Oxysterols and neurosteroids are important signaling molecules produced by monooxygenases of the cytochrome P450 family that realize their effect through nuclear receptors. CYP7B1 catalyzes the 6- or 7-hydroxylation of both steroids and oxysterols and thus is involved in the metabolism of neurosteroids and bile acid synthesis, respectively. The dual physiological role of CYP7B1 is evidenced from different diseases, liver failure and progressive neuropathy, caused by enzyme malfunction. Here we present biochemical characterization of CYP7B1 at the molecular level to understand substrate specificity and susceptibility to azole drugs. Based on our experiments with purified enzyme, the requirements for CYP7B1 hydroxylation of steroid molecules are as follows: C5 hydrogen in the α-configuration (or double bond at C5), a polar group at C17, a hydroxyl group at C3, and the absence of the hydroxyl group at C20–C24 in the C27-sterol side chain. 21-hydroxy-pregnenolone was identified as a new substrate, and overall low activity toward pregnanes could be related to the increased potency of 7-hydroxy derivatives produced by CYP7B1. Metabolic conversion (deactivation) of oxysterols by CYP7B1 in a reconstituted system proceeds via two sequential hydroxylations. Two mutations that are found in patients with diseases, Gly57Arg and Phe216Ser, result in apo-P450 (devoid of heme) protein formation. Our CYP7B1 homology model provides a rationale for understanding clinical mutations and relatively broad substrate specificity for steroid hydroxylase.
Human cytochrome P450 CYP7B1 (CYP7B1) is a microsomal enzyme that utilizes NADPH-cytochrome P450 reductase as an electron donor and catalyzes 7α-hydroxylation of oxysterols and steroids . CYP7B1 is associated with several physiological functions depending on tissue localization, including bile acid biosynthesis , metabolism of steroid hormones (including neurosteroids) , regulation of immunoglobulin production  and metabolism of estrogen and androgen receptor ligands [5-7]. The expression of human CYP7B1 is detected in the brain, testes, ovary, prostate, liver, colon, kidney and small intestine, being more abundant in the liver and brain [8, 9].
CYP7B1 catalyzes the hydroxylation of pregnenolone (PREG), dehydroepiandrosterone (DHEA) , 25-hydroxycholesterol  and 27-hydroxycholesterol  at the 7α-position. The product of 5α-androstane-3β,17β-diol metabolism has a hydroxyl group at the 6α-position . Among these steroids, PREG and DHEA are neurosteroids, a group of steroids with a broad regulatory role and whose physiological function is not fully understood [12, 13]. It was found that the primary site for 7α-hydroxylation is the hippocampus, which is also the main localization of CYP7B1 in the brain . DHEA, the most abundant neurosteroid, improves hippocampus-associated learning and memory in rodents . Interestingly, 7β-hydroxy products of PREG and DHEA are also found in the brain in large amounts. The 7β-hydroxylation reaction is presumably catalyzed by CYP7B1; however, levels of 7β-hydroxy-PREG and 7β-hydroxy-DHEA detected in the brain cannot be directly associated with CYP7B1 activity . CYP1A1 participates in 7β-hydroxylation of PREG but not DHEA in mouse brain , while other cytochromes P450 tested in this reaction showed no activity . The current hypothesis is that in humans 7β-hydroxy-DHEA is derived from 7α-hydroxy-DHEA through the 7-oxometabolite in the reaction catalyzed by 11β-hydroxysteroid dehydrogenase type 1 .
CYP7B1 is also involved in an alternative bile acid biosynthesis pathway in liver . There are three enzymes involved in 7α-hydroxylation in human bile acid synthesis: CYP7A1 (cholesterol as substrate), CYP7B1 (oxysterols as substrates)  and CYP39A1 (24-hydroxycholesterol as substrate) . CYP7B1 dysfunctions are associated with a number of genetic disorders, such as liver failure in newborns due to accumulation of toxic oxysterols, and with neuropathy in adults . Mutations in the CYP7B1 gene lead to spastic paraplegia type 5 (SPG5), an autosomal recessive disorder characterized by lower limb spasticity and weakness caused by the degeneration of motor neuron axons in the spinal cord . Currently it is unclear whether this disorder is caused solely by a malfunction of steroid or oxysterol metabolism. Low expression of CYP7B1 in dentate neurons correlates with Alzheimer's disease progression .
