Unraveling and engineering the production of 23,24-bisnorcholenic steroids in sterol metabolism

The catabolism of sterols in mycobacteria is highly important due to its close relevance in the pathogenesis of pathogenic strains and the biotechnological applications of nonpathogenic strains for steroid synthesis. However, some key metabolic steps remain unknown. In this study, the hsd4A gene from Mycobacterium neoaurum ATCC 25795 was investigated. The encoded protein, Hsd4A, was characterized as a dual-function enzyme, with both 17β-hydroxysteroid dehydrogenase and β-hydroxyacyl-CoA dehydrogenase activities in vitro. Using a kshAs-null strain of M. neoaurum ATCC 25795 (NwIB-XII) as a model, Hsd4A was further confirmed to exert dual-function in sterol catabolism in vivo. The deletion of hsd4A in NwIB-XII resulted in the production of 23,24-bisnorcholenic steroids (HBCs), indicating that hsd4A plays a key role in sterol side-chain degradation. Therefore, two competing pathways, the AD and HBC pathways, were proposed for the side-chain degradation. The proposed HBC pathway has great value in illustrating the production mechanism of HBCs in sterol catabolism and in developing HBCs producing strains for industrial application via metabolic engineering. Through the combined modification of hsd4A and other genes, three HBCs producing strains were constructed that resulted in promising productivities of 0.127, 0.109 and 0.074 g/l/h, respectively.

Hsd4A is a putative gene located in the sterol catabolic gene cluster, which was deduced to encode a short chain dehydrogenase (Hsd4A). As Hsd4A shares intriguing similarity with the N-terminal domain of eukaryotic 17β -hydroxysteroid dehydrogenase IV (17β HSD4), Hsd4A has been speculated to be a 17β -hydroxysteroid dehydrogenase 15,23 . However, the hsd4A gene in R. jostii RHA1 was highly up-regulated along with other genes that are involved in the side-chain degradation of sterols, and thus was speculated to be essential for sterol side-chain degradation 15 . Therefore, Hsd4A was further proposed to be a β -hydroxyacyl-CoA dehydrogenase in the degradation of the sterol side chain because the N-terminal domain of 17β HSD4 also could act as a D-3-hydroxyacyl-CoA dehydrogenase in degrading branched fatty acids and bile acids 15,23,24 . Although these two possible biochemical functions have been assigned to Hsd4A, neither have been validated 1 . As these two putative functions may have underlying associations in the production of C19 or C22 steroids, the exact function of Hsd4A requires elucidation.
While investigating the role of hsd4A in sterol metabolism, we discovered that hsd4A might be a key gene relating to the formation of C22 steroids. In this study, therefore, we attempted to investigate the physiological role of Hsd4A and its possible mechanism in the conversion of sterols to C22 steroids, based on which we also tried to establish a rational strategy to modify hsd4A and other key genes to develop attractive HBCs producing strains.

Results
Hsd4A MN gene from M. neoaurum ATCC 25795. The sterol catabolic gene cluster in M. neoaurum ATCC 25795 is described schematically in Fig. 2a. In the gene cluster, a putative gene hsd4A (hsd4A MN ) was annotated as an ortholog of the hsd4A gene (Rv3502c) from M. tuberculosis H37Rv. Its encoded protein, Hsd4A MN , was estimated to be 30.9 kDa and shares high similarities with its counterparts, such as 97%, 71% and 67% amino acid identities with the Hsd4As from Mycobacterium sp.VKM Ac-1815D, M. tuberculosis H37Rv and R. jostii RHA1, respectively (Fig. 2b). As shown in Fig. 2, hsd4A MN is located between the mce4 operon and fadE26-27. The mce4 operon encodes a multicomponent ABC-like ATP-dependent transport system, which plays an important role in the cellular uptake of steroids 1,15,34 . The fadE26-27 locus encodes acyl-CoA dehydrogenase enzymes that catalyze the dehydrogenation of 3-oxo-cholest-4-en-26-oyl CoA in the first cycle of cholesterol side-chain β -oxidation 35 . Obviously, hsd4A MN and its genetic organization are highly conserved in some mycobacteria and rhodococci (Fig. 2b), indicating the conserved function of Hsd4A in these strains for sterol catabolism. Hsd4As from M. tuberculosis H37Rv and R. jostii RHA1 have been proposed to be a β -hydroxyacyl-CoA dehydrogenase or a 17β -hydroxysteroid dehydrogenase 1,15,23,36 , indicating a possibly similar function for Hsd4A MN . However, the exact role of Hsd4A in these microorganisms has not been characterized 1 .

