Suspect and Target Screening of Natural Toxins in the Ter River Catchment Area in NE Spain and Prioritisation by Their Toxicity

This study presents the application of a suspect screening approach to screen a wide range of natural toxins, including mycotoxins, bacterial toxins, and plant toxins, in surface waters. The method is based on a generic solid-phase extraction procedure, using three sorbent phases in two cartridges that are connected in series, hence covering a wide range of polarities, followed by liquid chromatography coupled to high-resolution mass spectrometry. The acquisition was performed in the full-scan and data-dependent modes while working under positive and negative ionisation conditions. This method was applied in order to assess the natural toxins in the Ter River water reservoirs, which are used to produce drinking water for Barcelona city (Spain). The study was carried out during a period of seven months, covering the expected prior, during, and post-peak blooming periods of the natural toxins. Fifty-three (53) compounds were tentatively identified, and nine of these were confirmed and quantified. Phytotoxins were identified as the most frequent group of natural toxins in the water, particularly the alkaloids group. Finally, the toxins identified to levels 2 and 1 were prioritised according to their bioaccumulation factor, biodegradability, frequency of detection, and toxicity. This screening and prioritisation approach resulted in different natural toxins that should be further assessed for their ecotoxicological effects and considered in future studies.


Introduction
Natural toxins in the aquatic ecosystem can be produced by different organisms, including bacteria, plants and fungi, thus grouping together a wide variety of structures and physicochemical properties and effects [1]. The risk of water contamination by natural toxins generates environmental and public health issues. In some cases, natural toxins can be accumulated in aquatic organisms and transferred throughout the aquatic food chain to humans [2].
However, if we consider freshwater environments, the primary route of human exposure includes the consumption of contaminated water, dermal exposure, and inhalation during recreational activities. Intoxication can include different symptoms, such as a severe headache, a fever, and respiratory paralysis, as well as a variety of possible effects that include hepatotoxicity, neurotoxicity, carcinogenicity, and dermal toxicity. Due to their diversity, toxicological assessment is still challenging and there is also

