Bacillus cereus Toxin Repertoire: Diversity of (Iso)cereulide(s)

The emetic Bacillus cereus toxin cereulide (1) poses a significant safety risk in the food industry, causing emesis and nausea after consumption of contaminated foods. Analogously to cereulide, the structures of various isocereulides, namely, isocereulides A–G, have been recently reported and could also be identified in B. cereus-contaminated food samples. The HPLC fractionation of B. cereus extracts allows us to isolate additional isocereulides. By applying MSn sequencing, post-hydrolytic dipeptide, amino acid and α-hydroxy acid analyses using UPLC-ESI-TOF-MS to purify the analytes, seven new isocereulides H–N (2–8) could be elucidated in their chemical structures. The structure elucidation was supported by one-dimensional and two-dimensional NMR spectra of the isocereulides H (2), K (5), L and N (6 + 8) and M (7). The toxicity of 2–8 was investigated in a HEp-2 cell assay to determine their respective 50% effective concentration (EC50). Thus, 2–8 exhibited EC50 values ranging from a 0.4- to 1.4-fold value compared to cereulide (1). Missing structure-activity correlations indicate the necessity to determine the toxic potential of all naturally present isocereulides as single compounds to be able to perform a thorough toxicity evaluation of B. cereus-contaminated foods in the future.


Introduction
The ubiquitous, endospore-forming, facultative anaerobe bacterium Bacillus cereus is commonly categorized as a food-borne pathogen, causing gastrointestinal diseases and emesis arising from the production of various enterotoxins and the emetic toxin cereulide (1) [1]. Cereulide (1) is composed of 12 alternatingly placed α-amino and α-hydroxy acids thought to form the three-time circularly repeating tetradepsipeptide unit D-O-Leu-D-Ala-L-O-Val-L-Val (Figure 1), leading to its characteristic dodecadepsipeptide structure in a rectangular cylindrical shape [2][3][4][5]. Because of its unique chemical structure, 1 was found to be inert to a wide range of temperatures, pH values and enzymes, making it a considerable safety risk to the food industry [6][7][8][9].
In recent years, screenings of B. cereus extracts have revealed a wide range of cereulide structure homologs, of which the seven isocereulides A-G could be identified and elucidated in their structure [10][11][12]. In addition, the chemical structure of isocereulide A has been recently revised [13]. Moreover, a cereulide isoform differing by +14 Da from the mass-to-charge ratio of 1, referred to as homocereulide, was synthesized as cyclo ( 2 ), even though no reference to the already known isocereulides was made [14]. The complex structure of naturally occurring 1 is biosynthetically assembled through nonribosomal peptide synthetases (NRPS), with its CesNRPS genes, cesA and cesB, located on a pXO1-related virulence megaplasmid and organized as an operon [15][16][17][18]. Other than the aforementioned cereulide synthetase genes cesA and cesB, the ces gene locus involves a phosphopanthetein transferase for activating NRPS machinery (cesP), a type II thioesterase (cesT) which has a proofreading function, and an ABC transporter (cesC/D), which is involved not only in cereulide export but also directly in cereulide biosynthesis [19][20][21]. By coupling in situ reduced D-Leu moieties with D-Ala, cesA generates a peptidyl carrier protein (PCP)-coupled D-O-Leu-D-Ala didepsipeptide intermediate, analogous to cesB assembling the corresponding L-O-Val-L-Val intermediate [22,23]. Subsequently, cereulide (1) biosynthesis was described to occur via a generated L-O-Val-L-Val-D-O-Leu-D-Ala-PCPcoupled intermediate, whereas the assembly of isocereulides F and G was proposed as occurring via an unexpected addition of a TE-bound didepsipeptide [24]. Recent model studies on the TE-dependent trimerization and macrocyclization of the tetradepsipeptide unit L-O-Val-L-Val-D-O-Leu-D-Ala by cesT found cereulide (1) synthesis in line with the biosynthesis pathway of the structurally similar toxin valinomycin [5,25].
