Occurrence and diversity of cyanotoxins in Greek lakes

Toxic cyanobacteria occur in Greek surface water bodies. However, studies on the occurrence of cyanotoxins (CTs) are often limited to mainly microcystins (MCs), with use of screening methods, such as ELISA, that are not conclusive of the chemical structure of the CT variants and can be subject to false positive results. A multi-lake survey in Greece (14 lakes) was conducted in water and biomass, targeted to a wide range of multi-class CTs including MCs, nodularin-R (NOD), cylindrospermopsin (CYN), anatoxin-a (ANA-a) and saxitoxins (STXs), using multi-class/variant LC-MS/MS analytical workflows, achieving sensitive detection, definitive identification and accurate quantitation. A wide variety of CTs (CYN, ANA-a, STX, neoSTX, dmMC-RR, MC-RR, MC-YR, MC-HtyR, dm3MC-LR, MC-LR, MC-HilR, MC-WR, MC-LA, MC-LY, MC-LW and MC-LF), were detected, with MCs being the most commonly occurring. In biomass, MC-RR was the most abundant toxin, reaching 754 ng mg−1 dw, followed by MC-LR (458 ng mg−1 dw). CYN and ANA-a were detected for the first time in the biomass of Greek lakes at low concentrations and STXs in lakes Trichonis, Vistonis and Petron. The abundance and diversity of CTs were also evaluated in relation to recreational health risks, in a case study with a proven history of MCs (Lake Kastoria).

SCIeNtIfIC REPORTS | (2018) 8:17877 | DOI: 10.1038/s41598-018-35428-x methods, using a variety of clean-up procedures prior to LC-MS/MS. Using this workflow, several cyanotoxins have been unambiguously identified and quantified for the first time in Greece. Additionally, this workflow provides a monitoring toolkit for laboratories around the world, offering a detailed analytical guide for different types of samples and CTs. The application of this toolkit enables the reliable and accurate determination of a wide diversity of CTs in surface water and biomass, aiming to shape future regulations and guidelines, towards common analytical protocols and standards.
In the case of method (C) for filtered water, a dual-cartridge SPE-LC-MS/MS analysis was performed 74 . Method trueness (% recoveries) and precision (%RSD) have already been reported 74 . Values have been confirmed for the present study and %recoveries ranged 62.3-97.9%, except for MC-LW (47%). LOD values ranged from 0.8 to 6.5 ng L −1 . The LODs of STX and neoSTX in method (D) for the determination of STX and neoSTX in filtered water, were 1.0 and 3.0 μg L −1 , respectively.
The observed effect of the cyanobacterial extracted matrix, on the analysis of the selected CTs (CYN, ANA-a, MC-RR, MC-LR, neoSTX and STX), is shown in Table S4, ranging 7.8-29.5%, except for CYN with an observed average matrix suppression of 68.0%. As far as specificity is concerned, blank samples were analyzed for every method and no interfering peaks were observed close to the retention times (t R ) of the analytes. Linearity was assessed with linear regression analysis giving coefficients of determination R 2 >0.980. (16) was identified in biomass samples collected from Greek lakes, as presented in Table 1 Figure 1a depicts the profile of different groups of CTs i.e. MCs, CYN, ANA-a, STXs identified in the biomass of the studied Greek lakes. Only Lakes Ismarida and Marathonas were free of CTs in the sampled biomass. All the other lakes contained MCs, except for Lake Vistonis (2014) and Lake Kerkini (10/2014). STX was also identified in the biomass of 5 lakes (Kastoria, Kerkini, Trichonis, Petron and Vistonis) and neoSTX was found in 3 biomass samples (Lakes Kerkini, Trichonis and Vistonis). This is the first report of such a wide diversity of CT groups and variants detected in Greek lakes. Lake Kastoria, presented the highest variety of CTs in its biomass (Table 1) throughout the study. In September 2014, 11 MC-variants and STX were detected in the biomass collected by the lake, which is the most diverse cocktail of toxin variants identified in Greece during the study. Lake Doirani (9/2008), also presented a large diversity of toxins in the biomass, including CYN, ANA-a and 4 MCs (dmMC-RR, MC-RR, MC-YR, MC-LR). In Lake Chimaditis during the same period (September 2008), 5 MCs were identified (dmMC-RR, MC-RR, MC-YR, MC-LR, MC-HilR), while Lake Trichonis presented biomass samples with 3 MCs (dmMC-RR, MC-YR, dm 3 MC-LR), STX and neoSTX. In Lake Kerkini (6/2008), 3 MCs (MC-RR, MC-LR, MC-LW), ANA-a, STX and neo-STX. In all other lakes a smaller number of toxins were identified.

