A Na+ leak channel cloned from Trichoplax adhaerens extends extracellular pH and Ca2+ sensing for the DEG/ENaC family close to the base of Metazoa

Acid-sensitive ion channels belonging to the degenerin/epithelial sodium channel (DEG/ENaC) family activate in response to extracellular protons and are considered unique to deuterostomes. However, sensitivity to pH/protons is more widespread, where, for example, human ENaC Na+ leak channels are potentiated and mouse BASIC and Caenorhabditis elegans ACD-1 Na+ leak channels are blocked by extracellular protons. For many DEG/ENaC channels, extracellular Ca2+ ions modulate gating, and in some cases, the binding of protons and Ca2+ is interdependent. Here, we functionally characterize a DEG/ENaC channel from the early-diverging animal Trichoplax adhaerens, TadNaC6, that conducts Na+-selective leak currents in vitro sensitive to blockade by both extracellular protons and Ca2+. We determine that proton block is enhanced in low external Ca2+ concentration, whereas calcium block is enhanced in low external proton concentration, indicative of competitive binding of these two ligands to extracellular sites of the channel protein. TadNaC6 lacks most determinant residues for proton and Ca2+ sensitivity in other DEG/ENaC channels, and a mutation of one conserved residue (S353A) associated with Ca2+ block in rodent BASIC channels instead affected proton sensitivity, all indicative of independent evolution of H+ and Ca2+ sensitivity. Strikingly, TadNaC6 was potently activated by the general DEG/ENaC channel blocker amiloride, a rare feature only reported for the acid-activated channel ASIC3. The sequence and structural divergence of TadNaC6, coupled with its noncanonical functional features, provide unique opportunities for probing the proton, Ca2+, and amiloride regulation of DEG/ENaC channels and insight into the possible core-gating features of ancestral ion channels.

DEG/ENaC 2 ion channels comprise a large family of animalspecific Na ϩ channels that operate either as constitutively open Na ϩ leak channels or as transiently open channels that are activated and/or modulated by distinct but sometimes overlapping stimuli. For example, epithelial Na ϩ leak channels (i.e. ENaC channels), important for Na ϩ re-uptake and homeostasis, are constitutively open but modulated by extracellular protons, peptide hormones, bile acids, proteolytic cleavage, and mechanical stress (1)(2)(3). The vertebrate bile acid-sensitive channels (i.e. BASIC or ASIC5 channels) conduct leak Na ϩ currents that are enhanced by bile acids and blocked by external Ca 2ϩ ions (4,5). Acid-sensitive ion channels (ASICs), important for extracellular pH sensing, are potently and transiently activated by extracellular acidification/protons, but are dynamically modulated by extracellular Ca 2ϩ , mechanical stress, proteolytic cleavage, and neuropeptides (1,6,7). The peptide-gated FaNaC channels from gastropod molluscs, which are directly activated by the ligand neuropeptide FMRF-amide (8), are also inhibited by protons and Ca 2ϩ (9). In Caenorhabditis elegans, the mechanosensory degenerin (DEG) and MEC channels are activated by mechanical stimuli (10), whereas ACD-1 acts as a Na ϩ leak channel that is blocked by external protons and Ca 2ϩ (11). Indeed, the extensive diversity in DEG/ENaC channel gating, coupled with their independent duplication in distinct animal lineages and incomplete lineage sorting apparent on phylogenetic trees (12)(13)(14), makes it difficult to conceptualize what the core functional attributes of the primordial DEG/ENaC channel might have been like. To date, the most early-diverging DEG/ENaC channels to be functionally characterized are from the cnidarian Hydra magnipapillata, where they were found to be similar to FaNaC channels in their activation by neuropeptides (Hydra-RF-amides) and blocked by external Ca 2ϩ (13) but, in contrast, insensitive to protons (14). Notably, DEG/ ENaC channels are present in the genomes and transcriptomes of the three most early diverging animal phyla: Placozoa, Porifera, and Ctenophora (12), but their functional features have not yet been explored. Thus, studies on DEG/ENaC homologues from these early diverging animals will help develop a better understanding about the core characteristics and evolution of this large family of ion channels.
Among these, placozoans such as Trichoplax adhaerens are particularly interesting in that, despite lacking synapses and a nervous system, they express a nearly complete set of genes involved in electrical neural signaling (15), including most major families and classes of ion channels. Furthermore, Trichoplax is able to conduct motile behaviors, including feeding, chemotaxis, and geotaxis (16 -18), employing secreted neuropeptides for roles in coordinating cellular activity (19,20). Notably, with the exception of the voltage-gated T-type calcium channel (21), the functional properties of Trichoplax electrogenic genes have yet to be reported. Here, we report the cloning and functional expression of a second Trichoplax ion channel from the DEG/ ENaC family, named TadNaC6 for Trichoplax adhaerens Na ϩ channel subtype 6.
Phylogenetically, all but one of the Trichoplax DEG/ENaC channels form a sister clade with deuterostome pH-sensitive ASIC channels, whereas the homologues from protostome invertebrates such as arthropods, molluscs, and nematodes form separate clades. In situ hybridization of TadNaC6 revealed expression within rows of cells that line the periphery on the animal, a region bearing cells thought to be involved in cellular communication (20,22). In vitro expression of Tad-NaC6 in Chinese hamster ovary (CHO) cells, coupled with whole-cell patch-clamp recording, revealed robust Na ϩ -selective leak currents that were blocked by both external protons and Ca 2ϩ , a functional profile also reported for the mouse BASIC and nematode ACD-1 DEG/ENaC channels (5,11). Proton block of TadNaC6 was biphasic in the presence of 2 mM external Ca 2ϩ , but monophasic in 0.1 mM Ca 2ϩ , revealing an interplay between proton and Ca 2ϩ block. In accordance, proton block was enhanced in low Ca 2ϩ , while Ca 2ϩ block was enhanced in low [H ϩ ], indicative of competition for channel block by these two extracellular ligands. Notably, an S353A mutation, within a homologous site shown to be a key determinant for differential Ca 2ϩ sensitivity in rodent BASIC channels (5), had no effect on Ca 2ϩ block for TadNaC6, but instead, it slightly reduced its proton sensitivity. Indeed, given its phylogenetic distance from the functionally similar mouse BASIC and nematode ACD-1 Na ϩ leak channels, and the absence of key amino acids involved in proton and Ca 2ϩ sensitivity in other DEG/ENaC channels, it is likely that the gating features of TadNaC6 evolved independently. Nevertheless, our work extends extracellular protons and Ca 2ϩ as modulatory ligands for DEG/ENaC channels close to the base of Metazoa. Finally, TadNaC6 was found to be activated by the general DEG/ENaC channel blocker amiloride, a rare feature only reported for rat ASIC3 (23) and a mutant version of human ASIC2 (24), but lacks key residues attributed to this phenomenon. Instead, diminazene, another general DEG/ENaC channel blocker, produced a low-affinity block corroborating that the two drugs have fundamentally different modes of action. The highly-divergent sequence and atypical properties observed for Tad-NaC6 provide some unique perspectives on the structure, function, and evolution of proton-and Ca 2ϩ -sensitive DEG/ENaC channels.

