First report of detection of microcystins in farmed Mediterranean mussels Mytilus galloprovincialisin Thermaikos gulf-Greece, using ELISA and liquid chromatography high-resolution mass spectrometry

Microcystins are emerging marine biotoxins, produced by potentially toxic cyanobacteria. Their presence has been reported in aquatic animals in Greek internal waters, while data are few in marine environments. Since the climate change induce eutrophication and harmful algal blooms in coastal marine ecosystems affecting the public health, further research on microcystins’ presence in marine waters is required. The aim of this study was to detect the potential presence of microcystins in mussels Mytilus galloprovincialis in the largest farming areas in Thermaikos gulf, in Northern Greece, and to investigate their temporal and spatial distribution, adding to the knowledge of microcystins presence in Greek Mediterranean mussels. in sampling in in Greece, where the 90% of mussels’ farming activities is located. The isolation of potentially toxic cyanobacteria was conrmed by molecular methods. An initial screening was performed with a qualitative and quantitative direct monoclonal (DM) ELISA and results above 1ng/g were conrmed for the occurrence of the most common microcystins -RR, -LR and -YR, by Ultra High Performance Liquid Chromatography (UHPLC) coupled with a high- resolution mass spectrometer (HRMS) (Orbitrap analyzer). Microcystin -RR and microcystin -LR were detected, while the intensity of microcystin-YR was below the method detection limit. Most samples that exhibited concentrations above 1ng/g were detected during the warm seasons of the year and especially in Spring. Results indicated an overestimation of the ELISA method, since concentrations ranged between 0.70 ± 0.15 ng/g and 53.90 ± 3.18 ng/g, while the conrmation declared that the levels of microcystins were 6 to 22 times lower. in Greece. Their presence was linked to potentially toxic cyanobacteria. Bioaccumulation in digestive gland was observed, while the concentrations in muscles were found extremely low. Samples with levels above 1ng/g were observed mostly during spring, conrming the seasonal distribution of microcystins. The comparison of the results by the ELISA and the LC-Orbitrap MS method indicated an overestimation of concentration by the ELISA method. 75% (75:25 v/v methanol:ultrapure water) was used to extract the microcystins from the soft tissues of the mussels, since it is proposed as the most suitable solvent for the recovery of the toxins under investigation [22, 24]. After the extraction, a clean-up procedure using solid phase extraction (SPE) was [25]. The extracts were evaporated at a The clean-up procedure was performed in a vacuum manifold system (Alltech® Rd, U.S.A), using OASIS HLB (Hydrophilic-Lipophilic Balance) cartridges (200mg/6cc). After the elution, 0.5mL from the fraction was received and dried under nitrogen stream, reconstituted with 2 mL Milli-Q water and analyzed by ELISA. The rest of the cleaned extraction was used for the conrmation by Ultra High Performance Liquid Chromatography - tandem mass spectrometry.


Conclusions
Microcystin RR and microcystin LR were detected for the rst time in mussel Mytilus galloprovincialis, harvested from farms in Thermaikos gulf, in Central Macedonia, in Greece. Their presence was linked to potentially toxic cyanobacteria. Bioaccumulation in digestive gland was observed, while the concentrations in muscles were found extremely low. Samples with levels above 1ng/g were observed mostly during spring, con rming the seasonal distribution of microcystins. The comparison of the results by the ELISA and the LC-Orbitrap MS method indicated an overestimation of concentration by the ELISA method.

