Structural characterization of human cholesterol 7α-hydroxylase.

Hepatic conversion to bile acids is a major elimination route for cholesterol in mammals. CYP7A1 catalyzes the first and rate-limiting step in classic bile acid biosynthesis, converting cholesterol to 7α-hydroxycholesterol. To identify the structural determinants that govern the stereospecific hydroxylation of cholesterol, we solved the crystal structure of CYP7A1 in the ligand-free state. The structure-based mutation T104L in the B′ helix, corresponding to the nonpolar residue of CYP7B1, was used to obtain crystals of complexes with cholest-4-en-3-one and with cholesterol oxidation product 7-ketocholesterol (7KCh). The structures reveal a motif of residues that promote cholest-4-en-3-one binding parallel to the heme, thus positioning the C7 atom for hydroxylation. Additional regions of the binding cavity (most distant from the access channel) are involved to accommodate the elongated conformation of the aliphatic side chain. Structural complex with 7KCh shows an active site rigidity and provides an explanation for its inhibitory effect. Based on our previously published data, we proposed a model of cholesterol abstraction from the membrane by CYP7A1 for metabolism. CYP7A1 structural data provide a molecular basis for understanding of the diversity of 7α-hydroxylases, on the one hand, and cholesterol-metabolizing enzymes adapted for their specific activity, on the other hand.

(codons  was replaced with an optimized sequence, MAKKTSS. The C-terminally His 5 -tagged protein was coexpressed with GroEL/ES (pGro12, Takara Bio Inc.) in Escherichia coli JM109. Cells were lysed by passing through a Microfl uidizer (Microfl uidics Corp.) at 18,000 psi. Sodium cholate was added to the lysate at a fi nal concentration of 23 mM, and the lysate was incubated at 4°C for 1 h. After centrifugation at 60,000 g for 1 h, protein was purifi ed using metal affi nity chromatography on a HiTrap chelating column charged with Ni 2+ (Amersham Biosciences) and cation-exchange chromatography using a Source 30S column (Amersham Biosciences). The protein storage buffer was 50 mM potassium phosphate buffer, pH 7.4, 20% glycerol, and 0.5 M NaCl. The molecular mass of purifi ed protein measured by ESI-MS was 56,255 Da (expected 56,257 Da). The spectrophotometric index OD 418/OD280 of purifi ed sample used in crystallization trials was 1.5. A total P450 protein concentration was determined from reduced CO difference spectra ( 21 ).

Protein crystallization
Purifi ed CYP7A1 was crystallized in the presence of 100 M cholesterol using the hanging drop vapor diffusion method at 18°C after mixing 1 µl of the protein solution with 1 µl of the reservoir solution containing 0.1 M sodium chloride, 0.1 M trisodium citrate pH 5.5, and 20% polyethylene glycol (PEG) 400. Crystals were soaked with 30% glycerol as cryoprotectant before fl ash freezing in liquid nitrogen. To obtain a complex with cholest-4-en-3-one, a T104L mutant was used, and the crystals were grown in conditions containing 0.1 M MES pH 6.0 and 18% PEG 550 monomethylether (MME) at room temperature and soaked with 20% glycerol for fl ash freezing in liquid nitrogen. For the T104L mutant-7KCh complex, crystals were grown in 0.1 M sodium chloride, 0.1 M tri-sodium citrate pH 5.6, and 20% PEG 400.

Structure determination and refi nement
Crystallographic methods and results are summarized in Table 1 . Diffraction data were collected at 100 K at synchrotron radiation sources. The ligand-free structure was solved by molecular replacement with PHASER ( 22 ) and a search model derived from Protein Data Bank (PDB ) entry 2IAG ( 23 ). Density-modifi ed maps from RESOLVE ( 24 ) guided pruning of the initial model. Molecular replacement with an intermediate model and MOLREP ( 25 ) produced the current orientation of the model in the unit cell. The ligand-free structure subsequently served as search model for molecular replacement solution of the complex structures with PHASER. Restraints for substrate geometry were calculated with PRODRG ( 26 ).

Spectral binding assay
Ligand binding affi nities were determined with the spectralligand binding assay as described previously ( 27 ).

