Bacillus cereus ATCC 14579 wild type (wt), its flhF (GeneBank ID: Bc1670) mutant derivative (ΔflhF, MP06), and complemented strain (ΔflhF-comp, MP07) (Salvetti et al., 2007) stocks from -80° C were streaked on brain heart infusion supplemented with 0.1% (w/v) glucose (BHIG; Becton Dickinson, Cockeysville, NJ, USA) plates and incubated at 37° C. BHIG was also used for liquid cultures. When required, 5 μg/ml erythromycin for strain MP06 or 30 μg/ml kanamycin for strain MP07 were added to the media. MP07 cultures were also supplemented with 4 mM isopropil-β-D-1-tiogalattopiranoside (IPTG; Sigma–Aldrich, St. Louis, MO, USA) in order to induce pspac dependent gene expression.

BLAST¹ was used for comparative analysis of nucleotide and protein sequences. Protein sequences in the FASTA format were retrieved from the UniProt database² (The UniProt Consortium, 2015). Functional domain analysis was performed using the ProDom Server³ (Bru et al., 2005). The presumptive secondary and tridimensional structure of proteins were generated using the Phyre2 web portal for protein modeling, prediction and analysis⁴ (Kelley et al., 2015) and the Raptor X Structure Prediction Server⁵ (Källberg et al., 2012), respectively.

For each experiment, swarm plates (TrA plates; 1% tryptone, 0.5% NaCl, 0.7% granulated agar) were prepared fresh daily and allowed to sit at room temperature overnight before use (Salvetti et al., 2011). Swarming was initiated by spotting 50 μl of a culture containing approximately 2×10⁴ cells/ml onto the center of TrA plates, and incubating cultures at 37° C. Swarming migration was evaluated by measuring colony diameters after 8 h. Since B. cereus flagella are very fragile, bacterial samples were taken by slide overlay of single agar blocks (5 mm ×5 mm) that contained different colony portions. Bacterial cells were stained with tannic acid and silver nitrate (Harshey and Matsuyama, 1994) for microscopy. Several samples were analyzed at 1000× magnification using an optical microscope (BH-2; Olympus, Tokyo, Japan). All experiments were performed in duplicate in three separate days.

Protein samples were prepared by growing bacterial cells to the late exponential growth phase in BHIG at 200 rpm for 6 h at 37° C. Culture supernatants were collected by high-speed centrifugation (10000 × g), added with the serine protease inhibitor phenylmethylsulfonyl fluoride 1 mM (PMSF; Sigma–Aldrich), and stored at -80° C until use. Protein concentration in supernatants was determined by the bicinchoninic acid (BCA) assay (Smith et al., 1985), using bovine serum albumin as a standard. The activity of fructose-1,6-bisphosphate aldolase in culture supernatants was spectrophotometrically measured at 30° C by following the rate of NADH oxidation at 340 nm according to the method described by Warth (1980).

Proteins contained in supernatants were precipitated using the TCA method (trichloroacetic acid, Sigma–Aldrich). Precipitated proteins were washed eight-times with cold 96% (v/v) ethanol, air-dried and suspended in sample rehydration buffer (7 M urea, 2 M thiourea, 4% CHAPS, 2.4% aminosulfobetaine-14; Invitrogen, Carlsbad, CA, USA). ZOOM^® IPG strips with a linear pH range of 4–7 (Invitrogen) were rehydrated for 16 h in 400 μl of sample rehydration buffer and 10 μg of protein samples were subsequently added by cup loading. Focusing was carried out at 200 V for 20 min, 450 V for 15 min, 750 V for 15 min, and 2000 V for 30 min using the Xcell SureLock^TM Mini-Cell system (Invitrogen). The IPG strips were then equilibrated for 10 min in 5 ml of equilibration solution (LDS 1X; DTT 0.5 mM). For the second dimension, samples were run at 200 V for 35 min on 4–12% gradient SDS-PAGE gel (Bis-Tris ZOOM^TM Gel, 1.0 mm IPG-well; Invitrogen). Three independent biological experiments were performed in separate days. Gels were silver or Coomassie Blue (Sigma–Aldrich) stained, according to standard procedures. Gels were scanned at 300 dpi and 8 bits depth on an Image Scanner equipped with a film-scanning unit and analyzed with the Image-Master 2D Platinum v.6 software (GE Healthcare, Little Chalfont, UK). The spots were quantified after normalization and spot volumes (pixel intensity × area) were expressed as percentage of the total volume of the spots on the gel. Presumptive analysis of protein gels was performed by comparison of silver stained 2-DE gels with gels available in the literature (Gohar et al., 2002, 2005) and then protein spots were identified by spot excision from Coomassie-Blue stained gels and subsequent identification. To this purpose, excided spots were digested with 0.1–0.5 μg of trypsin at 37° C for 6 h (Shevchenko et al., 1996). Digested proteins were analyzed by liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) at Université de Genève, Proteomics Core Facility (Genève, CH). Database searches were performed with the Mascot Server (Matrix Science Ltd; version 04_2014⁶) and results were analyzed and validated using the Scaffold software (Proteome Software Inc.). Searches were performed using trypsin for the fragments cleavage, allowing up to one trypsin miscleavage and without restriction on protein mass or isoelectric point (pI). Fixed modifications included carbamidomethylation of cysteine, while variable modifications comprised oxidation of methionine. Peptide masses ranged from 849.0 m/z to 4001.0 m/z with a charge (z) of 1 +. The mass tolerance was set to ± 25 ppm. Significance was established according to expectancy (e) value transformed into MS score (with significance at a P-value <0.05 at scores over 70), percentage coverage and theoretical pI and molecular weight (Mw) compared to the approximate experimental values observed on 2-DE gels.