Here, we used the purified enzyme for the evaluation of human CYP7B1 substrate specificity, including identification of a new steroid substrate, and susceptibility to the marketed azole drugs. Our results show that CYP7B1 is strict to steroids with a hydroxyl group at the C3-position, functioning as 3-hydroxysteroid 7α-hydroxylase. Several azoles, including those used for crop protection, show undesirable binding with CYP7B1. The structure–function relationship was studied using a constructed CYP7B1 model, which gave a rationale for the effect of disease-causing mutations as well as suggesting the nature of the dual specificity to steroids and oxysterols.
Purified human CYP7B1 has a typical P450 absorption spectrum with the heme iron in a low-spin state, with a Soret region maximum at 418 nm, α- and β-bands at 567 and 535 nm, respectively, and a spectrophotometric index A418/280 = 1.4 (Fig. 1). The protein molecular mass corresponds to the expected value of 55980 kDa. A ferrous-CO complex has a maximum at 450 nm (Fig. S1), indicating that the reduced substrate-free form is easily formed and stable (the enzyme preparation contained no more than 10% of inactive P420 form in the CO spectra after 6 months of storage at −70 °C). In order to understand the effect of mutations causing SPG5 on the protein function, Gly57Arg and Phe216Ser mutants were expressed in Escherichia coli but could not be obtained as heme proteins. Purified mutants were tested for heme content by the pyridine hemochromogen method, which shows trace amounts of heme present in mutant protein samples compared with the wild type-protein (Fig. S2). These results indicate an inability of mutants to properly bind the heme prosthetic group and therefore they cannot be considered as functional enzymes.
Purified wild-type CYP7B1 was used to evaluate enzymatic activity with oxysterols and steroids in the reconstituted system. The interaction with steroids was monitored based on a type I spectral response, i.e. a blue shift of the Soret band to 393 nm (Fig. 2). The requirements for the sterol side chain were tested with different derivatives, including 20S-, 22S- and 22R-, 24S-, 25-, 27-hydroxycholesterols. Only 25- and 27-hydroxycholesterols (Table 1) induce spectral changes and have similar Kd values, confirming the specificity of CYP7B1 in vivo only for these oxysterols. Activity in the reconstituted system with these oxysterols showed the same product profile (Table 1, Fig. 3 and Fig. S3). The major product of 25-hydroxycholesterol conversion was identified as 7α,25-dihydroxycholesterol by MS, which showed the base ion at m/z = 383.40 (MH+-3H2O), and by NMR analysis (Table S1). The minor product was a dihydroxy derivative of 25-hydroxycholesterol based on mass spectra, which showed the base ion at m/z =381 (MH+-4H2O). The formation of a minor product with 25-hydroxycholesterol is consistent with previous observations . The position of the extra hydroxyl group was not identified due to the small quantity of this compound.
Table 1. Spectrophotometric titration and catalytic activity of CYP7B1 with steroids.
