The strains Bacillus sp. C3 and P5 were isolated from puba as described below. The maintenance of bacterial strains was performed in Brain Heart Infusion broth (BHI, Difco, Sparks, NV, USA) containing 20% (v/v) glycerol at -20°C. The indicator strains used for evaluation of antibacterial and antifungal activity were selected for their importance as human and animal pathogens or relevance as food spoilage agents. The strains are listed in Table 1.

Bacterial strains were isolated from samples of puba that were obtained from different batches of the same producer. The samples were obtained in the District of Saco da Raiz in Estância, a town of Sergipe State, located in northeast of Brazil. The strains were selected among Gram-positive rods obtained from puba samples, where five isolates were presumptively identified as Bacillus species based on standardized methods including observation of cell morphology Gram-staining, phase-contrast microscopy for detection of parasporal crystal proteins formation, and catalase activity (Food and Drug Administration-Bacteriological Analytical Manual [FDA-BAM], 2012). Additional tests were conducted using the identification kits API 50 CHB and API 20E (BioMérieux SA, Marcy-l’Étoile, France). The results were analyzed by the API LAB Plus software for strain identification (BioMérieux SA). Following, the antimicrobial potential of these strains was tested against varied bacteria, yeasts and filamentous fungi (data not show). The strains C3 and P5, which presented the most promising results, were selected for the subsequent tests.

Total cellular fatty acids from the isolates Bacillus sp. P5 and C3 were analyzed using the MIDI Sherlock^® Microbial Identification System (Microbial Identification System, Microbial ID Inc., Newark, NJ, USA). Fatty acid extraction and methyl ester generation were performed with Instant FAME Method kit according to the manufacturer’s instruction. Gas chromatography (GC) was performed on an Agilent 6890N analyzer using calibration standards (#1300-AA; MIDI, Inc.). The Sherlock^® Microbial Identification (MIDI, Inc., version 4.5) software was used to assign GC peaks to individual fatty acid structures. The identification was made by comparative fatty acids with database to Instant Environmental TSA library (ITSA1) version 1.10. Samples with a similarity index (SI) ≥ 0.5 were considered as an acceptable FAME identification (Kunitsky et al., 2006).

Total DNA from Bacillus sp. C3 and P5 was extracted from overnight cultures of strains using the Promega Wizard SV Genomic DNA kit (Promega, USA) and the amplification of 16S rRNA gene was performed using the universal primers 27F (5′-GAGTTTGATCCTGGCTCAG-3′) and 1525R (5′-AGAAAGGAGGTGATCCAGC C-3′), according to Lisbôa (2006). The PCR conditions were: initial denaturation for 5 min at 95°C, 30 cycles of 30 s at 95°C for denaturation, 1 min 30 s at 46°C for annealing, 80 s at 72°C for extension and 7 min at 72°C for final extension (adapted from Horisawa et al., 2009). The amplicons were sequenced by the ATCGene Laboratory (Porto Alegre, Brazil). The sequences obtained were submitted to the BLAST search algorithm¹, edited using Bioedit software and aligned with Clustal X. For the construction of the dendrogram, sequences were checked for quality, aligned and analyzed using the software Phred v.0.20425 (Ewing and Green, 1998), Phrap v.0.900319 (Gordon et al., 2001) and Consed 12.0 (Gordon et al., 1998). The phylogenetic tree was developed using the Neighbor-joining method present in MEGA version 5.0 (Kumar et al., 2004). Genetic distance was calculated based on Kimura two-parameter model of nucleotide evolution. The support of nodes was estimated using 1000 bootstrap replicates. The sequences obtained were deposited in GenBank under the accession numbers JX456531 for Bacillus sp. C3 and JX456530 for Bacillus sp. P5.

