Introduction

Androst-4-ene-3,17-dione (AD) is an important precursor in the synthesis of pharmaceutical steroid, and it can be used to synthesize almost all kinds of steroidal medication (YuKimatsu and Kaji 1979). Mycobacterium neoaurum is the species that is protected by numerous patents to obtain AD from the side-chain cleavage of phytosterol (PS) (Fernandes et al. 2003; Szentirmai 1990; Wovcha et al. 1979). However, the production of AD is often associated with androst-1,4-diene-3,17-dione (ADD). Thus, separating these substances in downstream processing is difficult because of their high structural similarity, which reduces product yield and affects product quality (Malaviya and Gomes 2008; Wei et al. 2010; Zhang et al. 2013).

Current research has shown that it is 3-ketosteroid-Δ1-dehydrogenase (KsdD) that change AD into ADD by introducing a double bond into the C1–C2 position of the 3-ketosteroid A-ring. Several strains of bacteria (Donova 2007), such as Arthrobacter sp. (Fokina et al. 1997), Pseudomonas sp. (Adham et al. 2003; Plesiat et al. 1991), Rhodococcus (Geize et al. 2000, 2001) and Mycobacterium (Wei et al. 2010) exhibit C1,2 dehydrogenation. For the significant industrial applications, KsdD from Mycobacterium has long attracted wide attention, but the studies mostly focused on mutation breeding by physical and chemical methods (Marsheck et al. 1972; Zhong et al. 2009), exploring novel enzyme by bioinformation analysis (Bragin et al. 2013; Geize et al. 2007; Shtratnikova et al. 2014; Zhang et al. 2015) and engineering of Mycobacterium by molecular approaches (Wei et al. 2010), the study on the structure and active sites residues of KsdD from Mycobacterium is tedious.

The crystal structure of the KsdD enzymes of Mycobacterium is not yet available. Studies about the structure and active sites of KsdD mostly focused on the KsdD from Rhodococcus (Fujii et al. 1999; Geize et al. 2002; Matsushita and Itagaki 1992). Oosterwijk et al. (2012) reported the first crystal structure of KsdD4 from Rhodococcus jostii RHA1. Rohman et al. (2012, 2013) reported the crystallization and X-ray crystallographic analysis of KsdD from Rhodococcus erythropolis SQ1. These paper suggested that, different from elimination (E2) mechanism, the enzyme catalyzed dehydrogenation more likely follows a stepwise unimolecular elimination conjugate base (E1cB) mechanism, in which departure of the first hydrogen atom precedes that of the second hydrogen accompanied by the formation of an intermediate. During this progress, Y487 and G491 act as an electrophile to promote labilization of the C2 hydrogen atoms of 3-ketosteroid substrate. Meanwhile, Y318, which is assisted by Y119, as a general base abstract a proton from the C2 atom and result in either an enolate or a carbanionic intermediate; subsequently, a double bond between the C1 and C2 atoms is formed.

Sructural and active sites information on KsdD from Mycobacterium is still limited, which restricts the progress of structure-based engineering of KsdD. Meanwhile, the variation of active sites of KsdD from Mycobacterium have significant impact in the production of AD. Liu et al. (2015) revealed the mutant KsdD gene resulted in the change of AD/ADD molar ratios. Shao et al. (2016) reported that Val366 of KsdDI (from M. neoaurum ST-095) changed to Ser366 of KsdDII (from M. neoaurum JC-12) affected KsdD activity (Zhong et al. 2009). Bragin et al. (2013) and Shtratnikova et al. (2014) sequenced the full genome of Mycobacterium spp. VKM Ac-1815D and 1816D and pointed out that one of single nucleotide polymorphisms (SNPs), located in a single-copy ksdD gene, caused the major practical difference between VKM Ac-1815D and 1816D: production of AD and ADD, respectively, as major products from sterols (Marsheck et al. 1972).

In our previous report, Shen et al. (2012) found that M. neoaurum TCCC 11028 (MNR) and its derived mutant strain M. neoaurum TCCC 11028 M3 (MNR M3) significantly differ in the ADD/AD ratio, with 44.5:1 and 1:8.3, respectively. The change of S138 in KsdD-MNR to L138 in KsdD-MNR M3 because of C413T in the KsdD gene, significantly affected its activity (Xie et al. 2015).