T. adhaerens DEG/ENaC ion channels phylogenetically cluster with vertebrate ASIC channels and exhibit distinct spatial expression patterns
Previously, we reported the cloning of complete open reading frames for nine DEG/ENaC ion channel homologues identified in the transcriptome of the early-diverging animal T. adhaerens, while two others identified in the genome failed to amplify via RT-PCR from whole-animal total RNA. 3 A maximum likelihood phylogenetic tree of the Trichoplax channels, along with DEG/ENaC channel proteins from various representative animals, reveals that all but one of the 11 Trichoplax homologues form a sister clade with ASIC channels from vertebrates, whereas the BASIC channels (also known as ASIC5 or BLiNAC channels) form a separate node from these two groups (Fig. 1A). Similar to the nomenclature used to name the cnidarian (Hydra) DEG/ENaC Na ϩ channels (i.e. HyNaCs), we chose to refer to the Trichoplax channels as TadNaCs 1-11. Notably, no protostome invertebrate DEG/ENaC channels clustered between the Trichoplax and vertebrate ASIC clades, and generally, the phylogenetic positions of the various ion channels throughout the tree do not correspond with the metazoan phylogeny (Fig. 1A, inset). Whether this extensive incomplete lineage sorting is attributable to sequence divergence/convergence, or extensive gene gain/loss, is difficult to discern. However, it is notable that clades with strong bootstrap support tend to be populated by channels from a single species or phylum, consistent with proposed lineage-specific genetic expansion events for DEG/ENaC channels within the Metazoa (12). One exception is a node/clade that includes a TadNaC10, the FMRF-amide peptide-activated channels from gastropods (i.e. FaNaCs) (25)(26)(27), the vertebrate ENaCs (28), and the nematode mechanically-gated DEG/UNC/MEC/DEL channels (29). The gating modalities of this particular clade of channels are relatively well-understood and quite distinct from each other, underscoring that even closely-related channels at the protein sequence level can exhibit divergent functional profiles. The phylogenetic tree also depicts 14 DEG/ENaC ion channel homologues identified in an updated gene dataset for the sponge Amphimedon queenslandica (Fig. 1A), notable because these were previously reported absent (14). Finally, at least two distinct clades of ctenophore DEG/ENaC channels are apparent on the tree, none sharing strongly supported nodes with clades from other animals. Altogether, the DEG/ENaC ion channel family appears to have undergone extensive lineage-specific evolutionary change, a feature similarly observed for other gene families involved in sensory functions (30).
A recent single cell transcriptome study on T. adhaerens revealed that the DEG/ENaC channels TadNaC2 (TRIAD-DRAFT_58143), TadNaC3 (TRIADDRAFT_58141), and Tad-NaC8 (TRIADDRAFT_29260) are co-expressed in lipophil cells, which bear large acidophilic vesicles proposed to be secreted along the ventral epithelium for external digestion during feeding (22). A single channel, TadNaC10 was found to be expressed in ciliated epithelial cells, possibly the ventral cells involved in ciliary locomotion (31). Here, fluorescence in situ hybridization of TadNaC6, TadNaC5, and TadNaC2 revealed somewhat overlapping cellular expression, with TadNaC6 label being most prevalent in a narrow band at the edge, whereas the TadNaC5 label is abundant in a region extending ϳ100 m from the edge. TadNaC2 label is very sparse within 100 m of the edge but more prevalent farther in the interior (Fig. 1B), the zone occupied by lipophil cells.

TadNaC6 is a proton-inhibited cation leak channel in vitro, similar to vertebrate BASIC and nematode ACD-1 channels
We subcloned all nine cloned Trichoplax DEG/ENaC channel open reading frames into the mammalian expression vectors pIRES2-EGFP and pEGFP-C1 ( Fig. 2A), in order to attempt functional expression for in vitro electrophysiology. Although various expressed Trichoplax channels have produced functional currents in vitro, this study focuses solely on TadNaC6, which possesses some striking and unique functional attributes. The vector pEGFP-TadNaC6 permits ectopic expression of the channel with an N-terminal fusion tag of the green fluorescent protein (GFP), which was used to visualize successful channel expression in CHO cells via fluorescence microscopy (Fig. 2B), and as a full-length channel protein fused to GFP on Western blottings using polyclonal anti-GFP antibodies (Fig. 2C). Also evident is a fainter band on the blot with a molecular mass of roughly 200 kDa, absent in untransfected cell lysates, which possibly represents dimerized EGFP-TadNaC6 fusion proteins. Expression of the unfused TadNaC6 cDNA from the pTadNaC6 -IR-EGFP vector in CHO cells produced robust acid-sensitive depolarizing leak currents recorded under a whole-cell voltage clamp when the membrane voltage was held at Ϫ60 mV (Fig. 2D). Leak inward currents were considerably active upon perfusion with an extracellular saline of pH of 7.5, whereas an acidic external solution of pH 5 completely blocked channel currents, and pH 9 and 10 solutions elicited maximal inward currents. TadNaC6 thus forms a functional homomeric channel in vitro, with gating features similar to those of vertebrate BASIC and C. elegans ACD-1 channels, all conducting constitutive cation leak currents at rest that are not activated by extracellular protons (5,11). This is in contrast to the ASIC channels, which become transiently activated upon extracellular acidification (32,33). Furthermore, similar to ACD-1 and mouse BASIC, TadNaC6 is blocked by extracellular protons. To our knowledge, this is the first report of acid sensitivity for a DEG/ENaC channel from an early-diverging metazoan. T a d N a C 1 T a d N a C 3 T a d N a C 7 1 0 0 1 0 0 T a d N a C 1 1 9 9 T a d N a C 9 10 0 1 0 0 9 9 T a d N a C 8 9 9 A p A SI C b Sp  P P K 1 6 P P K 1 1 1 0 0 5 7 P P K 2 5 P P K 9 7 9 P P K 2 3 P P K 1 0 8 4 9 4 P P K 1 3 P P K 1 5

In vitro properties of a placozoan DEG/ENaC ion channel
Instead, the cloned channels from the cnidarian H. magnipapillata, which form a clade with TadNaCs and vertebrate ASIC and BASIC channels (Fig. 1A), were reported insensitive to pH and activated by RF-amide neuropeptides in vitro (14,34).
Evidently, the kinetics of TadNaC6 macroscopic currents change with pH, with desensitization becoming apparent after transition from a pH 5 to a pH 8.5 external solution, and more pronounced when transitioning from pH 5 to pH 10 ( Fig. 2D). In this regard, relief from proton binding/inhibition, which serves to activate and then desensitize TadNaC6, appears similar to proton activation of the ASIC channels. Thus, it appears as though activation and subsequent desensitization of Tad-NaC6 are strictly dependent on transition between singlechannel states, caused by dissociation of protons. We note that the current onset at pH 7 and 8 reflects slower channel activation compared with the intermediate pH of 7.5, as well as the more basic pH conditions Ն8.5. The slow activation kinetics at pH 8 mirror the steady-state current, where proton block of TadNaC6 is biphasic through this pH range, comprising a highaffinity proton block with an average IC 50 pH of 8.5 Ϯ 0.04 (plus or minus standard error, n ϭ 9) and a low-affinity block with an IC 50 of pH 7.12 Ϯ 0.03 (Fig. 2E). Therefore, under these exper-imental conditions having external solutions bearing 140 mM [Na ϩ ] and 2 mM [Ca 2ϩ ], TadNaC6 appears to have at least two extracellular binding sites for protons that affect both gating and activation kinetics.