CTs in cyanobacterial biomass. A wide variety of intracellular CTs
The most commonly found CT group in the biomass of Greek lakes was MCs (found in 11 out of 14 lakes). MC-RR was the most frequently detected MC-variant (10 lakes), followed by MC-LR, MC-YR and dmMC-RR, which were found in 8, 7 and 4 lakes, respectively.
It should be noted that most of the studies conducted in the past have focused on the analysis of CTs in lake samples containing both biomass and water. In most cases, a certain amount of the sample was filtered and the biomass collected was further extracted for CT analysis. The amount of CTs determined in the filtered biomass, was expressed as μg of intracellular toxins per volume (L) of filtered water. During a cyanobacterial bloom, the surface of water can be abundant with biomass (bloom/scum), however, as cyanobacteria are not homogenously distributed in the water volume, this practice can result in over-estimation of CTs per volume of water. Additionally, this approach gives no information regarding the extracellular (water-dissolved) amount of CTs. In the study presented here, we have selected to report separately the extracellular (free water soluble) toxins found in filtered water (μg L −1 ) and the intracellular (cell-bound) toxins found in lyophilized biomass. The content of intracellular toxins is also expressed in two ways, i.e. nanogram of toxins found per milligram of dry biomass (ng mg −1 dw) and microgram of toxins found in dry biomass collected by a certain volume of water that was filtered (μg L −1 ). In this way, results can be comparable to other studies. Furthermore, both fractions of the toxin content (free and cell-bound) can be clearly discriminated. Table 2 presents the content of intracellular CTs and their variants, found in Greek lakes (ranges in Figure S2). It is evident that biomass samples from Greek lakes are rich in MCs. The results are also graphically presented in Fig. 2a or <LOQ to 132 μg L −1 (Kastoria 9/2016). Nevertheless, some of the samples (i.e. Kastoria) presented extremely high content, which increased the overall mean content of the samples. Other MCs detected in the biomass  of Greek lakes (dmMC-RR, MC-YR, MC-HtyR, dm 3 MC-LR, MC-HilR, MC-WR, MC-LA, MC-LY, MC-LW,  MC-LF) were found in lower levels, as shown in Table 2.
Occurrence of MCs and more specifically the presence of MC-RR, MC-LR and MC-YR in increased abundance in Greek lakes, is in agreement with numerous studies worldwide, showing that the main MC-congeners produced by Microcystis spp. are MC-LR, MC-RR and MC-YR in varying proportions 2,10,20 , while more hydrophobic MCs (e.g. MC-LA, MC-LW, MC-LF) and some desmethyl variants are rarely dominant 45,77,78 . Most of the lakes studied were found to be dominated by Microcystis sp., Aphanizomenon sp., Anabaena sp. and Anabaenopsis sp. which are known MC producers.
Previous studies have also demonstrated the presence of MCs in Greek lakes. In a study carried out by Gkelis et al. 55 cyanobacterial biomass from several Greek lakes was extracted and analyzed, using ELISA, PPIA and LC-DAD 55 . Results were in agreement with the present study, showing that MCs were found in 95% of the samples, with MC-RR and MC-LR representing the predominant MC variants. In a previous study of six Greek lakes and reservoirs 59  The biomass samples that were obtained from Lake Trichonis in November 2008, contained only the desmethyl variants dmMC-RR (reaching 175 ng mg −1 ) and dm 3 MC-LR (reaching 4.93 ng mg −1 ), but not their most commonly-found methylated forms. As shown in Table 1, the lake was dominated by Planktothrix rubescens and Aphanizomenon flos-aquae in the shallow layer while the maximum abundance of P.rubescens was observed in deeper water 79 . To the best of our knowledge, this is the first report of a Greek lake presenting the maximum abundance of P. rubescens in deep waters. The presence of desmethyl variants is in agreement with past studies suggesting that Planktothrix and some Anabaena sp. tend to produce mainly desmethyl MC variants, namely  [80][81][82] . Nevertheless, the presence of the desmethyl MC variants could show periodic trends in a lake, alternating between periods in which they represent the only variants and periods when they coexist with other variants of MCs. These variations have been attributed to changes in chemotypes' composition of Planktothrix rubescens in relation to certain environmental variables 82 . The absence of the methylated variants could be attributed to the total absence of Microcystis sp., Cylindrospermopsis raciborskii or Anabaena flos-aquae, which are usually responsible for the production of methylated MC variants ( Table 1).
