On the Chemistry, Toxicology and Genetics of the Cyanobacterial Toxins, Microcystin, Nodularin, Saxitoxin and Cylindrospermopsin

The cyanobacteria or “blue-green algae”, as they are commonly termed, comprise a diverse group of oxygenic photosynthetic bacteria that inhabit a wide range of aquatic and terrestrial environments, and display incredible morphological diversity. Many aquatic, bloom-forming species of cyanobacteria are capable of producing biologically active secondary metabolites, which are highly toxic to humans and other animals. From a toxicological viewpoint, the cyanotoxins span four major classes: the neurotoxins, hepatotoxins, cytotoxins, and dermatoxins (irritant toxins). However, structurally they are quite diverse. Over the past decade, the biosynthesis pathways of the four major cyanotoxins: microcystin, nodularin, saxitoxin and cylindrospermopsin, have been genetically and biochemically elucidated. This review provides an overview of these biosynthesis pathways and additionally summarizes the chemistry and toxicology of these remarkable secondary metabolites.


Introduction
The heptapeptide hepatotoxin, microcystin, has been isolated from multiple genera of cyanobacteria, including Microcystis, Anabaena, Oscillatoria, Planktothrix, Chroococcus and Nostoc. Microcystin-producing strains such as Microcystis aeruginosa have a cosmopolitan distribution and thrive in a range of climates, making these organisms a global threat to human health. Consequently, significant research efforts have been directed towards their identification and eradication.

Chemistry
The microcystins comprise the largest and most structurally diverse group of cyanobacterial toxins. Around 90 microcystin isoforms varying by degree of methylation, hydroxylation, epimerization, peptide sequence and toxicity have been identified [1,2]. Underlying the extraordinary heterogeneity present among the microcystins is their common cyclic structure ( Figure 1) and possession of several rare, highly conserved amino acid moieties. Collectively, the microcystins may be described as monocyclic heptapeptides containing both D-and L-amino acids plus N-methyldehydroalanine and a unique ß-amino acid side-group, 3-amino-9-methoxy-2-6,8-trymethyl-10-phenyldeca-4,6-dienoic acid (Adda) [3]. The microcystin isoforms differ primarily at the two L-amino acids, and secondarily on the presence or absence of the methyl groups on D-erythro--methylaspartic acid (D-MeAsp) and/or N-methyldehydroalanine (Mdha) [4]. However, substitutions of all moieties within microcystin have been reported [5][6][7].  The mcyF ORF was originally predicted to encode a glutamate racemase, responsible for the epimerization of the L-Glu residue of microcystin [22,26]. A subsequent study by [27] contended this theory and offers evidence that McyF acts exclusively as an Asp racemase. The authors propose that the D-Glu residue is provided by an L-Glu racemase residing outside the mcyS gene cluster. Mutagenesis experiments in P. agardhii showed that the production of Adda also involves an O-methylation step catalyzed by the putative monofunctional tailoring enzyme, McyJ [23].
The remaining biosynthetic enzymes in the microcystin biosynthesis pathway (NRPSs) are putatively involved in the specific activation, modification and condensation of substrate amino acids onto the linear peptide chain, which is then cyclized to produce microcystin. Firstly, McyA adds L-Ser to the growing chain, followed by the addition of D-Ala. This step is followed by the addition of L-Leu and D-MeAsp residues (McyB) followed by the addition of L-Arg (McyC), and subsequent cyclization and release of the final peptide product (Figure 3b).
The remaining stand alone enzyme, the 2-hydroxy-acid dehydrogenase, McyI, is putatively involved in the production of D-methylaspartate at position three within the microcystin cyclic structure via the conversion of 3-methylmalate to 3-methyloxalacetate. It is hypothesized that a promiscuous aspartate aminotransferase then converts 3-methyloxalacetate to methylaspartate [28].
An ABC transporter gene, mcyH, is believed to be involved in the transport of microcystin [29]. This transporter may be responsible for the thylakoid localization of the toxin [30,31] or for the extrusion of the toxin under certain growth conditions, including exposure to high and red light [32].  Comparative studies of the mcyS gene clusters from M. aeruginosa, P. agardhii [23], and Anabaena sp. [24] have noted variation in the arrangement of mcyS genes between these different species of cyanobacteria, although the proposed toxin biosynthetic processes are thought to be similar. The M. aeruginosa and Anabaena sp. mcyS clusters are both arranged into two divergently transcribed operons, however, the arrangement of genes within these operons differs between the two species ( Figure 2). In P. agardhii, the mcyS cluster also has a distinctive arrangement and lacks mcyF and mcyI. Furthermore, the P. agardhii mcyS cluster contains an additional gene mcyT, upstream of the central promoter region. This gene is thought to encode a putative type II thioesterase enzyme, which may play an editing role by removing mis-primed amino acids from the NRPS and PKS enzymes. The characterization of mcyS in M. aeruginosa, P. agardhii and Anabaena sp. has important implications for understanding the origins and evolution of hepatotoxin biosynthesis in cyanobacteria. The identification of transposases associated with the mcyS and ndaS (nodularin) gene clusters and subsequent phylogenetic analysis has led to the theory that horizontal gene transfer and recombination events are responsible for the sporadic distribution of the mcyS gene cluster throughout the cyanobacteria and the various microcystin isoforms that have been identified to date [22,33,34].
