Geldanamycin (GDM) is a 19-membered macrocyclic lactam and is related to ansamycin antibiotics. GDM binds to heat shock protein 90, inhibiting its chaperone activity.1 GDM and its derivatives have the potential to serve as chemotherapeutic agents in cancer and virus-infected diseases.2 A clear understanding of GDM biosynthesis should help us to create novel GDM analogs with optimal properties by genetic manipulation of its biosynthetic pathway.

The biosynthetic gene cluster of GDM, including a set of type-I polyketide synthase (PKS) genes, a set of post-PKS tailoring genes and a separate set of genes for biosynthesis of 3-amino-5-hydroxybenzoic acid, the starter unit for GDM biosynthesis, had been cloned, sequenced and analyzed from several streptomycetes independently.3, 4, 5 The post-PKS tailoring genes, such as gdmN (encoding the carbamoyltransferase for 7-O-carbamoylation), gdmP (encoding the cytochrome P450 monooxygenase for C-4,5 oxidation) and gdmM (encoding the FAD-binding monooxygenase for C-21 hydroxylation), had been disrupted, respectively, in Streptomyces hygroscopicus 17997, S. hygroscopicus subsp. duamyceticus JCM4427 or other GDM producer(s), to confirm their biological functions in GDM biosynthesis.6, 7, 8 As a result, a brief post-PKS tailoring process of GDM biosynthesis was proposed (Figure 1, thick arrows).6

Figure 1
figure 1

The post-polyketide synthase tailoring process of geldanamycin biosynthesis.

The gel8 (identical to gdmN) disruption mutant of S. hygroscopicus subsp. duamyceticus JCM4427 produced 4,5-dihydro-7-O-descarbamoyl-7-hydroxygeldanamycin (CT-1-7) as major component.6 The gdmN disruption mutant of S. hygroscopicus 17997 produced, besides CT-1-7, 4,5-dihydro-7-O-descarbamoyl-7-hydroxy-19-O-glycylgeldanamycin (CT-1-1), which is a novel GDM analog.9

To examine whether there are other undiscovered GDM analogs in the gdmN disruption mutant of S. hygroscopicus 17997, a time-course monitoring of its fermentation culture was conducted. The fermentation supernatants at different times (36, 60, 84, 108 h, and so on) were extracted with equal volumes of ethyl acetate (EtOAc). The organic layers were concentrated, then subjected to silica-gel TLC (developed with a mobile phase of EtOAc/CH2Cl2/hexane/methanol, 9:6:6:1, v/v) and then alkaline color reaction for a preliminary discrimination of GDM analogs.10 A weak purple band appeared above the blue band of CT-1-7, which was supposed to be a novel GDM analog (at 84 h, Figure 2).

Figure 2
figure 2

Silica-gel TLC (GF254) of the ethyl acetate (EtOAc) extract of gdmN disruption mutant of Streptomyces hygroscopicus 17997 at different fermentation times (after the color reaction by 2 mol per liter NaOH). The gdmN disruption mutant of S. hygroscopicus 17997 was grown at 28 °C for 7–10 days on MY agar (yeast extract 0.4%, malt extract 1.0%, glucose 0.4%, agar power 1.5%) plates (apramycin, 50 μg ml−1) for sporulation, then inoculated into the fermentation medium (starch 2%, cotton seed power 0.5%, glucose 0.5%, corn steep liquor 1.0%, yeast powder 0.5%, CaCO3 0.2%) for shaking (200 r.p.m.) at 28 °C for different times, and the culture supernatants were extracted with EtOAc. CT-1-7x appeared at a culture time of 84 h with a production level of about 10 mg l−1.

Compounds corresponding to the weak purple band of alkaline treatment were eluted out with EtOAc from silica-gel TLC. A major peak appeared by HPLC analysis, with its retention time being very close to that of CT-1-7, and its UV spectrum profile similar to that of CT-1-7 and very similar to that of GDM (see Supplementary Figure S1). The peak contained a major compound with a molecular weight of 540.5 Da ([M+Na]+), which is two units less than that of CT-1-7 (542.4 Da, [M+Na]+), as analyzed by liquid chromatography-ESI(+)-MS. Based on these data, we suggested the chemical structure of the major compound as 7-O-descarbamoyl-7-hydroxygeldanamycin (designated as CT-1-7x), with a molecular formula of C28H39O8N1.

CT-1-7x was then purified for structural elucidation. An equal volume of EtOAc was used to extract the fermentation supernatant (about 10 l) of gdmN disruption mutant of S. hygroscopicus 17997 (culture time 84 h), then dried to a crude solid (about 3.042 g) by rotary evaporation at 37 °C. It was then fractionated by silica-gel column chromatography. The pool containing CT-1-7x (eluted by petroleum ether–EtOAc, 75:25, v/v) was dried (30.8 mg) and re-dissolved in methanol for preparative Sephadex LH-20 (ϕ1.8 cm × 150 cm) fractionation, yielding the refined preparation (17.9 mg). The refined preparation was used for preparative HPLC (Shimadzu LC-10ATVp, Shimadzu International Trading (Shanghai) Co., Ltd, Beijing, China; Agilent ZorBax SB-C18, Agilent Technologies Co. Ltd (China), Beijing, China, 5 μm, ϕ9.4 mm × 250 mm, methanol/water, 71:29 (v/v), 1.5 ml min−1), yielding a pure preparation of CT-1-7x (10.7 mg, purity⩾97%). The pure preparation, an amorphous yellow powder, was used for 1H- and 13C-NMR analyses.