The present study for Lake Pamvotis, showed very low intracellular MCs content, reaching a maximum of 7.51 ng mg −1 dw during October 2014 or 0.751 μg L −1 . Furthermore, CYN was detected for the first time in this lake at trace concentration. In the past 83 , cell bound MCs were found in the same lake to reach a maximum of 19 μg MC eq. L −1 , while free extracellular MCs were detected with the same method, reaching a maximum of 9 μg MC eq. L −1 . Another study for the same lake using ELISA for the detection of MCs has shown that extracellular MCs ranged from 0.012-7.88 μg MC-LR eq. L −1 , while intracellular MCs ranged 0.15-15.21 μg MC-LR eq. L −1 54 . Similar results were also obtained in an earlier study of the Lake Pamvotis 58 , where the measured concentration using ELISA ranged from 0.01 to 19.5 μg L −1 .
Marathonas, is a drinking water reservoir of Athens, which seasonally suffers from cyanobacterial blooms 84 . In the present study, filtered samples did not contain traceable amount of cyanotoxins, in contrast to past results from our research group 69 Table 3 shows the various CTs found to be present in water. Figure 1b depicts the diversity of different groups of CTs identified in the studied lakes. Most of the analyzed filtered water samples contained mainly MCs.
Lake Kastoria presented the highest number and diversity of CTs in nearly all the water samples obtained throughout different time periods (Fig. 1b). The highest variety of MCs in Lake Kastoria, was found during October 2015, when 8 different MCs were detected (dmMC-RR, MC-RR, MC-YR, dm 3 MC-LR, MC-LR, MC-HilR, MC-WR, MC-LA). In the water samples from Lake Kastoria obtained during 2014 and 2016, a large variety of MCs was also found. Table 3     MC-LR were detected. Lake Kerkini (October 2014), Marathonas reservoir (9/2013) and Lake Vistonis (9/2014) did not contain detectable amounts of the targeted CTs. MC-RR was detected in five lakes and was the most abundant toxin in the filtered water samples. The concentration of the water samples are given in Table 4 60 .

Sample STX neoSTX CYN ANA-a dmMC-RR MC-RR NOD MC-YR MC-HtyR dm 3 MC-LR MC-LR MC-HilR MC-WR MC-LA MC-LY MC-LW MC-LF
In Lake Pamvotis several MC variants (dmMC-RR, MC-RR, MC-YR, MC-LR) were detected in water samples at relatively low concentrations (MC-LR at 131 μg L −1 ). The filtered biomass collected from the same sample (10/2014) contained only MC-RR and MC-LR at low concentrations, indicating the presence of a late bloom in the lake with progressive cell lysis, releasing CTs to the aqueous environment. In the past, extracellular TMCs were detected ranging from <1 μg L −1 eq. to 9 μg L −1 eq. 83   The case of Lake Kastoria. In the case of Lake Kastoria, a series of samples were collected in different time periods, during the years 2007, 2014, 2015 and 2016. The analytical results reveal the consistent presence, increased diversity and abundance of various CTs in the biomass and filtered water of the lake, throughout the sampled periods (Fig. 3).
All the targeted toxins, except for CYN, ANA-a and neoSTX, have been identified throughout the different sampling periods in the biomass of the lake. MCs were the dominant CT group (Tables 1 and 2). The main MC present was MC-RR with a mean content of 516 ng mg −1 dw or 51.1 μg L −1 (range 24.7-754 ng mg −1 dw). The median content was 632 ng mg −1 or 63.2 μg L −1 (ranges in Fig. S4). MC-LR also occurred at high content levels (13.4-458 ng mg −1 dw) with mean and median values, 323 and 382 ng mg −1 , or 32.0 and 38.2 μg L −1 , respectively. MC-YR was present at increased levels (2.77-128 ng mg −1 dw) or 0.045-12.8 μg L −1 . The biomass samples from Lake Kastoria also contained significant amounts of the desmethyl variants dmMC-RR and dm 3  Apart from MCs, other CTs such as STX was detected in the biomass of Lake Kastoria, in the samples of September 2007 (6.10 ng mg −1 dw) and September 2014 (1.40 ng mg −1 dw). CYN and ANA-a were not detected in the biomass of lake Kastoria, although the presence of ANA-a was detected at trace concentration in filtered water during September 2016 (Table 4).