Similar to the antibiotic valinomycin, cereulide unfolds its toxicity potential due to a high affinity for complex formation with alkali metal ions (Li + , Na + and K + ) and ammonium ions, resulting in its biological function as a strong K + ionophore and as a K + transporter, resulting in cell membrane depolarization [26][27][28]. Using a porcine model, it was recently shown that 1 can cross the blood-brain barrier, possibly causing a disturbance of the potassium content of the cerebrospinal fluid, which might partially explain the cerebral effects reported from human intoxication cases [29]. In lower concentrations, 1 was found to provoke vomiting in an animal model via stimulation of the vagus afferent by binding to 5-HT 3 receptors located in the duodenum [30]. Additional effects observed for 1 are cellular damage and inhibition of natural killer cells of the immune system [31,32]. Cytotoxicity screening of pre-fractionated B. cereus cell extracts was performed using HEp-2 cells in parallel to structure determination of isocereulides A-G, showing increasing toxicity with growing analyte hydrophobicity [12]. Furthermore, synthetically produced homocereulide showed an approx. 2.8-fold increased effect in vacuolation assays [14].
Cereulide (1) and isocereulides formed in foods by food-borne pathogen B. cereus show a significant safety risk to the food industry [33], resulting in food wastage, thus economic and environmental damage [8]. The present study provides a more comprehensive insight into cereulide chemodiversity by unraveling the complex composition of the broad variation in isocereulide heterogeneity. Conducting UPLC-TOF-MS experiments, ion-trap MS n sequencing, post-hydrolytic dipeptide, enantioselective amino acid and α-hydroxy acid analyses, and 1D-and 2D-NMR experiments, the structures of seven additional isocereulides could be assigned after isolation from B. cereus cell extracts. Through a cytotoxicity screening on HEp-2 cells, the bioactivity of 1 and the newly identified isocereulides H-N (2-8) was investigated to estimate their contribution to the overall toxicity of B. cereus-contaminated food samples.

Structure elucidation of isocereulides H-N (2-8).
Following the recently reported structures for isocereulides A-G [12], the masses of [M + Na] + pseudomolecular ions of the newly detected isocereulides H-N (2-8) were found to vary by 14-28 Da from the [M + Na] + adduct of cereulide (1; m/z 1175.6679) ( Table 1) Figure S4.1c). This observation indicated the exchange of one Ala for one Gly unit but also gave room to the possibility of one O-Leu being exchanged by the isobaric O-Ile.
The eligible reference dipeptides for all isocereulide samples were synthesized through solid-phase peptide synthesis (Supplementary Materials, Section B) and compared to the dissenting dipeptide present in the respective isocereulide by applying co-chromatography. Therefore, the alkaline hydrolysate of 2 was spiked with L-O-Leu-Gly ( Figure 2c) and L-O-Ile-Gly (Figure 2d), respectively. Only L-O-Leu-Gly matched the retention time and increased the signal intensity of the candidate dipeptide, whereas L-O-Ile-Gly revealed a new analyte signal. These findings confirm the inclusion of the leucine-containing dipeptide in isocereulide H (2). The collected data on the released dipeptides of isocereulides H-N (2-8) are summarized in Table 2. Table 2. UPLC-ESI --TOF-MS data of dipeptides released after alkaline hydrolysis from cyclic cereulide (1) and isocereulides H-N (2-8).    This way, the structures of isocereulides I-N (3-8) were elucidated (Figure 3a,b). According to the literature [12], a change in the stereochemistry of the replaced amino acid or α-hydroxy acid was not observed for any isocereulide. The chemodiversity within the structures of 2 and 4 was well in accordance with that in the literature [12], where, for unidentified isocereulides, an exchange of one Ala for one Gly and one O-Val/Val for one O-Ala/Ala, respectively, was predicted. Therefore, isocereulide J (4) was identified as . Interestingly, the exchange of one amino acid by one non-proteinogenic amino acid has not been reported in a naturally occurring isocereulide so far. The structure of isocereulide K (5) was identified as , matching the originally reported chemical structure of isocereulide A [12], which has been recently updated [13]. Isocereulides L and N (6 + 8) were isolated as a mixture (approx. 42/58) and their structures were elucidated in parallel and compared with the isobaric isocereulide C [12].  (Table 1). Due to the cyclic arrangement of the described isocereulides and their close molecular weight (∆~1%), a high similarity in their threedimensional structure is assumed, leading to nearly identical CCS values. Keeping in mind a general deviation in the CCS of approx. 2% [34], a differentiation of 1 and 2-8, especially in their complex natural constitution, could not be observed on the basis of only their CCS.
