Improving the N-terminal diversity of sansanmycin through mutasynthesis

Background Sansanmycins are uridyl peptide antibiotics (UPAs), which are inhibitors of translocase I (MraY) and block the bacterial cell wall biosynthesis. They have good antibacterial activity against Pseudomonas aeruginosa and Mycobacterium tuberculosis strains. The biosynthetic gene cluster of sansanmycins has been characterized and the main biosynthetic pathway elucidated according to that of pacidamycins which were catalyzed by nonribosomal peptide synthetases (NRPSs). Sananmycin A is the major compound of Streptomyces sp. SS (wild type strain) and it bears a non-proteinogenic amino acid, meta-tyrosine (m-Tyr), at the N-terminus of tetrapeptide chain. Results ssaX deletion mutant SS/XKO was constructed by the λ-RED mediated PCR targeting method and confirmed by PCR and southern blot. The disruption of ssaX completely abolished the production of sansanmycin A. Complementation in vivo and in vitro could both recover the production of sansanmycin A, and the overexpression of SsaX apparently increased the production of sansanmycin A by 20 %. Six new compounds were identified in the fermentation culture of ssaX deletion mutant. Some more novel sansanmycin analogues were obtained by mutasynthesis, and totally ten sansanmycin analogues, MX-1 to MX-10, were purified and identified by electrospray ionization mass spectrometry (ESI-MS) and nuclear magnetic resonance (NMR). The bioassay of these sansanmycin analogues showed that sansanmycin MX-1, MX-2, MX-4, MX-6 and MX-7 exhibited comparable potency to sansanmycin A against M. tuberculosis H37Rv, as well as multi-drug-resistant (MDR) and extensive-drug-resistant (XDR) strains. Moreover, sansanmycin MX-2 and MX-4 displayed much better stability than sansanmycin A. Conclusions We demonstrated that SsaX is responsible for the biosynthesis of m-Tyr in vivo by gene deletion and complementation. About twenty novel sansanmycin analogues were obtained by mutasynthesis in ssaX deletion mutant SS/XKO and ten of them were purified and structurally identified. Among them, MX-2 and MX-4 showed promising anti-MDR and anti-XDR tuberculosis activity and greater stability than sansanmycin A. These results indicated that ssaX deletion mutant SS/XKO was a suitable host to expand the diversity of the N-terminus of UPAs, with potential to yield more novel compounds with improved activity and/or other properties. Electronic supplementary material The online version of this article (doi:10.1186/s12934-016-0471-1) contains supplementary material, which is available to authorized users.


Open Access
Microbial Cell Factories  [7]. The increasing emergence of multi-drug-resistant (MDR) and extensive-drug-resistant (XDR) tuberculosis make the treatment more difficult. So there is an urgent need to develop novel anti-TB drugs with no cross-resistance to current clinically used antibiotics. Sansanmycins and other UPAs are of interest, due that they inhibit a clinically unexploited target MraY (phospho-MurNAc-pentapeptide translocase, also known as translocase I) [8], which catalyzes the transfer of UDPMurNAc-L-Ala-γ-D-Glu-m-DAP-D-Ala-D-Ala onto lipid carrier undecaprenyl phosphate to give lipid intermediate I in the bacterial peptidoglycan biosynthetic pathway. The potential bioactivity against M. tuberculosis coupled to its intriguing structure made this natural product a fascinating anti-TB lead compound.
Recently, the biosynthetic gene clusters for pacidamycins [5,9], napsamycins [10], and sansanmycins [11] were identified and characterized, indicating that the assembly of the pseudo-tetrapeptide chain is catalyzed by nonribosomal peptide synthetases (NRPSs) with highly dissociated modules [12]. Besides, the biosynthesis of uridyl pentapeptide of pacidamycins was catalyzed by the tRNA-dependent aminoacyltransferase PacB, which transferred the alanyl residue from alanyl-tRNA to the N-terminus of the pseudo-tetrapeptide [13]. In contrast to ribosomal peptide synthesis, non-ribosomally assembled peptides contain not only the 20 proteinogenic amino acids but also many different building blocks, such as DABA, D-amino acids, hydroxyl amino acids, N-and C-methylated amino acids etc. Among them, non-proteinogenic amino acid meta-tyrosine (m-Tyr) is rarely found in bacterial secondary metabolites. UPAs and a potent cyclophilin inhibitor sanglifehrin A (SFA) are two examples containing m-Tyr as one of the building blocks. SfaA, identified in the SFA biosynthetic gene cluster, was speculated to catalyze the biosynthesis of m-Tyr [14]. As a homologue of SfaA in pacidamycin biosynthetic gene cluster, PacX was characterized as a phenylalanine 3-hydroxylase that catalyzed the synthesis of m-Tyr from L-phenylalanine (L-Phe) in vitro [15]. In sansanmycin biosynthetic gene cluster, SsaX is homologous to PacX with amino acid identity of 80 % across the whole protein, indicating that it is responsible for the biosynthesis of m-Tyr in Streptomyces sp. SS.