Tentatively Identified Compounds
In this study, after removal of the background and the very small signals under the minimum intensity threshold, 4404 suspect masses were detected in the 48 water samples by using Compound Discoverer 3.1 software. Among them, 381 compounds (8.6%) were assessed as suspect natural toxins that were included in the in-house database and finally selected for further screening. It is noteworthy that the compounds of the study were natural toxins pertaining to three major groups in water, phytotoxins, mycotoxins, and cyanotoxins. Other compounds, such as pesticides, were discarded in this study. Among these 381 structures, after filtering by way of the isotopic patterns, ionisation efficiency, and fragmentation patterns, the number of suspected identified compounds diminished to 191 structures (50.1% of the initial potential for natural toxins). Finally, the comparison with in-silico MS 2 patterns gave 50 structures that were tentatively identified at level 2 (25.7% of the initial potential for natural toxins) ( Table 1 and Figure 1). Finally, nine natural toxins were confirmed and quantified by injections of the standard.   Plant toxins were the most prominent group in the studied samples (73% of the tentatively identified compounds), with a prevalence of the alkaloids group. The most frequently identified phytotoxins were acetoxytropane, retronecine, and N-methyl pseudo conhydrine in 71%, 70%, and 46% of the samples, respectively. These results are in agreement with the diversity of endemic plants of the area [22], due to the different climatic zones of the occidental Pyrenees and the variation in dry and wet periods. The occurrence of some of these toxins was at a maximum in April, May, August, and September. These two peaks of natural toxins can be related to the leaching into the water immediately after the flowering period in the Mediterranean area, corresponding to April and May, and posteriorly the release of toxins from the dead plant with the consequent rain-washing effect into the river in August and September. For example, in Figure 2, the intensity of the signals of three alkaloids, acetoxytropane, anethole, and retronecine, which can be attributed to the Symphytum officinale, Pimpinella anisum [23], and Apiaceae families, are displayed. As can be seen, the maximum intensities of the toxins were between May and September. In addition to the alkaloids, some terpenes were also tentatively identified. A common species in this area and in the general region of the Iberian Peninsula is bracken (Pteridium aquilinum) [24], which produces ptaquiloside [15]. Ptaquiloside is a carcinogen norsesquiterpene glucoside that is responsible for haemorrhagic disease and bright blindness in livestock and can produce gastric cancer in humans [25]. As can be seen in Figure 1, in this study the degradation product of ptaquiloside, ptaquilosin B (PTB) [26], was identified in 33% of the samples, while ptaquiloside was not detected. The degradation of ptaquiloside in soils and the start of the rainy season explains the leaching of PTB into the water, which is coincident with the maximum intensities of the signals in the samples that were collected in August and September ( Figure 3). Another relevant group of phytotoxins, the phenolic group, was less represented in the samples that were identified, and the representatives of this group were present in a minor number of samples. An example was p-coumaric acid, which was found in only 8% of the samples. Plant toxins were the most prominent group in the studied samples (73% of the tentatively identified compounds), with a prevalence of the alkaloids group. The most frequently identified phytotoxins were acetoxytropane, retronecine, and N-methyl pseudo conhydrine in 71%, 70%, and 46% of the samples, respectively. These results are in agreement with the diversity of endemic plants of the area [22], due to the different climatic zones of the occidental Pyrenees and the variation in dry and wet periods. The occurrence of some of these toxins was at a maximum in April, May, August, and September. These two peaks of natural toxins can be related to the leaching into the water immediately after the flowering period in the Mediterranean area, corresponding to April and May, and posteriorly the release of toxins from the dead plant with the consequent rain-washing effect into the river in August and September. For example, in Figure 2, the intensity of the signals of three alkaloids, acetoxytropane, anethole, and retronecine, which can be attributed to the Symphytum officinale, Pimpinella anisum [23], and Apiaceae families, are displayed.
As can be seen, the maximum intensities of the toxins were between May and September. In addition to the alkaloids, some terpenes were also tentatively identified. A common species in this area and in the general region of the Iberian Peninsula is bracken (Pteridium aquilinum) [24], which produces ptaquiloside [15]. Ptaquiloside is a carcinogen norsesquiterpene glucoside that is responsible for haemorrhagic disease and bright blindness in livestock and can produce gastric cancer in humans [25]. As can be seen in Figure 1, in this study the degradation product of ptaquiloside, ptaquilosin B (PTB) [26], was identified in 33% of the samples, while ptaquiloside was not detected. The degradation of ptaquiloside in soils and the start of the rainy season explains the leaching of PTB into the water, which is coincident with the maximum intensities of the signals in the samples that were collected in August and September ( Figure 3). Another relevant group of phytotoxins, the phenolic group, was less represented in the samples that were identified, and the  Mycotoxins were marginally detectable in the samples, and 58% of the studied water samples did not present detectable concentrations. Alpha-zearalenol was the most prevalent suspect mycotoxin with an occurrence of 29%, followed by aflatoxin B2 (25%), aflatoxin B1 (12%), and averufin, which is an anthraquinoid precursor of aflatoxins [27,28]. Regarding the distribution during the study period, mycotoxins were almost exclusively detected in August and September when the rainy season started, indicating that their presence in water could be due to the washing effect of plants infected with Aspergillus flavus and Aspergillus parasiticus in the case of aflatoxins and Fusarium mycotoxins in the case of alpha-zearalenol. As can be seen in Figure 2, and on the principal component analysis (PCA) presented in Figure 4, the occurrence of natural toxins in natural waters is influenced by seasonality, and the months with a higher charge of natural toxins were in this case April, August, and September, while a very low presence of natural toxins was found at the end of winter and during the driest months. Contrary to what can be expected, the samples from May and July were almost free of cyanotoxins. Only in M1 and M2 during April, August, and September was the occurrence of cyanotoxins detected, in agreement with the two peaking algal blooms in the Mediterranean region. This site (M1) corresponded to the area of Pasteral dam, which is the reservoir that is located downstream of the other reservoirs and presenting slightly higher levels of eutrophication in comparison with the other three areas. The more frequently found cyanotoxins were anatoxin-a, which was present in four samples, followed by microcystin LR, LW, and YR. Mycotoxins were marginally detectable in the samples, and 58% of the studied water samples did not present detectable concentrations. Alpha-zearalenol was the most prevalent suspect mycotoxin with an occurrence of 29%, followed by aflatoxin B 2 (25%), aflatoxin B 1 (12%), and averufin, which is an anthraquinoid precursor of aflatoxins [27,28]. Regarding the distribution during the study period, mycotoxins were almost exclusively detected in August and September when the rainy season started, indicating that their presence in water could be due to the washing effect of plants infected with Aspergillus flavus and Aspergillus parasiticus in the case of aflatoxins and Fusarium mycotoxins in the case of alpha-zearalenol. As can be seen in Figure 2, and on the principal component analysis (PCA) presented in Figure 4, the occurrence of natural toxins in natural waters is influenced by seasonality, and the months with a higher charge of natural toxins were in this case April, August, and September, while a very low presence of natural toxins was found at the end of winter and during the driest months. Contrary to what can be expected, the samples from May and July were almost free of cyanotoxins. Only in M1 and M2 during April, August, and September was the occurrence of cyanotoxins detected, in agreement with the two peaking algal blooms in the Mediterranean region. This site (M1) corresponded to the area of Pasteral dam, which is the reservoir that is located downstream of the other reservoirs and presenting slightly higher levels of eutrophication in comparison with the other three areas. The more frequently found cyanotoxins were anatoxin-a, which was present in four samples, followed by microcystin LR, LW, and YR.
The concomitant presence of three MCs, both with anatoxin-a, at the sampling point M1, suggests this area is of a higher risk in terms of the occurrence of MCs, and therefore of MC producers. This is in line with the previous studies reporting benthonic species in the NE of Catalonia. Thirty-two different species have been identified as endemic in this area [29]. Toxins producing genera of freshwater cyanobacteria include Phormidium spp., Oscillatoria spp., Nostoc spp., and Pseudanabaena spp. [27]. These were considered to be the main producers of MC-LR, MC-YR, and -LW found in the M1 point in May and July. The occurrence of cyanotoxins can be related to increments in temperature and eutrophication, as was confirmed by the Catalan Water Agency [28] and CARIMED 2018 [30] for this area during the period studied. On the other hand, M1 is the downstream point of the studied area, which receives nutrients from areas in the upper river, with nitrate levels between 0.67 and 10 mg N-NO 3 − /L.  The concomitant presence of three MCs, both with anatoxin-a, at the sampling point M1, suggests this area is of a higher risk in terms of the occurrence of MCs, and therefore of MC producers. This is in line with the previous studies reporting benthonic species in the NE of Catalonia. Thirtytwo different species have been identified as endemic in this area [29]. Toxins producing genera of freshwater cyanobacteria include Phormidium spp., Oscillatoria spp., Nostoc spp., and Pseudanabaena spp. [27]. These were considered to be the main producers of MC-LR, MC-YR, and -LW found in the M1 point in May and July. The occurrence of cyanotoxins can be related to increments in temperature and eutrophication, as was confirmed by the Catalan Water Agency [28] and CARIMED 2018 [30] for this area during the period studied. On the other hand, M1 is the downstream point of the studied area, which receives nutrients from areas in the upper river, with nitrate levels between 0.67 and 10 mg N-NO3 -/L.

