Adrenodoxin Reductase-Adrenodoxin Complex Structure Suggests Electron Transfer Path in Steroid Biosynthesis*

,

In mitochondria of the adrenal cortex, the cytochrome P450 enzymes of the CYP11 family catalyze the side chain cleavage of cholesterol to form pregnenolone (P450scc, 1 CYP11A1) and are involved in the formation of cortisol (P45011␤, CYP11B1) and aldosterone (P450aldo, CYP11B2) (1). The enzymatic activity of the cytochrome P450-dependent steroid hydroxylases is based on their ability to activate molecular oxygen by reductive splitting of dioxygen. This multistep reaction requires the transfer of electrons from the flavoprotein adrenodoxin reductase (AR) via adrenodoxin (Adx) to the terminal cytochromes P450 as electron acceptors in dependence on the specific hydroxylation substrate (1)(2)(3). Several models for electron transfer have been discussed, including a shuttle model in which Adx forms consecutive 1:1 complexes (4) with AR and cytochrome P450scc and models requiring the formation of an organized 1:1:1 ternary complex (5) or a 1:2:1 quaternary complex (6) between AR, Adx, and cytochrome P450scc. Common to these models is a complex between AR and Adx during the first steps of electron transfer from the reductase to the cytochrome P450.
Recently, the crystal structures of two forms of bovine adrenodoxin (7,8) and of adrenodoxin reductase (9) were determined. These structures revealed the general topology of the two proteins and the molecular environments of the [2Fe-2S] cluster of Adx and the FAD moiety of AR. Here, we report the 2.3-Å resolution crystal structure of a cross-linked 1:1 complex of full-length Adx and AR. This structure shows the geometry of an electron transfer complex of soluble, freely dissociable proteins from a higher eukaryote for the first time, highlights structural adaptations that accompany the binding of AR to Adx, and permits us to predict electron transport paths in their complex.

EXPERIMENTAL PROCEDURES
Sample Preparation-Recombinant bovine Adx and AR were purified and crystallized as described (10). The synthesized Adx differs from the wild-type protein by the exchange of Ser 1 for glycine and is composed of 128 amino acids, including the N-and C-terminal residues missing in the truncated adrenodoxin, Adx-(4 -108), studied earlier (7). Crosslinking of AR to Adx has also been described (11,12). The native complex is formed at low ionic strength between the two proteins, and the cross-linking was carried out with the water-soluble coupling reagent 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC, purchased from Sigma). The mixture of Adx (1.2 mol) and AR (300 nmol) was dialyzed for 18 -20 h against 20 mM potassium phosphate, pH 7.2, followed by addition of an equal volume of fresh 8 mM EDC solution in distilled water and incubation at 4°C in the dark with occasional stirring. After 8 h, excess of the reagent was removed on a Sephadex G-25 column equilibrated with 10 mM potassium phosphate, pH 7.4. The colored fraction was pooled and applied on a 2.4 ϫ 10 cm DEAE-Fractogel column and washed with two gradient solutions as follows: 10 -50 mM potassium phosphate, pH 7.4, for 3 h; 50 -100 mM potassium phosphate, pH 7.4, for 3 h. The peak containing the covalent crosslinked complex of the recombinant Adx and AR was consequently purified on an AD-Sepharose column to remove residual AR and on an ADP-Sepharose column to remove unbound Adx. The cross-linking of AR to Adx with EDC was expected (11,12) to yield an amide bond between the ⑀-amino group of Lys 66 in Adx and the ␥-carboxyl group of Glu 4 in AR.
X-ray Data Collection-Four x-ray diffraction data sets from three crystals were collected at 100 K on MAR345 imaging plates at beam lines BW7B (EMBL Outstation at DESY, Hamburg) and BW6 (MPG-ASMB, c/o DESY). Due to problems with spatial reflection overlaps * This work was supported by Deutsche Forschungsgemeinschaft Grants He 1318/19-1 and WER436 and the Fonds der Chemischen Industrie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The caused by the long c axis of 607.85 Å, the data reached only 79% completeness (85% at 2.5 Å resolution), although several detector settings were used. The data sets were processed by DENZO/SCALEPACK (13) and contained 55,229 unique reflexes after merging (Table I).
Structure Determination and Refinement-The structure of the AR⅐Adx complex was solved by molecular replacement using the coordinates of AR and Adx-(4 -108) as deposited in the Protein Data Bank (codes 1cjc and 1ayf). Two complexes per asymmetric unit with a molecular mass of 64,927 Da each were assembled by placing the protein molecules into the unit cell using AMORE (14) and rigid-body refinement with the program CNS (15) resulting in R ϭ 0.386, R free ϭ 0.394 at 2.8 Å resolution. Application of a solvent mask, positional and atomic temperature factor refinement, several rounds of manual density fitting, and the addition of 5 sulfate ions and 277 water molecules reduced R to 0.223 and R free to 0.268 (Table I). 1132 residues out of 1176 could be localized within the electron density. Almost all modeled water positions are also occupied in the crystal structures of Adx-(4 -108) (7) and AR (9). The averaged main chain and side chain parameters are equal or better than those in a set of 118 structures used by PRO-CHECK (16), and the Ramachandran diagram is free of outliers.

