The association of bacterial C9-based TTX-like compounds with Prorocentrum minimum opens new uncertainties about shellfish seafood safety

In 2012, Tetrodotoxin (TTX) was identified in mussels and linked to the presence of Prorocentrum minimum (P. minimum) in Greece. The connexion between TTX and P. minimum was further studied in this paper. First, the presence of TTX-producer bacteria, Vibrio and Pseudomonas spp, was confirmed in Greek mussels. In addition these samples showed high activity as inhibitors of sodium currents (INa). P. minimum was before associated with neurotoxic symptoms, however, the nature and structure of toxins produced by this dinoflagellate remains unknown. Three P. minimum strains, ccmp1529, ccmp2811 and ccmp2956, growing in different conditions of temperature, salinity and light were used to study the production of toxic compounds. Electrophysiological assays showed no effect of ccmp2811 strain on INa, while ccmp1529 and ccmp2956 strains were able to significantly reduce INa in the same way as TTX. In these samples two new compounds, m/z 265 and m/z 308, were identified and characterized by liquid chromatography tandem high-resolution mass spectrometry. Besides, two TTX-related bacteria, Roseobacter and Vibrio sp, were observed. These results show for the first time that P. minimum produce TTX-like compounds with a similar ion pattern and C9-base to TTX analogues and with the same effect on INa.

neurotoxic symptoms 7,8 . This unknown compound was able to block sodium and calcium channels and was accumulated by shellfish 7 . However, since some of the toxic episodes related to P. minimum were also associated with the presence of other toxic microalgae such as Dinophysis spp., its toxicity was discussed 1 . This coexistence does not eliminate the potential risk of P. minimum blooms for human health, being evident that the reported toxicity of this dinoflagellate was clone-related and also related to environmental conditions 1 . In this sense, although most P. minimum clones were reported as non-toxic, and rare human toxicity episodes have been attributed to this specie, it should be considered for risk assessment.
Tetrodotoxin (TTX) is one of the most important neurotoxins known to block sodium channels and thus it inhibits the propagation of action potentials in muscle and nerve cells. The minimum lethal dose for TTX is 8 μg/Kg and the median lethal dose is 10 μ g/Kg 9 . Symptoms can appear between 10 and 45 minutes after exposure, although depending of the amount of the toxin ingested some manifestations could appear 6 hours later. This compound blocks site 1 of voltage dependent sodium channels 10 and it has interesting pharmacological applications in the study of excitable membranes as well as therapeutic uses in treating migraines, addictions, or as an anaesthetic agent for pain 11 . TTX has been described in different aquatic animals such as fish, arthropods, echinoderms, molluscs, worms, newts, frogs or toads as well as several bacteria species and dinoflagellates 11,12 . However, the origin of TTX is still unclear, although the presence of this toxin in such a diverse group of animals suggests that bacteria might be as the primary source 13 . In this sense, TTX can be produced without bacteria, but the presence of microorganisms could, in some way, simplify TTX biosynthesis 11 .
To date, few appearances of TTX in bivalve molluscs have been reported 14,15 . Recently, TTX has been detected in mussels and Pacific oyster from the English coast (2013-2014) 16 . In addition in 2012, in Greece, during routine controls of shellfish, an unexplained toxicity associated with neurological symptoms was observed in a series of mouse bioassays (MBA). This atypical toxicity coincided with the absence of known dinoflagellates and other toxins, while only P. minimum was present in seawater. After mass-spectrometry analysis, both TTX and TTX analogues were confirmed 17 . Therefore a potential TTX presence-dinoflagellate link was hypothesized. In this context the aim of this work was to further study the relationship between TTX and P. minimum blooms in order to identify and characterize the compounds produced by this dinoflagellate.

Results
To further study the occurrence and origin of TTXs in mussels collected in the unusual toxic episode in Greece, it was evaluated the presence of bacteria as TTX producers as well as the effect on Na v channels.
