Azide and acetate complexes plus two iron-depleted crystal structures of the di-iron enzyme delta9 stearoyl-acyl carrier protein desaturase. Implications for oxygen activation and catalytic intermediates.

Delta9 stearoyl-acyl carrier protein (ACP) desaturase is a mu-oxo-bridged di-iron enzyme, which belongs to the structural class I of large helix bundle proteins and that catalyzes the NADPH and O2-dependent formation of a cis-double bond in stearoyl-ACP. The crystal structures of complexes with azide and acetate, respectively, as well as the apoand single-iron forms of Delta9 stearoyl-ACP desaturase from Ricinus communis have been determined. In the azide complex, the ligand forms a mu-1,3-bridge between the two iron ions in the active site, replacing a loosely bound water molecule. The structure of the acetate complex is similar, with acetate bridging the di-iron center in the same orientation with respect to the di-iron center. However, in this complex, the iron ligand Glu196 has changed its coordination mode from bidentate to monodentate, the first crystallographic observation of a carboxylate shift in Delta9 stearoyl-ACP desaturase. The two complexes are proposed to mimic a mu-1,2 peroxo intermediate present during catalytic turnover. There are striking structural similarities between the di-iron center in the Delta9 stearoyl-ACP desaturase-azide complex and in the reduced rubrerythrin-azide complex. This suggests that Delta9 stearoyl-ACP desaturase might catalyze the formation of water from exogenous hydrogen peroxide at a low rate. From the similarity in iron center structure, we propose that the mu-oxo-bridge in oxidized desaturase is bound to the di-iron center as in rubrerythrin and not as reported for the R2 subunit of ribonucleotide reductase and the hydroxylase subunit of methane monooxygenase. The crystal structure of the one-iron depleted desaturase species demonstrates that the affinities for the two iron ions comprising the di-iron center are not equivalent, Fe1 being the higher affinity site and Fe2 being the lower affinity site.

⌬9 stearoyl-acyl carrier protein (ACP) desaturase is a -oxo-bridged di-iron enzyme, which belongs to the structural class I of large helix bundle proteins and that catalyzes the NADPH and O 2 -dependent formation of a cis-double bond in stearoyl-ACP. The crystal structures of complexes with azide and acetate, respectively, as well as the apo-and single-iron forms of ⌬9 stearoyl-ACP desaturase from Ricinus communis have been determined. In the azide complex, the ligand forms a -1,3-bridge between the two iron ions in the active site, replacing a loosely bound water molecule. The structure of the acetate complex is similar, with acetate bridging the di-iron center in the same orientation with respect to the di-iron center. However, in this complex, the iron ligand Glu 196 has changed its coordination mode from bidentate to monodentate, the first crystallographic observation of a carboxylate shift in ⌬9 stearoyl-ACP desaturase. The two complexes are proposed to mimic a -1,2 peroxo intermediate present during catalytic turnover. There are striking structural similarities between the di-iron center in the ⌬9 stearoyl-ACP desaturase-azide complex and in the reduced rubrerythrin-azide complex. This suggests that ⌬9 stearoyl-ACP desaturase might catalyze the formation of water from exogenous hydrogen peroxide at a low rate. From the similarity in iron center structure, we propose that the -oxo-bridge in oxidized desaturase is bound to the di-iron center as in rubrerythrin and not as reported for the R2 subunit of ribonucleotide reductase and the hydroxylase subunit of methane monooxygenase. The crystal structure of the one-iron depleted desaturase species demonstrates that the affinities for the two iron ions comprising the di-iron center are not equivalent, Fe1 being the higher affinity site and Fe2 being the lower affinity site.
⌬9 stearoyl acyl carrier protein desaturase (⌬9 desaturase) 1 from Ricinus communis is a soluble enzyme located in the plastid catalyzing the NADPH-and O 2 -dependent insertion of a cis-double bond between the C-9 and C-10 positions of saturated fatty acids (1)(2)(3). The natural substrate, stearic acid, is attached to acyl carrier protein (ACP) via a thioester linked to a pantetheine group (4). Desaturase isozymes act specifically on saturated fatty acids of varying chain length and differ in the insertion position of the double bond (5)(6)(7)(8)(9). The enzymes interact directly with molecular oxygen and reduced ferredoxin during catalysis, with the electrons ultimately coming from NADPH via ferredoxin reductase or directly from photosystem I (10).
