The First Peptide-Gated Ion Channel

Neuropeptide intercellular messengers occur widely in invertebrate and vertebrate species. They exert many different effects, generally slow or modulatory and frequently dependent on GTP-binding proteins (G-proteins). In some identified neurones of the the snail Helix aspersa, however, one endogenous neuropeptide, phenylalanyl-methionyl-arginyl-phenylalanyl-NH₂ (FMRFamide), exerts a fast excitatory response accompanied by an increase in membrane conductance (Cottrell et al. 1984). This fast response is readily distinguished from other, slower neuronal responses in Helix aspersa, such as those resulting from an increase in K+ conductance and the suppression of Ca²⁺ conductance (Cottrell et al. 1984; Colombaioni et al. 1985).

The peptide FMRFamide was discovered by Price and Greenberg (1977) in the clam Macrocallista nimbosa. The snail Helix aspersa also contains FMRFamide and in addition more than 10 related peptides, all of which are formed from two precursor proteins: one precursor incorporates FMRFamide, FLRFamide (phenylalanyl-leucyl-arginyl-phenylalanyl-NH₂) and pQFYRFamide (p-glutamyl-phenylalanyl-tyrosyl-arginyl-phenylalanyl-NH₂), the other many related peptides with N-terminal extensions (Lutz et al. 1992; Price et al. 1996b; Cottrell et al. 1994).

The fast FMRFamide response desensitizes rapidly and shows a marked preference for the tetrapeptide amides (FMRFamide and FLRFamide) rather than the N-terminally extended forms (Cottrell and Davies, 1987). Unlike similar depolarizing responses evoked by acetylcholine, 5-hydroxytryptamine (5-HT) and dopamine, it is not blocked by tubocurarine. The first evidence that the peptide directly gates a channel was obtained using outside-out membrane patches from the Helix C2 neurone (Cottrell et al. 1990). Inward unitary currents were observed when FMRFamide was applied to the external membrane surface of isolated outside-out patches, even when the recording pipette contained guanosine 5′-O-(2-thiodiphosphate), which blocks G-protein-coupled responses (Green et al. 1994). These results suggested that there is a tight receptor/channel coupling with no involvement of a G-protein. The channels are much more permeable to Na⁺; than to K⁺ or Ca²⁺ and have a relatively low conductance of approximately 4 pS. It was also observed that the response is blocked with amiloride. Often, in addition to the major fast component, a slower component of the total response is seen. This slower component is less susceptible to amiloride and is potentiated by extracellular CPT-c-AMP, but not extracellular cyclic AMP (Falconer et al. 1993; K. A. Green and G. A. Cottrell, unpublished data). The slower component is variable, but also probably results from the direct gating of an ion channel (Green et al. 1994).

The cDNA of an FMRFamide-gated receptor/channel protein (FaNaCh) was recently cloned and sequenced from a Helix aspersa cDNA library (Lingueglia et al. 1995). The observation that the neuronal peptide response is blocked by amiloride facilitated cloning, because cDNAs for the amiloride-sensitive epithelial Na⁺; channel proteins, ENaChs, had already been cloned from rat colon (Canessa et al. 1993, 1994; Lingueglia et al. 1993; and see Canessa et al. 1995). Sequence data from the ENaChs provided clues for designing degenerate polymerase chain reaction (PCR) primers to search for, and finally to sequence, the cDNA for the channel, FaNaCh. The predicted structure of the protein encoded by FaNaCh has a long extracellular loop (including a cysteine-rich region) and two membrane-spanning domains with both termini located intracellularly (Lingueglia et al. 1995). The predicted structure is different from the well-characterized class of ligand-gated channels typified by the nicotinic acetylcholine (ACh) receptors (each subunit of which possesses several transmembrane domains). A diagrammatic representation of the peptide channel (FaNaCh) is shown in Fig. 1.

Sections of the FaNaCh, notably the predicted membrane-spanning regions and part of the extracellular cysteine-rich domains, show similarity to the cloned epithelial Na⁺ channel subunits (ENaChs) (Canessa et al. 1993, 1994; Lingueglia et al. 1993). The naturally occurring channels, which occur on the apical surface of epithelial cells in the kidney, distal colon and lung of mammals, are thought to comprise three different subunits deriving from different messages: α-, β- and γ-ENaChs (Canessa et al. 1995; Lingueglia et al. 1994). Epithelial Na⁺; channels are highly selective for Na⁺ and are rate-limiting for Na⁺ reabsorption. Under normal conditions they are open, but they can be blocked by submicromolar concentrations of the diuretic amiloride. Biochemical analysis has provided evidence that the channel proteins comprise two transmembrane α-helices and a large extracellular domain of approximately 500 amino acid residues, and that both the amino and carboxy termini are located within the cytoplasm (Renard et al. 1994).