Target Analysis
A target analysis of 27 natural toxins was carried out using certified standards that are summarised in Table A1 of Appendix A. Matrix-matched calibration curves were used for the quantification of eight natural toxins. The limits of detection (LODs) were between 0.002 to 0.4 µg/L while the limits of quantification (LOQs) were between 0.07 and 1.5 µg/L. The analytical parameters are summarised in Table A3. Nine toxins were confirmed (Ana, AflB1, MC-LR, MC-LW, Nod, MC-YR, Kja, 7-methoxycoumarin, and umbelliferone). Concentrations were under the limit of 1 µg/L as proposed by the World Health Organisation [24] and they were used as an arbitrary reference limit in this work. MC-LR was confirmed in only two sampling points (April M1 and September M1), where the precursor ion [M + H] + 995.5560 m/z was detected for both with the fragment 135.0806 m/z, which is typically generated by the ADDA structure. Finally, MC-LR was confirmed with standards in these two samples. MC-LW and MC-YR were detected at the M1 point in September, August, and, surprisingly, in April, which correspond to the same months where the MC-LR was detected. Anatoxin-a was further detected in the same periods. 7-methoxycoumarin and umbelliferone were confirmed by certified standards. The concentrations of the detected natural toxins are reported in Table 2, showing their presence at relatively low levels in water.

Prioritisation
In this study, a scoring system was designed to highlight the most significant natural toxins of concern in drinking water reservoirs. The scoring system was in accordance with the previous protocol that was published by Choi et al. [31], which is based on the risk-relevant parameters such as the detection frequency in percentage, biodegradability, log BAF, and the toxicity values based on the 50% lethal dose (LD50) laboratory tests in mice. A score in the range of 0 to 100 for each parameter was used, and 100 points were additionally added if carcinogenicity or neurotoxicity was already reported for the substance as what happens, for example, with AflB 1 and AflB 2 . Thus, the maximum total for a given toxin can be 500. In Table 3, detailed information on the parameterisation and scoring is provided, and in Table 4, the parameters used for each tentatively identified substance are shown. It is noteworthy that the biodegradability and the bioaccumulation factor (BAF), used as log BAF, were calculated using EPI Suite TM software (United States Environmental Protection Agency, U.S. EPA).     In Table 5, the ranking of the tentatively identified substances is presented. Four substances, namely, tetrahydrocannabivarin, MC-LW, aconosine, and MC-LR, were ranked with more than 300 points, and 13 toxins were ranked with more than 200 points. In this case, it was considered to be the frequency during the sampling period, which includes seasons with a lower incidence of the substances in water. However, following a month-by-month inspection, for certain substances the frequency was higher; hence, this ranking then varies a little and a higher number of toxins reaches 300 points.
For this reason, in spite of the low concentrations of the substances that are quantified as the top 12 toxins to be tentatively identified, Barcelona city water reservoirs should be monitored at least from May to September, which were the months with higher occurrences of natural toxins.