Firstly, the presence of bacteria was checked in contaminated mussel-samples, 1770/2012 and 1774/2012, from the Greek episode. Levels of 222.9 μ g/Kg and 206.3 μ g/Kg of TTXs respectively were reported when these samples were previously analysed by mass spectrometry 17 . Sequence reads were clustered at 97% identity resulting in 297 unique Operational Taxonomic Units (OTUs). A rarefaction plot of observed bacterial species in control (white non-contaminated tissue) and TTX-mussel samples showed higher species diversity in the uncontaminated material (Fig. 1A). Although high species diversity was achieved, uncontaminated sample was under represented when compared to TTX-mussel samples. Jackknifed hierarchical clustering by UPGMA (unweighted pair group method with arithmetic mean) showed a high support for the depurated samples clustering separate from the non-depurated one, while a lower support was observed for contaminated samples (Fig. 1B). In this way, the identified OTUs corresponded to 9 bacterial phyla (Supplementary Table 2 and Sub-Table 2). Higher percentages of proteobacterial OTUs, the phylum containing Pseudomonas, Pseudoalteromonas, and Vibrio, two known TTX producers, were detected in TTX-contaminated mussels. However these genera were not represented in the Proteobacteria of non-contaminated sample. To better identify Vibrio and Pseudomonas species, OTUs were re-picked with 100% homology. When compared to the GreenGenes 16 S RNA database 18 , several potential TTX  Table 3). The distribution of these bacteria showed that in both samples Vibrio parahaemolyticus and Vibrio alginolyticus were the dominant species, while Pseudomonas appeared at lower levels as well as other Vibrio species.
Next, to further support the existence of TTX in mussel extracts, the effect over Na v channels, the main cellular target of TTX, was checked by patch clamp measurements in stable transfected hNa v 1.6 HEK cells 19,20 . Aliquots of 2 g of DG of sample 1786/2012 collected in Greece were extracted as previously described 17 . As Fig. 2 shows, serial dilutions of the TTX-positive mussel extract 1786/2012 were tested (0.4, 2 and 4 μ l) and a dose response effect was reported. At the highest concentration tested, an inhibition of 54.65 ± 14.08% versus control current was observed ( Fig. 2A). In the same cells, the control of TTX standard elicited a dose response sodium current (I Na ) inhibition, IC 50 0.72 nM (95% confidence interval 0.43 to 1.2 nM) (Fig. 2B).
Although the source of TTXs in contaminated mussels seemed to be related to bacteria, the dinoflagellate P. minimum was present in water in high abundance when toxic mussels were harvested, therefore the presence of toxins and bacteria in this microalgae was studied. Three strains of P. minimum from different collection sites, ccmp1529 from Ecuador, temperature range 18 °C to 22 °C, ccmp2811 from Sarasota, Florida (USA), temperature range 18 °C to 22 °C, and ccmp2956 from Johor Strait, between Singapore and Malaysia, temperature range 22 °C to 28 °C, were cultured in L1 medium in different growth conditions, varying light intensity (785, 840 and 865), salinity (ranging from 31‰ to 37‰) and temperature (19 °C or 24 °C), with each condition being carried out in triplicate. After 15 days of growth, cells were harvested and the extracts obtained dissolved in acetic acid 0.03 M and analysed for the presence of active compounds with similar characteristics to TTXs. The same rate of growth was obtained for ccmp1529 strain either at 19 °C and 24 °C, slightly less growth was obtained when the ccmp2956 strain was grown at 19 °C compared to 24 °C and no growth was observed when the ccmp2811 strain was cultured at 24 °C. Therefore ccmp1529 and ccmp2956 strains were analysed both at 19 °C and 24 °C, while only cultures obtained at 19 °C were analysed in the case of ccmp2811. No significant effect, in terms of growth, was observed when light was modified and only ccmp2956 strain was able to grow at 37‰ of salinity. Then, the effect of dinoflagellate extracts on I Na was checked. Electrophysiological recordings showed that while ccmp1529 extracts did not produce any effect on I Na when dinaflagellates were grown at 19 °C, the addition of serial dilutions of 24 °C extract culture reveal a dose dependent current reduction with a maximal inhibition of 49.75 ± 8.04% versus control current (Fig. 3A). This effect was not dependent on light intensity. Likewise, the ccmp2956 strain cultured at 19 °C had no effect on I Na but when it was grown at 24 °C and 37‰ salinity, a maximal current inhibition of 47.65 ± 10.6% versus control was observed (Fig. 3B). When salinity was 31-33‰, I Na inhibition was dependent on light intensity, since no effect was observed with dinoflagellates cultured at 18 W/840, while 39.69 ± 12.03% inhibition was obtained at 18W/865. ccmp2811 extracts from cultures grown at 19 °C did not produce a significant I Na inhibition when consecutive dilutions were added (Fig. 3C). For all these assays TTX standard was used as control of I Na inhibition (Fig. 3D). Therefore, these results indicate the presence of some compound produced by P. minimum with a similar effect than TTX on I Na . The profile of compounds present in cultures grown at 19 °C or 24 °C seems to be different, at least with regard to potency over I Na inhibition.
