Biosynthesis of the Enterotoxic Pyrrolobenzodiazepine Natural Product Tilivalline

Abstract The nonribosomal enterotoxin tilivalline was the first naturally occurring pyrrolobenzodiazepine to be linked to disease in the human intestine. Since the producing organism Klebsiella oxytoca is part of the intestinal microbiota and the pyrrolobenzodiazepine causes the pathogenesis of colitis it is important to understand the biosynthesis and regulation of tilivalline activity. Here we report the biosynthesis of tilivalline and show that this nonribosomal peptide assembly pathway initially generates tilimycin, a simple pyrrolobenzodiazepine with cytotoxic properties. Tilivalline results from the non‐enzymatic spontaneous reaction of tilimycin with biogenetically generated indole. Through a chemical total synthesis of tilimycin we could corroborate the predictions made about the biosynthesis. Production of two cytotoxic pyrrolobenzodiazepines with distinct functionalities by human gut resident Klebsiella oxytoca has important implications for intestinal disease.


S4
For mutagenesis, competent K. oxytoca cells containing pKOBEG were induced with 0.2 % arabinose and transformed with 100 ng of PCR-amplified linear recombination fragment after DpnI digestion. Transformants were selected with 8 µg/mL tetracycline or 50 µg/mL kanamycin and incubated at 32 °C overnight, then cured of the temperature-sensitive pKOBEG by growth at 42 °C. Successful disruption of the bacterial gene was verified via PCR and DNA sequencing.
Our previous study showed that aroX, npsA, thdA and npsB are essential for tilivalline biosynthesis. [4] In this study the remaining genes dhbX, icmX, adsX and hmoX and the chromosomal tryptophanase gene tnaA were inactivated to elucidate their role in tilivalline biosynthesis ( Figure S1 ΔadsX: Both 500 bp recombination flanks were amplified from the genome (HindIII-adsX_upfw and SacI-adsX_up-rev and KpnI-adsX_dw-fw plus SalI-adsX_dw-rev) and inserted up-or downstream of tetA/R and a rrnB terminator sequence.
ΔΔhmoX: For the PAI hmoX homology arms were amplified from the genome (SacI-hmoX_up-fw and HindIII-hmoX_up-rev and SalI-hmoX_dw-fw plus Kpn-hmoX_dw-rev) and inserted up-or downstream of tetA/R and a rrnB terminator sequence. For the second hmoX homologue the 500 bp recombination flanks were amplified (SacI-hmoXg_up-fw and HindIII-hmoXg_up-rev and SalI-hmoXg_dw-fw plus Kpn-hmoXg_dw-rev) and inserted up-or downstream of aphA and an rrnB terminator sequence. S6

SI-2. Isolation and characterization of metabolites of K.oxytoca AHC-6 strains General Information
Analytical thin layer chromatography (TLC) was carried out on Merck TLC silica gel aluminum sheets (silica gel 60, F 254 , 20 x 20 cm) and spots were visualized by UV light (λ = 254 nm and/or λ = 366 nm) and by staining with cerium ammonium molybdate solution (50 g (NH 4 ) 6 Mo 7 O 24 were dissolved in 400 mL H 2 O and 50 mL conc. H 2 SO 4 was added followed by 2.0 g Ce(SO 4 ) 2 ) and developed by heating with a heat gun.
Column chromatography was performed on silica gel 60 from Acros Organics with particle sizes 35-70 µm. A 30-to 100-fold excess of silica gel was used with respect to the mass of dry crude product, depending on the separation problem. The crude material was dissolved in MeOH and subsequently adsorbed on the 2.5-fold excess of Celite™. Afterwards the solvent was removed in vacuum and the adsorbed crude material was dried in oil pump vacuum. The dimension of the column was adjusted to the required amount of silica gel and formed a pad between 25 cm and 40 cm. In general, the silica gel was mixed with the eluent and charged into the column before equilibration. For smaller column dimensions the dry silica gel was filled into the column and was equilibrated by forcing an appropriate amount of eluent through by over-pressure. Subsequently, the dissolved or adsorbed crude material was loaded onto the top of the silica gel and the mobile phase was forced through the column by pressure exerted by a rubber bulb pump. The volume of each collected fraction was adjusted between 20 % and 30 % of the silica gel volume, according to the separation problem. Purification via preparative RP-HPLC of tilimycin (2), culdesacin (3) and 9-deoxy-tilimycin (4) was performed on a Thermo Scientific Dionex UltiMate 3000 system with UltiMate 3000 pump, UltiMate 3000 autosampler, UltiMate 3000 column compartment, UltiMate 3000 diode array detector (deuterium lamp, λ = 190-380 nm) and a UltiMate 3000 automatic fraction collector. The components were separated on a RP Macherey-Nagel 125/21 Nucleodur ® 100-5 C18ec column (21 × 125 mm, 5.0 µm). Signals were detected at 210 nm and 254 nm. As mobile phase acetonitrile (VWR HiPerSolv, HPLC grade) and water (Barnstead NANOpure ® , ultrapure water system) were used. The following method was used

