As observed for cyanotoxins in general, there is a variation of several orders of magnitude in the concentration of MCs, the most studied toxins, both within and between taxonomic groups of invertebrates (Figure 1 and Supplementary Table S1). In spite of the high variability, zooplankton is clearly the best bioaccumulator of MCs, reaching values over 1000 μg g⁻¹ DW and an average value of about 383 μg g⁻¹ DW. Gastropods may contain concentrations up to 436 μg g⁻¹ DW (Sinotaia histrica hepatopancreas [82]), with an average of 58 μg g⁻¹ DW. Although bivalves are efficient filter-feeders, maximum concentration was 630 μg g⁻¹ DW (Unio douglasiae hepatopancreas [83]), and the average was only 57 μg MC g⁻¹ DW. Larger crustaceans (i.e., Decapoda) seem to be the less efficient accumulators of MCs, with an average of 4.2 μg g⁻¹ DW and a maximum concentration of 12.4 μg g⁻¹ DW (Macrobrachium nipponensis stomach [84]).

Even if high concentrations of MCs have been measured in aquatic consumers of different trophic levels, a meta-analysis based on the Biomagnification Factor (BMF) quantitatively confirms biodilution of MCs, not biomagnification [76,85], as the dominant process in aquatic food webs [86].

BMF is herein defined as the ratio between the concentration of MCs in aquatic consumers and the concentration of MCs of their food, both with same unity. Although we will discuss some aspects of BMF in the next sections, here we focused primarily on the potential for accumulation of cyanotoxins and their effects on aquatic invertebrates and vertebrates. For further details on data selection, BMF calculation, analysis and discussion see Kozlowsky-Suzuki et al. [86].

Watanabe et al. [73] were the first in demonstrating the accumulation of MCs by zooplankton. They found values up to 202% above phytoplankton, but their method of estimation was indirect, using fractionated samples of phytoplankton and zooplankton still contaminated by phytoplankton and calculating the difference in concentration between these samples. The dominant species were the cladocerans Bosmina falalis and Diaphanosoma brachyurum which have very small body size (~1 μg DW animal⁻¹), which makes very difficult to separate from phytoplankton using nets. The amount required for analysis (~10 mg) would require about 10,000 individuals of both species, which would be impossible to separate by trialing animals from the samples. A more powerful detection method, available after their work, the Enzime Linked Immunosorbant Assay (ELISA) allowed the use of very small of biomass (~50–100 µg) [79].

Using ELISA, Thorstrup and Christoffersen [79] showed experimentally that Daphnia magna was able to accumulate up to 24.5 μg MC g⁻¹ DW when submitted to toxic cells of the Microcystis aeruginosa strain CYA228/1. These authors also concluded that if the amount of toxins was originated only from the content of the cells in the gut of the animals, calculated by the biovolume as 0.002 ng MC Daphnia⁻¹, the amount of toxins actually measured (0.19 ng MC Daphnia⁻¹) was 100 times higher, indicating that this toxins was assimilated and stored in other parts of the animal.

Trophic transfer of MCs in invertebrates has been rarely demonstrated. With a simple food chain model (M. aeruginosa→Daphnia pulex→Chaoborus), Laurén-Määtä et al. [80] observed a higher mortality of Chaoborus after preying on Daphnia fed toxic Microcystis than that fed non-toxic algae, suggesting that Daphnia transferred toxins from Microcystis to Chaoborus. However, as they did not detect MCs in the midge larvae they suggested that most of the toxin was metabolized or excreted in this two-step food chain.

Ferrão-Filho et al. [75] measured up to 16.4 μg MCs g⁻¹ DW in zooplankton and estimated BMFs in the order of 0.7 to 43. In comparison, Ibelings et al. [76] showed higher accumulation of MCs in zooplankton (up to 1226 μg MCs g⁻¹ AFDW), but lower BMF (0.25). In lab experiments, Oberhaus et al. [81] showed that D. pulicaria was able to accumulate up to 1090 μg MC g⁻¹ DW by consuming small filaments of Planktothrix. Despite the fact that the above mentioned studies indicated that zooplankton can be an efficient accumulator of MCs and acts as vectors of these toxins to higher levels, few studies found BMF values higher than unity [75,76,77].

Bioaccumulation of MCs in mollusks allows us to study the distribution of toxins in different organs and tissues. Figure 2 shows the mean of maximum concentrations found in organs or tissues of several species of bivalves and gastropods. Hepatopancreas is the organ presenting the highest MCs concentrations, followed by the intestines. Concentrations up to 22.7 μg g⁻¹ DW for the gastropod Sinotaia histrica [87]. Low concentrations of MCs are, however, found in muscles, and as soon as other organs are removed before food preparation and consumption the risk of contamination is significantly lower. Nevertheless, consumption of mollusks in a regular basis may represent a health threat for humans [88,89,90,91], especially in countries such as China where it is common to consume soup prepared with whole snails [91].

Most studies on NOD accumulation were carried out in the Baltic Sea, mainly in the Gulf of Finland, where Nodularia spumigena blooms heavily every year. Both field and laboratory studies estimated the transfer of NOD via zooplankton. Engström-Öst et al. [92] verified the bioaccumulation and transfer of NOD in mysidaceans (Mysia relicta) and fish (Gasterosteus aculeatus) via copepods. Karjalainen et al. [93] estimated uptake rates of dissolved NOD of 0.37–1.55 μg gC⁻¹ in the calanoid copepods Acartia tonsa and Eurytemora affinis, and in the ciliate Strobidium sulcatus, resulting in bioconcentration factors (BF) of 12, 18 e 22, respectively. Some studies revealed that NOD can be excreted through faecal pellets of copepods [94] and faeces of Mytilus edulis [95,96], while others showed that the concentration of NOD in the organisms is positively related with direct consumption of toxic cells [58,96] and transferred to copepods through the microbial food web [97].

Total hepatotoxins including NOD or a NOD-like compound [98] have also been measured in other invertebrates such as the crustaceans Neomysis integer [99], Panaeus monodon [100], Gammarus zaddachi [101] and the bivalve Macoma balthica [102]. Although some of these studies indicated trophic transfer of NOD along the food chain, not only low retention of ingested toxins [58] and low toxin concentration of copepod faecal pellets [94] have been observed, but also low bioconcentration of dissolved NOD [93,103] and BMF [87]. In addition, the concentration of NOD in the faecal material of blue mussels can be decreased up to 99% after the process of coprophagy [95] and NOD depuration from grazer tissues takes place rapidly despite incomplete removal [96,104] even after 168 h of cleaning-up period [103].

