Bacillibactins E and F from a Marine Sponge-Associated Bacillus sp.

Abstract

Chemical investigation of a marine sponge-associated Bacillus sp. led to the discovery of bacillibactins E and F (1 and 2). Despite containing the well-established cyclic triester core of iron-binding natural products such as enterobactin, bacillibactins E and F (1 and 2) are the first bacterial siderophores that contain nicotinic and benzoic acid moieties. The structures of the new compounds, including their absolute configurations, were determined by extensive spectroscopic analyses and Marfey’s method. A plausible biosynthetic pathway to 1 and 2 is proposed; this route bears great similarity to other previously established bacillibactin-like pathways but appears to differentiate itself by a promiscuous DhbE, which likely installs the nicotinic moiety of 1 and the benzoic acid group of 2.

Graphical Abstract

Iron is an essential element for all life forms on Earth. Although it is the fourth most abundant element in the Earth’s crust, iron cations are found in diminishingly low concentrations in marine environments, a result, in part, of generally poor aqueous solubility.¹ Consequentially, marine organisms have evolved a number of strategies by which to acquire iron despite its limited availability in marine settings. Microbial production of iron-chelating agents, known as siderophores, represents one of the most well-studied strategies for microbial iron acquisition.^2–4 Siderophores are generally small molecules (500 to 1500 Da in size), and, to date, over 500 siderophores have been identified predominantly from bacteria,^5,6 fungi,⁷ and cyanobacteria.⁸ Their novel structural features⁴ and promising medicinal applications,⁹ as well as their ecological significance,⁶ make siderophores attractive targets for natural product chemists.^10–12 Herein, we report two new members of the well-known siderophore class, the bacillibactins, from a marine sponge (Cinachyrella apion)-associated Bacillus sp. (strain WMMC1349).

Marine-derived Bacilli represent important sources of biologically active and/or structurally fascinating natural products including polyketides, peptides, macrolactones, and isocoumarins.¹³ Notably, many of these metabolites function as siderophores by virtue of their Fe-chelating 2,3-dihydroxybenzoic acid (DHB) moieties.^14,15 DHB represents a vital structural feature of many amino acid-based siderophores; examples include Bacilli-derived enterobactin (2,3-DHB-^LSer)₃, paenibactin (2,3-DHB-^LAla-^LThr)₃, griseobactin (2,3-DHB-^LArg-^LThr)₃, many other cyclic and acyclic nonribosomal peptide synthetase (NRPS) products.¹⁶ Notably, there have been no reports of bacillibactins containing heterocyclic catecholate residues. The pyridinyl moiety of 1 is, thus, particularly noteworthy, as is the simple benzoate (instead of DHB) seen in compound 2.

The scaffold, and its presumptive biosynthetic origin, for both 1 and 2 shares strong similarity to those previously established for bacillibactin.¹⁷ Briefly, biosynthesis of bacillibactin (3) begins with production of 2,3-DHB by isochorismate synthase DhbC, bifunctional isochorismatase/aryl carrier protein DhbB, and 2,3-dihydro-2,3-dihydroxybenzoate dehydrogenase DhbA. This is then activated via conversion to an aryl-adenylate by the stand-alone adenylation domain DhbE and tethered as a thioester to the 4′-phosphopantetheinyl group of bifunctional isochorismatase/aryl carrier protein DhbB. Glycine and threonine are similarly activated and tethered to the first and second modules of the NRPS DhbF, respectively. The condensation domains of DhbF then catalyze amide bond formation to yield DHB-Gly-Thr, which is then transferred to the thioesterase domain of DhbF. By two additional iterations of this process, DHB-Gly-Thr is effectively trimerized and compound 3 is liberated from the NRPS via cyclization to install the 12-membered triester core. Structurally similar to the well-known bacillibactin 3, other newly discovered bacillibactins include bacillibactins B and C¹⁴ and linbacilibactins A–C¹⁵ from marine Bacilli.