Background
Global climate change, nutrient enrichment of water bodies and eutrophication has induced cyanobacterial blooms worldwide [1]. The abundance of potential toxic cyanobacteria species is prone to the production of secondary metabolites, the cyanotoxins. Microcystins (MCs) are the most common group of cyanotoxins, with high toxicity [2] and they are considered as emerging toxins [3]. They are produced by different genera of freshwater cyanobacteria, such as Microcystis, Anabaena (Dolichospermum), Nostoc, Planktothrix, Chroococcus [4], as well as by marine picoplanktonic species, as Synechococcus and Synechocystis [5]. Microcystins are monocyclic heptapeptides, having a general structure of cyclo-(D-alanine-R1-DMeAsp-R2-Adda-D-glutamate-Mdha), where R1 and R2 are variable L-amino acids [1]. Up to date more than 279 congeners have been recognized [2]. Adda, [(2S,3S,8S,9S)-3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic acid], is a unique amino acid, reported only in microcystins and in a similar group of pentapeptides, the nodularins [6]. The presence of Adda de nes the toxicity of microcystins [7]. Microcystins' toxic effects have been reported in humans, domestic and wild animal and in aquatic organisms [7]. The inhibition of the protein phosphatases PP1 and PP2A, has been described as the primary mechanism of toxicity expression, inducing hepatotoxicity, apoptosis, and necrosis of the hepatocytes [2,8]. Moreover, microcystins have been related to toxic effects in other organs, such as heart [9], kidney [2], intestine [10], lung [11] and brain [12]. Microcystin -LR is considered the most common analogue with the highest toxicity, while microcystin -RR (MC-RR) and microcystin -YR (MC-YR) are coming after [13]. Moreover, it is reported that chronic exposure to MC-LR can lead to primary liver cancer [14,15], and it is considered as potential carcinogenic (group 2B) [16]. The detection of microcystins has been reported mostly in freshwater, as well as in brackish and marine environments [1]. Moreover, their presence in food web and especially in seafood such as shell sh, sh and crustaceans is considered hazardous for the public health [1]. In Greece, the detection of microcystins have been reported in the lake sh Cyprinus carpio, freshwater mussel Anodonta sp., in the freshwater European cray sh Astacus astacus and in amphibian's Rana epirotica skin [17,18]. Regarding the presence of microcystins in marine aquatic organisms, there is only one report in Mediterranean mussels Mytilus galloprovincialis from Amvrakikos gulf, a closed embayment in Ionian Sea [5]. Exports of shell sh and especially Mytilus galloprovincialis mussels, contribute signi cantly to Greek economy. The main farming activity (almost 90%) of Mytilus galloprovincialis, is practiced in the Thermaikos gulf areas of Chalastra, Imathia and Pieria in northern Greece [19]. Thermaikos is a semi-closed gulf, (90 meters maximum depth, surface of 5.100 km 2 ), located in the north-west Aegean Sea, in Central Macedonia Greece. It is enriched by four rivers (Axios, Loudias, Aliakmonas and Gallikos), often inducing eutrophication and harmful algal blooms [20,21]. In this study two methods were applied to detect microcystins in farmed mussels of Thermaikos gulf, ELISA and mass spectrometry. The aim of the study was to investigate the temporal and spatial distribution of the microcystins in mussels Mytilus galloprovincialis, in these mussels' farming areas, as well as to compare the afore mentioned detection methods proposed by the scienti c literature. The results would contribute to the knowledge of microcystins presence in Greek mussels, to the proposed methodology and to protection of public health.

Sampling area
The three territories with the largest farming activity of Mytilus galloprovincialis, Chalastra, Imathia and Pieria were chosen for the investigation. Five sampling points (Kavoura Chalastra − 40 o 32΄20, 12΄΄N '46,13"E) were selected in regard to their environmental and economic importance and also due to the potential sources of pollution from industrial, agricultural, veterinary, and urban waste (Fig. 1). These farms applied long line farming system. Mussels Mytilus galloprovincialis allowed to grow on the long lines for 15 to 18 months and then seven hundred and fty (750) mussels were sampled during 2013 and 2016. Two hundred and twenty samples (220) were collected in spring and in summer, two hundred and ten (210) in autumn and one hundred (100) in winter. During the sampling, physical parameters of the water (temperature, salinity, pH and dissolved oxygen) were measured, using a handheld multiparameter instrument (YSI 556, YSI Incorporated, Ohio, USA). The samples were transported in a portable refrigerator at 4 ± 1 o C and forwarded within 4 hours to the Laboratory of Microbiology and Infectious Diseases, Faculty of Veterinary Medicine, Aristotle University of Thessaloniki. Before processing the shell was cleaned and dead mussels were rejected. The mean length of the sampled mussels was 5.9 ± 0.50 cm (n = 25) and their mean body weight was 6.1 ± 0.64 g (n = 25).