cDNA for an amiloride-sensitive channel has been cloned from human and rat brain neurones. This cDNA was designated BNC1 by Price et al. (1966c) and MDeg by Waldmann et al. (1996). When expressed, the channel was closed, but it was activated by a point mutation in an equivalent location that results in neuronal death with the degenerins of Caenorhabditis elegans (i.e. glycine-430 is replaced with a valine or phenylalanine) (Waldmann et al. 1996). This mammalian neuronal sequence (BNC1 i.e. MDeg) shows the greatest sequence similarity to FaNaCh. Waldmann et al. (1996) suggest that it too may be a ligand-gated channel, but the natural ligand is not known. García-Añoveros et al. (1997) have also cloned a related mammalian sequence (BNaC2) as well as BNaC1 (i.e. BNC1 or MDeg). Both sequences appear to be widely expressed in neurones of the central nervous system, although their patterns of expression vary slightly. The most recent development is the discovery of a related rat sequence, ASIC, which has the properties of a proton-gated channel when expressed in Xenopus laevis oocytes (Waldmann et al. 1997). ASIC is expressed in neurones of the dorsal root ganglia and different regions of the brain. It appears to be important in sensing extracellular pH variations and translating them into depolarizations.

Genetic analysis has identified a group of related genes, e.g. Mec-4 and Mec-10 (presumed mechano-sensory) and Deg-1 (all of which, when mutated, result in neuronal degeneration – hence ‘degenerins’) in C. elegans (see García-Añoveros et al. 1995). On the basis of genetic data from mutations affecting degeneration of specified mechano-sensory neurones in C. elegans, and some analogy with the epithelial Na⁺; channel, Huang and Chalfie (1994) have proposed that Mec-4, Mec-10 and another sequence, Mec-6, form a complex in the membrane of the mechano-sensitive neurones needed for mechano-sensation. Several other related sequences occur, some of which may also play a role in mechano-sensation (see Tavernarakis et al. 1997). So far, none of the C. elegans sequences has been expressed to test directly the view that they encode Na⁺ channels. However, Waldman et al. (1995) have expressed chimeric proteins from synthetic RNAs that incorporate the MI and/or the MII (presumed transmembrane domains) from Mec-4 within synthetic RNAs of the α-ENaCh (from which the corresponding regions had been deleted) and shown that they conducted Na⁺ and Li⁺.

Hydropathicity plots of predicted proteins of the recently cloned cDNAs of ATP-P2X receptor subunits (Brake et al. 1994; Valera et al. 1994; Chen et al. 1995; Lewis et al. 1995) are similar to those of the FaNaCh protein, suggesting that they all have two membrane-spanning domains. Further, the putative extracellular loop of P2X receptor proteins is rich in cysteine residues, as is the FaNaCh protein. However, there is little or no similarity in the P2X sequences with FaNaCh. Unlike FaNaCh, P2X receptors are not selective for Na⁺; or blocked by amiloride. Thus, the ATP-P2X receptors do not fall into the same group as those described above, although they too are predicted to have two membrane-spanning domains (see North, 1996).

It is likely that the neuronal peptide-gated channel incorporates the protein derived from the nucleotide sequence FaNaCh. If this assumption is correct (it can be verified by determining whether the message for FaNaCh occurs in responsive neurones), does the natural receptor comprise just one type of subunit (i.e. FaNaCh)? Evidence for the existence of three different subunits in epithelial Na⁺ channels was obtained by observing increased Na⁺ conductance by co-expression of the three different subunits (Canessa et al. 1994). It is well known that receptors for ACh and γ-aminobutyric acid (GABA) comprise many different subunits (see Barnard, 1996). However, only one subunit has been found for the 5-HT₃ receptor (Maricq et al. 1991).

One way to address the question of subunit composition is to compare the properties of expressed FaNaCh with those of the neuronal peptide-gated channel. Marked differences would suggest that the native receptor consists of more than one type of subunit (compared with the native epithelial Na⁺ channel; see Canessa et al. 1994). Pharmacological experiments on the neuronal peptide-gated channel are technically more difficult than studies on the expressed FaNaCh in Xenopus laevis oocytes. Thus, pharmacological studies have proceeded on the oocyte-expressed receptor, and selected comparisons have been made with responses in the C2 neurone.