Enzymatic characteristics of Hsd4A MN in vitro.
To explore the possible function, hsd4A MN was expressed heterologously in Escherichia coli BL21(DE3) (E-hsd4A, Table S1). The proposed activities of Hsd4A MN were then investigated using purified protein (Fig. S1). As indicated in Table 1, Hsd4A MN catalyzed the dehydrogenation of testosterone (TS) and boldenone (BD) in the presence of NAD + to produce AD and ADD, respectively. Interestingly, it did not catalyze the 17β -reduction of AD and ADD in the presence of NADH, demonstrating that Hsd4A MN does have 17β -hydroxysteroid dehydrogenase function without reversible activity. As for the proposed β -hydroxyacyl-CoA dehydrogenase activity of Hsd4A MN , two commercially available fatty acyl-CoA esters, acetoacetyl-CoA and DL-β -hydroxybutyryl-CoA, were tested as substrates for Hsd4A MN . The results showed that Hsd4A MN catalyzed the conversion of β -hydroxybutyryl-CoA to acetoacetyl-CoA using NAD + as coenzyme and also carried out the reverse reaction in the presence of NADH, which unambiguously demonstrated that Hsd4A MN also had β -hydroxyacyl-CoA dehydrogenase function with reversible catalytic activity (Table 1). According to the above results, therefore, Hsd4A MN was characterized as a dual-function enzyme in vitro, possessing both the activities of β -hydroxyacyl-CoA dehydrogenase and 17β -hydroxysteroid dehydrogenase.
Physiological role of Hsd4A MN in vivo. Hsd4A MN acts as a dual-function enzyme in vitro, but its physiological role in vivo is unknown. As hsd4A MN is highly conserved in the sterol catabolic gene cluster, the investigation mainly focused on its role in sterol metabolism. Corresponding to the enzymatic activities in vitro, Hsd4A MN were deduced to have two roles in sterol metabolism in vivo: the oxidation of 17β -hydroxysteroids by 17β -hydroxysteroid dehydrogenase and the β -hydroxyacyl-CoA dehydrogenation in the β -oxidation of the sterol side chain by β -hydroxyacyl-CoA dehydrogenase. Both of these reactions can occur before the cleavage of the steroid nucleus in sterol metabolism. Given that the natural sterols can be completely degraded by M. neoaurum ATCC 25795 to CO 2 and H 2 O without obvious accumulation of intermediates (Fig. 3a), the function of Hsd4A MN is difficult to determine via identifying the metabolites of sterols after genetic manipulation of hsd4A MN in M. neoaurum ATCC 25795. To facilitate the functional analysis of Hsd4A MN , therefore, it was necessary to build a model strain which could accumulate suitable metabolites with intact steroid skeletons before the genetic manipulation of hsd4A MN . 3-Ketoseroid-9α -hydroxylase (encoded by two genes, kshA and kshB) is one of the key enzymes in dissociation of the steroid skeleton, and inactivating this enzyme would block the cleavage of the steroid skeleton 4,28 . Based on the genome sequence, two putative kshA (kshA1 and kshA2) genes were identified in M. neoaurum ATCC 25795 (Fig. 2a). Therefore, a model to investigate the function of Hsd4A MN in vivo was constructed by the successive deletion of kshA1 and kshA2 in M. neoaurum ATCC 25795, which is a kshAs-null mutant (NwIB-XII, Table S1). As expected, sterol catabolism in strain NwIB-XII was blocked, accumulating multiple metabolites with intact steroid skeletons, including ADD (I), AD (II), and two 17β -hydroxysteroids TS (III) and BD (IV) (Figs 3a,b and S2).