Isolation of cis/trans-tilimycin (2) from conditioned medium
K. oxytoca AHC-6 ΔtnaA was cultivated in 100 mL CASO medium in 300 mL Erlenmeyer flask (40x flasks were prepared in parallel) at 37 °C with orbital shaking at 180 rpm for 16 h. A total volume of 4 L culture broth was centrifuged at 8000 rpm for 30 min at 4 °C. The supernatant was filtered through a 0.45 µm cellulose-acetate filter followed by a 0.2 µm cellulose-acetate filter to remove residual bacterial cells. Metabolites were extracted from the conditioned medium with n-butanol (≥99.5%, for synthesis, Roth) applying a 3:2 mixing S8 ratio (two times). The organic extract was concentrated to dryness under high vacuum at 40 °C and was adsorbed on Celite™ and then purified via flash column chromatography (150 g silica gel, size: 420 x 30 mm, CH 2 Cl 2 /MeOH = 20:1 (v/v)). The obtained crude product (109 mg, colorless solid) was further purified by preparative RP-HPLC (Method A), followed by a second flash column chromatography (10 g silica gel, size: 210 x 11 mm, CH 2 Cl 2 /MeOH = 20:1 (v/v)).

NMR-data of cis/trans-tilimycin (2)
Natural tilimycin (2) was obtained as an inseparable ~1:1 mixture of the cis-and transisomer. The assignment of the chemical shifts to the atoms in the mixture was made by 1D and 2D experiments (HSQC, HMBC) (Full spectra . Differentiation between the cis-and the trans-isomer of tilimycin (2) was obtained by NMR analysis of the J-coupling between H-8 to H-9. Only the trans-isomer gives a doublet (8.8 Hz) due to coupling between H-8 and H-9, whereas the protons in the cis-isomer show no coupling ( Figure S2). Additional confirmation about the cis/trans geometry came from a 2D NOESY experiment. The F1 traces of protons H8/H8' are shown in Figure S3. The trans-isomer (H8') shows a stronger NOE to H9', while in the cis-isomer (H8) a stronger NOE to H10b is observed.

3-Hydoxy isatoic anhydride
A dry 20 mL Schlenk tube with magnetic stirring bar was charged with 250 mg (1.63 mmol, 1.0 eq) 3-hydroxyanthranilic acid and 5.0 mL THF abs. in an Ar counter-stream. 165 mg (555 µmol, 0.3 eq) triphosgene were added to the red suspension. The reaction mixture was heated to 45 °C (oil bath) and stirred for 1.5 h until TLC indicated full conversion. The gray suspension was concentrated under reduced pressure and 10 mL n-hexane was added. The precipitate was collected by filtration and dried in vacuum to obtain the crude title compound as a grayish-red powder, which was used without further purification. [7] Yield: 263 mg (1.47 mmol, 90 %), grayish-red powder.  The obtained crude product was purified via flash column chromatography (100 g silica gel, size: 140 x 50 mm, CH 2 Cl 2 /MeOH = 20:1 (v/v)).