The 13C-NMR and DEPT spectra displayed 28 carbon signals, including 7 methyls, 2 methylenes, 11 methines and 8 quaternary carbons. The 13C-NMR of CT-1-7x showed strong similarities to that of CT-1-7 except for C-4,5. The carbon signals of CH2-4 (δC 23.84) and CH2-5 (δC 29.19) in CT-1-7 were absent, while two additional olefinic methyl signals at δC 131.163 and 127.340 emerged in the 13C-NMR of CT-1-7x. The hydrogen signals of CH2-4 (δH 2.24/2.43) and CH2-5 (δH 1.74/1.89) of CT-1-7 disappeared in CT-1-7x, while two additional hydrogen signals of CH- (δH 6.590, 6.075) appeared in the 1H-NMR of CT-1-7x. The two newly appeared olefinic methyl signals (CH-4 and CH-5) in CT-1-7x were also confirmed by the DEPT spectra of CT-1-7x. The NMR chemical shifts were assigned completely for CT-1-7x from HSQC, COSY and HMBC (Table 1). Thus, the chemical structure of CT-1-7x was determined to be the C-4,5 olefinic form of CT-1-7, that is, 7-O-descarbamoyl-7-hydroxygeldanamycin. CT-1-7x shared the same chemical structure as the compound designated as CP 144365,11 whose spectroscopic data and references, if any, were not disclosed.

Table 1 NMR data of CT-1-7xa

Shin et al.7 proved that a cytochrome P450 monooxygenase encoded by gdmP of the GDM biosynthetic gene cluster was involved in the formation of C-4,5 double bonds of GDM in the post-PKS tailoring process of GDM biosynthesis. The discovery of CT-1-7x from the gdmN disruption mutant of S. hygroscopicus 17997 indicated that the cytochrome P450 monooxygenase encoded by gdmP had a relaxed substrate specificity, which is common to enzymes for microbial secondary metabolite biosynthesis.

It seems that the carbamoyltransferase in the post-PKS tailoring process of GDM biosynthesis exhibited also a relaxed substrate specificity, as several GDM biosynthetic intermediates or their analogs with 7-O-carbamoylation, such as KOS-1806, reblastatin and 8-demethylgeldanamycin, were discovered or isolated.12, 13 To confirm whether CT-1-7x could be further 7-O-carbamoylated (producing GDM), a bioconversion experiment was conducted (Figure 3). Contrary to our expectation, incubation of CT-1-7x with GDM-pks−,8 a GDM's PKS gene disruption mutant of S. hygroscopicus 17997 (it lost the ability to produce GDM, but still had a full complement of post-PKS tailoring process genes), resulted in no production of GDM, while incubation of CT-1-7 (as the positive control) with GDM-pks− resulted in the production of GDM as had been reported earlier by Hong et al.6 This result led us to conclude that the carbamoyltransferase in the post-PKS tailoring process of GDM biosynthesis showed a special substrate specificity, that is, the C-4,5 of its substrate(s) being single-bonded. Therefore, 7-O-carbamoylation must take place before C-4,5 oxidation in the post-PKS tailoring process of GDM biosynthesis, but the two steps are not necessarily close to each other.

Figure 3
figure 3

Bioconversion of CT-1-7 and CT-1-7x. CT-1-7 was converted to geldanamycin (GDM) without difficulty by GDM-pks-, while CT-1-7x failed to be converted to GDM by GDM-pks-. Left: 1, CT-1-7; 2, GDM-pks-; 3, CT-1-7+GDM-pks-; 4, GDM. Right: 5, CT-1-7x; 6, GDM-pks-; 7, CT-1-7x+GDM-pks-; 8, GDM. For bioconversion, flasks of 250 ml, each containing 40 ml fermentation medium (starch 2%, glucose 0.5%, cottonseed cake power 0.5%, corn steep liquor 1.0%, yeast power 0.5% and CaCO3 0.2%) were inoculated with spores of GDM-pks-, shaken at 28 °C with 200 r.p.m. for 60 h and then treated with 4 mg CT-1-7 or CT-1-7x (in 100 μl ethyl acetate (EtOAc)) for each flask, and shaken for another 12 h (or for an even longer time of 24–48 h). The culture broth was extracted once with 40 ml EtOAc, the organic phase was evaporated in vacuo, and the residue dissolved in 200 μl EtOAc. About 10 μl of each sample was used for silica-gel TLC and then alkaline color reaction to check whether GDM was produced.