Although natural UPAs have potential to treat refractory infections, there is no UPAs entering clinical trials until now mainly due to their relatively poor physicochemical property. In previous studies, the N-terminal amino acid of the tetrapeptide of UPAs was supposed to be important functional group for the inhibition of MraY [16,17]. It was proposed that the protonated ammonium ion binds in place of the Mg 2+ cofactor at the MraY active site via cis-amide linkage [17]. The N-terminal amino acid of known UPAs is almost dominated by m-Tyr or different bicyclic acids (possibly derived from m-Tyr), except that some pacidamycins possess Ala instead (Fig. 1). In this study, we focus on the substitution of the N-terminal amino acid to get novel sansanmycin analogues by mutasynthesis. Mutasynthesis is a useful method in the generation of new antibiotic derivatives [18]. This approach could expand the chemical diversity of secondary metabolites and produce novel compounds with improved physicochemical properties or altered bioactivity. For example, it has been successfully employed to get novel nucleoside antibiotics such as nikkomycin analogues [19] and new ansamycin derivatives [20]. However, mutational biosynthesis has not been employed to obtain UPA derivatives so far.
Here, we demonstrated that SsaX is responsible for the biosynthesis of m-Tyr in vivo by gene deletion and complementation and the sansanmycin production could be increased through the overexpression of ssaX. Six new sansanmycin analogues were purified and characterized in ssaX deletion mutant, indicating the substrate flexibility of the responsible NRPS. To expand the diversity of sansanmycins by mutasynthesis, different types of substrates were fed to the ssaX deletion mutant and some novel sansanmycin derivatives were obtained. These compounds were purified and structurally identified, some of which exhibited improved antibacterial activity or stability.

In-frame deletion of ssaX and its complementation
In order to investigate the contribution of ssaX to sansanmycin biosynthesis, an ssaX deletion mutant SS/XKO was constructed from Streptomyces sp. SS by PCR targeting [21] using cosmid 13R-1 [11] which contains ssaX and the majority of other biosynthetic genes. Cosmid 13R-1-SCP2KO was firstly constructed from cosmid 13R-1 with the minimal replicon of SCP2* replaced by ampicillin resistance gene in order to promote homologous recombination for the disruption of ssaX. Then the ssaX gene in 13R-1-SCP2KO was in-frame deleted and the resulted 13R-1-SCP2KO-XKO was introduced into the wild type strain by conjugation (Fig. 2a). The ssaX in-frame deletion mutant SS/XKO was verified by PCR using primers PT-X-7 and PT-X-8 (Fig. 2a) and further confirmed by southern blot analysis (Fig. 2b, c). The coding region of ssaX was cloned into a pSET152 [22] -derived expression plasmid, pL646 [23], under the control of a strong constitutive promoter, ermE*p, to give pL-ssaX. The plasmid was introduced into SS/XKO strain by conjugation and the complementary strain SS/XKO/pL-ssaX was resulted to exclude the potential polar effect in SS/XKO. Metabolites of constructed strains were scrutinized to compare with that of the wild type strain by HPLC, LC/MS and bioassay analysis. The major product in the wild type strain, sansanmycin A with m-Tyr at the N-terminus of tetrapeptide chain, was not detected in SS/XKO (Fig. 2d). The culture broth of SS/XKO showed no antibacterial activity against P. aeruginosa 11, which was used as the test microorganism for the bioassay of sansanmycins (Fig. 2e). Meanwhile, the complementary strain SS/ XKO/pL-ssaX completely recovered the production of sansanmycin A detected by HPLC analysis (Fig. 2d) and antibacterial bioassay (Fig. 2e).
