All 13 strains produced the purple antibacterial pigment, violacein (Table 1). Five strains produced pentabromopseudilin (chemotype 1), three strains produced indolmycin (chemotype 2) and four strains did not produce other known antibiotics (chemotype 3). No strain produced both indolmycin and pentabromopseudilin. Hence, the sub-division previously seen with respect to pentabromopseudilin and indolmycin was confirmed in a larger selection of strains. We have attempted to further broaden our collection; however, we were unable to do so as P. luteoviolacea under certain conditions is autoinhibitory [30] and several laboratories no longer had stock cultures.

Despite the violacein production by all strains, just 11 of the 13 strains inhibited V. anguillarum and S. aureus in the live cell assay (Table 1). This indicates that violacein is not a major antibacterial compound under these conditions. Instead, violacein has been suggested to act as a cell-associated anti-predation compound [31] which is more in line with its low solubility in water and distinct cell-association and offers an ecological role for this compound which may explain the ubiquitous production within strains of P. luteoviolacea. Sterile filtered culture supernatants were tested in well diffusion agar assays. Sterile filtered supernatants were antibacterial only if harvested from strains producing either pentabromopseudilin (6 strains, inhibition profile A) or indolmycin (3 strains, inhibition profile B). Both indolmycin and pentabromopseudilin containing supernatants inhibited S. aureus, whereas only indolmycin containing supernatants inhibited the Gram-negative V. anguillarum (Table 1). This is in agreement with previous studies describing pentabromopseudilin as a compound targeting Gram-positive bacteria, whereas indolmycin, previously only isolated from Streptomyces species [32,33], inhibits both Gram-positive and Gram-negative bacteria and is very potent against staphylococci [34]. The supernatants from the remaining four strains were not inhibitory against either of the target organisms and these four strains produced neither indolmycin nor pentabromopseudilin. Discrepancies among live cell and sterile supernatant inhibition profiles (inhibition profile C) suggest that additional antibiotic compounds may be produced, for instance macromolecular antibiotics such as L-amino acid oxidases which are produced by some P. luteoviolacea strains [35].

Phylogenetic trees were created based on 16S rRNA gene sequences alone or concatenated 16S rRNA, gyrB, and recA gene sequences. The 16S rRNA sequences were identical at a level of >99% and the phylogenetic reconstruction showed no correlation to production of antibiotics within clusters (Figure 1). The resolving power of the 16S rRNA gene is limited when performing phylogenetic analyses below the species level [36] whereas studies have shown the usefulness of housekeeping genes such as recA and gyrB [37,38,39] in resolving sub-species phylogenetic relations. Seperate analyses of recA and gyrB genes were carried out to verify that each gene contributed to phylogenetic delineation (data not shown).

A phylogenetic analysis of concatenated 16S rRNA, gyrB and recA genes led to identification of three clusters (Figure 2) as indicated in Table 1 under genotypes. Genotype I represents 4 of the 6 pentabromopseudilin producing strains of chemotype 1 and inhibition profile A. Genotype II is identical to chemotype 2 and inhibition profile B. Genotype III is heterogeneous, covering two strains of chemotype 1 with inhibition profile A, two strains of chemotype 3 with inhibition profile C and two strains of chemotype 3 with inhibition profile D. Within genotype III, sub-clusters tightly reflected associations to chemotype and inhibition profile.

A correlation among taxonomic units and secondary metabolite synthesis is well established in the world of fungal natural products [40], and was also seen among marine actinomycete Salinispora spp. Indeed, within actinobacteria bacterial systematics are emerging as a tool to aid in biodiscovery by focusing on the use of bacterial systematics to dereplicate actinobacterial species in order to focus on novel diversity and its potential for novel chemistry. To our knowledge, this approach has not been tested within the γ-Proteobacteria, or at the infraspecies level within a group of closely related strains as presented here for P. luteoviolacea.

To investigate the potential role of indolmycin and violacein as antibacterial compounds, the MICs against V. anguillarum, S. aureus, and the 13 P. luteovioloacea strains were determined. Pentabromopseudilin was not included as the compound was unavailable as commercial reference standard. Under the experimental conditions in this study violacein did not inhibit any of the included strains at tested concentrations; hence the MIC of violacein for all tested strains was >128 μg/mL. This is in agreement with the lack of inhibition observed in some violacein producing strains (Table 1). The MIC of indolmycin to V. anguillarum was 2 μg/mL whereas S. aureus was inhibited by indolmycin at all tested concentrations, resulting in a MIC of ≤0.5 μg/mL. None of the P. luteoviolaceae, irrespective of antibiotic profile, were inhibited by indolmycin leading to a MIC of >128 µg/mL.

Biosynthetic pathways are subjects of horizontal gene transfer events and their organization in genetic clusters facilitates the exchange of entire biological pathways among bacterial strains [41,42,43]. This intuitively supports the common expectation that secondary metabolite production is strain specific [44] and hence bacterial systematics would not be expected to be related to the secondary metabolome. In contrast, distinct species-specific secondary metabolite profiles were observed within the genus Salinispora and it was suggested that biosynthetic gene clusters are stably maintained in these species due to the distinct competitive advantage obtained through antibiosis and may in fact represent niche-specific adaptations [45]. Further studies of the biosynthetic pathways behind the antibiotic production of P. luteoviolacea and their genomic context are likely to add to our understanding of these concepts.