When the cDNA sequence FaNaCh is transcribed and the message injected into Xenopus laevis oocytes, applied FMRFamide evokes an inward Na⁺; current. The properties of this current (Lingueglia et al. 1995) resemble in many respects those of the native channel in Helix neurones (Green et al. 1994). However, the expressed FaNaCh does not show the same degree of desensitization as the faster (i.e. major and amiloride-sensitive) component of the FMRFamide response of the C2 neurone. Activation of the expressed channel by FMRFamide (and also by FLRFamide) appears to be cooperative, with an apparent Hill coefficient of 1.6 (Lingueglia et al. 1995), indicating that at least two molecules interact with the receptor complex (presumably a homo-oligomer in the oocyte membrane). Lingueglia et al. (1995) also showed that some other molluscan peptides (SDPFLRFamide, GDPFLRFamide) and the mammalian peptide FLFQPQRFamide did not activate the channel.

The effects of several other peptides have recently been tested on FaNaCh expressed in Xenopus oocytes [see Table 2; G. A. Cottrell, M. C. Jeziorski, P. A. V. Anderson, D. A. Price, K. E. Doble, M. J. Greenberg, R. M. Greenberg and A. Edison, unpublished data; further data from the Helix C2 neurone (K. A. Green and G. A. Cottrell, unpublished data) are included in Table 2 and also shown in Fig. 5].

cDNA for the FMRFamide precursor protein also encodes the peptides FLRFamide and pQFYRFamide (Lutz et al. 1992; Price et al. 1996b). Extensive comparison of the potencies of FLRFamide and FMRFamide on oocyte-expressed FaNaCh confirmed that FLRFamide has a lower apparent affinity than FMRFamide, as described by Lingueglia et al. (1995) and, in addition, when both were tested on the same oocytes, showed that the maximum effect of FLRFamide is always smaller (40–60 %) than that of FMRFamide (Fig. 2A; Table 1). These data strongly suggest that FLRFamide is a partial agonist on expressed FaNaCh (Cottrell, 1997). Hill plots constructed for FMRFamide, FLRFamide and also for FnorLRFamide (see below) gave values ranging from 1.4 to 1.7, providing additional evidence that at least two molecules of each peptide bind to the oocyte-expressed channel.

Unexpectedly, pQFYRFamide (the other neuropeptide encoded on the precursor protein) did not evoke the Na⁺ current, but antagonized both FMRFamide and FLRFamide (Cottrell, 1997). An example of this antagonistic effect is shown on the FLRFamide response (Fig. 2C). Such receptor antagonism by an endogenous ligand has not been described before; its possible significance is considered in the Discussion.

Table 1 shows the relative potencies of several peptides on FaNaCh expressed in Xenopus oocytes. Some tentative conclusions can be drawn from the data.

• There is a requirement for a non-polar aromatic amino acid (phenylalanine or tryptophan) at amino acid positions 1 and 4. Biological activity is markedly reduced with peptides containing non-aromatic amino acids or a more polar aromatic amino acid, such as tyrosine, in either of these positions. The reduced activity of D-FMRFamide also suggests that the aromatic ring needs to be correctly orientated. Non-amidated analogues are much less active (compare FMRF with FMRFamide), as first shown on muscle responses by Price and Greenberg (1977).

• Replacing methionine by leucine (FLRFamide) results in considerable loss of activity. There is reduction in the maximum response as well as a reduction in apparent affinity. FnorLRFamide, in contrast, is as potent as FMRFamide, and the dose–response relationship is very similar (see also WnorLRFamide). Thus, the chain of four carbons (with nor-leucine) appears to be as effective as the similar chain incorporating one atom of sulphur (methionine) (see Fig. 3).

• Other amino acid substitutions in position 2 cause pronounced changes in biological activity. Some substitutions (e.g. with aspartic acid, arginine or lysine) result in antagonist activity. The most potent competitive antagonist detected so far is FKRFamide (phenylalanyl-lysyl-arginyl-phenylalanyl-NH₂) with an ID₅₀ of approximately 7 μmol l⁻¹. FKRFamide also blocks the neuronal response (see below). A stick-figure representation of FKRFamide is compared with FMRFamide in Fig. 3.

• Substituting the positively charged arginine in amino acid position 3 with isoleucine, D-arginine or even lysine also results in considerable loss of activity, suggesting that a positively charged residue is required in precisely the correct location and orientation. Some ionic interaction may occur between the amino acid in this position and the ligand-recognition site of the channel.

