Chimeric Mutations in the M2 Segment of the 5-Hydroxytryptamine-gated Chloride Channel MOD-1 Define a Minimal Determinant of Anion/Cation Permeability*

The ionic selectivity of ligand-gated ion channels (LGICs) determines whether receptor activation produces an excitatory or inhibitory response. The determinants of anion/cation selectivity were investigated for a new member of the LGIC superfamily, MOD-1, a serotonin-gated chloride channel cloned from the nematode Caenorhabditis elegans. In common with other anionic LGICs (glycine receptors and GABAA receptors), the selectivity triple mutant in the pore-forming M2 segment (proline insertion, Ala → Glu substitution at the central ring, and Thr → Val at the hydrophobic ring) converted the selectivity of MOD-1 from anionic to cationic. Unlike other LGICs, however, this mutant in MOD-1 was highly selective for K+ over other cations. Subsets of this selectivity triple mutant were studied to define the minimal change required for conversion from anion-permeable to cation-permeable. The double mutant at the central ring of charge (ΔPro-269/A270E) produced a non-selective cation channel. Charge reversal at the central ring alone (A270E) was sufficient to convert MOD-1 to cationpermeable. These results refine the determinants of ion-charge selectivity in LGICs and demonstrate the critical role of the central ring of charge formed by the M2 segments.

The ionotropic receptor MOD-1 (for modulation of locomotion defective 1) has been cloned in the nematode Caenorhabditis elegans (1). This homomeric assembly has been characterized as a 5-hydroxytryptamine (5-HT) 1 ligand-gated chloride channel with a predicted protein structure similar to that of the members of the nicotinic acetylcholine receptor family of ligand-gated ion channels (LGICs). These receptors, which also include ␥-aminobutyric acid (GABA), glycine, and cationic 5-HT3 channels, are structurally similar allosteric membrane proteins (2, 3) that mediate fast synaptic transmission (4,5). They are believed to assemble as pentamers, allowing formation of a central water-filled pore (6) and are differentially selective; the 5-HT 3 receptor and the nicotinic acetylcholine receptor (nAChR) are cation-selective, whereas the glycine re-ceptor (GlyR) and the GABA A receptor are anion selective. Individual subunits are predicted to contain a large N-terminal domain and four transmembrane-spanning domains (M1-M4), with the M2 domain forming the walls of the pore (7,8). Agonist binding to the N-terminal domain of the LGICs promotes a rearrangement that converts the conformation of the M2 domains from a non-conducting to a conducting open state (9 -11). Identification of the molecular determinants of ion selectivity (12) has been based in part on a comparison between anionic and cationic selective ionotropic receptors in such a closely related family (for review, see Ref. 13). Site-directed mutagenesis studies performed on nAChRs have identified rings of residues that alter channel gating (14), conductance (15), or the selectivity for monovalent (15)(16)(17) or divalent cations (18,19). Studies of chimeric substitutions between the cationic M2 regions allowed the identification of three essential residues for determining selectivity of the pore (19 -21). Subsequently, in the ␣1 GlyR a single mutation has also been identified that converts the selectivity of this channel (19).
MOD-1 offered an advantage to test the basis for anion selectivity over previously studied LGICs because it is a homomeric receptor of identical subunits (1). We analyzed the basis for ionic selectivity of MOD-1 using different mutants that explored the roles of the extra-and intra-cellular rings of charge. A minimal substation sufficient to convert the selectivity from anionic to cationic was identified at A270E. In addition, we also demonstrated that in MOD-1 the triplet of mutations not only induced a conversion in charge selectivity but also gave rise to a highly K ϩ -selective channel (Fig. 1).

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis-Complementary DNA encoding the wild type (WT) MOD-1 was subcloned into the pGEMHE expression vector (22). Site-directed point mutations were constructed by oligonucleotidedirected PCR mutagenesis using Pfu Turbo polymerase (Stratagene). The mutations were confirmed by sequencing with at least two different primers.
Expression in Vitro--For transcription, the expression construct was first linearized by PST1. RNA transcripts encoding MOD-1 and mutants were synthesized from linearized constructs using the T7 mMessage mMachine kit (Ambion, Austin, TX). The quality of RNA was confirmed by agarose gel analysis. Stage V oocytes were removed from adult female Xenopus laevis frogs and prepared as described previously (23). Oocytes were incubated in ND-96 medium that consisted of 96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 1 mMMgCl 2 , and 5 mM HEPES, pH 7.6, supplemented with 0.1 mg/ml gentamycin and 0.55 mg/ml pyruvate. RNA encoding WT or mutant versions of the MOD-1 channel was injected at ϳ50 ng/oocyte. Oocytes were incubated at 18°C for Ͼ24 h in ND-96 medium before voltage clamp experiments.
