A Novel Thermostable Cytochrome P450 from Sequence-Based Metagenomics of Binh Chau Hot Spring as a Promising Catalyst for Testosterone Conversion

: Biotechnological applications of cytochromes P450 show di ﬃ culties, such as low activity, thermal and / or solvent instability, narrow substrate speciﬁcity and redox partner dependence. In an attempt to overcome these limitations, an exploitation of novel thermophilic P450 enzymes from nature via uncultured approaches is desirable due to their great advantages that can resolve nearly all mentioned impediments. From the metagenomics library of the Binh Chau hot spring, an open reading frame (ORF) encoding a thermostable cytochrome P450—designated as P450-T3—which shared 66.6% amino acid sequence identity with CYP109C2 of Sorangium cellulosum So ce56 was selected for further identiﬁcation and characterization. The ORF was synthesized artiﬁcially and heterologously expressed in Escherichia coli C43(DE3) using the pET17b system. The puriﬁed enzyme had a molecular weight of approximately 43 kDa. The melting temperature of the puriﬁed enzyme was 76.2 ◦ C and its apparent half-life at 60 ◦ C was 38.7 min. Redox partner screening revealed that P450-T3 was reduced well by the mammalian AdR-Adx 4-108 and the yeast Arh1-Etp1 redox partners. Lauric acid, palmitic acid, embelin, retinoic acid ( all-trans ) and retinoic acid ( 13-cis ) demonstrated binding to P450-T3. Interestingly, P450-T3 also bound and converted testosterone. Overall, P450-T3 might become a good candidate for biocatalytic applications on a larger scale.


Introduction
Cytochromes P450 (CYPs) belong to one of the largest enzyme superfamilies, which is widely distributed in all living organisms like bacteria, fungi, plants and animals [1,2]. They play a crucial role in life by catalyzing more than 20 types of reactions in regio-and stereoselective manners, such as hydroxylation, dealkylation, epoxidation, oxidation, dehalogenation, dehydrogenation and located in the Ba Ria province and is the second hottest hot spring in Vietnam. The average opened mouth temperature is 80 • C, hence, this location is a potential resource for finding thermophiles and thermostable enzymes. The sequencing and analysis of 9.4 Gb DNA metagenomics from the Binh Chau hot spring by Illumina Hiseq Platform revealed the presence of 68 putative ORFs encoding for putative monooxygenase enzymes, which belong to 36 different P450 subfamilies [21]. Based on the melting temperature index program (http://tm.life.nthu.edu.tw/), a novel sequence, soluble P450-T3, was selected due to its predicted melting temperature of >65 • C. Subsequently, its cDNA was synthesized and expressed, and the protein was purified for characterization. The purified recombinant enzyme showed thermal stability, coupled with several artificial electron transfers and bound a broad spectrum of substrates.
The phylogenetic tree obtained with MEGA X revealed that P450-T3 formed a cluster with other members of the CYP109C subfamily, which were identified in S. cellulosum (Supplementary Figure S1), suggesting that P450-T3 belongs to the CYP109C subfamily.
Multiple amino acid sequence alignments of P450-T3 with its closest homologs reveal the presence of the three conserved domains of cytochrome P450 (Figure 1). The oxygen binding and activation motif AGXXTT is located in I-helix (residues 222-227) [22], of which the highly conserved threonine probably plays an important role in catalysis [23]. The EXXR motif (residues 261-264) contains the conserved glutamic acid and arginine in all the cytochromes P450 that build a set of salt-bridge interactions to form the final P450 tertiary structure [24]. The heme-binding domain motif FXXGXXXCXG (residues 325-334) contains the three most conserved residues in the cytochrome P450 superfamily, including phenylalanine, glycine, and cysteine, of which the conserved cysteine plays the role of the axial ligand to the heme [25].

