The actual number of different substances that comprise the “prymnesins” is not presently known, but their broad range of biological activities support the notion that extracts from cells and from cell-free supernatants are composed of a complex and diverse mixture of toxic metabolites [30–32,44–47]. For example, Ulitzur and Shilo [48] isolated 6 hemolytic fractions that were potentially chemically distinct from a hemolysin extracted by Kozakai et al. [49] and the toxic compound isolated by Paster [39]. Discrepancies between published investigations are likely due to the techniques used for the extraction of prymnesins, resulting in the isolation of unique sets of compounds that separate differently based on their affinity for different solvent(s). Therefore, caution must be exercised when interpreting data and comparing studies. Variations in toxicity reported for different extracts may well be attributed to the extraction methods resulting in extracts containing only a subset of all the possible toxic substances and activities.

Prymnesins have been characterized as glycolipids, galactolipids, proteolipids and lipid-carbohydrate compounds [32,39,48–50]. Crude extractions have been performed with both whole cells and culture filtrates of P. parvum and many different approaches have been taken to separate the toxic principles. All of these methods are founded on the differential extraction of prymnesins using an organic mobile phase [32,47]. As a rule, the toxic fractions have been found to be the most soluble in polar organic solvents, and they demonstrate low affinity (and extremely low solubility) in highly non-polar organics. Toxins can be extracted, albeit very slowly, from cells using aqueous solutions; however, extracts from cells are greater than three times more concentrated when extracted with an organic solvent [32]. Shilo and Rosenberger [46] reported that prymnesins were not retained by either anion or cation exchange resins. However, Yariv and Hestrin [31] were able to obtain a concentrated sample of prymnesins by passing culture filtrates over precipitated Mg(OH)₂, which suggests that some prymnesins demonstrate an affinity for this basic adsorbant material.

Regarding the extraction of prymnesins from whole cells, cells are typically washed several times with cold acetone [34,35,39,46,48]. Shilo and Rosenberger [46] found that prymnesins are not soluble in acetone, which makes this an ideal organic solvent to eliminate interfering compounds including chlorophylls and accessory pigments. Afterwards, polar organics such as methanol, ethanol and/or n-propanol are usually employed to extract prymnesins from the cell remains [34,35,39,46,51]. In some cases, multiple rounds of precipitation (e.g., using diethyl ether) are utilized to separate the toxic fractions; further purification of the target compounds can be achieved using column chromatography and/or additional rounds of partitioning between biphasic solvents (e.g., n-butanol) [32,39]. Prymnesin extracts have also been prepared from whole algal cell suspensions by liquid-liquid chromatography using a CHCl₃:MeOH phase system, wherein the toxins are extracted into the lower phase that consists of a suspension of CHCl₃/MeOH [48,52,53]. It is reported that prymnesins are most stable when they are stored at cooler temperatures as a dry powder; similarly, these extracts are more soluble in organic solvents than in water, and are less labile under acidic conditions near pH 5 [32,54].

It wasn’t until the mid-1990’s that we had the first glimpse of what two of these compounds actually look like. Prymnesin-1 (prym1) and prymnesin-2 (prym2) were meticulously isolated from whole algal cells by a group in Japan and these were the first toxic metabolites to be chemically characterized from any isolate of P. parvum using modern analytical methods (Figure 3) [34,35]. Purified prym1 and prym2 were shown to be potent polyketides with ichthyotoxic and hemolytic activities when present at nanomolar concentrations [34,35]. Prym1 and prym2 are ladder-like, polycyclic ethers that posses several noteworthy features. These complex compounds are remarkable in that they have double and triple carbon-carbon bonds in the unsaturated head and tail regions, an amino group, several chlorines, four 1,6-dioxadecalin units, and an assortment of sugar moieties [34,35]. Structurally, prym1 and prym2 are homologous compounds with a common head and backbone: they differ only in the number and in the type of sugars in the tail region, with prym2 containing a rare L-xylose. Prym1 was shown to be slightly more polar (due to the additional sugar residues) and it elutes ahead of prym2 in reversed-phase C-18 chromatography [34,35].

