Determining the rate-limiting step in producing 9OHAD with KSH
KSH is a two-component Rieske oxygenase (KshAB). The reaction cycle requires two reducing equivalents , which originate from NADH, and the resulting electrons are transferred from KshB to KshA (Fig. 1). Over the course of the steroidal transformation, reductases play a crucial role in the electron conversion using NAD(P)H . KSH activities should, therefore, be abolished in the absence of the reductase KshB, which implies that KshB might affect the productivity of 9OHAD in this KSH system [4, 8]. Therefore, the rate-limiting step in producing 9OHAD was investigated using purified KshA and KshB. In the full pathway of this KSH reaction system, a 31.9% yield of 9OHAD was obtained. When only a 20% concentration of KshA enzyme was used in this reaction system, only a subtle effect on the yield of 9OHAD was observed (Fig. 2A). However, when the concentration of KshB was decreased to 20%, a yield of only 14.5% 9OHAD was obtained, meaning a significant decrease of 55% compared with the standard full pathway reaction system. We further investigated the effect of the KshB concentration (10-100%) on this reaction system. The yields of 9OHAD decreased from 31.9% to 5.5% when the concentration of KshB decreased from 100 to 10% (Fig. 2B), which demonstrated that the reductase KshB was a potential rate-limiting enzyme in the hydroxylation reaction for producing 9OHAD.
Construction of an NADH regeneration system
KshB requires the cofactor NADH as an electron donor (Fig. 1) . KSH showed no hydroxylation activity toward AD without the NADH in a reaction mixture . NADH regeneration often results in a significant improvement of whole-cell conversion . The addition of the FDH enzyme was crucial to lower the industrial cost by recycling NADH efficiently in this reaction system. Therefore, the consumption of NADH was determined in this reaction system. The concentrations of NADH were observed to drop significantly from 500 μM and maintained at a low level of 40.3 μM after reaction for 20 min. A fed-batch mode with the addition of NADH, unfortunately, did not work continuously in this reaction system (Fig. 3A). Furthermore, NADH often takes over a considerable proportion of the production cost of steroidal medicines . Therefore, the reconstruction of an NADH regeneration system was required, steady-state kinetic analysis of selected FDH toward NAD+ was carried out, and the resulting Km and kcat were 32.02 μM and kcat/Km 0.34 s-1 μM-1, respectively (Fig. S1). FDH and sodium formate were thus added to this multienzyme cascade catalysis in vitro to regenerate NADH as an electron donor (Fig. 1). The rate-limiting step was also investigated via decreasing FDH to 10% (0.025 μM) (Fig. 3B). The concentrations of NADH were increased gradually and reached the same level as that of the original reaction conditions. This indicated that FDH would not affect this NADH regeneration system. Therefore, we used 0.025 μM FDH in the following multienzyme cascade catalysis.
Screening reductases to improve the conversion of AD
As KshB was the rate-limiting step in producing 9OHAD, therefore, we screened the optical ferredoxin reductases by determining activity toward NADH to improve the efficiency of electron transfer in our reaction system. We further screened the reductase genes of KshB, DMR1, DMR2, DMR3, and TDO, which all have potential for the hydroxylation reaction of AD. The kinetic parameters of these reductases were determined with NADH as a substrate (Table 2 and Fig. S2). The affinity (Km) values were between 55.72 and 182.1 μM, and the catalytic efficiencies (kcat/Km) were between 0.11 and 0.43 s-1 μM-1, respectively. The reductase of TDO showed a higher activity than other reductases, with Km of 65.37 μM and kcat/Km of 0.43 s-1 μM-1. TDO contains the reductase TDO-R and the Rieske [2Fe-2S] cluster TDO-F . Therefore, the combination of TDO-R and TDO-F may improve the efficiency of electrons transfer for our catalysis scheme. The yield of 9OHAD reached 54.8% using TDO as a reductase compared with that by KshB (42%) with an NADH regeneration system (Fig. 3C).
