Steroid Hydroxylation by Basidiomycete Peroxygenases: a Combined Experimental and Computational Study

Abstract

The goal of this study is the selective oxyfunctionalization of steroids under mild and environmentally friendly conditions using fungal enzymes. With this purpose, peroxygenases from three basidiomycete species were tested for the hydroxylation of a variety of steroidal compounds, using H₂O₂ as the only cosubstrate. Two of them are wild-type enzymes from Agrocybe aegerita and Marasmius rotula, and the third one is a recombinant enzyme from Coprinopsis cinerea. The enzymatic reactions on free and esterified sterols, steroid hydrocarbons, and ketones were monitored by gas chromatography, and the products were identified by mass spectrometry. Hydroxylation at the side chain over the steroidal rings was preferred, with the 25-hydroxyderivatives predominating. Interestingly, antiviral and other biological activities of 25-hydroxycholesterol have been reported recently (M. Blanc et al., Immunity 38:106–118, 2013, http://dx.doi.org/10.1016/j.immuni.2012.11.004). However, hydroxylation in the ring moiety and terminal hydroxylation at the side chain also was observed in some steroids, the former favored by the absence of oxygenated groups at C-3 and by the presence of conjugated double bonds in the rings. To understand the yield and selectivity differences between the different steroids, a computational study was performed using Protein Energy Landscape Exploration (PELE) software for dynamic ligand diffusion. These simulations showed that the active-site geometry and hydrophobicity favors the entrance of the steroid side chain, while the entrance of the ring is energetically penalized. Also, a direct correlation between the conversion rate and the side chain entrance ratio could be established that explains the various reaction yields observed.

INTRODUCTION

Steroids represent an important class of natural compounds that are widespread in nature and have a multitude of pharmacological properties. Indeed, steroids are ranked among the most marketed medical products and represent the second largest category next to antibiotics. The physiological activity of steroids depends on their structures, including the oxidation state of the rings, and the type, number, and regio- and stereoposition of the functional groups attached (1). It is known that even minor changes in the structure of steroids can highly affect their biological activity, which has promoted countless studies on the modification of naturally occurring steroids in search of new and more active compounds (2). Among these modifications, hydroxylation is one of the most important reactions in steroid oxyfunctionalization.

Hydroxylation serves to increase the polarity of the rather hydrophobic steroids, and hydroxylated steroids often express a higher level of biological activity than their less polar nonhydroxylated analogs. Moreover, hydroxylation can be used to develop intermediates for further synthesis by offering access to otherwise inaccessible sites of the steroid molecule. These modifications have been obtained mostly by applying microorganisms (1), with only a few examples where isolated enzymes were used. Members of the cytochrome P450 monooxygenase (P450) superfamily are notable examples of such enzymatic catalysts, since, because of their catalytic versatility, they would perfectly meet the requirements of chemical synthesis (3, 4). However, large-scale applications are not yet feasible due to the intrinsic properties of these oxygenases (5), whose activation requires NAD(P)H as an electron donor and auxiliary flavin-reductases (or a second flavin domain) for electron transfer to O₂ (6).

A few years ago, a new peroxidase type was discovered in the wood-dwelling basidiomycete Agrocybe aegerita (7), which turned out to be a true peroxygenase efficiently transferring oxygen from peroxide to various organic substrates (8). This peroxygenase is able to catalyze reactions formerly assigned only to P450s (9). However, unlike P450s, which are intracellular enzymes whose activation often requires an auxiliary enzyme and a source of reducing power, the A. aegerita enzyme is a secreted protein; therefore, it is far more stable and requires only H₂O₂ for activation (8). The A. aegerita peroxygenase has been shown to catalyze numerous interesting oxygenation reactions on aromatic compounds (10), and recently, the action of this enzyme on aliphatic compounds was demonstrated, expanding its biotechnological interest (11, 12). The enzyme, first described as a haloperoxidase and later mostly referred to as aromatic peroxygenase (APO), now is named “unspecific peroxygenase” (UPO; EC 1.11.2.1).

After the first peroxygenase of A. aegerita (7), similar enzymes also have been found in other basidiomycetes, such as Coprinellus radians (13) and Marasmius rotula (14), and there are indications for their widespread occurrence in the fungal kingdom (15). Recently, over 100 peroxygenase-type genes (encoding enzymes of the heme-thiolate peroxidase superfamily) have been identified during the analysis of 24 basidiomycete genomes (16), including Coprinopsis cinerea (17). The wild-type C. cinerea peroxygenase has not been isolated to date from this fungus, but one of the peroxygenase genes from its genome (protein model 7249, available from the Joint Genome Institute [JGI; http://genome.jgi.doe.gov/Copci1]) was heterologously expressed by Novozymes A/S (Bagsvaerd, Denmark). This first recombinant peroxygenase can be a powerful biocatalyst for synthetic applications because of the potentially high expression yield and the possibility of tuning its catalytic and operational properties using protein engineering tools. Recently, the ability of this recombinant peroxygenase in the oxyfunctionalization of several aliphatic compounds was demonstrated (18).

Here, the efficiency of three different peroxygenases, namely, the wild enzymes from A. aegerita (AaeUPO) and M. rotula (MroUPO) and the recombinant enzyme from C. cinerea (rCciUPO), in oxyfunctionalization of a variety of steroidal compounds (including free and esterified sterols, steroid hydrocarbons, and ketones) is evaluated for the first time. All three fungi are agaric basidiomycetes, but they belong to different families of the order Agaricales, and they also differ to some extent from an ecophysiological point of view. M. rotula (family Marasmiaceae) is a fungus that stands between white rotters and litter decomposers, because it preferably colonizes small twigs of broad-leaved trees (Fagus, Quercus, and Acer). C. cinerea (family Psathyrellaceae) naturally dwells on older dung and in soils rich in organic nitrogen (i.e., it is a specialized litter decomposer). A. aegerita (family Bolbitaceae) is a wood-dwelling fungus that causes an unspecific white rot on trunks and stumps of broad-leaved trees (e.g., Populus and Acer species). However, it would be daring to conclude from these evolutionary and ecological differences that the UPOs of these fungi must be different.

