A 63-year-old male patient was admitted at a French University Hospital with Crohn’s disease and chronic renal failure. The patient further developed sepsis and a central venous catheter infection. B. cereus isolates were isolated from various blood samples from this patient.

The patient underwent intensive rounds of antibiotic treatment including amoxicillin, ciprofloxacin (21 days), gentamycin (3 days), imipenem (18 days) followed by ciprofloxacin and vancomycin (10 days) (Glasset et al., 2018). Three positive blood cultures yielding B. cereus isolates (12CEB42BAC_S94, 12CEB40BAC_S20, 12CEB43BAC_S95, further named S94, S20 and S95, respectively) were obtained during a span of 87 days. The first sample (S94) was collected before the patient received antibiotic treatments, the second sample (S20) was collected after 23 days of treatment, and the last sample (S95) was collected 64 days after the previous one. The isolated strains were confirmed as being B. cereus by MALDI-TOF and no other bacteria were isolated in the blood cultures. The patient was later released from hospital, presenting no further signs of infection.

The bacterial strains were grown overnight in BHI medium at 37°C, 200 rpm until mid-exponential growth phase and bacteria were pelleted. Total DNA was extracted as previously described (Cadot et al., 2010; Kavanaugh et al., 2021) and quantified with the Qubit^® Fluorimeter. DNA concentrations were adjusted to 30 ng/ml and sequenced by the MiSeq Illumina platform hosted at the Pasteur Institute, giving 2x150 bp paired-end reads.

Sequencing analysis was performed as previously described (Kavanaugh et al., 2021) and included quality control using FastQC and MultiQC, de novo assembly of draft genomes with SPAdes version 3.13.0. Raw WGS data are accessible through https://www.ebi.ac.uk/ena/ (accession number PRJEB46455).

The genome sequences of the 3 strains were aligned against the pAH187_270 plasmid from the strain B. cereus AH187 using NUCmer for sequence alignment and MUMmer/MUMmerplot for visualisation/coverage plots. Further analysis was performed using the Proksee server (https://beta.proksee.ca/) to create circular alignments of reads to the reference plasmid.

MLST was determined for the 3 strains using the online MLST tool available from the Centre for Genomic Epidemiology (Larsen et al., 2012). Genome annotation was carried out using the Prokka automatic annotation tool v [1.13] (Seemann, 2014). Core and accessory genomes were determined by Roary with default settings, with the core genome defined as genes present in at least 3 of 4 samples, including B. cereus AH187 as reference strain.

SNPs were identified using Snippy (Seemann, 2014), which infers polymorphisms at the nucleotide level by aligning the sequencing reads against the reference plasmid pAH187_270.

M13 sequence-based polymerase chain reaction (M13-PCR), derived from a RAPD technique, allows differentiating between various strain patterns. M13 typing was performed as described (Glasset et al., 2016). The DNA profiles were analyzed with BioNumerics 7.1 software (Applied Maths). The software compared the DNA profiles and clustered the strains according to their similarity.

The toxin gene profiles were identified by assessing the presence of the cytK-1, cytK-2, HBLA, HBLC, HBLD, NHEA, NHEB, NHEC, hlyII and ces genes by PCR using specific primers (Glasset et al., 2016). The strains were then clustered into genetic signatures (GS) according to their different combinations of presence/absence patterns (Glasset et al., 2021).

The strains were affiliated to one of the seven known phylogenetic groups according to the partial sequencing of the panC gene (Guinebretière et al., 2008). The production of the enterotoxins NHE and HBL was tested with the immunological tests BCET-RPLA Toxin Detection (Oxoïd) and Tecra (BDE VIA, 3M-Tecra) kits, respectively (Guinebretière et al., 2002).

Plasmid-curing was tempted to determine the influence of the 270 kb pAH187 plasmid on strain characteristics. Plasmid-curing was investigated through culturing the strains at increased temperature (41°C and 43°C), addition of antibiotics (ampicillin – 50µg/ml or novobiocin - 2ug/ml), or ethidium bromide (15, 100, 125 µg/ml) during repeated or prolonged culture between 5-10 days.

Knock-out of the region corresponding to the PAI on the pAH187 plasmid carried by the S94 strain was accomplished by double-cross over region substitution. Briefly, using the available sequencing information of the pAH187 plasmid (NC_011655), 1 kb regions upstream (region 130550 to 131543) and downstream (region 174739 to 175734) of the identified PAI were synthesized by the Genecust company (Boynes, France). A spectinomycin-resistance cassette was obtained from pAT28 (Trieu-Cuot et al., 1990). The two fragments upstream and downstream of the PAI were cloned at each side of the cassette into the pAT113 vector (Trieu-Cuot et al., 1991) by the Genecust company. The constructed plasmid was named as pAT113 Δpai-pAH187 and transformed into chemically competent E. coli ET12567 and then by conjugation into B. cereus S94 strain (Trieu-Cuot et al., 1987; Huys et al., 2004). Briefly, the donor strain and receptor strains were mixed at ratio 14:1 on a sterile membrane filter with a pore size of 0.05 µm (VMWP02500 Milipore) deposited on BHI agar plates for 18 h at 37°C. After mating, bacteria were resuspended from filter with 1.5 mL of colicin solution (Hill and Holland, 1967). The transconjugates were selected on BHI agar plates supplemented with 300 µg/mL of spectinomycin at 37°C. The deletion of the PAI by double recombination event was verified by PCR using primers located upstream and downstream of the cloned region. The corresponding mutant was named S94 Δpai. Due to the size of the entire PAI, it was technically not possible to obtain a complemented strain.

