A simple and sensitive method for the determination of methylene blue and its analogues in fish muscle using UPLC-MS/MS

Abstract A new, simple and sensitive method for determining and confirming methylene blue and its analogues such as azure A, azure B, azure C, thionine, and new methylene blue in fish muscles have been developed. The method is based on acetonitrile extraction followed by extract purification using dispersive solid-phase extraction (dSPE) with basic aluminium oxide (ALN) and solid-phase extraction (SPE) using primary and secondary amines (PSA) sorbent in matrix adsorption mode. The separation and detection of the dyes in the fish extract are achieved within 5 min by ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) using an octadecyl analytical column with a mixture of acetonitrile, methanol and 0.1% formic acid as a mobile phase in gradient elution. The developed method has been in-house validated according to European law. The method recovery for fish muscle was 98.3–103.1%, whereas the decision limit (CCα) was from 0.45 to 0.49 µg kg−1.


Introduction
Methylene blue (MB) is a cationic phenothiazine dye. The analogues of MB are azure A (AZA), azure B (AZB), azure C (AZC), thionine (TH), and new methylene blue (NMB) (Figure 1). They all are widely used as textile and paper dyes. The dyes also have many applications in biology and chemistry, mainly as bacteriological stains and redox indicators. Moreover, MB has diverse medical applications, the most common of which is the treatment of methemoglobinemia and ifosfamide-induced encephalopathy. MB and its analogues are known to have antimalarial and antimicrobial properties (Vennerstrom et al. 1995;Thesnaar et al. 2021). In veterinary medicine, MB is used as a disinfectant, as an agent controlling fish ectoparasites and bacterial infections and for treating nitrite toxicity (Noga 2010). The dye is used primarily in freshwater aquarium fish, but there are some suspicions that MB can also be used in farmed fish to replace malachite green or crystal violet. As a result of N-demethylation, MB is metabolised to its analogues, i.e. azure compounds (AZB, AZA, AZC) and TH (Figure 1) (Verdon and Andersen 2017). Metabolic studies indicated the possibility of forming the colourless leuco-forms of methylene blue and its N-demethylated metabolites however, as it has been documented by Turnipseed et al. (1997) that leuco methylene blue (LMB) is unstable and readily converts back to its coloured parent form. Because of its instability, a standard of LMB is not commercially available. MB can be made colourless by reducing it to LMB in a solution of ascorbic acid; LMB has been detected in standard solution using LC-MS/MS by Xu et al. (2009); however, isolating LMB from fish muscle is impossible because it is reconverted to MB during the extraction and analysis. Residues of MB and its metabolites in edible fish tissues are of food safety concern for the consumer because, as indicated on the European Chemicals Agency (ECHA) website, MB and its N-demethylated metabolites are suspected carcinogens, as shown in different models (ECHA 2023). MB has been classified by the International Agency for Research on Cancer (IARC) as Group 3, i.e. not classifiable as to its carcinogenicity to humans because of inadequate evidence of carcinogenicity in humans and limited evidence in experimental animals (IARC 2016). In addition, the European Food Safety Authority (EFSA) decided that methylene blue should be regarded as genotoxic (EFSA, 2017). NMB is structurally close to MB, bearing two methyl substituents at 2, 8 positions. The dye is a hazardous water pollutant affecting fish and other aquatic life (Tkaczyk et al. 2020). That is why MB and other dyes have never been authorised as pharmacologically active substances for use in veterinary medicinal products in foodproducing animals in the EU, and no Maximum Residue Limits (MRLs) have been established for these compounds (European Commission 2010). It can be used only in ornament fish disease treatment. The group of dyes has been placed on the list of substances to be mandatory monitored in aquaculture animals (European Commission 1996, 2022. Until now, the Reference Point for Action (RPA) in the EU has been defined only for malachite green (European Commission 2019), while for crystal violet and brilliant green Minimum Method Performance Requirements (MMPR) have been recommended by EURLs (EURL 2022). For the rest of the dyes determined in aquaculture products the ALARA (As Low As Reasonably Achievable) approach has to be applied.