Determination of EC 50 of cereulide (1) and isocereulides H-N (2-8) via HEp-2 cell assay. The testing principle of the HEp-2 cell assay is based on the uncoupling of ATP synthesis, effectuated by supplemented ionophores, resulting in a facilitated ion influx into the mitochondria, thus disrupting the membrane potential. As a result, the 50% effective concentration (EC 50 ) gives the amount of toxin required to inactivate half of the viable cells. The obtained EC 50 for cereulide (1) and the structurally similar reference toxin valinomycin exhibited a discrepancy in respect to each other, displaying that 1 spawned a 13.7-times higher toxic effect than valinomycin at an equally applied concentration (data not shown). This result agrees well with literature findings, where 1 is described to hold an approx. 15-fold increased toxicity compared to valinomycin [35]. The results of an HEp-2 cell screening of pre-fractionated B. cereus cell extracts showed an increase in the effect on the cells with the growth of analyte hydrophobicity, resulting in a coherence of the toxin's effect on the cells with the grade of its ionospheric properties [12].
While the naturally most abundant cereulide (1) showed an absolute EC 50 of 2.44 ng/mL, the isocereulides H-N (2-8) exhibited significantly deviating concentrations between 1.75 (I, 3) and 6.62 ng/mL (H, 2; exact sample composition and EC 50 values in Supplementary Materials, Table S2), indicating a toxicity range from 0.4-to 1.4-fold of 1 s toxicity ( Figure 4). Especially, 2 stood out with a 63% decreased toxicity effect on HEp-2 cells. Compound 3, despite its slightly more hdrophilic properties, exhibited a 1.4-fold higher toxic effect than 1 and constituted an exception to that hypothesis. Similarly, the EC 50 for 5, which is more hydrophobic than 1, deviated from the hypothesis-regarding the analyte hydrophobicitywith a cytotoxicity of 69%, compared to 1. The ethanolic solutions (1%) of the toxins were tested on 1 × 10 5 HEp-2 cells. Statistical difference between samples calculated based on a two-sided t-test with * being significant at α = 0.05 and ** being highly significant at α = 0.01.

Discussion
The production of variants of non-ribosomally produced peptide toxins, known from a variety of cyanobacteria, is generally based on the molecular diversity of the underlying structural genes [36]. However, isocereulides A-G are reported to be produced simultaneously by one single non-ribosomal synthetase, Ces-NRPS (encoded on a mega plasmid), predicted by the full sequencing of the genome of the B. cereus reference strain F4810/72 and excluding the presence of paralogous ces genes [12,16]. UPLC-TOF-MS studies on the bio-synthetic production of 1 via structure elucidation of depsipeptide intermediates proposed that the general cereulide formation occurs in the form of tetradepsipeptides. D-O-Leu-D-Ala and L-O-Val-L-Val are separately pre-formed by cesA and cesB, which are aligned to form a tetradepsipeptide intermediate, followed by the subsequent addition of further didepsipeptide units, with the possibility of "false" chain elongation, thus shedding light on the structure formation of isocereulides E-G [24]. Literature findings show a selectivity of the specialized A domains in CesA1 and CesB1 toward α-keto groups over α-amino or α-hydroxy groups [22]. Furthermore, CesA1 exhibits a high side-chain selectivity and CesB1 shows a loosened side-chain selectivity, which is possibly advantageous to the formation of cereulide homologs [23]. Recent model studies on the thioesterase-dependent macrocyclization of tetradepsipeptide subunits [5] explain the synthesis pathway of isocereulides A-D generated by a single misincorporation of an α-hydroxy acid or an α-amino acid by the subunits A1 and A2 of CesA or CesB. Probably, these tetradepsipeptide-based observations also build the backbone for the biosynthesis pathway The occurrence of isoforms of the emetic toxin cereulide (1) has been reported in various B. cereus strains over the last years. A wide range of structurally similar cereulide variants has been mass-spectrometrically detected [10][11][12][13]. The isocereulides A-G are predicted to be generated simultaneously by Ces-NRPS [12,16], while, in general, the subunit CesB1 is reported to exhibit a loosened side-chain selectivity [23]. Together with the knowledge that, for most isocereulides, only a single misincorporation of an α-hydroxy acid or an α-amino acid leads to isocereulide formation [5], these circumstances might not only foster the formation of the isocereulides A-G, but of all cereulide homologs detected so far. The very close structural relation between cereulide and the isocereulides and their biosynthesis by the same synthetase gene cluster might build the base for a wide-spread presence of isocereulides in B. cereus contaminated samples. Through their varying m/z, their similar polarity and-compared to cereulide-lower concentrations, the isocereulides might well be not detected during routine analyses due to lacking chromatographic separation or mass spectrometric detection.
