Traceability of potential enterotoxigenic Bacillus cereus in bee-pollen samples from Argentina throughout the production process

Bee-pollen is a functional food sold for human and animal consumption but also is a favorable microhabitat for many spore-forming bacteria. Among them, Bacillus cereus can produce several toxins and other virulence factors, causing an emetic or diarrheal syndrome after ingestion. The study involved 36 bee-pollen samples obtained from different sampling points throughout the production process (collecting, freezing, drying, and cleaning) in Argentina. Fifty isolates of B. cereus yielded 24 different fingerprint patterns with BOX and ERIC primers. Only three fingerprint patterns were maintained throughout the production process. In contrast, others were lost or incorporated during the different steps, suggesting that cross-contamination occurred as shown by differences in fingerprint patterns after freezing, drying, and cleaning steps compared to the initial collection step. Genes encoding for cereulide ( ces ), cytotoxin K ( cyt K), sphingomyelinase ( sph ), the components of hemo-lysin BL ( hbl A, hbl B, hbl C, hbl D) and non-hemolytic complex ( nhe AB) were studied. All the isolates displayed one or more enterotoxin genes. The most frequent virulence genes detected belong to the HBL complex, being the most abundant hbl A (98%), followed by hbl D (64%), hblB (54%), and hbl C (32%), respectively. Ten strains (20%), present at all sampling points, carried all the subunits of the HBL complex. The non-hemolytic en-terotoxic complex ( nheAB ) was found in 48 strains (96%), while seven strains (14%) present at all sampling points showed the amplification product for sphingomyelinase ( sph ). One cereulide-producer was isolated at the cleaning step; this strain contained all the components for the hemolytic enterotoxin complex HBL, the NHE complex, and cytotoxin K related to the foodborne diarrhoeal syndrome. In total, 11 different virulence patterns were observed, and also a correlation between rep-fingerprint and virulence patterns. The results suggest that bee-pollen can be contaminated at any point in the production process with potential enterotoxic B. cereus strains, emphasizing the importance of hygienic processing.


