8-Deoxy-Rifamycin Derivatives from Amycolatopsis mediterranei S699 ΔrifT Strain

Proansamycin X, a hypothetical earliest macrocyclic precursor in the biosynthesis of rifamycin, had never been isolated and identified. According to bioinformatics analysis, it was proposed that RifT (a putative NADH-dependent dehydrogenase) may be a candidate target responsible for the dehydrogenation of proansamycin X. In this study, the mutant strain Amycolatopsis mediterranei S699 ΔrifT was constructed by deleting the rifT gene. From this strain, eleven 8-deoxy-rifamycin derivatives (1–11) and seven known analogues (12–18) were isolated. Their structures were elucidated by extensive analysis of 1D and 2D NMR spectroscopic data and high-resolution ESI mass spectra. Compound 1 is a novel amide N-glycoside of seco-rifamycin. Compounds 2 and 3 feature conserved 11,12-seco-rifamycin W skeleton. The diverse post-modifications in the polyketide chain led to the production of 4–11. Compounds 2, 3, 5, 6, 13 and 15 exhibited antibacterial activity against Staphylococcus aureus (MIC (minimal inhibitory concentration) values of 10, 20, 20, 20, 40 and 20 μg/mL, respectively). Compounds 14, 15, 16, 17 and 18 showed potent antiproliferative activity against KG1 cells with IC50 (half maximal inhibitory concentration) values of 14.91, 44.78, 2.16, 18.67 and 8.07 μM, respectively.

Rifamycins were first reported in 1957 from Amycolatopsis mediterranei S699 [8][9][10]. They have good antibacterial activity against G + bacteria and some G − bacteria [11]. The semi-synthetic rifamycin derivatives such as rifampicin are clinically used for the treatment of tuberculosis, adhesion and leprosy infection caused by Staphylococcus and other G + bacteria [12,13]. In recent years, due to its widespread use, pathogens, especially Mycobacterium tuberculosis, have gradually developed resistance to rifampicin [14,15]. In order to increase the structural diversity of rifamycins, the mechanism of rifamycin biosynthesis has been continuously studied [16][17][18].

Strains and Plasmids
Amycolatopsis mediterranei S699 strain was a gift from Prof. Linquan Bai at Shanghai Jiaotong University. A. mediterranei S699 ΔrifT strain was constructed by deleting the rifT gene which was predicted to be responsible for the dehydrogenation of putative proansamycin X [20][21][22]. The strain was initially propagated on ISP2 agar medium (4 g/L yeast extract, 10 g/L malt extract, 4 g/L glucose and 20 g/L agar). Then, a single colony of each strain was inoculated in 50 mL of ISP2 medium with 8 g of glass beads (Ø 3 ± 0.3 mm) in a 250 mL baffled flask and cultivated at 28 °C in a shaking incubator at 200 rpm. Escherichia coli DH5α strain was used as the general cloning host and grown in Luria-Bertani (LB) medium. Cell stocks were prepared with 20% glycerol and stored at −80 °C. Apramycin was added into media as an antibiotic at a final concentration of 50 μg/mL for all strains in this study.

Strains and Plasmids
Amycolatopsis mediterranei S699 strain was a gift from Prof. Linquan Bai at Shanghai Jiaotong University. A. mediterranei S699 ∆rifT strain was constructed by deleting the rifT gene which was predicted to be responsible for the dehydrogenation of putative proansamycin X [20][21][22]. The strain was initially propagated on ISP2 agar medium (4 g/L yeast extract, 10 g/L malt extract, 4 g/L glucose and 20 g/L agar). Then, a single colony of each strain was inoculated in 50 mL of ISP2 medium with 8 g of glass beads (Ø 3 ± 0.3 mm) in a 250 mL baffled flask and cultivated at 28 • C in a shaking incubator at 200 rpm. Escherichia coli DH5α strain was used as the general cloning host and grown in Luria-Bertani (LB) medium. Cell stocks were prepared with 20% glycerol and stored at −80 • C. Apramycin was added into media as an antibiotic at a final concentration of 50 µg/mL for all strains in this study.

