Báo cáo khoa học: Three-step hydroxylation of vitamin D3 by a genetically engineered CYP105A1 doc

In this study, we expressed the R73V⁄ R84A variant in Streptomyces lividans cells, and used the recombi-nant cells for the bioconversion of vitamin D3 to its hydroxylated metabolites.. l

engineered CYP105A1

Enzymes and catalysis

Keiko Hayashi1, Kaori Yasuda1, Hiroshi Sugimoto2, Shinichi Ikushiro1, Masaki Kamakura1, Atsushi Kittaka3, Ronald L Horst4, Tai C Chen5, Miho Ohta6, Yoshitsugu Shiro2and Toshiyuki Sakaki1

1 Department of Biotechnology, Faculty of Engineering, Toyama Prefectural University, Kurokawa, Imizu, Toyama, Japan

2 RIKEN SPring-8 Center, Harima Institute, Sayo, Hyogo, Japan

3 Faculty of Pharmaceutical Sciences, Teikyo University, Sagamiko, Kanagawa, Japan

4 Heartland Assays Inc., Ames, IA, USA

5 Boston University School of Medicine, Boston, Massachusetts, USA

6 Department of Food and Nutrition Management Studies, Faculty of Human Development, Soai University, Nanko-naka, Suminoe-ku, Osaka, Japan

Introduction

1a,25-Dihydroxyvitamin D3(1a,25(OH)2D3), the active

form of vitamin D3, mediates its biological effects

by binding to the vitamin D receptor (VDR) [1–4]

Activation of the VDR leads to the expression of genes

involved in bone and calcium metabolism, cellular prolif-eration and differentiation, immune responses, etc [5] Thus, 1a,25(OH)2D3and its analogs have been developed for the clinical treatment of rickets, osteoporosis,

Keywords

crystal structure; cytochrome P450; electron

transport chain; protein engineering;

vitamin D

Correspondence

T Sakaki, Department of Biotechnology,

Faculty of Engineering, Toyama Prefectural

University, Kurokawa, Imizu, Toyama, Japan

Fax: +81 766 56 2498

Tel: +81 766 56 7500

E-mail: tsakaki@pu-toyama.ac.jp

Database

Structural data are available in the Protein

Data Bank database under the accession

numbers 3CV8 and 3CV9

(Received 14 May 2010, revised 1 July

2010, accepted 27 July 2010)

doi:10.1111/j.1742-4658.2010.07791.x

Our previous studies revealed that the double variant of cytochrome P450 (CYP)105A1, R73V⁄ R84A, has a high ability to convert vitamin D3 to its biologically active form, 1a,25-dihydroxyvitamin D3 [1a,25(OH)2D3], suggesting the possibility for R73V⁄ R84A to produce 1a,25(OH)2D3 Because Actinomycetes, including Streptomyces, exhibit properties that have potential advantages in the synthesis of secondary metabolites of industrial and medical importance, we examined the expression of R73V⁄ R84A in Streptomyces lividans TK23 cells under the control of the tipA promoter

As expected, the metabolites 25-hydroxyvitamin D3 [25(OH)D3] and 1a,25(OH)2D3 were detected in the cell culture of the recombinant

S lividans A large amount of 1a,25(OH)2D3, the second-step metabolite of vitamin D3, was observed, although a considerable amount of vitamin D3 still remained in the culture In addition, novel polar metabolites 1a,25(R),26(OH)3D3and 1a,25(S),26(OH)3D3, both of which are known to have high antiproliferative activity and low calcemic activity, were observed

at a ratio of 5 : 1 The crystal structure of the double variant with 1a,25(OH)2D3 and a docking model of 1a,25(OH)2D3 in its active site strongly suggest a hydrogen-bond network including the 1a-hydroxyl group, and several water molecules play an important role in the substrate-binding for 26-hydroxylation In conclusion, we have demonstrated that R73V⁄ R84A can catalyze hydroxylations at C25, C1 and C26 (C27) positions of vitamin D3to produce biologically useful compounds

Abbreviations

CYP, cytochrome P450; 25(OH)D 3, 25-hydroxyvitamin D 3 ; 1a(OH)D 3 , 1a-hydroxyvitamin D 3 ; 1a,25(OH) 2 D 3 , 1a,25-dihydroxyvitamin D 3 ; 1a,25,26(OH)3D3, 1a,25,26-trihydroxyvitamin D3; FDR, ferredoxin reductase; FDX, ferredoxin.

psoriasis, secondary hyperparathyroidism, autoimmune

diseases and cancers [6,7]

The large-scale production of 1a,25(OH)2D3 from

vitamin D3 uses a bioconversion system of

Amycola-ta autotrophica, which is one of the successful

applica-tions of the P450 enzymatic reaction, on an industrial

scale [8] The primary structure of the cloned gene

reveals that the 25-hydroxylase of vitamin D3 is a

water-soluble cytochrome P450 named CYP105A2 [9]

