Total Synthesis of the Bacterial Diisonitrile Chalkophore SF2768

Abstract

Chalkophores are bacterial natural products that chelate and transport extracellular copper. The diisonitrile natural product SF2768 was first isolated from a Streptomyces species as an antifungal antibiotic and has more recently been characterized as a bacterial chalkophore and potential virulence factor. Herein, we report a modular synthesis of SF2768 and related acyclic analogues, allowing assignment of syn-stereochemistry across the central lactol ring. The copper-binding properties of these diisonitriles have also been studied.

Graphical Abstract

Copper is an essential element that is used as a cofactor for a variety of enzymes that are important for cell growth and survival.^1,2 Conversely, excess copper can lead to cellular damage due to the formation of reactive oxygen species. Accordingly, copper homeostasis is highly regulated. Bacteria use chalkophore natural products to chelate and transport extracellular copper into the cell, and to sequester copper that can be toxic at high concentrations due to formation of reactive oxygen species.^3–5 The best-studied chalkophores are members of methanobactin family, which are comprised of ribosomally-produced, post-translationally modified peptides^6,7 that chelate copper in 1:1 stoichiometry using two heterocycle–thioamide motifs. Recently, He and coworkers identified the diisonitrile natural product SF2768 (1, Figure 1) as a new class of chalkophore that is structurally unrelated to the methanobactins and produced by a non-ribosomal peptide synthetase (NRPS^8,9) pathway.¹⁰ SF2768 was produced from an NRPS gene cluster from Streptomyces thioluteus expressed heterologously in S. lividans. It was shown to form copper complexes selectively when various metals were included in the fermentation broth, to reduce Cu(II) to Cu(I) upon binding, and to bind Cu(I) in 2:1 stoichiometry. Treatment of wild-type S. thioluteus with Cu(SF2768)₂ led to increased intracellular copper concentration. The relevance of putative chalkophore transporters was also demonstrated in the S. lividans model.

Figure 1.

Structures of diisonitrile natural products.

SF2768 was originally isolated from Streptomyces sp. SF2768 as an antibacterial and antifungal natural product.¹¹ Subsequently, SF2768 was identified as a cryptic antibiotic whose production in a S. griseorubiginosus strain was triggered by monensin.¹² The related acyclic, ornithine-based natural product SF2369 (2) had also been identified earlier as an antifungal from Actinomadura sp. SF2369.¹³ Notably, two lysine-based congeners (3,4) have been identified recently by Zhang and coworkers via heterologous expression of a biosynthetic gene cluster from S. coeruleorubidus having homology to a cryptic virulence factor gene cluster in Mycobacterium tuberculosis.^14,15 Related diisonitriles with longer side chains (not shown) were also produced by heterologous expression of the homologous gene cluster from M. marinum. Interestingly, He and coworkers found that the corresponding copper complexes of 3 and 4 did not significantly increase intracellular copper concentrations in S. thioluteus.¹⁰ Moreover, Zhang and coworkers showed that deletion of the entire biosynthetic gene cluster in M. marinum decreased the intracellular concentration of Zn but not Cu.¹⁴ Thus, it remains to be determined if these diisonitriles are generally involved in the regulation of Cu or other metals as well. To enable further chemical, structural, and biological studies, we report herein a modular synthesis of SF2768 and linear congeners 3 and 4. This work establishes the absolute and relative stereochemistry of the central lactol motif. We also report preliminary studies of the Cu- and Zn-binding properties of these diisonitriles.

At the outset of our studies, the absolute and relative stereochemistry of the central lactol moiety of SF2768 had not been established. However, Zhang and coworkers had determined that the NRPS adenylation domains from S. coeruleorubidus and M. marinum were selective for l-lysine over d-lysine, leading them to assign a 2l-configuration in linear congeners 3 and 4.¹⁴ Thus, we presumed in our retrosynthetic analysis that the central lactol motif of SF2768 is also derived from l-lysine (Figure 2). As the relative configuration at C5 was unknown, we pursued the synthesis of both diastereomers. It should be noted that SF2768 was isolated as a mixture of α and β anomers.¹¹ Zhang and coworkers had also determined an R-configuration for the β-isocyanobutyrate side chains of linear analogue 4,¹⁴ and we assumed the same for SF2768, to be derived from N-hydroxysuccinimide ester 5. We envisioned that the central lactol motif 6 could be accessed from a protected 5-hydroxylysine intermediate 7,^16,17 which could be synthesized via azido alcohol 8 as a mixture of syn and anti diastereomers by epoxidation and ring opening of butenylglycine precursor 9.¹⁸ The 2l-configuration would be set using Williams’ diphenyloxazinone auxilliary-based method.^19–21 We envisioned that the linear diisonitriles 3 and 4 could be synthesized analogously by acylation of a protected l-lysine derivative (not shown) with NHS ester 5.

