Comparative Bioinformatics and Experimental Analysis of the Intergenic Regulatory Regions of Bacillus cereus hbl and nhe Enterotoxin Operons and the Impact of CodY on Virulence Heterogeneity

Bacillus cereus is a food contaminant with greatly varying enteropathogenic potential. Almost all known strains harbor the genes for at least one of the three enterotoxins Nhe, Hbl, and CytK. While some strains show no cytotoxicity, others have caused outbreaks, in rare cases even with lethal outcome. The reason for these differences in cytotoxicity is unknown. To gain insight into the origin of enterotoxin expression heterogeneity in different strains, the architecture and role of 5′ intergenic regions (5′ IGRs) upstream of the nhe and hbl operons was investigated. In silico comparison of 142 strains of all seven phylogenetic groups of B. cereus sensu lato proved the presence of long 5′ IGRs upstream of the nheABC and hblCDAB operons, which harbor recognition sites for several transcriptional regulators, including the virulence regulator PlcR, redox regulators ResD and Fnr, the nutrient-sensitive regulator CodY as well as the master regulator for biofilm formation SinR. By determining transcription start sites, unusually long 5′ untranslated regions (5′ UTRs) upstream of the nhe and hbl start codons were identified, which are not present upstream of cytK-1 and cytK-2. Promoter fusions lacking various parts of the nhe and hbl 5′ UTR in B. cereus INRA C3 showed that the entire 331 bp 5′ UTR of nhe is necessary for full promoter activity, while the presence of the complete 606 bp hbl 5′ UTR lowers promoter activity. Repression was caused by a 268 bp sequence directly upstream of the hbl transcription start. Luciferase activity of reporter strains containing nhe and hbl 5′ IGR lux fusions provided evidence that toxin gene transcription is upregulated by the depletion of free amino acids. Electrophoretic mobility shift assays showed that the branched-chain amino acid sensing regulator CodY binds to both nhe and hbl 5′ UTR downstream of the promoter, potentially acting as a nutrient-responsive roadblock repressor of toxin gene transcription. PlcR binding sites are highly conserved among all B. cereus sensu lato strains, indicating that this regulator does not significantly contribute to the heterogeneity in virulence potentials. The CodY recognition sites are far less conserved, perhaps conferring varying strengths of CodY binding, which might modulate toxin synthesis in a strain-specific manner.

Bacillus cereus is a food contaminant with greatly varying enteropathogenic potential. Almost all known strains harbor the genes for at least one of the three enterotoxins Nhe, Hbl, and CytK. While some strains show no cytotoxicity, others have caused outbreaks, in rare cases even with lethal outcome. The reason for these differences in cytotoxicity is unknown. To gain insight into the origin of enterotoxin expression heterogeneity in different strains, the architecture and role of 5 intergenic regions (5 IGRs) upstream of the nhe and hbl operons was investigated. In silico comparison of 142 strains of all seven phylogenetic groups of B. cereus sensu lato proved the presence of long 5 IGRs upstream of the nheABC and hblCDAB operons, which harbor recognition sites for several transcriptional regulators, including the virulence regulator PlcR, redox regulators ResD and Fnr, the nutrient-sensitive regulator CodY as well as the master regulator for biofilm formation SinR. By determining transcription start sites, unusually long 5 untranslated regions (5 UTRs) upstream of the nhe and hbl start codons were identified, which are not present upstream of cytK-1 and cytK-2. Promoter fusions lacking various parts of the nhe and hbl 5 UTR in B. cereus INRA C3 showed that the entire 331 bp 5 UTR of nhe is necessary for full promoter activity, while the presence of the complete 606 bp hbl 5 UTR lowers promoter activity. Repression was caused by a 268 bp sequence directly upstream of the hbl transcription start. Luciferase activity of reporter strains containing nhe and hbl 5 IGR lux fusions provided evidence that toxin gene transcription is upregulated by the depletion of free amino acids. Electrophoretic mobility shift assays showed that the branched-chain amino acid sensing regulator CodY binds to both nhe and hbl 5 UTR downstream of the promoter, potentially acting as a nutrient-responsive roadblock repressor of toxin gene transcription. PlcR binding sites are highly conserved among all B. cereus sensu lato strains,