Recently, Fujii et al [10] demonstrated that

Pseudono-cardia autotrophica P450, a member of the CYP107

family, could convert vitamin D3 to 1a,25(OH)2D3

We found that Streptomyces griseolus CYP105A1,

which has 55% amino acid homology to CYP105A2,

also has weak activities of both 25-hydroxylation

and 1a-hydroxylation of vitamin D3 to produce

1a,25(OH)2D3 [11] Recent crystal-structure analysis of

CYP105A1 revealed three arginine residues (Arg73,

Arg84 and Arg193) within the substrate-binding

pocket of CYP105A1 [12] The Ala-scan mutation

analysis indicated that the variant with R73A and

R84A showed much higher activity than the wild type,

suggesting that Arg73 and Arg84 residues may have an

inhibitory effect on activity, while Arg193 may be

essential for activity We therefore attempted to further

enhance the vitamin D hydroxylation activity of

CYP105A1 by mutating these two inhibitory Arg

resi-dues into various amino acids The resulting

double-variant R73V⁄ R84A exhibited 435- and 110-fold

higher kcat⁄ Kmvalues, respectively, for

1a-hydroxyvita-min D3 [1a(OH)D3] 25-hydroxylation and

25-hydrox-yvitamin D3 [25(OH)D3] 1a-hydroxylation, compared

with the wild-type enzyme [13] These values notably

exceed those of CYP27A1, which is a physiologically

essential vitamin D3 hydroxylase The results suggest that the R73V⁄ R84A variant could be useful for the bioconversion process to produce 1a,25(OH)2D3 from vitamin D3

In this study, we expressed the R73V⁄ R84A variant

in Streptomyces lividans cells, and used the recombi-nant cells for the bioconversion of vitamin D3 to its hydroxylated metabolites As expected, the cells pro-duced 1a,25(OH)2D3 Furthermore, a polar peak was also detected In this article we describe the identifica-tion of the novel metabolites, and the mechanism of a three-step hydroxylation by the R73V⁄ R84A variant

In addition, we evaluated the biological significance of the recombinant S lividans cells expressing the R73V⁄ R84A variant from the viewpoint of industrial and medical applications

Results

Expression of the R73V⁄ R84A variant of CYP105A1 in the recombinant S lividans TK23 cells

The R73V⁄ R84A variant of CYP105A1 was expressed

in the recombinant S lividans cells (Fig 1) The expression level of R73V⁄ R84A, 72 h after addition of the substrate, was estimated to be approximately

3 lmolÆL)1 of culture based on western blot analysis using R73V⁄ R84A purified from recombinant Escheri-chia colicells as a standard (Fig 2)

Metabolism of vitamin D3in the recombinant

S lividans cell culture Figure 3 shows HPLC profiles of vitamin D3 and its metabolites Their retention times and mass spectra (data not shown) strongly suggest that the two major metabolites are 25(OH)D3 and 1a,25(OH)2D3, These metabolites were not observed in the control S livi-dans cells, suggesting that they were produced by the R73V⁄ R84A variant expressed in the recombinant

S lividans TK23 cells These results are consistent with the results of our previous studies showing that the R73V⁄ R84A variant has the capability to convert vitamin D3 into 1a,25(OH)2D3 through 25(OH)D3 However, a novel metabolite, which is more polar than 1a,25(OH)2D3, was observed in this study The conversion ratios of 25(OH)D3, 1a,25(OH)2D3 and the more polar metabolite at 24 h were 38.9, 10.6 and 2.7%, respectively, as shown in Fig 3 The amount of 25(OH)D3 showed a maximum at 24h, and then decreased, while the amount of 1a,25(OH)2D3 gradually increased The conversion ratios of

Fig 1 Structure of the expression plasmid for the CYP105A1

vari-ant (R73V ⁄ R84A), FDX1 and FDR1 The DNA fragment harboring

three genes was inserted into HindIII and EcoRI sites of the vector

pIJ6021, as described in the Materials and methods.

1a,25(OH)2D3 at 48 and 72 h were 11.8% and 15.2%,

respectively (Fig 4) The most polar metabolite

increased linearly, and its conversion ratios at 48, 72

and 96 h were 4.7%, 8.2% and 10%, respectively

Judging from the time course of these metabolites,

the most polar metabolite appears to be a final

product derived from vitamin D3 via 25(OH)D3 and

1a,25(OH)2D3

Identification of the most polar metabolite

As shown in Fig 5, a molecular ion of m⁄ z 433

(M+H) is very small compared with major

frag-ment ions of m⁄ z 415 (M+H-H2O) and m⁄ z 397

(M+H-2H2O), suggesting that this metabolite has a

1a-hydroxyl group Based on these results, it appears

that this metabolite occurs through the further

hydrox-ylation of 1a,25(OH)2D3 1NMR studies revealed that

a signal at d1.2 ppm (6H, singlet) derived from

26,27-CH3 observed in 1a,25(OH)2D3 was not observed in the metabolite M3, suggesting that C26 or C27 of 1a,25(OH)2D3 was hydroxylated to yield M3 (data not shown) These results led us to speculate that M3 could be 1a,25(R),26(OH)3D3 or 1a,25(S),26(OH)3D3

To confirm this, we examined periodate oxidation that would cleave the C25–C26 bond of M3 to yield 25-oxo-27-nor-1a (OH)D3 After the periodate treat-ment of M3, the product was eluted at 24.1 min, while M3 was eluted at 18.9 min under the HPLC conditions described in the Materials and methods The product recovered from the HPLC eluents was analyzed using LC-MS As shown in Fig 5, its mass spectrum showed

a molecular ion of m⁄ z 401 (M+H) and fragment ions

of m⁄ z 383 (M+H-H2O) and m⁄ z 365 (M+H-2H2O), indicating that this compound is 25-oxo-27-nor-1a

Fig 3 HPLC profiles of vitamin D 3 and its metabolites formed in the recombinant Streptomyces lividans TK23 cells After culture for

48 h with 20 mgÆL)1(0.05 m M ) of vitamin D3and 0.2% 2-hydroxy-propyl-b-cyclodextrin, the cell suspension was extracted and analyzed by HPLC, as described in the Materials and methods.