Considering the above, it is essential, for the proper protection of consumer health, to have a simple, fast and sensitive method for detecting and confirming residues of MB and other phenothiazine dyes such as AZA, AZB, AZC, TH and NMB in fish muscles. Methods for determining malachite green and crystal violet residues and their metabolites in fish are frequent research topics, but MB in fish has been poorly studied. Often, single phenothiazine dyes are additional analytes for the detection of the triphenylmethane dye (Tarbin et al. 2008;Xu et al. 2012;Chen et al. 2013;Reyns et al. 2014;Amelin et al. 2017;Dubreil et al. 2019;Park et al. 2020;Touchais et al. 2021). The analytical range of existing methods does not cover all phenothiazine dyes and have only focused on MB (Turnipseed et al. 1997;Xu et al. 2009;Chen et al. 2013), MB, AZB and NMB (Tarbin et al. 2008;Reyns et al. 2014;Dubreil et al. 2019;Touchais et al. 2021), MB and AZA, AZB, AZC (Xu et al. 2012;Amelin et al. 2017;Wang et al. 2020;Park et al. 2020;Zhang et al. 2021). Most methods are based on acetonitrile extraction with the following purification of extracts using solid-phase extraction (SPE). Liquid chromatography (LC) coupled with mass spectrometry (MS) is the most important technique for identifying and quantifying phenothiazine dyes in fish tissues. However, there are also methods based on LC with visible detection (VIS) (Turnipseed et al. 1997;Wang et al. 2020). Although LC-MS, compared to LC-VIS, is an expensive option and running an instrument requires expert knowledge, it gives the ability to analyse a sample simultaneously for many compounds and provides information on the structural chemical composition of the analyte which is desirable for confirmation purposes (European Commission 2021).
To date, only one method has been published for the determination of phenothiazine dyes that includes the ability to detect thionine in milk (Munns et al. 1992), but there is no such method for fish muscle. Therefore, the objective of the study was to develop a simple, fast and sensitive method for determining and confirming residues of methylene blue and other phenothiazine dyes, including thionine, in fish muscle using ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS).

Standard solutions
Stock solutions of all dyes and IS were prepared separately at 1 mg ml À1 in methanol, taking into account the purity of the active substances. A mixed standard solution of all six dyes (MB, AZA, AZB, AZC, TH, NMB) and a standard solution of the IS, both at 10 ng ml À1 were prepared in methanol. All standard solutions were prepared in amber glass and remained stable for 3 months when stored at À20 C and 1 month at 4 C (Xu et al. 2012).
The calibration solution standards were prepared at 0, 0.04, 0.08, 0.2, 0.4, 0.8, and 2 ng/ml in methanol, containing 0.4 ng ml À1 of IS. The matrix-fortified standards were prepared at 0, 0.2, 0.4, 1, 2, 4, and 10 mg kg À1 , containing 2 mg kg À1 of the IS by adding the right amount of mixed standard solution of all dyes and the IS to blank fish samples prior to extraction.

Fish samples
Fish used to produce blank samples for method development and validation studies were farmed carp and trout; the fish had never been treated with any phenothiazine dyes. Fish muscles with adhering skin in natural proportions were cut into small pieces, blended and frozen until needed at À20 C. Live trout used for incurred muscle fish analysis were from the same source.