Given the significance of the sites of KsdD from Mycobacterium in the production of AD and to have a direction in the engineering of KsdD from Mycobacterium, we focused on the structure and active site residues of KsdD from Mycobacterium.

In the present study, we aimed to obtain the structure information of the active sites of KsdD enzymes from Mycobacterium. Thus, mutant genes KsdD Y125F, Y365F, Y541F, Y122F, S138A, E140V, and Y472F were constructed based on in silico protein docking modeling using site-directed mutagenesis and exogenous expression. Biotransformation experiments were also performed. Changes in the enzyme activity of mutants were determined by the difference of the conversion rate. The structural mechanism of mutations in KsdD of Mycobacterium was analyzed based on this process. The information presented facilitate the manipulation of the catalytic properties of the enzyme to improve its application in the pharmaceutical steroid industry.

Materials and methods

Chemicals

AD and ADD were obtained from Sigma-Aldrich Co. (USA). All of the utilized chemical solvents and salts were of analytical grade or higher.

Bacterial strains and culture conditions

Mycobacterium neoaurum TCCC 11028 (MNR) used in this study was obtained from Tianjin University of Science and Technology Culture Collection Center (TCCC), Tianjin, China. The methods of microbial cultivation and bioconversion were established by Shen et al. (2012). Escherichia coli DH5α and E. coli BL21 (DE3), which express wild-type KsdD and KsdD with mutated amino acids, were grown at 37 °C overnight in LB medium containing 50 μg/mL ampicillin and shaken at 180 rpm. This 5 mL pre-culture was seeded by 1–100 in 800 mL of fresh LB medium with 25 μg/mL ampicillin. The culture was grown at 37 °C and shaken at 200 rpm. When optical density at 600 nm reached 0.8, 100 mM isopropyl-β-d-thiogalactopyranoside was added to trigger the expression of KsdD. E. coli cells were grown in 3 L flasks at 16 °C for 18 h, with shaking at 200 rpm. The cultures were centrifuged at 5000×g for 10 min. The cell pellets were washed three times with 20 mM PBS buffer (100 mM dipotassium hydrogen phosphate-potassium dihydrogen phosphate, 1% glycerol, pH 7.3) and resuspended in the same buffer to an optical density of 10 at 600 nm using a Shimadzu model UV-1601PC spectrophotometer (Kyoto, Japan).

In silico generation of protein structure and molecular structures

For the reason of KsdD from R. erythropolis SQ1 shared 40.83% sequence identity and 42.60% sequence similarity to KsdD from MNR (sequence coverage is 98%), the crystal structure of KsdD from R. erythropolis SQ1 could be used as the template in the homologous modeling. Based on the known crystal structure of KsdD from R. erythropolis SQ1 (PDB:4C3Y, 2.3 Å resolution), the 3D structure of KsdD from MNR was homologically modeled using MODeller 9.15 program (http://salilab.org/modeller/). The 3D structural information of AD, which was used as the ligand, was handled using Sybyl2.0 program.

Docking studies

To understand the binding model of compound AD with KsdD, compound AD was docked into the active site by Sybyl2.0 program. The molecular construction and energy minimization of compound AD was prepared by Sybyl2.0 program. Based on the conformation of KsdD docking with AD, all amino acids in the substrate-binding groove of KsdD and in proximity to AD within 5 Å were located by PyMol software (http://pymol.org/). Swiss-model (http://swissmodel.expasy.org/) was used to simulate the conformation of the site mutants.

