Introduction

Paralytic shellfish poisoning (PSP) is a seafood illness associated with the consumption of shellfish contaminated with a group of compounds known as paralytic shellfish toxins (PSTs) that were accumulated in filter-feeding organisms, such as bivalve molluscs. The PSTs found in the marine environment can be produced by the dinoflagellate microalgae Gymnodinium catenatum, Pyrodinium bahamense var compressum, Pyrodinium bahamense var bahamense and several species from the genus Alexandrium (FAO 2004; Usup et al. 2012). These PSTs comprise more than 30 different analogues that are water-soluble tetrahydropurine compounds and depending on the modification of their functional groups results in different toxicity (Fig. 1) (Vale 2010). PSTs commonly found in contaminated marine bivalves and included in food safety regulations are divided into three sub-groups. The group of carbamoyl compounds includes the most potent toxins like saxitoxin (STX), the double sulphated C-toxins are the least toxic, and the decarbomoyl analogues have intermediate toxicity (EFSA 2009). Contamination above the regulatory limit of 800 µg STX diHCl eq/kg originates the closure of classified production areas (Anonymous 2004).

Fig. 1
figure 1

Structures of the principal PSP toxins found in marine bivalves derived from microalgae

Full size image

The occurrence of PSTs in bivalve molluscs from the Portuguese coast is mainly associated with blooms of Gymnodinium catenatum (Vale et al. 2008). The monitoring programme for PSTs on bivalve molluscs started in 1986 (Vale et al. 2008). Unlike other marine biotoxins, such as diarrhetic shellfish poisoning (DSP) toxins and amnesic shellfish poisoning (ASP) toxins, which contaminate bivalves above the maximum permitted levels every year without exceptions, PST occurrence is more variable, only occurring in discrete years (Vale et al. 2008; IPMA 2024a).

PSTs are unstable under alkaline conditions, especially in the presence of hydrogen peroxide, periodic acid or t-butyl hydroperoxide, yielding fluorescent derivatives. This has been exploited to develop two types of liquid chromatographic (LC) methods: with pre-column derivatisation or post-column derivatisation. A method with pre-column derivatisation and fluorescence detection (FLD), based on Lawrence and Ménard (1991), was introduced in 1996 for the complementary testing of PSTs in Portuguese bivalves (Vale and Sampayo 2001). In 2007, this method was upgraded to the validated version of the method: AOAC Official Method 2005.06 (AOAC 2005). It was later in-house validated during 2017–2018 in a novel equipment set (Vale 2019). Since 2017, the so-called Lawrence method has become the official testing method for PSTs in the EU, fully replacing the live mouse bioassay (Anonymous 2017).

Application of this method for routine monitoring purposes in official control laboratories is labour-intensive. It involves a double extraction and one solid-phase extraction (SPE). When samples are contaminated with PSTs, the Lawrence method can involve another SPE partitioning, one hydrolysis for C3 + 4 toxins, two distinct oxidation steps and multiple analytical runs (up to four or five) to detect all relevant toxins, which are time-consuming and demanding in terms of personnel, consumables and reagents (certified reference materials), when a full quantitation is needed.

From our own historical experience acquired with the application of the Lawrence method to bivalve samples with PST’s profiles derived from G. catenatum ingestion, data was recently evaluated to optimise a high-throughput scheme for the application of this method for monitoring purposes (Vale et al. 2021). This scheme is based on our in-house screening that uses in the first place hydrogen peroxide oxidation instead of periodate oxidation as recommended in the Lawrence method. The peroxide approach allows rapid, immediate and sensitive determination of all four N1-H toxins present (C1 + 2, GTX5, dcSTX and dcGTX2 + 3), out of seven individual toxins and/or epimeric pairs commonly found in shellfish after contamination with G. catenatum (which includes additional dcNEO, C3 + 4 and GTX6 [B2]).