Figure 2.

Retrosynthesis of SF2768. The approach assumes the lactol is derived from l-lysine, by analogy to diisonitriles 3 and 4.

Synthesis of the side-chain NHS ester 5 was readily achieved in multi-gram quantities from R-3-aminobutyrate (10) by formylation²² to formamide 11, activation to NHS-ester 12, and conversion of the formamide to the isonitrile^23,24 in 5 (Figure 3). The key δ-hydroxylysine derivative 8 was synthesized essentially as previously described,^18,25 as a separable 1:1 diastereomeric mixture, providing convenient access to both diastereomers at what would eventually be the C5 position of SF2768.

Figure 3.

Synthesis of NHS ester 5.

Next, we investigated conversion of (2l,5S)-8 to the key lactol intermediate 6 (Figure 4). Attempted direct hydrogenation–hydrogenolysis of 8 led only to azide reduction. It has been reported that removal of the Boc group facilitates hydrogenolysis of the auxilliary.²⁶ Thus, treatment of 8 with TFA afforded oxazinone 13, which then underwent hydrogenolysis–hydrogenation in MeOH to provide diamino ester 14 as its bis·TFA salt.²⁷ Careful reaction monitoring minimized formation of a caprolactam byproduct (≈20%). The crude material was carried directly onto Boc protection to afford isolable bis(Boc) diamino ester 7. Lactonization with PPTS afforded 15.²⁸ Initial attempts to acylate the corresponding Boc-deprotected lactone resulted in caprolactam formation. Thus, DIBAL reduction²⁹ of 15 afforded lactol 16 as a mixture of anomers. Finally, TFA deprotection generated the diamino lactol intermediate 6, which underwent bisacylation with NHS ester 5 to afford the syn-substituted diisonitrile (2l,5S)-1. The lactol exhibited a 1.2:1 α/β ratio in DMSO-d₆, but an inverted 1:1.2 α/β ratio in D₂O. Overall, the synthesis provided the diisonitrile in 7 steps (longest linear sequence) and 16.3% yield from known azidoalcohol 8. We also synthesized the corresponding anti diastereomer (2l,5R)-1 by the analogous route from diastereomeric azidoalcohol (2l,5R)-8.²⁵

Figure 4.

Synthesis of SF2768. The epimer (2l,5R)-1 was also synthesized by the analogous route from (2l,5R)-8 (not shown).²⁵ The (2l,5S) configuration was assigned by comparison to the ¹H-NMR and ¹³C-NMR spectra of the natural product.²⁵

Comparison of the ¹H-NMR and ¹³C-NMR spectra of the synthetic syn-diastereomer (2l,5S)-1 and anti-diastereomer (2l,5R)-1 to those reported for the natural product SF2768 isolated from Streptomyces sp. SF2768¹¹ and S. griseorubiginosus strain 574,¹² as well as from S. thioluteus via heterologous production in S. lividans¹⁰ clearly established the syn-diastereomer as corresponding to the natural product stereochemistry (Figure 5).²⁵ The matching NMR spectra of synthetic (2l,5S)-1 and the natural products also support assignment of the l-configuration to the lysine-derived lactol, although the diastereomeric d-lysine-derived diisonitrile was not synthesized for direct comparison.

Figure 5.

Selected regions of ¹H-NMR (left) and ¹³C-NMR spectra (right) of natural product SF2768 (top),¹² synthetic syn-diastereomer (2l,5S)-1 (middle), and synthetic anti-diastereomer (2l,5R)-1 (bottom) in D₂O.²⁵

We also synthesized the corresponding acyclic diisonitriles by bisacylation of l-lysinol acetate³⁰ with NHS ester 5 to generate diisonitrile acetate 4 (i-Pr₂NEt, THF, 82%), followed by deacetylation to afford diisonitrile alcohol 3 (K₂CO₃, MeOH, H₂O, 90%).²⁵ The ¹H- and ¹³C-NMR spectra matched those reported for the natural products produced by heterologous expression.^10,14,25