The structural elucidation of prym1 and prym2 has been especially challenging due to their poor solubility in the deuterated organic solvents used with analytical methods such as proton nuclear magnetic resonance (NMR). However, chemical derivatives (e.g., N-acetylation) did demonstrate improved solubility and also provided a quantifiable shift in the mass/charge (m/z) for negative ion fast atom bombardment-mass spectrometry (FAB-MS) spectra [34]. However, chemical structures that have been elucidated from spectral analyses are often inaccurate and reassignments are common due to the complexity of many secondary metabolites [55]. The backbone of prym1, comprised of rings A-N, was characterized using a computational approach [56]. Partial (in some cases, total) synthesis of compounds is also useful to reproduce structures and to determine the correct conformations. Recently, investigators have synthesized truncated ring models of prym1 and prym2. This work has led to the verification of the HI/JK rings, the conformation of the chiral centers at C14 and C85, and the reassignment of the juncture between rings E-F (Figure 3) [36,37].

Very little is known about the synthesis of prymnesins in vivo. However, it is very likely that prym1 and prym2 are derived from acetate-related metabolism based on what we know about their structures. In other organisms, the acetate pathway is responsible for the generation of fatty acids, as well as numerous secondary products that include polyketides and non-ribosomal peptides (Figure 4) [57,58]. Both primary and secondary metabolites synthesized in this pathway utilize 2-, 3-, and 4-carbon basic skeletal types coupled to coenzyme-A (CoA) [58,59]. Subunits are subsequently combined by successive decarboxylative condensations resulting in chain elongation.

The biosynthesis of polyketides has many features in common with fatty acid synthesis and it uses similar, perhaps some of the same, enzymatic domains. Cane and Walsh [57] described polyketide synthase (PKS) as a modular “megaprotein” possessing a core catalytic domain and auxiliary (carrier) domains. The core domain contains the ketosynthase (KS), acyltransferase (AT) and the acyl carrier protein (ACP)—to which the growing polyketide molecule is tethered via an acyl-S-enzyme of the 20 Å long phosphopantetheine arm. The AT domain catalyzes priming of the donor ACP with monomers, and the carbon-carbon bond-forming step is performed by the KS domain. The auxiliary domains encode for ketoreductase (KR), dehydratase (DH), and enoyl reductase (ER). The growing polyketide is handed over from one reactive thiol group to the next by transacetylation reactions. This process is repeated until the ultimate chain length has been achieved, whereupon thioesterase (TE) facilitates the removal of the polyketide by hydrolysis at the distal end of the pantetheine prosthetic group. The individual enzymes function to modify the elongating molecule before adding the next subunit. Chain elongation ceases when all of the “programmed” subunits are incorporated, resulting in a single-chain carbon backbone. Some PKSs also contain enzymes for cyclization (CYC) in addition to other secondary modifications (e.g., methylation) [58,59]. Note that cyclization may occur prior to or after the release of the polyketide from TE.

A minimal PKS operon contains only the gene for KS, whereas diverse PKSs may contain all, parts or none of the respective domain sequences [59,60]. Structurally, prym1 and prym2 resemble other heterocyclic algal toxins produced by Type I PKSs, such as okadaic acid produced by marine diatoms, and brevetoxin and maitotoxin synthesized by the dinoflagellates Karenia brevis and Gambierdiscus toxicus, respectively [61,62]. These polyether polyketides are non-aromatic, which is characteristic of the Type I modular (non-iterative) PKSs found in bacteria, fungi, animals and some monocotyledonous plants. There is still some debate as to whether PKS associated with dinoflagellates is actually of bacterial origin [63]; in fact, the Type I PKS genes attributed to Prorocentrum have been isolated from associated bacteria (Roseobacter) [64]. Nevertheless, Type I PKS genes were recently isolated and sequenced from axenic cultures of P. parvum [41] and the related haptophyte Emiliania huxleyi (Haptophyta) [65], as well as from the green alga Chlamydomonas reinhardtii (Chlorophyta) [65].

