Venom toxins are invaluable tools for exploring the structure and mechanisms of ion channels. Here, we solve the structure of double-knot toxin (DkTx), a tarantula toxin that activates the heat-activated TRPV1 channel. We also provide improved structures of TRPV1 with and without the toxin bound, and investigate the interactions of DkTx with the channel and membranes. We find that DkTx binds to the outer edge of the external pore of TRPV1 in a counterclockwise configuration, using a limited protein-protein interface and inserting hydrophobic residues into the bilayer. We also show that DkTx partitions naturally into membranes, with the two lobes exhibiting opposing energetics for membrane partitioning and channel activation. Finally, we find that the toxin disrupts a cluster of hydrophobic residues behind the selectivity filter that are critical for channel activation. Collectively, our findings reveal a novel mode of toxin-channel recognition that has important implications for the mechanism of thermosensation.DOI: http://dx.doi.org/10.7554/eLife.11273.001

Voltage-activated potassium (Kv) channels open to conduct K+ ions in response to membrane depolarization, and subsequently enter non-conducting states through distinct mechanisms of inactivation. X-ray structures of detergent-solubilized Kv channels appear to have captured an open state even though a non-conducting C-type inactivated state would predominate in membranes in the absence of a transmembrane voltage. However, structures for a voltage-activated ion channel in a lipid bilayer environment have not yet been reported. Here we report the structure of the Kv1.2–2.1 paddle chimera channel reconstituted into lipid nanodiscs using single-particle cryo-electron microscopy. At a resolution of ~3 Å for the cytosolic domain and ~4 Å for the transmembrane domain, the structure determined in nanodiscs is similar to the previously determined X-ray structure. Our findings show that large differences in structure between detergent and lipid bilayer environments are unlikely, and enable us to propose possible structural mechanisms for C-type inactivation.