Development of a Taxon-Specific Real-Time PCR Method Targeting the Bacillus subtilis Group to Strengthen the Control of Genetically Modified Bacteria in Fermentation Products

Most of the bacteria that are used to produce fermentation products, such as enzymes, additives and flavorings, belong to the Bacillus subtilis group. Recently, unexpected contaminations with unauthorized genetically modified (GM) bacteria (viable cells and associated DNA) that were carrying antimicrobial resistance (AMR) genes was noticed in several microbial fermentation products that have been commercialized on the food and feed market. These contaminations consisted of GM Bacillus species belonging to the B. subtilis group. In order to screen for the potential presence of such contaminations, in this study we have developed a new real-time PCR method targeting the B. subtilis group, including B. subtilis, B. licheniformis, B. amyloliquefaciens and B. velezensis. The method’s performance was successfully assessed as specific and sensitive, complying with the Minimum Performance Requirements for Analytical Methods of GMO Testing that is used as a standard by the GMO enforcement laboratories. The method’s applicability was also tested on 25 commercial microbial fermentation products. In addition, this method was developed to be compatible with the PCR-based strategy that was recently developed for the detection of unauthorized GM bacteria. This taxon-specific method allows the strengthening of the set of screening markers that are targeting key sequences that are frequently found in GM bacteria (AMR genes and shuttle vector), reinforcing control over the food and feed chain in order to guarantee its safety and traceability.

The current strategy that has been implemented by enforcement laboratories in order to control genetically modified organisms (GMO) in the food and feed chain is mainly targeting genetically modified (GM) plants. This strategy is composed of two main successive steps using real-time PCR methods. More precisely, a first-line screening analysis step, one that is targeting the taxon-specific markers and genetic elements that are commonly found in GM plants, is applied in order to screen for the potential presence of GM plants. A second-line analysis step is then applied in order to identify the specific GM plant events in targeting their unique junctions between the host genome and the transgenic cassette. The second-line methods are provided by the GM plant developer companies [21][22][23]. However, in order to control GMM contaminations, all of the sequences that are related to GMM are confidential and, in contrast to the GM plants that are intended for human and animal consumption, no methods that are specific to GMM strains are provided by manufacturers to the enforcement laboratories, therefore hampering the the establishment of a similar workflow for GMM testing [5][6][7][8][9][10][11][12]15].
In order to overcome these issues, publicly available data from patents and scientific literature were recently used in order to develop a set of first-line screening real-time PCR methods targeting the genetic elements that are frequently found in the GM bacteria that are used to produce microbial fermentation products [5][6][7]9]. More precisely, these methods target three AMR genes, being cat (GenBank: NC_002013.1), aadD (GenBank: M19465.1) and tet-L (GenBank: D00946.1) conferring a resistance to chloramphenicol, kanamycin and tetracycline, respectively, as well as a shuttle vector pUB110 carrying aadD (GenBank: M19465.1) [5][6][7]9]. If at least one of these screening markers is detected, a potential GMM contaminations can be suspected. A second-line analysis step was also developed in order to prove the presence of specific GM bacterial strains using real-time PCR methods targeting their unnatural associations of sequences. These sequences, which are publicly unavailable, were previously characterized by sequencing the DNA that was extracted either from GM bacterial isolates or from commercial microbial fermentation products with GMM contaminations [3,[8][9][10][11]13]. Currently, these second-line methods allow the identification of the GM Bacillus velezensis strain that produces protease, the GM B. amyloliquefaciens strain that produces alpha-amylase, the GM B. subtilis strain that produces vitamin B2 and the GM B. amyloliquefaciens strain that produces protease. All of these first-and second-line real-time PCR methods were developed in order to comply with the Minimum Performance Requirements for Analytical Methods of GMO Testing of the European Network of GMO Laboratories, being used as a standard in GMO enforcement laboratories [5][6][7]9,24].
However, in contrast to the GM plant detection strategy, no taxon-specific method for GMM detection is currently available in the first-line screening analysis step. PCRbased methods, including real-time PCR and conventional PCR followed by sequencing, that are specific to key markers such as the bacterial genus (e.g., 16S-rRNA gene) [25,26] and bacterial species of interest (e.g., B. subtilis, B. licheniformis) [12,[27][28][29][30] do exist. Nonetheless, the real-time PCR technology is usually preferred as it has been widely mastered by enforcement laboratories which are controlling GMO. Given that the enforcement laboratories carry out their analyses commonly under accreditation, the presently used realtime PCR methods should also comply with the Minimum Performance Requirements for Analytical Methods of GMO Testing of the European Network of GMO Laboratories. Moreover, in order to efficiently strengthen the GMM detection strategy, the taxon-specific method needs to be compatible with the set of methods that are used in the first-line screening analysis step (e.g., real-time PCR technology, TaqMan chemistry and PCR conditions) [5][6][7]9,24].
To this end, a taxon-specific real-time PCR method was developed in this study in order to target Bacillus species belonging to the B. subtilis group. More precisely, this B. subtilis group is composed in total of 10 [3,6,[9][10][11]13,28,[31][32][33]. The performance of the BSG method was assessed at the specificity and sensitivity levels, according the Minimum Performance Requirements for Analytical Methods of GMO Testing of the European Network of GMO Laboratories [24]. The applicability of the BSG method was also tested using several commercial microbial fermentation products. In addition, this taxonspecific method was designed to be compatible with the first-line screening analysis step of the GMM detection strategy. Therefore, by completing the set of real-time PCR screening methods, the BSG method can strengthen the control over GM bacterial contaminations in microbial fermentation products that are commercialized in the food and feed chain.

