The toxins active on voltage-gated sodium channels are abundant in all sea anemone venoms assayed to date, and are present even in rare and highly unique species (Ishida et al., 1997; Moran and Gurevitz, 2006). Their abundance of these toxins and the fact that these toxins they constitute a major fraction of the proteinaceous content of the venom (Beress et al., 1975) indicate that they have a point for their major role in predation and defense.

The initial release of the genome sequence draft of the starlet anemone, N. vectensis (Sullivan et al., 2006) provided the first opportunity to explore the toxin-related genomic landscape of a cnidarian. Fourteen putative genes were identified by homology searches, but unexpectedly they were all found to encode the same mature toxin, Nv1, which highly resembles Type I and Type II sea anemone toxins (Moran and Gurevitz, 2006). The Nv1 gene structure is similar to that of the C. parasitica toxin genes, with two exons separated by an intron. Astoundingly, the gene region that encodes the mature toxin is devoid of any synonymous substitutions (codon substitutions that do not change the amino acid sequence) and the genes in general are strikingly similar (identity >97.5%). This is in sharp contrast to the previously described heterogeneity of animal peptide neurotoxins (Nakashima et al., 1993; Duda and Palumbi, 1999; Froy et al., 1999).

The official release of the more complete genome sequence of N. vectensis (Putnam et al., 2007) was useful in exploring the genomic organization of the Nv1 genes. At least eight Nv1 genes are located in a narrow genomic region of ~30,000 bp at the same transcriptional orientation (Moran et al., 2008a; Fig. 4A). The adjacent genomic regions of all genes are also highly similar (Fig. 4B; Identity >92.5% for at least 1000 bp). This unusual nucleotide conservation is reminiscent of the genes encoding ribosomal RNA (rDNA). The rDNA genes were suggested to evolve in a rare evolutionary mode known as ‘concerted evolution’ (Brown et al., 1972). This suggestion has been proven in yeast to be based on unequal crossover (Szostak and Wu, 1980). Due to their concerted evolution the rDNA copies and their surrounding regions in one species are all more similar to one another than to rDNA genes of other species. Indeed, when the transcripts and genes of Type I toxins from A. viridis and Actinia equina were amplified, the deduced toxins of each species were highly similar to one another, but less similar to those of other species (Fig. 5). A remarkable example was found in A. viridis, where six at least seven genes encode Av2, with noticeable insertions/deletions (indels) in their introns, but still with highly similar (>95%) exons (Moran et al., 2008a).

The selective advantage of concerted evolution of toxin genes may relate to a 'dosage' effect of gene expression. This suggestion is supported by a strong statistical positive correlation found between gene expression level and concerted evolution of duplicated genes in yeast (Sugino and Innan, 2006). Unlike many other poisonous animals, sea anemones lack a venom gland or a stinging organ. Instead they use for venom delivery and use instead nematocytes, spread all over their body, distributed throughout the integumentary tissues (Kass-Simon and Scappaticci, 2002). Since these cells and since these cells are replaced after discharge are able to discharge toxins (via their discharging nematocysts; see Fautin paper in this issue) into a victim only once, and must be replaced, swift production of copious toxin in young nematocytes is seemingly advantageous. and the Such production of toxin may be accelerated by the presence of multiple copies of the same toxin gene. While this strategy for large-scale toxin production may appear primitive or ineffective compared to the acquisition of very strong promoters and enhancers, concerted evolution may also be useful for the rapid “transmission” of advantageous mutations from a single toxin gene locus to the other loci or in preventing the loss of a highly effective toxin.

In addition to the multiple genes encoding highly similar toxins in A. viridis and A. equina, genes encoding novel putative toxins were also identified in these species (Moran et al., 2008a). These genes evolved in an adaptive manner under positive Darwinian selection (‘accelerated evolution’) as shown for toxin genes of other poisonous animals (Nakashima et al., 1993, Duda and Palumbi, 1999; Zhu et al., 2004; Moran et al., 2008a). This mode of evolution can be highly advantageous as the evolution of pharmacological targets of neurotoxins might have been enhanced under the selective pressure exercised by the neurotoxins, resulting in an evolutionary “arms race”. Accelerated evolution driven by positive Darwinian selection and gene variability seems to contradict the proposed mechanism of concerted evolution. Still, variability is also found in rDNA, and several cases of rDNA that “escaped” concerted evolution were documented (Carranza et al., 1999; Keller et al., 2006). Accordingly, in A. viridis and A. equina it is possible that some toxin genes "escaped" the concerted process, diverged rapidly and were fixed depending on the selective value and neutral drift of these loci.