Thirteen P. luteoviolacea strains were used in this study (Table 2). They originated from distinct geographical areas and were primarily isolated from surface water or algae. The strains were cultured in a marine minimal medium (MMM) [46] with 4 g/L mannose and 3 g/L casamino acids. All strains were incubated at 25 °C and 200 rpm agitation.

Vibrio anguillarum 90-11-287 [47] and Staphyloccoccus aureus 8325 were cultured in tryptic soy broth (Difco, USA). Well diffusion agar assays were carried out as previously described. The assay substrate contained 30 g/L Sea Salts (Sigma, USA) and 10 g/L agar. To support growth of V. anguillarum 4 g/L glucose and 3 g/L casamino acids was added. An additional 5 g/L peptone was added to S. aureus agar. Antibiosis by live cells was tested by spotting colony mass directly onto the assay plates and observing if clearing zones had formed following incubation for 24 h.

Samples for LC-MS analyses were prepared from cultures in MMM extracted with ethyl acetate (EtOAc). Extracts were dried under nitrogen and redissolved in methanol (MeOH). LC-MS samples were analysed using an Agilent 1100 HPLC system with a diode array detector (Waldbronn, Germany) coupled to an LCT TOF mass spectrometer (Micromass, Manchester, UK) using a Z-spray electrospray (ESI) source. A Phenomenex Luna II C₁₈ column (50 mm × 2 mm, 3 μm) was used for separation, applying a linear acetonitrile (MeCN)-water (20 mM formic acid) 0.3 mL∙min⁻¹ gradient (15%–100%) over 20 min at 40 °C. For all LC-MS analyses, violacein and indolmycin were detected in positive ionisation mode (ESI⁺), while pentabromopseudilin was detected in negative mode (ESI⁻).

DNA was purified from overnight P. luteoviolacea cultures using the NucleoSpin Tissue kit (Machery-Nagel, Germany) or QIAGEN Genomic-Tip G/100 (QIAGEN, USA) following the manufacturer’s protocol. 16S rRNA genes from strains H33, H33S, NCIMB 1942, NCIMB 1944 and NCIMB 2035 and gyrB and recA genes from all strains were amplified by PCR. One reaction consisted of 2.5 µL 10× Hot Start PCR buffer (Fermentas, Canada), 2.5 µL 2 mM dNTP mix, 4 µL 25 mM MgCl₂, 0.8 µL 12.5 µM forward primer, 0.8 µL 12.5 µM reverse primer, 12.28 µL MilliQ H₂O, 0.2 µL Maxima Hot Start Taq DNA polymerase (Fermentas, Canada) and 1 µL DNA template at 50 ng/µL for a total volume of 25 µL. The reactions were performed on an Applied Biosystems Veriti 96 well cycler. 16S rRNA genes were amplified according to. The primers and reaction conditions used for gyrB amplification were as described in [48]. Primers and conditions for amplification of recA fragments were according to [49]. Sequencing was done by Eurofins MWG Operon, Germany. Nucleotide sequences generated in this study were deposited in GenBank under accession numbers: 16S rRNA JQ250820-JQ250824, gyrB JQ280430-JQ280442 and recA JQ280417-JQ280429. Accession numbers for 16S rRNA nucleotide sequences previously available from GenBank: NR_026221, FJ457234, FJ457187, FJ457230, FJ457238, EU158365, EU158366 and FJ952782.

Analysis of concatenated gene sequences, termed multi locus sequence analysis, has succesfully been used to infer robust phylogenies within, e.g., the Vibrionaceae [49], and a similar approach was used in this study. The 16S rRNA gene sequences were obtained from GenBank or by sequencing and aligned in MEGA5 using MUSCLE [50]. The evolutionary history was inferred using the maximum likelihood method based on the Kimura 2-parameter model [51]. The phylogenetic tree was constructed using MEGA5 default settings. For analysis of recA and gyrB sequences, an alignment was created for each gene in MEGA5 using MUSCLE. The alignments were curated manually and trimmed to be of equal length and in-frame. 16S rRNA, gyrB and recA alignments were concatenated and phylogenetic analyses were performed in MEGA5. Evolutionary relationships based on the nucleotide sequences were inferred using the maximum likelihood method with the general time reversible model [52], assuming a gamma distributed substitution rate with five discrete categories. Neighbor-joining phylogenetic trees were generated and tested with 1000 bootstrap replications.

MICs of commercially available violacein (Sigma, USA) and indolmycin (Bioaustralis, Australia) standards to P. luteoviolacea strains, V. anguillarum 90-11-287, and S. aureus 8325 were tested. MIC assays were carried out in 96-well microtiter plates according to the guidelines by the clinical and laboratory standards institute [53], with minor modifications. P. luteoviolacea was cultured in MMM with 4 g/L mannose and 3 g/L casamino acids, V. anguillarum in MMM with 4 g/L glucose and 3 g/L casamino acids, and S. aureus in MHB. Overnight bacterial cultures were diluted to 10⁵ CFU/mL and 90 µL added per well. 10 µL of antibiotic solution was added to each well. The antibiotics were tested in serial two-fold dilutions at final concentrations of 0.5 µg/mL to 128 µg/mL. Controls were included for no antibiotic and ethanol solvent. The well containing the lowest concentration of antibiotic that had no visual bacterial growth after 48 h corresponded to the MIC.