Architecture of a single membrane spanning cytochrome P450 suggests constraints that orient the catalytic domain relative to a bilayer

Significance

The absence in the Protein Data Bank of full-length structures of bitopic membrane proteins with one transmembrane helix, probably because of difficulties with ordered crystallization, has limited understanding of how single-transmembrane helices orient enzymes and sensors at the bilayer surface. X-ray crystal structures of full-length yeast lanosterol 14α-demethylase, a cytochrome P450, show how a helix spanning a single transmembrane may lead to constraints on the orientation of the putative substrate entry portal from within the bilayer. The crystal structures also locate the substrate lanosterol, identify putative substrate and product channels, and reveal constrained interactions with triazole antifungal drugs that are important for drug design and understanding the drug resistance associated with orthologs of the enzyme found in fungal pathogens.

Abstract

Bitopic integral membrane proteins with a single transmembrane helix play diverse roles in catalysis, cell signaling, and morphogenesis. Complete monospanning protein structures are needed to show how interaction between the transmembrane helix and catalytic domain might influence association with the membrane and function. We report crystal structures of full-length Saccharomyces cerevisiae lanosterol 14α-demethylase, a membrane monospanning cytochrome P450 of the CYP51 family that catalyzes the first postcyclization step in ergosterol biosynthesis and is inhibited by triazole drugs. The structures reveal a well-ordered N-terminal amphipathic helix preceding a putative transmembrane helix that would constrain the catalytic domain orientation to lie partly in the lipid bilayer. The structures locate the substrate lanosterol, identify putative substrate and product channels, and reveal constrained interactions with triazole antifungal drugs that are important for drug design and understanding drug resistance.

Membrane proteins that span the lipid bilayer once constitute around 50% of all integral membrane proteins (1). Although monospanning membrane proteins carry out numerous key biological functions, including environmental sensing, organelle-specific catalysis, and the regulation of cell morphology, only individual domains or subdomains are currently represented in the Protein Data Bank, and structural information about interactions between their transmembrane domains and extramembranous components is lacking. Cytochrome P450 proteins are prominent enzymes with orthologs found in all kingdoms of life. In eukaryotes, microsomal members of this major family of mixed-function mono-oxygenases contain a single transmembrane helix and can be grouped in two broad functional categories: biodefense, such as the first phase of xenobiotic detoxification, and core metabolism including reactions in sterol biosynthesis and fatty acid oxidation (2).

The lanosterol 14α-demethylases or CYP51 enzymes, probably the most genetically ancient of the cytochrome P450 families, play a central role in cholesterol or ergosterol biosynthesis (3). CYP51s carry out three consecutive mono-oxygenase reaction cycles to remove the 14α-methyl group from lanosterol to yield 4,4-dimethyl-cholesta-8,14,24-trienol, a key precursor in cholesterol and ergosterol biosynthesis, releasing water and formic acid (3). Because of the key roles that CYP51s play in yeast, filamentous fungi, and some parasitic protozoa, these enzymes are therapeutic targets for antimicrobial agents, including fluconazole (FLC), voriconazole (VCZ), and itraconazole (ITC) (4). Fungal infections play an increasingly significant role in disease, impacting agriculture ecosystems and human health, especially in immunocompromised individuals (5–7) for whom antifungal resistance continually poses a threat (8). In humans CYP51 is being tested as a target for cholesterol-lowering drugs (9) and in antiangiogenic cancer therapies (10). A limited set of cytochrome P450 isoforms (1A2, 2C8, 2C9, 2C19, 2D6, and 3A4) metabolize about 75% of the drugs currently used in medicine (11).