In addition to the experimental assays, where the products from the reaction of a variety of steroids with the three above-mentioned peroxygenases were analyzed by gas chromatography-mass spectrometry (GC-MS), a set of representative molecules was selected for computational simulations with AaeUPO (for which a crystal structure is available) to get further insights into the molecular determinants that affect the reactivity of the different steroid types. All simulations followed an identical protocol where, after appropriate preparation of the protein with heme as compound I [the two-electron oxidized cofactor containing an Fe(IV)=O porphyrin cation radical complex], each substrate was placed at the entrance of the heme access channel, still in the solvent. From there the substrate explored both the entrance pathway and the active site using Protein Energy Landscape Exploration (PELE) software (19, 20). From the information provided by the energy profiles and trajectories, analyses of the effect of the structural differences between the substrates and their peroxygenase reactivity could be rationalized.

MATERIALS AND METHODS

Enzymes.

rCciUPO was provided by Novozymes A/S (Bagsvaerd, Denmark). This recombinant enzyme corresponds to the protein model 7249 from the sequenced C. cinerea genome, available at the JGI (http://genome.jgi.doe.gov/Copci1), expressed in Aspergillus oryzae (40). The protein was purified using a combination of S-sepharose and SP-sepharose ion-exchange chromatography. The recombinant peroxygenase preparation is an electrophoretically homogeneous glycoprotein with a molecular mass of around 44 kDa (a nonuniform glycosylation pattern was observed), a typical UV-visible (UV-Vis) spectrum with a Soret band at 418 nm, and the ability to oxygenate different aromatic compounds with a specific activity of approximately 100 U · mg⁻¹ (measured as described below).

The AaeUPO and MroUPO enzymes are two wild-type peroxygenases isolated from cultures of A. aegerita DSM 22459 and M. rotula DSM 25031, deposited at the German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). AaeUPO was produced in suspensions of soybean meal and purified by several steps of fast protein liquid chromatography (FPLC) using different ion exchangers (SP-sepharose, MonoQ, and MonoS) with size exclusion chromatography (Superdex 75) as the final isolation step. The final protein fraction had a molecular mass of 46 kDa with one enriched isoform (AaeUPO II). This isoform was used in the present study (7, 21). MroUPO was purified by FPLC to apparent homogeneity, confirmed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) under denaturing conditions, and showed a molecular mass of 32 kDa and an isoelectric point of pH 5.0 to 5.3. The UV-visible spectra of the enzymes showed a characteristic maximum around 420 nm (Soret band of heme-thiolate proteins) (14).

All media and columns used for enzyme isolation were purchased from GE Healthcare Life Sciences. One UPO activity unit is defined as the amount of enzyme oxidizing 1 μmol of veratryl alcohol to veratraldehyde (ε₃₁₀, 9,300 M⁻¹ · cm⁻¹) in 1 min at 24°C, pH 7 (the optimum for peroxygenase activity), after addition of H₂O₂ (0.5 mM H₂O₂ in rCciUPO reaction mixtures and 2.5 mM H₂O₂ in AaeUPO and MroUPO reaction mixtures).

Steroids.

The model steroid compounds (from Sigma-Aldrich) used in the enzymatic reactions include (i) free sterols, such as cholesterol, campesterol, ergosterol, sitosterol, and stigmasterol; (ii) steroid ketones, such as cholestan-3-one, 4-cholesten-3-one, cholesta-3,5-dien-7-one, and testosterone; (iii) steroid hydrocarbons, such as cholestane, cholesta-3,5-diene, and pregnane; and (iv) sterol esters, such as cholesteryl acetate, cholesteryl butyrate, and cholesteryl caprylate (see Fig. 1).

FIG 1.

Chemical structures of the different steroid compounds tested as substrates of the peroxygenases, including (i) free sterols, such as cholesterol (A), campesterol (B), ergosterol (C), sitosterol (D), and stigmasterol (E); (ii) steroid ketones, such as cholestan-3-one (F), 4-cholesten-3-one (G), cholesta-3,5-dien-7-one (H), and testosterone (I); (iii) steroid hydrocarbons, such as cholestane (J), cholesta-3,5-diene (K), and pregnane (L); and (iv) sterol esters, such as cholesteryl acetate (M), cholesteryl butyrate (N), and cholesteryl caprylate (O).

Enzymatic reactions.

Reactions of the model steroids (0.05 mM concentration) with the three peroxygenases (1 U) were performed in 5-ml vials containing 50 mM sodium phosphate (pH 7 in AaeUPO and rCciUPO reactions and pH 5.5 in MroUPO reactions) at 40°C and a 60-min reaction time in the presence of H₂O₂ (0.5 mM in rCciUPO reactions and 2.5 mM in AaeUPO and MroUPO reactions). Prior to use, the substrates were dissolved in acetone and added to the buffer to give a final acetone concentration of 40% (vol/vol) in most cases. In control experiments, substrates were treated under the same conditions (including 0.5 mM and 2.5 mM H₂O₂) but without enzyme. Products were recovered by liquid-liquid extraction with methyl tert-butyl ether and dried under N₂. N,O-Bis(trimethylsilyl)trifluoroacetamide (Supelco) was used to prepare trimethylsilyl (TMS) derivatives that were analyzed by GC-MS.

GC-MS analyses.

The GC-MS analyses were performed with a Varian 3800 chromatograph coupled to an ion-trap detector (Varian 4000) using a medium-length fused-silica DB-5HT capillary column (12 m by 0.25 mm internal diameter, 0.1 μm film thickness) from J&W Scientific (22). The oven was heated from 120°C (1 min) to 300°C (15 min) at 10°C · min⁻¹. The injector was programmed from 60°C (0.1 min) to 300°C (28 min) at 200°C · min⁻¹. The transfer line was kept at 300°C, and helium was used as the carrier gas (2 ml · min⁻¹). For some analyses, a Shimadzu GC-MS QP2010 Ultra with a fused-silica DB-5HT capillary column (30 m by 0.25 mm internal diameter, 0.1 μm film thickness) also was used. The oven was heated from 120°C (1 min) to 300°C (15 min) at 5°C · min⁻¹. The injection was performed at 300°C, and the transfer line was kept at 300°C. Compounds were identified by mass fragmentography and comparing their mass spectra with those of the Wiley and NIST libraries and standards. Quantification was obtained from total-ion peak area, using response factors of the same or similar compounds. Data from replicates were averaged, and in all cases (substrate conversion and relative abundance of reaction products) the standard deviations were below 3.5% of the mean values.