In general, Type I PKS genes encode multiple catalytic domains organized into modules and each complete subunit may be composed of 2 or more modules (Figure 5) [64]. Each module contains all functional domains required for exactly one round of chain extension and modification. The complexity and diversity among polyketides and PKSs are a result of mutation, multiple gene duplications, deletions, and horizontal gene transfer [59]. We postulate that these modules might also potentially operate independently of one another in P. parvum to generate a diverse array of smaller, yet related polyether molecules, as was demonstrated in mutational analyses using a bacterial PKS [66].

Type I PKS enzymes are analogous to the Type I FAS, and polyketide elongation is performed on an enzymatic framework prearranged into open reading frame modules encoding the mulitmeric protein complex (Figure 6). Because of this, it is likely that polyketide synthesis occurs in the stroma of plastids, which is known to be the case for FAS in green algae [67]. It should be noted that these mechanisms have not yet been elucidated for P. parvum. Glycosylation is a common modification of secondary metabolites, and this might be a mechanism for tagging prym1 and prym2 for transport. This process changes the physical and chemical properties of the compounds, their bioactivity and their sub-cellular localization [68]. Post-synthesis modifications (such as glycosylation) probably occur at the Golgi apparatus or endoplasmic reticulum, where many different molecules are tagged and packaged for transport [69]. More information regarding cellular location of polyketide synthesis, storage and transport would be of great importance in determining the biological significance of prymnesins to the alga (see later).

Until recently, there were very few gene sequences available for P. parvum. An expressed sequence tag analysis of P. parvum identified a number of gene products that may be involved with acetate metabolism and polyketide synthesis [41]. Interestingly, these transcripts have a high sequence similarity to bacterial Type I PKS genes. More specifically, individual Type I PKS genes in P. parvum have their nucleotide sequence identities most similar to Roseovarius (a bacterium) and Nostoc (a cyanobacterium), and one PKS subunit is most closely related to that of a eukaryotic stramenopile [41]. In addition, there are at least 43 different unique genes whose products have been classified as being involved with secondary metabolite synthesis, catabolism and transport, including ACP synthase, oligoketide cyclase, and DH that are potentially relevant to polyketide synthesis. Another 40 unigenes are classified as encoding proteins involved with lipid synthesis and transport, which could also have parallel roles in the manufacture of polyketides. Finally, there are a number of genes expressing vesicle transport-related proteins. Interestingly, the most abundantly expressed gene transcripts have no known sequence homology [41]. More work is needed in this area to determine which of these genes’ products directly participate in the synthesis or transport of prym1, prym2 and other toxic metabolites of P. parvum.

The vitamins thiamine (B₁) and cobalamin (B₁₂) are required in culture media and are considered essential for the growth of most known P. parvum strains [22,32]. Several genes have been elucidated whose products are related to thiamine and cobalamin metabolism [41]. On the other hand, biotin (B₇) and pantothenate (B₅) have not been shown to be essential for growth, yet both vitamins are prosthetic groups essential to the synthesis of primary and secondary metabolites. Acetyl-CoA carboxylase is a biotin-dependent enzyme in the acetate pathway and an essential component in the synthesis of fatty acids and polyketides. Similarly, pantothenate is required for the production of CoA, and in turn ACP. EST analyses indicate that P. parvum may possess several genes encoding enzymes involved in these B-vitamin synthetic pathways, and it is therefore likely that P. parvum is able to synthesize biotin and pantothenic acid de novo [41].

P. parvum can thrive in a wide range of physical conditions, but nutrient availability has been shown to greatly influence HAB and toxin formation in this organism. Agricultural effluents and eutrophication of waters are commonly implicated in a growth surge in P. parvum populations [11,91,92]. High nitrogen (in the form of nitrate or nitrite) and phosphorus (phosphate) concentrations support the rapid growth and formation of blooms that subsequently result in an imbalance of N and P sources. These limiting conditions ultimately slow the growth of P. parvum, which is usually accompanied by an increase in toxicity (as noted later) [5,30,93]. Mesocosm experiments over the course of five years indicated a reduction in toxicity in nutrient-replete conditions [43]. These observations suggest that extracellular prymnesins decrease appreciably when growth and nutrient conditions are favorable; therefore, it has been proposed that prymnesin toxicity can be controlled by nutrient manipulation [94]. However, a decrease in exotoxins may also be concomitant with an increase in intracellular toxins (see later).

