Multiplex Detection of 24 Staphylococcal Enterotoxins in Culture Supernatant Using Liquid Chromatography Coupled to High-Resolution Mass Spectrometry

Staphylococcal food poisoning outbreaks are caused by the ingestion of food contaminated with staphylococcal enterotoxins (SEs). Among the 27 SEs described in the literature to date, only a few can be detected using immuno-enzymatic-based methods that are strongly dependent on the availability of antibodies. Liquid chromatography, coupled to high-resolution mass spectrometry (LC-HRMS), has, therefore, been put forward as a relevant complementary method, but only for the detection of a limited number of enterotoxins. In this work, LC-HRMS was developed for the detection and quantification of 24 SEs. A database of 93 specific signature peptides and LC-HRMS parameters was optimized using sequences from 24 SEs, including their 162 variants. A label-free quantification protocol was established to overcome the absence of calibration standards. The LC-HRMS method showed high performance in terms of specificity, sensitivity, and accuracy when applied to 49 enterotoxin-producing strains. SE concentrations measured depended on both SE type and the coagulase-positive staphylococci (CPS) strain. This study indicates that LC-MS is a relevant alternative and complementary tool to ELISA methods. The advantages of LC-MS clearly lie in both the multiplex analysis of a large number of SEs, and the automated analysis of a high number of samples.


Introduction
Staphylococcus enterotoxins (SEs) are potent gastrointestinal exotoxins produced by coagulase-positive staphylococci (CPS), including Staphylococcus aureus (S. aureus). They are resistant to harsh conditions such as heat treatment and low pH that easily destroy CPS, as well as to proteolytic enzymes, and therefore retain their activity in the digestive tract after ingestion [1]. Consumption of food contaminated by a few nanograms of SEs rapidly and abruptly induces symptoms such as nausea, copious vomiting, abdominal cramps, and diarrhea. Reported dose levels of SEs involved in foodborne outbreaks are scarce and variable. The Benchmark Dose lower limit (BMD10) was estimated to be 6.1 ng for SEA type [2]. Since 2011, the annual zoonosis reports of the European Food Safety Authority This approach has been used in different studies to analyze the cytoplasm of various mutants of Staphylococcus aureus and Bacillus subtilis, allowing for the absolute quantification of several proteins of the cytosolic proteome [40,41].
In this study, a label and antibody-free LC-MS method based on the bottom-up proteomics approach was optimized for the multiplex quantification of 24 SE types in medium culture, i.e., SEA, SEB, SEC, SED, SEE, SEG, SEH, SEI, SElJ, SEK, SEL, SEM, SEN, SEO, SEP, SEQ, SER, SES, SET, SElU, SElX, SEY, SElZ, and TSST1. A signature peptide database was established for 24 SE types, considering 162 sequence variants. The LC-MS method was tested on 24 SEs produced in medium culture by a panel of 49 representative CPS strains previously characterized by a WGS tool for their content of se genes. The LC-MS method was evaluated in terms of specificity, sensitivity, and accuracy. Finally, SE quantification is essential to study ES toxicity and determine the emetic dose as Benchmark Dose (BMD50). In this work, SE quantification was performed by the label-free Hi3 method to overcome the absence of SE standard solutions. The types of SEs and the concentrations secreted by each CPS strain were estimated, showing the ability of each strain to simultaneously produce several types of SEs at different concentration levels, depending on the SE type.

Se Gene Characterization
Reference CPS strains, types of se genes, and their variants are presented in Table 1. More detailed information on CPS strains, such as origin, date of isolation, and type of food from which the strain was isolated, are reported in Supplementary Materials Table S1.
Among the 49 CPS strains, 42 (86%) were isolated from food samples, most of which were involved in food poisoning outbreaks, indicating their ability to express SEs. They were sequenced and analyzed using the NAuRA [16] genomic tool. The results showed that each CPS strain harbors between two and 11 se genes. The 11 se genes contain between two and five genetic variants: sea (4), seb (4), sed (4), see (3), seh (4), selj (4), sep (3), seq (5), ser (4), ses (2), and selz (4). Eight enterotoxin genes contain between six and ten variants: seg (6), sec (10), sek (8), sem (9), sen (9), seu (10), sey (6), and tsst1 (6). According to gene sequencing data, the four se enterotoxin genes sel, sei, seo and selx contain 11, 12, 13 and 20 variants, respectively. Finally, the set type has no variants. The sev, sew, sel26 and sel27 genes were not detected in the strains analyzed in this work. This may be explained by the rarity of these genes in the CPS strains of this study that were isolated from food samples [16]. In total, the 49 CPS strains that were used to prepare naturally SE-contaminated medium samples harbored 24 se genes and their 162 variants.