PELE computational analyses.

For the computational study, eight representative compounds (cholesterol, sitosterol, cholestan-3-one, cholesta-3,5-dien-7-one, cholestane, cholesta-3,5-diene, cholesteryl acetate, and cholesteryl caprylate) were prepared and PELE simulations performed.

The starting structure for all PELE simulations was the AaeUPO 2.19-Å-resolution crystal (PDB entry 2YOR) (23). As the optimum pH for peroxygenase activity is 7, the structure was prepared accordingly using Schrodinger's Protein Preparation Wizard (24) and H++ web server (25). Histidines were δ-protonated, with the exception of His-82 (ε-protonated) and His-118 and His-251 (double protonated). All acidic residues were deprotonated, except Asp-85, which was kept in its protonated state. The eight ligands studied in this section were optimized with Jaguar 7.8 (Schrödinger, LLC, New York, NY) at the DFT/M06 level with the 6-31G** basis and a PBF implicit solvent in order to obtain their electrostatic potential atomic charges. Finally, the heme site was modeled as compound I after being fully optimized in the protein environment with quantum mechanics/molecular mechanics (QM/MM) using QSite 5.7 (Schrödinger).

Once the initial protein structure was prepared, the optimized ligands were placed manually in identical positions at the entrance of the protein's binding pocket and PELE (20) simulations were performed. PELE is a Monte Carlo-based algorithm that produces new configurations through a sequential ligand and protein perturbation, side-chain prediction, and minimization steps. New configurations then are filtered with a Metropolis acceptance test, where the energy is described with an all-atom OPLS force field (26) and a surface-generalized Born solvent (27). In this way it is possible to locate and characterize local and global minima structures for the most favorable protein-ligand interactions. PELE has been used successfully in a number of ligand migration studies with both small and large substrates (28–30). In this work, PELE was set up first to drive the ligands inside the protein until the center of mass was less than 10 Å from the compound I oxygen. After that, simulation proceeded by allowing a free exploration of the enzyme active site by the ligand. The results presented are based on ∼350 48-h trajectories for each ligand.

RESULTS

The efficiency of three peroxygenases (AaeUPO, MroUPO, and rCciUPO) hydroxylating 15 steroidal compounds (including free and esterified sterols, steroid hydrocarbons, and ketones) was experimentally evaluated, and the reaction products were analyzed by GC-MS. To rationalize the results obtained with the different steroid types, substrate diffusion simulations were performed with PELE (20), providing energy profiles and ligand trajectories, as described in the final section.

Free sterol reactions.

Free C₂₇ (cholesterol), C₂₈ (campesterol and ergosterol), and C₂₉ (sitosterol and stigmasterol) sterols (Fig. 1A to E, respectively) were tested as UPO substrates (Table 1). Cholesterol was converted to a larger extent than the other sterols, although the individual enzymes showed differences in their performance (after 60 min): rCciUPO completely (100%) transformed the substrate, followed by AaeUPO (64%) and MroUPO (23%). The reaction products of cholesterol were oxygenated derivatives at the C-24, C-25, and C-26/C-27 positions of the side chain (Fig. 1 depicts atom numbering), with the monohydroxylated C-25 derivative predominating (Table 2). Interestingly, a strict regioselectivity was observed in the reaction of cholesterol with rCciUPO, yielding 25-hydroxycholesterol as the sole reaction product.

TABLE 1.

Conversion of different types of steroid compounds by the rCciUPO, AaeUPO, and MroUPO peroxygenases within a 60-min reaction

Substrate	% substrate transformed by:
	rCciUPO	AaeUPO	MroUPO
Free sterols
Cholesterol	100	64	23
Campesterol	30	47	10
Ergosterol	6	10	7
Sitosterol	6	13	4
Stigmasterol	2	2	5
Steroid ketones
Cholestan-3-one	54	42	12
4-Cholesten-3-one	100	67	14
Cholesta-3,5-dien-7-one	97	57	39
Testosterone	0	0	0
Steroid hydrocarbons
Cholestane	14	3	1
Cholesta-3,5-diene	34	18	23
Pregnane	0	0	37
Sterol esters
Cholesteryl acetate	34	16	5
Cholesteryl butyrate	7	1	1
Cholesteryl caprylate	0	0	0

TABLE 2.

Abundance of the different oxygenated derivatives identified by GC-MS in the reactions of free sterols with rCciUPO, AaeUPO, and MroUPO peroxygenases

Derivative	Abundance (relative %) of derivative with:
	24-OH	25-OH	26/27-OH	28-OH	26/27-COOH
Cholesterol
rCciUPO		100
AaeUPO	6	80	9		5
MroUPO		89	6		5
Campesterol
rCciUPO		85	15
AaeUPO		99	1
MroUPO		98	2
Ergosterol
rCciUPO		100
AaeUPO		100
MroUPO		100
Sitosterol
rCciUPO		52		48
AaeUPO		94		6
MroUPO		100
Stigmasterol
rCciUPO		100
AaeUPO		100
MroUPO		100

The position of the hydroxyl group was determined by the mass spectra of the TMS derivatives, as found for 25-hydroxycholesterol (see Fig. S1 in the supplemental material). The spectrum includes the molecular ion (m/z 546) and a characteristic ion at m/z 131, resulting from C-24–C-25 bond cleavage, together with additional fragments. Likewise, the spectra of monohydroxylated derivatives at C-24 and C-26 (not shown) showed, in addition to the molecular ion at m/z 546, the characteristic fragment at m/z 131, described above, and that at m/z 145 from the C-23–C-24 bond cleavage. In the AaeUPO and MroUPO reactions, further oxidized compounds (carboxycholesterol) at C-26/C-27 were identified (with a molecular ion at m/z 560 and the above-described characteristic fragment at m/z 145), although they were in small amounts.