Gene amplification of sfp, ituD, sboA, and spaS was performed by PCR with specific primers, as described by Hsieh et al. (2008) and Velho et al. (2011). The following parameters were used: for ituD (iturin A), denaturation at 94°C for 1 min., annealing for 1 min at 50°C, elongation at 1.5 min at 72°C, in a total of 30 cycles; for sfp (surfactin), denaturation for 1 min at 94°C, annealing for 30 s at 46°C for 1 min and elongation at 72°C for a total of 25 cycles. For sboA (subtilosin A), the following parameters were used: 1 min denaturation at 94°C, annealing for 30 s at 50°C for 1 min and elongation at 72°C for a total of 35 cycles; for spaS (subtilin), 1 min denaturation at 94°C, annealing for 30 s at 55°C for 1 min and elongation at 72°C for a total of 35 cycles. The amplified products were sequenced by ATCGene Laboratory (Porto Alegre, Brazil). The sequences obtained were submitted to the BLAST search algorithm² and edited using Bioedit software for contigs assembly.

Erlenmeyer flasks of 125 ml containing 30 ml of BHI broth were inoculated with a loop of Bacillus strains cultivated on BHI agar. The inoculum was pre-incubated for 24 h at 125 rpm and 37°C. An aliquot of 1% (v/v) of this culture was transferred to a 500 ml Erlenmeyer flask containing 200 ml of BHI broth. The culture was incubated for 48 h at 42°C with constant stirring at 125 rpm. After this period, the culture was centrifuged for 15 min at 10,000 g. The supernatant was sterilized by filtration through a cellulose filter with a pore size of 0.22 μm (for small volumes) or by vacuum filtration through a 0.22 μm silica filter (for larger volumes). The filtrates were kept at 4°C.

The antimicrobial activity of crude supernatants was detected by a modified diffusion assay (Kimura et al., 1998; Motta and Brandelli, 2002). Aliquots (20 μl) of the crude supernatants of Bacillus sp. C3 and Bacillus sp. P5 were applied onto BHI agar plates previously inoculated with a cell suspension (10⁸ colony forming units CFU/ml – corresponding to 0.5 of McFarland scale) of the indicator microorganism. Zones of inhibition were measured after incubation for 24–48 h under optimal growth conditions for the indicator strain. The inhibitory zones were measured with a digital pachymeter. The test was performed in duplicate with two supernatants obtained from different culture. For this assay, the indicator strains are shown in Table 1.

The fungal strains selected as indicators for this experiment were Fusarium oxysporum f. sp. lycopersici, Aspergillus fumigatus, Aspergillus flavus, Aspergillus niger, and Candida tropicalis (Table 1). They were inoculated on potato dextrose agar (PDA) plates and incubated for 72 h at 30°C. Spore suspensions were prepared for each fungus according to Lopes et al. (2011), with the exception of the yeast C. tropicalis, which was prepared by the same method used for the antibacterial activity assay. After the preparation of the suspensions, a final concentration of 10⁶ spores/ml was mixed with melted PDA at 45°C. Then, 15 μl of the filtrates were added, and the plates were incubated at 30°C for 48 h (Rouse et al., 2008). The tests were performed in triplicate.

The method used by Hsieh et al. (2004) was modified as follows. Bacillus sp. C3 and Bacillus sp. P5 were cultivated for 24 h in tryptic soy broth medium (TSB, Difco) at 37°C and inoculated later in Tryptic casein Soy Agar (TSA, Difco) plates containing 5% (v/v) sheep blood and incubated at 37°C for 24 h. The presence of a zone of hemolysis around the colony was observed for strains producing a biosurfactant.

The evaluation of emulsification (Cooper and Goldenberg, 1987) was performed using cultures and growth culture supernatants of Bacillus sp. C3 and Bacillus sp. P5 grown in BHI medium at 37°C for 24 h. For this test, 2 ml of culture or growth culture supernatant were mixed with 3 ml of hydrophobic compounds (mineral oil, xylene, toluene or soybean biodiesel) in test tubes with flat bottom (100 mm × 15 mm), the mixture was vortexed for 2 min and the flasks left to stand for 24 h. After this period, the emulsifying index was calculated by the following formula (Velho et al., 2011):

where E₂₄ = emulsification index, H_e = height of the emulsified column, H_t = total height.