So far, MCs have been the most extensively studied cyanotoxin regarding its effect on zooplankton [71,105,106,107,108]. MC-producer strains of cyanobacteria may be toxic to a variety of zooplankton species, including rotifers [109,110], cladocerans [71,72], copepods [60,111], and Artemia salina [107,112], but the toxicity has hardly been directly related to the production of MCs. Toxicity to zooplankton may, instead, be related to other compounds produced by cyanobacteria. Jungmann and Benndorf [105] isolated a compound from extracts of natural blooms that was more toxic to Daphnia than MC-LR and found no correlation between the concentration of MCs from different strains of cyanobacteria and the toxicity to Daphnia. Compounds produced by some cyanobacteria that are neither hepatotoxins nor neurotoxins may also cause toxic effects to Artemia salina and to the Aedes aegypti mosquito larvae [112,113].

Recent studies comparing wild and genetically engineered or spontaneous mutant genotypes of Microcystis aeruginosa strains have provided good tests to support the ecological role of MCs [66,67,68,69]. For example, the wild type of the strain PCC7806 is able to produce MCs, while the mutant genotype was engineered to knock out the MC synthetase gene cluster—mcyA-J—and does not produce MCs [66]. Nevertheless, the strain MRC, which has the mcyA-J cluster, is unable to produce MCs [67]. From the same bloom, and sub-cultured from the same original toxic isolate (MR), M. aeruginosa strain MRD produces MCs [67]. Thus, the availability of these mutants provides an opportunity to test the hypothesis that the negative effects of Microcystis on Daphnia growth and reproduction is caused uniquely by the presence of MCs in the cells, because the wild-type and mutant strains only differ in their ability to produce this toxin. Rohrlack et al. [66] showed that both ingestion rates and survival were largely inhibited by the wild genotype of the strain PCC7806, while the mutant genotype inhibited ingestion rates but caused no death in D. galeata. Kaebernick et al. [67] also found higher mortality with the MC-producing mutant (MRD), but ingestion rates were inhibited in a similar manner by both genotypes, showing that MCs were not the factor causing feeding inhibition. These authors found, however, that the wild MC-free genotype (MRC) caused a disruption in the molting process and in the midgut epithelia of two Daphnia species. On the other hand, Lürling [68] reported similar growth rates (in body size) for D. magna feeding on both genotypes, but a low survivorship with both genotypes, although the wild genotype was a little more toxic. Even though these studies represent good tests of the role of MCs, the incongruence among results suggests that other compounds than MCs may be involved in the toxicity of M. aeruginosa to Daphnia. For instance, while MCs seem to be the cause of death, the feeding inhibition and disruption in the molting process seem to be caused by a bioactive compound other than MCs.

Rohrlack et al. [69] also showed that the wild genotype of the strain PCC7806 caused a reduction in the heartbeats and in the movements of the thoracic limbs, mandibles, and second antennae of Daphnia galeata as well as a decrease in the activity of the foregut and stimulation of the midgut muscles. The mutant genotype caused, however, none of these effects. Both genotypes caused, nevertheless, a disruption of the midgut epithelium of D. galeata. Hence, all those studies together suggest that MCs can be considered a toxic metabolite to Daphnia, but cannot be considered responsible for all effects observed.

Although the mechanism of action of MCs in invertebrates has been poorly investigated, damage to the cells of the midgut of the Aedes aegypti mosquito larvae exposed to MCs [114] and hystopathological alterations in the midgut of Daphnia magna exposed to MC-LR [115] and of D. galeata exposed to intact cells of toxic Microcystis [69] have been demonstrated. DeMott and Dhawale [116] also demonstrated that MC-LR inhibited the activity of protein phosphatases 1 and 2A from bruit extracts of D. pulex, D. pulicaria and Diaptomus birgei, corroborating the results of previous studies in which purified MC-LR and a strain of Microcystis aeruginosa (PCC7820) showed acute toxic effects on the same zooplankton species [71]. Biochemical alterations in the activity of other important enzymes have been also demonstrated, such as the inhibition of glutathione (GSH) and glutathione-S-transferases (GST) in D. magna [115], tripsyns and chymotrypsins in Moina macrocopa and D. magna [117,118], soluble GST in Artemia salina [107], microsomal GST in D. magna [119], acetylcholinesterase (AChE) in D. pulicaria [120] and a stimulation of the enzyme lactate-dehydrogenase in Daphnia magna [115].

MCs and toxic cyanobacteria can also exert effects on other aquatic invertebrates such as crustaceans, gastropodes and insect larvae. The standard test with the anostracoid crustacean Tamnocephalus platyurus has been applied in some studies in a tentative to verify the toxicity of cyanobacterial blooms, providing results comparable to mouse bioassays [121,122,123,124,125]. Vasconcelos et al. [126] tested the effects of living cells of toxic and non-toxic strains of M. aeruginosa on Procambarus clarkii (Crustacea, Decapoda) and showed that juveniles were more susceptible to non-toxic strains, suggesting that other toxins or compounds, more potent than MCs, may have being produced. Montagnolli et al. [127] tested the toxicity of extracts of a MC-producer strain of M. aeruginosa on Kalliapseudes schubartii (Crustacea, Tanaidacea) and estimated a LC₅₀ (96 h) of 1440 mg L⁻¹ (equivalent to a 1.58 mg MC L⁻¹), and an increase in the oxygen consumption. Pinho et al. [128,129] studied the response of hepatopancreas and gills of the crab Chasmagnathus granulatus (Decapoda, Brachyura) to the oxidative stress after exposure to MC-aqueous extracts, through the activity of catalase (CAT), superoxidase dismutase (SOD) and glutatione-S-transferase (GST), as well as with measures of oxygen (O₂) consumption, formation of sulfidril groups (NP-SH) and lipid peroxides (LPO). They observed increase in the consumption of O₂, and in the activity of CAT and GST, as well as oxidative damage by the increase in LPO. In addition, extracts of M. aeruginosa administered orally on the same crab caused a decrease of the content of glycogen, which indicates that the increase in O₂ consumption may be associated to the glycolitic pathway [130].