Chemicals and reagents
Methanol water and acetonitrile of LC-MS grade, as well as formic acid 99-100% a.r. were purchased from Sigma Aldrich Co. Canada (Newland and Labrador, Canada). The LC separation of the microcystins was carried out on a C18 reverse-phase column ACCLAIM POLAR ADVANTAGE II 3µm 3×150mm with guard column 2/pk and holder-coupler.
Sample preparation Sampled mussels were deshelled and then their soft tissues were collected.The digestive glands and the muscles were separated, and one hundred and fty grams (150g) of each were weighted. The samples were placed in a sieve to drain, homogenized and 5 ± 0.1g were transferred in a 50mL polypropylene centrifuge tube. An aqueous solution of methanol 75% (75:25 v/v methanol:ultrapure water) was used to extract the microcystins from the soft tissues of the mussels, since it is proposed as the most suitable solvent for the recovery of the toxins under investigation [22,23,24]. After the extraction, a clean-up procedure using solid phase extraction (SPE) was performed [25]. The extracts were evaporated at a nal volume of 2mL. The clean-up procedure was performed in a vacuum manifold system (Alltech® Vacuum Manifold, 2051 Waukegan Rd, Bannockburn, U.S.A), using OASIS HLB (Hydrophilic-Lipophilic Balance) cartridges (200mg/6cc). After the elution, 0.5mL from the fraction was received and dried under nitrogen stream, reconstituted with 2 mL Milli-Q water and analyzed by ELISA. The rest of the cleaned extraction was used for the con rmation by Ultra High Performance Liquid Chromatography -tandem mass spectrometry.

Microcystins' analysis
Detection by ELISA Initially screening was performed by ELISA [26,27,28]. Among several commercial kits, a direct competitive ELISA for microcystins and nodularins, ([(2S,3S,4E,6E,8S,9S)-3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic acid, ADDA)-direct monoclonal (DM) ELISA, was chosen [29]. The analysis was performed in triplicate according to the manufacturers' instructions. The reproducibility and the recovery were evaluated by the analysis of a spiked sample of uncontaminated mussel tissue with microcystin -LR (2µg/g). The analysis of the spiked sample was repeated ten times and the obtained recovery was 89%.The limit of detection (LOD) according to manufacturer was set at 0.10 ng/g of microcystin-LR. All standards (0.15ng/g to 5ng/g) were measured in duplicate, as well as control (0.75 ± 0.185 ng/g), calibrator (standard 0) and the spiked sample. The absorbances in the microplate of the kit were measured at 450nm with a spectrophotometer (DAS model A3, Italy). A standard curve was used to calculate the values and the results were expressed as microcystin-LR equivalents (µg/kg wet weight). Samples with concentrations higher than 1 ng/g were further analyzed by LC-Orbitrap HRMS.

