Block of an ether-a-go-go-Like K⁺ Channel by Imipramine Rescues egl-2 Excitation Defects inCaenorhabditis elegans

An egl-2 gain-of-function mutation causes multiple behavioral phenotypes

Two independently isolated dominant gain-of-function (gf) mutations in the C. elegans egl-2 gene (n693 and n2656) inhibit egg-laying and enteric muscle contraction, leading to Egl and Exp phenotypes. These mutant phenotypes are rapidly rescued by exposure to tricyclic antidepressants such as imipramine, suggesting that theegl -2 gene product is a tricyclic target. Pharmacological and cell ablation manipulations combined with behavioral assays suggest that the Egl and Exp defects are likely in the egg-laying and enteric muscles and not in the motor neurons that innervate them (Reiner et al., 1995; Weinshenker et al., 1995).

We tested egl-2 mutants for chemotaxis to volatile odorants, a behavior that is mediated by amphid sensory neurons.egl-2(gf) mutants were defective in chemotaxis to both isoamyl alcohol and benzaldehyde, which are sensed by the AWC amphid neuron (Bargmann et al., 1993). Similar to the muscle defects, the isoamyl alcohol defect was rescued by imipramine (Fig. 1). Althoughegl-2(gf) has a weak locomotory defect, they were outperformed in this assay by the more severely locomotion-defective unc-25 mutant (Fig. 1), suggesting that the egl-2(gf) chemotaxis defect is caused by sensory and not motor defects.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Volatile chemotaxis response ofegl-2 mutants to isoamyl alcohol. Chemotaxis indexes (CI) were calculated in the following way: CI = (number at attractant − number at control)/(total number) (Bargmann et al., 1993). Error bars indicate SD among individual trials. Theegl-2(gf) allele was egl-2(n693sd), theegl-2(lf) allele was egl-2(n693sd sa236), and the unc-25 allele was unc-25(e156). Considering p < 0.05 as significant,egl-2(lf) is not significantly different from wild type at any dilution, egl-2(gf) is significantly different from wild type at every dilution, unc-25(e156) is significantly different from egl-2(gf) at all isoamyl alcohol dilutions except 10⁻⁴, andegl-2(gf) + imipramine is significantly different fromegl-2(gf).

egl-2(gf) mutants also had an imipramine-sensitive anterior mechanosensory-defective (Mec) phenotype, which is mediated by the ALM mechanosensory neurons (Chalfie and Sulston, 1981) (Table 1). Previously characterized male mating and locomotion defects inegl-2(gf) mutant are likely neuronally mediated, and these are also rescued by imipramine (Trent et al., 1983; Weinshenker et al., 1995; data not shown). These results demonstrate that an egl-2(gf) mutation can compromise neuronal and muscle function, and that its interaction with tricyclics is similar in the two tissues.

egl-2(gf) revertants define three classes of eag mutations

To determine the loss-of-function (lf) phenotype for egl-2, we mutagenizedegl-2(gf) animals and isolated revertants that were no longer Egl or Exp. All revertants were tightly linked to egl-2 (Materials and Methods), suggesting that they contain second site suppressors within the egl-2gene.

To investigate the genetic properties of the revertants, we used the semidominance of the egl-2(gf) alleles in gene dosage tests. The enteric muscle contraction (EMC) defect of egl-2(gf) intrans to a deficiency of egl-2 is more severe than the gf in trans to a wild-type copy ofegl-2 (Table 2). A null mutation of egl-2 in trans to the gfwould be expected to behave like a deficiency, whereas non-null mutations might give different results. We put each revertant intrans to egl-2(gf) and measured EMC defects (Table 2). The revertants fell into three classes. Class I alleles (20 alleles), such as sa236, behaved like anegl-2 deficiency and probably are strong lf or null mutations. As homozygotes,egl-2(sa236) and other members of this class of revertants appear wild-type, having no gross defects in movement, feeding, fertility, defecation, egg laying, or chemotaxis to volatile odorants (Tables 1, 2, Fig. 1; data not shown). Class II alleles (four alleles) behaved like a wild-type copy ofegl-2. Class III alleles (eight alleles) suppressed thegf to a greater degree than a wild-type allele, and thus are probably dominant-negative alleles.

