Poultry and beef meat as potential seedbeds for antimicrobial resistant enterotoxigenic Bacillus species: a materializing epidemiological and potential severe health hazard

Although Bacillus cereus is of particular concern in food safety and public health, the role of other Bacillus species was overlooked. Therefore, we investigated the presence of eight enterotoxigenic genes, a hemolytic gene and phenotypic antibiotic resistance profiles of Bacillus species in retail meat samples. From 255 samples, 124 Bacillus isolates were recovered, 27 belonged to B. cereus and 97 were non-B. cereus species. Interestingly, the non-B. cereus isolates carried the virulence genes and exhibited phenotypic virulence characteristics as the B. cereus. However, correlation matrix analysis revealed the B. cereus group positively correlates with the presence of the genes hblA, hblC, and plc, and the detection of hemolysis (p < 0.05), while the other Bacillus sp. groups are negatively correlated. Tests for antimicrobial resistance against ten antibiotics revealed extensive drug and multi-drug resistant isolates. Statistical analyses didn’t support a correlation of antibiotic resistance to tested virulence factors suggesting independence of these phenotypic markers and virulence genes. Of special interest was the isolation of Paenibacillus alvei and Geobacillus stearothermophilus from the imported meat samples being the first recorded. The isolation of non-B. cereus species carrying enterotoxigenic genes in meat within Egypt, suggests their impact on food safety and public health and should therefore not be minimised, posing an area that requires further research.

The presence of environmental isolates of Bacillus spp. harboring one or more enterotoxin gene holds crucial importance for the food safety, however, evaluation of toxin gene presence and toxin activity in Bacillus spp. other than B. cereus has not been thoroughly investigated. The diarrheal enterotoxins, Cytotoxin K (encoded by the cytK gene) and enterotoxin FM (encoded by the entFM genes), hemolysin BL (encoded by the hbl operon) and non-hemolytic enterotoxin (encoded by the nhe genes) are regarded as the main enterotoxins involved in foodborne illness studied in B. cereus 7,8 . There is a lack of information on the presence of these enterotoxins in the non-B. cereus groups.
As Egypt depends on imported beef meat, which reached 200,200 metric tons in the year 2016 9 , to complement the animal protein demands, it is crucial to assess the safety of both imported meat and local products. There is increasing concern about imported meat and question whether Egypt's food safety system can protect them from tainted foreign products. Compared to the U.S.A., EU or Australia, the food safety standards in Egypt and other underdeveloped countries are not as high. The aim of this study was to determine if the presence of enterotoxins and other putative virulence factors are contributing to the total cytotoxicity of the enterotoxic Bacillus spp. isolated from retail meat in the Egyptian market. Specifically, we have determined incidence and level of contamination with B. cereus and non-B. cereus groups in retail chicken and local and imported beef. We performed a comparative overview on the observed hemolysin BL production, cytotoxic activity, swarming, ability to form biofilms, phenotypic antibiotic resistance profile and enterotoxin gene profile (NHE, HBL and CytK) 10-12 . There is an intricate link between metabolism, swarming, biofilm production, antibiotic resistance and virulence. Two soluble virulence factors, lecithinase and amylase production, were noted as phenotypic virulence markers 13,14 . Swarming motility was tested because it is generally considered a precursor step to biofilm formation and a major factor in pathogenesis by many human pathogens 15 . The presence of swarming generally increases virulence as movement over surfaces enable bacteria to migrate from sites of infection, offers protection from macrophages as swarming cells were shown to have enhanced resistance to engulfment, and become resistant to a broad range of antibiotics during swarming [16][17][18] . Biofilms are considered to promote adhesion and protect cells from antimicrobials and other external insults 19 . Amyloid-like fibrils have been shown to increase the biofilms' structural integrity and allow staining with certain dyes such as Congo red [20][21][22][23][24] . However, results depending on phenotypic markers alone lack reliability in determining the virulence of Bacillus strains.