NMR-data of culdesacin (3)
Culdesacin (3) was obtained as a single isomer and the signals were assigned to the atoms by 1D and 2D NMR experiments. The configuration of the hydroxyl group 8OH was determined by NOESY experiments (see SI-7).

SI-4. Total Synthesis of tilivalline (1)
Tilivalline (1) used for this study was synthesized according to the following procedure. The transformations from 11 to 1 were adapted from a previous total synthesis by Nagasaka. [8] S20
The solvent was removed in vacuum and the crude product was purified via column chromatography (300 g silica gel, size: 20 x 6 cm, cyclohexane/EtOAc = 3:1 (v/v), 200 mL fractions). [9] Yield: 16.5 g (69.4 mmol, 83%) light-yellow liquid. The spectra are in accordance with previously reported data. [10] Diethyl were dried over Na 2 SO 4 , filtrated and the solvent was removed in vacuum. The brown liquid was dried in oil pump vacuum and was used in the following steps without further purification. [11] Yield: 13.4 g (56.1 mmol, 81%).
Step The spectra are in accordance with previously reported data. [12] S23
The spectra are in accordance with previously reported data. [8]

SI-6. Feeding experiments for confirmation of biosynthetic pathway
In simple CASO medium tilimycin (2) was degraded to culdesacin (3), whereas in ∆npsA supernatant and culture, tilivalline (1) was also formed due to presence of indole produced by the bacteria. The addition of synthetic indole (500 µM) and 2 to CASO medium, ∆npsA supernatant and culture led to similar levels of 1 and 3 for all three conditions. Spiking S35 experiments for 1 or 3 showed that a back reaction to 2 did not occur under biological conditions (culture or supernatant) or in simple medium (see Figure S6).

Epimerization of tilivalline (1) and conversion of tilimycin (2) and indole to tilivalline (1) in vitro
With the in vitro conversion experiments above, we could show the spontaneous conversion of tilimycin (2) and indole to tilivalline (1). Natural and synthetic tilimycin (2) were found to exist as a ~1:1 of trans/cis-isomers. In contrast tilivalline (1) was isolated from bacteria as the trans-isomer. In the published synthesis of tilivalline (1) by Shioiri et al., which used a strategy similar to the biosynthetic pathway proposed in this manuscript to access the PBD-S36 backbone, namely the nucleophilic attack of indole at the imine, also the trans-isomer was isolated as the exclusive product. [14] .
In order to verify the hypothesis that tilivalline (1) is formed spontaneously in a stereoselective way we conducted the following experiment: The spontaneous in vitro conversion of tilimycin (2) and indole was repeated using a 200x excess of indole to increase the conversion rate. Extracts were analyzed with HPLC-MS for presence of tilivalline (1) and epi-tilivalline (1a). The HPLC-reference spectrum was achieved by epimerization of tilivalline (1) according the Matsumoto-protocol [15] and the tilivalline (1)/epi-tilivalline (1a) mixture was measured via HPLC-MS with the previously described method to detect the two distinct peaks for 1 and 1a.

SI-7. Cell culture assay Culturing methods, growth conditions and media for cell cultures
Human cell lines used in this study were HeLa cells (American Type Culture Collection). The cells were cultivated in DMEM medium (4500 mg/L glucose), supplemented with 10 % FBS, 100 μg/mL penicillin and 100 μg/mL streptomycin, at 37 °C with 5 % CO 2 . All media and additives were purchased from Invitrogen.

Cytotoxicity assay of bacterial supernatant and IC 50 determination of metabolites 1-3
A modified MTT cytotoxicity assay applied conditioned medium of K. oxytoca strains to HeLa cells as described previously. [4,16] Variation in cell viability was normalized to CASO medium as control.
For IC 50 measurements HeLa cells were treated with serial dilutions of tilivalline (1), tilimycin represent the mean ± SD of three independent experiments (see Figure S9).