The Evolution of Azole Resistance in Candida albicans Sterol 14α-Demethylase (CYP51) through Incremental Amino Acid Substitutions

Purification and enzyme catalysis of CaCYP51 proteins.

The yield of the CaCYP51 proteins after purification on Ni²⁺-nitrilotriacetic acid (NTA) agarose varied from 32 nmol per liter of culture (I471T) to 235 nmol per liter (F126V) (Table 1), indicating a 7-fold variability in expression levels in E. coli between recombinant CaCYP51 proteins. The same expression and cell breakage/solubilization conditions were used for all the CaCYP51 proteins. The CaCYP51 amino acid substitutions that gave low yields of native protein gave cell debris pellets after ultracentrifugation that were notably more red-brown in color than the pellets of the wild-type expression clone, suggesting the increased occurrence of inclusion bodies. Optimization of the expression and solubilization conditions for each CaCYP51 protein should increase native protein yields. The absolute absorbance spectra of all the CaCYP51 proteins were similar (Fig. 1; see also Fig. S1 in the supplemental material) with α, β, Soret (γ), and δ spectral bands at 566, 536, 418, and 360 nm, respectively, indicative of low-spin ferric cytochrome P450 enzymes (30, 31). Variability in the Soret peak between the CaCYP51 proteins was only ±1 nm. Dithionite-reduced carbon monoxide difference spectra for all the CaCYP51 proteins yielded a characteristic red-shifted heme Soret peak at ∼448 nm (Fig. S2), indicating that the CYP51 proteins were isolated in their native form. The main difference between the CaCYP51 proteins was the rate at which the heme Soret peak at ∼448 nm formed. There appeared to be little inactivated CaCYP51 (indicated by the appearance of a heme Soret peak at ∼420 nm for P420) present in the purified proteins. For K143R, approximately 15% of the native form of the protein was converted to the P420 inactive form after 6 min, suggesting instability in the presence of sodium dithionite.

Catalytic turnover assayed by using lanosterol as the substrate and in the presence of a cognate human cytochrome P450 reductase (HsCPR) varied between purified CaCYP51 proteins (Table 1), with the highest turnover numbers of 0.896 and 0.919 min⁻¹ being observed with the V456I and Q474K proteins, respectively. In contrast, the G307S+G450E, Y132H+K143R, G307S, and Y118A proteins had the lowest catalytic turnover numbers of 0.044, 0.039, 0.018, and 0.016 min⁻¹, respectively, whereas the value was 0.574 min⁻¹ for wild-type CaCYP51, using HsCPR as the redox partner. Most (20/23) enzymes with CaCYP51 single amino acid substitutions and all enzymes with double substitutions were catalytically less efficient than the wild-type enzyme (Table 1).

Azole IC₅₀ studies.

Several CaCYP51 amino acid substitutions conferred tolerance to inhibition of in vitro CYP51 activity by fluconazole (Table 2). Among the single substitutions, K143R, S279F, and G450E appeared to confer the greatest catalytic tolerance to inhibition by fluconazole (Fig. 2, 3, and 4), yielding values of the concentration causing 50% enzyme inhibition (IC₅₀) 3-fold higher than the IC₅₀ for the wild-type enzyme. The D278N, S405F, and G448E enzymes gave fluconazole IC₅₀ values 2-fold higher than the IC₅₀ for the wild-type enzyme. In addition, significant residual CYP51 activities (3% to 28%) were observed in the presence of 4 μM fluconazole for the enzymes with the Y132F, K143R, F145L, Y257H, D278N, S279F, S405F, Y447H, G448E, G450E, V456I, G464S, R467K, and I471T substitutions (Table 3; Fig. S4), suggesting that these substitutions confer catalytic tolerance to fluconazole. The double substitutions conferred the greatest catalytic tolerance to fluconazole in vitro compared to the IC₅₀ for the wild-type enzyme, with 3-, 10-, 13-, 15-, and 22-fold increased IC₅₀ values for fluconazole (Table 2) being observed with the D278N+G464S, Y132F+F145L, G307S+G450E, Y132F+K143R, and Y132H+K143R enzymes, respectively. This was also reflected in the high residual CYP51 activities of 18% to 75% observed with the double substitution CaCYP51 enzymes in the presence of 4 μM fluconazole (Table 3; Fig. S4). The IC₅₀ experiments indicated that nine single substitutions (Y118A, F126V, Y132H, P230L, G307S, V437I, F449Y, Q474K, and V488I) appeared to confer no increase in catalytic tolerance to fluconazole. The ineffectiveness of the V437I and V488I proteins was expected, as these substitutions occur in both azole-sensitive and azole-resistant C. albicans strains (23).