Membrane-bound cytochrome P450s catalyze reactions involving many hydrophobic substrates, including lipophilic drugs and sterols that partition from the lipid bilayer into the active site. Identification of enzyme orientation relative to the lipid bilayer therefore is critical for understanding substrate entry into and egress from the heme-containing active site (12). For the eukaryotic cytochrome P450 enzymes that contain transmembrane segments, most X-ray structures have been obtained after removal of the N-terminal transmembrane domain to facilitate expression and crystallization (13–18), as is the case with many other monospanning membrane protein structures. Resolved CYP51 structures include the soluble enzyme from Mycobacterium tuberculosis (19) and N-terminal truncated enzymes from Homo sapiens (20), Trypanosoma cruzi (21), Trypanosoma brucei (22), and Leishmania infantum (23). The available eukaryotic cytochrome P450 structures show the catalytic subunit has a conserved fold surrounding a central prosthetic heme group. A proposed substrate channel to the membrane as well as interactions with several inhibitors, including ketoconazole, posaconazole, FLC, and substrate analogs such as 14α-methylenecyclopropyl-Δ7–24,25-dihydrolanosterol, have been visualized (19, 20, 22, 24).

Although bacterial and mitochondrial orthologs of cytochrome P450s and the M. tuberculosis CYP51 are soluble enzymes, bioinformatic analysis of many microsomal cytochrome P450s led to the detection of a hydrophobic surface in the catalytic domain and prediction of an N-terminal monospanning transmembrane domain that has been shown to be important for subcellular targeting and membrane insertion (25–28). Multiple in vitro approaches have suggested that a transmembrane anchor is not always required for the catalytic activity of some microsomal preparations of cytochrome P450s (29, 30). For example, M. tuberculosis CYP51 specifically converts detergent-solubilized lanosterol to its demethylated product (31). However, the in vivo role of the membrane anchor in catalysis by other CYP51s that use membrane-associated lanosterol has yet to be established. Antibody-binding accessibility assays showed that most of the surface of CYP2B4 (32) was solvent exposed with the exception of the putative transmembrane helix and the loop between the F and G helices. About 30% of the catalytic domain of CYP11A1 was inaccessible to chemical modification by fluorescent probes (33). The physical displacement of phospholipids in Langmuir–Blodgett monolayers indicated a considerably greater area of protein insertion than could be accounted for by a single N-terminal transmembrane helix (34). This finding was supported by atomic force microscopy, which showed the catalytic domain protrudes only 3.5 ± 1 nm above the membrane (35). A crystal structure of full-length aromatase at 2.9-Å resolution has been modeled, but weak electron density meant that the structures of 44 N-terminal and seven C-terminal residues could not be determined (36). Models that place the mouth of the substrate channel in direct contact with the hydrophobic environment of the lipid bilayer required the catalytic domain helices A′ and A to be imbedded and partially imbedded, respectively, in the lipid bilayer (28, 37).

Results

Architecture of the Cytochrome P450 Saccharomyces cerevisiae Lanosterol 14α-Demethylase.

We report X-ray crystal structures of full-length Saccharomyces cerevisiae lanosterol 14α-demethylase (ScErg11p-6xHis) with the membrane domain structurally discernable and two ligands cocrystallized: the substrate lanosterol and the triazole inhibitor ITC. Clear electron density for the 50 N-terminal amino acids revealed two helices oriented at ∼60° to each other (Fig. 1 A and B). The N-terminal helix [membrane helix 1 (MH1), residues 6–23] is amphipathic and makes extensive crystal contacts with symmetry-related molecules (Fig. 2 A and B). The distribution of hydrophobic and hydrophilic side chains suggests that the hydrophobic face may contact the inner leaflet of the endoplasmic reticulum (ER) membrane (Fig. 1B). MH1 connects to TMH1 via a short proline-containing turn (residues 24–26), a slightly kinked helix 37.5 Å long (residues 27–51) that is of sufficient length to traverse the lipid bilayer. The side chains of amino acids R30 and Q29, adjacent to the turn, form a C-terminal cap for helix MH1 and an N-terminal cap for TMH1, respectively.

Fig. 1.

Overall fold of ScErg11p and predicted orientation in the lipid bilayer. (A) Helices are colored from the N terminus to the C terminus with a gradient from blue to red. The heme is shown as yellow sticks; lanosterol in the active-site cavity is shown in orange. (B) The structure of the amphipathic helix (MH1) and the transmembrane helix (TMH1). Hydrophobic side chains are colored blue, and polar residues colored magenta, with oxygen atoms colored red, and nitrogen atoms colored blue. Difference electron density (Fo–Fc) contoured to 2σ is depicted as a green mesh. (C) Interaction of the transmembrane domain with the catalytic domain. Ordered water molecules are depicted as red spheres. Hydrogen bonding interactions are shown as dashed lines; distances are in Å. (D) Distribution of hydrophobic and charged residues in the F/G loop.