Gene Knock-Out
The knock-out plasmid for the rifT gene was generated as the following steps. The~2 kb upstream and downstream homologous arms of the target genes were amplified by polymerase chain reaction (PCR) using A. mediterranei S699 genomic DNA as a template, respectively. Purified PCR fragments were inserted into the linearized pOJ260 by Gibson assembly [24]. The assembled product was then transformed into 100 µL DH5α-competent cells. Positive clones were identified by restriction enzyme digestion (Supplementary Figure S1) and DNA sequencing. The knock-out plasmid was propagated in DH5α and transformed into A. mediterranei S699 competent cells by electroporation, as described by Ding et al. [25]. The apramycin-resistant recombinants resulting from the homologous recombination between the knock-out plasmid and genomic DNA of A. mediterranei S699 were selected (Supplementary Figure S2) and transferred to ISP2 agar for several rounds of nonselective growth. Apramycin-sensitive recombinants derived from double-crossover recombination were screened, from which the targeted gene knockout mutant was verified by PCR (Supplementary Figure S3). The specific process is shown in the Supplementary Information (Supplementary Figure S4).

Gene Complementation
The integrating vector pSET152 [23] was used for gene complementation in the A. mediterranei S699 ∆rifT strain. Synthesized rifK promoter fragment was digested with NdeI and XbaI, and inserted into XbaI-pretreated pSET152 vector to yield pSET152-rifKp. The rifT gene was amplified by PCR using the genomic DNA of A. mediterranei S699 as a template. The NdeI/EcoRI rifT PCR fragment was inserted into the downstream of the rifKp promoter in pSET152. Positive clones were identified by restriction enzyme digestion and DNA sequencing. The resultant plasmid pSET152-rifKp-rifT was transformed into A. mediterranei S699 ∆rifT-competent cells by electroporation to obtain the rifT gene complementation mutant.

Detection and Analysis of the Metabolites in Mutants
For rifamycins production, A. mediterranei S699 mutants were inoculated on ISP2 agar media (100 mL) and cultivated for 7 days at 28 • C. The culture was diced and extracted overnight with EtOAc/MeOH (4:1, v/v) at room temperature. The concentrated crude extract was dissolved in 1 mL MeOH, and analyzed by high-pressure liquid chromatography (HPLC; Agilent 1200). Chromatographic conditions were as follows: solvents: (A) water, and (B) CH 3 CN, samples were eluted with a linear gradient from 20% to 35% B in the first 5 min, increased to 55% B at 19 min, to 65% B at 20 min, to 100% B at 23 min, followed by 4 min with 100% B, flow rate was 1 mL/min, and UV detection at 254 nm. In order to specify compound peaks, the concentrated crude extract was analyzed by liquid chromatography-electrospray ionization-high-resolution mass spectrometry (LC-ESI-HRMS; Finnigan). Chromatographic conditions were as follows: solvents: (A) water, and (B) CH 3 CN, samples were eluted with a linear gradient from 30% to 45% B in the first 10 min, increased to 65% B at 15 min, to 90% B at 19 min, followed by 5 min with 100% B, flow rate: 1 mL/min, and UV detection at 254 nm.

General Experimental Procedures
The NMR spectra were recorded on Bruker 400 MHz and/or AVANCE 600 MHz NMR spectrometers with tetramethylsilane (TMS) as an internal standard. HRESIMS analyses were carried out on a LTQ-Orbitrap XL (Thermo Scientific, Waltham, MA, USA). Silica gel GF 254 for thin-layer chromatography (TLC) was purchased from Qingdao Marine Chemical Ltd. (Qingdao, China). Column chromatography (CC) was performed on reversed-phase (RP) C 18 silica gel (Merck, Darmstadt, Germany) CC and Sephadex LH-20 (GE Amersham Biosciences, Piscataway, NJ, USA) stationary phases. High-performance liquid chromatography (HPLC) was performed on an Agilent 1200. Semi-preparative HPLC was performed on a Waters 1525 Binary HPLC Pump (Agilent Eclipse XDB-C 18 , 5 µm, 9.4 × 250 mm) and a Waters 996 Photodiode Array Detector. Compounds were visualized under UV light and by Iodine vapor. Optical rotations were measured on an Auton Paar MCP200 Automatic Polarimeter. IR spectra (KBr) were obtained on a Nicolet 6700 FT-IR spectrometer.

