Ctenophores and parahoxozoans independently evolved functionally diverse voltage-gated K⁺ channels

Voltage-gated K⁺ (Kv) channels participate in a wide diversity of animal physiological systems but are most noted for shaping and patterning electrical responses in neurons and other excitable cells. Animal Kv channels display a high degree of functional diversity and are encoded by three structurally unique families: Shaker (Kv1–6, 8, and 9), ether-a-go-go (EAG, Kv10–12), and KCNQ (Kv7). All these gene families encode subunits that share a common core of six transmembrane domains (S1–S6), comprised of a voltage-sensor domain (VSD) and a K⁺-selective pore domain (Long et al., 2005; Whicher and MacKinnon, 2016; Sun and MacKinnon, 2017). Functional channels assemble as tetramers with a single pore surrounded by four independent VSDs. What structurally distinguishes the three gene families are unique family-specific intracellular domains that regulate gating and assembly. Shaker family channels have an N-terminal cytoplasmic T1 domain based on the BTB/POZ fold that self-tetramerizes and governs assembly (Pfaffinger and DeRubeis, 1995; Kreusch et al., 1998). In contrast, the cytoplasmic N-terminal EAG domain of EAG family channels has homology to Per-Arnt-Sim domains (Morais Cabral et al., 1998) and works in tandem with the cytoplasmic C-linker/cyclic nucleotide-binding homology domain to regulate gating (Gustina and Trudeau, 2011; Gianulis et al., 2013; Whicher and MacKinnon, 2016; Wang and MacKinnon, 2017) and assembly (Lin et al., 2014). KCNQ family channels lack these domains but instead have a unique C-terminal coiled-coil domain that regulates channel assembly (Schwake et al., 2006; Sun and MacKinnon, 2017). The Shaker family encodes a broad diversity of depolarization-gated K⁺ currents, including classic neuronal A-currents and delayed rectifiers. KCNQ family channels require PIP2 for gating and encode neuronal M-currents that regulate action potential threshold (Wang et al., 1998). The neurophysiology of the EAG family is less well characterized, but EAG family channels tend to activate with low voltage thresholds and influence subthreshold excitability (Saganich et al., 1999, 2001; Zou et al., 2003; Zhang et al., 2009, 2010; Kazmierczak et al., 2012; Li et al., 2015c; Hermanstyne et al., 2023).

Phylogenetic comparisons of diverse bilaterian and cnidarian species indicate that eight Kv channel types spanning these three gene families were present in the last common ancestor (LCA) of Bilateria and Cnidaria (Jegla et al., 1995, 2009, 2012; Jegla and Salkoff, 1997; Martinson et al., 2014; Li et al., 2015b, 2015c; Lara et al., 2023). The cnidarian/bilaterian Shaker family is comprised of four functionally independent gene subfamilies (Shaker or Kv1, Shab or Kv2, Shaw or Kv3, and Shal or Kv4) (Salkoff et al., 1992), while the cnidarian/bilaterian EAG family has three functionally independent gene subfamilies (Eag or Kv10, Erg or Kv11, and Elk or Kv12) (Warmke and Ganetzky, 1994). The KCNQ family has two sub-types widely conserved in bilaterians (Jegla et al., 2009), but only one ancestral lineage shared between cnidarians and bilaterians (Li et al., 2015b). Surprisingly, the Kv channel complement of the last common cnidarian ancestor appears to have been far more extensive than that of the last common bilaterian ancestor. The entire complement of bilaterian Kv channels is predicted to have descended from only eight to nine Kv channels present in the last common bilaterian ancestor (one from each EAG/Shaker subfamily and possibly two KCNQs) (Jegla et al., 2009; Li et al., 2015b, 2015c; Lara et al., 2023). In contrast, extant cnidarians share 28 Kv channel gene lineages, with most of the diversification occurring within the Shaker family (Lara et al., 2023). Despite this large number of ancestral Kv channels, all cnidarian Kv channel families and subfamilies are shared with bilaterians.

Functional phenotypes of Kv channels are highly conserved between cnidarians and bilaterians (Jegla et al., 1995, 2012; Jegla and Salkoff, 1997; Li et al., 2015b, 2015c). However, cnidarian-specific gene duplications within the Kv1 and Kv4 subfamilies do expand the functional phenotypes of these subfamilies beyond what is typically seen in bilaterians. For example, bilaterian Kv4 subfamily channels all seem to encode classic A-type currents with closed state inactivation, whereas the large cnidarian Kv4 subfamily encodes functional orthologs of the bilaterian Kv4 channels, fast-inactivating Kv1-like currents, and delayed rectifiers (Jegla and Salkoff, 1997; Li et al., 2015b). Similarly, the large cnidarian Kv1 subfamily includes channels with highly diverse kinetics and a broad range of activation thresholds (Jegla et al., 2012). Thus, phylogenetic lineage alone might not be sufficient to predict precise gating phenotypes of Kv channels over long evolutionary times or in the face of large, lineage-specific gene expansions.