INTRODUCTION
Bacillus cereus is an opportunistic pathogen and food contaminant that produces several toxins causing gastrointestinal illness in humans. B. cereus sensu stricto is closely related to B. thuringiensis and B. anthracis, which are regarded to be one species on the basis of high core-genomic relatedness (Ash et al., 1991). Together with B. mycoides, B. pseudomycoides, B. weihenstephanensis, B. toyonensis, and B. cytotoxicus they form the B. cereus sensu lato group. Due to their diverse lifestyles, differences in plasmid content, and varying toxinogenic potentials, some strains of B. cereus are considered to be probiotics (Hong et al., 2005), while others are opportunistic pathogens, causing several human infections such as endophthalmitis (David et al., 1994), meningitis (Barrie et al., 1992) and periodontitis (Helgason et al., 2000) or foodborne illness (Stenfors Arnesen et al., 2008). Occasionally, B. thuringiensis strains were reported to be responsible for human infections resembling B. cereus infections (Damgaard et al., 1997;Kuroki et al., 2009). B. cereus spores are frequently detected in food originating from soil, dust and plant material. They are transferred through air as well as by cross-contamination from food and food-processing equipment (Frankland and Frankland, 1887;Daelman et al., 2013). Two different forms of food poisoning are recognized: Emesis is caused by ingestion of the small, cyclic and heat stable dodecadepsipeptide cereulide with a short incubation period of 0.5-6 h. The diarrheal type is caused by single or combined action of heat-labile enterotoxins acting on epithelial cells of the gastro-intestinal tract with an incubation period of 8-16 h (Shinagawa, 1990;Ehling-Schulz et al., 2004). The three most important pore-forming B. cereus cytotoxins that have been linked to diarrheal disease are hemolysin BL (Hbl), non-hemolytic enterotoxin (Nhe) and cytotoxin K (CytK). Hbl consists of a binding component B and the two lytic components L 1 and L 2 . These proteins are encoded in one operon hblCDAB by the genes hblA, hblD, and hblC, respectively (Guinebretiere et al., 2002). Nhe is a tripartite toxin encoded by the operon nheABC and acts cytolytic on erythrocytes and epithelial cells due to its ability to form pores in the plasma membrane . All three components NheA, NheB, and NheC are required for full toxic activity, although NheC is only expressed in small amounts due to translational repression (Lindbäck et al., 2004). The third diarrhea causing agent is the single-component toxin CytK (34 kDa), which is a hemolytic and dermonecrotic β-barrel pore-forming enterotoxin (Lund et al., 2000).
It was shown that toxicity varies greatly between strains and that the sole presence of enterotoxin genes is not sufficient Abbreviations: BCAA, branched-chain amino acid; CAAs, casamino acids; 5 IGR, 5 intergenic region; 5 UTR, 5 untranslated region; TSS, transcription start site.
for the classification of a B. cereus strain as being pathogenic or apathogenic (Dietrich et al., 2005;Jeßberger et al., 2015). Indeed, enterotoxin expression is highly complex and strainspecifically regulated (Jeßberger et al., 2015). The promoter regions of both nhe and hbl were described to harbor binding sites for a variety of transcriptional regulators such as Fnr, ResD, SinR, CcpA and PlcR, which supposedly control enterotoxin expression in B. cereus in a concerted action (Esbelin et al., 2008(Esbelin et al., , 2009Gohar et al., 2008;van der Voort et al., 2008;Fagerlund et al., 2014). These regulators control the expression of either broad or narrow-spectrum regulons responding to oxygen tension (ResD, Fnr), carbohydrate availability (CcpA) and to the B. cereus group specific quorum sensing peptide PapR (PlcR). However, enterotoxin expression is additionally regulated by nitrogen source availability and the general energetic cell status, sensed by the pleiotropic regulator CodY (Frenzel et al., 2012). CodY is a global regulator of adaptation to unfavorable environments, sensing nutrient availability through interaction with GTP and the BCAAs isoleucine, leucine and valine (Ratnayake-Lecamwasam et al., 2001;Shivers and Sonenshein, 2004;Sonenshein, 2005). In the majority of low GC gram-positive bacteria, the CodY regulon controls profound cellular functions, such as motility, chemotaxis, catabolism, production of proteases and virulence (Shivers and Sonenshein, 2004;Sonenshein, 2005;Handke et al., 2008;Majerczyk et al., 2010). In B. cereus, CodY indirectly controls the expression of virulence genes via the activation of the PlcR regulon (Frenzel et al., 2012;Lindbäck et al., 2012;Slamti et al., 2015), which includes the enterotoxin operons nhe, hbl, and cytK (Agaisse et al., 1999;Lund et al., 2000).
To decipher the basis of differing enterotoxin expression potentials, we compared 5 IGRs of the most prominent enterotoxins (Nhe, Hbl, and CytK) by means of bioinformatics and in vitro protein-DNA binding experiments. The contribution of the strikingly long stretches of 5 IGRs preceding the nhe and hbl operons to enterotoxin expression was studied in detail with consecutively trimmed 5 IGR-luxABCDE fusions. We provide evidence that the hbl 5 UTR naturally attenuates promoter activity, while the entire nhe 5 UTR is necessary for maximal promoter activity. Our results show that CodY may modulate hbl and nhe expression by acting as a downstream roadblock repressor of toxin gene transcription in a strain-dependent manner.

Microbial Strains
In order to compare intergenic regulatory regions upstream of the enterotoxin operons hbl, nhe, and cytK strains representative for the seven phylogenetic groups were selected from a set of 142 B. cereus sensu lato strains  Table 1). Out of these, 29 nhe, 23 hbl, 3 cytK-1, and 15 cytK-2 5 IGRs were examined. Enterotoxin promoter activity was studied experimentally in B. cereus INRA C3 (IV) and gel mobility shift assays were performed with both B. cereus INRA C3 and B. cytotoxicus CVUAS 2833 (VII).

Bioinformatic Analyses
Multiple sequence alignments were computed using Clustal 1 and sequence conservation was graphically represented as sequence logo 2 . The web-based program ORF Finder 3 was used to detect potential open reading frames (ORFs) embedded in the 5 UTRs of enterotoxin genes. The 5 UTR RNA sequences were further analyzed for similarity to known RNA families with Rfam v.12.0 (Nawrocki et al., 2015) and potential RNA secondary structures were predicted at default settings (linear RNA, 37 • C, 1 M NaCl) with Mfold v.4.6 (Zuker, 2003).