Establishment of a Signature Peptides Database for Each Enterotoxin Type
The brain heart infusion (BHI) broth samples contaminated by SEs were prepared for each CPS strain and were then submitted to the acid precipitation protocol, tryptic digestion, and LC-MS analysis, as described in Figure 1.
SEs peptides released by trypsin in the contaminated BHI samples were analyzed by LC-MS in data-dependent acquisition (DDA) mode (non-targeted analysis) and were processed by the X!Tandem Pipeline 0.4.3 [42] using both genomic sequences obtained by NAuRA genomic software [16] and DDA data. This step is used to determine all peptides and their corresponding retention times (RT). In total, for the 24 SEs and their variants, a preliminary list of 207 peptides was established (Supplementary Materials Table S2). Between four and 13 peptides were detected for each SE type: SEA (12), SEB (12), SEC (14), SED (9), SEE (8), SEG (7), SEH (10), SEI (10), SElJ (6), SEK (10), SEL (8), SEM (7), SEN (6), SEO (8), SEP (5), SEQ (13), SER (11), SES (7), SET (9), SElU (4), SElX (9), SEY (8), SElZ (5) and TSST1 (9). These 207 peptides were fragmented with three different collision energies: 15, 18, and 21% CE. The best collision energy was selected on the extracted-ion chromatogram (XIC) of the three most intense fragments for each peptide. Fragment selection was performed using Skyline version 21.1 software (https://skyline.ms/; accessed date 25 June 2021); only the y ions were analyzed. The fragmentation energies and the selected fragments for further quantification are shown in Supplementary Materials Table S2. An example of optimized collision energy for SEA is shown in Table 2. These 207 peptides were analyzed in parallel reaction monitoring (PRM) mode for identification of the best responder peptides, as described in   [29]. Among the best peptides, the most intense and non-interfered y-ion fragments were selected for peptide detection and quantification. At least three specific peptides were selected for each type of enterotoxin, considering SEs variants. Therefore, 93 specific signature peptides covering the 24 SEs and their variants were selected and included in a signature peptide database. Enterotoxins (and their number of specific peptides) are: SEA (5) ). An example of the optimization and identification of specific peptides for enterotoxin SEA is described in Table 2. Only one common specific peptide (QNTVPLETVK, m/z = 564.8166) was detected in the four variants. To avoid negative deviation, this result highlights the importance of optimizing several specific peptides for each type of SE, considering sequence variants. SEs peptides released by trypsin in the contaminated BHI samples were analyzed by LC-MS in data-dependent acquisition (DDA) mode (non-targeted analysis) and were processed by the X!Tandem Pipeline 0.4.3 [42] using both genomic sequences obtained by NAuRA genomic software [16] and DDA data. This step is used to determine all peptides and their corresponding retention times (RT). In total, for the 24 SEs and their variants, a preliminary list of 207 peptides was established (Supplementary Materials Table S2). Between four and 13 peptides were detected for each SE type: SEA (12), SEB (12), SEC (14), SED (9), SEE (8), SEG (7), SEH (10), SEI (10), SElJ (6), SEK (10), SEL (8), SEM (7), SEN (6), SEO (8), SEP (5), SEQ (13), SER (11), SES (7), SET (9), SElU (4), SElX (9), SEY (8), SElZ (5) and TSST1 (9). These 207 peptides were fragmented with three different collision energies: 15, 18, and 21% CE. The best collision energy was selected on the extracted-ion chromatogram (XIC) of the three most intense fragments for each peptide. Fragment selection was performed using Skyline version 21.1 software (https://skyline.ms/; accessed date 25 June 2021); only the y ions were analyzed. The All peptides reported for SEA to SED in the literature were also detected in the list of 93 peptides [28,[43][44][45][46]. For enterotoxins SEA, SEB and SED, the authors determined between two and four specific peptides that we confirmed in this study. Muratovic et al. 2015 [43] reported a set of eight peptides for SEA and eight for SEB. These SEB peptides were identified in our set. For SEA, six of the eight peptides were available in our set.
As a result, the established list of 93 specific signature peptides can be considered as a reference database for further work on the detection of SEs using an LC-MS method, and could be used for further applications in food matrices.