The eight Kv channel lineages shared between cnidarians and bilaterians do not appear traceable to the origins of the nervous system. Recent genomic analyses place ctenophores, or comb jellies, as a sister lineage to all other animals (Ryan et al., 2013; Moroz et al., 2014; Whelan et al., 2015, 2017; Schultz et al., 2023). They can therefore provide insights into Kv channel diversity in the LCA of all animals at a time when the nervous system was first evolving. Examination of the first ctenophore genome (Mnemiopsis leidyi) revealed a large set of Kv channel genes, but these represented only two channel lineages shared with cnidarians and bilaterians: Kv1-like Shaker family genes (>40 genes) and the EAG family (2 genes forming an outgroup to the cnidarian/bilaterian Kv10–12 subfamilies) (Li et al., 2015b). Recently, Kv2–4-like Shaker family channels were discovered in choanoflagellates, the protozoan sister lineage of animals (Jegla et al., 2024), but the Kv2–4 lineage is surprisingly not found in Mnemiopsis. These two lineages, Kv1-like and Kv2–4-like, can be differentiated based on the unique presence of a highly conserved Zn²⁺-binding site in the T1 interface of Kv2–4-like channels (Bixby et al., 1999; Jahng et al., 2002). No Mnemiopsis Shaker family channels have the Kv2–4-like Zn²⁺-binding site.

Comparison of choanoflagellates and Mnemiopsis with cnidarians and bilaterians therefore suggests the last common animal ancestor had as few as three Kv channel lineages (Kv1-like, Kv2–4-like, and EAG), one of which was lost in the lineage leading to Mnemiopsis (Li et al., 2015b; Jegla et al., 2024). This raises the possibility that the complexity of electrical signaling, which requires functionally diverse Kv channels, was low at the time ctenophores diverged from the rest of the animals. But does this mean electrical signaling in ctenophore nervous systems lacks some of the features present in nervous systems of cnidarians and bilaterians? Or just that signaling complexity simply evolved independently in ctenophores? The large expansion of Kv1-like Shakers in Mnemiopsis provides an opportunity to gain insight into how much independent functional diversification of voltage-gated ion channels has occurred in ctenophores. However, the functional significance of this expansion to ctenophores is unclear because its evolutionary origins are unknown and only five of the channels have been functionally expressed. MlShak1 and MlShak2 do not form functional channels as homomultimers, although they can heteromultimerize with a cnidarian Shaker (Li et al., 2015b). MlShak3–5, in contrast, do function as homomultimers but encode fast (MlShak4 and 5) or slow (MlShak3)-inactivating K⁺ channels that are highly similar to cnidarian and bilaterian Kv1 currents (Jegla et al., 2012; Simonson et al., 2024). Fast inactivation in MlShak4 and MlShak5 even occurs by the same N-type ball-and-chain mechanism first described for Drosophila Shaker (Hoshi et al., 1990; Zagotta et al., 1990).

Here, we examine the evolutionary origins of Kv channel diversity in Mnemiopsis Shaker family expansion by phylogenetically comparing voltage-gated K⁺ channel sets from diverse ctenophores and functionally characterizing a broader phylogenetic representation of Mnemiopsis Shakers. We find that most of the Mnemiopsis Shaker and EAG family expansion predates the radiation of three major ctenophore lineages: Beroe, Mnemiopsis, and Hormiphora. In contrast, we found no evidence for the KCNQ family, the Kv2–4 Shaker subfamilies, or the Kv10–12 EAG subfamilies in any ctenophore species. Functional expression of 16 Mnemiopsis Shaker channels representing conserved ctenophore Shaker lineages revealed an unexpectedly broad diversity of gating phenotypes. These results suggest that ctenophores have indeed independently evolved a functionally diverse set of Kv channels, primarily via expansion of the Kv1-like lineage of the Shaker gene family.

We identified Kv channel sequences in Beroe ovata (Vargas et al., 2024) and Hormiphora californensis (Schultz et al., 2021) protein predictions with BLASTP searches using M. leidyi (Shaker and EAG families) (Li et al., 2015b, 2015c) and Nematostella vectensis (KCNQ and HCN) (Li et al., 2015b) family members as queries. Protein family identity was confirmed with reciprocal BLASTP searches against N. vectensis proteins. Reciprocal best BLAST strategies have been very effective for comprehensively identifying Kv channel sets from Mnemiopsis, cnidarians, placozoans, and choanoflagellates (Jegla et al., 2012, 2018, 2024; Li et al., 2015b, 2015c; Lara et al., 2023). We used a similar pipeline to search the transcriptome of the ctenophore Euplokamis dunlapae (Whelan et al., 2017), except we started with a TBLASTN search and performed a reciprocal BLASTP using the translated results. Sequences were used for phylogenetic analysis if they were >90% complete with respect to conserved domains.