Production and Purification of P450-T3
For expression purpose, the P450-T3 gene was successfully cloned into the pET17b vector and expressed in the Escherichia coli C43(DE3) strain. The carbon monoxide (CO) difference spectrum was recorded according to the Omura and Sato methods [26]. The results showed a typical peak with a maximum at 450 nm without another peak at 420 nm (Figure 2a), indicating that the protein was produced in an active conformation. The yield of recombinant P450-T3 was 585 nmol/L bacterial cell culture before purification. The purified enzyme was shown to have a molecular weight of approximately 43 kDa on SDS-PAGE as predicted (inset of Figure 2a). UV-Vis spectroscopy revealed that P450-T3 contains a heme molecule as a prosthetic group, which is involved in the generation of the Soret (γ) band at 418 nm (Figure 2b), and the smaller α and β bands at 567.5 and 536 nm, respectively (inset of Figure 2b), indicating a low-spin state of heme iron.

Production and Purification of P450-T3
For expression purpose, the P450-T3 gene was successfully cloned into the pET17b vector and expressed in the Escherichia coli C43(DE3) strain. The carbon monoxide (CO) difference spectrum was recorded according to the Omura and Sato methods [26]. The results showed a typical peak with a maximum at 450 nm without another peak at 420 nm (Figure 2a), indicating that the protein was produced in an active conformation. The yield of recombinant P450-T3 was 585 nmol/L bacterial cell culture before purification. The purified enzyme was shown to have a molecular weight of approximately 43 kDa on SDS-PAGE as predicted (inset of Figure 2a). UV-Vis spectroscopy revealed that P450-T3 contains a heme molecule as a prosthetic group, which is involved in the generation of the Soret (γ) band at 418 nm (Figure 2b), and the smaller α and β bands at 567.5 and 536 nm, respectively (inset of Figure 2b), indicating a low-spin state of heme iron. The secondary structure of an enzyme can be identified rapidly by using circular dichroism (CD) spectroscopy. The CD spectra were recorded in the far-UV and near UV visible region. The far-UV CD spectrum showed double bands with minima at 211 and 222 nm (Figure 2c), a representation of combined contribution of helical and β-strand structures [27]. P450-T3 was predicted to consist of 64.16% regular secondary elements (53.77% the α-helix and 10.39% strand of β-sheet) and 35.84% coil as an irregular secondary element (Supplementary Figure S2). In the near UV-visible region, P450-T3 displayed two large negative signals at 350 nm and at 408 nm ( Figure 2d). These are in correspondence with the characteristic peaks for other bacterial P450s [28,29]. The secondary structure of an enzyme can be identified rapidly by using circular dichroism (CD) spectroscopy. The CD spectra were recorded in the far-UV and near UV visible region. The far-UV CD spectrum showed double bands with minima at 211 and 222 nm (Figure 2c), a representation of combined contribution of helical and β-strand structures [27]. P450-T3 was predicted to consist of 64.16% regular secondary elements (53.77% the α-helix and 10.39% strand of β-sheet) and 35.84% coil as an irregular secondary element (Supplementary Figure S2). In the near UV-visible region, P450-T3 displayed two large negative signals at 350 nm and at 408 nm ( Figure 2d). These are in correspondence with the characteristic peaks for other bacterial P450s [28,29].

Thermal Stability of P450-T3
The effect of temperature on P450-T3 structural integrity was investigated using various spectral methods. P450-T3 showed the highest integrity at 50-60 °C and lost it at 70 °C (Figure 3a). The far-UV CD spectra were recorded between 190-260 nm every 10 °C to determine the changes in their secondary structure as a function of the temperature [30]. In the inset of Figure 3b, the appearance of an intense negative minimum around 208 and 222 nm and a positive maximum around 197 nm