Hepatotoxin production in cyanobacteria is thought to be influenced by a number of different physical and environmental parameters, including nitrogen, phosphorous, trace metals, growth temperature, light, and pH [35][36][37][38][39][40][41]. However, due to the fact that most regulatory investigations have not been standardized, and the data have not been interpreted against the same specific growth controls, the subject of hepatotoxin regulation remains a somewhat contentious issue. While most toxin regulation studies have focused on direct measurements of cellular toxin, the description of the mcy gene cluster by Tillett and co-workers [22] enabled a closer examination of microcystin regulation at the molecular level [32]. Kaebernick et al. used the RNase protection assay to measure the transcription of mcyB and mcyD under a variety of different light conditions. High light intensities and red light were correlated with increased transcription, while blue light led to reduced transcript levels. Interestingly, the authors observed two light thresholds, between dark and low light (0 and 16 µmol photons m -2 s -1 ), and medium and high light (31 and 68 µmol photons m -2 s -1 ), at which a significant increase in transcription occurred. The same group later found that transcription of mcy genes occurs via two polycistronic operons, mcyABC and mcyDEFGHIJ, from a central bidirectional promoter between mcyA and mcyD [42]. Interestingly, alternate transcriptional start sites were identified for both operons when cells were cultured under different light intensities. For example, under low light conditions, the polyketide and tailoring genes mcyD-J are transcribed as part of a polycistronic message (mcyDEFGHIJ) from a central (mcyD) promoter, while under high light conditions, the genes are transcribed from an alternative up-stream promoter. It is thought that initiation from the alternate promoters under high light conditions may lead to increased transcription, as previously observed for mcyB and mcyD. Many of the tailoring enzymes (mcyF, G, H, I and J) also possess their own individual promoters [32].
Interestingly, light intensity also appears to favor the production of certain microcystin variants over others. For example, Tonk et al. (2005) found that the cellular content of total microcystin remained constant, independent of the irradiance. However, of the two main microcystin variants detected in P. agardhii, the microcystin-DeRR content decreased two-fold with increased photon irradiance, whereas the microcystin-DeLR content increased three-fold. Since microcystin-DeLR is considerably more toxic than microcystin-DeRR, this implies that P. agardhii becomes more toxic at high light intensities [40].
Other factors such as nutrient content and temperature have also been demonstrated to affect mcyS expression and toxin biosynthesis. For example, Sevilla et al. (2008) investigated the effect of iron on mcyS expression and toxin biosynthesis in M. aeruginosa PCC7806 [41]. Real-time PCR analysis and HPLC were used to measure transcription of mcyD and the synthesis of microcystin-LR, respectively. The results of this study suggested that iron starvation causes an increase in mcyD transcription, correlative to the increase of toxin levels [41]. Davis et al. (2009) investigated the effects of temperature on Microcystis growth and toxin genes and found that elevated temperatures yield more toxic Microcystis cells and/or cells with more mcyD copies per cell, with either scenario potentially yielding more toxic blooms [39].

Introduction
The cyclic pentapeptide nodularin is most commonly isolated from the filamentous, planktonic cyanobacterium, Nodularia spumigena. This species generally forms toxic blooms in brackish and estuarine environments. Blooms of toxic N. spumigena occur annually during summer months in the Baltic Sea [43] resulting in nodularin being one of the most abundant naturally occurring compounds in the Baltic Sea. Blooms are also particularly common within the estuaries and coastal lagoons of Australia [44,45], and have been reported worldwide, including on the German North Sea coast [46], New Zealand [47], and North America [48]. N. spumigena scums have also been reported in the fresh to brackish/saline lakes at the mouth of the lower River Murray, South Australia, which is a major source of both potable and irrigation water [49]. Nodularin is structurally similar to microcystin and can induce similar toxic effects. The toxin has been reported to have detrimental effects on numerous organisms within the ecosystem, including invertebrates and fish, but may have no effect on other organisms [50].
