The genes hblC, hblD and hblA, which encode Hbl L2, L1 and B, respectively, are organised in an operon (Ryan et al., 1997[197]) (Figure 3A(Fig. 3); References in Figure 3: Buchacher et al., 2023[29]; Dietrich et al., 2021[56]; Duport and Armengaud, 2025[57]; Fagerlund et al., 2010[71]; Ghelardi et al., 2007[84]; Mazzantini et al., 2020[162]; Salvetti et al., 2007[200]; Sauciuc et al., 2024[202]; Tsukazaki, 2018[221]; Vörös et al., 2014[225]; Waterhouse et al., 2018[231]). The majority of hbl-positive isolates also bears hblB, encoding a fourth protein component, Hbl B' (Heinrichs et al., 1993[100]). Generally, sequence similarities point to the development of the whole hbl operon from duplication of a single precursor gene (Økstad et al., 1999[181]). Its transcription begins upstream of hblC and hblCDA are co-transcribed (Ryan et al., 1997[197]). Upstream of hblB, a stem loop as possible transcriptional terminator and a possible new promoter structure were identified (Clair et al., 2010[45]; Sastalla et al., 2013[201]). The nheABC operon encodes Nhe A, B and C (Granum et al., 1999[89]) (Figure 3A(Fig. 3)). All three genes are co-transcribed (Lindbäck et al., 2004[141]), but similarly to hblB, stem loops were identified in the intergenic region between nheB and nheC as well as downstream of nheC (Granum et al., 1999[89]). Upstream of nheA, two promoter regions containing two PlcR binding sites exist. These are part of uncommonly long 5' untranslated regions present upstream of the hbl and nhe operons but lacking upstream of cytK (Böhm et al., 2016[24]) (Figure 3B(Fig. 3)). The latter encodes the single protein toxin CytK (Figure 3A(Fig. 3)). Sequence analyses revealed two highly homologue variants, cytK-1 and cytK-2 (Fagerlund et al., 2004[73]; Guinebretiére et al., 2006[92]). Only few strains, mainly isolated from potato products or flour and just recently, from edible insects, produce the toxic CytK-1 variant, which are classified as a separate species, B. cytotoxicus (Cairo et al., 2022[32]; Contzen et al., 2014[48]; Etter et al., 2024[68]; Guinebretiére et al., 2013[91]; Heini et al., 2018[99]; Kone et al., 2019[125]; Mohammadpour et al., 2025[172]). Although cytK-1 is highly conserved within this species, recent whole genome sequence analyses reveal considerable chromosomal and plasmidial intra-species diversity (Fayad et al., 2021[75]). The cytK-2 gene is frequently found in B. cereus isolates; however, it seems not to be relevant to the diarrhoeal disease (Castiaux et al., 2015[38]; Jessberger et al., 2019[115]; Jessberger et al., 2015[116]).

