Lichenysin is produced by most Bacillus licheniformis strains

Authors

  • E.H. Madslien,

    1. Forsvarets Forskningsinstitutt FFI, Norwegian Defence Research Establishment, Kjeller, Norway
    2. Department of Food Safety and Infection Biology, Section for Food Safety, Norwegian School of Veterinary Science, Oslo, Norway
    Search for more papers by this author
  • H.T. Rønning,

    1. Department of Food Safety and Infection Biology, Section for Food Safety, Norwegian School of Veterinary Science, Oslo, Norway
    Search for more papers by this author
  • T. Lindbäck,

    1. Department of Food Safety and Infection Biology, Section for Food Safety, Norwegian School of Veterinary Science, Oslo, Norway
    Search for more papers by this author
  • B. Hassel,

    1. Forsvarets Forskningsinstitutt FFI, Norwegian Defence Research Establishment, Kjeller, Norway
    Search for more papers by this author
  • M.A. Andersson,

    1. Department of Food and Environmental Sciences, University of Helsinki, Helsinki, Finland
    Search for more papers by this author
  • P.E. Granum

    Corresponding author
    1. Department of Food Safety and Infection Biology, Section for Food Safety, Norwegian School of Veterinary Science, Oslo, Norway
    • Forsvarets Forskningsinstitutt FFI, Norwegian Defence Research Establishment, Kjeller, Norway
    Search for more papers by this author

Correspondence

Per Einar Granum, Department of Food Safety and Infection Biology, Section for Food Safety, Norwegian School of Veterinary Science, P. Box 8146 Dep., N-0033 Oslo, Norway.

E-mail: PerEinar.Granum@nvh.no

Abstract

Aims

The aim of this study was to elucidate the prevalence of lichenysin production in Bacillus licheniformis and to see whether this feature was restricted to certain genotypes. Secondly, we wanted to see whether cytotoxicity reflected the measured levels of lichenysin.

Methods and Results

Fifty-three genotyped strains of B. licheniformis, representing a wide variety of sources, were included. lchAA gene fragments were detected in all strains by polymerase chain reaction (PCR). All 53 strains produced lichenysins with four molecular masses as confirmed by LC-MS/MS (liquid chromatography–tandem mass spectrometry) analysis. The amounts of lichenysin varied more than two orders of magnitude between strains and were irrespective of genotype. Finally, there was a strong association between lichenysin concentrations and toxicity towards boar spermatozoa, erythrocytes and Vero cells.

Conclusions

Lichenysin synthesis was universal among the 53 B. licheniformis strains examined. The quantities varied considerably between strains, but were not specifically associated with genotype. Cytotoxicity was evident at lichenysin concentrations above 10 μg ml−1, which is in accordance with previous studies.

Significance and Impact of Study

This study might be of interest to those working on B. licheniformis for commercial use as well as for authorities who make risk assessments of B. licheniformis when used as a food and feed additive.

Introduction

Bacillus licheniformis is a facultative anaerobic, Gram-positive, rod shaped, endospore-forming bacterium, closely related to Bacillus subtilis (Logan and De Vos 2009). Bacillus spp. are found almost everywhere in the environment where they survive harsh conditions in the form of highly resistant endospores (Nicholson et al. 2000). Great fermentative capacity combined with low toxicity has made B. licheniformis a popular choice for the industrial production of enzymes, antibiotics (Schallmey et al. 2004) and probiotics (Cutting 2011).

Despite being generally regarded as safe, sporadic cases of B. licheniformis-associated systemic infections in humans have been reported (Sugar and McCloskey 1977; Kramer et al. 1982; Logan 1988; Santini et al. 1995). However, B. licheniformis has not been shown to be able to invade the outer barriers of the body, for example, the mucosal lining of the gastrointestinal tract, respiratory tract or the skin without previous lesions. Nevertheless, abortions in cattle and sheep have been reported (Agerholm et al. 1995; Syrjälä et al. 2007), and it has been experimentally demonstrated that B. licheniformis is able to infect the bovine placenta (Agerholm et al. 1999).

Bacillus gastrointestinal disease is most often caused by Bacillus cereus which, depending on the toxins present, is classified into two different types: the diarrhoeal type and the emetic type, reviewed by Stenfors Arnesen et al. 2008;. Foodborne disease caused by Bacillus species outside the B. cereus-group is less common (Logan 2012). Reported cases of B. licheniformis-associated food poisoning have been characterized by a relatively short incubation time (2–14 h) and high infective dose (>10CFU g−1) followed by mild gastrointestinal symptoms lasting for 6–24 h (Kramer and Gilbert 1989; Salkinoja-Salonen et al. 1999).

