| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Biochemistry, Queen's University, Kingston, Ontario, Canada
Correspondence: Address reprint requests to Dr. Glenville Jones, Queen's University, Dept. of Biochemistry, Botterell Hall, Room 277, Kingston, Ontario, Canada K7L 3N6. Tel: 613-533-2498; Fax: 613-533-2987; E-mail: gj1{at}post.queensu.ca.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
-hydroxyvitamin D3 to favor 25- or 27-hydroxylation. The results suggest that conserved hydrophobic residues in the ß-5 hairpin help define the shape of the substrate binding cavity and that this structure interacts with Phe-248 in the F-helix. Mutations directed toward the ß-3a strand suggested a possible heme-binding interaction centered on Asn-403 and a structural role for substrate contact residues Thr-402 and Ser-404. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
-hydroxylase (CYP27B1) into the active hormone, 1,25-(OH)2D3. Subsequent catabolism in vitamin D target tissues by the 1,25-(OH)2D3-inducible 24-hydroxylase (CYP24A1) produces the biliary excretory product calcitroic acid (1
). The active hormone and its precursors are transported between tissues complexed to vitamin D-binding protein. In the nucleus of cells in vitamin D target tissues, 1,25-(OH)2D3 binds vitamin D receptor to transactivate, the expression of specific gene products affecting calcium homeostasis and the proliferation and differentiation of certain specialized cell types (2). Thus the structure of the vitamin D molecule helps define crucial interactions with at least five proteins, thereby influencing transport, activation, biological action, catabolism, pharmacokinetics, and the precise sequence of these events (3). However, research into the design of vitamin D prodrugs has shown that the activation sequence can be manipulated, and that a nonphysiological, 1-hydroxylated vitamin D analog can be 25-hydroxylated and activated in a single cytochrome-mediated step (4).
The circulating level of 25-OH-D3 is a measure of vitamin D status (5), but the CYP enzymes regulating its formation are still not completely defined. The best studied 25-hydroxylase, CYP27A1, is well conserved across species and is a bifunctional enzyme that also expresses 27-hydroxylase activity toward cholesterol and related sterols in two separate bile acid biosynthetic pathways (6). Despite being the affected target in cerebrotendinous xanthomatosis (CTX) (7,8), a lipid storage disorder characterized by the accumulation of abnormal metabolites of bile acid precursors (9), the only indication of vitamin D deficiency associated with the loss of CYP27A1 is an increased risk of osteoporosis and bone fractures and in low 25-OH-D3 levels that fail to normalize with bile-acid replacement therapy (10). This suggests a redundant or compensatory 25-hydroxylase activity that is in agreement with one of the earliest studies of vitamin D 25-hydroxylation, which inferred the existence of a high specificity, low capacity (microsomal) enzyme and a low specificity, high capacity (mitochondrial) enzyme (11). Two of the candidates for the human microsomal 25-hydroxylase include CYP3A4, a hepatic drug metabolizing enzyme with broad substrate specificity (12,13), and CYP2R1, recently shown to have activity (14) and to be associated with a mild form of rickets (15). A similar microsomal 25-hydroxylase redundancy is seen in the rat enzymes CYP2J3 (16), CYP2R1, and CYP2C11 (17), the pig CYP2D25 (18), and the mouse CYP2R1 (19).
Since there are no crystal structures of mitochondrial cytochrome P450s, any model for CYP27A1 must be based upon one or more existing prokaryotic or microsomal CYP crystal structures. Although an important model of bovine mitochondrial CYP11A1 (side-chain cleavage enzyme, P450scc) (20) is available in the Protein Data Bank, there is disagreement concerning the existence of an ad hoc surface loop between the E- and F-helices hypothesized to mediate membrane binding in it since subsequent models of bovine CYP11A1 (21), human CYP11A1 (22), human CYP11B1 and CYP11B2 (23), and all cytochrome P450 crystal structures, including 14-sterol demethylase (CYP51) from Mycobacterium tuberculosis (24,25) and the mammalian enzymes CYP2B4 (26), CYP2C5 (27,28), CYP2C8 (29), CYP2C9 (30), and CYP3A4 (31,32) show no evidence of this surface loop. Thus the quality of a previous model for CYP27A1 based on the original bovine model (7) remains in doubt.
It is generally recognized that the multisequence alignment is a basic prerequisite for homology modeling. In the case of the cytochrome P450 superfamily, obtaining a comprehensive alignment is difficult because sequence homologies between families are typically 10–20%. We have examined many P450 crystal structures, including a variety of mammalian cytochromes from related families, and used the observed secondary structure to refine the multisequence alignment process. This simplified the assignment of secondary structure in CYP27A1 to a degree that appears superior to theoretical prediction. On further examination of the crystal structures, we observed an extensive conservation of hydrogen bonding patterns and structural motifs acting to stabilize the folded structure. Consequently, this helped to align less conserved residues that merely contribute to local structure through backbone interactions. The cumulative effect of well-defined structural elements, extensive ß-sheets and conserved helices enabled an unprecedented refinement of the multisequence alignment.
We have used this systematic analysis of P450 crystal structures to homology model CYP27A1 and to investigate substrate contact residues in the binding cavity. Using structure-guided, site-directed mutagenesis, we report the effects of mutations in the F-helix, ß-3 sheet, and ß-5 hairpin on the hydroxylation of the vitamin D prodrug 1-OH-D3. We have also used the model to examine substrate recognition, the location of access channels in CYP27A1, and the structure-function relationships associated with the polymorphisms giving rise to CTX (CYP27A1) and VDDR-I (in CYP27B1) (1).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
) and the alignment edited manually using GeneDoc (34), according to observed secondary structure. Structural coordinates retrieved from the Brookhaven Protein Data Bank listed in Table 1 were aligned to the -carbons of the heme-thiolate cysteine and I-helix of P450cam using the LSQKAB structural alignment utility (CCP4, Collaborative Computational Project, No. 4 Suite: Programs for Protein Crystallography, University of York, Heslington, UK). The aligned coordinates of P450bm3, P450cam, P450eryF, and P450terp were superimposed to give a cage-like consensus framework into which the human CYP27A1 sequence was threaded using Turbo Frodo (Architecture et Fonction des Macromolécules Biologiques-Centre National de la Recherche Scientifique, Marseille, France). This initial structure was annealed through a series of extensively constrained geometry minimizations using Gasteiger-Marsili atom charges in SYBYL v6.8 (Molecular Simulations, San Diego, CA). Constraints applied to maintain proper hydrogen-bond lengths in the -helices, ß-sheets, structural motifs, and hydrogen-bonding networks seen in 11 crystal structures were set to 2 Å and gradually relaxed during extensive geometry optimizations. Molecular dynamics simulations were performed using Insight II v2000 (Accelrys, San Diego, CA) and RMSD calculations were made using PyMOL (DeLano Scientific, San Carlos, CA). Ramachandran parameters were calculated with PROCHECK (European Bioinformatics Institute, Cambridge, UK). Crystal structures and models were visualized for detailed study using custom script files written for Chime (Molecular Simulations) and PyMOL.