In this study, we carried out a m-PCR to detect the genes for three pore-forming enterotoxins, responsible for the diarrheal type of food poisoning: hemolysin BL (HBL), non-hemolytic enterotoxin (NHE), cytotoxin K (CytK) and enterotoxin FM (EntFM), and the hemolytic gene encoding phospholipase (plc) 1,10-12 in B. cereus (n = 24) and non-B. cereus isolates (n = 97) collected from retail chicken local and imported beef meat. In addition, we calculated the level of antibiotic resistance and the multiple-antibiotic resistance indices of isolates present in meat samples.
The other virulence factors as indicated in Table 1, reveals that 92/124 (74.2%) of the total Bacillus species isolated were cytotoxic to the Vero cells. In addition, β-hemolysis activity was observed on sheep blood agar in 77.4% of isolates, α-hemolysis activity in 16.1%, and the γ-hemolysis activity in 6.5% of the isolates.

Distribution of virulence genes among the B. cereus and non-B.cereus isolates. Eight virulence
genes encoding enterotoxins were targeted in the Bacillus isolates based on PCR detection (Table S1). PCR detection included genes encoding hemolytic (hblA, hblC, and hblD) and non-hemolytic (nheA, nheB, and nheC) enterotoxin complexes, cytotoxin K (cytK), enterotoxin FM (entFM) and one hemolytic gene encoding phospholipase (plc) (Fig. 3). At least one of the nine genes were detected in all the isolates except one B. sphaericus isolated from the frozen imported beef meat. The isolated Bacillus commonly possessed cytk (58.9%) as the most prevalent toxin gene followed by hblA (45.2%), plc (40.3%) and entFM (35.5%) genes while the nheB gene was the least present (12.9%).
The non-hemolytic genes nheA and nheC, and the hemolytic enterotoxin HBL complex gene hblC were significantly more prevalent in the B. cereus group isolates compared to the B. subtilis group, and the other Bacillus spp. isolates (p < 0.05). There was no significant difference between the number of isolates carrying the hblA hemolytic enterotoxin gene in the B. cereus group compared to B. subtilis group (p = 0.055). The other Bacillus spp. isolates showed a significantly lower abundance of the hblA gene compared to the B. cereus and B. subtilis groups (p < 0.05). There was no significant difference in the presence of the nheB gene between the groups (p > 0.13). In regard to the source of isolated Bacillus spp., there were significantly fewer incidences of the enterotoxin FM genes (entFM) within the isolates from raw chicken compared to fresh local or imported frozen meat (p < 0.05).
There were no significant difference in slime (CR) and biofilm (CV) production of the isolates in the different Bacillus spp. groups (p > 0.05). More isolates from the B. cereus group had a gene encoding phospholipase (plc), lecithinase activity, amylase activity, and significantly fewer isolates showing swarming ( Fig. 3; p < 0.05).
As for the differences between the Bacillus groups, prevalence of antibiotic resistance in P. alvei and G. stearothermophilus are notably lower compared to other species. Resistance prevalence in P. alvei, G. stearothermophilus did not significantly differ compared with B. cereus, B. thuringiensis, B. mycoides, B. licheniformis, and B. coagulans, although this observation may be due to the low number of these species isolated. Resistance to the antibiotics oxacillin, cephalothin, and the antibiotic combination sulfamethazole/trimethoprim was significant more prevalent within isolates of the Bacillus cereus group compared to the B. subtilis group and other Bacillus spp. isolates (Fig. 4, p < 0.05). There was no significant difference in the percentage of isolates with antibiotic resistance to vancomycin, nalidixic acid, chloramphenicol, tetracycline, and erythromycin between the Bacillus

Abundance of Antibiotic resistant Bacillus isolates from different meat sources. Bacillus spp.
antimicrobial susceptibility results were stratified by isolate, source (raw chicken, local beef meat and imported beef meat) and Bacillus spp. group (Table S3).