Assessment of Qualitative Performance
The PRM method implemented with the peptide database was selected for further evaluation, considering the higher sensitivity of the targeted acquisition mode [47,48]. The sensitivity, specificity, and accuracy of the developped LC-MS method were calculated for each type of SE based on an analysis of the contaminated BHI samples. The qualitative performances obtained for each type of SE are reported in Table 3.

LC-MS Specificity
Specificity describes the ability to obtain a negative result for a sample that does not contain SEs, regardless of the SE type and the analyzed variant (blank sample). In this work, specificity was determined by comparing the signal detected by PRM with the WGS results. Therefore, the absence of a PRM signal was expected when the se gene was not detected by the genomic tool. This was calculated as follows: Number of samples correctly analyzed as negative for type x Number of blank BHI samples × 100 where the number of blank BHI samples is the culture samples prepared with a CPS strain negative for the targeted se gene. For example, when the SEA type is analyzed, the blank sample must be negative for the sea gene even if it contains other types of se gene. The calculate specificity was 100%, irrespective of the type of the SE, meaning the absence of positive deviation. This highlighted the absence of an interference between different enterotoxin types. In fact, the obtained results were negative for the targeted enterotoxin despite the presence of other SE types.

LC-MS Sensitivity
Sensitivity describes the ability to detect the correct type of SE for a sample that contains the analyzed SE type. In this work, sensitivity was calculated assuming that all se genes detected by the genomic tool expressed the corresponding SEs in BHI supernatant. For example, if the CPS strain added to and incubated in the sample contains the sea gene, the enterotoxin SEA must be detected by LC-MS, even if this CPS is able to produce several types of SE. Therefore, sensitivity for the LC-MS method is calculated as follows: Number of SE types correctly detected Number of BHI samples containing the type of se targeted × 100 The LC-MS method had a sensitivity of 100% for 20 SEs. However, for the SEL, SEN, SEO and SElU types, method sensitivity was 93.8% (15/16), 76.0% (19/25), 88.5% (23/26) and 90.0% (18/20), respectively. For SEN, SEO and SElU, the observed negative deviations could be due to the very low amount of enterotoxins expressed by egc operon. In fact, this operon harbors six types of SEs: SEG, SEI, SEM, SEN, SEO, and SElU ( Table 4). The sensitivity of the method was 100% for types SEG, SEI and SEM. The measured concentrations were the lowest and were close to the limit of detection (1 ng/mL), Table 5. This observation was reported by Schwendimann et al. 2021 [49] and confirmed the expression of egc enterotoxins at the ng/mL level. Finally, for type SEL, only one negative deviation was observed (15/16), which could be due to lack of expression of the sel gene by this strain or to the very low amount produced in the BHI.
It is also possible to evaluate overall sensitivity for the 24 SEs. In fact, in the 49 analyzed BHI samples, 335 enterotoxins were expected to be detected, assuming that all se genes expressed their corresponding SEs in the BHI supernatant. The LC-MS method detected 323 SEs among them, providing an overall sensitivity of 96.3%.
The sensitivity values that were obtained in our study could be considered satisfactory, considering the multiplex detection of 24 SE types and the bias that can be introduced by the non-expression of se genes.  1  3  1  15  3  12CEB496  3  3  3  16  12CEB512  5  6  4  5  3  4  10  1  13CEB193  2  2  2  7  2  13CEB437  3  3  3  10  3  3  3  6   14A/FRI361  3  1  3  3  1  5  8  3  3  1  5  14SBCL1004  3  10  4  5  14SBCL770  1  4  10  5  4  4  1  3  15SBCL1151  1  3  4  7  4  3  3  5  9  15SBCL1428  3  4  4  3  3  4   (1) fragments selected for detection. (2) peptide was detected in variant (number of variant is available in Table 1). (3) peptide was not detected in variant (number of variant is available in Table 1). (4) peptide detected in another type of SEs. (5) five specific signature peptides were selected for the detection of type SEA, taking into account the four variants.   (%)  100  100  100  100  100  100  100  100  100  100  100  100  100  100  100  100  100  100  100  100  100  100  100  100   where the number of SE types that were correctly detected or not detected is the sum of the number of correct positive results when SE was expected to be "detected" and the number of correct negative results when the SE was expected to be "not detected", and total number of analyses corresponds to the sum of the number of strains containing the targeted se genes and number of strains negative for the targeted se gene (see Table 3). The overall accuracy of the LC-MS method can be calculated for the multiplex detection of 24 SEs. Therefore, 1164 correct analyses were obtained in the 49 BHI samples over the 1176 analyses. Therefore, the overall accuracy of the multiplex LC-MS method was 99.0%.
Qualitative performance criteria highlighted the high specificity of the optimized LC-MS method dedicated to the detection of 24 SEs. Specificity, sensitivity, and accuracy values can be considered satisfactory regarding the large number of targeted enterotoxins. In the literature, these qualitative performance criteria were mainly reported for ELISA methods and for a very limited number of SEs. Nia et al. 2021 [18] calculated performance criteria for the standard EN ISO 19020 for five classical enterotoxins, SEA, SEB, SEC, SED and SEE, in five categories of food matrices. Specificity varied between 90 and 100%, and sensitivity between 87 and 100%, for the two most efficient immuno-enzymatic assays. Féraudet-Tarisse et al. 2021 [27] developed immuno-enzymatic tests for enterotoxin types SEA, SEG, SEH, and SEI in medium and evaluated specificity on only eight CPS strains. The results showed an absence of cross-reactions and were not interpreted in terms of specificity, sensitivity, and accuracy. For LC-MS methods, specificity and sensitivity were only evaluated in the case of three types of SEs (SEA, SEB, and SED). In this work, performance criteria were calculated using data obtained from a large set of 24 SEs and their variants naturally produced in medium. This work is the first comprehensive evaluation of LC-MS performance for the detection of 24 SEs in medium culture. Table 5. SE concentrations measured in the 49 contaminated BHI samples using the LC-MS method, the signature peptides database and the Hi3 procedure. LOQ = 1 ng/mL; (1) blank box: se gene not detected by genomic analysis and SEs not detected by LC-MS; (2) <1: below LoQ; (3) ND: se gene detected by genomic analysis but corresponding SE not detected by LC-MS.