Enzyme assays
The 7 ␣ -hydroxylase activity of CYP7A1. Activity was measured in the reconstituted system at 37°C in 25 mM Hepes buffer pH 7.2 containing 0.1 mM DTT, 0.1 mM EDTA, 4 mM MgCl 2 , and 0.15% sodium cholate. The aliquots of concentrated recombinant proteins were mixed and preincubated for 5 min at room temperature. The fi nal concentrations of CYP7A1 and NADPHcytochrome P450 reductase (CPR) were 0.5 and 1.0 M, respectively. Cholest-4-en-3-one (10 mM in 45% hydroxypropyl-beta-cyclodextrin, HPCD ) was added to the reaction mixture at a fi nal concentration of 50 . After 10 min of preincubation at 37°C, the reaction was started by adding NADPH to a fi nal concentration of 0.25 mM. Aliquots (0.5 ml) were taken from the incubation mixture at selected time intervals. Steroids were extracted with 5 ml pathways in the retina, 7KCh might be a pathogenetic factor in age-related macular degeneration ( 9 ). The major routes for 7KCh metabolism include a ) conversion into less toxic 27-hydroxylated 7KCh by CYP27A1/CYP46A1 ( 10-13 ) and further to more water-soluble metabolites, thus protecting mitochondria from reactive oxygen species; b ) esterifi cation including sulfonation ( 14 ); c ) lipoprotein-mediated elimination ( 15 ); and d ) interconversion to 7 ␤ -hydroxycholesterol by 11 ␤ -hydroxysteroid dehydrogenase type 1 ( 16,17 ). However, the mechanisms of 7KCh formation and elimination are still not fully understood.
Conversion of cholesterol into 7 ␣ -hydroxycholesterol by CYP7A1 represents the fi rst and rate-limiting step in the classic pathway of bile salts biosynthesis. Hydroxylation of the ring system of cholesterol in a regio-and stereospecifi c manner with further oxidation and shortening of the side chain produces water-soluble bile acids with powerful detergent properties to emulsify dietary lipids ( 4 ). Bile acids also serve as signaling molecules that bind to G-protein-coupled receptors (GPCRs) and nuclear hormone receptors that regulate lipid, glucose, and energy metabolism ( 18 ). A modulation of both oxysterol and bile acid signaling pathways has recently emerged as a source of promising novel drug targets to treat common metabolic and hepatic diseases ( 19 ).
Bile acid synthesis is tightly regulated through the transcriptional regulation of CYP7A1 ( 1 ) and possibly by the availability of substrate to the enzyme, which is located in the cholesterol-poor endoplasmic reticulum ( 20 ). In humans, three cytochrome P450 enzymes perform the 7 ␣hydroxylation reaction: CYP7A1 is specifi c for cholesterol, CYP7B1 for oxysterols and steroids, and CYP39A1 for 24( S )hydroxycholesterol. None of these 7 ␣ -hydroxylases are structurally characterized, hindering understanding of the molecular mechanisms of their substrate selectivity. Here we present crystal structures of CYP7A1, both unliganded and bound to either the substrate cholest-4-en-3-one or the inhibitor 7KCh. Ligands are bound deep in the active site cavity, isolated from the bulk solvent, in a previously unobserved orientation complemented by unusual structural features of CYP7A1. An asparagine (Asn ) residue in place of the highly conserved threonine (Thr) in I helix does not directly interact with cholest-4-en-3-one. Instead, the Asn residue appears to interact with the sixth ligand of the heme iron in the ligand-free and 7KCh structures. Maintaining the network involving the 7-keto group and interaction with the protein's active site residues along with closed conformation of the access channel explain a competitive inhibition by 7KCh. Combining these new fi ndings with previous data prompted us to suggest a working model for cholesterol binding by CYP7A1 from the membrane.