CYP7B1 is also active with 3β-C19 and 3β-C21 steroids, including DHEA and PREG . PREG, a C21 steroid, has ~ 10 times lower affinity than 25- and 27-hydroxycholesterols (Table 1), suggesting that strong nonpolar contacts between the cholesterol side chain and hydrophobic active site residues with possible anchoring of the hydroxyl group by a hydrogen bond play a significant role in enzyme specificity. The relatively low catalytic efficiency with PREG is consistent with the results of spectrophotometric titrations (Table 1). Among other tested pregnanes, only 21-hydroxy-PREG binds and can be metabolized by CYP7B1 (Tables 1 and S2). Activity with 21-hydroxy-PREG is significantly lower than that with PREG. Microsomal cytochrome b5, known as an activator of some reactions catalyzed by microsomal cytochrome P450-dependent enzymes , did not change the reaction rates when included in the reconstituted system (data not shown). Human CYP7B1, in contrast to the mouse enzyme , does not bind and metabolize 17α/β-OH-estradiol. CYP7B1 shows similar activity with different C19 steroids, including 5-androsten-3β-ol-17-one (DHEA), 5α-androstane-3β,17β-diol and 5-androsten-3β,17β-diol (Table 1). In our reconstituted system, only one monohydroxylated product was detected with these steroids. However, with 5α-androstane-3β-ol-17-one (EpiA, epiandrosterone) two monohydroxylated products, P1 and P2, were detected (Fig. S4), with retention times of 9.3 and 9.8 min and reaction rates of 1.5 and 5.1 min−1, respectively. Based on the metabolism of 5α-androstane-3β,17β-diol, which produces a 6α-hydroxy derivative , a second site of EpiA hydroxylation more likely occurs at the C6α-position. Notably, purified CYP7B1 does not bind 5β-androstanes (Tables 1 and S2). Taken together, CYP7B1 substrate requirements for androstanes are: C5 hydrogen in the α-configuration (or double bond at C5), a polar group at C17 and a hydroxyl group at C3.
Azole drugs are known inhibitors of cytochrome P450 enzymes. Although the antifungal azole drug target is a sterol 14α-demethylase (CYP51) involved in the biosynthesis of ergosterol, cross-inhibition of human CYPs, especially those involved in steroid synthesis and metabolism, can lead to hormone imbalance . It should be noted that several azole drugs can cross the blood–brain barrier and in this regard brain-specific P450s have the potential for undesired interactions with these drugs. Azole binding to CYP7B1 was monitored based on a type II spectral response, i.e. a red shift of the Soret band to 422–425 nm, and the observed Kd values are presented in Table 2. Various imidazole- and triazole-based drugs bind tightly to CYP7B1. The highest affinities were observed for the industrial pesticides tebuconazole (0.11 μm), propiconazole (0.13 μm) and the antifungal drugs tioconazole (0.15 μm) and miconazole (0.23 μm), whereas fluconazole showed weak binding (26.0 μm). Voriconazole and metyrapone (non-azole compound) also interact with CYP7B1 (Table 2). An in vitro inhibition assay showed that in the reconstituted system containing 60 μm DHEA as substrate and 30 μm voriconazole or metyrapone as inhibitor, CYP7B1 retains only 30% and 15% of initial activity, respectively. Azoles that do not bind to the CYP7B1 are listed in Table S3.
Table 2. Azole and other active site ligands binding to CYP7B1.
To understand the structural aspects of CYP7B1 substrate specificity we modeled the three-dimensional structure based on the crystal structures of CYP7A1 (PDB IDs: 3DAX and 3SN5). Previously published models [25, 26] did not consider common requirements in the P450 enzyme family in that the position of substrate oxidation should be located within a certain distance (~ 4–5Å) of the heme iron. In contrast, we used steering molecular dynamics (SMD) and molecular dynamics (MD) structure refinement in physical potential. Two models built on different templates with the same substrate docked showed RMSDCα = 4.9 Å for steroids and RMSDCα = 5.1 Å for 25- and 27-hydroxycholesterols. The differences are observed in the F-G loop region, and also at the N-terminus of the molecule (A′ helix and β1-sheet). The latter is consistent with structural changes observed upon ligand binding in CYP7A1 crystal structures. In both models, movements of amino acid side chains contacting substrate also occur.
An SMD study with the CYP7B1 model constructed based on the CYP7A1 structure (PDB ID: 3DAX) showed that C27 sterols may occupy two different positions: one parallel to the heme plane and one characterized by a dihedral angle value over 60°, a so-called ‘vertical’ orientation (Fig. 4A). The difference between the calculated energy of these two binding modes (parallel and vertical) was not statistically significant. We cannot exclude the possibility that this alternative vertcal orientation might be transient. However, during SMD we did not observe further rearrangements probably because of the high energy barriers surrounding this local minimum of the phase space. It should be mentioned that the ‘vertical’ orientation did not emerge during docking using ligand-free protein conformation, but it can still provide access to C7 from the α-face of the sterol molecule with the distance to the heme iron not exceeding 6 Å; therefore this mode of binding was considered. In this orientation the OH group of 25- and 27-hydroxycholesterols can form a hydrogen bond with R242 of the G helix (Fig. 4A). R242 is invariant in CYP7A1 but varies from R to Q among species in CYP7B1.