Con rmation by LC-Orbitrap HRMS
The con rmation of the ELISA results was obtained by using certi ed standards of microcystin -RR (MC-RR), microcystin -LR (MC-LR) and microcystin -YR (MC-YR). The analysis was performed using a high-resolution Q Exactive Focus Orbitrap mass spectrometer equipped with a heated electrospray ion source (HESI) operating in positive ionization mode, coupled to an Ultimate 3000 ultra-high-performance liquid chromatography system (Thermo Scienti c, USA). A reverse phase C18 column Acclaim Polar Advantage II 3 µm3×150 mm in conjunction with a guard column 2/pk and a Holder-Coupler was used to separate the analytes (Thermo scienti c, USA). The mobile phases were water (A) and acetonitrile (B), both acidi ed with 0.1% formic acid. The chromatographic conditions were the following: column temperature 40 o C, injection volume 20µL, ow rate 0.3mL/min and total run time 11min. The gradient elution program is shown in Table 1.  The HESI source parameters were the following: capillary temperature 320 o C, spray voltage 3.5 kV, auxiliary gas heater 413 o C, sheath gas ow rate 48 au, auxiliary gas ow rate 11 au, and sweep gas ow rate 2 au. The S-lens RF level was set at 50.0. A full-scan MS acquisition was carried out in a mass range of 400-1,100 (m/z). The resolution was set at the maximum available value (70,000) and the spectrum data were obtained using the centroid algorithm. The maximum injection time (IT) was set to auto and the AGC target (automated gain control) to 1e6 ions. For the ddMS2 (data-dependent) fragmentation, a stepped collision energy (30, 50, 70 eV) was applied. The instrumental limit of quanti cation (LOQ) was 0.017 µg/L for microcystin-RR, 0.084 µg/L for microcystin -LR and 1.495 µg/L for microcystin -YR. Raw data were processed with the aid of Xcalibur software, version 4.1. (Thermo scienti c, USA) and the results were reported as total microcystin concentration (µg/kg wet weight).

Molecular detection of bacterial species
The potentially toxic cyanobacteria that were cultured from water samples of the sampling areas as described previously [30], were subjected in DNA isolation using the QIAamp DNA Mini Kit (Qiagen, USA) following the manufacturer's protocol. The quality and quantity of the extracted DNA was evaluated in a NanoDrop spectrophotometer (Shimadzu, Japan). Approximately 50 ng of extracted DNA were utilized in a PCR using the MyTaq™ Red Mix (Bioline, UK) and the modi ed primers 27f-CM and 1492r [31] that amplify a part of the 16Sr RNA bacterial genome. PCR

Statistical analysis
Both parametric and non-parametric statistical methods were applied for the statistical evaluation of the data. The assumptions of normality and homogeneity of variances for the microcystins' concentrations of digestive gland were tested using the Shapiro Wilk and Levene's test, respectively. In case of normality and variances' homogeneity, one way ANOVA was performed to evaluate possible mean effects of area and season on the microcystins' concentrations. In cases where the assumptions of variability and/or normality of the populations' distribution were seriously violated the Kruskal-Wallis non -parametric test was applied to evaluate group differences, while differences between speci c groups were evaluated using the Mnn-Whitney U-test. All analysis were conducted using the IBM SPSS statistical software (version 25.0). Signi cance was declared at p-value ≤ 0.05.

Results
Microcystins above 1 ng/g were detected by ELISA in 52 samples out of 750. On the other hand, only 15 of them were con rmed by LC-Orbitrap HRMS. The identi cation was based on the presence of at least one peak corresponding to the pseudomolecular ion of the microcystin, at a retention time (t R ) matching to that of the certi ed standards with a drift < 0.2 min, as well as the determination of the exact mass with a mass error (Δ) < 5 ppm. The number of samples above 1 ng/g for both methods are shown in Table 2. No microcystins were detected during winter (levels lower than the LOD of both methods). Moreover, concentrations higher than 1 ng/g were detected only at the digestive gland samples, on the contrary all muscle samples were below the limit of detection. Table 2 Temporal and spatial distribution of microcystins in farmed mussels in the sampling areas in Thermaikos gulf, analyzed by ELISA and LC-Orbitrap HRM. Numerators declare the number of the samples, where microcystins above 1 ng/g were detected.
Denominators declare the total number of the samples per area and season.
Also, the presence of cyanobacteria was con rmed molecularly, by the visualization of one clear band of approximately 1400 bp in size in all tested samples.
The quantitative and qualitative analysis with ELISA showed statistically signi cant difference between areas at the same season and also in the same area in different seasons. The temporal and spatial distribution of microcystins' concentrations are given in Table 3. Values with different small superscript in the same row indicate difference at the 5% signi cance level.
Values with different capital superscript in the same column indicate difference at the 5% signi cance level.