Regarding the extracellular MCs determined in the filtered water from Lake Kastoria, the main MC variant found was MC-RR, at concentrations ranging from 0.092 to 338 μg L −1 (ranges in Fig. S5). The mean concentration of the toxin was 75.0 μg L −1 and the median was 24.4 μg L −1 . MC-LR was also present consistently in all samples at concentrations ranging from 0.072 to 354 μg L −1 . The mean and median concentrations of MC-LR are 66.7 and 14.0 μg L −1 , respectively. MC-YR was present at concentrations ranging 0.013-80.7 μg L −1 . The desmethyl forms dmMC-RR and dm 3 MC-LR were present at lower concentrations ( Table 4).

Ratio of concentrations (intracellular MC-LR/MC-RR).
The ratio of MC-LR/MC-RR is known to change in relation to the available total phosphorus concentrations, light intensity and NO 3 -N concentration 86 . The effect of various environmental parameters on the ratio of the produced MC congeners and their relative abundances has been evaluated by a few studies, suggesting that the cellular composition of MC variants may change in response to changing environmental conditions, such as temperature 87 , light intensity and nutrient supply 88 , photon irradiance 47 or amino acid availability (leucine and arginine) 89 . The ratio of different MC-congeners is strongly related to the dominant environmental parameters of a surface water body, the dominant cyanobacterial species and their growth stage. Nevertheless, the precise mechanisms determining the composition of MC   47 . The toxicity of MC-LR is far higher than that of MC-RR 90 , therefore low values of MC-LR/MC-RR ratios in combination with low TMC concentrations are desirable. Content ratios of two common MCs (intracellular MC-LR/MC-RR) were calculated for Lake Kastoria samples, collected throughout the sampling periods and found to range from 0.5 to 1.1 (Fig. 4). The situation was similar in other Greek lakes (Pamvotis, Mikri Prespa, Vegoritis and Doirani), where both these MC variants were detected. The production of certain MC variants and their ratio is largely related to the toxin producing cyanobacterial species present in the sample. In contrast, Lakes Kerkini (2008), Chimaditis and Petron presented different ratios (4.1, 0.2 and 2.3, respectively), shown in different colour in Fig. 4, which was accompanied by more diverse cyanobacterial assemblancies ( Table 1). The highest intracellular MC-LR/MC-RR ratio values were recorded in lakes Kerkini and Petron and coincided with the highest cyanobacterial species diversity per sample. More specifically, in Lake Kerkini the dominant cyanobacteria were Aphanizomenon flos-aquae, Microcystis aeruginosa, Microcystis panniformis and several species of Nostocales (Cylindrospermopsis raciborskii, Anabaena cf. viguieri, Anabaenopsis elenkinii, Anabaena flos-aquae, Anabaena aphanizomenoides) and Oscillatoriales (Pseudanabaena limnetica) (Fig. S6), while Lake Chimaditis was dominated by Microcystis panniformis, Microcystis aeruginosa, Microcystis wesenbergii, Microcystis flos-aquae, Cylindrospermopsis raciborskii and Anabaena flos-aquae. Finally, Lake Petron was dominated by Planktothrix agardhii, Cylindrospermopsis raciborskii, Cyanodictyon imperfectum, Aphanizomenon gracile, Anabaenopsis elenkinii, Planktolyngbya limnetica and Planktolyngbya microspira (Fig. S7). The differences in cyanobacterial diversity could possibly explain the variation of MC-LR/MC-RR content ratios of these lakes. Also, light conditions in shallow lakes have been known to influence certain cyanobacterial species, e.g. Planktothrix agardhii, is enhanced at higher irradiance conditions. This species is known to produce less arginine-based MCs (MC-RR) and more leucine-based MCs 47 . This could possibly explain the different content ratios in a shallow lake, such as Lake Petron.
In the past, Gkelis et al. 55    Based on the probability of adverse health effects of the TMC content, each lake sample was categorized according to the classification proposed by WHO (Table S5). Figure 5 shows that seven of the analyzed samples contained MCs at a concentration that poses a high risk of adverse health effects, including lakes Kastoria, Pamvotis, Zazari, Mikri Prespa and Vegoritis. The sample from Lake Trichonis posed low to moderate risk of inducing adverse health effects.