All strains were inoculated into BHI broth and grown at 37°C, 200 rpm. Bacterial growth was followed by measuring the OD at 600 nm at regular intervals.

To determine cellular morphology and size, bacteria were observed on BHI plates. The size of 25 colonies per plate was measured with a graduated scale. Alternatively, bacteria harvested at the end of exponential growth phase were observed under an AxioObserver. ZI Zeiss inverted microscope.

The sporulation efficiency of the strains was determined in HCT plus 0.3% glucose as previously described (Ramarao and Lereclus, 2005). For germination assays, the spores were incubated in BHI medium for 55 min. Samples were taken at 0, 7 min, 16 min, 25 min, 40 min, 55 min. The number of remaining spores was determined as heat-resistant (85°C for 15 min) CFU on BHI plates and normalized to the initial spore value at T0.

Spores were injected at various concentrations between the second and third body segment from the rear of 10 last instar Galleria mellonella larvae as previously described (Buisson et al., 2019). The mortality rate was measured after 24 h of infection at 37°C.

The Minimum Inhibitory Concentrations (MICs) of selected antimicrobial agents were measured by using concentration gradient strips (Etest^®, BioMerieux) (Glasset et al., 2018). The following agents were tested: ampicillin^$, cefotaxime, imipenem^$, vancomycin^$, gentamicin^$, rifampicin^$, tetracycline^$, ciprofloxacin^$, chloramphenicol^$, azithromycin, sulfamethoxazole/trimethoprim^$ and clindamycin^$. Due to scarce availability of interpretative criteria in the literature, clinical breakpoints were used when available (^$) (Wayne and Clinical and Laboratory Standards Institute (CLSI), 2010).

Three B. cereus isolates were isolated from three different blood cultures of the same patient within a period of 87 days and following intensive antibiotic treatments. No other bacteria were isolated from the blood cultures.

The M13 profiles of the 3 isolates were compared and were highly similar for the 3 isolates, suggesting that the 3 isolates are identical or very similar strains ( Figure 1A ).

In order to further compare the isolates, their genomes were sequenced. Mean size of the draft genomes was 5,533,542bp (range 5318760 [S95] - 5643975 [S94]). Mean GC % was 35.3%. In silico MLST analysis determined each strain to belong to the same MLST type 26.

The strains were further characterized for the presence of 10 genes implicated in virulence, the production of Nhe and the phylogenic group. The three isolates possess the nhe gene and were high Nhe producers. The hbl and cytK genes were absent in all isolates. The 3 isolates belong to phylogenetic group III. However, the ces gene was present in the two first isolates (S94 and S20) and absent in the last isolate (S95).

The patient received an intensive antibiotic treatment. To assess the potential impact of this treatment on the antibiotic resistance of the strains, the CMI against major antibiotic were measures for the first S94 and the last S95 isolated strains. The two isolates were resistant to ampicillin and cefotaxime. The strains were sensitive to the antibiotics that were administered to the patient. Strikingly, the 94 strain was susceptible to rifampicin whereas S95 strain displayed resistance to rifampicin ( Table 1 ).

Mash analysis of the 3 strains identified the closest strain to S94 and S20 to be AH187 with an average nucleotide identity of 99.9%. The AH187 strain possesses the pAH_187- 270 kb mega-plasmid. Analysis of the core and accessory genomes reveals that nearly all variability among the 3 patient strains results from the accessory genes located on this mega-plasmid ( Figure 1B ).

To further assess the differences among the patient-isolated strains, snps were examined via Snippy using the plasmid, pAH187_270, as reference. A significant amount of variation is evident when comparing the S95 patient strain against the mega-plasmid, pAH187_270 ( Table 2 ).

Consistently, the pAH187_270 plasmid sequence resulted in nearly 100% coverage for S20 and S94 strains. By contrast, the S95 strain did not demonstrate significant alignment to the pAH187_270 plasmid ( Figure 1C ).

Altogether, the 3 isolates have an almost identical chromosome but differed by the absence of the pAH187_270 plasmid in the S95 isolate compared to the S94 and S20 isolates.

A closer look at the pAH187_270 plasmid revealed that some of the biomarkers previously identified as characteristic of clinical strains (Kavanaugh et al., 2021) were present on this plasmid (i.e., araC, gshAB, BCQ_PI180) ( Figure 1C ).

We confirmed by WGS that S20 and S94 carried these biomarkers, whereas the S95 strain did not. In addition, S94 and S20 carried the ces gene located in the pAH187_270 plasmid whereas S95 did not (not shown).