SEs Quantification in BHI Culture Supernatant
Only five purified SE standards are commercially available (SEA, SEB, SEC, SED, and SEE). To estimate the concentration levels of the 24 SEs produced in the BHI and to overcome the absence of SE standard solutions, an unlabeled quantification method, known as the "Hi3" procedure, was used in accordance with Silva et al. 2006 [39]. In brief, protein concentration can be estimated from the MS signal of the sum of the three most reactive peptides of each protein compared to the best three peptides of a standard at a known concentration [40,41].
The calibration curve, established according to the Hi3 procedure, was tested individually for each type of SE for which standard solutions were available, over the 12 concentrations, in duplicate. Then, biases were calculated and are reported in Table 4 and Figure 2.
The R 2 of each curve is above 0.95 and the CVs of replicates were below 20% overall. Calculated bias can be interpreted as follows: In conclusion, the calibration curve obtained according to the Hi3 procedure can be considered representative of the four SE types: SEA, SEB, SED, and SEE. For SEC, the calibration curve can be used semi-quantitatively due to the considerable bias observed.
The limit of quantification (LOQ) was determined for the five SEs, based on bias below 20%. For each SE, we observed at least two best peptides showing an LOQ at 40 fMol, corresponding to 1 ng/mL.
In the absence of a standard solution for the other 19 SEs studied in this work, it is difficult to extrapolate the observations obtained for the 5 SEs (SEA to SEE) to the 19 SEs. However, we assume that the standard curve obtained according to Hi3 can be implemented to obtain a semi-quantitative SEs concentration in BHI medium.