Media and Growth Conditions
All cloning steps were performed in Escherichia coli TOP10 grown in lysogeny broth (LB: 5 g/l yeast extract, 10 g/l tryptone, 10 g/l sodium chloride) at 150 rpm or on LB agar plates at 37 • C. B. cereus INRA C3 was grown in modified MOD (Frenzel et al., 2012) or CGY medium (Beecher and Wong, 1994) at 30 • C unless stated otherwise. When appropriate, cultures (16 h, 150 rpm) were supplemented with 120 µg/ml ampicillin or 5 µg/ml chloramphenicol. Modified MOD medium was used for the determination of promoter activity. Stock solutions of 2 M glucose and trace elements were prepared with ddH 2 O, filter sterilized (0.22 µm pore size), and added to the MOD medium to a final concentration of 20 mM. To obtain MOD + 1% CAAs or MOD + 1% tryptone, 10 g/l CAA, or 10 g/l tryptone were dissolved and autoclaved together with the MOD medium components (110 • C, 10 min). CGY medium was prepared in a volume of 900 ml ddH 2 O. After autoclaving 100 ml of filter sterilized glucose were added to a final concentration of 1%. The media compositions are listed in detail in Supplementary  Table S1.

Determination of Transcription Start Sites with 5 RACE
Bacilli were cultured in CGY medium supplemented with 1% (w/v) glucose as described previously (Jeßberger et al., 2015) to an OD 600 of 4. Six milliliter of the cultures were harvested (10.000 g, 4 • C, 10 min), cell pellets were snap-frozen in liquid nitrogen and stored at −80 • C. Total RNA was extracted by resuspending the pellet in 1 ml TRIreagent (Sigma-Aldrich) followed by cell disruption using a Fastprep 24 instrument (M. P. Biomedicals, 0.1 mm zirconia beads). DNase I digestion and RNA isolation were performed as previously described (Dommel et al., 2010). DNA-free RNA was used as template for 5 RACE (5 RACE system for rapid amplification of cDNA ends, version 2.0, Invitrogen). Gene specific primers to detect TSSs of hbl and nhe are listed in Supplementary Table S2.

Construction of Bioluminescent B. cereus Reporter Strains
To study gene regulation and promoter activities, bioluminescent B. cereus reporter strains were constructed. The promoter region of interest was inserted into the E. coli/Bacillus shuttle vector pXen1 (Francis et al., 2001), which contains the luciferase gene cassette luxABCDE by using primers comprising restriction sites for EcoRI and BamHI (Supplementary Table S2). Fragments were amplified from genomic DNA using a proof reading Pfu polymerase (Promega). The resulting plasmid was propagated in non-methylating E. coli INV110 and introduced in B. cereus by electroporation as described previously (Ehling-Schulz et al., 2005). Promoter fusions were verified by amplification of the insertion region with primers pXen for and pXen rev (Supplementary Table S2) and by sequencing (GATC Biotech).
Promoter fusions containing internal deletions were constructed by amplification of two regions adjacent to the target region for deletion by introducing XmaI restriction sites up-and downstream of the target region using the primer pairs A-B and C-D (Supplementary Table S2). After digestion and ligation of the two fragments AB and CD, a nested PCR with primers containing restriction sites for EcoRI and BamHI (Nhe for/Nhe rev or Hbl for/Hbl rev, respectively) allowed directional insertion of the promoter regions containing internal deletions into pXen1 as described above.

Luciferase Assay
Enterotoxin promoter activity of B. cereus strains grown in different media at 37 • C was monitored by detecting the luminescence signals at 490 nm using the Victor3 TM multilabel plate reader (Perkin Elmer). Precultures grown for 16 h at 150 rpm were diluted 1:1000 in their respective medium supplemented with 5 µg/ml chloramphenicol, distributed into a 96-well microtiter plate (µClear white, Greiner Bio-One) and incubated at 800 rpm and 37 • C. Each condition was tested in quadruplicates and for each strain three independent biological replicates were analyzed. Cell density (OD 600 ) and luminescence (490 nm, 0.1 s) were measured in hourly intervals for 19 h to detect the maximal promoter activity. Significance of differences between measured activities was calculated in the R free statistical software (version 3.1.1) 4 using Welch's t-test.

CodY Overexpression and Purification
Heterologous expression of CodY from B. cereus INRA C3 and B. cytotoxicus CVUAS 2833 was performed in E. coli BL21(DE3) as a soluble N-terminal His 6 -tag fusion protein using the plasmid pET28b(+) as described previously (Frenzel et al., 2012). The primers CodY-C3-for, CodY-C3-rev, CodY-CVUASfor and CodY-CVUAS-rev (Supplementary Table S2) containing restriction sites for NdeI and XhoI, respectively, were used to construct the overexpression plasmid. Protein expression was induced in LB medium at an OD 600 of 0.6 with 1 mM IPTG and cells were harvested after 5 h. Cell disruption and protein purification were performed as described (Frenzel et al., 2012) with the exception of using the Äkta purifier (Amersham Biosciences) with a Frac-950 fractionator.
A step-wise elution was performed by increasing imidazole concentration from 10 mM to 83.5 mM, 304 mM, and 402 mM to a final concentration of 500 mM. CodY-containing fractions were pooled and then dialyzed and concentrated in buffer BS using ultrafiltration columns with a 10 kDa cut-off (Amicon Ultra-15, Merck Millipore). Protein purity was analyzed on a 15% SDS-polyacrylamide gel with Coomassie staining.

