Neuropeptides and degenerin/epithelial Na+ channels: a relationship from mammals to cnidarians

Ion channels of the degenerin (DEG)/epithelial Na+ channel (ENaC) family serve diverse functions ranging from mechanosensation over Na+ reabsorption to H+ sensing and neurotransmission. However, several diverse DEG/ENaCs interact with neuropeptides; some are directly activated, whereas others are modulated by neuropeptides. Two questions arise: does this interaction have a common structural basis and does it have an ancient origin? Current evidence suggests that RFamide neuropeptides activate the FMRFamide‐activated Na+ channels (FaNaCs) of invertebrates via binding to a pocket at the external face of their large extracellular domain. It is likely that RFamides might activate DEG/ENaCs from the freshwater polyp Hydra (the HyNaCs) via binding to a similar pocket, although there is not yet any experimental evidence. In contrast, RFamide neuropeptides modulate acid‐sensing ion channels (ASICs) from vertebrates via binding to a central cavity enclosed by β‐sheets of the extracellular domain. Dynorphin opioid peptides, for their part, bind to the acidic pocket of ASICs, which might be evolutionarily related to the peptide binding pocket of FaNaCs, but instead of opening the channels they work as antagonists to stabilize its closed state. Moreover, peptides interacting with DEG/ENaCs from animals of different phyla, although having similar sequences, are evolutionarily unrelated to each other. Collectively, it appears that despite a seemingly similar interaction with similar peptides, the interaction of DEG/ENaCs with neuropeptides has diverse structural bases and many origins.


Introduction
Ion channels of the degenerin (DEG)/epithelial Na + channel (ENaC) family are typically Na + -selective cation channels. They assemble as homo-or heterotrimers of individual subunits. The subunits share a structure consisting of two transmembrane domains (TMDs), a large extracellular domain (ECD) and short intracellular termini. Despite this conserved structure, DEG/ENaCs serve surprisingly diverse functions ranging from mechanosensation over Na + reabsorption to H + sensing and neurotransmission. A common theme uniting several diverse DEG/ENaCs is the tuning of their channel properties by neuropeptides, a diverse group of neuromodulators. Some DEG/ENaCs are directly gated by neuropeptides and represent the only known examples of peptide-gated ion channels, whereas others are modulated by neuropeptides. The DEG/ENaCs interacting with neuropeptides occur in diverse groups of animals spanning the eumetazoan tree and are not directly related to each other (they are not species orthologues). Based on this pattern, two questions arise: first, is there a common structural basis for the regulation by neuropeptides and, second, does this relationship have an ancient origin? These are the two main questions that we cover in this review. In addition, we briefly touch upon the physiological functions of the channel-neuropeptide relationship.
Before discussing the regulation of DEG/ENaCs by neuropeptides, we need briefly to consider the structure of these channels. The best known structure of a DEG/ENaC is that of chicken acid-sensing ion channel 1 (cASIC1), for which several high-resolution crystal structures (Baconguis et al., 2014;Gonzales et al., 2009;Jasti et al., 2007;Yoder et al., 2018) and more recent cryogenic electron microscopic structures are available (Yoder & Gouaux, 2020). The three subunits assemble symmetrically around the central ion pore, which is surrounded by the six TMDs, in particular TMD2. The large ECD has an intricate structure that has been compared to a clenched hand made up of a large palm domain, the upper part of which provides, together with the knuckle domain, the scaffold for the thumb, the finger and the lower part of the palm domain, which together surround the central β-ball domain (Fig. 1A). The β-sheets of the lower palm domains encapsulate a central, negatively charged cavity, the central vestibule, which contracts during desensitization (Baconguis & Gouaux, 2012). Lateral fenestrations provide access for cations to the extracellular vestibule and the ion pore ( Fig. 1C; Yoder et al., 2018). At the interface of two adjacent subunits and formed by intrasubunit contacts between the thumb, the β-ball and the finger domains, together with residues from the palm domain of an adjacent subunit, is a pocket with a suspiciously large number of negatively charged amino acids, the so-called acidic pocket (Fig. 1B;Jasti et al., 2007). It has been proposed (Jasti et al., 2007;Yoder et al., 2018), and there is also experimental evidence, that protons bind to residues of the acidic pocket to neutralize negative charges and thereby induce conformational changes. But it appears that protons also bind outside the acidic pocket to open an acid-sensing ion channel (ASIC; Krauson et al., 2013;Liechti et al., 2010;Paukert et al., 2008;Vullo et al., 2017). Upon opening, the acidic pocket collapses and, via the palm domain, this movement expands the lateral fenestrations and opens the channel gate at the extracellular end of TMD2 (Yoder et al., 2018).

