Detection and Profiling of Diarrheic Marine Biotoxins in Shellfish by mRNA Analysis of Exposed Caco-2 Cells Using qRT-PCR and Multiplex Magnetic Bead-Based Assays

The mouse bioassay for the detection of marine biotoxins in shellfish products is 40 years old and still in use. A full ban or total replacement of this in vivo test has been postponed because of the fear that current chemical-based detection methods could miss a new emerging toxin. In order to fully replace the mouse bioassay, more efforts are needed in the search for functional assays with specific endpoints. Gene expression elicited by diarrheic shellfish poisons (DSP) in Caco-2 cells allowed us to determine three “DSP profiles”, i.e., OA/DTX, AZA-YTX, and PTX profiles. Twelve marker genes were selected to represent the three profiles. qRT-PCR is relatively cheap and easy, and although its multiplex capacity is limited to 5 genes, this was sufficient to show the three expected profiles. The use of the multiplex magnetic bead-based assay was an even better alternative, allowing the detection of all 12 selected marker genes and 2 reference genes, and resulting in clear profiles with for some genes even higher induction factors than obtained by qRT-PCR. When analyzing blank and contaminated shellfish samples with the multiplex magnetic bead-based assay, the contaminated samples could easily be distinguished from the blank samples, and showed the expected profiles. This work is one step further towards the final replacement of the mouse bioassay. We suggest to combine the neuro-2a bioassay for screening with detection by analytical chemical analyses and with the multiplex magnetic bead-based assay for confirmation of known and unknown toxins. This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International license (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium, provided the original work is appropriately cited.