To determine whether Hsd4A MN possesses 17β -hydroxysteroid dehydrogenase activity in vivo, hsd4A MN was augmented in NwIB-XII using plasmid pMV261-hsd4A, resulting in the XIIp261hsd4A strain. The vacant plasmid pMV261 was introduced into NwIB-XII as a control (designated as XIIp261). In contrast to NwIB-XII and XIIp261, strain XIIp261hsd4A transformed sterols to AD and ADD without any detected TS and BD (Figs 3a and S3), which could be ascribed to the conversion of TS and BD to AD and ADD, respectively, by the enhanced Hsd4A MN activity (Fig. 3b). Therefore, this suggests that Hsd4A MN can also act as a 17β -hydroxysteroid dehydrogenase to irreversibly catalyze the oxidation of TS to AD and BD to ADD in vivo.
To further characterize the function of Hsd4A MN , hsd4A MN was deleted in NwIB-XII, leading to the XIIΔ hsd4A strain (Table S1). Surprisingly, the metabolic products of sterols in XIIΔ hsd4A were significantly different from those in NwIB-XII. XIIΔ hsd4A converted sterols to two novel products, V and VI (Fig. 4a). The major product V was identified as 1,4-HBC and the minor product VI was identified as 4-HBC (Figs 4b and S4). Both 1,4-HBC and 4-HBC are the incompletely degraded products of the sterol side chain, which still contain three carbon atoms at carbon 17. To confirm whether the production of these two compounds could be attributed to the deletion of hsd4A MN , hsd4A MN was then complemented in XIIΔ hsd4A (strain Cp-4A, Table S1), which restored the same metabolic phenotype observed in NwIB-XII (Fig. S5). These results confirmed that deletion of hsd4A MN resulted in incomplete degradation of the sterol side chain, demonstrating that Hsd4A MN plays a significant role in sterol side-chain degradation. Therefore, Hsd4A MN exhibited two enzymatic activities in vivo, simialr to results observed in vitro. According to the characterized enzymatic properties of Hsd4A MN in vitro, the involvement of Hsd4A MN in side-chain degradation should be attributed to its β -hydroxyacyl-CoA dehydrogenase activity, which has been proposed in previous studies 1,15,36 .
Although 1,4-HBC and 4-HBC were the products derived from the deletion of hsd4A MN , they were verified to be not the substrates of Hsd4A MN in vitro ( Table 1), indicating that the two compounds are further derivatives of the physiological substrate of Hsd4A MN . According to the postulated β -oxidation of the sterol side chain, the most probable precursor of 1,4-HBC and 4-HBC is 22-hydroxy-3-oxo-25,26-bisnorchol-4-en-24-oyl CoA (22HOBNC-CoA), which contains a β -hydroxybutyryl-CoA moiety that has been determined to be the active substrate of Hsd4A MN in vitro ( Table 1).
Engineering of the metabolism of sterols to produce HBCs. In the industry, HBCs (including 1,4-HBC, 4-HBC and 9-OHHBC, Fig. 1) can be used as precursors to produce progestational and adrenocortical hormones. HBCs are mainly produced from the modified metabolism of sterols in microorganisms. However, the formation mechanism of HBCs in sterol metabolism has not been clarified, and industrial strains used to produce HBCs are usually achieved by mutation breeding 11 . Therefore, it is a great challenge to rationally develop HBCs producing strains. In this study, the production of HBCs in the hsd4A MN -deleted strain opened the door to unravel the production mechanism of HBCs and paved the way to rationally construct HBCs producing strains.