|
-hydroxyvitamin D3, cholesterol, and bile acid precursor 5ß-cholestane-3,7,12-triol were retrieved from the Cambridge Crystallographic Data Centre and modified and geometry optimized as required. The vitamin D structures were manually docked with c25 and c27 constrained at 4 Å above the heme A-ring nitrogen and cholesterol structures with c27 at 4 Å above the heme A-ring nitrogen according to the position of c5 in camphor bound to P450cam. For comparison, these structures were also docked in the binding cavity using AutoDock v3.05 (Scripps Research Institute, La Jolla, CA). Wild-type and mutant CYP27A1 structures used to assess the impact of each mutation on substrate binding were energy minimized under identical conditions (1000 iterations) from a single energy minimized structure.
Expression vectors and site-directed mutagenesis
Human CYP27A1 in pSG5 (4) was subcloned into the EcoRI-NotI site of the C-terminal tag vector pcDNA3.1 His-V5-"B" (Invitrogen, Carlsbad, CA) after introduction of these "in frame" restriction sites by amplification with Taq polymerase (Qiagen, Mississauga, Ontario, Canada) and oligonucleotide primers (Cortec DNA Service Laboratories, Kingston, Ontario, Canada) on a PerkinElmer (Wellesley, MA) Thermocycler. Restriction enzymes and other molecular biological reagents were obtained from Promega (Madison, WI) and Amersham-Pharmacia (Baie d'Urfe, Quebec, Canada). After confirmatory sequencing (Cortec DNA Service Laboratories), both vectors were used in site-directed mutagenesis reactions using a Chameleon Site-Directed Mutagenesis Kit (Stratagene, LaJolla, CA). The list of primers is as follows:
COS-1 transfection
African green monkey kidney COS-1 cells (ATCC CRL-1650) were grown on Nunc cell culture dishes using Dulbecco's modified Eagle's medium and 1x antibiotic-antimycotic (Gibco-BRL Canadian Life Technologies, Burlington, Ontario, Canada) supplemented with 10% CELLect Gold fetal bovine serum (ICN, Costa Mesa, CA) at 37°C. COS-1 cells were transiently cotransfected on 150 mm plates by the DEAE-dextran method (35) using a plasmid cocktail containing 10 µg each of CYP27A1 wild-type or mutant vector and vectors expressing bovine ferredoxin (pEAdx) (36), and ferredoxin reductase (pEAR) (37); 3 µg of a firefly luciferase expression plasmid (pRSVLuc) were included as a transfection control in each cocktail. Transfected cells were allowed to recover overnight in supplemented medium and subcultured onto triplicate 100 mm plates 1 day before CYP27A1 activity assay.
Vitamin D compounds
1-OH-D3 and 1,25-(OH)2D3 were gifts of Drs. Ernst Binderup and Anne-Marie Kissmeyer (Leo Pharmaceuticals, Ballerup, Denmark). Reference compound 1,26-(OH)2D3 was synthesized and generously provided by Dr. Martin Calverley (Leo Pharmaceuticals).
CYP27A1 activity assay
Transfected COS-1 cells were cultured to late log phase, washed with 10 mL phosphate-buffered saline, and then incubated with 10 mL incubation medium/dish containing Dulbecco's modified Eagle's medium, 1% bovine serum albumin (Boehringer-Mannheim/Roche Diagnostics, Laval, Quebec, Canada), 1x antibiotics, 100 µM N,N-diphenylethylenediamine as an antioxidant (Sigma-Aldrich, Oakville, Ontario, Canada), 40 µg (10 µM) 1-OH-D3, and 0.02% ethanol for 48 h. No-cell and dead-cell controls consisted of, respectively, an empty dish or dead cells (microwave treated) incubated as above. The incubation medium was extracted to isolate vitamin D compounds in the total lipid fraction using a modification of the method of Bligh and Dyer (38) in which chloroform was replaced by dichloromethane. The organic layer was evaporated to dryness under a stream of nitrogen, redissolved in hexane/isopropanol/methanol (91:7:2 v/v/v) and subjected to separation by HPLC (39).
High performance liquid chromatography
Straight-phase HPLC (40,41) was performed using Millennium32, an Alliance 2690 Separations Module, and a model 990 photodiode array detector (Waters Associates, Milford, MA). Vitamin D metabolites were separated on Zorbax-SIL (3 µ, 6.2 x 80 mm) columns using hexane/isopropanol/methanol 91/7/2 at a flow rate of 1 mL/min. and identified by the characteristic ultraviolet absorption spectrum of the triene chromophore (
max = 265 nm). Retention times of metabolites varied on the different columns used. Metabolite peak areas were normalized to an HPLC internal standard (2 µg 25-OH-D3) and to the cell transfection efficiency, the latter determined from luciferase activity measurements acquired using cell culture lysis buffer, single luciferase assay reagents (Promega), and a Microlumat LB96V microplate luminometer (EG&G Berthold, Bad Wildbad, Germany).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
-helices, ß-sheet, and loop secondary structure in eleven CYP crystallographic templates was used to manually parse the sequences in the alignment shown in Fig. 1. Gaps were introduced into the alignment to highlight these secondary structures. This revealed many functionally equivalent residues and conserved backbone interactions within secondary structures and tertiary structural motifs. Further refinements were made by systematically cataloging conserved hydrogen bonding patterns in the templates as shown in Figs. 2 and 3. As described below, there was extensive conservation of ß-sheet structure and hydrogen bonding networks across all templates. The refined alignment was used to assign secondary structure in seven uncharacterized CYPs, including CYP27A1, using size of structural element, sequence homology, and side-chain polarity.
