Bacillimidazoles A−F, Imidazolium-Containing Compounds Isolated from a Marine Bacillus

Chemical investigations of a marine sponge-associated Bacillus revealed six new imidazolium-containing compounds, bacillimidazoles A–F (1–6). Previous reports of related imidazolium-containing natural products are rare. Initially unveiled by timsTOF (trapped ion mobility spectrometry) MS data, extensive HRMS and 1D and 2D NMR analyses enabled the structural elucidation of 1–6. In addition, a plausible biosynthetic pathway to bacillimidazoles is proposed based on isotopic labeling experiments and invokes the highly reactive glycolytic adduct 2,3-butanedione. Combined, the results of structure elucidation efforts, isotopic labeling studies and bioinformatics suggest that 1–6 result from a fascinating intersection of primary and secondary metabolic pathways in Bacillus sp. WMMC1349. Antimicrobial assays revealed that, of 1–6, only compound six displayed discernible antibacterial activity, despite the close structural similarities shared by all six natural products.


Introduction
Heterocyclic scaffolds are commonly encountered in natural products isolated from both terrestrial and marine organisms [1][2][3][4]. Their vast structural diversity, drug-like features, and biological properties have inspired both intensive efforts to discover new heterocyclic compounds, as well as imaginative total syntheses [4]. Nitrogen-containing heterocycles, such as pyrroles, imidazoles, oxazoles, pyridines, and quinolones, exhibit a diverse array of biological activities; these include, but are by no means limited to, antibacterial [5], antifungal [6], and anticancer [7] activities. Marine-derived imidazole alkaloids have been one of the most fruitful families of bioactive compounds giving rise to many pharmaceutical leads [8]. Historically, marine-derived imidazole alkaloids have been most often isolated from sponges; imidazole alkaloids [9] featuring bromopyrrole-imidazoles [10], indole-containing imidazoles [11], and 2-aminoimidazoles are among the species most often identified from marine sponges [12]. Furthermore, in recent years, marine microorganisms have come to be viewed as sustainable and productive sources of new bioactive imidazole-containing natural products [9]. However, reports of positively charged imidazolium natural products remain relatively rare. Most reports of imidazoliumcontaining compounds feature 1,3-dimethyl-5-methylthiol [13][14][15] or 2-aminoimidazolium containing structures (guanidinium-like) [10,11]. A wide array of applications in ionic liquids [16], important biological activities [17], and their amenability to further structural modifications have made imidazolium salts an attractive target of contemporary research [18]. Therefore, discovering new imidazolium-based species and gaining further insight into their biosynthetic origins has become an interesting, yet challenging, task for natural product scientists.

Results & Discussion
As part of our ongoing efforts to discover new natural products from marine invertebrateassociated bacteria [19], we developed a streamlined discovery platform that includes strain prioritization by metabolomics [20] and an LC/MS fractionation platform to generate screening libraries [21]. Strain WMMC1349, a marine Bacillus sp. cultivated from the sponge Cinachyrella apion, drew our attention since one of its fractions displayed activity against methicillin-resistant Staphylococcus aureus (MRSA). Further purification of the active fraction by HPLC resulted in enrichment of an active subfraction. Interestingly, subsequent analytical HPLC revealed this fraction to be a represented by a broad peak ( Figure S1) despite the clear presence of a mixture of compounds as revealed by 1 H NMR ( Figure S1). Fortunately, Bruker timsTOF (trapped ion mobility spectrometry) MS data for this subfraction indicated a series of new molecules with m/z values of 305.2, 319.2, 344.4, 358.2, 383.2, and 397.2, respectively ( Figure 1). Exhaustive attempts to separate each different m/z species (see Figures S1-S55), as initially visualized by timsTOF MS, ultimately afforded HPLC conditions amenable to clean separation and isolation of each discreet compound. The six new imidazolium-containing compounds, now differentiated from each other, were termed bacillimidazoles A-F (1-6, Figure 2), and all were assessed for in vitro activity against MRSA, B. subtilis, and E. coli; only compound 6 was found to be active (MRSA). Isotopic labelling of these metabolites using isotopically enriched culture media, and bioinformatic analysis were conducted to decipher the means by which the bacillimidazoles are biosynthesized. Marine-derived imidazole alkaloids have been one of the most fruitful families of bioactive compounds giving rise to many pharmaceutical leads [8]. Historically, marinederived imidazole alkaloids have been most often isolated from sponges; imidazole alkaloids [9] featuring bromopyrrole-imidazoles [10], indole-containing imidazoles [11], and 2-aminoimidazoles are among the species most often identified from marine sponges [12]. Furthermore, in recent years, marine microorganisms have come to be viewed as sustainable and productive sources of new bioactive imidazole-containing natural products [9]. However, reports of positively charged imidazolium natural products remain relatively rare. Most reports of imidazolium-containing compounds feature 1,3-dimethyl-5-methylthiol [13][14][15] or 2-aminoimidazolium containing structures (guanidinium-like) [10,11]. A wide array of applications in ionic liquids [16], important biological activities [17], and their amenability to further structural modifications have made imidazolium salts an attractive target of contemporary research [18]. Therefore, discovering new imidazoliumbased species and gaining further insight into their biosynthetic origins has become an interesting, yet challenging, task for natural product scientists.