Sample preparation
A 2 g portion of blended fish muscle was weighed in a 15 ml centrifuge tube, and 6 ml of acetonitrile was added. After vortexing, the tube was placed into an ultrasonic water bath at 40 C for 10 min. Next, the tube was centrifuged at 2200 g for 5 min. The supernatant was transferred to another 15 ml centrifuge tube, and an extra 3 ml of acetonitrile was added to the sample for the second extraction. The vortexing, sonication, and centrifugation steps were repeated, and both supernatants were combined. Subsequently, 1 g of ALN was added to the whole supernatant, and vortexing and centrifugation were performed. The supernatant was diluted to 10 ml with acetonitrile, and 2.5 ml of the final extract was loaded on an SPE column filled in with 0.1 g of PSA previously conditioned with 3 ml of acetonitrile. The obtained eluate was then passed through a syringe PTFE 0.2 mm filter and transferred to a vial for the UPLC-MS/MS analysis.

UPLC-MS/MS analysis and quantitation
The UPLC-MS/MS analysis was performed on a Shimadzu Nexera X2 (Shimadzu, Japan) system connected to the QTRAP 4500 triple quadrupole mass spectrometer (AB Sciex Framingham, the USA) operated in positive electrospray ionisation (ESI þ ) mode. The following mass spectrometry conditions were used: Q1 and Q3 at unit resolution; curtain gas set at 20; ion source temperature 400 C, ion source gas (GS1) at 50 and ion source gas (GS2) at 60; ion spray voltage at 1000 V; and dwell time 20 ms for all transitions. Two MRM transitions for each dye were monitored, except for IS, where one MRM transition was used. The compound-dependent parameters for the compounds are shown in Table 1. The chromatographic separation of the analytes was performed on an Agilent InfinityLab Poroshell 120 EC-C18 column (3.0 Â 50 mm, 2.7 mm) with a mixture of acetonitrile and methanol (50:50) (mobile phase A) and 0.1% formic acid in water (mobile phase B) at a flow rate of 0.45 ml min À1 . The gradient conditions were set as follows: 0 min, A ¼ 10%; 2 min, A ¼ 90%; 3 min, A ¼ 90%; 3.01 min, A ¼ 10%; 5 min, A ¼ 10%. The analytical column was maintained at room temperature, and the injected volume was 10 ll.

Method validation
The method was in-house validated according to Commission Decision 2002/657/EC (European  Commission  2002) and Commission Implementing Regulation 2021/808 (European Commission 2021). The following parameters were assessed: working range of calibration curves, selectivity/specificity, repeatability, withinlaboratory reproducibility, recovery, decision limit (CCa), relative matrix effect, and ruggedness. Quantification was based on the ratios between the areas of the analytes and IS peaks. For both standard solution and matrix-fortified calibration curves the working range was estimated and the coefficient of determination (R 2 ) was calculated. To assess selectivity/specificity, extracts of 20 blank fish samples originating from different sources were injected into the UPLC systems. The blank fish samples were fortified with malachite green and crystal violet and next analysed to verify whether they could interfere with the analysis of the target dyes. The repeatability was determined by fortifying six blank samples with the analytes at three concentration levels (0.4, 1, and 2 mg kg À1 ), including the lowest reasonably achievable concentration. The samples were analysed on the same day with the same instrument and by the same operator, and the CV (%) of the fortified samples were calculated. This was repeated on two other days with the same instrument but by different operators and with different batches of reagents, and the within-laboratory reproducibility was evaluated by calculating the CV (%) of the fortified samples on three separate days. The percentage recovery was determined in the same experiment as repeatability by comparing the mean measured concentrations with the fortified concentrations of the samples at each concentration level. The decision limit (CCa) was calculated with a statistical certainty of 1-a (a ¼ 0.01). The CCa was determined by calibration curve procedure according to ISO 11843-1:1997 and corresponded to the concentration at the y-intercept plus 2.33 times the standard deviation of the within-laboratory reproducibility at the intercept. In addition, the CCa was verified by analysing blank samples fortified at the calculated CCa. The relative matrix effect expressed as the matrix factor (MF) was evaluated by comparing a peak area of 20 matrix-matched standards (blank samples fortified after extraction) with a peak area of solution standards at the same concentration (1 mg kg À1 ) with a correction using IS, and the CV (%) of MF (standard normalised for IS) was calculated. To assess ruggedness, the One Factor At a Time (OFAT) approach was used (Kruve et al. 2015), and minor changes to the method were introduced such as different batches of stock solutions, different LC column batches and ages, and different types of fish used as blank samples.

