Two herbimycin analogs, 4,5-dihydro-(4S)-4-hydroxyherbimycin B and (15S)-15-hydroxyherbimycin B, from Streptomyces sp. CPCC 200291

Herbimycin A (Figure 1) is a benzoquinone ansamycin, discovered by Satoshi Ōmura in 1979, with herbicidal activity.¹ Herbimycin A possesses strong cytotoxicity against tumor cells, as it is a specific inhibitor of heat-shock protein 90, whose client proteins have important roles in cell-cycle regulation.^{2, 3} Chemical derivatives of herbimycin A were synthesized for improved activity against Ehrlich’s carcinoma.^{4, 5}

Figure 1

The chemical structures of herbimycins A/B/C, 4,5-dihydro-(4S)-4-hydroxyherbimycin B (1) and (15S)-15-hydroxyherbimycin B (2). A full color version of this figure is available at The Journal of Antibiotics journal online.

Streptomyces are well known for their potential to produce secondary metabolites with antibacterial, antifungal and cytotoxic activities. Streptomyces sp. CPCC 200291, with strong antibacterial and antifungal activities, was a soil isolate of Hainan Province, China. In the course of studying the secondary metabolites of the isolate, which is grown by solid-state fermentation, we identified herbimycins A and C as the principal components by TLC and LC–MS, and discovered two new herbimycin analogs as minor components. Herein, we report the identification, structure determination and probable biosynthetic routes of the two new herbimycin analogs.

Spores of Streptomyces sp. CPCC 200291 were spread on an ISP2 medium (composition: 0.4% yeast extract; 1.0% malt extract; 0.4% glucose; and 1.5% agar powder) and the plate was incubated at 28 °C for 7 days to produce an even layer of mycelial lawn. The agar culture was cut into small pieces and extracted with ethyl acetate (EtOAc). The EtOAc extract was concentrated and then analyzed by silica gel TLC developed with EtOAc–n-hexane–CH₂Cl₂–CH₃OH, 6:4:4:1, v/v. The two dominant yellow bands with R_f 0.75 and 0.52 were identified as herbimycins A and C, respectively (Supplementary Figure S1). There were two minor yellow bands at the lower part of the silica gel TLC plate. Similar to herbimycins A and C, they turned purple/blue upon spraying with 2 m NaOH, a diagnostic color reaction of benzoquinone ansamycins (Supplementary Figure S1).⁶ HPLC revealed peaks with UV–vis spectra similar or identical to that of herbimycins A or C, but with molecular ions [M+Na]⁺ at m/z 571 and 569 (Supplementary Figures S2–S6), which differentiated these metabolites from any known natural or biotransformation herbimycin analogs.^{7, 8, 9, 10, 11, 12}

To purify the two new herbimycin analogs, stock spores of Streptomyces sp. CPCC 200291 were spread onto 8–10 ISP2-medium plates and then incubated at 28 °C for 9–10 days for spore maturation. Fresh spores were recovered from these ISP2 plates with sterile water, and these were used to inoculate ~800 ISP2 (20 l) plates. After an incubation period of 6–7 days at 28 °C, the agar cultures were pooled and extracted three times with equal volume of EtOAc. The EtOAc extract was evaporated under reduced pressure to yield a dark brown solid residue (10.98 g). This was resuspended in 200 ml EtOAc, and 200 ml aqueous FeCl₃ (10%) was added and the mixture was stirred for 3 h. This procedure was to facilitate the subsequent separation and purification by oxidation of hydroquinone herbimycins to the quinones. The EtOAc layer was separated and concentrated at 35 °C to afford a brown oil, which was fractionated by normal-phase flash chromatography with a petroleum ether–EtOAc step-wise gradient 25–100% on a silica gel column, 35 × 460 mm (Büchi Sepacore C-620 from BUCHI Labortechnik AG, Flawil, Switzerland). Fractions eluting at ~90% EtOAc contained the new herbimycin analogs were pooled, dried (1.7 g), and further fractionated by Sephadex LH-20 chromatography (28 × 1400 mm; CH₃OH–CH₂Cl₂–petroleum ether, 5:3:1, v/v). This resulted in a yellow fraction containing the two new herbimycin analogs (150 mg, dried). These were separated by preparative HPLC on a Shimadzu (Kyoto, Japan) Prominence System with a LC-20AP pump and an SPD-20AV UV detector; using a YMC (Kyoto, Japan) C₁₈ column (21.4 × 150 mm, CH₃OH-H₂O, 58–100% in 50 min, 10 ml min⁻¹). Each of the new herbimycins was further purified by semi-preparative HPLC on a YMC C₁₈ column: 10.0 × 250 mm, with isocratic ACN-H₂O elution (30:70 for 1 and 34:66 for 2), yielding 1 (3.8 mg) and 2 (9.3 mg) as yellow amorphous powders.