Results & Discussion
As part of our ongoing efforts to discover new natural products from marine invertebrate-associated bacteria [19], we developed a streamlined discovery platform that includes strain prioritization by metabolomics [20] and an LC/MS fractionation platform to generate screening libraries [21]. Strain WMMC1349, a marine Bacillus sp. cultivated from the sponge Cinachyrella apion, drew our attention since one of its fractions displayed activity against methicillin-resistant Staphylococcus aureus (MRSA). Further purification of the active fraction by HPLC resulted in enrichment of an active subfraction. Interestingly, subsequent analytical HPLC revealed this fraction to be a represented by a broad peak ( Figure S1) despite the clear presence of a mixture of compounds as revealed by 1 H NMR ( Figure S1). Fortunately, Bruker timsTOF (trapped ion mobility spectrometry) MS data for this subfraction indicated a series of new molecules with m/z values of 305.2, 319.2, 344.4, 358.2, 383.2, and 397.2, respectively ( Figure 1). Exhaustive attempts to separate each different m/z species (see Figure S1-S55), as initially visualized by timsTOF MS, ultimately afforded HPLC conditions amenable to clean separation and isolation of each discreet compound. The six new imidazolium-containing compounds, now differentiated from each other, were termed bacillimidazoles A−F (1−6, Figure 2), and all were assessed for in vitro activity against MRSA, B. subtilis, and E. coli; only compound 6 was found to be active (MRSA). Isotopic labelling of these metabolites using isotopically enriched culture media, and bioinformatic analysis were conducted to decipher the means by which the bacillimidazoles are biosynthesized.  The molecular formulae of bacillimidazoles A (1) and B (2) were determined to be C21H25N2 + (m/z = 305.2029, M + , calcd 305.2012, Figure S37) and C22H27N2 + (m/z = 319.2171, M + , calcd 319.2169, Figure S38), respectively, based on HRESIMS data. In the 13 C NMR spectra of 1 and 2 in Table 1, only 11 and 12 carbon signals were observed, respectively, suggesting symmetric scaffolds for both 1 and 2. Furthermore, comparisons of 1 H and 13 C NMR data revealed a high degree of similarity between compound 1 and lepidiline A, an imidazolium-containing alkaloid isolated from the South American plants Lepidium meyenii Walp [22]. These similarities suggested the presence of a 4,5-dimethyl imidazolium cyclic structure and two phenyl-ring containing substituents in compound 1. More highly refined datasets revealed that H2-7 (δH 3.06) and H2-8 (δH 4.35) showed COSY correlations ( Figure 3) to each other. The HMBC correlations ( Figure 3) were also observed from H2-7 to C-2 (δC 130.0) and from H2-8 (δH 4.35) to C-10 (δC 128.5), C-11 (δC 135.5), suggesting that -CH2CH2-groups linked the central imidazolium ring to two terminal phenyl rings, one on either side of the imidazolium. Therefore, the structure of 1 was assigned as a 1,3-difunctionalized imidazolium-containing structure. The NMR dataset of 2 was compared to that obtained for 1 and the only observable difference was an additional methyl group (δH, 2.01, H3-13; δC 9.7, CH3) substitution on C-11 (δC 144.0), which was determined by careful interpretation of the well resolved HMBC correlation from H3-13 to C-11 ( Figure  3).   Figure S38), respectively, based on HRESIMS data. In the 13 C NMR spectra of 1 and 2 in Table 1, only 11 and 12 carbon signals were observed, respectively, suggesting symmetric scaffolds for both 1 and 2. Furthermore, comparisons of 1 H and 13 C NMR data revealed a high degree of similarity between compound 1 and lepidiline A, an imidazolium-containing alkaloid isolated from the South American plants Lepidium meyenii Walp [22]. These similarities suggested the presence of a 4,5-dimethyl imidazolium cyclic structure and two phenyl-ring containing substituents in compound 1. More highly refined datasets revealed that H 2 -7 (δ H 3.06) and H 2 -8 (δ H 4.35) showed COSY correlations ( Figure 3) to each other. The HMBC correlations ( Figure 3) were also observed from H 2 -7 to C-2 (δ C 130.0) and from H 2 -8 (δ H 4.35) to C-10 (δ C 128.5), C-11 (δ C 135.5), suggesting that -CH 2 CH 2 -groups linked the central imidazolium ring to two terminal phenyl rings, one on either side of the imidazolium. Therefore, the structure of 1 was assigned as a 1,3difunctionalized imidazolium-containing structure. The NMR dataset of 2 was compared to that obtained for 1 and the only observable difference was an additional methyl group (δ H , 2.01, H 3 -13; δ C 9.7, CH 3 ) substitution on C-11 (δ C 144.0), which was determined by careful interpretation of the well resolved HMBC correlation from H 3 -13 to C-11 ( Figure 3).  C-6, δC 120.1; C-7, δC 122.8; C-8, δC 112.6; C-9, δC 138.1) were characteristic of the 3-indoyl structural motifs in 3; the validity of this idea was verified by the observation of HMBC correlations from H-2 to C-3/C-4/C-9 and from H-5 to C-3/C-9 ( Figure 3). Similar to 1 and 2, the connection between -CH2CH2groups and the aromatic building blocks in 3, as well as the central imidazolium ring, was deduced by HMBC correlations from H-3 to C-10 and from H2-11 to C-13/C-14 ( Figure 3), respectively, establishing the full structural assignment of bacillimidazole C (3). Finally, in a fashion similar to that applied to 3, the structure of 4, with its additional imidazolium-linked methyl group, was elucidated.