The surface tension was measured in the absence of microbial cells, which were removed by centrifugation at 10,000 g for 15 min. The samples were maintained for 30 min at room temperature and surface tension was determined using a digital tensiometer (Gibertini, Milan, Italy) using the Wilhelmy plate method (Biswas et al., 2001). Distilled water (72 mN m^-1) and ethanol (24 mN m^-1) were used as standards (Cerqueira et al., 2012).

Cell-free supernatants were processed in two different ways: lipopeptides were isolated by a combination of acid precipitation and solvent extraction procedure following Cooper et al. (1981) and Vater et al. (2002), and bacteriocin were extracted with n-butanol as described by Kawulka et al. (2004). In brief, cells were removed from the 6, 30, and 36 h growing culture in BHI broth by centrifugation (13,000 g) for 15 min at 4°C. For the acid extraction the supernatant was adjusted to pH 2.0 by addition of HCl and allowed to precipitate at 4°C for 16 h. Precipitate was collected after centrifugation (13,000 g) for 20 min at 4°C and extracted with dichloromethane. The lipopeptide containing dichloromethane fraction was collected after filtration and vacuum-dried. Alternatively, the supernatant was extracted by adding one-quarter the volume of n-butanol, shaked for 1 h, and then poured into a separator funnel and allowed to stand overnight. The organic layer was separated, concentrated under vacuum and the residue suspended in methanol (10 ml per liter of cell culture) (Kawulka et al., 2004).

Twenty grams of puba were weighted and suspended in 40 ml of distilled water and homogenized. Then, this suspension was submitted to extraction with n-butanol as described previously. Zip Tip^® pipette was standardized with 10 μl acetonitrile, 10 μl1:1 (v/v) acetonitrile/0.1% trifluoroacetic acid (TFA) and twice with 10 μl of 0.1% (v/v) TFA. The solution with the lipopeptides extracted from puba was loaded to the pipette and the eluate was discarded. The retained material was washed three times with 10 μl of 0.1% (v/v) TFA and the sample was eluted with 5 μl of acetonitrile/0.1% TFA (60:40, v/v). The eluate was analyzed by mass spectrometry.

The samples were dissolved in methanol and fractionated by reversed phase chromatography. High-performance liquid chromatography (HPLC) was performed on a Shimadzu Prominence equipment (Department of Biochemistry, Immunology, Institute of Biological Sciences, Federal University of Minas Gerais) with a SPD-20A-UV/VIS detector. The column was a Sephasil^TM Peptide C18 (5 μm ST 4.6/250 C18, 100-Å pore size, 5 μm particle size). The elution condition was: 0–10 min mobile phase A (0.05% TFA); 10–40 min a gradient of 0–100% mobile phase B (acetonitrile + 0.05% TFA); and 40–50 min mobile phase B. A flow rate of 1 ml/min was used. Column effluent was monitored at 220 and 280 nm, and fractions were checked for antimicrobial activity against L. monocytogenes ATCC 7644. The active peaks were selected for further study, including chemical and mass spectrometry analyses. All HPLC solvents were prepared fresh daily and filtered under vacuum before use. All aqueous solutions were prepared with ultrapure water.

Active fractions against L. monocytogenes ATCC 7644 were freeze-dried and suspended in 50 μl of ultrapure water, and then 0.7 μl of this solution was mixed with 0.7 μl of a saturate matrix solution of α-cyano-4-hydroxy-cinnamic acid (Sigma-Aldrich, St Louis, MO, USA) or 2-5-dihydrobenzoic acid (DHB, Fluka). The matrix solutions were prepared in 1:1 (v/v) CH₃CN:H₂O containing 0.1% TFA. The mixtures were spotted onto a MALDI-TOF sample plate (Bruker Daltonics, Inc., Department of Biochemistry, Immunology, Institute of Biological Sciences, Federal University of Minas Gerais), at room temperature. Analysis was performed in a mass spectrometer by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF, AUTOFLEX III, Bruker Daltonics) in the positive reflective mode, using the Flex Control 3.3 software (Bruker Daltonics, Inc.). The calibration was performed using Peptide Calibration Standard II (Bruker Daltonics, Inc.). MALDI-MS/MS peptide fragmentation patterns were compared using commercial standards of surfactin and iturin A (Sigma–Aldrich).