Two-electrode Recordings-Oocytes were placed in the recess of a small chamber and continuously superfused by gravity-driven flow (10 ml/min) with control or test solutions selected by computer-controlled * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The reversal potential (E rev ) was measured by continuously recording the membrane current as the voltage was ramped from Ϫ70 mV to ϩ50 mV over 50 ms and then back again. This triangular voltage stimulus was applied repeatedly as the agonist (1 M 5-HT) was washed over the oocytes. Ligand-activated channel opening results in more current flow whenever the membrane potential is away from its reversal value. Thus the E rev of 5-HT-activated channels was readily apparent from a superposition of current responses before and after the application of 5-HT. The data in Figs. 2 and 3 show MOD-1-specific currents, computed by subtracting the total current recorded before application of 5-HT. Ionic selectivity was determined by measuring shifts in E rev as the anionic composition of the bath was altered. Bath solutions contained 100 mM of the test ions (NaCl, KCl, choline chloride, or sodium gluconate) plus 0.2 mM calcium gluconate and 5 mM HEPES. The relative permeability of two cations, a and b (or two anions), was estimated from the shift in the reversal potential according to the Goldman-Hodgkin-Katz equation shown here as Equation 1, where R, T, and F have their usual meaning, E rev a is the reversal potential obtained with external solution a, and E rev b is obtained with solution b. This computation assumes that the permeability of the counter ion to a or b is negligible. For the determination of agonist dose-response relationships (Fig. 5), an amplitude normalization was required because channels did not completely recover from the desensitization induced by high concentrations (Ͼ3 M) of 5-HT. The peak current was measured in response to brief (5 s) applications of 5-HT beginning at 0.03 M, followed by a 5-min wash and then subsequent test concentrations up to 1 M. Complete recovery from desensitization was verified by repeating the measurement at 1 M after a 20-min wash. Finally, a single response at 3, 10, or 30 M 5-HT was measured, but no additional concentrations could be tested because of desensitization. The responses for each egg were normalized to an amplitude of 1 at a 5-HT concentration of 1 M. Equation 2, shown here, was fitted with normalized data pooled from multiple eggs. Finally, the data were renormalized by dividing by I max to generate the amplitudenormalized curves shown in Fig. 5. Data were analyzed using Clampfit (Axon Instruments, Foster City, CA) and Origin (MicroCal, Northampton, MA), and all parameter estimates are given as the mean Ϯ S.E. for n independent experiments.  Table II. The changes in E rev demonstrate that MOD-1 is a chloridepermeant anion channel. The reversal potential of WT MOD-1 was relatively insensitive to the external cation; E rev was Ϫ27.5 Ϯ 1.1 mV (n ϭ 11) for NaCl, Ϫ31.9 Ϯ 2.1 mV (n ϭ 11) for KCl, and Ϫ24.3 Ϯ 2.1 mV (n ϭ 11) for choline chloride. All of these values nearly matched the typical equilibrium potential for Cl in oocytes of about Ϫ25 mV in 100 mM external Cl Ϫ . In contrast, switching from NaCl to sodium gluconate shifted E rev by ϩ66 mV. The interpretation is that MOD-1 is much more permeable to Cl Ϫ than gluconate (20).

Determinants of Ionic Selectivity for MOD-1 Are Conserved Among the LGIC Family Members-WT
To test whether MOD-1 shares structural determinants of ion selectivity in common with other LGICs, several sets of mutations were introduced in the M2 segment based on conserved motifs in the pore region (Table I). Galzi et al. (19) showed that a triplet of mutations ( Fig. 1) is required to change the selectivity of the ␣7 nAChR channel from cationic to anionic. We tested the hypothesis that the inverse mutation in the anion-selective channel MOD-1 will convert this channel into a cation-selective one. Fig. 3 shows I-V relationships measured for the selectivity triple mutant (mutant 1) using different external solutions, and the mean E rev values are listed in Table  II. As expected, mutation 1 converted the selectivity from an anionic to a cationic channel. Quite unexpectedly, however, mutant 1 was highly selective for K ϩ over other cations (Fig. 3). The reversal potential was insensitive to an extracellular anion (E rev for NaCl was Ϫ68.8 Ϯ 1.8 mV, n ϭ 8; for sodium gluconate E rev was Ϫ70.4 Ϯ 1.4 mV, n ϭ 8) and remained near Ϫ70 mV for sodium or choline used as the cation, but it shifted to Ϫ6.0 Ϯ 1.1 mV (n ϭ 8) for KCl. Based on this 70-mV shift in E rev , the relative permeability for P K ϩ/P Na ϩ or P K ϩ/P choline is 12:1. This observation was surprising, because the comparable mutation in GlyR channels results in a cation non-selective channel (25).