Neuropeptides as agonists of DEG/ENaCs
The DEG/ENaC family contains the only known ion channels directly opened by a neuropeptide. We will first consider these channels in bilaterian invertebrates before we turn to such channels in a cnidarian, the freshwater polyp Hydra magnipapillata. So far, no peptide-gated channels have been identified in deuterostome animals, including vertebrates, but many deuterostome DEG/ENaCs have not yet been deorphanized (Elkhatib et al., 2022), meaning that we cannot exclude their presence in deuterostomes.
Peptide-gated ion channels from invertebrates: FMRFamide-gated Na + channels and Wamide-gated Na + channels First characterized in the garden snail Helix aspersa (now Cornu aspersum; Cottrell, 1997;Cottrell et al., 1990), the FMRFamide-gated Na + channel (FaNaC) was also the first peptide-gated channel to be cloned (Lingueglia et al., 1995). It is directly activated by the small neuropeptide FMRFamide, which is widely used by invertebrates. In Helix, the prepropeptide containing FMRFamide also contains FLRFamide (Cottrell, 1997), which is a partial agonist of FaNaC with lower potency (Table 1; Lingueglia et al., 1995). The FMRFamide-gated Na + channel is a homomer, is Na + -selective and, upon prolonged exposure to the peptide ligand, it desensitizes slowly (Lingueglia et al., 1995).
In Helix and Lymnea neurons, FMRFamide elicits fast depolarizing inward currents, suggesting that FMRFamide acts as a fast excitatory neurotransmitter via FaNaCs. Upon repeated application, the response desensitizes (Cottrell et al., 1990;Green et al., 1994;Perry et al., 2001). In addition to FaNaCs, in molluscs FMRFamide is also likely to act via a type 2 FMRFamide-activated G protein-coupled receptor (GPCR; Bauknecht & Jekely, 2015;Thiel et al., 2021), showing that this peptide mediates rapid signalling via ion channels, in addition to slow, modulatory signalling via GPCRs. Although protons do not open FaNaCs, at higher concentrations (pH < 6) they reduce FaNaC current amplitude (Furukawa et al., 2006; Perry Yang et al., 2017), and for the FaNaC from Lymnaea they prevent the reduction in current upon repeated stimulation (Perry et al., 2001). Initially, the FMRFamide-gated Na + channel was described only in snails (Furukawa et al., 2006;Jeziorski et al., 2000;Lingueglia et al., 1995;Perry et al., 2001), but broader sampling in a larger number of lophotrochozoans (the clade including annelids, molluscs, phoronids and other phyla) has recently revealed a hitherto hidden diversity of peptide-gated DEG/ENaCs throughout these animals, including cephalopods, annelids and brachiopods (Dandamudi et al., 2022). It appears that peptide-gated DEG/ENaCs diversified through several duplications within these phyla. Testing the ligand specificity of several of these channels revealed that FMRFamide represents the broadest and, probably, the ancestral ligand for the subfamily. Most of these DEG/ENaCs are activated with different sensitivities not only by FMRFamide and FLRFamide, but also by FMKFamide. In one annelid branch, however, the channel also became sensitive to FVRIamide peptides (Dandamudi et al., 2022), derived from a distinct precursor conserved across lophotrochozoans (Thiel et al., 2021), while retaining activation by FMRFamide. In another branch, channels evolved sensitivity to a group of Wamides called myoinhibitory peptides (MIPs) and lost FMRFamide sensitivity (Dandamudi et al., 2022). Thus, despite their sequence homology, FaNaCs diversified to respond to a range of short neuropeptides.
The Wamide-gated channels have been named WaNaCs and form a subgroup within FaNaCs that seems to be restricted to annelids (Dandamudi et al., 2022;Schmidt et al., 2018). The best-characterized WaNaC is the MIP-gated ion channel (MGIC) from the marine annelid Platynereis dumerilii, which is activated by 11 different MIPs that share extensive sequence identity, are nine or 10 amino acids long and are released from a single MIP propeptide . Table 1 shows that the potency of neuropeptides for FaNaCs and WaNaCs varies from the high nanomolar to the low micromolar range. Like Helix FaNaC, MGIC desensitizes slowly and partially and is modestly Na + selective . Like FMRFamide, MIPs can also activate a GPCR (Conzelmann et al., 2013). The MIP-gated ion channel was the only peptide-responsive channel in a screen involving four Platynereis DEG/ENaCs and 121 synthetic peptides. This high specificity and low hit rate suggest that peptide-gated channels are not as numerous in annelids as peptide-activated GPCRs (Bauknecht & Jekely, 2015).