The promoter regions of the hbl and nhe operons as well as cytK-1 and cytK-2 bear binding sites (known as PlcR-boxes with the conserved sequence TATGNANNNNTNCATA) for the pleiotropic transcriptional regulator PlcR (Figure 3B(Fig. 3)), which acts as an activator for its regulon at the transition from exponential to stationary growth phase via a well-investigated mechanism responding to cell density (Agaisse et al., 1999[2]; Gohar et al., 2002[86], 2008[85]; Lereclus et al., 1996[134], 2000[135]). Herein, the signalling peptide PapR plays a particularly important role as part of the quorum-sensing system (Bouillaut et al., 2008[25]; Slamti and Lereclus, 2002[211]). This peptide is exported, and with the help of secreted neutral protease B (NprB) and oligopeptide permease OppABCDF subsequently processed and re-imported as heptapeptide PapR₇. PlcR-PapR complexes can then bind to the PlcR-boxes and activate gene expression (Bouillaut et al., 2008[25]; Gominet et al., 2001[87]; Pomerantsev et al., 2009[190]; Slamti and Lereclus, 2002[211]). hbl, nhe and cytK-1 promoter regions furthermore show binding sites for Fnr and SinR (Böhm et al., 2016[24]) (Figure 3B(Fig. 3)). Fnr (fumarate and nitrate reductase) is a redox regulator activating enterotoxin gene expression in response to carbohydrates and oxidoreduction potential (Messaoudi et al., 2010[167]; Zigha et al., 2007[241]). SinR, which - together with its antagonists - is primarily involved in biofilm formation and motility, was proven to influence hbl transcription (Fagerlund et al., 2014[70]; Milton et al., 2020[171]; Newman et al., 2013[178]). Additionally, the hbl and nhe upstream regions harbour different ResD binding sites (Figure 3B(Fig. 3)), which is the response regulator interacting with sensor kinase ResE in response to oxidoreduction potential (Duport et al., 2006[60]). ResD and Fnr have mutual influence on each other (Esbelin et al., 2009[67]). Palindromic catabolite responsive elements (CRE) have also been identified upstream of hbl and nhe (Böhm et al., 2016[24]; Stenfors Arnesen et al., 2008[214]; van der Voort et al., 2008[222]) (Figure 3B(Fig. 3)). These are binding sites to CcpA, a transcriptional regulator repressing hbl and nhe expression under different conditions including glucose surplus (Duport et al., 2004[59]; Ouhib-Jacobs et al., 2009[183]; Ouhib et al., 2006[182]; van der Voort et al., 2008[222]). Finally, one (hbl) and two (nhe) CodY binding site were also found (Böhm et al., 2016[24]) (Figure 3B(Fig. 3)). Under nutrient surplus, the regulon of this pleiotropic transcriptional regulator is repressed. As nutrients become limited, derepression occurs. Affected genes are involved in catabolism, chemotaxis and motility, as well as virulence (Brinsmade, 2017[28]; Lindbäck et al., 2012[143]). CodY affects the PlcR-PapR quorum sensing system as well as the hbl and nhe operons by direct binding (Böhm et al., 2016[24]; Slamti et al., 2015[210] ). The regulator is phosphorylated at serin 215 by the Hanks kinases PrkC and YbdM, which impedes DNA binding (Kortebi et al., 2025[127]). Böhm et al. (2016[24]) found that, in contrast to PlcR, the CodY binding sites show a much higher heterogeneity within the B. cereus group. This might lead to varying binding intensities of this nutrient-responsive roadblock repressor, and thus, to strain-specific intensities of enterotoxin production (Böhm et al., 2016[24]). Furthermore, Prince and Kovac emphasise in their minireview, that although the upstream regions of the hbl and nhe operons show binding sites for the same transcriptional regulators, they are differentially regulated. This might be due to different modes of action of regulators, once as activators for nhe, once as repressors for hbl, stronger binding of repressors upstream of hbl, stronger binding of activators upstream of nhe, or further unidentified transcriptional regulators (Prince and Kovac, 2022[192]). Further studies suggest an involvement of the OhrRA (organic hydroperoxide resistance) system (Clair et al., 2013[44] and the transcriptional regulator RpoN (sigma 54) (Hayrapetyan et al., 2015[97]) on hbl and nhe expression, as well as a new, yet unknown regulatory mechanism for hblB (Clair et al., 2010[45]; Jessberger et al., 2019[112]). In summary, it has been shown that enterotoxin gene expression depends on cellular growth phase and energy status, temperature, oxygen tension, nutrient availability, and host factors (Dietrich et al., 2021[56]; Jessberger et al., 2019[112]; Jessberger et al., 2017[117]; Prince and Kovac, 2022[192]; Vyklicka et al., 2025[226]). Further, Meng et al. (2025[166]) found that enterotoxin gene expression can be decreased by environmental stress factors such as temperature, pH and salt concentrations.

On the one hand, it is believed that the Hbl and Nhe enterotoxin components as well as CytK are released from the cell via the Sec translocation pathway, as all of them contain N-terminal signal peptides for secretion. SecA (ATPase), SecDF (translocation factor) and SecYEG (channel) assemble this pathway (Beckwith, 2013[14]; Tsukazaki, 2018[221]) (Figure 3C(Fig. 3)). Experimentally, a loss of Hbl B secretion was shown after signal peptide modification, as well as reduced enterotoxin secretion after addition of SecA inhibitor sodium azide (Fagerlund et al., 2010[71]). A secDF knock out mutant showed lower amounts of Hbl, Nhe, CytK and phospholipase C in the secretome as well as reduced virulence and motility compared to the wild type (Vörös et al., 2014[225]). On the other hand, especially Hbl secretion has been linked to the flagellar secretion system (Figure 3C(Fig. 3)). A connection was established between swarming motility and Hbl secretion (Ghelardi et al., 2007[84]), and an flhA (integral transmembrane protein) deletion compromised Hbl secretion as well as hblC transcription (Bouillaut et al., 2005[26]; Ghelardi et al., 2002[83], 2007[84]). Impaired Hbl secretion was also observed after deletion of flhF (signal recognition particle-like GTPase). Interestingly, enhanced levels of Nhe and further virulence factors were detected (Mazzantini et al., 2016[161]; Salvetti et al., 2007[200]). It has further been shown that FlhF homodimers can bind Hbl L2 with the purpose of directing it to the plasma membrane for secretion (Mazzantini et al., 2020[162]). In contrast to that, it was recently proven that a fliF (flagellar MS-ring) deletion compromises motility and adhesion, but not Hbl secretion (Mazzantini et al., 2024[163]).