Conclusions
The method described in this article is a good alternative for tentatively identifying suspect natural toxins in surface water. We have shown that the presence of organic matter near the river can potentially cause the leaching of mycotoxins. Moreover, in this study, plant toxins were mostly spread across different points in relation to the presence of different endemic plants. Notwithstanding, the botanical diversity influences the presence of natural toxins as equally as the precipitation and dry periods. The concentrations of natural toxins were not determined due to the lack of certified standards; however, a correlation between the rain and the leaching in water was described and assessed.
Thanks to these results, we report on the importance of the suspect screening for the identification of natural toxins and their final inclusion in prioritisation lists in order to control their presence in water environments, in particular in drinking water reservoirs. It is also important to increase the amount of data, to help scientists identify environmental compounds when no standards are available, or where they are excessively expensive. Many MC congeners are still not included in databases such as MzCloud and Chemspider. Hence, the retrieval of MS 2 spectrums for the MC congeners is an issue that is being solved with the efforts of the scientific community via the constant updating of data in dedicated databases for environmental research. For comparison purposes, future works should apply this method of analysing natural toxins across different climates worldwide.

Chemicals and Reagents
Twenty-seven (27) natural toxin standards with a maximum purity between 95 and 99% were selected for the targeted analysis. In Table A1 of Appendix A, the list of standards, their main chemical parameters, and providers are listed. Methanol (MeOH), acetone, and acetonitrile (ACN) of HPLC grade were from Merck (Darmstadt, Germany). HPLC water grade was from Baker (Madrid, Spain). The sampling was carried out from March to September 2018, except for June, twice per month, in order to study the prior, during, and after blooming periods, when higher concentrations of natural toxins are expected [77]. In each sampling site, the pH, conductivity, and pO 2 were measured. Water samples were collected in amber glass bottles that had previously been rinsed, transported at 4 • C, and maintained frozen at −40 • C until the start of the analytical process.

Sample Pre-Treatment
Sample pre-treatment was based on the generic methodology to isolate natural toxins from water, as recently developed by Picardo et al. [20]. Briefly, each sample was processed in an ultrasonic bath for 20 min to disrupt the microbial cells and to release the intracellular toxins. Then, the sonicated samples were filtered through a glass microfibre filter of GF/B grade (Sigma Aldrich, Steinheim, Germany). Natural toxins were isolated from the filtrate via a three-step solid-phase extraction (SPE) method, using a hand-made cartridge that had been prepared with 200 mg of a porous graphitised carbon (PGC) 120 mesh (Sigma Aldrich, Steinheim, Germany) and 200 mg of a Bond-Elut PPL (PPL) 120 mesh (Agilent, Santa Clara, CA, USA), coupled to an HLB plus cartridge (225 mg sorbent) (Waters Corporations, Milford, MA, USA).
Then, water samples, each of 100 mL, were loaded into the cartridges at a flow rate of 2 mL/min, previously conditioned with 10 mL of MeOH and 10 mL of water, and both solvents were acidified with 0.5% of formic acid (FA). After loading, the cartridges were dried and switched to elute the analytes in the backflush mode. The PGC/PPL cartridge was reversed, while the HLB cartridge maintained the same position. The toxins were eluted with 15 mL of water/MeOH 2:8 (v/v), followed by 15 mL of MeOH and 15 mL of acetone/MeOH 50:50 (v/v). All the solvents were previously warmed at 45 • C before each elution. The eluate was evaporated almost to dryness and re-dissolved in 1 mL of the mobile phase.