Electrophoretic Mobility Shift Assays
DNA fragments containing the promoter regions PcytK-2 (B. cereus INRA C3), PcytK-1 (B. cytotoxicus CVUAS 2833, which is identical to PcytK-1 of the type strain B. cytotoxicus NVH 391-98), parts of Phbl (B. cereus INRA C3) and Pnhe (B. cereus INRA C3 and B. cytotoxicus CVUAS 2833) were amplified by PCR. Primer pairs and fragments ranging from 241 to 568 bp are listed in Supplementary Table S3. A fragment of the 16S rRNA gene rrn was used as negative control, since it lacks any similarity to the CodY consensus sequence. Electrophoretic mobility shift assays were performed as described previously (Frenzel et al., 2012) at 4 • C with varying amounts of protein and 100 ng of target DNA. The molarity of DNA fragments is given in Supplementary Table  S3. The equilibrium dissociation constant K D was estimated on the basis of three independent replicates of each electrophoretic mobility analysis as described earlier (Belitsky and Sonenshein, 2011b).

Length of 5 Intergenic Regions (5 IGRs)
To investigate the origin of differences in the regulation of enterotoxin expression, the 5 IGRs of cytK-1, cytK-2, nhe and hbl of 142 B. cereus sensu lato strains  were compared in a multiple sequence alignment approach (data not shown). Out of these, 29 nhe, 23 hbl, 3 cytK-1, and 15 cytK-2 toxin genes of B. cereus strains representing the diversity of the seven phylogenetic groups of B. cereus sensu lato (Guinebretiere et al., 2008;Böhm et al., 2015) were selected and compared (Figure 1).
The promoter containing intergenic regions of cytK-1 and cytK-2 ( Figures 1A,B) were found to be relatively short (∼100 bp) in comparison to the nhe and hbl 5 IGRs. CytK-1 is exclusively found in rare B. cytotoxicus isolates, the most distant cluster of species within the B. cereus group (Guinebretiere et al., 2013). However, despite the limited strain number, the high similarity of 5 cytK-1 IGRs analyzed in this study is in agreement with the potentially highly clonal structure of this species (Guinebretiere et al., 2013). The cytK-2 5 IGR, in contrast, consists of a differently sized variable region and a highly conserved promoter.
Some strains contain short insertions within their nhe 5 IGR, such as B. pseudomycoides DSM 12442 and other strains of the phylogenetic cluster I, which possess insertions downstream of each PlcR binding site ( Figure 1C). The nhe 5 IGR is approximately 350 bp longer in B. pseudomycoides strains compared to that of all other B. cereus group strains investigated. In contrast, B. cytotoxicus (cluster VII) contains an nhe 5 IGR which is ∼50 bp shorter than the other 5 IGRs, thus lacking the second PlcR binding site (data not shown). B. weihenstephanensis WSBC 10204 (cluster VI) contains a 14 bp insertion upstream of the hbl ribosomal binding site ( Figure 1D). Strains of clusters III and IV, which harbor many pathogenic B. cereus isolates, lack 11 bp close to each ResD binding site within the hbl 5 IGR. These missing regions might be used as an additional target to discern cluster III and IV from other B. cereus strains by molecular techniques. However, there is no correlation to high toxin production, since both high and low enterotoxigenic strains were found to harbor this deletions (Jeßberger et al., 2015). Taken together, all 5 IGRs comprise well-conserved regions, but lengths of hbl and cytK-2 5 IGRs are highly variable. With the exception of B. pseudomycoides, the insertions or deletions within 5 IGRs are not specific at the species level.

Transcription Start Sites and 5 Untranslated Regions (5 UTRs)
The TSSs of nhe in B. cereus strains NVH 0075-95 and NVH 1230-88 are localized 66 and 62 bp, respectively, upstream of the nheA startcodon (Lindbäck et al., 2004), while in B. thuringiensis Bt407 the TSS was reported to be located 331 bp upstream of nheA (Agaisse et al., 1999). This might indicate strain-specific differences in the promoter architecture. However, a 5 RACE analysis revealed that the TSS of the nhe operon in B. cereus INRA C3 is identical with the TSS in B. thuringiensis Bt407 ( Figure 1C). The 5 UTR is approximately 350 bp long and shows a high degree of sequence similarity to all other analyzed nhe 5 UTRs.
The TSS of the hbl operon in B. cereus ATCC 14579 was reported to localize 606 bp upstream of hblC (Lindbäck et al., 1999), which is identical to the transcription start of hblC in B. cereus INRA C3 (our data, Figure 1D). The TSS of the hbl cluster in B. thuringiensis Bt 407 is located 605 bp upstream of hblC (Agaisse et al., 1999). Thus, the 5 IGR of hbl comprises an exceptionally long, generally conserved 5 UTR of approximately 660 bp.
This study shows for the first time that the nhe and the hbl toxin operons are preceded by extended and widely conserved 5 UTRs in their upstream intergenic sequences.