One case of fatal illness connected to B. licheniformis-contaminated baby milk powder has been described (Salkinoja-Salonen et al. 1999). Cell extracts from this particular strain were found to be toxic towards boar spermatozoa. The only cytotoxic substance detected was lichenysin A (Yakimov et al. 1999; Mikkola et al. 2000), a small cyclic lipopeptide (CLP) structurally very similar to surfactin (Arima et al. 1968) and pumilacidin (Naruse et al. 1990). The ability of surfactin to form ion-specific channels in lipid bilayers was demonstrated in 1991 by Sheppard and colleagues (Sheppard et al. 1991). Also, an unspecific, ‘detergent-like’ membrane-disrupting effect of surfactin has been described (Hoornstra et al. 2003; Shaligram and Singhal 2010). Several in vitro studies have indicated a strong correlation between surfactin and lichenysin and their toxicity towards different mammalian cells (Mikkola et al. 2000; Nieminen et al. 2007; Apetroaie-Constantin et al. 2009). From and colleagues (From et al. 2007) showed that deletion of the sfp gene (encoding the enzyme that converts the surfactin synthetase complex into an active holoform) in Bacillus mojavensis B31 led to termination of both surfactin synthesis as well as cytotoxicity towards erythrocytes, boar spermatozoa and Vero cells. Further, Nieminen et al. 2007 found that purified lichenysin exerted a similar biologic effect towards boar spermatozoa as seen for surfactin, such as loss of motility, damaged plasma membrane and swelling of the acrosome. Despite the high association between surfactin and cytotoxicity in vitro, studies in rats and mice have shown a generally low toxicity in vivo (Park et al. 2006; Hwang et al. 2009). The effect of lichenysin in mammals has, to our knowledge, not been demonstrated.

The frequency of cytotoxic, lichenysin-synthetizing B. licheniformis strains has previously been found to be low. In a Finnish study, only 4% of 239 foodborne and natural isolates of B. licheniformis were toxic towards boar spermatozoa (Salkinoja-Salonen et al. 1999). In a similar study, the cytotoxicity of 333 B. subtilis-group isolates from various sources was examined, and 3% of the isolates were found to be cytotoxic towards boar spermatozoa and Vero cells (From et al. 2005).

Lichenysin and other Bacillus CLPs are synthesized nonribosomally by multimodular peptide synthetases, so called NRPSs (Stachelhaus and Marahiel 2006). Genes encoding NRPSs are organized into large (20–30 kb) gene clusters and have been detected in many Bacillus species (Tapi et al. 2009). Surfactin (SrfA) and lichenysin (LchA) synthetase are responsible for the synthesis of surfactin and lichenysin, respectively. PCR-based methods targeting NRPSs genes and the sfp gene have been developed in order to detect surfactin and lichenysin producers (Turgay and Marahiel 1994; Hsieh et al. 2004; Tapi et al. 2009). Although these assays have revealed a high overall distribution of NRPSs genes, little is known about the distribution within one single species.

The genetic relationship of 53 B. licheniformis strains was recently described using a novel multilocus sequence typing (MLST) scheme (Madslien et al. 2012). In order to clarify the association between lichenysin production and toxicity, we have examined the 53 B. licheniformis strains for a) the presence and distribution of the lichenysin synthetase gene (lchAA), b) the presence of lichenysin in bacterial cell extracts and c) the cytotoxicity of bacterial cell extracts towards boar spermatozoa, erythrocytes and Vero cells.

Materials and methods

Strains

Fifty-three strains of B. licheniformis, representing various sources, were included in this study (Table 1). Their internal phylogenetic relationship was recently characterized by multilocus sequence typing (MLST) (Madslien et al. 2012).

Table 1. Lichenysin levels and cytotoxicity of cell extracts from 53 genotyped strains of Bacillus licheniformis
IsolateSourceMLST*Group*lchAA allelβ haemolysisMethanol extracts from 60 mg bacterial biomass
LichenysinOil displacementHb releaseVero cellBoar sperm
  1. MLST (group): * Genotypes represent results from Madslien et al. 2012.

  2. lchAA: Allel type was determined on the bases of partial sequence of the lchAA gene encoding lichenysin synthetase A.

  3. β-haemolysis: Evaluated after 48 h incubation on sheep blood agar.

  4. Lichenysin levels: Were determined by mass spectrometry from the average of two independently prepared extracts. The values represent the total quantities of lichenysin [sum of mol.mass (Da) 993·3; 1007·3; 1021·3; 1035·4].

  5. The quantities were categorized into three levels (μg ml−1): <1: ≤3; 1: [1–9]; 2: [10, 99]; 3: [100, 999]; 4: ≥1000.

  6. Oil displacement test: The ability of the extracts to spread oil on a watery surface was categorized into three levels: ‘Weak’ (<50% displacement), ‘moderate’ (>50% displacement) and ‘strong’ (complete displacement).