• As observed with the neuronal response (Cottrell and Davies, 1987), none of the peptides with an N-terminal extension is as active as FMRFamide. The synthetic peptide tyrosyl-glycyl-glycyl-phenylalanyl-methionyl-arginyl-phenylalanyl-NH₂ (YGGFMRFamide, related to met-enkephalin) was approximately 300 times less active than FMRFamide, and no effect could be detected with another endogenous peptide, asparagyl-glycyl-histidyl-tyrosyl-methionyl-arginyl-phenylalanyl-NH₂ (NGHYMRFamide), which is encoded by the cDNA sequence for the ‘heptapeptide’ FMRFamide-like peptides (Cottrell et al. 1994). Several other related peptides did not activate the current (see Lingueglia et al. 1995).

The conductance of FaNaCh expressed in Xenopus oocytes was 13 pS in cell-attached patches (with 140 mmol l⁻¹ Na⁺ in the recording pipette); the channel showed a slight increase in open probability as the potential on the inner surface of the membrane was increase from −60 to −100 mV. With a concentration of FMRFamide of 30 μmol l⁻¹, the frequency distribution of the open state was fitted with a time constant of 4 ms, and that of the closed state with two time constants of 8 ms and 30 ms. It was noted that open states lasting several seconds occurred rarely. 100 μmol l⁻¹ amiloride effectively blocked the unitary currents in outside-out patches (Lingueglia et al. 1995).

The single-channel conductance was similar with FLRFamide on oocyte-expressed FaNaCh, but at all concentrations tested the open-state time constant was shorter; 1.9 ms for FLRFamide compared with 3.9 ms for FMRFamide with the concentrations used in Fig. 4 (A. B. Zhainazarov and G. A. Cottrell, unpublished data), which at least partly accounts for its smaller maximum effect on the intact oocyte. With both peptides, open states lasting for several seconds were occasionally observed, as noted by Linglueglia et al. (1995) for FMRFamide.

Some properties of the neuronal- and oocyte-expressed channel are compared in Table 2. The most notable difference is the rate of desensitization in response to FMRFamide; the C2 neurone response rapidly desensitized with repeated FMRFamide applications, but desensitization of the oocyte-expressed FaNaCh is very slow. There is also a difference in potency of approximately threefold. This concentration difference may, however, be exaggerated because the neuronal response could decay in accessible regions of the large neurone (as a result of the marked desensitization), while access is being gained to more remote regions. This could influence the apparent affinity of the peptide. Peptide solutions are usually applied locally from a fine-tipped pipette onto the neurone, to reduce desensitization, whereas bath application is used with the oocytes. At present, the basis of neuronal desensitization is not known. Desensitization is far less pronounced, and sometimes not seen, in isolated outside-out neuronal patches. Thus, the observed differences in potency and also desensitization do not necessarily represent major differences between the neuronal channel and oocyte-expressed FaNaCh. The single-channel conductance of the neuronal channel is 4 pS, and it is 11–13 pS for the expressed FaNaCh. Can this difference be accounted for in terms of a difference in environment within the lipid membrane of the Helix C2 neurone compared with the oocyte? We have recently determined that the conductance of FaNaCh is 12 pS in a cell line permanently expressing FaNaCh, kindly supplied by Drs M. Lazdunski and E. Lingueglia. Thus, in the cell membrane of two different cell types, the heterologously expressed FaNaCh protein has a similar conductance, highlighting the difference from the native channel. Perhaps the difference is due to a different subunit composition. Alternatively, there may be a factor within the Helix neurones that influences the channel’s conductance.

With both the expressed FaNaCh and the neuronal channel, maximal responses to FLRFamide were always smaller than to FMRFamide, suggesting that FLRFamide is a partial agonist. The time constant of the open state of the oocyte-expressed channel with FLRFamide was approximately half that of FMRFamide. Some data recently obtained from the neuronal channel also indicate that the open-state time constant for FLRFamide is much less than for FMRFamide (Fig. 5).

Accurate determination of the amiloride ID₅₀ is more difficult on the neurone than the oocyte because of the technical problem of maintaining long-term recordings from the C2 neurone and also because of the presence of the second slower component of the neuronal response. Earlier experiments suggested an ID₅₀ of approximately 10 μmol l⁻¹; more recently, data have yielded a value of 1 μmol l⁻¹. The latter is the same as that for amiloride on the heterologously expressed FaNaCh.

FKRFamide (phenylalanyl-lysyl-arginyl-phenylalanyl-NH₂) antagonised the action of FMRFamide on the Helix C2, as observed on the expressed FaNaCh. A blocking effect has been observed both on the whole neurone response and on activated unitary currents (see Fig. 5C). FKRFamide is the first effective competitive antagonist found for an FMRFamide receptor. Further ‘molecular roulette’ is likely to yield even more effective antagonists now that more is known about the structural requirements for activation.

