips used to process the protein samples. Each sample and the corresponding blank of the same weight (ranging from 0.5 to 1.3 g) were placed in acid-cleaned quartz tubes and dry-ashed (23). Samples were diluted to a final volume of 3 ml using 18 megohm-cm purity water and analyzed by inductively coupled plasma atomic emission spectrometry (ICP-AES). A Leeman PS3000 spectrometer was used (combination simultaneous and sequential high resolution echelle-based ICP-AES system). Detailed instrumental operating conditions as well as calibration information are summarized elsewhere (24). NIST-SRM 1640 Trace Elements in Natural Water was run as standard along with all samples to ensure analytical accuracy; iron concentrations determined were in good agreement with the certified value of 34.4 Ϯ 1.6 ng/g.
Miscellaneous-Total protein concentration was measured routinely using the Advanced protein assay reagent (Cytoskeleton, Denver, CO). The NanoOrange protein quantification kit (Molecular Probes, Inc., Eugene, OR) was used for measurement of diluted protein solutions (Ͻ1 g/ml).

Identification of Perfectly Conserved Amino Acid Residues-
Alignment of all known family members from taxonomically diverse organisms revealed ϳ40 residues to be absolutely conserved not only in BCMO1 and RPE65, but also in ACO and BCMO2 proteins (Fig. 1). Included among these are four histidines and five acidic residues (not including two in the signature sequence) that we propose to be putative ironbinding residues in all family members. To determine the role of these four histidines and five acidic residues in the cata-lytic mechanism of BCMO1, we constructed a set of alanine substitution mutants (Fig. 2). We also replaced each of four non-conserved histidines and one non-conserved acidic residue (Glu 450 ) with alanines. These residues are present only in the mammalian members of the family (BCMO1, BCMO2, and RPE65). If our hypothesis is correct, these latter residues FIG. 1. Alignment of carotenoid oxygenase family members. Histidine and acidic residues conserved throughout the superfamily are in boldface; the signature sequence is underlined. Asterisks, identity; double dots, strong similarity; single dots, weak similarity. The chloroplast recognition signal was removed from the amino acid sequence of VP14 dioxygenase, and the last 24 amino acids of mouse BCMO1 were removed. GenBank TM /EBI accession numbers are as follows: mouse BCMO1, AF271298; mouse BCMO2, AJ290392; mouse RPE65, NM_029987; Zea mays VP14, ZMU95953; Synechocystis lignostilbene dioxygenase (ACO), D90914.
should not be crucial for activity or participate in iron binding. We also replaced the first conserved residue in the very acidic signature sequence (Glu 469 ) with alanine. Because we propose that the signature sequence might take part in electron transfer and that overall negative charge is more important than individual residues, this replacement should reduce but not eliminate the catalytic activity.
Production of Mutant Forms of BCMO1 and in Vivo Enzymatic Activity-We made use of ␤-carotene-accumulating pAC-BETA-transformed E. coli cells (20) as an in vivo system for measurement of enzymatic activity. In these experiments, pBAD mutant constructs were transformed into competent ␤-carotene-producing E. coli cells. Cultures were split at midlog phase; one portion was induced, and the other was not induced. After several hours, induced cultures expressing the wild-type protein and some of the mutants were completely bleached, whereas cultures expressing other mutants did not show bleaching (data not shown). To quantify the amount of all-trans-retinal formed or ␤-carotene consumed by the enzyme, all-trans-retinal and ␤-carotene extracted from induced cells were analyzed by reverse phase HPLC. Cells were grown in the dark at 30°C for 3 h after induction, which was the bleach end point for the wild-type protein as judged visually. However, the amount of all-trans-retinal extracted from cells did not linearly correlate with the amount of ␤-carotene cleaved or the time after induction. This was probably due to further metabolism of all-trans-retinal in the bacterial cells. Therefore, all-trans-retinal production could not be used as a measure of activity in this system. On the other hand, the loss of ␤-carotene could be used to determine activity by comparing the ␤-carotene produced by the uninduced portion of the same culture used as a control with that in the induced portion of the culture. The catalytic activity is expressed as the ratio (%) of ␤-carotene extracted from the induced culture to ␤-carotene extracted from the uninduced culture, normalized to cell density. It should be emphasized that the presence of all-trans-retinal in cell extracts is an important enzymatic characteristic of the functionally active expressed protein and is an important qualitative correlate of activity.