The comparison of the cytotoxicity of the purified isocereulides H-N (2-8) found deviating EC 50 values ranging from 0.4-to 1.4-fold of the concentration observed for 1. These findings demonstrate that the cytotoxicity of individual isocereulides cannot be deduced from a sole structure nor from other physicochemical characteristics, such as hydrophobicity. At this point, we can only speculate about the magnitude of the impact that isocereulides have on the overall toxicity of (iso)cereulide-containing B. cereus cell cultures. However, keeping in mind that isocereulide A was shown to exhibit an approximately 8-fold and homocereulide a 2.8-fold higher toxicity than cereulide [12,14], it is essential to individually determine the toxic potential of every single isocereulide.
In summary, our study highlights the importance of the isolation and characterization of the structure and bioactivity of cereulide homologs. Furthermore, it shows the importance of including not only cereulide, but also all naturally occurring isocereulides in routine diagnostics to achieve a realistic toxicity evaluation of emetic B. cereus in contaminated foods. Bacterial cultures and growth conditions: B. cereus strains F4810/72 and F4810/72/ SCV/AN were used for isolating 1-8. The bacterial cultures were prepared as previously described [37,38]. In brief, pre-cultures were prepared from 3 mL of lysogeny broth, kinetically inoculated with 10 3 cfu/mL, incubated (24 h, 30 • C, 120 rpm) and harvested by centrifugation (2 min, 8000 rpm). The supernatant was discarded and the remaining cell pellets were autoclaved (15 min, 121 • C). The pellets were stored at −20 • C until further use.

Experimental Section
Solvent extraction of B. cereus culture pellets. Solvent extracts of the cell cultures were prepared as reported recently [13]. The pellets of strains F4810/72 and F4810/72/SCV/ AN were thawed, extracted with EtOH by shaking (3 × 30 mL, 1 h, RT, 400 rpm) and centrifuged (10 min, 4000 rpm). The supernatants were membrane-filtrated (0.2 µm; PTFE; Phenomenex, Aschaffenburg, Germany) and all liquids combined; then, the solvent was reduced using a rotary evaporator and stored at −20 • C until further use.

Purification of cereulide (1) and isocereulides H-N (2-8):
The purification of 1-8 from the reduced ethanol extract was performed according to the literature [12,13]. In brief, the extract was diluted with H 2 O (1:10) for better retention of the target analytes on the column material and pre-fractionated via C18 SPE cartridges (60 mL, 10 g; Chromabond, Macherey-Nagel, Düren, Germany). All obtained methanolic fractions were combined and the solvent was reduced to approx. 150 mL using a rotary evaporator and stored at −20 • C until further use for compound isolation. The sample material was separated via semipreparative HPLC into 10 fractions (Supplementary Materials, Figure S1), for which the respective eluates were combined and their solvents removed through a rotary evaporator. The obtained HPLC fractions were screened for cereulide (1) and its isocereulides using a UPLC-TOF-MS system Synapt G2-S (Waters, Manchester, UK) in the positive electrospray mode equipped with a 2.1 × 150 mm, 1.7 µm BEH-C18 column (Waters, Manchester, UK) at a flow rate of 0.4 mL/min at 45 • C using aqueous HCOOH (0.1%) as solvent A and MeCN with HCOOH (0.1%) as solvent B. Elution was performed by starting at 93% B, increasing to 100% B in 4 min, holding at 100% B for 2 min, followed by decreasing to 93% in 0.1 min and holding at 93% B for 0.9 min.