We made alignments from the ctenophore EAG and Shaker family sequences in MEGA 11 (Tamura et al., 2021) using MUSCLE (Edgar, 2004) with default parameters. Gaps in conserved regions for a few gene predictions were evident in these alignments. When possible, these gaps were filled with additional sequence data from genomic and transcriptomic databases. Transcriptomic and genomic data used to fill gaps (1) were confirmed to be on the same contig as the original predicted protein and (2) were located between the predicted sequences flanking the gaps. We removed poorly conserved linker regions with repeated independent length variations from the alignment and manually adjusted the alignment around gaps for consistency prior to phylogenetic analysis. We included Kv genes from Trichoplax adhaerens (Placozoa), N. vectensis (Cnidaria), Drosophila melanogaster (Bilateria and Protostomia), humans (Bilateria and Deuterostomia), and choanoflagellates (Shaker family only, Salpingoeca dolichothecata, Salpingoeca helianthica, and Mylnosiga fluctuans) for comparison. We performed phylogenetic reconstruction of the channel families through Bayesian inference using MrBayes v3.2.7a (Ronquist et al., 2012) in combination with BEAGLE 3 (Ayres et al., 2019) under a mixed amino acid model. Phylogenies were generated with MCMC sampling for 3,000,000 generations using 2 independent runs of 6 chains sampled every 5,000 generations; the first 25% of samples were discarded to focus the analysis on the best trees. SDs of split frequencies converged to 0.0003 and 0.001 for the EAG and Shaker phylogenies, respectively. We ran a second phylogenetic analysis for each alignment using maximum likelihood in IQ-TREE 2 (Minh et al., 2020) with an LG+F+R5 model (with 1,000 ultrafast bootstrap replications) to corroborate results. To ascertain the number of Kv channels in the LCA of Beroe, Mnemiopsis, and Hormiphora, we identified clades with strong statistical support (posterior probability >0.95 for Bayesian phylogenies or bootstrap support >0.95 for the maximum-likelihood phylogeny) that contained at least one sequence from H. californensis and at least one sequence from either B. ovata or M. leidyi. Statistically supported clades containing E. dunlapae sequence(s) along with at least one sequence from the other three species were inferred to have been present in the LCA of these lineages. Because the E. dunlapae transcriptome almost certainly represents an incomplete subset of genes in the species, this study is unable to comprehensively predict all channel clades that are traceable to this early ctenophore ancestor.

Coding sequences for 14 Mnemiopsis Shaker genes (MlShak3–16) were synthesized using Xenopus laevis–optimized codons (Twist Bioscience) and cloned into the pET21(+) vector (Twist Bioscience) between the EcoRI and NotI sites. We included 5′ and -3′ Xenopus β-globin UTR sequences from the pOX oocyte expression vector surrounding the ORFs to further optimize expression in oocytes (Jegla and Salkoff, 1997). The MlShak13 Δ2–18 transcription template was generated by PCR from the MlShak13 WT expression plasmid. Capped run-off transcripts were synthesized either from NotI-linearized templates using the mMESSAGE mMACHINE T7 Transcription Kit (Thermo Fisher Scientific) or from PCR templates using 100U T7 polymerase (Takara), 4 mM CleanCap-AG (Trilink Biotechnologies), 5 mM each nucleotide, 5 mM DTT, 20U rnase inhibitor, and 0.02U yeast inorganic pyrophosphatase (NEB). RNA synthesis using the latter method, which we switched to partway through this study, was cheaper, more reliable, and increased protein expression in oocytes (though relative expression differences between constructs were maintained); it is our recommended method.

Xenopus oocytes (Xenopus 1) were defolliculated using 0.5–1 mg/ml type II collagenase (Sigma-Aldrich) in calcium-free ND98 solution and cultured in ND98 with 2 mM Ca²⁺ supplemented with 100 U/ml/100 µg/ml/50 µg/ml of penicillin/streptomycin/tetracycline and 2.5 mM Na-pyruvate (Li et al., 2015a; Simonson et al., 2024). Optimal RNA injection amounts were empirically determined and ranged from ∼0.05 to 25 ng/oocyte. Oocytes were incubated for 1–4 days at ∼18°C, and then two-electrode voltage-clamp recordings were made under a constant flow of low chloride solution (98 mM Na⁺, 2 mM K⁺, 2 mM Cl⁻, and 5 mM HEPES, titrated to pH 7.2 with methanesulfonic acid) to reduce native Cl⁻ currents (important when recording small currents [Baker et al., 2015]). Electrodes were made from borosilicate glass (Sutter Instrument) and filled with 3 mM KCl to tip resistances of ∼0.4–1.5 MΩ. Bath pellet electrodes were separated via a 1 M NaCl 1% agarose bridge. Recordings were made using a pClamp10/Digidata 1440a digitizer (Molecular Devices) and a CA-1B amplifier (Dagan Instruments) run in TEV mode (Thermo Fisher Scientific). Data were low-pass filtered at 2–5 kHz and digitized at 4–10 kHz. We analyzed the data using OriginLab (Northampton, MA) and Clampfit (Molecular Devices). Boltzmann distributions were fit using the equation $f(V)=A2+(A1−A2)/{1/[1+e(V−V50)/s]}$⁠, where s is the slope, A₁ and A₂ are the lower and upper bounds, and V₅₀ is the half-maximal activation/inactivation voltage. The inactivation and recovery time course of MlShak13 was fit with a single exponential function $(It=Ito+Ae−t/τ)$⁠, where I_t is the current at time t, I_to is the initial current, A is the amplitude, and τ is the time constant.