Maršálek and Bláha [124] tested the toxicity of crude extracts and fractions of purified toxins from natural assemblages of M. aeruginosa with a series of aquatic invertebrates such as the protozoans Tetrahymena pyriformis, T. termophyla and Spirostomum ambiguum, the shrimp Artemia salina, the cladocerans D. magna, D. pulex and Ceriodaphnia dubia, the anostracoid crustacean T. platyurus, the rotifer Brachionus calyciflorus, the nematode Panagrellus redivivus, the Culex pipiens mosquito larvae, the cnidarian Hydra tenuata and the oligochaeta Tubifex tubifex. However, there was a large variability in the sensitivity of these organisms and results were not always correlated with MC concentration. For some invertebrates, such as A. salina, P. redivivus, T. pyriformis and T. platyurus there was a significant difference in the toxicity of crude extracts and corresponding toxin fractions (i.e., crude > toxin) while for all others no significant differences in the toxicities of both extract preparations (crude vs. toxin fraction) were found.

Mosquito larvae, such as that of Aedes aegypti also show sensitivity to cyanobacteria [113,114,131]. For example, a MC-producing strain of M. aeruginosa led to significantly longer development times than the controls or those grown with the non MC-producing strain [131].

Despite their high resistance to cyanotoxins, mollusks may also be affected by these compounds. The pulmonate gastropod Lymnaea stagnalis, besides accumulating MC-LR, presented a reduction in the egg production and in locomotion when exposed to purified toxin [132]. The zebra mussel, Dreissenia polymorpha, may show a large reduction in the food ingestion and absorption rate, leading to a reduction in the net energy balance, especially when exposed to single diets or mixtures of a MC-LF-producing M. aeruginosa and the diatom Asterionela formosa [133]. Toxic cells may also cause an acute irritant response manifested by the production of an unusually fluid pseudofaeces (“pseudodiarrhoea”) rich in mucus and Microcystis cells, ejected through the pedal gape of the mussels [134]. This selective rejection of MC-producing cyanobacteria by zebra mussels may enhance the presence of toxic M. aeruginosa in mixed cyanobacterial blooms and in the benthos [133].

The effects of NOD and/or toxic N. spumigena have been studied on different invertebrates. Koski et al. [57] studied the feeding and egg production of calanoid copepods (Acartia bifilosa and Eurytemora affinis) treated with water from a mesocosm experiment containing N. spumigena. Although copepods preferred ciliates, significant amounts of cyanobacteria were consumed with no adverse effect in the survivorship and egg production. Despite high ingestion rates by A. bifilosa and E. affinis on toxic N. spumigena offered alone or in mixtures with a green flagellated alga (Brachimonas submarina) or natural community, gross growth efficiency was low suggesting a high metabolic cost of coping with toxins [58].

Karjalainen et al. [135] reviewed the effects of Baltic Sea cyanobacteria including hepatotoxic N. spumigena on aquatic organisms including invertebrates. More recent studies have observed no apparent negative effect of N. spumigena on the survival of Baltic Sea meiofaunal assemblages under experimental conditions [136]. In fact, cyanobacterial labelled carbon was taken up by major meiofauna taxa living in the first centimeter layer indicating that cyanobacteria may be an important food source for the Baltic Sea meiobenthos. Nevertheless, even if cyanobacterial carbon can for instance be assimilated by the amphipod Monoporeia affinis, this does not translate into significant growth [137]. These authors also observed that less cyanobacterial carbon from toxic N. spumigena was mixed down in the sediment indicating lower bioturbation activity.

Evidence on CYN accumulation in aquatic invertebrates is scarce and a broader review of the bioaccumulation and effects of this toxin can be found in Kinnear [138] (this volume). Cherax quadricarinatus (Decapoda) from aquaculture ponds and in laboratory conditions may accumulate CYN in higher concentrations in the hepatopancreas [139] (Table S1). Saker et al. [140] found CYN concentrations of 2.9 and 5.9 μg g⁻¹ in whole body and viscera, respectively, of the bivalve Anodonta cygnea, but higher values (61.5 μg g⁻¹) in the haemolymph, corresponding to 408 μg CYN L⁻¹, when the concentration in the medium was 34 μg CYN L⁻¹, which gives a bioaccumulation factor (BF) of 12. In both studies, no adverse effects were observed. Despite toxin transfer, low BF (0.41–0.76) of CYN in relation to the surrounding medium was estimated for D. magna suggesting no evidence biomagnnification [141]. In the same study, D. magna was exposed to a CYN-producer strain (Cylin-A) and to a non CYN-producer strain (Cylin-P) of Cylindrospermopsis racibosrkii and both strains reduced the survivorship and growth, with the Cylin-A being more effective, suggesting that CYN was the toxic agent. In another study, however, two CYN-producer strains (C. raciborskii and Aphanizomenon ovalisporum) caused a reduction in the survivorship and growth of D. magna, but only the C. racibosrskii strain caused damage in the midgut and diverticula epithelium [142]. This result suggests that not all adverse effects on Daphnia can be attributable to CYN.

White et al. [143] detected CYN bioconcentration and bioaccumulation in Melanoides tuberculata (Gastropoda) after exposure to freeze-thawed cell extracts and live cells of toxic C. racibosrskii, with average BF and bioaccumulation (BAF) factors up to 1.5 and 124.4, respectively. Exposure to live cells resulted in higher toxin burden (up to 250 µg g⁻¹ DW) than exposure to freeze-thawed solutions indicating that despite uptake from both extracellular and intracellular pathways, the availability of intracellular toxin is critical. Berry and Lind [144] presented evidence for CYN bioaccumulation in a freshwater lake in Mexico. They measured appreciable concentrations of both CYN and STXs in “tegogolo” snails (Pomacea patula catemacensis), despite low concentrations in the seston, and calculated BAFs of 157 and 196, respectively. These studies suggest that CYN and STX may have a higher (or at least equal) potential for biomagnification than MCs.

Contrary to the vast literature from marine studies [11], there is very little evidence of bioaccumulation and effects of STXs produced by cyanobacteria in freshwaters. Sasner et al. [145] was one of the first studies showing the accumulation of STXs (PST) produced by Aphanizomenon flos-aquae in the bivalves Ellipio campanatus and Corbicula fluminea. Negri and Jones [146] showed the accumulation of STXs produced by the cyanobacterium Anabaena circinalis in the mollusk Alathyria condola. They reported an “adverse effect on feeding behavior” of the mollusks but did not quantify any physiological parameter. Nogueira et al. [147] showed not only that D. magna accumulated STXs from Aphanizomenon issatschenkoi, but also that both fitness and somatic growth of this daphniid were negatively affected, and that the activity of the cytosolic glutathione-S-transferases (cGSTs) decreased. Although STXs detoxification via GST activity is possible [35], there is no available data about the conjugation of SXTs and GSH.