As has been observed in other LIGCs (15,26), the internal and external rings of charge were not primary determinants of cation versus anion permeability. Mutant MOD-1 receptors in which the positively charged lysine in the inner ring was replaced by negatively charged aspartate remained anion-per-  Table II. NaGluc, sodium gluconate. meant and impermeant to cations (Fig. 3, Mutant 6). One difference from WT channels was that the E rev in sodium gluconate was more positive by 15 mV (Table II), which implies that the selectivity for chlorine over gluconate is increased in mutant 6. The external ring of charge in MOD-1 is positive because of Lys-292, whereas in cation-selective channels this residue is negative (Asp or Glu; Fig. 1). The introduction of a negative charge in the external ring of MOD-1 (K292D, mutant 7) did not change the selectivity from anion to cation. The reversal potentials remained relatively insensitive to external cations, whereas for anions switching from NaCl to sodium gluconate the E rev shifted by ϩ90 mV (Table II). Thus, in comparison to WT, replacing the positive charge in either the inner (mutant 6) or the outer (mutant 7) ring by inserting a negative aspartate residue increases the selectivity for chlorine over gluconate. The permeability ratio for P Cl Ϫ/P Gluc is 12 for WT, 30 for mutant 6, and 34 for mutant 7.
A Minimal Set of Mutations for Determining Anion/Cation Selectivity-Mutational studies of nAChR channels identified a triplet of substitutions required to convert selectivity from cationic to anionic (19). More recently, Keramidas et al. have shown in the GlyR that part of this triplet, a single Ala-to-Glu mutation in the intermediate ring of charge in M2, is sufficient to change selectivity from anionic to cationic (20). We constructed subsets of the selectivity triple mutant (mutants 2, 3, 4, and 5 in Table I) to look for the minimal set of residues in MOD-1 required to determine anion/cation selectivity. First, we eliminated the polar to hydrophobic change of the triple mutant. The resulting double mutant at the intermediate ring (⌬Pro-269/A270E, mutant 2 in Table I) was cation-permeant but, unlike the triple selectivity mutant, was not K ϩ -selective (Fig. 3). The modest 7-mV shift of E rev in switching from NaCl to sodium gluconate shows that the permeability to chlorine is relatively low. Conversely, the Ϫ36 mV shift of E rev from NaCl to choline chloride demonstrates substantial sodium permeability. Mutant 2 does not strongly discriminate cations tested in these experiments, and the E rev values show a permeability sequence of potassium Ϸ sodium Ͼ choline. We interpret these data as evidence that mutant 2 is a non-selective cation channel. The complementary mutation, wherein Pro-269 and Ala-270 remained unchanged but the third element of the triple selectivity mutant (T284V) was studied in isolation, did not alter the ionic selectivity from that of WT MOD-1 channels (mutant 5; Table II).

Differences in the intermediate ring of charge between anion-permeant and cation-permeant
LGICs are totally conserved. In MOD-1 and all other anion permeant channels, a PA occurs at this ring (Fig. 1). For cation-permeant LGICs the proline is absent and Ala is replaced by Glu. Deletion of this proline alone in MOD-1 (mutant 4) resulted in nonfunctional channels with no detectable 5-HT-activated current. The charge substitution alone, A270E (mutant 3), produced a cat-ion-permeant channel as demonstrated by a negative E rev of Ϫ38.5 Ϯ 0.2 mV (n ϭ 2) in a N-methyl-D-glucamine bath. Mutant 3 did not discriminate between Na ϩ and K ϩ and was only modestly less permeant to choline (Table II).
