Stereochemical Assignment of Intermediates in the Rifamycin Biosynthetic Pathway by Precursor-directed Biosynthesis

Natural and semisynthetic rifamycins are potent inhibitors of bacterial DNA-dependent RNA polymerase and are used for the treatment of tuberculosis and AIDS-associated mycobacterial infections.¹ The parent compound, rifamycin B 1, is biosynthesized by the soil bacterium Amycolatopsis mediterranei employing the assembly-line methodology of a modular polyketide synthase (PKS).² The first of five PKS enzymes (RifA-E) is primed with 3-amino-5-hydroxybenzoic acid (AHBA, Figure 1), which is derived from the amino-shikimate pathway. The last step of this pathway comprises the aromatization of amino-DHS catalyzed by AHBA synthase (RifK).³ Knock-out of RifK activity delivers a mutant A. mediterranei strain (HGF003) which is incapable of rifamycin B production unless AHBA is supplemented. Formation of the naphthalene core of rifamycin B likely occurs after three rounds of C₂-elongation. The absolute and relative configuration of both stereocenters introduced by module 1 is obscured by the naphthalene aromatization. Premature chain release provides the shunt product P8/1-OG (or the derived lactone 2), which was isolated from mutated A. mediterranei strains as well as from engineered heterologous producing organisms.⁴ Feeding of 3,5-dihydroxybenzoic acid or 3-hydroxybenzoic acid instead of AHBA to the rifK(-) mutant of A. mediterranei produced analogs of P8/1-OG but no higher polyketides.⁵ After a total of ten rounds of C₂ elongation the fully assembled polyketide chain is cyclized to the macrolactam proansamycin X, which undergoes post-PKS modifications to finally yield rifamycin B 1.

Figure 1.

Biosynthesis of Rifamycin B. RifA is primed with AHBA, derived from amino-DHS by action of AHBA synthase (RifK). The loading module consists of an adenylation domain (A) for AHBA activation and a peptidyl carrier protein (PCP). Subsequently, ten cycles of C₂-elongation are carried out by RifA-RifE. Each cycle plus reductive modifications is catalyzed by one module being comprised of a β-keto synthase domain (KS), an acyl transferase domain (AT), an acyl carrier protein (ACP) and optionally a keto reductase domain (KR) or a dehydratase domain (DH, inactive in module 1). Premature chain release after module 3 of RifA gives rise to the shunt product P8/1-OG (or the derived lactone 2). Macrolactamization of the fully assembled polyketide chain is catalyzed by amide synthetase RifF leading to proansamycin X, which undergoes post-PKS modifications finally yielding rifamycin B.

Modified rifamycins are to-date only accessible by semi-synthesis from natural rifamycins. Precursor-directed biosynthesis, in which synthetic compounds are converted into modified natural products by engineered organisms,⁶ might complement semisynthetic approaches in opening access to currently inaccessible rifamycins. An understanding of specificity barriers within the rifamycin PKS constitutes one prerequisite for such an approach. The promiscuity of the loading module and module 1 of RifA has been investigated previously.^5,7 We have extended these studies now by synthesizing all four diastereomers of the biosynthetic substrate for module 2 of the rifamycin PKS in form of their N-acetylcysteamine (SNAC) thioester (3, Figure 1). Only one diastereomer was turned over in vivo into rifamycin B thus establishing the obscured absolute and relative configuration of the native biosynthetic intermediates.

The synthesis of syn-AHBA diketide 10 is shown in Scheme 1. 3,5-Dinitrobenzoic acid 4 was converted into phenol 5 applying the Rickards’ procedure.⁸ NaBH₄-I₂ reduction,⁹ MnO₂ oxidation and TBS protection provided carbaldehyde 6 as common precursor for all four diastereomers. Syn-Evans aldol conditions¹⁰ employing the (4R)-benzyloxazolidinone delivered aldol 7. TBS-protection and LiOOH-mediated cleavage of the auxiliary (accompanied by cleavage of the phenolic TBS group) was followed by formation of the N-acetylcysteamine (SNAC) thioester 9. Finally, HF-mediated TBS cleavage and hydrogenation in the presence of HCl gave rise to ammonium hydrochloride 10.

Scheme 1.