|
|
|
) and in rat CYP2C11 where it may have been inadvertently disrupted experimentally due to a sequence alignment error (41). It is also the site of two P450 deficiency polymorphisms (P42S in CYP11B1 and P30L in CYP21B1). As shown in Fig. 2, the proline amide carbonyl forms a conserved hydrogen bond to the same ß-1 sheet amide nitrogen position in 7/11 templates or an equivalent interaction to the same ß-1 residue position in 2/4 of the remaining templates. This helps orient a hairpin structure between the PGP motif and the A'-helix region in 7/11 templates. Postulated to form a folding nucleus (42), the PGP motif and surrounding structure may also influence active site structure and a substrate access channel through interactions with the F-G loop and the A'-helix region. Shielding and contacting part of the substrate binding site, the sequence along the 14–15 amino acid length of the A-helix is, for a given CYP, well conserved across species. A conserved glutamine in the middle of the CYP27A1 and CYP27B1 A-helix corresponds to a VDDR-I mutation (Q65H in CYP27B1). The orientation of the A-helix depends on CYP-specific polar interactions with the ß-3 and ß-5 sheets and the presence of a tyrosine-glycine pair at the A-helix terminus. The ß-1 sheet begins at the end of the A-helix and is distinguished by a glycine at the ß-1 hairpin and the pattern of hydrogen bond conservation seen in Fig. 3. Although conserved in the smaller ß-1 sheet of the thermophile CYP119, what is especially remarkable about this pattern is how it integrates seamlessly through the ß-3 sheet and B-B' loop to two major heme-binding residues in all templates. In fact, this extended ß-sheet encompasses conserved connections through the ß-5 sheet to the D-E loop and across one side of the active site through the B-B' loop to the B'-C loop.
Incorporating the heme into a structural network
The heme-binding region is a rigid network of 20 or more hydrogen bonds and structural waters seen in Fig. 4, a–c, to connect the heme A- and D-ring propionates to the B-/B' loop, B'-C loop, C-helix, ß-3 sheet, and heme-thiolate loop. First described in P450Bm3, P450cam, and P450terp (43), the heme-binding region is remarkably conserved in all templates with functionally equivalent residues occupying an exclusive space and making an equivalent contribution. Careful maintenance of these heme-binding interactions during energy minimization stabilizes important substrate contact residues in adjacent structures.
|
). The mutation of the analogous residue in CYP51 (F105L) arose clinically in five Candida albicans isolates as fluconazole resistance (44). That structural integrity in this region is important is further illustrated by a CAH polymorphism (G424S in CYP21B1) at the –4thiolate position.
The heme iron is axially coordinated by the conserved cysteine thiol on the proximal side of the heme. This thiolate bond is stabilized by the surrounding residues, with the sulfur atom being nearly equidistant from the +1thiolate, +2thiolate, and +3thiolate amide nitrogens, and the tight turn at +2thiolate into the L-helix probably accounts for the near absolute conservation of glycine at this position. As shown in Fig. 4 a, the heme-thiolate loop is stabilized by four conserved hydrogen bonds: –1thiolate amide nitrogen to –4thiolate amide carbonyl; –6thiolate amide nitrogen to –9thiolate amide carbonyl; cysteine amide carbonyl to +3thiolate amide nitrogen; and cysteine amide nitrogen to –7thiolate amide carbonyl, the latter being a highly conserved phenylalanine influencing heme reduction potential (45). The heme-thiolate loop is further constrained by hydrogen bonds between a C-helix arginine (Arg-158 in CYP27A1) and the –1thiolate amide carbonyl in 9/11 templates (Fig. 4 c). This C-helix arginine also hydrogen bonds with the D-ring propionate, but the main C-helix to D-ring propionate link involves tryptophan one turn earlier (Trp-154 in CYP27A1) in the mitochondrial CYPs and 6/11 templates. In the remaining five templates, an equivalent polar side chain (arginine, glutamine, or histidine) is hydrogen bonded to the D-ring propionate, and in three of these it is also hydrogen bonded to –3thiolate amide carbonyl.
Another hydrogen bond to the D-ring propionate is made by the –2thiolate arginine (or histidine in CYP119, CYP51, P450cam, P450eryF, and P450terp) and mutation of this side-chain disrupted heme incorporation and enzyme activity in CTX (R474Q, R474W), VDDR-I (R453C), and P450 deficiency polymorphisms (R448H in CYP11B1, R440H in CYP17A1, and R426H in CYP21B1). The –2thiolate arginine also hydrogen bonds to amide carbonyls of key substrate contact residues in the B'-C loop and to the side chain at the start of the C-helix (Glu-150 in CYP27A1; Fig. 4 c). The latter is the site of a P450 deficiency polymorphism (N133H in CYP11B1) and the aspartate, glutamate, or asparagine side chain found there is bonded to the backbone of the B-B' loop at precisely the location of a conserved arginine (asparagine, lysine, or serine in the prokaryotic templates). This B-B' loop arginine hydrogen bonds to the A- and D-ring propionates, a fact that was not fully appreciated until CYP2C5 and other eukaryotic structures became available. B-B' loop arginine is the site of CTX (R127Q and R127W), VDDR-I (R107H), and CAH (R96W in CYP17A1) polymorphisms. In all other templates, the hydrogen bonding of the aligned residue (asparagine, lysine, or serine) to the A-ring propionate is mediated through the nearby ß-3 "heme-binding" arginine. The B-B' loop is connected to the +1ß3Arg by two ß-sheet bonds in all templates. Polymorphisms affecting ß-3 arginine cause CTX (R405Q, R405W), VDDR-I (R389C, R389G, R389H), and P450 deficiency (R384P in CYP11B2) because they disrupt a major structural link (usually a bidentate salt bridge) to the A-ring propionate. Conserved in the mitochondrial, bacterial (except P450bm3), and CYP3A4 enzymes, ß-3 arginine is replaced by proline + histidine in the CYP2 templates. In many microsomal CYPs, this histidine binds heme by bridging the A-ring propionate to the B-B' loop. In all enzymes with ß-3 arginine though, the side chain is hydrogen bonded to both –1ß3Arg amide carbonyl and via structural water to the penultimate B-helix carbonyl.