Comparison of Av2 effects at the Drosophila melanogaster sodium channel DmNa_v1 (Para) and rNa_v1.2a (rat brain channel subtype) expressed in Xenopus laevis oocytes has shown clear preference of Av2 for the insect sodium channel (Warmke et al., 1997). However, comparison of Av2 action on DmNa_v1 and rNa_v1.5 clearly indicated that this toxin is equally active at both channels (Fig. 6A, B). A sharp difference in selectivity was observed when the effects of Bg2, a toxin of Bunodosoma granulifera, were evaluated at insect versus mammalian Na_vs expressed in Xenopus oocytes (Bosmans et al., 2002). Bg2 affected DmNa_v1 in concentrations of 5 nM, but affected the mammalian Na_vs only in 50-fold higher concentrations. However, this reported insect-selectivity of Bg2 is in contrast to previous reports about anti-mammalian activity of this toxin when examined in binding competition assays against the scorpion α-toxin Aah2 (of Androctonus australis hector) at rat brain synaptosomes, and in electrophysiological assays performed on rat neurons (Loret et al., 1994; Salceda et al., 2002).

The profound activity of sea anemone toxins on insects is puzzling considering the fact that the two groups of organisms live in separate habitats. Still, taken that insects and crustaceans are both arthropods suggests that their ion channels are similar. This assumption is corroborated at the bioinformatic level when the sequence encoding Na_v derived from the initial genome draft of the crustacean Daphnia pulex (genome draft available at http://genome.jgi-psf.org/Dappu1/Dappu1.home.html) is compared to that of DmNa_v1 (GenBank accession P35500). Moreover, although most insects are terrestrial, in their larval stage some of them could encounter sea anemones in shallow water. Indeed, N. vectensis was reported to feed on insect larvae (Frank and Bleakney, 1978). At present, no Type I sea anemone toxin that is truly invertebrate selective has been reported, although recent studies have raised Av1 of A. viridis and Nv1 of N. vectensis as two promising candidates for invertebrate-selective Type I sea anemone toxins: Av1 of A. viridis and Nv1 of N. vectensis (Moran et al., 2008 and Unpublished data).

Although the study of Type II sea anemone toxins has not been as extensive as that of Type I toxins, some of them exhibit high preference for invertebrates. While Rp4 of H. paumotensis is more active on mice than on crabs, Rp2 from the same anemone is 280-fold more active on crabs than on mice (Schweitz et al., 1985). Electrophysiological analysis of the toxin Sh1 of Stychodactyla helianthus on crayfish and cockroach axons, as well as its injection to crabs, cockroaches and mice has shown no effect on the injected mice, a much stronger toxicity on crabs compared to cockroaches, and higher activity at the crayfish over cockroach axons (Salgado and Kem, 1992). In this study, Calliactis toxin was also shown to have an almost identical action on sodium channels. Evidently, The Type II toxins still await further characterization of toxin effects, especially on heterologously expressed channels, may contribute to the knowledge about selectivity of Type II sea anemone toxins.

The Type I sea anemone toxin that was most extensively studied during the last decade was ApB (Gallagher and Blumenthal, 1992). With the advent of its recombinant production using an Escherichia coli expression system, Blumenthal and his colleagues were able to modify the toxin residue by residue and to assess the effect of these substitutions on bioactivity. All ApB mutants were assayed either on cardiac rNa_v1.5 and brain rNa_v1.2a channels transiently expressed in HEK cells, or mammalian cell lines that constitutively express these channels among a minority of other channel subtypes. These studies have suggested a bioactive role for Arg-12, Asn-16, Leu-18, Ser-19, Trp-33, Lys-37 and Lys-49, of which most bioactive residues were associated with the Arg-14 loop and with the substitutions of Leu-18 having the strongest effect on ApB activity (Blumenthal and Seibert, 2003; Khera and Blumenthal, 1996; Dias-Kadambi et al., 1996a; Dias-Kadambi et al., 1996b; Kelso et al., 1996; Seibert et al., 2004; Fig. 2 and Fig. 7). In contrast to the results obtained by chemical modifications of Av2 (Barhanin et al., 1981), this mutagenic analysis has indicated that Arg-14, His-34 and His-39 have no bioactive role (Khera and Blumenthal, 1994, 1996).