Thermal Stability of P450-T3
The effect of temperature on P450-T3 structural integrity was investigated using various spectral methods. P450-T3 showed the highest integrity at 50-60 • C and lost it at 70 • C (Figure 3a). The far-UV CD spectra were recorded between 190-260 nm every 10 • C to determine the changes in their secondary structure as a function of the temperature [30]. In the inset of Figure 3b, the appearance of an intense negative minimum around 208 and 222 nm and a positive maximum around 197 nm confirmed the existence of a predominant α-helical secondary structure [27]. The presence of an isodichroic point at 203 nm (inset of Figure 3b) indicated a dominant α to β structural thermal transition. The thermal stability curve of P450-T3 was evaluated by recording the increment in the CD signal at a 211 nm wavelength over the temperature range from 25-95 • C. At 70 • C, a little loss of magnitude of the negative CD signal was observed, compared to those at temperatures of 30-60 • C. It means that, at a temperature of ≥70 • C, P450-T3 lost helical secondary structure. This phenomenon was clearly observed at 80 • C and 90 • C. The overlap of CD signals of protein P450-T3 at temperatures ≤60 • C suggested that the helical secondary structure of P450-T3 was stable at 60 • C. Moreover, in the thermal Catalysts 2020, 10, 1083 6 of 17 stability curve determined by CD (Figure 3b), we observed that P450-T3 started to change the helical structure at 67 • C. A rapid inactivation was shown after incubation at 80 • C. The calculated Tm was 76.2 ± 0.05 (R 2 = 0.99).
transition. The thermal stability curve of P450-T3 was evaluated by recording the increment in the CD signal at a 211 nm wavelength over the temperature range from 25-95 °C. At 70 °C, a little loss of magnitude of the negative CD signal was observed, compared to those at temperatures of 30-60 °C. It means that, at a temperature of ≥70 °C, P450-T3 lost helical secondary structure. This phenomenon was clearly observed at 80 °C and 90 °C. The overlap of CD signals of protein P450-T3 at temperatures ≤60 °C suggested that the helical secondary structure of P450-T3 was stable at 60 °C. Moreover, in the thermal stability curve determined by CD (Figure 3b), we observed that P450-T3 started to change the helical structure at 67 °C. A rapid inactivation was shown after incubation at 80 °C. The calculated Tm was 76.2 ± 0.05 (R 2 = 0.99).  Measuring the half-life (T 1/2 ) at optimal temperature is another parameter for evaluating thermal stability. The T 1/2 of P450-T3 was determined by incubating the enzyme at 60 • C and measured by CO difference spectroscopy every 15 min to calculate the relative stability ( Figure 3c). The T 1/2 of P450-T3 was 38.7 min at 60 • C.

Identification of Electron Transfer Partners
Cytochromes P450s are heme-containing monooxygenases which require a coupled and stepwise supply of electrons to start oxidation and hydroxylation reactions [3]. For efficient catalysis, P450s require an electron transfer chain (in bacteria mostly a ferredoxin reductase and ferredoxin). In nature, microorganisms can use autologous electron transfer partners. In this study, under laboratory conditions, we tested the compatibility of P450-T3 with known ferredoxin reductase/ferredoxin partners, including BmCPR-Fdx2, BmCPR-Fdx3, Arh1-Etp1 and AdR-Adx 4-108 . The diflavin reductase BmCPR and ferredoxin Fdx2 are electron transfer partners of Bacillus megaterium DSM319. The BmCPR-Fdx2 system efficiently supported the activity of CYP106A1 [31] and CYP107DY1 [32]. The bovine adrenodoxin reductase homologue 1 Arh1 and its natural redox partner Etp1 are originated from Schizosaccharomyces pombe [33]. The Arh1/Etp1 system was demonstrated to be efficient for a biotechnological application with CYP105A1 from Streptomyces griseolus [34,35]. The bovine adrenodoxin reductase (AdR) and adrenodoxin (Adx 4-108 ) were identified as the most efficient redox partners for several myxobacteria cytochrome P450s [36,37].
The redox partners were compared on the base of the spectra of the reduced CO-complexed enzyme of P450-T3 ( Figure 4). The redox partners BmCPR-Fdx3 did not produce a significant Soret peak at 450 nm. The redox partners BmCPR-Fdx2 reduced by~15% in the CO complex sample. The systems AdR-Adx 4-108 and Arh1-Etp1 showed a higher efficiency with~30% of peak recovered (compared to dithionite). The redox system AdR-Adx 4-108 was selected for further investigation of P450-T3.
Measuring the half-life (T1/2) at optimal temperature is another parameter for evaluating thermal stability. The T1/2 of P450-T3 was determined by incubating the enzyme at 60 °C and measured by CO difference spectroscopy every 15 min to calculate the relative stability (Figure 3c). The T1/2 of P450-T3 was 38.7 min at 60 °C.

