Main

The resistance of microorganisms to existing antimicrobial medicines poses an increasing public-health threat worldwide5,6. To address this problem, chemical agents that operate through novel modes of action—especially those that are unique to the offending microorganism—are particularly attractive7. For example, iron is essential for the growth and survival of organisms, and is carried into microbial cells with the aid of compounds known as siderophores; inhibiting the biosynthesis of siderophores provides a mechanism of action that is unique to the pathogen, because mammalian host cells use other ways to regulate iron concentration8. From a library of 19,855 marine-microbe-derived natural products, and using an assay to target siderophore biosynthesis, two natural products have been identified—baulamycin A and B—that are active against the superbug methicillin-resistant Staphylococcus aureus (MRSA) and Bacillus anthracis (for baulamycin A, the half-maximal inhibitory concentrations are 69 μM and 110 μM, respectively)4. However, the limited availability of these products (just 3.6 mg of baulamycin A and 2.1 mg of baulamycin B were obtained from 39 litres of culture) precluded their chemical modification as a means to structural assignment, which was thus guided entirely by isotropic solution-state NMR spectroscopy—principally, empirical J-based configurational analysis3.

Study of the biology of these products was similarly limited by their supply, so we decided to synthesize them. Below, we report a ten-step synthesis of the proposed structures of baulamycin A (compound 1) and B (compound 2), but unfortunately we found that the data did not match those for the natural products. During the course of our work, a 17-step synthesis of the proposed structure of baulamycin A was reported9. These authors similarly found that the data did not match, and further attempts to prepare the correct diastereomer on the basis of re-examination of the J-based configurational analysis were also unsuccessful.

Elucidating the structures of natural products that are available only in very small quantities is often exceptionally difficult, highlighted by the high number of structure revisions reported every year in the chemical literature2. This challenge is epitomized by the baulamycins, which contain seven stereogenic centres (128 possible stereoisomers), of which only three are contiguous, distributed along a 14-carbon-long chain. The flexibility of the chain leads to experimental NMR parameters that are a weighted average from an ensemble of conformations. This severely complicates configurational analysis to the extent that standard approaches based on empirical NMR analysis3, or even recent quantum-chemical approaches using chemical-shift data10,11,12, are insufficient. However, combining accurate quantum-chemical prediction of the ensemble-averaged NMR parameters with analysis of encoded stereoisomeric mixtures—which were accessible through the power of iterative, reagent-controlled homologation of boronic esters (assembly-line synthesis)13—has enabled us to correct five of the seven stereogenic centres and to finally establish the correct structure of these important natural products (see below).

In order to plan a synthesis of baulamycin A (1) and B (2), we first carried out a retrosynthetic analysis of the compounds and identified a disconnection at C11–C12, which divides the target molecule into equally complex halves—fragments A and B (Fig. 1a). We considered joining them through a late-stage lithiation–borylation reaction, a process that uses our recently reported regioselective and stereoselective homologation of 1,2-bis(boronic esters)14. Fragment A could be obtained through a Morken hydroxy-directed diboration of homoallylic alcohol 6 (ref. 15), which itself could be obtained by Antilla allylboration16,17. We envisaged fragment B coming from our recently developed assembly-line synthesis13, which lends itself readily to the synthesis of analogues.

Figure 1: Retrosynthetic analysis and synthesis of the originally proposed structures of baulamycin A and B.
figure 1

a, Retrosynthetic analysis. Disconnection across the C11–C12 bond through lithiation and borylation gives two fragments, which can be recombined using a late-stage lithiation–borylation reaction. Fragments A and B can be assembled from smaller building blocks, as shown at the right. P, protecting group; TIB, 2,4,6-triisopropylbenzoyl; pin, pinacolato. b, Synthesis of fragment A (R,S,R)-17. c, Synthesis of fragment B, (R,S,R)-22 and (R,S,R)-23, for baulamycin A and B, respectively. d, Fragment union and deprotection to give the proposed structures for baulamycin A (1) and B (2). cat, catecholato; cod, cyclooctadiene; MOM, methoxymethyl; sp., sparteine; TES, triethylsilyl; THF, tetrahydrofuran; TRIP-PA, 3,3′-bis(2,4,6-triisopropylphenyl)-1,1′-binaphthyl-2,2′-diyl hydrogenphosphate.

