Two approaches were used to isolate bacteriocin-producing marine bacteria from a range of seaweeds, as well as one sample each of sand and seawater. The first yielded a total of 303 isolates, which were selected based on colony morphology and subsequently screened for antimicrobial activity against Staphylococcus aureus, Salmonella Typhimurium, Listeria innocua, Escherichia coli and Bacillus subtilis. Only seven of these (2.3%) displayed consistent activity against at least one of the target strains and were selected for further analysis. Using the second approach, more than 6000 bacterial colonies were screened against the same targets using a high throughput, combined isolation and antimicrobial detection assay. This led to the recovery of eight isolates (0.1%) with confirmed antimicrobial activity against at least one indicator organism. Half of these were generated by heating the samples prior to plating to select for spore-forming bacteria. To establish if the antimicrobial agents were released into the culture supernatant and to determine the inhibitory spectra, activity of the cell-free supernatants (CFS) was tested against the initial indicator strains as well as Listeria monocytogenes, Lactococcus lactis, Lactobacillus bulgaricus, Enterococcus faecium, Enterococcus faecalis, Pseudomonas aeruginosa, methicillin-resistant S. aureus (MRSA), Cronobacter sakazakii and Clostridium difficile. In total, the CFS of all 15 isolates displayed consistent antimicrobial activity against at least two indicator strains. Nine of these were isolated from marine agar (MA) (four of which were obtained by selecting for sporeformers), three from low nutrient agar (LNA), two from Actinomycete isolation agar (AIA) and one from laboratory-prepared marine agar (LPMA) (Table 1). The green seaweed Ulva (both U. lactuca and Ulva spp.) yielded the greatest number of antimicrobial producers (6), followed by Polysiphonia lanosa (3), Fucus vesiculosus (2), Palmaria palmata (1) and Fucus serratus (1) (Table 1).

All antimicrobial-producing isolates were Gram-positive rod-shaped, sporulating bacteria (large central spores) which were catalase positive, had variable oxidase activity and displayed a range of colony morphologies (data not shown). 16S rRNA gene sequencing revealed that the isolates were either Bacillus pumilus (7 isolates) or licheniformis (8 isolates) (Table 1; Figure S1). Comparison of PFGE fingerprints generated with NotI demonstrated that B. pumilus WIT 560 and WIT 561 were essentially the same strain (Figure S2), whereas, the remainder of the marine isolates differed from one another. These findings were confirmed by further PFGE typing using XbaI- and in some cases ApaI-digested DNA, as not all isolates were digested by the latter (data not shown).

All 15 of the selected antimicrobial-producing marine isolates inhibited the growth of Lc. lactis HP and Lb. bulgaricus LMG 6901 when the CFS was assessed by the agar well diffusion assay (WDA) (data not shown). Moreover, 13 isolates had antimicrobial activity against S. aureus, 12 against L. innocua, 13 against L. monocytogenes, six against B. subtilis and three against E. faecium in the WDA and eight inhibited E. faecalis in the deferred antagonism assay (Table 1). Nine isolates displayed activity against MRSA, five of which (WIT 560, 561, 570, 571 and 572) produced a zone of inhibition with a radius of up to 3 mm (Table 1). All of the isolates that were active against MRSA also inhibited S. aureus DPC 5246. None of the CFS from any of the marine isolates inhibited any of the Gram-negative indicator bacteria tested but nine isolates displayed activity against E. coli, S. Typhimurium and/or C. sakazakii when assessed using the deferred antagonism assay (Table 1). WIT 560 and WIT 561 had the broadest spectra of inhibition, as they were active against 10 different indicator bacteria, both Gram-positive and -negative (Listeria, S. aureus, MRSA, Lc. lactis, Lb. bulgaricus, E. faecalis, C. sakazakii, E. coli and S. Typhimurium). Similarities between the inhibitory spectra of these isolates support our assertion that they represent the same strain of B. pumilus.