The isolates P5 and C3 showed cellular morphology typical of spore-forming Gram-positive bacteria and were positive for catalase test and motility, and negative to rhizoid growth. The ability of strain C3 for parasporal crystal protein formation was confirmed by phase contrast microscopy examination (Supplementary Figure S1). The API 50CHB and 20E tests showed 99.1% ID and 0.56 T for strain C3 with Bacillus cereus/thuringiensis. For strain P5 API test showed 99.8% ID and 0.59 T for Bacillus subtilis/amyloliquefaciens.

Similarity index values of 0.675 and 0.563 were calculated for Bacillus thuringiensis-GC subgroup A and B. cereus GC subgroup A, respectively, based on MIDI-FAME profile of Bacillus sp. C3. According to the criterion established by FAME analysis, when SI is larger than 0.5 and separated from other organisms from the library by at least 0.100, the isolate is considered identified, in this case as B. thuringiensis. For Bacillus sp. P5 FAME analyses revealed a SI of 0.452 for B. subtilis GC subgroup A and 0.407 for B. subtilis subsp. spizizenii. When the test was repeated, the SI values were 0.463 for B. subtilis subsp. spizizenii and 0.458 to B. subtilis GC subgroup A. In this case, the SI values were lower than 0.5 and this isolate could not be reliably identified at the species level by MIDI-FAME analysis.

The phylogenetic reconstruction of the 16S rDNA sequences is shown in Figure 1. The strain Bacillus sp. P5 was recovered in a node with 92% of support with B. amyloliquefaciens, while the strain Bacillus sp. C3 was clustered together with the members of B. cereus group, with a similarity of 100%, showing a major identity (76%) with B. thuringiensis.

The crude supernatants of Bacillus sp. C3 and Bacillus sp. P5 were tested for antimicrobial activity against Gram-positive and Gram-negative bacteria and fungi. The inhibitory activity of Bacillus sp. P5 was observed against most of the bacteria, including major pathogens and food spoilage organisms such as Bacillus cereus, Listeria monocytogenes, Staphylococcus haemolyticus, and Staphylococcus saprophyticus, as well as against filamentous fungi and the yeast Candida tropicalis. On contrast, Bacillus sp. C3 showed a narrow spectrum of antimicrobial activity (Table 1).

The antifungal activity of Bacillus sp. P5 was notorious, and clear inhibitory halos were observed against filamentous fungi growing onto PDA agar plates (Figure 2). Bacillus sp. P5 showed a broad spectrum of antifungal activity, inhibiting the growth of all the fungi when tested as whole culture and three of five strains when it was tested as culture supernatant (data not shown).

PCR assays followed by sequencing were conducted to investigate the occurrence of essential genes for production of antimicrobial peptides. The genes of iturin A (ituD), surfactin transcriptional terminator (sfp), and subtilosin A (sboA), all with 99% of coverage and identity, were detected for Bacillus sp. P5 (Table 2). For Bacillus sp. C3, the identification of the genes encoding subtilosin A (sboA) and subtilin (spaS) was possible with 99 and 100% identity, respectively (Table 2). The sequences of PCR products obtained for sboA, spaS, ituD and sfp are provided in the Supplementary Table S1.

The ability of strains to produce biosurfactant was first checked on blood agar plates. Bacillus sp. C3 and Bacillus sp. P5 growth resulted in clear rings of erythrocyte lysis around each colony (data not shown), which indicates the production of highly surface active compounds for both strains.

The emulsification index (E₂₄) was determined for strains C3 and P5. The results showed that Bacillus sp. C3 emulsified only mineral oil, presenting an E₂₄ of approximately 40%. Bacillus sp. P5 produced biosurfactants with emulsifying index of 24–57% for soybean biodiesel and toluene, respectively (Table 3). Emulsification of xylene was not observed.