Table S1 includes supporting analysis of published transcriptome data (see Sebé-Pedrós et al. [2018] for source data). Data S1, S2, S3, S4, S5, S6, and S7 contain the Shaker family protein sequences, sequence alignments, and tree files underlying Table 1 and Figs. 2 and 3.

To explore the evolutionary diversification of Kv channels within ctenophores, we identified Kv channel sequences from two newly available high-quality ctenophore genomes and transcriptomes from B. ovata (Vargas et al., 2024) and H. californensis (Schultz et al., 2023) using a comprehensive BLAST search strategy (see Materials and methods). These species are believed to have shared a common ancestor with M. leidyi (Fig. 1, A and B) ∼290 million years ago (Liu et al., 2024) and provide insights into the ancestral channel set of these three major ctenophore lineages. We also mined the transcriptome of E. dunlapae, which is a member of a lineage that branched off from the rest of ctenophores an estimated 480 million years ago (Liu et al., 2024). In comparison to the genomes we examined for other species, the E. dunlapae transcriptome is unlikely to capture the complete set or even the majority of Shaker genes present in these species. It can nevertheless provide some insight into how Shaker channels present at an earlier time point in ctenophore evolution are related to the Shaker channels in Beroe, Hormiphora, and Mnemiopsis. The phylogenetic relationship between ctenophores, parahoxozoans, and choanoflagellates is shown in Fig. 1 C. Ctenophora and Parahoxozoa contain all animal lineages in which nervous systems are found, and choanoflagellates are the closest protozoan relatives of animals. Sponges, which do not have nervous systems, have been eliminated for visual simplicity but are the sister lineage of Parahoxozoa (Dunn et al., 2008; Ryan et al., 2013; Moroz et al., 2014; Li et al., 2021; Schultz et al., 2023).

Kv channel gene counts from four ctenophore species (E. dunlapae, H. californensis, B. ovata, and M. leidyi) are compared with bilaterians (human and fly), a cnidarian (N. vectensis), a placozoan (T. adhaerens), and three choanoflagellates (S. helianthica, S. dolicothecata, and M. fluctuans) in Table 1. Sequences for the B. ovata, H. californensis, and E. dunlapae Kv channels described here are provided in Data S1. These are lower bound estimates of the gene numbers, as we did not include a small number of incomplete Shaker family sequences, and the E. dunlapae transcriptome is unlikely to be comprehensive. We did not find any ctenophore KCNQ family genes and found no evidence for novel ctenophore-specific Kv channel families. The ctenophore EAG family was comprised of two ortholog pairs shared between H. californensis and M. Leidyi in our phylogenetic analyses (Table 1 and Fig. 2), indicating that both genes were likely present in the LCA of Beroe, Mnemiopsis, and Hormiphora despite their absence from the draft B. ovata genome. We used the HCN family of hyperpolarization-gated cation channels as an outgroup for the EAG family phylogeny since both families are members of the cyclic nucleotide-binding domain (CNBD) superfamily of voltage-gated cation channels (Baker et al., 2015; Jegla et al., 2018). A single HCN channel has previously been reported for the cydippid ctenophore Dryodora glandiformis (Baker et al., 2015), and we show here that H. californensis, M. leidyi, and B. ovata also have one HCN channel (Table 1, Fig. 2, and Data S1). Ctenophore EAGs form an outgroup to the cnidarian/bilaterian Kv10–12 subfamilies as previously described (Li et al., 2015c). The EAG/HCN sequence alignment and tree file for Fig. 2 are provided as Data S2 and S3, respectively. We repeated the phylogenetic analysis using a maximum-likelihood approach (tree file provided as Data S4) and reached the same conclusions: (1) Ctenophore and parahoxozoan EAG family channels group separately, and (2) there were two EAG channels in the LCA of Beroe, Mnemiopsis, and Hormiphora. We did not find EAG or HCN family genes in E. dunlapae, but we can infer at least one of each must have been present in the ctenophore common ancestor from the presence of both gene families in ctenophores and parahoxozoans.