In addition to genetic complementation in vivo, a chemical complementation by feeding the substrate m-Tyr (3 mM) to the fermentation medium of the ssaX deletion mutant SS/XKO was also performed. The result of HPLC showed that the major product of this culture broth was sansanmycin A and the production level was significantly higher than the wild type strain (Fig. 3a). Consistent with this result, the inhibition zone was bigger than that of wild type strain in the bioassay (Fig. 3b). All of these results demonstrated that SsaX was responsible for the biosynthesis of m-Tyr in sansanmycin biosynthetic pathway.
To further investigate the role of ssaX in sansanmycin biosynthesis, the plasmid pL-ssaX was transferred into the wild type strain by conjugation to give the ssaX overexpression strain SS/pL-ssaX, with pSET152 transferred strain SS/pSET152 as a control. With the same growth curves, the overexpression of SsaX apparently increased the production of sansanmycin A by 20 % from HPLC analysis (Fig. 2d). The antibacterial activity against P. aeruginosa 11 also showed that SS/pL-ssaX exhibited bigger inhibition zone than the control strain (Fig. 2e). This result, together with the result of chemical complementation, suggested that the biosynthesis of m-Tyr catalyzed by SsaX is at least one of the rate-limiting steps in the sansanmycin production [24].

Isolation and structural determination of sansanmycin analogues in ssaX deletion mutant
Through the HPLC and LC/MS analysis, a series of minor components of sansanmycin analogues were detected in the cultivated broth of ssaX deletion mutant SS/XKO. In order to characterize these compounds, SS/XKO fermentations were scaled up to obtain enough amount of material for further analysis. The target compounds were enriched by macroporous absorbant resin from fermentation broth, then isolated using DEAE-Sephadex A25 guided by HPLC-UV to yield the crude sansanmycin analogues. Subsequently, the crude compounds were purified by preparative HPLC. As a result, six new sansanmycin analogues were obtained and designated as (See figure on next page.) Fig. 2 Effects of the deletion of ssaX on sansanmycin production. a Construction of the ssaX deletion mutant (SS/XKO) by PCR targeting and its PCR verification (using primers PT-X-7 and PT-X-8). Primers PT-X-1 and PT-X-2 were used to amplify a streptomycin resistance cassette (aadA gene), which substituted ssaX on cosmid 13R-1-SCP2KO. bla, ampicillin resistance marker; aac (3) Fig. 4) showed the same fragment patterns with sansanmycin A. All of these four compounds showed the same diagnostic fragment with sansanmycin A, m/z 701, corresponding to the loss of the N-terminal amino acid (AA 1 ), suggesting they varied at the N-terminal amino acid. In support of this hypothesis, the 1 H NMR data for MX-1 ( Fig. 5a; Additional file 1: Table S1), MX-2 (Additional file 1: Table S2, Figure S4), MX-4 (Additional file 1: Table S3, Figure S6) and MX-6 (Additional file 1: Table S4, Figure S8 The obtained six compounds MX-1-6 were new members of sansanmycin family. Compared with SS-A, MX-1, short of the N-terminal m-Tyr, bears a tri-pseudopeptide backbone that was found in the family of UPAs for the first time. It is the accumulated precursor when there is no m-Tyr present in SS/XKO, which is also the obvious evidence that SsaX catalyzes the biosynthesis of m-Tyr. The other five compounds were different from the known sansanmycins by the virtue of bearing Tyr, Phe and Met at the N-terminus, which were firstly reported in the family of UPAs. The presence of the new sansanmycin analogues with various N-terminal amino acids hinted that the NRPS responsible for the incorporation of the N-terminal amino acid into the tetra-pseudopeptide backbone has moderate substrate promiscuity, suggesting that certain amounts of sansanmycin analogues might be able to be generated by mutational biosynthesis using ssaX deletion mutant SS/XKO.