Apart from these two secretion systems, Duport et al. (2020[58]) compared the dynamics of the proteome and exoproteome in B. cereus AH187, an emetic strain which also produces Nhe, at 16 and 30 °C growth temperature. The authors concluded that exotoxins such as Nhe or cereolysin O are a major part of temperature-induced proteome changes and are important in controlling homeostasis of the cellular proteome and exoproteome during cold stress (Duport et al., 2020[58]). In a similar approach, comparing B. cereus ATCC 14597 and a Δtspo mutant, it was found that this translocator protein (BcTSPO, Figure 3C(Fig. 3)) significantly influences cellular metabolism and stress response by re-structuring the cellular proteome and exoproteome. It was particularly noticeable that the enterotoxins Hbl, Nhe and CytK were significantly reduced in the exoproteome in the absence of BcTSPO, while the level of the above-mentioned cereolysin O increased (Duport and Armengaud, 2025[57]). It has also recently been discovered that extracellular vesicles can play an important role in virulence of B. cereus. It has been shown for strain NVH0075-95 that these sphere-shaped, membranous entities, which are released into the extracellular space by a yet not fully understood mechanism (Briaud and Carroll, 2020[27]), contain virulence factors such as sphingomyelinase, phospholipase C, and the three Nhe components (Buchacher et al., 2023[29]) (Figure 3C(Fig. 3)). Overall, B. cereus not only exhibits highly variable enterotoxin gene transcription, but also various, potentially interrelated secretion mechanisms, which may further account for the high toxin- and strain-specific differences.

The first studies conducted on Hbl activity described dermonecrotic, haemolytic and cytotoxic properties, activity in vascular permeability tests, as well as fluid accumulation in rabbit ileal loops (Beecher and Macmillan, 1991[15]; Beecher et al., 1995[18], 2000[17]; Beecher and Wong, 1994[19]; Lund and Granum, 1997[153]). The three Hbl components have a size of 45 kDa (L2), 36 kDa (L1) and 35 kDa (B) (Beecher and Macmillan, 1991[15]) (Figure 3D(Fig. 3)), but variable homologues can exist (Beecher and Wong, 2000[20]; Schoeni and Wong, 1999[203]). All three components are required for biological activity (pore-formation) of the toxin complex, as well as a specific binding order and a defined concentration ratio. Even before reaching the target cells, Hbl components can form complexes in solution, mainly B-B, B-L1, L2-L1, and only weakly, B-L2 (Jessberger et al., 2020[113]; Tausch et al., 2017[216]). This preparatory configuration facilitates rapid pore formation (Jessberger et al., 2014[109]; Jessberger et al., 2019[114]). The required binding order is B-L1-L2 (Jessberger et al., 2019[114]; Sastalla et al., 2013[201]), whereby the binding of Hbl B to the target cell surface is enhanced by its complexation with Hbl L1 (Jessberger et al., 2019[114]). In addition, the concentration ratio of the components plays a decisive role. Ratios of L2:L1:B = 1:1:10 and 10:1:10 lead to most rapid pore formation, thus, excess of Hbl B accelerates, excess of Hbl L1 decelerates pore formation, while the Hbl L2 levels seem not to be crucial (Jessberger et al., 2019[114]) (Figure 3E(Fig. 3)). Another study, using a Chinese hamster ovary cell-free system for the synthesis of the Hbl components, mainly confirmed these results, but further claimed pore-forming activity of a “Hbl L1-L2 pre-pore-complex” on Caco-2 cells (Ramm et al., 2021[195]). Contrary to that, a further study involving the production of His-tagged recombinant Hbl components in a prokaryotic expression system, did not determine haemolytic or cytotoxic activity of any two components (Chen et al., 2023[42]).

The Hbl pores in the plasma membrane are rather small with a probable channel diameter of approx. 1.2 nm (Beecher and Wong, 1997[21]; Jessberger et al., 2020[113]). Further, these moderately cation selective pores are rather unstable and of limited lifespan (Jessberger et al., 2020[113]). The putative fourth component of the Hbl complex, Hbl B', has first been detected in the secretome of B. cereus ATCC 14579 (Clair et al., 2010[45]), and later, using Western blot analyses, in concentrations from 0.5 to 4.8 ng/µl in 10 further enteropathogenic strains (Jessberger et al., 2022[111]). With its approx. 50 kDa, the protein is larger than Hbl B. Despite sequence similarities, it is not able to perform or enhance the function of Hbl B. On the contrary, when used in excess, Hbl B' reduces cytotoxic and haemolytic activity of Hbl (Jessberger et al., 2022[111]). Further, the protein shows no significant influence on the binding of Hbl B to the target cell surface; Hbl B' rather interacts directly with L1, thereby inhibiting its complex formation with B. Hbl B' can therefore be attributed an important regulatory function in Hbl complex formation, which is the balanced orchestration of Hbl B-L1 complexes and the corresponding free subunits (Jessberger et al., 2022[111]).