Based on the IC₅₀ values (Table 2), catalytic tolerance to voriconazole was less pronounced than that to fluconazole among the CaCYP51 proteins, with only 2- to 3-fold increases in IC₅₀s being observed with the Y132F, K143R, D278N, S279F, G307S, G450E, V456I, Y132F+K143R, D278N+G464S, and G307S+G450E enzymes. The greatest tolerance to voriconazole was observed with the Y132F+F145L and Y132H+K143R enzymes, which gave IC₅₀ values 6- and 16-fold higher than the IC₅₀ observed for the wild-type CaCYP51 enzyme. In addition, residual CYP51 activities of over 3% were observed with the K143R, S279F, G450E, Y132F+K143R, Y132H+K143R, Y132F+F145L, and D278N+G464S enzymes in the presence of 4 μM voriconazole (Table 3; Fig. S4), relative to 0.11% residual activity for the wild-type enzyme.

The IC₅₀ values obtained for itraconazole with the CaCYP51 proteins with amino acid substitutions ranged from 0.15 and 0.60 μM (Table 2), whereas the IC₅₀ for the wild-type enzyme was 0.39 μM. However, residual CYP51 activities of 3% to 8% in the presence of 4 μM itraconazole (Table 3; Fig. S4) indicate limited catalytic tolerance to itraconazole due to the K143R, D278N, S279F, G450E, V456I, and D278N+G464S substitutions.

The IC₅₀ values obtained for posaconazole varied between 0.2 μM (for the wild-type enzyme) and 0.93 μM (for the D278N+G464S enzyme) (Table 2), suggesting poor catalytic tolerance to posaconazole among the CaCYP51 proteins. Residual CYP51 activities of 3% to 21% in the presence of 4 μM posaconazole (Table 3; Fig. S4) confirmed the limited catalytic tolerance to posaconazole for the enzymes with the K143R, Y257H, D278N, S279F, S405F, Y447H, G448E, F449Y, G450E, V456I, R467K, I471T, Q474K, and D278N+G464S substitutions. P230L, previously predicted to confer itraconazole and posaconazole resistance (32, 33), did not confer catalytic tolerance to either itraconazole or posaconazole using in vitro CaCYP51 reconstitution assays.

Azole ligand binding.

Fluconazole, voriconazole, itraconazole, and posaconazole bound to all CaCYP51 proteins, producing type II difference spectra indicative of the triazole ring N-4 nitrogen coordinating as the sixth ligand with the heme iron (34) to form the low-spin CYP51-azole complex, resulting in a red shift of the heme Soret peak. The fluconazole binding difference spectra obtained with the CaCYP51 proteins with double substitutions identified a number of differences from the wild-type protein (Fig. 5). Both the D278N+G464S and G307S+G450E proteins gave type II difference spectra that were similar to the spectrum of the wild-type CaCYP51 protein with peaks at 428 to 430 nm and troughs at 412 nm. However, both the Y132F+K143R and Y132H+K143R proteins gave less intense difference spectra (change in absorbance [ΔA]), with a blue-shifted peak at 421 nm and a blue-shifted trough at 367 to 371 nm. Likewise, the Y132F+F145L protein gave a less intense blue-shifted spectrum, with a peak at 423 nm and a broad trough extending from 371 to 411 nm relative to the wild-type protein. These relatively weak blue-shifted type II difference spectra (ΔA) suggest that the Fe-N coordination between the triazole N-4 atom and the heme ferric ion has been weakened (35) because either the orientation of the triazole moiety is unfavorable or the length of the Fe-N bond has been increased. The fluconazole binding difference spectra of the enzymes with the single substitutions Y132F, Y132H, K143R, and F145L (Fig. S3) were all type II difference spectra, albeit with slightly blue-shifted peak maxima at 425 to 427 nm compared to the spectrum of the wild-type protein (429 nm). Therefore, the altered fluconazole difference spectra observed with the Y132F+K143R, Y132H+K143R, and Y132F+F145L enzymes appeared to be a cumulative effect rather than an effect attributable to one particular amino acid substitution.