In order to investigate if the marine bacteria produce a previously identified bacteriocin, PCR analysis was performed using primer pairs specific for known Bacillus bacteriocins. Nine of these were designed in the present study and six had previously been used by others (Table 3). The marine isolates were also screened for the presence of three lantibiotic-associated genes. PCR analysis using primers designed to amplify the bli04127 gene, which is the structural gene of the Bliα lichenicidin peptide and part of the lichenicidin gene cluster, yielded amplicons of the appropriate size when B. licheniformis WIT 562, 564 and 566 were tested (Figure 1a). This was in agreement with results obtained for the known lichenicidin producer, B. licheniformis ATCC 14580. Sequencing of the amplicons from WIT 562 and 566 revealed 100% homology to the bli04127 gene (Figure 1b). The same was true for the ATCC 14580 strain (the known lichenicidin producer). However, analysis of the WIT 564 sequence revealed one nucleotide difference. This difference, however, does not result in any amino acid change and the amplicons from all three marine isolates, as well as the lichenicidin-producing control strain, are predicted to encode the Bliα propeptide. This confirms that the three marine isolates harbor the lichenicidin-encoding operon. All other PCR assays for other known bacteriocins failed to generate appropriately sized amplicons from any of the marine isolates. The relevant positive control strains did, however, yield amplicons of the expected size (data not shown), indicating that the PCR conditions were appropriate to amplify the genes, had they been present. These data indicate the absence of the associated bacteriocin gene clusters from the marine isolates.

In order to confirm that B. licheniformis WIT 562, 564 and 566 produce lichenicidin, the two-peptide bacteriocin was extracted from all three marine isolates, purified and compared to lichenicidin produced by B. licheniformis ATCC 14580. HPLC fractions of extracts from the marine-derived bacteria with masses corresponding to those previously determined for the Bliα and Bliβ lichenicidin peptides (~3021 Da and ~3251 Da, respectively [12]) did not show activity when each was tested separately (Figure 2a). However, activity was detected when the fractions were combined. This contrasted somewhat with the corresponding fractions from the positive control strain B. licheniformis ATCC 14580, each of which had antimicrobial activity, which was enhanced when tested in combination (Figure 2b).

All chemicals were purchased from Sigma Aldrich (Dublin, Ireland) unless otherwise stated. Bacterial strains used as indicators for antimicrobial characterization and as positive controls for bacteriocin production and their respective growth conditions are listed in the table in Table S3.

Samples of seven macroalgae, namely Ascophylum nodosum, F. serratus, F. vesiculosus, P. palmata, P. lanosa, U. lactuca and Ulva spp. were collected during low tide from October 2009 to April 2010 from two locations in Fethard Bay, Co. Wexford, on the south east coast of Ireland (N52°10′, W06°50′ and N52°11.5′, W06°49.3′). One sample of seawater and one sample of sand were also collected from one from these locations on one occasion (March 2010). Samples were stored in the dark during transport to the laboratory and were processed within 2 h of sampling.

Approximately 10 g of each seaweed or sand sample was washed twice in sterile artificial seawater [3.33% (w/v) Instant Ocean Salt, Aquarium Systems, Sarrebourg, France] for 2 min at 100 rpm to remove loosely attached bacteria and debris. Samples were then homogenized as 10-fold dilutions in sterile artificial seawater for 4 min in a stomacher (Masticator, IUL Instruments, Barcelona, Spain). Ten-fold serial dilutions of these homogenates and the seawater sample were then prepared in artificial seawater and appropriate dilutions were spread-plated on the following media: (1) MA [Difco marine broth 2216 (Becton, Dickinson and Company (BD), Franklin Lakes, USA) with 1.5% (w/v) agar (Oxoid, Basingstroke, Hampshire, UK)], (2) Gram-negative isolation agar (GNA) [MA with 100 µg/mL novobiocin], (3) LNA [3.33% (w/v) Instant Ocean Salt, 0.05% (w/v) tryptone (BD), 0.005% (w/v) yeast extract (Merck, Darmstadt, Germany), 0.01% (w/v) β-glycerol phosphate disodium salt, 1.5% (w/v) agar], (4) LPMA [0.5% (w/v) peptone (BD), 0.1% (w/v) yeast extract, 3.33% (w/v) Instant Ocean Salt, 0.01% (w/v) ferric citrate, 1.5% (w/v) agar], (5) AIA and (6) starch yeast peptone seawater agar (SYP-SW) [32]. Nystatin (50 units/mL) was added to all media to inhibit the growth of yeasts and molds. Spore-forming bacteria were also isolated by heating the seaweed homogenates for 15 min at 80 °C prior to spread-plating on MA. Five sets of spread plates were prepared in this way on each medium (except for GNA for which three sets were prepared) so that each set could be overlaid with a different indicator strain, as outlined below. All plates were incubated at 28 °C for up to 6 weeks, during which they were checked periodically for growth.

Two approaches were used to detect and isolate bacteriocin-producing bacteria. In the first approach, for each sample up to 10 colonies with different morphologies were picked from each medium (except for the MA on which spore-forming bacteria were selected) and re-streaked at least twice onto the corresponding medium to obtain pure cultures. Colony morphologies were noted and isolates were stocked at −20 °C in marine broth supplemented with 40% (v/v) glycerol. These isolates were subsequently screened for antimicrobial activity using the deferred antagonism assay outlined below.

