Three hundred and one Vibrionaceae strains were isolated during a global marine expedition (http://www.galathea3.dk/uk) based on their ability to antagonize the fish pathogen Vibrio anguillarum strain 90-11-287 [8]. After being stored at −80 °C for between six and 12 months, all strains were retested for antibacterial activity against V. anguillarum strain 90-11-287 and the human pathogen Staphylococcus aureus strain 8325 by spotting colony mass on pathogen-seeded agar [8]. Activity was assessed by the formation of clearing zones around spotted colony mass. From 301 strains, only 138 retained antibacterial activity, being a small fraction compared to other antagonistic marine bacteria [32,33]. One hundred strains causing pronounced inhibition (diameter of clearing zones larger than 10 mm) were retested using the same set-up, resulting in a subselection of 39 strains with reproducible strong antibacterial activity when spotted on pathogen-seeded agar. This subselection was inoculated in liquid cultures and extracted with ethyl acetate to determine if antibacterial compounds were extractable with organic solvent. Activity was seen in ethyl acetate extracts from five strains, which were selected for further analyses. The five bioactive strains originated from different surface samples collected in distant oceanic regions (Figure 1).

All strains investigated in the present study had previously been assigned to the Vibrionaceae family based on 16S rRNA gene similarities [8]. However, the 16S rRNA gene is highly conserved among the Vibrionaceae and is not well suited for identification to the species level [34]. Therefore, additional sequence analyses of three housekeeping genes (recA, rpoA, and toxR) were performed. These genes encode constitutively expressed proteins and are suitable for phylogenetic studies of Vibrionaceae [34,35]. On the basis of recA and rpoA sequence similarities, strains S2052 and S4053 were identified as Vibrio coralliilyticus, S2394 and S4051 as Vibrio neptunius, and S2753 as Photobacterium halotolerans (Figure 1). The toxR gene was less suited for general species identification due to its high variability even in closely related vibrios, as well as comparatively few toxR sequence data available in public gene libraries [36]. However, multiple alignments and neighbor-joining analyses of toxR sequences provided the best phylogenetic resolution for determining the relationship between the five strains (Figure 2). The usefulness of toxR for species discrimination was consistent with previous reports [35]. LC-UV/MS metabolite profiles underlined the close relationship between V. coralliilyticus S2052/4053 and V. neptunius S2394/4051, respectively. The evolutionary distance of P. halotolerans S2753 to the other strains was reflected by a unique metabolite profile (Figure 2). All five strains were consistent in their metabolite production in separate cultivations over a one-year interval.

Several metabolites were produced by all V. coralliilyticus and V. neptunius strains, for instance those related to the peaks at retention times Rt = 11.08 and 12.03 min (Figure 2). Although they are different species, V. coralliilyticus and V. neptunius are closely related vibrios with only 2–3% sequence variation in the recA and rpoA genes (data not shown), signifying why biosynthetic pathways are shared between the species. Based on their molecular formulas, UV, and MS characteristics [37], most of the metabolites produced by both species were assigned as smaller peptides (m/z 300–500), a class of molecules commonly produced by marine culturable bacteria [38,39]. Despite the presence of shared metabolites between V. coralliilyticus and V. neptunius, clearly distinguishable peaks were seen as well. For instance, the major peak at retention time Rt = 10.02 min (MW 479 Da) was only seen in the two V. coralliilyticus, and the peak at Rt = 10.11 min (MW 493 Da) only in the two V. neptunius strains.

The metabolites produced by P. halotolerans S2753 comprised a series of larger peptides (m/z 500–900) [40]. The large peak at Rt = 1.70 min (MW 213 Da) displayed a unique UV spectrum characteristic of that of a highly conjugated system. However, this peak could not be ascribed to any known compound or compound class based on LC-UV/MS data alone.

Several metabolites (Rt = 4.70, 7.41, 8.60, 9.60, and 10.50 min) were found in all five strains and assigned as poly-β-hydroxybutyric acid polymers (PHB) of varying lengths (repeating unit n = 86 Da). This was verified by NMR for some of the compounds (data not shown). PHB are common bacterial storage compounds accumulated when growing on an excess carbon source [41].