Extraction and Isolation
The 15 L culture was diced and extracted overnight with EtOAc/MeOH (4:1, v/v) at room temperature three times. The crude extract was partitioned between H 2 O and EtOAc (1:1, v/v) until the H 2 O layer was colorless. The EtOAc extract was partitioned between 95% aqueous MeOH and petroleum ether (PE) to afford the defatted MeOH extract. The MeOH extract was fractionated by medium-pressure liquid chromatography (MPLC) over RP C 18 silica gel (130 g) eluted with gradient aqueous CH 3 CN (30%, 50%, 70% and 100% CH 3 CN, 500 mL each) to give six fractions (Fr.), A-F.
The microbroth dilution method [27] was applied to determine the MIC value of active compounds against the growth of Staphylococcus aureus ATCC 25923. Kanamycin was used as a positive control. Microorganisms were cultured in LB (tryptone 10 g, yeast extract 5 g, NaCl 10 g, ddH 2 O 1000 mL, pH 7.2) media in 96-well plates at a concentration of 1 × 10 6 CFU/mL, and the MIC values were obtained after incubating for 12 h at 37 • C with the tested compounds.

Cytotoxicity Assay
The in vitro antiproliferative activity against KG1 cells was measured: the cells were purchased from Cell Bank of the Institute of Biochemistry and Cell Biology, China Academy of Sciences (Shanghai, China), and cultured in RPMI 1640 (Roswell Park Memorial Institute 1640) media with 10% fetal bovine serum (Biological Industries), incubating at 37 • C in a humidified atmosphere containing 5% CO 2 . Cell-grown inhibition was determined using the Cell Counting Kit-8 (CCK-8) (Bimake, Houston, TX, USA) according to the manufacturer's instructions [28]. Briefly, cells were seeded in 96-well plates at 7 × 10 3 cells/well and treated with different concentrations of compounds 1-18 for the indicated 48 h. Cytosporone B (VP16) was used as the positive control. Then, 10 µL CCK-8 was added to each well and incubated for another 4 h. The absorbance was read at 480 nm by Spark 30086376 (TECAN, Austria). Growth inhibition (%) was calculated at each concentration and the IC 50 was calculated by software Prism 7 (GraphPad Software, Inc., San Diego, CA, USA).

Anti-Type III Secretion System (T3SS) Assay
Salmonella enterica is the major cause of foodborne illness and typhoid fever [29,30], and uses a type III secretion system (T3SS) to translocate the virulence factors into host cells [31]. These virulence factors include specific effector proteins encoded by the S. enterica pathogenicity island 1 (SPI-1) [32]. T3SSs are highly conserved among Gram-negative bacteria [33]. Compounds 1-18 were assayed for their anti-T3SS activity of S. enterica Typhimurium UK-1 χ8956 in vitro, as previously described in our laboratory [34][35][36]. S. enterica Typhimurium UK-1 χ8956 was cultured in the LB media (tryptone 10 g, yeast extract 5 g, NaCl 10 g, ddH 2 O 1000 mL, pH 7.2) supplemented with 0.2% L-arabinose at 37 • C in the presence of a solvent control or the tested compounds at the final concentration of 100 µM [37], respectively. Cytosporone B (Csn-B) was used as the positive control [38]. Salmonella enterica Typhimurium UK-1 χ8956 was a gift from Roy Curtiss III (School of Life Sciences, Arizona State University) [37].