Results for all 15 mutant proteins expressed in the in vivo system are presented in Fig. 3. pBAD/BCMO1 (wild-type) was used as the positive control and pBAD/LacZ as the negative control in our scale of enzymatic activity. We found that all four mutants with replaced conserved histidines (H172A, H237A, H308A, and H514A) neither lost ␤-carotene nor accumulated all-trans-retinal and thus failed to convert ␤-carotene to all-trans-retinal, as detected in cell extracts. Mutation of Glu 405 to Ala also eliminated catalytic activity (ϳ100%, no loss of ␤-carotene). Replacement of the conserved Glu 457 led to a substantial reduction in activity (ϳ80% of ␤-carotene remaining). On the other hand, mutation of any of the four non-conserved histidines (H49A, H58A, H174A, and H309A) as well as the conserved Glu 314 and non-conserved Glu 450 did not lead to a dramatic decrease in activity in comparison with wild-type BCMO1. The amount of ␤-carotene extracted from induced cells compared with uninduced cells was 17.6 Ϯ 1.2% for wild-type BCMO1 and up to 42.6 Ϯ 2.2% for the H309A mutant (Fig. 3). Although these mutations influenced the catalytic activity of the enzyme, they clearly did not abolish it. It is likely that these amino acids are not necessary for metal ion coordination or can be compensated for. The change in the signature sequence Glu 469 to Ala led to a significant decrease in activity (52.3 Ϯ 4.9%) in 3 h, and cells expressing the E469A construct were bleached if left growing overnight. In cells expressing either the D52A or E140A mutant, all-trans-retinal was still produced, but the amount of ␤-carotene degraded was Ͻ50%. To investigate further the importance of the last two acidic residues, we constructed two double mutants (D52A/E140A and E140A/ H49A) and tested them for activity. The loss of ␤-carotene was insufficient to make a conclusive comparison with single mutants; however, we found that the E140A/H49A mutant was still able to produce all-trans-retinal, whereas the D52A/ E140A mutant failed to do so (Fig. 4). Therefore, these two acidic residues in combination are required for catalytic activity.
To confirm that the different mutants produce comparable amounts of protein in vivo, immunoblots of extracts from 13 cell lines transformed with mutant BCMO1 constructs were analyzed. All cells except for those transformed with the empty pBAD/LacZ construct expressed comparable amounts of pro- teins of appropriate size (Fig. 5). Similar results were obtained with the additional mutant cell lines (data not shown). Therefore, the change in the ␤-carotene extracted reflects a change in the catalytic activity of the mutant enzymes.
Kinetic Parameters of Altered Enzymes-To compare the biochemical characteristics of the mutant proteins, we developed an in vitro assay based on the protocol of Lindqvist and Andersson (5). In recombinant mutant proteins purified by metal affinity (Co 2ϩ ) chromatography, we could maintain activity for several weeks with storage at Ϫ20°C in 40% glycerol with protease inhibitors. As this procedure resulted in an only partially purified preparation (10 -15%), the specific concentration of BCMO1 in the preparation was determined by ELISA using highly purified Rapid Translation System-expressed recombinant BCMO1 as a standard. Recombinant wild-type protein had similar characteristics as reported earlier (3), with K m ϭ 1.2 M and V max ϭ 104.8 pmol of all-trans-retinal/g of BCMO1 protein/h. K m and V max values were calculated from the averaged Hanes plot (s/v versus s) (Fig. 6). Comparison with published data demonstrated that all previously described singlestep affinity-purified preparations of BCMO1 have comparable K m values (1-7 M), whereas V max values based on total protein concentration from the studies of Redmond et al. (3) and Paik et al. (7) were considerably lower compared with our data ( Table  I) and those of Lindqvist and Andersson (5). Because more extensive purification procedures lead to a total loss of activity, using insect cells to express protein (5) and quantifying the BCMO1 protein by a specific ELISA method are the only practical ways to characterize this enzyme so far. Formation of product was linear up to 60 min (data not shown) and for 25-1500 ng of enzyme protein (data not shown). The results from kinetic analyses of the six active mutant proteins are summarized in Table I. The biochemical characteristics of enzymatically active mutants revealed that the decrease in activity compared with the wild-type enzyme was primarily due to changes in V max , the largest difference being observed for the D52A and E140A mutants. The greatest increase in K m (4.7 times) was also observed for the D52A mutant, although the significance of this finding is still to be shown. We were not able to obtain kinetic curves for the E457A and E469A mutants. As expected from the in vivo experiments, no activity was detected for mutants with replaced conserved histidines or for the E405A mutant. The H49A mutant was active but unstable upon purification, so we were unable to determine reliable biochemical parameters for this enzyme. The H174A and H309A mutants demonstrated low activity in the in vivo system, probably due to their close location to conserved histidines (His 172 and His 308 ), and were not assayed in vitro. These changes most probably affected protein folding but not iron binding directly.