Ligand binding studies with fluconazole gave dissociation constant (K_d) values ranging from 7 to 90 nM for all CaCYP51 proteins with single amino acid substitutions, except for K143R (K_d, 118 ± 24 [standard deviation] nM), D278N (K_d, 120 ± 35 nM), G307S (K_d, 133 ± 13 nM), and S279F (K_d, 232 ± 36 nM). The double substitutions D278N+G464S, G307S+G450E, Y132F+F145L, Y132F+K143R, and Y132H+K143R gave K_d values for fluconazole of 15 ± 9, 158 ± 38, 202 ± 57, 15,600 ± 3,900, and 18,100 ± 4,800 nM, respectively. The ligand binding studies accurately predicted the fluconazole tolerance conferred by the double substitutions Y132F+K143R and Y132H+K143R but not the catalytic tolerance to fluconazole conferred by Y132F+F145L, G307S+G450E, D278N+G464S, K143R, D278N, S279F, S405F, G448E, and G450E in the IC₅₀ studies.

Preliminary ligand binding studies with voriconazole and the CaCYP51 proteins gave K_d values ranging from 3 to 100 nM, with only the Y132F+K143R and Y132H+K143R enzymes having substantially higher K_d values of 2,300 ± 310 and 8,000 ± 1,200 nM, respectively. Similarly, ligand binding studies for the CaCYP51 proteins with itraconazole and posaconazole gave K_d values ranging from 1 to 70 nM, with the exception of those for enzymes with Y132F+K143R (K_d for itraconazole, 127 ± 33 nM) and Y132H+K143R (K_d for itraconazole, 387 ± 117 nM; K_d for posaconazole, 162 ± 59 nM). Direct ligand binding studies with voriconazole, itraconazole, and posaconazole did not predict the observed catalytic tolerance to these three azoles conferred by several CaCYP51 amino acid substitutions in the IC₅₀ experiments, as evident by the residual activities in the presence of 4 μM triazole antifungal (Table 3; Fig. S4).

In vitro assessment of CaCYP51 substitutions in C. albicans.

In order to determine the impact of CYP51 amino acid substitutions on in vitro azole susceptibility in C. albicans, we selected several mutations representative of those that led to the greatest biochemical effect on the CYP51 enzyme. When mutant alleles were introduced into the fluconazole-susceptible parent strain SC5314, CaCYP51 amino acid substitutions displayed variable effects on the azole drug MIC (Table 4). The effects of nine single CYP51 amino acid substitutions and three double CYP51 amino acid substitutions on the fluconazole, voriconazole, itraconazole, and posaconazole MICs were tested. The G450E and K143R amino acid substitutions produced the greatest relative increase in the MIC of fluconazole, with each producing a 16-fold increase in the MIC compared to that for the reference strain, SC5314, which lacks any substitution in CYP51. The strains containing the CYP51 single amino acid substitutions Y132F, G448E, and G464S also demonstrated resistance to fluconazole, displaying an 8-fold increase in MIC over that for SC5314. The overall strongest effect on the fluconazole MIC occurred with the double substitutions Y132F+K143R and Y132F+F145L, each conferring a 32-fold increase in the fluconazole MIC. Against voriconazole, the single substitutions Y132F, K143R, F145L, and G450E had the greatest effect on the MIC, and the strongest in vitro effect was observed with the Y132F+K143R and Y132F+F145L double substitutions. When tested against itraconazole, most of the strains with a CYP51 substitution showed a small increase in MIC, with the strain with the strongest single amino acid substitution, K143R, displaying an MIC (0.5 μg ml⁻¹) greater than that for any of the strains with the double amino acid substitutions. In contrast to the other azoles, posaconazole appeared to be very effective against all CYP51 mutant strains, as the MICs for the reference strain and all other strains tested were identical, with the exception of the strain containing the single G448E substitution, which displayed a slight 2-fold increase in MIC over that for SC5314.