The C₂₈ sterols, especially ergosterol, were transformed to a lesser extent than cholesterol (Table 1). AaeUPO showed more reactivity toward campesterol and ergosterol than the other peroxygenases. The reaction of ergosterol was strictly regioselective, with 25-hydroxyergosterol being the only reaction product identified in all cases (Table 2). The reaction with campesterol also was highly regioselective, giving the 25-hydroxyderivative as the main product, although hydroxylation at C-26 also was observed (especially in the rCciUPO reaction). Likewise, the three peroxygenases showed low reactivity with the C₂₉ sterols, and only very low conversion rates (<13%) were attained (Table 1). In the reaction with stigmasterol, only the monohydroxylated derivative at C-25 was observed, whereas with sitosterol, nearly the same amount of 28-hydroxysitosterol was formed by rCciUPO (and a small amount by AaeUPO) (Table 2).

Steroid ketone reactions.

Cholestan-3-one, 4-cholesten-3-one, cholesta-3,5-dien-7-one, and testosterone (Fig. 1F to I, respectively) were tested as UPO substrates (Table 1). rCciUPO was the most efficient enzyme in transforming steroid ketones except testosterone, which was not modified by any of the three peroxygenases, followed by AaeUPO and MroUPO. The unsaturated 4-cholesten-3-one and cholesta-3,5-dien-7-one were completely transformed by rCciUPO, and the saturated cholestan-3-one also was significantly transformed (54%).

The three ketones were predominantly monohydroxylated at C-25 (Table 3). In the case of cholestan-3-one, the regioselectivity was strict, since only the 25-monohydroxylated derivative was formed by the three enzymes, and the same was observed in the reactions of 4-cholesten-3-one with rCciUPO and MroUPO. However, in the reaction with AaeUPO, monohydroxylation at the 4-cholesten-3-one terminal positions (C-26/C-27) also was observed, although as a minor proportion. Interestingly, in the reaction of the diunsaturated cholesta-3,5-dien-7-one, hydroxylation in the steroidal ring also was produced, being especially pronounced in the MroUPO reaction.

TABLE 3.

Abundance of the different oxygenated derivatives identified by GC-MS in the reactions of steroid ketones with rCciUPO, AaeUPO, and MroUPO peroxygenases

Derivative	Abundance (relative %) of derivative with:
	25-OH	26/27-OH	x-OH	x,25-diOH
Cholestan-3-one
rCciUPO	100
AaeUPO	100
MroUPO	100
4-Cholesten-3-one
rCciUPO	100
AaeUPO	93	7
MroUPO	100
Cholesta-3,5-dien-7-one
rCciUPO	93			7
AaeUPO	87	5	4	4
MroUPO	56		39	5

The mass spectra of the TMS derivatives of the three 25-monohydroxylated ketones (not shown) showed the above-described characteristic fragment at m/z 131 (from C-24–C-25 bond cleavage). No molecular ions for the saturated and monounsaturated ketones were observed, although the molecular masses could be determined from the [M-15]⁺ fragments (at m/z 459 and m/z 457, respectively). However, the mass spectrum of the 25-hydroxyderivative of the diunsaturated ketone showed a molecular ion (m/z 470) in addition to the m/z 131 and [M-15]⁺ fragments. Likewise, the 26-hydroxyderivative of 4-cholesten-3-one showed a molecular ion (m/z 472) in addition to the m/z 131 and [M-15]⁺ fragments. The diunsaturated cholesta-3,5-dien-7-one yielded the mono- and dihydroxyderivatives, with one hydroxylation at the steroidal ring. The most prominent fragment in the mass spectrum of the monohydroxylated derivative (see Fig. S2A in the supplemental material) corresponded to the molecular ion at m/z 470. Other characteristic fragments were observed, but the position of the hydroxyl group in the ring could not be determined. The mass spectrum of the dihydroxylated derivative (see Fig. S2B) showed the characteristic fragment at m/z 131. Additionally, a molecular ion (m/z 558) and other characteristic fragments were found, but, like in the monohydroxylated derivative, the position of the hydroxyl group in the ring could not be determined.

Steroid hydrocarbon reactions.

Cholestane, cholesta-3,5-diene, and pregnane (Fig. 1J to L, respectively) were tested as UPO substrates (Table 1). Cholestane, and especially cholesta-3,5-diene, were transformed by the three peroxygenases, with rCciUPO being the most efficient one. In contrast, pregnane was modified only by MroUPO.

The reaction products of these hydrocarbons are shown in Table 4. Whereas the reaction of cholestane was 100% regioselective at C-25, cholesta-3,5-diene gave several di- and trihydroxylated derivatives at the steroidal ring, in addition to the 25-hydroxyderivatives (mainly formed by rCciUPO). Among the dihydroxylated derivatives at the ring core, the ones with hydroxylation at C-3 and C-6 are noteworthy (Fig. 2A and B). In these reactions, the saturation of one double bond and displacement of the other one (to C-4) took place. Therefore, the 3,5-diene structure in the substrate was transformed into a 3,6-dihydroxy-4-ene structure. Likewise, the C-3 and C-4 dihydroxylated derivatives lost the double bond at C-3, yielding a 3,4-dihydroxy-5-ene structure. The reaction of pregnane with MroUPO gave several monohydroxylated derivatives at the steroid ring (97% of reaction products), together with a further oxidized derivative (3% 3-ketopregnane) (see Fig. S3 in the supplemental material), although the position of the hydroxyl group could not be determined.

TABLE 4.