Binding Sites of Transcriptional Regulators
Many regulator binding sites within the promoter regions Pnhe, Phbl, and PcytK have been confirmed by experimental studies in individual strains during recent years. In this study, putative recognition sites of all known transcriptional regulators demonstrated so far to be involved in B. cereus enterotoxin expression were predicted by 5 IGR alignments (Figure 1). In the nhe and hbl promoter regions putative binding sites for an overlapping set of transcription factors were identified. Two PlcR binding sites are localized within the nhe 5 IGR. PlcR 1 is 16 bp long and highly conserved, while the less conserved PlcR binding site 2 contains a 2 bp central insertion compared to the consensus sequence TATGNAN 4 TNCATA . Our sequence comparison-based analysis additionally predicted the stabilizing Shine-Dalgarno sequence in Phbl and putative binding sites for the master regulator of biofilm formation SinR in PcytK-1, Pnhe, and Phbl. In contrast to the tripartite enterotoxin operons nhe and hbl, cytK-1 and cytK-2 are not preceded by 5 UTRs. Apart from the detection of the PlcR binding site immediately upstream of the -35 element no further transcription factor binding site was described so far (Lund et al., 2000;Brillard and Lereclus, 2004). Our bioinformatical analysis of the 5 IGR  Structure and sequence of intergenic regions containing enterotoxin promoters in Bacillus cereus sensu lato. Sequence motifs were determined by sequence comparison. Promoter regions (−35, −10) and TSSs (+1, vertical arrow; Agaisse et al., 1999;Lindbäck et al., 1999Lindbäck et al., , 2004Brillard and Lereclus, 2004), CodY binding sites (den Hengst et al., 2005;Wray and Fisher, 2011;Frenzel et al., 2012;Belitsky and Sonenshein, 2013), catabolite responsive element (Cre; van der Voort et al., 2008), PlcR binding sites (Agaisse et al., 1999;Brillard and Lereclus, 2004;Lindbäck et al., 2004;Gohar et al., 2008), ribosomal binding site (RBS) of hbl (Ryan et al., 1997), ResD and Fnr binding sites (Geng et al., 2007;Esbelin et al., 2008Esbelin et al., , 2009, SinR binding sites (Kearns et al., 2005;Chu et al., 2006), and stabilizing Shine-Dalgarno sequence (Stab-SD; Agaisse and Lereclus, 1996). Conservation of the sequences is depicted as logo and based on a multiple sequence alignment of strains representative for the seven phylogenetic groups (see Table 1 and Böhm et al., 2015). TSSs in B. cereus INRA C3 are marked by an asterisk. Colored arrows indicate gene function and transcriptional direction of genes. AraC: AraC family transcriptional regulator TrrA. Black arrow: transcriptional direction of a potential open reading frame (ORF). (A) CytK-1 5 IGR of 3 B. cytotoxicus strains. Cluster VII strains contain a non-annotated ORF upstream of cytK-1 (unknown function: UF). (B) CytK-2 5 IGR of 15 B. cereus sensu lato strains. All strains are affiliated to phylogenetic clusters II-V. With the exception of the promoter, the intergenic region upstream of cytK-2 is not conserved. 80% of all cytK-2 strains possess a phosphoesterase gene directly upstream of cytK-2. In the remaining strains an insertion of 500-2000 bp length separates the two genes. (C) Nhe 5 IGR of 29 B. cereus sensu lato strains. All clusters except I and VII contain a hypothetical amino acid permease gene upstream of nheA. The cluster VII (B. cytotoxicus) intergenic region contains the same promoter elements as the other strains with an overall sequence identity of 70-90%, but is ∼50 bp shorter. The intergenic region of cluster I strains (B. pseudomycoides) consists of the same promoter elements, but is ∼350 bp longer. Clusters I and VII strains were excluded from the analysis due to distortion of the multiple sequence alignment. (D) Hbl 5 IGR of 23 B. cereus sensu lato strains. In 95% of all investigated hbl strains araC appears 1600-1200 bp upstream of hblC. The intergenic region upstream of hbl varies in size. Presented is the entire region of which up to 500 bp are lacking in several strains. Insertions occur in clusters II, V, and VI (nucleotides 1182-1192, 1651-1661) and in strains B. weihenstephanensis WSBC 10204 (nucleotides 696-720, 1706-1720) and B. cereus BAG5X1-1 (nucleotides 696-720, 1604-1610). A putative ORF starting with an alternative start codon (in most strains TCA or TAT) is noted. The hbl operon is part of a degraded, in most cases no longer functional transposon . As a remnant, a transposase (pseudogene) occurs in two cluster VI strains (B. cereus BAG5X1-1) instead of araC.
of cytK-1 and cytK-2 predicted putative SinR and Fnr binding motifs in PcytK-1. It must, however, be mentioned that an in silico analysis of regulator binding sites is always putative and, finally, needs experimental confirmation.
While cytK is expressed independently of CcpA-mediated catabolite control (van der Voort et al., 2008), the expression of nhe and hbl is regulated by carbon catabolite repression (the cre box of hbl is located downstream of the start codon). Our analysis additionally revealed that nhe and hbl 5 IGRs contain motifs which may be relevant for in vivo binding of CodY (for further analysis and discussion see below).

Other Potential Functions of 5 IGRs
There is increasing evidence that intergenic regions in prokaryotes code for unknown small proteins (Neuhaus et al., 2016). Considering the size of several 100 bp, it is possible that the intergenic sequences upstream of nhe and hbl might encode small proteins or peptides. We could not detect any obvious ORFs upstream of nhe, but the 5 UTR of hbl contains a putative ORF of varying size (180-192 nucleotides, Figure 1D). This ORF appears in all clusters, but not in all investigated strains, and is thus not restricted to particular phylogenetic groups. A BLASTP analysis did not reveal any homology to proteins of known function, thus the expression and function of this ORF remains to be studied. 5 UTRs occasionally contain temperature sensitive RNA thermometers (Narberhaus et al., 2006) or metabolite responsive riboswitches Winkler and Breaker (2005). In Listeria monocytogenes long 5 UTRs were found to regulate virulence gene expression, e.g., the prfA 5 UTR is a thermosensor allowing transcription of the transcriptional activator at 37 • C and blocking it at lower temperatures (Johansson et al., 2002;Loh et al., 2006). The activating and temperature-independent function of listerial actA and hly 5 UTRs was shown (Wong et al., 2004;Shen and Higgins, 2005), but the mechanism of expression enhancement is not clear yet.
Recently, several repeat regions that might encode novel riboswitches have been identified in B. cereus (Kristoffersen et al., 2011), but none of them is located in the 5 UTR of the nhe or hbl operons. Our Rfam analysis of the 5 UTR sequences revealed no similarities with known RNA families and no conserved RNA secondary structures (data not shown). Nonetheless, the long 5 IGRs could encode yet unknown functions or they may interact with different regulators to allow for differential expression of enterotoxin genes.