Next, we proceeded to the identification of compounds produced by P. minimum cultures using mass spectrometry technology. The first screening of culture extracts was done by Multiple Reaction Monitoring (MRM) mode, searching for the most common TTX analogues described in the literature (Supplementary Table 1). In this way, 5 compounds with same product ions, but different retention times than TTXs standards were obtained, m/z 320 > 302, m/z 304 > 162, m/z 302 > 162, m/z 290 > 272, m/z 272 > 254. Then, by operating in the Product Ion Scan mode, the fragmentation pathway of these five compounds was checked following the pattern fragmentation observed in TTX, 4,9-  The five compounds initially observed in P. minimum cultures did not follow this fragmentation pathway (data not show), therefore these five compounds could not be identified as TTX analogues. However, since P. minimum extracts had some TTX activity, mass spectrometry operating procedure was changed from Product to Precursor Ion Scan Mode. Thus the search of compounds was addressed from the common product ion for all TTX analogues, m/z 162, to precursor ions. In this way, operating in the Precursor Ion Scan mode, the extract from ccmp2956 strain (24 °C and 37‰) with high inhibitory effect on I Na , was checked. In this sample two precursor ions, m/z 265 and m/z 308, were observed (  To characterize and predict the molecular formula of these two new compounds, m/z 308 and m/z 265, and their product-ions, a high-resolution mass spectrometry Ion Trap-Time of Flight (IT-TOF) was used. This technology allows us to predict the formula using ad oc predictor software based on the accurate mass data recorded. TTX standard was employed as control to predict the molecular formula of precursor and ion products. In this way, elemental composition, accurately measured mass/charge ratio, theoretical value m/z, and mass errors expressed in ppm and mDa for TTX, m/z 308 and m/z 265 were obtained after MS n spectra. These data are summarized in Table 1 Fig. 7). Since these compounds are present in low concentration in P. minimum extracts only MS 2 spectra were obtained. The errors between these measured product-ions and the theoretical values calculated are shown in Table 1. In this case the higher range observed was probably related to the low concentration. In addition the predicted structure proposed for each compound based on the high resolution masses of fragment ions are shown in Figs 5, 6 and 7. Next, the presence of this characterized compounds, m/z 308 and m/z 265, was checked in all P. minimum extracts, using [M + H-H 2 O] + ions for identification and m/z 162 ions for quantification. Since no standards of these compounds were available, results were obtained as peak area and also as ng of TTX equivalent/mL (see Supplementary Table 4). Both compounds were present in all P. minimum extracts. While the amount of compound 308 did not show a direct relation with the effect on I Na , the extracts with higher amount of compound 265 produced higher I Na inhibition.