HR-ESI(+)MS (Supplementary Figure S7) established the molecular formula of 1 as C₂₈H₄₀N₂O₉ (measured 571.26340, calculated 571.26260 for C₂₈H₄₀N₂O₉Na), which is -C₂H₂ less than herbimycin A (C₃₀H₄₂N₂O₉). The unsaturation degree of 1 is 9, whereas that of herbimycin A is 10, suggesting that 1 may contain one olefinic bond less than herbimycin A. The UV absorption profile of 1 seemed to support this prediction, as it lost the shoulder peak at 240–250 nm ascribed to the dienamide moiety of herbimycin A.^{1, 13} On the basis of the molecular formula and saturation degree differences, we speculated that 1 lacked two of the methyl groups in herbimycin A. The two methyl groups may be the O-methyl groups at C-11 and C-15, which were added in the post-PKS (polyketide synthase) tailoring process in the herbimycin A biosynthesis. The other two O-methyl groups (at C-6 and C-12) should be present in 1, as they were incorporated into the molecule in the form of methoxymalonyl-CoA at the PKS stage of biosynthesis.¹⁴

The NMR spectra (Supplementary Figures S8–S13) of 1 were very similar to herbimycin A.^{4, 15} They showed signals characteristic of benzoquinone ansamycins, such as the 2,5-disubstituted-p-benzoquinone ring. Compound 1 revealed resonances for two tri-substituted double bonds at δ_H 6.03 (dd, J=9.0, 1.2 Hz, H-3) and 5.53 (d, J=9.6 Hz, H-9), whereas herbimycin A displayed resonances for one tri-substituted conjugated double bond (that is, two double bonds) and one tri-substituted double bond. On the basis of these differences, together with the aid of ¹H-¹H COSY correlations of H-3/H-4/H₂-5/H-6/H-7 (Figure 2) and their chemical shifts (δ_H 6.03 (1H, dd, J=9.0, 1.2 Hz, H-3), 4.74 (1H, m, H-4), 1.93 (2H, m, H-5), δ_C 136.6 (C-3), 65.2 (C-4), 37.1 (C-5)), we concluded that 1 had a C-4,5 single bond and a hydroxyl group at C-4, rendering C-4 chiral.

Figure 2

¹H-¹H COSY (thick lines) and main HMBC (blue arrows) correlations of 1 and 2. A full color version of this figure is available at The Journal of Antibiotics journal online.

The ¹H-¹H COSY correlations of H-10/H-11/H-12, H₂-13/H-14/H₂-15 and HMBC correlations from H-9, CH₃-10 and H₂-13 to C-11, and from H-21 to C-15 (Figure 2) established that 1 had a hydroxyl group bonded to C-11 (δ_C 73.7) and a methylene at C-15 (δ_C 38.7), which implied that 1 had lost the C-11 O-methyl group and the C-15 methoxyl group in herbimycin A. The NMR spectra of 1 indicated only two methoxy singlet peaks (δ_H 3.45 and 3.33) assigned to C-6 and C-12, respectively, by HMBC correlations from 6-OCH₃ to C-6, 12-OCH₃ to C-12, and COSY correlations of H-4/H₂-5/H-6/H-7 and H-9/H-10/H-11/H-12/H₂-13 (Figure 2).

As 1 was co-produced with herbimycin A and shared the PKSs in herbimycin A biosynthesis, the absolute configurations (except at C-4) of 1 and herbimycin A are presumed to be the same. The absolute configuration of C-4 in 1 was determined to be S by Mosher’s method (Figure 3; Supplementary Figures S14–S19).^{6, 16} Therefore, the structure of 1 was determined to be 4,5-dihydro-(4S)-4-hydroxy-11-O-demethyl-15-demethoxyherbimycin A, or 4,5-dihydro-(4S)-4-hydroxyherbimycin B (Figure 1). The NMR spectra of 1 are assigned completely in Table 1.

Figure 3

The Δδ^SR values of 1 R and 1 S from Mosher’s reactions and the absolute configuration of Sec-OH at C-4 in 1. A full color version of this figure is available at The Journal of Antibiotics journal online.