Phosphorus and (to a lesser degree) nitrogen availability are probably among the most widely-studied factors related to prymnesin toxicity. Shilo [74] was the first to demonstrate elevated toxicity of P. parvum in association with low phosphorus conditions, and this trend has been confirmed in subsequent studies [5,77,82,85,95]. Shilo [30] reported a 10- to 20-fold increase in the amount of intracellular prymnesins when grown in P-limited conditions; a rise in toxin concentration was also documented for culture supernatants. Johansson and Granéli [95] analyzed the effects of N and P levels on cell density, chemical composition and toxicity of P. parvum, being the first to show increased toxicity under N-limiting conditions. Hemolytic activity was observed regardless of the nutrient conditions, so they proposed that cell stress (resulting in lysis and death?) was the cause for increased toxicity rather than the direct involvement of either N or P in toxin synthesis per se [96]. This is also in agreement with analyses confirming increased toxicity in senescing/dying cultures of P. parvum (see later).

The addition of a carbon source has also been shown to augment toxicity in populations of P. parvum. Dark-incubated cultures were able to grow in glycerol-supplemented media [84], and the addition of glycerol has been demonstrated to cause an increase in toxicity [97]. We postulate that this increased toxicity might be correlated to a surge in glycolytic activity from the exogenous carbon source. This would provide precursor molecules for the acetate pathway that might be used for polyketide synthesis. Conversely, glycerol and glucose have been shown to interfere with the formation of secondary metabolites in some organisms [86]. It is obvious that more information is needed to determine the roles of various nutrients in prymnesin toxicity and possible inhibitory pathways.

Beyond being photosynthetic, P. parvum is also known to phagocytose bacteria and other photosynthetic algae, which may provide exogenous N, P and carbon sources in times of nutrient limitation [98,99]. Legrand et al. [94] demonstrated that P. parvum f. patelliferum utilizes various sources of phosphate including inorganic, dissolved, organic and particulate forms; similarly, P. parvum has been shown to utilize ammonia, methionine and ethionine as alternative N-sources [22,100]. Recent findings by Carvalho and Granéli [101] indicate that P. parvum is truly mixotrophic and therefore feeds heterotrophically under both nutrient-limiting conditions and in nutrient-replete cultures. Interestingly, bacterial suspensions of Proteus vulgaris and Bacillus subtilis decreased toxicity by 50% in one hour [83], suggesting that these bacteria may be capable of metabolizing at least some of the toxins [102].

Due to the predominance of blooms in low salinity and brackish water systems, many studies have examined the effects of salinity on prymnesin toxicity. The broad osmotic tolerances of P. parvum have been well-documented [5,22,97,103]. In most cases, P. parvum has been recorded in highly-mineralized freshwaters between 3 and 8 ppt (e.g., [28]); some assert that P. parvum can grow almost anywhere, in salinities ranging from 1 to 45 ppt [22,40,103]. However, low salinity conditions have been shown to have a negative effect on growth rates of P. parvum. Baker et al. [104] illustrated that there is a decrease in growth rate concurrent with a decrease in salinity, even when grown in nutrient-sufficient conditions. Interestingly, Dickson and Kirst [27] postulated that the broad osmotic tolerance of P. parvum is related to their ability to synthesize a range of compatible solutes coupled to a partial exclusion of toxic ions. The fact that gene transcripts encoding various ion transporters were among the most highly-expressed ones in log-phase cultures of P. parvum, may also help to explain how this alga is able to handle such a broad range of salinities [41].