The different source of meat showed few statistically significant variations in abundance of isolates resistant to the 10 antibiotics tested (Fig. 4). The Bacillus spp. isolates from raw retail chicken showed significantly fewer incidences of resistance to the antibiotic cephalothin (p < 0.05) compared to the isolates from local beef and imported beef. Isolated Bacillus spp. from local beef meat showed significantly more isolates resistant to oxacillin (p < 0.05) compared to isolates from raw chicken and no significant difference compared to imported beef (p > 0.15). The Bacillus spp. isolated from imported beef showed significantly fewer isolates with nalidixic acid resistance compared to isolates from raw chicken and local beef meat (p < 0.01). Isolates from local beef meat showed significantly more isolates with kanamycin resistance compared to isolates from raw chicken and imported beef meat (p < 0.02). There were significantly more isolates (p = 0.04) from imported beef with resistance to the drug combination sulfamethazole/trimethoprim compared to isolates from local beef, but there was no significant difference compared with raw chicken (p = 0.07).
Bacillus spp. resistance profile. Multiple drug resistance (MDR) for Bacillus spp. was defined as a strain non-susceptible to ≥1 antibiotic in ≥3 antibiotic classes, extensive drug-resistance (XDR) was defined as a Bacillus spp. strain non-susceptible to ≥1 antibiotic in all but ≤2 antibiotic classes and pan drug resistance (PDR) for those Bacillus spp. resistant to representatives of the 8 classes of antibiotics tested. (The antibiotics and their classes are listed in Table S2). In general, the strains were resistant to 1-7 antibiotics represented in 1-5 classes. Overall, the least number of classes was one represented by penicillin and the highest number of classes was five. Bacillus subtilis group licheniformis (n = 9)

Source of samples
Other Bacillus spp

Frozen beef meat (n = 28)
Bacillus cereus group Bacillus subtilis group licheniformis (n = 5) Other Bacillus spp coagulans (n = 5) The total number of MDR isolates were 85 (68.5%) and no XDR nor PDR were detected (Table S3). The MDR profiles of the Bacillus spp. collected from chicken and beef (local and imported) meat samples, resistant to the ten tested antibiotics are summarized in Table S3.  Correlation analysis of measured variables: virulence, antibiotic resistance and meat source.
Permutational MANOVA analysis using the isolated Bacillus spp. groups, isolate meat source, and the interaction term as terms within the model indicated significant difference in all three terms (p = 0.001, 0.019, and 0.002, respectively). Pairwise comparisons showed no significant difference between sources of isolates (p > 0.081) indicating little variation between meat sources. Comparisons between groups showed significant difference between all three groups (p ≤ 0.023) suggesting variation between isolates related to phenotypic and biochemical testing were related to the Bacillus group. Associations of variables were identified within the hierarchical clustering of the heatmap (Fig. 5), PCA analysis (Fig. 6), and correlation matrix (Fig. 7). The correlation matrix indicated a stronger association of antibiotic resistance with one another than other variables such as hemolysis genes and biochemical tests. The lack of  Binary factors (such as antibiotics or genes) indicating presence as green (relative response 1) or absence as red (relative response 0). Factors containing greater ranges were adjusted to range from 0 to 1 as indicated by color key.
correlations between antibiotic resistance and other phenotypic traits or virulent genes suggest independence of antibiotic resistance to virulence.
The PCA analysis (Fig. 6) also revealed a low association of antibiotic resistance to virulence factors. The variables were also unable to separate the isolates of the Bacillus spp. groups into very distinct groups indicating no clear delineation of trait to a specific group. However, increased number of isolates from groups and sources may yield greater separation of the groups.
