**4. The Ubiquitous Hemolytic Activity of Fish Cytolysins: From the Definition of the Pore-Forming Mechanism to its Structural Confirmation**

The cytolytic activity of almost all fish cytolysins isolated so far has been extensively characterized against erythrocytes, hence their being referred to as hemolysins in some instances. In spite of some divergence in the literature as to the potency and the species specificity of the hemolytic activity of fish cytolysins, which can by at least partly explained by different experimental conditions—they are all most potent against rabbit erythrocytes [13,24,25], although the reason for such marked sensitivity remains unknown. In addition, a weak hemolytic activity against rat, mice, and cattle erythrocytes has been reported for some fish cytolysins, while those from human, sheep, pig, and chicken appear to be resistant to the lytic activity of these toxins [12,13,25].

There is also some evidence as to the lytic activity of fish cytolysins against other cell types. For instance, the prolonged exposure of neuro-glioblastoma NG108-15 cells to TLY led to an irreversible increase in membrane permeability [57]. In addition, SNTX lysed the membranes of platelets in rabbit blood in a dose-dependent way [58].

Cytolytic activity is often associated with the action of enzymes such as proteases and phospholipases and, in some cases, of C-type lectins such as CEL-III from the sea cucumber *Cucumaria echinata* [59,60]. However, fish venoms lack phospholipase activity [24,26,27,61–63] and, although enzymatic activity such as proteolytic and hyaluronidase has been described in these venoms, no such activity was associated with the purified lethal/hemolytic factors [12]. As to C-type lectins, although such molecules have been described in fish venoms [64], they are not associated with the hemolytic activity induced by these venoms [65]. Therefore, fish cytolysins were very soon believed to lyse erythrocytes through a direct, non-enzymatic mechanism [26].

The formation of non-selective transmembrane pores was soon shown to be the mechanism by which fish cytolysins destroy cells [42,57,66]. The pore-forming activities of SNTX and Sp-CTx were demonstrated through osmotic assays, in which the kinetics of the hemolysis was evaluated in the presence of osmotic protectants of different sizes [42,66]. While the smaller compounds raffinose and saccharose failed to prevent hemolysis, polyethylene glycols (PEGs) of ∼1000–2000 reduced the rate of hemolysis induced by the toxins and PEGs > 3000 conferred almost full protection against it [42,66]. The pores formed by SNTX in the membrane of rabbit erythrocytes were estimated to be 2.5–3.2 nm in diameter [66].

The search for what mediates the interaction between fish cytolysins and the cell membrane began as soon as the first such toxins were purified. For instance, the hemolytic activity of the *T. draco* venom was found to be preceded by the binding of its hemolytic component to a protein receptor (glycophorin) on the surface of erythrocytes [13]. From then on, it became clear that cationic amino acid residues present in fish cytolysins are essential for their interaction with anionic, neutral or zwitterionic lipids in the membrane of erythrocytes, eventually leading to hemolysis.

In 1997, it was observed that SNTX no longer induced hemolysis when positively charged lysine and arginine residues in its surface were chemically modified, although the toxin's secondary structure remained unaffected [66]. The modification of cationic residues also inhibited the lethal activity associated with SNTX [67]. These data are in agreement with structural motifs observed in these toxins, for instance, an amphiphilic, ∼20-residues long α-helix flanked by regions rich in basic residues was predicted in both α- and β-subunits of SNTX, and it is believed to be the cationic site crucial for the toxin's hemolytic activity [52].

As expected, the hemolytic activity of SNTX and neoVTX against rats and rabbit erythrocytes, respectively, was competitively inhibited by negatively charged lipids, most potently by cardiolipin and less so by phosphatidylserine and the gangliosides GM1 and GM2 [39,66]. Although cardiolipin was the most potent inhibitor, it is not present in the erythrocyte membrane, thus the hemolytic activity must be actually triggered by the electrostatic interaction between the toxins' cationic residues and anionic phospholipids such as phosphatidylserine, which is abundant in the membrane of erythrocytes [39]. It stands to reason that the species-specificity of the hemolytic activity of fish cytolysins could be related to the density of these lipids in the erythrocyte membrane.