The three Nhe components have a size of 41 kDa (A), 39.8 kDa (B) and 36.5 kDa (C) (Granum et al., 1999[89]) (Figure 3D(Fig. 3)), but other variants may also occur (Stenfors Arnesen et al., 2008[214]). Due to structural similarities, both Hbl and Nhe were allocated to the Cytolysin A (ClyA) superfamily of α-helical pore forming toxins, which was ultimately confirmed by their crystal structure (Fagerlund et al., 2008[72]; Ganash et al., 2013[82]; Madegowda et al., 2008[154]; Phung et al., 2012[187]; Worthy et al., 2021[234]). Despite all similarities, the components of the two enterotoxin complexes cannot be interchangeably used or complement the respective other pore (Sastalla et al., 2013[201]). In contrast to Hbl, Nhe is known to form highly stable complexes of approx. 620 kDa between B and C, which can form pre-pores, transmembrane channels of approx. 2 nm diameter, in a suitable concentration ratio (NheB:C = 10:1) (Heilkenbrinker et al., 2013[98]; Zhu et al., 2016[240]). Nevertheless, all three Nhe components are necessary for maximum biological activity, with an optimum concentration ratio of NheA:B:C = 10:10:1 and a specific binding order (Lindbäck et al., 2004[141], 2010[142]) (Figure 3E(Fig. 3)). For completion of the whole pore, NheA, which cannot interact with the other components in solution, binds to cell-bound NheB and undergoes massive conformational changes (Didier et al., 2012[53], 2016[54]). The complete pore is large, > 2.8 nm, and stable, similarly to ClyA, and has a conductance of approx. 18 nS on planar lipid bilayers (Fagerlund et al., 2008[72]; Ramm et al., 2020[194] ). Similarly to Hbl and ClyA, it is cation-selective (Fagerlund et al., 2008[72]; Jessberger et al., 2020[113]; Ludwig et al., 1999[150]).

In contrast to Hbl and Nhe, CytK is a single protein toxin of approx. 34 kDa (Figure 3D(Fig. 3)). It is assigned to the family of β-barrel pore-forming toxins, which are typically present as monomers in solution, attach to the target cell membrane, oligomerise into heptamers, undergo conformational changes, and ultimately form the mushroom-shaped pore (Andreeva et al., 2006[9], 2007[8]; Cairo et al., 2022[32]; Menestrina et al., 2001[165]) (Figure 3E(Fig. 3)). CytK proved to be pore-forming, haemolytic, dermonecrotic and cytotoxic, with CytK-1 showing an approx. 80 % enhanced activity compared to CytK-2 (Fagerlund et al., 2004[73]; Hardy et al., 2001[95]). The pores are weakly anion selective with a predicted minimum diameter of approx. seven Å (Hardy et al., 2001[95]). Recently, Sauciuc et al. (2024[202]) showed that during electroosmotically driven translocation of polypeptides, blob-like structures are shaped inside nanopores, including those of CytK. These blobs seem to interfere with polypeptide transportation and inhibit the targeting of certain amino acids (Sauciuc et al., 2024[202]). However, it has recently been shown that CytK-1 appears to be less toxic than previously thought. Various B. cytotoxicus isolates, mainly obtained from potato products, showed little to no cytotoxicity towards Vero cells (Burtscher et al., 2021[30]; Etter et al., 2024[68]). Nevertheless, cytotoxicity tests on Caco-2 cells using cytK-1 knockout mutants proved that CytK-1 remains the major pathogenicity factor in B. cytotoxicus (Kone et al., 2021[126]).

The three main B. cereus enterotoxins act on a wide range of target cell lines and tissue with variable strength (Dietrich et al., 2021[56]). This and the non-interchangeability of the Hbl and Nhe protein components (see above), suggest different receptor types on the target cell surface. For Hbl, lipopolysaccharide (LPS)-induced tumour necrosis factor (TNF)-α factor (LITAF) was identified as the main receptor, using a genome-wide CRISPR-Cas9 knockout screen in RAW276.4 mouse macrophages. Additionally, the LITAF-like protein cell death-inducing P53 target 1 (CDIP1) was determined as a second, alternative receptor in LITAF-deficient cells. Cell lines as well as mice lacking LITAF and/or CDIP1 showed resistance towards Hbl-mediated toxicity (Liu et al., 2020[144]). Since LITAF functions as an activator of inflammation and is present in various human cell types, tissues, and organs, it is speculated that targeting this receptor via Hbl is part of the strategy of B. cereus to cause cell death responses in the host (Enosi Tuipulotu et al., 2021[66]). The receptor for Nhe, which likely differs from LITAF and CDIP1, has not yet been identified. Instead, it has been shown that the above-mentioned Nhe-containing extracellular vesicles use cholesterol-rich domains and dynamin-mediated endocytosis to connect with intestinal epithelial cells, resulting in internalisation of Nhe and prolonged cytotoxicity. These vesicles also seem to provoke an inflammatory response in human monocytes (Buchacher et al., 2023[29]). A specific receptor has not yet been identified for CytK either. However, it is understood that β-barrel pore-forming toxins typically bind to the target cell membrane through interaction with liposomes (Andreeva et al., 2007[8]; Menestrina et al., 2001[165]; Song et al., 1996[212]).