  7. Hb release: Maximum release (%) of haemoglobin (Hb) was calculated relative to 100% lysis (1% Triton-X100) when incubated for 2 h with 3·3% methanol extract. ND: not determined.

  8. Vero cell toxicity: Positive: >20% inhibition of protein synthesis after 2 h incubation with 3·3% methanol extract. ND: not determined.

  9. Boar spermatozoa toxicity: Positive: <20% motile spermatozoa after 72 h incubation with 1% methanol extract.

749Industrial15A5Yes1Weak1NegNeg
S172Unknown1B2No1Weak0NDNeg
LMG17659Horse blood7B1No1Weak8NegNeg
S170Unknown1B2No1Weak0NegNeg
ATCC14580Type strain1B2No1Weak0NegNeg
ATCC9945AIndustrial1B2No1WeakNDNegNeg
ATCC8480Unknown4A3Yes2Weak39NegNeg
LMG7559Food4A3Yes3Weak0NegNeg
LMG7633Chincilla feces4A3Yes3Weak0NegNeg
LMG7558Industrial7B1No3Weak0NegNeg
NCTC6346Industrial7B1Yes3Moderate0NegNeg
NCTC962Food7B1No3Weak1NegNeg
B317Ovine abortion5B2No3Moderate1NegNeg
CCUG26008Food11A6Yes3Weak2NegNeg
ATCC10716Industrial4A3Yes3Weak1NegNeg
NVH1090Industrial7B1No3Moderate1NegNeg
NVH1077Ovine abortion6A5Yes3Weak0NegNeg
NVH1112Bovine abortion17A4Yes3Moderate0NegPos
LMG6934Soil18A6Yes3Moderate0NegPos
MB1Air21A5Yes3Moderate3NegNeg
553/1Food poisoning, fatal2B2No3Moderate1NegNeg
NVH1111Bovine abortion3B2No3Strong34PosPos
CCUG44767Human blood14A6Yes3Strong35PosPos
NVH1032Food8A3Yes3Moderate1PosPos
NVH622Unknown1B2Yes4Strong39PosPos
NVH1109Bovine abortion16A6Yes4Moderate33PosNeg
NVH1078Ovine abortion26B2Yes4Moderate1NegPos
NVH1079Ovine abortion2B2Yes4Strong19PosPos
BAS50Oil reservoir1B2Yes4Strong63PosPos
NVH1113Bovine abortion3B2Yes4Strong66PosPos
F231Food poisoning23B2Yes4Strong0NegPos
NVH800Food27A6Yes4Moderate50PosPos
B316Bovine abortion10B2Yes4Strong60PosPos
B357Water5B2Yes4Strong3PosPos
LMG17661Food3B2Yes4Strong84PosPos
CCUG41412Food poisoning3B2Yes4Strong72PosPos
M3Air3B2ND4Strong86PosPos
M46Air19B2Yes4Strong92PosPos
NVH1023Food3B2Yes4Strong63PosPos
NCIB7224Industrial9B2Yes4Strong71PosPos
CCUG43512AFood13B2Yes4StrongNDNegPos
F2943Food poisoning22B2No4Strong57PosPos
F5520Food poisoning3B2Yes4Strong89PosPos
NVH1110Bovine abortion9B2Yes4Strong90PosPos
M55Air20B2Yes4Moderate71PosPos
CCUG31354Water2B2Yes4Strong97PosPos
Koskio52Bovine mastitis24B2Yes4Strong92PosPos
F287Food poisoning9B2Yes4Strong92PosPos
M23Air2B2Yes4Strong91PosPos
CCUG43486Water12B2Yes4Strong44PosPos
Koskio51Bovine mastitis24B2Yes4Strong85PosPos
NVH1123Bovine abortion2B2Yes4Strong97PosPos
NVH1115Bovine abortion25B2NoNDStrong1NegPos

β-haemolysis

Freeze bacteria cultures were grown at 37°C on agar plates supplemented with 5% bovine blood and Columbia agar plates supplemented with 5% sheep blood (Oxoid Limited, Hampshire, UK).

The presence or absence of a zone of clearing surrounding the colonies (β-haemolysis) was recorded after 24 and 48 h.

DNA extraction, primer design and PCR against the lichenysin synthetase A (lchAA) gene

Strains were stored at −70°C and plated on sheep blood agar (Columbia blood agar, Oxoid Limited) and grown at 30°C for 24 h. Single-colony material was inoculated in 20 ml Luria broth (LB), grown overnight at 30°C and centrifuged at 3000 g for 10 min. The pellet was resuspended in 1 ml lysis buffer (20 mmol l−1 Tris–Cl, pH 8·0, 1·2% Triton X-100, 20 mg ml−1 lysozyme) (Sigma, Steinheim, Germany), and DNA was isolated using the DNeasy Blood and Tissue Kit (QIAGEN Hamburg GmbH, Hamburg, Germany) according to the manufacturer's instruction. DNA concentrations were measured using a Nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). PCR primers targeting the B. licheniformis ATCC14580 lichenysin synthetase A (lchAA) gene sequence NC_006270·3 (378877–389619) were designed using PRIMER 3 (Rozen and Skaletsky 1999). The primers (F: 5′-ACTGAAGCGATTCGCAAGTT-3′, R: 5′–TCGCTTCATATTGTGCGTTC-3′) complementary to the B. licheniformis ATCC14580 genome [NC_006270.3; (Rey et al. 2004)] positions 379037–379018 and 379470–379489 were synthetized by Invitrogen Life Sciences (Life Technologies Ltd., Paisley, UK).