In pursuing novel secondary metabolites from marine-animal-associated bacteria,^18,19 the well-established workflow for siderophore discovery in our lab^20,21 was applied to the production, isolation, and identification of putative bacillibactins from Bacillus sp. WMMC1349. Two iron-limited media (ASW-A and RAM-2) were used for siderophore production. The acetone eluents of Diaion HP-20 resin were subjected to LC-MS analyses. As tracked by LC-MS, the known metabolite bacillibactin (C₃₉H₄₂N₆O₁₈, m/z 883.2634, [M + H]⁺, calcd 883.2628, Supporting Information, Figure S2) and two related metabolites with similar MS profiles were found only in the iron-limited RAM-2 media extracts.

Subsequently, guided by the Chrome Azurol S (CAS) assay^20–22 and LC-MS, the CHCl₃-soluble portion of the HP-20 acetone extract (harvested from 10 L of iron-limited RAM-2 media) was chromatographed sequentially over Sephadex LH-20 and then RP-HPLC to yield pure bacillibactins 1 (2.5 mg), 2 (10.2 mg), and 3 (22.1 mg); analytically pure 1–3 each retained iron-binding activity as confirmed by another round of CAS assays (Figure S23). The known compound was readily identified as bacillibactin (3) by analyzing its 1D NMR spectroscopic data and HRESIMS data.^{14 1}H NMR spectra for new compounds 1 and 2 bore great similarity to those of 3 (methyl groups and threonine/glycine proton signals), whereas significantly different signals could be seen in the aromatic region consistent with a cyclic peptide core and variations in the catecholate residues of 1–3.

Bacillibactin E (1) was obtained as an optically active white powder. The HRESIMS analysis was consistent with the molecular formula of 1 as C₃₈H₄₁N₇O₁₆ (m/z 852.2675 [M + H]⁺), requiring 22 degrees of unsaturation. The ^13C resonances from δ_C 167.7 to δ_C 172.6 indicated the presence of amide and ester carbonyls. Three methyl resonances at δ_H 1.29 (H₃-4/4′/4″, d, J = 6.6 Hz), glycine resonances (H₂-6, δ_H 3.92, m; δ_H 4.42, m; H₂-6′/6″, δ_H 3.99, m; δ_H 4.42, m), and threonine resonances (H-2/2′/2″, δ_H 4.50, m; H-3/3′/3″, δ_H 5.51, m) strongly indicated the presence of the same cyclic peptide core found in bacillibactin (3); this was verified by COSY correlations from H-2/2′/2″ to H₃-4/4′/4″ through H-3/3′/3″ and the HMBC correlations from H-2/2′/2″ and H₂-6/6′/6″ to C-5/5′/5″ and H-3/3′/3″ to C-1″/1/1′. The two typical DHB moieties in parts B and C shown in Figure 1 were recognized by the identical proton and carbon resonances compared to 3. Furthermore, the existence of a nicotinic acid moiety in part A (Figure 1) was determined by the clear COSY correlations of H-11 (δ_H 8.66, d, J = 4.8 Hz)/H-12 (δ_H 7.50 dd, J = 8.4, 4.8 Hz)/H-13 (δ_H 8.29, d, J = 8.4 Hz) and key HMBC correlations from H-9 (δ_H 9.03, s)/H-13 to C-7 (δ_C 173.0, C) and from H-9/H-12 to C-8 (δ_C 130.9, C) along with the HMBC correlation from H-9 to C-11 (δ_C 152.5, CH). The connections between the aromatic rings (nicotinic acid and two DHB moieties) and the peptide core were established by HMBC correlations from H₂-6 to C-7 and H₂-6′/6″ to C-7′/7″, respectively, enabling assignment of the 2D structure of 1. Finally, the 2D structure of 1 was consistent with the MS/MS analysis and the observation of daughter ions at m/z 716, 659, and 690 (Figure S13).

Figure 1.

COSY and key HMBC correlations for compounds 1 and 2.