Fig. 2.

Crystal packing of ScErg11p-6xHis. (A) ScErg11p-6xHis crystals show alternating layers of transmembrane (TM) segments and the soluble segments (S). (B) Crystal packing of two symmetry-related copies of ScErg11p-6xHis with the amphipathic helix S6–H21. The amphipathic helix is colored green, and the symmetrically related copies are colored salmon or blue. (C) Overlay of lanosterol-bound (gray) and ITC-bound (salmon) ScErg11p-6xHis.

Seventeen of the 25 residues in TMH1 are hydrophobic, with a high concentration in the N-terminal portion, as is consistent with immersion of TMH1 in the lipid bilayer. Beyond the proline kink at P38, TMH1 makes extensive polar contacts with the catalytic domain (Fig. 1C). The helix then consists of opposing hydrophilic and hydrophobic faces that interact with the catalytic domain and lipid bilayer, respectively. An ionic interaction between the guanidinium group of R52 and the carboxylate of D54 enables a sharp turn between TMH1 and the first loop of the catalytic domain.

Paired polar residues constrain the orientation between TMH1 and the catalytic domain. The ε-amide group of Q46, which hydrogen bonds with the main-chain carbonyl and amide of Y61, is absolutely conserved in the fungal CYP51 family. The hydroxyl of Y61 also makes a hydrogen bond with the δ-amide group of N42. Water molecules bridge the ε-amide group of Q46 with the main-chain amide oxygen of L58 and the ζ-hydroxyl of Y49 with the hydroxyl of S92. Other polar contacts in this region include the guanidinium group of R55 and the carbonyl of A399; the main-chain amide of K53 and the hydroxyl of S397; and the main-chain carbonyl of D54 and the amide of A399. These extensive contacts bury a surface area of 51.8 Ų on TMH1 and appear to form a rigid architecture that is important for establishing the pose of the enzyme in the bilayer (Figs. 1A and 2 A and C). The energetic stabilization conferred by the extensive contacts between TMH1 and the catalytic domain make it unlikely that the crystal contacts between the amphipathic MH1 (residues S6, E10, E13, and H21) and residues K270 and S267 of one symmetry-related molecule and G532 and E354 of the other (Fig. 2B) would have distorted the overall structure of ScErg11p significantly. The ScErg11p-6xHis structures show few significant differences between structures, with the most extensive changes (0.36 Å rmsd on Cα) being in the F/G loop (Fig. 2C) of the lanosterol and ITC costructures.

Inferred Orientation of ScErg11p Relative to the Membrane.

To establish the membrane orientation, ScErg11p-6xHis was inserted into preformed lipid vesicles and labeled from the outside at free carboxylates with glycinamide in the presence of N′N′-Dicyclohexylcarbdiimide (DCCD) at pH 6 (Fig. 3A) (38). Tandem mass spectrometry of tryptic digests identified labeled residues only on the catalytic domain, with no labeling on the N-terminal MH1 (residues 1–26), as is consistent with topological labeling of other cytochrome P450s and with MH1 and the catalytic domain associating with opposite sides of the bilayer.

Fig. 3.

Sequence and predicted membrane interactions of ScErg11p. (A) Regions highlighted in yellow are predicted by OPM to lie inside the lipid bilayer. Red squares indicate acidic residues that were accessible to the water-soluble glycinamide probe when ScErg11p-6xHis was incorporated into artificial liposomes. (B) Electrostatic surface map of ScErg11p with cutaway showing the primary and secondary vestibules. Surface electrostatic potential is displayed with the molecular graphics software Chimera (59) with blue as positive and red as negative.

The orientation of ScErg11p also was explored using two algorithms for membrane insertion: the all-atomistic positioning of proteins in the membrane (PPM) procedure (39) and the knowledge-based Ex3d procedure (40). Both calculated poses were very similar to those inferred from our crystallographic analysis, i.e., TMH1 residues 27–50 link the N-terminal amphipathic helix MH1 and the C-terminal catalytic domain that are partially submerged in opposite leaflets of the lipid bilayer (Fig. 1A). The OPM analysis, which calculated a partitioning free energy of −31.3 kcal/mol into the lipid bilayer, was the best fit with our topological labeling, including the labeling of residue E249 on the F/G loop, and was used in this analysis (Fig. 3).