The correlation matrices (Fig. 7) indicate the B. cereus group positively correlated with the presence of the genes hblA, hblC, and plc, and the detection of hemolysis (p < 0.05), while the other Bacillus sp. group were negatively correlated (p < 0.05). Vero cell toxicity was significantly positively correlated with the presence of nheA, nheB, and nheC genes (p < 0.05) suggesting these gene products contributed to the toxicity observed. The presence of the cytK gene was positively correlated with the presence of nheA and plc genes (p < 0.05). Hemolysis was also positively correlated with the presence of the hblA and hblC genes. Additionally, the B. cereus group was positively correlated with oxacillin resistance and negatively correlated with swarming motility. The isolates of the B. subtilis group were negatively correlated with the presence of the plc gene and cephalothin resistance (p < 0.05). The correlation matrix (Fig. 7) did not identify significant correlations between antibiotic resistances and other variables with the exception of a negative correlation of erythromycin resistance and the Casein-Mannitol (CM) methods for biofilm assessment. Chloramphenicol resistance positively correlated with vancomycin, tetracycline, and erythromycin resistance (p < 0.05). There was a strong significant positive correlation of oxacillin resistance and cephalothin resistance (p < 0.05) which was evident in the MDR strains (Table S2). The CM biofilm assessment method correlated significantly with CR as expected as both are associated with degrees of biofilm formation.

Discussion
The widespread detection of Bacillus spp. carrying enterotoxin-enconding genes in both chicken and beef implicates the possible hazardous impact on public health of these bacteria. We show for the first time non-B. cereus strains bearing an inventory of enterotoxigenic genes similar to B. cereus strains that could be the cause of diarrheal illness related to non-B. cereus strains as previously reported 13,25 and thus should be considered as a risk for consumers. Further in-depth study of the sequences of the enteroxigenic genes distributed in the Bacillus genus may yield an explanation of the origin of the genes in non-B. cereus strains and the possibility of horizontal gene transfer between strains and species.
The Bacillus cereus sensu lato group are known to be involved in food poisoning 26 although not always a notifiable disease in the majority of the international locations; therefore, incidence statistics is restrained even though fatal incidences were pronounced. Food poisoning caused by Bacillus cereus occurs year-round diffusely over a geographic area 27 . Bacillus cereus induced 0.7% of foodborne disease of the 31 primary pathogens within the US 28 while in the Netherlands B. cereus was reported as a major causative agent in 5.4% of the foodborne outbreaks in 2006 and 32% of foodborne outbreaks in Norway in 2000 28 . B. cereus is reported as the fourth and second major cause of notified FoodBorne Organisms in the European Union and France, respectively 26 . Meat dishes were the second most commonly implicated foods in B. cereus outbreaks 29 and 24% of B. cereus outbreaks were associated with meat or poultry dishes 27,28,30 .
In addition to the potential risk of non-B. cereus strains to consumers the specific isolation of Paenibacillus alvei (formerly Bacillus alvei) from the imported frozen meat samples is of concern. P. alvei has been one of 22 Paenibacillus species to be recognized as responsible for transitory or authentic infections in human clinical samples 31 . Our study showed a prevalence at 25.8% of B. cereus group and 48.6% Bacillus spp. of 255 meat samples consistent with earlier studies reporting rates of incidences in raw beef and chicken in different countries between 23.5% and 80% 25,30,[32][33][34][35][36][37][38][39][40] . This variation might be due to differences in the hygienic practices executed in the meat shops in different countries and even in the same country. It should also be emphasized that, a critical control point for microbial contamination is the storage of food at ambient temperature (room temperature) of about 5-6 h and improper cooking of food before consumption which favors endospore germination producing an increase in the total Bacillus spp. count 39,40 . B. cereus survives not only at room temperature 41 , but it is also found in heat treated meat 33 . The B. cereus spores are resistant to heat used in cooking leading to inadequate meat preparation allowing these pathogens to germinate, replicate, and potentially produce heat stable toxins 39,42 . The initial contamination of the meat might have occurred in the environment as Bacillus spp. are ubiquitous bacteria in soil, in intestinal tracts of animals, and in a variety of foods and ingredients 39 . Bacterial contamination of meat can also occur during processing at the slaughter house, from the water, air, soil, the workers and equipment involved or the carcass itself [43][44][45] .