As indicated in Fig. 4a, strain XIIΔ hsd4A could transform sterols to HBCs. To evaluate the production performance, a resting cell biotransformation of physterols was employed. When 40 g/l of phytosterols was fermented by XIIΔ hsd4A resting cells for 144 h, a 39-40% molar yield (12.48-13.04 g/l) of 1,4-HBC was achieved, along with an 18-20% yield (6.02-6.47 g/l) of 4-HBC and a 1-2% yield (0.41-0.53 g/l) of ADD (Table 2). This performance made the strain XIIΔ hsd4A not ideal for industrial application because the low yield ratio of 4-HBC or 1,4-HBC among the products would lead to a low recovery rate at the industrial scale. Therefore, a combined strategy to further modify the metabolic pathway of sterols was necessary to develop promising strains for industrial application.
1,4-HBC and 4-HBC differ at the C1, 2 double bond, which is catalyzed by KstDs 4,6,25 . There are three KstD homologues in M. neoaurum ATCC 25795, and KstD1 was characterized as the major enzyme in our previous work 4 . Therefore, strain XIIΔ hsd4A could be further optimized to better produce 1,4-HBC or 4-HBC by strengthening or weakening the activities of KstDs (Fig. 5a,b).
The compound VII was identified as 9-OHHBC (Fig. S4), which is a useful precursor to synthesize adrenocorticoids. Therefore, the feasibility in developing a 9-OHHBC producing strain was investigated. According to the steroid nucleus metabolic mechanism (Fig. 5c), kstD1, kstD2, kstD3 and hsd4A MN were progressively deleted in M. neoaurum ATCC 25795, resulting in the final strain MNΔ hsd4AΔ kstD123. In the transformation of 40 g/l of phytosterols, strain MNΔ hsd4AΔ kstD123 produced a 30-32% molar yield of 9-OHHBC (10.25-10.72 g/l) ( Table 2), demonstrating that a 9-OHHBC producing strain could be developed by combined modification of the  key enzyme genes hsd4A MN and kstDs. In terms of industrial applications, strain MNΔ hsd4AΔ kstD123 should be further optimized because of the massive accumulation of the byproduct 9-OHAD.
In conclusion, hsd4A MN plays a key role in the production of HBCs during sterol metabolism, and the combined modification of hsd4A MN and other key genes is a useful strategy to develop HBCs producing strains. Compared to C19 steroids producing strains in the industry (typically produce 12-13 g/l of C19 steroids from 30 g/l of phytosterols after 120 h), these HBCs producing strains displayed to be promising for industrial applications.

Discussion
The catabolism of sterols is a long and complex process. Although components of the metabolic process have been identified, several key steps remain unknown, seriously limiting our understanding of sterol metabolism. Hsd4A MN is a conserved gene in the gene cluster of sterol catabolism, the function of which has been speculated   but not verified. This gene has been speculated as a 17β -hydroxysteroid dehydrogenase or a β -hydroxyacyl-CoA dehydrogenase in sterol metabolism 15,23,36 . In this study, Hsd4A MN was first characterized as a dual-function enzyme that possesses both 17β -hydroxysteroid dehydrogenase and β -hydroxyacyl-CoA dehydrogenase activities in vitro. Subsequently, the dual-role of Hsd4A MN in the metabolism of sterols in vivo was also confirmed using a kshA-null strain NwIB-XII as a model. The 17β -hydroxysteroid dehydrogenase activity of Hsd4A MN in vivo was identified as the augmentation of Hsd4A MN in NwIB-XII significantly decreased the production of TS and BD (Fig. 3a). However, the identification of