In contrast to the EAG, KCNQ, and HCN families, we identified numerous Shaker family channels in all ctenophore species, and the Shaker family comprises 112/116 of the Kv channels collectively found in E. dunlapae, H. californensis, B. ovata, and M. leidyi (Table 1). Ctenophore Shaker gene numbers in the three genomes range from 27 in B. ovata up to 42 in M. leidyi, indicating expansion of the Shaker family is widespread in ctenophores. In addition, we found 11 Shaker family channels in the draft transcriptome of E. dunlapae, all of which form clades with at least one other ctenophore species. A Bayesian phylogeny of the Shaker gene family sequences from Table 1 is shown in Fig. 3 with the alignment and tree file provided as Data S5 and S6. The phylogeny indicates that much of the ctenophore Shaker family expansion predates the radiation of the Beroe, Mnemiopsis, and Hormiphora. We found 30 highly supported clades (posterior probability of 1) that contained sequences from H. californensis and at least one of M. leidyi and B. ovata, indicating that these 30 clades were present in the LCA of Beroe, Mnemiopsis, and Hormiphora. We repeated the phylogenetic analysis using a maximum-likelihood approach and recovered the same 30 ancestral clades (tree provided as Data S7) with high (>0.95) bootstrap support. In both phylogenies, the 11 E. dunlapae Shaker sequences fell into 9 different ancestral clades and one additional clade containing only E. dunlapae and H. californensis sequences. We interpret this clade as a 31st ancestral channel clade that must have also been present in the LCA of Beroe, Mnemiopsis, and Hormiphora. These E. dunlapae sequences indicate that much of the diversification of the Shaker family occurred early in ctenophore evolution, though the E. dunlapae data alone do not necessarily give us a comprehensive view of Shaker family and EAG family Kv channel genes present in the earliest ctenophores.

The ctenophore Shaker family sequences formed a single highly supported clade in unrooted phylogenies separate from all parahoxozoan and choanoflagellate sequences, indicating that the gene expansions observed in ctenophore species are fully ctenophore specific. The Shaker family phylogenies presented in Fig. 3 and Data S6 contain three major clades: (1) the ctenophore clade, (2) the parahoxozoan Kv1 subfamily, and (3) a clade comprised of the parahoxozoans Kv2–4 subfamilies together with the choanoflagellate Kv2–4-like Shakers. Channels in the latter clade share the T1 Zn²⁺-binding site that clearly indicates a single common origin. This site is absent from all Kv1 subfamily channels and all ctenophore Shaker family channels, but the phylogeny does not conclusively resolve the relationship between ctenophore Shakers and the Kv1 subfamily (see Discussion).

We did not attempt to functionally express ctenophore EAG or HCN channels in this study and instead focused on the functional characterization of the Shaker family expansion that represents the bulk of the ctenophore Kv channel set. MlShak1–5, the previously expressed ctenophore Shakers, represent only a small fraction of the phylogenetic diversity of ctenophore Shakers shown in Fig. 3 and are thus unlikely to represent the broader functional diversity of ctenophore Shakers. We therefore synthesized codon-optimized Xenopus oocyte expression vectors for 11 additional M. leidyi Shaker genes from across the breadth of the phylogeny, which we name MlShak6–16 here. These 11 genes were selected from the full gene set because (1) we had high confidence in the full coding sequence predictions, including sequence at the N- and C-termini, from transcriptome data and/or cross-species conservation, (2) they had coding sequence lengths compatible with efficient gene synthesis, and (3) they each represented one of the conserved Shaker channel clades present in the LCA of Beroe, Mnemiopsis, and Hormiphora. MlShak1–16 collectively represent 15 of these ancestral clades (Fig. 3). Only the MlShak5 clade has a topology suggesting it may have arisen after Mnemiopsis and Beroe split from Hormiphora, although it may only represent a loss of this family in Hormiphora.

Fig. 4 A shows example traces recorded in response to step depolarizations from oocytes expressing MlShak1–16. The MlShak1–5 current phenotypes shown here are consistent with previous reports (Li et al., 2015b; Simonson et al., 2024). A total of 9/16 channels (MlShak3–9, MlShak13, and MlShak15) consistently produced Kv currents (Fig. 4, A and B) when expressed in isolation and thus are able to form functional homomeric channels. We first injected RNAs at high concentration (∼10–25 ng/oocyte), which in our historical lab experience with >200 different ion channels is sufficient to see even low-efficiency functional channel formation in oocytes. We therefore interpret the absence of detectable K⁺ currents for MlShak1, 2, 10–12, 14, and 16 (Fig. 4, A and B) to mean they are either unable to assemble as homomultimers or unable to gate by voltage alone. Each of these non-expressing channels was injected in multiple oocyte batches to confirm the phenotype, but Fig. 4 B reports current size data from 1 to 2 batches. These channels may be equivalent to cnidarian/bilaterian silent or regulatory subunits that require heteromeric assembly for functional expression (Jegla and Salkoff, 1997; Jegla et al., 2012, 2024; Bocksteins, 2016; Pisupati et al., 2018), but we did not test heteromeric combinations in this study. MlShak3–6 currents were among the best-expressing channels we have observed in the oocyte system, and the data reported in Fig. 4 B were recorded from oocytes injected with ∼0.05–0.1 ng RNA/oocyte. What is immediately apparent from Fig. 4 A is the diversity of gating phenotypes, ranging from rapid inactivation (MlShak4, MlShak5, and MlShak13) through slow inactivation (MlShak3 and MlShak15) to almost no inactivation during 400 ms test steps (MlShak6–9). Most of the channels displayed fast activation typical of bilaterian and cnidarian Shaker family channels, with the exception of MlShak8 and MlShak9. These two channels required multiple seconds for full activation, a phenotype that has not previously been observed in the Shaker family to our knowledge. For MlShak9, activation was so slow it limited our ability to collect data at high voltages. We do not quantify kinetics for these channels in detail here (except for MlShak13 inactivation, below), but the properties of currents were extremely consistent from oocyte to oocyte.