Fig 2 Western blot analysis of R73V ⁄ R84A expressed in S

livi-dans TK23 cells After 72 h of culture, cell lysate prepared from

1 lL of whole-cell culture was separated by SDS ⁄ PAGE and

reacted with an antiserum against CYP105A1 after electrophoretic

transfer to a nitrocellulose filter Lane 1, purified sample of

R73V ⁄ R84A (1 pmol) prepared from the recombinant Escherichia

coli cells; lane 2, the above-mentioned cell lysate The numbers

indicate the migration points of the prestained protein marker

pro-teins (Cell Signaling Technology Inc., MA, USA): fusion of

maltose-binding protein (MBP) and paramyosin, M r 80,000; fusion of MBP

and chitin-binding protein (CBD), Mr58,000; rabbit muscle aldolase,

Mr 46,000; E coli triosephosphate isomerase, Mr 30,000; and

CBD-BmFKBP13, M r 25,000.

Fig 4 Time courses of the vitamin D 3 metabolites, 25(OH)D 3 ( ), 1a,25(OH) 2 D 3 (d), and the metabolite M3 ( ) in the recombinant Streptomyces lividans TK23 cells expressing R73V ⁄ R84A.

(OH)D3, derived from 1a,25(R), 26(OH)3D3 or

1a,25(S),26(OH)3D3 Finally, we performed

co-chroma-tography of M3 with authentic standards of

1a,25(R),26(OH)3D3 and 1a,25(S),26(OH)3D3 using a

chiral column As shown in Fig 6, the retention time

of M3 coincides with that of 1a,25(R),26(OH)3D3,

and co-chromatography of M3 with authentic

stan-dards of 1a,25(R),26(OH)3D3 and 1a,25(S),26(OH)3D3

confirmed that the dominant component of M3 is 1a,25(R),26(OH)3D3 However, the peak of M3 shows

a shoulder, probably because of the presence of 1a,25(S),26(OH)3D3 The ratio of 1a,25(R),26(OH)3D3

and 1a,25(S),26(OH)3D3 was estimated to be 5 : 1, based on HPLC data containing authentic standards

of 1a,25(R),26(OH)3D3 and 1a,25(S),26(OH)3D3 at various ratios (Fig 6F)

Fig 5 Mass spectra of the metabolite M3 (A) and its periodate-oxidation product (B) The putative structure of the periodate-oxi-dation product is shown.

Fig 6 HPLC analysis of M3 using a chiral b-cyclodextrin column The metabolite M3, authentic standards of 1a,25(R),26(OH)3D3 (R-form), and 1a,25(S),26(OH) 3 D 3 (S-form) were analyzed alone or in combination, as follows: R-form (400 pmol) (A); S-form (400 pmol) (B); R-form (200 pmol) and S-form (200 pmol) (C); R-form (200 pmol) and S-form (200 pmol) + M3 (80 pmol) (D); M3 (150 pmol) (E); and R-form (200 pmol) (A) and S-form (40 pmol) (F).

Metabolism of 1a,25(OH)2D3by the R73V⁄ R84A

variant expressed in E coli cells

To confirm the 26-hydroxylation activity of the

R73V⁄ R84A variant, the in vitro reaction, including

the R73V⁄ R84A variant and 1a,25(OH)2D3, was

examined In this experiment, overexpressed

R73V⁄ R84A variant in E coli and the reconstitution

system was used The reaction mixture was extracted,

and then analyzed by HPLC using both an ODS

column and the chiral column As expected, the

metab-olite showed the same retention time as the metabmetab-olite

M3 prepared from the recombinant S lividans TK23

cell suspension, suggesting that the R73V⁄ R84A

variant catalyzes hydroxylation of 1a,25(OH)2D3

(Fig 7) The HPLC analysis using the chiral column

showed nearly the same peak with a shoulder, as

shown in Fig 7B These results strongly suggest

that the R73V⁄ R84A variant has the capability to

convert 1a,25(OH)2D3 to 1a,25(R),26(OH)3D3 and

1a,25(S),26(OH)3D3at a ratio of 5 : 1

Comparison of kinetic parameters of the

R73V⁄ R84A variant-dependent 25-, 1a and

26-hydroxylation activities towards vitamin D3,

25(OH)D3, 1a(OH)D3and 1a,25(OH)2D3

The Km and kcat of the R73V⁄ R84A variant for

vita-min D3 25-hydroxylation were estimated to be 3.5 lm

and 0.141 min)1, respectively (Table 1) A small

amount of 1a(OH)D3 was detected, demonstrating that

the R73V⁄ R84A variant has vitamin D3

1a-hydroxyl-ation activity However, the kinetic parameters were

not estimated correctly, because the resultant

1a(OH)D3 is readily converted to 1a,25(OH)2D3, as

suggested by a much higher kcat⁄ Km value for

1a(OH)D3 25-hydroxylation than for any other

reac-tions detailed in Table 1 The Km and kcat of the

R73V⁄ R84A variant for 25(OH)D3 1a-hydroxylation

were estimated to be 2.2 lm and 0.136 min)1,

respec-tively [12] (Table 1), while 25(OH)D3 26-hydroxylation

activity was not observed Regarding the metabolism

of 1a(OH)D3, the R73V⁄ R84A variant prefers

25-hydroxylation to 26-hydroxylation, judging from

no detection of 1a,26(OH)2D3 (Table 1) Therefore,

among the four substrates examined in this study,

26-hydroxylation was only observed in the metabolism

of 1a,25(OH)2D3 The kinetic parameters Kmand kcat

of the R73V⁄ R84A variant for hydroxylation at C26

or C27 of 1a,25(OH)2D3 to yield 1a,25,26(R)(OH)3D3

and 1a,25,26(S)(OH)3D3 were estimated to be 2.2 lm

and 0.136 min)1, respectively (Table 1) It should be

noted that the kcat⁄ Km value for 1a,25(OH)2D3

26-hydroxylation is similar to those for vitamin D3

25-hydroxylation and for 25(OH)D31a-hydroxylation

Comparison of the stability of the oxygenated form of wild type and the double variants of CYP105A1

