Total Synthesis of Malacidin A by β-Hydroxy Aspartic Acid Ligation Mediated Cyclization and Absolute Structure Establishment

Abstract

The development of novel antibiotics is critical to combating the growing emergence of drug-resistant pathogens. Malacidin A is a new member of the calcium-dependent antibiotic (CDAs) family with activity against antibiotic-resistant pathogens. Its mode of action is distinct from classical CDAs. However, the absolute structure of malacidin A has not been established. Herein, the total syntheses of malacidin A and its analogues are reported by a combination of Fmoc-based Solid-Phase Peptide Synthesis (SPPS) and β-hydroxy aspartic acid ligation mediated peptide cyclization. The total synthesis enabled us to establish the absolute configuration of malacidin A, which is in agreement with those for natural malacidin A confirmed by advanced Marfey's analysis in our study.

Graphical Abstract

More than ninety years since the discovery of penicillin, developing new classes of antibiotics remains an urgent need with the increasing emergence of bacterial pathogens with multi-drug resistance.¹ Among the different antibiotic skeletons, (e.g., β-lactam, tetracycline, aminoglycoside, etc.), cyclic peptides represent an important class of antibiotics. This includes bacitracin, vancomycin, daptomycin and polymyxins which are all used in the clinic. Known cyclic peptide antibiotics often work through non-protein targets. With emerging technologies, such as environment-mimic cultivation method² and genome mining of Biosynthetic Gene Clusters (BGCs)³, several cyclic peptides with new structural motifs have been recently discovered and exhibited promising potency. Through genome mining of BGCs from environmental DNA, malacidins A and B were identified as a new class of calcium dependent antibiotics (CDAs).⁴ Malacidin A exhibits broad activity against Gram-positive bacteria including methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant Enterococcus (VRE), and has potent MIC values (0.2-2 μg/mL) in the presence of the divalent cation calcium.

Interestingly, the malacidin family contains a calcium-binding motif (HyAsp-Asp-Gly) that differs from the canonical sequence (Asp-Xaa-Asp-Gly) observed in other known CDAs⁵. In addition to the lack of the variable spacer amino acid (Xaa), a non-proteinogenic hydroxyl aspartic acid (HyAsp) replaces the first Asp residue in the calcium-binding motif of malacidin. Unlike the described modes of action (MOAs) of CDAs, which typically depolarize the bacterial cell membrane⁶ or binds to the undecaprenyl phosphate (C₅₅-P)⁷, malacidin A exerts bactericidal activity by binding to lipid II. As no cross resistance with vancomycin was observed, the interaction between malacidin A and lipid II should be different and comprehensively studied in the future. Moreover, bacterial resistance to malacidin A has not been observed under the laboratory resistance-inducing conditions.

Malacidin A consists of a cyclic nonapeptide anchored with an amino-acid-linked unsaturated C9-lipid (Figure 1), as well as one d-Val and five non-proteinogenic amino acids, including β-methyl-aspartic acid at position 1 and 8 (MeAsp-1 and −8), β-methyl-diaminopropionic acid-2 (MeDap-2), β-hydroxyl-aspartic acid-5 (HyAsp-5) and γ-methyl-proline-10 (MePro-10). Based on NMR characterization, the configuration of the lipid tail was readily determined as 2E,4Z-methyl-nonadienoic acid. Combining NMR technique with the Marfey’s analysis, configurations of MePro-10 and HyAsp-5 were suggested as l-(4R)-MePro and l-HyAsp-5. Bioinformatically, MeDap-2, MeAsp-1, and MeAsp-8 were predicted as l-MeDap-2, l-(3S)-MeAsp-1 and d-(3S)-MeAsp-8 respectively.⁴ However, the β-stereochemistry of l-MeDap-2 and l-HyAsp-5 remained uncertain and needed further investigation. This left four possible diastereomeric structures: 1a, 1b, 1c and 1d. If a chemical synthetic route could be established, malacidin A’s structure could be confirmed by NMR comparison to 1a-1d. The final route would also provide opportunities for structure-activity relationship (SAR) studies and the development of the next generation of CDAs. To this end, we initiated a program on the total synthesis of malacidin A.

Figure 1.

Structure of malacidin A. Residues with asterisk are predicted by bioinformatics.

Prior to the total synthesis attempt, building blocks of non-proteinogenic amino acids with proper protecting groups were required. Although there are several strategies reported to construct HyAsp, few of them could be conveniently utilized to prepare suitable building blocks for this synthetic route. This includes the tedious route starting from tartaric acid⁸, non-suitable side chain protection in the masked acyl cyanide reaction⁹ and difficult observation of trans product in the Sharpless aminohydroxylation reaction¹⁰. Inspired by the Sardina’s work¹¹, herein we developed a concise synthetic route for selectively preparing 2a or 2b in five steps with 37~40% overall yield (Scheme 1.).

Scheme 1.

Concise Syntheses of 2a and 2b.