The virulence factors responsible for enterotoxins production in B. cereus are primarily HBL, NHE and cytotoxin K 11,12 . Diarrhea can be caused by the production of one or more enterotoxin by vegetative cells in the small intestine 1 . High detection rates between 40 and 70% in B. cereus isolated from food origin of the HBL gene complex have been reported [46][47][48][49] . Similarly, high prevalence among B. cereus isolates from food and environmental sources [49][50][51][52][53] , as well as in some reference strains 54 have previously been reported for the nheABC and entFM gene complexes. All isolates of the B. cereus group carried one or more enterotoxin gene of the nine investigated in this study indicating all B. cereus group isolates could cause diarrheal illness. A similar results was obtained by Smith et al. 36 and Yang et al. 51 . Of the non-B. cereus group there was a 72% detection of one or more enterotoxin gene(s) indicating that this group may also participate in the cause of diarrheal illness 1 . One or more gene of the NHE complex (nheABC) was detected in 22 (81.5%) of the B. cereus isolates indicate the presence of NHE enterotoxin. A previous study recorded that, almost all isolates contained at least one gene of the NHE complex 55 . The isolates in the current study indicated the presence of at least one gene of the HBL complex (hblDAC) in 19 isolates (70.4%), and similarly 73% presence of HBL complex was reported in food-poisoning strains 50 , whereas others found HBL complex in 65.5% and 55.2% in B. cereus isolates 30,56 . In the present study, fewer isolates contained HBL complex genes compared to NHE complex genes, which is consistent with previous findings 30 . Although Ngamwongsatit et al. 57 and Vyletelova and Banyko 58 indicated that the three genes of HBL and NHE complexes form an operon, the present and previous report by Tewari et al. 30 indicate that the structural organization of these genes can be different. Interestingly, 32.3% of isolates did not show the presence of any of the HBL complex genes and produced positive β-hemolysis on blood agar. The observed hemolytic activity might be produced by other toxins such as hemolysin I (Cereolysin O) 59 , hemolysin II 60 , hemolysin III 61 or cytK 62 . The enterotoxin cytK gene was found in 81.5% of the B. cereus isolates which is similar with the results reported by Ngamwongsatit et al. 57 , whereas others detected it in smaller percent of their isolates 63,64 . On the other hand, the entFM gene was found in 35.5% of our Bacillus spp. isolates which is far different from the 100% recorded by Ngamwongsatit et al. 57 and the 93% in their B. cereus tested isolates. Smith et al. 36 isolated 27 B. cereus strains out of 60 poultry samples, which contained the gene(s) for at least one of the toxins (bceT, nheABC, hblACD), although none of the strains contained the cytK gene. This was also consistent with our results with the exception with the cytK as previously recorded.
The hazardous impact on the food production lines as a consequence of biofilm formation comes from the fact that the biofilms are dynamic structures that can release planktonic cells able to spread and colonize new areas of the production line and equipment [65][66][67][68][69][70] . The members of Bacillaceae family 65,66 have the property to attach to biotic (meat surfaces) and abiotic (meat processing equipment such as conveyor belts, tables, knives) surfaces 67 , thus the removal of biofilms in the food processing environment is critical [65][66][67][68][69][70] . In the present study, 84.7% of the Bacillus isolates analyzed, were able to form biofilms indicating a high potential for attachment to food processing surfaces and equipment which may lead to increase spread and incidence of illness. These results agree with previous reports that foodborne B. cereus and B. thuringiensis isolates are variable (from no or weak to strong) biofilm producers 65 . The current study highlighted that the non-B. cereus strains isolated from different meat sources were also variable biofilm producers. To our knowledge, the capacity of these species to form biofilm have not been previously studied. The disruption of biofilm formation is a key control point in eliminating the spread of Bacillus to food and decrease illness.