Other heme-binding interactions in the network are directed toward the A-ring propionate. Hydrogen bonds from the –1ß3Arg amide nitrogen are seen in CYP51 (Met-325 mediated by Tyr-76), P450bm3 (Ser-332 via water), P450eryF (Thr-292 via water), and CYP3A4 (Glu-374 via water), through the –1ß3Arg amide carbonyl in P450bm3 (Ser-332 via water) and CYP2 templates (proline amide carbonyl mediated through B-B' loop arginine). In the other templates, the –2ß3Arg side chain hydrogen bonds to the A-ring propionate, and this interaction is further stabilized by a side chain in the second ß-3 strand. In CYP2C5, this corresponds to Asn-362 and +21ß3Arg side chain (Thr-386), in P450cam as Asp-297 and +23ß3Arg (Gln-322), and in P450eryF +23ß3Arg (Asn-316). This –2ß3Arg and +23ß3Arg side-chain interaction with the A-ring propionate is especially relevant since the corresponding residues, Asn-403 and His-428 in CYP27A1, are completely conserved across species. In P450terp, the +23ß3Arg side chain (Tyr-342) hydrogen bonds to the A-ring propionate directly. In all templates, another hydrogen bond to the A-ring propionate is made from the heme-thiolate loop by –5thiolate amide nitrogen. This is mediated by a structural water in all templates except P450terp (which bonds directly) and the CYP2 templates (mediated by the –6thiolate serine side chain).
The rigid structure of the heme-binding region is continuous with the B'-C loop in all templates (except CYP51, which is atypical), and this helps to position a critical substrate contact residue (Phe-147 in CYP27A1) above the D-ring propionate. Irrespective of B'-helix variability, the B'-C loop assumes a well-defined structure in the vicinity of a conserved glycine in 6/11 templates and at similarly aligned residues in other CYPs. The structural importance of the B'-C loop glycine in orienting the substrate contact residue is revealed in the existence of CTX (G145E) and VDDR-I (G125E) polymorphisms and in a study of inactivating site-directed mutants: G145A and G125A/V, respectively (46). The B'-C loop is internally stabilized through hydrogen bonding of the glycine amide carbonyl to +4B'Cglycine amide nitrogen in P450bm3, CYP2B4 and CYP3A4, to +4B'Cglycine side chain in the CYP2C templates, and to +3B'Cglycine amide nitrogens in 4/6 prokaryotic templates. In the same prokaryotic templates, stability is further enhanced by a +1B'Cglycine amide carbonyl to +4B'Cglycine amide nitrogen interaction. Spatially, the B'-C loop is oriented through conserved hydrogen bonds with adjacent structures. This includes B-B' loop arginine (or equivalent) to +2B'Cglycine amide carbonyl (9/11 templates), –2thiolate arginine (or histidine) to +1B'Cglycine and +4B'Cglycine (10/11 templates), and an I-helix aspartate in CYP2C templates hydrogen bonded to +1B'Cglycine and +2B'Cglycine amide nitrogens. The latter I-helix interaction was deduced in CYP2D6 (47), and a functionally equivalent link between the I-helix and B'-C loop is seen in CYP3A4 and at least 3/6 prokaryotic templates. In fact, the alignment of a conserved glutamate at this position (Glu-331 in CYP27A1) in the mitochondrial vitamin D hydroxylases and aspartate or glutamate in many other mammalian CYPs (e.g., CYP2C11, CYP2J3, CYP2R1, CYP11, CYP17, and CYP21) further emphasizes the importance of a structural link between the I-helix and the B'-C loop.
Building the EH-gap motif, E-F loop, and a putative disulfide
Adjacent to the C-D loop is a highly structured region encompassing the end of the E-helix and the start of the H-helix. As shown in Fig. 4 d, two hydrogen bonds bridge the gap to join aligned residues and create the effect of a continuous helix in 10/11 templates. Centered on the interaction of a conserved aspartate at the start of the H-helix with a basic side chain four residues from the end of the G-helix, this EH-gap motif established these two residues as important landmarks in the alignment and positions the end of the G-helix. The EH-gap also contributed to the presence of a small cavity in the model similar to one recognized as a xenon binding site in P450cam (48,49).
In the CYP2 templates, where the E-F loop is long enough to parallel the D- and E-helices, hydrogen bonding residues near the middle of each helix interact with the second residue in the E-F loop (usually arginine) to further stabilize the region. Fig. 4 d shows this interaction in CYP27A1 with Asp-164 (D-helix), Arg-225 (E-F loop), and Glu-215 (E-helix). These residues are very well conserved in the vitamin D hydroxylases where the aspartate is the site of a VDDR-I mutation (D164N) and the glutamate is adjacent to a CTX mutation (A216P). Although the latter may be splicing mutant (50), it is also adjacent to Ile-217, which aligns with the most common mutation site in CYP21B1-related CAH deficiency (I172N) (51). The E-helix backbone near this Ile-217 is part of a small hydrogen bonded network in all templates that joins the I-helix oxygen binding site to a conserved L-helix glutamate (asparagine or glutamine) through a structurally conserved water molecule. Other structurally important landmarks nearby include the aspartate at the beginning of the E-helix (8/11 templates) and the conserved hydrogen bonds between the ß-5 sheet and D-E loop.
The establishment of hydrogen bonding interactions to anchor the E-F loop in the model of CYP27A1 exposed a possible disulfide bond with the E-helix involving Cys-218 and Cys-228 (Fig. 4 d). These cysteines are conserved in CYP27A1 from human, dog, pig, rat, rabbit, baboon, and Xenopus, but the latter cysteine has been lost in bovine, chicken, mouse, and fugu. In addition to CYP27A1, there may be two disulfides in CYP27B1 (human) and three in CYP24A1 (human, pig, and rat). The possibility of native disulfide bonds in cytochrome CYPs has never been addressed before despite multiple possibilities associated with the D- and E-helices in CYP2C5, CYP2C8, and CYP2C9. Interestingly, one introduced into P450cam by site-directed mutagenesis appears to have connected the B'- and G-helices to affect substrate access and egress channels (49).
