In silico analysis of the putative KstDs
By whole-genome sequencing, five putative encoding sequences of KstD genes were found and designated as kstD1, kstD2, kstD3, kstD4, and kstD5. The physicochemical properties of five KstD proteins are displayed in Table S2. All kstDs had more than 65% of GC content and were hydrophilic and contained one or two transmembrane regions, and no distinct signal was predicted in their leading region. It was found that four key residues in KsdD1 from R. erythropolis SQ1 [29] were highly conserved in five KsdDs. The N-terminal flavin adenine dinucleotide (FAD)-binding motif, GSG(A/G)(A/G)(A/G)X17E [30], for characterized KstD proteins, was fully conserved in the five KstDs in Fig. S1.
Phylogenetic analysis was performed based on the experimentally validated KstD proteins from typical Mycobacteria and Rhodococci to further disclose the evolutionary relationships among these KstDs (Fig. 2). These representative KsdDs were majorly located into four clusters but excluded KstD5 homologues. The sequence of KstD1 shared a high aa identity with KstD1 of M. neoaurum DSM1381(88%) [9], KstD1 of M. tuberculosis H37Rv (81%) [31], KstD1 of M. smegmatis MC2155(88%) [32], and KstD3 of R. erythropolis SQ1 (65%) [33]. KstD2 showed 65.8%, 66.5%, 64.5%, and 68% of identity with the KstD2 of M. neoaurum DSM1381, M. neoaurum ATCC 25795, R. erythropolis SQ1, and R. ruber Chol-4, respectively [9, 15, 18, 26]. It has been proved that these homologs had high Δ1-dehydrogenation activity. KstD3 shared 66.5% and 66.7% of aa identity with the KstD3 of M. neoaurum DSM1381 and M. neoaurum ATCC 25795, but 46.1%, 70% and 66.4% of aa identity with the KstD1 of R. erythropolis SQ1, R. ruber Chol-4 and KstD2 of M. smegmatis MC2155, respectively. In addition, KstD4 and KstD5 homologues from M. smegmatis MC2155[34] were closely related to KstD4 and KstD5 of ATCC 35855 in the phylogenetic branch and the protein identity was up to 80.7% and 83.2%, respectively.
The insights into their genomic allocation distinguished the status of various KstDs homologs are shown in Fig. 3. Five kstDs were in four different steroid degradation gene clusters and far away from each other. The loci of kstD2 and kstD3 were far away in the characterized steroid catabolism gene cluster and each of them accompanied by an oxygenase gene of KshA was organized in a neighboring community on the genome, which suggested that KstD2, KstD3 and KshA probably played a vital role in the metabolism of closely-related steroid nucleus [18]. KstD1 is surrounded by hsaE, hsaG, hsaF and hsd4B, and a similar structure can be found in M. neoaurum DSM 1381, M. smegmatis MC2155 and M. tuberculosis H37Rv. However, kstD4 and kstD5 were located in one gene cluster closely, and no sterol degradation-related genes with established functions were found in their vicinity. Moreover, only one highly conserved orthologous counterpart of kstD4-5 was found in M. smegmatis MC2155. This indicated that KstD4 and KstD5 are substitutable in steroid metabolic engineering.
In general, the analysis of kstDs genome allocation and phylogenetic tree showed that there are similarities and differences in steroids metabolic pathway compared with the reported strains.
Heterologous Expression Of Kstd Homologues
The expression of KstDs in E. coli BL21 (DE3) was identified through SDS-PAGE (Fig. S3). The enzyme activities of these KstDs in E. coli BL21 (DE3) were investigated using crude cell-free extracts listed in Table 2(). Although all KstDs can catalyze the Δ1-dehydrogenation of steroids, such as AD, 9-OHAD, HP and 9-OHHP, some significant differences were observed in their substrate preferences. The lowest catalytic activity of Kstd4 and KstD5 implied their negligible role in steroid metabolism. However, KstD1 showed higher activity for AD, 9-OHAD, 4-HP, and 9-OHHP, with the activities of 2.92, 3.43, 3.17 and 2.75 mU/mg, respectively. KstD2 displayed the highest specific activity toward four substrates: 261.75, 533.41, 2,104.38 and 1,203.52 mU/mg, and kstD3 displayed a weak dehydrogenation catalytic activity: 60.61, 51.78, 19.03 and 13.77 mU/mg. The specific enzyme activity of kstD2 and kstD3 for substrates was more than 10 folds higher than that of kstD1, kstD4 and kstD5, which indicated that the two KstDs may play a major role in the degradation of 9-OHAD. Interestingly, the specific enzyme activity of KstD2 to 4-HP was twice than that of AD. On the contrary, compared with substrate 4-HP, KstD3 showed higher specific activity for AD. For C9 hydroxylated steroids, the specific activity of KstD2 and KstD3 to 9-OHAD was 1.5 folds higher than that to 9-OHHP. Most importantly, KstD2 had higher specific enzyme activity for C9 hydroxylated steroids than C9 non-hydroxylated steroids as substrate, while KstD3 had completely opposite effects. The physicochemical properties of the two enzymes were consistent with those KstDs from other strains, such as Mycobacterium sp. VKMAc-1817D [14], M. fortuitum ATCC 6842 and its mutant HA-1[23] and some Rhodococcus [15, 35]. The distinct preferences of substrates demonstrated that these KstD homologs may play diversified roles in the catabolic pathway of sterols.