Our previous studies revealed that the double variants

of CYP105A1 – R73A⁄ R84A and R73V ⁄ R84A – have much higher activity than the wild-type CYP105A1 One of the most essential reasons for the high activity appears to be a significant increase in the coupling effi-ciency between product formation and NADPH oxida-tion Therefore, we examined the stability of the oxygenated forms of the wild-type CYP105A1 and its variants

Figure 8 shows spectral analysis of the NADPH-dependent reduction of heme iron of the P450s Although the wild-type CYP105A1 showed a rapid conversion of P450 to P420, the double variants R73A⁄ R84A and R73V ⁄ R84A showed a peak at

450 nm, suggesting that the oxygenated forms of these

Table 1 Kinetic parameters of the R73V ⁄ R84A variant of CYP105A1 VD3, vitamin D3. ND, kinetic parameters could not be determined although the activity was detected; –, the activity was not detected.

Substrate Hydroxylation K m (l M ) k cat (min)1) k cat ⁄ K m

VD3 25 3.5 0.141 0.040

25(OH)D3 1a 2.2 0.136 0.062

1a(OH)D3 25 6.5 2.12 0.33

1a,25(OH) 2 D 3 26 1.2 0.051 0.043

Fig 7 HPLC analysis of the metabolite M3 produced by R73V ⁄ R84A expressed in Escherichia coli cells 1a,25(OH) 2 D 3 and its metabolite, M3, were analyzed by RP-HPLC (A), and the chiral b-cyclodextrin column (B) as described in the Materials and methods.

variants are more stable than the wild type This

sta-bilization was also observed in the single variant

R84A, whereas R73A showed no such stabilization

Thus, the stabilization of the oxygenated form

origi-nates from the mutation of Arg84

Discussion

Our new discoveries in this study are summarized as

follows First, the double variants of CYP105A1 have

26-hydroxylation activity towards 1a,25(OH)2D3

Thus, they catalyze a three-step hydroxylation at C25,

C1 and C26 (C27) positions of vitamin D3 to yield

1a,25(R),26(OH)3D3 and 1a,25(S),26(OH)3D3 Second,

the putative reasons why the double variants have

much higher activities than the wild type were

demon-strated (a) The mutation of Arg84 to alanine stabilizes

the oxygenated form of P450, thus increasing the

cou-pling efficiency between product formation and

NADPH oxidation and (b) the mutation of Arg73 to

alanine or valine produces a hydrogen-bond network formed by the re-arrangement of water molecules sur-rounding the 1a-hydroxyl group Third, the recombi-nant S lividans cells expressing the R73V⁄ R84A variant of CYP105A1 are promising candidates for producing the active forms of vitamin D3 as useful drugs

In this study, we selected the R73V⁄ R84A variant as

a catalyst and S lividans as a host for the following reasons (a) S lividans has been utilized as a potential host for protein production, (b) the codon usage and

GC content of the CYP105A1 gene are similar to those of S lividans genomic DNA, (c) any ferredoxins

of S lividans and their reductases may act as an elec-tron donor for the R73V⁄ R84A variant based on the fact that CYP105A2 showed vitamin D3 25-hydroxyl-ation activity in S lividans cells [9] and (d) Actinomy-cetes, including Streptomyces, exhibit potential advantages in the synthesis of secondary metabolites

of industrial and medical importance in the bioconver-sion processes Consequently, S lividans TK23 cells and the vector pIJ6021 were used as a host–vector system, and the R73V⁄ R84A variant was expressed under the control of the tipA promoter induced by thiostrepton

The metabolites of vitamin D3 were detected in the recombinant S lividans cell culture Unexpectedly,

a large amount of 1a,25(OH)2D3, the second-step metabolite of vitamin D3, was observed, although most

of the added vitamin D3 still remained in the culture

In addition, a novel metabolite peak was observed (Fig 3) These results appear to be inconsistent with our previous in vitro studies using recombinant enzymes in a reconstituted system [13] However, it is possible that the efficiency of cell uptake is different among vitamin D3, 25(OH)D3 and 1a,25(OH)2D3, based on their hydrophobicity and affinity for 2-hydroxypropyl-b-cyclodextrin added to the culture Among the three compounds, the uptake of vitamin

D3 into S lividans cells appears to be the least efficient By contrast, 1a,25(OH)2D3 appears to read-ily cross the cell membrane and enter the cell In addition to 25(OH)D3 and 1a,25(OH)2D3, we found novel metabolites – 1a,25(R),26(OH)3D3 and 1a,25(S),26(OH)3D3– which were identified by LC-MS analysis with periodate treatment, NMR studies and HPLC analysis using a chiral column It should be noted that 1a,25(R),26(OH)3D3 and 1a,25(S),26(OH)3D3 can be separated by using the chiral column, although the previous separation method required acetylation [14] R73V⁄ R84A and R73A⁄ R84A prepared from the recombinant E coli cells also showed the conversion of 1a,25(OH)2D3 to