This approach is well suited to identify the source of algal populations from the chemotaxonomic analysis of PST profiles in shellfish, as described by Lewis et al. (2022). In the Portuguese bivalve monitoring programme, occasional trace levels of toxins typical from Alexandrium spp. have been found, which present mainly a gonyautoxin 2 + 3 profile, clearly distinct from the more complex G. catenatum’s profile (Vale and Sampayo 2001; unpublished monitoring data).

Our peroxide screening has been run with a new fluorescence detector since 2017: the Agilent 1260 Infinity series (Agilent 2013). Fluctuations in the fluorescent response of PST’s analogues occurred, despite using modern state-of-the-art equipment. This contrasted with our experience of using an ultraviolet detector for the quantification of the amnesic shellfish poisoning toxin domoic acid, where the response is stable all year round, and variations depend more on the stability of calibration solutions rather than the equipment itself.

The patterns of variation of fluorescence yields of PSTs were analysed here to refine and deepen our understanding and ruggedness of the method. Variation was found to be associated not only with chemical stability but mainly with ambient temperature at the time of the oxidation step.

Materials and Methods

Reagents and Chemicals

Certified reference toxins (dcGTX2 + 3, C1 + 2, dcSTX, GTX2 + 3, GTX5 [B1], STX, dcNEO, GTX1 + 4, GTX6 [B2] and NEO) were obtained from the Institute of Marine Biosciences, National Research Council Canada (IMB, NRCC, Halifax, Nova Scotia, Canada). The standard for C3 + 4, despite being commercially available for some time, had not been yet purchased at the time of these experiments. C3 + 4 were measured indirectly through hydrolysis, and quantified as GTX1 + 4. Stock solutions and calibration curves were prepared in toxin-free oyster matrix (C18 or COOH cleaned), circumventing the low recovery issue for dcGTX2 + 3 (Ben-Gigirey, Rodríguez-Velasco and Gago-Martínez, 2012).

For N1-H analogues, equimolar concentrations were used in the stock solution (2 µM), except for C1 + 2 and GTX5 (6 and 4 µM, respectively), to match the contamination profile derived from G. catenatum. Five-point calibration curves were prepared for each toxin: 0.07 to 0.50 µM for dcGTX2 + 3, dcSTX, GTX2 + 3 and STX; 0.21 to 1.51 µM for C1 + 2; and 0.14 to 1.0 µM for GTX5. For the N1-hydroxylated analogue dcNEO, the stock solution was 2 µM, and the curves were between 0.19 and 1.1 µM (Vale 2019).

Acetonitrile was of HPLC grade, and analytical grade chemicals were used for the remaining reagents. Ultrapure water was obtained with a Select Neptune Ultimate system (Suez-Purite, UK).

Tissue Preparation

Samples of bivalves were collected as part of the ongoing national monitoring programme for bivalve molluscs, covering the continental Portuguese coast. Soft tissues from live shellfish samples were removed from the shell, rinsed with water and homogenised with a blender. Five grammes of each shellfish homogenate was extracted twice with 3 mL 1% acetic acid, the first extraction performed with heating for 5 min in a boiling water bath, and the second only by vortex mixing. Both extractions were followed by centrifugation at 2500 × g for 10 min; the supernatants were combined and the volumes adjusted to 10 mL.

A 3 ml/500 mg SPE Supelclean C18 cartridge (Supelco, USA) was conditioned with 6 mL methanol and 6 mL water. One millilitre of the acetic acid extract was passed through, after which toxins were eluted with 2 mL water and collected in a 10-ml test tube. After a pH increase to circa 6.5 with 0.2 M NaOH solution, the volume was adjusted to 4 mL (‘C18’ fraction). An additional fractioning of toxins was prepared using a COOH cartridge as preconised in (AOAC 2005).

The toxins dcGTX2 + 3, C1 + 2, dcSTX, GTX2 + 3, GTX5 and STX present in this fraction were determined using peroxide reaction. To accomplish this, 100 μL of the cleaned extract was added to a mix of 25 μL aqueous H2O2 (10%) solution and 250 μL of 1 M NaOH solution. It was then mixed in a vortex mixer and left reacting for 2 min at room temperature, and 20 μL of glacial acetic acid was added to stop the oxidation and vortex mixed (AOAC 2005).