Shortly summarised, the B. cereus enterotoxins form pores leading to damage and osmotic lysis of intestinal epithelial cells, injury of microvilli, and consequently diarrhoea (Enosi Tuipulotu et al., 2021[66]). The cellular biological processes underlying this are vastly more complex and have not yet been fully elucidated. First, it has been demonstrated that Nhe pore formation causes calcium influx, oxidative stress and activation of apoptosis signal-regulating kinase 1 (ASK1) and p38 MAPK in Vero cells. Subsequent activation of caspase-8 and caspase-3 triggers apoptosis (Liu et al., 2017[145]). Newer studies showed that potassium efflux, caused by both Hbl and Nhe pores, leads to inflammasome-mediated mortality of macrophages (Fox et al., 2020[76]; Mathur et al., 2019[160]). As part of the innate immune system, this protein complex consists of a sensor protein, an apoptosis-associated-speck-like adaptor protein with a caspase recruitment and activation domain, and the cysteine protease caspase-1 (Enosi Tuipulotu et al., 2021[66]). In this case, the sensor protein is NLRP3 (NOD-, LRR- and pyrin domain-containing protein 3). Activation of the inflammasome leads to secretion of the proinflammatory cytokines interleukin-1β and interleukin-18, as well as cleavage of the pore-forming protein gasdermin D, causing pyroptosis, an inflammatory form of cell death (Enosi Tuipulotu et al., 2021[66]; Fox et al., 2020[76]; Mathur et al., 2019[160]). Not surprisingly, Hbl-induced NLRP3 inflammasome activation occurs quicker compared to Nhe (Fox et al., 2020[76]; Mathur et al., 2019[160]). This corresponds with the above-mentioned pore sizes as well as with the higher velocity of Hbl pore formation (Jessberger et al., 2014[109]). It is further speculated that different receptor densities on the target cell surface may also play a role (Enosi Tuipulotu et al., 2021[66]). Based on these studies, Zhao et al. also demonstrated induction of NLRP3/gasdermin D-dependent pyroptosis by pathogenic B. cereus H2 from a deep-sea cold seep as well as by purified, recombinant CytK (Zhao et al., 2021[237]; Zhao and Sun, 2022[238]).

As a further consequence of B. cereus infections, Jovanovic and Rajkovic (2023[122]) showed bioenergetic changes in Caco-2 cells, described as a more energetic phenotype caused by increased flux through oxidative phosphorylation and glycolysis, as well as higher ATP levels. As our own latest intestinal model, we exposed a 9:1 Caco-2/HT29-MTX co-culture to B. cereus culture supernatants as well as purified enterotoxin leading to enterotoxin-, dose-, differentiation- and time-dependent decline in trans-epithelial electrical resistance (TEER), as well as increased short-circuit currents (Isc) consistent with toxin-induced alterations in epithelial ion transport. Furthermore, the co-culture demonstrated partial recovery after toxin removal. Total transcriptome analyses revealed a specific time- and dose-dependent host response programme. Lethal toxin doses reduced proliferation- and cell cycle-related gene expression, while TNF/NF-κB-driven inflammation, MAPK-associated stress signalling, and cell-fate programmes involving the p53-mediated DNA-damage/stress response as well as regulated cell death (apoptosis) were up-regulated (Albahri et al., 2026[4]). Using a similar methodology but pursuing different goals, Fang et al. (2025[74]) found 937 differentially expressed genes in splenic cells of Chinese soft-shelled turtles after B. cereus infection and identified relevant pathways involved in the response to this infection.