Pereira et al. [148] estimated the accumulation of PST from A. issatschenkoi in the bivalve Anodonta cygnea and their BMF estimates varied from 0.035 to 2.2, thus in the range observed for MCs [86]. PST also affected the filtering behavior of mussels, though low mortality rates were observed.

Few studies have demonstrated the effects of STXs in freshwater zooplankton species. Reduction in the thoracic appendages beating rate and an increase in the rejection rate of particles by the post-abdomen of Daphnia carinata was observed after exposure to a filtrate of Aphanizomenon flos-aquae and purified STX [149]. Also, the intrinsic rate of population increase (r) of D. pulex and Moina micrura was negatively affected by a STXs-producer strain of C. raciborskii (T3), but not by a non STXs-producer strain (NPLP-1) of the same species [150]. In contrast, D. gessneri was stimulated by strain T3 and depressed by the strain NPLP-1, suggesting resistance to STXs produced by the strain T3 and sensitivity to some bioactive compound(s) produced by the strain NPLP-1. D. magna had slightly longer survivorship in suspensions with another saxitoxin-producer strain of C. raciborskii (CYRF-01) as the sole feed than in medium without food, showing that starvation was not the cause of death [151]. However, these authors showed that the fitness (i.e., body growth and population growth rates) and clearance rates were depressed only in high proportions (75–100%) of this cyanobacterium in the diet, and concluded that energy limitation, not toxicity, might be the dominant factor affecting growth of large-bodied cladocerans. These contradictory results exemplify how generalizations on cyanobacteria-zooplankton interactions are prone to failure and that species-especific interactions may occur.

STXs can also affect the swimming behavior of zooplankton. Using an image analysis system to track the swimming activity of Daphnia exposed to a STXs-producer C. raciborskii strain (CYRF-01) and to raw water samples from a eutrophic reservoir, Ferrão-Filho et al. [152] showed a decrease in the mean time spent in swimming and in the mean velocity of D. pulex, which was attributed to the presence of STXs in field samples and in the strain CYRF-01. Swimming movements of D. pulex and M. micrura were also inhibited by the strain T3 in 24 h exposures, but recovered within 48 h after the animals were transferred to a medium with only highly nutritious food [10]. A recent study [153] has, however, shown that both paralysis and recovery can occur even faster (paralysis in less than 1 h and recovery in less than 24 h). Cladocerans basically swim by the movements of the second antennae, thus it is likely that STXs are acting by the classic mechanism of action (i.e., blockage of sodium channels in the neuro-muscular junction), preventing the propagation of the nerve impulse in this appendage. Apparently, the filtration of food particles was not altered allowing the animals to survive for long periods [10]. This curious phenomenon was also observed by Costa [150] with the T3 strain during chronic exposures. Animals could even grow and reproduce, though in a lower rate, while paralyzed on the bottom of the test tube. It is unclear how those animals cope with paralysis of the second antennae while keeping the movement of the thoracic appendages.

Fewer studies were performed with other neurotoxin-producer cyanobacteria. Living cells of an ANTX-a(s)-producer strain of Anabaena flos-aquae (NRC525-17) had acute toxic effect on the copepod Diaptomus birgei and on the cladoceran D. pulicaria [71]. Another ANTX-a(s)-producer strain of A. flos-aquae (IC-1) was tested on rotifer species, being Brachionus calyciflorus the most susceptible with suppressed fecundity at concentrations of 0.5 mg DW L⁻¹ of living cells [154]. Acute toxicity tests with living cells and extracts of lyophilized cells of an ANTX-a(s)-producer strain of A. spiroides (ITEP-024) were carried out with D. gessneri, D. pulex and Moina micrura, but little effect was observed [150].

High concentrations of BMAA in aquatic food webs from South Florida in water-bodies where cyanobacteria bloom have recently been found [155]. However, unlike the biomagnification pattern observed for a terrestrial food web [27], no tendency for biomagnification was detected. They observed, instead, higher levels of BMAA in invertebrates such as pink shrimp (Panaeus duorarum; ~3000 µg g⁻¹ WW) and blue crabs (Callinectes sapidus; ~7000 µg g⁻¹ WW), and fish species feeding on benthos than those with planktonic feeding habit. Nevertheless, the still scarce knowledge about BMAA indicates the potential for biomagnification with serious implications for human contamination and association with neurodegenerative diseases.

Despite large data variability (Figure 3), the potential for accumulation of MCs in fish seems to be lower than for invertebrates, being on average 3.5 times lower in planktivores than in zooplankton (Figure 2 and Figure 3). The highest concentrations were found in planktivorous fish reaching up to 874 µg MC s g⁻¹DW (liver of smelt Osmerus eperlanus [76]). Other planktivorous fish, especially carp that feeds on phytoplankton cells, have also a great potential for MCs accumulation. Xie et al. [167] found up to 137 µg MC g⁻¹ DW in the intestines of silver carp (Hypophthalmichthys molitrix) from Lake Chaohu, in China. The maximum concentration of a carnivorous fish was 51 µg MC g⁻¹ DW in the liver of perch (Perca fluviatilis) of Lake IJsselmeer in the Netherlands [76].

Overall carnivorous fish, as top predators, had lower average MC contents than other trophic guilds (Figure 3), which suggests no tendency of biomagnification along the food chain. In fact, Ibelings et al. [76] showed generally low BMFs for a real food chain in Lake IJsselmeer, in the Netherlands. The BMF for the planktivorous smelt reached, however, 286% of its diet (i.e., MCs content of planktivores were ~2.9 times higher than that of zooplankton), while the BMF for the following trophic level (planktivores→carnivores) was only 11%. Surprisingly, the benthivorous ruffe (Gymnocephalus cernua), which feeds on pseudofaeces of the zebra mussel (Dreissena polimorpha), accumulated high amounts of MCs (up to 194 µg g⁻¹ DW), reaching a BMF of 1.2. Furthermore, both a qualitative assessment [85] and meta-analyses of a large dataset on aquatic consumers of different trophic levels and guilds [86] pointed out to biodilution, not biomagnification, of MCs as the predominant process in aquatic food webs. However, zooplanktivorous fish was an exception to this pattern having the highest BMF mean (1.8) among fish guilds, suggesting possible biomagnification of MCs [86].