MC-LF
Regulations for the presence of CTs in surface and drinking water nowadays, mainly accept the oral route (ingestion of toxins via drinking-water, recreation or consumption of fish) as the main vehicle of CT exposure 39,40,42 . The Tolerable Daily Intake (TDI) for the average adult or child, is described as the amount of a potentially harmful substance (in this case MCs) that can be consumed daily over a lifetime, with negligible risk of adverse health effects 14 . WHO derived that the TDI of MCs for humans should be 0.04 μg kg −1 bw 43,92,93 . TDI values for other CTs have also been set, e.g. by the OPHD which has also set TDI values for CYN, ANA-a and TMCs (0.03, 0.1 and 0.05 μg kg −1 bw, respectively) 42 (Table S5). Table 5 shows the minimum volume of surface water (mL) that has to be accidentally consumed by a child or adult swimmer at the sampling points, in order to reach the tolerable daily intake TDI set by WHO and OPHD. In Lake Kastoria during 2014, as low as 2.55 mL of lake water was enough to reach the TDI (for MCs) by an adult swimmer and only 0.42 mL by a child. Similar is the case of lakes Pamvotis, Zazari, Mikri Prespa and Vegoritis, where consumption of only a few mL is enough to reach the maximum TDI for MCs. This means that if an average adult swimmer accidentally consumes 200 mL of lake water in the sampling point, then the water of Lake Kastoria (10/2014) would provide 7859% of the TDI (Table 6). These values indicate the risk associated to the use of this water for daily recreational activities.

Diversity of Cyanobacteria in Greek lakes. Microscopic analysis revealed the diversity and abundance
of cyanobacteria in Greek lakes throughout the course of the study (Figures S6 and S7). In total, 55 different cyanobacterial species were identified, originating from the three main orders (Chroococcales, Oscillatoriales and Nostocales) with representatives in the plankton, thoroughly described in Table 1.
The largest diversity of cyanobacterial species was found in the shallow Lake Petron (20 different species), followed by Kerkini Reservoir (19 species) and Lake Chimaditis (13 species), while the lowest diversity were observed in Lake Mikri Prespa (3 species) and Lake Vistonis (3 species). One of the most commonly occurring More specifically, in Lake Kastoria, blooms were dominated by Μicrocystis aeruginosa, Microcystis panniformis as well as Anabaena (e.g. A. flos-aquae) which could be responsible for the production of the highly diverse and abundant CT content ( Table 1). The co-existence of various other cyanobacterial species, could have contributed to the increased production of MCs as well as STX. The dominant cyanobacterial species however belonged to Microcystis and Anabaena genera, which are expected to produce a high diversity of MC variants. However, the analysis of biomass samples including mixed cyanobacterial species, does not allow the safe discrimination and univocal correlation of the species responsible for the production of each CT.
Lake Doirani also presents a high diversity of MC variants (Fig. 1a) as well as CYN and ANA-a. Cyanobacterial species Μicrocystis aeruginosa, Anabanena flos-aquae, Α. aphanizomenoides were dominant in the lake bloom, while Cylinrospermopsis raciborskii, Cuspidothrix (Aphanizomenon) issatschenkoi and Aphanizomenon gracile were also found in lower biovolume. These species' association could be responsible for the diversity of CTs present in the bloom (Table 1).
In Lake Chimaditis, a wide variety of MCs was observed (dmMC-RR, MC-RR, MC-YR, MC-LR, MC-HilR). The species Μicrocystis aeruginosa, Microcystis panniformis and Microcystis flos-aquae were dominant in the sample and possibly contribute to the production of MCs, while Cylindrospermopsis raciborskii, Aphanizomenon issatschenkoi and Anabaena flos-aquae were also present in lower quantities. The occurrence of Merismopedia could also contribute to the production of various MCs 97 .
In Lake Petron, only MC-RR and MC-LR were identified, although the lake water presented the highest cyanobacterial diversity. MC-producing species such as Μicrocystis aeruginosa, Microcystis panniformis and Microcystis flos-aquae were absent from the lake bloom. Microcystis wesenbergii contributed to the cyanobacterial biomass, although it has been reported as a non-producing MC species 96 . Cylindrospermopsis raciborskii, Cyanodictyon imperfectum, Planktolyngbya limnetica, Planktolyngbya circumcreta and Aphanizomenon gracile were found to be the dominant species (Fig. S7). In the past Cylindrospermopsis raciborskii, Aphanizomenon, Anabaenopsis, Planktothrix as well as Aphanocapsa and Merismopedia have been related to the production of several MCs 97 .