UPLC-MS/MS optimisation
All mass spectrometric experiments were conducted using ESI mode. MS/MS parameters were optimised manually by direct infusion of a 100 ng ml À1 individual standard solution of each dye in methanol. Since all analytes are already positively charged, they were detected directly in the positive ion mode resulting in intensive [M] þ precursor ions. Fragments were identified using a product ion scan in range m/z of 100-350 Da and ramping collision energy with Multi Channel Analysis (MCA). In addition, we used the software ACD/MS Fragmenter to predict mass spectral fragmentation of the dyes and compare the predicted fragments to peaks in an experimental spectrum (ACD/MS Fragmenter 2021).
Considering the selected ionisation type and applying established literature rules, the software generates a list of the predicted fragments. When fragments detected in the experimental data are selected from the list, the software displays detailed information, including molecular structure and monoisotopic mass of the protonated species and information on the neutral loss. For each dye in this study, the two most intense fragments were selected, and their suggested molecular formulae are presented in Figure 2. In the case of MB, the ion at m/z 268 was formed by the elimination of CH 4 from m/z 284, while the ion at m/z 252 corresponded to the loss of C 2 H 8 from m/z 284. The expected fragmentation of MB has already been discussed in the literature (Kim et al. 2013), and this work's results agree with the previous study. Similarly, the loss of CH 4 and C 3 H 5 from m/z 270 was observed for AZB leading to the ions at m/z 254 and m/z 228, respectively. For AZA and AZC, a common mass loss of CH 2 N 2 was observed, which led to the ions at m/z 214 and m/z 200, respectively. TH generated the ions at m/z 211 and m/z 196 corresponding to the elimination of NH 3 and N 2 H 4 , respectively. To meet the confirmation criteria defined in Commission Implementing Regulation 2021/808 (European Commission 2021), multiple reaction monitoring (MRM) modes were employed. The presence of dyes was confirmed through two product ion transitions, more intense for quantification and less intense for confirmation, while one transition was monitored for isotopically labelled internal standard (Table  1). Ion source parameters were optimised manually using split tee infusion. To ensure cleanliness and sensitivity, the curtain gas, temperature, and ion source gas 1 and 2 were set as high as possible with minimal signal loss, while IonSpray Voltage was chosen using the lowest setting that provides the maximum signal. Consequently, the settings presented in the subsection UPLC-MS/MS analysis and quantitation were selected. Moreover, the impact of dwell time was tested and finally set as 20 ms due to the most effective results for each analyte. The chromatographic separation of phenothiazine dyes has generally been performed in reserved phase mode with C18-based stationary phase columns (Tarbin et al. 2008;Xu et al. 2009;Xu et al. 2012;Reyns et al. 2014;Amelin et al. 2017;Dubreil et al. 2019;Wang et al. 2020;Park et al. 2020;Touchais et al. 2021;Zhang et al. 2021), though C8 (Chen et al. 2013), HILIC (Kim et al. 2013) or CN (Turnipseed et al. 1997) were occasionally used. In the study, several experiments were conducted using different brands of C18 columns and different mobile phase compositions to elute phenothiazine dyes in the shortest possible time. An Agilent InfinityLab Poroshell 120 EC-C18 column and a mobile phase containing acetonitrile/methanol (50:50) and 0.1% formic acid were subsequently found to give the most reliable results. Adding formic acid enhancing analyte ionisation resulting in better chromatographic peak shapes. It was also found that methanol added to the organic phase strengthened the retention of the analytes and reduced peak tailing, compared to when only acetonitrile was used. Our experiments are consistent with previous results (Zhang et al. 2021;Touchais et al. 2021). Several tests have been carried out on the influence of the injection volume (between 5 and 20 ll) on the S/N ratio. The results showed an increase in S/N when up to 10 ml was injected, above which this ratio did not significantly improve and a significant loss of peak symmetry was observed. Therefore, 10 ml was decided to be the optimal injection volume. With the optimised chromatography conditions described in the subsection UPLC-MS/MS analysis and quantitation, all compounds eluted in less than 3 min and the overall run time was under 5 min ( Figure 3).