Docking of different substrates into the CYP7B1 homology model constructed based on the CYP7A1 structure (PDB ID: 3SN5) gave only one orientation of all substrates: a steroid skeleton parallel to the heme plane (Fig. 4B). Such positioning is consistent with the orientation of the steroid in the crystal structure of the substrate-bound CYP7A1 and enables contacts with active site residues L118, L119 (B′ helix), G288, W291 and A292 (I helix). CYP7B1-specific G288 (V281 in CYP7A1) (Fig. S5) creates additional space for the sterol aliphatic side chain to accommodate the C25/27-hydroxyl group in oxysterols. Our model shows that the 25- or 27-hydroxy group forms contacts with H285, while orientation of the corresponding T278 in CYP7A1 does not allow that owing to its involvement in a rigid hydrogen bond network (Fig. 4C) described below. The bulkier Y275 (S268 in CYP7A1), located in the loop between I and H helices, is also able to stabilize the OH-group at the sterol side chain terminus. Unfavorable interactions between the OH-group and hydrophobic amino acids (including L119) could explain the inability of 22- and 24-hydroxycholesterols binding, whereas C22–C24 hydroxyls of the substrate would be in a hydrophobic environment and unmodified C25–C27 carbons in polar surroundings. The shorter ethanone group of PREG (C21 steroid) has more rotational freedom, thus avoiding clashes with L119, and its carbonyl points to the CYP7B1-specific G288. C19 steroids lacking an aliphatic side chain do not cause any steric strain with L119 making them better ligands than C21 steroids. In this case, the additional water molecule could play the role of mediator of polar interactions in the vicinity of G288.
The absence of catalytic activity of CYP7B1 toward 5β-androstanes (Tables 1 and S2) is explained by the active site architecture, which cannot accommodate a bent conformation of the steroid skeleton due to the cis conformation of the A/B-ring junction. This conformation would lead to substantial changes in the positioning of the 3-hydroxyl group, which is a critical requirement for the substrate.
All substrates used for docking (25- and 27-hydroxycholesterol, PREG and DHEA) are in close proximity to N296, which appears to play a crucial role in catalysis similar to the conserved threonine in other P450s . It is of interest to compare the hydrogen bonding network at the N-terminus of the I helix, which is kinked to accommodate the cholesterol side chain in the CYP7A1 structure (Fig. 4C). Hydrophobic residues L271 and I282 of CYP7B1 cannot participate in hydrogen bonding in contrast to corresponding residues N264 and K275 of CYP7A1, thus decreasing the number of hydrogen bonds between the I and H helices. In such surroundings, CYP7B1-specific H285 is poised to interact with the substrates. Altogether these structural features provide specificity of CYP7B1 to both oxysterols and steroids.
Purified recombinant human CYP7B1 hydroxylates neurosteroids (DHEA, PREG), oxysterols (25- and 27-hydroxycholesterol), EpiA and 5α-androstane-3β,17β-diol as substrates in our experiments, which is consistent with previously published data . Notably, in the reconstituted system CYP7B1 performs a second hydroxylation step with both 25- and 27-hydroxycholesterol substrates. The first hydroxylation leads to the formation of 7α-hydroxy derivatives, metabolites of the acidic bile acid pathway of oxysterol elimination from the body . 7α,25-dihydroxycholesterol has recently been identified as a potent and selective activator of the G-protein-coupled receptor EBI2 (GPR183) [28, 29]. The physiological implication of produced tetraols is not clear; conversion of oxysterols into more polar metabolites could be a way to deactivate them. We found that CYP7B1 could also metabolize 21-hydroxy-PREG, a steroid detected in human adrenals and plasma and recently shown as a competitive inhibitor of CYP17 . Both the production of 21-hydroxy-PREG (increased in the case of 21-hydroxylase deficiency)  and its interference with androgen synthesis (neurosteroid DHEA) suggest an additional role of CYP7B1 in bioactivation or inactivation of this steroid metabolite.