We next investigated the copper-binding properties of these diisonitriles. Previously, He and coworkers reported that SF2768 binds Cu(I) and Cu(II) selectively over 14 other metals, in 2:1 stoichiometry based on HR-ESI-MS-based titration experiments.¹⁰ Notably, the MS results indicate that SF2768 binding to Cu(II) results in in situ reduction to Cu(I). Analogous Cu(II) reduction is also known for the methanobactins, although the mechanism is not understood in either case.^3–5,31,32 To complement these studies, we carried out NMR titrations of linear diisonitrile acetate 4, SF2768 (1), and its unnatural anti-diastereomer (2l,5R)-1, using Cu(I) as it is diamagnetic. The titration with linear diisonitrile acetate 4 revealed clear shifts of the C3´-proton adjacent to the isonitrile motif, as well as the diastereotopic C6-protons on the lysinol chain, with a maximal change at 2:1 ligand-to-copper stoichiometry (Figure 6). Disappearance of the isonitrile ¹³C signal was also observed, consistent with Cu-quadrupole coupling.^25,33 A characteristic IR shift caused by the inductive effect of the metal³⁴ was observed for the isonitrile group from 2414 cm^–1 to 2184 cm⁻¹.²⁵ Taken together, these results are consistent with direct Cu binding via the isonitrile motifs of diisonitrile acetate 4, resulting in conformational changes throughout the rest of the lysinol backbone. Intepretation of NMR spectra of the Cu(I) complexes with SF2768 or its unnatural anti-diastereomer were confounded by multiple overlapping peaks, as well as significant peak broadening, possibly arising from the anomeric mixture as well as increased spectral distinctions between diastereomers at the Cu center. However, similar downfield shifts of the H3´ resonances were observed, consistent with similar binding modes.²⁵

Figure 6.

(a) Formation of 2:1 complex of diisonitrile acetate 4 with Cu(I). (b) ¹H-NMR titration (DMSO-d₆) of diisonitrile acetate 4 with varying amounts of Cu(I).

To investigate further the Cu binding affinity of these diisonitrile compounds, we attempted competitive UV-Vis titration with disodium bathocuproinedisulfonic acid (BCS). BCS is known to bind Cu(I) as a stable 2:1 complex Cu(BCS)₂^3–, with an overall two-step association constant of β₂ = 10^19.8 M^–2 (λ_max = 483 nm, ε = 13300 M⁻¹ cm⁻¹).³⁵ However, no formation of the Cu(BCS)₂ complex was observed (A₄₈₃) when the Cu(I) complex of diisonitrile acetate 4 (40 μM in 25 mM HEPES buffer) was treated with up to a 100-fold excess of BCS (Figure 7a).²⁵ Conversely, treatment of the preformed Cu(BCS)₂ complex with diisonitrile acetate 4 led to a completely linear (hockey stick-shaped) decrease in absorbance at 483 nm, reaching A = 0 when ≈1 equiv of 4 had been added (Figure 7b).²⁵ Taken together, these data suggest that the Cu(I) binding affinity of the diisonitrile acetate 4 is much stronger than that of BCS, but preclude calculation of that binding affinity because an equilibrium could not be established. Similar results were obtained in titrations with SF2768 or its unnatural anti-diastereomer.²⁵

Figure 7.

(a) UV-Vis titration of Cu(4)₂ (40 μM) by up to 100 equiv. BCS, indicating no formation of Cu(BCS)₂ (λ_max = 483 nm). (b) UV-Vis titration of Cu(BCS)₂ (40 μM) by diisonitrile acetate 4, indicating complete competition at 1 equiv.²⁵

Cellular studies of Zhang and coworkers¹⁴ suggest that corresponding diisonitriles in M. marinum may be involved in Zn rather than Cu homeostasis. To assess this possibility, we attempted to form the Zn(II) complex with diisonitrile acetate 4 (1 equiv ZnBr₂, 1:1 CH₂Cl₂/THF).²⁵ Both free diisonitrile 4 and the corresponding 1:1 Zn complex were detected by ESI-MS analysis, but not the 2:1 Zn complex. Only the free diisonitrile could be readily assigned by ¹H-NMR analysis due to peak broadening. Taken together, these results suggest that diisonitrile acetate 4 complexes with Zn(II) much more weakly than with Cu(I). However, because the complete structures of the M. marinum natural products have not yet been determined, it is still possible that those compounds may play a role in Zn homeostasis.

In conclusion, we have developed a modular synthesis of the diisonitrile natural product SF2768 as well as two acyclic analogues (3, 4). We have determined that the central lactol motif of SF2768 has syn stereochemistry, based on comparison of NMR spectra observed for the corresponding anti-diastereomer and reported previously for the natural product. ¹H NMR titration experiments confirm that diisonitrile acetate 4 binds Cu(I) in a 2:1 ratio and, in conjunction with IR analysis, are consistent with direct copper binding by the isonitrile motifs. This work sets the stage for further evaluation of the physiologic roles of these novel metal chelators.