Based on the NUCmer alignment results, the localisation of the plasmid-based biomarkers of the S94 strain was determined. These biomarkers were located within a 43 kb span of the pAH187_270 plasmid, being distant from the CES operon for cereulide production ( Figure 1C ). We named this region the pathogenicity island (PAI). At first glance, it was observed that the PAI is flanked by resolvases and transposons indicating that they may have been obtained during horizontal gene transfer ( Figure 1D ). Further investigation of the plasmid pAH187 and the PAI revealed typical features used in defining PAIs. First, there is an abundance of direct repeats found in the plasmid. When a higher upper limit is set for maximum size (10,000 bp), 57 repeats of varying size, with 10 repeats of 396 bp clustered in the 175,466 to 189,105 region. However, with a lower limit of 200 bp, 20 hits are detected, with repeats detected on either side of the PAI at approx. 135,000 bp and 185,000 bp. Interestingly, pAH187 possesses a large inverted repeat 2,654 bp long, located 127,621-130,299 and 163,537-160,858, (99% match) found within the PAI. Furthermore, within the PAI, there is an increased presence of resolvases and transposases, which diminish in prevalence immediately outside the PAI. These are believed to play a role in the mobility of the DNA elements and facilitate their transfer and integration. Terminating one end of the PAI, a region, which spans from 175,000bp to 190,000 bp contains numerous tandem repeats, and finally terminates with a Holliday junction resolvase, further confirming the presence of the PAI.

Lastly, PAIs are often accompanied by changes in GC content in the inserted region, and this may be attributed to the DNA arriving from a different species through horizontal gene transfer. The GC content of the PAI was above average, compared to the immediately surrounding regions. It is to note that B. cereus is a low-GC-content bacteria and an increase in the GC content may thus indicate an integration of the PAI from a different species. However, as the difference in GC content is not a stark change, it may also represent an insertion or even a fusion event, as B. cereus megaplasmids (>100kb) have previously been hypothesized to result from fusion of smaller plasmids given the presence of multiple minireplicons (Zheng et al., 2013).

Several genes are present in the PAI: araC, BCAH187_C0208, BCAH187_C0212 being found at one end, BCN_P218 and BCAH187_C0212_2 towards the middle, and BCQ_PI180, BCQ_PI181, gshAB, 3 germination proteins encoded (including Ger(x)C), and a single protein (DUF11 domain-containing protein) being found on the opposing end of the pathogenicity island ( Table 3 ).

The first and last isolates, S94 and S95, were evaluated in various phenotypical and functional assays. The two strains have similar growth capacity over time although the S95 strain was slightly impaired ( Figure 2A ). The two strains show no apparent difference in morphology. However, the colony morphology on plate is different with S95 colonies being wider ( Figure 2B ). In average, the size of the colonies was 3 times higher for the S95 strain compared to the S94 strain. In addition, S95 strain showed a drastic diminution in its germination rate compared to the S94 strain ( Figure 2C ). Finally, the S95 strain was severely impaired in its virulence potential compared to the S94 strain in the insect model of infection ( Figure 2D ). The lethal doses 50 (LD50) were 1,16.10⁶ CFU/mL for the S94 strain and 5,29.10⁸ CFU/mL for the S95 strain, thus displaying a 456-fold difference.

Taken together, these data strongly suggest that the plasmid pAH187_270 carried by the S94 strain and lost by the S95 strain plays an essential role during B. cereus virulence.

To confirm that the pAH187_270 plasmid has an influence on the strain pathogenicity, we first tried to cure the S94 strain from this plasmid. However, all our attempts to cure the plasmid were unsuccessful. Thus, we focussed on the PAI region containing most of the markers previously identified as characteristic of clinical strains. To assess whether the genes located in the PAI may explain the difference in the strain pathogenicity, the S94 strain was deleted for its PAI and the Δpai-mutated strain was compared with the wild type strain in various phenotypical and functional assays.

The strains were observed under the microscope and bacterial morphology showed that the two strains were almost similar in cellular shape and size although the Δpai mutant was slightly impaired in its growth capacity ( Figures 2A, B ).

As at least two germination genes were identified within the PAI ( Figure 1 ), the capacity of the wt and mutant strain to sporulate and germinate was assessed. No difference in the sporulation efficiency was observed (not shown). However, the Δpai mutant showed a drastic diminution in its germination rate ( Figure 2C ). The same difference was observed for the S95 and the Δpai mutant compared to the S94 strain, strongly suggesting that the PAI is responsible for the germination efficiency of the initial S94 strain.

Finally, to assess whether the genes located in the PAI may explain the difference in the strain pathogenicity, the S94 and the Δpai mutated strains were evaluated in the Galleria infection model ( Figure 2D ). Strikingly, the Δpai mutant was severely impaired in its virulence potential compared to the wild type strain. The lethal doses (LD) was 3.80E+07 CFU/mL for the Δpai mutant strain, thus displaying a 34.5 fold difference, indicating that the PAI plays an important role in B. cereus virulence capacity.

The S95 strain showed morphological difference to the Δpai mutant and was even more severely impaired in its virulence capacity. This implies that the genes located within the PAI play an important role during B. cereus virulence, but that other genes located elsewhere in the plasmid are also required.