Abundance of the different oxygenated derivatives identified by GC-MS in the reactions of steroid hydrocarbons with rCciUPO, AaeUPO, and MroUPO peroxygenases

Derivative	Abundance (relative %) of derivative with:
	25-OH	x-OH	x,25-diOH	3,4-diOH-5-en	3,6-diOH-4-en	3,6,25-triOH-4-en
Cholestane
rCciUPO	100
AaeUPO	100
MroUPO	100
Cholesta-3,5-diene
rCciUPO	36		14	3	18a	29
AaeUPO	3		20	35	31b	11
MroUPO			1	36	61c	2
Pregnane
rCciUPO
AaeUPO
MroUPO		100d

^aIncluding 3α,6α-diOH-4-en (14%) and 3β,6β-diOH-4-en (4%).

^bIncluding 3α,6α-diOH-4-en (28%) and 3β,6β-diOH-4-en (3%).

^cIncluding 3α,6α-diOH-4-en (50%) and 3β,6β-diOH-4-en (11%).

^dIncluding 97% hydroxy and 3% keto derivatives (see Fig. S3 in the supplemental material).

FIG 2.

Mass spectra of 3,6-dihydroxycholest-4-ene (A), 3,6,25-trihydroxycholest-4-ene (B), and 3,4-dihydroxycholest-5-ene (C) from peroxygenase reactions, with cholesta-3,5-diene (structure K in Fig. 1) as a TMS derivative.

The mass spectra of the 25-hydroxycholestane and 25-hydroxycholesta-3,5-diene TMS derivatives (not shown) showed the characteristic fragment at m/z 131, as well as other characteristic ([M-15]⁺, [M-90]⁺, and [M-90-15]⁺) fragments. The mass spectrum of 3,6-dihydroxycholest-4-ene (Fig. 2A), whose formation is described above, showed a molecular ion (m/z 546) and a base peak at m/z 403 considered characteristic of the 4-ene-6-hydroxy structure (31) together with other fragments (including [M-15]⁺, [M-90]⁺, and [M-15-90]⁺). The mass spectrum of 3,6,25-trihydroxycholest-4-ene (Fig. 2B) showed a molecular ion (m/z 634) and the characteristic fragment at m/z 131, together with the three additional fragments mentioned above. The mass spectrum of 3,4-dihydroxycholest-4-ene (Fig. 2C) showed a molecular ion (m/z 546) and the characteristic fragment at m/z 327, together with other fragments (three additional fragments mentioned above plus [M-90-90]⁺). Finally, the mass spectrum of x,25-dihydroxycholesta-3,5-diene (not shown; where “x” denotes the unknown position of hydroxylation in the ring) showed a molecular ion (m/z 544) and the characteristic fragment at m/z 131, together with other fragments ([M-15]⁺ and [M-90]⁺), but the position of the hydroxyl in the steroid ring could not been determined.

Sterol ester reactions.

Cholesteryl acetate, butyrate, and caprylate (Fig. 1M to O, respectively) were tested as UPO substrates (Table 1). The best results were obtained with rCciUPO and cholesteryl acetate, reaching a conversion yield of 34%. Cholesteryl caprylate was not modified by any of the three peroxygenases. In the reactions of sterol esters, the 25-hydroxyderivative was the only product identified. The mass spectra (not shown) showed the characteristic fragment at m/z 131 and molecular ions at m/z 544 and m/z 428 for cholesteryl butyrate and acetate, respectively.

Computational analyses.

The molecular structures of eight representative steroids, and that of the AaeUPO (PDB entry 2YOR) enzyme, were prepared at the optimal pH for peroxygenase reactions. PELE simulations were performed, providing energy profiles and ligand trajectories. The driving criterion for the simulations was to reduce the distance between the oxygen atom of the peroxygenase compound I and the center of mass of the ligand. From the results of the simulations, we computed the fractions of ligand that entered by the C-25 end and the fractions that entered by the C-3 end. The structures closest to the reactive oxygen then were analyzed as described below. Further detail on energy profiles and interaction maps for each studied steroid in the heme active site are available in the supplemental material (including supplemental computational results and Fig. S4 to S9). A direct correlation between the percentage of substrate conversion (Table 1) and the percentage of entrance by C-25 was observed (Fig. 3). In the case of cholestane, the fact that it remains 3.6 Å away from the reactive oxygen (see Fig. S7) justifies its reduced reactivity.

FIG 3.

Correlation plot for steroid conversion rate (Table 1) and steroid (side chain) entrance by C-25 to the enzyme active site. Computational results were computed from PELE diffusion simulations (20) of eight selected steroids on the molecular structure of AaeUPO (PDB entry 2YOR). For steroid (A to K) identification, see the legend to Fig. 1. Cholestane (compound J) has been removed from the correlation plot (discussed in the text).

A comparison of PELE simulations (binding energy versus distance between substrate C-25 and compound I oxygen) for side-chain hydroxylation of two free sterols and one esterified sterol (cholesterol, sitosterol, and cholesteryl acetate) is shown in Fig. 4, which also includes the main interactions in a representative structure with the ligand at reaction distance from compound I. Cholesterol has the best fraction (82%) of entrance trajectories (Fig. 3). The reason for this is a very hydrophobic heme access channel of peroxygenase, which favors the entrance of the apolar C-25 side. With the side chain in the active site, the molecule then is further stabilized by a hydrogen bond with Glu245 on the surface of the protein (Fig. 4A). Reaction at C-24 and C-26/C-27 also is possible, despite the lower reactivity of secondary and primary carbons compared to that of tertiary C-25, due to the favorable minima with these carbons well positioned for reaction (see Fig. S4 in the supplemental material).

FIG 4.

Results from PELE (20) simulations for cholesterol (A), sitosterol (B), and cholesteryl acetate (C) diffusion at the active site of the AaeUPO, including (left) plots of the energy profile versus the distance between the steroid H25 and the oxygen atom in enzyme compound I and (right) main interactions between the substrate and the enzyme in a representative structure for each of the three steroids analyzed. The yellow circles identify structural differences relative to the structure of cholesterol.