Approaches employed to solve the structures of 1-4 also were applied to determine the structures of 5 and 6 in Table 2. Analysis of HRMS data for bacillimidazole E (5) and bacillimidazole F (6) 13 C NMR data suggested asymmetrical structures for both 5 and 6 since 23 13 C NMR signals were observed in 5 (24 signals for 6). Additionally, comparisons of NMR data for 5 to those of 1 and 3 suggested the presence of an imidazolium ring, a phenyl ring, and a 3-indole ring in 5. Furthermore, H-2 (δ H 7.03), assigned to the indole ring, showed an HMBC correlation to C-10 (δ H 26.5); and H-22 (δ H 7.04), assigned to the phenyl ring, showed an HMBC correlation to C-20 (δ C 36.7) (Figure 3) .4), which was assigned to the imidazolium ring ( Figure 3). Therefore, 5 was believed to contain an indole ring, an imidazolium ring, and a phenyl moiety (Figure 2). In addition, two -CH 2 CH 2 -groups were found to intervene the three different cyclic structural motifs. In applying the established correlation data, we thus elucidated structure 5 as shown in Figure 2. The structure of 6 was determined in a fashion similar to that employed for 5, and was ultimately identified as an analog of 5 bearing a methyl group on the central imidazolium ring. Following their structural elucidation, bacillimidazoles A-F (1-6) were tested for antibacterial activity against MRSA, B. subtilis, and E. coli (Table S1). All compounds failed to show any significant activity against B. subtilis and E. coli, although bacillimidazole F (6) did display weak activity against MRSA with an MIC of 38.3 µM.
To better understand the biosynthetic mechanisms involved in generating the uncommon imidazolium structures found in the bacillimidazoles, isotopic enrichment studies were carried out. In particular, we employed 13 C enriched culture media to glean vital insight into how the bacillimidazoles are constructed. By substituting two carbon sources, starch and D-glucose, with 13 C 6 -D-glucose, we found that the four carbons composing the vicinal dimethyl olefin of bacillimidazoles C (3) and E (5) (Figures S43-S48) underwent substantial 13 C enrichment relative to all other carbons (Figure 4a). On the basis of these findings, we theorized that all carbons of the basic framework originate from amino acids whereas the dimethyl olefin elements of the bacillimidazoles are derived from glucose, presumably via a glycolytic process (Figure 4b). This logic is, of course, buoyed by the resemblance of the bacillimidazole sidechains to those found in phenylalanine and tryptophan. On the basis of these findings, we investigated the possibility of a biosynthetic pathway, as shown in Figure 4. It is well known that glucose is readily converted to 2,3-butadione along the canonical glycolytic pathway to acetolactate and subsequent processing of acetolactate to the essential 2,3-butadione, which can spontaneously react with either tryptamine or phenethylamine, both of which are common bacterial metabolites formed from the aromatic amino acids tryptophan and phenylalanine. These three components-two amines and a dione-provide most of the atoms of the imidazole moiety tethered through both nitrogens to various side chains. This diimine could undergo further spontaneous reactions with either one or two carbon carboxylic acids, or derivatives, and appropriate redox agents (or enzymes) to form the central imidazolium ring. Acetyl-CoA, or an equivalent, would produce the methylated compounds and formate, or a derivative, would produce the unmethylated bacillimidazoles. There is an alternative possibility in which the relevant amino acids condense with the butadione moiety prior to decarboxylation, and that the decarboxylation(s) of such intermediates might expedite imidazolium ring formation. While it is not possible at this stage to propose a detailed stepwise biosynthesis for the bacillimidazoles, an overall path with both enzymatic and spontaneous steps is likely, and the 13 C-glucose feeding studies clearly indicate the importance of glucose processing en route to imidazole assembly. That the production of bacillimidazoles by WMMC1349 is driven by secondary metabolism biosynthetic machineries and is not relegated only to primary metabolic events and/or extract workup conditions, is supported by the fact that, of six bacilli strains evaluated, only two (including WMMC1349) proved to be bacillimidazole producers (Figures S4-S6, S49-S54 and Table S2).