Entire nhe 5 IGR is Necessary for Maximal Promoter Activity
Our further experimental analysis focused on B. cereus INRA C3 due to the presence of all three main enterotoxins nheABC, hblCDAB, and cytK-2 in this highly toxic strain (Jeßberger et al., 2015). For a detailed analysis of the regulator recognition sites in B. cereus INRA C3 nhe and hbl 5 IGRs see Supplementary Figure S1.
The functionality of the unusually long nhe 5 UTR was investigated using several partial deletions of 5 UTR sequences and luciferase as a reporter (Figure 2A). The promoter activities were assessed in the B. cereus INRA C3 background. The nhe operon was reported to contain two promoters and two PlcR binding sites (Figures 1C and 2A; Agaisse et al., 1999;Lindbäck et al., 2004). The 554 bp full-length sequence showed the highest activity of all Pnhe constructs, while the 406 bp Pnhe-s1 (Pnhe lacking PlcR binding site 1) showed a strongly decreased activity (Figures 2A,B). This strongly indicates the activating role of PlcR binding site 1, as confirmed by an 11-fold decrease of promoter activity ( Figure 2B). Thus, the highly conserved PlcR 1 site is responsible for PlcR-dependent nhe expression, while the less conserved PlcR 2 site contains a 2 bp central insertion ( Figure 1C). Deletion of the region containing the upstreamlocated promoter abolished transcription (construct Pnhe-s2, Figures 2A,B), indicating that this sequence is the active and essential promoter in B. cereus INRA C3. The luminescence  Figure S2A. Luminescence signals were generated by the transcription of lux genes located downstream of the complete or partial 5 IGR tested for promoter activity. Statistically significant differences in transcriptional activity are marked by an asterisk (p < 0.05). intensity of a promoter fusion lacking both PlcR binding sites (Pnhe-s3) was similar to the intensity of the Pnhe-s2 construct, showing that the PlcR binding site 2 does not have an activating function by itself (Figure 2B). A promoter fusion lacking the second putative promoter (Pnhe-5 UTR) resulted in an activity similar to the control plasmid pXen1 without any active promoter. This demonstrates that this fragment of the promoter region also contains one or more essential activating regulatory elements, which might include a ResD, cre, Fnr, PlcR 2 and a putative SinR binding site. In summary, transcription of the nhe operon strongly depends on the presence of the entire IGR, which underpins the necessity of a concerted interaction of all regulatory elements to regulate nhe expression.

Hbl 5 UTR Represses hbl Transcription
To study the function of the hbl 5 UTR lux reporter fusions including partial or complete deletion of the 606 bp 5 UTR were constructed and transferred into B. cereus INRA C3 ( Figure 3A). The full-length construct contains the highly conserved PlcR binding site 1 (Figures 1D and 3A). Transcription of the hbl operon was previously shown to be PlcR-dependent (Agaisse et al., 1999). Interestingly, deletion of the entire 5 UTR (Phbl-5 UTR) led to an increased promoter activity compared to the full-length construct ( Figure 3B). In contrast, deletion of the downstream half of the 5 UTR (Phbl-5 UTR-down) resulted in luminescence levels comparable to the wild type situation. Therefore, the putative ResD, SinR, and Fnr binding sites do not influence promoter activity under our experimental conditions. Deletion of the upstream half of the 5 UTR (Phbl-5 UTR-up) containing putative binding sites for CodY, ResD, SinR, and Fnr led to a stimulation of transcription, which was less pronounced compared to the deletion of the entire 5 UTR. We conclude that the 268 bp region designated 5 UTR-up ( Figure 3A) partially represses hbl transcription in B. cereus INRA C3. It has been shown previously that the deletion of the two-component system yvfTU leads to a reduced plcR, papR, nhe, nprB and plcB transcription but an increased hbl transcription (Brillard et al., 2008). This confirms not just differences in the regulation of nhe and hbl expression despite regulation by the same factors, but also an additional PlcR-independent regulation of hbl transcription.