Bacillibactin F (2) was isolated as a white amorphous powder. The molecular formula of 2 was identified as C₃₉H₄₂N₆O₁₆ (m/z 851.2731 [M + H]⁺) on the basis of HRESIMS data, one mass unit less than that of the co-occurring bacillibactin E (1), indicating 22 degrees of unsaturation. Further MS/MS analysis indicated the only significant difference between 1 and 2 to be in part A (Figure 1). More specifically, the typical daughter ion at m/z 690 and varying daughter ions at m/z 715 and 658 (Figure S22) strongly suggested the same structural residues in parts B and C of 1 and 2. Thus, the remaining residue in part A was assigned as a benzoyl group attached to a glycine in 2. The presence of a monosubstituted benzene ring was confirmed with the aid of ¹³C NMR and HSQC data; the connectivity of the benzoyl group and glycine was further verified by the well-established HMBC correlations from H₂-6 (δ_H 4.11, m; δ_H 4.40, m)/H-13 (δ_H 7.89, d, J = 7.8 Hz) to C-7 (δ_C 169.8, C).

With the 2D structures of bacillibactins E and F (1 and 2) in hand, we next sought to determine their absolute configurations. The similarity of the ¹³C NMR data and a shared biogenesis of the three isolates supported the speculation that all three compounds possess the same absolute configuration. To further validate the configurations of 1 and 2, we utilized the advanced Marfey’s method.²³ By LC-MS analysis of the 2,4-dinitrophenyl-l-leucine amide (DLA) derivatives (Figure S4), the threonines of 1 and 2 were unambiguously determined to be l-threonine (2S,3R,2′S,3′R,2″S,3″R).

Given our structure elucidations for 1 and 2, we hypothesized that WMMC1349 might have a promiscuous enzyme responsible for the installation of the nicotinic moiety of 1 and the benzoate of 2. As such, we wanted to examine, in particular, the sequence for the putative DhbE. To test this hypothesis, the whole genome of Bacillus sp. WMMC1349 was sequenced (GenBank accession number GCA_013155385.1; Biosample: SAMN14918578, Bioproject: PRJNA632728) and analyzed using BLAST and antiSMASH (version 5.1.2) software. The bacillibactin BGC of WMMC1349 spans 15.93 kb and includes a typical dhb operon with an average of 66% identity at the protein level with that of B. subtilis. The architecture and annotation of the bacillibactin cluster are shown in Figure 2A and Table S3, respectively. This includes a trimodular NRPS encoded by dhbEBF (red), responsible for incorporating DHB, Gly, and Thr; dhbCBA (dhbA and dhbC in yellow), responsible for synthesis of DHB from chorismate; several transport-related genes (blue); a hydrolase involved in iron release, encoded by besA and an MbtH family accessory protein (gray). We propose that 1 and 2 arise simply from promiscuous activation of nicotinate and benzoate by DhbE. DhbE belongs to a family of standalone, 2,3-DHB-activating A domains that includes its ortholog of the same name in Bacillus subtilis (74% amino acid identity). The DhbE of B. subtilis is known to promiscuously activate salicylate, 3-hydroxybenzoic acid, and 2-aminobenzoic acid.²⁴ Additionally, the well-characterized EntE of enterobactin biosynthesis in Escherichia coli has 49% identity to DhbE.²⁵ EntE has been shown to activate structurally related compounds including benzoic acid derivatives functionalized with methyl, ethynyl, or halogen moieties at the 3 position²⁶ or hydroxylated at the 2, 2,4, 2,5, or 2,3,4 positions, although, simple benzoic acid failed to serve as an EntE substrate.²⁷ The WMMC1349 DhbE also bears homology to a group of 3-hydroxypicolinic acid (3-HPA)-activating enzymes including many involved in the synthesis of the streptogramin B group of antibiotics.^28–33 Notably, two of these, PyrA of the pyridomycin biosynthetic pathway (55% ID) and SgvD1 from viridogrisein (etamycin) biosynthesis (54% ID), have been shown to promiscuously activate nicotinate.^28,32 The strong homology of the WMMC1349 DhbE with that of B. subtilis and their identical specificity codes hint at the possibility of promiscuous activation of nicotinate and benzoate during assembly of 1 or 2.³⁴ In the absence of other candidates within the cluster, the WMMC1349 DhbE represents the most likely candidate enzyme for carrying out nicotinic and benzoic acid activations. However, the DhbE orthologs in particular have a percent identity of 74.3% and identical A domain specificity codes of NYSAQGVVNK. As such, there are no obvious sequence differences that might predictably confer promiscuity. Hence, our sequencing data for the WMMC1349 genome fails to definitively correlate nicotinic and/or benzoic acid activation events to DhbE. In addition, the biosynthesis of nicotinate from aspartate is feasible in WMMC1349, as proposed for pyridomycin²⁸ since genes encoding enzymes in this pathway were found elsewhere in the genome (Table S4). Benzoate is believed to originate from tryptophan and/or phenylalanine³⁵ despite BLAST searches of the genome that failed to reveal a clear biosynthetic route to benzoate in 2.