Mutations in the M2 Selectivity Filter Affect Gating but Not Ligand Sensitivity-The view that ion channels are built from modular assemblies of ion-conducting and channel-gating domains is a useful first approximation, but detailed examination of channel function or a mutational analysis often show interactions between permeation and gating. Therefore, we examined the macroscopic current kinetics and sensitivity to 5-HT of the different mutants. Representative examples of currents recorded in response to prolonged application of 1 M 5-HT are shown in Fig. 4. MOD-1 rapidly desensitized under these conditions, whereas the cation-permeant mutants desensitized more slowly and with a single exponential decay (Fig. 4). Quite remarkably, the single Ala-to-Glu mutation in the inner ring (the minimal change sufficient to convert MOD-1 to cation-permeant) produced a non-desensitizing receptor. No decay in 5-HT-activated current was detected even after 20 min (Fig. 4). For each of these mutations that slowed desensitization, the rate of activation appeared to be slowed as well (Fig. 4); but this measurement was limited by the sluggishness of whole-oocyte perfusion. Mutations in the inner or outer rings of charge did not affect the desensitization rate. When either of these Lys residues in MOD-1 was mutated to Asp (mutants 6 and 7), the 5-HT activated current rapidly desensitized, similarly to WT. The EC 50 values for mutant channels were within a factor of 3, higher or lower, than that of WT channels (Table I) and did not show a pattern between change in selectivity and relative change in EC 50 . DISCUSSION Many experiments made on nAchRs have demonstrated that the channel pore of the LGICs was mainly localized in the putative M2 intermediate ring of charge, with the most profound effects on conductance and selectivity being obtained with mutations done in this region (15,16). The organic cation permeability has been demonstrated to be dependent on the size and charge of these residues (29) and, furthermore, an adjacent ring of polar amino acids affected the monovalent cation permeation (30,31). The relation between gating and accessibility to M2 residues in the aqueous inner vestibule of the pore has been demonstrated by cysteine-scanning mutagenesis coupled with cross-linking by thiol reagents (27,28).
In more recent work on the focus of cation/anion selectivity, Keramidas et al. (20) demonstrated that introducing both a negative charge (mutation A251E) and a conformational change to displace the positively charged residue Arg-252 caused primarily by the Pro-250 deletion in this constricted region was required to induce a selectivity change in the GlyR and that even the T265V mutation may not be required. The A270E mutation was enough to convert the selectivity of the GlyR, suggesting that the introduced negatively charged glutamic acid residue overrode the adjacent positively charged arginine and confirming the relevance of the electrostatic effect in charge selectivity. In our experiments we showed that, like the GlyR, the charge selectivity conversion was achieved by mutations made in the intermediate ring of the MOD-1R. Our results demonstrated that two adjacent point mutations (⌬Pro-269 and A270E) produced a channel selective for cations. The proline residue had been shown to be implicated in local geometrical change to the pore constriction in the ␣7 nAChR triple mutant (19,24) or in WT GlyR (32). We thus can hypothesize that the deletion of this residue has induced modifications in the pore region at the conducting state level of the channel. Cation permeation should be provided by a suitable electrical potential and geometric environment, allowing the substituted glutamate (Ala-270) to be exposed to cationic ions (19). Because the single mutation A270E produced a cation selective channel but the P269D failed to form functional channels, we can conclude that electrostatic properties and pore geometry are mainly responsible for the conversion of selectivity. The introduction of a negatively charged residue (A270E) is also neces-sary for obtaining a functional cationic channel. The mutant 3 was a cationic non-selective channel for small cation ions. More precisely, we noticed that the selectivity was higher for K ϩ ions with a rightward shift in E rev values than that for the Na ϩ ions. Na ϩ ions were the most heavily hydrated in contrast with Cl Ϫ ions, which were the most easily dehydrated and require closer interactions with the charged residues of the selectivity filter. In both cases, the diameter of the selectivity filter sets constraints on how easily the charged residues of this filter can compensate adequately for the hydration energy of the ions. A similar precise fit hypothesis has been put forward to explain the high selectivity for K ϩ ions in the KscA K ϩ channel (33). Dutzler et al. (2002) (34) studied the ClC Cl Ϫ channel and, as for the KscA K ϩ channel, ␣-helix dipoles in both cases contribute significantly to the charge selectivity filters. For the LGIC family, it is strongly suggested that the pore helix does not have a great effect on the charge selectivity. The changes in charge of the intracellular and extracellular rings from positive to anionic charge reduced anion selectivity to render the channel only mildly anion-selective. This suggested that both extracellular and intracellular rings have an effect on the selectivity of the channel. Taken separately, however, no conversion in selectivity was observed (mutants 6 and 7). Modifications in the intermediate ring were required to convert the selectivity of the channel. Here we showed the drastic effect of the mutation T284V in mutant 1 induced selectivity for K ϩ ions. At that point, we can argue that the geometrical changes in the pore induced by the introduction of a valine residue known to affect the geometry of the protein gave rise to a conformational change of the pore sufficient to enhance the selectivity for K ϩ ions.