The pose we infer for ScErg11p relative to the membrane buries portions of the loop connecting helices F–F′ and G in the lipid bilayer (Figs. 1D, 2C, and 3). It is consistent with mutagenesis experiments showing this region is important for membrane binding (33, 41, 42), fluorescence quenching analysis (5), the predicted pose of aromatase (36), and computational simulations for human cytochrome P450s performed in the presence of diolyeolphosphatidylcholine (DOPC) (37) or 1-palmitoyl-2-oleoyl-sn-glycerol-2-phosphocholine (POPC) (27, 32–34, 37, 43). Several aspects of the structure support the proposed orientation. Hydrophobic residues are distributed in a single band across the surface of the structure. Poisson–Boltzmann electrostatics calculations using Delphi software (44) reveal an electrostatic potential surface showing a distinct hydrophobic/hydrophilic boundary (Fig. 3B). Ordered water molecules are present only in proposed solvent-accessible areas of the model and are absent from the bilayer-accessible surface of the structure (Fig. 1 B and C). As a corollary to the positioning of the lipid bilayer, the protein crystallized with alternating layers of proposed membrane and soluble segments (Fig. 2A) corresponding to the placement of hydrophobic and hydrophilic elements in the crystal structure, including the surface-associated ligands.

We hypothesize that a cluster of charged residues found at the N-terminal end of the G helix forms a major membrane-binding contact with the negatively charged phosphate head groups of the lipid bilayer (Fig. 1D). This orientation places the end of a substrate channel in direct contact with the cytoplasmic surface of the lipid bilayer (Figs. 1A, 2C, 3B, and 4C). This substrate channel is bifurcated near its mouth, yielding a secondary vestibule (Fig. 3C) not previously identified in CYP51s. Although an N-decyl-β-d-maltoside (DM) detergent micelle differs from an intact lipid bilayer, the proposed orientation supports a model in which the substrate first is abstracted directly from the surface of the membrane via the juxtaposed channel mouth. After heme reduction and complex formation with oxygen, an oxygen rebound mechanism converts the C14 methyl to the first of three products that result in 4,4-dimethyl-cholesta-8,14,24-trienol (2).

Fig. 4.

Lanosterol and ITC binding in ScErg11p. (A) Lanosterol is depicted with carbon atoms colored cyan and heme with carbon atoms colored yellow. Selected oxygen atoms are colored red, nitrogen blue, sulfur gold, and iron brown. Electron density is depicted as a simulated annealing 2Fo–Fc omit map contoured to 0.7σ with the ligand, heme, and oxygen omitted from the electron density calculation. A bond links the heme–oxo complex. (B) ITC in a coordination complex with active-site heme. The color scheme is as in A, with ITC carbons colored magenta. Electron density is depicted as an Fo–Fc map contoured to 2.5σ with the ligand density omitted from the calculation. (C) Conserved amino acids in fungi that are commonly mutated in antifungal resistance are depicted with carbon atoms colored orange. Amino acids conserved in S. cerevisiae and in the main fungal pathogens of humans but not in human CYP51, and not mutated, are depicted with carbon atoms colored green. Amino acids that are similar only in the pathogenic fungi are shown in cyan. Other atoms are colored as in A.

Substrate and Product Binding.

The enzyme was purified and cocrystallized with the substrate lanosterol or with four other ligands (itraconazole, estriol, FLC, or VCZ). The lanosterol (Fig. 4A) and estriol (Fig. S1A) cocrystal structures showed density consistent with the common four-ring sterol in the active-site cavity. A mixed state probably involving both endogenous lanosterol and added ligand complicated interpretation of electron density in the active site. This mixed state did not occur when ScErg11p was purified in the presence of ITC and crystallized. As a result, only the lanosterol and ITC structures are detailed here with the four others reported as models in SI Discussion.