FMRFamide and MIPs belong to two different families of neuropeptides that diverged (in stem Bilateria) before the divergence of FaNaCs and WaNaCs (within the Lophotrochozoa). Myoinhibitory peptide belongs to a protostome-specific proneuropeptide family of RF/Wamides including RGWamide, proctolin and myo-suppressin/myomodulin, whereas FMRFamide is more related to tachykinin, luqin, leucokinin, short neuropeptide F and other RF/RYamides (Thiel et al., 2021). The evolution of a new ligand specificity in WaNaCs and in the FVRIamide-and FMRFamide-gated FaNaC subclade suggests a more relaxed evolution of ligand specificity in DEG/ENaCs. This contrasts with the pattern seen in GPCRs, where long-term conservation and co-evolution is the norm (Jekely, 2013;Mirabeau & Joly, 2013;Thiel et al., 2021).
Why did a group of FaNaCs evolve sensitivity to Wamide peptides deep in the history of annelids (Dandamudi et al., 2022)? In annelids, MIP also signals via a metabotropic receptor belonging to a well-conserved protostome-specific clade of GPCRs (MIP or sex-peptide receptors) (Kim et al., 2010). The MIPs induce larval settlement in Platynereis by rapidly inhibiting ciliary activity and triggering surface-exploratory behaviour (Conzelmann et al., 2013). The MIPs also have long-term effects on behaviour, including the activation of feeding (Williams et al., 2015). Measuring the affinity of all 11 MIPs from the same precursor to the two receptors identified peptides that activate only MGIC and not the GPCR. The two receptors are also expressed in different cells in Platynereis larvae . Experiments with knockout larvae or treatments with MIPs that preferentially activate the MGIC channel, but not the GPCRs, could be used to tease apart the physiological contributions of the two receptor types to these behavioural responses.
The detailed structure of FaNaCs remains elusive, but sequence identity with ASICs is high enough to assume a similar overall structure. Using chimeras of FaNaCs from Helisoma and Helix, the first 120 amino acids of the ECD, composing part of the palm, the β-ball and, most noticeably, the whole finger domain, have been identified as determining differences in apparent affinity to FMRFamide between these two FaNaCs (Cottrell et al., 2001). In agreement, another study identified the upper finger domain and FaNaC-specific insertions in the upper thumb domain and the upper knuckle domain of Helix FaNaC as important for apparent FMRFamide affinity (Niu et al., 2016). In combination with a homology model of FaNaC, together with substituted cysteine accessibility assays, Niu et al. (2016) proposed that these three regions form the FMRFamide binding pocket at the top of the ECD, far away from the channel gate and, although adjacent, not in the same position as the acidic pocket of ASICs ( Fig. 2A).
More recently, based on comparative analysis of the broader FaNaC family and using extensive site-directed mutagenesis, a different pocket in the ECD has been identified and proposed to mediate activation of FaNaCs by neuropeptides. This pocket lies at the interface of adjacent subunits and is formed mainly by residues from the palm and the β-ball (Dandamudi et al., 2022), thus below the other putative peptide binding pocket (Fig. 2B). Although mutations of individual amino acids within this pocket profoundly reduce apparent FMRFamide affinity (Dandamudi et al., 2022), independent functional data, for example from substituted cysteine accessibility assays, or biochemical data are needed to confirm the conclusion that this is the binding pocket for FMRFamides on FaNaC. Interestingly, the defining C-terminal Trp residue of an MIP is vital for activation of MGIC , suggesting that this Trp residue makes important contacts to the binding pocket of the receptor. Likewise, non-amidated FMRF is a much less potent agonist of FaNaC than FMRFamide (Cottrell, 1997), suggesting that the C-terminal, amidated Phe residue of FMRFamide also makes important contacts to the binding pocket of FaNaC.