Specific ions may play a role in prymnesin toxicity, but these assessments are limited and conflicting data have been reported. Paster [32] noted that the optimum NaCl concentration for growth and toxin production in P. parvum is between 0.3–5% (=3–50 ppt) (see also [1]) Salt begins to inhibit toxicity at levels greater that 5–12% NaCl [32]. This is not surprising as this represents salt concentrations of 50–120 ppt, exceeding two to four times the salinity of open ocean seawater. Paster [32] found elevated ichthyotoxicity for cultures grown in 5% (0.75 ppt) artificial seawater (ASW that contained ~15 g/L salts) and the lowest toxicities are documented for cultures grown in 30% (4.5 ppt) ASW. Interestingly, there appeared to be an absence of ichthyotoxicity in salinities >30% seawater. However, a lack of extracellular bioactivity is not directly related to intracellular toxin production, as there are still large quantities of toxins able to be extracted from the cells themselves [32,76,105]. Shilo and Rosenberger [46] reported that the hemolytic property of prymnesins was inversely related to salinity with the highest hemolysis observed for cultures grown in 1% salt (=10 ppt). In contrast, Larsen and Bryant [40] found no significant differences in toxicity for P. parvum when grown at different salinities, which they felt could indicate metabolic differences among different strains of the alga. Since a decrease in toxicity in the culture media was reported to coincide with an increase in intracellular toxins [76], we conclude that both the cells and the supernatants need to be carefully analyzed in order to determine any relationships between salinity and toxicity.

Light is a factor that is often associated with prymnesin toxicity, but published observations are somewhat inconsistent. P. parvum has been documented as naturally occurring in moderate-to-low light conditions, but some isolates have shown a preference for (and grow better with) higher illumination [40,106,107]. However, where P. parvum actually exists in the environment may be somewhat dependent on local community structure, which may also play a role in the stratification and distribution of P. parvum in the water column.

Many studies have shown that both light quality and light intensity are physical properties that are essential for the synthesis and activity of prymnesins [28,47,74]. On the other hand, some have concluded that light is not necessary for toxin production in P. parvum [108]. Dark-grown cultures reportedly have a reduced amount of intracellular toxins, which supports the latter claim; but an increase in toxin concentration was detected later in the stationary phase of culture [89]. This finding is in agreement with growth studies that observed differences in toxicity and it suggests that toxic compounds are accumulating in the cells steadily throughout the exponential growth curve (see later). Since P. parvum is mixotrophic and can feed in the absence of light, these accounts suggest that light may not be necessary as long as there is sufficient food to provide carbon skeletons for prymnesin synthesis.

Interestingly, cultures that were grown under constant illumination demonstrated no ichthyotoxic activity, but did retain the hemolytic activity [106]. It was postulated that the ichthyotoxic principles may either be inactivated or they may degrade more rapidly than the compounds responsible for hemolysis [45]; alternatively, it is possible that multiple factors are responsible for hemolysis, some of which remain to be characterized. In accordance, exotoxins were inactivated after 90 min when treated with UV and visible wavelengths of light (255 nm and 400–520 nm, respectively) [107,109]; however, intracellular toxins were not affected, suggesting that these compounds are being protected by some mechanism or they may be stored in vesicles [97].

P. parvum is eurythermal, being capable of growing at temperatures ranging from 2–30 °C [22,28,39,83]. This has made it difficult to determine whether ambient temperature plays a significant role in prymnesin presence or toxicity. Globally, recurring HAB of P. parvum do not always occur at the same time of the year. Blooms in the southern United States appear throughout the fall and winter months [26]; in China, blooms form in the late spring and early fall [28]; and a summer bloom in Finland was characterized as having water surface temperatures around 20–25° C [53].

Similarly, optimum temperatures for growth and toxicity vary greatly among different geographic isolates of the alga. The optimum temperature of a strain from Denmark was around 26 °C, and growth was limited by temperatures near 10 °C [11]. In contrast, Larsen and Bryant [40] found that most (but not all) of the strains they studied had maximum growth rates at 15 °C. Given the broad temperature tolerances of P. parvum, there are no clear-cut correlations between specific temperatures and toxicity [30].