Chemotyping of prokaryotes has mostly been restricted to analyses of fatty acids and sugars [42], but we show that also the profiling of small molecules can be used for species discrimination. This highlights the usefulness of metabolomics for bacterial classification, adding to recent work of whole-cell laser desorption MALDI-TOF mass spectrometry for characterization of vibrios [43] and secondary metabolite profiling to assess the biosynthetic potential of marine Pseudoalteromonas [32]. While our study is limited to the analysis of only three species from the Vibrionaceae family, the isolation of two genetically and chemically closely related “strain siblings” from distant oceanic regions indicated that production of certain secondary metabolites is a preserved trait. Similar secondary metabolite profiles were also shown for marine actinomycetal Salinispora spp. [44] from distant habitats. Also, all Ruegeria mobilis strains from worldwide locations produced the same antibiotic, tropodithietic acid [33].

V. coralliilyticus (strains S2052 and S4053) and P. halotolerans (S2753) inhibited both V. anguillarum and S. aureus, whereas V. neptunius (strains S2394 and S4051) only inhibited V. anguillarum (Table 1). Antibacterial activity was highest in aerated cultures and detected after one, three, and five days of incubation. No significant difference in activity was seen between the tested culture media.

The finding of bioactivity among marine Vibrionaceae underlined marine microorganisms being a source of antimicrobials. To our knowledge, none of the species investigated here have previously been studied with respect to their secondary metabolome including antibacterial compounds.

To identify the compounds responsible for the observed activity, large-scale cultivations and fractionations were undertaken for V. coralliilyticus S2052 and P. halotolerans S2753, representing two distant Vibrionaceae species with different metabolite profiles. All fractionation steps were guided by activity testing against V. anguillarum strain 90-11-287.

Initial dereplication of S2052 by LC-UV/MS [37] and explorative solid-phase extraction (E-SPE) [45] indicated that andrimid (Rt = 10.02 min; Figure 2) could be responsible for the antibacterial activity. This compound (Figure 3a), a hybrid nonribosomal peptide-polyketide antibiotic, was first described from an insect endosymbiont [46] and later found in other microbial species [47,48] including marine vibrios [29,31]. Pure andrimid was isolated for NMR analysis, and our data was in accordance with literature data [47]. Andrimid acts as an acetyl-CoA carboxylase inhibitor [49], and we extended its broad antibiotic spectrum [50] by showing inhibition of the bacterial pathogens Salmonella Enteritidis, Bacillus cereus, Yersinia enterolitica, Yersinia ruckeri, Vibrio harveyi, and Vibrio vulnificus (data not shown). Production of andrimid was also confirmed for the other isolated V. coralliilyticus strain, S4053. We furthermore speculate whether a recent report of antagonism in the V. coralliilyticus type strain [12] was also attributed to this compound. Previous studies have revealed almost identical andrimid gene clusters and a transposase pseudogene in two producer species, suggesting horizontal gene transfer as the most likely explanation behind the cosmopolitanism of the antibiotic [51]. We hypothesize that such transfer is also the reason for its presence in V. coralliilyticus S2052 and S4053. Our study is the first linking andrimid production to a specific Vibrio species, with production occurring in two strains isolated from very different geographical regions and sources.

The antibacterial compound of P. halotolerans S2753 was identified as holomycin (Rt = 1.70 min; Figure 2), a compound belonging to the pyrrothine class of antibiotics acting by interference with RNA synthesis [52]. Our NMR data (Figure 3b) was consistent with previous reports [53]. Holomycin has until now only been found in Gram-positive Streptomyces [54,55], and the present study is the first demonstrating production of this antibiotic in a Gram-negative heterotrophic bacterium. While parallel evolution of this trait is possible, horizontal gene transfer is the more likely explanation for its occurrence in both Vibrionaceae and actinomycetes. We extended the broad-spectrum activity of holomycin [52] by showing inhibition of the bacterial pathogens Listeria monocytogenes, Serratia marcescens, S. Enteritidis, B. cereus, Y. enterolitica, Y. ruckeri, V. harveyi, V. vulnificus and V. parahaemolyticus, as well as of several marine strains from the Roseobacter and Pseudoalteromonas groups (data not shown).

Neither andrimid nor holomycin were produced by V. neptunius S2394 and S4051, and further fractionation and purification is needed to identify the compound(s) responsible for their antibacterial activity. Interestingly, V. neptunius S4051 and V. coralliilyticus S4053 were isolated from the same seaweed sample, showing that two antagonistic Vibrio species whose antibacterial activity is based on different compounds co-occur in the same microenvironment. Moreover, the same sample also contained an antibiotic-producing Pseudoalteromonas strain [8].