⌬9 Desaturase is a homodimer with each mature subunit of 41.6 kDa containing a binuclear iron site (11). It belongs to the structural class I of large helix bundle proteins, which also includes the R2 subunit of ribonucleotide reductase (R2) and the hydroxylating subunit of methane monooxygenase (MMOH) (12). Both oxidized and fully reduced ⌬9 desaturase have been characterized spectroscopically (11,13,14). It was shown that the oxidized iron center contains a -oxo bridge, which is lost upon reduction, in agreement with the crystal structure of photoreduced ⌬9 desaturase, determined to 2.4-Å resolution (15).
The ⌬9 desaturase monomer consists mainly of ␣-helices with the catalytic di-iron center buried within a four-helix bundle. Two pairs of anti-parallel helices provide ligands to the iron ions: ␣ 3 -Glu 105 plus ␣ 4 -Glu 143 and -His 146 and ␣ 6 -Glu 196 plus ␣ 7 -Glu 229 and -His 232 (Fig. 1). The distance between the two 5-coordinated iron ions is 4.1 Å, and they have distorted square pyramidal coordination geometry. The structure of the cluster is highly symmetric. Glu 105 is a bidentate ligand to one iron ion (Fe1), and correspondingly, Glu 196 is a bidentate ligand to the second iron ion (Fe2). Glu 143 and Glu 229 both act as bridging ligands of the iron center. Besides these carboxylate ligands, each iron ion has one nitrogen atom ligand, N␦1 of His 146 and His 232 , respectively. A water molecule is loosely coordinated to the iron center at a distance of 3.0 and 3.3 Å to the Fe1 and Fe2 ions, respectively. A narrow, bent, hydrophobic cavity, expected to bind the saturated fatty acid substrate, extends from the surface down into the protein for ϳ20 Å. Where it passes the di-iron cluster, the shape favors a gauche substrate conformation, which predisposes the formation of a cis-double bond in the product. Based on this crystal structure, the substrate fatty acid chain length specificity was altered in desaturase by making series of mutations guided by the structure (16).
The catalytic mechanism of soluble desaturases remains to be fully elucidated but is expected to have intermediates in common with other di-iron enzymes of class I, especially methane monooxygenases, since these can catalyze desaturase reactions with some substrates (17). To react with molecular oxygen, the resting ferric di-iron center needs to be reduced by ferredoxin in two single-electron transfer steps. ⌬9 desaturase has not been observed in a one-electron reduced form, so the addition of the first electron seems rate-limiting for reduction (11). The binding surface on ⌬9 desaturase for ferredoxin remains to be determined. Reaction of the reduced ferrous center with molecular oxygen gives rise to a peroxo intermediate. This is followed by a highly reactive intermediate, able to abstract hydrogen atoms from the saturated fatty acid bound in the cavity, which has been proposed to be similar to the ferryl "Q" intermediate in MMOH (18). However, since the outcome of the reaction with natural substrates is different for ⌬9 desaturase and MMOH, desaturation versus hydroxylation, this hypothesis remains to be tested. Modeling stearic acid in the hydrophobic cavity of ⌬9 desaturase places the C-9 -C-10 carbon atoms near the di-iron center. Recent data give some indication that the initial hydrogen abstraction takes place at carbon 10 (19,20). However, the detailed mechanism for the desaturation remains to be elucidated.
A stable peroxo intermediate can be obtained by mixing chemically reduced ⌬9 desaturase with stearoyl-ACP under anaerobic conditions and then exposing the sample to one atmosphere of O 2 (21,22). Chemical reduction with dithionite reduces both subunits of the desaturase dimer in a 4e reduction in contrast to the biological 2e reduction, where one subunit of ⌬9 desaturase is reduced at a time. This peroxo-diferric intermediate of desaturase is more stable than similar peroxo intermediates of either MMOH or of mutants of R2 (23)(24)(25)(26). It decays without formation of the product oleoyl-ACP or of hydrogen peroxide but through an oxidase reaction forming water (22).