Table 1 The NMR spectra data of 1 and 2

HR-ESI(+)MS (Supplementary Figure S20) established the molecular formula of 2 as C₂₈H₃₈N₂O₉ (measured 569.24777 and calculated 569.24695 for C₂₈H₃₈N₂O₉Na), which was -C₂H₄ less than herbimycin A (C₃₀H₄₂N₂O₉). Compound 2 had the same unsaturation degree as herbimycin A. In addition, 2 exhibited nearly identical UV–vis absorption profile to that of herbimycin A. Therefore, 2 should be a close analog of herbimycin A, probably lacking two methyl groups.

A comparison of the NMR data of 2 (Supplementary Figures S21–S26) with that of herbimycin A confirmed the speculation on the structure of 2. The ¹H-NMR spectrum indicated the presence of only two methoxyl groups in 2 rather than four in herbimycin A. These were assigned as attached to C-6 and C-12 by ¹H-¹H COSY correlations of H-5/H-6/H-7, H-9/H-10/H-11/H-12 and by HMBC correlations of 6-OCH₃/C-6, 12-OCH₃/C-12 (Figure 2). Thus, 2 is assigned the structure without the two O-methyl groups linked to C-11 and C-15, which were added in the post-PKS tailoring in the biosynthesis of herbimycin A. The ¹H-¹H COSY correlations of H-9/H-10/H-11/H-12/H₂-13/H-14, together with their chemical shifts, indicated that C-11 of 2 carried a hydroxyl group (Figure 2). The C-15 of 2 is also associated with a hydroxyl group from its δ_C 71.7 (78.7 in herbimycin A), although the proton signal of C-15 was not observed. Other sections of the ¹H-NMR of 2 were very similar or identical to that of herbimycin A. As 2 was co-produced with herbimycin A and shared the same PKS for biosynthesis, the absolute configurations of chiral carbon atoms (including C-11 and C-15) of 2 are presumed to be the same as in herbimycin A. Thus, the structure of 2 was assigned as 11,15-di-O-demethylherbimycin A, or (15S)-15-hydroxyherbimycin B (Figure 1). The NMR spectra of 2 are assigned completely as Table 1.

Nine natural herbimycin analogs (herbimycins A–F, 17,19-dimethylthioherbimycin A, 1 and 2) have been identified.^{9, 11} Of them, herbimycin A is the primary major component. Herbimycins B and C could arise from incomplete post-PKS tailoring modifications in the biosynthesis of herbimycin A. Herbimycin D is the result of the addition of cysteine to herbimycin A,¹⁷ and 17,19-dimethylthioherbimycin A is the result of the addition of two methanethiols to herbimycin A.¹⁰ Herbimycins E and F may be shunt products in the biosynthesis of herbimycin A, with some new modifications (Figure 4).

Figure 4

Compounds 1, 2 and other natural herbimycin analogs (B–F) as shunt or as intermediate products of herbimycin A biosynthesis.

Compound 1 is probably a shunt metabolite in the biosynthesis pathway to herbimycin A without the benefit of the normal post-PKS tailoring enzymes, which would result in C-15 hydroxylation, and C-11 and C-15 O-methylations. The mechanism of the apparent C-4,5 stereospecific hydration is more difficult to speculate, as the position of the hydroxyl is such that it cannot have arisen in the normal functioning of a PKS. However, in our previous study of the biosynthesis of geldanamycin (a close analog of herbimycin A), we identified 4,5-dihydro-(4R/S)-4-hydroxygeldanamycins as a pair of minor products from the geldanamycin producer, Streptomyces hygroscopicus 17997.⁶

Compound 2 is presumably an intermediate of herbimycin A biosynthesis, lacking both C-11 and C-15 O-methylations (Figure 4).

The relative polarity of 1 and 2 compared with other herbimycins suggests greater water solubility. We assayed their water solubility, and found that 1 and 2 had slightly higher water solubility than herbimycin C (0.9422, 0.9418 and 0.8943 mg ml⁻¹, respectively), and >17-fold higher than herbimycin A (0.0512 mg ml⁻¹). This suggests better bioavailability of 1 and 2 than the water-insoluble herbimycin A.

A preliminary cytotoxicity comparison of 1, 2 and herbimycin A against HCT116 (colon carcinoma cell line), Hela (cervical cancer cell line), A549 (lung cancer cell line) and HepG2 (liver hepatocellular carcinoma cell line) was conducted. Except for 1 showing slightly higher cytotoxicity against HCT116 than herbimycin A (IC₅₀: 24.1 compared with 36.6 μM), 1 and 2 displayed less cytotoxicity than herbimycin A against all the cancer cell lines tested (Supplementary Table S1).