Introduction
Marine biotoxins are toxins produced by phytoplankton and/or bacteria, that can accumulate in several types of marine animals, e.g.shellfish, crabs and fish (FAO, 2004).Bivalve molluscs feed through filtration and concentrate these toxins in their bodies and digestive glands (Botana, 2012;Gerssen et al., 2010a).Mussels for example filtrate 7.5 litres of seawater per hour, leading to the accumulation and concentration of pollutants and toxins (Ciminiello and Fattorusso, 2006).When humans consume seafood with toxins in amounts that exceed the established safety levels, it may lead to intoxication.Marine biotoxins can cause clinical features that vary from diarrhoea and amnesia to paralysis and even death.Five groups of marine biotoxins are regulated within the European Union, i.e. okadaic acid group (OA), which includes the dinophysistoxin analogues (DTXs), this group is also known as the diarrheic shellfish poisons (DSPs); azaspiracid group (AZAs), which can also cause diarrhoea; yessotoxin group (YTXs); domoic acid group (DA), also known as amnesic shellfish poisons (ASPs); and saxitoxin group (STXs), also known as paralytic shellfish poisons (PSPs) (EFSA, 2009).Within these groups, several different types and analogues can be found.It is described that worldwide algae toxins are responsible for approximately 60,000 human intoxications annually (Van Dolah and Ramsdell, 2001).
In order to prevent intoxications, several countries have legislation regarding permitted levels of the different marine biotoxins in shellfish that need to be checked by monitoring programs.These monitoring programs to detect marine biotoxins make use of different methods (Gerssen et al., 2010a).The method of surveillance most used worldwide is the mouse bioassay (MBA), where an extract of shellfish is injected intraperitoneal into a number of mice and death is the endpoint to determine whether the sample is safe to consume or not (Garthwaite, 2000;Stewart and McLeod, 2014).Besides ethical issues regarding the use of laboratory animals, the MBA gives high rates of false-positive and false-negative outcomes (EFSA, 2009;Hess et al., 2006).In Europe, the use of the MBA is banned since 2015, but not for PSP analysis and not for the control of production areas, aiming at detection of possibly unknown toxins (EU, 2011).The EU reference method for the detection of lipohilic shellfish toxins (mainly DSPs and AZAs) is the LC-MS/MS method of the European Reference Laboratory (EURL) on marine biotoxins (Gerssen et al., 2010b;EU, 2011).However, the use of the MBA has been kept over time and is not fully replaced by analytical chemical methods.This is mainly due to a lack of standards for the known toxins, and because toxin patterns might change, which generates a concern about new toxins appearing that would be missed by such chemical analysis and not by the MBA (Rossini, 2005;Campbell et al., 2011).
Nowadays, when possible, toxicity testing should comply with the so called 3R principle, i.e. to refine, reduce and replace experiments with animals (Denisson and Anderson, 2007;Hess et al., 2006;Combes, 2003).Some cell-based assays have been tested in order to obtain information about the mode of action or biological activity of the marine biotoxins, and to replace the MBA (Bovee et al., 2011;Bodero et al., 2018b;Rossini, 2005).The neuro-2a assay is considered as one of the most promising cell-based bioassays for the broad screening of marine biotoxins, i.e.DSPs, neurotoxic shellfish poisons (NSPs), and PSPs (Cañete and Diogène, 2008;Ledreux et al., 2012;Serandour et al., 2012;Nicolas et al., 2014).The readout in this neuro-2a assay is reduction of MTT (decrease of cell viability) and suspect screened samples should be confirmed by additional LC-MS/MS analysis (Bodero et al., 2018a).However, in case a suspect screened sample is not confirmed by analytical chemical methods, it might contain a known unknown or yet unknown toxin.In these cases, an additional cellbased bioassay confirming the presence of a toxin and that is also able to determine the type of toxin present would be very helpful, e.g. a bioassay based on gene expression (Botana, 2012).To do so, previously a whole genome mRNA expression analysis was performed with the human intestinal Caco-2 cell line exposed to DSPs using DNA microarrays.Patterns obtained for toxins or other bioactives are specific and commonly used to characterise new compounds, i.e. compare the profile of the compound with those available in data banks (read across).Exposure to the five regulated DSPs, i.e.OA, DTX-1, AZA-1, PTX-2 and YTX, yielded specific gene expression patterns.From the information provided by these microarray analyses, insights on mode of action were described for OA, DTX-1 and AZA-1 (Bodero et al., 2018b).In summary, OA and DTX-1 induced almost identical mRNA expression patterns, in agreement with the fact that both molecules are analogues that belong to the same toxin group and cause similar effects.For instance, OA and DTX-1 increased expression of genes involved in the hypoxia induced factor pathway/process (HIF), in line with the inhibition of phosphatases and a subsequent activation of the Akt/mTOR pathway, which is involved in the activation of the HIF.OA and DTX-1 also affected pathways like unfolded protein response (UPR) and endoplasmic reticulum (ER) stress.The mRNA expression pattern from AZA-1 was different, where an increase of genes involved in cholesterol biosynthesis and glycolysis pathways was observed, suggesting a different mode of action (Bodero et al., 2018b).Since full genome microarray analysis is not suitable for rapid screening, alternative platforms to detect gene expression levels of highly up-or down-regulated genes as markers for detection and identification were evaluated.A first approach involved a so-called tube array with a limited number of selected marker genes.Although promising, the test was rather expensive, labour intensive and long (it took about 3 days).Moreover, the sensitivity of several of the 17 selected genes on this dedicated array was limited (Bovee et al., 2011).
The present study describes two new approaches for detecting marker mRNAs in exposed Caco-2 cells, i.e. the development of a multiplex qRT-PCR and a multiplex magnetic bead-based assay.The newly developed multiplex qRT-PCR was performed successfully with 5 markers (using the maximum number of 6 fluorescent markers resulted in interference).The newly developed multiplex magnetic bead-based assay was able to correctly quantify the expression levels of 12 selected marker genes.The present study shows that detection of marker mRNAs in exposed Caco-2 cells could be a promising tool to confirm the presence of yet unknown DSPs in mussel samples screened suspect in the neuro-2a assay which cannot be confirmed by LC-MS/MS.We thus propose a strategy where the neuro-2a assay is used as a screening method, LC-MS/MS for confirmation of suspects, and a second cell-based bioassay to confirm the presence of a toxin and detect a toxin profile related to gene expression when suspects from the neuro-2a cannot be confirmed by LC-MS/MS.This work will contribute to the search for new endpoints to detect known and yet unknown marine biotoxins, will help in the identification of unknown toxins, and does so without the need for animal testing.

Samples
In-house samples, both blank mussel samples from the Netherlands, and contaminated samples obtained from various locations in the EU and used for previous validation studies of the LC-MS/MS method (van den Top et al., 2011) were tested.