Application of the LC-MS Method for Multiplex Analysis of 24 Enterotoxins Produced by CPS Strains in Culture Supernatant
SEs were analyzed in the 49 BHI samples prepared for each CPS strain by the LC-MS method developed using the signature peptides database and the Hi3 calibration curve implemented in this study. Measured concentrations and median obtained for each type of SE are reported in Table 5. In conclusion, the calibration curve obtained according to the Hi3 procedure can be considered representative of the four SE types: SEA, SEB, SED, and SEE. For SEC, the calibration curve can be used semi-quantitatively due to the considerable bias observed.
The limit of quantification (LOQ) was determined for the five SEs, based on bias below 20%. For each SE, we observed at least two best peptides showing an LOQ at 40 fMol, corresponding to 1 ng/mL. The obtained results showed that individual strains were able to produce between 2 and 11 SEs. Measured concentrations varied between LoQ (<1 ng/g) and 21,943.4 ng/g. Among the 335 enterotoxins that were expected to be produced by strains in supernatant, 323 enterotoxins (96%) were detected by LC-MS. Among them, 215 (67%) were quantified and 108 (33%) were detected below the LoQ (<1 ng/g) (Table 5). However, concentration levels depend strongly on the type of SEs. Therefore, four SE groups can be distinguished depending on their concentration expressed by CPS strains: -Very low level: SEG, SEI, SEN, SEM, SEO and SElU, expressed by the genomic cluster egc, were detected at very low levels compared to other SEs. Their concentrations were below 1 ng/g except for SEM (median 2.4 ng/g, n = 24). Also, SElJ was produced at very low concentrations. -Low level: the six enterotoxins SEL, SEP, SES, SElX, SEY and SElZ were weakly produced with a median between 5.0 ng/g (SEY, n = 10) and 11.6 ng/g (SElX, n = 42).
It should be noted that type SES was analyzed only twice, at 6 and 18 ng/g. In this group, the concentration measured also depended on the strain. -Medium level: SEA, SED, SEE, SEH, SEK, SEQ, SER, SET, and TSST1 were produced at relatively similar concentrations, with a median between 29.6 ng/g (SER, n = 9) and 60.5 ng/g (SEA, n = 14). SET was very rarely detected and was analyzed in only one The total amounts of SE produced by each strain are reported in Table 5. The lowest value was 3.4 ng/g and the highest 22,103 ng/g. Generally, strains containing seb or sec types were the most toxin-producing of this types of SE. However, strains containing only egc operon se genes seemed to produce the least enterotoxins. This was observed by several authors who concluded that egc enterotoxins were less expressed than the classical enterotoxins [27,[49][50][51].
The five types, SEH, SEK, SEQ, SER, and SET, exhibited comparable concentration levels to the classical enterotoxins SEA, SED, and SEE. Type SEH was reported only once as being involved in a food poisoning outbreak due to the consumption of contaminated mashed potatoes. However, SEK, SEQ, SER, and SET were never detected in the context of food poisoning outbreak investigations due to the lack of methods targeting these enterotoxins [52]. Therefore, the present work showed that SEK, SEQ, SER, and SET should be studied for their toxicity and their possible involvement in SFPO.

Conclusions
This work highlights the usefulness of LC-MS methods for SEs multiplex analysis without the need for specific reagents, as is the case for immuno-enzymatic assays. A highly specific and sensitive LC-MS method was developed for the multiplex detection of 24 SE types in culture medium. The protocol was based on acid precipitation and enzymatic digestion followed by specific peptide detection. To enhance the accuracy of the method, a database of 93 signature peptides was optimized using the complete sequences of the 24 SEs, including their 162 variants.
The LC-MS method was successfully tested on 49 naturally contaminated BHI samples. The Hi3 procedure was adopted to overcome the absence of SE standard solutions for 19 SEs. The results showed the ability of CPS strains to simultaneously produce between 2 and 11 types of SEs. However, the concentrations expressed depended strongly on the type of SE. The lowest concentrations were obtained for enterotoxins SEG, SEI, SEM, SEN, SEO, and SElU expressed by egc operon (<1 ng/g). SEB and SEC were expressed at very high concentrations and reached the µg/g level.
The present study indicates that the newly described SEs should be taken into consideration during foodborne outbreak investigations, as CPS strains are able to simultaneously express several SEs, including these new SEs. LC-MS is a promising method for the multiplex detection of SEs and deserves further development to be applied to food.
Highly purified, freeze-dried SEs were purchased from Toxin Technology, Sarasota, FL, USA (batch 120794A for SEA, 61499B1 for SEB, 113094C2 for SEC, 12802D for SED, and 70595E for SEE) and were rehydrated according to the manufacturer's instructions to obtain stock solutions. In brief, 1 mL of osmosed water was added to 1 mg of SE powder to obtain a theoretical concentration equal to 1 mg/mL.  Table S2).
Safety Precaution: due to the high toxicity of SE, experiments were performed in accordance with safety rules for the handling of toxic substances and using personal protection equipment. SE-contaminated solutions and consumables were inactivated overnight with 2 M NaOH before elimination.

