Objective: Scorpion α-neurotoxins (ɑ-NaTx) inhibit the inactivation of voltage-gated sodium channels (Nav) with variable efficiency between organisms and channel isoforms. Based on our previous results, we hypothesized that the derivative of the ɑ-NaTx called BeM9 with two amino acid residues substituted with glycine (A4G and Y17G; BeM9GG) should be more selective for the mammalian channels. Surprisingly, BeM9GG lost its activity on the cardiac isoform Nav1.5. We provide a possible explanation of this effect, taking into account the published structures of ɑ-NaTx–Nav complexes. Methods: We produced BeM9GG in Escherichia coli as a fusion protein with thioredoxin, which was cleaved by cyanogen bromide. We purified BeM9GG using chromatography and measured its activity on Nav isoforms expressed in Xenopus laevis oocytes by the two-electrode voltage clamp technique. We performed computer modeling of ɑ-NaTx–Nav complexes to figure out the peculiarities of toxin interactions with Nav1.5. Results and Discussion: Using electrophysiology, we tested BeM9GG on Nav, and unexpectedly observed compromised activity on Nav1.5. We compared the structures of available ɑ-NaTx–Nav complexes to explain this effect. In case of Nav1.5 the toxin-binding site is immersed deeper in the membrane. We explain this by sequence variations in two positions of the voltage-sensing domain IV of Nav: in Nav1.5 corresponding residues have shorter side chains, permitting the toxin to sit deeper. At the same time, residue Y17 in BeM9, which is missing in BeM9GG, interacts with the hydrophobic core of the membrane and may play a significant role in its activity against Nav1.5. Conclusions: The role of the membrane in ɑ-NaTx interactions with Nav has been overlooked. It is especially important in case of Nav1.5, where the tripartite toxin–ion channel– membrane complex is formed. Our study will be of help for the future design of selective Nav ligands.