Identification of Electron Transfer Partners
Cytochromes P450s are heme-containing monooxygenases which require a coupled and stepwise supply of electrons to start oxidation and hydroxylation reactions [3]. For efficient catalysis, P450s require an electron transfer chain (in bacteria mostly a ferredoxin reductase and ferredoxin). In nature, microorganisms can use autologous electron transfer partners. In this study, under laboratory conditions, we tested the compatibility of P450-T3 with known ferredoxin reductase/ferredoxin partners, including BmCPR-Fdx2, BmCPR-Fdx3, Arh1-Etp1 and AdR-Adx4-108. The diflavin reductase BmCPR and ferredoxin Fdx2 are electron transfer partners of Bacillus megaterium DSM319. The BmCPR-Fdx2 system efficiently supported the activity of CYP106A1 [31] and CYP107DY1 [32]. The bovine adrenodoxin reductase homologue 1 Arh1 and its natural redox partner Etp1 are originated from Schizosaccharomyces pombe [33]. The Arh1/Etp1 system was demonstrated to be efficient for a biotechnological application with CYP105A1 from Streptomyces griseolus [34,35]. The bovine adrenodoxin reductase (AdR) and adrenodoxin (Adx4-108) were identified as the most efficient redox partners for several myxobacteria cytochrome P450s [36,37].
The redox partners were compared on the base of the spectra of the reduced CO-complexed enzyme of P450-T3 (Figure 4). The redox partners BmCPR-Fdx3 did not produce a significant Soret peak at 450 nm. The redox partners BmCPR-Fdx2 reduced by ~15% in the CO complex sample. The systems AdR-Adx4-108 and Arh1-Etp1 showed a higher efficiency with ~30% of peak recovered (compared to dithionite). The redox system AdR-Adx4-108 was selected for further investigation of P450-T3.

Substrate Screening for P450-T3
Seeking for the substrate specificity of an individual cytochrome, P450, especially of a novel enzyme, is a significant effort that helps to expand its potential application. Since P450-T3 was identified as a close relative of CYP109C2 from S. cellulosum So ce56-which is a fatty acid hydroxylase [38], lauric acid, palmitic acid-and other substances sharing a structural part of fatty acids, such as retinoic acid (all-trans), and retinoic acid (13-cis) and embelin were used to evaluate their ability to bind to P450-T3. The results showed a transition from low-spin state to high-spin state of the heme pocket for all the substances used (Table 1), indicating a broader potential substrate spectrum for P450-T3 in comparison with CYP109C2. Enzyme members of the CYP109 subfamilies have different substrate specificities. For example, CYP109A2 from B. megaterium DSM319 was involved in vitamin D3 hydroxylation [39], whereas CYP109D1 from S. cellulosum So ce56 was demonstrated to oxidize norisoprenoids [40]. CYP109E1 from B. megaterium DSM319 [41] and CYP109B1 from B. subtilis W23 [42] showed the capacity of steroid conversions. Besides steroid conversion, CYP109B1 was reported to hydroxylate αand β-ionone [43]. To date, CYP109E1 has shown the broadest substrate spectrum, being involved not only in steroid conversion, but also in the hydroxylation of vitamin D3, vitamin D2, cholesterol, statins, and terpenes [39,41,44,45]. Therefore, P450-T3 might also exhibit novel characteristics, including an altered substrate specificity. In this study, nootkatone and several steroids, such as 11-deoxycorticosterone (DOC), 11-deoxycortisol (RSS), progesterone, and testosterone were tested. Interestingly, testosterone was observed to bind to P450-T3 (Table 1). It was further investigated in in vitro conversion assays. Using the redox system AdR-Adx 4-108 for in vitro reconstitution, P450-T3 showed the ability to convert testosterone into two products at a retention times of 13.5 and 15 min, respectively ( Figure 5). The conversion ratio of 100 µM testosterone was~11% in total after 30 min.