The consumption of water containing toxic N. spumigena blooms has led to the death of domestic and native animals by massive liver haemorrhage [46,47,51,52]. In sub-acute doses, nodularin, like microcystin, is thought to act as a liver tumor initiator and promoter [53]. N. spumigena blooms are also of importance to the seafood industry. Nodularin has been shown to accumulate in shellfish and other seafood. Surveys of mussels, prawns, flounder both in Australia and the Baltic Sea report that nodularin accumulated to levels of concern with the potential to cause hepatotoxicity [54-56].

Chemistry
Nodularin is a cyclic pentapeptide with a similar structure to microcystin, consisting of Adda, Seven naturally-occurring isoforms of nodularin have been reported to date. Two of these isoforms, produced by a New Zealand Nodularia sp. bloom, have variations within the Adda residue, which reduces or abolishes the toxicity of the compound [5]. The D-Glu residue is essential for toxicity of nodularin, as esterification of its free carboxyl abolishes toxicity, however, substitution at position 1 has little effect on toxicity. The other two isoforms, nodularin-Har and motuporin, are variable at position 2. Nodularin-Har is produced by the strain N. harveyana PCC7804, with the L-Arg, replaced with L-Homoarginine (L-Har) [58,59]. Motuporin has been isolated from the Papua New Guinea sponge Theonella swinhoei, and may be synthesized by an associated cyanobacterium. The L-Arg residue of nodularin is replaced by L-Val in motuporin [60]. The L-Val residue is responsible for additional cytotoxicity of motuporin against cancer cell lines.

Toxicology
Nodularin is a potent hepatotoxin in humans and other animals. Nodularin induces liver hemorrhage in mice and has a lethal dose 50 (LD 50 ) of 50 g.kg -1 (intra-peritoneal route of injection) in mice [61]. At doses below this concentration, nodularin may act as a carcinogen via the initiation and promotion of liver cell division [53].
The hepatotoxicity and carcinogenicity of nodularin is associated with the inhibition of eukaryotic protein phosphatase (PP) catalytic subunit types 1 and 2A [62]. The toxin inhibits the activity of PP2A to a greater extent than PP1. Inhibition of PP2A by nodularin occurs at relatively the same concentration (IC 50 ) as that of microcystin (~0.1 nM) [62]. The hydrophobic C 20 amino acids Adda, present in both toxins, blocks PP enzyme activity by interacting with the hydrophobic groove and obstructing substrate access to the active site cleft [63][64][65]. MeDhb binds to Cys273 of PP2A in a similar fashion to the MeDha residue in microcystin which binds to Cys273 of PP1 and Cys266 of PP2A [64,66], however, binding of the toxin does not occur covalently and may be the reason for its additional carcinogenic properties [67].
The toxic effects of nodularin are primarily associated with the hepatocytes due to active transport of the toxin to the liver via the bile acid multi-specific organic anion transporters [10]. To date, studies have been unable to identify the specific mechanism of transport.
Nodularin has been observed to accumulate on different trophic levels, in numerous organisms including waterfowl, fish, mysid shrimp, zooplankton and benthic organisms [68]. Mesozooplankton in particular seem to play a major role in nodularin transfer to planktivorous fish [68]. The toxin appears to cause oxidative stress in the tissues in which it accumulates. In the case of the flounder (platichthys flesus L.), this oxidative stress occurs in the liver by way of reduced GST and CAT activities [69]. However, recent studies suggest that the toxin is rapidly detoxified and broken down or excreted [70].

Biosynthesis and Genetics
The nodularin biosynthesis gene cluster ndaS, from Nodularia spumigena NSOR10, was sequenced and characterized in 2004 by Moffitt and Neilan [71]. The 48 kb region of the genome consists of nine ORFs (ndaA-I) transcribed from a bidirectional regulatory promoter region ( Figure 2). While most of the ndaS encoded genes have homologs in the mcyS cluster, their arrangement adheres more closely to the 'co-linearity' rule of NRPS pathways that predicts the order of catalytic processes involved in the biosynthesis of a non-ribosomal metabolite is generally the same as the order of the genes which encode their catalytic enzymes [72].
The proposed pathway for nodularin biosynthesis is similar to that for microcystin. Functional assignment of the enzymes was based on bioinformatic analysis and homology to the microcystin synthetase enzymes. The Adda side-chain is produced via a mixed NRPS/PKS pathway from a phenylacetate starter unit and several malonyl-CoA extensions (NdaC, D and F) ( The ndaS cluster also encodes several putative monofunctional tailoring enzymes that may play a role in the modification and transport of nodularin. ndaE encodes an O-methyltransferrase, ndaG encodes a putative L-Asp/L-Glu racemase, and ndaI encodes an ABC transporter. Also encoded within the ndaS cluster is a D-3-PGDH homolog, NdaH, which shares 71% identity with McyI. It is likely therefore, that NdaH may be involved in the production of D-MeAsp [28].