Compared with cholesterol, sitosterol contains an extra ethyl group. This large substituent impedes the access of the substrate to the active site, which is reflected by the lower fraction of structures entering the protein through the C-25 side (39%) compared with cholesterol (Fig. 3). In particular, at the active site the C-28 and C-29 groups clash with Ala77 and Thr192 and reduce considerably the number of structures approaching the heme (Fig. 4B). In addition to C-25, some reaction can be expected at C-28, since the ligand also approaches a favorable distance for the reaction of this additional secondary carbon (data not shown). In spite of its similar structure, the fraction of cholestan-3-one entrance trajectories (49%) also is lower than that found for cholesterol (Fig. 3), and the distance between C-25 and the compound I oxygen is longer (see Fig. S5 in the supplemental material). This is most probably because cholestan-3-one is incapable of hydrogen bonding to Glu245. In the case of cholesta-3,5-dien-7-one, a better approach from the ring side, due to the two double bonds present (see Fig. S6), results in the observed ring hydroxylation.

Cholesteryl acetate has an ester bond at the C-3 hydroxyl, and as in the case of the ketones described above, binding for C-25 oxidation no longer is favored by the hydrogen bond with Glu245 (Fig. 4C), resulting in lower entrance trajectories (35%) than those for cholesterol (Fig. 3); however, an approach from the opposite side did not result in ring hydroxylation. Cholesteryl caprylate has a longer acyl chain than the cholesteryl acetate described above, and (as seen in Fig. S9 in the supplemental material) it does not approach the heme group from either side (the large hydrophobic chains would favor interactions with the protein to avoid solvent exposure).

Finally, cholesta-3,5-diene has a very interesting reactivity, due to the presence of two double bonds at C-3 and C-5 and the lack of the C-3 hydroxyl mentioned above, that resulted in similar entrance by both C-25 and C-3 positions (Fig. 5A and B, respectively). This explains the fraction of hydroxylated products at the steroid rings, with many of them being additions to its double bonds. In contrast, the bulkier saturated C-5 and C-6 in cholestane (and consequent loss of planarity) impedes a suitable binding of the ligand when it enters by C-25, and a very small number of structures come close enough to the compound I oxygen (see Fig. S7 in the supplemental material), explaining the low reactivity of this compound (cholesterol is converted by 64%, while cholestane is converted by only 3%) (Fig. 3).

FIG 5.

Results from PELE (20) simulations for cholesta-3,5-diene diffusion at the active site with C-25 (A) or C-3 (B) approaching the heme cofactor of the AaeUPO.

DISCUSSION

Despite significant progress in the development of efficient biocatalysts, there is still a great demand for cost-efficient and economical biotechnologies to produce valuable steroids. Most work has been dedicated to steroid modifications catalyzed by whole microbial cells (1). The present work deals with the oxyfunctionalization of steroids using fungal peroxygenases, relatively young representatives of the superfamily of heme-thiolate peroxidases that are characterized by the presence of a cysteine residue as the fifth ligand of the heme iron and their peroxygenase activity (8). Several steroids differing in their structures, including free and esterified sterols and steroid ketones and hydrocarbons, were tested as substrates of peroxygenases from three basidiomycete species. The relationships between the structure of the different substrates, the enzyme reactivity, and the reaction products obtained are discussed below. A combined experimental and computational approach shows that there is not a unique factor that determines the hydroxylation profile; instead, it is a result of the sum of many structural factors, concerning both the steroid molecule and the enzyme active site. These include the polar/apolar character of C-3/C-7 ring groups, the presence of double bonds in the steroidal rings, and the length and character of the alkyl C-17 chain, which affect the ligand entrance ratio on the C-25 side.

Relationships between steroid structure and conversion yield.

The influence of the C-17 alkyl chain, oxidation/unsaturation of the rings, and esterification of the C-3 hydroxyl group on the steroid conversions is discussed below.

(i) Alkyl chain at C-17.

The influence of the steroid alkyl side chain substitution and unsaturation was shown by the different extents of transformation obtained for C₂₇, C₂₈, and C₂₉ sterols. The C₂₇ cholesterol was completely transformed by rCciUPO, while the other sterols were transformed to a lesser extent by this and the two other peroxygenases. The presence of a methyl or ethyl group at C-24 makes it progressively more difficult to hydroxylate campesterol and sitosterol, as illustrated in Fig. 6. Moreover, the double bond at C-22, in addition to the methyl or ethyl group at C-24, made ergosterol and stigmasterol more difficult to be hydroxylated (by the three peroxygenases) than the above-described campesterol and sitosterol. The influence of a shorter side chain was shown by comparing the hydroxylation of pregnane (with an ethyl group at C-17) and other steroid hydrocarbons (e.g., cholestane). The presence of a two-carbon (instead of eight-carbon) alkyl chain prevents hydroxylation by rCciUPO and AaeUPO and, in contrast, promotes hydroxylation by MroUPO, which takes place in the steroidal ring (not in the side chain). Finally, the absence of an alkyl chain at C-17, like in testosterone (having instead a hydroxyl group), completely prevents its hydroxylation by the three fungal peroxygenases (compared with up to 100% conversion of 4-cholesten-3-one).

FIG 6.

GC-MS analysis of the rCciUPO reaction (at 60 min) with cholesterol (A), campesterol (B), and sitosterol (C) (depicted in Fig. 1) showing the remaining substrates and the monohydroxylated derivatives at C-25, C-26, and C-28 of the side chain.

The good conversion produced in the reactions of peroxygenases with cholesterol and the differences observed in the conversion of sterols with different alkyl chain sizes, like sitosterol, are explained by computational results. In general terms, the polarity of the substituent group at C-3 has a positive correlation with the ligand rate entrance by the C-25 side. This is due to the hydrophobic character of the protein access channel to the heme that contains numerous phenylalanine residues. Moreover, the presence of a hydroxyl group in C-3 (of cholesterol and other sterols) increases the ligand rate entrance by the C-25 side and stabilizes the ligand in a suitable catalytic position (a good catalytic position is assumed when the distance between the hydrogen in C-25 and the compound I oxygen is around 2.4 Å) due to a hydrogen bond with Glu245. These two factors correlate perfectly with the high activity shown for cholesterol. However, although sitosterol also has a hydroxyl group in C-3, it is not able to access the heme site by the C-25 end in the same way as cholesterol due to the bulkier alkyl chain in C-17 that clashes with residues in the heme access channel (namely, Thr192, Ala77, Phe69, Phe121, and Phe199).