We have now determined the structures of an azide complex of ⌬9 desaturase to 2.4-Å resolution and an acetate complex to 2.4-Å resolution. We propose that these azide and acetate complexes represent models for the -1,2-peroxide intermediate in desaturase catalysis and a model for inhibition by hydrogen peroxide. In addition, we determined the structure of the ironfree apo form of ⌬9 desaturase to 3.2-Å resolution and with a single iron ion (Fe1) present in the di-iron center to 2.8-Å resolution. These are compared with crystal structures of other iron-depleted di-iron enzymes, and we discuss the implications for iron insertion in ⌬9 desaturase.

MATERIALS AND METHODS
Protein Preparations-The mature castor protein, lacking the 33amino acid transit peptide, was expressed under the control of the T7 promoter in Escherichia coli strain BL21 gold (Novagen, Madison, WI). Cells were grown in a Bioflow 3000 fermenter (New Brunswick Scien-tific, New Brunswick, NJ), in Luria-Bertani broth supplemented with 1% (w/v) glucose to an A 600 of ϳ3 at which time the cells were induced by adding lactose to 0.4% (w/v) and incubating for 4 h at 30°C. Dissolved oxygen was maintained above 15% and pH was maintained at 6.5. Cell densities at harvest were ϳ12 A 600 . Cells were resuspended 1:2 (w/v) in 50 mM HEPES, 2 mM phenylmethylsulfonyl fluoride, pH 7.5, containing 1 mg/g fresh weight of DNase I and were disrupted by passage through a French pressure cell with a 70-megapascal pressure drop. The lysate was clarified by centrifugation at 250,000 ϫ g for 30 min and applied to a Poros20CM cation exchange column (Applied Biosystems, Foster City, CA). The column was developed with a linear gradient of 20 column volumes of 20 mM HEPES, pH 7.5, containing 0 -600 mM NaCl. Fractions enriched in desaturase were identified by SDS-PAGE, pooled, and concentrated with the use of an Amicon PM30 ultrafilter. The concentrate was subjected to size exclusion chromatography with a TSK G3000SW column (Mac Mod Analytical, Chadds Ford, PA) developed with 20 mM HEPES, pH 7.0, 70 mM NaCl. Desaturase enriched fractions were identified by SDS-PAGE and concentrated as before to ϳ0.15 mM desaturase (dimer) prior to crystallization.
Crystallizations-The hanging drop vapor diffusion method has been used for all crystallization experiments. The azide complex was prepared by adding 70 mM sodium azide to the protein solution prior to crystallization. Crystals were obtained in 0.08 M cacodylate buffer, pH 5.4, 200 mM magnesium acetate, 75 mM ammonium sulfate, 0.2% octyl glucoside, and 12-15% polyethylene glycol 4000 as precipitant as published for holoenzyme (15). The crystals were cryoprotected by a short soak in well solution with water exchanged for 20% (v/v) 2-methyl-2,7pentandiol. The cell parameters in P2 1 2 1 2 1 were as follows: a ϭ 192.9 Å, b ϭ 145.2 Å, and c ϭ 81.8 Å (i.e. ϳ3% shorter than the original cell due to the use of cryo conditions described above). They contain three dimers in the asymmetric unit. Data were collected at beamline 711 at MAX-LAB in Lund and processed with DENZO (27) and SCALEPACK (27) (Table I).
Crystals of the acetate complex were prepared as described earlier for holoenzyme, except for the addition of 20% (v/v) glycerol to the well solution. This results in a different space group, P3 1 12, with cell parameters a ϭ b ϭ 94.1 Å and c ϭ 81.7 Å. They contain only one subunit in the asymmetric unit. Data were collected at cryo temperature at beamline X12B at NSLS (Department of Biology, Brookhaven National Laboratory) and processed with MOSFLM (28) and SCALA (29) from the CCP4 suite (30).
The material used for crystallization of the iron-depleted form of desaturase consisted of a 2:2 complex (2 subunits of ⌬9 desaturase plus 2 molecules of stearoyl-ACP from spinach) verified by electrospray ionization (data not shown). The initial objective was to crystallize the complex between these two protein molecules. The iron depletion was achieved unintentionally by the use of N-(2-acetamido)-2-iminodiacetic acid buffer, which acted as a chelator. Crystals were obtained at 4°C with 8 -10% (w/v) polyethylene glycol 6000 as the precipitant, 60 mM strontium chloride, 0.1 M N-(2-acetamido)-2-iminodiacetic acid buffer, pH 6.7, and 15% (v/v) glycerol as cryoprotectant. The crystals, which were very pH-sensitive, belong to space group P3 1 with cell axes a ϭ b ϭ 188.0 Å and c ϭ 82.1 Å, and they contain three desaturase dimers in the asymmetric unit. Data were collected at cryotemperature at beamline X25 at NSLS and processed with DENZO (27) and SCALEPACK (27).
