Chemical exchanges between multilateral symbionts

Abstract

Herein is a report on the molecular exchange occurring between multilateral symbiosis partners – a tit-for-tat exchange that led to the characterization of two new metabolites, conocandin B (fungal-derived) and dentigerumycin F (bacterial-derived). The structures were determined by NMR, mass spectrometry, genomic analysis, and chemical derivatizations. Conocandin B exhibits antimicrobial activity against both the bacterial symbionts of fungus-growing ant and human pathogenic strains by selectively inhibiting FabH, thus disrupting fatty acid biosynthesis.

Graphical Abstract

Chemists have studied naturally occurring small molecules before the concepts of molecule or function existed, and the studies have grown along with the science. These molecules fascinate chemists for their complex architectures; however, the study of what these molecules do for their producers have been less frequently explored.

A particularly informative system is that of the chemical ecology that regulates the microbial symbionts found in the nests of fungus-growing ants.^1-6 These nests house fungal gardens in which plant material is brought to the nest by foraging ants deposited and converted into the food that feeds the colony.^7-9 This beneficial ant-fungus relationship is threatened by specialized fungal pathogens that can overwhelm the gardens and eliminate the nest.^10,11 In response to this existential threat, the ants have evolved specialized crypts on their exoskeleton to harbor symbiotic bacteria that produce chemical defenses to control the fungal pathogens.^12,13 In return for producing chemical defenses, the ant host provides the symbiotic bacteria with essential nutrients.¹⁴ Our previous studies focused on the molecular diversity of the antifungal agents produced by the bacterial symbionts,^6,15-17 but in this study, we sought to examine how specialization is reflected in the small molecule exchanges between the fungal pathogen and bacterial symbiont.

We focused, after some experimentation, on a single ant nest, a Trachymyrmex sp. nest designated ICBG2300 that was collected in the Adolpho Ducke Forest Reserve in the Amazon region of Brazil in 2017. The bacterial isolate from the nest, Pseudonocardia sp. ICBG1122 (Ps1122), is matched with the pathogenic fungus, Escovopsis sp. ICBG729 (Es729) from the same nest (Table S2,3). In order to study the symbiosis in a format that was both faithful to the natural relationship but amenable to efficient bioassay-guided fractionation, we used a trans-well assay system (Fig. S1, Table S4).¹⁸ Our study revealed two previously unreported molecules: dentigerumycin F (1) from Ps1122, which inhibits the growth of Es729; and conocandin B (2) from Es729, which inhibits the growth of Ps1122. Further, conocandin B induces the production of dentigerumycin F, and dentigerumycin F induces the production of conocandin B – a fully articulated example of a molecular tit-for-tat strategy (Fig. S2-4).

When Ps1122 was exposed to the supernatant of Es729, one particular Ps1122 metabolite was strongly upregulated. Analysis of this peak using high-resolution mass spectrometry (HRMS), MS/MS fragmentation patterns, and UV spectral data revealed that the induced metabolite was a member of the dentigerumycin family, a highly modified cyclic peptide that exhibits antifungal properties.² The dentigerumycins have been isolated from ant-associated Pseudonocardia sp. strains throughout the neotropics – Costa Rica, Panama, and Brazil – and they had been identified as the chemical defenses that protected the ant nests from Escovopsis sp. infections.^5,6 The molecular formula of the induced metabolite, as determined by HR-ESI-qTOFMS data ([M+H]⁺ m/z 868.4786), did not match any of the previously reported dentigerumycin analogs, but only differed by the loss of one methylene unit from dentigerumycin. NMR analysis (COSY, TOCSY, HSQC, and HMBC) confirmed this suggestion by identifying dentigerumycin F (1) differing from dentigerumycin by the loss of an aliphatic chain methylene (Fig. 1, Table S5,6 and Fig. S7-12). The results from the ROESY (Fig. S13), J-based (Fig. S14), and Marfey’s analyses, along with the identified biosynthetic gene cluster (BGC) exhibiting 99% similarity to known dentigerumycin BGCs (Fig. S15 and Table S7), strongly supports both the amino acid sequence and assigned stereochemistry to be consistent with dentigerumycin. Induced concentrations of 1 reach 46 μM when grown with the spent supernatant of Es729 and this is well above the concentration needed to inhibit the growth of Es729 (MIC - 8 μg/mL; Fig. S5).

Figure 1.

Chemical structures of the isolated dentigerumycins and conocandins.