The amplification reactions were performed in a LightCycler® 480 System using LightCycler® 480 SYBR Green I Master (Roche Diagnostics GmbH, Mannheim, Germany) according to recommendations given by the manufacturer of the kit. The amplification programme was as follows: 5 min initial denaturation at 95°C followed by 35 cycles of denaturation at 95°C for 10 s, annealing at 56°C for 10 s and primer extension at 72°C for 30 s. The amplifications were terminated after a final elongation step of 7 min at 72°C. DNA from the type strain ATCC14580 was used as positive control. The PCR fragments were verified by electrophoresis using Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). PCR products were purified and sequenced by Eurofins MWG Operon (Ebersberg, Germany), using the same primer pair as for amplification.

The Staden Package (Staden 1996) was used for alignment, editing and construction of consensus sequences based on the ABI sequence chromatograms. Consensus sequences were entered into the MEGA5 (Tamura et al. 2011) software and aligned by CLUSTALW (Thompson et al. 1994). Dendograms were constructed in MEGA5 using the neighbour-joining (NJ) (Saitou and Nei 1987) algorithm with branch lengths estimated by the maximum composite likelihood method (Tamura et al. 2007). Branch quality was assessed by the bootstrap test using 500 replicates. Ratio of dN/dS (nonsilent vs silent substitutions) was calculated in S.T.A.R.T. 2 (Jolley et al. 2001) using the method by Nei and Gojobori (1986).

Preparation of bacterial methanol extracts

All strains were grown for 10 days at 37°C on trypticase soy agar plates (TSA) (Merck KGaA, Darmstadt, Germany). 60 mg of bacterial biomass, equivalent to 109–1010 CFU, was extracted with 1·0 ml methanol and heated for 30 min at 80°C. The dry pellet was resuspended in 0·5 ml of methanol, vortexed and centrifuged for 3 min at 13 000 g. The supernatant was collected and heated at 80°C until complete evaporation (30 min). The dry residue was dissolved in 200 μl methanol and stored in dark glass vials at 4°C before use. Three independently prepared extracts were made from all strains.

Molecular mass determination and quantification of lichenysin by LC-MS/MS

Chemicals, reagents and sample preparations

All chemicals were of at least HPLC (high-performance liquid chromatography) grade and supplied by VWR (West Chester, PA, USA) except surfactin (Sigma-Aldrich, St Louis, MO, USA) and heptafluorobutyric acid (Fluka, Buchs, Switzerland). The water used was grade 1 water purified with a Milli-Q water purification system from Millipore (Bedford, MA, USA).

Bacterial methanol extracts were diluted 1 : 10 in methanol/water (1/1) before aliquots (1 μl) were injected into the LC-MS/MS apparatus.

Liquid chromatography

The instrumentation used was an Agilent 1200 SL system (Agilent Technologies, Waldbronn, Germany) consisting of a binary pump, thermostatted autosampler and column compartment. The separation was performed at 35°C on an RRHD Zorbax Eclipse Plus C18 column, 100 × 2·1 mm id, with 1·8 μm particles (Agilent Technologies, Santa Clara, CA, USA). Mobile phase A consisted of 2 mmol l−1 ammonium acetate and 0·2% heptafluorobutyric acid in water, whereas mobile phase B was acetonitrile and methanol (1/1). The flow rate was 0·4 ml min−1 with a linear gradient from 90–93% B in 4 min. Total time of analysis was 8 min. The temperature of the autosampler was 4°C.

Mass spectrometry

The instrument used was G6490 triple quadrupole mass spectrometer (Agilent Technologies, Singapore) equipped with a Jet Stream electrospray ion source. The LC-MS/MS was operated in full-scan mode (from 950 to 1150 Da) to determine the molecular masses of surfactin/lichenysin present in the methanol extracts, whereas data were acquired in positive multiple reaction monitoring mode, MRM, for quantification. Total ion chromatogram and extracted ion chromatograms with mass spectra are presented in Fig.  2.

A calibration curve of surfactin was used for external calibration; hence, all concentrations are given as surfactin equivalents. Six concentration levels were diluted in methanol from a stock solution of 1 mg ml−1 surfactin in ethanol.