Fig 8 NADPH-dependent reduced CO-difference spectra of

CP105A1 (A) and the R73A ⁄ R84A variant (B) NADPH-dependent

reduced CO-difference spectra were measured at 15-second

inter-vals in the presence of 2 l M wild-type CP105A1 (A) or the

R73A ⁄ R84A variant (B), 0.02 mgÆmL)1 of spinach ferredoxin,

0.02 UÆmL)1 of spinach ferredoxin-reductase, 10 l M of 1a(OH)D 3 ,

and 1 m M NADPH Very rapid conversion from P450 to P420 was

observed in the wild-type CP105A1 By contrast, R73A ⁄ R84A

showed a P450 peak, indicating that the oxygenated form of

R73A ⁄ R84A is more stable than the wild type.

1a,25(R),26(OH)3D3 and 1a,25(S),26(OH)3D3 These

results confirm that these variants have

26-hydroxyl-ation activity towards 1a,25(OH)2D3 In our previous

studies, 1a,25(OH)2D3 was considered to be the final

product, but 1a,25(OH)2D3 was found to be a

sub-strate for 26-hydroxylation in this study Figure 9

shows the reported crystal structure with

1a,25(OH)2D3 [Protein Data Bank (PDB) code 3cv9]

[13] which is superimposed by a docking model of

1a,25(OH)2D3 in the active site for 26-hydroxylation

The distance (13.5 A˚) from C26 to heme iron in the

crystal structure appears to be too far to simulate the

reactive structure of the enzyme–substrate for

26-hydroxylation Thus, we need to construct a docking

model for the 26-hydroxylation As shown in Fig 9,

the distance (3.6 A˚) from C26 to heme iron in the

docking model is suitable for the 26-hydroxylation

A small rotation of a secosteroid skeleton, and a

considerable conformational change of the side chain

in the heme pocket, will make the 26-hydroxylation possible It is noted that Fig 9 shows a docking model to yield more 1a,25(R),26(OH)3D3 than 1a,25(S),26(OH)3D3 It is possible that the distance from C26 or C27 to heme iron, and the orientations of the hydrogen atoms at C26 or C27, decide the ratio between 1a,25(R),26(OH)3D3and 1a,25(S),26(OH)3D3 Regarding the metabolism of 25(OH)D3, we did not detect any 25,26(OH)2D3, and only 1a,25(OH)2D3 was observed as a metabolite This strongly suggests that the 1a-hydroxyl group plays an important role in the 26-hydroxylation As shown in Fig 6, the 1a-hydroxyl group forms a hydrogen bond with a water molecule with the distance of 2.9 A˚ In addition, the bound water molecule forms a hydrogen-bond network involving the A92 amide group, the Q93 side chain and four other waters in the heme pocket Therefore,

it is possible that this hydrogen-bond network, includ-ing the 1a-hydroxyl group, is responsible for the stable binding of 1a,25(OH)2D3 for 26-hydroxylation In a previous report, we proposed that the reason for increased activity by the mutation on R73 might be the direct effect of its size and charge on the confor-mation of the hydrophobic substrate in the large active-site pocket Because the single variant R84A (PDB code 2zbz, 1.9A˚ resolution) does not have this hydrogen-bond network [12], it implies that the hydro-gen-bond network formed by the re-arrangement of the water molecule surrounding the 1a-hydroxyl group

is associated with the high activity of R73 variants Figure 10 summarizes the metabolic pathways of vitamin D3 by the R73V⁄ R84A variant Although the first step contains both 25-hydroxylation and 1a-hydroxylation, R73V⁄ R84A prefers the former The minor product, 1a(OH)D3, is a good substrate of R73V⁄ R84A for the 25-hydroxylation to yield 1a,25(OH)2D3 It seems likely that the 1a-hydroxyl group of 1a(OH)D3 forms a hydrogen bond with the bound water molecule as well as 1a-hydroxyl group of

Fig 9 Hydrogen-bond network involving the 1aOH group of

1a,25(OH)2D3 and water molecules observed in the crystal

struc-ture of the R73A ⁄ R84A variant The docking model of

1a,25(OH)2D3 for 26-hydroxylation is superposed and shown in

green The water molecules are shown as red spheres The

hydro-gen bonds are shown as broken lines.

Fig 10 Metabolic pathways of vitamin D3

catalyzed by the R73V ⁄ R84A variant of

CYP105A1.