The characteristic function of CYP7B1, in contrast to cholesterol 7α-hydroxylase (CYP7A1), is in the metabolism of neurosteroids. Neurosteroids are synthesized de novo in the central nervous system from cholesterol [32, 33] by steroidogenic cytochrome P450s and steroid dehydrogenases, i.e. cholesterol side chain cleavage enzyme (P450scc, CYP11A1), oxysterol 7α-hydroxylase (CYP7B1), 3β-hydroxysteroid dehydrogenase/Δ5,Δ4-isomerase (3β-HSD), steroid 17α-hydroxylase/17,20-lyase (CYP17) and 17β-HSD. PREG and DHEA are well known for their neuroprotective properties and ability to enhance memory . The product of PREG conversion by CYP7B1, 7α-hydroxy-PREG, is also classified as a neurosteroid . Its physiological function is more clearly understood in birds, newts  and rats  than in humans. The 7α-hydroxy-PREG mediates melatonin action on diurnal locomotor rhythms in quail , acts as a neuronal activator to stimulate locomotor activity via the dopaminergic system in breeding newts [39, 40] and improves spatial memory retention in cognitively impaired aged rats . These studies show that 7-hydroxy-PREG is a more potent neurosteroid than PREG, thus indicating an activating function of CYP7B1, which is consistent with the low activity of CYP7B1 toward this steroid (Table 1). Another CYP7B1 product, 7α-hydroxy-EpiA, has been reported to reduce ischemia-induced neuronal damage . However, the nature of such neuroprotective properties is unknown.
The role of the 7α-hydroxylated product of DHEA is less studied but has been suggested to have immunomodulatory and antiglucocorticoid effects similar to DHEA . Interestingly, both products of CYP7B1, 7α-hydroxy-PREG and 7α-hydroxyDHEA, may exert antiglucocorticoid effects in target tissues by competing with 11-ketoglucocorticoids for access to 11β-HSD1, thus attenuating regeneration of active glucocorticoids . The 11β-HSD1 has also been shown to carry out the interconversion of 7α-hydroxy-DHEA into 7β-hydroxy-DHEA, at least in vitro . The same localization of 11β-HSD1 and CYP7B1 in the hippocampus, on one hand, and the rotational symmetry between the 7α- and 11β-positions of the steroid nucleus, on the other hand, suggest a functional overlap between glucocorticoid and neurosteroid metabolism.
An important physiological role of CYP7B1 in steroid metabolism is evident from the mutations in the CYP7B1 gene which cause liver failure in newborns  and/or motor neuron degenerative disease, SPG5A (Table S4) . Our CYP7B1 structural model allows rationalizing the effect of amino acid substitutions leading to the disease. The R417H mutation is located in a highly conserved ERR triad, which acts as a folding motif and could result in inactive protein. Our experimental data show that both Gly57Arg and Phe216Ser mutations also lead to the formation of misfolded protein, indicating that substitution of Gly57Arg in close proximity to the N-terminal proline-rich region is critical for the proper assembly of the protein, while the aromatic ring of Phe216 is crucial for maintaining the I helix kink, which is important for catalysis (Fig. 5). The phenotype of some mutations is similar to the effect of pharmaceutical drugs. Several drugs which effectively bind and inhibit CYP7B1 (Table 2) show side effects on the nervous system associated with prolonged therapy. For example, most common side effects associated with voriconazole include dizziness, transient visual disturbances, nausea, headache, tremor, confused mental state, anxiety and depression [45, 46]. Metyrapone was found to reduce the recollection of emotional memories in healthy volunteers . Both voriconazole and metyrapone are effective inhibitors of CYP7B1 and are able to cross the blood–brain barrier. Thus, CYP7B1 acting on important endogenous substrates in the brain should be taken into consideration during the development of new drugs and in a combinatorial drug therapy.
