Site-Specific Modification of the Anticancer and Antituberculosis Polyether Salinomycin by Biosynthetic Engineering

The complex bis-spiroacetal polyether ionophore salinomycin has been identified as a uniquely selective agent against cancer stem cells and is also strikingly effective in an animal model of latent tuberculosis. The basis for these important activities is unknown. We show here that deletion of the salE gene abolishes salinomycin production and yields two new analogues, in both of which the C18=C19 cis double bond is replaced by a hydroxy group stereospecifically located at C19, but which differ from each other in the configuration of the bis-spiroacetal. These results identify SalE as a novel dehydratase and demonstrate that biosynthetic engineering can be used to redirect the reaction cascade of oxidative cyclization to yield new salinomycin analogues for use in mechanism-of-action studies.


General analytical procedures
NMR data were collected using Bruker Avance spectrometers using either a 500 DCH cryoprobe operating at 500.05 MHz for 1 H and 125.7 MHz for 13 C ( 13 C, DEPT, TOCSY) or a 500 TCI cryoprobe operating at 500.13 MHz for 1 H and 125.8 MHz for 13 C ( 1 H, DQF-COSY, HSQC, HSQC-TOCSY, HMBC, NOESY) in the Chemistry Department, University of Cambridge.
Chemical shifts were recorded using an internal deuterium lock for 13 C and residual 1 H in CD 3 CN (δ H 1.94, δ C 118.26) or CD 3 OD (δ H 3.31, δ C 49.00), and are given in ppm on a scale relative to δ TMS = 0. NMR spectra were processed using Bruker Topspin (v. 3.2).
DQF-COSY spectra were acquired with 2k data points in F 2 and 360 increments with 2 scans per increment. TOCSY spectra were acquired using DIPSI2 modulation and a mixing time of 120 ms, 8k data points were acquired in F 2 and 360 increments with 2 scans per increment. The edited HSQC spectra were optimized for 145 Hz with 2k data points in F 2 , 256 increments and 2 scans per increment. HMBC spectra optimized for 8 Hz with a three-fold low pass J filter to suppress one bond couplings, 4k data points were acquired in F 2 with 360 increments and 16 scans per increment. Edited-HSQC-TOCSY spectra were optimized for 145 Hz using DIPSI2 modulation and a mixing time of 120 ms, 8k data points were acquired in F 2 , 360 increments and 8 scans per increment. NOESY spectra were recorded using a mixing time of 0.5 s, 4k data points in F 2 with 512 increments and 16 scans per increment. All data were zero filled to 1K in F 1 for processing.
HPLC-MS analysis was performed using an HPLC (Hewlett Packard, Agilent Technologies 1100 series) coupled to a Finnigan MAT LCQ mass spectrometer fitted with an electrospray ionization (ESI) source. The HPLC was fitted with a Prodigy 5µ C18 column (4.6 × 250 mm, Phenomenex) column. A solvent system of CH 3 OH and H 2 O both containing 0.1% formic acid (v/v) was used. Samples were eluted with a linear gradient of 85 to 100% of CH 3 OH over 20 min, then 100% CH 3 OH over 10 min at a flow rate 0.7 mL min -1 (method A). Alternatively a linear gradient of 85 to 100% of CH 3 OH over 15 min, then 100% CH 3 OH over 8 min at a flow rate 1 mL min -1 was used (method B). The mass spectrometer was run in positive ionization mode, scanning from m/z 150 to 1800, and the collision energy was set to 35%. Production of salinomycin (1), E15 (2), and E16 (3) metabolites was verified by LC-MS 2 analysis on [M+Na] + ions at m/z 773.5, 791.5 and 791.5 respectively with a normalized collision energy of 35%.
ESI high resolution MS (ESI-HR-MS) was carried out on a Thermo Fisher Orbitrap with 60,000 resolution and normalized collision energy of 15%.
Samples were eluted with a gradient of 85 to 99% MeOH over 40 min at a flow rate of 15 mL min -1 .
DNA sequencing was carried out by the DNA Sequencing Facility in the Department of Biochemistry, University of Cambridge.

