Assignment of the absolute configuration of aurantoside G and J
Graphical abstract
Computational chemistry identifies the correct relative configuration of two naturally occurring aurantosides (G and J). In addition, based on the comparison of calculated and experimental specific optical rotation at 589 nm, the absolute configuration of both compounds is determined.
Introduction
Tetramic acids are partial structures of natural products [1], [2] such as the mycotoxin tenuazonic acid and a family of coloured compounds, the aurantosides. Since the isolation of its first members aurantoside A and B in 1991 by Matsunaga et al. (1991) [3], the following aurantosides have been isolated and characterized: C [4]; D, E, and F [5]; G, H and I [6]; J [7]; K [8]. Interestingly, the formulas of aurantosides G and J show a diastereomeric relationship. However, the configurational analysis of the family of aurantosides is based on traditional chemical procedures [3] (oxidation and derivatization to obtain the putative amino acid component of the tetramic acid part for GC-analysis; acidic methanolysis and subsequent derivatization for GC-analysis of the sugar component) which are typically prone to artefact generation such as racemization of aspartic acid [9]. In fact, epimerization at C4 might occur in the organism, during the isolation/storage, and synthesis [10] (as an example: the two C4 epimers harzianic and isoharzianic acid were isolated from the same organism.) [11].
In case of aurantosides G and J no proof for d-xylose instead of l-xylose, and no proof for the stereodescriptor at C4 were given. Instead, the assignments were based on biogenetic reasoning [6]. The proposed formulae of Fig. 1A and B are mainly based on spectroscopic NMR data (3JHH couplings, 1H/13C chemical shifts) [6], [7]. Throughout this publication, we refer to the numbering of Fig. 1A and B for interpretating literature data of related compounds. In particular the numbering in the tetramic acid part occasionally starts at the nitrogen position, however, we adopt the numbering of the original publication of the isolation of aurantoside G [6] for better comparison.
The first total synthesis of an aurantoside was recently presented by the Schobert group [12]. The authors claimed to have synthesized aurantoside G with the formula of Fig. 1A based on the comparison of the spectroscopic data presented in the isolated natural product [6] with the data for the synthesized compound [12]. Concerning this claim we want to illustrate and emphasize the restrictions under which a structural proof of an isolated natural product can be given by synthesis. The importance of this point was shown by Nicolaou and Snyder (2005) [13] describing the role of chemical synthesis in modern structure elucidation with special attention to molecules that were never there, meaning that the proposed formulae were incorrect. Consequently, it is not suprising that structure revisions can be found and can be expected for the future [14], [15].
Interestingly, in the case of the isolation and characterization of aurantoside G [6] neither the source nor the isolated compound are available any more. In this case which in our opinion could be quite common, a molecule existed for which a formula was proposed, and the proof of its existence can be given by its total synthesis according to the published formula and subsequent comparison of the experimental structural data. In addition, computational chemistry allows sufficiently accurate predictions of experimental spectroscopic data [16], rendering it possible to complement and/or substitute synthesis. However, the fundamental sine qua non condition for both approaches is the comparison with ALL compounds that could reasonably represent the initially isolated molecule. In this regard, the claim of Petermichl et al. (2016) [12] to have synthesized aurantoside G while referring to the proposed formula of Ratnayke et al. (2005) [6] is incorrect as long as not a comparison of the structural data of all reasonably possible molecules has been performed, using synthesis and/or computational chemistry. This problem was encountered by the same group in the synthesis of epicoccamide D for which only two diastereomers were prepared and compared to the original natural product [17]. Taking further into account that small differences in chemical shifts and specific optical rotations exist, the claim to have synthesized the originally isolated compound represented by the published formula [18] is again not justified. As an example, the non-synthesized diastereomer (5R,7R)-1d could have similar 13C chemical shifts and a similar specific optical rotation compared to the synthesized diastereomer (5S,7S)-1d (Table 2 in Ref. [17]) leading to an ambiguous assignment. In addition, the chemical shifts were measured in CDCl3 for the naturally occurring compound [18], whereas the synthetic compound was measured in methanol-d4. Since it is known that the NMR spectra of 2-acylated tetramic acid derivatives are highly solvent dependent (e.g. harzianic acid: 13C chemical shift differences of 2 ppm in met-d4 compared to CDCl3) [19] it is not justified to base the diastereomeric discrimination on the comparison of NMR resonances obtained in two different solvents. In addition, a closer look at the resonances revealed that differences in 13C chemical shifts of more than 1 ppm between natural and synthetic product occur (e.g. C1, C3) which are much larger than the differences used for diastereomeric discrimination (largest difference of 0.7 ppm) [17].