Assessment of kstD transcription
In order to understand the expression of kstDs in vivo, the transcriptome analysis of ATCC 35855 induced by cholesterol, AD and 9-OHAD was performed. Figure 4 illustrates that the transcription levels of kstD2 and kstD3 increased 5.74 and 2.46-folds when being induced with AD. Obviously, kstD2 and kstD3 involved in steroids metabolism might be more important. However, the transcription levels of kstD1, kstD4 and kstD5 decreased. Although 9-OHAD was the optimum substrate of kstD2, it did not cause a significant increase in the transcriptional level. Similarly, all kstDs were also not significantly up-regulated under the treatment of cholesterol, and the transcription levels increased less than 50%. This performance of kstDs was significantly different from that of other homologous analogues in steroid metabolizing strains such as M. neoaurum DSM1381 and M. neoaurum ATCC 25795.
Improving The 9-ohad Accumulation By Knocking Out Kstds
Compared with other steroid transformation strains [36–38], ATCC 35855 has the advantages of short growth cycle and fast degradation of phytosterols (Fig. 5 AB). Obviously, the degradation of 9-OHAD indicates that KstD genes exists in the strain (Fig. 5B). To elucidate the roles of kstDs in 9-OHAD degradation, the single and multiple deletions of ksdDs mutants were constructed. The phytosterols transformation was performed with these mutants to determine the function of five ksdDs in the degradation of 9-OHAD. Within 72 h of fermentation, 10 g/L phytosterols were completely consumed by all mutants. The yield of 9-OHAD reached the maximum after about 48h of inoculation with each of the mutants. Among all single kstD mutants (Fig. 5D), MFΔkd2 remarkably produced 9-OHAD to a maximum of 4.94 g/L at 72h, but a slower degradation occurred and decreased to 4.23 g/L after 120h. The fastest degradation occurred in MFΔkd4 and MFΔkd5 after 48h, which showed the same phenomenon with ATCC 35855, and all of them reached the peak of 9-OHAD yield (about 3.9 g/L) at 48 hours, and then decreased rapidly, which suggested that kstD2 plays the major role in the A-ring dehydrogenation of 9-OHAD combined with enzyme activity data. Compared with ATCC 35855, MFΔkd1 and MFΔkd3 showed a weak improvement on the degradation of 9-OHAD in the middle of fermentation which indicated that KstD1 and KstD3 also contribute to the degradation of 9-OHAD although their dehydrogenation activity is not high. Considering that the other four KstD homologs also displayed some dehydrogenase activity, they were successively knocked out in MFΔkd2. In Fig. 5E, all the different multiple kstD deficiency mutants showed that the characteristics of 9-OHAD were not being degraded even if the fermentation time was extended to 168 h. The maximum yield of 9-OHAD of MFΔkd23 was 5.05 g/L; MFΔkd123, MFΔkd123 and MFΔkstD accumulated 5.14, 5.18 and 5.29 g/L 9-OHAD, respectively. Obviously, all the five kstDs deletion mutant, MFΔkstD, seems to be superior with the extension of fermentation time (Fig. 5E). Single kstD compensation was performed in MFΔkstD to further verify the performance of each KstDs in the accumulation of 9-OHAD from phytosterols (Fig. 5F). The yield of 9-OHAD reached a maximum of 5.19 g/L, and 9-OHAD began to decrease slightly at 72 h with MFΔkstD-3. MFΔkstD-2 only accumulated 1.4 g/L 9-OHAD, and it was completely degraded within 72 h soon. However, the accumulation of 9-OHAD did not show a significant decrease in MFΔkstD-1, MFΔkstD-4 and MFΔkstD-5. These results proved that kstD2 and kstD3 have dominant contributions to 9-OHAD degradation, but the other three kstDs seem to play an optional role in this process.