Antibiotic resistance, precipitated by the overuse of antimicrobials, may rise up from a variety of mechanisms, in particular horizontal gene transfer of virulence and antibiotic resistance genes, that is frequently facilitated with the aid of biofilm formation [71][72][73] . Antimicrobial resistance is a growing problem around the world and is associated with increasing mortality and medical costs 73 . Determining the resistance of B. cereus to antimicrobial agents is critical for treatment during outbreaks. The isolates from beef and chicken in this study showed resistance profile highly consistent with previous reports 49,74 . While β-lactams have become ineffective, chloramphenicol, an antibiotic in use in Egyptian clinics, would still be effective against the isolates in this study. MAR indices calculated in this study for meat are similar to values of E. coli from poultry farms 75 but are approximately double the values for Enterococcus isolates 76 . The high variability in MAR index and the multi-antibiotic resistance profiles of the isolates indicate a variability in effective treatment and the importance on early determination of antibiotic resistance during infection 77 . The introduction of resistant B. cereus into the raw meat likely indicates contamination inputs from livestock and poultry operations, during transport of the meat or meat handling at the distribution outputs and poor hygienic practices in meat shops and restaurants 29,30,43,44 . Because there are no criteria for MAR index for B. cereus, it is difficult to assess human health risks due to presence of antimicrobial resistant B. cereus in the meat.
The deficit of correlation or association of virulence factors and antibiotic resistances tested within the Bacillus spp. isolated in this study are consistent with observations in other species examining the correlation of virulence and antibiotic resistance 78,79 . The negative correlation of biofilm formation and resistance to the macrolide erythromycin might be selected for in part due to increased resistance to antimicrobial within the biofilm 80 . The biofilm formation in vivo maybe sufficient for resistance to macrolides at concentrations the isolates are exposed to, but not during our in vitro assay where biofilm formation is unlikely. Interestingly, He et al. 81 showed that some strains of Staphylococcus epidermidis when exposed to sub-inhibitory concentrations of erythromycin showed increased expression of the resistance gene ermC with decreased biofilm formation. A similar complex SCIEnTIFIC REPORTS | (2018) 8:11600 | DOI:10.1038/s41598-018-29932-3 relationship of biofilm formation and antibiotic resistance maybe present in the isolates in this study. The variability of virulence and antibiotic resistance profiles within the isolates are usually associated with genomic plasticity 79,82 and Bacillus spp. are considered to have a plastic genome 83,84 . While plasmids may contain both virulence and antibiotic resistance genes 85,86 , the majority of plasmids currently sequenced and described within Bacillus spp. do not [87][88][89][90] . This lack of co-location on plasmids of virulence and antibiotic resistance genes may explain the lack of association of these factors within the isolates. Increased number of isolates and tested food products is needed to fully understand the relationship between virulence and antimicrobial resistance in Bacillus spp. and other pathogenic bacteria and further gene expression studies could also contribute to better understand the relationship between virulence and resistance.

Conclusion
Considering food safety issues, this provides confirmation that B. cereus and non-B. cereus must be considered important food-borne pathogens and underlines the need to improve monitoring. The present investigation highlights for the first time an initial finding that the non-B. cereus group poses a public health risk to consumers as a consequence to their carriage of the enterotoxigenic virulence genes and exhibiting phenotypic virulence characteristics (cytotoxicity and haemolytic activity, as well as MDR and biofilm formation) as the B. cereus. In addition, the importation of meat involves a degree of disease risk to the importing country through the introduction of pathogens previously uncommon in Egypt 91 . Further research is needed to better assess the risk pose to public health by non-B. cereus species.

Materials and Methods
Isolation and identification of Bacillus spp. A total of 255 retail meat samples, comprising 156 raw chicken meat and 99 beef meat (59 local and 40 imported) were purchased from different retail outlets in the local markets of Cairo and collected and transported to the laboratory following aseptic and safety precautions.