In the second approach, plates with <300 colonies, from which colonies were not selected in the first approach, were overlaid with soft BHI agar (0.75% w/v agar) seeded with a 0.25% inoculum of an overnight culture of either S. aureus DPC 5246, S. Typhimurium LT2, L. innocua WIT 361 or E. coli DSM 10720 or a 0.75% inoculum of an overnight culture of B. subtilis ATCC 6633. Plates were incubated overnight at 37 °C. Any colony which was surrounded by a clear zone of inhibition was picked off from under the agar overlay and inoculated into marine broth to kill off any contaminating indicator bacteria. Following overnight incubation at 28 °C, these cultures were then streaked in duplicate onto MA. One plate was overlaid with the same indicator strain against which inhibitory activity was initially observed. If activity was confirmed, the isolate was re-streaked onto MA from the duplicate plate and stocked at −20 °C in marine broth supplemented with 40% (v/v) glycerol. Antimicrobial production was then confirmed using the deferred antagonism assay outlined below.

The deferred antagonism assay was used to screen marine bacteria isolated using the first approach outlined above for antimicrobial production and also to confirm antimicrobial production in bacteria recovered using the second approach. For this assay, the marine isolates were grown overnight at 28 °C in marine broth agitated at 200 rpm. Five microliters of each culture was spotted onto MA and incubated at 28 °C until culture spots were approximately 0.5–1 cm in diameter. A number of sets of plates were prepared in this way and each set was overlaid with soft BHI agar seeded with one of the five indicator strains, as outlined above, as well as Lc. lactis HP, Lb. bulgaricus LMG 6901, P. aeruginosa PA01 or L. monocytogenes WIT 041 seeded into the appropriate soft agar (Table S3) at a rate of 0.25% (Lb. bulgaricus, P. aeruginosa, L. monocytogenes) or 0.5% (Lc. lactis). This assay was also used to test activity against E. faecalis ATCC 19433 and C. sakazakii ATCC 12868, except that the marine isolates were grown in BHI broth, spotted onto BHI agar and the E. faecalis indicator was seeded at a rate of 0.25% into soft MRS agar supplemented with 0.05% (w/v) L-cysteine hydrochloride. Any isolate producing a clear zone of inhibition against any of the indicators was investigated further.

To determine if antimicrobial compounds were released into culture supernatants, CFS from marine isolates with confirmed antimicrobial activity were tested using the WDA, as described by Begley et al. [12] except that wells were 6 mm in diameter and CFS was obtained by centrifuging cultures at 18,620× g for 10 min, removing the supernatant and centrifuging again. The CFS was tested against all of the indicator strains listed in Table S3 (except for B.megaterium, thuringiensis, licheniformis, halodurans, cereus and subtilis A1/3, 168 and HIL Y85,54728) in order to determine the spectra of inhibition. All indicator strains were used at an inoculum of 0.25% (v/v) except for Lc. lactis (0.5%), B. subtilis (0.75%) and Cl. difficile (1.25%). Plates were incubated for 20 h at the appropriate temperature and examined for zones of inhibition. Zones, where present, were measured. Cross sensitivity assays were performed to test the marine isolates for activity against each other and also against known bacteriocin-producing strains (B. licheniformis ATCC 14580, B. subtilis ATCC 6633, B. megaterium 216, B. cereus CECT 5148, B. halodurans ATCC BAA-125D-5 and Lc. lactis NZ9700). Well diffusion assays were employed, using MA for marine isolates or BHI for bacteriocin-producing strains, both seeded with a 0.75% (v/v) inoculum of the test culture.

Sensitivity of the antimicrobial compounds to proteolytic enzymes and catalase was determined as outlined by O’Shea et al. [15] except that the enzyme solutions were 10 mg/mL (giving a final enzyme concentration of 5 mg/mL), protease Type I and pronase E were also used and enzymatic treatments were performed at 37 °C for 2 h. Lc. lactis NZ 9700 CFS was used as a positive control and Lc. lactis HP was used as the indicator organism. The heat stability of each antimicrobial was determined by assaying the antimicrobial activity of CFS by WDA following heating to 40, 50, 60, 70, 80, 90 or 100 °C for 30 min in a heating block and to 121 °C for 15 min in an autoclave. Sensitivity to pH was investigated by adjusting the pH of the CFS to 2, 3, 5, 7, 9, 11 or 12, incubating at 28 °C for 2 h and assessing activity using the WDA. For all of the above assays, antimicrobial activity was expressed in arbitrary units (AU) as defined by Ryan et al. [33] and the percentage of residual activity was calculated, where 100% activity was the activity measured prior to treatment.