Positioning the F-helix
The alignment of the F-helix benefited significantly from detailed examination of the templates. Although the F-helix begins with an aspartate in 8/11 templates, neither it nor an apparently conserved phenylalanine in 9/11 templates was spatially conserved outside of P450bm3, CYP51, and the CYP2 templates. Thus the alignment of the F-helix for CYP3A4 and the prokaryotic templates in Fig. 1 is based on a spatially conserved -carbon corresponding to CYP2 template phenylalanines. On the basis that the phenylalanine is conserved across species in the mitochondrial CYPs, the alignment of Phe-240 in CYP27A1 was used to model the height and spin orientation of the F-helix.
Linking the J-, K-, and L-helices
The alignment and modeling of the I- and J-helices was essentially unremarkable. The abrupt turn of the I-helix into the J-helix, however, is punctuated by sites of CTX (D354G) and VDDR-I (R335P) mutations. A second VDDR-I locus in the J-helix (L343F) is directed toward the K-helix. The structural disruption mediated by these mutations became apparent only after a conserved glutamate side chain (aspartate in P450eryF and P450terp) was found to be hydrogen bonded with conserved amide nitrogen(s) near the start of the K-helix in 9/11 templates. Disruption in CYP27A1 of this J- to K-helix link (involving the J-helix Glu-364 side chain and the K-helix Leu-385 and Leu-386 amide nitrogens), shown in Fig. 4 e, by an adjacent CTX mutation (P384L) probably destabilizes the J- and K-helices and J-K loop.
Two of four CAH mutations (P342T, R347H in CYP17A1 and R339H, R341W in CYP21B1) seen in the J-K loop correspond to a residue involved in a four-part hydrogen bonding network connecting the K- and L-helices in the CYP2 templates (three-part in CYP2C9). Three side chains of this four-part network in CYP2B4, CYP2C5, and CYP2C8 align with essential cationic charges in CYP17A1 necessary for lyase activity (52). Three side chains in this four-part network also align with residues that appear to provide a functionally equivalent four-part interaction in the vitamin D hydroxylases (Lys-387, Lys-391, Glu-483, and Gln-487). Overall, functionally equivalent KL-motifs are seen in 10/11 templates and appear to be very well conserved in the mitochondrial vitamin D hydroxylases and may stabilize a P450cam-type xenon binding pocket above it (49).
Building the ERR-triad
Integral to the interactions of the K-helix with surrounding secondary structures is its participation in the ERR-triad. Continuous with the ß-3a strand, the K-helix also anchors the ERR-triad to the heme-region, the substrate-binding site, and the extended ß-sheet. At the center of the ERR-triad, the conserved glutamate (Glu-392) and arginine (Arg-395) appearing as a ExxR motif in the K-helix (11/11 templates) are hydrogen bonded to an arginine (Arg-448) in the meander region of 7/11 templates or histidine in CYP2 templates. Originally described as "a structureless wandering of 
20 amino acids" (43), the meander region contains some of the 16 residues preceding the arginine (Arg-448) that enveloped the ERR-triad in a highly coordinated hydrogen bonding network first described in P450Bm3, P450cam, and P450terp (43) and viewed in the CYP27A1 model shown in Fig. 4 e and depicted in Fig. 5. A pervasive structural element seen in all P450s, the ERR-triad may act as a folding motif, stabilize heme-binding, and play a role in redox-partner binding (43). It also anchored the start of the B-helix from a conserved aspartate numbered –14 from the meander arginine in all templates except in CYP51 where the aspartate is at position –11.
|
,25-(OH)2D3 to 1,27-(OH)2D3 product ratio shifted from wild-type 49:36 to mutant 42:43. Interestingly, Pikuleva et al. also experimented with Phe-248 (their numbering; F215) and Ile-244 (I211) using lysine scanning and demonstrated that either mutant decreased turnover number and caused a minor accumulation of novel metabolites from the bile acid precursor substrate 5ß-cholestane-3,7,12-triol (53). Annalora et al. experimented with a nearby phenylalanine in rat CYP24A1 (Phe-249), which aligns with Ser-251 in CYP27A1, and concluded that mutation to alanine, threonine, or tyrosine disrupted the orientation and affinity of 24-oxidation intermediates in the multicatalytic oxidation sequence converting 1,25-(OH)2D3 to calcitroic acid. It was postulated that phenylalanine contact with the vitamin D cis-triene was the determining factor in the accumulation of pathway intermediates (54). Overall, these studies suggest that the part of the F-helix around Phe-248 influences the orientation of bound substrate, possibly through interaction with Leu-516 as described below.
|
|
-OH-D3 metabolites shown in Fig. 6 b revealed that 27-hydroxylation was preferentially reduced. The hydroxylation product ratio shown in Table 2 was shifted from 14:49:36 (wild-type) to 21:52:27 (Leu-516-Val) with only an 11% loss of total activity. This was notably opposite to the effect of the F-helix Phe-248-Met mutation and suggests that these two residues act together to shape a substrate contact surface. The minor increase in 24-hydroxylation was also in agreement with an altered substrate orientation in the Leu-516-Val mutation. The residue preceding Leu-516 is less conserved and seen as valine in CYP27 isoforms and isoleucine in CYP24A1 and the CYP11 isoforms. Two conservative mutations at Val-515-Ile and Val-515-Leu were selected to probe the effect of a larger side chain at this position. These mutations dramatically shifted the products toward 25-hydroxlation at the expense of 27-hydroxylation (Fig. 6 c) and caused a 40% increase in total enzyme activity (Table 2). This indicated that the Val-515 side chain made a large contribution to the orientation of bound 1-OH-D3, and the docked substrate shown in Fig. 6 c suggests that this occurs through contact near the substrate D-ring. The third residue in the ß-5 hairpin is conserved as isoleucine in many CYP27A1 homologs, threonine in CYP27B1, glycine in CYP24A1, leucine in CYP11A1, and phenylalanine in CYP11B isoforms, and may represent a finely tuned determinant of substrate specificity. The mutation of this residue (Ile-514-Phe), completely shifted activity toward 25-hydroxylation (Fig. 6 d), but abolished 94% of the total activity. Western analysis of the Ile-514-Phe mutant revealed that it disrupts protein expression (data not shown), suggesting that it may destabilize correct folding of the ß-5 hairpin and a functional enzyme. Coincidentally, a similar effect was seen in the Thr-402-Phe mutant described below.