butadione along the canonical glycolytic pathway to acetolactate and subsequent processing of acetolactate to the essential 2,3-butadione, which can spontaneously react with either tryptamine or phenethylamine, both of which are common bacterial metabolites formed from the aromatic amino acids tryptophan and phenylalanine. These three components-two amines and a dione-provide most of the atoms of the imidazole moiety tethered through both nitrogens to various side chains. This diimine could undergo further spontaneous reactions with either one or two carbon carboxylic acids, or derivatives, and appropriate redox agents (or enzymes) to form the central imidazolium ring. Acetyl-CoA, or an equivalent, would produce the methylated compounds and formate, or a derivative, would produce the unmethylated bacillimidazoles. There is an alternative possibility in which the relevant amino acids condense with the butadione moiety prior to decarboxylation, and that the decarboxylation(s) of such intermediates might expedite imidazolium ring formation. While it is not possible at this stage to propose a detailed stepwise biosynthesis for the bacillimidazoles, an overall path with both enzymatic and spontaneous steps is likely, and the 13 C-glucose feeding studies clearly indicate the importance of glucose processing en route to imidazole assembly. That the production of bacillimidazoles by WMMC1349 is driven by secondary metabolism biosynthetic machineries and is not relegated only to primary metabolic events and/or extract workup conditions, is supported by the fact that, of six bacilli strains evaluated, only two (including WMMC1349) proved to be bacillimidazole producers ( Figures S4-S6, S49-S54 & Table S2). Carbons highlighted with bold bonds and red spheres showed high levels of 13 C incorporation. (b) Proposed biosynthetic pathway to compounds 1-6 calls for enzymatic production of tryptamine, phenethylamine and 2,3-butanedione (from glucose); all subsequent steps may proceed spontaneously.
Based on the biosynthetic insights gained from labelling experiments, we set out to identify the gene cluster responsible for the biosynthesis of 1−6. We sequenced and assembled the genome of the producer, Bacillus sp. WMMC1349 [23]. The genome sequence was analyzed with state-of-the-art biosynthetic pipelines, yet no likely biosynthetic pathway for the biosynthesis of 1−6 was identified. We therefore mined the genome for the presence of the genes involved in acetoin biosynthesis, of which 2,3-butanedione, the proposed building block of 1−6, is a precursor. We identified all of the genes responsible for acetoin biosynthesis in the genome of Bacillus sp. WMMC1349. Acetoin is biosynthesized from two molecules of pyruvate that are condensed to generate acetolactate. Acetolactate can either be enzymatically transformed into acetoin by a decarboxylase or can undergo Based on the biosynthetic insights gained from labelling experiments, we set out to identify the gene cluster responsible for the biosynthesis of 1-6. We sequenced and assembled the genome of the producer, Bacillus sp. WMMC1349 [23]. The genome sequence was analyzed with state-of-the-art biosynthetic pipelines, yet no likely biosynthetic pathway for the biosynthesis of 1-6 was identified. We therefore mined the genome for the presence of the genes involved in acetoin biosynthesis, of which 2,3-butanedione, the proposed building block of 1-6, is a precursor. We identified all of the genes responsible for acetoin biosynthesis in the genome of Bacillus sp. WMMC1349. Acetoin is biosynthesized from two molecules of pyruvate that are condensed to generate acetolactate. Acetolactate can either be enzymatically transformed into acetoin by a decarboxylase or can undergo spontaneous decarboxylation to yield the building block of 1-6, 2,3-butanedione. A diacetyl reductase subsequently converts 2,3-butanedione into acetoin. Whole genome alignments of different Bacillus spp. revealed that the region upstream of the acetoin biosynthetic genes differs significantly from other Bacillus spp., including the model strain B. subtilis 168 (insertion of a large low-density coding region in Bacillus sp. WMMC1349), while the downstream region is homologous in all analyzed