The various modes of prymnesin toxicity are also restricted to different pH ranges, with hemolysis and ichthyotoxicity highest during acidic and alkaline pH conditions, respectively. Prymnesins are also acid- and base-labile, which means that they are only functional under certain pH conditions [52]. Shilo and Rosenberger [46] demonstrated that an increase in pH to alkaline levels eliminated the hemolytic activity, but did not appear to affect ichthyotoxicity. In fact, these authors state that toxicity in fish requires a pH > 7.0, which is in agreement with the observations that toxic P. parvum occurs in Chinese waters ranging from pH 7.2 to 9.3 [28]. Ulitzur and Shilo [88] reported that activation of the ichthyotoxic property is dependent on cationic activation at a pH greater than 8.0 with maximum toxicity observed at pH 9.0 (see below). Conversely, hemolysis can occur in conditions as low as pH 5, with decreasing hemolytic activity with increasing alkalinity. These studies show that pH can greatly affect the range, nature and intensity of prymnesins’ toxic properties, which also suggests that numerous (and probably some unknown) compounds may be at work here.

The primary reason why the presence of prymnesins and actual ichthyotoxicity are not necessarily related is because co-factors are required to “activate” prymnesins [5,30,31,85]. Furthermore, the type of activator determines the extent and severity of the toxic effects [89]. Toxicity to fish is greatly enhanced by the presence of Na⁺, Ca²⁺, and Mg²⁺, as well as by antibiotics (e.g., streptomycin) and polyamines (e.g., spermine) [1,30,88,89,105]. Shilo and Rosenberger [46] established that divalent cations were more effective than monovalent ones in promoting ichthyotoxicity. These ion interactions may play a role in the aggregation, complexation and structural conformation of prymnesin molecules (as are common with polyether ionophores, e.g., the monensin-Na⁺ complex [110]), which may in turn affect the ability of prymnesins to interact with specific components of the plasma membrane [1,70].

Toxicity in this alga and in culture filtrates also varies with growth phase, i.e., lag phase, exponential phase and stationary phase cultures, in addition to the abiotic and community interactions mentioned [30,83]. While Paster [32] found a significant positive correlation between the concentration of P. parvum cells and the degree of exotoxins, some have suggested that factors necessary for growth and toxicity are regulated by different parameters [74,111]. In fact, Reich and Aschner [10] noted that concentration of prymnesins in the surrounding media is not always related to the intracellular concentrations of prymnesins. Extensive mortalities have been observed with low cell concentrations, and dense blooms of P. parvum have been documented as having no toxic effects [74,83]. Guo et al. [28] confirmed that the appearance of extracellular toxins from P. parvum has been shown to be independent of cell density, with the two parameters having no clearly-defined relationship.

A reduction in ichthyotoxicity is typically observed during periods of rapid growth (e.g., exponential phase) and the effects of prymnesins are most obvious when populations of P. parvum reach limiting conditions (e.g., stationary phase) [28]. Similarly, Houdan et al. [4] reported that P. parvum cultures were less toxic to cultures of the brachiopod crustacean Artemia while in the exponential, growing phase. During the stationary phase (which happens to correspond to the highest cell concentrations and limiting conditions), they found complete mortality of Artemia nauplii. This is not surprising since secondary metabolism increases during periods when balanced growth has ceased [87]. This is often the case with eutrophication leading to surge in algal growth, which is subsequently followed by nutrient imbalance and limiting conditions (e.g., [93,95]).

Berman [111] followed the total endo- and exotoxin content in growing cultures. In this study, it was noted that the concentrations of both hemolytic and ichthyotoxic compounds increased with culture age on a per-cell basis, implying that these compounds might be synthesized constitutively and accumulate as a result. In agreement, Paster [32] reported that prymnesins are synthesized throughout the alga’s life cycle, and are only released after cell death and disintegration of the cell components. This also lends more support to prymnesins serving a possible intrinsic function, because it would seem unlikely that cells would store waste products long-term (see later). Superficially, it would appear that there is a difference in the presence of endo- and exotoxins, but more information is needed regarding the mechanism of release to make further conclusions on the roles of these compounds.

Differences among published studies could be attributed to the methods of extraction, how cells and cell-free supernatants were being compared, how toxicity was quantified, the presence of cofactors, examining cultured vs. natural samples, differing geographic isolates, etc.