Substrate contacts and heme-binding in the ß-3 sheet
The ß-3a strand connects the ERR-triad and ß-5 hairpin to the heme-binding region. Along its length are structural, heme-binding, and potential substrate contact residues that are generally specific to a given CYP subfamily and conserved across species. In the CYP27 family, the ß-3a strand starts with a PVVP motif in which the position of the first valine amide carbonyl hydrogen bonds to the amide nitrogen of an aligned ß-5 position (Val-517) in all 11 templates. This backbone interaction helps position the side chain at the second valine position toward a substrate contact near the heme A-ring. In the CYP2 templates, the amide carbonyl at the second valine position also hydrogen bonds with the +3 amide nitrogen to structurally stabilize the heme-binding histidine at the +6 position. Following the PVVP motif in CYP27A1 are three residues, Thr-402, Asn-403, and Ser-404, which are as well conserved as Gly-Asn-Ser in CYP27B1 and Phe-Thr-Thr in CYP24A1. The first of these in CYP27A1, Thr-402, was seen in the model to make a potential hydrophobic contact through its methyl group with the vitamin D A-ring. Conserved in human, mouse, and rat, this interpretation is also consistent with valine in pig, bovine, and zebrafish. Additionally, computational docking of 1-OH-D3 in the binding cavity suggested that the threonine hydroxyl interacts with the substrate 1-hydroxyl and its methyl interacts with the substrate C- and D-rings. Interaction of a threonine hydroxyl in this vicinity with a substrate hydroxyl has also been implicated in substrate recognition in CYP27B1 (55). However, given the absolute conservation of phenylalanine in CYP24A1, the Thr-402-Phe mutation was chosen to see if a possible contact with the cis-triene or changes in the shape of the binding cavity would shift CYP27A1 activity toward 24-hydroxylation.
Similar to the effect of Ile-514-Phe, the Thr-402-Phe mutation abolished 98% of the activity toward 1-OH-D3 (Table 2) and produced only a trace of 25-hydroxylated product (Fig. 6 e). As with the Ile-514-Phe mutant, Western analysis revealed that the Thr-402-Phe mutation also disrupts protein folding (data not shown). Given the proximity of Thr-402 and Ile-514 and their location in the model, the residual activity of these mutants may be attributed to effects on substrate orientation, forcing bound substrate deep into the cavity. The loss of detectable protein for these mutants suggests that either phenylalanine mutation may sterically disrupt a structural hydrogen bond between the ß-5 and ß3a sheets. A structural interaction here between the ß-sheets is likely to be important in CYP27A1 since hydrogen bonding between these two loci is found in 8/11 templates. This is best seen from position 402 side chain (threonine) through a crystallographic water to position 515 amide carbonyl in CYP2C5, CYP2C9, and probably CYP2C8 and from position 402 side chain (serine) to position 514 amide carbonyl in P450terp. In other templates, this interaction involves the position 402 amide nitrogen to position 514 amide carbonyl in P450eryF, to position 515 amide carbonyl in P450cam and P450terp, and through a crystallographically conserved water to position 515 amide carbonyl in CYP51, CYP2C5, CYP2C9, and probably CYP2C8. In CYP119, an analogous hydrogen bond links position 401 amide nitrogen to position 515 amide carbonyl. Given that the ß-5 hairpin is also secured by a hydrogen bond between the PVVP motif and the position 517 amide, the principal effect of hydrogen bonding at the Thr-402 position would be in the positioning of Ile-514 residue. However, the involvement of the side-chain hydroxyl (threonine) in this interaction is not crucial in P450eryF or possible in half of the CYP27A1 species homologs. Interestingly, the presence of threonine at position 514 in fish CYP27A1s and in the CYP27B1 homologs may provide the reciprocal hydrogen bond interaction to the ß-3a strand.
All evidence points to the next ß-3a residue, Asn-403, being an accessory heme-binding residue. Designated earlier as –2ß3Arg, heme-binding with the A-ring propionate is clearly evident in CYP2C5 (Asn-362 stabilized by +21ß3Arg Thr-386) and in P450cam (Asp-297 stabilized by +23ß3Arg Gln-322). In CYP27A1, heme-binding by Asn-403 would be stabilized by +23ß3Arg His-428, the latter being conserved across 15 species. Structurally, the –2ß3Arg position is constrained in the ß-sheet through two backbone hydrogen bonds (in 7/11 templates including CYP3A4) or a single backbone hydrogen bond within the ß-3a strand in the CYP2 family (4/11 templates). In all templates except P450terp, this hydrogen bonding oriented the –2ß3Arg side chain downward and negated any possible substrate contact. Although a serine or threonine at this position is only seen in 3/11 templates and only bound heme through a crystallographic water in P450eryF, the Asn-403-Thr mutation in CYP27A1 was selected to examine whether threonine (conserved in CYP24A1) could assume a heme-binding role. The resulting loss of activity, shown in Fig. 6 e, suggests that Asn-403-Thr cannot support a structurally viable heme-binding environment. Western analysis confirmed that Asn-403-Thr disrupts protein expression, and this is similar to the effect of a mutation at an adjacent heme-binding and CTX residue, Arg-405-Lys, which was shown spectrally to disrupt heme incorporation (46).
The third ß-3a residue examined in this study was –1ß3Arg Ser-404. From its proximity to the binding cavity and the heme A-ring, it is clear that Ser-404 is a potential substrate contact residue. Computational docking of 1-OH-D3 in the binding cavity also suggested that the serine hydroxyl interacts with the substrate 3ß-hydroxyl. Ser-404 also appears to be an integral structural component at the confluence of ß-1, B-B' loop, F-G loop, ß-3a, and ß-3b strands. This region is structurally constrained by extensive ß-sheet structure and by heme-binding interactions through the ß-3 arginine (Arg-405) and B-B' loop arginine (Arg-127). Rooted in this well-defined structure, the B-B' loop made additional contacts with the binding cavity and supported important determinants of substrate specificity and substrate access in the B'-helix, which are poorly represented in the crystallographic templates. Interactions of the B-B' loop with the ß-3a strand (in the vicinity of Ser-404) and ß-3b strand occur in a number of templates. This is best seen in P450bm3 as a triad (–1ß3Arg Ser-322, +1BB'Arg Asn-70, and +18ß3Arg Glu-352), which is functionally conserved in CYP27A1 (across at least seven species) as Ser-404, Asn-128, and Gln-423. Interestingly, functionally similar interactions anchor the B-B' loop in CYP3A4 (–1ß3Arg Glu-374 to +1BB'Arg Arg-106), CYP2B4 (+2BB'Arg Lys-100 amide nitrogen to +18ß3Arg Glu-387), and CYP2C5 (+2BB'Arg Ser-99 amide to +18ß3Arg Asp-383). The hydrogen bonding pattern was more elaborate in P450cam (+2BB'Arg Cys-85, +4BB'Arg Phe-87 amide carbonyl, +18ß3Arg Gln-317), and P450terp (+4BB'Arg Ser76 amide carbonyl, +1ß3Arg Thr-320, +18ß3Arg Arg-337).