TadNaC6 bears hallmark structural features of the DEG/ENaC superfamily
Structural modeling of TadNaC6, as well as the other Tad-NaC channel subunits, predicts conserved transmembrane helices that combine with helices from two additional subunits in the trimeric channel complex to form a functional six-helix ion-conducting pore ( Fig. 3 and Fig. S1). In addition, the Trichoplax channels all are predicted to possess the six hallmark extracellular "ball-in-hand" structures of the DEG/ENaC channel family: 1) a palm domain made up of an array of closelypacked ␤-strands; 2) a wrist domain, which connects the palm domain to the transmembrane helices; 3) a thumb domain, important for channel ligand gating and made up of two tightly disulfide-linked ␣-helices; 4) a finger domain, comprising three ␣-helices and thought to regulate gating via interactions with the thumb domain; 5) a ␤-ball, made up of four ␤-strands; and 6) a knuckle domain, made up of two closely-linked ␣-helices  (35). Models of TadNaC3, TadNaC6, and TadNaC7 lack predictions for one of the three ␣-helices in the finger domain (helix 2), similar to models of the vertebrate BASIC channels that bear a reduced or absent helix in this region ( Fig. 3 and Fig.  S2). In contrast, TadNaC2, TadNaC5, TadNaC8, TadNaC9, and TadNaC10 all bear three predicted finger helices, whereas TadNaC2, TadNaC9, and TadNaC10 bear an additional predicted helix between the thumb and wrist domains, absent in the other analyzed channels except for nematode ACD-1. Notably, the finger domain is one of the most variable extracellular regions among the different classes of the DEG/ENaC channels (35). In accordance, a protein alignment reveals poor sequence homology in this region and, furthermore, a variability in sequence length with respective gaps in the alignment of 26 for TadNaC6, 24 for TadNaC5, 23 for BASICs and ASIC3, as compared with the longer sequences for ASIC1 (only 16 gaps) and ACD-1 (10 gaps; Fig. 4). Also notable is that most TadNaCs bear insertions between the two knuckle helices, predicted to form loops that are absent in models of all vertebrate ASIC and BASIC channels, and C. elegans ACD-1 ( Fig. 3 and Figs. S1 and S2).

In vitro properties of a placozoan DEG/ENaC ion channel
Consistent with an independent radiation of DEG/ENaC channels within the Placozoa, aligned Trichoplax TadNaC6 and TadNaC5 proteins share more sequence features with each other than they do with homologues from vertebrates and C. elegans (Fig. 4). Nevertheless, both Trichoplax channels share with the other channel predictions for two conserved transmembrane helices, the second bearing more sequence homology, including the canonical "GXS" selectivity filter motifs crucial for Na ϩ selectivity and, overall, similar predicted secondary structures throughout the alignment (Fig. 4). Extracellular cysteine residues, which are important for stabilizing extracellular secondary structures, are particularly conserved, and of the 33 identical amino acids that span the alignment between transmembrane helices 1 and 2, 11 (33%) are cysteines. Glycines, which promote formation of loops between secondary structures, are also highly conserved at six positions (i.e. 18%), whereas 11 of the remaining 16 amino acids are hydrophobic (ϳ69%). Finally, we note the absence of key residues in TadNaC6 that are conserved in deuterostome ASIC channels, which are crucial for proton activation, including a histidine located between transmembrane helix 1 and ␤-strand 1, and a lysine located in the loop between ␤-strands 5 and 6 (36,37).

Ion selectivity profiling of TadNaC6 reveals a moderate preference for Na ؉ over K ؉ ions, similar to other DEG/ENaC channels
DEG/ENaC ion channels are highly to moderately selective for Na ϩ over K ϩ ions (38). To evaluate the ion selectivity of TadNaC6, we conducted bi-ionic reversal-potential experiments using 150 mM [Na ϩ ] in the internal solution and external perfusion of solutions bearing either 150 mM [Li ϩ ], 150 mM [Na ϩ ], 150 mM [K ϩ ], or 150 mM [Cs ϩ ] (Fig. 5). A voltage step to Ϫ100 mV produced the largest amplitude inward current when Na ϩ was present in the external environment and was sequentially decreased in amplitude upon replacement of Na ϩ with external Li ϩ , K ϩ , and Cs ϩ . Sequential voltage steps between Ϫ100 and ϩ80 mV allowed determination of the macroscopic current reversal potential for each bi-ionic condition, which is the voltage at which inward X ϩ ion currents reverse to outward Na ϩ currents, determined by the pore's preference for X ϩ ion over Na ϩ . Not surprisingly, equimolar concentrations of Na ϩ across the cell membrane produced a reversal potential close to 0 mV (Ϫ5.6 Ϯ 0.5 mV, n ϭ 8), with a slight outward rectification at voltages more positive than ϩ20 mV (Fig. 5B). In these experiments, a rightward shift in reversal potential relative to exter- Confidence in the models is shown as follows: 403 residues (82%) modeled at Ͼ90% accuracy for TadNaC6; 440 residues (89%) modeled at Ͼ90% accuracy for MmuASIC5; 403 residues (74%) modeled at Ͼ90% accuracy for TadNaC5; and 415 residues (79%) modeled at Ͼ90% accuracy for MmuASIC1.

In vitro properties of a placozoan DEG/ENaC ion channel
nal Na ϩ reflects an increased permeability for the X ϩ ion relative to Na ϩ , whereas a leftward shift indicates decreased permeability. Despite external Li ϩ producing smaller amplitude inward and outward macroscopic currents compared with external Na ϩ , the reversal potential for Li ϩ is about 7 mV shifted toward the right, similar to what was reported for the vertebrate ASIC1 channel (39). Comparison of the Na ϩ and Li ϩ reversal potentials with the Goldman-Hodgkin-Katz equation : : . : .: *. ::* : : * . .*.::**::*: * *.:*: *: : : Na + S353A Figure 4. MUSCLE protein alignment of TadNaC6 and TadNaC5 with C. elegans ACD-1, ASIC, and BASIC channels from human (Hsa), mouse (Mmu), rat (Rno), and rat ASIC3. Black arrow heads denote amino acids positions previously shown to be important for proton activation of ASIC channels (36,37,64,70,82,83), and gray arrow heads amino acid positions involved in proton-inhibition of ACD-1 (11). Arrow heads with black outlines denote residues associated with Ca 2ϩ block of ASIC channels (orange fill) (66,67) and BASIC channels (no fill) (5). Pink arrow heads denote amino acids previously shown to be required for the activation of rat ASIC3 by amiloride (23); purple arrow heads denote glycine amino acids in ASIC2 that rendered the channel amiloride-activated upon mutation to valine (G430V) and cysteine (G437C) (24); and green arrow heads denote amino acids associated with sensitivity to block by diminazene (75). The bold pink residues in transmembrane helix 2 denote the degenerin locus (74). Cysteines, which can form extracellular disulfide bridges, are shown in bold orange. Colored underlines represent amino acid sequences predicted with Phyre2 to form ␣-helices (green) and ␤-strands (blue), and black lines denote transmembrane helices (TM1 and TM2).