The microcystins that were identi ed by LC-Orbitrap HRMS were microcystin -RR and microcystin -LR in traces. The identi cation of the compounds was based on the detection of the molecular ion(s) that were the monocharged and bicharged ions of the initial molecule (Table 4) as well as the isotopic pattern matching, while the con rmation was based on the existence of a characteristic fragment. In the case of microcystins, the typical fragment corresponds to ADDA, with a theoretical accurate mass 135.0806. Figure 2 represents the steps for the identi cation and con rmation. It is noteworthy that the toxicity of microcystins is manly attributed to the presence of the uncommon amino acid ADDA. No microcystin -YR was detected in all samples. The (t R ) of the certi ed standards were 6.63 min for MC-RR, 7.56 min for MC-LR and 7.46 min for MC-YR (Table 4). Figure 3 shows the chromatograms of the certi ed standards, used in the method. Concentration of the con rmed samples were lower than the ones found by ELISA ( Table 5). The highest levels were observed in mussels harvested near the Delta of Axios river in May and in Klidi (Imathia) in June (Fig. 4, 5). In the rst case concentrations in the digestive gland were 6.25 ng/g of MC-RR and 1.35 ng/g of MC-LR and in the second 1.08 ng/g of MC-RR and 0.15 ng/g of MC-LR. Regarding the comparison between the two methods, an overestimation was observed by ELISA. Ions from which the fragments were generated are shown in bold Table 5 Microcystins' concentration in the digestive gland and in the esh of farmed mussels Mytilus galloprovincialis in Thermaikos gulf.

Sampling areas Digestive gland Flesh
Mean ± SD (ng/g). Mean ± SD (ng/g Nd: not detected at levels higher than limit of detection (LOD).
Values with different superscript indicate difference at the 5% signi cance level.

Discussion
The obtained results revealed the presence of microcystins in farmed mussels Mytilus galloprovincialis, in Thermaikos gulf for rst time, while concentrations above 1 ng/g were found in digestive gland. Their detection was linked to the presence of potentially toxic cyanobacteria. Microcystins in mussels and other seafood have been detected globally [32]. Microcystins -LR and -RR are considered the most common [33], while MC-LR has been described as the toxin most detected in mussels, with MC-RR and MC-YR following [13,34]. In a survey in freshwater bivalves Sinanodonta woodiana, Sinanodonta arcaeformis, and Unio douglasiae from a wetland in South Korea [35], it was found that the concentrations of MC-RR, MC-LR and MC-ΥR, were 11.2 to 70.1 µg/g dry weight in the muscles and 168.9 to 869 ng/g dry weight in the digestive gland. The same microcystins were detected in a similar study in freshwater mussels (Cristaria plicata, Hyriopsis cumingii and Lamprotula leai) in lake Taihu in China [36], at high concentrations mostly in the digestive gland, reaching up to 38,478.1 ng/g dry weight.
Also, MC-RR, MC-LR and MC-YR were detected for the rst time in lake Dau Tieng in Vietnam, in mussels Corbicula sp. and Ensidens sp. at concentrations 1.54 ± 0.21 and 3.15 ± 0.65 µg MC/g dry weight, respectively [37]. Another study in North Baltic Sea, showed that the green mussels Mytilus edulis examined by ELISA, accumulated microcystins up to 21.50 ± 60 ng/g, expressed as equivalents of microcystins and nodularins. At the same study it was reported that these toxins were detected in the liver of the sh Platichthys esus, at concentrations up to 99 ± 5 ng/g; hence no toxins were detected in the muscle [38]. A similar study [39] in mussels harvested in the North-East Paci c Ocean of British Columbia, reports the presence of MC-LR at levels up to 600 ng/g, as measured by LC/MS, and also in mussels imported from Canada and the Netherlands.