In the present publication we put forward evidences that literature contains the correct assignments for both aurantoside G and J, and that the publication of the Schobert group [12] can be correctly regarded as total synthesis of aurantoside G as represented by the formulae proposed by Ratnayke et al. (2005) [6]. Due to the restriction that only published data can be used in this and comparable cases, recent examples of a combined approach computational chemistry and experimental NOE-derived distances [20], electronic circular dichroism (ECD) spectra [21], or residual dipolar couplings (RDCs) [22], are not appropriate since neither NOE nor RDC/ECD data have been presented in the publications about aurantoside G and J [6], [7], [12].
The present publication is organized as follows: first, structural ensembles for each of the reasonably possible isomers are created; second, using a reference compound regression parameters for chemical shift predictions are obtained; third, the comparison between calculated and experimental chemical shifts of the isolated compounds allows a discrimination among diastereomers; fourth, the comparison between calculated and experimental specific optical rotation discriminates among enantiomers. Since only a specific optical rotation at 589 nm was presented, it is clear that the absolute configuration (AC) can be predicted only on this level of reliability [23], [24].
Throughout this study, the comparisons were made between calculated and experimental data obtained from the isolated natural products. This is justified since only small differences exist between the experimental data of isolated aurantoside G [6] and synthesized compound [12], and a proof of the correct formula of the isolated natural product is the first intention of this study.
Section snippets
Results and discussion
Based on the 3JHH couplings pattern, the sugar of aurantoside G and J was determined as xylose (excluding arabinose and lyxose: Table S1A, supporting information), and from the specific couplings of 3JH1′,H2′ = 8.8 Hz (aurantoside G) and 3JH1′,H2′ = 2.6 Hz (aurantoside J), the glycosidic linkages were derived as β and α, respectively [6], [7]. No NOE information was presented for both compounds.
A comparison between the experimental proton and carbon chemical shifts of the isolated aurantosides
Conclusion
Analysis of 13C and 1H chemical shifts support the claim of the Schobert group [12] that the isolated compound named aurantoside G [6] belongs to one member of the 4S-β-d-xylose/4R-β-l-xylose aurantoside enantiomeric pair. Our conclusion also supports a β-linkage of the xylose moiety which was proposed on the basis of JHH couplings [6].
Since sign and magnitude of the calculated optical rotation of 4S-β-d-xylose aurantoside match with experiment, we can conclude that the isolated aurantoside G
Computation
Models of aurantoside G and J, and their 4R epimers were built in GaussView 5.0. A systematic conformational search with three torsional angles τ1: C3C4C5C6; τ2: C4C5C6N; τ3: C4NC1′C2′ as variables was performed on a semiemperical level of theory (PM6, in vacuo, Gaussian09) [43]. Conformers lowest in energy were geometry-optimized on the mpw1pw91/cc-pvdz (IEFPCM: methanol as solvent) level of theory [52]. On the same level spectroscopic properties (NMR shieldings and specific optical rotations)
Acknowledgements
We want to thank the DFG (FOR 934, RE 2612/5-1) and the MPG (MPIBPC: Department for NMR-based structural biology, Prof. Griesinger) for support.
References (52)
- et al.
Bioorg. Med. Chem.
(2011) - et al.
Org. Lett.
(2015) - et al.
Org. Biomol. Chem.
(2015) - et al.
J. Mol. Struct.
(2016) - et al.
Chem. Phys.
(2003) - et al.
Org. Biomol. Chem.
(2013) - et al.
J. Antibiot.
(1988) Theor. Chem. Acc.
(2007)- et al.
Chem. Comm.
(2015) - et al.
Tetrahedron
(2000)