For mutants 2 and 3 K ϩ was more permeant than Na ϩ , even though it is a larger ion. These results imply that the channel walls were lined with polar residues (threonine and serine), which provide a hydrogen bond acceptor. For the mutants 8, 9, and 10, the changes of charge in extracellular or intracellular rings had an influence on the selectivity, and we hypothesized that the changes in charge have modulated the diameter of the pore.
In addition to the change in selectivity, MOD-1 mutants presented changes in their gating behavior, with mutants activating more slowly than the WT. Presumably, changes in the slower component of desensitization were responsible for the different gating compared with WT.   (35) suggested that desensitization was not involved in the reduced rate of current decay in the triple mutant compared with WT. We showed here that a single exchange of one amino acid (mutant 3) prevented the desensitization of the MOD-1 receptor even if the process of LGIC desensitization was thought to be coupled to binding of the agonist. However, little is known about the structure and the mechanisms underlying the gating process.
A loss of function has also been described by Keramidas et al. (35). It indicated that the threonine-to-valine mutation would impair the gating of the open or activate state. We did not observe spontaneous opening, in contrast to Corringer et al. (36) who showed a gain of function with an effect on the openclosed equilibrium by the threonine-to-valine mutation. For all the mutants except mutant 3, we noticed that these mutations resulted in a "loss of function." This property was characterized by an increase of affinity in response to 5-HT compared with WT, as shown by the sustained electrophysiological response following a 10-s application of 5-HT (Fig. 2, A and B), and an increase in EC 50 for these properties has led to the suggestion that the mutant receptors may attain a desensitized but conducting state in the presence of an agonist (19, 37, 38). Mutant 3 did not seem to be a conducting state. The channel closed after the application of the ligand and displayed a loss of function with an increase in the affinity for the ligand, (EC 50 of 0.75 as compared with 1.23 for WT). Hille (39) has demonstrated that the ligand-gated ion channels induced a rapid open state in response to an application of an agonist. Also, in the continuous presence of the agonist these channels may enter an agonist-bound but non-conducting "desensitized" state with a higher affinity for agonist than the closed resting or open states of the channel. The marked reduction in the rate of activation and removal of desensitization as shown by the double electrode recordings could therefore be due to a similar phenomenon. If one or more of the desensitized states of the receptor has been changed into a conducting one by the mutation, then the reduced rate of current decay may reflect passage from the open state to a conducting desensitized state before subsequent entry into a further (non-conducting) desensitized state. Such a model may explain the long duration open times observed even in responses to short applications of agonist, where the receptor may remain in a high affinity desensitized (conducting) state until agonist unbinding occurs. CONCLUSION The present study elucidates several key determinants of ion-charge selectivity in the MOD-1 receptor, a 5-HT ligand-gated chloride channel. We demonstrate that a single point mutation, A270E, was required for the change in the selectivity (19) that also dramatically affected the desensitization of this receptor. The ability to convert an ionic channel into a potassium-conducting channel (triplet of mutations P269D, A270E, and T284V) implies that homologies in the selectivity features of the voltage-gated potassium channels and MOD-1 receptor may exist. Further experiments using electron microscopy may improve the resolution of secondary structures of the LGIC receptors and improve understanding of the selectivity mechanisms of these channels. Ϫ61.9 Ϯ 9.6 (n ϭ 3) Mutant 2 Ϫ15.4 Ϯ 3.2 (n ϭ 6) Ϫ9.4 Ϯ 3.2 (n ϭ 3) Ϫ10.0 Ϯ 1.7 (n ϭ 7) Ϫ28.0 Ϯ 3.1 (n ϭ 3) Mutant 3 Ϫ16.5 Ϯ 0.86 (n ϭ 9) Ϫ7.7 Ϯ 2.4 (n ϭ 9) Ϫ11.1 Ϯ 1.5 (n ϭ 9) Ϫ20.8 Ϯ 0.95 (n ϭ 4) Mutant 5 Ϫ28.6 Ϯ 2.7 (n ϭ 6) 42.3 Ϯ 4.4 (n ϭ 5) Ϫ30.0 Ϯ 2.0 (n ϭ 7) Ϫ22.9 Ϯ 1.1 (n ϭ 7) Mutant 6 Ϫ31