CYP7A1 protein expression and purifi cation
The CYP7A1 cDNA was purchased from Origene [accession code TC123882 (NM_000780)] and subcloned into a modifi ed pCW-LIC vector. The N-terminal transmembrane anchor domain heme plane; a profound I helix kink disrupted the G helix at C term to form a separate G'' helix; an Asn residue instead of the conserved Thr, which is involved in oxygen activation in most P450s; and a long meander region on the proximal surface, a site for the interaction with the redox partner ( Fig. 1A, B ). CYP7A1 was structurally less similar to CYP51, the evolutionary oldest P450, with the preserved function of sterol 14 ␣ -demethylation in the cholesterol biosynthetic pathway.
Ligand-free CYP7A1 crystallized in the presence of cholesterol adopted a closed conformation with no obvious access channels. It is unclear whether the rigid and closed active site was a crystallization artifact or refl ected the in vivo state of the enzyme in the absence of the substrate and/or certain physiological stimuli. Electron density maps showed some peaks in the putative substrate binding site, but density was too weak to enable placement of the substrate in the model. Additional efforts to cocrystallize and analyze WT CYP7A1 with a substrate failed to produce interpretable electron density for the cholesterol molecule.
An analysis of the closed conformation shows the location and orientation of residues lining the active site cavity and those residues that are likely to interact with the substrate. Structure-based sequence comparison of 7 ␣hydroxylases reveals two polar residues, Thr104 and Ser 105 of B' helix, as potential substrate recognition residues. The B' helix is known to control substrate specifi city and generally displays considerable deviation in sequence and structural organization within the P450 family ( 28 ). The active site was not as nonpolar as expected for the binding of the hydrophobic cholesterol substrate. Therefore, we hypothesized that mutation of these active site polar residues to the corresponding nonpolar residues of CYP7B1 of methylene chloride. After vigorous mixing, aqueous and organic layers were separated by centrifugation at 3,000 rpm for 10 min. The organic layer was carefully removed and dried under argon fl ow. Methanol (100 l) was added to the resulting pellet, and steroids were analyzed on an HP1090 liquid chromatograph equipped with a Bondapak C18-column (3.9 × 300 mm) or an LC-MS system (Thermo Accela HPLC system with LCQ-Fleet MS system). The 7 ␣ -hydroxylase activity in the presence of 7KCh was measured in the reconstituted system, which includes 0-100 M of 7KCh as described previously.

RESULTS AND DISCUSSION
Overall structure of CYP7A1 CYP7A1 exhibited a typical cytochrome P450-fold with the conserved structural core of helices D, E, I, and L around the heme prosthetic group ( Fig. 1 ). The most similar structures available in the PDB, identifi ed using the DALI server, were prostacyclin synthase, CYP8A1 (PDB code 3B6H, 34% identity, 2.0 Å rmsd), and lanosterol 14-demethylase, CYP51 (PDB code 3JUS, 22% identity, 3.5 Å rmsd). Notably, CYP8A1 is a self-suffi cient P450 (i.e., requires neither molecular oxygen nor NADPH reductase to perform its distinct isomerase reaction, as opposed to the typical monooxygenase reaction). However, another member of the CYP8 family, CYP8B1, as well as CYP7A1 and CYP7B1, is involved in bile acid biosynthesis and requires a cytochrome P450 reductase as an electron donor. Despite the divergent chemistry of their respective catalyzed reactions and different substrates, CYP7A1 shared common structural features with CYP8A1 ( 23 ). Specifically, the B' helix was approximately perpendicular to the nucleus facing the heme plane at a distance of ‫ف‬ 3.5 Å ( Fig. 1C ). The substrate was positioned above the heme propionates rather than the pyrrol rings. This "pushing" of the substrate toward the B' helix is caused by W284 and N289 of I helix that protrude into the active site. As a result, the aliphatic chain of cholest-4-en-3-one was bound in the hydrophobic pocket formed by H helix, bent I helix, and the extended B-C loop; whereas the C7 carbon of the steroid B-ring was 5 Å away from the heme iron for stereospecifi c oxidation. The hydrophobic side chain of the T104L mutant in the B' helix provided additional favorable van der Waals contact with the C19 methyl group of cholest-4-en-3-one. Notably, T104L was located in the hydrophobic environment of V220 and L486, residues known to affect substrate binding could be favorable for interaction with cholesterol and thus render the protein more amenable to cocrystallization. We designed and analyzed mutant protein T104L and solved crystal structures of its complexes with cholest-4-en-3-one and 7KCh.