Finally, the presence of TTX-producing bacteria was checked in P. minimum cultures. For this experiment, ccmp1529 strain was cultured both at 19 and 24 °C, ccmp2956 strain was cultured at 19 °C, and ccmp2811 cultures were obtained at 24 °C all 31-33‰ salinity. Firstly, we checked for the presence of bacteria described in contaminated mussels, Vibrio and Pseudomonas spp. P. minimum cultures were analysed by PCR using specific primers for Vibrio alginolyticus, Vibrio parahaemolyticus, Vibrio vulnificus, and Vibrio cholerae. After PCR amplification, V. alginolyticus was observed in ccmp2811 and ccmp1529 cultures, while none were detected in sample ccmp2956 (Supplementary Figure 2). When the same assay was performed using specific primers for Pseudomonas, no positive amplification product was detected (results not shown). In addition, P. minimum cultures were analysed for  Table 5).

Discussion
The presence of TTX in mussels is an important issue from both a toxicological and human-health risk standpoint. Although this compound is a regulated toxin in Europe and around the world, it is not included in regular monitoring programs because the consumption of species that usually accumulate the toxin is prohibited. Before 2014 only one highly toxic TTX-accumulation in scallops from Japan had been reported 14 . However, from 2014 to date three reports have identified the presence of TTX in shellfish and it was also detected in several edible gastropods [15][16][17]22,23 . Therefore, the presence of TTXs in shellfish as well as the source of these toxins should be taken into account, since TTX-contaminated samples will not be detected by official analytical methods, such as LC-MS/MS or HLPC with fluorescence detection, because TTX is not required to be included in these methods. In addition, the MBA, that could detect all toxins, is no longer used in Europe as an official assay. In this paper the presence of TTX in contaminated samples collected in Greece was confirmed through the effect on I Na . In addition, V. parahaemolyticus and V. alginolyticus species, two known TTX-producers, were identified in these samples 11 . Besides this, several species from Pseudomonas genus were also observed. This genus had previously been reported in TTX production, although the species were not specified 24,25 . From our data, both Vibrio and Pseudomonas sp. could be detected in TTX-contaminated mussel samples. In the case of Pseudomonas, several species are confirmed by 99% sequence homology and for the first time they can be clearly linked to TTX and contaminated mussels 17 .
Therefore, the connection between shellfish, bacteria and TTX in the toxic episode earlier described in Greece was clear, however, a P. minimum bloom was occurring at the same time and it could also be connected. Since the toxin production by this dinoflagellate has not to date been clarified, we studied its growth and toxicity in different conditions. In this way, depending on the salinity, temperature and light, P. minimum strains produce some compounds with an inhibitory effect on I Na , the same as TTX. The inhibitory effect on transfected hNa v 1.6 HEK cells was higher when P. minimum strains from Ecuador and Malaysia were grown at 24 °C, while the Florida strain was not able to grow at this temperature. In the Malaysia strain the effect was also dependent of light and salinity conditions. These results confirm previous data where a water-soluble component produced by P. minimum was able to block sodium channels 7 . The variations of P. minimum toxicity were previously linked to the clone studied and to environmental circumstances 1,8 . Our results also confirm these hypotheses, since depending on the strain and on the growth conditions the effect on I Na is different, as well as the toxin production.