Acidic hydrolysis and analysis of amino acid and α-hydroxy acid entities: Aliquots (~200 µL) of the alkaline hydrolysates of 1-8, respectively, were further used for acidic hydrolysis (3 mL of 6 M HCl, 24 h, 110 • C), adjusted to pH 7.0 with NaOH (1 M) and freeze-dried [12]. The lyophilisates were separated into the amino acid-containing and α-hydroxy acid-containing phases, as has been reported recently [13]. Both aliquots of the acidic hydrolysate of 1-8, respectively, were derivatized enantioselectively to determine the stereochemistry of the single amino acid and α-hydroxy acid units contained in the chemical structures. The derivatization of the amino acids in the sample hydrolysates (1)(2)(3)(4)(5)(6)(7)(8) and the corresponding reference amino acids Gly, D/L-Ser, D/L-Ala, D/L-2-Abu and D/L-Val (50 µmol, each) was performed using OPA and IBLC, followed by SPE purification [12,13,39]. The methanolic eluate from the SPE purification was analyzed through UPLC-TOF-MS using the mass spectrometric parameters described for alkaline hydrolysis. For 2-5, chromatography was performed according to the literature [13] at 45 • C and a flow rate of 0.4 mL/min on a 2.1 × 150 mm, 1.7 µm BEH-C18 column (Waters) with aqueous HCOOH (0.1%; solvent A) and a mixture of MeCN/HCOOH (99.9/0.1; v/v; solvent B) while starting the gradient at 30% B for 3 min, increasing to 33% B in 3 min, 50% B in 6 min and 100% B in 1 min, followed by holding 100% B for 1 min, decreasing to 30% B within 0.5 min and ending with an equilibration at 30% B for 1.5 min. For 6-8, the analysis was operated by starting the gradient at 10% B for 3 min, increasing to 25% B in 3 min, 40% B in 6 min and 100% B in 2 min, followed by holding 100% B for 1 min, decreasing to 10% B within 0.1 min and ending with an equilibration step at 10% B for 0.  100; MeCN), the solution was analyzed via UPLC-TOF-MS using MS parameters identical to those described for analyzing dipeptides with the following solvent gradient that started at 30% B, held the initial conditions for 1 min, then increased to 50% B in 10 min, 53% B within 5 min and 100% B in 1 min, held at 100% B for 1 min, decreased in 0.5 min to 30% B and held isocratically for 1. HEp-2 cell assay: The cytotoxic activity of 1-8 was determined by quantifying their EC 50 using HEp-2 cells, as described previously [41,42]. In a 96-well microtiter plate assay, a 2% ethanolic minimum essential medium (MEM-Earle's), containing 2% FCS, 1% sodium pyruvate and 0.4% penicillin-streptomycin (v/v each), was provided to perform stepwise dilution of each toxin solution. As an overall toxicity reference and cell control, the structurally related toxin valinomycin was used analogously, in a range of approx. 2-500 ng/mL. The composition of each toxin sample is listed in the Supplementary Materials, Table S2. The HEp-2 cells were freed from their medium and washed with 5 mL of PBS twice and 5 mL of 0.05% Trypsin/EDTA solution was added. After incubation (8 min, 37 • C, 5% CO 2 ), 10 mL of MEM was added and the cells were resuspended, transferred to a falcon tube and centrifuged (RT, 7 min, 700 rpm). The supernatant was discarded, the cell pellet was suspended in 5 mL of MEM and the cell count was determined via Trypan blue. Finally, the cell count was set to 6.7 × 10 5 cells/mL MEM (1 × 10 5 cells/well); a total of 150 µL of this solution was spiked to each well and the plate was incubated (48 h, 37 • C, 5% CO 2 ). After removing 100 µL of the medium, 10 µL of the cell proliferation reagent WST (cell counting kit) was added and shaken for 1 min (400 rpm) and the plate was incubated (20 min, 37 • C, 5% CO 2 ). To determine cell viability, the extinction at 450 nm/620 nm (TECAN infinite F200) was measured. The respective EC 50 values for each sample were then calculated from the obtained dose-response curves and the corresponding applied sample dilution and concentration.