Substrate recognition
CYP7A1 can bind and hydroxylate cholesterol and cholest-4-en-3-one with high effi ciency as well as oxysterols ( Table 2 ) ( 29 ). The T104L mutant was similar to WT in binding and catalytic properties ( Table 2 ) and readily crystalized with substrates. The structure of the T104L mutant in complex with cholest-4-en-3-one explicitly identifi ed key residues involved in 7 ␣ -hydroxylation. Cholest-4-en-3one bound parallel to the heme with the ␣ -side of the steroid Fig. 1. CYP7A1 structure. A: Overall structure of CYP7A1 rainbow-colored from blue N terminus to the red C terminus. B: A sequence alignment and superposition of I helix in CYP7A1 (yellow), CYP19A1 (green, PDB code 3EQM), human CYP51A1 (blue, PDB code 3LD6), Mycobacterium tuberculosis CYP51 (pink, PDB code 1EA1) highlighting active site residue Asn289. C: CYP7A1 bound to cholest-4-en-3-one. Cholest-4-en-3one (orange) is surrounded by 16 residues within 4 Å from different structural elements: B' helix (H101, T104L, S105), B-C loop (I114), C helix (I125 and F129), H helix (R260), I helix (V281, W284, A285, and N289), the loop between the K helix and ␤ 1-4 strand (S360 and L361), and the loop between two ␤ 4 strands at the C terminus (G485, L486, and G487). 485, and this interaction would likely be preserved in cholesterol binding. This fi nding suggests that the presence of a hydroxyl or keto group at the C3 position is a signifi cant determinant for substrate binding, consistent with previous studies ( 31 ).
Binding of cholest-4-en-3-one was accompanied by structural changes at the entrance of the access channel, mostly at the N terminus of the protein, where transition from a loop to A' helix conformation and the movement of the loop between ␤ 1 and ␤ 2 strands result in channel opening ( Fig. 2A ). and metabolism ( 30,31 ), thus stabilizing the T104L mutation.
As noted previously, the active site volume and positioning of the substrate was controlled by bulky W284, positioned above the C6-C7 bond of cholest-4-en-3-one. The indole moiety of W284 was sandwiched between F216 (F' helix) and A288 (I helix) and interacted with D213 (F helix), which also formed a salt bridge with R483 near the protein surface. This interaction network seems to be important for the proper function, as evident from the R486C mutation of homologous CYP7B1, which is linked to neurodegenerative disease ( 32 ). In CYP7A1, the mutation W284F but not W284Y has been shown to abolish a spectral response to cholesterol ( 31 ), indicating the requirement of this residue to maintain the local network for proper positioning of the steroid ring scaffold near the heme. Residue W284 is conserved among microsomal sterol 7 ␣ -hydroxylases (CYP7B1, CYP8B1, and CYP39A1), as well as in CYP8A1, but is absent in P450s that catalyze cholesterol side chain hydroxylation (CYP11A1, CYP27A1, and CYP46A1). This fi nding is consistent with the role of the active site tryptophan as a "low ceiling" keeper ( 33 ), which is required for both monooxygenase (CYP7B1, CYP8B1, and CYP39A1) and isomerase (CYP8A1) activities.
Due to steric hindrance by W284, a steroid substrate cannot move any closer to the heme iron. Substitution of a "canonical" Thr for Asn in I helix may enable C7 regioselectivity of CYP7A1. From the multiple sequence alignment of >180 CYP7 proteins, we can deduce a new motif of L W Axxx N T in place of (A/G)Gxx T , which is present in most P450s, and a different proton transfer pathway/ mechanism is anticipated. A stable CO complex and the lack of heme movement upon substrate binding are typical features of P450s including CYP7A1, but not of CYP8A1 with P450-fold but non-P450 chemistry ( 34 ), which exhibits the highest structural homology with CYP7A1. The atypical Asn of CYP7A1 may have a slightly different function from that of CYP8A1. In CYP7A1, N289 was positioned within 3.8 Å of C6 of cholest-4-en-3-one and may assist O-O bond cleavage in later steps of the catalytic cycle. On the other hand, the corresponding Asn of CYP8A1 stabilizes stereospecifi c substrate binding by a hydrogen bond with one of two chemically equivalent endoperoxide oxygen atoms, based on analysis of an analog. Moreover, binding of the substrate to the heme iron in CYP8A1 is accompanied by heme movement ( 34 ).
Cholest-4-en-3-one adopted an elongated conformation, and its keto oxygen was located ‫ف‬ 3.4 Å away from the amide link between S360 and L361. In the 7KCh complex structure, the 3 ␤ -hydroxyl hydrogen bonded with glycine  cause opening of the substrate access channel, into which cholesterol enters with its side chain leading. Further positioning in the active site would ensure that the position of cholesterol hydroxylation approaches the heme iron. Consistent with a generally accepted model for cholesterol substrate delivery directly from the membrane, this relatively straightforward binding model is supported by mutagenesis, chemical modifi cation studies ( 28,30,38 ), and membrane penetration depth calculations (39)(40)(41). Interestingly, all available mammalian P450s structures of enzymes that hydroxylate sterol substrates (CYP11A1: PDB code 3N9Y, CYP46A1: PDB code 2Q9F, and CYP7A1: PDB code 3SN5), independent of the cholesterol hydroxylation position on the aliphatic chain or the ring system, show a conserved cholesterol orientation (side chain near the heme). The possibility of a cholesterol fl ip-fl op from the membrane into the channel similar to cholesterol diffusion within the membrane ( 42,43 ) cannot be excluded and warrants further investigation.