The lanosterol (Fig. 4A) and estriol (Fig. S1A) cocrystal structures showed density consistent with the common four-ring sterol in the active-site cavity. Electron density consistent with the tail of lanosterol is located in the primary vestibule that leads up to the active site (Fig. 5B). Similar electron density was observed in the same locations when no ligand was supplied (Fig. S1D), indicating that significant amounts of substrate copurify with the enzyme. HPLC-MS in positive ion mode (45) showed that lanosterol copurifies with ScErg11p-6xHis in the absence of added ligand, supporting this interpretation (Fig. S1E).

Fig. 5.

Multiple ligands bind to ScErg11p. (A) Electron density of lanosterol-like molecule. Fo–Fc density contoured to 2.0σ is shown in green with a theoretical model of 4,4-dimethyl-cholesta-8,14,24-trienol placed. The 4,4-dimethyl-cholesta-8,14,24-trienol is depicted with carbon atoms colored cyan. (B) Hypothetical model of ligand binding with surface representation of the primary and secondary vestibules. Lanosterol bound in the active site is depicted with carbon atoms colored cyan. The second lanosterol-like molecule is depicted with carbon atoms colored purple. Oxygen atoms are colored red, nitrogen blue, and iron brown. The outside surface of the vestibules and active site are depicted as a light mesh surface colored dark gray, and their visible inside surface is colored light gray.

Electron density proximal to the heme iron, consistent with the presence of an oxygen molecule coordinated to the heme iron, fits with the ferrous dioxo intermediate predicted for the CYP51 reaction cycle. This feature also was seen in the crystal structure without added ligand, which also had bound lanosterol (Fig. S1D). To interpret this peak, the electron density near the heme was assessed using crystallographic difference-map refinement. Fo–Fc maps calculated after several rounds of refinement showed that a single water (10 electrons) would generate insufficient positive density, whereas imidazole (32 electrons) would give a strong negative density peak. Placing O₂ (16 electrons) gave a flat Fo–Fc electron density map consistent with O₂ being at the heme site. Although unusual for P450s, this result is not without precedent (46) and may be caused by reduction after X-ray radiolysis (47) or purification of ScErg11p-6xHis trapped in the oxygen-bound state. The distance of ∼8 Å between oxygen and C14 methyl is consistent with a lanosterol binding in a precatalytic rather than a catalytically active state. Visual inspection of the electron density maps and a high real-space R-factor indicated the need for considerable care in placement of the ligand in the lanosterol costructure. Minimally biased Fo–Fc electron density omit maps, calculated with lanosterol omitted to place the lanosterol correctly, resulted in a ligand placement with reasonable correlation coefficient (0.813). Bound lanosterol is oriented differently from other lanosterol-like molecules in related CYP51s (24), and the electron density offers no evidence of these alternate orientations.

Density consistent with a second lanosterol-like molecule projects from the secondary vestibule (Fig. 5 A and B). In all ScErg11p structures examined, a hydrophobic tail fills the narrow channel from the secondary vestibule to the surface of Erg11p, whereas a multiring head that projects into solution appears to lack the C14 methyl group. Lack of density for the C14 methyl group suggested that the lanosterol-like molecule was the reaction product 4,4-dimethyl-cholesta-8,14,24-trienol or a subsequent metabolite that likely copurified with the enzyme. Mass spectrometry of the purified ScErg11-6xHis detected lanosterol, larger quantities of zymosterol, and traces of 4,4-dimethyl-cholesta-8,14,24-trienol (Fig. S1E). If a product egress channel exists, the narrow channel off the secondary vestibule may provide a selectivity filter for the product; that is, removal of the C14 methyl group of lanosterol may be required for the multiring head group to traverse the channel (Fig. 5B). Erg11p has been shown to interact physically with Erg25p-Erg27p and possibly with Erg24p (48). Presentation of 4,4-dimethyl-cholesta-8,14,24-trienol on Erg11p then would allow concerted reduction by Erg24p (C14 sterol reductase) and demethylation at carbon 4 by Erg25p (C4 sterol methyl oxidase), Erg26p (C3-sterol dehydrogenase), and Erg27p (3-keto sterol reductase) in complex with the scaffold protein Erg28p to give zymosterol in the ergosterol biosynthesis pathway (48). Alternatively, the retention of 4,4-dimethyl-cholesta-8,14,24-trienol and zymosterol by Erg11p may play roles that include conferral of structural stability or a feedback control mechanism. Two regions of weak excess electron density located near the heme and adjacent to K220 bordering helix G are likely hydrophobic molecules, possibly detergent or phospholipid copurified with the enzyme, (Fig. S2 A and B).