the β -hydroxyacyl-CoA dehydrogenase activity of Hsd4A MN in vivo was a difficult process. The deletion of hsd4A MN in NwIB-XII resulted in the production of 4-HBC and 1,4-HBC, rather than TS and BD, suggesting that Hsd4A MN could play another enzymatic function other than as a 17β -hydroxysteroid dehydrogenase in vivo. As 4-HBC and 1,4-HBC do not act as substrates for Hsd4A MN , these must be further derivatives of the physiological substrate of Hsd4A MN during sterol side-chain degradation. Given that sterol side-chain degradation is a consensus β -oxidation, according to the characterized enzymatic properties of Hsd4A MN in vitro, the observed β -hydroxyacyl-CoA dehydrogenase activity of Hsd4A MN is likely responsible for in vivo sterol side-chain degradation. In the side-chain β -oxidation of sterols (Fig. 5d), there are two steryl fatty acyl-CoA esters that are considered as the substrates of β -hydroxyacyl-CoA dehydrogenase: 24-hydroxy-3-oxo-chol-4-en-26-oyl CoA (24HOC-CoA) and 22HOBNC-CoA. Compared to 24HOC-CoA, 22HOBNC-CoA seemed to be the most likely substrate of Hsd4A MN , as 22HOBNC-CoA could theoretically be converted to HBCs via a common retro-aldol cleavage, while the conversion of 24HOC-CoA to HBCs would be a complex and difficult biochemical process (Fig. 5d). Furthermore, the fact that Hsd4A MN catalyzes the dehydrogenation of β -hydroxybutyryl-CoA in vitro further suggested that 22HOBNC-CoA would be a suitable substrate for Hsd4A MN , as β -hydroxybutyryl-CoA is an intrinsic moiety of 22HOBNC-CoA. If this was the case, 3,22-dioxo-25,26-bisnorchol-4-ene-24-oyl CoA (22OBNC-CoA) should be the direct derivative of 22HOBNC-CoA after catalysis by Hsd4A MN , which is a reaction known to be catalyzed by a β -hydroxyacyl-CoA dehydrogenase in the side-chain β -oxidation of sterols (Fig. 5d). Acetoacetyl-CoA and β -hydroxybutyryl-CoA are the side chain moieties of 22OBNC-CoA and 22HOBNC-CoA, respectively. As β -hydroxybutyryl-CoA and acetoacetyl-CoA could be reversibly interconverted by Hsd4A MN , it is conceivable that 22OBNC-CoA might also be a possible substrate of Hsd4A MN . 22OBNC-CoA has been characterized as a substrate of FadA5, a β -ketoacyl-CoA thiolase that catalyzes the thiolytic cleavage of 22OBNC-CoA to 3-oxo-22,23-bisnorchol-4-ene-22-oyl-CoA (OBNC22-CoA) 2 . To help decipher the above, fadA5 was deleted in NwIB-XII (XIIΔ fadA5, Table S1), which resulted in a similar phenotype to XIIΔ hsd4A, producing 1,4-HBC and 4-HBC from the metabolism of sterols (Fig. 4a). These results demonstrated that 22OBNC-CoA similar to 22HOBNC-CoA is also a precursor of HBCs, indicating a close relationship among 22HOBNC-CoA, 22OBNC-CoA, Hsd4A MN and HBCs. Structurally, the transformation of 22HOBNC-CoA or 22OBNC-CoA to HBCs requires a cleavage between the carbon 22 -carbon 23 bond of the side chain. Biochemically, the carbon 22 -carbon 23 bond cleavage of 22HOBNC-CoA could be performed by an aldolytic reaction, and the cleavage of 22OBNC-CoA could be achieved by a β -ketoacyl-CoA thiolysis. As the β -ketoacyl-CoA thiolysis of 22OBNC-CoA is carried out by FadA5 2 , the direct carbon 22 -carbon 23 cleavage of 22OBNC-CoA was impossible in XIIΔ fadA5. However, the direct carbon 22 -carbon 23 cleavage of 22HOBNC-CoA was feasible in XIIΔ fadA5 as a very similar reaction has been identified in cholate side-chain degradation in Pseudomonas sp strain Chol1 (Fig. 5e) 37 . In contrast to 22OBNC-CoA, therefore, the conversion of 22HOBNC-CoA