Addition of dentigerumycin F (1) to Es729 cultures at a sub-lethal concentration (5 μg/mL) significantly upregulated the production of two metabolites that could be visualized by their extracted molecular ions (Fig. S6). The two compounds are structural isomers with molecular ions of 311.2235 amu. As the two compounds could not be identified using various databases, a more traditional mass-guided purification scheme was initiated. Conocandin B (2) was isolated as amorphous powder having molecular formula of C₁₈H₃₀O₄ determined by HR-ESI-qTOFMS analysis. ¹H NMR data for the compound revealed an allylic methyl, three oxygenated methines, two of which had the characteristic chemical shifts of an epoxide (δ_H 3.52 and 2.65), an vinylic hydrogen (δ_H 5.14), an exomethylene (δ_H 5.80 and 5.31), and a large aliphatic CH₂ envelope (δ_H 1.58-1.30). The ¹³C NMR spectrum displayed corresponding features and, in addition, an ester carbonyl (δ_C 174.4), along with two quaternary olefinic carbons (δ_C 146.4 and 136.6).

Comprehensive analysis of 2D NMR spectra (COSY, HSQC, HMBC, and ROESY) of 2 established the backbone to consist of an 18-membered fatty acid, but with a high degree of post-modifications (Fig. 2a, Table S8, Fig. S16-21). Key COSY and HMBC correlations established two partial structures. COSY correlations from the oxygenated methine (H-14; δ_H 3.43) to several aliphatic methylenes – H-15 (δ_H 1.48 and 1.39), H-13 (δ_H 1.43 and 1.30), H-12 (δ_H 1.58 and 1.45), and H-11 (δ_H 2.01 and 1.98) – and the terminal methyl H-16 (δ_H 0.92), led to the assignment of the first partial structure as a hexane chain bearing a branch hydroxyl group. Additional COSY correlations between the triplet olefinic proton, H-9 (δ_H 5.14), and consecutive aliphatic protons from H-8 to H-5 (δ_H 1.72 and 1.70) resulted in the expansion of an unmodified fatty acid aliphatic chain. The characteristic features of an epoxide, H-3 (δ_H 3.52) and H-4 (δ_H 2.65), were connected to the aliphatic chain at C-5 by a COSY correlation between H-4 and H-5. Finally, the characteristic protons of an exomethylene, H-2’ (δ_H 5.80 and 5.31), exhibited a COSY correlation with H-3 and a HMBC correlation to both C-2 (δ_C 146.4) and C-1 (δ_C 174.4), thus establishing a second partial structure as a fatty acid chain composed of 10 carbons bearing an epoxide ring and an exomethylene. Connection of the two partial structures was possible using key HMBC correlations from the singlet methyl group H-10’ (δ_H 1.59) to the quaternary carbon C-10 (δ_C 136.6), C-9 (δ_C 126.3), and C-11 (δ_C 41.3), and completing the assignment of the planar structure of conocandin B (2) as shown in figure 1.

Figure 2. Structural characterization of conocandin B and C.

a. Key 2D NMR correlations for conocandin B and C (2 and 3). b. Δ_S-R values of MTPA ester (4 and 5) of conocandin B. c. Δ_S-R values of MTPA ester (6 and 7) of conocandin C.

The relative configuration of 2 was established using J-based methods, ROE analysis, and optical rotation data. The small ³J_HH coupling (2.0 Hz) between H-3 and H-4 suggests that the epoxide is of the trans configuration, and this was supported by the key ROESY correlation between the same protons. The absolute stereochemistry was assigned by comparing the [α]_D data of 2 to published data of related analogs, which established the assignment as 3S and 4S (See methods; Fig. S22).¹⁹ The final stereocenter, the isolated hydroxyl, posed much greater challenges. Using a modified Mosher’s method – dichloromethane as the solvent instead of pyridine – allowed for the successful derivatization of 2 while limiting degradation and the stereocenter was determined to be 14S (Fig. 2b, Fig. S29-32, and Table S10).²⁰ Thus, establishing the absolute stereoassignments of conocandin B as 3S, 4S, and 14S.

The remaining structural isomer had an identical UV spectrum, however based on ¹H NMR data it was evident that conocandin C (3) lacked both the epoxide and exomethylene functionality. Comprehensive analysis of the NMR spectroscopic data (Table S9 and Fig. S23-28) revealed that the epoxide protons were replaced by signals at δ_H 4.26 and 4.11 and the exomethylene signals were replaced with a doublet methyl. Furthermore, HMBC correlations from H-4 (δ_H 4.26) to C-1 (δ_C 181.6) established the presence of a heterocyclic lactone ring in 3. The hydroxyl group at C-14 in 2 has been oxidized to a ketone (δ_C 215.0), yielding the planar structure of 3 as shown in figure 1. The relative and absolute configuration of the heterocyclic lactone ring was determined using similar methods described above. The coupling between H-3 and H-4 was 5.5 Hz indicating a trans orientation and a key NOESY correlation between 2-Me and H-4 allowed for the relative assignment of the 5-membered lactone ring as 2R*, 3S*, and 4R* (Fig. 2a). Using the Mosher’s method²⁰ to derivatize the secondary hydroxy on C-3 allowed for the assignment of the absolute stereochemistry of 3 as 2R, 3S, and 4R (Fig. 2c, Fig. S33-36, and Table S11).