Discussion

Azole Binding and Antifungal Resistance.

Our structures identify important ligand-binding constraints. The antifungal triazole drug ITC extends from the active site to just beyond the mouth of the entry channel, similar to posaconazole in the T. brucei CYP51 structure (49). The triazole group of the ITC makes a coordination bond with the heme iron. The space occupied by ITC (Fig. 4B) fits closely with that occupied by lanosterol and bound O₂ (Fig. 4A), with the triazole head group displacing the O₂ and the di-halogenated headgroup replacing the first sterol ring. The expected chlorine atoms on the dichlorophenyl group consistently showed strongly negative Fo–Fc density, suggesting that the chloro groups had been damaged through photoreduction upon X-ray data collection. As a result, the chloro groups have been refined to zero occupancy in the model. In confirmation of our findings, the triazole head groups in the smaller triazole drugs FLC and VCZ coordinate with the heme iron and displace the presumed O₂, and their unmodified difluorinated head groups fill the space occupied by the first sterol ring of lanosterol and the sterol analog estriol (Fig. 4 A and B and Fig. S1 A–C). The long tail of ITC fills the entry channel and appears to exclude all but one water molecule. The tail of ITC lies within a hydrophobic pocket lined by helices F and F′ (F236, P238, and F241), consistent with previous structures of truncated CYP51 family members (50), and by the residues L380HSL383 and F506TSMV510 (Fig. 4B and see Fig. S4A).

Although less than 24% of the amino acids in ScErg11p are conserved among fungi (50), there is an especially high concentration of conserved residues near the heme (Fig. S4A) and active-site cavity. These conserved residues include the absolutely conserved coordination of the C470 sulfur to the heme iron and the ionized K151, R385, and H468 side chains. The sequence G310VLMG314 in helix I is a consensus sequence (GXXXG) for sterol (cholesterol) binding, with the carbonyl of G310 making the only polar interaction with lanosterol and a slightly kinked face toward ligand.

The structure of the active site and substrate channel may explain the susceptibility of some azole-resistant clinical isolates, including Candida albicans mutants F126L, Y132F/H, P230L, G307S, and F380S, Aspergillus fumigatus G54E/F and I301, Cryptococcus neoformans Y145F, and Ajellomyces capsulatus Y136F (51, 52) to ITC or FLC (Fig. S3). For example, Y140 in S. cerevisiae (Fig. S3C) is equivalent to Y132 in C. albicans, Y145 in C. neoformans, and Y136 in A. capsulatus. The absence of the heme-binding hydrogen bond in S. cerevisiae Y140F/H mutations could modify the orientation of the heme and reduce susceptibility to both FLC and ITC. In contrast, G73E/F mutations in ScErg11p, equivalent to G54E/F in A. fumigatus Cyp51A, are in the mouth of the substrate channel and might be expected to modify susceptibility to the long-tailed triazoles ITC and posaconazole but not short-tailed VCZ.

Numerous azole-resistance mutations occur outside the active site. Significantly, the N22D azole-resistance mutation in TMH1 of A. fumigatus Cyp51A (ScErg11p N42D) supports the proposition that interactions between TMH1 and the catalytic domain can affect active-site function. CYP51s from S. cerevisiae and the fungal pathogens of humans (Fig. S4) contain a group of conserved/similar aromatic residues (F60, Y61, W62, Y72, Y77, F79, F80, Y87, F90, H381) bordered by TMH1 that form a compact, buried structure. In concert with neighboring conserved/similar hydrophobic residues (V59, L95, L96 L383, L407) and the polypeptide backbone (A69, G73, S382), this grouping supports a narrow entry channel into the active site and egress vestibules. The tight vestibules orient the lanosterol and may provide gated entry of demethylated reaction product into the egress channel. Finally, mutation conferring azole resistance at residues homologous to ScErg11p L95, F241, H381, and S382 has yet to be reported. These residues along with the chemically similar residue I139 could provide fungal-specific side-chain contacts for the design of novel antifungals directed against the major fungal pathogens of humans (Fig. 4C and Fig. S3B).