Figure 2.

Putative biosynthetic pathway to the bacillibactins in Bacillus sp. WMMC1349. (A) Organization of the bacillibactin biosynthetic gene cluster (BGC) with the core NRPS components shown in red, the genes required for DHB synthesis in yellow (other than bifunctional dhbB, which is also a core component), transport-related genes in blue, and others involved in NRPS function and release of iron from the ferrisiderophore in gray. (B) Proposed biosynthetic pathway to 1–3 with domain annotations as ArCP, aryl carrier protein; C, condensation domain; A, adenylation domain; T, thiolation domain; TE, thioesterase domain. (C) Proposed biosynthesis of aromatic building blocks including DHB. Nicotinic acid is plausibly synthesized from aspartate, whereas benzoic acid is likely synthesized from tryptophan and/or phenylalanine.

CONCLUSIONS

In this study, we report the isolation and identification of two uncommon marine siderophores, bacillibactins E and F (1 and 2); 1 and 2 possess nicotinic and benzoic acid moieties, respectively, that are unprecedented among natural product siderophores. In addition, Marfey’s method was employed to unambiguously determine the absolute configuration of the unknown stereogenic centers of 1 and 2 as (2S,3R,2′S,3′R,2″S,3″R). Since bacillibactins and enterobactins have been studied for many years, we hypothesize that the discovery of 1 and 2 supports the notion that perhaps WMMC1349 encodes a DhbE enzyme with sequence variability enabling promiscuity. Unfortunately, whole genome sequencing showed that there was high sequence homology across DhbE variants. Therefore, the sequence of the WMMC1349 DhbE does not explain why we observed siderophores containing benzoic acid and nicotinic acid. Overall, the discovery of two new bacillibactins helps to expand our knowledge base of marine siderophores.

EXPERIMENTAL SECTION

General Experimental Procedures.

NMR spectra were obtained in CD₃OD (δ_H 3.34 ppm, δ_C 49.0 ppm) with a Bruker Avance 600 III MHz spectrometer. HRESIMS data were acquired with a Bruker MaXis 4G ESI-QTOF mass spectrometer. Reversed-phase HPLC was performed using a Shimadzu Prominence HPLC system, a Phenomenex Gemini C18 column (250 × 30 mm), and a Phenomenex Luna C18 column (250 × 10 mm, 5 μm). UHPLCHRMS was acquired using a Bruker MaXis 4G ESI-QTOF mass spectrometer coupled with a Waters Acquity UPLC system operated by Bruker Hystar software and a C18 column (Phenomenex Kinetex 2.6 μm, 2.1 mm × 100 mm).

Biological Material.

Sponge specimens (Cinachyrella apion, NCBI Taxonomy ID: 263142) were collected on May 27, 2015, near the west shore of Ramrod Key (24°39′38.1″ N, 81°25′25.0″ W) in Florida. A voucher specimen is housed at the University of Wisconsin–Madison. For cultivation, a sample of sponge (1 cm³) was ground in 500 μL of sterile seawater and dilutions were made using 500 μL of sterile seawater. Subsequently, 400 μL of diluted sponge sample was added to 200 μL of sterile artificial seawater, and 100 μL was plated using a sterile L-shaped spreader. A diluted sample was plated on Gauze 1 media supplemented with artificial seawater. Each medium was supplemented with 50 μg/mL cycloheximide, 25 μg/mL nystatin, and 25 μg/mL nalidixic acid. Plates were incubated at 28 °C, and colonies were isolated over the course of two months.