Rapid inactivation in MlShak4 and MlShak5 occurs by a classic N-type ball-and-chain mechanism shared with Drosophila Shaker (Simonson et al., 2024). Here we identify MlShak13 as a third rapidly inactivating ctenophore Shaker and tested it for N-type inactivation by expressing a truncated version lacking the native N terminus (MlShak13 Δ2–18). This truncation removed fast inactivation from the channel, implicating the same classic N-type inactivation mechanism (Fig. 5 A). The MlShak13 inactivation time course is similar to rapid inactivation in MlShak4 and MlShak5 in that all have time constants faster than 6 ms above 0 mV (Fig. 5 B and [Simonson et al., 2024]). However, MlShak13 differs markedly from MlShak4 and MlShak5 in having rapid recovery from inactivation (Fig. 5 C) with a time constant of 35.5 ± 0.7 ms at −120 mV (n = 4) compared with 552 ± 116 and 2,822 ± 452 ms previously reported for MlShak4 and MlShak5 (Simonson et al., 2024), respectively. N-type inactivation has evolved multiple times within the Shaker family and is the mechanism for inactivation in all fast-inactivating Shaker family channels described in cnidarians (three Kv1 channels and one heteromeric Kv4 channel), ctenophores (MlShak4, 5, and 13), and choanoflagellates (two Kv2–4-like channels) (Jegla et al., 1995, 2012, 2024; Jegla and Salkoff, 1997; Simonson et al., 2024).

We generated conductance–voltage (GV) curves and steady-state inactivation (SSI) curves for the nine MlShak channels that expressed as homomultimers (Fig. 6). GV data were calculated from isochronal tail currents following step depolarizations, while SSI data were recorded from isochronal measurement of currents during a depolarizing test step after 6-s preconditioning pulses to various voltages. Average V₅₀ and slope values calculated from single Boltzmann fits of data from individual eggs are shown in Table 2 and were used to generate the smooth curves in Fig. 6. These average data are replotted for direct comparison of voltage activation and SSI ranges in Fig. 7. MlShak3–5 GV data in Table 2 were previously reported (Simonson et al., 2024), and the curves shown in Fig. 6 and Fig. 7 A were generated with the V₅₀ and slope values from that study. GV data for all three fast-inactivating channels (MlShak4, 5, and 13) were generated using N-terminal truncated constructs that remove fast N-type inactivation to reveal tail currents. The V₅₀ values of the GVs for the 9 MlShak homomultimers spanned a range of ∼65 mV from −45.7 ± 0.8 mV in MlShak9 to 18.9 ± 1.1 mV in MlShak7. This range is comparable with the ∼58 mV V₅₀ range for GVs of all reported homomeric cnidarian Kv1–4 channels combined (Jegla et al., 1995, 2012; Jegla and Salkoff, 1997; Bouchard et al., 2006; Sand et al., 2011; Li et al., 2015b). SSI V₅₀ values ranged from −105.8 ± 0.7 mV in MlShak13 to −28.5 ± 1.1 mV in MlShak7. This range of SSI V₅₀s (∼77 mV) is again comparable with the entire range of SSI V₅₀s reported for cnidarian Shaker family channels spanning the Kv1–4 subfamily (∼73 mV) (Jegla et al., 1995, 2012; Jegla and Salkoff, 1997; Bouchard et al., 2006; Li et al., 2015b).

Fig. 8 show mRNA expression in transcripts/million transcripts for MlShak1–16 and MlEAG1 and 2 in deep coverage public transcriptome data (Babonis et al., 2018) for three Mnemiopsis tissues enriched in excitable cells: aboral organ (contains neurons), tentacle bulb (contains neurons), and comb rows (contains comb cells) (see Fig. 1). All the genes are expressed above baseline in at least one of the tissues, supporting a functional role for all of the channels. Notable examples of high expression include MlShak3 in aboral organ; MlShak4, MlEAG1, and MlEAG2 in tentacle bulb; and MlShak15 in comb rows. All three tissues express at least one of the fast-inactivating N-type channels (MlShak4, 5, and 13) and multiple candidate silent subunits. It is interesting that the two M. leidyi EAG genes are expressed almost exclusively in tentacle bulbs and that the EAG genes have been lost in Beroe, which itself has lost its tentacles and tentacle bulbs. Single-cell Mnemiopsis transcriptome data from another study corroborate and extend these results by showing expression of MlShaks in identified classes of neurons, muscle, comb row cells, photocytes, and colloblasts (Sebé-Pedrós et al., 2018; Hehmeyer et al., 2024) (summarized in Table S1). These two data sets are not sufficient to define the sets of channels expressed in single cells with certainty because the first study lacks cell-level resolution and the second lacks the necessary depth of coverage. Nevertheless, both studies suggest a potential for significant co-expression of Kv channels in individual cells.