RESULTS AND DISCUSSION
Architecture of the AR⅐Adx Complex-The hexagonal crystals used in this analysis are formed by cross-linked 1:1 complexes between AR and Adx, both in their oxidized form. Two complexes related by a noncrystallographic screw rotation are present in the asymmetric unit. Complex I contains residues 5-117 of Adx and 4 -460 of AR, and complex II contains residues 5-110 of Adx and 5-460 of AR. Electron density for both complexes clearly reveals the [2Fe-2S] cluster of Adx and the FAD moiety of AR. As in the crystal structure of the free protein (9), no NADP is bound to AR. With the exception of a small domain rearrangement in AR (see below), protein conformation in these independent copies of the complex is generally similar allowing least squares superpositions of the Adx C␣ atoms with a root mean square deviation (r.m.s.d.) of 0.46 Å and of the AR C␣ positions with a r.m.s.d. of 0.95 Å. The description of the crystal structure of the AR⅐Adx complex will focus on complex I, where the C terminus of Adx is ordered up to Ser 117 , reveal-ing a number of residues neither observed in the structure of Adx-(4 -108) (7) nor in that of full-length Adx (8).
By fitting the globular Adx molecule into a prominent depression on the AR surface, a compact AR⅐Adx complex is formed (Fig. 1). A total of 580 Å 2 of solvent-accessible surface are buried between the protein molecules. This AR-Adx interface contains many polar residues. In the complex, Adx contacts both AR domains. In a primary interaction region, polar contacts are formed between residues of the NADP domain of AR and the Adx side chains belonging to the interaction domain (7) of the protein. Further polar interactions take place in a secondary interaction region where the core domain (7) of Adx contacts the FAD domain of AR and the covalent cross-link is formed linking Adx Asp 39 with AR Lys 27 . In a third interaction region, the C-terminal polypeptide stretch of Adx dips into a deep cleft between the two globular domains of AR. Adx residues Asp 113 to Arg 115 of this region adopt 3 10 -helical conformation. This contact is assumed to be rather loose, since the C-terminal residues of Adx adopt high atomic displacement factors up to 80 Å 2 indicating flexibility, and the interaction is not observed in complex II of the crystal where no electron density is seen beyond Ala 110 of Adx. Finally, further hydrogen bonding and van der Waals interactions are observed between residues bridging the [2Fe-2S] cluster of Adx and the isoalloxazine ring of the FAD of AR.
Two electron density peaks near the AR N terminus were assigned as sulfate ions. Since AR is a membrane-associated protein (18,19), it is tempting to speculate that these sulfate positions mark interaction sites of AR with phospholipids in the membrane. This hypothesis is supported by the observed arrangement of hydrophobic (e.g. Trp 420 ) and basic residues (Arg 31 , Arg 70 , Arg 73 , Lys 411 , Lys 429 , and Arg 456 ) around the sulfate ions that might interact with lipid or phosphate moieties of the membrane.
Reorientation of AR Domains during Complex Formation-Both Adx and AR are two-domain proteins. Adx consists of a core domain containing the [2Fe-2S] cluster and a small interaction domain (7), and AR contains a FAD domain and a NADP domain of about equal size (9). Whereas no significant difference between Adx-(4 -108) and Adx as present in the complex is detected, the two AR domains show a slightly different orien-  tation with respect to each other when the complex is compared with free AR (Fig. 2). After superposition of the FAD domains of AR (r.m.s.d. ϭ 0.38 Å), optimal fit of the NADP domains requires a 3.7°rotation (7.2°for complex II). Considering Arg 240 and Lys 27 of AR as reference contact points with Adx (see below), this domain reorientation results in a narrowing of the distance between these two anchor points by 2.4 Å (4.2 Å in complex II). AR thus has the ability to adapt to the Adx molecule in the binary complex by domain reorientation to various degrees.
Complex Formation by Electrostatic Interactions-The AR⅐Adx complex displays a highly charged surface (Fig. 3, top) arising from interacting surfaces that are predominantly acidic (Adx) or basic (AR). Of the 580 Å 2 of solvent-accessible surface buried in the complex, 325 Å 2 are from hydrophobic side chains. Nearly half of the AR-Adx interface is composed of polar and charged residues engaging in a large number of hydrogen bonds and salt links. Hence, electrostatic interactions may be considered the primary driving force for complex formation in agreement with chemical modification (21) and site-directed mutagenesis experiments (1,(22)(23)(24) of AR and Adx.