Bitopic Membrane Proteins.

Full-length structures of bitopic membrane proteins with one transmembrane helix are absent in the Protein Data Bank, probably because of the difficulties with ordered crystallization. This deficiency has limited understanding of how single-transmembrane helices orient enzymes/sensors at the bilayer surface (1). The X-ray structure of ScErg11p-6xHis, a full-length fungal cytochrome P450 enzyme, shows that the transmembrane domain not only tethers the protein to the ER but also may orient it relative to the bilayer. The surprisingly rigid interaction between the transmembrane domain and catalytic domain may establish the pose required for catalysis and interaction with other protein partners such as Erg24p and Erg25p–Erg27p. With monospanning membrane proteins constituting about half of all membrane proteins, structural information supporting properties beyond tethering and localization in ScErg11p provides clues about constraining polar residues in other monospanning enzymes. It also may define a motif that may be a feature of other membrane-bound enzymes.

The ScErg11p structures identify lanosterol binding in the active site of a CYP51, a substrate channel linking to the lipid bilayer, and a secondary vestibule with a product exit channel. The model also more fully maps how triazole antifungals interact with a fungal cytochrome P450 to block catalysis and identifies possible interactions that confer azole resistance applicable to pathogenic fungal species. Like other CYP51 structures, the intact enzyme displays less conformational heterogeneity between structures than seen in many truncated ectodomain eukaryotic cytochrome P450s, suggesting that the transmembrane domain itself or associated hydrophobic molecules such as lipids or reaction product may decrease conformational heterogeneity and protect the active site from bulk water. The structures shown here may enable the development of new molecular models to facilitate drug design that targets fungal CYP51s or other eukaryotic cytochrome P450s and may provide a practical basis for the design of therapeutics with minimized off-target effects.

Methods

S. cerevisiae ScErg11p was overexpressed homologously as a 6×His derivative (ScErg11p-6xHis) from the PDR5 locus in the host yeast strain ADΔ (Table S1) (55). ScErg11p-GFP expressed in ADΔ was found in nuclear-associated ER (Fig. S5A). ScErg11p-6xHis was purified from strain MMLY941 by extraction of a crude membrane fraction with the detergent DM, Ni-NTA-agarose affinity chromatography, and size-exclusion chromatography (SEC) in the presence or absence of ITC (Fig. S5B). The purified protein behaved as a detergent micelle-protein monomer during SEC, as is consistent with a molecular mass for the protein of ∼62 kDa (Fig. S2C). It bound ergosterol and showed type II binding of known antifungals (Fig. S5D). The purified protein was authenticated by mass spectrometry of tryptic fingerprints at the Stanford University Mass Spectrometry facility (Fig. S5E).

Red crystals were obtained initially by hanging-drop vapor diffusion at 18 °C in 200-nL drops with a MemGold (Molecular Dimensions Limited) screen set using a Mosquito robot (TTP Labtech). Crystallization conditions were optimized, and large, red, boat-shaped crystals (100 µm) were obtained in 2- to 6-µL drops with the purified protein at ∼20 mg/mL in 10% (wt/vol) glycerol, 150 mM NaCl, 4× critical micellar concentration DM, and 20 mM Hepes (pH 7.5) mixed 1:1 with 43–45% (wt/vol) PEG 400 in 100 mM glycine (pH 9.3–9.5) (Fig. S5F).

Complete datasets were obtained for crystals at the Berkeley Advanced Light Source BL 8.3.1 with a beam at wavelength 1.1587 Å (Table S2). Data were processed, integrated, and scaled using the HKL2000 (53) package. An initial model was determined by molecular replacement using the soluble domain of the HsCYP51 model 3LD6 (20) in Phaser (54). Refinement was performed using both Refmac5 in the CCP4 (55) suite of programs and Phenix (56), and model building was performed using Coot (57). Model validation was performed in Phenix and with Molprobity (55) and PROCHECK (58). Display of electrostatic maps was performed in Chimera (59).