Like mcyS, the ndaS gene cluster is transcriptionally regulated by a bi-directional promoter region. Analysis of transcription of the ndaS cluster found that it is transcribed as two polycistronic mRNA, ndaAB, ORF1, and ORF2, and ndaC [71]. The two genes downstream of ndaAB, ORF1 and ORF2, encode a putative transposase and a putative high light-inducible chlorophyll-binding protein, respectively. It is not clear why the putative transposase and the putative high light-inducible chlorophyll-binding protein are also co-transcribed with the ndaS gene cluster. ORF2 has been identified in all strains of toxic Nodularia and the association between ORF2 and nodularin biosynthesis may suggest a physiological function associated with high-light stress in the cells producing it. A putative heat shock repressor protein, encoded by the gene ORF3, was also identified downstream of ORF2, which may be involved in the transcriptional regulation of the ndaS genes in response to heat stress. More recently, elucidation of the nda cluster has provided an opportunity to monitor transcriptional regulation of the biosynthetic pathway [74]. The effects of ammonia and phosphate starvation were analyzed. While expression of the nda cluster appears to be constitutive, phosphate starvation resulted in an approximately two-fold increase in expression, while ammonia supplementation decreased expression two-fold. Despite the changes to expression, intracellular and extracellular nodularin concentration remained stable [74].

Introduction
Saxitoxin and its analogs, collectively termed paralytic shellfish poisons (PSPs), are highly potent neurotoxins. Several freshwater species of cyanobacteria and marine dinoflagellates are known to produce saxitoxins. Blooms of these toxic species have led to mass kills of fish, native animals and livestock, as well as the contamination of freshwater resources [75][76][77]. Saxitoxins typically accumulate through the food chain, in organisms consumed as seafood [78]. Marine shellfish are particularly resistant to the toxins, and can therefore accumulate dangerously high levels of PSPs by ingesting toxic plankton [79]. Saxitoxin and its analogs cause an annual estimated 2000 cases of PSP globally, with a mortality rate of 15% [80].
The capacity to synthesize saxitoxins and other PSPs has an unusually wide phylogenetic distribution, including both marine and freshwater organisms from two kingdoms of life (Eubacteria and Protista). Typically, there is little convergence between the structures of secondary metabolites from marine and freshwater organisms [81], or from organisms belonging to different phylogenetic groups. Although several studies suggest that other bacteria are capable of synthesizing PSPs [82,83], the findings are controversial, as the analytical methods used were not definitive and could not be readily repeated by other researchers [84]. On the other hand, PSP biosynthesis in cyanobacteria and dinoflagellates has been extensively verified and shown to be a consistent and heritable genetic trait [85][86][87][88][89][90].
In the marine environment, dinoflagellate species capable of producing saxitoxins belong to the genera Alexandrium, Pyrodinium and Gymnodinium [91][92][93]. While in freshwater systems, several filamentous species of cyanobacteria, such as Anabaena circinalis, Aphanizomenon sp., Aphanizomenon gracile, Cylindrospermopsis raciborskii and Lyngbya wollei are also known to produce saxitoxins [94][95][96][97]. Saxitoxin production is varied among dinoflagellate and cyanobacterial producer genera, with not all species in a toxigenic genera being toxic, and different isolates of the same species having differential toxicity. Furthermore, cyanobacterial isolates of the same species from geographically distant locations provided different toxin profiles [89].

Chemistry
Saxitoxin is a trialkyl tetrahydropurine and the parent compound of more than 30 naturally occurring derivatives that differ structurally at four positions [98] (Figure 6.). The variable positions may be hydroxylated, sulfated or carbamoylated. Most of these analogs have been detected in both cyanobacteria and dinoflagellates [95,96,[98][99][100][101]. In addition to the usual carbamate and decarbamoyl toxins, six new saxitoxin derivatives have been isolated from the cyanobacteria Lyngbya wollei, including derivatives which provided an acetyl side-chain instead of the typical carbamate, or lacked one of the C-12 hydroxyl moieties [102]. Further unusual modifications, such as a C-11 ethanoic acid side-chain [103], or the N-hydroxylation of the carbamoyl side-chain [104], have also been described in xanthid crabs. In addition, the freshwater puffer fish was shown to produce a unique saxitoxin analog where the carbamoyl side-chain is N-methylated [105]. Furthermore, the Panamanian golden frog Atelopus zeteki produces zetekitoxin AB, which was recently confirmed to be a novel analog of saxitoxin, with sodium channel blocking activity 580-fold higher than that of saxitoxin [106]. Whether or not these complex eukaryotic organisms actually produce PSPs or simply accumulate the toxins in their tissues and organs is an issue of contention.