Peptide-gated ion channels from Cnidaria: Hydra Na + channels
The genome of the freshwater polyp Hydra contains 11 DEG/ENaC subunits, the Hydra Na + channels (HyNaCs). In a heterologous expression system, 10 HyNaCs assemble into 13 different heterotrimeric ion channels that are directly activated by two neuropeptides of the Hydra nervous system: Hydra-RFamide I (pQWLGGRF-NH 2 ) and Hydra-RFamide II (pQWFNGRF-NH 2 ) (Assmann et al., 2014;Dürrnagel et al., 2010;Golubovic et al., 2007), which are both released from two different proneuropeptides; Hydra-RFamide I is additionally released from a third proneuropeptide (Darmer et al., 1998). Functional trimeric HyNaCs differ in their apparent ligand affinity by more than two orders of magnitude (Table 1; Assmann et al., 2014). Moreover, HyNaCs do not desensitize and, uncommonly for DEG/ENaCs, are unselective cation channels that also conduct Ca 2+ (Dürrnagel et al., 2012). Some HyNaCs are very sensitive to their peptide ligands and start to open at low nanomolar concentrations of RFamides (Assmann et al., 2014). Although also activated by RFamide neuropeptides, HyNaCs have only a distant relationship to FaNaCs (Aguilar-Camacho et al., 2022;Assmann et al., 2014;Golubovic et al., 2007). It is also becoming clear that W/Famides in cnidarians, including Hydra-RFamides, are not one-to-one orthologues of any of the bilaterian W/F/Yamide families. This conclusion is suggested by the history of GPCR diversifications and the lack of direct receptor orthologues between Cnidaria and Bilateria (Quiroga Artigas et al., 2020;Thiel et al., 2021).
Co-expression of the mRNA coding for HyNaC subunits in Hydra tissue suggest that at least six different functional HyNaC trimers exist in situ, which localize to the base of the tentacles and the peduncle (Assmann et al., 2014). Given that two blockers of HyNaCs, amiloride and diminazene, delay tentacle contractions during the feeding response, it has been proposed that HyNaCs are involved in neuromuscular transmission (Assmann et al., 2014;Dürrnagel et al., 2010). Their Ca 2+ permeability might even be sufficient to elicit muscle contractions directly (Gründer & Assmann, 2015). The facts that HyNaCs do not desensitize and that the effect of the peptide can be long lived, however, suggest that the peptide functions through volume transmission (Hawk et al., 2021) and that HyNaCs modulate neuromuscular transmission rather than eliciting contractions directly. Therefore, the large range of apparent peptide affinities (Table 1)  physiological role of HyNaCs remain hypothetical at present.
The binding pocket of peptides on HyNaCs is unknown. The HyNaCs are obligate heterotrimers, all of which contain HyNaC2 as an invariable subunit (Assmann et al., 2014). It might be that HyNaC2 is necessary for the assembly or trafficking of HyNaCs. It might also be that it is involved in ligand binding and that the other subunits only fine-tune ligand binding, giving rise to the large variety of apparent affinities. Of note, Hydra-RFamide III (KPHLRGRF-NH 2 ) and Hydra-RFamide IV (HLRGRF-NH 2 ) do not activate HyNaCs (Golubovic et al., 2007), although they have the same C-terminal sequence as Hydra-RFamide I and II, suggesting that the N-terminal part of the peptides makes important contributions to binding or activation of HyNaCs. Given that HyNaCs have a relatively close relationship to ASICs (Aguilar-Camacho et al., 2022;Assmann et al., 2014;Golubovic et al., 2007) homology models can reveal whether they possess a cavity corresponding to the acidic pocket of ASICs. Although the acidic residues of this pocket are not conserved in HyNaCs, it would still be a candidate for the ligand binding pocket of HyNaCs. It will be interesting to see whether the peptide binding pockets are conserved between HyNaCs and FaNaCs, which would have important implications for the evolution of peptide gating in this gene family.
Like Hydra, it appears that other non-bilaterian organisms also use neuropeptides extensively as transmitters. Ctenophores, for example, mainly use glutamate and neuropeptides (Sachkova et al., 2021) and also have a large variety of DEG/ENaCs (Moroz et al., 2014). Likewise, the sea anemone Nematostella vectensis also has a large variety of DEG/ENaCs, the NeNaCs, two of which are closely related to HyNaCs (Aguilar-Camacho et al., 2022), making neuropeptides candidate ligands for these NeNaCs. A first screen with neuropeptides from Nematostella, however, did not identify a neuropeptide directly activating NeNaCs (Aguilar-Camacho et al., 2022). Future studies will reveal whether DEG/ENaCs of Ctenophores and other Cnidaria are also activated by neuropeptides and whether these are related to Hydra-RFamides. Currently, it appears that the deuterostome lineage of bilaterian animals, which includes vertebrates, did not develop or, alternatively, lost peptide-gated DEG/ENaCs.