In vitro properties of a placozoan DEG/ENaC ion channel
reveals a slight preference for Li ϩ over Na ϩ (P Li /P Na ϭ 1.46 Ϯ 0.07, n ϭ 6; Fig. 5C). However, the smaller amplitude macroscopic currents with Li ϩ in the external environment suggest that although this smaller radius ion is better at occupying the pore, its conductance through the pore is not as efficient as it is for Na ϩ , similar to ASIC1 (39). In summary, TadNaC6 is similar to other DEG/ENaC channels, exhibiting a roughly 7-fold preference for Na ϩ over K ϩ (P K /P Na ϭ 0.14 Ϯ 0.02, n ϭ 5), and a 29-fold preference for Na ϩ over the large radius Cs ϩ ion (P Cs / P Na ϭ 0.03 Ϯ 0.00, n ϭ 4).

Ca 2؉ block experiments uncover an interplay between proton and Ca 2؉ sensitivity for TadNaC6 and a unique residue involved in proton block
Many DEG/ENaC channels, including ASIC and BASIC channels and nematode ACD-1, are blocked by external Ca 2ϩ ions. Notably, the highly homologous BASIC channels from mouse and rat differ in their sensitivity to Ca 2ϩ when expressed in Xenopus oocytes, where physiological external [Ca 2ϩ ] of ϳ2 mM almost completely blocks rat BASIC, but has little effect on the mouse channel (5). However, this difference was not observed when the channels were expressed in the human cell line HEK293 (4). Nevertheless, given the similar functional properties of TadNaC6 compared with BASIC channels, we thought it instructive to assess its sensitivity to external [Ca 2ϩ ]. Increasing external [Ca 2ϩ ] from 0.1 to 2 mM roughly halved leak inward current amplitude for TadNaC6 at Ϫ60 mV, which could be completely blocked by perfusion of a pH 5 solution (Fig. 6A). Thus, Ca 2ϩ is an external blocker of TadNaC6. To generate a dose-response curve, we ran a voltage ramp protocol on patched cells while perfusing external solutions containing increasing [Ca 2ϩ ] from 1 M to 30 mM (Fig. 6B)

In vitro properties of a placozoan DEG/ENaC ion channel
cellular channel surface. In this regard, TadNaC6 is different than ACD-1, for which proton block was shown to be insensitive to [Ca 2ϩ ] ext (11). Notably, the differences in external Ca 2ϩ sensitivity between rat and mouse BASIC channels expressed in oocytes was attributed to a single amino acid, where an alanine to serine mutation at this position caused the rat channel to lose its micromolar sensitivity to calcium. Indeed, all TadNaC channels, including TadNaC6 and TadNaC5, bear a serine at this position similar to the less calcium-sensitive mouse ASIC5 (Fig. 4). We therefore reasoned that the reverse mutation, serine to alanine, might render TadNaC6 more sensitive to Ca 2ϩ and perhaps alter the differences in proton block observed in high versus low [Ca 2ϩ ] ext . At pH 8.5, increasing [Ca 2ϩ ] ext from 0.1 to 2 mM blocked peak inward current at Ϫ60 mV by 67.94 Ϯ 1.36% S.E. for the WT channel (n ϭ 6) and 62.93 Ϯ 3.42% S.E. for the S353A mutant (n ϭ 4) (Fig. 6E). At the more acidic pH of 7.5, both channel variants were less sensitive to Ca 2ϩ , with respective average block in response to increased calcium of 47

In vitro properties of a placozoan DEG/ENaC ion channel
pH range examined here, a feature similarly reported for the ASIC1 channel (40) that once again reveals an antagonistic interplay between proton block and Ca 2ϩ block for TadNaC6.

Amiloride, a general DEG/ENaC blocker, is a strong agonist of TadNaC6
DEG/ENaC ion channels are often referred to as amiloridesensitive sodium channels, attributable to their near-ubiquitous sensitivity to blocking by this drug. Interestingly, the rat ASIC3 channel was found to be directly activated by amiloride, where application of the drug produced large, nondesensitizing inward currents at a physiological pH of 7.4, quite distinct from the rapidly activating and desensitizing proton-activated currents observed at pH 5 (23). This atypical sensitivity to amiloride was attributed to two glutamate amino acids, one adjacent to transmembrane helix 1, and the second preceding transmembrane helix 2 (Fig. 4), since mutation of either of these two residues to alanine nearly abolished amiloride activation. At these same positions, TadNaC6 bears a lysine and threonine, respectively. The human ASIC2 channel, which is normally blocked by amiloride, also becomes amiloride-activated upon mutation of two glycines within transmembrane helix 2, G430V and G437C (24). Notably, although Gly-437 appears to be conserved even in the Trichoplax channels and ACD-1, the Gly-430 position bears a serine for TadNaC6, TadNaC5, and ACD-1 and an alanine for vertebrate BASIC channels (Fig. 4).
Perfusion from a pH 5 solution (where all channels would be closed) to pH 7.5 solutions bearing 10 -300 M amiloride produced sustained nondesensitizing inward currents that increased slightly with increasing concentrations of the drug (Fig. 7). However, transition from pH 5 to 7.5 would activate TadNaC6 currents such that the effect of amiloride is masked by channel activation elicited by relief from the proton block. Nevertheless, perfusion of 1 mM amiloride at pH 7.5 produced large amplitude inward currents, considerably larger than those elicited by pH 9 solution without amiloride. Furthermore, 3 mM amiloride produced even larger TadNaC6 currents, roughly 12-fold greater than those elicited by pH 9. Thus, similar to ASIC3, amiloride is a nonproton ligand of TadNaC6. Unlike amiloride-activated currents for ASIC3, TadNaC6 currents

In vitro properties of a placozoan DEG/ENaC ion channel
were marked with a moderate desensitization component (Fig.  7A), resembling amiloride-activated currents of the mutant human ASIC2 channel (24). Instead, TadNaC6 currents were weakly blocked by the phenylhydrazine compound diminazene (Fig. 7C), another general blocker of DEG/ENaC channels, with an IC 50 of 578.1 Ϯ 69.4 M. The HyNaC channels from Hydra were found to be much more sensitive to diminazene, with IC 50 values ranging from 0.05 to 31.4 M (13). Similarly, vertebrate ASIC and BASIC channels are also blocked by micromolar concentrations of diminazene, with reported IC 50 values ranging from 2 to 16.8 M (4, 33). Thus, whereas diminazene blocks TadNaC6 (albeit with a low affinity), amiloride acts as a potent agonist, indicating that the two compounds elicit different allosteric changes in the channel's structure. Furthermore, the lack of conserved amino acid determinants associated with amiloride activation suggests that different mechanisms are responsible for this phenomenon in TadNaC6.