Construction of the mutant KsdD genes and expression of KsdD

KsdD gene were codon-optimized for expression in E. coli and synthesized by BGI-Shenzhen. To amplify KsdD, forward primer KsdD exp-F (5′-GGAATTCCATATG GTGTTTTATATGACCGCGCAGG-3′) was design, including start codon, Nde I restriction site (underlined), and reverse primer KsdD exp-R (5′-CGGGATCCGACGCTTTGCCCGCTAAATGTA-3′) with removed stop codon and included a BamH I restriction site (underlined). PCR product KsdD (1701 bp) was cloned into Nde I/BamH I-digested pET-22b(+) to introduce a C-terminal His tag into this protein. The resulting construct pET-22b(+)-KsdD was transformed into E. coli BL21 (DE3). Mutagenesis was constructed by overlap PCR. Two-step PCR was essential to introduce a single amino acid mutation. For each mutation, a new pair of mutagenic primer was designed according to the nucleotide sequence of KsdD (Table 1). The mutated nucleotides are shown as capital letters. Reverse mutagenic primer sequences were complementary to the forward primers and had the same lengths. First, the mutagenic primers were compared with outside primers to introduce base mutation. A pair of DNA fragment, in which one side overlapped with the other, was amplified. Mutation was then introduced into the overlapped region. After purification, the PCR product was used as the template to amplify the total KsdD using outside primers.

Table 1 primers used in this study
Full size table

PCR product–whole KsdD (1701 bp) was inserted into Nde I/BamH I-digested pET-22b(+), followed by the transformation of ligation product into E. coli DH5α. Individual clones of these E. coli variants were grown, and their plasmids were isolated to identify the result of the clones digested by Nde I/BamH I. After confirmation through DNA sequencing, the recombinant vector was transformed into E. coli BL21 (DE3) for exogenous protein expression.

Biotransformation reactions

The resuspension was incubated with AD at a final concentration of 1.2 g/L at 20 °C with shaking at 200 rpm. The samples were withdrawn at various time intervals and extracted with the same volume of ethyl acetate. The organic phases were collected and detected through thin layer chromatography (TLC) and HPLC. Wild E. coli BL21 (DE3)/pET-22b(+)-KsdD was used as negative control. The transformation products detected by TLC and HPLC were prepared and analyzed following the procedures described by Shen et al. (2012). Solid extracts which was dried in vacuum from organic phases, were redissolved in the HPLC mobile phase and filtered through a 0.22 μm filter before HPLC analysis. HPLC analysis was performed on a Kromasil C18 column (250 × 4.6 mm) with methanol/water (4:1, v/v) as mobile phase at 30 °C with UV absorbance detection at 254 nm. The concentrations of AD and ADD were determined from the calibration curves of eluent solutions of standard AD and ADD.

Results

In silico design of mutant KsdD

KsdD is a FAD dependent enzyme. The putative N-terminal FAD-binding motifs in the KsdD was consistent with the sequence as previously described (Geize et al. 2000; Wei et al. 2010; Rohman et al. 2013) GSG(A/G)(A/G)(A/G)X17E (Fig. S1). During the homological modelling, FAD in 4c3y was reserved. As shown in Fig. S3, this model was accepted and can be used for the following research.

As shown in Fig. 1, KsdD can be divided into FAD-binding domain and catalytic domain. FAD was bound in the pocket and the isoalloxazine ring pointed to catalytic domain. AD was in the middle of catalytic domain and FAD-binding domain. Figure 2 showed that carbonyl of C3 sterone’s was in the hydrogen bonding distance with Y365, Y541, and G545 of KsdD from MNR, similar to the corresponding amino acid residues Y318, Y487, and G491 in the active center of KsdD from R. erythropolis SQ1 (Rohman et al. 2013). The change of the ligand from ADD in 4c3y to AD in docking removed the H bond between the N5 atom of the isoalloxazine ring of FAD with C1 atom of A ring of AD. However, the isoalloxazine ring of FAD was still closed to the A ring of AD, which was essential for the transfer of α-hydrogen of the C1 atom of A ring as a hydride ion.

Fig. 1
figure 1

The model of KsdD with AD and FAD. KsdD can be divided into FAD-binding domain (green) and catalytic domain (blue)

Full size image
Fig. 2
figure 2

Local model of the residues of KsdD around AD within 5 Å

Full size image

Based on the docked conformation of KsdD and AD, 27 residues of KsdD in proximity to AD within 5 Å were located using PyMol software (Fig. 2). According to the results of Rohman et al. (2013), who recognized the necessity of the general base in the catalysis of KsdD, seven potential mutation sites, namely, Y122, Y125, S138, E140 and Y541 from the FAD-binding domain and Y365 and Y472 from the catalytic domain, were determined. G545 from the FAD-binding domain was ignored because it was a backbone amide in the hydrogen bonding distance of the C3 carbonyl group. An amino acid sequence alignment of the available KsdDs from different bacteria revealed that, among these amino acids at the active site of KsdD, Y125, Y541, G545 and Y365 from the catalytic domain were strictly conserved (Fig. S1). As shown in the phylogenetic tree (Fig. S2) which all KsdD proteins were majorly clustered into some branches, Y122, S138, E140 and Y472 were conserved (Fig. S1) in the branch of MNR.