Discussion
Recently, considerable attention has been paid towards identification and characterization of new thermostable P450s due to their potential use in industrial processes. Until now, only few thermostable P450s were found from thermophilic microorganisms in nature, including CYP119 from

Discussion
Recently, considerable attention has been paid towards identification and characterization of new thermostable P450s due to their potential use in industrial processes. Until now, only few thermostable P450s were found from thermophilic microorganisms in nature, including CYP119 from S. solfataricus [46], CYP119A1 from T. thermophilus [14], CYP119A2 from S. tokodaii [47], CYP175A1 from T. thermophilus [13], CYP154H1 from Thermobifida fusca [48], CYP231A2 from P. torridus [49], and CYP116B subfamilies from several archaea species [15,16]. However, the number of thermostable cytochrome P450s is still insufficient for versatile biocatalysis on a larger scale. To enhance the catalytic efficiency, different strategies have been used, such as the engineering of proteins, redox-partner interactions, substrates, and electron sources [50], of which the improvement in cytochrome P450s is the main effort and challenge. Besides the molecular evolution of these enzymes, another approach-metagenomics-was introduced in 2004 by Handelsman et al. [51]. Until now, metagenomics is a well-accepted power tool for exploiting the microbial genome in a natural environment. Several enzymes have been successfully discovered using this method, such as hydrolases [52][53][54][55][56], oxidoreductases [57], dehydratases [58], and cytochrome P450s (e.g.,CYP153A members) [18]. Recently, 264 putative CYP153 proteins were identified from the metagenomic dataset obtained from 23 sediment samples [59]. Their characteristics were computed to point out a highly promising application for catalysis using a broad variety of substrates under cold conditions. The advantage of using metagenomics is especially visible in extreme habitats, such as geothermal ecologies (volcanoes, hot spring) and offshore oil platform. Most of the thermostable enzymes have been produced by thermophilic microorganisms in hot springs or through metagenomic analysis of this environment [60][61][62]. Vietnam has a diverse hot spring system which allows us to find desired enzymes. To the best of our knowledge, this is one of the first examples of identifying thermostable cytochrome P450s that rely on metagenome-based sequences in the Binh Chau hot spring. Here, we report on the characterization of one of them, P450-T3. Bioinformatic analyses of the amino acid sequence of P450-T3 indicated that it shares the highest identity with CYP109C2 of S. cellulosum So ce56. However, P450-T3 does not belong to the CYP109C2 subfamily due to a moderate bootstrap value. This finding might indicate a novel CYP109C subfamily, which is one of the smallest families among bacterial P450s with so far only eight subfamily members (https://cyped.biocatnet.de/sFam/109). Moreover, analyses of 16S rRNA database in the metagenomics of the Binh Chau hot spring revealed that S. cellulosum accounts for~0.00438% (data not shown). Therefore, it was supposed that P450-T3 might have originated from this myxobacterial species.
On the other hand, the half-life (T 1/2 ) of an enzyme is an important index to evaluate its stability. Protein engineering has a great impact on enzyme stability. CYP102A1 from B. megaterium was engineered to strengthen the thermal stability of the reductase domain, which led to an increase in the optimal temperature from 25 to 40 • C and prolonged its half-life at 50 • C ten-fold [68]. Moreover, Urlacher et al. [69] created a library of chimeric fusion proteins by exchanging the unstable reductase domain of CYP102A1 with the more stable reductase domain of CYP102A3, resulting in a chimeric protein displaying a wider temperature range and broader substrate specificity. The T 1/2 of the new chimeric protein at 50 • C was 100 min, which was more than ten-fold longer than that of the wild type. P450-T3 displayed a T 1/2 at 60 • C of 37.5 min, which is a moderate thermostability as compared to other CYPs. However, as described for CYP102A1, it might be further increased by using protein engineering techniques, such as site-directed mutagenesis or directed evolution.