On the phylogenetic relationships of DEG/ENaC channels
Here, we describe the in vitro characteristics of TadNaC6, one of 11 DEG/ENaC ion channel subunit genes identified for the early-diverging animal T. adhaerens. Our phylogenetic analysis corroborates previous reports that all but one of the Trichoplax channels form a sister clade relationship with the acid-activated ASIC channels from chordates (Fig. 1A). Instead, the vertebrate sodium leak channels BASIC and ENaC are each phylogenetically separate, whereas the remaining Trichoplax channel, TadNaC10, forms a clade with ENaCs, mollusc FaNaCs, and nematode DEG/MEC/DEL channels. Two prominent features of phylogenetic trees inferred from DEG/ENaC channel protein alignments are as follows: 1) extensive incomplete lineage sorting, where phylogenetic relationships of channel homologues do not correspond with the species phylogeny, and 2) frequent and independent expansion of DEG/ENaC channels within distinct animal groups. With respect to feature 1, it is striking that the major clade of Trichoplax channels is not separated from vertebrate ASIC channels by homologues from protostome invertebrates such as arthropods, nematodes, or molluscs, although they appear broadly in the deuterostomes such as early-diverging chordates (i.e. Branchiostoma belcheri and Oikopleura dioica), hemichordates (Ptychodera flava), and echinoderms (Strongylocentrotus purpuratus and Acanthaster planci) (Fig. 1A) (36). Thus, either the ancestor of all protostome invertebrates lost an ASIC-like homologue, there was lateral gene transfer, or the Trichoplax and chordate channels converged at the amino acid sequence level. With respect to feature 2, it is notable that the majority of clades bearing strong bootstrap support consists of channels from a single phylum/ species, consistent with the notion that the DEG/ENaC channel family has undergone extensive independent expansion within different animal lineages (12). Interestingly, similar expansion events are evident for transient receptor potential channels (41) and ionotropic glutamate receptors (42, 43), which like DEG/ ENaC channels were frequently adapted for sensory functions. Extensive gene gain/loss is also evident for chemosensory G-protein-coupled receptors, where it has been proposed that a combination of adaptation to different sensory environments, coupled with random gene gain/loss, accounts for observed phylogenetic differences (44).
In situ hybridization of TadNaC6 revealed mRNA expression enriched along the periphery of the animal (Fig. 1B), a region occupied by cells that secrete mucus and label for the endomorphin-like peptide (TaELP) that causes Trichoplax ciliary locomotion to cease (19,45). TadNaC5 was expressed near the outer edge of the animal; however, its expression reached deeper into the center, whereas TadNaC2 appeared absent along the outer rim but more prevalent in the interior, a region enriched with lipophil cells (45). The localization of TadNaC2 is consistent with the single-cell transcriptome study revealing co-expression of TadNaC2, TadNaC3, and TadNaC8 in digestive/lipophil cells (31). Indeed, the sensitivity of TadNaC6 to external protons begs the following question: when might the animal experience change in extracellular pH along its periphery? One obvious possibility is during feeding, where lipophil cells in the ventral epithelium are proposed to secrete large acidic vesicles that contain digestive enzymes into the extracellular environment upon detection of food microalgae, for the purpose of hydrolytic breakdown and digestion (16). Given the small volume between the ventral epithelium and the underlying substrate, it is conceivable that the contents of these secreted vesicles temporarily but significantly lower extracellular pH. By extension, cells expressing TadNaC6 would undergo membrane hyperpolarization due to proton block of the channel. Alternatively, protons could serve modulatory functions for yet undiscovered ligands for TadNaC6, such as secreted peptides, bile-like compounds, mechanical stress, and/or chemicals in the environment. Clearly, additional studies on this and other Trichoplax DEG/ENaC channels are needed to further characterize their intrinsic functional properties and furthermore to elucidate their roles in physiology and behavior.

Structural and functional properties of TadNaC6
Modeling the tertiary structures of the nine cloned Trichoplax DEG/ENaC channels revealed predicted extracellular "ballin-hand" structures common to all DEG/ENaC channels ( Fig. 3 and Fig. S1). Although these predictions are to be interpreted lightly, it is interesting that the Trichoplax channels distinguish themselves from the vertebrate ASIC and BASIC channels, as well as the nematode ACD-1 channel, with notable insertions between the two ␣-helices that make up the knuckle region (Figs. 3 and 4 and Fig. S1). Based on the crystal structure of the functional trimeric chick ASIC1 channel, the knuckle domains from three separate subunits come together at the extracellular face to give rise to an upper vestibule of the channel pore (46). For ENaC channels, the knuckle domain has been implicated in sodium self-inhibition (47), indicating that structural changes in this region have the capacity to regulate DEG/ENaC channel gating. Another region with notable variability is the finger domain, where one of the three expected ␣-helices was not predicted for TadNaC3, TadNaC4, TadNaC6, and TadNaC7 (i.e. helix 2). It is interesting that the vertebrate ASIC and BASIC channels also differ in this region, with ASIC channels bearing strong predictions for all three helices, while the BASIC channels lack predictions for helix 2. Indeed, the finger domain

In vitro properties of a placozoan DEG/ENaC ion channel
has been flagged as the most divergent extracellular region for DEG/ENaC channels and hence a proposed locus for functional variability between different channel classes (35).
Functionally, TadNaC6 most resembles mouse BASIC and C. elegans ACD-1, conducting constitutive Na ϩ leak currents at resting membrane voltages that are blocked by external H ϩ and Ca 2ϩ ions (Figs. 2 and 6) (4, 5, 11). Its ion preference is similar to that of most other DEG/ENaC channels, being roughly 7-fold selective for Na ϩ over K ϩ (Fig. 5). Accordingly, TadNaC6 bears a GAS selectivity filter motif in transmembrane helix 2 and a key glutamate just downstream of that, both important for Na ϩ selectivity (Fig. 4) (38, 48). Other reported DEG/ENaC Na ϩ leak channels include ENaC and the human/rat BASIC channels, which differ in that extracellular protons either potentiate (ENaC) or do not affect (BASICs) their macroscopic currents (49 -51). ENaC is also considerably more selective for Na ϩ compared with other channels, with a P Na /P K ratio of roughly 100 (52). In Drosophila, the DEG/ENaC channels ripped pocket (RPK) and pickpocket (PPK), associated with sensory functions are Na ϩ leak channels, of which PPK1 was shown to be activated by extracellular protons (53,54). Phylogenetically, it would appear as though TadNaC6, ENaC, BASIC, ACD-1, and PPK/RPK each evolved the capacity to conduct constitutive Na ϩ leak currents independently, because they are separated by clades of channels that exhibit transient activation in response to various stimuli (Fig. 1A). For example, although TadNaC6 and BASIC channels sit relatively close to each other on the phylogenetic tree, they are separated by the proton-activated ASIC channels, as well as several acid-activated Trichoplax channels that we are currently characterizing in the laboratory. Thus, assuming that the apparent phylogenetic relationships within this group are true, the ancestral channel was either (i) a transiently-activatable channel, such that constitutively open leak channels evolved independently in BASICs and TadNaC6, or alternatively, (ii) the ancestral channel was a leak channel, such that proton activation evolved independently in ASICs and the Trichoplax acid-activated channels. More phylogenetically distant to TadNaC6 are the vertebrate ENaC channels, which are nestled within a group of transiently activated mechanosensitive and neuropeptide-gated channels, and unlike TadNaC6, BASICs, and ACD-1, ENaCs do not form functional homotrimers but instead operate as obligate heterotrimers of ␣, ␤, and ␥ subunits (55). Finally, ACD-1 and RPK/PPK channels are located within separate clades with no clear phylogenetic link to TadNaC6 and other Na ϩ leak channels. Interestingly, ASIC channels have recently been proposed to exhibit two distinct conductance states (56), a fast-desensitizing conductance triggered by extracellular protons, and a slow-nondesensitizing conductance triggered by nonproton ligands such as the coral snake venom MitTx (57), neuropeptides (58,59), and the polyamine 2-guanidine-4-methylquinazoline (GMQ) (60). Similarly, GMQ potently activates the mollusc FaNaC channels through a distinct mechanism than their natural ligand FMRF-amide (9), pointing to a conserved and ancient duality in DEG/ENaC channel gating. Considering this, the apparent homoplasy in Na ϩ leak current activity for phylogenetically distant members of the superfamily might instead be attributable to adaptive changes that independently enhanced sustained conductance properties among distinct channel lineages.