Directed evolution of TDO by error-prone PCR
The reductase TDO was chosen for random mutation by error-prone PCR to improve its catalytic activity. The improved mutants from the mutant library (over 3,000 clones) were screened by assaying the consumption of NADH using 96-well plates. Nine mutants were obtained with the improved activity toward NADH compared to the wild-type TDO (Fig. S3), and the purified mutants of MT2 and MT9 showed a 1.8- and 2.25-fold increase of relative activity compared with that of wild-type (WT), respectively (Fig. 4A). The MT2 (L127F/S139C/L316Q/V422A/D461G) and MT9 (G199S/G205R/S223N/N396S/P441S) mutants improved the activity toward NADH. MT2 and MT9 also showed the higher activity, with an affinity of 56.63 and 48.13 μM and catalytic efficiencies of 1.19 and 1.62 s-1 μM-1, respectively (Table 1 and Fig. 4B). The yields of 9OHAD by MT2 and MT9 were significantly improved to 69.3 and 74.8%, respectively (Fig. 4C), which indicated that the efficiency of electron transfer in MT2 and MT9 had been strengthened compared with WT. The structure of TDO showed that G205, S139, L127, G199, and S223 were located in the NADH binding domain. L316 and N396 were located at the FAD-binding domain. P441 and D461 were located at [2Fe-2S] cluster domain of TDO-F (Fig. 4D)
Structural analysis of iron–sulfur clusters in KshB and TDO
Both the KshB and TDO reductases contain FAD and NADH binding domains in the N-terminal region and iron–sulfur clusters at the C terminus (Figs. 5A and B). The iron–sulfur clusters are crucial in the mediating electron transfer processes in forms of [2Fe-2S], [3Fe-4S], and [4Fe-4S] clusters [22, 23]. TDO belongs to Rieske [2Fe-2S] (Fe2S2Cys2His2), and the cluster was composed of H44, H64, C42, and C61 (Fig. 5A), whereas each iron atom was coordinated by one sulfur from two Cys thiolates and two His residues. Meanwhile, KshB is a plant-type one (Fe2S2Cys4) and is composed of C306, C310, C314, and C344. The iron atoms are coordinated by four Cys residues (Fig. 5B). The reaction depends on an electron transport chain that transfers electrons from the NADH to FAD and then to the [2Fe-2S] cluster of KshB (Fig. 5B). The electrons continue to be transferred to [2Fe-2S] cluster of the terminal oxygenase KshA for hydroxylation of steroidal substrates. When removing both [2Fe-2S] cluster domains in KshB and TDO reductases, there was no activity detected toward AD (Figs. 5C and D), which indicates that the [2Fe-2S] cluster domain plays an important role of in the electron transfer chain. Interestingly, the addition of the Rieske [2Fe-2S] cluster from TDO-F to the N terminus of KshB and TDO showed remarkably improved yields (TDO-F+KshB: 56.1% and TDO-F+TDO: 74.5%) toward AD. Meanwhile, the addition of the plant-type [2Fe-2S] cluster from KshB (pFeS) showed yields (pFeS+KshB: 47.3% and pFeS+TDO: 65.6%) compared with that of wild-type of KshB (42%) and TDO (54.8%), respectively (Fig. 4D). Therefore, the Rieske [2Fe-2S] cluster used in this electron transfer system showed a higher bioconversion yield of 9OHAD than that of the plant-type cluster. This also implied that the Rieske [2Fe-2S] cluster made the electron transfer more efficient in this KSH system.
Optimizing reaction conditions to improve 9OHAD yield
To increase the 9OHAD yield, the reaction conditions (organic co-solvents, CDs, biomass, and concentrations of AD) were further optimized by whole-cell biotransformation using the BLKA-KB-F strain with NADH regeneration; various organic co-solvents including methanol, ethanol, isopropyl alcohol, acetone, DMSO, and DMF and the surfactants Tween 80 and Triton X-100 were used to improve AD solubility in the aqueous phase in the range of 1-100 μM. When 3 vol% methanol and DMSO were used as an optimal co-solvent, the higher yields of 9OHAD were obtained with 54% and 56%, respectively. However, the 9OHAD yield decreased to 48% and 39.5% when 5 vol% co-solvents were used (Fig. 6A). This may be due to the cell membrane swelling and denaturation of membrane-associated proteins caused by high organic solvent concentrations .
CDs have been used widely to improve steroid solubility in aqueous media due to their hydrophilic outer surfaces and hydrophobic cavities [24, 25]. The effect of different molar ratios of CDs and AD was investigated, and the highest yields of 9OHAD were obtained when Me-β-CD (67%) and γ-CD (65.5%) were used (Fig. 6B). Therefore, Me-β-CD was used for further research due to its lower cost compared with γ-CD .
To investigate the influence of biomass on the yield of 9OHAD, various concentrations of recombinant cells were used, from 10 to 100 g/L. As shown in Fig. 6C, the yield of 9OHAD rose with increasing of the concentrations of cells. After 6 h, the highest yield of 9OHAD was obtained from AD at 50 g/L wet cells. The 9OHAD yields were not increased by the addition of higher concentrations of biomass. This was also consistent with previous reports for biotransformation of cortisone to 11β-hydrocortisone using E. coli as a whole-cell biocatalyst , and phytosterols to 5α-AD by recombinant M. neoaurum cells . This may have been to the insufficient oxygen input and distribution in reactions with high cell mass, and oxygen availability playing a crucial role in monooxygenase-catalyzed reactions .
Excessive substrates are difficult to dissolve even using Me-β-CD. The influence of the concentrations of AD for the bioconversion of 9OHAD was also investigated. The yield of 9OHAD decreased gradually with a concentration increases of AD, and a maximum yield was achieved at 89% when 1 g/L AD was used. However, the maximum production of 9OHAD (4.11 g/L) was obtained when 5 g/L AD was used compared with that when 1 g/L AD (0.94 g/L) and 3 g/L AD (2.72 g/L) were used. Furthermore, when 7 and 9 g/L AD were used, the yields of 9OHAD were reduced to 46% (3.40 g/L) and 32.9% (3.12 g/L), respectively (Fig. 6D). Therefore, 5 g/L AD was used for further study.