The toxin is synthesised by a non-ribosomal peptide synthetase (NRPS). All genes involved in this process are encoded in what is known as the ces gene cluster, which is located on a pX01-like megaplasmid called pCER270 (Ehling-Schulz et al., 2005[64], 2006[62]). The 24 kbp gene cluster encodes seven enzymes, including a hydrolase/esterase (CesH), a phosphopantetheinyl transferase (CesP), a type II thioesterase (CesT), the structural NRPS core (CesA and CesB, assembling amino acids into the cyclic peptide) as well as an ABC transporter consisting of CesC and CesD (Ehling-Schulz et al., 2005[64], 2006[62]; Gacek-Matthews et al., 2020[81]). From the main promoter P1, six of the seven genes, cesPTABCD, are transcribed as a single 23-kb polycystron, mostly at late exponential growth phase. The role of additional promoters is not yet understood (Frenzel et al., 2012[79]; Lücking et al., 2009[148]) (Figure 4A(Fig. 4)). cesH, whose corresponding protein inhibits cereulide synthesis by a rather indirect mechanism, is expressed from its own promoter structure (Lücking et al., 2015[149]; Tian et al., 2019[219]). The ces gene cluster is not controlled by PlcR, but indeed by CodY (see above) and the transition state transcriptional regulator AbrB, as well as Spo0A (Frenzel et al., 2012[79]; Lücking et al., 2009[148]). Additionally, the plasmid-encoded, ArsR/SmtB family transcriptional regulator PagRBc was identified, which represses cereulide synthetase gene expression (Kalbhenn et al., 2022[123]). Moreover, posttranscriptional regulation plays an important role in cereulide synthesis, for example activation of NRPS by plasmid (CesP) and chromosomal 4´-phosphopanthetheinyl transferases (Ehling-Schulz et al., 2006[62]; Lücking et al., 2015[149]; Walsh, 2016[229]), or the essential part of ABC transporter CesCD in cereulide biosynthesis (Gacek-Matthews et al., 2020[81]). A role in cereulide production was also demonstrated for a flagellar hook protein after the corresponding gene flgE was knocked out (Li et al., 2022[139]). Transferring pCER270 from emetic to phylogenetically distant B. cereus group strains demonstrated transcriptional cross-regulation between the megaplasmid and the host genome, for instance regarding carbohydrate metabolism and sporulation genes expression, as well as pCER270 encoded genes (Nevers et al., 2023[176]).

Several studies have further shown that extrinsic parameters, such as temperature, oxygen content, a_W, pH, food matrix or food additives influence levels and even isoforms of produced cereulide (Frenzel et al., 2011[80]; Häggblom et al., 2002[94]; Kranzler et al., 2016[131]; Messelhäusser et al., 2014[168]; Rouzeau-Szynalski et al., 2020[196]; Wang et al., 2024[230]; Yang et al., 2023[235]). It is known that branched-chain amino acids enhance binding of CodY to the ces promoter region (Frenzel et al., 2012[79]) and just recently, Li et al. (2026[137]) demonstrated that isoleucine regulates cereulide biosynthesis via folate metabolism. Some phytochemicals, such as gingerol and curcumin, were shown to inhibit not only growth, but also cereulide production of emetic B. cereus at concentrations ≤ 0.01 % (Kranzler et al., 2021[130]). It has further been suggested that lactose can function as a signal molecule to regulate cereulide production by repressing ces gene expression, possibly due to inhibition of the Spo0A regulator (Zheng et al., 2024[239]). Modelling growth and cereulide production at different temperatures for microbiological food safety assessment, Ellouze et al. (2021[65]) concluded that cereal-based food matrices highly support cereulide production, followed by culture medium and dairy-based matrices with intermediate influence, and meat- and vegetable-based matrices with lowest cereulide formation. Similarly, Buss da Silva et al. (2022[31]) modelled time to first cereulide quantification, maximum growth rates and cereulide production rates as a function of temperature in an intermediate dairy wet-mix. They admitted, however, that the models cannot be easily transferred to unrelated food matrices, particularly with regard to cereulide production (Buss da Silva et al., 2022[31]). Moreover, cereulide production in an insect cadaver has been shown, allowing the conclusion that the toxin might be of ecological relevance for a saprophytic lifestyle (Jenull et al., 2023[108]), but also raising the question if edible insects might be a source of cereulide intoxication. Apart from different food matrices and temperatures, it is known that cereulide production ability differs widely among B. cereus isolates (Carlin et al., 2006[34]; Frentzel et al., 2024[78]), and three classes of cereulide producers have been defined: low (< 1-50 ng/g bacterial mass), medium (> 50-500 ng/g), and high (> 500-1600 ng/g) (Stark et al., 2013[213]). So far, this could be explained neither by sequence alterations in cesHPTABCD or the genes encoding the corresponding regulatory proteins, nor population structure and phylogeny of isolates subjected to whole genome sequencing. Nevertheless, the involvement of further associated proteins was suggested, which still needs to be verified (Frentzel et al., 2024[78]).

Of the seven Ces enzymes (see above), CesA and CesB comprise the actual cereulide synthetase, while CesA is responsible for D-Ala-D-O-Leu synthesis, and CesB for L-Val-L-O-Val (Marxen et al., 2015[159]; Rouzeau-Szynalski et al., 2020[196]) (see also Figure 4B(Fig. 4)). First, 4-methyl-2-oxopentanoic acid is converted into d-2-hydroxy carboxylic acid by the ketoreductase domain of CesA. Second, the acid is activated as adenosine monophosphate ester by the adenylation domain. The epimerisation domain epimerises L-Ala and another adenylation domain activates it. Subsequently, both monomers are attached to a peptidyl carrier protein and the free amine of D-Ala attacks on the D-O-Leu thioester. The condensation domain catalyses dimer formation (Jovanovic et al., 2021[121]). Similarly, L-Val-L-O-Val is synthesised by CesB. Then, another condensation domain facilitates the attack of D-O-Leu on the thioester-activated L-Val-L-O-Val, which transfers a tetradepsipeptide onto the PCP domain of CesA. In a last step, the thioesterase domain of CesB takes up the tetradepsipeptide and further chain elongation and macrocyclisation take place (Alonzo et al., 2015[5]; Jovanovic et al., 2021[121]; Magarvey et al., 2006[155]). Further, 4′-phosphopantetheinyl-transferase CesP activates NRPS and thioesterase CesT most likely removes misprimed monomers from the complex as part of its proofreading function. As mentioned above, CesCD is an ABC transporter most likely responsible for cereulide export, but also for self-resistance towards the toxin. CesH is probably involved in the right timing of cereulide synthesis (Lücking et al., 2015[149]; Rouzeau-Szynalski et al., 2020[196]).