General absence of determinant residues associated with proton sensitivity
Initially, it was thought that proton-activated ASIC channels evolved strictly in the bony fish (61). However, a subsequent in vitro study on the shark homologue ASIC1b revealed proton activation, attributed to a conserved set of histidine residues adjacent to transmembrane helix 1 (Fig. 4), where similar to rat ASIC1a, mutation completely disrupted proton sensitivity (62). Nevertheless, acid activation was not observed for the cloned ASIC1 channel from lamprey (61) nor a tunicate ASIC-like homologue (63), suggesting proton activation is unique to vertebrates. More recently, a comprehensive study determined that acid-activated channels exist more broadly, where specifically cloned homologues from various basal deuterostome animals conducted proton-activated currents in vitro (36). Indeed, numerous amino acids have been implicated in proton activation of ASIC channels (Fig. 4); however, this recent study perhaps identified the most crucial and conserved: 1) one of the histidines adjacent to transmembrane 1 mentioned above (i.e. His-73 in mouse ASIC1a), and 2) a lysine located in a linker between the ␤-ball and palm domains (Lys-211). In accordance, mutation of either of these two residues in the divergent deuterostome channels severely disrupted or completely abolished proton sensitivity (36). The question whether the mechanisms for proton activation of ASIC channels relates to proton block of other channels was addressed for ACD-1. Notably, the proton block was found to involve three amino acids (11), only one of which, an aspartate located in the thumb domain, is conserved with proton-sensitive acidic residues in ASIC channels (Fig. 4). Indeed, outside of an additional conserved tryptophan at the base of the thumb domain that is thought to interact with a tyrosine in transmembrane helix 1 of ASIC1 channels (64), ACD-1 lacks most amino acids implicated in proton sensitivity for ASIC channels. Hence, the mechanisms by which protons affect the gating of this particular channel are quite distinct and likely independently evolved from those of ASIC channels. Similarly, TadNaC6, which is considerably sensitive to protons, lacks all but two amino acids associated with proton sensitivity in other channels, sharing none with ACD-1. As mentioned above, the mouse BASIC channel is also blocked by protons; however, it is much less sensitive than ACD-1 and TadNaC6, where the highly-acidic extracellular pH of 4.0 only reduces the amplitude of macroscopic current by about 40% (5). To our knowledge, the amino acids associated with proton block of mouse BASIC have not yet been identified. Nevertheless, given the general lack of conservation of amino acid residues associated with proton sensitivity among these phylogenetically distant Na ϩ leak channels, it seems likely that they evolved proton sensitivity independently from each other, as well as ASIC channels.

In vitro properties of a placozoan DEG/ENaC ion channel Interplay between Ca 2؉ block and proton block
An early study into the mechanisms for proton activation of ASIC channels (i.e. rat ASIC3) identified an interplay between external protons and Ca 2ϩ , where reducing external Ca 2ϩ allowed protons to more potently activate the channel (65). Furthermore, removal of external Ca 2ϩ , at a constant pH, directly activated ASIC3, indicative of the block of ASIC currents by external Ca 2ϩ ions at physiological concentrations. These observations led to the proposal that proton activation of ASIC channels involves displacement of blocking Ca 2ϩ ions by protons that effectively reduce the binding affinity of Ca 2ϩ to a site on the extracellular surface of the channel. However, unlike ASIC3, Ca 2ϩ depletion does not significantly activate rat ASIC1a (66). Furthermore, the mechanisms for proton and Ca 2ϩ sensitivity for this channel appear to be functionally distinct, because mutation of two amino acids in transmembrane helix 2, Glu-425 and Asp-432 (Fig. 4), completely abolished Ca 2ϩ block but did not disrupt proton activation nor create constitutively open channels, as would be expected based on the proposed Ca 2ϩ displacement model of ASIC channel gating (66). A subsequent study identified an additional glutamate residue in transmembrane helix 2, unique to ASIC3, that accounts for its unique activation in response to Ca 2ϩ depletion (Fig. 4), where a corresponding mutation in the rat ASIC1a channel (G429E) rendered it activatable by Ca 2ϩ depletion similar to ASIC3 (67). Thus, it appears as though ASIC3 is unique among ASIC channels in its ability to become activated by external Ca 2ϩ depletion. Nevertheless, it is interesting to note that currents elicited by Ca 2ϩ depletion at a fixed pH of 8.0 where sustained and nondesensitizing (65), indicating that at basic pH, ASIC3 conducts constitutive Na ϩ leak currents that are sensitive to Ca 2ϩ block, similar to those of TadNaC6, ACD-1, and BASIC channels. Also interesting is that the aforementioned glutamate in transmembrane helix 2 of ASIC3 is also present in BASIC channels (Fig. 4); however, whether this residue is involved in Ca 2ϩ block for BASIC channels has not been explored, to our knowledge.
As mentioned above, in vitro studies on rodent BASIC channels expressed in Xenopus oocytes revealed a differential sensitivity to Ca 2ϩ , where the rat channel was almost completely blocked by 1.8 mM [Ca 2ϩ ], while the mouse channel was minimally blocked at this same concentration (5). Instead, in a more recent study where the rat, mouse, and human BASIC channels were expressed in HEK293 cells, all three homologues were found to be highly sensitive to Ca 2ϩ block, indicating that Ca 2ϩ sensitivity might depend on the properties of the membrane environment that are likely different between mammalian cells and frog oocytes (4). In accordance, exploration into the mechanisms by which bile acids activate BASIC channels has revealed that the mode of action of these and other membraneactive compounds is likely through alterations in membrane environment, and not necessarily binding to channels per se (68). Nevertheless, we sought to determine whether the single serine amino acid associated with reduced Ca 2ϩ sensitivity in mouse BASIC was involved in the low Ca 2ϩ sensitivity observed for TadNaC6. Our hypothesis was that mutation of this residue to alanine (S353A), making it similar to rat BASIC, would increase its sensitivity to Ca 2ϩ block. We saw no change in Ca 2ϩ block at pH 7.5 and 8.5 for the S353A mutant compared with WT, but instead, we observed a reduced sensitivity to proton block in both low and near physiological Ca 2ϩ (Fig. 6E). Indeed, this study was not exhaustive, and the mechanisms by which Ca 2ϩ and H ϩ ions bind to the TadNaC6 channel to regulate its gating need to be explored further. However, it can be stated that similar to proton sensitivity, it is likely that the mechanisms for Ca 2ϩ sensitivity of TadNaC6 are highly divergent from other channels, because it lacks key amino acid determinants identified for other channels (Fig. 4).
Comparing the Ca 2ϩ block of various Ca 2ϩ -sensitive Na ϩ leak DEG/ENaC channels reveals that TadNaC6 is the less sensitive to Ca 2ϩ than BASIC channels expressed in mammalian cells (4), and it is similar to mouse BASIC and nematode ACD-1 expressed in oocytes (11). Although TadNaC6 was expressed strictly in mammalian cells, making it difficult to derive conclusions based on these comparisons, it is worth noting that the Trichoplax channel is adapted to a marine extracellular saline environment, in which [Ca 2ϩ ] is roughly 10 mM. At this concentration, Ca 2ϩ ions might therefore cause TadNaC6 to remain predominantly in a closed/blocked state under physiological conditions, similar to what might be expected for BASIC channels. Something yet to be determined is whether Ca 2ϩ modulation of TadNaC6, BASIC, and ACD-1 channels serves a function in vivo. For ASIC channels, lactate released during muscle stress reduces the concentration of divalent cations in the extracellular environment and potentiates ASIC channel sensitivity to protons in ischemia-sensing neurons (69). Mechanistically, the interplay between Ca 2ϩ block and proton block for TadNaC6 appears to be quite unique. At 2 mM external Ca 2ϩ , proton block was biphasic, suggesting the existence of two distinct proton-binding sites (Fig. 2E), which to our knowledge has never been reported for a WT DEG/ENaC channel. Furthermore, the activation kinetics of the channel are slower than expected at pH 8 compared with pH 7.5 and 8.5, where, possibly, the biphasic dose-response observed for steady-state currents might be due in part to slowed activation of the macroscopic current when transitioning from pH 5 to 8 and/or accelerated activation from pH 5 to 7. Previously, several mutations of acidic residues in ASIC1a (E79K/E79Q, D345K/D345R, and E416K; Fig. 4) were found to cause the channel's normally monophasic proton activation curve to become biphasic, which the authors proposed reflects the existence of two distinct proton-binding sites that become apparent through mutagenesis (70). What is particularly notable is that one of these mutants, with a negatively-charged aspartate changed to a positivelycharged lysine (E79K), resembles TadNaC6 bearing an arginine in this position. For TadNaC6, a reduction of external Ca 2ϩ caused proton block to resolve to a monophasic response (Fig.  6D); however, whether this occurs for the mutant ASIC channels was not addressed. Regardless, as noted above TadNaC6 lacks most residues associated with proton sensitivity in ASIC channels, and thus it is likely that the occurrence of two distinct proton-binding sites in ASIC and TadNaC6 channels emerged through convergence. Similarly, the proton and Ca 2ϩ -gating dynamics of TadNaC6 are likely fundamentally distinct from those of ACD-1, because proton block of the nematode channel