Mass spectrometry: High-resolution mass spectrometry (UPLC-ESI-TOF-MS) was performed on a Waters Synapt G2-S HDMS spectrometer combined with an Acquity UPLC core system (Waters, Milford, MA, USA) and MS n measurements were conducted on a Bruker Daltonics HCTultra PTM Discovery System (Bruker Daltonics Billerica, MA, USA) according to literature protocol [12].
An aliquot of 2 µL of the respective samples with 13 C 6 -cereulide as an internal standard (100 ng/mL) was applied for the quantitation of 1-8 in the samples, later applied in the HEp-2 cell assay according to the literature [33], performed on a Waters Xevo TQ-S mass spectrometer (Waters) coupled to an Acquity UPLC i-class core system (Waters) with a binary solvent manager, sample manager and column oven. The system was operated with MassLynx 4.1 SCN 813 Software (Waters) and data processing and analysis were performed with TargetLynx (Waters). The recently reported quantitation method for cereulide and isocereulides A-G [33] was extended by implementing the ammonium adducts of 2-8 using the multiple reaction monitoring mode with the following parameters. The calibration for the quantitation of the analytes consisted of mixtures of cereulide (0.1-1000 ng/mL in EtOH) and 13 C 6 -cereulide (100 ng/mL) as an internal standard. Mixtures were prepared from stock solutions and analyzed in triplicate by UPLC-MS/MS. The calibration curve was obtained by plotting the peak area ratios of analytes to internal standard against the concentration ratios of analyte to internal standard for each solution and applying linear regression (origin excluded) to give the equation y = 2.5174x + 0.0061 with R 2 = 0.9998. For 1 and 13 C 6 -cereulide, the quantifier mass transitions were selected as m/z 1170.7 → m/z 357.2 and m/z 1176.7 → m/z 358.2, respectively. All mass spectrometric and chromatographic parameters were used as reported [33].
The CCS values were determined using UPLC-ESI-TWIMS-TOF MS on a Waters Vion HDMS mass spectrometer (Waters) coupled to an Acquity i-class UPLC system (Waters) according to reference [43], whereas we used chromatographic parameters equal to the quantitation of 1-8. The scan time for the HDMS e method was set to 0.2 s and the analyses were performed in positive ESI sensitivity mode by applying the following ion source parameters: capillary voltage of 1.5 kV, source temperature of 150 • C, desolvation temperature of 550 • C, cone gas flow of 50 L/h, desolvation gas of 1100 L/h and the collision energy ramp for HDMS e was set from 20 to 60 eV. The calibration of the mass spectrometer was performed in the m/z range from 20 to 2000 using MajorMix™ (Waters). All data were lock-mass corrected on the pentapeptide leucine enkephalin (Tyr-Gly-Gly-Phe-Leu, m/z 556.2771; [M + H] + ) supplied as a solution (50 pg/µL) of MeCN/0.1% HCOOH (1/1, v/v). The scan time for the lock mass was set to 2.0 s with an interval of 0.5 min. The operation of the UPLC and Vion systems, as well as data processing, was performed with UNIFI™ Software version 3.1.2 (Waters).
Nuclear magnetic resonance spectroscopy: NMR spectra were recorded on a 400 MHz Avance III spectrometer with a Broadband Observe BBFO plus, a 500 MHz Avance NEO spectrometer with a cryo probe CTCI ( 1 H/ 13 C/ 15 N) and a 600 MHz Avance NEO spectrometer with a cryo probe TCI 600S3 H-C/N-D-05 Z XT (Bruker, Rheinstetten, Germany). The chemical shift was referenced to the solvent signal, MeOH-d 4 and pyridine-d 5 , respectively. Data processing and evaluation were performed using Topspin Software Version 4.0.7 (Bruker, Rheinstetten, Germany).