Bacterial strains and culture conditions
Streptomyces albus strains were grown in TSBY liquid medium (3% tryptone soy broth, 10.3% sucrose, 0.5% yeast extract) for isolation of genomic DNA, and on SFM solid medium (2% mannitol, 2% soya flour, 2% agar) for conjugation and strain maintenance. For liquid cultures, the strains were grown at 30°C with shaking at 220 rpm in a rotary incubator for 36-44 h. For solid culture, the strains were grown at 30°C for 10-12 days.

Materials, DNA isolation and manipulation
Bacterial strains, plasmids and oligonucleotides (Invitrogen) used in this work are summarised in Tables S1, S2 and S3 respectively. Restriction endonucleases and alkaline phosphatase were purchased from New England Biolabs. T4 DNA ligase was purchased from Fermentas. All chemicals were from Sigma-Aldrich. All organic solvents used were HPLC grade.
Plasmid DNA was isolated from an overnight culture using the Plasmid Mini Kit I (Omega BioTek) according to the manufacturer's protocol.
High-molecular weight PAC DNA was isolated from E. coli overnight culture by alkaline lysis. [1] High molecular weight genomic DNA from Streptomyces strains was isolated using the salting out procedure. [2] Purification of DNA fragments from agarose gels was performed using the Anachem Gel  For analysis of S. coelicolor M1154 metabolite profiles a 35 mL R2YE agar plate was extracted twice with 35 mL of ethyl acetate. The organic phase was evaporated, the residue was dissolved in 1 mL of methanol, the mixture was centrifuged, and 5 µL of the supernatant was analyzed by HPLC-MS.

Purification of S. albus ∆salE metabolites 2 and 3
A 7-day-old 6 L culture broth of S. albus DSM 41398 ∆salE mutant strain was extracted twice with 4 L of ethyl acetate. The combined organic layers were dried over MgSO 4 and the solvent was evaporated, yielding 153 g of an oily residue. The latter was redissolved in 2 L of hexane and extracted three times with 1 L methanol/water (4:1, v/v) mixture. Methanol-water fractions were combined, and methanol was removed by evaporation at reduced pressure. The remaining water layer was extracted three times with 500 mL of ethyl acetate. The combined extracts were dried with anhydrous MgSO 4 and the solvent was removed in vacuo. 8.4 g of oily residue was obtained.
To remove the remaining oil, the sample was further purified by flash chromatography on a column (3 cm × 20 cm), of silica gel, 60 µm particle size. The column was washed with 300 mL of 1:1 (v/v) mixture of hexane/ethyl acetate, and compounds eluted with 2 L of ethyl acetate/methanol (19:1, v/v). Combined fractions contained 1.4 g of a mixture containing compounds 2 and 3.
Further purification was achieved by repeated rounds of preparative HPLC; fractions were collected at 0.9 min intervals. Final fractions were combined and desalted using Chromabond C18 EC column (Macherey-Nagel) yielding around 8 mg of 2 and 4 mg of 3 ( Figure S9). All stages of purification were monitored by direct injection into the Finnigan MAT LCQ mass spectrometer or by HPLC-MS analysis.

Gene disruption in S. albus DSM 41398
In-frame deletion or replacement of the salE gene was carried out as follows. Recombinant plasmids were constructed by ligating DNA fragments (about 2 kb) PCR-amplified from the upstream and downstream flanks of the target gene into vector pYH7, previously digested with NdeI and gel purified. To ligate the fragments the isothermal Gibson assembly method was used as described. [3] The assembly mixture was incubated at 50°C for 60 min, and then was used to transform E. coli DH10B.
To disrupt salE by in-frame deletion, two flanking fragments to be used for homologous recombination were amplified from S. albus genomic DNA (gDNA) by PCR using the following two pairs of primers: SalE_1F, SalE_1R and SalE_2F, SalE_2R. The integrity of recombinant plasmids was checked by restriction digestion and sequencing. To verify the in-frame deletion in the construct and in the mutant by PCR, a pair of primers PCR_salE_f and PCR_salE_r was designed. The approach is schematically depicted in Figure S5.
The construct obtained, pHL∆salE, was introduced by conjugation into S. albus DSM 41398.
The donor strain was E. coli ET12657/pUZ8002, and conjugation was carried out on 25 mL SFM plates. After incubation at 30°C for 20 hours, exconjugants were selected with 5 µg mL -1 apramycin and 25 µg mL -1 nalidixic acid. Exconjugants were transferred to an SFM plate containing 50 µg mL -1 apramycin and 25 µg mL -1 nalidixic acid to double check for antibiotic resistance. Loss of recombinant plasmid pHL∆salE with consequent formation of potential double-crossover mutants was initiated by streaking exconjugants on SFM agar medium for up to 12 rounds of nonselective growth. Single colonies were patched onto both SFM agar and SFM agar containing apramycin (50 µg mL -1 ) in parallel to check for apramycin sensitivity; colonies with the correct phenotype (Apr S ) were selected and further grown in TSBY medium, gDNA was purified. Potential mutants were checked by PCR and Southern blot analysis ( Figure S6).