Development and Validation of the BSG Method
Based on previous studies [28,30], the 16S-23S region was selected in order to develop a taxon-specific real-time-PCR method allowing the targeting of the B. subtilis group, including B. subtilis, B. licheniformis, B. amyloliquefaciens, B. pumilus and B. velezensis [28,30,32,33]. The targeted 16S-23S region from all of these species was aligned using the Clustal Omega software with the default parameters as well as a set of primers and a probe that was designed using the software Primer3 (Table 1). Table 1. Oligonucleotides of the real-time PCR BSG method targeting bacterial species belonging to B. subtilis group that are used to produce microbial fermentation products. On the targeted sequences, the location of the used oligonucleotides is indicated in bold. This taxon-specific real-time-PCR method, named BSG, was developed in order to be integrated into the first-line screening step of the current GMM detection strategy. Therefore, each real-time PCR assay was applied on a 25 µL reaction volume containing 1X Taq-Man ® PCR Mastermix (Diagenode, Liège, Belgium), 400 nM of each primer (Eurogentec, Liège, Belgium), 200 nM of the probe and 5 µL of DNA. The real-time PCR program consisted of a single cycle of DNA polymerase activation for 10 min at 95 °C followed by 45 amplification cycles of 15 sec at 95 °C (the denaturing step) and 1 min at 64 °C (the annealing-extension step). All of the runs were performed on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). A no template control (NTC) was included in each assay.

Specificity Assessment
First, an in silico analysis of the developed BSG method was performed using SCREENED v1.0 [34,35]. The parameter settings were maximum 10% of mismatches in the annealing site, minimum 90% of alignment length in the annealing site and no mismatch in the last five nucleotides at the 3′ end for the primers. This tool uses a two-step BLAST approach. First, the sequence(s) that were targeted by the developed BSG method were extracted from a dataset belonging to the NCBI Nucleotide RefSeq database (accessed on March 2021; filter: genome AND refseq AND complete). Subsequently, the hybridization properties of the targeted regions and the developed set of primers and probe were investigated. As the input for the BSG method, the sequences from the designed set of primers and probe and the expected amplicons that were generated from the targeted Bacillus species were used (Table 1 and Supplementary Tables S1 and S2).
Second, the developed BSG method was tested in triplicate on 10 ng of DNA from positive and negative controls ( Table 2). For the positive controls, 24 wild-type strains belonging to B. amyloliquefaciens, B. licheniformis, B. pumilus, B. subtilis and B. velezensis as well as 2 GM strains corresponding to B. subtilis (RASFF2014.1249) and B. velezensis (RASFF2019.3332) were used. For the negative controls, 75 wild-type microbial strains belonging to fungal and bacterial species that are frequently used by the food and feed industry in order to produce microbial fermentation products were used. In addition, DNA from an animal (Homo sapiens) and a plant (Zea mays) was included. Table 2. Specificity assessment of the developed real-time PCR BSG method. The presence and absence of amplification are symbolized by "+" and "−", respectively. For each result, the experiment was carried out in triplicate on 10 ng of each sample. The means of the observed Cq are indicated in brackets.