The crystals with a single iron present in the active site were made at room temperature with pH 4 using 8 -12% (v/v) polyethylene glycol 8000, 50 mM potassium dihydrogen phosphate, and 20% (v/v) glycerol as cryoprotectant. The crystals belong to space group P2 1  by molecular replacement using the program AMORE (31) with the original dimeric structure of ⌬9 desaturase (15) (Protein Data Bank accession number 1AFR) as a search model. Initial refinement of all four structures was performed with the use of the CNS program package (32). This refinement consisted of anisotropic scaling, bulk solvent correction, simulated annealing, conjugate gradient minimization, and isotropic B-factor refinement against the maximum likelihood target. Tight non-crystallographic symmetry restraints were used whenever possible to improve the data to parameter ratio. Averaged maps, calculated by density modification (33), were also utilized for model building carried out in O (34). Final refinement was performed with REFMAC5 (35) using the maximum likelihood residual, anisotropic scaling, bulk solvent correction, and atomic displacement parameter refinement with the "translation, liberation, screw rotation" (TLS) method (36). Each subunit was treated as a rigid group, but all water molecules were excluded. TLS refinement decreased the R-factors by ϳ3-4%. Annealed omit maps were used to confirm the binding mode of azide and acetate, respectively. To verify the significance of the shift in position of the side chain of Glu 196 in the acetate complex, it was perturbed to its original position in the structure of the reduced holoenzyme. Performing 60 cycles of minimization in REFMAC5 (35) confirmed the carboxylate shift. The geometry of the structures was checked with PROCHECK (37), and the final statistics are presented in Table I. Model coordinates and structure factors have been deposited in the Protein Data Bank with the following accession codes: azide complex, 1OQ4; iron-free form, 1OQ7; acetate complex, 1OQ9; one-iron form, 1OQB. The figures were prepared with Bobscript (38) and Raster3D (39).

RESULTS
Crystal Packing and Space Groups-Crystals from three space groups, P2 1 2 1 2 1 , P3 1 12, and P3 1 , with approximately the same length of their shortest cell axis have been obtained during this work (Table I). During crystallization, the desaturase dimers have a strong tendency to pack along a 3-fold screw axis, being noncrystallographic in the P2 1 2 1 2 1 crystal form, and are described in more detail in the original structure determination. The packing of desaturase dimers along the 3-fold axis gives rise to "rodlike" structures, and the three different space groups are created when these rods pack differently with respect to each other.
Quality of the Models-No large conformational changes are observed for the four models presented here compared with the original model (1AFR) of ⌬9 desaturase determined at room temperature (Table II). Minor differences are localized to three areas where all of the structures show very high B-factors and in some cases badly defined electron density. These areas comprise residues 18 -50, residues 205-215, and in particular residues 338 -346, which have been omitted in the acetate complex in P3 1 12, where they appear disordered. In regions of badly defined electron density, side chain occupancies have been set to zero.
For the azide complex in P2 1 2 1 2 1 and the acetate complex in P3 1 12, the electron density maps are of excellent quality in the active site. However, due to the absence of iron ions, the two iron-depleted models show considerable disorder of the carboxylate side chains in the active site. In all structures, the iron center bridging Glu 229 has more flexibility than other iron center ligands, as evidenced by higher B-factor and less well defined electron density.