Neuropeptides as modulators of DEG/ENaCs
Although neuropeptides do not activate ASICs directly (Kuspiel et al., 2021;Vyvers et al., 2018), it appears that they modulate different ASICs. To appreciate this modulation, we first need briefly to introduce proton gating of ASICs. For our purpose, it is sufficient to consider only the two most proton-sensitive ASICs: ASIC1a and ASIC3. Increasing the concentration of extracellular protons opens ASICs. Both ASIC1a and ASIC3 start to open when the proton concentration is increased rapidly from its resting concentration of 40 nm (corresponding to pH 7.4) to ∼100 nm (pH 7.0). At 1 μm protons (pH 6.0), they are fully activated (Fig. 3A). After opening, ASICs enter a desensitized state (Fig. 3B), ASIC1a with a time constant (τ des ) = 1.2-3.5 s and ASIC3 with τ des = ∼0.3 s (for a review, see Gründer & Pusch, 2015). ASIC1a exhibits a relatively slow recovery from desensitization, with a time constant (τ ) = ∼10 s, and ASIC3 more rapidly, with τ = ∼0.5 s. Slightly and slowly increasing the proton concentration, in the range between pH 7.4 and 6.9, transfers ASICs into a desensitized state instead of opening them and prevents their subsequent opening by a fast pH drop, a process that is referred to as steady-state desensitization (SSD; Fig. 3A; Gründer & Pusch, 2015).

. Proton gating of acid-sensing ion channels
A, schematic drawing of pH-dependent steady-state desensitization (SSD) and pH-dependent activation curves of an acid-sensing ion channel (ASIC). B, ASIC3 is opened by a reduction in extracellular pH. During prolonged acidification, the channel enters a desensitized state, and the pore is shut again. Right, pre-application of RFamide peptides, such as FMRFamide, slows current decay. C, Big Dyn shifts the SSD curve to lower pH. Likewise, FRRFamide shifts the SSD curve of ASIC1a (not shown).

RFamide neuropeptides and ASIC3
Although FMRFamide is not produced by vertebrates, there are several RFamide-related neuropeptides in vertebrates, including neuropeptide FF (NPFF; FLFQPQRF-NH 2 ), neuropeptide AF (NPAF) and neuropeptide SF (NPSF; SLAAPQRF-NH 2 ), all of which are derived from a common precursor. The mRNA for the precursor is upregulated by inflammation (Vilim et al., 1999), and NPFF elicits hyperalgesia (Askwith et al., 2000), probably by activating GPCRs. Given that ASICs respond to the tissue acidosis that is associated with inflammation and because they are related to RFamide-gated DEG/ENaCs, it has been hypothesized that RFamide neuropeptides might also modulate ASICs. It has been found that FMRFamide slows current decay of typical ASIC currents in dorsal root ganglion neurons and that it induces a sustained current in some neurons (Askwith et al., 2000). Although it was initially reported that FMRFamide slowed current decay of human ASIC1a with low potency (EC 50 ∼33 μm; Askwith et al., 2000), another study found that FMRFamide has only modest effects on rat ASIC1a and primarily modulates rat ASIC3 and the ASIC3/1b heteromer (EC 50 ∼24 μm; Chen et al., 2006). The different results are attributable, at least in part, to species differences (human vs. rat and mouse; Sherwood & Askwith, 2008). The small synthetic peptide FRRFamide, which has two positive charges, robustly modulates ASIC1a (Sherwood & Askwith, 2008). In contrast, the mammalian neuropeptide NPFF has only modest effects on ASIC1a, whereas NPFF and NPSF have a stronger effect on ASIC3 (Askwith et al., 2000;Deval et al., 2003). In addition, deletion of the ASIC3 gene attenuated the response of ASICs in mouse dorsal root ganglion neurons to FMRFamide-related peptides (Xie et al., 2003), further suggesting that ASIC3 might be the main ASIC target for NPFF and related RFamide peptides in situ.