Elimination of by-products by deletion of Opccr and overexpression of hsd4A and fadE28-29
Compared with ATCC 35855, the output of 9-OHAD increased from 3.87 g/L to 5.29 g/L in MFΔKstD. However, the purity of 9-OHAD in the fermentation was only 80.24% due to the existence of by-products such as 9-OHHP (5.46%) and 9,24-DHC (5.24%) (Fig. 6A). The presence of C22 intermediate 9-OHHP indicated that C22 steroidal metabolic pathway may play a weak role in cholesterol metabolism (Fig. 1). Hsd4A, a key enzyme located at the bifurcation of phytosterol metabolism pathway, could manipulate the phytosterols metabolic flux to accumulate either C19 or C22 steroid products through enhancement or loss of catalytic function. However, the exogenous hsd4A was over-expressed in MFΔkstD, without obvious effects on the reduction of the accumulation of 9-OHHP (Fig. 6A). In the AD pathway, mnOpccr has been proved to be able to convert 3-OPA and 3-OPC into 4-HP [22]. It was speculated whether there is an enzyme with the same function in the genome of ATCC 35855 and contributes to the accumulation of 4-HP or 9-OHHP. Subsequently, opccr, the homologous gene (76.3% identity) of mnOpccr was found in ATCC 35855 genome by BLAST and then knocked out in MFΔkstD. Compared with MFΔkstD, the ratio of 9-OHHP in the product had not changed significantly with MFΔkstDΔopccr (Fig. 6A). The overexpression of hsd4A or knockout of Opccr alone could not effectively reduce the accumulation of 9-OHHP. One of the reasons may be that Hsd4A is a bidirectional functional enzyme. Even if the activity was increased, the metabolic flow to the 9-OHHP pathway was not changed. Besides, 9-OHHP pathway was not completely blocked through the deletion of Opccr that only undertook part of the catalytic function due to some possible isoenzymes. Therefore, MFΔkstDΔOpccr_Hsd4A, in which the exogenous hsd4A was overexpressed in MFΔkstDΔOpccr, was constructed and the by-product 9-OHHP was significantly reduced. The proportion of 9-OHHP in the product was only 0.89%, compared with MFΔKstD, and decreased by 83.70%
The accumulation of 9,24-dihydroxychol-4-en-3-one(9,24-DHC), an intermediate in the incomplete degradation of side chain, may be caused by insufficient activities of side chain degradation related enzymes. Therefore, its accumulation can be reduced by increasing the activity of side chain degradation related enzymes theoretically. Some enzymes related to side chain degradation, such as fadE26-27, fadE28-29, fadE34, etc., were knocked out in MFΔkstD to verify the possible functions in the process of generating 9,24-DHC. Surprisingly, a significant yield change of 9,24-DHC was found in FadE26-27 and FadE28-29 deletion mutants, separately. As showed in Fig. 6B, the proportion of 9,24-DHC increased from 5.24–9.78% after fadE26-27 was deleted, while increased to 16.79% after fadE28-29 was knocked out. Since fadE28-29 has more significant contribution to the accumulation of product 9,24-DHC than fadE26-27, fadE28-29 was overexpressed in MFΔkstDΔopccr to try to eliminate the accumulation of by-product 9,24-DHC. It is satisfactory that the proportion of 9,24-DHC in the product decreased from 5.24–2.04%, and the purity of 9-OHAD in the product increased from 78.6–85.12%.
Based on the above results, MF-FA5020, a mutant that makes hsd4A and fadE28-29 were co-express in MFΔkstDΔopccr, was constructed to attempt to remove by-products 9-OHHP and 9,24-DHC simultaneously. According to Fig. 6B, the proportions of 9-OHHP and 9,24-DHC in the product were reduced to less than 3% and the proportion of 9-OHAD was 90.14%.
Evaluation Of 9-ohad Producer
To evaluate the ability of MF-FA5020 of transforming phytosterols into 9-OHAD, higher concentrations of phytosterols were incubated with the mutants. Figure 7 illustrates that the conversion rates of 10, 20 and 30 g/L phytosterols were over 168 h of fermentation, respectively. Specifically, the mutant can completely convert 10 g/L phytosterols into 6.12 g/L 9-OHAD within 96 h, and the molar yield of 9-OHAD was 84.18%. Compared with ATCC 35855, the yield of 9-OHAD of MF-FA5020 raised 53.85%. 12.21 g/L 9-OHAD was accumulated within 144 h from 20g/L phytosterols and the molar yield of 9-OHAD was 83.68%. Importantly, phytosterols were also completely consumed. However, even if the fermentation process was prolonged, the phytosterols were not exhausted when they increased to 30 g/L. Finally, 17.62 g/L 9-OHAD was still produced. The molar yield of 9-OHAD is 80.53%, and the molar conversion of phytosterol is 84.46%. What’s more, the production of 9-OHAD did not increase with the increase of phytosterols concentration at the early stage of transformation. This phenomenon was probably caused by the pH change, insufficient oxygen content, the inhibition of high concentrations of sterols on the growth of bacteria, and the reduction of nutrients in the later stage of fermentation. Although 30g/L sterol could not be completely transformed, MF-FA5020 still had great advantages compared with many reported 9-OHAD producing strains.