Preparation of extracts
Prior to extraction of the blank samples and the ones containing lipophilic marine biotoxins, shellfish material was homogenized with a T25 Ultra Turrax mixer at 24,000 rpm (IKA® Works Inc., Wilmington, NC, USA).One gram of shellfish homogenate was vortex mixed with 3 mL methanol for 1 min and centrifuged for 5 min at 2000 × g.The supernatant was transferred to a volumetric flask and the residue was extracted twice more with 3 mL methanol.After the third extraction the volume of the collected supernatant was adjusted to 10 mL with methanol.For exposure of Caco-2 cells, additional clean-up steps using n-hexane and solid phase extraction (SPE) were applied.
Clean-up by n-hexane wash step followed by SPE A 4.8 mL aliquot of the crude methanol shellfish extract was diluted with 1.2 mL Milli-Q water and extracted twice with 6 mL n-hexane in order to remove matrix substances that led to false-positive test outcomes (Bodero et al., 2018a).The hexane layer was discarded and the aqueous methanolic extract was further diluted by adding 10 mL Milli-Q water and the total extract of 16 mL was transferred to an SPE StrataTM-X cartridge (200 mg/6 mL; Phenomenex, Utrecht, the Netherlands), previously conditioned with 4 mL methanol/water (30:70 v/v).Subsequently, the cartridge was washed with 8 mL methanol/water (20:80 v/v) and the toxins were eluted with 4.8 mL methanol.The eluate was evaporated to dryness under a stream of nitrogen gas and reconstituted in 20 µL DMSO.

Exposure, RNA isolation and cDNA synthesis for multiplex qRT-PCR analysis
Caco-2 cells were exposed to the standards and sample extracts.For this, 600 µL of Caco-2 cell suspension were seeded in 24 well plates (Ref.Number 3524, Corning, NY), using 8 x10 4 cells per mL and incubated for 48 h at 37 °C and 5% CO2 to reach 80-90% confluence.DMSO 0.25% (v/v) was included as vehicle control.Exposures were performed in triplicate.Cells were exposed to the standards and samples for 24 h, medium was removed, cells were washed with PBS and lysed with 600 µL of RTL buffer with 1% β-mercaptoethanol.RNA was extracted using the QIAshredder and RNeasy Mini Kit (Qiagen, Venlo, the Netherlands) followed by a DNAse treatment with RNAse free DNAse (Qiagen, Venlo, the Netherlands), both by following the instructions of the manufacturer.After the extraction, the amount and quality of the RNA were evaluated by UV spectrophotometry (260 and 280 nm wavelength) on the Nanodrop spectrophotometer (Nanodrop technologies).cDNA was synthetized using 1 µg of RNA per sample and from an 'RNA pool mix' of all treatments with and without reverse transcriptase using the Biorad iScript cDNA Synthesis Kit with iScript and reverse Transcript (Biorad, 170-8891) in the BioRad iCycler (Biorad, Veenendaal, the Netherlands).The program used was 5 min at 25 °C, 30 min at 42 °C, 5 min at 85 °C, after which the samples were put on ice for 5 min.After the cDNA synthesis, the samples were diluted 10 and 100 times and the pool was diluted 10, 31.6,100, 316, 1000 and 3160 times and used to make a calibration line.The samples were stored at -20 °C.

Singleplex qRT-PCR method
Singleplex qRT-PCR was performed for the selected marker genes with certified QuantiTect primers from Qiagen (Venlo, the Netherlands) using 15 µL of final volume containing: 8.5 µL SYBR green (BioRad 170-8880), 2.5 µL of the QuantiTect forward/reverse primer mix, 2 µL RNAse free water and 2 µL of 100x diluted cDNA.Reactions were performed in a BioRad HSP9645 PCR plate.Water and 'RNA pool mix without reverse transcriptase' were used as negative controls.The plate was covered with a microseal and centrifuged for 1 min.Thermal cycling was performed in a CFX96 Real-Time System (Biorad), starting with a denaturation step at 95 C for 3 min, followed by 45 cycles at 65 C with 35 s for annealing, 10 s at 95 C for denaturation, and 1 min at 65 C for extension.Data were analysed using BioRad software.Expression ratios of the genes were calculated for exposures versus DMSO control.