From electrophysiological experiments with P. minimum extracts, the presence of some active compound produced by dinoflagellates was evident. To determine its structure, UPLC-MS/MS technology with Electrospray Ionization (ESI) source was employed to follow the fragmentation pathway of TTX and analogues earlier described 21,26 . In this case a column for Hydrophilic Interaction Chromatography (HILIC) with polar analytes was used and 5 compounds with apparently the same product ions but different retention times to TTX and analogues were obtained 22,26,27 . However, these compounds did not follow the characteristic fragmentation pathway of TTXs. TTX and analogues follow the same pattern, first dehydrated ions by losing a molecule of water [M + H-H 2 O], and sometimes a second one [M + H-2H 2 O], and then the m/z 162 and m/z 178 or 176 ion products. These conditions are deemed necessary to affirm that the compounds are TTX analogues 27,28 . Therefore, P. minimum extracts were checked for ion precursors of the common ion product for all TTX analogues, m/z 162, and two peaks m/z 265 and 308 were obtained. To determine its structure, high-resolution UPLC-IT-TOF technology was performed using TTX as model. Based on the predicted elemental composition and the structure of a parent compound, the structures of product ions can be obtained with a high degree of confidence 21,29-32 . As mentioned, it is well known that TTX can easily lose a molecule of water (− 18 Da), yielding the abundant ion m/z 302. Taking into account the structure for the TTX, the loss of water can occur in different carbons but it is considered that the easiest loss is in C4 21 . This fragmentation was confirmed with a MS 3 spectrum. In this spectrum, the most intense product ions were m/z 284. 0877  162.0842, [C 9 H 12 N 3 + H] + , are also obtained by elimination of C 2 H 8 O 6 and of C 2 H 10 O 7 respectively. The ion m/z 162.0662 identified, as product ion of TTX is different from m/z 162.1026 identified as product ion m/z 265 and m/z 308. In the first case it is a C8 molecule while in the second it is a C9. These structures and molecular formula were also proposed as fragment ions of 5,6,11-trideoxyTTX 21 . In addition, a similar molecule C 9 H 9 N 3 O 2 , molecular weight 191.1, is produced after strong base treatment of TTX 15 . When the errors between theoretical and experimental data are compared some differences are observed. In the case of TTX, both precursor and product ions have a narrow mass difference (0.1-5.7 mDa). In the case of compound m/z 265, mass differences for precursor and 3 of 4 product ions are within the range (3.7-9.9 mDa). While in the case of compound m/z 308, due to the low amount, the mass differences are wider (3-26.9 mDa). Therefore the increase in error should be related to the decrease in the amount of compound used. In addition the sample were not purified, and some matrix effect is also present. However, the errors observed do not invalidate these results since the fragmentation pattern of the TTX analogues observed show the loses of water (− 18 Da), m/z 162 and m/z 179-180. Therefore, m/z 265 and 308 compounds could be precursors, maybe important molecules for later TTXs synthesis. In this regard, bacteria associated with TTX production, Roseobacter genus and V. alginolyticus specie, have also previously been identified in P. minimum cultures 11 . So far, the biosynthetic pathway in organisms that produce TTX is not known and some TTX analogues have been proposed as intermediates of TTX in microorganisms 21 . Therefore, compounds m/z 308 and m/z 265 identified in P. minimum cultures are TTX analogues produced in or by the dinoflagellate, with the same activity on I Na as TTX. These molecules could be TTX precursors in mussels.
In terms of activity it seems that the effect on Na v channels is directly related to m/z 265 presence. However, the quantification of both molecules was done with TTX standard and can be not accurate. In addition, as happens with other TTX analogues, the effect on I Na can be different 22 . In this sense, it has been described that the hydroxyl group at C6 and a CH 2 OH group in the same carbon, are necessary for TTX binding to Na v channels 33 . These groups are also present in compounds m/z 308 and m/z 265. It is important to emphasize that the presence and above all the position of CH 2 OH group is important for TTX-Na v channels binding, and the compounds reported in this work have the same group. Thus, although when compared to TTX structure, some groups are missing, these molecules show inhibition of I Na since the C9 base of TTX is maintained. This structure is common to all TTX active analogues 21 . In this sense the C9 base of TTX is often used to determine the total amount of TTX analogues in samples 15,16 . However, the affinity and toxicity of analogues should be different 22,34 .