For the periodate oxidation reaction, a solution of 0.3 M ammonium formate, 0.3 M Na2HPO4 and 0.03 M periodic acid (1:1:1, v:v:v) was prepared on a daily basis and its pH adjusted to 8.2 with 0.2 M NaOH. In a vial containing 100 μL of a blank matrix (oyster) and 100 μL of the sample extracts/fractions prepared above, 500 μL of the oxidant reactant was added. After vortex mixing, it was permitted to react during 1 min at room temperature, and then 5 μL of glacial acetic acid was added and vortex mixed (AOAC 2005). The C18 fraction was used to determine dcNEO, C3 + 4 in the COOH fraction #1 was determined indirectly by hydrolysis, and GTX6 [B2] was determined in the COOH fraction #2.

Liquid Chromatography Analysis

HPLC analyses were carried out using an Agilent Technologies (USA) system comprising: a quaternary pump with built-in degasser and a refrigerated autosampler from the 1290 Infinity series plus, a column oven and a fluorescence detector from the 1260 Infinity I series. The OpenLAB CDS Chemstation Edition (Rev. C) software performed data acquisition and peak integration. The separation was performed on a Supelcosil LC-18, 150 × 4.6 mm, 5 µm column (Supelco, USA), equipped with a C18 guard column (SecurityGuard™, 4 × 3 mm, Phenomenex, USA), and the column was maintained at 30 °C. Mobile phase A was made of 0.1 M ammonium formate, pH = 6, and mobile phase B was made of 0.1 M ammonium formate in 5% acetonitrile, adjusted to pH = 6 with 0.1 M acetic acid. The flow rate was 1 mL/min. at 100% A, with the following linear gradient during each run: 5% B at 0.25 min, 10% B at 4.0 min, 90% B at 9.0 min, 10% B at 11 min and 0% B at 13.0 min. Run time was 13 min followed by 1.5 min post-run. The auto-sampler was kept at 8 °C, and 40 µL was injected for peroxide reactions or 80 µL for periodate reactions. The detection wavelengths were set at 340 nm for excitation, 395 nm for quantitation and 430 nm for confirmation.

Study of Temperature Dependence

A stock comprising all N1-H toxins (at 0.25 µM dcGTX2 + 3, dcSTX, GTX2 + 3 and STX; 0.77 µM C1 + 2; and 0.51 µM GTX5) was spiked in oyster matrix and oxidised in triplicate in January 2022 at room temperature (18 °C). For heating, the reactives (NaOH and H2O2) were preheated to 22 °C or 26 °C in a dry heating block (Grant Instruments), before the 2-min oxidation taking place also inside the heating block. Another stocks, comprising dcNEO or GTX1 + 4 plus GTX6, were submitted to periodate oxidation at room temperature (18 °C) or at 22 or 26 °C as described above.

Additional experiments were performed in another equipment just recently acquired: the Agilent Technologies 1260 Infinity II series, controlled with OpenLab CDS Workstation Software (version 2.7). A stock comprising all N1-H toxins was spiked in mussel matrix at LOQ level and oxidised with peroxide at room temperature (20 °C) for 2 min or at 50 °C for 4 min in a dry heating block. A 0.1 M stock of GTX2 + 3 was heated with peroxide at 60° between 2–10 min in the same conditions.

Air Temperature

As a detailed record of the laboratory air temperature or of laboratory water available at room temperature was not available for the period of the chemical data (2018–2022), the average air temperature for Lisbon’s airport was used. Data was retrieved from IPMA’s monthly climatological bulletin for continental Portugal (IPMA 2024b). The monthly average temperature of the previous month was used to relate with chemical data obtained in the first week of the following month, and the current monthly average was used for the remaining weeks.

Statistical Treatment

Statistical calculations were performed using KyPlot 6.0 (kyensLab Inc., Japan). Third-degree polynomial regression analysis was used for the variation of calibration slopes with air temperature.