Quantification of lichenysin was performed on three independently prepared extracts.

Oil displacement

Presence of surfactant in the bacterial methanol extracts was determined semi-quantitatively by the ability to displace oil from a watery surface (Morikawa et al. 2000). Petri plates (diameter 90 mm) were filled with 30 ml of distilled water. 20 μl of sterile filtered (0·45 μm) corn oil (Mills, Norway) was carefully layered on top of the water phase with a micropipette. 10 μl of bacterial methanol extracts was applied in the centre of the oil surface. Immediately after application, the occurrence and size of a clearing zone appearing around the applications site were observed. The results were classified into three categories (‘weak’, ‘moderate’ and ‘strong’) based on the size of the clearing zone that appeared around the application site (Fig. S1). The oil displacement test was performed in duplicate on three independently prepared extracts.

Commercially available B. subtilis surfactin (Sigma) was dissolved in methanol at a concentration of 1 mg ml−1 and diluted 1:2; 1:4; 1:10 and 1:100 in methanol. The oil displacement ability of the different dilutions was tested as above, and the presence and size of the clearance zone were recorded (Fig. S1).

Haemolytic activity of methanol extracts

Haemolytic activity of the methanol extracts was measured according to Fagerlund et al. 2008 with minor modifications. Bovine blood cells were washed three times in PBS. 5 μl crude bacterial methanol extract was incubated with 150 μl 2% (v/v) bovine blood cells for 2 h at 37°C. After incubation, the sample was centrifuged and haemolysis was determined from the A540 of the supernatant. The percentage of haemolysis was calculated by comparing the A540 of the sample with positive [100% lysis by 1% Triton X-100 (Sigma)] and negative (5 μl methanol) controls. All experiments were performed in duplicates.

Boar spermatozoa motility test

The boar sperm motility test was performed according to the protocol of Andersson et al. 1998 with modifications. Briefly, 20 μl of crude methanol extract was mixed with 2 ml of extended boar semen (Norsvin LE, Norway) containing approximately 2 × 107 cells ml−1 and 0·25 mg ml−1 gentamycin to inhibit bacterial growth. The samples were incubated at 18°C for 72 h before readings. Evaluation of motility of preheated (37°C for 5 min) samples was performed in phase-contrast microscope (100 x objective) counting approximately 20 spermatozoa from five different fields (minimum 100 spermatozoa in total). The percentage of immotile spermatozoa was recorded. Loss of motility in >80% of the cells was considered indicative of cytotoxicity. The percentage of motile cells of control samples (20 μl methanol per 2 ml boar semen) was always >80%. The assay was performed in duplicates on two independently prepared extracts.

Commercially available surfactin from B. subtilis (Sigma) was dissolved in methanol (1 mg ml−1) and diluted 1:2; 1:4; 1:10; 1:100 and 1:1000. The effect of surfactin on boar semen motility was tested as above, on two different batches of boar semen and compared against the toxicity of the B. licheniformis extracts.

Vero cell assay

Cytotoxicity was determined using a Vero cell test (Sandvig and Olsnes 1982) that monitors the inhibition of protein synthesis by measuring the reduction of incorporated 14C-leucine in the Vero cells (ECACC-European Collection of Cell Cultures, Public Health England, Salisbury, UK) upon addition of toxins. Briefly, 10 μl of bacterial methanol extract was applied to 300 μl of low-leucine medium (Gibco, Life Technologies Ltd.) covering a confluent monolayer of Vero cells. The cells were incubated for 2 h followed by washing and addition of 300 μl of low-leucine medium with 14C-leucine (0·2 nCi μl−1) (Hartmann Analytic GmbH, Braunschweig, Germany) and incubated further for 1 h. The radioactive medium was removed, and the cells were washed once in low-leucine medium before radioactivity (c.p.m.) in the Vero cells was counted in a scintillation counter (Pacard Liquid Scintillation Analyzer, Perkin Elmer, MA, USA) using Ultima Gold scintillation fluid (Perkin Elmer, MA, USA). The percentage of inhibition of protein synthesis was calculated using the formula [(negative control – sample)/negative control] × 100. Vero cells incubated with 10 μl methanol for 2 h prior to incubation with 14C leucine were used as negative control. Samples were considered toxin positive when >20% of protein synthesis was inhibited. Assays were performed in at least two independent assays with two technical replicates in each assay.

Statistical analysis

Spearman correlation coefficients (ρ) were calculated in GraphPad Prism 6 Software package. P-values were determined by a two-tailed test using a 95% confidence interval.