More recently, in a screen for toxins from cone snail venoms targeting ASICs, a small cono-RFamide, RPRFamide, has been identified that targets rat ASIC3 with a relatively high potency of 3−4 μm; RPRFamide does not modulate rat ASIC1a or ASIC1b (Reimers et al., 2017). RPRFamide increases excitability of mouse dorsal root ganglion neurons to slight acidosis, and injection of RPRFamide into the gastrocnemius muscle increases muscle pain in an ASIC3-dependent manner (Reimers et al., 2017), suggesting that this cone snail uses RPRFamide to induce pain via ASIC3.
However, the fact that endogenous RFamide neuropeptides modulate ASIC3 with lower potency than FMRFamide or RPRFamide calls into question that RFamide neuropeptides are endogenous modulators of ASICs. In agreement, subcutaneous injection of NPSF into the area innervated by the saphenous nerve elicited equal nociceptive behaviour in wild-type and ASIC3 −/− mice (Yudin et al., 2006), showing that at least the pain-inducing actions of NPSF in the skin do not rely on ASIC3. The two short endogenous opioid peptides endomorphin-1 (YPWFamide) and endomorphin-2 (YPFFamide), which have a high affinity for μ opioid receptors, also modulate ASIC3 and slow its desensitization (Farrag et al., 2017;Vyvers et al., 2018).
Irrespective of the physiological importance, it is clear that diverse RFamides, in particular short peptides such as RPRFamide, FRRFamide and FMRFamide, can modulate ASIC3. Modulation of ASIC3 by FMRFamide (Askwith et al., 2000) and even Hydra-RFamides I-IV (Golubovic et al., 2007) seems to indicate a conserved binding pocket for RFamide peptides on ASIC3 and on peptide-gated FaNaCs and HyNaCs. A recent study, in which some of the residues corresponding to amino acids of the putative external FMRFamide binding pocket of FaNaC at the interface of adjacent subunits (see above) were mutated in ASIC1a, concluded that this binding pocket is conserved in ASICs (Dandamudi et al., 2022). The mutation that had the strongest effect on peptide modulation of ASIC1a (K246A), however, might have done so unspecifically, by altering the proton sensitivity of ASIC1a. In addition, mutation of the corresponding amino acid in ASIC3 (R239A) did not abolish modulation by RPRFamide (Reiners et al., 2018). Thus, the experimental evidence for a conserved binding site for FMRFamides is not convincing. Moreover, the mechanism of RFamides on these channels has major differences. First, RFamides do not activate ASICs; their main effect is a slowing of the current decay (Fig. 3B). Second, the modulation by FMRFamide has a slow time course, of the order of several seconds (Askwith et al., 2000;Chen et al., 2006), whereas RFamides activate FaNaCs and HyNaCs rapidly, within 1 s.
Recently, a modified RPRFamide has been coupled covalently to ASIC3 via ultraviolet crosslinking, revealing that RPRFamide prevents desensitization (Reiners et al., 2018). Thus, when RFamide is not covalently coupled, the slow time course of current decay is attributable to its slow unbinding; desensitization proceeds with an unaltered time course. Moreover, extensive mutagenesis revealed the negatively charged central vestibule that is encapsulated by the three lower palm domains (Baconguis & Gouaux, 2012) as the likely binding pocket for RPRFamide on ASIC3 (Reiners et al., 2018) and for FRRFamide on ASIC1a (Bargeton et al., 2019); mutations in the external acidic pocket, in contrast, did not affect modulation of ASIC3 by RPRFamide (Reiners et al., 2018). As the central vestibule contracts during desensitization of ASICs (Baconguis & Gouaux, 2012;Baconguis et al., 2014), its identification as the binding pocket of RFamides provides a simple mechanistic explanation for their effect on ASIC3: RFamides stabilize the open conformation J Physiol 601.9 and prevent desensitization by preventing contraction of the central vestibule, such that their unbinding determines the time course of current decay ( Fig. 4; Reiners et al., 2018). Moreover, binding of RFamide to the negatively charged central vestibule explains why peptides with a high positive charge, such as RPRFamide and FRRFamide, strongly modulate ASICs (Reimers et al., 2017;Sherwood & Askwith, 2008). Interestingly, although strongly modulating ASIC3, FRRFamide has low potency and even behaves like an antagonist of FaNaC (Cottrell, 1997).