Results

Two distinct sets of calibration curves were obtained with the Agilent 1260 series HPLC equipment between 2018 and 2021. During 2018 and 2019, the new 1260-Series detector was used. Due to its malfunction, it was replaced by another equipment of the same model, previously refurbished in the factory by Agilent, which was used in 2020 and 2021. Despite the same configuration settings being used, the fluorescence response was higher with the second fluorescence unit (Fig. S1a).

The fluorescence response yield seemed to present a seasonal fluctuation, with responses increasing during the summer, as exemplified for the toxins C1 + 2 and GTX5 in Fig. S1a. This pattern repeated every summer during the 4 years of using an Agilent 1260 Infinity detector. The fluctuations in the calibration’s slopes were further analysed for the years 2020 and 2021, by averaging the slopes obtained in each HPLC run. The slope of individual N1-H toxins was normalised against this average. Figure 2 exemplifies the normalised response for C1 + 2 and GTX5 across these 2 years. The 11-hydroxysulphate C1 + 2 presented a larger seasonal fluctuation than the non-11-hydroxysulphate GTX5. This was also observed for the remaining toxins, with the 11-hydroxysulphate dcGTX2 + 3 and GTX2 + 3 presenting higher seasonal fluctuation than the corresponding non-11-hydroxysulphate analogues GTX5 and STX, respectively (data not shown).

Fig. 2
figure 2

Normalised fluorescence response of C1 + 2 and GTX5 (B1) calibration curves obtained with the Agilent 1260 Infinity FLD model between March 2020 and December 2021 (N = 90)

Full size image

An environmental factor that varies across seasons is air temperature, which presents the lowest levels in January, slowly increasing until July (IPMA 2024b). The normalised responses were analysed against the monthly average air temperature in Lisbon during 2020 and 2021 (Fig. S2). Fluorescent responses increased with ambient temperature for all toxins, but all the 11-hydroxysulphate analogues suffered larger variations with air temperature than their non-11-hydroxysulphate counterparts (Fig. 3).

Fig. 3
figure 3

Normalised response of calibration curves (slope) obtained in 2020–2021 distributed accordingly to monthly average air temperature: a dcGTX2 + 3, b C1 + 2, c GTX2 + 3, d dcSTX, e GTX5, f STX. Line represents third-degree polynomial fitting; N = 90; p < 0.001 for all regressions)

Full size image

Average daily temperature can vary by circa 5 °C over 1 month, either upward or downward. Even in months when a downward trend was expected, the opposite behaviour could be registered, with a 5 °C upward trend, as observed recently in November–December 2021 (Fig. S2; IPMA 2024b). This coincided with the reversal of fluorescence response at the end of 2021 (Fig. 2 and S1a). In January 2021, air temperature was 1° below average (10.3 °C versus 11.3 °C), and this coincided with the minimal fluorescence responses observed for the period (Fig. 2 and S1a).

In 2022, an older Agilent 1200 FLD model was used in the monitoring programme combined with the remaining 1260 series modules. With this detector, the same temperature-dependent fluctuation in calibration slopes was observed (Fig. S3). The average increase in the fluorescence response above 20 °C of monthly average air temperature was reported in Table 1. All the 11-hydroxysulphate analogues presented the greatest increase with ambient temperature.

Table 1 Variation of the fluorescence response depending on the exterior air temperature and the number of times of use of the calibration stock mix after multiple freeze–thaw cycles
Full size table

Due to the extensive usage of the peroxide screening and the scarcity of PST episodes, few curves of N1-OH toxins are run in the monitoring programme. A major G. catenatum bloom event took place across the south-west coast of Portugal during September–October 2018, which intensified toxin testing that autumn and the following winter (IPMA 2024a). The best consecutive set of curves of N1-OH toxins was from dcNEO and presented also maximal response during summer (Fig. S1b). The normalised fluorescence responses of dcNEO increased with ambient temperature (Fig. 4a), similar to those of N1-H toxins (Fig. 3).