Optimisation of the sample preparation
Current trends in sample preparation techniques for chemical analysis are focused on simplifying this step to reduce the amounts of reagents, costs, and turnaround time. To determine residues of methylene blue and other phenothiazine dyes in fish muscle, we decided to use a solvent extraction for sample pretreatment. According to the literature, acetonitrile is usually selected to extract phenothiazine dyes from fish (Amelin et al. 2017;Dubreil et al. 2019;Zhang et al. 2021). In the context of green analytical chemistry aiming to eliminate and reduce the consumption of organic solvents, the aim of sample preparation optimisation was to use only the necessary amount of acetonitrile to achieve the highest recoveries. Thus, we tested different amounts of acetonitrile (10, 15 and 20 ml) in a single and two-step extraction to extract the dyes from 2 g of fish muscle. It was observed that amounts of acetonitrile over 10 ml did not improve the recovery percentages, and two-step extraction with smaller volumes of solvent was more effective than a single extraction with the same volume of solvent (Supplementary Figure S1). To improve the extraction of the dyes, the sample was sonicated (Xu et al. 2012;Chen et al. 2013;Touchais et al. 2021;Zhang et al. 2021), and different temperatures (20, 30, and 40 C) and extraction time (10, 15 and 20 min) were sequentially tested. The results of these experiments indicated that the most efficient extraction in terms of recovery of all tested dyes was achieved when 2 g of fish muscle was extracted with 6 ml of acetonitrile, followed by vortexing and sonication at 40 C for 10 min and centrifugation at 2200 g for 5 min and reextracted with another 3 ml of acetonitrile, followed by combining both extracts and diluting to 10 ml with acetonitrile. As fish muscle is a complex matrix with many compounds in addition to target analytes, a further clean-up stage was needed to reduce possible sample-related matrix effects. ALN has previously been shown to be effective in isolating phenothiazine dyes from fish muscle (Turnipseed et al. 1997;Tarbin et al. 2008;Xu et al. 2009;Zhang et al. 2021). Compared to classical SPE, dispersive solid-phase extraction (dSPE) allows more contact area between the sorbent and the analyte, and requires lower solvent consumption, permitting more samples to be analysed in a shorter period. Thus, dSPE with different amounts of ALN (0.2, 0.5, 1 g) was tested for acetonitrile extract purification. The best extraction efficiency, ranging from 81 to 98% and the lowest matrix effects, ranging from 78 to 130%, were achieved when 2.5 ml of acetonitrile extract was cleaned-up using dSPE with 1 g of ALN. To further reduce the matrix's fatty undesired components and allow the dyes to remain in the liquid phase, we used PSA which is an alkylated amine sorbent that contains two different amino groups: a primary and a secondary amine, and thus retains fatty and organic acids. We decided to use PSA in matrix adsorption SPE mode as chemical filtration to retain impurities and allow the analytes to pass through the cartridge unretained. We tested different amounts of PSA (0.05, 0.1, 0.2 g) and the results indicated that 0.1 of PSA added to the acetonitrile extract after ALN purification was enough to reduce the matrix effect to a negligible level (93-108%) without further loss of the analytes. The final step of optimising the sample preparation procedure was the selection of the filter to prevent UPLC backpressure problems and to extend the column lifetime. The experiments with different types of filters (nylon, PTFE, PVDF) confirmed observations by Park et al. (2020) that PTFE membrane shows the lowest analyte binding and the highest recoveries due to very few functional groups interacting with analytes.