Generation of structurally diverse sansanmycin analogues using ssaX deletion mutant
Initially twenty proteinogenic amino acids including Phe, Tyr and Met were used to probe the feasibility of mutasynthesis. The production of sansanmycin MX-2 which bears Tyr as its N-terminus was nearly doubled when fed SS/XKO with Tyr (3 mM) (Fig. 3). Similarly, the production of sansanmycin MX-4 which bears Phe as its N-terminus was increased to two to three times when fed with Phe (3 mM) (Fig. 3). But the HPLC profile of the fermentation broth of ssaX deletion mutant fed with Met, as well as other proteinogenic amino acids (3 mM) had no obvious changes (data not shown). This may be explained by the substrate preference of the NRPS, which preferred to select Phe or Tyr rather than any other proteinogenic amino acids. This result is consistent with the production level of sansanmycin analogues in SS/XKO. The improved production of sansanmycin MX-2 and MX-4 by feeding substrates Phe and Tyr suggested that mutasynthesis might be suitable to produce sansanmycin analogues with alternate N-terminal substrate.
Incorporation of halogens into the molecules might exert a significant effect on their physicochemical properties of the products [26,27]. Halogenated Phe had been fed to get sansanmycin analogues in the wild type strain by Xie et al. and they were preferably incorporated at the C-terminus (AA 4 ) [28]. In order to incorporate halogenated Phe into sansanmycin analogues at the N-terminus efficiently, the ssaX deletion mutant SS/XKO was evaluated. Some new sansanmycin analogues were produced by SS/XKO and their structures were determined by ESI-MS/MS. These sansanmycin analogues may be divided into three main groups (Fig. 6a). Compounds in the first and second groups were with halogenated Phe incorporated into the N-terminus (e.g., compounds 1 and 2) or both the N-terminus and the C-terminus (e.g., compounds 3 and 4) respectively. The third group was with halogenated Phe incorporated into the C-terminus and with Phe at the N-terminus instead of m-Tyr (e.g., compounds 5 and 6). What's more, the production level of the analogues with halogenated Phe at the N-terminus was increased remarkably compared with the wild type strain with the same growth curves. For example, when fed with 2-chloro-and 2-bromophenylalanines (3 mM), the production level of compound 1 and 2 was 11-and 16-fold higher in SS/XKO respectively compared with the wild type strain (Fig. 6b). Similarly, compound 3 and 4 were increased 10-and 13-fold respectively (Fig. 6b). In a word, the ssaX deletion mutant was a better host to get sansanmycin analogues with N-terminus substitution, for its absence of the endogenous optimal substrate m-Tyr.
Motivated by the results of above feeding experiments and in order to expand the diversity of sansanmycins, more than 20 commercially available non-proteinogenic amino acid analogues were fed to SS/XKO (3 mM), including α-amino acids and β-amino acids with diverse side chains, such as aliphatic groups, substituted aromatic phenyl groups, as well as aromatic and nonaromatic heterocyclic groups (Fig. 7). The results showed that β-amino acids detected here couldn't be incorporated into the peptidyl chain of sansanmycins. When feeding methyl substituted phenylalanines, the major products were the C-terminal derivatives with Phe at the N-terminus. When supplemented with para-amino-phenylalanine, two products with the fed precursor incorporated at the N-(MX-9) and the C-terminus (MX-10) respectively were obtained. Unexpectedly, two aromatic heterocyclic α-amino acids, 2-furylalanine and 2-thienylalanine, can be incorporated into the N-terminus to produce corresponding derivatives MX-7 and MX-8, respectively. The structures of MX-7-10 were determined by ESI-MS/MS (Fig. 4). MX-7, MX-8 and MX-9 showed the expected molecular weights and the same diagnostic fragment with sansanmycin A, m/z 701 for the loss of the N-terminal amino acid, suggesting the precursors added were incorporated into the N-terminus (Fig. 4). MX-10 displayed the same fragment with MX-4, m/z 644 corresponding to the loss of the C-terminal amino acid, which suggested the p-amino-phenylalanine administrated was incorporated into the C-terminus (Fig. 4).
The yield of each sansanmycin derivative produced by mutasynthesis was estimated by the peak areas from HPLC ( Table 1). The production level of sansanmycin MX-4 in SS/XKO without supplementation was considered as 100 %. When fed with m-Tyr, production of sansanmycin A was the highest. Surprisingly, the production level of MX-7 bearing 2-furylalanine was much higher than that of MX-2 bearing Tyr. All of these results suggested that other unnatural substrates might be acceptable for the mutant strain, especially aromatic α-amino acids.

Antibacterial activity and stability of novel sansanmycin analogues
The obtained sansanmycin analogues MX-1-10 were tested for their antibacterial activity against different bacteria including gram-negative bacteria, gram-positive bacteria as well as M. tuberculosis H 37 Rv and clinically  Table 2).