Toxicology
Saxitoxin and its derivatives are the causative agents of the common seafood poisoning, paralytic shellfish poisoning (PSP). PSP symptoms generally onset within 30 minutes of ingestion, and invariably begin with a tingling or burning of the lips, tongue and throat, increasing to total numbness of the face [98]. Further symptoms may include perspiration, vomiting and diarrhea. In cases of acute poisoning, numbness may spread to the neck and extremities and progress to muscular weakness, loss of motor coordination, and finally paralysis. A lethal dose of saxitoxin usually results in cardiovascular failure due to respiratory muscle paralysis [98]. There is no clinically approved antidote to saxitoxin poisoning, and treatments during early stages of PSP include removal of unabsorbed toxin with activated charcoal, and artificial respiration. The half-life of saxitoxin in the body is approximately 90 minutes, and survival chances increase significantly after 12 hours from initial exposure [107]. It is a highly potent blocker of voltage gated sodium channels present in neuronal cell membranes [108]. In addition, saxitoxin has also been shown to block calcium channels [109], and to prolong the gating of potassium channels in heart muscle cells [110]. Saxitoxin is a highly potent phycotoxin with an intraperitoneal LD 50 of 10 μg/kg body weight in mice [111], while human death has occurred following the ingestion of as little as 1 mg of the toxin [112].
Mammals usually contain multiple isoforms of sodium channels, with different sensitivity to saxitoxin. This difference in sensitivity is due to variation in the amino acid sequence of the channel alpha subunit, whereby a single mutation may confer insensitivity to saxitoxin, though usually sacrificing speed of gating in the process [113]. The guanidinium groups and the carbon 12 hydroxyls in STX have been shown to be critical for the binding of the sodium channel, while the carbamoyl side chain also appears to be involved in the binding process, whereby one saxitoxin molecule binds one channel by lodging itself in the ion-conducting pore [113][114][115][116].
The toxicity of saxitoxin derivatives varies greatly, with the carbamate toxins being 10-100-times more potent than the N-sulfo-carbamoyl derivatives [98,117]. N-sulfo-carbamoyl analogs are, however, labile and may easily be converted to the more toxic carbamate derivatives [118,119].

Biosynthesis and Genetics
Using a reverse genetic approach, Kellmann and co-workers [120] identified the gene cluster putatively responsible for the biosynthesis of saxitoxin in Cylindrospermopsis raciborskii T3 (sxt) (Figure7c). The sxt gene cluster is encoded by more than 35 kb and comparative sequence analysis assigns 30 catalytic functions to 26 proteins. Bioinformatic analysis of this cyanobacterial saxitoxin gene cluster, coupled with identification of novel biosynthetic intermediates enabled a revision of the previously proposed saxitoxin biosynthesis pathway ( Figure 8).
The first step in the revised saxitoxin biosynthesis pathway involves a Claisen condensation reaction catalyzed by SxtA. This unique enzyme possesses a polyketide synthase (PKS)-like structure composed of four catalytic domains, SxtA1-4: SxtA1 is homologous to SAM-dependant methyltransferases; SxtA2 is related to GCN5-related N-acetyl transferases (GNAT) that transfer acetate from acetyl-CoA to various heteroatoms [121]; SxtA3 is related to ACPs and also provides a phosphopantetheinyl-attachment site; SxtA4 is homologous to class II aminotransferases and is most similar to AONS (8-amino-7-oxononanoate synthase). The predicted reaction sequence of SxtA, based on its primary structure, is the loading of the ACP (SxtA3) with acetate from acetyl-CoA, followed by the SxtA1-catalyzed methylation of acetyl-ACP, converting it to propionyl-ACP. SxtA4, the class II aminotransferase domain, then performs a Claisen condensation reaction between propionyl-ACP and arginine. The putative product of SxtA is thus 4-amino-3-oxo-guanidinoheptane, which Kellmann et al. designated compound A.
The sxtG gene encodes a putative amidinotransferase, with highest amino acid sequence similarity to L-arginine/L-lysine amidinotransferases. SxtA is the putative substrate for SxtG, which transfers an amidino group from arginine to the α-amino A' group, thus producing 4,7-diguanidino-3-oxoheptane (designated compound B'). SxtB, an enzyme similar to the cytidine deaminase-like enzymes from gammaproteobacteria, then catalyzes a retroaldol-like condensation in the conversion from B' to C'.
The putative sterol desaturase, SxtD is predicted to introduce a double bond between C-1 and C-5 of C', resulting in the 1,2-H shift between C-5 and C-6 (compound D'). The gene product of sxtS, which has sequence homology to nonheme iron 2-oxoglutarate-dependent dioxygenases, is predicted to perform the consecutive epoxidation of the new double bond and opening of the epoxide to an aldehyde with concomitant bicyclization. SxtU has sequence similarity to short-chain alcohol dehydrogenases and is therefore predicted to reduce the terminal aldehyde group of the saxitoxin precursor forming compound E'. The concerted action of SxtD, SxtS, and SxtU is therefore responsible for the hydroxylation and bicyclization of compound C' to E'.