The concentrations of MCs in the different organs/tissues by fish guild (Figure 4) reveal that, as predicted for hepatotoxins, the liver of planktivores (0.06–874 µg g⁻¹ DW); and omnivores (0.003–321 µg g⁻¹ DW) accumulates most of these toxins, followed by intestines (Planktivores: 5.4–137 µg g⁻¹ DW; Omnivores: 0.03–678 µg g⁻¹ DW). Other tissues such as kidney, bile and blood can also accumulate relatively high concentrations of MCs [168]. Low concentrations of MCs can be found in the gills of carnivores (up to 0.1 µg g⁻¹ WW), omnivores (up to 1.4 µg g⁻¹ WW) and planktivores (0.56 µg g⁻¹WW) [170]. Papadimitriou et al. [169] found low concentrations of MCs in the brain (0.044 µg g⁻¹ WW) and ovaries (0.009 µg g⁻¹ WW) of the omnivorous fish Carassius gibelio in Greek lakes. Despite low concentrations of MCs in fish muscles (up to 14.3 µg g⁻¹ DW), great concern is given to this tissue as it is part of human diet in many countries. The World Health Organization (WHO) has established guidelines for the consumption of fish containing MCs. A Tolerable Daily Intake (TDI) value of 0.04 µg MC kg⁻¹ of body weight day⁻¹ was adopted as a provisional guideline value [3]. Magalhães et al. [170] found a maximum daily intake of 1.68 µg MC kg⁻¹ in fish from Jacarepaguá Lagoon, which is 42 times above the proposed TDI. The authors pointed out that this form of human exposure must be monitored, and that the quality control of fish from water bodies presenting cyanobacterial blooms should be carried out.

Few experimental studies showed that primary consumers may act as an efficient vector of toxins to upper trophic levels such as fish. Pumpkinseed sunfish (Lepomis gibosus) accumulated MCs (up to 11.2 ng MC g⁻¹ WW and only 250 pg MC g⁻¹ WW in liver and muscles, respectively) from lyophilized zooplankton from the eutrophic Barbaboes Pond (New Hampshire, USA) [171]. Transfer of free and bound MCs from intoxicated snail Lymnaea stagnalis to stickleback Gasterosteus aculeatus has been shown, with MCs accumulating in fish liver, kidneys, gills and muscle [172]. Induction of oxidative stress, liver histopathology and behavioral disturbances were also observed [172].

NOD or a NOD-like compound has been measured in different fish species of the Baltic Sea, such as flounder and cod [173], sea trout [174], pike larvae [99], roach [165], three-spined stickleback, herring and sprat [175] from field or experiment samples. Persson et al. [176] provided the first data on NOD accumulation in flounder liver tissue from the Öresund (Swedish west coast) despite non-contaminated primary food source (blue mussels).

Although mass mortality of fish may be related to harmful algal blooms [177,178,179,180] factors such as high pH due to photosynthetic activity and low O₂ concentrations associated to the decay of algal biomass, may as well be related to these events [35]. Toxic cyanobacteria or their toxins may, nevertheless, exert adverse effects on fish, including damage to several organs such as liver, heart, gills and kidneys, besides ionic equilibrium disturbance, behavioral changes, growth inhibition and mortality [181]. However, as observed for other aquatic organisms results are generally non-conclusive and contradictory. For instance, Beveridge et al. [182] showed an increase in the opercular beating rate of the Nile tilapia (Oreochromis niloticus) and an increase in the ingestion rate of cyanobacteria with a non-toxic strain of M. aeruginosa but no effect with a toxic strain (MC-LR-producer), whereas Keshavanath et al. [183] observed an inverse response for the same fish species. The European roach Rutilus rutilus fed with Aphanizomenon or Microcystis cells collected from eutrophic lakes presented a low growth rate [184]. Nevertheless, growth rate with Aphanizomenon was significantly higher than without any food and low growth with Microcystis could not be attributed to the presence of MCs, but to the low digestibility of this cyanobacterium and low assimilation efficiency by the roach [184].

Best et al. [185] found that the rainbow trout (Oncorhynchus mykiss) exposed to intact or broken cell extracts of MCs-producer Microcystis (PCC 7813) had an increase in liver mass (hepatosomatic index, HSI) and in the water content in the gut, but only when administered together with heterotrophic bacterial lipopolysaccarides (LPS). They concluded that exposure of fish to toxic Microcystis promoted osmoregulatory imbalance resulting from stimulation of the drinking response, increased volume of fluid in the gut and inability to remove water excess. Bacterial LPS is known to affect numerous systems, including the liver detoxication system [186,187] and it is likely that cyanobacterial LPS exhibits similar effects to their bacterial counterparts [188], including the inhibition of glutathione S-transferases [189].

Ibelings et al. [76] reported histological abnormalities such as inflammation, degeneration and necrosis, potentially caused by MCs, in 50% of the smelt livers in Lake IJsselmeer. For the other species, especially ruffe, this percentage was lower. However, none of the livers were enlarged or darker as was observed in gavage experiments with perch exposed to MCs.

Histopathological alterations in liver, gastrointestinal tract and kidney, and also immunopositivity for MCs, heve been observed in the European whitefish Coregonus lavaretus when administered an oral dose (i.e., gavage) of a MC-producer strain culture of Planktothrix rubecens [190], suggesting a relationship between tissue damage and the presence of MCs. In addition, exposed fish showed behavioral alterations, such as increased opercular beat rates and elevated plasma glucose levels, possibly representing a physiological stress response. In another study, the same histopatological changes in the liver, kidney, and in the gastrointestinal tract of C. lavaretus sub-chronically exposed to ambient water containing low, medium and high P. rubescens densities were found [191]. However, gill pathology appeared to result primarily from mechanical abrasion and irritation due to ectoparasitic infestation. These studies confirmed the hypothesis that sub-chronic and chronic exposure to low cyanobacterial cell densities and hence MCs can exacerbate physiological stress and sustained pathological alterations in exposed coregonids. They also corroborated the evidence that P. rubescens blooms may be the cause of the observed weight reduction (i.e., fitness) of this fish in European lakes.