Recombinant expression of novel neurotoxin transcripts from Anthopleura of unknown species and activity assays using the contractile force of isolated rat atria, suggested that residues at sites 14, 22 and 25 might have a bioactive role (Wang et al., 2004). Nevertheless, these results await further validation using electrophysiological, binding and toxicity assays more common in the analysis of toxins (Blumenthal and Seibert, 2003; Moran et al., 2006).

Functional expression of Av2 enabled thorough alanine scanning and elucidation of its bioactive surface toward insects (Moran et al., 2006). Analysis of all mutants in toxicity to blowfly larvae and binding assays on cockroach neuronal membranes revealed a functional role for Val-2, Leu-5, Asn-16, Leu-18 and Ile-41 (Fig. 2 and Fig. 7). Further electrophysiological analysis of these mutants against at hNa_v1.5, indicated a bioactive surface toward the mammalian channel similar to that toward insect channels with the exception of Ser-19, whose substitution affected only the activity only at the cardiac channel. Av2 and ApB are 79% identical in sequence and excluding Arg-12 and Lys-49, all other residues shown to be important in Av2 appear also in ApB (Fig. 2). However, a major variation between both toxins is the functional role proposed for Trp-33 and Lys-37 in ApB, whereas substitution of their Trp-31 and Lys-35 equivalents in Av2 had no effect on the insecticidal activity, and affected only slightly the activity at Na_v1.5 (Moran et al., 2006). Moreover, of the four functional aliphatic residues on the Av2 surface, only Leu-18 was ascribed to the anti-mammalian bioactive surface of ApB (Dias-Kadambi et al., 1996b). These glaring differences suggest that despite some commonality in the bioactive surface of Av2 and ApB, the two toxins may interact differently with receptor site-3 on insect and mammalian Na_vs. This assumption is corroborated by findings of Benzinger and co-workers (1997) that have suggested about differences in the binding sites of ApA and ApB on rat skeletal muscle and cardiac Na_vs.

Mutagenesis of the glycine residues spanning the flexible Arg-14 loop severely hampered the ability of ApB and Av2 to affect Na_vs (Seibert et al., 2003; Moran et al., 2006). On the basis of these results and the fact that glycines usually are found in protein sequences at positions that may enlarge structural flexibility, it was proposed that the freedom of the Arg-14 loop to tilt is an important feature for toxin interaction with the receptor site. Another residue whose substitution by alanine affects toxin action is the conserved Asp-9 (Khera and Blumenthal, 1996; Moran et al., 2006; Fig. 2). Because substitution D9N had only little effect on toxicity, the carboxyl group of Asp-9 is not critical for Av2 and ApB activity. As may be inferred from the NMR structures of ApA and ApB, a hydrogen bond exists between the carbonyl group at the side chain of Asp-9 and the amine of Cys-6 (Pallaghy et al., 1995; Monks et al., 1995). Theoretically this bond should be maintained when the aspartate is substituted by asparagine. Khera and Blumenthal (1996) reported that Asp-9 is critical for proper folding of recombinant ApB on the basis of the prominent decline in polypeptide yield when mutated at this position. However, the same phenomenon was not observed in Asp-9 mutants of Av2 (Moran et al., 2006), suggesting that the role of the hydrogen bond formed by the carbonyl of Asp-9 with Cys-6 is to limit the free tilting of the Arg-14 loop. Assuming that the binding of sea anemone toxins to receptor site-3 includes a step of induced fit (Koshland, 1958), which involves the Arg-14 loop, it is possible that its flexibility enables adoption of a shape necessary for binding. Yet, it cannot be ruled out that the carbonyl of Asp-9 simply interacts with the Na_v receptor site.

Although Type II sea anemone toxins have been studied at the physiological level, none of them was analyzed using a molecular approach. their study at the molecular level was limited and focused on the charged residues in Sh-I of Stichodactyla helianthus (Pennington et al. 1990). It was shown that substitutions of Lys-4, Asp-6, Asp-7, Glu-8 and Asp-11 impaired severely the activity of Sh-I with no effect on its structure, suggesting that the bioactive surface of this toxin is highly charged. Although diverged from a putative common ancestor (Ishida et al., 1996; Moran and Gurevitz, 2006), sequence comparison of Type II toxins with Type I toxins reveals less than 50% identity, and the bioactive residues in both toxin types differ determined in Type I to be involved in bioactivity are not conserved in Type II toxins (Fig. 2). Despite the difference, both toxin types exert similar effects on a crustacean axon (Rathmayer et al., 1976; Salgado and Kem, 1992). Even more confusing is the fact that CLX-I (also known as CpI) of C. parasitica, which hardly shares any sequence similarity with other Type I and Type II sea anemone toxins, affects the crustacean axon in a very similar fashion (Cariello et al., 1989; Salgado and Kem, 1992). It seems therefore that the similar general structure, maintained by a conserved framework of 3 three disulfide cysteine bonds, and the flexibility of the Arg-14 loop, are of major importance in the ability of sea anemone toxins to bind at receptor site-3 of Na_vs. However, the question of how this receptor is recognized by a variety of toxin faces is still unclear.