PowerPoint slide

Full size image

We began with the synthesis of fragment A (Fig. 1b). Rhodium-catalysed hydroboration of alkyne 13 gave Z-vinyl boronic ester 14 (ref. 18), which was homologated with chloromethyl lithium19 to give Z-allyl boronic ester 8. Antilla allylboration of the methoxymethyl (MOM)-protected aldehyde gave the homoallylic alcohol 16 with good yield, high diastereoselectivity, and high enantioselectivity16,17. Subsequent Morken hydroxy-directed diboration15 gave the 1,2-bis(boronic ester) (R,S,R)-17 (after protecting the hydroxy group with triethyl silyl ether) in a diastereomeric ratio (d.r.) of 98/2, thus completing a short, highly stereoselective synthesis of fragment A. The synthesis of fragment B commenced with rhodium-catalysed hydroboration of allyl benzoate 19 to give boronic ester 12 (refs 20, 21)—the starting material for assembly-line synthesis (Fig. 1c). Then, sequential treatment of boronic ester 12 with the carbenoids (S)-11, chloromethyl lithium (10), (R)-11, 10 and finally (S)-11 gave the target boronic ester 9 in 64% yield. To complete the synthesis of fragment B, we decided to introduce an enol ether as a masked ketone, through a Zweifel olefination22. Thus, boronic ester 9 was treated with lithiated MOM vinyl ether 20 and the ethenyl variant 21 (ref. 23), and then with molecular iodine, giving the desired variants of fragment B—namely (R,S,R)-22 and (R,S,R)-23—at 72% and 86% yield, respectively. Joining fragments A and B (Fig. 1d) involved lithiation of the benzoate ester (22 or 23) followed by regioselective homologation of the primary boronic ester moiety of 1,2-bis(boronic ester) (R,S,R)-17 (ref. 14) to give the 1,3-bis(boronic ester), which, after oxidation, gave the desired diols 24 and 25 with high stereoselectivity (d.r. 95/5). Finally, treatment with aqueous hydrochloric acid in a mixture of tetrahydrofuran and methanol effected removal of the silyl and MOM groups, completing a short synthesis of the proposed structures of baulamycin A (1) and B (2)4.

Unfortunately, however, the 1H- and 13C-NMR spectra of the synthetic samples did not match those of the reported natural products, leading us to conclude that one or more of the stereogenic centres of the natural products had been misassigned4. But which of the seven stereogenic centres was incorrect? It was clearly not practical to make all 128 stereoisomers, so further analysis of the NMR data was required. As noted, the molecule can be divided into two halves and we initially considered fragment A, the C10–C1′ portion (Fig. 2a). Comparing the observed coupling constant for the vicinal protons of C1′–C14 in our synthetic material (3.6 Hz) with that of the natural product (7.2 Hz), we reassigned the syn configuration of C1′–C14 (originally proposed for the natural product) to anti: the small and large coupling constants are characteristic of syn and anti motifs, respectively, in similar fragments reported in the literature24,25,26. Then, we computed the NMR parameters for the C10–C1′ region of the four remaining diastereomers, 26–29 (which arise from varying the configuration at C13 and C11; Fig. 2b), and compared them with the parameters of the isolated natural material. For this computational analysis, we maintained the relative stereochemical configuration in fragment B as anti–anti, as reported in the original paper4.

Figure 2: Stereochemical analysis and synthesis of fragment A.
figure 2

a, The proposed structure of baulamycin A (1) from ref. 4, highlighting fragment A. b, Structures of fragment A diastereomers 26–29 used in our DFT predictions of NMR parameters for C10–C1′. c, Computed C10–C1′ NMR parameters (1H–1H coupling constants, given as a J value, and interproton NOE distances) of 26–29 and 1 were compared with the corresponding experimental data for the natural product, producing a χ2 (reduced) value. Diastereomer 29 is the only viable fit (χ2 (reduced) is about 1) to the corresponding experimental data. d, An overlay of all conformers of 29 analysed through DFT calculations. e, Synthesis of revised fragment A, (R,R,R)-34. DMSO, dimethylsulfoxide; [Pd(dmba)Cl]2, di-μ-chlorobis{2- [(dimethylamino)methyl]phenyl-C,N}dipalladium(ii).