Fig. 4
figure 4

a Fluorescence of dcNEO calibration slopes versus monthly average temperature (line represents third-degree polynomial fitting; N = 30; p < 0.001). b Relative response of five slopes after first usage of the mix (session nº 1) and after 1–2 weeks (session nº 2)

Full size image

Other possible sources of variation were considered, like detector stability or calibrant stability. Data obtained with the second Agilent 1260 Infinity I FLD (2020–2021) was assessed by comparing the highest calibration standard injected at the beginning of each sequence with its second injection, run circa 10 h later (Fig. 5). The average response losses were for the 11-OSO3− analogues (dcGTX2 + 3, 3%; C1 + 2, 2%; GTX2 + 3, 2%) and for the 11-H analogues: dcSTX, 4%; GTX5, 2%; STX, 3%. The average loss of the 20 sequences and the 6 toxins analysed here was 3%.

Fig. 5
figure 5

Relative fluorescence response of several toxins from the highest calibration standard, injected in duplicate circa 10 h apart from its initial injection (N = 20)

Full size image

The calibration curve standard’s mix was prepared by diluting a calibration stock, which was designed to be used twice only (approximately 2 months’ duration, 2 freeze–thaw cycles). The calibration mix itself was designed to be used a maximum of five times (approximately 1-month duration, 5 freeze–thaw cycles), and it was often observed that the first usage produced the largest fluorescence response. Each usage of the mixes implied imposing a new freeze–thaw cycle on the solutions prepared in the oyster matrix. Twenty sets of consecutive oxidation sessions obtained between 2018 and 2021, when the calibration mix was used five times within 1 month, were normalised against the first usage (first session) (Fig. 6).

Fig. 6
figure 6

Relative response (slopes) of twenty calibration mixes after first usage of the mix (session nº 1) and the fifth session, all obtained within one month: a dcGTX2 + 3, b C1 + 2, c GTX2 + 3, d dcSTX, e GTX5, f STX

Full size image

The largest drop in the fluorescence response of the calibration slopes was already obtained in the second session, with a much slower decline up to the fifth session (Fig. 6 and Table 1). The decline on the second session was more pronounced with the 11-hydroxysulphate analogues (dcGTX2 + 3, 94%; C1 + 2, 94%; GTX2 + 3, 95%), than for the 11-H analogues: dcSTX, 96%; GTX5, 98%; STX, 98%. Regarding the stability of the dcNEO calibration curve standards, the fluorescence response declined after the first usage of the calibration mix down to 82% (Fig. 4b).

None of these drops in fluorescence from detector usage or stability of calibration solutions could explain the much larger fluctuations in fluorescence response across the year. To further study the effect of temperature on the oxidation reaction, a set of three temperatures was chosen: the ambient temperature in winter of 18 °C and an increase to 22° and 26 °C, by pre-heating the reactants and also the reaction mix (Fig. 7). The N1-H toxins, oxidised with peroxide, presented a progressive increase in fluorescence yield with temperature. The N1-OH toxins, oxidised with the periodate mixture, did not present a progressive increase under the same temperature conditions. Also, all the N1-H toxins which possess the 11-OSO3− group presented a stronger fluorescence increase than their 11-H counterparts (Fig. 7).

Fig. 7
figure 7

Normalised average fluorescence response of several toxins oxidised at ambient temperature (18 ºC) or heated in a heating block (22ºC, 26 ºC)

Full size image

As PSTs can be oxidised at even highest temperatures, a setpoint of 50 °C was chosen. There was an increase in fluorescence for all 11-OSO3− group as obtained at 26ºC, but dcGTX2 + 3 presented an increase of up to 800% (Fig. 8a–b). In addition, increased temperature caused a reduction of impurities at the front of the chromatogram, which otherwise had not been removed by the solid-phase extraction step (Fig. 8a–b). The increase in fluorescence yield, besides increasing the detection level, improves spectral confirmation of PSTs. For the example of dcGTX2 + 3 in Fig. 8c–f, this improvement was more relevant in the emission than in the excitation spectra. The spectra obtained from the highest level of the peroxide calibration standards were supplied in Fig. S4 for comparison.