P450-T3 is an external monooxygenase. Therefore, it requires an external electron donor in the electron transfer chain to transfer electrons from NADPH to its heme for substrate hydroxylation. In this study, the used redox pairs BmCPR-Fdx2 and BmCPR-Fdx3, originating from B. megaterium, were not or only partly able to interact with P450-T3. Similar to CYP109C1 and CYP109C2 from S. cellulosum So ce56, P450-T3 accepted the reduction equivalents from bovine AdR-Adx 4-108 and yeast Arh1-Etp1 [38]. However, the reduction yield of P450-T3 was low, (Figure 5). Hence, a suitable ratio between P450-T3 and the redox partner may be identified in the future to optimize the electron transfer.
The next important question was to identify possible substrates for P450-T3. The similarity to CYP109C1 and CYP109C2 from S. cellulosum So ce56 indicated the ability to bind to medium-chain fatty acid (lauric acid) and unsaturated long-chain fatty acids (e.g., palmitic acid). As shown in Table 1, the substrate spectrum binding to P450-T3 is more diverse than that of the thermostable CYP175A1, which only showed the ability to bind to some unsaturated monoenoic fatty acids but not to saturated fatty acids [70]. It has to be demonstrated in future studies whether the bound substrates can also be converted by P450-T3, since this not always the case [71]. Lauric acid and palmitic acid are the main fatty acids in coconut oil and palm kernel oil; therefore, P450-T3 might be used in oil production in the future. Moreover, long aliphatic chain (embelin) and tretinoins (retinoic acid (all-trans), as well as retinoic acid (13-cis)), also bind as potential substrates to P450-T3, suggesting that P450-T3 may be used for the production of biopharmaceuticals. Embelin and its derivatives are used in cancer treatments and display other biological activities [72]. Retinoic acid (all-trans) and retinoic acid (13-cis) are two main substances used in dermatology for acne as well as steroid-damaged skin treatment, despite of some negative effects [73][74][75]. It will, therefore, be of high interest to investigate their conversion by P450-T3 and to identify the formed products.
Since steroids are broadly marketed drugs, together with antibiotics and antibodies, their derivatives have attracted much attention. Steroid hydroxylation results in hydroxylated products with high level of biological activity [76]. Interestingly, P450-T3 was able to convert testosterone into two products. However, under the conditions used here, the ratio of conversion only reached~11% in total. A variety of factors may affect testosterone conversion, such as an insufficient redox partner, improper temperature, and/or the inappropriate ratio of the components in the reconstituted system. The role of P450-T3 in Binh Chau hot spring environment, as well as the sequences of putative natural reductases and ferredoxins, which can serve as redox partners of P450-T3, is still unexplored. It is thus of high interest to identify a thermostable redox partner to finally optimize the condition for substrate conversion and to be able to characterize the obtained products of substrate conversion. Despite the presently low catalytic efficiency of testosterone conversion, this is, to the best of our knowledge, the first report of testosterone conversion of a thermostable P450 and of a member of the CYP109C subfamily.
Taken together, our findings demonstrate that P450-T3 is a novel thermostable P450 with an astonishing broad ability to bind different kinds of substrate. It might be a potential biocatalyst for versatile drug production. These results also are very useful for further biochemical studies and biotechnological applications of P450-T3.

Materials
The 1158 bp ORF encoding for P450-T3, comprising 385 amino acids, was subjected in this study. This ORF was synthesized and ligated into pUC19 vector by Phusa Biochem Ltd. (Cantho, Vietnam). The nucleotide and amino acid sequences of P450-T3 were deposited in the NCBI GenBank under the accession number MT232930.
The E. coli strain Top 10 was purchased from Invitrogen (San Diego, CA, USA). The E. coli strain C43(DE3) was obtained from Agilent Technologies (Santa Clara, CA, USA).