A DNA encoding human CYP7B1 without the N-terminal membrane segment was synthesized with codon optimization for E. coli expression (Codon Device) and subcloned into the pCW-LIC vector. The resulting plasmid vector pCW-LIC_CYP7B1 was used for protein expression and as a template for construction of Gly57Arg and Phe216Ser mutant forms. CYP7B1 was expressed in the E. coli DH5α strain (Stratagene, La Jolla, CA, USA). Expression of the wild-type and the mutants was induced by the addition of 0.5 mm isopropyl thio-β-d galactoside and 0.4 mm δ-aminolevulinic acid as a precursor for heme biosynthesis. Cells were grown at 25 °C for 24 h after induction. Expression of mutant forms Gly57Arg and Phe216Ser was additionally carried out in the presence of different ligands (DHEA, 25-hydroxycholesterol, 5α-androstane-3β-ol-17-one, 5α-androstane-3β-ol, miconazole or tebuconazole) in TB medium (final concentration 20 μm) in an attempt to improve their stability. Harvested cells were resuspended in 50 mm Tris/HCl buffer, pH 7.5, containing 20% glycerol.
Cytochrome P450 and heme content determination
The concentration of CYP7B1 was determined spectrophotometrically using a molar extinction coefficient of 91 mm−1·cm−1 . Determination of the heme content of the CYP7B1 preparations was measured by the pyridine hemochromogen method. A sample of CYP7B1 was diluted with 0.2 m NaOH followed by pyridine (final concentration 20% pyridine). After reduction with sodium dithionite, the absorbance at 557 nm minus 575 nm was measured and the concentration of heme was calculated using an extinction coefficient of 32.4 mm−1·cm−1 [48, 49].
Cells were disrupted by passing through an Emulsiflex C5 homogenizer (Avestin, Canada) and solubilized by CHAPS (final concentration 1%). After centrifugation of the lysate at 35 000 g, the supernatant was loaded onto an Ni2+–nitrilotriacetic acid Sepharose 6B column (1.5 × 10 cm). The column was washed with 300 mL of 50 mm Tris/HCl buffer, pH 7.5, containing 300 mm NaCl, 0.1% CHAPS, 20% glycerol and 5 mm imidazole. Protein was eluted with 50 mm Tris/HCl buffer, pH 7.5, 300 mm NaCl, 0.1% CHAPS, 20% glycerol, 200 mm imidazole. The resulting fractions were analyzed by SDS/PAGE (10%) and by recording CO difference spectra of reduced cytochrome P450 and absorption spectra to estimate the purity by heme/protein (418/280) ratio. Homogeneous fractions were combined and applied to the hydroxyapatite column (1.5 × 4 cm). The column was washed with 300 mL of 50 mm Tris/HCl buffer pH 7.5, 300 mm NaCl, 0.1% CHAPS, 20% glycerol. Protein was eluted from the column with 500 mm potassium phosphate buffer, pH 7.5, containing 0.1% CHAPS, 20% glycerol. The purified protein was stored at −70 °C. In an attempt to stabilize mutants during purification different ligands (DHEA, 25-hydroxycholesterol, 5α-androstane-3β-ol-17-one, 5α-androstane-3β-ol, miconazole or tebuconazole) were added to all buffers (final concentration 20 μm).
Ligand-induced spectral changes were analyzed using spectrophotometric titration as a shift of the heme Soret peak in 50 mm sodium phosphate buffer (pH 7.5) with CYP7B1 concentration 1 μm. Ligand solution was added to the experimental cuvette and an equal volume of the solvent (ethanol) was added to the control cuvette. Type I (blue shift from 418 nm) indicates displacement of distal water from the heme iron coordination and type II (shift to 425 nm) indicates coordination by azole nitrogen. For determination of the dissociation constant of the enzyme–ligand complex (Kd), an equation for tight binding was used. Titration data were approximated using the Levenberg–Marquart algorithm:
where A is the amplitude of the spectral change at ligand concentration [L], Amax is the amplitude of the spectral change at [L] ligand at ligand saturation, [L]t is the total ligand concentration and [R]0 is the total protein concentration.