Liver histopathology has been examined in sea trout [174] and flounder [192,193] either from Baltic Sea samples or experimentally after oral administration of toxic N. spumigena containing slurry. Flounder liver histopathology from field specimens had no connection with toxin concentrations and many changes and lesions were likely related to the presence of parasites or caused by combined factors such as various anthropogenic pollutants [100,192]. Both flounder [193] and sea trout [174] liver were affected after oral administration, but the effect on sea trout liver was more severe and liver architecture was completely lost after 1–2 days after a single dose despite low ELISA signal. Sea trout liver partially recovered after 4–8 days, indicating that in spite of the apparent high NOD toxicity, the effect was severe but reversible.

Effects of hepatotoxins or toxic cyanobacteria at the biochemical level were also observed in fish. Pflugmacher et al. [194] found a reduction in GST activity with MC-LR in several organisms, including fish. The uptake of MC-LR in larval stages of Danio rerio increased the activity of two detoxification enzymes (GST e glutatione-peroxidase, GPx), probably indicating an ability of this organism in metabolizing MC-LR to a less toxic compound [195]. Therefore, chronic effects such as growth inhibition in embryonary stages, when organogenesis is still not completed could be caused by energetic demands of the detoxification process [195]. Bury et al. [196,197] observed, however, that an extract of Microcystis inhibited the ionic pump of the gills of tilapia (Oreochromis mossambicus) more effectively than purified MC-LR, showing that this effect was due to fatty acids present in cells of Microcystis, which interact in the membranes of the gill epithelium.

Using intraperitoneal injections of MC-LR and MC-RR, Prieto et al. [198] studied the response of the antioxidant enzymes superoxide dismutase (SOD), catalase (CAT), GPx and glutathione reductase (GR) as well as lipid peroxidation (LPO) as a biomarker of oxygen-mediated toxicity in liver, kidney and gill of tilapia (Oreochromis sp.). Their results showed that MCs induce adaptive responses such as increased activity of the antioxidant enzymes, mainly SOD and CAT, as well as of LPO. Regarding LPO, the liver was the most affected organ by MC-LR, but was not affected by MC-RR.

Cazenave et al. [199] assessed the oxidative stress in embryos of zebra fish (D. rerio) to the exposure to purified MC-LF and MC-RR by determining both the production of hydrogen peroxide and the activity of the antioxidative enzymes guajacol peroxidase (POD), CAT, GPx, and GR. Energetic costs were evaluated by determining the content of fat, carbohydrates, and proteins in exposed and control embryos. They evaluated also the attenuating effect of exposures in combination with natural organic matter (NOM). They verified less pronounced teratological effects within 24 h as well as non-significant increase in the activity of sGST in embryos exposed to either MC-LF or MC-RR in combination with organic matter compared to embryos exposed to pure toxins. NOM also diminished oxidative effects caused by MC-LF, while it failed to attenuate oxidative stress caused by MC-RR. Energetic costs of detoxification were detected by a significant decrease in lipid content, especially in embryos exposed to pure toxins, following a similar trend to that obtained with teratological and enzymatic assays, confirming the attenuating effect of NOM. They concluded that physiological responses to MCs and NOM required energetic costs, which were compensated to the expense of the energy resources of the yolk, which in turn might affect the normal development of embryos.

Responses of the antioxidant enzymes GST, CAT and the formation of malondialdehyde (MDA) in liver tissue of flounder (Platichthys flesus) were monitored during 14 days after two doses of intraperitoneal injection of pure NOD [176]. GST and CAT activity decreased significantly in the highest concentration (50 µg NOD kg⁻¹) after 7 days, restoring to control levels after 10 days of recovery. No significant changes were, however, seen in MDA levels. Flounders have also been orally exposed to toxic Nodularia spumigena-containing slurry either as a single high dose or as three repeated lower doses 3 days apart [193]. Liver GST activity of flounders that received repeated doses was significantly higher than with single dose or control. This result together with a lower NOD concentration in the liver of the repeatedly dosed individuals suggest that repeated doses may increase detoxification efficiency, thus NOD is rapidly detoxificated.

Behavioral changes were also reported in fish exposed to hepatotoxins. Baganz et al. [200] showed that the swimming activity (i.e., the average swimming velocity and the average number of turns) of D. rerio and Leucaspius delineatus exposed to MC–LR presented a dose-dependent response, but that it was dependent on the time of the day (daytime or night-time). During the day, the activity of both species increased in the lower concentrations, while the higher concentrations led to a significant reduction in activity. These results suggest both fish have comparable susceptibility to MC-LR and may indicate some adverse consequences for fish populations.

Saker and Eaglesham [139] were the first to show CYN accumulation in fish (viscera of the Rainbow fish; Melanotaenia eachamensis) from an aquaculture pond in Australia (ref. 22 in Supplementary Table S2). Since then, few studies about CYN accumulation in aquatic vertebrates have been published (see Kinnear [138] in this volume for a complete review on CYN accumulation).

Berry et al. [201] tested the effects of purified CYN and extracts of CYN-producer C. raciborskii and A. ovalispurum on zebrafish. They found that purified CYN was toxic only when injected directly into embryos, but not by direct immersion at doses up to 50 mg L⁻¹. Despite the dose dependency of toxicity after injection of CYN, no consistent patterns of developmental alterations were observed, suggesting that toxic effects of CYN may not target specific developmental pathways. In contrast, direct immersion of embryos in all of the extracts resulted in both increased mortality and developmental dysfunctions, such as undeveloped eyes, truncated body axis, severe oedema and curved spine. Interestingly, there was no correlation of developmental toxicity observed for these extracts with the presence of CYN or with previously reported toxicity for these strains. These results suggest that CYN is lethal to zebrafish embryos, but apparently inhibits no specific developmental pathways, whereas other unknown metabolites from C. raciborskii and A. ovalisporum may be involved in developmental dysfunctions in zebrafish.