In Lake Vegoritis, MC-RR and MC-LR as well as CYN, were detected, while the dominant species was Μicrocystis aeruginosa (Table 1). Aphanizomenon and Anabaena were also present in the bloom.
Desmethyl MCs have been reported in the past as the main MC variants produced during blooms dominated from Planktothrix and Microcystis 80 . In lake Trichonis, a characteristic bloom of Planktothrix rubescens in deep water (23-40 m) was observed, while 2 out of 3 MC-variants identified in the bloom were desmethyl MCs (Table 1). Planktothrix rubescens has been known to produce [D-Asp 3 ]MCs 98 , which is in agreement with the present study.
In Lake Volvi, only Nostocales taxa were observed (Anabaena, Aphanizomenon and Cylindrospermopsis). The occurrence of these species is not necessarily related to the production of MCs 99 . Anabaenopsis was also present in  (Table 1) in the bloom of Lake Zazari was accompanied by the presence of MC-RR and MC-YR. Also present in the lake were Microcystis wesenbergii, which is known as a MC non-producting species, Anabaena spiroides which is also known for its non-toxic strains 101 and Merismopedia warmingiana 97 .
Μicrocystis aeruginosa, Microcystis panniformis as well as Anabaena, Aphanizomenon and Planktothrix were identified in Lake Pamvotis, which are able to produce multiple MC-variants. LC-MS/MS analysis revealed the presence of MC-RR, MC-YR and MC-LR in the biomass.   Similarly, in Lake Mikri Prespa, Microcystis aeruginosa, Anabaena cf. lemmermanii and Aphanocapsa sp. were the dominant species, with dmMC-RR, MC-RR and MC-LR identified in the sampled biomass.
No MCs were identified in Lake Vistonis, where Αphanizomenon favaloroi, Pseudanabaena limnetica and Limnothrix sp., were the most abundant species. This is in accordance with past studies in Mediterranean lakes, where the production of MCs is mainly linked to the presence of Microcystis. In a recent study in Spanish lakes, 31 Nostocales strains were isolated and analyzed with negative results for MCs 99 .
In Lake Kerkini, during June 2008, MC-RR, MC-LR and MC-LW were identified in the biomass, which mainly contained the MC-producing cyanobacterial species Μicrocystis aeruginosa, Microcystis panniformis, and Anabaena (Table 1), as well as other cyanobacterial species known to produce MCs, such as Cylindrospermopsis raciborskii, Planktothrix agardhii and several Aphanizomenon species (Fig. S8).The production of the identified MCs cannot be safely attributed to these species. It is noteworthy, that the cyanobacterial bloom, which was sampled from the same lake during October 2014, showed no evidence of MC content. The sample contained mainly Cylinrospermopsis raciborskii and Planktothrix agardhii. No Microcystis species were identified.
ANA-a in relation to species. ANA-a was identified only in Lake Doirani (September 2008, 65.5 ng mg −1 dw) and Lake Kerkini (June 2008, 61.7 ng mg −1 dw) as shown in Table 2. The production of ANA-a has been reported in the past by the following cyanobacterial species: Anabaena spp., Anabaena flos-aquae, Aphanizomenon flos-aquae, Cylindrospermum sp., Anabaena planctonica, Anabaena crassa, Planktothrix rubescens, Raphidiopsis mediterranea, Phormidium favosum, Aphanizomenon issatschenkoi, Αrthrospira fusiformis and Phormidium autumnale 102 . The bloom from lake Doirani (September 2008) was dominated by species Aphanizomenon issatschenkoi, Anabaena flos-aquae, Aphanizomenon flos-aquae, Raphidiopsis mediterranea and Cylindrospermopsis raciborskii, directly related to the production of ANA-a. The sample from Kerkini Reservoir (June 2008) contained mainly Aphanizomenon flos-aquae, Microcystis aeruginosa, Cylindrospermopsis raciborskii, Anabaena cf. viguieri, Anabaenopsis selenkinii, Anabaena flos-aquae, Microcystis panniformis, Anabaena aphanizomenoides, some of which are closely related to the production of ANA-a. Water samples from these two lakes did not contain detectable amounts of ANA-a. It is noteworthy, that although several other bloom samples from Greek lakes in this study contained some of those ANA-a producing cyanobacterial species, their analysis indicated negative results for the presence of ANA-a.  Table 4. In the past the production of CYN has been related to species such as Cylindrospermopsis   30 . In the case of Lake Doirani, CYN could be related to the dominance of Cylindrospermopsis raciborskii/Raphidiopsis mediterranea as well as Aphanizomenon gracile, while in lakes Vegoritis and Pamvotis, the occurrence of CYN in trace amounts could be related to the presence of Aphanizomenon flos-aquae, which usually forms blooms in those lakes during the summer period. Although Cylindrospermopsis raciborskii/Raphidiopsis mediterrane is usually abundant in Lake Vegoritis throughout various periods, it was not detected during the specific sampling period.  (2014), neoSTX was also found to be present. STX and neoSTX production has been linked in the past with Αphanizomenon flos-aquae, Planktothrix sp., Anabaena circinalis, Anabaena lemmermannii, Aphanizomenon sp., Aphanizomenon gracile, Cylindrospermopsis raciborskii and Lyngbya wollei 38,48,49 .