The gene product of sxtI is most similar to a predicted O-carbamoyltransferase from Trichodesmium erythraeum and other cyanobacteria. Kellmann et al.'s data [120] indicate that SxtI may catalyze the transfer of a carbamoyl group from carbamoylphosphate to the free hydroxy group of E'. Adjacent to sxtI are two short ORFs of unknown function, sxtJ and sxtK. While sxtJ and sxtK homologs are available in the databases, none of these genes have been functionally characterized.
sxtH and sxtT, each encode a terminal oxygenase subunit similar to those found in bacterial phenylpropionate and related ring-hydroxylating dioxygenases. SxtH and SxtT may therefore perform the consecutive hydroxylation of C-12, converting F' into saxitoxin. Members belonging to bacterial phenylpropionate and related ring-hydroxylating dioxygenases are multicomponent enzymes, as they require an oxygenase reductase for their regeneration after each catalytic cycle. The sxt gene cluster provides a putative electron transport system, which would fulfill this function in the form of SxtV and SxtW. SxtV, a 4Fe-4S ferredoxin, could putatively extract an electron pair from succinate, converting it to fumarate [122]. SxtW a fumarate/reductase/succinate dehydrogenase homolog could then transfer the electrons via ferredoxin to SxtH and SxtT.
sxtX encodes an enzyme with homology to cephalosporin hydroxylase. sxtX was detected only in C. raciborskii T3, Aphanizomenon flos-aquae NH-5, and Lyngbya wollei, which produce N-1-hydroxylated analogs of saxitoxin [94,108,123], such as neoSTX. This component of the gene cluster was not present in any strain of A. circinalis, and therefore probably represents the reason why this species does not produce N-1-hydroxylated PSP toxins [89,99]. The predicted function of SxtX is therefore the N-1 hydroxylation of saxitoxin.
A. circinalis AWQC131C and C. raciborskii T3 also produce N-and O-sulfated analogs of saxitoxin [GTX-5, C-2/C-3, (dc)GTX-3/GTX-4]. The activity of two 3'-phosphate 5'-phosphosulfate (PAPS)-dependent sulfotransferases, which were specific for the N-21 of saxitoxin and GTX-3/GTX-2 and the O-22 of 11-hydroxy saxitoxin, respectively, has been described previously in studies of the PSP toxin-producing dinoflagellate Gymnodinium catenatum [124,125]. A putative sulfotransferase encoded by sxtN is predicted to transfer a sulfate group to either N-21 or O-22. Interestingly, the sxt gene cluster also encodes an adenylylsulfate kinase (APSK), SxtO, putatively involved in the formation of PAPS. Other biosynthetic gene clusters that result in sulfated secondary metabolites also contain genes required for the production of PAPS [126].
Decarbamoylated analogs of STX could be produced via either of two hypothetical scenarios. Enzymes that act downstream of SxtI, the carbamoyltransferase, in the biosynthesis of PSP toxins are proposed to exhibit broad substrate specificity, processing both carbamoylated and decarbamoylated precursors of STX. Alternatively, hydrolytic cleavage of the carbamoyl moiety from STX or its precursors may occur. SxtL is related to GDSL lipases, which are multifunctional enzymes with thioesterase, arylesterase, protease, and lysophospholipase activities [127]. The function of SxtL could therefore include the hydrolytic cleavage of the carbamoyl group from STX analogs.
Kinetic studies of PSP toxin accumulation in producing cells and the media of cyanobacterial cultures suggest that there is an active transport mechanism for these toxins [85]. In addition, variations in the concentration of sodium in culture media are known to affect the accumulation of PSP toxins in producer cells [128]. sxtF and sxtM encoded two proteins with high sequence similarity to sodiumdriven multidrug and toxic compound extrusion (MATE) proteins of the NorM family. Members of the NorM family of MATE proteins are bacterial sodium-driven antiporters that export cationic substances [129]. All of the PSP toxins are cationic substances, except for the C toxins, which are zwitterionic. It is therefore probable that SxtF and SxtM are also involved in the export of PSP toxins.