Southern hybridisation
7 µg of genomic DNA from S. albus WT and S. albus ΔsalE strains was digested for four hours with PvuII. After separation by agarose gel electrophoresis, PvuII-digested S. albus WT and ΔsalE gDNA were probed, using the Roche DIG system according to the manufacturer's instructions, with a DIG-labelled 3654 bp fragment recovered from pHLΔsalE plasmid by digestion with NdeI. pHLΔsalE plasmid digested with NdeI used as a reference in gel electrophoresis alongside S. albus WT and ΔsalE gDNA.

Construction of a plasmid for complementation
The complementation plasmid pIB-salE was constructed based upon integrative vector pIB139 placing salE under the ermE * promoter. The PCR product (primers salE_NdeI_F and salE_EcoRV_R) and the pIB139 vector were digested with NdeI and EcoRV and the target fragments were recovered from a 0.7% agarose gel. The ligation of the digested and purified PCR product with the vector (treated with alkaline phosphatase after restriction step) was performed using T4 DNA ligase followed by transformation into DH10B competent cells using heat shock at 42°C for 55 sec. Cells were transferred onto LB plates containing 50 µg mL -1 apramycin. After incubation overnight, transformants were picked and inoculated into 10 mL LB broth containing isolated and their identity confirmed by restriction analysis and DNA sequencing ( Figure S7). pIB-salE was integrated into the S. albus ΔsalE mutant by conjugation as described above.

Heterologous expression of salinomycin in S. coelicolor M1154
Heterologous production of salinomycin was carried out by expression of the entire sal The screening of the library was performed as follows. Individual PAC clones were grown in 96 deep-well plates at 37°C, 300 rpm overnight. For each row (8 wells), the cultures (800 µL) were pooled into a 15-ml plastic centrifuge tube (Greiner) and centrifuged (4,600 × g, 10 min, 4°C).
After DNA purification by alkaline lysis PCR reactions with primer pairs (salPAC cen_F, salPAC cen_R) complementary to the centre region of the salinomycin biosynthetic cluster were performed.
Those samples that gave a band of the correct size were subjected to another round of PCR -with 2 primer pairs, complementary to the very beginning (salPAC beg_F, salPAC beg_R) and very end (salPAC end_F, salPAC end_R) regions of the cluster. Clones corresponding to the positive hits were grown again, DNA was isolated from each individual clone and subjected to PCR analysis with beginning and end primer pairs together. One out of the positive clones -PAC-sal -was chosen for further sequencing analysis to define exact insert boundaries and for heterologous expression experiments. The sequencing results showed that the exact insert size is 136,770 bp ( Figure S19).

E. coli triparental mating and conjugation of S. coelicolor M1154
This is a two-step protocol for transferring individual PAC clones (derivatives of pESAC13 vector) into the S. coelicolor strain [4] (Figure S18a).
E. coli cells ET12567 (Cam R ), TOPO10/pR9604 (Carb R ), DH10B/PAC (Kan R ) were inoculated into 5 mL LB medium containing appropriate antibiotic and incubated overnight at 37°C, 250 rpm. From the overnight culture 500 µL was inoculated into 10 mL LB medium containing half of the working concentration of appropriate antibiotic and incubated at 37°C, 250 rpm until A 600 reached 0.4. The cells were harvested by centrifugation at 2,200 × g for 5 min and washed twice with 20 mL of LB medium. The supernatant was discarded and the pellet was resuspended in 500 µL of LB medium. 20 µL of each strain was dripped onto the same location on the LB agar plate lacking antibiotics so that the three strains were mixed together. After drying the plates were incubated at 37°C overnight for tri-parental conjugation. Next day, to select for E. coli ET12567 derivatives containing the PAC clone and the helper plasmid (pR9604) the cells from the spot were streaked onto fresh LB agar plates containing kanamycin, chloramphenicol and carbanicillin antibiotics and incubated at 37°C overnight. Single colonies were used to inoculate LB medium containing the antibiotics. PCR analysis was carried out to confirm the presence of the PAC clone in ET12567 cells.
The standard conjugation protocol was followed. [2] Mixtures of Streptomyces and E. coli were plated on SFM agar plates, and overlaid after 20 h with thiostrepton (50 µg mL -1 ) and nalidixic acid (25 µg mL -1 ). After observing putative exconjugants they were streaked onto an SFM plate containing thiostrepton (50 µg mL -1 ) and nalidixic acid (25 µg mL -1 ). Thiostrepton resistant colonies were grown in TSBY medium for genomic DNA purification and PCR analysis to confirm that the entire PAC clone has been transferred to the S. coelicolor recipient ( Figure S20).
Integration of an empty pESAC13 vector was done in parallel as a negative control.