The central vestibule in the lower palm domain has previously been found to bind diverse modulators of ASIC3: the synthetic ligand 2-guanidine-4-methlyquinazoline (GMQ; Yu et al., 2010), the arginine metabolite agmatine Yu et al., 2010), serotonin (Wang et al., 2013) and the pruritogenic peptide SLIGRLamide (Peng et al., 2015). Interestingly, like ASIC3, FaNaCs can be opened directly by millimolar concentrations of GMQ, and it is likely that GMQ binds to the central vestibule in the lower palm domain of FaNaCs (Yang et al., 2017). Mutations within the central vestibule prevent activation by GMQ but not by FMRFamide, indicating that GMQ and FMRFamide bind to two separate binding pockets on FaNaC (Yang et al., 2017). Collectively, it thus appears that the GMQ binding pocket in the lower palm domain (the central vestibule) is conserved in ASIC3, where it binds RFamides, whereas the external RFamide binding pocket of FaNaCs/WaNaCs is not conserved in ASIC3. We can conclude that RFamides activate FaNaCs by binding to a different pocket than RFamides, which modulate ASICs.

Big dynorphin and ASIC1a
Collectively known as dynorphins, α-neoendorphin, dynorphin A (Dyn A), dynorphin B (Dyn B) and big dynorphin (Big Dyn) compose a class of endogenous opioid neuropeptides released from the propeptide prodynorphin (Schwarzer, 2009). Prodynorphin belongs to the family of opioid peptides that also includes enkephalin, nociceptin and endorphin (POMC). These precursors and their receptors diverged during the two rounds of genome duplications that occurred during early vertebrate evolution. Endorphin (POMC) evolved later, probably through the duplication of the nociceptin precursor and a fusion with an adrenocorticotrophic hormone/melanocyte-stimulating hormone gene (Elphick et al., 2018;Larhammar et al., 2015). The processing of prodynorphin commences with the action of prohormone convertase (PC) 1, which cleaves at the carboxyl-terminus of three pairs of basic amino acids (KR, KK and RR) to yield a 10 kDa intermediate that is subsequently processed into α-neodynorphin and Big Dyn. Big Dyn, a highly basic peptide comprising 32 amino acids, can be processed proteolytically into Dyn A (amino acids 1-17) and Dyn B (amino acids 20-32) by carboxypeptidase E (Schwarzer, 2009). Prodynorphin is widely distributed in the CNS. Dynorphins activate G protein-coupled opioid receptors with high potency, mainly the κ opioid receptors, but also μ and δ opioid receptors, to modulate neurotransmitter release and postsynaptic activity, which is important in a variety of physiological and pathophysiological processes, such as nociceptive transmission, stress-induced responses,  epilepsy and psychotic disorders (Ferre et al., 2019;Schwarzer, 2009).
In addition to the interaction with GPCRs, Dyn A and Big Dyn modulate ASIC1a. In fact, Big Dyn is one of the most potent modulators of ASIC1a, suggesting that ASIC1a is a specific target of Big Dyn. Homomeric ASIC1a and heteromeric ASIC1a/2a are the most prevalent ASICs in the CNS. Besides its role in several physiological processes, including fear-related behaviours, synaptic plasticity, memory and learning (Wemmie et al., 2002, ASIC1a seems to play an important role in several neurodegenerative diseases, such as multiple sclerosis (Friese et al., 2007) and brain ischaemia (Xiong et al., 2004), which are associated with a sustained acidosis. During ischaemic stroke, for example, the disruption of the blood flow generates hypoxia, anaerobic metabolism and, consequently, acidosis, with a reduction in the tissue pH to values as low as 6.5 (Pignataro et al., 2007), sufficient to activate ASIC1a. Pharmacological blockade or genetic ablation of ASIC1a has a profound neuroprotective effect in animal models of stroke (Chassagnon et al., 2017;McCarthy et al., 2015;Pignataro et al., 2007;Xiong et al., 2004), supporting a pivotal role of ASIC1a in acid-induced neuronal injury.
It has been shown that Big Dyn and, with much lower potency (Table 1), Dyn A shift SSD curves of homomeric ASIC1a and heteromeric ASIC1a-containing channels to higher proton concentrations, allowing the channels to remain in a closed state and available for activation during slight extracellular acidification ( Fig. 3C; Borg et al., 2020;Leisle et al., 2021;Sherwood & Askwith, 2009;Sherwood et al., 2011). In cultured neurons, this interaction promotes ASIC1a-mediated neuronal death in response to slight but sustained acidification, similar to the conditions that occur during ischaemic stroke (Sherwood & Askwith, 2009). Given that dynorphin levels are increased in areas of neuronal injury or ischaemia (Faden, 1990;Hauser et al., 2005;McIntosh et al., 1987), this modulation of ASIC by Big Dyn might be pathologically relevant.