(ii) Oxidation/unsaturation degree of the rings.

Steroids differing in the oxidation state of the ring were tested, including steroid hydrocarbons, sterols, and steroid ketones. Whereas rCciUPO completely converted both cholesterol and 4-cholesten-3-one, AaeUPO and MroUPO oxidized the ketone with lower efficiency. The contrary was reported for P450s (e.g., CYP27A1), which hydroxylated steroids with a 3-oxo-Δ⁴ structure at a much higher rate than those with a 3β-hydroxy-Δ⁵ structure (32). Therefore, the structural differences between 4-cholesten-3-one and cholesterol, which are the planarity of the sterol A-ring and the electronic properties resulting from a keto versus hydroxyl substituent at C-3, seemingly affect the peroxygenase activity. The computational results revealed that the polar character of the hydroxyl and oxo groups in the sterols and steroid ketones affects in a positive way the ligand entrance rate by C-25, which correlates with the high activity that these compounds show compared with the corresponding steroid hydrocarbons.

On the other hand, although differences were observed between the three peroxygenases, the presence and number of double bonds in the rings of steroid hydrocarbons and ketones significantly increases their hydroxylation degree. This is illustrated in Fig. 7, where better AaeUPO conversion of cholesta-3,5-diene and cholesta-3,5-dien-7-one, compared with that of the corresponding saturated molecules (cholestane and cholestan-3-one), is shown. The C-25 side entrance and the placement of the steroid hydrocarbons in a proper catalytic position are determined by the presence or absence of double bonds. These reduce the volume of the steroid ring, facilitating its entrance in the active site, and provide proper curvature to the substrate for an optimal approach to the heme. Therefore, both the apolar nature of the two extremes of the molecule and the presence of double bonds in cholesta-3,5-diene facilitate access to the heme from both sides, C-3 and C-25, and its correct placement for reaction. However, neither the substrate entrance nor its placement in the active site are as suitable as they are in sterols and steroid ketones. In the case of cholestane, the entrance to the heme site is hampered by the larger sp3 groups at C-3, C-4, C-6, and C-7. This can be better observed by comparing the energy profiles of this compound (see Fig. S7 in the supplemental material) to those of cholesta-3,5-diene (see Fig. S8) and observing the very different occupation of the binding site. In addition, the higher reactivity of alkenes than alkanes leads to addition of hydroxyl groups to the double bonds in the steroidal rings in cholesta-3,5-diene, which is not observed in cholestane. Concerning the two ketones compared in Fig. 7C and D, although both have a similar ratio for C-25 entrance, cholestan-3-one can be hydroxylated at only C-25, while cholesta-3,5-dien-7-one shows some additional hydroxylation due to the higher reactivity of the alkenes mentioned above. Moreover, as in the case of the steroid hydrocarbons, the two double bonds in the steroid ketone ring will facilitate the entrance and proper positioning of the substrate at the peroxygenase active site.

FIG 7.

GC-MS analysis of the rCciUPO reaction (at 60 min) with cholestane (A), cholesta-3,5-diene (B), cholestan-3-one (C), and cholesta-3,5-diene-7-one (D) (depicted in Fig. 1) showing the remaining substrates and the mono-, di-, and trihydroxylated derivatives included in Table 4 (x denotes the unknown position of hydroxylation in the ring).

Due to the above-described effects, although the conversion rates of cholestan-3-one are lower than those of cholesterol (the best steroid substrate of the basidiomycete peroxygenases), they are increased by the presence of double bonds in the steroid ring, and cholestan-3,5-dien-7-one hydroxylation rates are in the same order of those obtained for cholesterol.

(iii) Esterified hydroxyl group at C-3.

With the aim of investigating whether the hydroxyl group at C-3, in free form, is necessary for maximal peroxygenase activity, three esters of cholesterol with organic acids of different chain lengths were assayed as substrates. The experimental results revealed that the presence of an esterifying group hampers substrate conversion, and that this effect is directly correlated with increasing chain length of the fatty acid. In sterol esters, hydroxylation (at C-25) only of the side chain of the sterol was observed and not at the acyl moiety, despite AaeUPO and rCciUPO being found to efficiently hydroxylate fatty acids (11, 18).

The computational analyses revealed that in the sterol esters, the C-25 entrance rate is decreased with respect to cholesterol due to the presence of hydrophobic alkyl chains at the ester site. In the case of cholesteryl caprylate, which is not transformed by any of the peroxygenases, the long unsubstituted hydrocarbon chain tends to “anchor” the substrate away from the heme site by interacting with different groups on the protein surface. Since both ends are hydrophobic, interactions with the protein minimize the solvent exposure but restrain the ligand from reaching the heme, which explains the lack of activity for this compound.

Relationships between steroid structure and regioselectivity.

In general, the reactions of the three fungal peroxygenases with the different steroids proceed with remarkable regioselectivity and result in the formation of the 25-hydroxyderivatives as the main products. The complete conversion of cholesterol by rCciUPO to produce 25-hydroxycholesterol is noteworthy, since this compound is widely used, displaying an array of pharmacological actions in vitro and in cell-based systems (33), as discussed below. Different alkyl chains at C-17 seemingly do not influence the regioselectivity in AaeUPO and MroUPO reactions (see below) but affect that of rCciUPO, since cholesterol is C-25 hydroxylated with a strict (100%) regioselectivity, whereas for campesterol and sitosterol, hydroxylation at terminal C-26/C-27 (15%) and subterminal C-28 (48%), respectively, also was observed. On the other hand, the double bond at C-22 in ergosterol and stigmasterol seemingly influences the strict regioselectivity of the hydroxylation at C-25. Minor hydroxylation at the terminal positions of the steroid branched side chain by AaeUPO and MroUPO also leads in some cases (e.g., cholesterol) to formation of the C-26/C-27-carboxylated derivatives, in addition to the C-26/C-27-hydroxylated ones. This is because of the further oxidation (successive hydroxylation and dehydration reactions) of the monohydroxylated products, as previously demonstrated in ¹⁸O-labeling oxygenation studies with AaeUPO and H₂¹⁸O₂ (11).