As expected, all the tested compounds (except MX-3) displayed different degrees of activity against E. coli ΔtolC mutant strain, which was consistent with previous results that UPAs can be exported by the AcrAB-TolC efflux system in E. coli [29]. Among them, sansanmycin MX-2 and MX-6 remained potency against P. aeruginosa equivalent to sansanmycin A. Interestingly, compound MX-6 exhibited antibacterial activity against gram-positive B. subtilis, which was not found in natural UPAs. Sansanmycin MX-1, MX-2, MX-4, MX-6 and MX-7 showed potency against M. tuberculosis H 37 Rv comparable to sansanmycin A. More noticeably, these tested compounds exhibited equivalent potency against clinically isolated MDR and XDR strains. These results maybe lie in the fact that the target of UPAs is clinically unexploited.
During the early stage of drug development of sansanmycin A, we found that the structure of sansanmycin A was not stable at room temperature (Fig. 8a). As it is easier to get enough amount of sansanmycin MX-2 and MX-4, the stability test of sansanmycin A, MX-2 and MX-4 was performed in KH 2 PO 4 buffer (pH 6.0) at room temperature. Under this condition, sansanmycin MX-2 and MX-4 kept mostly unchanged after 6 days of incubation whereas less than 10 % sansanmycin A was remained (Fig. 8b), which suggested that sansanmycin MX-2 and MX-4 are more stable than sansanmycin A.

Discussion
During the past decade, considerable efforts have been made to exploit new UPA derivatives. Seventeen sansanmycin analogues were semi-synthesized with sansanmycin A as the starting material, but most of them exhibited less anti-mycobacterial activity in comparison with parent natural product [30]. Strategy of precursordirected biosynthesis was employed to get pacidamycin analogues with modified C-terminal amino acid through feeding Trp derivatives [26]. The same strategy was also applied to sansanmycin-producing strain, resulting sansanmycin analogues with the C-terminus substituted by Phe derivatives [28]. Although some UPA analogues were produced, few of them had significantly improved antibacterial activity and/or physicochemical property. In the past 5 years, the biosynthetic pathways have been studied extensively at the genetic, enzymatic and regulatory levels [9][10][11][12][13], and bioengineering approaches are available to be used in producing novel UPA derivatives. In this work, mutational biosynthesis is employed by blocking the biosynthesis of m-Tyr and then feeding variety of alternative substrates to produce novel sansanmycin derivatives. This strategy is efficient to obtain novel sansanmycin analogues, creating a great structural diversity at the N-terminus. In most of the reported UPAs, the N-terminus of the tetra-pseudopeptide (AA 1 ) was occupied by m-Tyr and its related bicyclic acids, except that pacidamycin D and S have an Ala at the N-terminus (Fig. 1). Two NRPSs responsible for selection of amino acid to incorporate into the N-terminus of pacidamycin have been reported. PacU was demonstrated to specifically activate Ala [5], and PacW was identified to activate m-Tyr [12]. In sansanmycin-producing strain, there are also two homologues of PacU, SsaU and SsaW, existed in sansanmycin biosynthetic gene cluster [11], but their amino acid sequences were almost exactly the same, with only one alteration of Lys to Arg [11]. Though they exhibited high overall homology with both PacU and PacW, the main residues in the binding pocket (specificity-conferring code) in the adenylation (A) domain [31,32] were same with PacW but different from PacU. This is consistent with the fact that m-Tyr is the optimum substrate for the production of sansanmycin A in the wild type strain, and all the natural sansanmycin derivatives have m-Tyr and its related bicyclic acids at the N-terminus. When the biosynthesis of m-Tyr is blocked, Phe, Tyr, and small amount of Met may incorporate into the polypeptide of sansanmycin in ssaX deletion mutant, showing the potential of this mutant as a cell factory to expand the chemical diversity of AA 1 .