Liquid Chromatography Coupled with High-Resolution Mass Spectrometry
According to the method described by Picardo et al., 2020 [20], the chromatographic separation was carried out using a C18 reversed-phase Lichrosphere (125 mm × 2 mm i.d., 5 µm) column (Merck, Barcelona, ES) connected to an Acquity high-performance liquid chromatography system (Waters Corp, Milford, MA, USA). The binary mobile phase was composed of water (solvent A) and acetonitrile (solvent B) and both had been acidified with 0.1% of FA. The elution gradient was as follows: from 0-3 min, 10% B; from 3-13 min, B was linearly increased to 90%; 13-15 min, stabilised at 90% B; 15-16 min B decreased linearly to 10%; 16-20 min, column stabilisation with 10% of solvent B. A 20 µL injection volume was used with a mobile phase flow rate of 0.25 mL/min.
The HPLC system was coupled to a Thermo Scientific Orbitrap Q-Exactive mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA) equipped with a heated electrospray ionisation source (HESI), and used in the positive and negative ionisation modes. The acquisition was performed using a full-scan and data-dependent analysis (FS-DDA) from m/z = 75 to m/z = 1100, with a resolution of 35,000 full widths at half maximum (FWHM) for the FS and 17,500 FWHM for the DDA There was a spray voltage of 3.75 kV (+) and −3.25 kV (−), a sheath flow gas of 20 a.u., an auxiliary gas of 20 a.u., and a sweep gas of 5 a.u. Heater and capillary temperatures were set at 300 • C with an S-lens RF level at 60%. An inclusion list of the 100 most probable suspect compounds was used (Appendix A  Table A2).

Data Processing: Suspect Screening of Natural Toxins
The suspect screening procedure that was previously described by Picardo et al. [20] was employed with minor changes. Briefly, the FS chromatograms that were obtained with the acquisition software Xcalibur Qual Browser (Thermo Fisher Scientific) were processed, using an automated screening with Compound Discoverer software version 3.1 v. x86 (Thermo Fisher Scientific, San Jose, CA, USA). The first screening steps included peak picking, RT alignment, and grouping of isotopes and adducts (to form compounds), as well as the grouping of compounds across samples. Suspect compounds were marked as background if their peak area in the samples was less than three times larger than the maximum peak area in the blanks. Suspects were tentatively identified using the exact mass with a mass error of 5 ppm. This created a first list of suspect compounds that were further filtered by comparison with a homemade database containing the exact mass of more than 2384 natural toxins. Further filtering steps consisted of the comparison of isotopic patterns, ionisation efficiency, and fragmentation patterns. In Figure 5, the general workflow is summarised, which is similar to the workflows of Krauss [78] and Schymanski [21]. Finally, the MS/MS spectrum was compared with the spectrum of a standard or the predicted fragmentation pattern using the ChemSpider and MzCloud online databases. Unequivocal confirmation was only possible when a reference standard was available (identification at level 1). compared with the spectrum of a standard or the predicted fragmentation pattern using the ChemSpider and MzCloud online databases. Unequivocal confirmation was only possible when a reference standard was available (identification at level 1).

Accuracy, Precision, Limits of Detection, and Quantification
Quantification was achieved through calibration curves that were prepared in an artificial freshwater matrix (AFW). The AFW was prepared using the same ingredients that were reported by Lipschitz and Michel [79]. Briefly, the organic matter was simulated with 10 mg/L of technical grade humic acid (Sigma-Aldrich, reference 53,680), and the pH was adjusted to 6.5 with 1.0 M formic acid. Matrix-matched calibration curves were produced using spiked samples from 0.5 to 100 μg/L. Intraassay precision, accuracy, LOD, and LOQ for the confirmed toxins were calculated according to the EURACHEM guidelines [80]. The instrumental limits of detection (iLOD) were obtained by progressive dilution to the lowest concentration, whereby each compound could be detected. Instrumental reproducibility (inter-day precision) was calculated as the average percentage of the relative standard deviation (RSD%) of the standard solutions (six replicates) at seven concentration levels on three consecutive days.

Accuracy, Precision, Limits of Detection, and Quantification
Quantification was achieved through calibration curves that were prepared in an artificial freshwater matrix (AFW). The AFW was prepared using the same ingredients that were reported by Lipschitz and Michel [79]. Briefly, the organic matter was simulated with 10 mg/L of technical grade humic acid (Sigma-Aldrich, reference 53,680), and the pH was adjusted to 6.5 with 1.0 M formic acid. Matrix-matched calibration curves were produced using spiked samples from 0.5 to 100 µg/L. Intra-assay precision, accuracy, LOD, and LOQ for the confirmed toxins were calculated according to the EURACHEM guidelines [80]. The instrumental limits of detection (iLOD) were obtained by progressive dilution to the lowest concentration, whereby each compound could be detected. Instrumental reproducibility (inter-day precision) was calculated as the average percentage of the relative standard deviation (RSD%) of the standard solutions (six replicates) at seven concentration levels on three consecutive days.

Conflicts of Interest:
The authors declare no conflict of interest.