Metal Analyses of BCMO1 and Its Inactive Mutant Versions-To address directly whether enzymatically inactive mutants lose the ability to bind iron, preparations of these proteins were probed for the presence of iron by ICP-AES. Purified BCMO1 without added iron showed much less activity (25%) than when iron was added into the assay mixture. Therefore, purified proteins were recharged with 100 nM FeSO 4 in 20 mM imidazole buffer and tested for activity prior to ICP-AES anal-  6. Hanes plot for wild-type BCMO1 kinetics. The points on the plot represent the average of three independent substrate curves obtained with different enzyme preparations. Variation among triplicates within each experiment averaged 15%. a Initial activity was comparable to that of wild-type BCMO1; however, the enzyme was too unstable, and the activity was lost within several hours of preparation.
ysis. This iron concentration was found to fully restore the activity of the enzyme (data not shown). Samples were measured against filtrates containing the same amount of imidazole, and therefore, the iron presence in the samples could be explained only by protein binding. The polyhistidine tag was present in all mutants as well as in wild-type BCMO1, and therefore, its inherent chelating effect does not explain the dramatic changes in iron binding between the mutants and wild-type protein. Also, we found that identically prepared untransformed E. coli TOP10 samples did not bind iron. To minimize experimental variation in absolute iron determinations, we present values for iron binding as relative to wild-type values (% of wild-type levels). For the H514A and H237A mutants, the amount of iron bound was not significantly different from that in the measured blanks and water. The E405A and H308A mutants also failed to bind any biologically significant amount of iron (Ͻ10% of the wild-type mol of iron/mol of enzyme), and the iron bound to the H172A mutant was substantially less (45.2 Ϯ 11.9%) than that bound to wild-type BCMO1 (Table II). Therefore, it appears that that the inactive mutants fail to bind as much iron as wild-type BCMO1. DISCUSSION Iron has long been known to be an essential cofactor in the enzymatic cleavage of ␤-carotene. Goodman and Huang (17) and Olson and Hayaishi (18) demonstrated in 1965 that chelators of ferrous iron inhibit the ␤-carotene cleavage activity of the enzyme in intestinal extracts. The results presented herein identify the residues of BCMO1 responsible for coordination of the iron cofactor recognized by these workers and underscore the importance of iron in the catalytic mechanism of this enzyme. Given the absolute conservation of these residues among its paralogs, this characteristic of BCMO1 is likely to be paradigmatic for all members of this family.
Heme or iron-sulfur clusters were not found in BCMO1 (17,18,25), and thus, it was proposed that iron binds directly to functional groups in the enzyme. A variety of oxygen-activating non-heme iron-containing enzymes have been described recently (26). Two basic classes include mono-and dinuclear iron centers, and both heavily implicate histidines and acidic functionalities (27,28). A common motif in oxygen-activating mononuclear iron centers consists of a two-histidine/one-carboxylate facial triad found in a variety of unrelated enzymes (29). A second class of enzymes employs binuclear non-heme iron clusters (26,27). So far, most of the oxygen-activating non-heme diiron enzymes contain a pair of (D/E)XXH motifs in their amino sequences, but the integral membrane oxygenase alkane hydroxylase (AlkB), another diiron enzyme, lacks this motif and appears to use eight conserved histidines as ligands (30).
However, the recent solution of the ACO crystal structure, employing four histidines at the axis of a seven-bladed ␤-propeller chain fold, demonstrates yet another type of mononuclear iron center (14). Iron(II) is coordinated in a nearly perfect octahedral arrangement by four histidines. Such coordination is also known for 15-lipooxygenase (31) and certain non-heme centers of photosystem type II and photosynthetic bacterial reaction centers (32)(33)(34). The fifth position in the crystal structure is occupied by a water molecule, and it is proposed that the sixth position is assumed by dioxygen. The positions of three of the four histidines (His 238 , His 304 , and His 484 ) are fixed by glutamates. Presumably, all members of this family, including BCMO1, share this iron center arrangement. In this study, we have presented the first direct empirical evidence supporting this crystal structure prediction.