A stomacher was used to homogenize 10 g of each sample in 90 mL of buffered peptone water (BPW) for 2 min. Heat treatment of all samples at 70 °C for 15 min to was used to eliminate vegetative cells and allow the isolation of spores 92 . The pasteurized samples were immediately placed in ice to prevent spore germination. An amount of 100 μl was spread on Mannitol-Egg Yolk-Polymyxin (MYP) agar plates and incubated at 37 °C for 24-h both aerobically and anaerobically. The plates were examined and presumptive Bacillus spp. were confirmed based on microscopy of Gram-stained preparations and biochemical tests 93 . A number of 15-20 colonies were randomly selected and analyzed by cell morphology, Gram staining, ability to form endospores, growth in the presence of sodium chloride, anaerobic growth, catalase and oxidase activity, Voges-Proskauer test and growth at pH 5.7 92,93 . The ability to ferment carbohydrates, starch hydrolysis, use of citrate as a carbon source, lecithinase activity, and growth inhibition by lysozyme were implemented according to methods described previously 92,93 . Reference strains for the phenotypic tests were Bacillus cereus ATCC 11778 and B. cereus ATCC 14579. Detection of phenotypic virulence factors and toxin encoding genes. Biofilm production. Biofilm production was assessed by the Congo red agar test and Microtiter Plate method as previously described in detail 98,99 . Staphlococcus epidermidis ATCC 35983, a strong slime producer, was used for positive control.
Congo red agar test. The 124 B. cereus and non-B. cereus isolates were cultured on brain heart infusion (BHI) agar augmented with 36 μg/ml of glucose and 0.8 μg/ml of Congo red. The plates were incubated at 37 °C for 24-h. After an additional 12-h at room temperature, slime production was assessed by colony color. Black colonies are strong slime producers, while red colonies lack slime production.
Microtitre Plate method. A 200 µl of the Bacillus isolates suspension equivalent to 0.8 McFarland (yielding 105 cfu/ml) in tryptone soy broth (TSB) was transferred to a 96-well polystyrene microtiter plate. The bacterial cell suspension was incubated for 18-h at 37 °C. The plates were decanted to remove well contents, washed with running tap water, allowed to dry for 30 min and stained with crystal violet (CV; 25%) for 5 min at room temperature before decantation washing and drying for 30 min. The plates were incubated for 1 min with 200 μl HCl 25% in each well. The absorbance at 570 nm of the CV stain were read for each well. Uninoculated wells subjected to the same procedures were used as negative controls.
Swarming activity. Swarming motility was observed as previously described 100,101 . Briefly, an overnight culture (2 × 10 8 cells/ml) was spotted (0.5 μl) onto the center of TrM plates (1% tryptone, 0.5%, NaCl, 0.25% agar). After 6-8 h incubation at 37 °C in a humidified chamber, the diameter of halos generated by growth were measured. demonstrated by discontinuous double hemolysis pattern on blood agar plates 102 . Overnight cultures of Bacillus isolates in BHI broth with 0.1% glucose (BHIG) were used to inoculate blood agar plates (Columbia agar +5% sheep-blood, Oxoid) by spot inoculation. The sheep blood agar plates were incubated at 24 °C and were frequently observed between 12 and 72 h. B. cereus ATCC 14579 and B. cereus INRA C15 were used as positive controls. B. cereus NC 1291 was used as a negative control.
Vero cell cytotoxicity assay. Bacillus spp. isolates were inoculated into 10 ml BHIG and incubated at 30 °C for 24-h at 150 rpm on a rotary shaker. Following incubation, 100 µl of culture was diluted into fresh 10 ml BHIG and re-incubated. The supernatant was harvested by centrifugation at 5000 g for 5 minutes and filtered with a 0.22 μm syringe filter (Millipore). Vero cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum (FBS) at 37 °C. Trypsinized Vero cells were diluted to 106 cells/ml in normal growth medium and 100 µl was placed in the wells of a Falcon 96-well flat bottom cell-culture plates. Plates were incubated until a confluent monolayer of cells was formed, approximately 24-h, at 37 °C 103 . Isolate-free supernatant was added (100 μl) to the first column of the 96-well plate and serially diluted by 2-fold across the columns of the plate. The plates were incubated for 18-h at 37 °C. Reactions were considered positive if greater than 50% of Vero cells showed detachment from the plate when examined under light microscopy.