Bioinformatics Analysis
The DNA sequence was translated using the ExPaSy Translate tool (https://www.expasy.org/). Identification of close homologs was performed using the Basic Local Alignment Search Tool (BLAST, NCBI) and Dr. Nelson's Cytochrome P450 Homepage [77]. Alignment of multiple amino acid sequences was performed with Clustal Omega [78]. Evolutionary analyses were conducted in MEGA X [79]. The evolutionary history was inferred with the maximum likelihood approach and Le_Gascuel model [80]. The content of secondary elements of proteins was calculated using PRIPRED tool [81].

Production and Purification of P450-T3
The ORF P450-T3 was amplified using the pair of primers described above and cloned into the pET17b vector (Novagen, Darmstadt, Germany) with the NdeI/HindIII restriction sites. The clones (pET-T3) were checked with both restriction enzymes before sequencing by MWG Biotech AG (Ebersberg, Germany).
The vector pET-T3 was transformed into E. coli C43(DE3), competent cells for heterologous gene expression. The main culture was inoculated at 37 • C in 250 mL Terrific Broth (TB) medium containing 100 µg/mL ampicillin to achieve an OD 600 of 0.8-1. Protein production was induced by adding 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) and 0.5 mM δ-aminolevulinic acid as a precursor for heme synthesis. The cells were harvested after 48 h expression at 30 • C and shaken at 150 rpm.
The cell pellets were resuspended and sonicated in 50 mL lysis buffer (50 mM potassium phosphate buffer, pH 7.4, containing 20% glycerol, 0.1 mM dithioerythritol, 500 mM sodium acetate, and 1 mM phenylmethanesulfonyl flouride) [32]. The lysate was centrifuged at 35,000 rpm and 4 • C for 30 min. Cell debris was removed and the supernatant was loaded onto a Ni-NTA agarose column equilibrated with lysis buffer. After washing with 100 mL of the equilibration buffer containing a gradient of 20-50 mM imidazole, protein was eluted with 20 mL elution buffer (50 mM potassium phosphate, pH 7.4, containing 1% Tween 20, 10% glycerol, 0.1 mM dithioerythritol, 0.1 mM phenylmethanesulfonyl fluoride, and 200 mM imidazole). Fractions with A 417 :A 280 >1.6 were collected and dialyzed overnight to remove imidazole. The protein was concentrated by Centriprep (Millipore, MA, USA) with a pore size of 50 kDa and 30 kDa to the expected volume before storing at −80 • C.
The concentration of P450 enzyme was estimated by CO difference spectroscopy assuming ∆ε (450-490) = 91 mM −1 * cm −1 as described by Omura and Sato [26]. UV-visible spectra for the purified P450 enzyme was recorded at room temperature from 200 to 700 nm on a double-beam spectrophotometer (UV2101PC, Shimadzu, Kyoto, Japan).

Circular Dichroism Spectroscopy
The far-UV CD spectra (195-260 nm) and the near-UV spectra (300-450 nm) were recorded with 4 and 20 µM purified P450-T3, respectively, in 10 mM potassium phosphate buffer pH 7.4 at 25 • C on a JASCO J715 spectropolarimeter, as described by Khatri et al. [38]. Molar ellipticity [θ] was calculated in deg*cm 2 /dmol using the following equation: [θ] = m o * M/(10 * L * C) (1) where m o is the CD millidegree, M is the average molecular weight of P450-T3 (g/mol), L is path length of cell (cm), and C is concentration of sample.