salE gene inactivation in E. coli
To inactivate the salE gene in E. coli the λ-RED recombination approach was used ( Figure   S22).
The vector pIJ790 was transformed into E. coli DH10B with PAC clone PAC-sal and the cells were grown under chloramphenicol and kanamycin selection at 30°C. A single colony was grown overnight at 30°C in 5 mL of LB medium supplemented with the same antibiotics. 10 mL of fresh LB medium containing chloramphenicol and kanamycin was inoculated with 100 µL of overnight culture and grown at 30°C to OD 600 reached 0.3. After 0.1% of L-arabinose was added to induce expression of exo, bet, gam and the cells were grown at 37°C for 45 min. Cultures then were centrifuged and electrocompetent E. coli cells (DH10B/PAC) were prepared as described by Dower. [5] Linear DNA fragments containing acc(3)IV gene PCR were amplified from plasmid pIJ773 using primer pairs salE_to_Apr cas_PAC_F and salE_to_Apr cas_PAC_R to knockout salE gene.
E. coli DH10B/PAC were electroporated with 100 ng of the linear fragment. For electroporation, the mixture of E. coli cells with DNA was immediately transferred into an ice-cold 2 mm electroporation cuvette and electroporated at 2.5 kV (25 µF, 200 Ω, t const ~ 5 ms) using a Bio-Rad Gene Pulser II. LB medium (750 µL) was added and the cells were incubated at 37°C, 250 rpm for an hour. The culture (100 µL) was streaked onto LB plate containing apramycin and kanamycin and incubated overnight at 37°C. Verification of positive transformants was performed using primers PCR salE_f and PCR salE_r ( Figure S23).

NMR Analysis
Careful analysis of the edited HSQC, TOCSY, HSQC-TOCSY and COSY spectra identified 10 individual spin systems for both compounds 2 and 3 and enabled the assignment of the proton and carbon NMR spectra. COSY and HMBC correlations then enabled the 2D structure to be determined as shown in Figure S30 where the C18/C19 double bond of salinomycin had been hydrated as proposed.