Evidence of the accumulation and effects of cyanobacterial neurotoxins on fish is still scarce. Osswald et al. [202] was the only study that showed the accumulation of ANTX-a in juveniles of the common carp (Cyprinus carpio) exposed to freeze-dried cells of Anabaena sp. (ANA 37), calculating a maximal theoretical bioaccumulation factor (BAF) of 2.65. They also observed a high lethality in concentrations of 10⁷ cells mL⁻¹, with 100% mortality after 26–29 h exposure. In another study, Osswald et al. [203] exposed embryos of the same fish to extracts of the same Anabaena strain and pure ANTX-a and verified the effects on several parameters of the developmental stages such as time to mortality, mortality rate, time to hatching, hatching rate, skeletal malformations rate, and larval standard length. At all tested concentrations of ANTX-a, the pure toxin was almost harmless to carp early stages of development, contrarily to cell extracts that were highly toxic. Only an adverse effect on the larval length was found at the highest concentration of pure toxin, while increasing concentrations of cell extracts caused increasing adverse effects in all parameters. Authors argued that other bioactive substances could be responsible for the extracts toxicity such as the lipopolysaccharides (LPS), mueggelone and lupenyl acetate, all found in strain ANA 37.

Oberemm et al. [204] reported that pure STX delayed hatching of zebrafish embryos and led to malformations (i.e., lateral and ventral body curvature and edema) and mortalities, while ANTX-a altered the heart rate in zebrafish, but no chronic effects were observed. Compared to results with pure toxins, much more pronounced effects were observed inembryos when exposed to crude aqueous extracts of cyanobacteria from field samples and batch cultures, but based on the toxin analysis they concluded that these effects cannot be explained by known toxins alone. Therefore, these studies emphasize that, even when using ecologically relevant toxin concentrations, the effects of pure toxins do not mimic the effects in the natural environment. The synergistic actions of different toxins and of toxins in complex matrix samples may increase toxicity considerably.

Clemente et al. [205] found GTX variants in fish muscle (0.012–0.02 µg STX eq. g⁻¹ WW) of Geophagus brasiliensis in Alagados Reservoir (Southern Brazil), along with histopathological and several biomarkers alterations such as LPO, genetic damage (comet assay), GST, AchE and EROD activities. In experiments with the carnivorous fish Hoplias malabaricus, Silva et al. [206] offered contamined prey (Astyanax sp.) that were intraperitonealy inoculated with a lysate of C. raciborskii culture containing PSP constituted by 97% STX, 3% neosaxitoxin (NeoSTX) plus gonyautoxin (GTX2). Although they did not find any PST analogue in fish muscle, the activity of the enzyme SOD increased, suggesting the generation of free radicals, whereas the activities of CAT, GST and GPx decreased, suggesting an inhibition of the antioxidant system in the brain. In addition, lipoperoxidation (LPO) and protein carbonylation (PCO) increased along with damages in DNA (comet assay), suggesting the occurrence of oxidative stress, genotoxic damage and apoptotic processes in fish exposed to STX.

Along with the biochemical and physiological effects, neurotoxins may cause behavioral changes in fish. Lefebvre et al. [9] showed that STXs can alter the sensorimotor function of herring (Clupea harengus pallasi), decreasing its response to spontaneous and touch-activated swimming behavior. However, the normal motor function recovered after 4–24 h of continuous exposure. In contrast, Ferrão-Filho et al. [10], showed that D. rerio exposed to a STXs-producer strain of C. raciborskii (CYRF-01) and to water samples from a eutrophic reservoir containing this and other cyanobacteria, had a significant increase in swimming activity (i.e., mean distance performed and mean velocity). The result obtained by Lefebvre et al. [9] is more compatible with the mechanism of action of STXs (i.e., paralysis), suggesting that other toxins or compounds, which may have irritant properties (e.g., LPS), probably in contact with the gills, might be involved in the stress response of D. rerio [10]. Osswald et al. [202] showed that intact cells of an ANTX-a producer strain of Anabaena caused an increase in the opercular movement and abnormal swimming of juveniles of the common carp (Cyprinus carpio). Despite contradictory results, all these effects on fish behaviour by neurotoxic strains may have negative consequences on the fitness of fish populations by changes in reproductive aspects and on predator—prey interactions.

Cyanobacteria may also cause adverse effects to other aquatic vertebrates. Oberemm et al. [204] tested the effects of purified MCs, STXs and crude extracts of cyanobacterial strains on the caudate amphibians axolotl (Ambystoma mexicana) and smooth newt (Triturus vulgaris), and on marsh frog (Rana ridibunda). Although no effects of MCs were observed during short-term embryonic development test for all amphibian species, prolonged exposure (35 days) of axolotl to the highest concentration of MC-LR resulted in less developed forelimbs of larvae. STX caused no acute effects on axolotl, but hatching started slightly earlier at lower STX concentrations (10 µg L⁻¹), whereas higher concentrations (50–500 µg L⁻¹) led to a minor delay in hatching. Crude extracts, however, lead to much more pronounced effects. Extracts of Anabaena floes-aquae (ANTX-a producer), for example, caused a dorsal curvature of the axolotl’s tail, along with cranio-facial malformations and gill hyperplasia.

Dvořáková et al. [156] studied the effects of purified MC-LR and samples of natural water blooms of Microcystis that contained or not MCs on the African clawed frog (Xenopus laevis) embryos in a 96 h teratogenesis assay (FETAX). They showed that pure MC-LR caused weak lethality at the highest concentration tested (LC₂₅ = 380 µg L⁻¹) and that cyanobacterial biomass caused significant lethality to frog embryos, but biomass containing MC-LR had a higher lethality (LC₂₅ = 125 mg DW L⁻¹ or 103 µg MC-LR L⁻¹) compared to biomass without MC-LR (LC₂₅ = 232 mg DW L⁻¹). Purified MC-LR as well as both samples of biomass (with and without MC-LR) induced frequent developmental abnormalities in the spinal area (backbone, tail) and malformations of other organs, such as the cranio-facial segment (head, brain, and eyes). Purified MC-LR, however, induced more frequent improper gut coiling and malformations in the abdomen and occurred even at very low concentrations (10–25 µg L⁻¹). The teratogenic potential (TI), calculated as the ratio of lethal concentration (LC₂₅) to effective concentration (EC₂₅) that cause 25% malformations, indicated that purified MC-LR, as well as both types of biomass, can impose a significant embryotoxic risk to freshwater vertebrates. Additional experiments with external additions of purified MC-LR to both types of biomass revealed little additive effect. The authors explained the results either by a possible decreased bioavailability of MC-LR in the complex biomass (sorption to proteins or membrane lipids) or by an overlapping effect of other unidentified metabolites present in the complex cyanobacterial bloom.