Within this context, two site-directed mutations were engineered to explore the role of Ser-404 in CYP27A1. The first, Ser-404-Ala, was selected to structurally disrupt the hydrogen bond with the B-B' loop without sterically interfering with substrate binding. As shown in Fig. 6 f, the Ser-404-Ala mutation preferentially suppressed the 27-hydroxylation of 1-OH-D3 and reduced the total enzyme activity by 37%. This is consistent with steric effects of B-B' loop on the bound substrate resulting from the loss a structural hydrogen bond at Ser-404. The second mutation, Ser-404-Thr, was selected to probe the steric effect of a methyl group with the substrate without disturbing the local structure of the cavity. Surprisingly, this reduced overall enzymatic activity by 64% (Fig. 6 f) without changing the product ratio (wild-type 7:46:46 versus mutant 9:46:45). This suggests that the unfavorable steric effect in Ser-404-Thr is exerted during substrate entry, probably with the substrate C- and D-rings, and not during the final orientation of bound substrate. These findings suggest that Ser-404-Thr retains the substrate recognition function and structurally stabilized the B-B' loop and the bound substrate contact residues therein.
Substrate docking and molecular dynamics
The binding cavity was adequately proportioned to permit manual docking of vitamin D3, 1-OH-D3 A-ring conformers, cholesterol, or the bile acid precursor 5ß-cholestane-3,7,12-triol with little energy minimization. This was confirmed by computational docking that placed 1-OH-D3 in the binding cavity with c25 positioned 0.97 ± 0.37 Å from its average position at a distance of 3.79 ± 0.42 Å from the heme iron (9/10 structures). The substrate A-ring was especially invariant in 8/10 of the docked structures and not significantly different from the manually docked position, reflecting the fact that much of the substrate is a structurally rigid seco-steroid. Most interestingly, the computationally docked substrate sampled the conformational space above the heme by rotating the substrate side-chain bonds at c20-c22, c22-c23, c23-c24, and c24-c25. This allowed c24, c25, and c26/27 to sample metabolically relevant orientations near the hydroxylation site. The absence of the 1-hydroxyl group in vitamin D substrates did not appear to affect binding orientation (data not shown), but a comparison of A-ring conformers suggested that the more planar ß-chair conformer, displaying an axial 1-hydroxyl, gave a better fit (data not shown). The computational docking of 1-OH-D2 was more precise, reflecting the smaller conformational space afforded by the 
22–23 in this substrate side chain. Nine docked structures positioned c24 in a physiologically relevant orientation 0.56 ± 0.31 Å from its average position at a distance of 4.37 ± 0.36 Å from the heme iron. This was especially exciting from a vitamin D standpoint since the 24-hydroxylation of vitamin D2 is a unique enzymatic property of the enzyme (4). Overall, we are confident that the manually docked substrate adequately exemplifies a metabolically relevant conformation in the active site, especially in the placement of the A-ring hydroxyls as molecular "handles" interacting with Thr-402, Ser-404, and other polar residues in the binding site.
We performed molecular dynamics simulations using Insight II on our docking model and observed structural convergence as early as 100 ps with an RMSD of 1.22 Å averaged over 479 alpha carbons. In light of the quick convergence, there is no need to continue the molecular dynamics calculations. We subsequently docked 1-OH-D3 into this structure and subjected it to an additional molecular dynamics simulation; only after 50 ps, it converged completely and the RMSD decreased to 1.13 Å. Using PROCHECK, we found that 96.2% of residues are in allowed Ramachandran plots.
The inner part of the binding cavity lay close to the ß-3a strand and was defined by residues in the oxygen binding site in the F- and I-helix, B-B' and B'-C loops, and nonpolar residues in the ß-5 hairpin. The outer part of the binding cavity and substrate access channel are defined by residues in the B-B' loop, A'-, A-, B'- and F-helices, F-G loop, and the ß-1 and ß-5 hairpin. Proton acceptor residues there are presumed to help partition substrate into the active site and orient it during the reaction cycle through interactions with A-ring hydroxyl groups in vitamin D and 1-OH-D3 and with the 3-, 7-, and 12-hydroxyl moieties of bile acid precursors. Additionally, a proton donor may be present to coordinate with the c3 carbonyl of cholestenone, a preferred substrate of CYP27A1 (6).
The outer part of the binding cavity near the ß-sheets contains an abundance of asparagines, glutamines, and histidines, which suggests a structural network of hydrogen-bonded residues with a multifaceted capacity to recognize substrate. Quite intriguingly, the contact residues in the ß-sheets and A-helix are located at discrete positions (shown in Fig. 1) and have identical side-chain orientations that are independently determined by cytochrome P450 structure. The cross species conservation of these discretely positioned residues in all known vitamin D hydroxylases is summarized in Fig. 7. What contributes to substrate specificity here is the identity of the residue at each position. What determines substrate contact is the size of the substrate, size of the residue, and the extent to which side chains are displaced by adjacent residues. What determine substrate recognition are proximity, side-chain polarity, and the presence of functional groups in the substrate. There are other substrate contact and substrate recognition residues in the B'- and F-helices and in the B-B' and F-G loops and in the substrate access channel, but in CYP27A1, there are no proton acceptor residues in the F-G loop other than backbone carbonyls and a nonionized Lys-259 or Arg-262 side chain.
