The S6 gate in regulatory Kv6 subunits restricts heteromeric K⁺ channel stoichiometry

The Shaker-like family of voltage-gated K⁺ channels comprises four functionally independent gene subfamilies, Shaker (Kv1), Shab (Kv2), Shaw (Kv3), and Shal (Kv4), each of which regulates distinct aspects of neuronal excitability. Subfamily-specific assembly of tetrameric channels is mediated by the N-terminal T1 domain and segregates Kv1-4, allowing multiple channel types to function independently in the same cell. Typical Shaker-like Kv subunits can form functional channels as homotetramers, but a group of mammalian Kv2-related genes (Kv5.1, Kv6s, Kv8s, and Kv9s) encodes subunits that have a "silent" or "regulatory" phenotype characterized by T1 self-incompatibility. These channels are unable to form homotetramers, but instead heteromerize with Kv2.1 or Kv2.2 to diversify the functional properties of these delayed rectifiers. While T1 self-incompatibility predicts that these heterotetramers could contain up to two regulatory (R) subunits, experiments show a predominance of 3:1R stoichiometry in which heteromeric channels contain a single regulatory subunit. Substitution of the self-compatible Kv2.1 T1 domain into the regulatory subunit Kv6.4 does not alter the stoichiometry of Kv2.1:Kv6.4 heteromers. Here, to identify other channel structures that might be responsible for favoring the 3:1R stoichiometry, we compare the sequences of mammalian regulatory subunits to independently evolved regulatory subunits from cnidarians. The most widespread feature of regulatory subunits is the presence of atypical substitutions in the highly conserved consensus sequence of the intracellular S6 activation gate of the pore. We show that two amino acid substitutions in the S6 gate of the regulatory subunit Kv6.4 restrict the functional stoichiometry of Kv2.1:Kv6.4 to 3:1R by limiting the formation and function of 2:2R heteromers. We propose a two-step model for the evolution of the asymmetric 3:1R stoichiometry, which begins with evolution of self-incompatibility to establish the regulatory phenotype, followed by drift of the activation gate consensus sequence under relaxed selection to limit stoichiometry to 3:1R.