Various substrates including Phe, Tyr, halogenated Phe and other non-proteinogenic amino acids were fed to the ssaX deletion mutant and about 20 novel sansanmycin analogues with different N-termini were produced according to the substrate promiscuity of the A domain of SsaU and SsaW. Ten of these compounds were purified and structurally determined by ESI-MS, ESI-MS/ MS (Fig. 4) and NMR. According to the production level of the compounds in ssaX deletion mutant fed with corresponding substrate (Table 1), the preference of the A domain in charge of activating the sansanmycin N-terminal amino acid is m-Tyr > Phe ≥ ortho-halogenated phenylalanines ≥ 2-furylalanine > Tyr > 2-thienylalanine ≥ p-amino-phenylalanine > Met. Out of our expectation, 2-furylalanine and 2-thienylalanine could be incorporated into the sansanmycin, suggesting the diversity of AA 1 might be worth to explore further by trying more structurally diverse substrates. The compounds produced by mutasynthesis in this study retained the anti-TB activity of sansanmycin A in vitro, and more encouragingly, they showed similar activity to MDR and even XDR strains isolated from patients. Meanwhile, the stability of MX-2 and MX-4 was demonstrated to be greatly improved compared to sansanmycin A. From the result of previous researches on the UPA derivatives of the wild type strains [8,25] or by precursor-directed biosynthesis [26,28], the A domain responsible for activating the C-terminal amino acid (AA 4 ) has relatively high substrate promiscuity from Trp, Tyr to Phe and substituted Phe and Trp. On the other hand, the amino acid at the position of AA 3 also can vary from Met and Ala [8] to Leu and Phe in sansanmycins [25,33]. Now, together with the diversity at AA 1 produced in this study, the combination of the variations in these three parts of the polypeptide backbone could be potentially expected that hundreds of new sansanmycin analogues might be obtained. Recently, crystal structure of Aquifex aeolicus MraY has been published [34], and the residues (Asp 117 , Asp 118 , Asp 265 , and His 324 ) important for the activity of MraY in the active site have been elucidated [34]. The structural information of MraY from A. aeolicus sets foundations for homologous modeling of MraY from M. tuberculosis [34,35], which will facilitate the study on structure activity relationship (SAR) of novel chemically diverse UPA derivatives obtained by further rationale genetic engineering manipulation.

Conclusion
It is demonstrated that SsaX is responsible for the biosynthesis of m-Tyr in vivo by gene deletion and complementation and the sansanmycin production could be increased through the overexpression of ssaX. Six new sansanmycin analogues were purified and characterized in ssaX deletion mutant, indicating the substrate flexibility of the responsible NRPS. The diversity of sansanmycin was further expanded by mutasynthesis, in which different types of substrates were fed to the culture of ssaX deletion mutant. Totally ten compounds were purified, structurally identified and firstly reported. Five of them displayed anti-mycobacterial activity comparable to sansanmycin A and especially, they are active to MDR and even XDR M. tuberculosis clinical strains. In addition, sansanmycin MX-2 and MX-4 displayed significantly improved stability than sansanmycin A. These improved properties may promote the novel anti-TB drug investigation targeting a clinically unexploited target MraY.

Strains, plasmids and growth conditions
The wild-type Streptomyces sp. SS strain, obtained from China Pharmaceutical Culture Collection (CPCC 200442), was used as a host strain for the propagation and disruption of genes, as described previously [11]. Streptomyces sp. SS and its derivatives were grown at 28 °C on solid S5 medium [36] for sporulation and in the liquid fermentation medium [1] for the production of sansanmycins. Mannitol soya flour (MS) agar [37] and liquid phage medium [38] were used for conjugation and isolation of genomic DNA respectively. Escherichia coli DH5α was used as a host for general cloning experiments [39]. E. coli ET12567/pUZ8002 [40] was used for conjugal transfer according to the established protocol [37]. E. coli BW25113/pIJ790 was used as the host for Red/ETmediated recombination [21]. E. coli DH5α containing the temperature sensitive FLP recombination plasmid BT340 was used as the host to remove the central part of the disruption cassette [21]. E. coli DH5α and ET12567/ pUZ8002 were incubated in Luria-Bertani medium (LB) [39] at 37 °C. E. coli BW25113/pIJ790 and DH5α/BT340 were grown at 30 °C in LB medium. P. aeruginosa 11, the indicator strain for the antibacterial bioassay of culture broth of different Streptomyces sp. SS strains [6], was grown on F403 agar [36]. E. coli ΔtolC mutant, P. aeruginosa 11, Bacillus subtilis CMCC (B) 63501, M. tuberculosis H 37 Rv and clinically isolated strains were used for testing antimicrobial activity for the compounds. When required, strains were incubated with apramycin (Am, 50 μg/ml), ampicillin (Amp, 100 μg/ml), spectinomycin (Spec, 50 μg/ml), kanamycin (Km, 50 μg/ml) and chloramphenicol (Cm, 30 μg/ml). All the strains and plasmids used in this study are listed in Table 3.