Two of the antagonistic Vibrionaceae species harbor pathogenic strains, with V. coralliilyticus being pathogenic to corals [56] and V. neptunius being pathogenic to oysters [57]. While we do not know whether V. coralliilyticus S2052 has pathogenic potential, the V. coralliilyticus type strain has both antagonistic and pathogenic traits [12]. Hence, our results suggest that some vibrios possess a dual physiology, being antagonistic against other prokaryotes but pathogenic towards higher organisms. Moreover, the production of antibiotics in several species suggests that these compounds may be of ecological importance [1].

This study highlights one of the challenges in natural product discovery. Despite major screening efforts for novel antimicrobials to be used in pharmaceutical, food, and aquaculture industries, only a limited amount of compounds have been discovered in recent years [58]. While the isolation of culturable bacteria remains a promising approach [42] and the secondary metabolome of marine vibrios has not been extensively studied, we only isolated known compounds despite careful dereplication prior to any compound purification. Dereplication is apparently troubled by the high degree of gene transfer between distantly related bacteria such as Gram-positive actinomycetes and Gram-negative Proteobacteria [3]. Many compounds in natural product databases such as AntiBase [59] have similar masses (<5 ppm difference), so even the combination of UV/VIS spectra, accurate mass data (<5 ppm), and E-SPE [45] is not sufficiently discriminatory for these organisms. To avoid isolation of redundant chemistry, dereplication by NMR [60] or ultra high-resolution mass spectrometry (<1 ppm) with high isotope accuracy ratios for correct elementary composition determination [61] is imperative to exclude previously isolated compounds.

During a global research expedition (http://www.galathea3.dk/uk), marine bacterial strains were isolated from environmental samples and screened for antagonistic activity against a pathogenic Vibrio anguillarum, strain 90-11-287. Three hundred and one bioactive strains were identified as Vibrionaceae based on 16S rRNA gene similarities [8]. Pure cultures of strains were stored in cryoprotectant solution at −80 °C until being analyzed in the present study.

All 301 strains were retested for antibacterial activity by spotting colony mass on agar seeded with either V. anguillarum strain 90-11-287 or S. aureus strain 8325. Activity was assessed by the formation of clearing zones around spotted colony mass. Selected active strains were grown both stagnant and aerated (200 rpm) in 30 mL Marine Broth 2216 (Difco 279110) for 3 days at 25 °C in 250 mL glass bottles. Cultures were extracted with an equal volume of HPLC-grade ethyl acetate (EtOAc) for 30 min. The organic phase was transferred to fresh sample vials and evaporated under nitrogen until dryness. Extracts were redissolved in 1 mL of EtOAc and stored at −20 °C until further analysis. EtOAc extracts were tested in a well diffusion agar assay [62] for activity against V. anguillarum strain 90-11-287 and S. aureus strain 8325.

Genomic DNA was extracted from 1-day cultures using the NucleoSpin Tissue Kit (Macherey-Nagel, Düren, Germany) according to the manufacturer’s instructions. PCR for recA and toxR gene fragments was performed according to [35], and PCR for rpoA gene fragments according to [34]. PCR products were checked by agarose gel electrophoresis and purified using the Wizard PCR Preps DNA Direct Purification System (Promega, Madison, USA) according to the manufacturer’s instructions. Obtained nucleotide sequences were edited using Chromas Lite (Technelysium, Australia) and aligned to its closest sequence relative [36]. The phylogenetic relationship between the five isolates was determined by neighbor-joining analyses (1000 bootstrap replicates) of nucleotide and amino acid alignments (translated using EMBOSS Transeq, http://www.ebi.ac.uk/Tools/emboss/transeq/) done in ClustalX. Gene sequences have been deposited at GenBank under the accession numbers HQ452614–452618 (toxR), HQ452619–452623 (recA), and HQ452624–452628 (rpoA).

The five strains with strongest antibacterial activity (S2052, S2394, S2753, S4051, and S4053) were grown both stagnant and aerated (200 rpm) at 25 °C in either Marine Broth (MB) or Marine Minimal Medium [63] containing 0.4% glucose and 0.3% casamino acids (MMM). Per strain and culture condition, three bottles were inoculated with 30 mL of medium each, of which each one was sampled after 1, 3, and 5 days of incubation. In addition, strains were grown in 30 mL sea salt solution (Sigma S9883; 40 g L⁻¹) with 0.4% glucose and 0.3% casamino acids for 3 days (200 rpm) at 25 °C. EtOAc extracts were prepared as described above, and tested in a well diffusion agar assay [62] for activity against V. anguillarum strain 90-11-287 and S. aureus strain 8325.