Determination of the catalytic activity of recombinant protein
The steroid hydroxylase activity of human CYP7B1 was measured in the reconstituted system at 37 °C in 100 mm sodium phosphate buffer, pH 7.4, containing 10 mm MgCl2 and 0.1% sodium cholate. Aliquots of concentrated recombinant proteins were mixed and preincubated for 5 min at room temperature. Substrate (10 mm stock solution in ethanol) was added to the reaction mixture at a final concentration of 50 μm. Inhibitors (dissolved in ethanol) were added to a final concentration of 0–10 μm. The final reaction mixture (0.5 mL) contained 0.5 μm P450, 1.0 μm rat recombinant NADPH-cytochrome P450 reductase, 8 mm sodium isocitrate, 10 U·mL−1 isocitrate dehydrogenase, 200 μm NADPH. Aliquots (0.5 mL) were taken from the incubation mixture at time intervals 0, 5, 15 and 30 min and subjected to solid phase extraction using a Waters SepPackC18 1 cc column.
Products of the CYP7B1 reaction were further analyzed by adapted LC/MS methods. For the analysis, an Accela HPLC system with integrated LCQ Fleet Ion Trap Mass Spectrometer (Thermo Sci, San Jose, CA, USA) was used. Chromatography was performed by gradient elution on a Cosmosil 5C18-MS-II column (4.6 mm × 150 mm, Nacalai Tesque Inc., Japan). Aliquots of 25 μL were injected and eluted from the column at a flow rate of 500 μL·min−1. The column temperature was maintained at 30 ± 1 °C. The mobile phase consisted of water (solvent A) and methanol (solvent B).
Cholestane derivatives and hydroxylation products were analyzed by the following program: 60% (v/v) of solvent B for 4 min, linear gradient from 60% to 100% of solvent B for 5 min, 100% of solvent B for 7 min and then held at 60% of solvent B for 4 min before new injection.
Androstane derivatives and hydroxylation products were analyzed by the program: 50% (v/v) of solvent B for 3 min, linear gradient from 50% to 100% of solvent B for 4 min, 100% of solvent B for 8 min and then held at 50% of solvent B for 2 min before new injection.
Pregnane derivatives and hydroxylation products were analyzed by the program: 50% (v/v) of solvent B for 3 min, linear gradient from 50% to 100% of solvent B for 5 min, 100% of solvent B for 5 min and then held at 50% of solvent B for 2 min before new injection.
The inlet capillary voltage was fixed at 18 V and its temperature was maintained at 275 °C. The vaporizer temperature was set to 350 °C. The tube lens off-set was set at 80 V. Effluent from the HPLC column (Cosmosil 5C18-MS-II from Waters, 4.6 × 150 mm) was nebulized with N2 as sheath, auxiliary and sweep gas with flow rates of 50, 5, 5 arbitrary units, respectively.
MS experiments were performed with atmospheric pressure chemical ionization in positive ion mode. Quantitative detection was performed with progesterone as internal standard (most abundant peak m/z 315). Activity quantification was done with the most abundant peak for the substrate (385 m/z for 25- and 27-hydroxycholesterols, 299 m/z for PREG, 273 m/z for 5-androsten-3β-17β-diol, 315 m/z for 21-hydroxy-PREG, 271 m/z for 3β-hydroxy-5-androsten-17-one (DHEA), 271 m/z for 5α-androstane-3β-ol-17-one). Activity was quantified by substrate consumption.
Purification of 25-dihydroxycholesterol hydroxylation product from the reaction mixture
The reaction was carried out as described above (reaction mixture was scaled up to 20 mL). 200 μL of 5 m sodium hydroxide was added to the reaction mixture to stop the reaction. The product was purified on a Waters SepPack C18 3 cc column and eluted with 3 mL of 100% methanol. The solution was evaporated under argon flow and dissolved in CDCl3 for NMR analysis.