Azide Binding to the ⌬9 Desaturase Di-iron Center-Occupancy of azide in the desaturase active site was high, and difference electron density corresponding to the azide ion was found in all six subunits of the asymmetric unit. No significant structural shifts have been introduced upon azide binding. The azide ion bridges the two iron ions of the desaturase di-iron center with no significant changes in the geometry of the coordinating carboxylate or histidine ligands compared with the original structure of the ferrous di-iron center. Therefore, this structure most likely represents azide bound to the reduced state of desaturase with an iron-iron distance of 4.1 Å. The azide ion binds the iron cluster in a -1,3 bridging mode, with iron-nitrogen distances of 2.5-2.6 Å (Fig. 2). Azide has displaced the weakly coordinated water molecule axial to the histidines, and the carboxylates are left in the plane of a distorted octahedron (Figs. 1 and 2). The orientation of the azide ion is clearly defined in the 2F o Ϫ F c electron density map, but some remaining positive electron density is observed in some subunits in the F o Ϫ F c map. This could reflect a small population of a second azide conformation that differs from that modeled in our 2.4-Å electron density map. A pathway of eight water molecules (four from each subunit) across the dimer interface connects the iron centers, which are ϳ23.5 Å apart (Fig. 3a). These water molecules form hydrogen bonds to main and side chain residues of the protein, and they are well defined in our electron density maps; three of them were also present in the original holo structure. The distance between the azide and the closest water molecule in the pathway is 3.6 Å. This water channel is present in both the azide-P2 1 2 1 2 1 and the acetate-P3 1 12 structure.
Acetate Binding to the ⌬9 Desaturase Di-iron Center-The overall structure of desaturase is very similar between this P3 1 12 structure and the P2 1 2 1 2 1 original structure (15). The root mean square deviation of 0.50 Å for 337 ␣-carbon atoms (Table II) is only slightly higher than for the P2 1 2 1 2 1 azide complex, 0.36 Å over 345 C-␣ atoms. There is a very slight reorientation of the subunits in the dimer in this space group; alignment of the dimer to the original holo structure gives a root mean square value of 0.66 Å versus 0.39 Å for the dimer azide complex.
Since the distance between the iron ions is long, 4.0 Å, we presume that the metal center was reduced during x-ray data collection as was previously reported for the holodesaturase crystals. Acetate bridges the iron center, displacing a water molecule, weakly coordinating the two irons in the structure of the reduced enzyme. The oxygen to iron distance is 2.6 and 2.5 Å, respectively. When acetate binds to the desaturase di-iron center (Fig. 4), Glu 196 , a bidentate ligand in the original structure of the reduced di-iron center (Fig. 1), appears as a monodentate ligand to Fe2. The distance between Fe2 and the oxygen atoms of Glu 196 are unchanged for O⑀1 (2.2 Å), whereas O⑀2 increases from 2.5 to 3.0 Å between the structures. The 2.4-Å electron density map of the acetate complex is of high quality in the active site, and the small shift observed in Glu 196 is therefore considered a valid feature of the structure. The orientation of Glu 196 was confirmed by calculation of an annealed omit map, and furthermore, the shift is reproducibly obtained on refinement after perturbation of the side chain to its bidentate configuration. Acetate binding in conjunction with this carboxylate shift results in a distorted octahedral coordi-   The Di-iron Center of ⌬9 Stearoyl-ACP Desaturase nation of Fe1 (five protein ligands plus one acetate ligand) and a distorted trigonal bipyramidal coordination of Fe2 (four protein ligands and one acetate ligand).
The One Iron Structure of ⌬9 Desaturase-The electron density maps showed unambiguously that one of the iron ions had been lost from the cluster at pH 4. Fe1 remains in its position in the active site with unchanged 5-coordination by ligands from the protein, whereas Fe2 is lost to the surrounding solvent (Fig. 5). No large structural perturbations are observed. Compared with the original ferrous di-iron center structure (Fig. 1), the absence of Fe2 causes some flexibility in its bidentate ligand Glu 196 and in the normally bridging Glu 229 . The positions of the other two carboxylate iron ligands, Glu 105 and Glu 143 , are not changed compared with the reduced di-iron center. Glu 229 and Glu 196 appear to form a 2.6-Å hydrogen bond between each other, probably caused by protonation of Glu 196 O⑀2 at pH 4. We are unable to distinguish whether the remaining Fe1 is in the ferrous state or in the oxidized ferric state.