Summarising, the most efficient extraction for all tested phenothiazine dyes was achieved using 2 g of fish muscle and 10 ml of acetonitrile, followed by vortexing, sonication at 40 C for 10 min and centrifugation at 2200 g for 10 min. The most effective clean-up procedure was obtained when 2.5 ml of acetonitrile extract was subjected to dSPE with 1 g of ALN and SPE in matrix adsorption mode with 0.1 g of PSA followed by PTFE filtration. It must also be pointed out that the evaporation of the extraction solution and the following reconstitution steps were omitted to minimise the loss of the analytes and to save time.

Quantification and method validation
The calibration curves were constructed based on the response of the corresponding ratio of the analyte peak area to IS. Due to the lack of stable isotopically labelled analogues of the analytes, MG-d 5 was chosen as IS as a substance not contained in the sample and having physicochemical properties as similar as possible to phenothiazine dyes. The standard calibration curves generated from the analysis of calibration solution standards for all analytes were linear from 0.08 to 2 ng ml À1 , with R 2 higher than 0.98. In the presence of matrix compounds, matrix-fortified calibration curves (blank samples fortified before extraction) for all analytes showed satisfactory linearity from 0.4 to 10 mg kg À1 with R 2 higher than 0.96. Although no significant matrix influences were observed in the preliminary experiments, the slopes of the two calibration curves differed statistically due to the significant differences in the physicochemical properties of the analytes and IS used. It should be remarked that as the high accuracy was achieved using matrix-fortified calibration curves, it was not necessary to use IS, but in most of the validation guidelines, its use is recommended to take into account any analyte losses during sample preparation. Thus the quantification was carried out using matrix-fortified calibration curves with IS. Selectivity was tested by analysis of 20 blank trout and carp samples originating from different sources and demonstrated no significant chromatographic interference at the retention times of the analytes of interest. The fortification of blank fish samples with the most frequently occurring pharmacologically active dyes in aquaculture products, such as malachite green and crystal violet, did not influence the analysis of the target dyes. The trueness of the method expressed as percentage recoveries, repeatability, and within-laboratory reproducibility in terms of CV, was evaluated for all analytes at three concentrations (0.4, 1, and 2 mg kg À1 ) ( Table 2). Mean recoveries for all dyes were in the range from 98.3 to 103.1% and were all within the acceptable criteria; for mass fractions <1 lg kg À1 CV should be in the range from 50 to 120%, while for mass fractions >1 lg kg À1 to 10 lg kg À1 it should be in the rage from 70 to 120% (European Commission 2021). The CV for all repeated analysis of fortified samples were < 13% for all analytes and met performance criteria; the CV under repeatability conditions shall be equal to or below 20%, while CV under within-laboratory reproducibility conditions shall not be greater than 30% (European Commission 2021). For unauthorised substances, the CCa, as the limit at and above which it can be concluded with an error probability of a (0.01) that a sample is non-compliant, should be as low as reasonably achievable. The CCa determined by calibration curve procedure according to ISO 11843-1:1997 was in the range from 0.45 to 0.49 mg kg À1 (Table 2). A negligible matrix effect, expressed as the MF (normalised for IS), was observed for all dyes as it was in the range from 94 to 109%, with CV not greater than 20% (Table 2). A value of 100% indicates no matrix effect, a value below 100% ion suppression and a value above 100% indicates ion enhancement. Most dyes showed very low ion enhancement, except for MB and NMB, which exhibited slight ion suppression. The method's ruggedness was assessed by introducing minor changes to the method (using OFAT approach), such as different batches of stock solutions, different LC column batches and ages, and different types of fish used as blank samples. These variations did not affect method trueness, precision or sensitivity for any of the analytes, demonstrating the excellent ruggedness of the method.