Identified proteins were classified based on their biological functions using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database resource⁷. Protein sequences were analyzed using the SIGNALIP 4.1 Server⁸, TATP 1.0⁹, SecretomeP 2.0 Server¹⁰ and TMpred program¹¹ in order to evaluate the presence of Sec-type signals, Tat-type signals, non-classically secretion signals or transmembrane domains, respectively.

Galleria mellonella larvae, obtained from Mous Live Bait (Balk, Netherland), were selected by weight (0.2–0.3 g) and absence of dark spots on the cuticle. Bacteria were grown to the late exponential growth phase in BHIG at 37° C for 6 h and harvested by centrifugation at 5000 × g for 10 min. Bacterial suspensions were prepared in phosphate buffered saline (PBS: 1 M KH₂PO₄, 1 M K₂HPO₄, 5 M NaCl, pH 7.2) and bacteria counted using a hemocytometer. Three infectious doses (about 10³, 10², and 10¹ CFU per larva) were used to infect three groups of 20 larvae. Larva abdomen was accurately cleaned with 70% ethanol and 10 μl of bacterial suspension was injected into the hemocoel through the last right pro-leg, with a sterile Hamilton syringe (Sigma–Aldrich) via a 26-gage needle. A control group of larvae was injected with PBS only. Infectious doses were checked by CFU count after plating appropriate dilutions and incubating 16 h at 37°C. Infected larvae were kept at 37°C and mortality was recorded after 24 h. Each experiment was performed three times in separate days.

To assess bacterial ability to multiply in vivo, groups of 20 animals were infected with 10⁴ CFU/larva of B. cereus wt, ΔflhF, or ΔflhF-comp. Groups of three insects were collected at different times post infection (2, 4, 6, 8, and 24 h) and their surfaces were disinfected with 70% ethanol. Larvae were homogenized in 2 ml of PBS using a Stomacher 400 Circulator (Seward, Worthing, UK). Serial dilutions of the homogenates were plated on LB agar and colonies were counted after incubation at 37°C for 24 h. Non-infected larvae were used as negative control. Three independent biological replicates were performed.

Data were expressed as the mean ± SD. A P-value <0.05 was considered significant. For 2-DE experiments, the One-way Anova analysis and the two-tailored Student’s t-test for equal or unequal variance were applied. For G. mellonella experiments, the 50% lethal dose (LD₅₀) values were estimated by probit analysis (Finney, 1971). Differences in mortality rates obtained for each infectious dose and in the CFU number for each time tested were evaluated by the One-way Anova analysis.

By BLAST alignments, the flhF nucleotide sequence resulted to be highly conserved (from 90 to 100%; Score ≥200) among different B. cereus strains. In a previous work (Salvetti et al., 2007), we showed that B. cereus FlhF possesses a C-terminal G domain, that is strongly conserved among SRP-GTPases (Zanen et al., 2004; Bange et al., 2007; Salvetti et al., 2007; Balaban et al., 2009; Green et al., 2009; Schniederberend et al., 2013; Schuhmacher et al., 2015), and a less conserved N-terminal B domain. Herein, revisited analysis with the ProDom program reveals a nucleoside-triphosphatase (amino acids 176–219), an SRP receptor (amino acids 220–302), an SRP54-type protein GTPase (amino acids 303–353), and an SRP-dependent cotranslational protein-membrane targeting (amino acids 354–439) subdomain inside the NG domain of B. cereus FlhF. Unlike the FlhF proteins of Vibrio cholerae. P. aeruginosa, and B. subtilis in which the N-terminus comprises a GTPase activity subdomain, the N-terminus (B domain) of B. cereus FlhF is unique and functionally unknown.