Interestingly, S138 of the KsdD of MNR, which spontaneously mutated to L138 of the KsdD of MNR M3 and triggered the decline in activity of KsdD (Xie et al. 2015), was also included. This finding implied that 5 Å was appropriate. Sites located at an appropriate distance to the ligand may promote keto-enol tautomerization and labilization of C2 hydrogen atoms.

Construction of the mutations

As shown by SDS–PAGE analysis in Fig. 3a, KsdD was successfully expressed by increased yield in the host with the expected molecular weight of 63.4 kDa, which was similar to the report of Shao et al. (2016). This protein was also identified by mass spectrometry. Overlap PCR was used to implement the point mutations. After mutagenesis, through the same process of the wild KsdD, the mutations were constructed and identified using the result of clones that were digested by Nde I/BamH I before sequencing. SDS–PAGE analysis revealed that (Fig. 3b) Y122F, Y125F, S138A, E140V, Y365F, Y472F, and Y541F were expressed as well. Gray level difference analysis of protein bands implied that no significant changes in KsdD expression were caused by the mutations.

Fig. 3
figure 3

Heterologous expression of KsdD and mutations in E. coli. a SDS–PAGE analysis of E. coli BL21 (DE3)/pET-22b(+)-KsdD-MNR Lane 1 E. coli BL21 (DE3)/pET-22b(+)-KsdD-MNR (before codon optimization); lane 2 E. coli BL21 (DE3)/pET-22b(+)-KsdD-MNR (after codon optimization). b SDS–PAGE analysis of mutations of KsdD in E. coli. Lane 1 Y122F; lane 2 Y122F induced; lane 3 Y125F; lane 4 Y125F induced; lane 5 S138A; lane 6 S138A induced; lane 7 E140V; lane 8 E140V induced; lane 9 Y365F; lane 10 Y365F induced; lane 11 Y472F; lane 12 Y472F induced; lane 13 Y541F; lane 14 Y541F induced

Full size image

Effect of amino acid mutations on KsdD activity

Under the same conditions, wild KsdD and mutant KsdD were used to transform AD to ADD. The results of TLC (Fig. 4a) showed the different activities of KsdD in wild type and mutant. The conversion rate of AD by wild type was nearly 50%, and mutant strains could be divided into two types according to the conversion rate (Fig. 4b). The conversion rate of Y125F, Y365F, and Y541F declined remarkably, whereas those of Y122F, S138A, and E140V decreased at varying degrees. Compared with the wild type, the conversion rate of AD by each mutation was 1.44%Y125F, 0.52%Y365F, 2.52%Y541F, 39.93%Y122F, 23.64%S138A, 38.99%E140V, and 98.24%Y472F. Thus, the activity of KsdD was almost abolished in Y125, Y365, and Y541. Meanwhile, Y122F, S138A, and E140V contributed differently to the activity of the enzyme, whereas Y472F did not affect enzyme activity.

Fig. 4
figure 4

AD conversion by mutation strains. a TLC analysis of AD conversion Lane 1 standard AD and ADD; lane 2 MNR; lane 3 Y122F; lane 4 Y125F; lane 5 S138A; lane 6 E140V; lane 7 Y365F; lane 8 Y472F; lane 9 Y541F. b HPLC analysis of AD conversion. Lane 1 E. coli BL21 (DE3)/pET-22b(+)-KsdD-MNR; lane 2 Y122F; lane 3 Y125F; lane 4 S138A; lane 5 E140V; lane 6 Y365F; lane 7 Y472F; lane 8 Y541F

Full size image

Structural mechanism analysis of mutations

Pymol software (http://pymol.org/) was used to analyze the local structure of those mutations that aroused changes in KsdD activity. The hydroxyl group of Y365 was situated close to the C2 atom of the bound AD and served as the catalytic base that abstracted a proton from the C2 atom of AD (Fig. 5a). Y125 had no close contacts with the bound AD in the conformation of KsdD docking with AD, but its phenolic hydroxyl group was of hydrogen-bonding distance from the phenolic hydroxyl group of Y365 (Fig. 5b) and serve to increase the basic character of the catalytic base and facilitate proton relay.