Sensitivity Assessment
Serial dilutions of the DNA from the control plasmid carrying one copy of the targeted sequence from B. subtilis, going from 100 to 0.1 estimated target copy number, were prepared and each dilution point was tested in 12 replicates (Table 3). Table 3. Sensitivity assessment of the developed real-time PCR BSG method using a control plasmid carrying one copy of the targeted sequence from B. subtilis. The presence and absence of amplification are symbolized by "+" and "−", respectively. For each target copy number, 12 replicates were tested. The number of positive replicates out of the 12 replicates tested is indicated and the means of the observed Cq are indicated in brackets. The calculation of the estimated target copy number was based on the control plasmid size (2957 bp), as previously described [5][6][7][8][9][10][11]. The limit of detection LOD95%, defined as the number of copies of the target that were required to ensure a 95% probability of detection (POD), was determined by using the QuoData web application (Table S3). The plausibility check of the POD curve indicated no irregularities and that the POD curve was associated to a LOD95% below 25 estimated target copy numbers [24,36,37].

Applicability Assessment
Several commercial microbial fermentation products, including enzymes and additives, were used in order to evaluate the applicability of the developed BSG method (Table  4). Table 4. Applicability assessment of the developed real-time PCR BSG method using commercialized microbial fermentation products. The BSG method was applied on 25 samples. For each sample, labelled information related to enzyme/additive microbial production sources, forms and application fields is indicated. Each sample was tested at a concentration of 10 ng in duplicate. The presence or absence of PCR amplification is symbolized by "+" or "−", respectively. The means of the observed Cq are indicated in brackets. Available information related to previously reported contaminations (GMM, Bacterial DNA) is also indicated for each sample. * Successful isolation of the GMM strain previously reported.

Samples
From each sample (n°1-25), 10 ng of the DNA was tested in duplicate (Table 4). These products (samples n°1-25) from different brands were labelled as containing vitamin B2, protease, alpha-amylase, cellulase, xylanase, beta-glucanase, transglutaminase, beta-glucosidase or rennet. These products, in liquid or solid forms, are intended for use in various sectors such as livestock farming, brewing, distillery, baking, diary processing, protein processing or fruit processing.
The selection of these samples was performed based on the available information that was related to previously reported contaminations, allowing the present study to cover three possible scenarios (Table 4). First, as was observed with the samples n°1-18, the sample is contaminated by a known GMM, including the vitamin B2-producing GM B. subtilis strain, protease-producing GM B. velezensis strain, alpha-amylase-producing GM B. amyloliquefaciens strain and protease-producing GM B. amyloliquefaciens strain. Second, the sample is contaminated with bacterial DNA that is not related to the currently known GM bacterial strains. Based on previously obtained results, the samples n°19-22 were selected for this scenario as presenting a positive signal for the 16S-rRNA marker and a negative signal for the second-line methods targeting the currently known GM strains. Third, the sample is not contaminated with bacterial DNA, as was observed for the samples n°23-25 for which no positive signal for the 16S-rRNA marker was previously observed.

Illustration of the Reinforcement of the First-Line Screening Analysis Step Using the BSG Method
Few commercial microbial fermentation products (samples n°2 -4, 6, 19-20, 22, 25) that are described in Table 4 were used in order to illustrate the reinforcement of the GMM detection strategy by completing the set of first-line screening methods with the BSG method (Table 5). Table 5. Illustration of the reinforcement of the first-line screening analysis step using the BSG method. The full set of first-line screening methods (BSG, cat, tet-l, aadD and pUB110 methods) was applied on commercialized microbial fermentation products described in Table 4. Each sample was tested at a concentration of 10 ng in duplicate. The presence or absence of PCR amplification is symbolized by "+" or "−", respectively. The results for the BSG method were generated in this study while the results for the other methods were previously generated.

Samples
First 19 The results for the BSG method that were generated in this study are as described in the Section 2.2, while the results for the cat, tet-l, aadD and pUB110 methods were previously generated [4,[8][9][10]12].