Burýšková et al. [157] also studied the effects of complex natural bloom samples on the African frog (X. laevis) embryos, but besides classical parameters (mortality, growth inhibition and malformations) evaluated in 96 h FETAX, they also assessed the effects on biochemical markers of oxidative stress and detoxification (glutathione pool, GSH; GPx activity, GR and GST activity). In general, complex biomass samples containing several genera of cyanobacteria (Fraction I) and aqueous extracts (Fraction III) were generally the most toxic fractions in terms of mortality and growth inhibition, whereas methanolic eluates passed through C18 column containing MCs (Fraction V) were less toxic. On the other hand, the same fraction (eluates) induced significant malformations in low concentrations but the effects were not related to the MCs content. Curiously, the strongest effect was caused by permeate from C18 column without detectable MCs (Fraction IV). Oedema, adverse curving of backbone, eye and mouth deformities were the most often observed malformations in cyanobacterial extracts. Also, significant growth reductions were recorded after exposure to all fractions of complex biomass, except for the eluate containing MCs. Biomarkers were affected in a variable manner and no significant or clear concentration-dependent effect was observed. Their data support the hypothesis that MCs are not the only or major toxic compounds in the complex cyanobacterial samples.

White et al. [158] presented the first evidence of bioaccumulation and effects of CYN-producing C. raciborskii in the developmental stages of an amphibian (Bufo marinus). Maximum average tissue concentrations of 895 and 61 µg free-CYN kg⁻¹ FW were measured after exposure to live culture containing cell-bound toxin and to whole cell extracts containing extracellular toxin, respectively, which resulted in up to 66% mortality of tadpoles exposed to live cultures, whereas tadpoles exposed to whole cell extracts containing similar toxin concentrations survived. Both types of exposure lead to decreases in relative growth rate and time spent for swimming. This study, similarly to others [71,111,201,203], indicates that absorption of dissolved toxins from cell extracts is relatively low compared to the ingested cell-bound toxins.

Evidence of bioaccumulation and toxicity of cyanotoxins to other aquatic vertebrates (i.e., reptiles and birds) are very scarce, and literature is mainly based on reports of mass mortality of birds and waterfowls, insufficiently linked to the presence of cyanotoxins (reviews about bird poisonings can be found in [31,85]).

Recently, Chen et al. [160] reported the presence of MCs in the domestic duck Anas platyrhynchos and in the black-crowned night heron Nycticorax nycticoraxs in Lake Taihu (China). Among the various organs analyzed, they found higher amounts of MCs in the intestines (51–82 ng g⁻¹ DW), liver (17–31 ng g⁻¹ DW) and stomach (10–21 ng g⁻¹ DW), and smaller amounts in the pancreas, gallbladder, kidney, heart, lung, spleen, gonads and muscles. They found also considerable amounts of MCs in the egg yolk (8–15 ng g⁻¹ DW), which indicates a possible risk for bird embryos. In spite of this, no bird mortality was reported in this study.

One of the first experimental studies using aquatic birds reported differences between mallard ducks and ring-necked pheasants for the ANTX-a produced by the strain of Anabaena flos-aquae NRC-44-1 [207]. More recent experimental work has used the Japanese quail (Coturnix coturnix japonica) as a model bird species [208]. In this study, male Japanese quails were offered Microcystis biomass containing increasing doses of MCs for 10 (acute) and 30 (sub-chronic) days, but showed no mortality. However, histopathological hepatic alterations in birds were reported after exposure, including cloudy swelling of hepatocytes, vacuolar dystrophy, steatosis and hyperplasia of lymphatic centers. On a subcellular level, shrunken nuclei of hepatocytes containing ring-like nucleoli, cristolysis within mitochondria and vacuoles with pseudomyelin structures were present. Vacuolar degeneration of the testicular germinative epithelium was also found in two exposed males. Evaluation of biochemistry and haematological parameters showed increased activities of lactate dehydrogenase (LD) and a drop in the blood glucose (GLU) on day 10, only for the highest dose (46 mg MC per 10 mL⁻¹ day⁻¹).

In another study with Japanese quails, Pašková et al. [163] studied the effects on detoxification and antioxidant enzymes as well as bioaccumulation of MCs after acute (10 days) and sub-lethal (30 days) exposure to natural cyanobacterial biomass containing MCs. Significant accumulation of MCs was observed in the liver for both test durations and slight accumulation also in the muscles of the highest treatment group from the acute test. The greatest accumulation was observed in the liver of the highest treatment group in the acute test reaching an average concentration of 43.7 ng MCs g⁻¹ FW. Cyanobacterial exposure caused an increase of activity representing the activation phase of detoxification metabolism (EROD), mainly in the heart and brain tissues in the acute exposure. Also the conjugation phase of detoxification, mainly the activity GST, was altered in the liver and heart in the sub-chronic exposure. Cyanobacterial exposure also modulated oxidative stress responses including the level of glutathione and activities of glutathione-related enzymes and caused increases in lipid peroxidation in most tissues, both in acute and sub-chronic exposures. Although the Japanese quail is not representative of wild aquatic birds, they may be considered as models for future investigations of wildlife mortalities related to toxic blooms of cyanobacteria.

The only evidence of mortality of reptiles is the study of Nasri et al. [159], who reported a case of mortality of the freshwater terrapin turtle species Emys orbicularis and Mauremys leprosa during a Microcystis bloom in Lake Oubeira (Algeria, Tunisia). MC-LR, -YR and -RR were detected in this bloom extract and also in fresh carcasses of terrapin liver, viscera and muscle tissues using the GC/MS after Lemieux oxidation method and the PP2A inhibition assay. The highest level of MC-LR equivalents detected using the Lemieux oxidation-GC/MS method was found in the liver (1193 µg g⁻¹ DW) and in the viscera tissues (37.2 µg g⁻¹ DW) of M. leprosa and E. orbicularis, respectively, and smaller levels in the muscle (9.4–10.1 µg g⁻¹ DW). The liver crumbling observed after the necropsy examination of the fresh carcasses of M. leprosa support the hypothesis that hepatotoxins contributed to the turtle mortalities. Although both species are prevalently carnivorous, M. leprosa diet is highly biased towards filamentous algae, insect larvae, earthworms, mollusks, small amphibians and tadpoles, and various aquatic plants. Thus, the most likely uptake of MCs by these turtles is from food chain. Chen et al. [160] also found MCs in several organs, including the liver and intestines, of the carnivorous turtle Pelodiscus sinensis in Lake Taihu (China), but no mortality was reported.