Multiplex qRT-PCR method
Multiplex qRT-qPCR was performed with primers as shown in Table 1, from Qiagen (Venlo, the Netherlands) and Biolegio (Nijmegen, the Netherlands).All the probes were provided by Biolegio.The sequences are confidential.The reactions were performed using 25 µL final volume containing 12.5 µL 2x Quantifast multiplex PCR master mix (Qiagen, Venlo, the Netherlands, cat number 204752), 1.25 µL of each primer probe mix, 3.75 or 6.25 µL RNAse free water and 2µL of 10x diluted cDNA.Reactions were performed in a BioRad HSP9645 PCR plate.Water and 'RNA pool mix without reverse transcriptase' were used as negative controls.The plate was covered with a microseal and centrifuged for 1 min.Thermal cycling was performed in a CFX96 Real-Time System (Biorad), starting with an initial denaturation step at 95 C for 5 min, followed by 44 cycles at 60 C with 45 s for annealing, 45 s at 95 C for denaturation, and 45 s at 60 C for extension.Data were analysed using BioRad software CFX manager v.3.0.Plate set up and standard curve were selected, and the results are shown as log2 values.Relative quantities (∆ Cq), which express the quantity of the gene under a certain treatment (toxin) vs the quantity under control treatment (vehicle) are plotted and expressed as log2 values versus the control.The expression of the reference gene TMEM179B was not affected by any treatment and left out in the newly developed multiplex qRT-PCR method.

Multiplex magnetic bead-based assay
Caco-2 cells were seeded in a 96 well plate (Ref.Number 3595, Corning, NY) using 100 µL of a suspension containing 8 x 10 4 cells per mL and incubated for 48 h at 37 °C and 5% CO2, to reach 80-90% confluence.Then, cells were exposed to samples and standards for 24 h.DMSO 0.25% (v/v) was included as vehicle control.After 24 h exposure of Caco-2 cells in the 96 well format (every exposure performed in triplicate, e.g. 3 wells per treatment), mild lysis of cells was achieved according to the manufacturer instructions (QuantiGene 2.0 plex assay user manual, Affymetrix, the Netherlands).Briefly, the lysis mixture was diluted in nuclease-free water and 100 µL were added per well.Plates were incubated for 18-22 hours at 54 °C ± 1 °C, at 600 rpm in a VorTemp™ 56 Shaking Incubator (Thermo Fischer, the Netherlands), previously validated with a QuantiGene Incubator Temperature Validation Kit (Isogen, the Netherlands).The assay procedure consists of several hybridization, incubation and washing steps, using a plate magnet to capture the beads (Affymetrix, the Netherlands).After the final binding step, 130 µL of washing buffer (provided in the kit) was added to the wells and plates were read in a xPonent ® 3D machine (Luminex corp).The protocol was defined using manufacturer instructions, i.e. sample size 100 µL, DD gate 5.000 -25.000, timeout 45 seconds and bead event 100.The total time needed from cell lysis to read out is about 30 h.Data analysis was performed as follows: MFI (median fluorescence intensity values) are provided from the xPonent ® 3D machine in a .cvsfile and were analysed using excel for calculating the average signal (avg MFI) for each gene (exposures were performed in triplicate).Then, the value obtained from each gene was divided by the value for the normalisation gene.
Here we used the CUL1 gene (avg MFI gene of interest/avg MFI CUL1).Finally, for each test gene, we calculated the fold change by dividing the normalised value for the treated samples by the normalised value for the untreated sample, i.e.DMSO ((avg gene/avgcul1)/avg DMSO).Values were plotted in GraphPad Prism.