Introduction
Bee-pollen is the result of the agglutination of pollen grains collected from flowers and mixed with nectar and salivary secretions by honeybees (Bertoncelj et al., 2018;Denisow and Denisow-Pietrzyk, 2016). Honeybees (Apis mellifera L.) collect pollen from different flowers during collecting trips and pack pollen grains into pollen pellets on their hind legs with the help of several combs and hairs. The pollen transferred to the hive in the form of pollen loads is called "bee-pollen" (Kieliszek et al., 2018), which is stored inside the hive separately from Manirajan et al., 2016;Gilliam, 1979;Gilliam et al., 1990;Kačániová et al., 2009), while post-harvest sources of bacteria are likely to be the same as those for other food products, and may include humans, equipment, containers, dust, insects, animals, and water (De-Melo et al., 2015, Estevinho et al., 2012. Also, bee-pollen is a favorable environment for spore-forming bacteria, among this group, Bacillus cereus sensu stricto (B. cereus s.s) is a ubiquitous bacterium found in soil, plants, and other niches such as enteric tracts of insects, honey and pollen (Ambika Manirajan et al., 2016;Gilliam, 1979;Gilliam et al., 1990;Heydenreich et al., 2012;Kačániová et al., 2009;López and Alippi, 2007;Moreno Andrade et al., 2019;Snowdon and Cliver, 1996). B. cereus s.s. belongs to the Bacillus cereus group consisting of Bacillus cereus, Bacillus anthracis, Bacillus cytotoxicus, Bacillus mycoides, Bacillus pseudomycoides, Bacillus thuringiensis, and Bacillus toyonensis (Guinebretière et al., 2013;Liu et al., 2018;Vilas-Boas et al., 2007).
The emetic syndrome is caused by the thermostable peptide cereulide, encoded by non-ribosomal peptide synthetase genes (ces) (Ehling-Schulz et al., 2005). Intoxication is usually caused by the ingestion of toxin pre-formed in food by ces-positive B. cereus strains (Ehling-Schulz et al., 2015).
The most recurrent toxins reported to produce diarrheic syndrome are hemolysin BL (Hbl), the non-hemolytic enterotoxin (Nhe), and cytotoxin K. Cytotoxin K, encoded by cytK, causes a diarrheal syndrome with necrotic, hemolytic, and cytotoxic effects on the intestinal epithelium . It has been found in B. cereus strains that cause severe necrotic enteritis (Lund et al., 2000). Hbl, encoded by the operon hblDAC, is a tripartite protein complex. All the parts are required to maximize their biological effects, i.e., hemolytic, cytotoxic, dermonecrotic, and vascular permeability activities of Hbl (Beecher and Wong, 1994;Carter et al., 2018;Ryan et al., 1997). Nhe is composed of three genes (nheA, nheB, and nheC) that constitute the nheABC operon, and it is responsible for the diarrheal food-poisoning syndrome when all the three components are present (Lindbäck et al., 2004). Nevertheless, other authors considered that it is not clear which combination and/or polymorphisms of enterotoxin genes are associated with B. cereus strains responsible for diarrhea illness (Ceuppens et al., 2011).
Finally, sphingomyelinase (SMase) has been reported as a virulence factor for septicemia (Hsieh et al., 1999;Oda et al., 2012); while other authors suggested that SMase acts in connection with Hbl (Beecher and Wong, 2000) and that Nhe and SMase complement each other significantly to cause B. cereus full virulence (Doll et al., 2013). The study of correlations between toxigenic profiles and other markers, e.g., rep-PCR fingerprinting, antibiotic sensitivity, RAPD patterns, and multilocus sequence typing, allows gaining further insight into the epidemiology of enterotoxic B. cereus strains (Ghelardi et al., 2002;Yu et al., 2020). Previous work evaluated which was the effect of the different stages of bee pollen production that lead to changes in the microbiota throughout the production process in the Southwest of Buenos Aires province, Argentina (Fernández et al., 2020). Within this context, the present work aimed to study the traceability of potential enterotoxic Bacillus cereus strains based on colony counts, rep-fingerprinting, and toxigenic profiles at four sampling points of the production process of pollen (collecting, freezing, drying, and cleaning).

Sampling
The bee-pollen production process designed by Cooperativa de Trabajo Apícola Pampero Limitada (Bahía Blanca, Argentina) involved four stages separated in time and space. This process comprises collection by using pollen traps, freezing (−10 °C), dehydration (drying at 40 °C, 48 h), cleaning, and packaging. The study involved 36 bee-pollen samples from the production line of three beekeepers from South West of Buenos Aires Province, Argentina. Samples were analyzed by testing three samples from each beekeeper at each sampling point (collecting, freezing, dehydration, and cleaning). All samples were aseptically collected in sterile 100 ml vials and stored refrigerated (4 °C) until processing within 30 days.

Isolation and identification of Bacillus cereus s.l.
For the isolation of Bacillus cereus s.l. (B. cereus group) a technique adapted and modified from López and Alippi (2007) was used. Briefly, 5 g of each bee-pollen sample was mixed with 5 ml of 0.01 M sodium phosphate buffer saline (pH 7.2) and submitted to agitation for 40 min. Each sample was filtrated to a new centrifuge tube and centrifuged at 6000 ×g for 30 min at 4 °C. Most of the supernatant was discarded to leave 3 ml of fluid that was vortex-mixed with the sediment and heated in a water bath at 80 °C for 10 min to kill bacterial vegetative cells and yeasts, and at the same time, activate Bacillus cereus spore germination. A.C. López, et al. International Journal of Food Microbiology 334 (2020) 108816 Samples were vortex-mixed again for 2 min, and 100 μl of the sedimentfluid mixture was poured over the surface of polymyxin-pyruvate-eggyolk-mannitol-agar (PEMBA) plates (Britania®, Argentina) and spread by using a sterile cotton swab. Plates were incubated at 32 °C and examined daily and up to 5 days for bacterial growth. Distinct colonies of B. cereus group, i.e., turquoise blue crenated colonies (mannitol negative) surrounded by a distinct opaque zone of egg yolk precipitation (lecithinase positive) were counted. Counts were expressed as colony-forming units per g of pollen (CFU.g −1 ). Colonies growing on PEMBA were identified by their shape, rhizoidal growth, and hemolytic activity as belonging to the B. cereus group. Bacterial smears were examined for the presence and location of spores within cells, as well as for the size and shape of vegetative cells (Parry et al., 1983;Priest et al., 1988). Also, the presence of both unstained globules in the cytoplasm and parasporal crystals were examined by using a phase-contrast microscope (1000×, oil immersion) (EFSA BIOHAZ, 2016; López and Alippi, 2007). The number of colonies submitted to phenotypic and genotypic tests corresponds to the square root of the total number of colony-forming units (CFU) per plate. These isolates were tested for catalase, production of lecithinase, Voges-Proskauer reaction, mannitol utilization, anaerobic utilization of glucose, hemolytic activity, and starch and gelatin hydrolysis according to standard protocols (Gordon et al., 1973;Lancette and Harmon, 1980;Pirttijärvi et al., 1996).