Starting from a commercially available building block of l-aspartic acid 7a, its carboxylic acid was protected by benzylation. After removal of Fmoc group, 8a was prepared by tritylation of the amine to shield the α-proton from epimerization in later steps. Then with LiHMDS at −78°C, stereo-selective hydroxylation of the enolate from 8a was performed by (+)-(camphorylsulfonyl)oxaziridine (CSO) to afford 9a. The d.r. was up to 10:1 in this oxidation and both diastereomers could be well separated by column chromatography. Likewise, using the oxaziridine with the opposite chirality, i.e., (−)-CSO, the stereochemistry of 8a was successfully reverted to produce 9b at −30°C with 1: 10 d.r.. The configurations of the β-position from 9a and 9b were confirmed through determining the coupling constants and nuclear overhauser effect (NOE) of the α and β-protons in their cyclic carbamate derivatives 10a and 10b. While 10a exhibited a weak NOE signal (J = 3.1 Hz), an apparent NOE and larger J value (9.2 Hz) was observed from 10b, indicative of a trans-proton pair in 10a (i.e. 2S, 3S) and cis-proton pair in 10b (i.e. 2S, 3R). Next, 9a or b was hydrogenated by catalytic palladium to remove the trityl and benzyl groups at the same time, followed by Boc installation in one pot to afford 2a or 2b, which were ready for Fmoc-based SPPS.

This synthetic route provided the following two advantages. Firstly, the two carboxylic groups were differentiated early on. Secondly, the protecting groups of 9 could be easily modified because trityl, benzyl and tert-butyl group could be removed under mild conditions with orthogonality. Through this facile and stereoselective introduction of the β-hydroxy group to aspartic acid, various protection strategies could be adapted to this key intermediate 9 to fulfill the requirements of SPPS.

As for the MeDap building block (Scheme 2A), 13 was first prepared from l-Thr or l-allo-Thr (11a or 11b) by the Mitsunobu reaction of 12a or 12b to introduce an azido group¹². After conversion of the trityl to an Fmoc group, followed by hydrolysis of the methyl ester, 3a or 3b was produced in high yield, respectively.

Scheme 2.

Synthesis of 3a, b; 4a, b; 5 and 6.

The scaffold of MeAsp 14 was generated from trityl protected 8a or 8b by methylation (Scheme 2B). Although there is no stereo-selectivity at this step, each diastereomer of 15 could be separated by column chromatography after hydrogenation of methylated mixtures¹³. Reinstallation of the Fmoc group gave high yield of 4a and 4b, which could serve directly as the building blocks for MeAsp-1 and −8.

Next, the synthetic route of the (2S, 4R)-methyl proline, building block 5, was developed using a combination of Li’s¹⁴ and Pedregal’s method¹⁵ (Scheme 2C). Briefly, the pyroglutamic acid 16 was protected as 17 and then methylated at its γ-position to generate the (2S, 4R) intermediate 18. After that, it underwent a chemoselective amide reduction and manipulation of protecting groups to afford compound 5.

For the lipid building block (Scheme 2D), the trans-iodoalkene 20 generated from 19 was coupled with the terminal alkyne of 5-methyl-1-hexyne through the Sonogashira cross coupling conditions to obtain 21. This trans-conformation maintaining intermediate was selectively hydrogenated using the Lindlar catalyst, followed by hydrolysis of methyl ester to give rise to the 2E, 4Z-lipid 6 (see SI for more details of building blocks syntheses).

With all the building blocks in hand, we proceeded towards the total synthesis of malacidin A. Our first attempt used solid phase peptide synthesis followed by in-solution peptide cyclization. In this route, the Gly7-MeAsp8 peptide bond (Figure 3, Scheme S6) was disconnected as the cyclization site of the linear peptide, as cyclizing the C-terminal glycine should avoid epimerization at the C-terminus due to the achirality of glycine. Unfortunately, during Fmoc-SPPS we observed significant aspartimide formation after the coupling of HyAsp as the third residue. All attempts to vary building blocks, deprotection methods or coupling conditions failed to resolve this issue. Furthermore, no cyclized product was found when the linear peptide (S7) was activated by PyBop with DIEA under highly dilute concentrations. Although changing coupling reagents offered trace amounts of product 1, by UPLC it eluted as a broad peak with overlapping aspartimide and related epimerization peaks, making it extremely difficult to purify this cyclized product.

Figure 3.

Three cyclization strategies of malacidin A.

Considering the facile formation of aspartimide at the junction of HyAsp5-Asp6 and suspected hindrance in cyclization, an on-resin cyclization synthetic plan was attempted in which the side chain of the HyAsp building block was directly anchored to the resin (Figure 3, Scheme S7). In theory, anchoring the side chain of HyAsp to resin would avoid aspartimide formation as no other residue would be linked to the C-terminus until the macrocyclization step. As outlined, peptide elongation proceeded smoothly without any trace of aspartimide. Unfortunately, the cyclization precursor S12 was found to fall off the resin under the C-terminus carboxylic acid allyl deprotection conditions. A similar problem has been reported previously when using Asp and 2-Cl-trityl resin¹⁶. More importantly, the cyclization of prematurely released peptide would be non-regioselective due to the presence of both the C-terminal and side-chain carboxylic acids and it would lose the pseudo-high-dilution advantage of on-resin cyclization.