1H NMR (500.03 MHz) spectra were recorded as CDCl3 and CD3OD solutions using the residual signal of the solvent (δ 7.26 ppm and 3.30 ppm, respectively) as an internal secondary standard on a Bruker AVANCE-500 instrument. COSY and TOCSY experiments were carried out using the standard Bruker software package. The small amount of product (~ 50 μg) did not allow collection of reliable 13C NMR data. The 1H NMR spectrum was recorded in CDCl3 and clearly shows the presence of two steroid components at an approximate ratio of 1 : 0.85 (Fig. S6). The minor steroid component was identified as starting material; the chemical shifts of the methyl groups and proton signals at C6 were identical to those for authentic 25-hydroxycholesterol. The major steroid component had characteristic chemical shifts corresponding to the B-ring protons, and a constant value for the spin–spin interaction of the proton at C6 can be ascribed to 7α,25-dihydroxycholesterol. A contribution due to the 7-hydroxyl  can be observed in the NMR spectra of several Δ5-marine sterols (topsentinols, Fig. S6E) bearing a 7α- or 7β-hydroxyl group in the structure. Based on these data, there is clear discrimination between 7α and 7β derivatives based on the chemical shifts and proton signal pattern at the C6 and C19 methyl groups. Chemical shift values obtained in our experiments for the characteristic protons of the A and B rings are consistent with those for 7α-topsentinols (Table S1). Partial assignment of characteristic protons was performed using COSY and TOCSY spectra. Cross peaks between protons at C6 and C7 were found in the NMR spectra. The latter peak is overlapped with a broad admixture signal and cannot be detected in the 1D spectrum. However, this proton is clearly observed at 3.76 ppm when the spectrum is recorded in MeOD.
Molecular modeling and docking
Spatial full atomic models of CYP7B1 were created by homology modeling with modeller 9v7 . Two CYP7A1 structures (PDB IDs: 3DAX and 3SN5) were used as separate templates for two models. Stability of the models was examined with 10 ns MD at constant temperature 300 K in the amber 10 software package (step of integration 2 fs). Water was modeled by means of generalized Born continuum solvation. The structures with the lowest potential energy were used for docking of the substrates.
CYP7A1 structure 3DAX represents the ‘closed’ ligand-free conformation and was used as a template for homology modeling of the closed structure for CYP7B1. The quality of the model was checked by procheck . In the model 99.5% of amino acids checked on their distribution on the Ramachandran plot are in the allowed region. The model does not contain d-conformers of amino acid residues. To dock the substrate we applied steered MD. An extra attracting force was used to speed up substrate arrangement in the active site. The force value was calculated according to the position of the center of mass for the substrate molecule. Boundaries of a harmonic potential well were moved gradually during 100 ps in order to bring the substrate molecule closer to the reactive center of the enzyme. As soon as an 8 Å distance between the center of mass and the iron atom was achieved, the extra force was removed and the structure underwent MD refinement during 1 ns moving freely near the heme moiety.
A procedure of docking to the active site of the ‘open’ structure was performed using CYP7A1 structure 3SN5. Unlike the 3DAX structure, 3SN5 accommodates a ligand in the active site and has an enlarged cavity near the heme. This feature was inherited by the model that facilitates docking. Docking was performed in the dock v6.5 program based on a physical potential. Molecular graphics images were produced using the UCSF chimera package from the Resource for Biocomputing, Visualization and Informatics . RMSD was measured in the vmd package . Structural alignment was made with the matchmaker function in UCSF chimera .
We thank T.S. Cherkesova and I.P. Grabovec for technical assistance. The Structural Genomics Consortium is a registered charity (No. 1097737) and receives funds from the Canadian Institutes for Health Research, the Canadian Foundation for Innovation, Genome Canada through the Ontario Genomics Institute, GlaxoSmithKline, Karolinska Institutet, the Knut and Alice Wallenberg Foundation, the Ontario Innovation Trust, the Ontario Ministry for Research and Innovation, Merck and Co. Inc., the Novartis Research Foundation, the Swedish Agency for Innovation Systems, the Swedish Foundation for Strategic Research, and the Welcome Trust.