Nutrient Deficiency Activates Transcription of Tripartite Enterotoxins
Bioluminescent reporter strains containing the wild type 5 IGRs of nhe or hbl (Pnhe and Phbl, Figures 2A and 3A) were used to compare the impact of a minimal medium (MOD) and a rich medium (CGY) on enterotoxin expression (Supplementary Table  S1). While growth of B. cereus INRA C3 was delayed and reduced in the minimal medium, we noticed a steep increase in nhe and hbl promoter activity in comparison to growth in nutrient and amino acid rich media such as CGY (Figures 4A,B). In CGY the maximal promoter activity was reached during stationary phase, while peak activity was detected in the late exponential phase in defined MOD medium.
To further analyze the impact of the availability and accessibility of nitrogen sources, maximal promoter activities were compared in defined MOD medium supplemented with either tryptone or CAAs. Tryptone represents enzymatically digested casein and is a mixture of differently sized oligopeptides (Wang et al., 2013), which are less accessible and need to be degraded by (exo-)proteases, but are thus longer available. In contrast, CAAs are free amino acids obtained by acid hydrolyzation of casein (Nolan, 1971). They are fast and easily taken up and metabolized. In MOD medium supplemented with 1% CAA, nhe and hbl enterotoxin promoter activity may show a trend to be slightly lower than in unsupplemented MOD ( Figure 4C). Free amino acids are also present in MOD, albeit in lower amounts, confirming that the depletion of free amino acids is a determining factor for enhanced enterotoxin promoter activity. When B. cereus INRA C3 was grown in MOD medium supplemented with 1% tryptone, promoter activities of Pnhe and Phbl were threefold and sevenfold lower than in MOD medium, respectively ( Figure 4C). These results point to an activation of enterotoxin promoter activity during unfavorable conditions, such as the absence of easily metabolizable amino acids. BCAAs and GTP activate the nutrient-sensitive repressor CodY (Ratnayake-Lecamwasam et al., 2001;Shivers and Sonenshein, 2004;Brinsmade et al., 2010). Thus, we hypothesized that CodYdependent promoter repression is prolonged in MOD + 1% tryptone, which continuously contains free amino acids from the bacterial degradation of oligopeptides. Detailed growth and transcription kinetics of B. cereus INRA C3 in media with different amino acid availability are shown in Supplementary Figure S3. The quorum sensing virulence regulator system PlcR-PapR is known to be indirectly positively controlled by CodY via enhancing the import of the signaling peptide PapR (Slamti et al., 2015) and the enterotoxin genes are under positive transcriptional control of PlcR Ramarao and Sanchis, 2013). In contrast to previous studies, which found the PlcR regulon to be activated by CodY in an unknown manner in an emetic B. cereus, an enterotoxic B. cereus and a B. thuringiensis strain (Frenzel et al., 2012;Lindbäck et al., 2012;Slamti et al., 2015), our results suggest a repression of enterotoxin gene transcription by CodY. In silico comparison of the CodY binding consensus sequence with the nhe and hbl promoter regions revealed several potential binding sites downstream of the promoters. This could result in premature termination of transcription by a roadblock mechanism, as described for the B. subtilis ybgE gene (Belitsky and Sonenshein, 2011b). Supplementary Figures S4 and S5 display alignments of the nhe and hbl promoter regions in B. cereus, B. thuringiensis and B. cytotoxicus strains. The first and second CodY binding sites in Pnhe are almost identical (Supplementary Figure S4). However, the second binding site in B. cytotoxicus CVUAS 2833 is different and located 20 bp farther downstream, which might explain the varying binding affinity. A potential CodY binding site was located in Pnhe of B. cereus F4810/72 (gray underlined), but no binding of CodY to Pnhe F4810/72 could be detected (Frenzel et al., 2012). This site is not present in Pnhe CVUAS 2833. Pnhe INRA C3 and Pnhe CVUAS 2833 were positive in electrophoretic mobility shift experiments. A possible direct regulation of hbl expression by CodY has not yet been analyzed. The hbl promoter    . (A) CodY consensus sequence of three CodY binding sites in 142 strains (based on 379 sequences: one potential site in 142 Pnhe, one potential site in 140 Pnhe (not present in the two cluster I B. pseudomycoides strains) and one site in 97 Phbl). This consensus sequence is highly similar to the CodY consensus binding sequence in B. subtilis (Belitsky and Sonenshein, 2013). regions and potential CodY binding sites are highly conserved (Supplementary Figure S5).

CodY binds to Phbl and Pnhe In vitro
Since the promoter activity studies indicated a strong activation of enterotoxin transcription after depletion of free amino acids, the affinity of CodY to enterotoxin promoter regions was analyzed in the highly enterotoxinogenic B. cereus INRA C3 and B. cytotoxicus CVUAS 2833 (Guinebretiere et al., 2013;Jeßberger et al., 2015).
Due to its size of over 600 bp Phbl was divided in five fragments each tested in electrophoretic mobility shift assays ( Figure 5A). Comparison with the consensus sequence (Belitsky and Sonenshein, 2013) identified putative binding sites with more than one mismatch to the consensus sequence in all tested sequences ( Figure 5A and Supplementary Figure S5). In vitro DNA affinity tests revealed that CodY shows a low affinity to the fragments Phbl-1, -3, and -4 with an estimated dissociation constant (K D ) of around 700 nM. One potential repressor binding site with only one mismatch to the consensus sequence was found in Phbl-2 and Phbl-5 downstream of the promoter (Figures 1 and 5A). Phbl-2 and Phbl-5 were both bound by CodY ( Figure 5B) with a K D of less than 250 nM. A gel retardation control experiment with the CodY-interacting inhA1 promoter region (Frenzel et al., 2012) also resulted in a K D of around 250 nM. It is thus likely that CodY binds to Phbl at a conserved binding site 84 bp downstream of the TSS (indicated by an ellipse in Figure 5A).
In addition, CodY binds to a stretch of the nhe 5 IGR of B. cereus INRA C3 (568 bp) and of B. cytotoxicus CVUAS 2833 (517 bp) at potential repressor binding positions with a K D of ∼125 and ∼330 nM, respectively (Figure 2A, data not shown). We additionally analyzed CodY affinity to the cytK promoter regions, but could neither detect specific interactions with PcytK-1 nor with PcytK-2 (K D values > 1000 nM, data not shown).