By using a variety of antibodies raised against synthetic peptides representing sequences of Na_v extracellular loops, the binding of the scorpion α-toxin Lqq5 of Leiurus quinquestriatus quinquestriatus was inhibited by those recognizing the S5–S6 loops of domain 1 and domain 4 (Thomsen and Catterall, 1989; Fig. 1). Unfortunately, these antibodies were not used in the context of sea anemone toxin binding. As determination of the structure of a channel-toxin complex is still an unachievable task, attempts to expose channel regions or residues involved in such interaction focus nowadays on mutagenesis accompanied by analysis of the effect on toxin binding and action.

With the assumption that scorpion α-toxins and sea anemone toxins interact with the Na_v via positively charged residues and on the basis of the results obtained with the antibodies (Thomsen and Catterall, 1989), Rogers and co-workers (1996) neutralized by mutagenesis acidic residues at the extracellular loops of domains 1 and 4 of rNa_v1.2a and examined the effects on Av2 and Lqq5 binding. These experiments highlighted a putative role for Glu-1613, at the loop connecting S3 and S4 of domain 4 (D4/S3-S4; Fig. 1). When the charge of Glu-1613 was inverted, the binding of both Av2 and Lqq5 was abolished. Substitutions E1616A, V1620A and L1624A also diminished the binding affinity of Av2, but had a minor effect on the binding of Lqq5 (Rogers et al., 1996). The difference in binding between the two toxins was further substantiated as substitutions L1614A and S1621A improved significantly the binding of only Av2. Thermodynamic mutant double cycle analysis of the K37A mutant of ApB against D1612N of rNa_v1.5 (Asp-1612 is the equivalent of Glu-1613 in rNa_v1.2a; Fig. 1) revealed coupling energy (ΔΔ G = 1.5 kcal/mol) between the two residues (Benzinger et al., 1998). This result implies that Lys-37 of the bound toxin is in close proximity to Asp-1612 of the channel, which therefore was suggested to be part of the receptor site of ApB on rNa_v1.5. Nevertheless, Lys-37 is not conserved in all Type I and Type II sea anemone toxins (Fig. 2) and therefore, despite the small effect on Av2 action when substituted (Moran et al., 2006), it could not be considered part of the bioactive surface of this sea anemone toxin. In light of the results obtained when the acidic residue at D4/S3-S4 was substituted in rNa_v1.2a (Rogers et al. 1996) and rNa_v1.5 (Benzinger et al., 1998), the equivalent residue in DmNa_v1, Asp-1701, was substituted to arginine and the activities of LqhαIT, Av2 and Av3 were tested on the mutated channel (Moran et al., 2007). While the scorpion and Type I sea anemone toxins lost their activity, the Av3 effect changed only slightly (Moran et al., 2007), indicating that its interaction with DmNa_v1 differs from that of LqhαIT and Av2. This result clearly indicated that Asp-1701 is not a common determinant for binding of all site-3 toxins. The data accumulated thus far question the extent by which the D4/S3-S4 loop is involved in receptor site-3. While it cannot be ruled out that for certain toxins, such as LqhαIT, ApB and Av2, this region is part of this receptor site, it might not be involved with the receptor site of Av3 (unpublished results). Another possibility is that D4/S3-S4 is in close proximity to receptor site-3 and therefore is involved indirectly in the binding of certain site-3 toxins. Despite the putative interaction of Lys-37 in ApB with Asp-1612 of rNa_v1.5 (Benzinger et al., 1998), further mutagenesis and thermodynamic mutant double cycle analysis performed in ApB and D4/S3-S4 of rNa_v1.5 failed to identify toxin-channel residue pairs that exhibit significant coupling energy (Siebert et al., 2004).