PowerPoint slide

Full size image

Conformational analysis of diastereomers 26–29 using molecular mechanics found between 650 and 2,455 conformations for each diastereomer. For each diastereomer, we submitted low-energy conformers (around 84–196 conformers) to sequential density functional theory (DFT) geometry optimization and free-energy calculations (Fig. 2d). The populations of the resulting conformers were refined further on the basis of quantitative interproton distances27 calculated from rotating-frame Overhauser spectroscopy (ROESY) nuclear Overhauser effect (NOE) measurements4. Next, we subjected the conformers that were predicted to make up more than 85% of the ensemble populations of diastereomers 26–29 (8–28 conformers) to DFT calculations, in order to predict coupling constants for the C10–C1′ regions. We then compared both the calculated 1H–1H coupling constants and the NOE-derived interproton distances for the C10–C1′ region of diastereomers 26–29 with the corresponding experimental data for the isolated natural product, baulamycin A, and applied a statistical analysis based on χ2 (reduced) values (Fig. 2c), whereby acceptable models must have values approaching 1.

On the basis of 1H–1H coupling constants alone, there is an excellent fit for the C10–C1′ region of structure 29 (χ2 (reduced) = 1.5), moderate fits but potentially viable matches for both 26 (χ2 (reduced) = 2.4) and 27 (χ2 (reduced) = 2.1), and no reasonable fit for 28 (χ2 (reduced) = 6.2). On the basis of NOE distances alone, there are good fits for 28 and 29 (χ2 (reduced) = 1.4 and 1.7, respectively) and poor fits for 26 and 27 (χ2 (reduced) = 5.3 and 5.4, respectively). Similarly, comparing the DFT-predicted NMR parameters of the structure originally proposed for baulamycin A (1) with the experimental NMR data of the isolated material revealed that both sets of NMR parameters did not fit simultaneously (χ2 (reduced) = 1.6 and 5.1 for 1H–1H coupling constants and NOE distances, respectively). Thus, diastereomer 29 was the only structure that fitted both experimental NMR parameters simultaneously for fragment A. Our analysis had thus identified the relative configuration of four of the seven stereogenic centres, thereby reducing the possible isomers of baulamycin from 128 down to 16. This assignment is also supported by χ2 (reduced) analysis of computed chemical shifts and 1H−13C coupling constants, although neither was as discriminating as the analysis based on 1H–1H coupling constants and NOE distances. Applying comparable analyses10,11 that are based on computed NMR chemical shifts for diastereomers 1 and 26–29 could not alone provide structural discrimination, thus highlighting the care that must be taken with NMR-based stereochemical analysis of flexible complex molecules. We synthesized the revised anti–anti fragment A, (R,R,R)-34, by using the same highly diastereoselective method used for (R,S,R)-17, but with the E-allyl boronic ester 32 (ref. 28) in place of the corresponding Z-isomer in the Antilla allylboration (Fig. 2e).

We also undertook computational analysis of fragment B (C4–C8) for diastereomers 29 and 35–37, but in this case we computed only 1H–1H coupling constants (Fig. 3c), owing to overlap in the critical region of the ROESY spectrum4. This analysis indicated that diastereomers 29 and 35 were very poor fits (χ2 (reduced) = 4.7 and 8.6, respectively) and could be excluded from consideration; however, the syn–anti and syn–syn diastereomers 36 and 37 could not be discriminated (χ2 (reduced) = 1.6 and 1.6, respectively) on the basis of 1H–1H coupling constants alone. Further analysis of the fits of chemical shifts, while less discriminating than the fits of coupling constants, suggested that syn–syn diastereomer 37 gave a good fit (χ2(reduced) = 0.2 and 0.6 for 1H and 13C chemical shifts, respectively), but that syn–anti diastereomer 36 fitted less well (χ2 (reduced) = 0.9 and 2.9, respectively).

Figure 3: Stereochemical analysis and synthesis of fragment B.
figure 3

a, Isomers of baulamycin A based on 29, highlighting fragment B. b, Fragment B diastereomers 29 and 35−37. c, Comparison of computed C4−C8 NMR parameters of 29 and 35−37 with experimental data. Structures 29 and 35 can be excluded from further analysis (because of χ2 (reduced) values being much greater than 1); 37 is the most viable fit; 36 may not be confidently excluded. d, Synthesis of an encoded mixture of baulamycin A diastereomers (by virtue of known but inequivalent amounts of each isomer). The experimentally measured integration of each diastereomer is given in bold blue type, and the expected (calculated) population is in parentheses. e, Comparison of the resulting 13C NMR spectrum (for the C6 and C21 bonds) of this mixture of diastereomers with that of the natural product, indicating a match for signals corresponding to 37.