Here, we show that the large set of Kv channel genes is a common feature of ctenophores, the earliest diverging animal lineage. Fig. 9 shows a comparison of our current view of the ancestral Kv channel sets for bilaterians, cnidarians, and ctenophores, the animals with nervous systems. It is remarkable that ctenophores and cnidarians, whose nervous systems are often described as simple relative to bilaterian animals, have much larger predicted ancestral channel sets than bilaterians. This suggests that a large Kv channel set was probably not a requirement of the origin of centralized nervous systems. It may be sufficient that the ancestral bilaterian channel set represents a functionally diverse group of independent (non-mixing) channel types. The ctenophore K⁺ channel set, while large, is markedly different than the K⁺ channel sets found in cnidarians and bilaterians in that it derives from just two ancestral lineages rather than eight (Jegla et al., 2009; Li et al., 2015b; Lara et al., 2023). Our analyses suggest that early ctenophores (at least as old as the LCA of Mnemiopsis and Hormiphora) had a complement of at least 31 Shaker family channels and 2 EAG family channels (Fig. 9). For both families, the ctenophore channels are more closely related to each other than they are to any of the cnidarian and bilaterian channels. This suggests that single EAG and Shaker family channels in the LCA of all animals independently radiated in each of these lineages to give rise to the full diversity of these channels in modern ctenophore and non-ctenophore animals. This same evolutionary pattern is observed in innexin channels (Ortiz et al., 2023). The cnidarian/bilaterian KCNQ family is entirely absent in ctenophores, and the ctenophore Shaker and EAG families lack the seven functionally independent cnidarian/bilaterian subfamilies. KCNQs and these Shaker/EAG subfamilies might have evolved after the divergence of ctenophores and parahoxozoans. All are also absent to date from sponges, and KCNQ channels and the Shaker Kv4 subfamily are missing in placozoans (Li et al., 2015b).

Given that EAG channels have not been identified in choanoflagellates or other nonanimals (Jegla et al., 2024), it is likely that this family arose in the animal stem after the split of choanoflagellates and animals. However, the absence of EAG channels in protozoans should be revisited as more species are sequenced to rule out selective loss in choanoflagellates. CNBD superfamily cation channels (to which the EAG family belongs) are widespread across the tree of life (Schachtman et al., 1992; Sentenac et al., 1992; Jegla and Salkoff, 1994, 1995; Brams et al., 2014; Jegla et al., 2018; Pozdnyakov et al., 2020), but the relationships between these channels in the major eukaryotic lineages have not yet been defined (Jegla et al., 2018; Pozdnyakov et al., 2020). The ancestral lineage from which the KCNQ family evolved is also not clear because there are no orphan Kv channels in early animals that fall outside the EAG and Shaker lineages. It is possible KCNQs evolved from Shakers via loss of T1 because both share a domain-swapped VSD arrangement not found in EAGs (Long et al., 2005; Whicher and MacKinnon, 2016; Sun and MacKinnon, 2017), but there is no smoking gun of sequence homology to favor this origin. Alternatively, they might have evolved from a ghost lineage of Kv channels that was lost in the early diverging animal lineages. This would be analogous to the Shaker Kv2–4-like lineage that we can infer was present in the animal’s LCA only from its presence in choanoflagellates (Jegla et al., 2024), the protozoan sister clade to animals. Choanoflagellates do have two Kv channel lineages in addition to Shaker (Jegla et al., 2024), but these have not been phylogenetically compared with KCNQs, and there is as yet no evidence for their presence in animals.

We call the large ctenophore Shaker family Kv1-like here because, like the bilaterian/cnidarian Kv1 subfamily, they lack the T1 Zn²⁺-binding site shared between the animal Kv2–4 and choanoflagellate Shakers (Bixby et al., 1999; Jegla et al., 2024). Given our phylogeny, it is likely that these two subfamilies each independently lost the T1 Zn²⁺-binding site, and we therefore cannot be certain of the presence of Kv1-like channels in the animal’s LCA. Ctenophore Kv1-like channels span a broad range of biophysical properties comparable with the Shaker families of cnidarians and bilaterians, including functional orthologs of classic delayed rectifiers (MlShak6), rapid N-type–inactivating Shakers (MlShak4, 5, and 13), and suites of channels tuned for activity at hyperpolarized (MlShak4–6, 9, and 13) and depolarized (MlShak7 and 15) voltages, as found in both cnidarians and vertebrates (Jegla et al., 1995, 2012; Jegla and Salkoff, 1997; Li et al., 2015b). Because ctenophores and parahoxozoans share a broad array of biophysical phenotypes across their Shaker families but share only one ancestral Shaker gene lineage, many of these biophysical similarities are likely to have evolved independently.