Statistical analysis
Data from colony counts of B. cereus between the three beekeepers at each sampling point throughout the production process were analyzed by one-way analysis of variance using Infostat software (Di Rienzo et al., 2013).

DNA preparation
Suspected Bacillus cereus isolates were cultured on PEMBA plates for 24 h at 32 °C under aerobic conditions. For the preparation of bacterial DNA template, a rapid procedure was used (Alippi and Aguilar, 1998). Briefly, bacterial colonies were picked up by using a 1-μl plastic disposable loop and suspended in 200 μl double-distilled sterile water. The sample was vortex mixed and centrifuged at 10,000 ×g for 4 min, the supernatant was removed, and the pellet was resuspended in 150 μl of an aqueous suspension of 6% resin Chelex 100 (Bio-Rad). The mixture of cells and resin was incubated at 56 °C for 20 min, vortex mixed, incubated at 99 °C for 15 min, and vortex mixed for 1 min. Finally, bacterial debris and resin were removed by centrifugation.

RFLP analysis of PCR-amplified 16S rDNA
Fifty isolates of the B. cereus group from bee-pollen were identified at the species level by RFLP analysis of PCR-amplified 16S rRNA as previously described (López and Alippi, 2019). Briefly, universal primers 27f and 1492r were employed (Yu et al., 2013). PCRs were carried out in a final volume of 25 μl (Yu et al., 2013). After amplification of the approximately 1492 bp PCR product, subsamples of 2 μl were incubated with endonucleases AluI and CfoI, according to the manufacturer's specifications (Promega®). RFLP analysis was performed by electrophoresis in a 1.6% agarose gel at 70 V for 2 h.

Analysis of the diversity of isolates by rep-PCR
The rep-PCR method with BOX (BOXA1R) and ERIC (ERIC1R and ERIC2) primers was used (Versalovic et al., 1994). PCR amplifications were done according to López and Alippi (2007). For the analysis of amplification products, 5 μl of each PCR reaction was run on a 1.6% (W/V) agarose gel in TBE buffer and visualized by using ethidium bromide and UV light. A digital image of each gel was analyzed using GelcomparII® software (v. 5.1, Applied Maths). Cluster analysis was performed using the DICE similarity coefficient and the UPGMA clustering algorithm with a band tolerance of 5% for a combined gel.