Fig. 8
figure 8

Fluorescent response of toxins spiked in mussel matrix at LOQ level: oxidised (a) at room temperature or heated (b) 4 min at 50 ºC. Quantification peaks for 1-dcGTX2 + 3; 2-C1 + 2; 3-dcSTX, 4-GTX2 + 3; 5-GTX5; 6-STX. Dashed line: impurities not removed by C18 SPE. Excitation and emission spectra of dcGTX2 + 3 at room temperature (c) and (e) or heated (d) and (f)

Full size image

The excessive use of heat during peroxide oxidation can alter the major oxidation products produced by some toxins. Namely, GTX2 + 3 oxidises as a single major peak (coded #8 according to the elution order proposed by Lawrence et al. 1996), while GTX1 + 4 after periodate oxidation originates two major peaks (coded #3 and #8 in the Lawrence method) (Fig. S5a). Heating with peroxide at 60 °C promotes GTX2 + 3 to produce progressively peak #3 instead of peak #8 (Figs. S5 and S6). Heating at 40 °C for 3 min minimises the production of peak #3 to only 8% of total fluorescence, keeping nevertheless peak #8 at maximum levels (Fig. S5b).

Discussion

The AOAC method advises that the stability of the fluorescent products of PSTs is reduced for NEO, B-2, and GTX1,4, where a slow degradation of the oxidation products occurs (about 30% over 8 h). However, the solutions are stable for about 1 day for the non-N1-hydroxylated toxins (AOAC 2005). Later studies analysed stability in a temperature-controlled HPLC autosampler (set at 4°C). These found that the oxidation products of all toxins were stable for up to 24 h and potentially longer (Turner et al. 2009).

Our analysis of stability found a consistent drop in fluorescence response of 3% after circa 10 h. The studies of PSTs carried out at our laboratory with HPLC started with the Hewlett Packard 1046A fluorescence detector, a precursor to the Agilent 1100, 1200 and 1260 series. The manual of this older model informed that the lamp intensity drift is less than 0.3% per hour (Hewlett Packard 1989) or about 3% in a 10-h period. In later models, the manufacturer did not provide details on the drift, such as for the Agilent 1260 FLD (Agilent 2013). The manuals alert only to “wait until the baseline stabilises” (Agilent 2013). For the initial stabilisation, our analysis sequences include an initialisation method to equilibrate the column, followed by a couple of blank runs to stabilise both the column and the detector, before injecting for quantification purposes, with a total of 100 min. The present results corroborate this range of detector drift, but cannot differentiate between the drift of the detector or the drift of the PSTs fluorescent products.

PSTs are stable when frozen in acidic solutions, but losses occur at higher pHs and temperatures (Louzao et al. 1994; Indrasena and Gill 2000). For incorporating recovery into the method, toxins standards were spiked in a bivalve matrix: the C18 cleaned extract. This extract was previously adjusted to pH ~ 6.5 due to method recommendations for improving the oxidation of N1-hydroxylated toxins (AOAC 2005). This higher pH can promote toxin loss in prolonged storage. After five sessions of usage, average toxin recovery was not below 90% for N11-sulphated toxins nor below 94% for non-N11-sulphated toxins (Table 1).

The proponents of this pre-chromatographic oxidation method for PSTs, in their initial reports, used room temperature for both the reactions required (Lawrence and Menard, 1991; Lawrence et al. 1991). Their recorded room temperature was 20 °C. The method was later refined for quantitative determination. It required the introduction of a second clean-up step using an ion exchange (-COOH) extraction cartridge that separated the PSP toxins into three distinct groups for quantitation, overcoming the issue of superposition of oxidation peaks originating from multiple toxins present in natural samples (Lawrence and Niedzwiadek 2001). This method also used room temperature, without any further specifications. The intercollaborative study that led to method validation also used simply ‘room temperature’ among the participants, without further specifications (Lawrence et al. 2005).