PCR detection of virulence genes for enterotoxins in B. cereus. DNA isolation. The 124 isolates of Bacillus spp. isolates were grown in 5 mL nutrient broth with shaking for 18 h at 30 °C and harvested at 5,000 g for 5 min. QIAamp DNA Mini Kit was used for genomic DNA extraction and purification. The concentration and purity of genomic DNA was measured using an Ultraspec 3000 spectrophotometer at the absorbance 260 and 280 nm. PCR was performed to detect eight enterotoxigenic encoding endotoxins genes (hblA, hblC, hblD, nheA, nheB, nheC, cytK and entFM) and one hemolytic gene encoding phospholipases (plc). A positive reference strain of B. cereus ATCC 14579 and sterile MilliQ water as a negative control was used in PCR analysis 104,105 . Table S4 provides details about the primers used.
Conditions for PCR amplification. PCR reactions were comprised of 25 ng genomic DNA, 10 mM Tris-HCl (pH 8.3), 10 mM potassium chloride, 2.5 mM magnesium chloride, 0.8 mM dNTPs, 1 μM each primer, and 0.5 U of Taq DNA polymerase (Promega Corporation, WI, USA). Ultrapure sterile water was used as non-template DNA control and for the PCR component preparation. A PTC-100 Programmable Thermal Controller was used for PCR amplification. The optimized multiplexPCR utilized a standard 3-step cycling: 3 min at 95 °C; 35 cycles of (1) 94 °C for 30 sec, (2) 54 °C for 45 sec, and (3) 72 °C for 1.5 min; and a final extension at 72 °C for 5 min. PCR reaction were performed in triplicate. After amplification, gel electrophoresis was used to analyze PCR fragments for presence and correct size compared to positive control. PCR runs where a negative control showed amplification or positive control did not amplify were ignored and repeated.
Statistical analyses. Numerical coding was used for antibiotic resistance phenotypic, biochemical results, and gene presence. Detection or absence of a specific gene (eg. hblA) was denoted as 1 and 0, respectively. Hemolysis, Vero Cell, Congo red (CR), and CM analysis results were exchanged with numerical values matching the degree of observed activity with 0 with absence followed by sequential integers (1, 2, etc.). For antibiotic resistance, antibiotic sensitive was denoted as 0 and resistance as 1. The open statistical program R was used for statistical analysis 106 . Multivariant statistical analysis were carried out using functions in the 'vegan' package 107 . Binomial similarity matrices were calculated for the isolate profiles using 'vegdist' function and used in permutational multivariate ANOVA (MANOVA) analyses using the 'adonis' function. Multiple pairwise comparisons were conducted using the 'pairwise.perm.manova' within the 'RVAideMemoire' R package 108 . A permutational ordianation method was used to determine variables statistical significant variables in further analysis utilizing 'ordiR2step' function. All variables, not related to species identification, were signification (p < 0.01). Principal component analysis (PCA) was performed and visualized with the R packages 'FactoMineR' 109 and 'factoextra' 110 . The 'cor' function was used to calculate correlations and 'cor.test' function was used to determine significance between variables. The 'corrplot' function from the 'corrplot' package was used to visualize significant correlations 111 . For multiple comparisons, False Discovery Rate was used to adjust p-values 112 . The function 'heatmap.3' in the 'GMD' package was used to generate heatmap representations a 113 . Significant difference between data shown as percentages such as biochemical tests or antibiotic resistance by source or group was determined by proportional Z test.
Availability of data and materials. All data generated or analyzed during this study are included in this published article.