We used the Phyre2 program (Kelley et al., 2015) and the Raptor X Structure Prediction Server (Källberg et al., 2012) in order to define the putative secondary and three-dimensional structure of B. cereus FlhF. The protein appeared to be constituted by 45% α-helices and 30% β-sheets and to consist of two distinct portions connected by an unstructured peptide linker (Figure 1A). The B domain was predicted by the Raptor X Structure Prediction Server only, with a P-value of 2.5 × 10^-2, while the NG domain was predicted by both programs with a good P-value (7.43 × 10^-7 for Raptor X Structure Prediction Server and 100.0% confidence by the single highest scoring template with Phyre2) and it is 100% similar in structure to B. subtilis FlhF.

In all described SRP-GTPases, five signature elements (G1–G5) interacting with the GDP/GTP-Mg²⁺ complex have been identified in the G domain (Anand et al., 2006; Bange et al., 2007). These elements have never been described in B. cereus SRPs. As shown in Figure 1B, the alignment between the FlhF G1–G5 signatures of B. subtilis, P. aeruginosa, and V. cholerae with the amino acid sequence of B. cereus FlhF, FtsY, and Ffh revealed that all these signatures are also present inside the G domain of B. cereus SRP-GTPases. In particular, G3 and G4 signatures are identical among FlhF, Ffh, and FtsY of B. cereus and G1 is highly conserved in all the analyzed bacterial species. The G2 signature of B. cereus SRP-GTPases are less conserved compared to the same sequence of B. subtilis. P. aeruginosa, and V. cholerae, while G5 is generally less conserved among all the species analyzed.

Ffh and FtsY of many organisms contain additional sequence motifs (ALLEADV in the N domain; GQ and DARGG in the G domain) that are essential for the communication between functional domains (Bange et al., 2007). Since these motifs have never been described in B. cereus SRPs, we first identified them in Ffh and FtsY of B. cereus, through amino acid alignments with the Escherichia coli and B. subtilis orthologs (data not shown). Figure 1C shows the alignment between the ALLEADV, GQ and DARGG motifs in B. cereus SRPs and the putative corresponding regions of FlhF in B. subtilis, P. aeruginosa, and V. cholerae.

Previous data indicated absence of swarm cells in the ΔflhF mutant of B. cereus grown on swarming-supporting media (Salvetti et al., 2007). In the same experimental conditions, we show that FlhF depletion also impairs cell migration. Indeed, as shown in Figure 2A, the ΔflhF mutant gave rise to colonies (4.2 ± 0.32 mm of diameter) that were significantly smaller than those produced by the swarming proficient wt (18.4 ± 1.28 mm of diameter; P = 0.0039) or ΔflhF-comp strain (14.2 ± 1.8 mm of diameter; P = 0.015). As expected, long and hyperflagellated swarmer cells were visualized in the colonies produced by the wt and the ΔflhF-comp strain only (Figure 2B).

To better characterize the effect of FlhF depletion on protein secretion by B. cereus, we performed comparative proteome analyses of the ΔflhF mutant, the wt, and the ΔflhF-comp strain grown in BHIG. There were no obvious growth differences between these strains in this medium, as already described (Salvetti et al., 2007). Additionally, as previously described (Salvetti et al., 2007), we again evaluated the activity of the cytosolic marker fructose-1,6-bisphosphate aldolase, a cytoplasmic enzyme that is not released by intact cells, in culture supernatants. No enzyme activity was detected. The bacterial supernatants were precipitated with TCA and analyzed using 2-DE. On average, 539 ± 103 spots were detected in the supernatant of the parental strain and 531 ± 99 spots in that derived from the ΔflhF-comp strain. This was not significantly different from the 430 ± 119 spots found in the mutant strain on average (P > 0.05). In the 2-DE gel analysis, 19 protein spots showed a different abundance of >1.5-fold (≥1.5 for upregulated proteins and ≤0.667 for downregulated proteins) in the ΔflhF mutant, compared to the wt (Table 1). These spots were identified by comparison with available gels (Gohar et al., 2002, 2005) and LC-MS/MS analysis. Three proteins were represented as multiple spots in the gel, having either similar mass but different isoelectric points or slight differences in the molecular mass due to putative post-translational modifications. No appreciable difference emerged between the wt and the ΔflhF-comp extracellular proteomes.