|
-OH-D3 and Fig. 8 c shows the contribution of binding cavity residues near the substrate side chain. It is interesting to note that, for clarity, His-105, Asn-107, and Cys-427 are not shown, but the histidine and cysteine are also well placed to hydrogen bond with the 3ß-hydroxyl and 1-hydroxyl groups, respectively. One of the more interesting residues, Gln-85, is a VDDR-I locus (Q65H) and was conserved in the A-helix across species in CYP27A1 and CYP27B1. It has been implicated in substrate binding in CYP27B1 (55), but given the opposite orientations of substrate in these two enzymes, it seems unlikely that Gln-85 side-chain placement would be optimal for substrate recognition in both. Instead, Gln-85 may fulfill a structural function, such as participation in the structural network in the outer part of the binding cavity or blocking an access channel between the A'- and A-helices. Alternatively, the Q65H polymorphism in CYP27B1 may disrupt a structural hydrogen bond to another discretely positioned residue, Thr-409, which is also the site of a VDDR-I polymorphism (T409I). If Gln-85 does coordinate with the 1-hydroxyl group on 1-OH-D3, the model suggests that this function may be shared with Cys-427 and His-82. In fact, His-82 may be the more interesting residue because it is positioned to simultaneously form a structural hydrogen bond with ß-5 Ile-514 amide carbonyl and contact the substrate 1-hydroxyl group. This may explain why His-82 is conserved across species in CYP11 isoforms, rat CYP2J3, CYP2R1, and all mitochondrial vitamin D hydroxylases (except murine Cyp27a1 where it is a proline).
|
,7,12-triol. Given its docked orientation, residues such as Gln-423 or His-105 (not shown) are well positioned to interact with the c3-hydroxyl, His-82 with the c7-hydroxyl, and possibly a B-B' loop backbone carbonyl (e.g., Asp-129) with the c12-hydroxyl. These same residues could also interact with the carbonyl of cholestenone.
Substrate access channel
Previously identified in P450cam (56), P450Bm3 (57), and P450terp (58) and modeled into aromatase (59), the substrate access channel connects the binding cavity to the protein surface. In all but a few crystal structures, this opening is hidden, even in the presence of bound substrate or inhibitor. Molecular dynamics simulations in P450Bm3, P450cam, and P450terp have examined the energetics of a number of molecular trajectories out of the binding cavity (60,61) to find concealed channels, and a recent review (49) discussed at least eight possible access routes or pathways: pw1, between the heme and I-helix (possible oxygen channel); pw2a, opening at F-G loop, B'-helix, ß-1; pw2b, parallel to ß-3a and opening between the B-B' loop and ß-1; pw2c, opening between the G- and I-helices and the B'-helix, B-C loop; pw2d opening between the A- and A'-helices; pw2e through the B-B' and B-C' loops; pw3, between the F- and G-helices or at the E-F loop; and a solvent channel between the E-, F-, and I-helices and ß-5. Unfortunately, the unveiling of a substrate access channel along one of these pathways may be a transitory event coincident with substrate penetration or the conformational changes associated with reaction cycling and membrane association. Although the most likely access route in most CYPs is pw2a (49), it is not known whether different CYPs use different pathways, whether a single CYP uses different channels for different substrates, or whether separate pathways are used for substrate access and product egress.
The geometry of the pw2a substrate access channel found in the CYP27A1 docking model and shown in Fig. 8 d is not ideal and exhibited some pw2b-type character similar to that seen in CYP2C8 and P450Bm3. It was primarily occluded by A'-helix Gln-75, F-helix Phe-256, and B-B' loop Met-130, possibly in part due to the uncertainties in modeling the F-G loop and the B'-helix. Although the structure of this channel does not affect how substrates are docked, it is desirable to objectively refine it to better understand substrate recognition during binding. Our model revealed a cluster of 10 aromatic residues above the heme that contributed a large hydrophobic surface to the binding cavity and substrate access channel. This cluster is surrounded by the F-G loop and included Trp-133 (B'-helix), Tyr-253, and Phe-256 (F-helix), Trp-260 and Trp-268 (F-G loop), Phe-267, Tyr-271, Trp-275 (G-helix), and possibly Phe-279 and His-136, and may interact with another cluster of aromatic residues in the A'-helix, the latter possibly analogous to hydrophobic surface patches in P450bm3 (57) and P450terp (58) postulated to be involved in substrate recognition. Curiously, the former cluster corresponded approximately to the location of a bound palmitic acid in CYP2C8 (29) and an allosterically bound progesterone near an analogous cluster of seven phenylalanine residues (Phe cluster) in CYP3A4 (31). Additional modeling may be necessary to fully understand the contribution of this aromatic cluster to pw2a geometry and CYP activity.
To further characterize the contribution of the F-G loop to the pw2a channel, we examined the model and site-directed mutagenesis and kinetic data of Murtazina et al., who showed that F-G loop residues in CYP27A1 influence cholesterol binding and 27-hydroxylation (62). Mutations in these sites were postulated to disturb membrane binding and proton availability during the reaction cycle and to restrict product egress. Our interpretation is that selected mutations might also disrupt the aromatic cluster above the heme and may variously affect the shape of the binding cavity and/or the pw2a access channel to affect the enzyme kinetics. Although the Murtazina et al. model shows some similarity to the model presented here, its major shortcoming is the lack of a binding cavity, which we attribute to an incorrect positioning of Gly-145 (a VDDR-I and CTX residue) in the B'-C loop, Phe-240 in the F-helix, and an improperly folded heme region.
The finding of an open pw3-type egress channel above the I-helix and centered on Met-334 was intriguing. That this channel is visible at all may result from uncertainty in the positions and separation of the F- and G-helices or the position of the E-F loop. If and how it should be closed in the model will permit refinement of local structure and has important implications for the movement of molecules through the enzyme. Synchronization of substrate binding and metabolite release through different channels is attractive because it would sequentially expose both ends of vitamin D to hydroxylation, in agreement with reports of CYP27A1 exhibiting both 25- and 1-hydroxylase activities (63,64).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
". Residues mutated in this study are indicated with white text on black and identified by lower case letters: a, Phe-248-Met; b, Thr-402-Phe; c, Asn-403-Thr; d, Ser-404-Ala, Ser-404-Thr; e, Ile-514-Phe; f, Val-515-Ile, Val-515-Leu; and g, Leu-516-Val. Sites of missense mutations causing cerebrotendinous xanthomatosis (CYP27A1), vitamin D-dependency rickets (CYP27B1), and autosomal deficiencies in CYP11B1, CYP17, and CYP21B1 are also indicated with white text on black shading. Gaps within secondary structures were added to emphasize structurally important landmarks.