A unique feature of CYP7A1, compared with other cholesterol-metabolizing P450s, is a long meander region on the proximal side. This region acts as the electron transfer interface and contains the positively charged residues that appear to contribute to reductase binding. The meander is located close to the Cys-binding loop, which can modulate electronic properties of the heme and ultimately the catalytic function of CYP7A1 ( 44 ). An alternative positioning of the meander is observed in full-length yeast CYP51 in complex with lanosterol ( 45 ) and suggests the following: a ) a signifi cant fl exibility of this region that enables its relocation for capping the Cys-binding loop and contacts with the C helix, thereby dramatically changing the electron transfer interface; and b ) that it may interact with other protein effectors, which stimulate/regulate substrate binding. Modulation of the interaction interface by the longer meander might also be a feature of CYP7/8 proteins evolutionary closest to CYP51 and might be further adapted for nonmonooxygenase chemistry (e.g., selfsuffi cient CYP8A1).
In conclusion, CYP7A1 structures identify residues involved in cholest-4-en-3-one binding and specifi c interactions with the inhibitor 7KCh. Some structural features that have evolved to hydroxylate the cholesterol ring system are not unique to CYP7A1 and enable different chemistry in other CYPs. Following the general concept for cholesterol-metabolizing cytochrome P450s to acquire a substrate molecule from the membrane, we suggest that orientation of cholesterol in the different leafl ets of the Structure in complex with 7KCh, a CYP7A1-specifi c inhibitor 7KCh is a strong competitive inhibitor of CYP7A1 (IC 50 ‫ف‬ 1 µM) ( 35 ) and recently has been reported to be the product of the CYP7A1 action on 7-dehydrocholesterol in vitro and possibly in vivo ( 36,37 ). Inhibition of CYP7A1 activity, estimated in our standard reconstituted enzymatic reaction, was similar for WT CYP7A1 and the T104L mutant with K i (app) at 5.6 M and 6.8 M, respectively.
To investigate the structural basis of CYP7A1 inhibition by 7KCh, we determined the crystal structure of the complex. Crystals were obtained with the T104L mutant under conditions similar to that for the ligand-free protein (see Materials and Methods). When compared with the ligandfree structure, the presence of 7KCh did not affect the overall conformation (rmsd 0.4 Å). The 7KCh-bound structure represents a closed conformation different from the semiopen cholest-4-en-3-one-bound structure ( Fig. 2A ). These ligand-specifi c structural differences suggest that conformational dynamics play an important role in substrate recognition. A possibility of resistance to conformational changes by the crystal packing in the 7KCh-bound structure, however, is not excluded.
The structure of the active site with 7KCh was virtually the same when compared with the substrate. The same set of residues packed against the steroid core and the side chain of 7KCh. In addition, the 3 ␤ -OH group of 7KCh formed hydrogen bonds with G485 and the hydroxyl group of S360 ( Fig. 2B ). A water molecule is coordinating the heme iron, consistent with the result of the ligand binding assay, which showed no transition from low spin of the Soret band. Notably, the iron-coordinating water formed a hydrogen bond to the 7-keto group of the inhibitor and also to the Ala285 backbone carbonyl and the side chain of Asn289. This network might resemble the state during proton delivery to the heme-bound dioxygen with Asn289 serving as the key proton donor, although unproductive in the presence of the 7-keto group. Overall, the structural data are consistent with 7KCh competitive inhibition in functional assays and demonstrate that the same hydrophobic and polar interactions are utilized as for the substrate.

Structural adaptations for cholesterol hydroxylation by CYP7A1
Both ligand-bound CYP7A1 structures suggest that the steroid entered the active site with the side chain fi rst. This binding orientation might be the evolutionary adaptation to access a cholesterol molecule from the membrane. We hypothesize that CYP7A1 is embedded into the outer leaflet of the membrane at a depth of approximately half of the lipid bilayer ( ‫ف‬ 10-15 Å). In this mode, the N-terminal transmembrane helix, the F-G loop, and the A' helix (membrane binding regions) face the cholesterol molecule from the inner leafl et of the membrane with cholesterol side chain oriented toward the entry point of the access channel ( Fig. 3 ). Either certain stimuli (e.g., changing in local lipid composition) or interactions with other proteins may induce the conformational changes that lipid bilayer can be recognized for further binding by CYP7A1 alone or in complex with protein partners.
The coordinates and structure factors of the CYP7A1 structures have been deposited in the Research Collaboratory for Structural Bioinformatics PDB under the accession codes 3DAX, 3SN5 (in complex with cholest-4-en-3-one), and 3V8D (in complex with 7KCh).