Fermentation, Extraction, and Isolation.

For siderophore production screening, 10 mL seed cultures in DSC medium (20 g soluble starch, 10 g glucose, 5 g peptone, 5 g yeast extract per liter of artificial seawater) were inoculated with strain WMMC-1349 and shaken (200 rpm at 28 °C) for 7 d. Each seed was then transferred into a 500 mL flask containing 100 mL ASW-A (20 g soluble starch, 10 g d-glucose, 5 g peptone, 5 g yeast extract, 5 g CaCO₃, 70 g Diaion HP-20 per liter of artificial seawater) and RAM2 medium (4 g corn meal, 10 g glucose, 15 g maltose, 7.5 g pharmamedia, 5 g yeast extract, 70 g Diaion HP-20 per liter of H₂O including 500 mL distilled H₂O and 500 mL artificial seawater) with or without iron. After inoculation, cultures were put into an incubator at 28 °C and agitated at 200 rpm for 7 d. Distilled H₂O was applied to wash filtered HP-20 (twice) followed by the extraction with acetone for 30 min. The acetone extract was dissolved in 1.0 mL of 10% aqueous MeOH and applied to an SPE column followed by washing with 1.0 mL of 10% aqueous MeOH (twice). Elution from SPE was achieved using 90% aqueous MeOH (1 mL), and the subsequent eluent was analyzed by LC-MS. Bacillibactin (3) was determined to be produced only in RAM2 iron-deficient media by the analysis of LC-MS (Figures S1 and S2). Detailed searching review of the LC-MS revealed at least two more bacillibactin analogues in the acetone extract (m/z 852 and 851).

To isolate sufficient quantities of material for structural elucidation, we scaled up production and isolated bacillibactins as guided by the MS and CAS assays. For the production of bacillibactins, 4 L flasks (10 × 1 L) containing iron-deficient medium RAM2 with Diaion HP-20 (7% by weight) were inoculated with 50 mL seed cultures (as described above) and agitated (200 rpm, 28 °C) for 7 d. Filtered HP-20 was washed with distilled H₂O (twice), and organics were then extracted from the HP-20 resin using 2.0 L of acetone (two 30 min washes carried out sequentially). The combined acetone solution was then concentrated to dryness, and the remaining extract syrup was subjected to liquid–liquid partitioning using 30% aqueous MeOH and CHCl₃ (1:1). The CHCl₃-soluble partition (2.5 g) was then fractionated by Sephadex LH-20 column chromatography (column size 500 × 40 mm, CHCl₃:MeOH = 1:1, 15 mL for each fraction). Fractions containing 1–3 (750 mg, fractions F3–F5) were subjected to Prep-RP-HPLC isolation (10%/90% to 100%/0% MeOH/H₂O (with 0.1% acetic acid), 20 min, 20 mL/min) using a Phenomenex Gemini C18 column (250 × 30 mm). Further semiprep-HPLC was applied for the purification of compounds (1, 2.5 mg, t_R = 15.6 min; 2, 10.2 mg, t_R = 17.4 min; 3, 22.1 mg, t_R = 16.6 min). All the eluates were tested for CAS activity,²² using MeOH as a negative control (Figure S3). For making artificial seawater (ASW), solutions I (415.2 g NaCl, 69.54 g Na₂SO₄, 11.74 g KCl, 3.40 g NaHCO₃, 1.7 g KBr,0.45 g H₃BO₃, 0.054 g NaF) and II (187.9 g MgCl₂·6H₂O, 22.72 g CaCl₂·2H₂O, 0.428 g SrCl₂·6H₂O) were made up separately using distilled water and combined to give a total volume of 20 L.

Bacillibactin E (1).

White powder; [α]²⁰_D −44 (c 0.1, MeOH); ¹H and ¹³C NMR data, see Table S1; HRESIMS m/z 852.2675 [M + H]⁺ (calcd for C₃₈H₄₂N₇O₁₆, 852.2683).