The surface tension was measured in the absence of microbial cells in a surface tension meter using the digital Wilhelmy plate method. The surface tension of the culture medium decreased from 48.4 ± 2.4 mN m^-1 (control medium) to 29.2 ± 0.2 mN m^-1 (supernatant after growth of Bacillus sp. P5), suggesting the production of surfactants by the bacteria. For Bacillus sp. C3, the reduction of surface tension of the culture medium was not significant, with a measurement of 46.2 ± 2.6 mN m^-1.

Based on these results, Bacillus sp. P5 was selected for additional characterization. The production of antimicrobials and biosurfactants was monitored during growth of Bacillus sp. P5. The microorganism reached the stationary growth phase after 12 h incubation and this condition was maintained until 48 h (Figure 3). The initial pH of the medium was 7.2, and during the growth of Bacillus sp. P5, an increase in pH was observed to reach 8.7 at the end of the culture (data not shown).

The antifungal and antibacterial activities were monitored during bacterial growth. The antibacterial activity was maximal at 6 h, coinciding with the exponential growth phase, then decreasing and showing another peak at 30 h during the stationary phase (Figure 3A). The antifungal activity was only observed after 24 h, with maximum values at 36 h (Figure 3B). The emulsifying activity was produced as the cells grew, but the maximum value was reached at 36 h coinciding with the late stationary phase and the maximum antifungal activity (Figure 3B).

Fractions showing antimicrobial activity were subjected to MALDI-TOF mass spectrometry. Molecular masses for bacteriocins and lipopeptides were searched in the m/z range 1000–4000. Molecular species of lipopeptides isolated from the culture medium were found in the m/z range 900–1600. The typical peak for subtilosin (m/z 3400.57) was not observed. A selected MS spectrum of an active fraction eluted from RP-HPLC of the butanol extract of culture supernatant is shown in Figure 4. A peak series in the m/z range 1030–1110 was observed, corresponding to the [M+H]⁺, [M+Na]⁺, [M+K]⁺ adducts for SrfA C₁₅ (1036.8, 1058.8, 1074.0) and SrfA C₁₆ (1050.8, 1072.8, 1088.8). Figure 4 also shows three additional peaks (m/z 1064.8, 1086.8 and 1102.8) that differ from a series of isoforms of SrfA C₁₆ by 14 Da, a putative methylene group, suggesting a series of homolog molecules possibly related to SrfA C₁₇. Comparisons of MS/MS spectra of an active fraction and commercial standards of surfactin or iturin confirm their identification. MS/MS spectra comparison of the peak m/z 1030 (SrfA C₁₃[M + Na]⁺or Srf B C₁₄ [M + Na]⁺) and peak m/z 1079 (ItrC₁₅ [M + Na]⁺) to standards showed the same fragmentation spectra, confirming the molecular identifications (Figure 5). The most intense fragmentation peak of m/z 1030 species was m/z 684.7.

Analyzing the lipopeptide profiles it was observed that surfactin appears in all times tested (6, 30, and 36 h), while fengycin and iturin were detected at 6 and 36 h, respectively (Table 4). In addition, the lipopeptide surfactin presented more abundant isoforms during cultivation, while fengycin isoforms were detected with higher abundance at 6 h. Comparing these results with the growth curves and production of antimicrobial activity, the highest antifungal activity coincided with the time in which greater abundance of iturin lipopeptides was detected in MALDI-TOF spectrometry.

An aqueous suspension of puba was extracted with butanol and concentrated in ZipTip^® C4, and molecular mass was determined by MALDI-TOF. Data were acquired in the m/z range 1000–4000 for the butanol extract. Two clusters were found, corresponding to iturin and fengycin. Expansion of the spectrogram of fengycin indicated protonated ion and sodium and potassium adducts, and MS/MS spectrum of precursor ion of m/z 1478.5 (fengycin C₁₇) generated product ions of m/z 1079.9 and 965.8, which are compatible with fengycin A with Ala at position 6.