Fig. 5
figure 5

Local model of the catalytic residues of KsdD-Tyr365 and Tyr125. a Local model of the catalytic residues of KsdD-Tyr365. b Local model of the catalytic residues of KsdD-Tyr125

Full size image

In Fig. 6a, the distance between the hydroxyl group of Y541 and the C3 carbonyl oxygen of ADD was 3.5 Å. But the Y541F mutation only showed a modest decline compared with Y365F and Y125F (Fig. 4b). This observation could be attributed to the backbone amide of G545, which within hydrogen bonding distance with the C3 carbonyl oxygen, together with Y541 to promote keto-enol tautomerization and labilization of the C2 hydrogen atoms.

Fig. 6
figure 6

Local model of the catalytic residues of KsdD-Tyr541/Tyr122/Ser138. a Local model of the catalytic residues of KsdD-Tyr541. b Superposition of residues Tyr122 and simulation of point mutations Phe122. c Local model of the catalytic residues of KsdD-Ser138

Full size image

The binding of AD molecule in the active site need a large number of van der Waals interactions (Rohman et al. 2013). As shown in Fig. 6b, Y122 is in a good position for hydrophobic stacking interactions with the B-ring of the 3-ketosteroid. This interaction is important for the binding of the steroid substrates(Matsushita and Itagaki 1992; Rohman et al. 2013). However, Y122F changed this position of Y122, which is detrimental to the binding of AD.

S138 located in the loop and in the hydrogen-bonding distance from V57, V139, and P544 (Fig. 6c). S138A and spontaneous mutation S138L (Xie et al. 2015) removed all the H bonds of S138, which could fix the movable loops. These loops may have a function in catalysis by shielding the FAD and substrate from the solvent (Oosterwijk et al. 2012).

In many flavoproteins, catalysis takes place at a location shielded from the solvent, which enhances the strength of polar interactions and is instrumental to substrate activation (Fraaije and Mattevi 2000; Mattevi 1998). E140 collaborated with five-membered D-ring of the 3-ketosteroid occupied a solvent-accessible pocket near the active site entrance (Fig. 7a). Shielded from the solvent to form hydrophobic platforms, which are important for substrate binding (Oosterwijk et al. 2012; Rohman et al. 2013).

Fig. 7
figure 7

Local model of the catalytic residues of KsdD-Glu140 and simulation of point mutations Va140. a Local model of the catalytic residues of KsdD-Glu140. b Simulation of point mutations Val140

Full size image

The substrate conversion of Y472F showed no significant differences from the wild type. Given that the long distance between the sites of Y472 and substrate and the mutation of Y472F did not initiate hydrogen bond changes, Y472 may have no contribution to the catalysis of KsdD.

Discussion

Site-directed mutagenesis is an effective tool to improve certain biochemical or catalytic properties of proteins and is being widely used in the molecular modifications of industrial enzymes (Li et al. 2014; Seo et al. 2013). While the key issue is that which basic residues should be considered for mutation. The most important for the site-directed mutagenesis is the selection of mutation sites, and structure analysis can provide useful information in this aspect (Yang et al. 2013). KsdD is generally accepted as a membrane protein (Molnar et al. 1995; Plesiat et al. 1991; Wei et al. 2010). In addition, many attempts have been made to get the crystal of KsdD from MNR in our group. But as reports, the progress of expression (Shao et al. 2017) purification (Forneris et al. 2009) and crystallization (Rohman et al. 2012) of KsdD is tedious. Considering that, site-directed mutagenesis under the direction of in-silico protein docking modeling KsdD from MNR could provide a driection in the engineering of KsdD from Mycobacterium.