The Structure of Apo ⌬9 Desaturase-Electron density for the iron cluster was completely absent in this structure and the normally liganding residues showed some disorder. The overall structure of the apo form of ⌬9 desaturase shows little difference to the ferrous enzyme at this resolution. The protein is thus able to form a stable structure also in the absence of the positive charges provided by the di-iron center. The buried carboxylate side chains of the empty di-iron center shift positions slightly and appear to be protonated, since they are at hydrogen bond distance to each other (Fig. 6); a similar observation was made for the apo form of R2 (40). The adjacency of the carboxylate side chains has increased their pK a , allowing them to be protonated at pH 6.7. In the absence of protonation of the di-iron ligands, the close proximity of negative charges would be expected to destabilize the desaturase four-helix bundle. The resulting pattern of hydrogen bonds could be different in the ensemble of protein molecules in the crystals and contributes to the observed disorder of the carboxylate residues in the di-iron center, especially Glu 229 , which is almost undefined in the 2F o Ϫ F c electron density map.
A peak of very high difference electron density, probably representing a novel metal ion-binding site, was found in the dimer interface of apodesaturase (Fig. 3b). Since strontium chloride (60 mM) was present during crystallization, we expect the bound metal ion to be strontium. We have observed the same metal-binding site but with bound iron ions in electron density maps of other desaturase crystals. 2 The metal ion has bidentate coordination by Glu 106 from both subunits of the dimer at distances in the range of 2.8 -3.2 Å. There are no water molecules modeled in this low resolution map of apodesaturase, but the metal in the dimer interface is located in the middle of the pathway of water molecules connecting the two metal centers of the ⌬9 desaturase dimer (Fig. 3a). The distance from the high affinity iron (Fe1) sites and the ion in the dimer interface is 12.2 Å.

DISCUSSION
The Azide Complex-Earlier spectroscopic studies have shown that when azide binds the resting oxidized state of ⌬9 desaturase, the -oxo bridge is either protonated or lost (41). Spectroscopy further predicted two different, pH-dependent, binding modes of azide (-1,3-bridging and 1 -terminal), with ϳ90% being in the bridging mode at pH 6.2, which is the pH used in our crystallization. The crystal structure of the ferrous azide complex shows the -1,3 binding mode but also shows some difference density, suggesting the presence of a small proportion of azide in a -1,1 binding mode. In fact, the spectroscopic data for the second conformation of the ferric azide complex could be interpreted as either 1 -terminal with a protonated -oxo-bridge or as -1,1 coordination, where the -oxobridge is lost. The data presented here support the latter model.
No structures of azide complexes from wild-type di-iron center enzymes of this class are known. However, they have been determined for the F208A/Y122F and E238A/Y122F mutants of the reduced R2 subunit of ribonucleotide reductase (42,43). The azide ion is close to -1,1-bridging in both R2 mutants occupying the position of the -oxo bridge (E238A/Y122F) or of a second solvent molecule (F208A/Y122F) coordinated by Fe2 in the oxidized di-iron center. In both mutants of R2, the azide extends away from the di-iron center, occupying the available space introduced by the mutations.
In contrast to these R2-mutant complexes, the desaturase azide complex is strikingly similar to the azide complex of reduced rubrerythrin (Fig. 7), showing virtually identical coordination (44). As in desaturase, binding of azide to the reduced state of rubrerythrin introduces no carboxylate shifts in the surrounding iron ligands, and it binds azide in the same fash- ion. An important difference in the structure of the oxidized iron center in rubrerythrin and in oxidized R2 and MMOH is the position of the -(hydr)oxo bridge (45)(46)(47). Whereas it replaces the position corresponding to Glu 229 in reduced R2 and MMOH, it is at the other side of the iron-iron axis (where MMOH has its second solvent bridge) in rubrerythrin, occupying the site where azide is bound. From spectroscopy, the -1,3 binding of azide is known to displace the -oxo-bridge in desaturase, and we thus suggest that the -oxo-bridge in oxidized desaturase is bound similarly as in rubrerythrin.
The Acetate Complex-Carboxylate shifts have been proposed to play an essential role during catalysis in di-iron enzymes through modifying the electronic structure of the iron center (48). The small carboxylate shift of Glu 196 observed in the acetate complex changes the low affinity iron ion (Fe2) geometry from a distorted octahedral coordination to a distorted trigonal bipyramidal coordination. This represents the first carboxylate shift identified in a ⌬9 desaturase crystal structure.