Next to the initially discovered cereulide, NRPS also synthesises different isoforms, known as isocereulides, of which various have been identified to date (Figure 4C(Fig. 4)). An emetic B. cereus isolate is able to produce all cereulide isoforms. Nevertheless, strain- as well as temperature-dependent differences in their ratios were detected (Kranzler et al., 2016[131] ; Marxen et al., 2015[158][159]; Rouzeau-Szynalski et al., 2020[196]). A recent study showed stable ratios of cereulide and isocereulides A-N of approx. 1:9 both under standard laboratory growth conditions and in contaminated fried rice balls. In contrast to that, the ratios between the isocereulides varied time-dependently, indicating different kinetics of isocereulide production (Kranzler et al., 2024[132]). The structure of the isocereulides A-G has been resolved first (Marxen et al., 2015[158]; Pitchayawasin et al., 2004[189]; Stark et al., 2013[213]; Walser et al., 2021[228]). In comparison to cereulide, [D-O-Leu-D-Ala-L-O-Val-L-Val]₃, the structure of highly toxic isocereulide A, for example, is [(D-O-Leu-D-Ala-L-O-Val-L-Val)₂(D-O-Leu-D-Ala-L-O-Ile-L-Val)] (Walser et al., 2021[228]). Since then, the structures of the newly discovered isocereulides H-N have also been elucidated, all of which exhibit similar structural modifications compared to cereulide (Walser et al., 2022[227]) (Figure 4C(Fig. 4)).

Cereulide was shown to restrain RNA and ATP synthesis, proliferation, as well as motility in eukaryotic cells. It was also suggested to interfere with the host's immune system (Agata et al., 1994[3]; Paananen et al., 2002[184]). Further, it is a potassium binding dodecadepsipeptide structurally resembling the ionophore valinomycin (Mikkola et al., 1999[170]; Suwan et al., 1995[215]; Teplova et al., 2006[217]) (compare Figure 4C(Fig. 4)). A classical ion-carrier transportation mechanism has been described: Cereulide-mediated transport across the lipid bilayer led to depolarisation and disturbance of the membrane potential in rat liver mitochondria by influx of K⁺ ions, which impairs ATP driving force and cellular respiration (Jääskeläinen et al., 2003[107]; Jovanovic et al., 2021[121]; Mikkola et al., 1999[170]; Teplova et al., 2006[217]) (Figure 4D(Fig. 4)). After slow release of K⁺ ions into the mitochondria, cereulide can diffuse back to the cytosol for continuous K⁺ uptake (Teplova et al., 2006[217]; Yang et al., 2023[235]). The different cereulide isoforms (see above) can vary significantly in their cytotoxic potential, reaching from 50 % to up to 8-fold compared to cereulide (Marxen et al., 2015[158]; Walser et al., 2022[227]). These variations can be explained by different membrane activity of the isocereulides (Marxen et al., 2015[158]). Constituting only 10 % of total toxins, isocereulides A-N were responsible for approx. 40 % of total cytotoxicity towards Hep2 cells in a recent study. Moreover, individual isocereulides showed higher or lower toxicity than cereulide, as well as additive and synergistic effects with the original toxin. This suggests that total toxicity originates from a mixture of cereulide and its variants, and that these variants must therefore also be included in routine food and clinical diagnostics (Kranzler et al., 2024[132]; Walser et al., 2022[227]). Until then, the effective dose leading to the onset of the emetic disease in humans is difficult to estimate, as it depends not only on cereulide, but also on its isoforms, as well as the susceptibility of the host, and the quantity of ingested food (Jovanovic et al., 2021[121]).

To better understand cereulide uptake, resorption and translocation within the host, the toxin was orally administered to piglets (Bauer et al., 2018[11]). While one part was rapidly excreted, another part was absorbed and accumulated in kidney, liver, muscles, fat and even brain tissues, leading to symptoms similar to critical human intoxications such as lethargy, seizures, and convulsions. It was concluded that the central nervous system is an important target for cereulide (Bauer et al., 2018[11]).