Results
Table 2 shows the twelve marker genes and three reference genes which were selected from the whole genome array studies, where undifferentiated Caco-2 cells were exposed to OA, DTX-1 and AZA-1 (Bodero et al., 2018b) and to YTX and PTX-2 (unpublished data).Genes were selected based on their response to the different toxins, e.g., NPPB is specifically downregulated by PTX-2 and to some extent by DTX-1, while RGS16 is specifically up-regulated by DTX-1 and to some extent by OA.TMEM179B, CUL1 and SH3BP2 were not affected and used as reference genes.Moreover, besides OA, DTXs, AZAs, YTXs and PTXs, other (nonregulated) marine biotoxins like the cyclic imines (CIs) might end up in the lipophilic extracts, however these toxins do not lead to clear effects on gene expression, even when tested at higher concentrations.Figure S1 1 is an example of the CIs PnTX-E and SPX, and shows that these toxins do not result in clear effects on gene expression in exposed Caco-2 cells.

Tab. 2: Selected genes and representation of their expression as determined in the whole genome array studies
Up-or down-regulated compared to a vehicle control: red arrows up represent genes that are up-regulated with log2 values higher than 0.7 and green arrows down are genes down-regulated with log2 values lower than -0.7 (Bodero et al., 2018b).The (*) represents up-regulation higher than a log2 value of 2.0 or down-regulation of a log2 value lower than -1.5.The (-) represents log2 values between -0.4 and 0.4, which are considered as no significant effects on gene expression.ND: not determined.

Development of a multiplex qRT-PCR detection method
First, singleplex qRT-PCRs were performed in order to confirm the results from the whole genome array studies.For that, Caco-2 cells were exposed to 3 and 9 nM of all the toxins, including the analogues, except for YTX, for which 12.5 and 37.5 nM were used.These concentrations were used for all following experiments, as these concentrations result from a newly developed clean-up procedure for DSPs from mussels in combination with the regulatory limits (160 µg/kg shellfish for OA, DTXs, PTX-2 and AZAs, and 3.75 mg/kg shellfish for YTXs).Assuming 100% toxin recoveries (Bodero et al., 2018a), the regulatory limits of OA, DTXs, AZAs and PTX-2 will result in a final concentration of about 12 nM in the well, while the regulatory limit of YTX would result in a final well concentration of about 200 nM.Figure 1 shows the relative expression level of each target gene for each toxin concentration.Showing that each toxin except OA can be detected at a concentration relevant for enforcement purposes, i.e. a lower concentration (using a newly developed clean-up procedure) than resulting from its regulatory limit.In general, genes in the singleplex qRT-PCR responded as expected from the whole genome array study (Tab.2).The responses observed for AZA-1 at only 3 nM are already clear and as expected, i.e. up-regulation of DDIT4 and downregulation of TGFB2 and no effects on NPPB, RGS16 and CXCR4.As expected, all 5 genes responded to DTX-1 and even did so in a dose related way when looking at the responses obtained with 3 and 9 nM.Unfortunately, there were no clear responses to OA at 3 and 9 nM.However, this was more or less anticipated, as 25 nM was used in the whole genome array study because at lower concentrations no effects of OA on gene expression could be observed (data not shown).In spite of this, it was worthwhile to test 3 and 9 nM OA using singleplex qRT-PCR to investigate if singleplex qRT-PCR would be sensitive enough to detect OA at concentrations relevant for enforcement purposes.It should also be emphasised that this singleplex qRT-PCR is mainly developed to confirm suspect screened samples in the neuro-2a bioassay.For further experimentation, the concentrations of OA were increased to 25 and 100 nM.PTX-2 is easily detected with the marker gene NPPB, that is specifically down-regulated by this toxin at low concentrations.As expected, YTX could be detected by the upregulation of DDIT4.It was remarkable that the YTX singleplex qRT-PCR profile is similar to that of AZAs, and also showed a down-regulation of TGFB2, which was not expected from the gene expression analysis.Another interesting finding is that this TGFB2 gene is downregulated by all toxins, except for OA at the (low) concentrations tested.The reference gene TMEM179B did not show any relevant expression (data not shown).It was decided to skip the TMEM179B as a reference gene as it turned out that when we designed the multiplex qRT-PCR, the use of the maximum amount of 6 fluorescent markers resulted in interference.
As the singleplex qRT-PCR results showed that the 5 selected marker genes responded as expected and in a sensitive way, a multiplex qRT-PCR was developed using primers, probes and dyes as shown in Table 1. Figure 2 shows that this newly developed multiplex qRT-PCR was able to detect all toxins, and all except OA, at concentrations that are relevant for enforcement purposes.It was anticipated that toxin analogues would result in similar expression profiles and that this multiplex qRT-PCR would thus be suited for detecting analogues as well.Therefore, also AZA-2, AZA-3, DTX-2 and hYTX were tested.Figure 2 shows that exposure to AZA-2 and AZA-3 indeed resulted in similar profiles as AZA-1 and at similar concentrations.Also, the profiles of OA, DTX-1 and DTX-2 are identical as are the profiles of YTX and hYTX.Tab.S1 1 shows a comparison of the expected and obtained results.
In order to further evaluate the performance of the newly developed multiplex qRT-PCR, ten blank mussel samples and a mussel sample contaminated with AZAs (1083 µg AZA-eq kg -1 ), all according to the EURL LC-MS/MS method, were tested.Figure 3 shows that the extracts of the blank mussel samples did not affect the expression of any of the selected marker genes, while mussel contaminated with AZAs resulted in a "perfect" AZA/YTX-profile.Standards of OA and PTX-2 were used as positive controls and also resulted in the expected profiles.To further increase the capacity to detect and identify the toxins by multiplex qRT-PCR, especially for the discrimination between the presence of AZAs or YTXs, a second multiplex qRT-PCR could be developed.However, in order to have one single test method to detect more than 5 marker genes, it was decided to analyse mRNA expression on another format: a multiplex magnetic bead-based assay.