Environmental factors such as nutrient (e.g., nitrogen and phosphate) content, salinity and temperature have been reported to regulate the production of PSP toxins in dinoflagellates and cyanobacteria [130][131][132]. Two transcriptional factors, sxtY and sxtZ, related to PhoU and OmpR, respectively, as well as a two-component regulator histidine kinase proximal to the 3' end of the sxt gene cluster in C. raciborskii T3 have been identified. PhoU-related proteins are negative regulators of phosphate uptake [133], whereas OmpR-like proteins are involved in the regulation of a variety of metabolisms, including nitrogen [134] and osmotic balance [135]. It is therefore likely that PSP toxin production in C. raciborskii T3 may be regulated at the transcriptional level in response to the availability of phosphate as well as other environmental factors. Following the identification and characterisation of the sxt cluster from C. raciborskii, Mihali et al. [136] have described similar gene clusters from an Australian isolate of Anabaena circinalis and an American isolate of Aphanizomenon sp. (Figure 7b and a, respectively). These saxitoxin gene clusters are slightly smaller than the C. raciborskii, spanning approximately 28 kb. The topology of all three sxt clusters is also varied which suggests the occurrence of multiple transposition events throughout the evolution of saxitoxin biosynthesis in the cyanobacteria. Phylogenetic analysis of the sxt O-carbamoyltransferase gene across several saxitoxin producing species indicated that the most likely origin of the gene was an ancestral a-proteobacterium and that the entire set of genes required for saxitoxin biosynthesis probably spread by horizontal gene transfer [137].

Introduction
The cyanobacterial alkaloid toxin, cylindrospermopsin, was first identified in 1979 when 148 people were hospitalized with symptoms of hepatoenteritis on Palm Island (Queensland, Australia). This outbreak was later linked to a bloom of Cylindrospermopsis raciborskii in a drinking water reservoir [138,139]. In addition to its impact on human health, cylindrospermopsin poisonings have been linked to the death of domestic animals [140].

Chemistry
Cylindrospermopsin is a polyketide-derived alkaloid with a central functional guanidino moiety and a hydroxymethyluracil attached to the tricyclic carbon skeleton [150] (Figure 9). The natural occurrence of an epimer at the hydroxyl bridge, 7-epicylindrospermopsin [151], and a cylindrospermopsin variant lacking the hydroxyl group at C7, 7-deoxycylindrospermopsin, have also been reported [143].

Biosynthesis and Genetics
The cylindrospermopsin biosynthesis (cyr) gene cluster from C. raciborskii AWT205 was recently sequenced [161] The cluster spans 43 kb and contains 15 ORFs, which encode all the functions required for the biosynthesis, regulation and export of the toxin (Figure 10). Biosynthesis is initiated via an amidinotransfer onto glycine followed by five polyketide extensions and subsequent reductions, rings are formed via Michael additions in a step-wise manner. The uracil ring is formed by a novel pyrimidine biosynthesis mechanism and tailoring reactions, including sulfation and hydroxylation that complete biosynthesis (Figure 11).  The first step in formation of the carbon skeleton of cylindrospermopsin involves the synthesis of guanidinoacetate via the transamidination of glycine [162][163][164]. An amidinotransferase encoded by cyrA, putatively transfers a guanidino group from arginine [163], to glycine thus forming guanidinoacetate. A mixed NRPS-PKS encoded by cyrB is thought to activate guanidinoacetate, which is then transferred via the swinging arm of the peptidyl carrier protein (PCP) to the KS domain. The AT domain of CyrB activates malonyl-CoA and attaches it to the ACP. This is followed by a condensation reaction between the activated guanidinoacetate and malonyl-CoA in the KS domain. The methyl transferase (MT) domain identified in CyrB is predicted to methylate C13. CyrB contains two reducing modules, KR and DH. Their concerted reaction reduces the keto group to a hydroxyl followed by elimination of H 2 O, resulting in a double bond between C13 and C14. A nucleophilic attack of the amidino group at N19 onto the newly formed double bond between C13 and C14 then putatively occurs via a 'Michael addition'. The cyclization follows Baldwin's rules for ring closure [165], resulting in the formation of the first ring in cylindrospermopsin. This reaction could be spontaneous and may not require enzymatic catalysis, as it is energetically favorable [165]. This is the first of three ring formations and is one of the principal differences between Mihali et al.'s [161] biosynthetic pathway and that previously proposed [163].
The third step in the biosynthesis of cylindrospermopsin involves cyrC, which encodes a PKS with KS, AT, KR, and ACP domains. The action of these domains results in the elongation of the growing chain by an acetate via activation of malonyl-CoA by the AT domain, its transfer to ACP and condensation at the KS domain with the product of CyrB. The elongated chain is bound to the ACP of CyrC and the KR domain reduces the keto group to a hydroxyl group on C12. Following the catalysis of enzyme CyrC is CyrD, a PKS. The action of this PKS module on the product of CyrC results in the addition of one acetate and the reduction of the keto group on C10 to a hydroxyl and dehydration to a double bond between C9 and C10. This double bond is the site of a nucleophilic attack by the amidino group N19 via another Michael addition that again follows Baldwin's rules of ring closure [165], resulting in the formation of the second ring, the first six-membered ring made in cylindrospermopsin. The intermediate produced by CyrD is the substrate for CyrE (step 5 in Figure 1). A PKS, CyrE, catalyzes the addition of one acetate and the formation of a double bond between C7 and C8. This double bond is attacked by N18 via a Michael addition and the third cyclisation occurs, resulting in the second 6-member ring. The cyrF gene encodes the final PKS module -a minimal PKS containing only a KS, AT, and ACP. CyrF acts on the product of CyrE and elongates the chain by an acetate, leaving C4 and C6 unreduced.