In the end, we turned to an unconventional peptide cyclization synthetic route that relied on chemical ligation¹⁷. We envisioned that linear peptide 28 with a C-terminal salicylaldehyde ester and an N-terminal β-hydroxyl-aspartic acid might undergo a similar reaction as Ser/Thr ligation^16-18. The advantages of this route, if successful, include avoiding epimerization during the peptide cyclization and using an unprotected linear peptide precursor. However, it remains to explore whether the hydroxylated amino acid could effectively proceed under the Ser/Thr ligation conditions. To this end, the Lys4-HyAsp5 linkage (marked as a red arrow in Figure 3) was chosen as the cyclization site.

As shown in Scheme 3, the crude peptide 26 with side-chain protection was prepared from d-Fmoc-Val-COOH linked trityl resin 22 via on-resin reduction of 23 to 24 and cleavage of 25 from resin. To enable the HyAsp ligation, the linear peptide with C-terminal salicylaldehyde ester 28 was prepared. Under Sakakibara conditions (EDC, HOOBt, TFE/CHCl₃), the side-chain-protected Lys salicylaldehyde semicarbazone 27 was coupled to peptide 26, followed by global deprotection in the presence of pyruvic acid. HPLC purification afforded the side chain unprotected peptide salicylaldehyde ester 28. With this ligation precursor in hand, different conditions for cyclization were screened. Gladly, the linear salicylaldehyde ester cyclized smoothly within a few hours in pyridine/acetic acid buffer at a concentration of 3 mM. Detection of the oxazolidine 29 by UPLC before acid addition to the ligation mixture suggests that the linear peptide underwent HyAsp ligation mediated cyclization by a mechanism similar to Ser/Thr ligation. After acidolysis and deprotection of the Fmoc group in one pot, HPLC purification gave malacidin A candidate 1a. Ultimately, this work established a robust synthetic route to the total synthesis of malacidin A.

Scheme 3. Assembly of 1a-1d. ^a.

^aConditions: (a) _D-Fmoc-Val-OH, DIEA, DCM, 2 h; (b) (i) 20% Piperidine in DMF, 10 min; (ii) Fmoc-AA-OH, HATU, DIEA, DMF, 1-2 h; (c) SnCl₂/PhSH/DIEA, DMF, 1 h; (d) HOAc/TFE/DCM, 2 h; (e) (i) EDC, HOOBt, CHCl₃/TFE, 6 h; (ii) Pyruvic acid, H₂O/TFA, 3 h; (iii) HPLC purification, 10% yield; (f) Acetic acid/pyridine, 3-6 h; (g) (i) TFA/TIPS/H₂O, 15min; (ii) 10% DEA in ACN, 1 h; (iii) HPLC purification, 50-60% yield.

The other three possible diastereomeric structures, 1b-1d, were synthesized following the same route. After careful comparison and analysis of spectral data from all four synthetic compounds, fortunately, 1b was found to be nearly identical to the reported NMR spectra for malacidin A (Table S1). Based on this analysis, we can conclude that 1b has the same chemical structure as malacidin A. Thus, the absolute stereocenters of its five non-proteinogenic amino acid residues are suggested as (2S, 3S)-MeAsp1, (2S, 3R)-MeDap2, (2S, 3R)-HyAsp5, (2R, 3S)-MeAsp8 and (2S, 4R)-MePro. To confirm these results, we carried out an advanced Marfey’s analysis on the natural malacidin A (Figures S10-S12). This analysis confirmed the stereochemical predictions suggested by our total synthesis studies. Our synthetic malacidin showed the same calcium dependent antibacterial activity against MRSA as the natural product. Interestingly, even minor changes in the structure, such as that seen in the (2S, 3S)-MeDap2 analog, abrogated antibiosis.

In summary, the total syntheses of malacidin A and its diastereomeric analogues was completed, from which we established its absolute configuration. Our synthetic route involved the HyAsp-mediated ligation as a key step for peptide cyclization. Compared with classical strategies that might be used to construct this natural product, the HyAsp ligation enables chemoselective and efficient macrolactamization free of epimerization and aspartimide formation during the precursor preparation. It is worth noting that HyAsp-mediated peptide cyclization was conducted at 3 mM concentration without observed dimerization and polymerization. Indeed, this outcome opens up new possibilities for the synthesis of cyclic peptides with HyAsp or similarly hydroxylated amino acids. Through such a practical protocol, analogues of malacidin A can be obtained routinely, providing the foundation for its SAR study.

Figure 2.

Three cyclization strategies of malacidin A.