Development of the BSG Method
Among the 37 GM bacterial strains that have been reported as being used by the food and feed industry to produce enzymes, additives and flavoring, 75.6% of them belong to the B. licheniformis and B. subtilis species [6,31]. These two bacterial species belong to the B. subtilis group, which is composed in total of 10 closely related Bacillus species, including B. amyloliquefaciens, B. pumilus and B. velezensis (which are also used to produce microbial fermentation products), as well as B. atrophaeus, B. mojavensis, B. sonorensis, B. tequilensis and B. vallismortis. These Bacillus species present, at minimum, 99% similarity at the 16S rRNA sequence level as well as very few or no phenotypic and biochemical distinctions [28,32,33]. To date, all of the GMM contaminations that have been observed on the food and feed market have concerned bacterial species belonging to the B. subtilis group, being a vitamin B2-producing GM B. subtilis strain, a protease-producing GM B. velezensis strain, a alpha-amylase-producing GM B. amyloliquefaciens strain and a protease-producing GM B. amyloliquefaciens strain [2,3,[8][9][10][11]13,38]. The presence of such bacterial species represents, thus, a strong indicator of the potential GMM contaminations in a given sample. Therefore, a real-time PCR method targeting bacterial species from the B. subtilis group was developed and designed in this study to be compatible with the first-line screening condition of the GMM detection strategy, allowing the strengthening of the latter.
To this end, this taxon-specific real-time PCR method, named BSG, was designed in this study on a part of the 16S-23S region (Tables 1 and S1). This sequence of interest was previously identified [28,30] as being specific to the B. subtilis group, including the bacterial species that are commonly used to produce microbial fermentation products (B. subtilis, B. licheniformis, B. amyloliquefaciens, B. pumilus and B. velezensis) [4][5][6][7]25]. The 16S-23S region belongs to the bacterial ribosomal RNA operon, playing a key role in protein synthesis. The number of copies per genome of this ribosomal RNA operon varies between and within bacterial species, ranging from 6 to 15 for Bacillus sp. The developed BSG method, generating PCR amplicons with a size range from 137 to 165 bp, is the first realtime PCR method that was designed to specifically target the B. subtilis group in order to control the GMM contaminations in microbial fermentation products [28,30,32,33,39,40] (Tables 1 and S1).

Specificity Assessment of the BSG Method
The specificity of the developed BSG method was first confirmed in silico using SCREENED on a dataset that was composed of animal (75548 items), archaea (54464 items), bacteria (11507253 items), fungi (86250 items), plant (46015 items), protist (89957 items) and virus (10108 items) sequences that were extracted from the NCBI RefSeq Nucleotide database, allowing the present study to theoretically determine the generation of a PCR amplification (Tables S1 and S2) [34,35]. On this basis, a PCR amplification was only predicted for 167 bacterial sequences from the Bacillus genus. Among these 167 sequences, 15 (Tables S1 and S2).
Second, the specificity of the BSG method was experimentally assessed. As positive controls, 26 bacterial strains from 5 Bacillus species belonging to the B. subtilis group (B.  amyloliquefaciens, B. licheniformis, B. pumilus, B. subtilis and B. velezensis) were used. Among these 26 bacterial strains, 2 GMM strains were included, namely a vitamin B2-producing GM B. subtilis strain (RASFF2014.1249) and a protease-producing GM B. velezensis strain (RASFF2019.3332). As negative controls, 29 bacterial species from 17 non-Bacillus genera (Arthrobacter, Cellulosimicrobium, Corynebacterium, Enterococcus, Escherichia, Geobacillus, Klebsiella, Lactobacillus, Lactococcus, Leuconostoc, Microbacterium, Paenibacillus, Protaminobacter, Pseudomonas, Pullulanibacillus, Streptomyces and Streptoverticillium), and 9 Bacillus species not belonging to the B. subtilis group (B. brevis, B. cereus, B. circulans, B. coagulans, B.  firmus, B. flexus, B. lentus, B. megaterium and B. smithii) were tested. In addition to the bacterial strains, 37 fungal species as well as plant and animal samples were used as negative controls ( Table 2). As expected, no amplification was observed for all of the negative controls while all of the positive controls presented a positive signal. Although all of the positive controls were tested at the same DNA concentration, a variation in their Cq values was observed. This could be explained by the natural copy number variation of the 16S-23S region that was targeted by the developed BSG method [28,32,33].
Based on these results, the developed BSG method was assessed as specific because, firstly, a positive signal was only observed in the samples that were presenting the targeted sequences and, secondly, no false positive and negative signals were observed. This method currently represents the first taxon-specific real-time PCR method targeting GMM contaminations in microbial fermentation products for which a large spectrum of targeted and non-targeted sequences that are related to microorganisms that are used to produce fermentation products were investigated.