The electrostatic interaction between anionic lipids and toxins is influenced by the number of negative charges and the pKa of acidic groups in the lipids, as indicated by the different inhibitory potencies of different lipids [66]. This conclusion is further supported by the fact that the maximal hemolytic activity of Sp-CTx is achieved between pH 8 and 9 [43]. Although the neutral lipids phosphatidylcholine, phosphatidylethanolamine, sphingomyelin, and cholesterol did not affect the hemolytic activity of SNTX and neoVTX [39,66], that which was induced by Sp-CTx in rabbit erythrocytes was inhibited by phosphatidylethanolamine, phosphatidylglycerol, and cholesterol [43]. Similar results regarding the role of differently charged lipids on the hemolytic activity were found for the venomous extract of the lionfish *P. volitans* [68], which is yet another indication that this venom contains cytolysins homologous to those identified in other fish species.

The presence of the MACPF/CDC domain in fish cytolysins [53,55] and the inhibitory effect of cholesterol on the hemolytic activity of Sp-CTx [43] suggest that this lipid is also required for the interaction between these toxins and the erythrocyte membrane, much like the MACPF/CDC proteins found in bacteria.

The hemolytic activity of fish cytolysins is influenced by factors other than the cationic residues present in their surfaces. That of Sp-CTx, for instance, is calcium dependent, being abolished by zinc ions [43]. Furthermore, five of the 15 cysteine residues and 10 of the 18 cysteine residues of SNTX and neoVTX, respectively, are free, and these free thiol groups, much like the tryptophan residues in the surface of SNTX, also play an important role in the hemolytic activity induced by these toxins [39,52,69].

All this evidence pointing to a pore-formation mechanism being responsible for the hemolytic activity of fish cytolysins was corroborated by the three-dimensional structure of SNTX. Pore-formation was shown to be the result of an interaction between the αand β-chains along their longitudinal axis through charge complementarity between the MACPF/CDC domains. These domains—which are formed by twisted four-stranded antiparallel β-sheets with two adjacent bundles of α-helices (transmembrane helices— TMH)—interact on the heterodimer to form a soluble early pre-pore. After that, the TMH from each chain unwind to form a large, continuous β-sheet that comprises the β-barrel of the pore (Figure 4).

**Figure 4.** Transmembrane pore formation mechanism. The SNTX heterodimer represents an early and soluble phase of the pore formation mechanism (**A**). After the interaction between chains α and β, the MACPF/CDC domain helices undergo a conformational change to form the continuous β-sheet of the transmembrane pore (**B**). (**A**) was produced on Pymol using the model 4wvm deposited in the Protein Data Bank [55] and (**B**) was adapted from [55]. TMH—transmembrane helices; M—membrane.

This arrangement is in agreement with the complementarity between MACPF/CDC monomers described for other perforins, which directs initial oligomerization events (prepore assembly). In SNTX, this pre-pore contains 20 SNTX subunits (or 10 SNTX-α/β heterodimers) that line up along a horizontal plane [55].

Furthermore, the PRYSPRY domain was shown to be responsible for the initial interaction of SNTX with the cell surface. Consistent with this function, this domain is located in the solvent-exposed face of SNTX, in a position analogous to the Ig and C2 domains of CDC and perforins, which mediate protein–protein and protein–lipid interactions in the recognition of pathogens by TRIM immune proteins [55].

The high sequence similarity and the presence of conserved domains in the other fish cytolysins identified so far allows us to predict that these proteins perform their cytolytic functions through the same mechanism. For instance, the residues G208 (α-chain) and G209 (β-chain) of SNTX, reported as necessary for pre-pore formation, are present in all cytolysins. So are the residues F206 (α-chain) and F54, I56, and S52 (β-chain), related to the formation of hydrogen bonds between the β4-strand of SNTX-α and the β1-strand of SNTX-β, which are also involved in pre-pore formation (Figure 3).

Moreover, the pairs K205 (α-chain)-E55 (β-chain) and K54 (α-chain)-E206 (β-chain) present in SNTX's β4 (α-chain) and β1 (β-chain) strands form ionic pairs analogous to a partially closed zipper, which contributes to pore formation. These ionic pairs are conserved in almost all cytolysins, in spite of exchanges as to which residue belongs to each chain.

In addition to the structural evidence, the formation of large ring-shaped pores was visualized in rabbit erythrocytes by transmission electron microscopy (outer dimension: 257 ± 5.7 Å (mean ± SEM); lumen i.d.: 117 ± 4.5 Å) [55], in agreement with what had been predicted for the pores formed by SNTX and Sp-CTx [42,66].