1a,25(OH)2D3 (Fig 9) However, it should be noted

that R73V⁄ R84A much prefers 25-hydroxylation to

26-hydroxylation towards 1a(OH)D3 Accordingly, both

pathways of the second steps produce 1a,25(OH)2D3

In this study, we found the third step, which

con-verts 1a,25(OH)2D3 into 1a,25(R),26(OH)3D3 and

1a,25(S),26(OH)3D3 Thus, the major pathway occurs

in the order of 25-hydroxylation, 1a-hydroxylation and

26-hydroxylation It is possible to assume that the first

product, 25(OH)D3, is released from the heme pocket

and re-enters the heme pocket in the opposite direction

for 1a-hydroxylation, and the resultant product,

1a,25(OH)2D3, is released from the heme pocket and

again re-enters the heme pocket in the opposite

direc-tion for 26-hydroxyladirec-tion

It is well known that 1a,25(OH)2D3has potent

anti-proliferative effects in many cancer-cell types,

includ-ing breast and prostate cancers However,

1a,25(OH)2D3is not suitable as a therapeutic agent for

cancer treatment, because its systemic administration

causes hypercalcemia and hypercalciuria Therefore, a

less calcemic vitamin D analog is promising for cancer

treatment As reported previously, the calcemic effect

of 1a,25(R),26(OH)3D3 or 1a,25,26-trihydroxyvitamin

D3 [1a,25,26(OH)3D3] is significantly lower than that

of 1a,25(OH)2D3, whereas its antiproliferative activity

is nearly the same [15–18] These findings suggest that

1a,25(R),26(OH)3D3 is suitable for anticancer

treat-ment Although conversion of 1a,25(OH)2D3 into

1a,25(R),26(OH)3D3 and 1a,25(S),26(OH)3D3 is

unfa-vorable for practical use of the recombinant S lividans

cells to produce 1a,25(OH)2D3, it is quite favorable for

the production of a promising anti-cancer compound

Recently, Peng et al [19] and Lou et al [20] claimed

that 25(OH)D3was able to act as a hormone in

mam-mary gland using CYP27B1 [25(OH)D3

1a-hydroxyk-ase] knockout mice They concluded that 25(OH)D3

could be developed as a nontoxic, natural,

chemoventive agent for further development for cancer

pre-vention Based on these results, all three compounds

produced in this study appear to be clinically useful

In this study, we attempted and succeeded in

estab-lishing a bioconversion system to convert vitamin D3

into its multiple hydroxylated metabolites, including

25(OH)D3, 1a,25(OH)2D3 and 1a,25,26(OH)3D3, by

using recombinant S lividans cells expressing a variant

of CYP105A1 It should be noted that we did not

con-firm the expression of ferredoxin 1 (FDX1) and

ferre-doxin reductase 1 (FDR1) genes on the expression

plasmid Recombinant S lividans cells harboring a

plasmid containing the R73V⁄ R84A variant of

CYP105A1, and FDX1 genes without the FDR1 gene,

showed activity that was not so different from those

harboring a plasmid containing three genes These results strongly suggest that some endogenous FDR(s)

of S lividans function as electron donors

In order to enhance the productivity of these metab-olites, we plan to further mutate CYP105A1, increase its expression level, optimize the electron-transfer sys-tem and modify culture conditions

Materials and methods

Materials

DNA-modifying enzymes, restriction enzymes and E coli SCS110 were purchased from Takara Shuzo Co Ltd (Kyoto, Japan) S lividans TK23 cells and the vector pIJ6021 [21] were kindly provided by Dr Hiroyasu Onaka

of Toyama Prefectural University The HPLC column SU-MICHIRAL OA-7000 (4.6· 250 mm) was purchased from Sumika Chemikcal Analysis Service, Ltd (Osaka, Japan) Ferredoxin and NADPH-ferredoxin reductase from spinach were purchased from Sigma (St Louis, MO, USA) Vitamin

D3, 1a(OH)D3, 1a,25(OH)2D3, glucose dehydrogenase, catalase, NZ-casein, NZ-amine, corn steep liquor and 2-hydroxypropyl-b-cyclodextrin were purchased from Wako Pure Chemical Industries, Ltd (Osaka, Japan) 25(OH)D3 was purchased from Funakoshi Co Ltd (Tokyo, Japan) 1a,25(R),26(OH)3D3and 1 a,25(S),26(OH)3D3were chemi-cally synthesized, as described by Partridge et al [14] Meat extract was purchased from Kyokuto Pharmaceuticals Co (Tokyo, Japan) Bacto-soytone, bacto-peptone and yeast extract were obtained from Difco Laboratories (Detroit,

MI, USA) S griseolus CYP105A1 and ferredoxin genes were kindly provided by Sumitomo Chemical Co Ltd (Takarazuka, Japan) NADPH was purchased from Orien-tal Yeast Co Ltd (Tokyo, Japan) Other chemicals used were of the highest quality commercially available

Expression of the R73V⁄ R84A variant of CYP105A1 in S lividans TK23 cells

Variants were generated using the Quick Change Site-directed Mutagenesis kit (Stratagene, La Jolla, CA, USA) using the CYP105A1 gene as a template The oligonucleo-tide primers used for mutagenesis to yield the variant R73V⁄ R8A were as described previously [13] For construc-tion of the expression plasmid, E coli SCS110, which has

no ability to methylate DNA, was used as a host The PCR fragment containing R73V⁄ R84A and FDX1 genes, with HindIII and PstI restriction sites, was prepared using the primers 5¢-ACCAAGCTTATGAAAAGATACCGCCAC-GACG-3¢ and 5¢-TTCTGCAGTCACCAGGTGACCGG-GAGTTCG-3¢, and the expression plasmid for R73V⁄ R84A in E coli cells as a template DNA The resul-tant PCR fragment was digested with HindIII and PstI,

and inserted into the HindIII and PstI sites of pUC19 The

PCR fragment containing the Streptomyces coelicolor

FDR1 gene [22], with PstI and EcoRI restriction sites,

which contains approximately 170 bp 5¢ upstream region

from the initial codon ATG, was prepared using the

prim-ers 5¢-AACTGCAGCCGTCCCCACGCCTGCGTCACC-3¢

and

5¢-CGAATTCTCAGGCGCCGCTCTCGCGGAGCA-3¢, and total DNA of S coelicolor cells as a template The