Microcystins in Greece have been detected mostly in freshwater mussels linked to the occurrence of cyanobacterial blooms [5,17]. In the freshwater mussel Anotonda sp. in Lake Kastoria, microcystins have been detected at the level of 3.271 ng/g equivalents of MC-LR dry weight, by the protein phosphatase 1 inhibition assay test (PP1IA) [17]. Regarding the Mediterranean mussels Mytilus galloprovincialis, there is only one report in Amvarkikos gulf, in Western Greece, where microcystins were detected after a Synechococcus sp. and Synechocystis sp. bloom [5]. The concentrations of the toxins were 45 ± 2 to 141.5 ± 13.5 ng/g equivalents of MC-LR (ELISA), values that according to the researchers were above the Tolerable Daily Intake (TDI), as it is set by the World Health Organization (WHO).
In Greece, microcystins have been also detected in other aquatic animals' tissues. A research in sh from the lakes Kastoria, Iliki, Kerkini and Pamvotis, and Gallikos river, showed that microcystins were accumulated in the sh Cyprinus carpio, Carassius gibelio, Silurus aristotelis and Perca uviatilis and in the amphibian Rana epirotica. The samples were analyzed by ELISA and the toxins' concentrations were 20 to 1,440 ng/g dry weight in the muscles and 25 to 5,400 ng/g dry weight in visceral tissue [17]. Microcystins were also detected in sh Carassius gibelio in lake Pamvotis. It was reported that the toxins were accumulated mostly in liver samples (124.4 ng/g) and less in muscle samples (7.1 ± 2.5 ng/g), analyzed by ELISA [40]. Similar study was contacted in Cyprinus carpio tissues in lake Karla in Thessaly. It was observed that the highest concentrations were accumulated in liver (732 ± 350 ng/g), while kidneys (362 ± 207 ng/g) and muscles (362 ± 207 ng/g) were following [41].
Our study indicates that microcystins can be bioaccumulated in mussels, a nding that is in accordance with the literature [42,43,44,45,46].
The fact that mussels are lter feeders, enables them to bioaccumulate environmental pollutants [47], like microcystins [48]. A study in the brackish waters of Curonian lagoon in Lithuania, revealed that mussels Dreissena polymorpha (zebra mussels), accumulated microcystins at high concentrations, up to 139 ng/g dry weight analyzed by ELISA and 284 ng/g dry weight by PPIA. The toxins were detected even at low abundances of potentially toxic cyanobacteria periods. It was assumed that the bioaccumulation in mussels could be explained by a secondary contamination by resuspended microcystins' residues in sediment particles [48]. Moreover, cyanobacteria are considered food for mussels, leading to the uptake of intracellular cyanotoxins. Although the detection of hepatotoxins in mussels at high levels due to the uptake of intracellular and free toxins, should be more clari ed. [38] According to our study, the target organ of microcystins bioaccumulation seems to be the digestive gland. Several studies have reported the bioaccumulation of microcystins in the digestive gland of freshwater, brackish and marine mussels, at concentrations higher than of other organs [35,36,42,43]. Moreover, during an experimental contamination, pearl mussels (Hyriopsis cumingii) from an aquaculture of Yueshan village, Ezhou city, in China, were exposed to Microcystis aeruginosa 905 for 15 days. After the analysis by HPLC-UV, it was found that MC-LR was accumulated in hepatopancreas at the highest level of 55.78 ± 6.73 µg/g dry weight, while the concentrations in other organs were 27.88 ± 2.22 µg/g dry weight in gonads, 5.66 ± 0.55 µg/g dry weight in gills and 5.17 ± 0.87 µg/g dry weight in muscles [50].