In a study of magnetic circular dichroism in chemically reduced ⌬9 desaturase, it has been shown that the coordination environment of the reduced di-iron center is influenced by the presence of stearoyl-ACP, changing it from two equivalent 5-coordinated irons in a distorted square pyramidal geometry into a geometry where Fe2 is 4-coordinated (49). This correlates to what is observed here, where both iron ions of the original reduced di-iron center are 5-coordinated, but in the acetate complex, Fe2 is 4-coordinated by ligands from the protein. It was suggested that the change in iron coordination, caused by binding of stearoyl-ACP, increases the reactivity against molecular oxygen (49). This is an attractive model, since it would prevent formation of reactive oxygen intermediates in the absence of stearoyl-ACP. In our case, the reactivity of the di-iron center toward binding of acetate is indeed increased by this change in coordination, since acetate does not bind to the di-iron center in P2 1 2 1 2 1 but only to the di-iron center in the P3 1 12 crystal form, where Fe2 is 4-coordinated. No significant shift in ligation is observed in the azide complex in P2 1 2 1 2 1 .
Acetate has been observed bound to the oxidized di-iron center of methane monooxygenase hydroxylase from M. capsulatus (47), and the two complexes are compared in Fig. 8. Acetate binds in the same location opposite to the histidine iron ligands. In the case of ⌬9 desaturase, we suggest that acetate has replaced the -oxo bridge, but in MMOH it has replaced a second solvent bridge. In both ⌬9 desaturase and MMOH, the histidine iron ligands are hydrogen bonding to carboxyl side chains. Conserved aspartate residues in di-iron hydroxylases (50) are structurally equivalent to Asp 228 and Glu 142 in ⌬9 desaturase. Both in desaturases and hydroxylases, the charges of these carboxylate side chains are compensated for by interaction with conserved arginine residues corresponding to Arg 231 and Arg 145 in ⌬9 desaturase. The difference is that the histidine ligand of the low affinity Fe2 forms a hydrogen bond to a conserved glutamate (Glu 142 ) in ⌬9 desaturase and a conserved aspartate (Asp 143 ) in MMOH (Fig. 8). The conservation of these histidine-liganding carboxylates underscores the importance of correct positioning of the histidines for the geometry and reactivity of the di-iron clusters.
Azide-and Acetate-bound Iron Centers as Analogues of a Peroxodiferric Intermediate of ⌬9 Desaturase-A stable -1,2peroxide intermediate species that decays with formation of water instead of oleoyl-ACP can be prepared (21,22). The azide ion bound with -1,3 coordination to the irons and the similarly bound acetate are assumed to mimic this unreactive intermediate. The formation of water instead of product from this stable peroxide intermediate is probably caused by an excess of electrons present due to chemical reduction. The extra electrons delivered from surplus dithionite or from the second subunit of desaturase can intercept formation of the reactive intermediate that follows the peroxo intermediate, preventing it from abstracting hydrogens from stearic acid. An attractive hypothesis is that the extra electrons are delivered via the water pathway that connects the two iron sites and that the required protons are delivered via the same path (i.e. a net transfer of two hydrogens).
Little information about the intermediates formed in the presence of the natural electron transport chain is available for ⌬9 desaturase, since most of the results obtained from spectroscopy have been made with chemically reduced enzyme. When the natural electron transport chain is used for catalysis, the peroxo intermediate, leading to formation of oleoyl-ACP, is short lived in contrast to the -1,2 peroxide intermediate that decays with formation of water (22). The subtle structural modifications giving rise to this difference in reactivity are unknown, but it seems reasonable to assume that the stable peroxo intermediate and the catalytically competent peroxo intermediate will be quite similar. We propose that the acetate complex, where a carboxylate shift has taken place, increasing the reactivity of the iron-center, is closer in structure to the catalytically competent -1,2-peroxo intermediate than the az-ide complex. The nature of the reactive intermediate that follows the peroxo intermediate in desaturase and abstracts hydrogen atoms from stearic acid remains to be characterized.