Construction of C. albicans CYP51 mutant strains.

The strains used in this study are shown in Table 5. Strains expressing seven single mutations (Y132F, K143R, F145L, S405F, G448E, G450E, G464S) and three double mutations (Y132F+K143R, Y132F+F145L, D278N+G464S) were selected from a previous study (36). Two additional strains expressing the Y132H and D278N CYP51 gene mutations were created utilizing the SAT1/FLP cassette described previously (44). Briefly, C. albicans isolate SC5314 was transformed via electroporation with inserts of a pSFS2-derived plasmid possessing the SAT1 nourseothricin resistance marker, FLP recombinase, and mutant CaCYP51 genes of interest. The 5′ end of the DNA fragment contained either the Y132H or D278N CYP51 open reading frame (ORF), and the 3′ end possessed homology downstream of the CaCYP51 stop codon to ensure homologous recombination at the CYP51 gene locus. Successful transformants were screened on yeast-peptone-dextrose agar plates containing nourseothricin and confirmed via Southern hybridization. Genomic DNA from nourseothricin-screened transformants was digested with the HindIII restriction enzyme, and the resulting DNA digests were loaded on 1% agarose gels containing ethidium bromide. Digested samples were visualized briefly under UV light after approximately 2 h of electrophoresis (120 V) to confirm appropriate band separation. Lanes containing digested DNA in the gel were transferred onto a nitrocellulose membrane via vacuum suction, and DNA was cross-linked to the membrane through UV exposure. PCR amplification of the genomic DNA from parent isolate SC5314 using primers ERG11_Probe_F (5′-AGTTCAATGGTGGTTTTACCTACT-3′) and ERG11_Probe_R (5′-ATTTCTGATTGAGTCATCCTAACA-3′) generated a 254-bp product, used as a probe for the ERG11 gene locus. The probe was labeled and prepared using an Amersham AlkPhos direct labeling and detection system (GE Life Sciences) and hybridized to the membrane through overnight incubation at 55°C. Bands containing the hybridized, alkaline phosphatase-labeled probe corresponding to either wild-type ERG11 alleles (2,678 bp) or successfully replaced alleles with the SAT1/FLP cassette included (4,318 bp) or recycled (3,119 bp) were visualized with the CDP-Star chemiluminescent detection reagent. To confirm the sequence of the replaced ERG11 allele, a DNA fragment containing the ERG11 ORF and approximately 435 bp upstream of the start codon and 229 bp downstream of the ERG11 stop codon was PCR amplified using primers CaERG11_A_(ApaI) (5′-GGGCCCGGGTTATTTGAGAACAGCC-3′) and CaERG11_E_(XhoI) (5′-CTCGAGCCAGTGGACAAAAACCATCA-3′). The resulting DNA fragment was purified and Sanger sequenced on an ABI model 3130XL genetic analyzer with the following sequencing primers: CaERG11SeqA (5′-GCCACCACACCCTATGGCTATT-3′), CaERG11SeqB (5′-TATTTTCACTGCTTCAAGATCT-3′), CaERG11SeqC (5′-CCAAAAGGTCATTATGTTTTAG-3′), CaERG11SeqD (5′-CATACAAGTTTCTCTTTTTTCC-3′), CaERG11SeqE (5′-CATTTAGGTGAAAAACCTCATT-3′), and CaERG11SeqF (5′-TACTCCAGTTTTCGGTAAAGGG-3′).

Susceptibility testing of C. albicans CYP51 mutant strains.

Azole MICs were determined using the broth microdilution method described by the Clinical Laboratory and Standards Institute (45, 46). Cells were incubated in 96-well microtiter plates containing RPMI 1640 medium and serially diluted concentrations of either fluconazole, itraconazole, voriconazole, or posaconazole. The concentrations for MIC determinations ranged from 0.125 to 64 μg ml⁻¹ for fluconazole and 0.03 to 16 μg ml⁻¹ for itraconazole, voriconazole, and posaconazole. Measurements were read visually at 24 h after incubation at 35°C for a 50% reduction in growth from that in the drug-free control wells. MIC measurements were performed in duplicate, and the MIC was reported as the higher of two values for all strains tested. In cases where the duplicate MIC values were not identical, the higher of the two MIC values is reported. Overall, the MIC values were consistent with 92% (184/200) of the MICs being within a single dilution of each other for a given strain and drug combination. Additionally, no MICs differed by more than 2 dilutions for a given strain and drug.