Development of a Multiplex magnetic bead-based assay for 14 genes
The multiplex magnetic bead-based assay enables the examination of up to 100 genes, and it is based on the direct detection of the mRNA present in the sample, making it less labour intensive, i.e. no need for RNA purification, reverse transcription or amplification.Besides the 5 genes selected for multiplex PCR, more genes were selected from the whole genome array experiments, i.e. 7 marker genes and 2 more reference genes (Tab.2).The multiplex magnetic bead-based assay uses magnetic beads coupled with DNA probes.These specific probes hybridise with a cognate mRNA present in the sample.The fluorescent signal associated with each specific bead is read on a Luminex ® flow cytometer, where the equipment detects the specific bead, representing the gene, and the fluorescent signal attached to that bead, indicating the amount of cognate mRNA in the sample.Median fluorescence intensities (MFIs) are measured and used to calculate relative gene expression levels.The same toxins and toxin analogues as described above for the multiplex qRT-PCR method (5-plex) were tested in this multiplex magnetic bead-based 14-plex assay, i.e.OA, DTX-1, DTX-2, AZA-1, AZA-2, AZA-3, YTX, hYTX and PTX-2.Pinnatoxin (PnTX-E) was used as a negative control, since PnTX-E hardly affects the gene expression levels in Caco-2 (Fig. S1 1 ). Figure 4 shows the results for all the analogues and the 12 marker genes selected for this method (see also Tab.S2 1 ).These data also revealed three clear profiles, i.e.OA/DTXs profile, AZAs/YTXs profile and a PTX-2 profile.As expected, PnTX-E did not elicit any specific responses at the gene expression level.When looking at more data in more detail (Fig. S2 1 ), it also becomes clear that in the concentration ranges tested, i.e.AZAs 3-9 nM; DTXs and PTX-2 3-27 nM; OA 3-100 nM and YTXs 12.5-37.5nM, the OA/DTX profile shows clear dose-response effects.It also shows that on this test format it is also not possible to detect gene expression at low concentrations of OA, i.e. 3-9 nM.OA starts to affect gene expression at 25 nM and resulting in a clear profile at 100 nM, similar to the profile obtained with DTX-1 at 9 nM, indicating that OA is about 4 times more potent than DTX-1 and in line with the relative potencies as observed in the neuro-2a bioassay (Bodero et al., 2018a;Bodero et al., 2018b).Thus, also on the bead-based format it is not possible to detect OA at concentrations relevant for enforcement purposes.The DTX-2 response is lower than that of DTX-1, this is expected, as DTX-2 is less potent than DTX-1 (FAO/WHO, 2016; Aune et al., 2007;Bodero et al., 2018a).Anyway, at a relevant level for enforcement purposes, 27 nM, the DTX-2 profile is identical to that of DTX-1 at 9 nM.Just as the DTXs, the AZAs, YTXs and PTX-2 can be detected at relevant concentrations for enforcement purposes.Unfortunately, the profiles for AZAs and YTXs are still identical.The TNS4, OSR2 and MT1H genes were especially added to distinguish the AZAs from the YTXs (Tab.2), but just like TGFB2 in the multiplex qRT-PCR, the YTXs cause the same effect on these three genes as the AZAs do.In order to rule out that YTX or AZA toxin standard were switched, the YTX, AZA-1 and DTX-1 stock solutions in DMSO were checked by LC-MS/MS analysis.Figure S3 1 shows the obtained mass chromatograms, demonstrating that YTX and AZA were not switched, and that these standards are of the quality as expected for certified reference standards.