The reaction that follows upon the addition of hydrogen peroxide to ⌬9 desaturase is currently not well documented, but it is known that the presence of catalase speeds up the desaturase reaction in the enzymatic assay (3,51). The hydrogen peroxide consumed by catalase was suggested to be formed by ferredoxin and ferredoxin reductase used in the desaturase assay (3). We propose that hydrogen peroxide inhibits ⌬9 desaturase by binding to the reduced di-iron center bridging the iron atoms in the same way as the azide and acetate ions presented here. Taken together, the striking similarity of the ⌬9 desaturase and rubrerythrin azide complexes and the ability of the -1,2 peroxodiferric intermediate in ⌬9 desaturase to decay with formation of water, raise the possibility that the reduced state of desaturase can act as a peroxidase, albeit at a low rate. Formation of water is slow, since a residue corresponding to Glu 97 of rubrerythrin, catalyzing the required proton transfers (44,52), is occupied by Thr 199 in ⌬9 desaturase. Experiments are currently under way to test this hypothesis.
If desaturases have a low peroxidase activity, they might have a secondary role contributing to protection against oxidative stress. It has been shown that a ⌬12 desaturase from C. elegans introduced and expressed in yeast (Saccharomyces cerevisiae) increased the tolerance against hydrogen peroxide in the yeast, although this was attributed to changes in membrane fluidity and not to peroxidase activity of the introduced ⌬12 desaturase gene (53).
Iron-deficient ⌬9 Desaturase-Since desaturase can be observed with only one of the iron sites specifically occupied at pH 4, the two binding sites must have different affinities for iron. This is a reflection of differences in pK a of the carboxylate ligands that will be manifested as a difference in affinity also at physiological pH. Fe1 is the high affinity and Fe2 is the low affinity binding site for iron in the di-iron center. Differential iron affinity has been observed for other di-iron proteins (e.g. MMOH, the R2 subunit of ribonucleotide reductase, and bacterioferritin) (54 -57). Desaturase has the same high affinity site (Fe1) as MMOH from M. capsulatus and bacterioferritin from Rhodobacter capsulatus, whereas R2 from both mouse and E. coli have Fe2 as their high affinity site.
From Mössbauer spectroscopy, it is known that ⌬9 desaturase, which has been reduced by dithionite, either by itself or in the presence of the substrate stearoyl-ACP, consists of two high spin ferrous iron atoms with slightly different Mössbauer parameters (21). Since no spectroscopy has been reported regarding desaturase reduced by the natural electron transport chain, the spin state of the catalytically competent di-iron center remains unknown. The high spin ferrous state has more loosely bound electrons than the low spin ferrous state; therefore, at least one of the irons could be expected to be in the high spin ferrous state when delivering the first electron to molecular oxygen. It is possible that a mixture of one ferrous iron high spin and one ferrous iron low spin is the optimal arrangement for activating molecular oxygen, since the low spin state has a higher affinity for oxygen at least in heme-containing globins (58,59). The low spin iron could be used for binding molecular oxygen, whereas the high spin iron delivers the first electron. The high spin iron in ⌬9 desaturase would be Fe2, since it is more loosely bound than Fe1, and the Fe1 would bind the oxygen.
To date, there is no evidence for a system for active insertion of iron into desaturases; indeed, in overexpression of the recombinant enzyme, iron insertion occurs spontaneously. The presence of an auxiliary binding site for metal ions at the subunit surface, as observed with a bound strontium ion in the iron-free structure and bound iron ions in other ⌬9 desaturase structures, 2 suggests a possible route for spontaneous iron insertion. Under this scenario, primary iron trapping would occur at this site, and the iron ions would then be transported along the water channel connecting this site to the iron center.
Conclusions-The ⌬9 desaturase acetate complex reveals the first observation of a carboxylate shift for a desaturase di-iron enzyme. This makes Fe2 4-coordinate with respect to protein ligands, increasing the binding affinity for acetate. The desaturase azide and acetate complexes are proposed to mimic a stable -1,2-peroxo intermediate closely related to the catalytically competent species present during productive turnover. The strong structural similarity to the reduced rubrerythrin azide complex suggests that desaturase might catalyze formation of water from hydrogen peroxide at a low rate. We also propose that the -oxo bridge in oxidized ⌬9 desaturase is bound as in rubrerythrin rather than as in R2 and MMOH.
The structure determinations of the apo and one-iron forms of ⌬9 desaturase show that the enzyme forms a stable structure upon removal of iron from the active site. They further show that the two iron atoms bind with different affinity to the di-iron center and identify Fe1 as the high affinity iron site.