Prevalence and levels of Bacillus cereus s.l.
All pollen samples analyzed (n = 36) contained spores of B. cereus s.l. A total of about 3 × 10 2 CFU of B. cereus equivalent to 2 × 10 3 CFU/ g were counted on PEMBA plates. Colony counts revealed no statistically significant differences among means at the different sampling points (p = 0.5). However, B. cereus incidence (total of CFU/g at each sampling point) increased from collection (7 × 10 1 CFU/g) to freezing (5 × 10 2 CFU/g) and drying (9 × 10 2 CFU/g) and slightly decreased at the final step of cleaning (3 × 10 2 CFU/g). In spite that B. cereus was present at the different sampling points, the spore countings complied with the food safety criteria for B. cereus in food (< 10 5 CFU/g) (EFSA BIOHAZ, 2016;Lücking et al., 2013).
The differences obtained in colony counts per g of pollen suggest that bee-pollen can be contaminated at any point in the production process, as shown in bacterial counts after freezing, drying, and cleaning steps compared to the initial collection step, where higher counts occurred at freezing and drying.
As reported by De Melo and co-workers (De-Melo et al., 2015), the presence of any microorganism in dehydrated bee-pollen is related to inadequate hygienic practices during manipulation at the steps of collection and processing, and contamination of floral pollen grains on the plant or by bees. In the production process, the time that bee-pollen remains in the collection traps is critical because pollen grains are in contact with air, dust, and other dirt. The presence of viable microorganisms in samples of dehydrated bee-pollen after the collection step could be related to two hypotheses: that the low temperature used in the dehydration process is insufficient to remove microorganisms or that contamination occurs during the processing steps after dehydration (De-Melo et al., 2015). Other authors (Estevinho et al., 2012) affirmed that freezing and drying steps allow the multiplication of microorganisms.
Fifty strains were selected from the isolation plates, corresponding to the square root of the total number of colony-forming units (CFU) per plate (Section 2.2), for further analysis (collecting = 8; freezing = 12; A.C. López, et al. International Journal of Food Microbiology 334 (2020) 108816 drying = 17 and cleaning = 13) ( Table 2).

Identification of selected isolates
Typical B. cereus isolates produced crenated colonies retaining the turquoise blue of the pH indicator because of their inability to ferment mannitol acidifying the medium, and generated an egg-yolk precipitation halo as a result of lecithinase activity. The 50 selected strains were Gram-positive, catalase positive, and showed ellipsoidal spores in a central position, not distending the sporangia. No parasporal crystals were detected, and the cytoplasm was filled with unstained globules. All the isolates were facultatively anaerobic, grew in 7% NaCl, hydrolyzed gelatin and starch, and were positive for Voges Proskauer, reduction of nitrates to nitrites, and hemolytic activity. These characteristics correspond with the phenotypic features typical for Bacillus cereus (Gordon et al., 1973;Parry et al., 1983;Priest et al., 1988). Besides, all the strains showed the expected restriction patterns for Bacillus cereus s.s. when using a combination of AluI and CfoI enzymes (López and Alippi, 2019).

Analysis of the diversity and traceability of isolates by rep-PCR
The fingerprints generated by BOX-and ERIC-PCR were analyzed by using GelcomparII software, being the number of polymorphic bands Fig. 2. Dendrogram from computer-assisted analysis of rep-PCR profiles of 50 B. cereus s.s. isolates obtained from the different sampling points. Cluster analysis was performed using the DICE similarity coefficient and the UPGMA clustering algorithm with a band tolerance of 5%.
A.C. López, et al. International Journal of Food Microbiology 334 (2020) 108816 between 5 and 14, ranging between 200 bp and more than 1500 bp for BOX (Fig. 1), and between 3 and 14 bands, ranging between 150 bp and more than 1500 bp for ERIC (Fig. 1). Among the 50 isolates of B. cereus s. s, 24 different rep-fingerprint patterns by using BOX and ERIC were identified ( Table 2). The BOX and ERIC data were combined and used to generate a dendrogram (Fig. 2). Fifteen strains yielded unique fingerprint patterns, while the rest (n = 35), were grouped into nine different clusters (Fig. 2). All the isolates clustered together at 46% similarity level, while clusters 1, 2, and 3 were separated at about 60%, 70%, and 65%, respectively. The results of the cluster analysis of fingerprints generated by rep-PCR with BOX and ERIC primers revealed a high genetic diversity among B. cereus strains in coincidence with results reported by other authors working with isolates from other foods (Chaves et al., 2011;Lee et al., 2012;López and Alippi, 2007).
The rep-fingerprint patterns obtained at the different sampling points throughout the production process were compared (Table 3). At collecting, eight patterns (A, B, C, D, E, F, G, and H) were observed, while at freezing, patterns A, C, D, E, and G were maintained; three new patterns (I, J, and K) were incorporated and three patterns (B, F, and H) were lost. Besides, at drying, 12 patterns were observed, where five patterns (A, C, G, I, and K) were maintained; seven new patterns were incorporated (L, M, N, O, P, Q, and R) and three were lost (D, E, and J). Finally, during the cleaning step, ten patterns were visualized, where four patterns were maintained (A, C, G, and K), six new were incorporated (S, T, U, V, W, and X), and the eight patterns which were incorporated at the drying step were lost.
Only three fingerprint patterns (named A, C, and G) were maintained throughout the production process, while others were lost or incorporated during freezing, drying, and/or cleaning (Table 3). These results suggested that cross-contamination occurred as shown by differences in fingerprint patterns after freezing, drying, and cleaning steps compared to the initial collection step.