Results

Screening for the presence of lichenysin synthetase A gene (lchAA)

The presence of the first structural gene (lchAA) of the lichenysin synthetase A operon was detected by PCR in all 53 strains. The region corresponded to amino acid positions 69 to 184 of lichenysin synthetase A (GenBank accession: YP_077640), part of the condensation domain involved in the formation of peptide bonds (Stachelhaus et al. 1998). All sequences have been submitted to GenBank (accession number: KC986877-KC986929). Alignment and cluster analysis of the 53-nucleotide sequences revealed six unique alleles (Fig. 1). The lchAA-based dendogram followed roughly the same pattern as seen for the individual MLST house-keeping gene sequences (Madslien et al. 2012). Base substitutions were spread over the entire sequence and the overall substitution frequency was 9·5%. The ratio of nonsynonymous (changes in amino acid sequence) vs synonymous (no change in amino acid sequence) substitutions (dN/dS) was 0·0780, which is slightly higher than the calculated values for the MLST loci (0·0043–0·0457) (Madslien et al. 2012).

Figure 1.

Cluster analysis of partial lchAA sequence in 53 strains of Bacillus licheniformis. The cluster analysis was constructed using the neighbour-joining method. The optimal tree with the sum of branch length = 0·06 is shown. Branch quality is estimated by using the bootstrap test (500 replicates) and is shown next to the branches. The tree is drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. The evolutionary distances were computed using the maximum composite likelihood method and are in the units of the number of base substitutions per site. The analysis involved 53 nucleotide sequences. Numbers in brackets represent MLST sequence type. There were a total of 348 positions in the final dataset. The analyses were conducted in MEGA5.

Molecular mass determination and quantification of lichenysin by LC-MS/MS

The results of the LC-MS/MS detection and quantification of lichenysin are summarized in Table 1. Four different isoforms of lichenysin with m/z [molecular mass (Da)]: 993·7 (993·3); 1007·7 (1007·3); 1021·7 (1021·3); 1035·7 (1035·4) were detected by LC-MS/MS. The isoforms were separated from each other and interfering matrix components as shown in the chromatogram (Fig. 2). The two peaks representing m/z 1021·7 had a different ratio in lichenysin compared to commercial surfactin. Hence, the double peak was integrated as one. Lichenysin was detected in methanol extracts from all of the 53 strains included, although seven of the strains produced lichenysins at concentrations below the quantification limit (LOQ, calculated as 10× standard deviation of the noise) at 1 μg ml−1 (Table 1). The amount of lichenysin in Table 1 is presented as the sum of the four isoforms. Log10–quantities of lichenysin from three independent extractions are given in Fig. S2.

Figure 2.

Chromatogram from a representative sample (strain LMG17661). Total ion chromatogram and extracted ion chromatograms with mass spectra for each of the four detected isoforms of lichenysin are displayed.

Oil displacement test

The results of the oil displacement test are listed in Table 1. We found a strong correlation (ρ = 0·8; < 0·0001) between the oil displacement ability and the quantity of lichenysin (Table 2). The effect of the different extracts varied from no displacement to complete displacement. Complete displacement (‘strong’) was equivalent to the effect of 1 mg ml−1 commercial surfactin, 50% displacement (‘moderate’) was observed at 0·5 mg ml−1, and <50% displacement (‘weak’) was observed at concentrations of ≤0·25 mg ml−1. The effect was observed immediately after application (Fig. S1).

Table 2. Correlation matrix showing the association between lichenysin concentration, oil displacement and Hb release
 LichenysinOil displacement
  1. The analysis was performed in GraphPad Prism v.6.01 using Spearman rho (ρ). **** 0·0001.

Lichenysin  
Oil displacement0·8**** 
Hb release0·8****0·7****

β-haemolysis and cytotoxicity

Thirty-nine strains of B. licheniformis caused β-haemolysis in sheep blood agar after 48 h of growth, whereas 13 strains did not. Because haemolytic activity on bovine blood agar plates was weak, diffuse and hard to classify compared to what observed on sheep blood agar, only the results from the sheep blood plates are included in Table 1. Generally, β-haemolysis was more evident after 48 h than 24 h. Methanol extracts from 33/53 strains caused cytotoxicity in boar spermatozoa as could be seen from a loss of motility in more than 80% of the sperm cells (Table 1). Motility loss seemed to occur after cell membrane disruption and was accompanied by swelling of the acrosome (as observed by phase-contrast microscopy). The overlap between β-haemolysis and sperm-toxicity was 75%. Extracts from 28/52 strains were toxic to Vero cells, causing loss of 14C -leucine incorporation into proteins. The overlap between sperm-toxic and Vero cell-toxic strains was 87%; only seven strains were toxic to only one of the cell types. 20/51 of the samples generated a haemoglobin (Hb) release of ≥50% (Table 1).