genomes. To our surprise, we were not able to identify a copy of the gene family encoding butanediol-dehydrogenases which was present in all other analyzed Bacillus spp. The absence of genes involved in acetoin catabolism would be expected to increase the concentration of the precursor 2,3-butanedione. Genome mining (Section 4.5 below) of the bacillimidazole producer revealed the presence of a gene encoding an aromatic-L-amino acid decarboxylase, which is the proposed second enzyme essential for 1-6 biosynthesis en route to building blocks tryptamine and phenethylamine, respectively. No homologs of the aromatic-L-amino acid decarboxylase gene were identified in any other Bacillus genome analyzed, including the model strain Bacillus subtilis 168, indicating that the gene is facultative for the genus Bacillus. Relative to the biosynthetic machineries of most bacterial secondary metabolites, genes involved in acetoin biosynthesis and the aromatic-L-amino acid decarboxylase gene are not clustered. This observation suggests that the bacillimidazoles may form spontaneously from high-abundance primary metabolites with congruent reactivities. The biosynthesis of the bacillimidazoles is yet another example of the growing number of natural products that are produced partly by genetically coded instructions and partly by spontaneous reactivity-the joining of reactive intermediates, or products, from different pathways [24][25][26]. The initial formation of the bis-imines represents, for example, the unsurprising coupling of highly reactive primary amines with a very reactive alpha-diketone.

Conclusions
The isolation and structural elucidation of bacillimidazoles A-F, six imidazoliumcontaining heterocycles from marine Bacillus sp. WMMC1349 represent the discovery of a new class of marine-derived heterocyclic natural products. Isotopic labeling of 1-6 using 13 C enriched culture media, biomimetic synthetic approaches and bioinformatic analyses were performed in order to gain insights into the biosynthetic assembly of these interesting compounds. To the best of our knowledge, there are only a few reports of natural products containing 1,3-difunctionalized imidazolium moieties, and this is the first report of naturally occurring imidazolium-containing heterocycles. We contend, based on these findings, that the discovery of 1-6 contributes to our advancing knowledge of significantly underexplored biosynthetic pathways, specifically those that intersect primary and secondary metabolic pathways. We anticipate that the lessons learned here will help to expedite efforts to more fully understand and exploit the full biosynthetic potential of Bacillus spp.

General Experimental Procedures
UV spectra were recorded on an Aminco/OLIS UV-Vis spectrophotometer. IR spectra were measured with a Bruker Equinox 55/S FT-IR spectrophotometer. NMR spectra were obtained in CD 3 OD (δ H 3.34 ppm, δ C 49.0 ppm) with a Bruker Avance 600 III MHz (Billerica, MA, USA) spectrometer equipped with a 1 H{ 13 C/ 15 N/ 31 P} cryoprobe, a Bruker Avance III 500 MHz (Billerica, MA, USA) spectrometer equipped with a 13 C/ 15 N{ 1 H} cryoprobe, and a Bruker Avance III HD 400 MHz (Billerica, MA, USA) spectrometer. HRMS data were acquired with a Bruker MaXis™ 4G ESI-QTOF (Billerica, MA, USA) mass spectrometer. RP HPLC was performed using a Shimadzu Prominence HPLC system and a Phenomenex Gemini C18 column (250 × 30 mm). UHPLC-HRMS was acquired using a Bruker MaXis™ 4G ESI-QTOF (Billerica, MA, USA) 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). Bruker timsTOF Pro instrument (Billerica, MA, USA) was used for the trapped ion mobility MS analysis using direct infusion with 0.003 mL/min of flow rate and ESI+ ionization source. Nebulizer gas 0.4 bar, dry gas 3.5 L/min, source temperature 220 • C, ESI voltage 4200V (+). MS spectra were collected using the following parameters: tims ramp time = 350 ms, PASEF on, scan range (m/z, 20-1000; 1/k 0 , 0.70-1.00 V·s/cm 2 . of Wisconsin−Madison. For cultivation, a sample of sponge (1 cm 3 ) was ground in 500 µL sterile seawater, and dilutions were made using 500 µL 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. Diluted samples were 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.