Comparison of the developed method with other approaches for the determination of phenothiazine dyes in fish muscle Table 3 summarises the methods for the determination of phenothiazine dyes in fish muscle. Compared to other methods, the presented method allows the detection and quantification of a broader range of phenothiazine dyes, including TH, which has never been analysed before in fish muscle. The optimised sample preparation procedure utilises solvent extraction with acetonitrile, followed by clean-up using dSPE with ALN and SPE-PSA in matrix adsorption mode, which has great potential to remove a large number of interferents, resulting in minimal matrix effect. Both sorbents, ALN (Turnipseed et al. 1997;Tarbin et al. 2008;Xu et al. 2009;Zhang et al. 2021) and PSA (Park et al. 2020), have already been applied to determine phenothiazine dyes in fish muscle, but never in the same sample preparation protocol. Moreover, most methods required evaporation and the following reconstitution steps, which have been omitted in the proposed method to minimise analyte losses and save time. The methods developed by Park et al. (2020) and Amelin et al. (2017) applied a similar approach with no evaporation step. Apart from phenothiazine dyes (MB, AZA, AZB, AZC) they covered other pharmacologically active dyes but, at the same time, are less sensitive. LC-MS/MS is the most often used technique for identifying and quantifying phenothiazine dyes in fish, but thanks to UPLC-MS/MS, the proposed method offers more rapid analysis compared to those previously published.
Application of the method to the analysis of incurred fish muscle Trout (n ¼ 3) were exposed to a 2 mg l À1 MB bath for 3 h and then transferred to fresh water for 24 h. At this point, fish were killed, filleted and skinned, and muscle samples were next analysed using the presented UPLC-MS/MS method. The mean concentration of MB in trout muscle samples collected at 24 h after the end of the treatment was 5.7 lg kg À1 . As was previously observed in the incurred muscle of crucian carp (Zhang et al. 2021) and Japanese eel (Wang et al. 2020), our results also indicated the metabolism of MB into AZB and AZA, which were present at that time in trout muscle samples at 5.6 lg kg À1 and 4.5 lg kg À1 , respectively (Figure 4). Although no other metabolites such as AZC and TH were detected in incurred trout muscle, there is still a possibility they might be present in other fish tissues and organs after exposing trout to a methylene blue treatment bath. No such study has been performed for NMB, but it is expected that when used therapeutically in fish treatment, it might result in the presence of the parent dye or its N- deethylated metabolites in incurred fish muscle, as in the case of MB.

Conclusions
A new method has been developed for determining methylene blue and its analogues azure A, azure B, azure C, thionine and new methylene blue in fish muscle with UPLC-MS/MS. The optimised sample preparation procedure utilises solvent extraction with acetonitrile, followed by clean-up using dSPE with ALN and SPE-PSA in matrix adsorption mode that removed a large number of interferents, resulting in minimal  matrix effects. Moreover, evaporation and reconstitution steps were omitted to minimise analyte losses and save time. The separation and detection of the dyes in the fish extract were achieved within 5 min by UPLC-MS/MS using a C18 column with a mixture of acetonitrile, methanol and 0.1% formic acid as a mobile phase in optimised gradient elution. To the authors' knowledge, this is the first method for quantifying thionine in fish muscle. The proposed method has been validated and has been shown to be accurate, precise and sensitive, with the CCa ranging from 0.45 to 0.49 mg kg À1 . In addition, the method has been successfully applied to analyse incurred fish muscle samples and is suitable for analysing fish exposure to methylene blue.

Disclosure statement
Xu YJ, Tian XH, Zhang XZ, Gong XH, Liu HH, Zhang HJ, Huang H, Zhang LM. 2012. Simultaneous determination of malachite green, crystal violet, methylene blue and the metabolite residues in aquatic products by ultra-performance liquid chromatography with electrospray ionization tandem mass spectrometry. J Chromatogr Sci. 50 (7)