Cytotoxicity was generally observed at lichenysin concentrations above 10 μg ml−1 (20 μg lichenysin in 2 ml cell suspension) for the boar spermatozoa and 33 μg ml−1 (10 μg lichenysin in 300 μl cell suspension) for the Vero cells. Hb release of ≥50% was only observed in samples containing >33 μg ml−1 (5 μg lichenysin in 150 μl cell suspension). A sigmoid relationship between haemolysis (Hb release) and lichenysin concentrations was observed (Fig. 3). Toxicity was absent in samples containing <29 μg ml−1 lichenysin (Vero cell assay) or <1·8 μg ml−1 lichenysin (boar spermatozoa assay).

Figure 3.

Hb release (%) from erythrocytes plotted against the lichenysin concentrations. The analysis was performed in the statistical software package GraphPad Prism v.6.01.

Discussion

Lichenysin synthesis is a common feature of Bacillus licheniformis

The type strain B. licheniformis ATCC14580/DSM13 and a few other B. licheniformis strains have previously been shown to harbour the lchAA gene (Nieminen et al. 2007). However, the presence of lichenysin synthetase genes in a large number of B. licheniformis strains is, to our knowledge, unknown. In this study, lchAA was detected in all 53 B. licheniformis strains examined, indicating that most, if not all, B. licheniformis strains are capable of producing lichenysin.

Four lichenysin isoforms with molecular masses (Da) of 993·3, 1007·3, 1021·3 and 1035·4 were detected by LC-MS/MS in all 53 B. licheniformis strains, showing that lichenysin synthesis is a common feature of this species. The three highest masses, 1007·3, 1021·3 and 1035·4, are identical to those classified as lichenysin A in strains BAS50 and 553/1 (Yakimov et al. 1995; Mikkola et al. 2000). Both of these strains have been included in our work. The isoform with the lowest molecular mass (993·3) was first defined as lichenysin G (Grangemard et al. 1999) with the C-terminal amino acid valine instead of leucine/isoleucine as in lichenysin A. On the other hand, the same isoform may also represent lichenysin A, as Yakimov et al. 1995 concluded that the fatty acid moiety of lichenysin might contain 12 to 17 methylene groups. To summarize, among the 53 strains included in this study, the 14 Da mass shift between the molecular masses detected might arise both from amino acid substitutions in the peptide moiety and different lengths of the fatty acid chain (Yakimov et al. 1995; Grangemard et al. 1999).

Although lichenysin was synthetized in all strains, the amounts varied more than two orders of magnitude between strains, ranging from <0·013 μg to >3·3 μg per mg biomass. The quantity produced by each strain was highly reproducible (Fig. S2) indicating that strain-dependent differences were not a random observation. The differences might arise at the a) nucleotide sequence level, b) transcriptional level, c) translational level or d) enzyme (lichenysin synthetase) activity level. In B. subtilis, several gene regulatory and environmental factors are known to affect the biosynthesis of surfactin, which could lead to 1) different yield from the same strain under different conditions or 2) different yield from different strains under identical conditions (Jacques 2011). Examples of important environmental factors that can affect surfactin synthesis are: pH, nitrogen/iron/manganese-ratio, carbon source and temperature. Less is known about conditions influencing lichenysin biosynthesis, although nutrient sources and temperature are thought to play essential roles (Yakimov et al. 1995; Joshi et al. 2008).

The 28 strains that produced high levels of lichenysin (above 3·3 μg mg−1; level 4) were from a variety of sources (Table 1). No obvious relationship between genotype and concentrations of lichenysin was observed. However, strains belonging to MLST group A generally produced lower quantities than strains of group B. Similar lineage-dependent differences have been reported for the nonribosomally synthesized peptide antibiotic bacitracin in B. licheniformis (Ishihara et al. 2002).

The six unique lchAA allels were unevenly distributed among the strains. Allel type 2 (lchAA2) was the most abundant, carried by 34 of the strains. Most of the level-4 lichenysin-producing strains carried lchAA2. However, this allel was also carried by four of the strains that produced the lowest levels of lichenysin (level 1), including the type strain ATCC14580 (Table 1). Thus, the cluster analysis based on partial NRPSs sequence (lchAA) presented in Fig. 1 cannot be used to generally distinguish between strains producing high levels of lichenysin vs low producers. Interestingly, some of the most potent lichenysin-producing strains (BAS50 and NVH622) were genetically undistinguishable from the low-producing type strain ATCC14580 (Fig. 1). Presumably, the distribution of lchAA sequences in our study reflects the genetic relationship between the strains rather than the capacity of lichenysin synthesis.

Cytotoxicity was observed in 64% of the extracts

Previous studies have indicated an association between cytotoxicity and the ability to produce lichenysin (Mikkola et al. 2000). A relatively low number (0·5–4%) of B. licheniformis strains have previously been found to synthesize lichenysin or other heat-stable toxins at cytotoxic levels (Salkinoja-Salonen et al. 1999; From et al. 2005; Taylor et al. 2005). In this study, methanol extracts from 64% of the strains were found to be cytotoxic. We suggest that this large divergence was mainly due to 1) higher growth temperature prior to extraction or 2) modifications of the extraction protocol (higher ratio of biomass vs solvent, different solvent) leading to a higher concentration of toxins rather than to difference in sensitivity of the cell assays.