The decline of the conversion of AD illustrated that Y125, Y365, and Y541 were essential to the function of KsdD. Y122, S138, and E140 contributed to the catalysis of KsdD. This indicated these residues could be potential sites to mutate in the engineering of KsdD. By combining the two-step mechanism of KsdD and the local structure, the mechanism of these amino acids could be explained.

The wild-type KsdD could introduce a double bond and then catalyze the ring-opening reaction of steroids. During this process, a concerted bimolecular elimination (E2) mechanism, in which the two hydrogens depart simultaneously without the formation of any intermediate, was considered less likely. In this study, we referred to the mechanism reported by Oosterwijk et al. (2012). Enzyme-catalyzed dehydrogenation likely followed a stepwise unimolecular elimination conjugate base (E1cB) mechanism, which was initiated by the interaction of the C3 carbonyl group of the 3-ketosteroid substrate with Y541 and G545 as an electrophile to promote labilization of the C2 hydrogen atoms. Subsequent abstraction of a proton from the C2 atom by Y365, which was assisted by Y125 as a general base, was proposed to result in either an enolate or a carbanionic intermediate; subsequently, a double bond was formed between the C1 and C2 atoms in synchrony with hydride transfer from the C1 atom of the substrate to the flavin coenzyme (Itagaki et al. 1990; Jerussi and Ringold 1965; Ringold et al. 1963). This process formed the indirect mechanism of Y122. In particular, the active site residues Y541 and G545 may act in tandem, so the absence of Y541 did not completely abolish the dehydrogenase activity of KsdD. Together, they promote keto-enol tautomerization and labilization of the C2 hydrogen atoms.

In addition, the hydroxyl groups of Y365 and Y541 of KsdD from MNR are in very similar positions to those of S468 and Y466 of Δ4-(5α)-KsdD, respectively. Whereas in Δ4-(5α)-KsdD, the S468 hydroxyl group abstracts a proton from the C4 carbon atom of the substrate, and in KsdD from MNR, the Y365 hydroxyl group is in a position to abstract a proton from C2. Finally, Y541 of KsdD from MNR can promote keto-enol tautomerization at C3 keto-group, similar to Tyr466 of Δ4-(5α)-KsdD.

In silico protein docking modeling showed that E140 was only 4 Å away from AD. This distance indicated that the carboxyl of Glu possibly reacted with the substrate. Although the conversion of AD decreased, the local structure of E140V showed that no H bonds were removed (Fig. 7b). Moreover, the mutation did not influence the key amino acid sites which as the general base to removed H of C2 of steroids. As shown in Fig. 7a, the residue E140 possibly maintained the hydrophobic environment of the active center.

In our previous study, we found that S138 of the KsdD of MNR, which spontaneously mutated to L138 of the KsdD of MNR M3, triggered the decline in KsdD activity (Xie et al. 2015). Furthermore, S138L was found to influence the stabilization of the loop. Considering the lack of the structure information of KsdD from MNR, the function of 138-amino acid was predicted to combine the crystal structure of the KsdD from R. erythropolis SQ1 and KsdD4 from R. jostii RHA1. For the C1,2-dehydrogenation reaction, we recognized that the fatty acid side chain of Leu may prevent the hydrophobic formation of the active center because of steric hindrance. In this study, in silico protein docking modeling revealed that AD in the loop, S138A, and S138L could remove the H bonds with V57, V139, and P544. Similar to the spontaneous mutations, the conversion of the substrate of S138A with the methyl side chain of Ala compared to S138L with iso-butyl side chain of Leu which has a higher steric hindrance declined, too. This phenomenon proved that these H bonds were essential for the activity of S138.

The catalytic function of KsdD was also partly clarified by docking screening and biotransformation assay. These results enabled us to identify the roles of the amino acid residues involved in catalysis, which will be useful in manipulating the catalytic properties of the enzyme and improving them for application in the pharmaceutical steroid industry. Nevertheless, in view of the importance of KsdD enzymes for industrial use, KsdD from MNR needs further investigation. With this information, we could establish a possible relationship between steroid and KsdD and clarify the reaction mechanisms of the enzyme. Therefore, future research should focus on the purification and structural characterization of KsdD.