Thermal Stability
The optimal temperature of P450-T3 was identified by CO difference spectroscopy. Enzyme samples (10 µM) were incubated for 15 min at 40, 50, 60, and 70 • C, respectively, and the retained amount of the active form of P450 (absorption in the CO spectrum at 450 nm) was estimated.
The enzyme melting curve was evaluated from its far-UV CD spectrum as an important measure of thermal stability. The purified protein was dissolved in 20 mM of mM potassium phosphate buffer pH 7.4 to obtain a concentration of 20 µM. The enzyme was scanned using a Chirascan CD spectrometer between 190-260 nm while recording every 1 nm for 0.5 s per nm with a bandwidth of 5 nm. Each spectrum was acquired independently three times. Then, the melting curve was obtained by monitoring the CD at 211 nm over a temperature range 25-95 • C at a rate of 1 • C per min. At each temperature, the enzyme was able to equilibrate for 30 s before recording the CD. Melting temperature was obtained from the second-derivative plots of the melting curve.
Another parameter related to enzyme stability is the enzyme half-life inactivation (T 1/2 ). To determine T 1/2 , the purified P450-T3 was dissolved in 20 mM potassium phosphate buffer, pH 7, and incubated at 60 • C. Samples were taken every 15 min, and the residual absorption at 450 nm in the CO difference spectrum was measured. The T 1/2 index was calculated by the following equation: where K d is the first-order rate constants determined by linear regression of ln (residual absorption at 450 nm) versus the incubation time (t) [82,83].

Investigation of Electron Transfer Partners
The functional interaction of the electron transfer partners for a P450-T3 was examined by recording the NADPH reduced CO-complex peak at 450 nm when P450-T3 was combined with the different ferredoxins(Fdx2/Fdx3/Adx 4-108 or Etp1)/ferredoxin reductase (BmCPR, AdR or Arh1) pairs in the absence of substrate, as described by Milmim et al. [32].

Monitoring of Substrate Binding
The substrate binding of the enzyme was followed by a transition of Soret bands from low-spin state which shows a maximal peak at 417 nm to high-spin state where the maximal peak raised at around 390 nm. The measurement was carried out by adding substrates in the excess ratio of 1:15 or 1:20 and monitored in the range of 200-700 nm at room temperature on a double-beam spectrophotometer (UV2101PC, Shimadzu, Kyoto, Japan). The kind of substrates varied from fatty acids (palmitic acid and lauric acid) to metabolites (embelin, retinoic acid (all-trans) and retinoic acid (13-cis)) and steroids (11-deoxycorticosterone (DOC), 11-deoxycortisol (RSS), progesterone, and testosterone).

In Vitro Conversion of Testosterone and HPLC Analysis
A reconstituted in vitro system was used in a final volume of 250 µL at 37 • C in a buffer (50 mM HEPES, pH 7.4, and 20% glycerol) which consisted of 1 µM P450-T3, 2.5 µM AdR, 20 µM Adx 4-108 , 1 mM MgCl 2 , 5 mM glucose-6-phosphate, 1 U glucose-6-phosphate dehydrogenase, and 100 µM testosterone. The addition of 200 µM NADPH started the conversion. After 30 min, the conversion was stopped by adding ethyl acetate (250 µL) and extracted twice. The negative control was carried out parallel in the absence of enzyme to verify the P450-dependent reaction. The organic phase was evaporated before resolving the precipitate by acetonitrile for HPLC analysis.

Conclusions
In this study, a novel thermostable cytochrome P450-T3 from the Binh Chau hot spring metagenomics database was identified, synthesized, expressed, and characterized. The enzyme had a higher optimal temperature (60 • C) compared to mesophilic cytochrome P450s with a melting temperature at 76.2 • C. P450-T3 exhibited a broad substrate binding ability, including lauric acid, palmitic acid, embelin, retinoic acid (all trans), retinoic acids (13-cis), and testosterone. We further identified heterologous redox partners for this P450 and investigated the conversion of testosterone, which was shown to be converted by P450-T3 into two products. Our results contribute to the database of natural thermostable cytochrome P450s, which is still limited so far. Furthermore, the enzyme toolbox can be expanded with thermostable P450s via an uncultured approach, which is useful for biotechnological applications in the future.