Discussion
Worldwide, the mouse bioassay has been the main method to detect shellfish poisons in samples for human consumption and still complete surveillance programs that rely on the use of this animal test in many countries.Chemical analytical methods have been developed and proven suitable for the detection of known toxins, but countries with relative high occurrences of shellfish toxins in their coastal waters are still afraid to rely solely on such analytical chemical methods.One concern is the lack of standards for the known toxins.Another serious concern is that new appear that would be missed by such chemical analysis.In that regard, cell-based bioassays are an additional promising alternative.Especially the neuro-2a bioassay has been shown to be very useful for the broad detection of marine biotoxins, i.e.DSPs, NSPs and PSPs (Nicolas et al., 2014;Reverte et al., 2014;Cañete and Diogène, 2010).When using the neuro-2a bioassay for the broad detection of DSPs, samples screened negative are safe to consume and suspect screened samples can be confirmed by analytical chemical methods.It has been demonstrated that this is a fruitful approach (Bodero et al., 2018a).However, in case a suspect screened sample cannot be confirmed by chemical analysis, indicating the presence of an unknown toxin, additional analysis is needed.In order to be successful, a second bioassay that is able to confirm the presence of such "DSP-like toxins" and also to identify the kind of DSP toxin present, would be very helpful.When this second bioassay also indicates the presence of a toxin, a bioassaydirected approach can be followed to identify this unknown active (Rijk et al., 2009).Figure 7 is a schematic view of the proposed strategy.The strategy uses the neuro-2a bioassay for broad screening/detection (nonspecific cytotoxicity), analytical chemical analyses (LC-MS/MS) for the confirmation of known toxins, and the multiplex magnetic bead-based assay (specific mRNAs) or qRT-PCR for the confirmation of unknown toxins in case suspect neuro-2a outcomes cannot be explained by LC-MS/MS analysis.When the presence of an unknown active is confirmed by the second bioassay, a bioassay directed fractionation approach can be used to identify the new toxin.
In the present study, specific effects of the DSPs on the gene expression in Caco-2 cells were used to develop a method that is able to distinguish these toxins.Previous gene expression studies envisioned three toxin profiles: i) OA/DTXs, ii) AZAs (and YTXs to some degree), and iii) PTX-2.Marker genes were selected, and two multiplex assays were developed, i.e. a multiplex qRT-PCR (5-plex) method and a multiplex magnetic bead-based assay (14-plex).
The multiplex qRT-PCR method, using 5 markers only, was able to determine the presence of each of all regulated DSPs, including their analogues and thus potentially also unknowns, in extracts prepared from mussel samples.The obtained profiles enabled the discrimination between the presence of OA/DTXs, AZAs and PTX-2, but unfortunately the toxin profiles were not specific enough to discriminate between the presence of AZAs and YTXs.As OA and DTX belong to the same group, have a similar mode of action, and only differ in their potency, it is not possible to distinguish them with these kinds of effect based bioassays (Bodero et al., 2018b;Ferron et al., 2014).A multiplex magnetic bead-based assay, i.e. using specific probes that hybridise with the selected marker mRNAs and which are attached to Luminex ® magnetic beads, allowed us to multiplex 14 genes in one reaction and resulted in more clear and complete toxin profiles, showing similar or higher induction factors as obtained by qRT-PCR.Unfortunately, the profiles did still not allow a discrimination between AZAs and YTXs, or a more sensitive detection of OA.The TNS4, OSR2 and MT1H genes were especially added to the multiplex magnetic bead-based assay to distinguish the AZAs from the YTXs (Tab.2), but just like TGFB2 in the multiplex qRT-PCR, the YTXs cause the same effect on these three genes as the AZAs do.The selection of those genes was done on gene expression analysis performed on a different platform, microarrays.Probably, this YTX array did not work as accurate as the ones used for OA, DTX-1, AZA-1 and PTX-2, as the standards were checked by LC-MS/MS and were pure and not switched.Anyway, both methods are able to confirm the presence of DSPs and also to (partly) identify the kind of DSP toxin present.