For complementation analysis, complete ssaX coding region was amplified using primers pL-ssaX-F (5′-CGCATATGCAAGGGCATCGCGAC-3′) and pL-ssaX-R (5′-ATAGGATCCTCAGCGCCGGGTG CC-3′), and then cloned into the NdeI and BamHI sites of a pSET152-derived expression plasmid, pL646 [23], under the control of a strong constitutive promoter ermE * p. The resulted expression vector pL-ssaX was transferred into SS/XKO and Streptomyces sp. SS by conjugation to give the complementation strain and ssaX overexpression strain respectively. The plasmid pSET152 [22] was transferred to SS/XKO and the wild type strain respectively as controls.

Analysis of sansanmycin production
Fermentation, isolation, and high-pressure liquid chromatography (HPLC) analysis of sansanmycins were carried out as described previously [1,25]. In brief, pieces of well-grown agar cultures of different strains were firstly inoculated in fermentation medium and cultured at 28 °C for 48 h at 200 rpm. The obtained seed cultures were trans-inoculated into three parallel 100 ml fermentation medium by 5 % inoculation and grown at 28 °C for 5 days at 200 rpm. In the feeding test, each exogenous substrate was added to the fermentation medium to the final concentration of 3 mM. At indicated time points, five-milliliter cell cultures were collected by centrifugation and dried at 60 °C to constant weight for monitoring the growth curve. The obtained supernatants were analyzed for antibacterial activity and production of sansanmycins by bioassay and HPLC. Antibacterial activity was measured by cylinder plate method using P. aeruginosa 11. For analyzing the expected analogues, the supernatant of fermentation broth was enriched by Sep-Pak C 18 Classic Cartridge (Waters Associates, Milford, MA, USA), eluted with 60 % methanol solution. The effluent was subjected to HPLC on an XBridge ™ C 18 column (4.6 × 150 mm, 3.5 μm, Waters, Dublin, Ireland) maintained at 40 °C, with a gradient of 80:20 0.1 % (w/v) (NH 4 ) 2 CO 3 -MeOH to 40:60 in 40 min as mobile phase at a flow rate of 1 ml/min. Absorbance was monitored at 254 nm. For the analysis of sansanmycin MX-3 and MX-6, the mobile phase was changed to 10:90 MeOH-H 2 O (pH adjusted to 12.0 with NH 3 ·H 2 O) in 40 min.

Purification of sansanmycin analogues
Isolation and purification of sansanmycin analogues was performed following the method of Xie et al. [25] with some modifications. Fifty liters of fermentation supernatant was obtained by centrifugation and then applied on a column of macroporous absorbant resin 4006. The active materials were eluted with 30 % aqueous acetone. Then the effluent was applied on Toyopearl DEAE-Sephadex A25 eluted with Tris-HCl (20 mM, pH 8.5) plus NaCl and monitored by HPLC-UV. The concentration of NaCl was adjusted with different compounds from 0.01 to 0.05 M. The effluent containing target compounds was collected and further purified by preparative HPLC (YMC-Pack ODS-A 5 μm, 250 × 20 mm column, 0.1 % (w/v) (NH 4 ) 2 CO 3 -MeOH; flow rate, 5 ml/min; UV detection at 254 nm and oven temperature at 40 °C). The ratio of 0.1 % (w/v) (NH 4 ) 2 CO 3 and MeOH was dependent on different compounds. The structures of obtained compounds were determined using ESI-MS and ESI-MS/ MS (ThermoFisher LTQ Orbitrap XL mass spectrometer) as well as NMR [Varian Mercury 600 spectrometers, in dimethyl sulfoxide (DMSO)-d 6 ].

Antibacterial assay
The minimum inhibitory concentrations (MICs) for M. tuberculosis strains were determined by the microplate Vector used as the template for amplifying aadA cassette, Spec r [21] pSET152 Streptomyces integrative vector, Am r [22] pL646 pSET152 derivative containing the constitutive promoter ermE*p, Am r [36] pL-ssaX pL646 derivative plasmid containing 843 bp complete coding region of ssaX This study