Site 3 sea anemone toxins modify inactivation of mammalian voltage-gated Na channels. One variant, anthopleurin A (ApA), effectively selects for cardiac over neuronal mammalian isoforms while another, anthopleurin B (ApB), which differs in 7 of 49 amino acids, modifies both cardiac and neuronal channels with high and approximately equal affinity. Previous investigations have suggested an important role for cationic residues in determination of toxin activity, and our single-site mutagenesis studies have indicated that isoform discrimination can be partially explained by the unique cationic residues Arg-12 and Lys-49 of anthopleurin B (ApB). Here, we have further investigated the role of cationic residues by characterizing toxin mutants in which two such residues are replaced simultaneously. The ApB double mutants R14Q-K48A (cationic residues identical in both ApA and ApB), R12S-K49Q (cationic residues unique to ApB), and R12S-R14Q (cationic residues located in the unstructured loop shared among anemone toxins) were constructed by site-directed mutagenesis and their biological activities characterized by sodium uptake assays in cell lines expressing the neuronal (N1E-115) or cardiac (RT4-B) isoform of the Na channel. Each double mutant displayed reduced activity compared with wild type, but none were completely inactive. Neutralization of the proximal cationic residues (R12 and R14) was the most effective, reducing affinity 72-fold (neuronal) and 56-fold (cardiac). Substitution of cationic residues that differed between ApB and ApA (R12S-K49Q) reduced affinity of the toxin for neuronal channels to a much greater extent than for cardiac channels, producing affinities only slightly lower than for ApA in each case. A structural model for ApB that takes available data into account is proposed. Electrophysiological comparison of ApA, ApB, and the R12S-K49Q mutant confirmed that these two residues are key to the isoform discriminatory ability of ApA. Surprisingly, toxin affinities estimated under voltage clamp, measured in the absence of veratridine and thus most closely reflecting affinity for the closed state of the channels, were greater than estimated by flux for all three toxin forms tested. Of particular note was ApB, which exhibits a 100-fold higher affinity for cardiac channels than estimated previously by flux. We conclude that (1) the residues at positions 12 and 49 are critical for both high affinity and isoform discrimination, (2) of the two arginines, Arg-12 is primarily responsible for maintaining the high binding affinity of ApB to neuronal Na channels, and (3) the cationic nature of Arg-14 is not essential since it can be at least partially compensated for by Arg-12 or one of the two C-terminal lysines. In addition, our data suggest that ApB may also discriminate between closed and open forms of the cardiac Na channel.