PowerPoint slide

Full size image

In parallel, we sought to determine the relative configuration of fragment B through synthesis. We anticipated that the most powerful aspect of assembly-line synthesis—that is, the exquisite reagent-mediated control over stereoselectivity—could be used to rapidly generate an encoded mixture of all four diastereomers of fragment B, the identity of each diastereomer being indicated by its relative population. The strategy is reminiscent of the fluorous-mixture synthesis that was used to rapidly identify stereoisomers of a given natural product29. Specifically, growing the C11–C1 carbon chain by using chiral carbenoids with moderate, yet accurately measured enantiomeric ratio (e.r.) values (for C8, C6 and C4) would lead to the synthesis of fragment B as a mixture of all four diastereomers, wherein the relative population of each diastereomer could be accurately predicted (Fig. 3d). Linking the revised anti–anti fragment A with this diastereomeric mixture would give a mixture of diastereomers 29 and 35−37. Comparing the resulting 13C-NMR spectrum with that of the isolated material and noting the relative integration of the matching peaks should immediately confirm the relative configuration of the methyl-group-rich region. At this stage, we decided to prepare only one enantiomeric series of the four diastereomers of fragment B. We would then resolve, at a later stage, the remaining stereochemical elements.

For this approach, we needed to ensure that the relative populations of diastereomers were such that all four diastereomers could be accurately quantified by NMR spectroscopy, and that they were maximally distributed in terms of peak intensity. We therefore selected e.r. values (ratios of S/R enantiomers) of >99.9/0.1, 72/28 and 64/36 for α-stannyl ethyl benzoate (the carbenoid precursor to be used in the assembly-line synthesis) at, respectively, the first (C8), third (C6), and fifth (C4) iterations. These e.r. values were calculated to lead to a 46/26/18/10 mixture of diastereomers 37/36/29/35. Using R/S mixtures of carbenoid precursors with accurately measured e.r. values, we obtained fragment B as a 47/25/18/10 mixture of diastereomers (Fig. 3d). The almost-perfect match between the expected and observed ratio of diastereomers showed that the homologation reactions operated under full reagent control. The mixture was carried forward for fragment coupling and removal of protecting groups (Fig. 3d) to give the desired mixture of baulamycin A diastereomers, the 47/25/18/10 ratio being retained with high fidelity. Comparing the 13C-NMR spectrum of the mixture of diastereomers (Fig. 3e) with that of the isolated natural product revealed that the chemical shifts of the diastereomer with the highest population matched those of the natural product closely; this finding, in agreement with the DFT calculations, established that the correct relative configuration of fragment B was indeed syn–syn, and not anti–anti as originally proposed.

With the relative configuration within each fragment established, we set out to determine which of the two remaining diastereomers (C11/C8 syn or C11/C8 anti) was baulamycin A (Fig. 4). The two enantiomers of the anti–anti diastereomer of fragment A [(R,R,R)-34 and (S,S,S)-38] were coupled to the syn–syn diastereomer of fragment B [(R,R,R)-39]. As expected, the resulting two diastereomers, 37 and 40, had almost identical 13C-NMR spectra, except for very small differences at C7, C9 and C11, with the C11/C8 syn diastereomer matching the natural product perfectly. Furthermore, the diastereomers exhibited marked differences in certain regions of the 1H-NMR spectra, the C11/C8 syn diastereomer 37 again matching the spectrum of the natural product perfectly. However, the optical rotation of the synthesized matching diastereomer 37 was positive, whereas that of the isolated natural product was negative, indicating that compound 37 was the enantiomer of the natural product. To prepare the correct enantiomer of baulamycin A, we linked together (S,S,S)-38 (fragment A) and (S,S,S)-41 (fragment B). Using the same protocol we also synthesized baulamycin B with the revised configuration, the analytical data fully matching that of the reported natural product.

Figure 4: Determination of the relative and absolute configurations of baulamycins A and B.
figure 4

a, Reaction of the two enantiomers of fragment A, (S,S,S)-38 and (R,R,R)-34, with one enantiomer of fragment B, (R,R,R)-39. Comparing the 1H- and 13C-NMR spectra of the resulting compounds 40 and 37 revealed that 37 had the same relative configuration as baulamycin A, but had the opposite optical rotation to the natural product. b, Coupling fragment A, (S,S,S)-38, to fragment B, (S,S,S)-41 and (S,S,S)-42, gives the correct structure of baulamycin A (ent -37) and B (43).

PowerPoint slide

Full size image