An in silico molecular model of the interaction between Dyn A and ASIC1a suggested that the highly basic peptide inserts deep into the acidic pocket to occlude it (Leisle et al., 2021). Experiments using site-directed mutagenesis of the peptide and the channel confirmed an extensive network of electrostatic interactions of acidic residues on ASIC1a and basic residues on Big Dyn (Borg et al., 2020;Leisle et al., 2021). Furthermore, covalent tethering of Big Dyn to ASIC1a using noncanonical amino acid-mediated photo-crosslinking confirmed the Big Dyn binding pocket in the ASIC1a acidic pocket (Borg et al., 2020;Leisle et al., 2021). The residues on ASIC1a mediating the interaction with Big Dyn seem to be clustered in three regions: helix α5 of the thumb domain, various stretches of the finger domain and a shorter stretch of the palm domain (Borg et al., 2020;Leisle et al., 2021). Cross-linking Big Dyn to ASIC1a also revealed a decreased proton sensitivity of ASIC1a activation (by almost 1 pH unit), an effect that is not observed during the application of soluble Big Dyn, suggesting that soluble Big Dyn unbinds from the acidic pocket during channel opening. Altogether, these findings suggest that in resting conditions Big Dyn interacts through electrostatic forces with acidic residues in the acidic pocket of ASIC1a, stabilizing this region in its extended conformation and shifting the SSD curve to a more acidic pH. As protons bind to the acidic pocket, Big Dyn is displaced and the pocket collapses, resulting in the opening of the channel (Fig. 5; Leisle et al., 2021). Collectively, it appears that the Big Dyn binding pocket might overlap with the putative binding pocket of FMRFamide on FaNaCs and of Hydra-RFamides on HyNaCs, but that this longer peptide uses more extensive contacts with the channel protein. Unlike FMRFamide on FaNaCs, Wamides on WaNaCs and Hydra-RFamides on HyNaCs, however, Big Dyn acts as an antagonist but not as an agonist of ASICs. Another difference is that Big Dyn binds to ASIC1a only slowly (Borg et al., 2020;Sherwood & Askwith, 2009).

Conclusions
Initially, we raised two questions. First, is there a common structural basis for the regulation of diverse DEG/ENaCs by neuropeptides?
We can now answer this question with 'No' . Current evidence indicates that RFamides modulate ASICs via binding to the central vestibule and activate peptide-gated DEG/ENaCs via binding to another, as yet ill-defined pocket on the exterior of the ECD, probably at subunit interfaces. Even for the two classes of peptide-gated channels, FaNaCs/WaNaCs and HyNaCs, the picture is ambiguous. Given that it seems that RFamides have a different origin in Cnidaria and lophotrochozoans, activation of these channels by RFamides must have arisen by convergent evolution. Whether a similar pocket in the ECD has been co-opted for this purpose in FaNaCs/WaNaCs and HyNaCs is an interesting question that awaits further research. The binding pocket of Big Dyn on ASIC1a might overlap with the binding pocket of peptide ligands on FaNaCs/WaNaCs and HyNaCs, but the binding mode is substantially different, in that Big Dyn stabilizes the closed state of ASIC1a and is displaced upon channel opening, whereas binding of peptide ligands opens FaNaCs/WaNaCs and HyNaCs.
Second, does the relationship of DEG/ENaCs with neuropeptides have an ancient origin? This is more difficult to answer but related to the first question. Independent evolution of neuropeptide ligands of HyNaCs, FaNaCs/WaNaCs and ASICs suggests an independent evolution of their interaction with their ion channel receptors. It might be, however, that an ancestral DEG/ENaC already interacted with peptides and that it has been peptide gated. Peptide binding might have been lost in some lineages and been switched to other peptides in others. Further testing of this scenario will require experimental analysis of a larger diversity of non-vertebrate DEG/ENaCs (including cnidarian, placozoan and bilaterian sequences) and determination of the structural basis of peptide recognition. But for now, we come to the somewhat surprising conclusion that the relationship of DEG/ENaCs with neuropeptides has many origins. Perhaps the large ECD of these channels was ideally suited for an interaction with the heterogeneous group of neuropeptides.