Strong association between lichenysin level and cytotoxicity

We found that toxicity of the methanol extracts was highly associated with the amount of lichenysin as determined by LC-MS/MS (Tables 1 and 2). Cytotoxicity towards boar spermatozoa, Vero cells and erythrocytes was observed at lichenysin concentrations above 10 μg ml−1, which is in agreement with previous reports (4–8 μg ml−1) (Mikkola et al. 2000; Hoornstra et al. 2003). Lichenysin is thought to interact with the spermatozoa plasma membrane in a surfactant-like fashion, as opposed to the emetic toxin cereulide that has been shown to act as a K-ionophore on the mitochondrial membrane (Mikkola et al. 1999). Our microscopic observations of membrane damage and swelling of the acrosome correspond to what has previously been described for lichenysin and other surfactants (Hoornstra et al. 2003). However, we did not specifically search for molecules other than lichenysin and surfactin in the extracts. Therefore, we cannot rule out that there were other methanol-soluble heat-stable toxic substances present in the extracts that could, at least partly be responsible for the observed cytotoxicity.

Although we found that more than half of the B. licheniformis strains produced lichenysin at cytotoxic concentrations in vitro, it does not necessarily explain the pathogenicity of B. licheniformis in vivo. This might be illustrated by the observation that extracts from strains associated with clinical disease in humans or animals were not always cytotoxic (Table 1). However, environmental factors are known to affect lichenysin production (Yakimov et al. 1995; Joshi et al. 2008). Therefore, the amounts of lichenysin detected from each strain in our study might differ from the situation in vivo and in foods. Interestingly, the extract from the only strain connected to the death of a human patient, 553/1 was not cytotoxic, although previous studies have found the opposite (Salkinoja-Salonen et al. 1999; Mikkola et al. 2000). Different growth temperature and modification of the extraction procedure might have generated lower toxin amount from 553/1 than previously detected. However, four other food poisoning-associated strains from the Finnish study were also included for comparison (F231, F2943, F5520 and F287). Among these strains, boar sperm cytotoxicity corresponded well to what has previously been found. We therefore speculate that the toxic ability of strain 553/1 might have been lost, possibly as a result of adaption to repeated cultivations in the laboratory (Maughan and Nicholson 2011). It is interesting that these toxins appear to be so widespread among B. licheniformis strains and still there are very few reports of B. licheniformis-associated disease. The reason for this is probably that, even for the most potent lichenysin producers, a very high number of cells are required in order to synthesize sufficient lichenysin to cause toxicity in humans.

The oil displacement test-a reliable tool for rapid detection of lichenysin at cytotoxic concentrations

Haemolysis on blood agar is a widely used method for screening strains for the presence of surfactins. However, the oil-clearance test has been proposed as a more accurate method for determination of surfactin (Youssef et al. 2004). We found a strong correlation between oil displacement and the amount of lichenysin (Table 2). Moderate and strong displacement always occurred in samples containing >1 mg ml−1 lichenysin. Interestingly, while all MLST group A strains were β-haemolytic on blood agar plates, only a few of them produced methanol-soluble, heat-stable substances at toxic levels. We could therefore speculate that, during vegetative growth, group A strains may produce haemolytic substances that are inactivated by the heat treatment. Regarding group B-strains, β-haemolysis on blood agar plates seemed to be a better indicator of lichenysin production. However, several of the nonhaemolytic strains produced considerable amounts of lichenysin. Hence, a rapid screening of lichenysin-producing strains on blood agar will most probably give false negatives. In this context, the oil displacement test seemed to be the most reliable tool for a rapid screening of strains producing lichenysin at concentrations associated with cytotoxicity.

We have shown that the ability to synthesize lichenysin is highly conserved among all 53 B. licheniformis strains used in this study. In vitro cytotoxicity of the methanol extracts was strongly associated with the presence of lichenysin. Further studies are needed to see if the in vitro pathogenicity of B. licheniformis is reflected by the in vivo situation in different tissues.

Acknowledgements

The work was supported by grants from the Norwegian Research Council (grant 178299/I10) and the Norwegian Defence Research Establishment (FFI). The authors wish to thank Kristin O'Sullivan (Norwegian School of Veterinary Science) for technical assistance with the toxicity assays and Avi Ring (FFI) for helpful comments to the design of this study and for kindly sharing his knowledge about bioactive compounds and their interactions with lipid membranes. We also want to thank Tone Normann Asp (Norwegian School of Veterinary Science) for valuable discussions.

Conflict of Interest

The authors declare no conflict of interest.

Ancillary