Results
The A. mediterranei S699 ∆rifT mutant shows different morphological characteristics compared with that of the wild-type strain. After analysis of the fermentation products cultivated for 7 days on ISP2 agar media at 28 • C by HPLC and LC-ESI-HRMS (liquid chromatography-electrospray ionization-high-resolution mass spectrometry), the ∆rifT mutant exhibited a completely different metabolic profile from that of the wild-type strain (Figure 2). In addition, a rifT gene complementation plasmid was constructed (Supplementary Figure S5) and introduced into the ∆rifT mutant to get the complementation mutant ∆rifT::rifT (Supplementary Figure S6). HPLC analysis indicated that the metabolites of the ∆rifT::rifT strain were almost identical to that of the wild-type one (Supplementary Figure S7), which definitely eliminated the polar effect caused by genetic manipulation.  Table S1 and green in Figure 3). The presence of deoxyhexapyranose moiety (Figure 3 The deoxyhexapyranose moiety attached to the amide nitrogen of ansamycin was determined based on the HMBC correlations from H-1′ to C-2 (δC 136.9). In order to determine its stereochemistry, the sugar was purified from the spontaneous hydrolysis products of 1 ( Supplementary Figures S15 and S16). The 1 H NMR spectroscopic data of the sugar was completely consistent with that of authentic α-L-rhamnose (Supplementary Figure  S17 Figure S18). The NMR data (Supplementary Figures S19-S23)   In order to explore the products accumulated by the A. mediterranei S699 ∆rifT mutant, 15 L fermentation was carried out and the fermented agar cakes were diced and extracted. The extract was subjected to column chromatography over Sephadex LH-20, MPLC over RP C 18 silica gel, and finally, HPLC, to yield compounds 1-18.  Figure S9). The 1 H and 13 C NMR spectroscopic data (Tables 1 and 2) (Supplementary Figures S10-S14) indicated that 1 had a structural skeleton of rifamycin, but the signals for a deoxyhexapyranose were also clearly observed. The presence of a naphthaquinone chromophore was indicated by the HMBC correlations from H-3 (δ H 7.30) to C-9 (δ C 137.1) and C-10 (δ C 122.6), and from H-5 (δ H 7.34) to C-4 (δ C 147.7), C-6 (δ C 153.4), C-7 (δ C 127.4) and C-8 (δ C 123.1), as well as from H-8 (δ H 7.68) to C-6 (δ C 153.4), C-8 (δ C 123.1), C-9 (δ C 137.1) and C-13 (δ C 12.5) (Supplementary Table S1 and Figure 3).  Table S1 and green in Figure 3). The presence of deoxyhexapyranose moiety (Figure 3, orange), was determined based on the 1 H-1 H COSY correlations of H-1 (δ H 4.82) with H-2 (δ H 4.09), H-5 (δ H 3.96) with H-4 (δ H 3.33) and H-6 (δ H 1.19), along with the HMBC correction from H-1 to C-2 (δ C 70.8), from H-4 to C-2 (δ C 70.8) and C-6 (δ C 18.0), and from H-6 to C-1 (δ C 105.1) and C-2 (δ C 70.8). The deoxyhexapyranose moiety attached to the amide nitrogen of ansamycin was determined based on the HMBC correlations from H-1 to C-2 (δ C 136.9). In order to determine its stereochemistry, the sugar was purified from the spontaneous hydrolysis products of 1 (Supplementary Figures S15 and S16). The 1 H NMR spectroscopic data of the sugar was completely consistent with that of authentic α-L-rhamnose (Supplementary Figure S17) Table  S1 and Figure 3). The stereochemistry of C-20 to C-28 was assumed to be same as that of protorifamycin I [39] on the basis of biosynthetic logic. Thus, compound 1 was determined to be N-α-L-rhamnosyl proansamycin B-M1, a novel rifamycin amide N-rhamnoside, named rifamycinoside C.
The molecular formula of compound 5 was elucidated as C37H47NO11 (Supplementary Figure  S36). A close NMR comparison with that of protorifamycin I [39] revealed that the evident difference was one or more acetyl signals coupling with H-34a, indicating the acetylation of H-34a in 5, which was confirmed by the HMBC correlations of H-34a (δH 4.01, 4.00) with C-35 (δC 173.0) and H-36 (δH 2.03) with C-35 (Supplementary Table S5 and Figures S37-S41). Thus, 5 was elucidated as 34a-acetylprotorifamycin I.
The molecular formula of 3 was determined to be C 34 Figure S24). The NMR spectra of 3 ( Supplementary Figures S25-S29) were similar to those of 2, except that the ansa chain was suggested to connect to C-5 of the naphthoquinoid via an ester bond between C-11 and C-28 on the basis of the HMBC correlations from H-28 (δ H 3.96) with C-12 (δ C 211.0), from H-29 (δ H 2.76, 2.65) with C-12 and C-28 (δ C 81.5) as well as the remaining degrees of unsaturation and the molecular formula (Supplementary Table S3 and Figure 3). Based on the NMR data comparison with those of rifamycinoside A [41], 3 was most likely the aglycone moiety of rifamycinoside A, both of them occurred at decarboxylation of C-34a and C-11/12 cleavage of ansa chain. Thus, compound 3 was determined to be 11,12-seco-28-desmethyl-28-hydroxyprotorifamycin I 11-carboxy-28-ester.  Table S4) demonstrated that the structure of 4 was similar to protorifamycin I, except that hydroxylation of Me-30 and Me-34a oxidized to an aldehyde group and formation of hemiacetal with a C-25 hydroxyl group, which was supported by 1 H NMR of H-30 (δ H 4.46, 4.24), H-34a (δ H 5.12) and HMBC correlation of H-34a and C-25 (δ C 73.7) (Tables 1 and 3) (Supplementary Figures S31-S35). Thus, 4 was determined as a new rifamycin hemiacetal derivative, named 30-hydroxy-protorifamycin I-hemiacetal.
The molecular formula of compound 5 was elucidated as C 37 H 47 NO 11 (Supplementary Figure  S36). A close NMR comparison with that of protorifamycin I [39] revealed that the evident difference was one or more acetyl signals coupling with H-34a, indicating the acetylation of H-34a in 5, which was confirmed by the HMBC correlations of H-34a (δ H 4.01, 4.00) with C-35 (δ C 173.0) and H-36 (δ H 2.03) with C-35 (Supplementary Table S5 and Figures S37-S41). Thus, 5 was elucidated as 34a-acetyl-protorifamycin I.
Compound 6 was confirmed to have the molecular formula of C 35 Figure S42), the same as that of protorifamycin I. The down-field chemical shifts of C-31 (δ H 3.52, 3.53, δ C 63.9) and up-field chemical shifts of C-34a (δ H 1.06, δ C 20.0) revealed the hydroxylation of C-31 (Tables 1 and 2) (Supplementary  Table S6 and Supplementary Figures S43-S47). Thus, the structure of compound 6 was determined and named as 31-hydroxyproansamycin B.