Except from bioaccumulation, the biomagni cation of microcystins in aquatic animals has been a eld of study for many researchers, leading to different opinions. According to some studies, no biomagni cation is observed, or it is not documented su ciently [36,37,51,52]. Moreover, biomagni cation has been observed more in lipophilic marine biotoxins rather than in hydrophilic ones, as microcystins [48]. On the other hand, biomagni cation was reported in farmed and wild bivalves (mussels, oysters and clams), where concentrations of microcystins up to 107 times higher than the ones in water were reported [53]. Also, it has been reported that biomagni cation of cyanotoxins in aquatic animals could be explained by their evolution in relation to cyanobacteria [54], leading to their ability to develop defense mechanisms against the cyanotoxins [55]. Through this the aquatic animals could bioaccumulate toxins at high levels and thus be carriers in the food chain, contributing to biomagni cation. In a recent meta-analysis of eld studies, biomagni cation and biodilution of microcystins in aquatic foodwebs (zooplankton, mollusks, sh, decapods, turtles and birds) was assessed [56]. The biomagni cation factor (BMF) for microcystins was calculated by the researchers as the ratio between their concentration detected in those organisms and their diet. Biodilution was su cient. Zooplankton showed potential for microcystins' biomagni cation, in uencing their concentrations in the liver of zooplanktivorous shes and carnivorous jelly sh, where high values of BMF were observed. There seemed to be a proportional increase between the duration of consumers' exposure to diet, having high microcystins' concentrations. It was concluded that the exposure to high populations of potential toxic cyanobacteria of aquatic animals, especially those that are part of the human food chain, could be related to biomagni cation.
In our study, results above 1 ng/g were observed mostly during warm seasons and especially in May. This could be explained by the increase of sunlight, the high temperatures in the water and the enrichment of Thermaikos gulf with nutrients from the adjacent rivers. A survey in Greek internal waters showed that the highest levels of MC-LR were observed during the warm months of the year [57]. Microcystin -RR, and microcystin -LR have also been detected at 79.4% and 73.5% respectively in lakes in Greece during warm seasons [17]. In another study by ELISA in lake Koronia, microcystins were detected at higher levels during spring [58]. Moreover, mussels Mytilus galloprovincialis in Amvrakikos gulf accumulated the highest concentrations of microcystins in the same season [5]. Same ndings have been reported in other Mediterranean countries. In Italy, a survey in lakes Garda, Como, Iseo, Lugano and Maggiore revealed that the highest concentrations of microcystins and anatoxin-a were observed during warm seasons and mostly in May and September [59]. Additionally, in South-East Adriatic, in mussels Mytilus galloprovincialis analyzed by ELISA, microcystins were detected up to 256 ng/g. Moreover, they were detected up to 2.3 ng/g in clams Chamelea gallina [60]. A similar survey in Portuguese recreational waters showed that the highest levels of microcystins were recorded during spring and summer, and mostly from April to September [61]. Moreover, it has been reported that the global climate change and the increase of the temperature in water environments induce potentially toxic cyanobacterial blooms and accumulation of microcystins in aquatic animals [1,7].
According to our results a spatial distribution of microcystins in the same area, in different seasons was detected. This could be explained by the different levels of enrichment in the coastal zones by the runoffs of the rivers, carrying nutrients. Axios is near Kavoura in Chalastra, Aliakmonas is near Makrigialos in Pieria and Loudias is near Klidi in Imathia. Their annual runoffs are 158 up to 279 m 3 /s, 73 up to 137 m 3 /sand 5 to 10 m 3 /s, respectively. Large agriculture activity is located in the adjacent areas with the transferred sediments, which are rich in nitrates and phosphates, are up to 500 t/km 2 [62]. The highest amounts of runoffs in Kavoura, Imathia and the estuaries areobserved during spring [62], while in Makrigialos the lowest are observed during summer [63].