Bioconversion of AD to produce 9OHAD
To improve the yield of 9OHAD, different recombinant E. coli BL21(DE3) cells harboring plasmids carrying the genes for FDH, KshA, and KshB or TDO were constructed (Table 1). For efficient co-expression of all three proteins in one host and two vectors (pET28a and pETDuet-1/pCold I) were used, resulting in an engineered strain carrying the genes for the oxygenase KshA for hydroxylation, ferredoxin reductases for electron transfer, and FDH for NADH regeneration. We constructed 10 recombinant strains as described in Table 1. The resting cells of BLKA-T-F showed the high yields of 9OHAD (84.5%), followed by BLKA-KB-F cells at 72.1% (Fig. S4). It indicated that co-expression of the reductase TDO/KshB and FDH genes in pETDuet-1 played an important role in the NADH regeneration system to produce 9OHAD. The pETDuet-1 plasmid with two promoters and multiple cloning sites, is often successfully used for the construction of whole-cell biocatalysts systems with cofactor regeneration systems . The introduction of reductase TDO/KshB and FDH genes in pETDuet-1 seemed to improve the efficiency of the electron transfer chain. Notably, co-expression of oxygenase KshA and reductase TDO/KshB genes in pETDuet-1 also showed the considerable yields of 9OHAD at 67.5 and 61.3%, respectively. It is obvious that the expression of the FDH gene in independently may slightly affect the efficiency of NADH regeneration. The resting cells of BLKA-T and BLKA-KB without NADH regeneration only had the yields of 9OHAD at 59.5 and 45.6%, respectively. However, there was little product of 9OHAD detected in the cells BLKA-KBF and BLKA-TF, where the reductase of KshB or TDO was linked with FDH in pCold I. This may have been due to an inappropriate linker peptide leading to the incorrect folding of KshB/TDO and FDH . The overexpression of the reductase KshB or TDO with FDH genes using one promoter may increase the cellular stress during cell growth, even as the cold shock gene of cspA was designed for use with pCold I.
The TDO mutant (MT9) gene was inserted into pETDuet-1 instead of TDO wild type and then transformed into BL21(DE3) together with KshA and the FDH gene. The resting cells of BLKA-T-F, BLKA-KB-F, BLKA-TM-F, and BLKA-RTM-F were then used to convert AD to 9OHAD. Comparing with the resting cells of BLKA-T-F and BLKA-KB-F with the space-time yields of 9OHAD at 0.72 g/(L·h) and 0.65 g/(L·h), the resting cells of the strain BLKA-TM-F further increased the yield of 9OHAD. It completely converted 5 g/L AD to 5.19 g/L 9OHAD within 6 h (Fig. 7 and Table 3), with a space-time yield of 0.87 g/(L·h). The construction of a modified TDO mutant with a Rieske [2Fe-2S] cluster at the N terminus (BLKA-RTM-F) showed a higher production and space-time yield of 5.24 g/L and 1.05 g/(L·h) of 9OHAD, respectively. An efficient electron transfer system has improved the yields of 9OHAD in these whole-cell catalysts.
In comparison, the biosynthesis of 9OHAD are extensively studied by the engineered Mycobacterium using phytosterol as substrate. Xiong et al. constructed a 9-OHAD-producing strain with productivity of 10.27 g/L (0.071 g/(L·h)) from 20 g/L phytosterol by deletion of a sigma factor D (sigD) and overexpression of a cholesterol oxidase ChoM2 . They further achieved the productivity of 0.114 g/(L·h) by deleting kasB gene encoding a β-ketoacyl-acyl carrier protein synthase in M. neoaurum ATCC 25795 strain to improve the cell permeability . Yao et al. demonstrated a stable accumulation of 9-OHAD with productivity of 7.33 g/L (0.051 g/(L·h)) from 15 g/L phytosterol by multiple deletions of three KstD genes and overexpression of the KshA gene in M. neoaurum ATCC 25795 strain . Furthermore, they improved the production of 9OHAD to 11.7 g/L (0.098 g/(L·h)) from 20 g/L phytosterol by overexpression of a mutant KshA1N in engineered M. neoaurum ATCC 25795 strain . The 9OHAD production and space-time yield reached 19.64 g/L and 0.82 g/(L·h) from 20 g/L AD using whole cells of modified R. erythropolis RG1-UV29 strains without KstD activity . Gao et al. reported that the 9OHAD production and space-time yield could reach 36.4 g/L and 0.379 g/(L·h) from 70 g/L phytosterol with resting cells of M. neoaurum NwIB-yV in a 5-L bioreactor, respectively . However, the production of 9OHAD could only reach 0.63 g/L (0.01 g/(L·h)) from 1 g/L AD using whole cells of recombinant E. coli , and 7.23 g/L (0.45 g/(L·h)) from 8 g/L AD using whole-cells recombinant B. subtilis, respectively . Thus, the constructed BLKA-RTM-F was a great 9OHAD producer with high production of 5.24 g/L and space-time yield of 1.05 g/(L·h), a considerable yield of 99.3% of theoretical without by-products.