Discussion
The biosynthesis of rifamycins has been extensively studied ever since the discovery of its biosynthetic gene cluster, and it can be divided into three stages: the first stage is the synthesis of the starting unit AHBA (3-amino-5-hydroxybenzoic acid) [45,46], the second stage is the extension of rifamycin polyketide [16,20,21] and the third stage is the rifamycin post-PKS modification [17]. The first two stages have been clearly studied, however, the formation process from the putative proansamycin X to rifamycin W, an important intermediate, is still unclear. According to previous research, there may be a C7/C8 dehydrogenation reaction in this progress. To investigate the dehydrogenation of proansamycin X, the mutant strain Amycolatopsis mediterranei S699 ∆rifT was constructed by deleting the rifT gene (putative NADH-dependent dehydrogenase gene).
The structures revealed that all eighteen compounds isolated from the ∆rifT mutant strain had undergone deoxygenation at C-8. However, we could not successfully obtain proansamycin X, which was possibly due to its instability of a hydroxyl group at C-8 within the conjugated system from C-1 to C-10. When the putative rifT gene-dependent dehydrogenation in rifamycin B biosynthetic route ceased, accumulated proansamycin X tended to undergo dehydration at C-7/C-8 to form a stable naphthalene ring and transformed to proansamycin B (Figure 4a), which can be subjected to sequent ansa polyketide chain post-PKS modifications to produce a series of 8-deoxy-rifamycin derivatives. Moreover, the metabolites of the complementation mutant ∆rifT::rifT were identical to that of the wild-type strain (Supplementary Figure S7), indicating that the rifT gene was involved in the biosynthesis of rifamycins.
Biomolecules 2020, 10, x 12 of 15 Moreover, the metabolites of the complementation mutant ΔrifT::rifT were identical to that of the wild-type strain (Supplementary Figure S7), indicating that the rifT gene was involved in the biosynthesis of rifamycins. The biosynthesis pathway of 8-deoxy-rifamycins demonstrated diverse cleavage patterns of ansa polyketide backbone, including 5,11 retro-Claisen cleavage, just like that observed in ansa biosynthesis of divergolides R and S [47], hygrocins I and J [48] and microansamycins G-I [49], which lead to protorifamycin I-M1 and proansamycin B-M1 (15) (Figure 4a), 12,19 double-bond cleavage and skeleton rearrangement lead to 8-deoxy-rifamycin B (18) (Figure 4b) and a novel 11,12-cleavage carried out by a typical Baeyer-Villiger oxidation and intramolecular transesterification formed 2 and 3 (Figure 4c) [41]. Compounds 5, 6, 7, 8, 9, 10 and 11 oxygenated at C-20, C-23, C-30 and C-31 suggested that the ansa chain is prone to be oxidized in the ΔrifT strain during fermentation. In addition, the oxidation process of C-34a from compounds 4 and 14 to compounds 2 and 3 represented that the oxidation of C-34a alcohol to the carboxyl group may occur before the 12,29-olefinic bond and 11,12-oxygen insertion cleavage.