Since B. cereus is a ubiquitous soil inhabitant, it is reasonable to assume that cereulide plays an important ecological role. In a recent study, antibacterial activity of cereulide was tested and proven to be species-specific as well as enhanced with rising environmental K⁺ levels. Antifungal activity was K⁺-independent and only observed for Candida albicans and Rhodotorula species, while cereulide seemed to have no effect on Nitrososphaera viennensis belonging to the group of ammonia-oxidising archaea. Mimicking intoxication, cytotoxic activity towards Hep-2 cells increased with rising external K⁺ concentrations, while only cereulide and not external K⁺ concentrations showed an effect on the development of Oesophagostomum dentatum (nematode) larvae (Jenull et al., 2023[108]). Via experiments in a Suncus murinus animal model, cereulide has been found to interact with serotonin dependent receptors (5-HT3) stimulating the afferent vagus nerve and activating the vomiting centre in the medulla oblongata (Agata et al., 1994[3]). In relation to this, the cereulide dose necessary to cause emesis in S. murinus and rhesus monkeys has been determined at approx. 10 μg/kg body weight (Shinagawa et al., 1995[208]). The corresponding dose necessary to cause emesis in humans is not yet known.

Several studies demonstrated effects on gastrointestinal cells such as Caco-2 and HT-29. Among others, these were reduced non-mitochondrial respiration, maximal respiration, high-density lipoprotein (HDL) and ATP production, as well as inhibition of cell proliferation and the disturbance of intestinal barrier function after chronical, low dose cereulide exposure (Decleer et al., 2018[51]; Lin et al., 2021[140]; Rajkovic et al., 2014[193]; Yang et al., 2023[235]). Intestinal inflammation as well as apoptosis have also been observed in a mouse model mimicking chronic low-dose cereulide exposure. These resulted from endoplasmic reticulum (ER) stress IRE1/XBP1/CHOP pathway activation (Lin et al., 2021[140]). Dose-dependently, HepG2 (liver) cells showed reduction in maximum respiration and ATP production (Decleer et al., 2018[51]). Moreover, upon intraperitoneal cereulide exposure, mice showed pathological changes in the liver, such as hepatocyte degeneration, inflammation and necrosis, culminating in death of the mice at high doses (Yokoyama et al., 1999[236]). Modelling long-time cereulide exposure, Li et al. (2021[136]) found induction of oxidative stress and inflammation in mice, promoting apoptosis of liver and kidney cells. In detail, endoplasmic reticulum (ER) stress was activated via the pathways of inositol-requiring enzyme 1α (IRE1α)/Xbox binding protein (XBP1) and PRKR-like ER kinase (PERK)/eukaryotic translation initiation factor 2α (eIF2α). The authors further showed that this was due to accumulation of reactive oxygen species (Li et al., 2021[136]). In mouse pancreatic islets as well as in foetal porcine Langerhans islets, results ranged from reduced insulin production to death of MIN6 cells, depending on cereulide concentrations added (Hoornstra et al., 2013[103]; Vangoitsenhoven et al., 2014[223]; Virtanen et al., 2008[224]). Additionally, interference of cereulide with serotonin synthesis has been observed, leading to depression in animals. For this, cereulide either acts in the intestine on the synthesis of its precursor or, after crossing the blood-brain barrier, directly in the central nervous system (CNS) by inducing apoptosis (Bauer et al., 2018[11]; Lin et al., 2021[140]). Just recently, a complex neuronal mechanism of cereulide-induced retching as defence mechanism was discovered, involving Calbindin1-positive (Calb1⁺) neurons of the nucleus of the solitary tract (NTS). It was shown that cereulide directly modulates vagal sensory neurons which innervate Calb1⁺ NTS neurons. Further, it was demonstrated that two pathways, Calb1^NTS-PBNel and Calb1^NTS-Amb/RVLM, mediate retching movement and nausea, respectively (Huo et al., 2024[104]). The toxin can also have a significant impact on the immune system, which was shown by loss of function of natural killer cells at low, and induction of apoptosis at high cereulide concentrations (Paananen et al., 2002[184]). These studies are also reflected in human cases of cereulide intoxication, in which liver (most commonly; Dierick et al., 2005[55]; Ichikawa et al., 2010[105]; Mahler et al., 1997[156]; Naranjo et al., 2011[175]; Shiota et al., 2010[209]; Tschiedel et al., 2015[220]), intestine (Naranjo et al., 2011[175]), pancreas (Pósfay-Barbe et al., 2008[191]; Thery et al., 2022[218]), kidney (Naranjo et al., 2011[175]; Pósfay-Barbe et al., 2008[191]) and/or CNS (Ichikawa et al., 2010[105]; Mahler et al., 1997[156]; Pósfay-Barbe et al., 2008[191]; Shiota et al., 2010[209]) were affected.

The authors declare that they have no conflict of interest.

The authors declare that no type of generative artificial intelligence has been used for the writing of this manuscript, nor for the creation of images, graphics, tables, or their corresponding captions.

This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - project number 407925087.