Step 7 in the pathway involves the formation of the uracil ring, a reaction that has been elusive so far and is required for the toxicity of the final cylindrospermopsin compound [151].
The cylindrospermopsin gene cluster encodes two enzymes CyrG and CyrH that are most similar to the enzyme family of amidohydrolases/ureases/dihydrotases, whose members catalyze the formation and cleavage of N-C bonds. Mihali and co-workers [161] propose that these enzymes transfer a second guanidino group from a donor molecule, such as arginine or urea, onto C6 and C4 of cylindrospermopsin resulting in the formation of the uracil ring. The first reaction consists of the formation of a covalent bond between the N of the guanidino donor and C6 of cylindrospermopsin followed by an elimination of H 2 O forming a double bond between C5 and C6. The second reaction catalyses the formation of a bond between the second N on the guanidino donor and C4 of cylindrospermopsin, co-committently with the breaking of the thioester bond between the acyl carrier protein of CyrF and cylindrospermopsin, causing the release of the molecule from the enzyme complex. The third reaction -if required -would catalyze the cleavage of the guanidino group from a donor molecule other than urea. The action of CyrG and CyrH in the formation of the uracil ring in cylindrospermopsin describes a novel biosynthesis pathway of a pyrimidine. Mihali and co-workers' [161] genetic analysis shows that cyclization may happen stepwise, with successive ring formation of the appropriate intermediate as it is synthesized. This mechanism also explains the lack of a thioesterase or cyclization domain, which are usually associated with NRPS/PKS modules and catalyze the release and cyclization of the final product from the enzyme complex.
The sulfation of cylindrospermopsin at C12 is likely to be carried out by the action of a sulfotransferase. The cyrJ gene encodes a protein that is most similar to human 3'-phosphoadenylyl sulfate (PAPS) dependent sulfotransferases. Similar enzymes have recently been implicated in the sulfation of other cyanotoxins [120]. The cylindrospermopsin gene cluster also encodes an adenylsulfate kinase (ASK), namely CyrN. ASKs are enzymes that catalyse the formation of PAPS, which is the sulfate donor for sulfotransferases. Mihali and co-workers [161] propose that CyrJ sulfates cylindrospermopsin at C12 while CyrN creates the pool of PAPS required for this reaction. Screening of cylindrospermopsin producing and non-producing strains revealed that the sulfotransferase genes were only present in cylindrospermopsin producing strains, further affirming the involvement of this entire cluster in the biosynthesis of cylindrospermopsin. The cyrJ gene might therefore be a good candidate for a toxin probe, as it is more unique than NRPS and PKS genes and would presumably have less cross-reactivity with other gene clusters containing these genes, which are common in cyanobacteria. The final tailoring reaction is carried out by CyrI. CyrI putatively catalyzes the hydroxylation of C7, a residue that, along with the uracil ring, seems to confer much of the toxicity of cylindrospermopsin [158,166].
The cylindrospermopsin gene cluster contains an ORF denoted cyrK, the product of which is most similar to sodium ion driven multi-drug and toxic compound extrusion proteins (MATE) of the NorM family. CyrK is hypothesized to function as a transporter for cylindrospermopsin, based on this homology and its central location in the cluster.
Cylindrospermopsin production has been shown to be highest when fixed nitrogen is eliminated from the growth media [167]. Flanking the cylindrospermopsin gene cluster are "hyp" gene homologs involved in the maturation of hydrogenases. In the cyanobacterium Nostoc PCC73102 they are under the regulation of the global nitrogen regulator NtcA, that activates transcription of nitrogen assimilation genes [168,169]. It is plausible that the cylindrospermopsin gene cluster is under the same regulation, as it is located wholly within the "hyp" gene cluster in C. raciborskii AWT205, and no obvious promoter region in the cylindrospermopsin gene cluster could be identified. Finally, the cylindrospermopsin cluster also includes an ORF at its 3'-end designated CyrO. By homology, it encodes a hypothetical protein that appears to possess an ATP binding cassette, and is similar to WD repeat proteins, which have diverse regulatory and signal transduction roles. CyrO may also have a role in transcriptional regulation and DNA binding. It also shows homology to AAA family proteins that often perform chaperone-like functions and assist in the assembly, operation, or disassembly of protein complexes. Further insights into the role of CyrO are hindered due to low sequence homology with other proteins in databases.