Conclusions
In summary, the results of in vivo gene inactivation and complementation indicated that the rifT gene is involved in the biosynthesis of rifamycins, and the 8-deoxy-rifamycin proansamycin B could undergo post-PKS modifications similar to that of its 8-hydroxyl analogue 34a-deoxyrifamycin W. The biosynthesis pathway of 8-deoxy-rifamycins demonstrated diverse cleavage patterns of ansa polyketide backbone, including 5,11 retro-Claisen cleavage, just like that observed in ansa biosynthesis of divergolides R and S [47], hygrocins I and J [48] and microansamycins G-I [49], which lead to protorifamycin I-M1 and proansamycin B-M1 (15) (Figure 4a), 12,19 double-bond cleavage and skeleton rearrangement lead to 8-deoxy-rifamycin B (18) (Figure 4b) and a novel 11,12-cleavage carried out by a typical Baeyer-Villiger oxidation and intramolecular transesterification formed 2 and 3 (Figure 4c) [41]. Compounds 5, 6, 7, 8, 9, 10 and 11 oxygenated at C-20, C-23, C-30 and C-31 suggested that the ansa chain is prone to be oxidized in the ∆rifT strain during fermentation. In addition, the oxidation process of C-34a from compounds 4 and 14 to compounds 2 and 3 represented that the oxidation of C-34a alcohol to the carboxyl group may occur before the 12,29-olefinic bond and 11,12-oxygen insertion cleavage.

Conclusions
In summary, the results of in vivo gene inactivation and complementation indicated that the rifT gene is involved in the biosynthesis of rifamycins, and the 8-deoxy-rifamycin proansamycin B could undergo post-PKS modifications similar to that of its 8-hydroxyl analogue 34a-deoxyrifamycin W. Accordingly, eleven new derivatives of 8-deoxy-rifamycin were isolated and characterized, including a novel amide N-glycoside of seco-rifamycin 1, 2 and 3, which featured the third ansa chain cleavage pattern of rifamycins [41]. Compounds 2, 3, 5, 6, 13 and 15 exhibited antibacterial activity against Staphylococcus aureus . Compounds 14, 15, 16, 17 and 18 showed potent antiproliferative activity against KG1 cells, respectively.