resultant PCR fragment was digested with PstI and EcoRI,

and inserted into the PstI and EcoRI sites of pUC19

Then, the HindIII⁄ PstI fragment containing R73V ⁄ R84A

and FDX1 genes, and the PstI⁄ EcoRI fragment containing

the FDR1 gene prepared from the cloning vectors was

dou-bly inserted into HindIII and EcoRI sites of the vector

pIJ6021

The resultant co-expression plasmid for R73V⁄ R84A,

FDX1 and FDR1 was introduced into protoplasts of S

liv-idansTK23 according to the method described by

Horinou-chi et al [23] After 3 days of incubation at 30C,

kanamycin-resistant colonies were obtained

Culture of the recombinant S lividans cells

For the preculture, the transformant on Bennett’s agar

plate containing 0.1% yeast extract, 0.1% meat extract,

0.2% NZ-amine, 1% maltose, 2% agar and kanamycin at

a final concentration of 15 lgÆmL)1 was inoculated into

medium (10 mL) containing 1% soluble starch, 0.5%

glucose, 0.3% NZ-casein, 0.2% yeast extract, 0.5%

Bacto-peptone, 0.1% KH2PO4, 0.05% MgSO4Æ7H2O, 0.3%

CaCO3 and kanamycin at a final concentration of

15lg⁄ ml, and then vigorously shaken at 30 C for 3 days

The preculture (2ml) was inoculated into the sterilized

medium (100 mL) containing 1.5% glucose, 1.5% Difco

bacto Soytone, 0.5% Corn steep liquor, 0.5% NaCl, 0.2%

CaCO3 and kanamycin at a final concentration of

15lg⁄ ml, and incubated on a rotary shaker for 3 days at

200 rpm and 30C Thiostrepton was then added into the

medium at a final concentration of 10lg⁄ ml to induce

expression of R73V⁄ R84A variant

HPLC analysis of vitamin D3metabolites

produced in the recombinant S lividans cells

Twenty-four hours after the addition of thiostrepton,

2-hy-droxypropyl-b-cyclodextrin solution (200 mgÆmL)1of water)

and vitamin D3-ethanol solution (20 mg in 0.5 mL of

etha-nol) were added to 100 mL of culture of the recombinant

S lividans cells Aliquots of the culture were extracted at

0, 24, 48 and 72 h with four volumes of chloroform⁄ methanol

(3 : 1, v⁄ v) The organic phase was recovered and dried

in a vacuum evaporator centrifuge (Sakuma Seisakusyo,

Tokyo, Japan) The resulting residue was solubilized with

acetonitrile and applied to HPLC under the following

con-ditions: column, YMC-Pack ODS-AM (4.6· 300 mm)

(YMC Co., Tokyo, Japan); UV detection, 265 nm; flow rate, 1.0 mLÆmin)1; column temperature, 40C; mobile phase, a linear gradient of 70-100% acetonitrile aqueous solution in 15 min followed by 100% acetonitrile for

25 min

The metabolite designated as M3, which was recovered from the RP-HPLC, was further analyzed by HPLC using

a chiral b-cyclodextrin column under the following condi-tions: column, SUMICHIRAL OA-7000 (4.6· 250 mm) (Sumika Chemikcal Analysis Service, Ltd); UV detection,

265 nm; flow rate, 0.7 mLÆmin)1; column temperature,

25C; mobile phase, methanol ⁄ water (85 : 15, v ⁄ v)

LC-MS analysis of metabolites produced in the recombinant S lividans cells

Isolated metabolites from HPLC effluents were subjected to mass spectrometric analysis, using a Finnigan LCQ ADVANTAGE MIX (ThermoFisher SCIENTIFIC, Wal-tham, MA, USA) with atmospheric pressure chemical ioni-zation in the positive mode The conditions of LC were: column, reverse-phase ODS column (2· 150 mm, Develosil ODS-HG-3; Nomura Chemical Co Ltd, Aichi, Japan); mobile phase, acetonitrile⁄ methanol ⁄ water (3 : 4 : 3,

v⁄ v ⁄ v); flow rate, 0.2 mLÆmin)1; UV detection, 265 nm

Periodate oxidation

Periodate oxidation was carried out to identify the presence

of vicinal diol or an a-keto group The metabolite M3 was dissolved in 15 lL of methanol, and 10 lL of 5% sodium metaperiodate aqueous solution was added as described by Reinhardt et al [24] After 30 min of incubation at 25C, the solution was dried in vacuo and the resultant residue was analyzed by HPLC and LC-MS The HPLC conditions were

as follows: column, YMC-Pack ODS-AM (4.6· 300 mm) (YMC Co.); UV detection, 265 nm; flow rate, 1.0 mLÆmin)1; column temperature, 40C; mobile phase, a linear gradient

of 20–100% acetonitrile aqueous solution per 25 min fol-lowed by 100% acetonitrile for 12 min

1H-NMR analysis of the metabolite produced by the recombinant S lividans cells

The 400-MHz 1H-NMR spectra of the metabolite M3 was measured on a BRUKER-400 (1H, 399.9 MHz) The metabolite M3 was dissolved in 500 lL of CD3OD and transferred into a probe

Preparation of anti-CYP105A1 serum

Purified wild-type CYP105A1 (0.5 mg) was mixed with an equal volume of Freund’s complete adjuvant, and then injected subcutaneously into a young male rabbit weighing