to HBCs was more plausible by carbon 22 -carbon 23 cleavage. In this way, the conversion of 22OBNC-CoA to HBCs in XIIΔ fadA5 could be reasonably attributed to the reversible conversion of 22OBNC-CoA to 22HOBNC-CoA by Hsd4A MN . The conversion from 22HOBNC-CoA to AD (designated as the AD pathway, Fig. 5d) has long been proposed as the sole metabolic pathway of sterol side-chain degradation. Based on the results in this study, an additional pathway was proposed: the conversion from 22HOBNC-CoA to HBCs (designated as the HBC pathway, Fig. 5d). Between the two pathways, 22HOBNC-CoA is the branching-node, which leads to AD via the catalysis of Hsd4A MN and leads to HBCs via an aldolytic reaction (Fig. 5d). Chemically, the aldolytic reaction of 22HOBNC-CoA to 3-oxo-23,24-bisnorchol-4-ene-20-carbaldehyde (OBNC20CA) could be plausibly speculated according to the similar reaction catalyzed by a sal-encoded aldol-lyase within the degradation of cholate side chain in Pseudomonas sp. strain Chol1 (Fig. 5e) 37,38 . Further, two aldolytic reactions have been proposed and genetically characterized in sterol side-chain degradation, including carbon17-carbon20 cleavage by aldol-lyase Ltp2 39 and C24-branched chain cleavage by aldol-lyases Ltp3 and Ltp4 40 . For many organisms, aldehydes are a group of active compounds that are readily converted to alcohols by reductive enzymatic reactions or to carboxylic acids by oxidative enzymatic reactions. Therefore, the terminal aldehyde group of OBNC20CA has two possible further conversions. One is the oxidative transformation to 3-oxo-23,24-bisnorchol-4-ene-20-formic acid (OBNC20FA), and the other is the reductive reaction to HBCs. Obviously, the reductive conversion correlates with the results of this study, suggesting the presence of this reaction in M. neoaurum ATCC 25795. It must be noted that the presence of an oxidative conversion could not be determined from the results of this study. However, it is the established conversion in cholate side-chain degradation in Pseudomonas sp. strain Chol1 as a subsequent reaction catalyzed by an aldehyde dehydrogenase encoded by sad after the aldolytic reaction (Fig. 5e) 37 . Under the conditions of this study, HBCs were shown to be the dead-end products of sterols in XIIΔ hsd4A and XIIΔ fadA5, suggesting that the reductive conversion is a terminal reaction, negating further degradation of the sterol side chain. In contrast, the oxidative conversion may not affect the complete side-chain degradation, as the terminal formic acid group can be further converted to its CoA ester (i.e. OBNC22-CoA) by acyl-CoA synthetases, which will return the metabolite to the AD pathway. As the reduction and oxidation of aldehydes are common reactions in cell metabolism, the existence of the reductive conversion does not exclude the existence of the oxidative conversion in M. neoaurum ATCC 25795. The minor yield of C19 steroids in HBCs producing strains, such as the 1-2% molar yield of ADD in XIIΔ hsd4A and a small amount of AD in XIIΔ hsd4AΔ kstD123, implied that the oxidative conversion may also be active in these strains ( Table 2). The Scientific RepoRts | 6:21928 | DOI: 10.1038/srep21928 molar yields of HBCs were much higher than the C19 steroids in these engineered strains; therefore, the reductive conversion was likely the dominant way to yield HBCs in M. neoaurum ATCC 25795, which may be closely related to the oxidation-reduction potential in mycobacterial cells.