In the ΔflhF mutant, the amount of six proteins increased and 12 proteins decreased in abundance compared to the parental strain (Table 1). The extracellular protein products significantly more abundant in the ΔflhF mutant included three predicted secretory proteins, cereolysin O (Clo; P = 0.007), the B component of NHE (NHE_B; P = 0.041), both contributing to B. cereus pathogenicity (Lund and Granum, 1997; Alouf, 2000), and a protein with unknown function (P = 0.017). Three other proteins that do not exhibit a conventional signal peptide, the SodA superoxide-dismutase (P = 0.048), enolase (Eno; P = 0.007), and the chaperone GroEL (P = 0.01), were also present at higher levels in the supernatant of the ΔflhF mutant. Despite these proteins are cytosolic in nature, their orthologs are commonly detected in an extracellular form in many bacteria, B. cereus and other Bacillus species included (Nouwens et al., 2002; Braunstein et al., 2003; Lenz et al., 2003; Gohar et al., 2005).

Six predicted secretory proteins playing a role in B. cereus pathogenicity, since they are able to degrade phospholipids (PC-PLC; P = 0.017. PI-PLC; P = 0.0034. Smase; P = 0.0072), proteins (bacillolysin, NprP2; P = 0.032 and P = 0.007. Collagenase, ColA; P = 0.0002), and to form pores in planar lipid bilayers (cytotoxin K; P = 0.034), were significantly less abundant in the extracellular proteome of the ΔflhF mutant. The membrane-anchored substrate-binding protein OppA_1 of the oligopeptide permease transporter (Slamti et al., 2015) was also significantly reduced (P = 0.01) in the mutant secretome. In agreement with our previous results (Salvetti et al., 2007), the flagellin proteins BC1657 (P = 0.0091) and BC1658 (P = 0.016), classified as non-conventionally secreted proteins by the SecretomeP 2.0 server, were found significantly less abundant in the ΔflhF strain. In addition, a significant reduction of the L₂ component of HBL (HBL-L₂; P = 0.023), apparently lacking a signal peptide by the SIGNALIP 4.1 but herein classified as secretory accordingly to Fagerlund et al. (2010), was demonstrated for the mutant strain. The mutant also displayed a reduction in the spots corresponding to aconitate hydratase (BC3616; P = 0.0048) and elongation factor-Tu (EF-Tu; P = 0.002), commonly found in the cytoplasm.

Based on the observation that several virulence proteins were less abundant in the supernatant of the ΔflhF mutant and that this strain was unable to swarm, we wondered whether the deficiency in FlhF would affect B. cereus pathogenicity in an in vivo model of infection. To this aim, larvae of the greater wax moth G. mellonella, which have been shown to provide useful insights into the pathogenesis of a wide range of microbial infections (Jander et al., 2000; Purves et al., 2010), were used to evaluate the pathogenicity of the wt, ΔflhF, and ΔflhF-comp strains. Larvae were intra-hemocoelically injected with three doses of mid-log phase bacteria and the number of dead larvae was recorded. As shown in Figure 3, a significant reduction in larvae mortality compared to the wt was observed when the ΔflhF mutant was inoculated at the three doses (P < 0.01 for the highest dose, and P < 0.05 for the other doses). No significant difference was observed in the mortality rates after injection of the wt or the ΔflhF-comp strain at different doses. As resulted from Probit analysis, the LD₅₀ of the ΔflhF mutant was higher (3.16 ± 0.83 × 10³ CFU per larva) than the value obtained for the wt (9.10 ± 6.73 CFU per larva). The LD₅₀ of the ΔflhF-comp strain was 6.27 ± 4.63 × 10¹ CFU per larva.

To test whether differences in larvae mortality could be attributed to a defect in bacterial growth in vivo, the bacterial load in infected larvae was quantified over time. Larvae were infected with 10⁴ CFU of wt, ΔflhF, or ΔflhF-comp. At selected time points, the number of CFU per larva was determined. Infection of G. mellonella with all B. cereus strains resulted in an initial 10-fold reduction compared to the infectious dose in the CFU recovered from infected larvae at 2-h post infection (wt: 1755 ± 214.6 CFU/larva; ΔflhF: 1673 ± 350.4 CFU/larva; ΔflhF-comp: 1780 ± 280.2 CFU/larva). Bacterial recovery progressively increased over time up to 10000-fold the infectious dose at 24- h post infection (wt; 3.45 ± 0.102 × 10⁸ CFU/larva; ΔflhF: 3.48 ± 0.856 × 10⁸ CFU/larva; ΔflhF-comp: 3.46 ± 0.103 × 10⁸ CFU/larva). No significant difference in the CFU recovery of different strains was observed at each time point considered (data not shown). B. cereus was never recovered from non-infected larvae. Our data demonstrate that B. cereus wt, ΔflhF, and ΔflhF-comp are able to replicate at the same rate in G. mellonella.