A symbiosis between bacteria and dinoflagellates has been often proposed, however little is known about this phenomenon. Sometimes the bacteria induce death of the algae, while in other situations it provides essential vitamins for microalgae. This is the case for Roseobacter sp and P minimum cultures, which the symbiotic association increases the production of B1 and B12 vitamins essential for the dinoflagellate 35,36 . Roseobacter sp have been associated with the production of TTX in the copepod Pseudocaligus fugu 37 . In this case, m/z 320 and m/z 302 ions were identified as TTX and anhydroTTX in Roseobacter sp. cultures. In our experiments both ions (m/z 320 and m/z 302) and other fragment ions were also detected, however since they did not follow the TTX characteristic fragmentation pathway they were not identified as TTXs. As it was reported before, TTX and analogues have been detected in many different animals and environments, in addition considerable differences were found in the analogues identified in pufferfish (TTX derivatives) and amphibians (chiquiritoxin) 21,38 . It has been suggested that this wide range of compounds identified is probably due to the different biosynthesis or metabolism of TTX between animals 39 . Therefore, our results point to compounds 307 (C 11 H 21 N 3 O 7 ) and 264 (C 9 H 20 N 4 O 5 ) as other TTX analogues produced by P. minimum associated with bacteria (Roseobacter and Vibrio sp). This symbiotic production may explain why the strain is not always toxic. As far as we know, this is the first time that the toxic compounds produce by P. minimum have been identified and related to symbiotic bacteria. The structures proposed for the C9 base in these two compounds are shown in Figs 6C and 7C. In the case of m/z 265, the position of the NH 2 group is not defined. For compound m/z 308 two structures could be proposed because the double bond can be located in several positions. To finally clarify this, the structural NMR determination and crystallographic analysis should be performed, although this would require a massive scaling of cultures to obtain enough pure compound. A consequence of TTX being produced by bacteria in algae is that this would potentially be highly influenced by environmental conditions, hence opening a justification as to why climate change provide a link to the recent increase of TTX presence in previously unreported areas.

Mussel samples treatment. Aliquots of 2 g of DG (digestive glands) of sample 1786/2012 collected in
Greece were extracted as previously described 17 . The extract obtained was used for patch clamp recordings.
For DNA extraction DG of samples 1770/2012 and 1774/2012 17 , and one white sample (WT) (not toxic) were separated from mussels and stored at − 80 °C until processed. At least 5 DG from each sample were mechanically homogenized, and DNA was extracted using the DNA extraction kit NucleoSpin Tissue DNA, RNA and protein purification kit (Macherey Nagel, forensic quality certified free of detectable DNA). The bacterial DNA extraction Scientific RepoRts | 7:40880 | DOI: 10.1038/srep40880 protocol was used, following the manufacturer instructions. After purification DNA concentration was determined with a NanoDrop (Thermo Scientific). To study bacteria presence, 16s DNA was amplified, sequenced and analysed (See Supplementary Information).

Automated Patch clamp electrophysiological recordings. For Na current measurements HEK-293
cells stably transfected with the Na channel hNa v 1.6 were used 20,[40][41][42] . This cell line was kindly provided by Dr Andrew Powell (GlaxoSmithKline R&D, Stevenage, UK). Cells were cultured in DMEM/F12 medium supplemented with 10% of fetal bovine serum, Glutamax and MEM NEAA (1% w/v). G418 was freshly added at a final concentration of 0.4 mg/ml in each cell passage. Cells were incubated in a humidified 5% CO 2 /95% air atmosphere at 37 °C until 80% of confluence. Then, cells were incubated at 30 °C for 24-48 h before electrophysiological measurements. Cells were split twice per week.
All the measurements were recorded in whole-cell patch clamp configuration using an IonFlux 16 system (Fluxion, California, USA) and the corresponding Ionflux 16 software for cell capture, seal formation, whole cell obtaining, data acquisition and analysis. After incubation at 30 °C for 24-48 h, cells were washed twice with Ca 2+ and Mg 2+ free phosphate buffered saline (PBS) and harvested with 5 ml of Detachin TM solution. Detached cells were resuspended in extracellular solution containing (mM): 2 CaCl 2 , 1 MgCl 2 , 100 Hepes, 4 KCl, 145 NaCl, 10 TEA-Cl and 10 Glucose. pH 7.4 and 320 mOsm. Electrophysiological recordings were carried out at room temperature (± 22 °C) in a 96-well IonFlux microfluidic plate 43 .