Besides the DSPs tested, also the neurotoxic brevetoxins will be present in the prepared extracts (lipophilic), while PSPs and ASPs (hydrophilic) will not end up in these lipophilic sample extracts.Effects on gene expression by brevetoxins and transcriptomics data are scarce (Walsh et al., 2003;Murrell and Gibson, 2011).However, the brevetoxins were not included in the present study, as the neuro-2a bioassay is already able to discriminate between the presence of DSPs and the neurotoxic brevetoxins, i.e.DSPs can be detected without the addition of ouabain and veratridine, while the brevetoxins can only be detected by the neuro-2a bioassay by adding low concentrations of ouabain and veratridine (Manger et al., 1995;Dickey et al., 1999).In addition, it is shown that the nonregulated CIs like PnTX-E and SPX, that can also end up in the prepared lipophilic extracts, do not elicit clear effects on gene expression in Caco-2 cells (Fig. S1 1 ).It can never be fully ruled out that an unknown compound leads to a positive neuro-2a bioassay outcome that cannot be confirmed by both the LC-MS/MS and the multiplex assays based on gene expression of Caco-2 cells.It is impossible to address upfront whether it could lead to human intoxications, e.g.diarrhoea.However, in case of an DSP outbreak (humans with e.g.diarrhoea), the bioassay-directed fractionation approach can still be used, despite the absence of a confirmed Caco-2 effect.
In a previous study, it was shown that the use of an additional n-hexane washing-step improved the clean-up of the lipophilic marine biotoxins by eliminating false-positives in the neuro-2a bioassay due to matrix effects (Bodero et al., 2018a).The present study shows that this clean-up also results in extracts that can be used to expose the Caco-2 cells, as blank samples did not affect the gene expression patterns, while contaminated samples resulted in the expected profiles.The use of the multiplex magnetic bead-based assay allowed us to multiplex 14 genes in one reaction.The results show that the method performs well and is less labour intensive than the multiplex qRT-PCR method, but the costs are higher.However, both multiplex methods work, and laboratories involved in monitoring can make their own choice, as the Caco-2 cells are easily available.Although more testing and validation are required, an approach where the neuro-2a bioassay is used for the broad screening of lipophilic marine biotoxins and LC-MS/MS analysis is used to confirm and identify the toxins present in the suspect screened samples, supplemented with a multiplex assay based on the expression of marker genes in Caco-2 cells in case suspect screened neuro-2a samples cannot be confirmed by LC-MS/MS analysis (Fig. 7), is very promising for ultimately replacing the rather cruel assay with mice.

Fig. 3 :
Fig. 3: Multiplex qRT-PCR profiles from blank shellfish extracts, standards and a naturally contaminated mussel sample Caco-2 cells were exposed to 10 blank shellfish extracts, OA and PTX-2 standards and an extract prepared from a mussel sample indicated as 'AZA m' that is naturally contaminated with AZAs (1083 µg AZA-1-eq kg -1 ) Bars represent log2 values of the relative expression levels (RQ) of the genes.

Fig. 7 :
Fig. 7: Proposed strategy for the broad screening of shellfish for the presence of DSPsThe strategy uses the neuro-2a bioassay for broad screening/detection (nonspecific cytotoxicity), analytical chemical analyses (LC-MS/MS) for the confirmation of known toxins, and the multiplex magnetic bead-based assay (specific mRNAs) or qRT-PCR for the confirmation of unknown toxins in case suspect neuro-2a outcomes cannot be explained by LC-MS/MS analysis.When the presence of an unknown active is confirmed by the second bioassay, a bioassay directed fractionation approach can be used to identify the new toxin.