In vitro properties of a placozoan DEG/ENaC ion channel
was found to be completely independent of external [Ca 2ϩ ] (11).

Atypical pharmacology of the TadNaC6 channel
The drug amiloride blocks ENaC channels with high affinity (IC 50 of ϳ100 nM), hence its clinical use for high blood pressure where it causes decreased Na ϩ reabsorption by renal epithelial cells and increased excretion in the urine (71,72). Amiloride also blocks ASIC channels and, in fact, most DEG/ENaC channels (73), such that the DEG/ENaC channel family is often alternatively referred to as the amiloride-sensitive ion channels. Interestingly, a study on a "degenerin" mutant of the human ASIC2a channel (i.e. G430V), a locus discovered in nematodes that when mutated renders DEG/ENaC channels constitutively open (Fig. 4) (74), revealed that introducing a second mutation adjacent to the selectivity filter (G437C) rendered the channel activated by amiloride rather than blocked (24). The authors proposed that ASIC channels might bear two distinct amiloride-binding sites, one in the pore where the drug blocks and another within an undiscovered location. The proposed model was thus that the G430V/G437C mutations disrupted amiloride binding in the pore and hence blocked and unmasked the capacity of amiloride to bind the second locus to trigger channel activation. A subsequent study on the WT rat ASIC3 channel revealed that it could be potently activated by amiloride, attributed to two glutamate residues in the so-called "nonproton ligand-sensing domain" (Glu-79 and Glu-423), located between the palm and wrist domains (60). Nevertheless, ASIC1 channels possess these same amino acids (Fig. 4) but are not activated by amiloride, indicating that additional structural features are involved. To our knowledge, whether there is an interplay between the G430V/G437C mutations in ASIC2a, and the Glu-79/Glu-432 residues in ASIC3, all conserved between ASIC1, ASIC2, and ASIC3 channels, remains unexplored. Notably, the Glu-79/Glu-432 residues were also shown to be involved in activation of ASIC3 by the nonproton agonist GMQ, which, like amiloride, generates nondesensitizing currents (60). Furthermore, GMQ was found to activate molluscan neuropeptide-gated FaNaC channels through amino acids within the nonproton ligand-sensing domain, independent of FMRF-amide activation (9), pointing to a deeply-conserved gating mechanism for DEG/ENaC channels that can be acted upon by GMQ and amiloride. Here, we provide a second report of a WT DEG/ENaC channel being activated by amiloride. Tad-NaC6 currents elicited by 1-3 mM amiloride were far larger in amplitude that those elicited by increasing pH (Fig. 7, A and B). We note that the channel lacks the two glutamate residues in the nonproton-sensing domain of ASIC3 associated with amiloride activation. However, the channel bears a serine in the degenerin position corresponding to the G430V mutation in ASIC2a, and furthermore, a cysteine just two amino acids downstream from the Gly-437 position, which as mentioned above, caused the human channel to become amiloride-activated (Fig. 4).
Finally, it is interesting to note that the Glu-79/Glu-432 residues required for amiloride activation of ASIC3 were also found to be important for block of ASIC1a by diminazene, another general DEG/ENaC channel blocker. Single glutamine mutations at either of these glutamate residues reduced the channel sensitivity to diminazene by roughly 7-fold. Moreover, a glycine to alanine mutation in the Gly-437 position, associated with amiloride sensitivity of ASIC2a, increased diminazene sensitivity of ASIC1a by 1 order of magnitude (75). In addition, this same locus bears serine and glycine residues in the three ENaC channel subunits that are critical for amiloride block (76), where altogether, it appears as though amiloride and diminazene block DEG/ENaC channels by binding similar regions of the pore. Here, we found TadNaC6 to be sensitive to block by diminazene, albeit with a relatively low-affinity IC 50 value of ϳ578 M (Fig. 7C), which is fundamentally different from the effect of amiloride on this channel. Considering the model for paradoxical activation of ASIC3 by amiloride, it seems plausible that TadNaC6 similarly possesses a secondary binding site for amiloride within the nonproton ligand-sensing domain, and poor affinity for the drug within the pore. Instead, diminazene might effectively bind within the pore to block the channel and/or lack the ability to bind the nonproton ligandsensing domain. Clearly, future studies will be required to make sense of how these drugs regulate the gating of TadNaC6, which might provide some unique perspectives on how these drugs modulate DEG/ENaC channel gating in general.