2.5 kg After 2 and 4 weeks, the rabbit was boosted with

1.0 mg of the antigen in emulsified Freund’s incomplete

adjuvant, and bled 1 week after the final injection

Western blot analysis of the R73V⁄ R84A variant

of CYP105A1 expressed in S lividans cells

Cytosolic fractions prepared from recombinant S lividans

cells were subjected to SDS⁄ PAGE on a 10% gel and

trans-ferred to a poly(vinylidene difluoride) membrane

Immun-odetection was performed using the anti-CYP105A1 serum

as the primary antibody, and alkaline

phosphatase-conju-gated goat anti-rabbit IgG as the secondary antibody

Spectral analysis

The reduced CO-difference spectra of the R73V⁄ R84A

vari-ant expressed in E coli cells were measured using a Hitachi

U-3310 spectrophotometer with a head-on photomultiplier

(Tokyo, Japan) [25] The P450 content of the purified

wild-type or variants of CYP105A1 was estimated using a molar

extinction coefficient of 110 mm)1Æcm)1at 417 nm [11]

To measure the NADPH-dependent reduction rate of

heme iron, reduced CO-difference spectra were measured at

15-second intervals in the presence of 2 lm wild-type or

variants of CYP105A1, 0.02 mgÆmL)1of FDX, 0.02 UÆmL)1

of FDR, 10 lm 1a(OH)D3and 1 mm NADPH

Metabolism of 1a,25(OH)2D3by the R73V⁄ R84A

variant expressed in E coli cells

The metabolism of 1a,25(OH)2D3was examined in a

recon-stituted system containing 0.2 lm R73V⁄ R84A,

0.1 mgÆmL)1 of spinach ferredoxin, 0.1 UÆmL)1 of spinach

ferredoxin reductase, 1 UÆmL)1 of glucose dehydrogenase,

1% glucose, 0.1 mgÆmL)1 of catalase, 1 mm NADPH,

0.5–10.0 lm of the substrate, 100 mm Tris⁄ HCl (pH7.4)

and 1 mm EDTA, at 30C, as described previously [13]

Other methods

The concentrations of vitamin D3 derivatives were

esti-mated by their molar extinction coefficient of 1.80·

104m)1Æcm)1 at 264 nm [26] Docking of 1a,25(OH)2D3to

the variant enzyme was performed by procedures described

previously [12] using the Lamarckian Genetic Algorithm

(LGA) method implemented in Autodock 4 [27] The model

of the R73V⁄ R84A variant was generated from the crystal

structure of R73A⁄ R84A (PDB code 3cv9), and used as a

template for the docking calculations The hydrogenated

protein and atomic charges of the residues were calculated

using the method described in the Autodock tools package

The 1a,25(OH)2D3was treated as a flexible ligand, and the

side chains of V73, F85, V88 and Q93 were treated as a

flexible residue Water molecules numbered Wat560, 587 and 593 in the binding site were included in the calculation Cluster analysis was performed on the docked results of

100 runs using a root-mean-square tolerance of 2.0 A˚ Other parameters were left at the default values The figure

of protein structure was prepared using pymol software (http://www.pymol.org)

Acknowledgement

This work was supported, in part, by a Ministry of Education, Culture, Sports, Science and Technology grant, and the Novozymes Japan Research Fund

References

1 Norman AW, Myrtle JF, Midgett RJ, Nowicki HG, Williams V & Popjak G (1971) 1,25-dihydroxycholecal-ciferol: identification of the proposed active form of vitamin D3 in the intestine Science 173, 51–54

2 Holick MF, Schnoes HK & DeLuca HF (1971) Identifi-cation of 1,25-dihydroxycholecalciferol, a form of vita-min D3 metabolically active in the intestine Proc Natl Acad Sci USA 68, 803–804

3 Lawson DE, Fraser DR, Kodicek E, Morris HR & Williams DH (1971) Identification of 1,25-dihydroxy-cholecalciferol, a new kidney hormone controlling calcium metabolism Nature 230, 228–230

4 Boyle IT, Gray RW & DeLuca HF (1971) Regulation

by calcium of in vivo synthesis of 1,25-dihydroxychole-calciferol and 21,25-dihydroxychole1,25-dihydroxychole-calciferol Proc Natl Acad Sci USA 68, 2131–2134

5 Jones G, Strugnell SA & DeLuca HF (1998) Current understanding of the molecular actions of vitamin D Physiol Rev 78, 1193–1231

6 Bouillon R, Okamura WH & Norman AW (1995) Structure-function relationships in the vitamin D endo-crine system Endocr Rev 16, 200–257

7 Masuda S & Jones G (2003) Vitamin D analogs – drug design based on proteins involved in vitamin D signal transduction Curr Drug Targets Immune Endocr Metabol Disord 3, 43–66

8 Sasaki J, Miyazaki A, Saito M, Adachi T, Mizoue K, Hanada K & Omura S (1992) Transformation of vita-min D3 to 1 alpha,25-dihydroxyvitavita-min D3 via 25-hydroxyvitamin D3 using Amycolata sp strains Appl Microbiol Biotechnol 38, 152–157

9 Kawauchi H, Sasaki J, Adachi T, Hanada K, Beppu T

& Horinouchi S (1994) Cloning and nucleotide sequence

of a bacterial cytochrome P-450VD25 gene encoding vitamin D-3 25-hydroxylase Biochim Biophys Acta

1219, 179–183

10 Fujii Y, Kabumoto H, Nishimura K, Fujii T, Yanai S, Takeda K, Tamura N, Arisawa A & Tamura T (2009)