Construction of pCWori⁺:CaCYP51 expression vectors for expression in E. coli.

Wild-type C. albicans CYP51 (UniProtKB accession number P10613) and 28 mutant C. albicans CYP51 genes, 23 coding for single amino acid substitutions and 5 coding for double substitutions, were synthesized by Eurofins-MWG-Operon (Ebersberg, Germany) and inserted into a shuttle vector with kanamycin selection using NdeI and HindIII cloning sites flanking the start and stop codons, respectively. All the CaCYP51 genes were modified so that the first 8 amino acids encoded MALLLAVF to facilitate expression in E. coli (47), and a 12-base sequence (CATCACCATCAC) encoding a 4-histidine tag was inserted immediately in front of the stop codon to facilitate protein purification by Ni²⁺-NTA agarose affinity chromatography. All 29 genes were codon optimized for expression in E. coli.

The CaCYP51 genes were excised by digestion with NdeI and HindIII, gel purified, and then ligated into the pCWori⁺ expression vector that had previously been cut with NdeI and HindIII. Positive recombinants were isolated by transformation into E. coli strain DH5α and selected on LB agar plates containing 0.1 mg ml⁻¹ sodium ampicillin. The sequence integrity of each pCWori⁺:CaCYP51 construct was checked by DNA sequencing in both directions (Eurofins-MWG-Operon) at this stage and then again from a plasmid preparation derived from the overnight cultures used to inoculate the protein expression culture flasks.

The CaCYP51 single amino acid substitutions synthesized were Y118A, F126V, Y132F, Y132H, K143R, F145L, P230L, Y257H, D279N, S279F, G307S, S405F, V437I, Y447H, G448E, F449Y, G450E, V456I, G464S, R467K, I471T, Q474K, and V488I, and the double substitutions were Y132F+K143R, Y132H+K143R, Y132F+F145L, D278N+G464S, and G307S+G450E.

Heterologous expression in E. coli and isolation of recombinant CaCYP51 proteins.

The pCWori⁺:CaCYP51 constructs were transformed into competent E. coli DH5α cells, and transformants were selected using 0.1 mg ml⁻¹ sodium ampicillin. One-liter volumes of Terrific broth (1.2% tryptone, 2.4% yeast extract, 0.4% glycerol, 0.09 M potassium phosphate, pH 7.5) containing 0.1 mg ml⁻¹ sodium ampicillin in 2-liter plastic baffled flasks were inoculated with 20 ml of an overnight culture of the expression clones. The cultures were shaken at 200 rpm and 37°C for 7 h, followed by induction with 1 mM isopropyl β-d-1-thiogalactopyranoside and 1 mM the heme precursor 5-aminolevulinic acid, prior to shaking for a further 16 h at 160 rpm and 27°C. Protein isolation was done according to the method of Arase et al. (48), except that 2% sodium cholate was used in the sonication buffer. The solubilized CYP51 proteins were purified by affinity chromatography using Ni²⁺-NTA agarose as previously described (30), with the modification that 0.1% l-histidine in 0.1 M Tris-HCl (pH 8.1) and 25% glycerol were used to elute nonspecifically bound E. coli proteins after the salt washes. The CaCYP51 proteins were recovered by elution with 1% l-histidine in 0.1 M Tris-HCl (pH 8.1) and 25% glycerol, followed by dialysis against 5 liters of 25 mM Tris-HCl (pH 8.1) and 10% glycerol. The Ni²⁺-NTA agarose-purified CaCYP51 proteins were used for all subsequent spectral and IC₅₀ determinations. Protein purity was assessed by SDS-polyacrylamide gel electrophoresis.

Determination of cytochrome P450 protein concentrations.