CPS Strains and Enterotoxin Sequences
A total of 49 CPS strains were used in this study: 6 reference strains obtained from the Institut Pasteur collection (France) and 43 CPS strains from the ANSES collection, including 40 strains isolated from different types of foods involved in staphylococcus food poisoning outbreaks. Three strains were isolated in the factory to check the contamination of the milk or cheese by CPS or SEs; this is called the own check. The origins of the strains are detailed in Supplementary Materials Table S1.
The enterotoxin gene profiles were analyzed for their staphylococcal enterotoxin genes using the genomic tool NAuRA (https://github.com/afelten-Anses/NAuRA, accessed date 25 June 2021). Screening of the enterotoxins was performed using the gene sequence and the relative protein sequence of the previously described 27 SEs and the estimated parameters of Blast in Merda et al. Sequence alignment is detailed in Supplementary Materials Figure S1.

SE Expression in Culture Supernatant and Sample Preparation
Strains were incubated for 24 h at 37 • C on milk plate count agar. Colonies were then transferred to 40 mL of BHI at 37 • C for 18 h and centrifuged at 3000× g for 15 min to discard the bacterial pellet. The supernatant was filtered through a 0.2 µm filter before SE analysis, as in Lefebvre et al. 2021 [29,53]. A total of 1.2 mL of the filtered supernatant was centrifuged at 15,000× g for 10 min to discard residues. Then, 1 mL was recovered and mixed with 50 µL of TCA 100% for 5 min, and toxins were precipitated by centrifugation at 15,000× g for 10 min. The supernatant was removed, and the protein pellet was resuspended with 20 µL of RapiGest 0.05% in AB500 (Ammonium bicarbonate 500 mM). 10 µL of 20 mM DTT were added and heated for 15 min at 95 • C to induce protein denaturation. Then, 10 µL of 45 mM IAA was added and incubated for 30 min in the dark at room temperature. Digestion was performed with 1 µg trypsin for 3 h at 37 • C. After incubation, digestion was stopped with 5 µL of 1 M HCl and 10 µL of the mixture of labeled peptides was added to the samples. A total of 10 µL were sampled for LC-MS analysis.

Liquid Chromatography−Mass Spectrometry Analysis
LC-MS analysis was performed using an Ultimate 3000 chromatography system coupled to an Q-Exactive (Quadrupole-Orbitrap) mass spectrometer (Thermo Fisher Scientific, Courtaboeuf, France). Peptide separation was performed on an Aeris peptide XB-C18 reverse phase column (150 mm × 2.1 mm; 1.7 µm; 100 Å; Phenomex, Torrance, CA, USA) at 60 • C for 30 min. The mobile phases were prepared in water with 0.1% formic acid (A) and ACN with 0.1% formic acid (B). Following an isocratic step with gradient from 5 to 40% phase B in 20 min was carried out at a mobile phase flow rate of 0.5 mL/min, before equilibration for 5 min. Eluted peptides were introduced into the Q-Exactive instrument by an electrospray ionization (ESI) source. Data were acquired in positive ion mode, capillary voltage was set at 4 kV, and temperature at 320 • C.
DDA mode (Top 5) was performed with a resolution of 70,000 (Full MS) or 17,500 (ddMS2) at m/z 200 (FWHM), and normalized collision energy was fixed to 18%. The automatic gain control (AGC) target was constantly set at 1 × 10 6 , fill time was adjusted at 128 ms. For PRM mode, the isolation mass window was fixed at 1.2 Da. Collision energy was optimized in the HCD for each peptide, as indicated in Section 2.2 and shown in Supplementary Materials Table S2. The AGC target was constantly set at 1 × 10 6 , fill time was adjusted at 128 ms, and Orbitrap resolution set at 35,000 at m/z 200 (FWHM) to reach at least 10 acquisition points per chromatographic peak.