Liquid chromatography-diode array/mass spectrometry (LC-UV/MS) analyses were performed on dried EtOAc extracts redissolved in methanol (MeOH) from all tested culture conditions to visualize the array of produced molecules. In addition, 3-day MMM cultures were extracted and analyzed in biological triplicate. LC-UV/MS was performed on an Agilent 1100 liquid chromatograph with a diode array detector (Agilent, Waldbronn, Germany) coupled to an LCT TOF mass spectrometer (Micromass, Manchester, UK) using a Z-spray ESI source. The separation was done on a Luna II C₁₈ column (50 mm × 2 mm, 3 μm) (Phenomenex, Torrance, CA) fitted with a security guard system using a linear gradient starting from 15% acetonitrile (MeCN) in water (H₂O) to 100% MeCN over 20 min at a flow rate of 300 μL min⁻¹. Both MeCN (HPLC grade) and H₂O were buffered with 20 mM HPLC-grade formic acid (FA).

Strains S2052 and S2753 were grown in 20 L sea salt solution (Sigma S9883; 40 g L⁻¹) with 0.4% glucose and 0.3% casamino acids for 3 days (100 rpm) at 25 °C. On day 3, sterile Dianion HP20SS resin (Sigma-Aldrich, St. Louis, MO) was added to the broth (12 g of resin L⁻¹). After 24 h, the resin was filtered off and washed with H₂O (2 × 1 L), followed by extraction with MeCN/H₂O (80/20 v/v; 2 × 1500 mL).

For S2052, all organic extracts were pooled, absorbed onto 90 g Sepra ZT C18 (Phenomenex), and dried before packing into a 100 g SNAP column (Biotage, Uppsala, Sweden) with pure resin (10 g) in the base. Using an Isolera flash purification system (Biotage), the extract was subjected to a crude fractionation using a MeCN/H₂O gradient (flow rate 30 mL min⁻¹) starting with 10% MeCN (10 min, isocratic), increasing to 100% MeCN (25 min) before washing with 100% MeCN (15 min). Fractions were automatically collected using UV detection (210 and 320 nm). The fraction with antibacterial activity (185 mg) was subjected to further purification on a Luna II C₁₈ column (250 × 10 mm, 5 μm) (Phenomenex) using a 45–70% MeCN/H₂O gradient (buffered with 20 mM FA, flow rate 5 mL min⁻¹) over 20 minutes on a Gilson 322 liquid chromatograph with a 215 liquid handler/injector (BioLab, Risskov, Denmark). This yielded 7.6 mg of pure andrimid.

For S2753, the MeCN/H₂O extract from Dianion HP20SS extraction was evaporated until dryness on a rotary evaporator. The extract was redissolved in EtOAc, absorbed onto 5 g Isolute diol (Biotage), and added to a glass column with pure diol (95 g). A total of 12 fractions were collected from the diol column (100 g, 20 × 350 mm) ranging from heptane, dichloromethane, EtOAc to pure MeOH, running under gravity. The fraction with antibacterial activity (172 mg, 100% EtOAc) was further separated on the Isolera flash purification system, on Sepra ZT C18 (10 g SNAP) using a MeCN/H₂O gradient (flow rate 12 mL min⁻¹) starting with 5% MeCN increasing to 30% MeCN (12 min), quickly increasing to 100% MeCN (10 min). Fractions were automatically collected using UV detection (210 and 380 nm). Pure holomycin (4.3 mg) was obtained after final purification on a Luna II C₁₈ column (250 × 10 mm, 5 μm) (Phenomenex) using a MeCN/H₂O (buffered with 20 mM FA) gradient from 7–37% MeCN over 17 min.

NMR spectra were recorded on a Bruker Avance 800 MHz spectrometer with a 5 mm TCI Cryoprobe at the Danish Instrument Center for NMR Spectroscopy of Biological Macromolecules, using standard pulse sequences. The NMR data used for the structural assignment of andrimid and holomycin were acquired in DMSO-d₆ (δ_H 2.49 and δ_C 39.5 ppm).

Optical rotation was measured on a Perkin Elmer Model 341 polarimeter (Perkin Elmer, Waltham, MA) (α_D at 589 nm).