In Lake Kastoria, the production of STX could be attributed to the presence of cyanobacteria Anabaena cf. circinalis, while in lakes Petron, Trichonis and Kerkini, the presence of STX concurs with the occurrence of cyanobacterial blooms dominated by Aphanizomenon gracile/Cylindrospermopsis raciborskii (Lake Petron), Αphanizomenon flos-aquae (Lake Trichonis) and Cylindrospermopsis raciborskii/Αphanizomenon flos-aquae (Kerkini Reservoir). In those water bodies, other cyanobacterial species could also be potentially responsible for the production of STX, like Planktothrix (P. agardhii and P. rubescens). Especially for Lake Trichonis, STX was detected in the bloom dominated by P. rubescens. Seven of the analyzed samples from Greek lakes (Kastoria, Pamvotis, Zazari, Mikri Prespa and Vegoritis), contained MCs at a concentration that poses a high risk of inducing adverse health effects, according to WHO guidelines. In those lakes, the accidental consumption of only a few milliliters of lake water by an average adult swimmer would be enough to reach the TDI set for MCs. Sigma-Aldrich (Steinheim, Germany). CYN was purchased from Abraxis (Warminster, UK) and a racemic mixture of (± ANA-a Fumarate from TOCRIS Bioscience (Bristol, UK). ANA-a concentration was calculated based on the exact concentration of fumarate, which was determined using high-purity standard solutions of fumaric acid, analyzed with a HPLC-UV system. Fumaric Acid (>99%) was provided by Sigma-Aldrich, Germany. Since most of the naturally occurring ANA-a is found in the (+) form 36

Microscopic analysis.
Water samples for microscopic analysis were collected. Fresh and preserved phytoplankton samples were examined under an inverted light microscope (Nikon TE 2000-S) and species were identified using taxonomic keys [104][105][106][107] . Phytoplankton counts (cells, colonies, and filaments) were performed using the Utermöhl's sedimentation method 108,109 . For biomass estimation (mg L −1 ), the dimensions of 30 individuals (cells, filaments, or colonies) of each species were measured using tools of a digital microscope camera (Nikon DS-L1), while mean cell or filament volume estimates were calculated using appropriate geometric formulae 84 . Species and taxonomical groups comprising more than 10% (w/w) of the total phytoplankton biomass were considered to be dominant. The cell volume and total phytoplankton biovolume estimates were converted to biomass (wet weight) by assuming a density of 1 g cm −3 .  Fig. 7. All methods were validated in order to assess specificity, linearity, precision (repeatability and reproducibility), accuracy (% recovery), as well as limits of detection (LODs)/quantification (LOQs). To assess specificity, blank samples were analyzed and no interfering peaks were observed close to the retention times (t R ) of the analytes.

Sample treatment for CT analysis -Analytical workflow.
The developed workflow provides a useful toolkit for the determination of various CTs in different matrices. After receiving the samples in the laboratory, alternative protocols were followed, according to the purpose and scope of the analysis. In cases where there was a surface bloom in the lake, cyanobacterial cells (biomass) were withdrawn from the surface layer of samples, using a glass pipette. The collected biomass was subsequently frozen and then transferred to a lyophilization apparatus (Martin Christ ALPHA 1-2, Vacuubrand HV Pump), where vacuum was applied for 24 h, at − 51 °C. Lyophilized biomass was further analyzed using method A. If the samples did not contain visible biomass, they were filtered through GF/F filters (previously dried and pre-weighed). Filters where then extracted and analyzed with method B. Selected filtrates from samples with known history of dissolved CTs were stored at 4 °C for further analysis of extracellular soluble CTs using methods C and D.