Regarding the comparison between ELISA and LC-Orbitrap HRMS, our results indicate an overestimation by ELISA. In particular, during spring the samples above 1 ng/g were 24 by ELISA and 7 by LC-Orbitrap HRMS, in summer 12 and 4 respectively and in autumn 16 and 4, respectively. Moreover, the concentrations obtained by ELISA were 6 to even 22 times higher. According to the manufacturer's instructions the microcystins' standards in ELISA were MC-LR, MC-RR, MC-YR, MC-LF, MC-LW, and the desmethylated [D-Asp3] MC-RR and [Dha7] MC-LR. On the other hand, the LC-Orbitrap HRMS analysis focused on the most common microcystins MC-LR, MC-RR and MC-YR. Although ELISA is considered a suitable screening method due to its high sensitivity, detection of multi analogues of microcystins, rapid results, low cost and lack of ethical issues [26,27,28,29], false positive results due to matrix effects cannot be excluded. Similar observations have also been reported in other studies. In a study in bovine's drinking water, Microcystis aeruginosa cells were added at the abundance of 10 5 cells/mL for 28 days. The microcystins' concentrations in the liver were 0.92 µg MC-LR equivalents/g fresh weight measured by ELISA, although no toxins were detected by HPLC/GC-MS [64]. It was concluded that the levels were 1000 times higher due to matrix effect. In another study during an expansion of a cyanobacterial bloom in farms of Mytilus galloprovincialis in Adriatic Sea, [60] the MC-LR equivalents were 256 ng/g (ELISA) in mussels' tissues. Although only the desmethylated derivative desMe-MC-RR was con rmed by LC/ESI-Q-ToF-MS/MS at the level of 39 ng/g. In other mussel samples an overestimation was also observed (1.5 to 6.5 times higher). It was mentioned that ELISA is a useful screening tool, but results should be con rmed by chromatographic analytical methods tandem to mass spectrometry. In another study in McMurdo Ice Shelf station in Antarctica, researchers from New Zealand examined water samples and cyanobacterial mats [65]. The concentrations of microcystins were up to 8 times higher by ELISA than LC-MS. This overestimation could lead to unnecessary bans in drinking water, so the results should be con rmed by mass spectrometry. A similar study evaluated the results obtained by ELISA and LC-MS/MS [66]. It was proposed that ELISA could be used for screening, since it provides rapid results. Still, it was reported that the nonlinear calibration curve could lead to false positive results, since small differences in absorbances might be translated into big concentration differences. The same authors report that LC-MS/MS provides higher speci city and sensitivity and concluded that in any case the abundance of cyanobacteria and the water treatment should be considered. Also, microcystins' concentrations have been assessed in nontoxic cyanobacterial food supplements, peals and capsules [67]. It was reported that no statistically signi cant differences were observed in capsules, since the concentrations were 43 to 410 ng/capsule and 40 to 425 ng/capsule by ELISA and LC/MS, respectively. In peals the levels by ELISA were lower, 200 to 960 ng/peal, when by LC/MS they were 280 to 1,310 ng/peal. So, an underestimation at the level of 27% was observed by ELISA. This was explained by the different sensitivity of the LC-MS method in some microcystins' analogues. Finally, in a recent study [68] water samples from 31 water ecosystems in Michigan, U.S.A., were examined by both methods, ELISA and LC-MS for microcystins, during July, August, September and October, when cyanobacterial blooms often occur. It was reported that no statistically signi cant differences were observed in July and August. On the other hand, an overestimation by ELISA was noticed in September and October. This was attributed to cross-reactivity with microcystins' degradation products and to the quanti cation using nonlinear calibration curve.

Conclusions
In our study microcystin -RR and microcystin -LR were detected for rst time in mussel Mytilus galloprovincialis, harvested from farms in Thermaikos gulf, in Central Macedonia, in Greece. The presence of potentially toxic cyanobacteria in those areas was con rmed molecularly. Bioaccumulation in digestive gland was observed, while the concentrations in muscles were extremely low. Samples with levels above 1 ng/g were observed mostly during spring, con rming the seasonal distribution of microcystins. Moreover, a spatial distribution among sampling areas was noticed, explained by the uctuations of the rivers' runoffs and the nutrients' enrichments. The analysis by ELISA and LC-Orbitrap HRMS, revealed overestimation of the rst method. In any case the use of ELISA as screening method is recommended to be followed by chromatographic analytical methods and mass spectrometry as con rmatory methods.

Declarations
Ethics approval and consent to participate: Not applicable Consent for publication :Not applicable Availability of data and materials: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no competing interests Funding: no funding Figure 1 Sampling area of the study, in Thermaikos gulf, in Central Macedonia