Sensitivity Assessment of the BSG Method
In order to assess the sensitivity of the developed BSG method, DNA from a control plasmid that was artificially synthetized to carry one copy of the targeted sequence from B. subtilis was used to prepare the serial dilution points, from 100 to 0.1 estimated target copy number (Tables 3 and S3).
A positive signal was observed as low as 20 estimated target copies for all 12 of the replicates. Moreover, a positive signal was detected as low as 1 estimated target copy. Based on these results, the LOD95% was subsequently determined at 22 estimated target copies, indicating that the developed BSG method is sensitive. This taxon-specific method represents the first real-time PCR method that is designed to specifically target the B. subtilis group with performance complying with the Minimum Performance Requirements for Analytical Methods of GMO Testing of the European Network of GMO Laboratories, being used as a standard in GMO enforcement laboratories, allowing to support enforcement laboratories for the control of GMM contaminations in microbial fermentation products [24,28,30,39,40].

Applicability Assessment of the BSG Method
The applicability of the developed BSG method was tested using several commercial microbial fermentation products that are intended for use in various sectors, such as livestock farming, baking, distillery, brewing, diary processing, protein processing or fruit processing (Table 4).
Among the 25 investigated samples, 18 of them were previously notified for a contamination with the vitamin B2-producting GM B. subtilis strain (samples n°1-2), the protease-producing GM B. velezensis strain (samples n°3-16), the alpha-amylase-producing GM B. amyloliquefaciens strain (samples n°3-17) or the protease-producing GM B. amyloliquefaciens strain (sample n°18). As expected, all of these samples presented a positive signal for the BSG method, highlighting the possibility of using this taxon-specific method as an indicator of potential GMM contaminations (Table 4).
Regarding the 7 remaining samples (n°19-25), no GMM contamination was previously reported based on the negative signals for the second-line methods that were targeting the currently known GM strains (Table 4). Among these 7 samples, a contamination with bacterial DNA was previously observed for the samples n°19-22 (a positive signal for the 16S-rRNA marker) and discarded for the samples n°23-25 (a negative signal for the 16S-rRNA marker) ( Table 4). Using the BSG method, a positive signal was detected for the sample n°19 while the samples n°20-25 presented a negative signal (Table 4). On this basis, the previously reported bacterial DNA contamination was associated to bacterial species that are related to the B. subtilis group, as in the case of the sample n°19, or not, as in the case of the samples n°20-22. In addition, given that no bacterial DNA contamination was previously observed in the samples n°23-25, the negative signal for the BSG method was expected. According to all of these results, the applicability of the developed BSG method was confirmed.