We examined only half of the Mnemiopsis orthologs representing conserved ctenophore Shaker family lineages, so it is quite possible that the Shaker family is even more functionally diverse than we show here, both within Mnemiopsis and across the ctenophore lineages. Cnidarians have a similarly large complement of ancestral Shaker family channels, and while there is very limited data on the conservation of functional phenotypes between the major cnidarian lineages, it can range from near identity (in Shak1 and Shaw1 orthologs) to substantively different (in Shal1 orthologs) (Lara et al., 2023). Furthermore, we did not examine whether alternative splicing further increases the functional diversity of ctenophore Shakers. Some bilaterian Shaker family genes have been shown to undergo alternative splicing, most notably Drosophila Shaker, which has an alternative splicing of the N terminus and S6 region that alter inactivation (Schwarz et al., 1988; Timpe et al., 1988; Hoshi et al., 1990, 1991). We did not observe any examples of alternative exons in the conserved T1 + S1–S6 channel core in our BLAST searches of genome drafts, and only found a single gene, MlShak3, with alternative transcripts in transcriptome data sets in the form of four alternative N-termini (all four predicted proteins are included in Data S1). These transcripts could potentially alter inactivation rate as seen for Drosophila Shaker, but we did not test their functional significance here. Transcriptome datasets for Mnemiopsis are unlikely to be sufficiently comprehensive for identification of all transcript variants, but they are large enough to suggest that alternative splicing is not a major contributor to the overall functional diversity of Shaker channels in this species.

In both cnidarians and vertebrates, silent/regulatory subunits that require heteromeric assembly further diversify the functional properties of Shaker family channels. For example, the vertebrate silent/regulatory subunit Kv6.4 introduces closed state inactivation into Kv2 subfamily delayed rectifiers (Ottschytsch et al., 2002; Pisupati et al., 2018), and Kv4 subfamily silent/regulatory subunits in cnidarians diversify the voltage-activation range and inactivation kinetics of classic A-type currents (Jegla and Salkoff, 1997; Li et al., 2015b). In fact, more than half of all cnidarian Shaker family genes are likely to encode silent/regulatory subunits (Lara et al., 2023). Seven out of the 16 Mnemiopsis channels we expressed failed to form functional homomeric channels and this set of channels could include silent/regulatory subunits. In support of this, MlShak1 and MlShak2 do form functional heteromultimers when co-expressed with a cnidarian Kv1 subfamily channel (Li et al., 2015b). However, proving the silent/regulatory phenotype requires finding native assembly partners in ctenophores, a potentially extensive combinatorial expression process that we did not undertake here. Co-expressed channels in transcriptomic data, like the data we show here (Fig. 8 and Table S1) (Babonis et al., 2018; Sebé-Pedrós et al., 2018; Hehmeyer et al., 2024), are candidates for native partners. However, these data do not provide a comprehensive view of cell-level expression overlap, and it would therefore be premature to eliminate other candidates. Alternative explanations for the lack of current include gating requirements other than voltage changes (rare for Shakers) or significant errors in the coding ORF predictions that block homomeric expression. The latter possibility seems unlikely because our ORF predictions were supported by cross-species conservation. It is also highly unlikely that these non-expressers are remnant pseudogenes because they represent conserved ancestral lineages (Fig. 3) and all are expressed in transcriptome data (Fig. 8 and Table S1).

One of the key underpinnings of complex electrical signaling in cnidarians and bilaterians is that their eight conserved Kv channel families and subfamilies are functionally independent because they are not assembly compatible. This allows for the expression of up to eight functionally distinct versions of these channels in single neurons, allowing complex spatiotemporal patterning of electrical signals. For example, vertebrates use distinct suites of Kv channels in axons and dendrites (Trimmer, 2015). While ctenophores have large numbers of Kv channel genes, and transcriptome data suggest the potential for significant co-expression, it is unknown whether ctenophores are more limited in the number of independent channel types they can express per cell. The key to answering this question is to determine whether ctenophore Shakers are broadly assembly compatible or whether ctenophores have independently evolved multiple-assembly incompatible subfamilies equivalent to the Kv1–4 subfamilies in cnidarians and bilaterians. Interestingly, MlShak4 and MlShak5 are not assembly compatible (Simonson et al., 2024) and thus may represent distinct subfamilies. We did not attempt to define ctenophore Shaker subfamily boundaries here because, much like confirming silent/regulatory subunit phenotypes, this is an enormous job to address in a comprehensive way given the size of the ctenophore Shaker family. Molecular dynamics modeling of T1 assembly domain interactions can be predictive of assembly compatibility in Shaker channels (Jegla et al., 2024; Simonson et al., 2024) and could aid in subfamily prediction. Future studies to address this question will provide key insights into the neuronal signaling potential of ctenophores.

Data shown in Figs. 2, 3, 4, 5, and 6 are available in the article and online supplemental materials. Electrophysiology recordings can be obtained from the corresponding author upon reasonable request.

Christopher J. Lingle served as editor.

This research was supported by the Huck Institutes of the Life Sciences and the Eberly College of Science at Penn State University and an Allen Distinguished Investigator Award (to J.F. Ryan), a Paul G. Allen Frontiers Group advised grant of the Paul G. Allen Family Foundation.

Author contributions: B.T. Simonson: conceptualization, data curation, formal analysis, investigation, methodology, project administration, software, visualization, and writing—original draft, review, and editing. Z. Jiang: investigation. J.F. Ryan: funding acquisition, resources, validation, visualization, and writing—review and editing. T. Jegla: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, supervision, visualization, and writing—original draft, review, and editing.