During method implementation, several experimental parameters were tested for ruggedness experiments by Turner et al. (2009). They found these parameters did not have a statistically significant effect on method performance, including a variation of ambient temperature during oxidation between 22 and 25°C.

The variations observed due to the stability of the oxidation products/fluorescence detector or, the freeze-thawing cycles, could not account for the higher variations observed across seasons. These were attributed to fluctuations in room temperature. These can pose problems to method validation, as method’s detection and quantification limits can then vary across seasons. It is then recommended that method validation must be performed during the expected lowest temperature season than can be observed throughout the year. Also, it is important for laboratories performing this method to routinely record the room temperature. Another recommendation to minimise the temperature-dependent variation of the oxidation reactions is that a minimal time be required to thaw to room temperature the calibration vials, such as a period of circa 1 h, for example.

The temperature dependence of the oxidation reaction was reported in the early PST studies when the goal was to optimise the fluorescence of N1-hydroxylated toxins, which present low fluorescent yields when oxidised with hydrogen peroxide (Indrasena and Gill 1998). High temperature (85 °C) continues to be used in post-column methods (van de Riet et al. 2011), but its relevance has been forgotten in pre-column methods.

The fluorescent yield of PSTs increases with time at ambient temperature up to 45 min (Indrasena and Gill 1998), but the requirement of tens of minutes is impractical for routine monitoring applications. A warming step of 100 °C produced better results in a shorter time, but is more complicated to manage due to the need for an ice bath to cold quickly because the fluorescent yield can rapidly decrease at this high temperature (Indrasena and Gill 1998). As these experiments used only fluorometric detection and no LC separation was obtained, no data was made available on the partition of the oxidation products by a single compound (as occurs mainly with N1-H toxins C1 + 2, GTX2 + 3, GTX5 or STX) of by two to three compounds (as occurs with most N1-hydroxylated toxins) (Lawrence et al. 1996). This partition between two oxidation products instead of one, as observed here with GTX2 + 3, decreases the sensitivity aimed here with the heating of the reaction.

Temperature dependence of PST’s oxidation is an approach that can be exploited in pre-chromatographic methods to increase sensitivity for trace analysis in natural phytoplankton samples or to improve spectral confirmation at low contamination levels in shellfish. The dual wavelength capability can be used to rule out interfering compounds at a glance (Lopes de Carvalho et al. 2019). Another option is spectral confirmation, and promoting fluorescence yield is a simple and inexpensive procedure to aid in confirmation. This is particularly important for spectral confirmation of dcGTX2 + 3 and C1 + 2, which present the lowest fluorescence response after oxidation at ambient temperature. Their LOQs were 0.06 µM, while for the remaining N-1H toxins, LOQs were between 0.006 and 0.014 µM with our newest 1260 Infinity II detector (Vale 2024).

However, incubation in a dry block heater, for example, is not practical for the quick reaction times used in the peroxide oxidation and for the high number of samples required in a high-throughput monitoring laboratory. The peroxide oxidation performed at room temperature within 2 min is already quite sensitive for routine purposes. For example, in a bivalve sample contaminated only with gonyautoxin-2 + 3, with the Agilent 1260 Infinity II FLD model, these toxins can be quantified at 0.014 µM in the C18 extract or 25 µg STX diHCl eq/kg in the bivalve soft tissues (Vale 2024). This is 30 times lower than the regulatory limit in force around the world (Anonymous 2004).

Despite the highly selective tandem mass spectrometry detectors being more common in food safety laboratories nowadays, these detectors are very expensive to acquire and to maintain and are also prone to matrix interference problems, and even the recently improved methods are not more sensitive than the fluorescence method implemented with the Agilent 1260 Infinity series (Boundy et al. 2015; Vale 2024). For low- and middle-income countries, a modern FLD detector can provide sensitive and selective detection of PSTs to protect public health.