4-HBC
As HBCs are valuable precursors in steroid synthesis, mycobacteria have been developed to produce HBCs 8,11 . Previously, 23,24-bisnorcholenic acids were considered as derivatives from the blocked AD pathway, and HBCs were speculated to be the reduced derivatives of 23,24-bisnorcholenic acids via 23,24-bisnorcholenic aldehydes (Fig. 5d) 32 . However, this study disclosed another possibility to produce HBCs in mycobacterial mutants, which has practical significance in clarifying the metabolic mechanism of C22 steroids and in the development of engineered strains for producing specific C22 steroids.
In this study, three HBCs producing strains were constructed and the productivity of each was assessed using a resting cell system that was supplemented with 40 g/l of phytosterols (Table 2). To the best of our knowledge, the production performances of these HBCs producing strains were significantly superior to other documented strains (Table 2), and the productivity of 4-HBC (0.106-0.109 g/l/h) and 1,4-HBC (0.122-0.127 g/l/h) was comparable to industrial productivity of C19 steroids (about 0.100-0.108 g/l/h), such as AD and 9-OHAD. For industrial application, the genetic and performance stability of engineered strains is highly important. In this study, gene inactivation was performed by in-frame deletion in the core region of gene, resulting in permanent inactivation of the target gene. Gene overexpression was performed using a universal mycobacterial plasmid pMV261 with kanamycin resistance. The plasmid stability of pMV261-kstD1 in strain XIIΔ hsd4A-p261kstD1 was tested under conditions lacking kanamycin addition. The results showed that the loss rate of pMV261-kstD1 averaged less than 7% during 144 h (Fig. S6). In addition, the engineered strains developed in this study, as well as our previous study 3,4 , displayed stable production performances after irregular subculture for years. These results clearly demonstrated that the metabolic engineering strategies used in this study were effective in developing HBCs producing strains that were promising for industrial application, although their performances require further improvement.
All strains used in this study are listed in Supplementary Table S1. Oligonucleotides and plasmids are listed in Table S2. Mycobacteria were cultivated aerobically in MYC/01 medium, as described previously 3 . The stock solutions of sterols were prepared as reported 3 . E. coli DH5α used for molecular cloning and E. coli BL21(DE3) used for heterologous expression were grown in Luria-Bertani medium with appropriate amounts of antibiotics at 37 °C with shaking at 200 rpm.
Gene deletion and complementation in M. neoaurum. The genome was isolated from M. neoaurum ATCC 25795 and sequenced as indicated previously 4 . Basic bioinformatics tools for sequence alignments and searching for protein homologues were performed as previously described 3,4 .
Unmarked in-frame gene deletion of M. neoaurum ATCC 25795 was based on a non-replicative suicide plasmid, p2NIL, and combined with the selectable marker cassette from pGOAL19, following a procedure described previously 3,41 . To delete kshA1 and kshA2 from M. neoaurum, pDELkshA1 and pDELkshA2 were constructed. M. neoaurum NwIB-XII was an engineered strain with the unmarked deletion of kshA1 and kshA2, which served as the host to generate further gene knock-outs. The deletion of hsd4A MN in M. neoaurum ATCC 25795 and NwIB-XII (designated as MNΔ hsd4A and XIIΔ hsd4A, respectively) was carried out using plasmid pDELhsd4A. A deletion mutant of fadA5 was constructed in NwIB-XII (XIIΔ fadA5) using plasmid pDEL-fadA5. Three kstD genes (kstD1, kstD2 and kstD3) were knocked out sequentially in XIIΔ hsd4A or MNΔ hsd4A using p2N-k1, p2N-k2 and p2N-k3, respectively. Functional complementation through recombinant pMV306 (pMV306-hsd4A) was established as indicated 3 . A strain harboring a vacant pMV306 was used as the blank control (XIIΔ hsd4A-p306). All desirable mutants and complements were validated by PCR and gene sequencing. The gene sequences of kshA1, kshA2, hsd4A MN and fadA5 have been deposited in the GenBank database with the accession numbers KF573736, KF573737, KP642512 and KP642515, respectively.