Reduced carbon monoxide difference spectra (49) were used to determine the cytochrome P450 concentration, with carbon monoxide being passed through the cytochrome P450 solution prior to addition of sodium dithionite to the sample cuvette. An extinction coefficient of 91 mM⁻¹ · cm⁻¹ (50) was used to calculate the cytochrome P450 concentrations from the absorbance difference between 445 to 447 nm and 490 nm. Absolute spectra at wavelengths between 700 and 300 nm were determined using 20-fold dilutions of the purified CaCYP51 proteins in 0.1 M Tris-HCl (pH 8.1) and 20% glycerol as previously described (30). Confirmation that the isolated CaCYP51 proteins were active was demonstrated by measuring the 14α-demethylation of lanosterol using the CYP51 reconstitution assay detailed below.

CYP51 reconstitution assay system.

IC₅₀ determinations were performed using the CYP51 reconstitution assay system (final reaction volume, 500 μl) previously described (51) containing 1 μM CaCYP51, 2 μM Homo sapiens cytochrome P450 reductase (UniProtKB accession number P16435), 50 μM lanosterol, 50 μM sn-dilaurylphosphatidylcholine, 4% (wt/vol) 2-hydroxypropyl-β-cyclodextrin, 0.4 mg ml⁻¹ isocitrate dehydrogenase, 25 mM trisodium isocitrate, 50 mM NaCl, 5 mM MgCl₂, and 40 mM MOPS (morpholinepropanesulfonic acid; pH ∼7.2). Azole antifungal agents were added in 2.5 μl dimethyl sulfoxide, followed by 10 min of incubation at 37°C prior to assay initiation with 4 mM β-NADPH-Na₄. The samples were then shaken for 15 min at 37°C. Sterol metabolites were recovered by extraction with ethyl acetate, followed by derivatization with 0.1 ml N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA)–trimethylchlorosilane (TMCS) (99:1) and 0.3 ml anhydrous pyridine (2 h at 80°C), prior to analysis by gas chromatography-mass spectrometry (GC-MS) (52). IC₅₀ determinations were performed in duplicate.

Azole binding studies.

Studies of azole antifungal binding to 4 μM CaCYP51 proteins were performed as previously described (53) using quartz split cuvettes with a 4.5-mm light path. Stock 1-, 0.5-, 0.2-, and 0.1-mg ml⁻¹ solutions of fluconazole, itraconazole, voriconazole, and posaconazole were prepared in dimethyl sulfoxide. Azole antifungals were progressively titrated against the CaCYP51 proteins in 0.1 M Tris-HCl (pH 8.1) and 25% glycerol at 22°C, with equivalent volumes of dimethyl sulfoxide also being added to the CYP51-containing compartment of the reference cuvette. The absorbance difference spectra between 500 and 350 nm were determined after each incremental addition of azole, with binding saturation curves constructed from the change in the peak and trough absorbance (ΔA_peak-trough) against the azole concentration. The dissociation constant (K_d) of the enzyme-azole complex for each azole was determined by nonlinear regression (Levenberg-Marquardt algorithm) using a rearrangement of the Morrison equation for tight ligand binding (54, 55). Tight binding is normally observed when the K_d for a ligand is similar to or lower than the concentration of the enzyme present (56). Where ligand binding was weak, the Michaelis-Menten equation was used to fit the data. Each binding determination was performed in triplicate. The chemical structures of the azole antifungals used in this study are shown in Fig. S5 in the supplemental material.

Data analysis.

Curve fitting of the ligand binding data was performed using the computer program ProFit (version 6.1.12; QuantumSoft, Zurich, Switzerland). Spectral determinations were made using quartz semimicrocuvettes with a Hitachi U-3310 UV/visible spectrophotometer (San Jose, CA).

Chemicals.

All chemicals, including all azole antifungals except voriconazole, were obtained from Sigma Chemical Company (Poole, UK). Voriconazole was supplied by Discovery Fine Chemicals (Bournemouth, UK). Growth media, sodium ampicillin, IPTG, and 5-aminolevulenic acid were obtained from Foremedium Ltd. (Hunstanton, UK). The Ni²⁺-NTA agarose affinity chromatography matrix was obtained from Qiagen (Crawley, UK).