Illustration of the Reinforcement of the First-Line Screening Analysis Step Using the BSG Method
The BSG method was developed in order to strengthen the GMM detection strategy by completing the set of first-line screening methods targeting the key sequences that are frequently found in GM bacteria. To illustrate this reinforcement, the presence or absence of PCR amplification that was observed for the full set of first-line screening real-time PCR methods (BSG, cat, tet-L, aadD, pUB110) were indicated for few commercial microbial fermentation products (samples n°2-4, 6, 19-20, 22, 25) ( Table 5).
For the samples n°2-4 and 6, positive signals for the cat, tet-L, aadD and/or pUB110 methods as well as for the BSG method were observed, indicating potential contaminations with GM bacteria belonging to B. subtilis group. This hypothesis was confirmed by the previous detection of the vitamin B2-producing GM B. subtilis strain (sample n°2), the protease-producing GM B. velezensis strain (samples n°3-4, 6) and the alpha-amylase-producing GM B. amyloliquefaciens (samples n°3-4, 6) ( Table 5). For the sample n°19, a negative signal was obtained for all of the methods that were targeting key genetic elements that are found in GM bacteria while the BSG method presented a positive signal, suggesting a contamination with bacteria from the B. subtilis group (Table 5). Regarding the sample n°20, the positive signal for the aadD method associated to the negative signal for the BSG method suggests a potential contamination that is not related to bacteria belonging to the B. subtilis group, including bacteria naturally carrying the detected AMR gene (e.g., Staphylococcus sp.) or a new unknown GMM (Table 5). Finally, for the samples n°22 and 25, only negative signals were observed for all of the screening markers, discarding contamination with bacteria from B. subtilis group as well as bacteria naturally carrying the targeted AMR genes and shuttle vectors (e.g., Enterococcus sp., Staphylococcus sp. or Streptococcus sp.) ( Table 5) [5][6][7][8][9]. Based on all of these results, the added value of the developed BSG method to strengthen the first-line screening analysis step was thus highlighted, allowing the indication of potential contaminations with bacteria from the B. subtilis group, GM or not.
The BSG method was developed in order to be used as an indicator of the potential presence of GMM contaminations. Similar to its application to GM plants, this taxon-specific BSG method can also be applied to specifically target GM Bacillus strains, supporting the identification of such GMM. However, as the targeted 16S-23S region is naturally present in several copies per bacterial genome, ranging from 6 to 15 copies for Bacillus species, another target will need to be selected in case a taxon-specific method is required in the future for the quantification of GMM [24,28,30,39,40]. In addition, given the high copy number of the targeted 16S-23S region that is naturally present in bacteria, potential issues for the interpretation of the results may be encountered in the case of GMM contaminations at trace level (e.g., a late positive signal for the BSG method and a negative signal for the method specifically targeting a GM Bacillus strain). It should also be mentioned that the sequences that are targeted by the other screening methods (cat, tet-L, aadD and pUB110) may also be present in a GMM at a high copy number. Indeed, these sequences were previously found as being carried by plasmids or introduced multiple times in the host genome [2][3][4][5][6][7][8][9][10][11][12][13]38]. Consequently, in case of late signals from screening methods, the need for further investigations will depend on the nature of the contaminated products as well as the final decision of the competent authorities.

Conclusions
In order to control GMO, including GMM, in the food and feed chain, an efficient screening tool covering a large spectrum of key targets is essential to allow the identification of samples with potential GMO contaminations. To this end, a combination of taxonspecific and genetic element-specific methods are needed.
Regarding the control of GM bacteria, only screening methods that were targeting key genetic elements that are commonly found in GMM were only available and no taxonspecific method, designed to be compatible with the technical requirements of the GMM detection strategy as well as harboring performance in line with the Minimum Performance Requirements for Analytical Methods of GMO Testing of the European Network of GMO Laboratories used as a standard in GMO enforcement laboratories, was available to target GMM contaminations.
In order to meet this need, the BSG method was therefore developed and validated in this sense in the present study. This taxon-specific real-time PCR method was designed to target bacterial species belonging to the B. subtilis group, covering most of the GM bacterial species that are used by the food industry to produce microbial fermentation products [3,6,[8][9][10][11]13,24,31]. The performance of the BSG method, successfully assessed as specific and sensitive, complies with the requirements of the GMO enforcement laboratories. In addition, the BSG method also represents an interesting tool for screening viable B. subtilis group strains isolable from the food-feed matrix.
By completing the set of screening methods of the first-line screening analysis step, the BSG method strengthens the GMM detection strategy that is used to control the GMM contaminations in the food and feed chain. Indeed, the sequences that are targeted by the cat, tet-L, aadD and pUB110 methods are usually not naturally carried by the bacterial species from the B. subtilis group [5][6][7][8][9]. Therefore, when combined with a positive signal for at least one of the cat, tet-L, aadD and pUB110 methods, the detection of a positive signal for the BSG method is considered as a strong indicator of potential GM bacterial contaminations, allowing the identification of suspicious samples for further analysis.
Supplementary Materials: The following are available online at www.mdpi.com/article/10.3390/fermentation8020078/s1, Table S1: In silico analysis for the specificity of the BSG method using SCREENED, Table S2: Misclassification of NZ_CLNG01000048.1, Table S3: Calculation of LOD95% based on the POD curve for the developed real-time PCR BSG method.