Representative synthesis of AHBA diketide 10^a

^a Reaction conditions: a) 1. LiOMe, HMPA, 85 °C, 20h, 86%; 2. AcOH, HBr, reflux, 20h, 95%. b) 1. NaBH₄, I₂, THF, rt, 6h, 79%; 2. MnO₂, CH₂Cl₂/acetone (10:1), rt, 24h, 75%; 3. TBSCl, imid., DMF, rt, 85–90%. c) (4R)-N-Propionyl-4-benzyloxazolidinone (1.1 equiv.), Bu₂BOTf (1.27 equiv.), DIPEA (1.43 equiv.), Et₂O, 97%, d.r. > 25:1. d) 1. TBSOTf, 2,6-lutidine, CH₂Cl₂, 0 °C to rt, 97%; 2. LiOH ·H₂O, H₂O₂, THF, 0 °C → rt, 91%. e) 1. CDI (2.1 equiv.), DMF, rt, 3h; 2. HSNAC, DMAP, DMF, rt, 84%. f) 1. HF (30%), CH₃CN, 0 °C to rt, 94%; 2. H₂, Pd/C, EtOH/HCl (1N) 10:1, rt, 99%.

The enantiomeric syn diketide 11 (Scheme 2) was accessible in analogous fashion from carbaldehyde 8 using the opposite enantiomer of the oxazolidinone. Anti-diketides 12 and 13 were prepared applying Heathcock’s anti-aldol conditions.¹¹

Scheme 2.

In vivo transformations

The hydrochlorides of all four diastereomers 10–13 were transformed into free anilines prior to administration to A. mediteranei strain HGF003 cultures by adjustment of substrate solutions to pH 7.2. Fermentation in the presence of AHBA was used as positive control. Analysis of the crude extracts by HPLC and LCMS using commercially available rifamycin B as reference showed that only anti-diketide 12 was transformed into rifamycin B while substrates 10, 11, and 13 did not deliver any detectable amount of 1. The turnover efficiency of the natural diketide analogue 12 into 1 is at least 100 times lower than that of AHBA, possibly reflecting a fundamental difference in the loading mechanism of SNAC substrate 12 compared to AHBA. To further underscore this finding, in vitro labeling and digestion experiments with RifA module 2 (RM2) were undertaken. A stand-alone protein construct of RM2 was incubated with substrates 10–13, digested with trypsin, and the resulting fragments were analyzed by LCMS. Substrate 12 was the only diketide to covalently label the KS domain of RM2, indicating the KS domain of RM2 provides a selectivity barrier to diketide incorporation.

Earlier studies have established that PKS modules possess intrinsic selectivity barriers discriminating between naturally and non-naturally configured substrates.¹² Therefore, substrate 12 likely contains the natural configuration of intermediates in the rifamycin PKS pathway. Additionally, the keto reductase domain of module 1 can be expected to be D-selective based on sequence analysis¹³ thus establishing that e.g. C7 in lactone 2 is S-configured (as in 11 and 12). The P8/1-OG derived lactone 2 shows a large ¹H-NMR coupling constant (> 9 Hz) for this benzylic proton.^4a Computational minimization of conformational energies for the syn- and anti-diastereomers of lactone 2 led to a dihedral angle of 173.86° for the anti-2, which is in agreement with the observed coupling constant (Karplus calculation: 9.1 Hz, measured: 9.0 Hz in d₆-DMSO). This additional evidence underscores that anti-diketide 12 contains the natural configuration of biosynthetic intermediates in the rifamycin PKS pathway.

Natural and semisynthetic rifamycins are clinically important inhibitors of bacterial DNA-dependent RNA polymerase. Although the polyketide-nonribosomal peptide origin of the naphthalene core of rifamycin B is well established, the absolute and relative configuration of both stereocenters introduced by the first polyketide synthase module is obscured by aromatization of the naphthalene ring. To decode the stereochemistry of the rifamycin polyketide precursor, we synthesized all four diastereomers of the biosynthetic substrate for module 2 of the rifamycin synthetase in the form of their N-acetylcysteamine (SNAC) thioester. Only one diastereomer was turned over in vivo into rifamycin B thus establishing the absolute and relative configuration of the native biosynthetic intermediates.