I Na (Na current) was evoked by depolarizing to − 10 mV for 50 ms after a 100 ms step to − 120 mV from − 90 mV holding potential (V h ). The intracellular solution composition for I Na recordings was (in mM): 100 CsF, 45 CsCl, 10 Hepes, 5 NaCl, 5 EGTA corrected to pH 7.1 using CsOH. The contaminating effects of resistance and capacitance currents were compensated electronically by the software. Leak resistance is measured by introducing a short 20 mV pulse at the beginning of each sweep and measuring the current difference 43 . A sampling frequency of 10 kHz was used.

UPLC Conditions.
Chromatographic separation was carried out using both a 1290Infinity ultra-highperformance liquid chromatography system coupled to a 6460 Triple Quadrupole mass spectrometer (Agilent Technologies, Waldbronn, Germany) and from Shimadzu (Kyoto, Japan), two pumps (LC-30AD), autoinjector (SIL-10AC) with refrigerated rack, degasser (DGU-20A), column oven (CTO-10AS) and a system controller (SCL-10AVP) coupled to IT-TOF. The toxins were separated using a column ACQUITY UPLC BEH Amide (2.1 × 100 mm, 1.7 μ m, Waters) at 35 °C. Mobile phase A was 100% water with 10 mM formic acid and 10 mM ammonium formate. Mobile phase B was acetonitrile-water (95:5), containing 5 mM formic acid and 2 mM ammonium formate. The gradient program with a flow rate of 0.4 mL/min was started with 100% B and then a linear gradient to 65% B in 7 minutes. After an isocratic hold time linear of 2 minutes at 65% B and return to the starting conditions of 100% B in 0.5 minutes. Finally, 100% B was kept for 3.5 minutes before the next injection. The injection volume was 5 μ L. MS detection. MS detection was performed using an Agilent G6460C Triple Quadrupole mass spectrometer equipped with an Agilent Jet Stream ESI source (Agilent Technologies, Waldbronn, Germany) and an IT-TOF-MS system with an electrospray ionization (ESI) interface (Shimadzu, Kyoto, Japan). The nitrogen generator is a Nitrocraft NCLC/MC from Air Liquid (Spain).
Agilent source conditions were optimized to achieve the best sensitivity for all compounds: drying gas temperature of 250 °C and flow of 5 L/min, nebulizer gas pressure of 55 psi (Nitrocraft NCLC/MS from Air Liquid), sheath gas temperature of 400 °C and flow of 12 L/min. The capillary voltage was set to 3000 V in positive mode with a nozzle voltage of 0 V. The fragmentor was 152 and the cell accelerator voltage was 2 for each toxin in this method.
IT-TOF-MS source conditions were nebulizing gas flow, 1.5 L/min, heat block temperature and CDL temperature, 200 °C and detector voltage, 1.65 kV.
MS/MS analysis. Initial simple analysis was done in MRM mode. The mass spectrometer was operated in positive mode and the collision energy optimized using MassHunter Optimizer software (Supplementary Table 1). Two ions product were analysed per compound, one for quantification and another for confirmation. IT-TOF-MS analysis. The UPLC from Shimadzu (Kyoto, Japan) mentioned above was connected with an Ion Trap-Time of Flight (IT-TOF) mass spectrometer from Shimadzu. The molecules were analysed using an ion accumulation time of 10 msec and the collision energy for MS n was adjusted to 50% in the analysis and the isolation width of precursor ions was 3.0Th. For full-scan MS analyses, the spectra were recorded in the range of m/z 150-500. Data-dependent acquisition was set such that the most abundant ions in full-scan MS would trigger tandem mass spectrometry (MS n , n = 2-3). Data were acquired and processed by LC/MS solution software including a formula predictor to calculate elemental compositions.
Formula assignments. Accurate masses of fragment ions were processed using the Lab Solution software supplied with the instrument. Any mass corresponding to the formula predictor also calculated particular compositions. To assign the elemental composition of fragment ions, the error ranges were set less than 1 Da as a limit to the calculation of possible elemental compositions using the formula predictor. The other conditions for calculating elemental compositions that were taken into account were the upper limits on the number of C, H, O, N atoms, C/H ratios and the range of double-bond-equivalent (DBE).