Antimicrobial-producing marine isolates were cultured overnight at 28 °C and agitated at 200 rpm in the following media: BHI, Difco Marine broth 2216, Luria-Bertani (LB) broth (Merck), tryptic soy broth (BD), nutrient broth (Oxoid), Bacillus production medium [34] and Actinomycete isolation broth [0.2% (w/v) sodium caseinate, 0.01% (w/v) L-asparagine, 0.4% (w/v) sodium propionate, 0.05% (w/v) dipotassium phosphate, 0.01% (w/v) magnesium sulphate, 0.0001% (w/v) ferrous sulphate and 0.5% (v/v) glycerol]. Antimicrobial production was assessed by the WDA using Lc. lactis HP as the indicator organism.

Marine isolates were initially characterized phenotypically by Gram staining, spore staining and oxidase and catalase tests. In addition, 16S rRNA gene sequencing was performed on each isolate, by amplifying the 16S rRNA gene in a 15 µL reaction volume using primers 63f and 1387r, purifying the PCR products and sequencing them on an AB 310 genetic analyzer (Applied Biosystems, CA, USA), as outlined by Coffey et al. [35]. Nucleotide sequences were compared with those in the GenBank database using the blast program [36] through the National Center for Biotechnology Information (NCBI) server [37]. TreeView 1.6.6 [38] was used to construct a phylogram from a ClustalW alignment (ClustalX) [39].

Molecular fingerprinting of the marine isolates was then performed by pulsed-field gel electrophoresis (PFGE), as described previously [40], using ApaI, XbaI and NotI. Electrophoresis conditions used were 200 V for 17 h with a pulse time from 1 to 15 s for ApaI; 10 h at 200 V with a pulse time from 1 to 15 s for XbaI and 15 h at 200 V with a pulse time from 1 to 12 s for NotI. A low-range molecular weight DNA marker (0.13–194.0 Kb; New England Biolabs, Beverly, MA) was used.

PCR was used to screen the marine isolates for the presence of genes encoding known Bacillus bacteriocins, as well as general lantibiotic genes. Genomic DNA was extracted from overnight cultures using a DNeasy 96 blood and tissue kit (Qiagen, Crawley, UK). The PCR primers used are listed in Table 3 and were synthesized by Eurofins (MWG Operon, Ebersberg, Germany). Some were designed in the present study based on gene sequences associated with previously identified bacteriocins obtained from Genbank. Sequences of the other PCR primers were obtained from previous studies (Table 3). Annealing temperatures were optimized using a gradient thermal cycler (Applied Biosystems). PCR conditions were as follows; initial denaturation at 95 °C for 5 min, followed by 35 cycles of 95 °C for 1 min, appropriate annealing temperature (Table 3) for 1 min, 72 °C for 1 min, followed by a final extension at 72 °C for 5 min.

PCR products obtained using lichenicidin primers were firstly cleaned using a DNA Clean and Concentrator™-5 kit (Zymo Research, Irvine, CA, USA) and then cloned using a PCR cloning kit (Qiagen) as per the manufacturer’s instructions. Recombinant plasmids were purified using the GenElute™ plasmid miniprep kit, as per the manufacturer’s instructions. T7 promoter and M13 forward (−20) primers were used for the sequencing reactions of the cloned PCR products. DNA sequencing was performed as described by Coffey et al. [35] and the sequences obtained were analyzed using BLAST software [36] from the Genbank (NCBI) database. Nucleic acid and deduced amino acid sequences were analyzed with DNASTAR software (DNASTAR Inc., Madison, USA).

Lichenicidin was extracted and purified from the marine isolates found to harbor the bli04127 gene that encodes one of the lichenicidin peptides and B. licheniformis ATCC 14580 (known lichenicidin-producing strain) using solid phase extraction followed by reverse-phase HPLC, as described by Begley et al. [12] with the following modifications; the HPLC column was developed in a gradient of 25% to 60% acetonitrile-0.1% (v/v) trifluoroacetic acid at a flow rate of 2.5 mL/min. HPLC fractions were analyzed by matrix-assisted laser desorption ionization time of flight (MALDI TOF) mass spectrometry, as outlined by Begley et al. [12]. Fractions with masses corresponding to those previously determined for lichenicidin [12] were tested for antimicrobial activity (individually and in combination) against Lc. lactis HP using the WDA.