Fluorescence in situ hybridization
Animals were prepared for FISH by freezing in tetrahydrofuran overnight on dry ice followed by fixation in 3% acetic acid in methanol (MeOH) for 30 min at Ϫ20°C and then 4% paraformaldehyde in methanol for 30 min at room temperature, as described (45). In situ hybridization was performed with RNAscope probes for ASIC6, ASIC5, and ASIC2 (#572821, #572811-C2, and 572801-C3, respectively) and Multiplex Fluorescent Assay reagents (#320850) from Advanced Cell Diagnostics (Hayward, CA). Fluorescence images were collected ϫ63 NA 1.4 objective on an LSM880 laser-scanning confocal microscope (Carl Zeiss Microscopy LLC, Thornwood, NY).

In silico analyses
The maximum likelihood phylogenetic tree was inferred from a MUSCLE protein alignment (77) trimmed with trimAl Mnemiopsis leidyi were obtained from a transcriptome assembly generated in-house, and those for Beroe ovata were obtained from Drs. Joseph Ryan and Mark Martindale (Whitney Laboratory, Florida) from gene model databases generated for genome assembly. All protein sequences are provided in FASTA format in File S1. Structural modeling of the tertiary structures of various DEG/ENaC channels homologues was achieved using the Phyre2 server intensive modeling mode (80). Confidence values for the resulting models are reported in the corresponding figure legends. All Protein Data Bank files for predicted tertiary structures are provided in the compressed File S2. Phyre2 was also used to predict secondary structures, depicted in a ClustalW protein alignment in Fig. 2 (81).

In vitro expression and Western blotting
We previously reported the cloning of various TadNaC ion channel cDNAs from whole-animal total RNA. 3 For functional expression, the TadNaC6 cDNA was PCR-amplified using primers TadNaC6-forward (5Ј-ATTATACTCGAGGCCG-CCACCATGATGGCCAAGGGAAAAATCGATCACG) and TadNaC6-reverse (5Ј-ATTATAGGATCCTCTTTACATAT-CTGGATTTGTTGCATCAGC) bearing respective N-and C-terminal XhoI and BamHI restriction endonuclease sites, permitting direct cloning into the mammalian expression vectors pIRES2-EGFP and pEGFP-C1 (both from Clontech). The forward primer TadNaC6-forward also contained the Kozak consensus sequence GCCGCCACC directly upstream of the start codon for efficient protein translation in vitro. The resulting expression vectors, named pTadNaC6 -IR-EGFP and pEGFP-TadNaC6-C1, differ in that the former expresses the channel protein separate from soluble GFP (EGFP) through a bicistronic internal ribosome entry site, whereas the latter expresses the channel with an N-terminally fused EGFP. The S353A variant of TadNaC6 was generated via site-directed mutagenesis using the pTadNaC6 -IR-EGFP vector as template and primers S353A-forward (5Ј-GCTATCCAACTGAA-ATCGCGCAATCATCACTTGGAAC) and S353A-reverse (5Ј-GTTCCAAGTGATGATTGCGCGATTTCAGTTGGAT-AGC). Chinese hamster ovary (CHO-K1) cells (Sigma) cells were cultured and transfected in 6-ml vented flasks incubated at 37°C within a CO 2 incubator, using Kaighn's modified Nutrient Mixture F-12 Ham's media (Sigma) supplemented with 10% fetal bovine serum (Wisent). For all in vitro expression experiments, 3 g of plasmid DNA was transfected into CHO-K1 cells at 80 -90% confluency with the transfection reagent PolyJet TM (FroggaBio). Post-transfection, cells were washed two times with 4 ml of serum-free media and incubated for 24 h to ensure adequate channel protein expression.

In vitro properties of a placozoan DEG/ENaC ion channel
clonal mouse anti-GFP IgG1 antibody (Sigma) diluted to 1:1000. After three washes in TBST, the secondary antibody goat anti-mouse IgG peroxidase conjugate (Sigma) was applied at 1:4000 dilution for 2 h at room temperature, and the membrane was washed again three times in TBST and then imaged on an ImageQuant TM LAS 500 imager using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific).

Patch-clamp electrophysiology
For electrophysiology, CHO cells transfected with the pTadNaC6 -IR-EGFP vector were trypsinized the following day and were plated onto glass coverslips in 35-mm culture dishes and incubated overnight at 37°C in a CO 2 incubator. On the day of recording, coverslips were transferred to new 35-mm culture dishes bearing ϳ3 ml of appropriate extracellular recording solution. For the pH dose-response curve shown in Fig. 1, the external solution contained 140 mM NaCl, 4 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, 5 mM MES (pH 5-7.5 with HCl/NaOH), or 140 mM NaCl, 4 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, 5 mM CHES (pH 8 to 9 with HCl/ NaOH). The intracellular solution consisted of 120 mM KCl, 2 mM MgCl 2 , 10 mM EGTA, 10 mM HEPES, pH 7.2, with KOH. All salts used in our experiments were purchased from Sigma and were of high purity (Ն99%). Similar solutions were used for the low Ca 2ϩ pH dose-response curve shown in Fig. 6, D and E; however, the external [Ca 2ϩ ] was reduced to 0.1 mM. For the Ca 2ϩ dose-response curve (Fig. 6, B and C), the external solution consisted of 60 mM NaCl, CaCl 2 (1, 10, 100, 1000, 3000, and 30,000 M), and 10 mM HEPES, pH 7.5, and the same internal solution as above was used. For the ion selectivity experiments (Fig. 5), the external contained 150 mM XCl ion where X ϭ Na, K, Li, or Cs, 10 mM TEA-Cl, and 10 mM HEPES, pH 7.4, with XOH, and the internal solution contained 150 mM NaCl, 10 mM EGTA, 10 mM TEA-Cl ,and 10 mM HEPES, pH 7.2, with NaOH. Whole-cell patch-clamp recordings were obtained using an Axopatch 200B amplifier coupled to a Digidata 1550A digitizer, using the pClamp 10 software (Molecular Devices). Patch pipettes were pulled using a P-1000 micropipette puller (Sutter), from thick-walled borosilicate tubing (1.5 and 0.86 outer and inner diameter, respectively), to a resistance of between 2 and 5 megohms. Series resistance was not compensated, and we only used data for analysis where the patch had minimal access resistance and tight capacitive transients upon voltage steps. Recordings were sampled at 2000 Hz and filtered offline at 500 Hz using the pCLAMP software. For all experiments that required perfusion of external solutions, we employed a Valvelink8.2 gravity flow Teflon perfusion system (AutoMate Scientific, Berkeley, CA). IC 50 values for proton, Ca 2ϩ , and diminazene dose-response curves were determined by fitting monophasic or biphasic dose-response curves over the data using the software package Origin 2016 (OriginLab). The relative permeability P X /P Na determined for TadNaC6 (Fig. 5), where X is a monovalent cation (Li ϩ , Cs ϩ , or K ϩ ), was determined using Equation 1, ⌬E Rev ϭ E Rev,X Ϫ E Rev,Na ϭ RT/zF ϫ ln P x /P Na (Eq. 1) Author contributions-W. E. and A. S. conceptualization; W. E. and A. S. data curation; W. E., C. L. S., and A. S. formal analysis; W. E., C. L. S., and A. S. investigation; W. E. and A. S. visualization; W. E., C. L. S., and A. S. writing-original draft; C. L. S. and A. S. resources; C. L. S. and A. S. funding acquisition; A. S. supervision; A. S. project administration.