Electrostatic interactions predominate in the two main interaction sites of the AR⅐Adx complex. In the primary interaction region (Fig. 3, bottom left), arginines 211, 240, and 244 of the NADP domain of AR are involved in numerous salt bridges with Adx carboxylate groups. Aspartates 72, 76, and 79 of the Adx interaction domain are binding partners to AR, whereas the acidic residues Glu 73 and Glu 74 of Adx are facing away from the interface. The electron density provides no evidence for a covalent cross-link formed at this interaction site. Acidic Adx residues located at the primary AR-Adx interaction region are known also to be involved in cytochrome P450scc binding (1). Given the participation of several of these side chains in contacts to AR, the formation of an organized 1:1:1 complex between AR, Adx, and cytochrome P450scc for electron transport during steroid biosynthesis must be regarded as very unlikely.
A secondary interaction region is centered around the Adx residues Asp 39 and Asp 41 contacting His 28 and Lys 27 of AR, respectively (Fig. 3, bottom right). Again, these contacts are polar and are mainly formed by charged side chains. Asp 39 and Asp 41 are located in the core domain of Adx at a surface region that has been implicated in cytochrome P450cam (CYP101) binding by the homologous putida redoxin (27). An involvement of Asp 39 in redox-partner binding has been suggested earlier (28) based on a comparison of the Adx structure with crystal structures of plant-type ferredoxins (29).
Covalent cross-linking of Adx and AR with carbodiimide prior to crystallization results in the formation of a peptide bond between the carboxylate function of Asp 39 and the primary amino group of Lys 27 as clearly revealed by electron density (see Fig. 3, bottom right). This finding is unexpected, FIG. 3. Electrostatic interactions between AR and Adx. Top, surface drawings of AR (right), the AR⅐Adx complex in the orientation displayed in Fig. 1 (center), and Adx (left). Adx and AR are rotated relative to their orientation in the complex as indicated to emphasize the interacting surfaces. Surfaces are colored corresponding to the electrostatic potential calculated by the program DELPHI (25)  since the cross-linking procedure employed was reported (11,12) to yield a covalent bond linking AR Glu 4 and Adx Lys 66 . Peptide sequencing and mass spectrometric analysis prove that the Glu 4 -Lys 66 cross-link is indeed not formed in the AR⅐Adx complex. 2 The suggested Glu 4 -Lys 66 cross-link is incompatible with the binding mode to AR of Adx reported here. It does not permit contacts between the proteins in the primary interaction region as supported by mutagenesis experiments (1,(22)(23)(24) and renders unlikely a close enough approach of the redox centers for electron transfer. For these reasons we are convinced that the reported complex, and not a complex cross-linked at AR Glu 4 / Adx Lys 66 , represents the functional interaction between AR and Adx.
Possible Electron Transfer Path-Efficient electron transfer between the redox centers requires spatial proximity. The closest approach of atoms belonging to the [2Fe-2S] cluster of Adx and the isoalloxazine ring of the FAD of AR is 10.3 Å (9.65 Å in complex II), well within the 14-Å threshold reported to define the limit of electron tunneling in a protein medium (30). The fractional packing density of protein groups between the redox centers is 0.61 (0.73 in complex II), again within the observed range of densities found in natural multiredox center oxidoreductases of known structure (30). Thus, from proximity and packing density considerations alone, one may conclude that the geometry of the AR⅐Adx complex will support electron tunneling between the redox centers. The observed geometry is calculated by ETUNNEL (30) to support electron transfer rates of 10 8 to 10 9 s Ϫ1 . This is orders of magnitude above the experimentally determined (4) flavin-to-iron transfer rate of 3-4 s Ϫ1 . By assuming that the covalent cross-link does not force an unnaturally tight AR-Adx interaction, it may thus be concluded that the rate of the redox reaction in which AR and Adx are involved is not limited by electron transfer within the AR⅐Adx complex.
The program HARLEM, analyzing distinct protein structures with respect to tunneling probabilities (31), was further used to compute possible electron transfer routes between AR and Adx (Fig. 4). According to this analysis, electrons would most likely travel along covalent bonds, requiring two throughspace jumps from the FAD isoalloxazine to AR Ile 376 and from AR Thr 377 to Adx Cys 52 , one of the [2Fe-2S] ligands. However, alternative transfer paths and a possible involvement of water molecules located at the interface region cannot be ruled out.
In summary, the 2.3-Å crystal structure of the AR⅐Adx redox complex suggests modes of electron transfer between a soluble [2Fe-2S] ferredoxin and its cognate reductase. It reveals the importance of electrostatic interactions in complex formation, in agreement with the concept of "electrostatic steering" (1, 7), and demonstrates that a slight domain rearrangement in AR is required for a tight AR-Adx interaction. FIG. 4. Electron transfer region between the [2Fe-2S] cluster of Adx and the FAD moiety of AR. The hypothetical electron pathway shown in red was calculated with the program HARLEM (31). Red dotted lines mark through-space electron jumps. The AR-Adx interface is stabilized by hydrogen bonds (blue dotted lines) and van der Waals contacts. Residues are labeled black for AR and red for Adx. The blue spheres are water molecules.