A detailed description of the methods is given below:  74 . Briefly, chromatographic separation was performed with reversed phase liquid chromatography (RPLC) column Atlantis T3 (2.1 mm × 100 mm, 3 μm) from Waters (Ireland). A gradient elution program was applied with solvents (A) ACN and (B) water, both containing 0.5% FA. The gradient started at 5% A (held for 3 min), which increased to 20% A in 1 min (held for 2 min), further to 35% A in 1 min (held for 7 min), 70% A in 14 min and finally 90% in 1 min (held for 3 min). An equilibration time of 10 min was kept after each sample run. Flow rate was set at 0.2 mL min −1 with 20 μL injection volume and column temperature was set at 30 °C.
STX and neo-STX were separated using a SeQuant ZIC-Hydrophilic Interaction Chromatography (HILIC) 150 mm × 2.1 mm, 3.5 mm column (Merck). A gradient elution program with water, ammonium acetate, acetonitrile and formic acid as solvents was used 76 . Selected molecular and fragment ion transitions, were according to Dell' Aversano et al. 111 .
Electrospray Ionization (ESI) in positive mode was used for ionization of all analytes. Multiple Reaction monitoring (MRM) mode was applied for the detection of CTs, using the three most intense and characteristic precursor/product ion transitions for each analyte. Identification of target CTs was based on three criteria: (1) retention time (t R ) of compounds (2) three characteristic precursor/product ion transitions and (3) two calculated ratios Matrix effects of extracted biomass. In order to assess the matrix effects of biomass extract on the quantitative determination of selected CTs, a series of experiments were performed. For biomass analysis, a sample which did not contain CTs was extracted according to the method previously described (method A), and an appropriate amount of analyte mixture containing CYN, ANA-a, MC-RR, MC-LR, STX and neoSTX was spiked in the final reconstituted solvent, in order to obtain a nominal concentration of 100 μg L −1 . The spiked samples were analyzed using the above mentioned LC-MS/MS methods and they were compared to standard mixture solutions of the compounds at the same concentration level.
Validation of methods. The methods were validated in order to assess specificity, linearity, precision (repeatability and reproducibility), accuracy (% recovery) and limits of detection (LODs)/quantification (LOQs). Blank samples were analyzed to assess method specificity. The linearity of each method was evaluated by analyzing in triplicates standard solutions at eight different concentrations (1, 2, 5, 10, 20, 50, 100, 250 μg L −1 ) for all CTs.
Limits of detection (LODs), accuracy and precision of methods C and D, used for the determination of CTs in water, are thoroughly described in past studies 74,76 . In the case of biomass/filter analysis (methods A and B, respectively), LOD calculation (expressed in ng mg −1 dw) was based on the LOD of each analyte in the extract (μg L −1 ), taking into consideration that 10 mg of biomass were extracted. Trueness and precision were evaluated by analyzing a toxins-free lyophilized biomass sample, spiked with a mixture of 12 MCs, CYN, ANA-a, STX and neoSTX at two content levels (3 and 30 ng mg −1 ), in six replicates at three different working days (n = 18).
Risk assessment. In order to evaluate the possible human health hazards related to the presence of CTs in lake water, a risk-assessment approach has been carried out, taking into consideration the possible exposure routes and available guidelines. The TMCs found in the samples of the present study were calculated and compared to the guidance values and thresholds, established by WHO 41 . Finally, the amount of surface water that has to be accidentally consumed during swimming, in order to reach the threshold tolerable daily intake (TDI) set by WHO 43 , was also calculated for an average adult and child, based on the following formula: where: V: amount of water (L) that have to be involuntarily ingested in order to reach each threshold. T: threshold (TDI) value (μg kg −1 body weight). bw (body weight): assuming child (10 kg) or adult (60 kg) average body weight (kg). C: TMCs concentration: sum of intracellular and extracellular (μg L −1 ). Assuming that each adult or child accidentally consumes 200 mL of water each day they swim 41 , the amount of orally ingested toxins and the percentage of TDI reached for a day's swim, were also calculated.

Data Availability
All data generated or analyzed during this study are included in this published article (and the Supplementary  Information files).