Bacillibactin F (2).

White powder; [α]²⁰_D −48 (c 0.1, MeOH); ¹H and ¹³C NMR data, see Table S1; HRESIMS m/z 851.2731 [M + H]⁺ (calcd for C₃₉H₄₃N₆O₁₆, 851.2730).

Sequencing and Identification of Candidate Bacillibactin Biosynthetic Genes.

16S rDNA sequencing was conducted as previously described.³⁶ WMMC1349 was identified as a Bacillus sp. The 16S sequence for WMMC1349 was deposited in GenBank (accession number MK892477). Genome sequence (GenBank accession number GCA_013155385.1; Biosample: SAMN14918578, Bioproject: PRJNA632728) was subjected to antiSMASH 5.0 analysis.³⁷ Results were analyzed by BLAST analysis. The bacillibactin BGC was identified by a combination of analysis of the WMMC1349 genome by antiSMASH and manual comparison of the ORFs to other related BGCs, including the enterobactin BGC, in SnapGene. The BGC-encoded enzymes, excluding the transport-related genes, have an average percent identity of 66% when compared to the dhb cluster of B. subtilis, and the DhbE orthologs in particular have a percent identity of 74.3% and identical A domain specificity codes of NYSAQGVVNK.

Determination of Amino Acid Configuration in Bacillibactins.

1-Fluoro-2,4-dinitrophenyl-5–5-l-leucine amide (l-FDLA) was synthesized as previously reported,³⁸ and this reagent was used to generate hydrolysate-derived diastereomers. The hydrolysate was mixed with 1 N NaHCO₃ (80 μL) and 150 μL of l-FDLA (10 mg/mL in acetone). The solution was stirred at 45 °C for 1 h, cooled to room temperature (rt), quenched with 1 N HCl (80 μL), and dried under vacuum. Similarly, the standard amino acids (l-threonine, l-allo-threonine, d-threonine, and d-allo-threonine) were derivatized separately. The derivatives of the hydrolysate of compounds 1–3 and the standard amino acids were subjected to LC-MS analysis with a Phenomenex Luna C18 reversed-phase column (250 × 4.6 mm, 5 μm) at a flow rate of 1.0 mL/min and with a linear gradient of H₂O (containing 0.1% formic acid) and MeOH (90:10 to 0:100 over 20 min and a hold at 100% MeOH for 5 min). The absolute configuration of the amino acid in 1–3 was determined by comparing the retention times of the amino acid derivatives, which were identified by LC/MS. The retention times for the l-DLA derivatives of the hydrolysate of compounds 1–3 and the standard l-threonine were the same (t_R = 8.73 min, Figure S4).

Chrome Azurol S Assay for Siderophore Activity of Analytically Pure 1–3.

Hexadecyltrimethylammonium bromide (CTAB) (21.9 mg) was dissolved in 25 mL of H₂O at 35 °C. To this solution was added 1.5 mL of 1 mM iron(III) chloride solution (prepared by dissolving anhydrous FeCl₃ in a 10 mM aqueous HCl solution) and 7.5 mL of a 2 mM aqueous CAS solution at rt. In a separate Erlenmeyer flask, 9.76 g of 2-(N-morpholino)ethanesulfonic acid (MES) was diluted in 50 mL of water, and a 50% KOH solution was used to adjust the pH of this solution to 5.6. The premade CTAB–CAS–Fe(III) solution was then poured into the MES buffer with stirring, and the final volume of the modified CAS assay solution was adjusted to 100 mL with H₂O. CTAB–CAS–Fe(III) with MES buffer solution (100 μL) was added to each well of the 96-well microplate. Each well was then treated with 100 μL of the dilute solution of thalassosamide or deferoxamine mesylate in H₂O to achieve final concentrations ranging from 1.28 mM to 1.25 μM. Each reaction was carried out in duplicate. After incubation at 37 °C for 3 h, the resulting color changes were observed by visual inspection, or the corresponding absorption changes were recorded. Data are shown in Figure S23.