Hydroxylation at positions other than C-25 occurs in cases that show a high (energetically favorable) occupation of binding positions near any of these centers, as demonstrated for AaeUPO by computational studies. In particular, the case of cholesterol hydroxylation at C-24 is due to interactions between the substrate and the protein that include a local minima for C-24, which is slightly more favorable than that for C-25. Hydroxylation at C-25 (tertiary C-H bond) is electronically preferred by its lower C-H bond strength compared with that of the secondary and primary C-H bonds (34). In contrast, steric constraints rather than the differential electronic properties of the C-H bonds have been reported to govern the regioselectivity of the reaction of steroids (cholesterol and 4-cholesten-3-one) with P450s (CYP125A1) where terminal oxidation was preferred (35).

Interestingly, in addition to hydroxylation at the side chain, some hydroxylated derivatives at the steroid core are formed when oxygen groups are lacking at C-3, such as in cholesta-3,5-dione, cholesta-3,5-diene, or pregnane. Indeed, the presence of a hydroxyl or oxo group at this position seems to prevent the hydroxylation of the steroidal core. On the other hand, the presence of two conjugated double bonds in the ring seems to favor the hydroxylation at the steroidal core, especially by AaeUPO and MroUPO, as evidenced in the reaction of cholesta-3,5-diene yielding 3,6-dihydroxycholest-4-ene and 3,4-dihydroxycholest-5-ene, as mentioned above. The mechanism of this double (at C-3 and C-6 or at C-3 and C-4) hydroxylation implies the displacement of one of the cholesta-3,5-diene double bonds to the intermediate (C-4 or C-5) position and the saturation of the other.

Significance of side-chain versus ring hydroxylation.

It has been shown that the reactions of the peroxygenases from the three different basidiomycetes occurred generally with noteworthy regioselectivity, and hydroxylation at the side chain over the steroidal ring was preferred, with the 25-hydroxyderivatives predominating. From a biochemical point of view, side-chain and steroid nucleus oxidation have been confirmed to be independent processes, at least in some microorganisms (1). Side-chain hydroxylation is usually a prestep in side-chain degradation in most microbial systems that oxidize steroids. The initial step of the side-chain oxidation of sterols (and other C₂₇ steroids) is hydroxylation at C-26 or C-27, followed by a complex sequence of reactions involving different enzymes (some of them still unidentified) that produce the elimination of side chain at C-17 and then steroidal nucleus oxidation. However, further oxidative degradation on the C-17 side chain may not occur when hydroxylation is produced at C-25 (instead of C-26 or C-27). This has been observed using 25-hydroxylated cholesterol as the substrate (36), concluding that C-25 hydroxylation must be a distinct reaction from that of the usual cholesterol side chain degradation pathway. Indeed, the authors do not consider that steroid metabolism is a specific function of UPOs, not as much as it is the oxidation of alkanes, alkenes, ethers, or aromatics. Rather, UPOs represent a catalytic system (“extracellular liver”) that, irrespective of a particular structure, can nonspecifically oxidize diverse compounds emerging in the fungus microenvironment. In this context, unspecific detoxification surely is more important than specific degradation.

Interestingly, a previously unrecognized biological (antiviral) role for 25-hydroxycholesterol, which is selectively formed by the rCciUPO as discussed above, has been highlighted recently (37). Indeed, that study revealed 25-hydroxycholesterol as the only oxysterol notably upregulated following macrophage infection or activation by interferons. This antiviral effect is specific to 25-hydroxycholesterol, because other related (enzymatically generated) oxysterols, such as 19-hydroxychoesterol and 7α-hydroxycholesterol, fail to repress viral infection. 25-Hydroxycholesterol has been reported to have a high potency to inhibit a broad spectrum of viruses from high to low physiological concentrations depending on lipid conditions and virus-host cell system. Likewise, other recent studies also revealed 25-hydroxycholesterol as a potential antiviral therapeutic (38). Independent of its known regulatory effect on metabolism, 25-hydroxycholesterol impairs viral entry at the virus-cell fusion step by inducing cellular membrane changes. Because 25-hydroxycholesterol can permeate through membranes, it likely modifies cellular membranes to exert its antiviral effect. Moreover, in addition to interfering with viral entry and replication, 25-hydroxycholesterol also amplifies the activation of immune cells and increases the production of immune mediators (39).

Final remarks.

The first crystal structure of a fungal peroxygenase (23), from the basidiomycete A. aegerita (AaeUPO), was used in the computational simulations included in the present study. The availability of more peroxygenase crystal structures, including those from C. cinerea and M. rotula, will provide the opportunity to correlate the different regioselectivities in steroid oxygenation described here with the architecture of the active site and other structural features. Such information will permit us to engineer these self-sufficient monooxygenases (whose activation depends on a peroxide source) as new and robust industrial biocatalysts for the pharmaceutical and fine-chemical sectors.

In conclusion, in the present work, the ability of three different fungal peroxygenases to catalyze the regioselective hydroxylation of a variety of steroids is shown for the first time. Generally, hydroxylation at the side chain over the steroid ring was preferred, with 25-hydroxyderivatives being the main products formed. It has been observed that some structural features of the steroid substrates, such as the presence and length of the alkyl chain at C-17 and the presence of oxygen groups and/or conjugated double bonds in the steroid ring system, influence substrate conversion and guide the regioselectivity of the reactions. A better understanding of the mechanistic aspects of hydroxylation in combination with the development of suitable biocatalysts would be the basis for efficient steroid hydroxylation bioprocesses using new enzymes with self-sufficient mono(per)oxygenase activities.