We have demonstrated an important role for the conserved histidines (His 172 , His 237 , His 308 , and His 514 ) and Glu 405 in the catalytic mechanism of BCMO1. Replacement of any of these residues individually led to total loss of enzymatic activity, and as we expected, replacement of any of the four non-conserved histidines (H49A, H58A, H174A, and H309A) did not eliminate catalytic activity, although mutation of the non-conserved His 174 and His 309 led to a substantial decrease in activity in the in vivo assay. A reasonable explanation is that they are located adjacent to the conserved His 172 and His 308 and thus may indirectly influence the functioning of those residues. On the other hand, only one of the predicted well conserved acidic residues (Glu 405 ) was crucial for catalytic activity, whereas replacement of Glu 314 had no significant effect on activity, and the Glu 457 , Asp 52 , and Glu 140 mutations had an intermediate effect in the in vivo assay. As we expected, replacement of the non-conserved Glu 450 did not lead to any change in activity, whereas mutation of the signature sequence conserved glutamic acid (E469A) had also an  intermediate effect on catalytic activity. The in vitro biochemical characterization of mutants was consistent with the in vivo results. Of those mutants with activity, the difference in catalytic activity compared with the wild-type enzyme was primarily due to a V max change, with the D52A and E140A mutants having the most substantial V max decrease (ϳ100-fold). A significant increase in K m was seen only in D52A (ϳ4.7-fold). The in vivo assay with the D52A/ E140A double mutant showed that the combination of the two was absolutely required for activity. The changes in K m for D52A and V max for D52A and E140A may indicate a role in substrate binding and turnover. On the other hand, it may be that these residues participate in iron binding by being replaceable ligands and can be replaced with a reaction intermediate or another amino acid during catalysis in the wild-type protein and in single mutants (35).
To visualize the spatial positions of BCMO1 replacements, we plotted the BCMO1 paralogs of ACO residues on the ACO three-dimensional ribbon model (Fig. 7). Our data are in good agreement with predictions from the crystal structure. The four conserved histidines required for BCMO1 activity are paralogs of the iron-coordinating histidines in ACO. Of three fixing glutamate paralogs in BCMO1, one (Glu 405 ) is absolutely crucial for enzymatic activity, and two (Glu 457 and Glu 140 ) are important but not essential for the activity. Although Glu 405 and its paralogs are shown to be absolutely conserved in our alignment (Fig. 1), this identity was overlooked in the ACO study (14). We found that mutations resulting in significant loss of activity correspond to residues in the substrate-binding tunnel of ACO (His 484 , His 304 , Glu 370 , His 238 , His 183 , Asp 69 , Glu 150 , and Glu 426 ), respectively. Only one glutamate (Glu 469 in BCMO1) that results in significant loss of activity is not located in the substrate-binding tunnel. This particular glutamate is part of the conserved acidic signature sequence. We speculate that it could be involved in electron transfer. All residues examined except two non-conserved histidines (His 58 and His 174 ) are conserved in ACO. The conserved Asp 70 (paralog of Asp 52 in BCMO1) is located on the upper side of the substrate-binding tunnel of ACO and is not involved in iron coordination. However, it is adjacent to the conserved Phe 69 (Phe 51 of BCMO1), which is proposed to be involved in hydrophobic substrate orientation. Our K m data suggest that Asp 52 may be involved in substrate binding in BCMO1 by itself or in influencing the adjacent phenylalanine involved in binding of the substrate.
It is of interest that the human RPE65 conserved Glu 417 (paralog of BCMO1 Glu 405 ) was found to be mutated in some cases of Leber's congenital amaurosis (36), a congenital disease of early-onset blindness. Based on the phenotype of the Rpe65 knockout mouse (12), these patients probably lack 11-cis-retinoids in their retinas. Thus, these perfectly conserved residues may also be required for the normal activity of RPE65, presumably for iron coordination. This fact strongly supports the idea that all family members share a common chain fold.
The depletion of protein-bound iron in the conserved histidine and Glu 405 mutants supports the hypothesis that these residues are involved in metal coordination in BCMO1. Our finding of partial iron content in the His 172 mutant can be explained by considering that, in the ACO structure, this histidine is not fixed by a glutamate and could be a less important ligand. Without this ligand, the iron might still be bound to the enzyme by the remaining ligands but less tightly. Such im-proper binding could shift the iron from its proper geometry and thus block enzymatic activity. A similar phenomenon of partial iron content in a histidine mutant was observed with human 5-lipooxygenase (37).
In summary, our mutagenesis and iron data together provide the first biochemical insight into the catalytic mechanism of BCMO1. It is likely that this role of the conserved histidines and glutamates in iron coordination is similar for the entire superfamily. Furthermore, our data provide clear experimental validation of the proposed structure of this family of enzymes.