Fine-Tuning of Enterotoxin Expression Through Variation of CodY Binding Sites?
In contrast to B. cereus INRA C3 (our data), the B. cereus F4810/72 nhe 5 IGR is not bound by CodY in vitro (Frenzel et al., 2012). These results indicate that CodY has a different affinity to the three putative binding sites in the 5 IGRs (Supplementary Figure S4).
One site is identical in all three strains, while differences in the second site might cause strain-specific deviation of binding affinity. Both sites are indicated in Figure 1C. DNA fragments containing only one of the two sites were negative in gel retardation experiments (data not shown). The third potential binding site (Frenzel et al., 2012) occurs in B. cereus, but not in B. cytotoxicus leading to the conclusion that it plays a marginal role in CodY-mediated repression of Pnhe activity. Thus, more than one target site might be necessary to allow effective binding of CodY. The electrophoretic mobility shift assays showed that CodY binds in addition to the main binding site with low affinity to several sites within the hbl 5 IGR (Figure 5). These might influence CodY binding in vivo. Our results fit to the previously stated hypothesis that the difference in CodY binding strength is one of the determinant factors of the hierarchical CodY regulon expression (Brinsmade et al., 2014). Hierarchical and DNA sequence-dependent binding of several molecules of a single transcriptional regulator was previously also shown for E. coli (Yoshida et al., 2006) and might similarly occur in CodYbinding to the enterotoxin 5 IGRs.
The two putative CodY binding sites in Pnhe and the confirmed site in Phbl were found in almost all of the 142 investigated B. cereus sensu lato strains ( Figure 6A). They show maximally one nucleotide mismatch to the B. subtilis consensus sequence (Belitsky and Sonenshein, 2013), but a comparison reveals considerable variability with only one completely conserved T at position 14 and an almost conserved A at position 10. Interestingly, variations at almost all positions within the binding motif have previously been shown to modulate the stringency of CodY binding (Belitsky and Sonenshein, 2011a). Our in vitro experiments indicate that a high affinity to the 5 IGR of nhe depends on the sequence of the second CodY binding site (Figure 1 and Supplementary Figure S4). The C-terminal DNAbinding domain of CodY as well as the dimerizing N-terminal cofactor binding domain (Levdikov et al., 2006) is highly conserved in B. cereus group strains (Supplementary Figure S6). The strainspecific differences in CodY affinity and the general lack of CodY binding site conservation support the hypothesis that the binding site sequence, or the combination of different motifs, is crucial for fine-tuning of enterotoxin gene transcription.
In contrast, PlcR binding sites within the enterotoxin promoters reveal a much higher degree of conservation ( Figure 6B). The sequence of the DNA-binding N-terminal domain of PlcR is conserved, while the regulatory C-terminal regions are variable to allow adaptation to changing conditions (Declerck et al., 2007;Böhm et al., 2015). Therefore, PlcRmediated activation of enterotoxin transcription is controlled by protein activity and environmental factors, while CodY-mediated repression additionally may depend on strain-specific 5 UTR sequences.
Being the major virulence regulator in B. cereus, PlcR is responsible for the general activation of enterotoxin transcription under unfavorable conditions . In contrast, fine-tuning of enterotoxin transcription and, thus, a differential toxin gene regulation might be mediated by CodY binding in a strain-specific manner via affinity to variable CodY binding motifs. Transcription of the construct Phbl-5 UTR still showed media-dependency (Supplementary Figure S7), indicating that not only regulators binding within the 5 UTR play a role to finetune transcription. All putative ResD, Fnr, CodY, cre and PlcR 2 sites are also variable, perhaps indicating that oxygen and other nutrient levels may affect enterotoxin transcription differently in each B. cereus strain.

CONCLUSION
We present evidence for a high CodY-mediated nhe and hbl promoter activity under nutrient, especially amino acid limited conditions. While PlcR is the main virulence activator in B. cereus sensu lato, CodY may be used for a strain-specific fine-tuning of enterotoxin transcription via repression in response to specific environmental conditions. The unusually long promoter regions of nhe and hbl might be important for a concomitant interaction of several global regulators. The redox regulators ResD and Fnr, for instance, interact not only directly with their DNA recognition sites, but are also capable of interaction with each other (Esbelin et al., 2009) and formation of a ternary complex with the virulence regulator PlcR (Esbelin et al., 2012).
However, the actual enterotoxin synthesis in B. cereus is rarely consistent with transcriptional activity and is, moreover, highly strain-specific (Ceuppens et al., 2013;Jeßberger et al., 2015). It might be speculated that the 5 UTRs, in addition, interfere with post-transcriptional and/or translational processes (Esbelin et al., 2008) to modify the efficiency of enterotoxin production according to the environmental conditions prevalent in the human intestine.

AUTHOR CONTRIBUTIONS
M-EB performed the experiments. VK helped with the construction of mutants and the lux reporter experiments, and also with the design of the study. NJ was involved in compiling up the strain collection and critically revised the work. EF helped with the EMSA assays and critically revised the paper. SS was involved in the design of the study and wrote the paper together with M-EB. All authors worked on the manuscript. All authors finally approved the version to be published and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

ACKNOWLEDGMENTS
This IGF Project of the FEI (Forschungskreis der Ernährungsindustrie e.V., Bonn) was supported via AiF (Project AiF 17506N) within the program for promoting the Industrial Collective Research (IGF) of the German Ministry of Economics and Energy (BMWi), based on a resolution of the German Parliament.
We thank Scharifa Bornschier, Christine Braig, and Romy Wecko for technical assistance and Felix Behr, Stephan Rambichler, Johanna Reiter, and Astrid Zinz for their experimental support.