Detection of virulence genes by PCR
An overview of all virulence genes detected by PCR is provided in Tables 2 and 4. All the isolates displayed one or more enterotoxin genes. The most frequent virulence genes detected belong to the Hbl complex, being the most abundant hblA (98%), followed by hblD (64%), hblB (54%), and hblC (32%), respectively. Besides, ten strains (20%), present at all sampling points, carried all the subunits of the Hbl complex. The non-hemolytic enterotoxic complex (nhe) was found in 48 strains (96%) that were present at all sampling points, while six strains (12%), isolated at the freezing, drying, and cleaning steps showed the amplification product for sphingomyelinase (sph) ( Table 4). One cereulide-producer (BCP40) was isolated at the cleaning step; this strain also contained all the components for the hemolytic enterotoxin complex Hbl, and cytotoxin K related to the foodborne diarrhoeal syndrome. Interestingly, strain BCP40 showed a unique fingerprint pattern named X (Figs. 1 and 2).
Most strains of B. cereus isolated from various food sources contained both hbl and nhe-encoding genes (Chaves et al., 2011;Lee et al., 2012;López and Alippi, 2010;Yu et al., 2020); similar results were obtained from strains isolated from bee-pollen in this study. The incidence of B. cereus strains from pollen carrying cytotoxin K (cytK) and cereulide (ces) genes was lower than reported by other authors for other types of foods (Guinebretière et al., 2002;Lee et al., 2012;Yu et al., 2020).
A certain degree of correlation between rep-fingerprinting and virulence gene patterns were found. For instance, all the isolates showing fingerprint pattern C correlated only with virulence pattern V, and the same situation occurs with fingerprint pattern G with virulence pattern VIII, K with virulence pattern XI, fingerprint pattern N with virulence pattern IV, fingerprint pattern O with virulence pattern VI, fingerprint pattern P with virulence pattern VII, and fingerprint pattern X with virulence pattern X, respectively. However, virulence patterns I, II, III, and IX correlated with more than one fingerprint pattern, being the most promiscuous, virulence pattern IV that were present in fingerprint patterns F, M, S, T, U, V, and W (Table 2).

Table 3
Traceability of rep-fingerprint patterns in Bacillus cereus strains isolated from bee-pollen throughout the different sampling points tested for all the beekeepers.

Collecting
Freezing Drying Cleaning 10 Totals A.C. López, et al. International Journal of Food Microbiology 334 (2020) 108816 The ubiquity, resistance, and persistence of B. cereus spores favor their survival from the environment to food processing facilities. Moreover, due to their strong adhering properties and ability to form biofilms, contamination of food products, including pollen, is almost impossible to avoid (Carlin, 2011).
The results obtained here emphasize the importance of hygienic processing to avoid spore contamination at all steps of the production process. This work represents the first analysis of potential enterotoxic B. cereus s.s. strains contamination at all steps during the production process of bee-pollen. It reinforces the idea that appropriate management and practices would improve the microbiological quality of beepollen for human consumption.

Declaration of competing interest
The authors declare that no conflict of interest exists.