Initial spot-on-lawn inhibitory assays against Ps1122 revealed that only conocandin B (2) possessed significant antimicrobial activity, while conocandin C (3) did not produce a visible zone of inhibition at concentrations up to 100 μg/mL. This suggests that the reactive epoxide functionality is an essential feature for antibacterial activity. Conocandin B exhibited minimum inhibitory concentrations (MICs) ranging from 0.5 - 2 μg/mL (Fig. 3a) against four Pseudonocardia sp. strains that were isolated in Brazil and Panama. Compound 2 was also evaluated against various human bacterial pathogens, including gram positive bacteria (S. aureus, MRSA, and VRE) and gram negative bacteria (Klebsiella salmonella, Enterococcus coli, and Pseudomonas aeruginosa). Conocandin B possesses potent activity against gram positive pathogens with MICs between 0.5 - 10 μg/mL but was ineffective against gram negative pathogens (Fig. 3a and Fig. S37). Compound 2 was also evaluated against fungus-growing ant’s crop fungus and various fungal pathogens, however it did not exhibit any inhibitory activity against any of the fungal strains (Fig. S38).

Figure 3. Antibacterial activity and mechanism of action of conocandin B.

a. MIC values for conocandin B against Pseudonocardia strains and select human pathogens. Abbreviations are as follow: GP, gram positive; and GN, gram negative. b. MIC values for conocandin B against strains of S. aureus that overexpress individual fatty acid biosynthetic genes.

Conocandin B loosely resembles the structure of cerulenin, an epoxide-containing C-12 fatty acid antibiotic as a fungal metabolite, and amycomicin, an epoxy-isonitrile containing C-18 fatty acid with potent activity against S. aureus. Both of these metabolites are known fatty acid synthase inhibitors,^18,21 although cerulenin also inhibits sterol biosynthesis.²² Therefore, we suspect that 2 may have a similar target. Screening 2 against a panel of S. aureus strains each possessing overexpression of different steps within fatty acid biosynthesis suggested that it was also targeting 3-ketoacyl-[acyl-carrier-protein] synthase 3 (FabH) as its MIC was increased 5-fold (Fig. 3b).^23-26 FabH initiates bacterial fatty acid synthesis by condensing branched-chain acyl-CoA derived from amino acid metabolism with malonyl-ACP to form ketoacyl-ACP. To further establish FabH as a target, 2 was tested in a FabH enzyme radiochemical assay. In this assay, the activity of purified FabH from S. aureus and B. subtilis is determined by calculating how much [¹⁴C]-ketoacyl-ACP is formed. Conocandin B inhibited the activity of both the S. aureus FabH and B. subtilis FabH at IC₅₀ of 18.8 ± 2.5 μM and 53.5 ± 10.4 μM, respectively (Fig. S39).

Dentigerumycin F and conocandin B were each evaluated in in vivo mouse models of infections using C. albicans K1 and methicillin-resistant S. aureus, respectively. Dentigerumycin F exhibited significant decrease of fungal colony forming units (CFU) in the kidney in dose-dependent fashion when compared to untreated mice (Fig. S40). Conocandin B exhibited potent in vitro efficacy, but only showed a modest decrease in CFU loads in the kidney in the in vivo model. We believe this discrepancy is due to stability liabilities.

Previous studies for fungus-growing ant’s symbiosis have focused on antifungal agents derived from Pseudonocardia sp., including highly-modified cyclic peptide dentigerumycins,^2,5,6 glycosylated indolocarbazole rebeccamycin,³ and macrocyclic polyene selvamicin.⁴ However, very few studies have analyzed the metabolites produced by the specific parasitic fungal pathogen of the ant’s society.²⁷ In this study, we used the trans-well system to unravel the intimate and specialized chemical interaction occurring between Pseudonocardia sp. and Escovopsis sp. that shared the same nest. Activation of cryptic chemical defenses by replicating environmental conditions led to the identification of three new metabolites – dentigerumycin F, conocandin B and C – each possessing toxic activity against the producer of the other metabolite. Their reciprocal activities, reciprocal induction, and limited distribution of both molecules argue that both the metabolites, their regulation, and their activities have co-evolved over a lengthy interaction.