Configuration of the bis-spiroacetal region in compounds 2 and 3
NOESY and 1 H-1 H coupling constant data was used to determine the stereochemistry of the newly formed hydroxyl centre at C19 and the configuration of the bis-spiroacetal region.
For compound 3 H15b appeared as an apparent quartet with a 3 J H-H of 13Hz. This implied a diaxial relationship to both H14 and H16 and suggested that the B ring should be in a chair conformation with the substituents on C13, C14 and C16 equatorial. An NOE between H13 and H15b supported this conformation ( Figure S31). A further NOE correlation between H20/H22a suggested that C22 should be equatorial in relation to the C ring and thus confirmed the configuration of C21.
For compound 2 a similar analysis of the B ring led to conclusion that it was again in a chair conformation with an axial orientation of the C13, C14 and C16 substituents. In this case however NOE's were observed between H16/H18b and H34/H18a suggesting that C18 was in an equatorial position relative to the B ring and that C17 had the opposite configuration to 3 and 1 (salinomycin sodium salt, purchased from Sigma Aldrich). It was also observed that the 13 C chemical shift for C18 changes from 26.00 in 3 to 40.10 in 2.
Analysis of the C ring was initially hindered by coalescence of the H18a and H18b signals, this problem was solved by switching the solvent to methanol. A series of 6 -8 Hz couplings in the C18-C20 region immediately indicated that this ring did not adopt a chair conformation. A twist boat type structure satisfied the observed NOE correlations; in particular a strong NOE between H13/H20 that could not be accounted for by any other conformation. This correlation also served to place the C20 hydroxyl group on the top face of the ring in an R configuration the same as that in 2 and 1. H19 exhibited a strong NOE to H18b and a weak one to H18a suggesting that the C19 hydroxyl was also R configured. This agreed with the observed coupling constants in this region all of which indicate torsion angles of around 30˚ or 150˚, see Figure S32. The possibility of the C19 hydroxyl being S configured cannot be entirely ruled out as no NOE was observed between H19/H20. This configuration however would be expected to lead to a large (8-12 Hz) and a small (0-4 Hz) coupling constant for H19-H18a/b which does not fit with the observed 8 Hz coupling observed. The absence of the H19-H20 NOE can possibly be rationalised by the proximity of the H19 and H20 signals meaning that the NOE cross peak could be obscured by the diagonal. A degree of flexibility in this region is also possible and the observed coupling constants could reflect the existence of multiple conformers although this is thought to be unlikely.
As with 3 an NOE correlation between H20/H22a and H20/H22b placed C22 on the same face of the C ring as H20 and confirmed the configuration at C21 as being the same as in 1. A further NOE between H33 and H34 confirmed the configuration of C24 in 2, again the same as that in 1.

Stereochemistry of the A ring
Analysis of this region was hampered somewhat by the convergence of the H4a and H5a protons meaning NOE's to these protons could be due to either or both. It is proposed that both are oriented axially and the NOE's observed are due to those shown in Figure S33. A series of NOE correlations between H2/H5a/H7 served to place these substituents in axial orientations on the bottom face of the ring. This implies a trans relationship across the tetrahydropyran oxygen between the C3 and C7 substituents. NOE's between H6/H7 and H4a/H40 along with the lack of an NOE between H7/H40 suggested that the C6 methyl group was axially oriented and thus the configuration of this ring is indeed the same as for 1. Analysis of the coupling constants that could be determined suggested that the 3 J H3-H2 coupling constant was 11.0 Hz in agreement with the conformation shown in Figure S33 where H2 is oriented into the ring and is hence anti to H3. H3 appears as a broadened doublet of doublets (J = 11.0, 4.4 Hz), so the H3-H4 coupling constants must both be small, again consistent with the equatorial orientation of H3.
H7 showed two coupling constants (10.1, 2.0 Hz), and the H6-H7 coupling was assigned as 2.0 Hz based upon the NOE's shown in Figure S33 and the appearance of H6 in rows extracted from TOCSY and NOESY spectra, a narrow multiplet with no large couplings apparent.
Comparison with the NOE's and coupling constants observed in the spectra for a sample of authentic 1 showed good agreement with the observed data for 2 and 3 supporting the conclusion that the configuration of the A ring is unchanged.
The C2 configuration is also assumed not to have changed from the natural product. The H2-H3 coupling constant is almost identical for 1 and both 2 and 3. This would be expected to change should the C2 configuration be different, as the conformation about this bond would be different in order for the C1 carboxylic acid to participate in intramolecular hydrogen bonding.  Table S9).

Stereochemistry of the C7-C13 region
Comparison of the NOE correlations and coupling constants in this region with an authentic sample of 1 helped determine that the stereochemistry of 2 and 3 was unchanged with respect to the natural product.
Conformations about each bond were determined based on the observed coupling constants and NOE correlations as shown in Figure S35 and were compared to an authentic sample of 1.
While the data do not rule out the existence of other conformers or configurations the good match for the data observed for 1, particularly the coupling constant data, suggests that the configuration in this region is the same.
The strong hydrogen bonding network present in 1 is the dominant factor in determining its conformation. As a result if the configuration of one of the centres were to have changed the conformation of that region is expected to change in order to maintain the hydrogen bonding network, and this would result in a significant change in the coupling constants and NOE's observed in this region.  Table S10).     All data recoreded in CD 3 CN. a 1 H data reported as δ H (ppm), multiplicity (coupling constants in Hz), δ H values for signals listed as multiplets were determined from the HSQC spectrum. b (w) denotes a weak coupling. c Protons 4a and 5a overlap and cross peaks are potentially due to either or both protons.