The FMRFamide-gated ion channel of Helix aspersa neurones is the first peptide-gated ion channel to be discovered. For a ligand-gated channel, it is unique in terms of its Na⁺-selectivity. Previous work on transmitter-gated channels shows that they are generally selective for small cations or Cl⁻. The significance of the Na⁺-selectivity is not known.

It is not yet clear whether the natural channel is a homo-oligomer of expressed FaNaCh or if it comprises other subunits, as with the epithelial Na⁺ channels. The comparisons described above, of the expressed FaNaCh and the neuronal responses, show that they are not identical. However, the differences are not great and some can be explained in terms of differences in desensitization (notably not seen in some isolated neuronal patches). The differences are very much less than those observed with the ENaCh subunits, where current was increased by a factor of 100 when three subunits (α, β and γ) were co-expressed rather than one alone. Considered in terms of such marked differences, there is little compelling evidence at present for more than one subunit of the peptide-gated Na⁺ channel. However, detailed studies on the response of the Helix C2 neurone have provided evidence for a second, slower component of the FMRFamide response and suggest that this component also results from the gating of a channel (Green et al. 1994). This channel may be related to FaNaCh. Is the peptide-activated channel involved in fast synaptic transmission? Evidence at present is indirect: (a) FMRFamide exerts a fast response; (b) the cerebral ganglia, in which the C2 neurones occur, are rich in FMRFamide; (c) the peptide is detected in neuronal somata and axon processes (Cottrell et al. 1983); (d) neuronal release of FMRFamide has been detected in the periphery (see Bewick et al. 1990). Nevertheless, blockade of the synaptically evoked fast FMRFamide reponse remains to be demonstrated. This may now be possible using FKRFamide or better antagonists, if they can be found.

Unlike the other proposed receptors for FMRFamide, the peptide-gated Na⁺; channel shows a high selectivity for FMRFamide. Such selectivity may partly explain the marked preponderance of FMRFamide (compared with the less readily oxidised FLRFamide) within precursor proteins from the molluscs so far studied (Greenberg and Price, 1992); the large number of FMRFamide copies could relate to the higher concentration of peptide required for activation of the ligand-gated channel compared with G-protein-coupled responses.

The physiological importance, if any, of the unique receptor-antagonism by the endogenous peptide pQFYRFamide is not known. The antagonism is weak and appears to be insignificant considered in terms of the many copies of FMRFamide encoded and processed from the precursor protein. However, some natural antagonistic role for pQFYRFamide can be envisaged if the precursor domain that encodes all the FMRFamide molecules (plus one of FLRFamide) is processed and targetted separately from the domain encoding the two pQFYRFamide molecules and the other molecule of FLRFamide. The two domains are separated by a furin-recognition sequence (see Hosaka et al. 1991), likely to be the first cleavage site, so such a scheme, although speculative, is feasible. Another identified Helix neurone, the C3 neurone, contains FMRFamide (Cottrell et al. 1983). It is interesting to note that all three peptides can be detected in the C3 neurone in approximately the ratio predicted from the cDNA for the precursor protein i.e. 10 FMRFamide: 2 FLRFamide: 2 pQFYRFamide (Price et al. 1996a). It may also be relevant that pQFYRFamide is more potent than FMRFamide in evoking the slow increase in K+ conductance.

Finally, the discovery of the Helix peptide-gated channel raises the question of whether there are similar ligand-gated channels in other groups of animals (others almost certainly occur in other molluscan species). At present, FaNaCh is the only example of a peptide-gated, amiloride-sensitive channel of the group typified by epithelial Na⁺; channels (see North, 1996; García-Añoveros et al. 1997). MNCh (i.e. MDeg or MNaC1) and MNaC2 are expressed in mammalian neurones, but their roles are not known. So far, the only related mammalian neuronal sequence (ASIC) whose role appears to be established is a proton-gated channel (Waldmann et al. 1997). The genetic studies on C. elegans predict many different sequences related to the amiloride-sensitive mammalian epithelial channels that are expressed in neurones and muscle cells, possibly at different stages of development. Other animals may similarly express many related sequences in neurones. It is, however, already clear that channels of this group serve diverse roles: in transporting epithelia, they regulate Na⁺ transport; in neurones, at least one (the Helix FaNaCh) is peptide-gated, another (ASIC) is a proton-gated channel and yet others are probably involved in mechano-sensation.

I thank Drs Michel Lazdunski and Eric Lingueglia for supplying clones of FaNaCh, the Wellcome Trust for financial support and Dr M. Lazdunski for helpful discussions.