In vivo expression of Hsd4A MN and KstD1. The coding sequences of hsd4A MN and kstD1 (GenBank ID: GQ411074.1) were amplified using PrimeSTAR ® High-fidelity DNA polymerase (Takara, Dalian, China) and the primers shown in Table S2, and then were digested by PstI-HindIII (for hsd4A MN ) or BamHI-HindIII (for kstD1) before being inserted into the corresponding sites in pMV261 42 . The resulting plasmids pMV261-hsd4A and pMV261-kstD1 were electro-introduced into NwIB-XII and XIIΔ hsd4A, respectively. The resulting strains were then selected under 30 μ g/ml kanamycin. A strain with a vacant pMV261 was used as the blank control. The plasmid stability of pMV261-kstD1 was assesssed as indicated in Supplementary Method. Expression and purification of Hsd4A MN . The coding region of hsd4A MN was ligated into vector pET-28a(+ ) via PCR amplification and restriction digestion (NcoI-XhoI). The resulting plasmid pET28-hsd4A was first replicated in E. coli DH5α and then transformed into E. coli BL21(DE3) to generate E-hsd4A for expression. E-hsd4A was cultured at 37 °C and shaken at 220 rpm in LB medium to OD 600 0.4-0.8 and IPTG at a concentration of 0.5 mM was added to induce Hsd4A MN expression. The induced cells were further cultured for 24 h at Scientific RepoRts | 6:21928 | DOI: 10.1038/srep21928 25 °C and shaken at 220 rpm until harvested by centrifuge at 8,000 × g for 20 min at 4 °C. The pellets were washed in buffer A (50 mM Tris-HCl pH7.5, 500 mM NaCl, 0.1 mM EDTA, 20 mM imidazole) plus 1 mM PMSF, centrifuged at 8,000 × g for 20 min at 4 °C and resuspended in 80 mL of buffer A. The cell-free extract of E-hsd4A was prepared by sonication for 15 min and centrifugation at 13,000 × g for 30 min at 4 °C. Purification of Hsd4A MN was conducted using an AKTA Prime system (GE Healthcare, Shanghai, China) as indicated in Supplementary Method. The standard Bradford method, with bovine serum albumin as the standard, was applied to determine the final protein concentration 43 . Characterization of Hsd4A MN . The 17β -hydroxysteroid dehydrogenase activity of Hsd4A MN was determined in triplicate as the rate of NAD + /NADH formation at 340 nm (ε = 6.22 mM −1 cm −1 ) and 30 °C, and recorded by a SpectraMax 190 (Molecular Devices; CA, USA) with the Soft-max PRO program. The assay mixture (200 μ l) was comprised of 100 mM Tris-HCl (pH 7.8 for oxidation or pH 6.6 for reduction), 100 μ M NAD + /NADH, 0.15 mM substrate and purified Hsd4A MN 44,45 . The β -hydroxyacyl-CoA dehydrogenase activity of Hsd4A MN was measured by monitoring the reduction of NAD + as described 46 . The assay mixture (200 μ l) was comprised of 125 mM glycine-NaOH buffer (pH 9.0), 125 μ M NAD + , 0.125 mM substrate (DL-β -hydroxybutyryl-CoA and DL-β -hydroxybutyric acid) and purified Hsd4A MN 46 . The acetoacetyl-CoA reductase activity of Hsd4A MN was assayed by measuring the oxidation of NADH 46 . The assay mixture (200 μ l) was comprised of 125 mM Tris-HCl (pH 8.0), 125 μ M NADH, 0.02 mM substrate (acetoacetyl-CoA and acetoacetate) and purified Hsd4A MN 46 . The β -hydroxyacyl-CoA dehydrogenase and acetoacetyl-CoA reductase activities were also recorded by the SpectraMax 190 (Molecular Devices). In these enzyme reactions, one unit of specific enzyme activity was indicated as the reduction or oxidation of 1 μ M NADH/NAD + within 1 min per