egl-2 encodes a voltage-gated K⁺ channel

To determine the role of egl-2 in cell excitation and the nature of the tricyclic antidepressant interaction, we clonedegl-2 (Materials and Methods). egl-2 encodes an eag-like voltage-gated K⁺ channel (VGK) (Fig. 2a) (GenBank accession number AF130443). The EGL-2 and Drosophila eag proteins are 54% identical in the N terminus, 75% identical in the transmembrane and pore domain, and 81% identical in the domain with homology to cyclic nucleotide-binding proteins (cNTP), but are divergent at the C terminus. The molecular lesion in the two independently isolatedegl-2(gf) alleles is an A478V change in the S6 domain. This alanine is conserved among nearly all VGKs described to date, suggesting that this residue is critical for normal channel function (Fig. 2b). The Class Isa236 mutation results in a stop codon at amino acid 84 (Fig. 2a), and thus is probably a null allele. We identified molecular lesions for four Class III dominant negative alleles. These lesions all cause single missense mutations affecting residues near S1 and S2: S213F (sa395), T230M (sa377), and T285I (sa378 and sa391) (Fig. 2a).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Alignment of EGL-2 with other K⁺ channels. a, Predicted amino acid sequence of the egl-2 gene product and the molecular lesions found in mutant alleles aligned with Drosophilaeag (EAG) and mouse eag (MEAG). The predicted transmembrane and pore domains are underlined, as is the region with homology to cyclic nucleotide-binding proteins (cNTP BINDING DOMAIN). Amino acid identities are boxed in black. Amino acids changed in mutant egl-2 alleles areoutlined in black and annotated on the line above. GFP fusion junctions are shown by vertical gray lines. b, Alignment of the sixth transmembrane domain (S6) of EGL-2 and other K⁺ channels. dEAG and dSHAKER areDrosophila channels, UNC-103 is a C. elegans erg homolog, HERG (human erg; mutated in long QT syndrome), KVLQT1 (mutated in long QT syndrome), and hSK1 (small conductance Ca²⁺-activated K⁺channel) are human channels, mSLO is a mouse large conductance Ca²⁺-activated K⁺ channel, and AKT1 is an Arabidopsis inward-rectifying K⁺ channel. The alanine that is mutated inegl-2(gf) alleles and the corresponding residues in other channels are boxed. GenBank accession numbers for sequences used: EGL-2 (AF130443), MEAG (UO4294),Drosophila EAG (M61157), DrosophilaShaker (M17211), UNC-103 (Z35596), HERG (UO4270), KVLQT1 (U89364), hSK1 (AF131938), mSLO (L16912), and AKT1 (X62907).

S6 mutation shifts voltage dependence of activation for EGL-2 and murine MEAG channels

To investigate the mechanism underlying the in vivophenotype of the egl-2(gf) mutations, we studied the properties of WT and gf mutant (A478V) EGL-2 channels expressed in Xenopus oocytes. Oocytes injected with wild-type egl-2 cRNA expressed voltage-dependent, noninactivating potassium currents characterized by unusually slow activation kinetics, requiring voltage steps of 5 sec to approach steady-state (Fig.3a). A conductance–voltage (G–V) plot of WT EGL-2 currents revealed a voltage of half-maximal activation (V₅₀) of −7 mV, centered within a voltage-operating range of ∼80 mV (Fig.3b). Oocytes injected with mutant egl-2(A478V) cRNA expressed similar nonactivating potassium currents, but with altered voltage dependence of activation (Fig. 3a).G–V plots for A478V EGL-2 channels revealed a V₅₀ of −40 mV, a shift of −33 mV relative to WT EGL-2 channels (Fig. 3b). This hyperpolarized shift in the V₅₀ of A478V EGL-2 channels allows a significant fraction of available mutant channels to be activated at hyperpolarized potentials. This is consistent with the observed mutant behavioral phenotypes of egl-2(gf) mutants, because a muscle or neuron expressing these mutant channels may be hyperpolarized and unable to respond appropriately to excitatory stimuli.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Functional properties of wild-type and gain-of-function EGL-2 and MEAG K⁺ channels inXenopus oocytes assayed by two-electrode voltage-clamp.a, Current records from oocytes injected with wild-typeegl-2, gain-of-function egl-2(A478V), and corresponding meag and meag(A492V) cRNAs. Currents evoked by families of voltage steps from a holding potential of −100 mV, in 10 mV increments [−90 to +40 mV foregl-2 and egl-2(A478V)); −90 to +50 mV for MEAG and MEAG(A492V)]. Outward tail currents observed from wild-type EGL-2 channels resulted from a repolarization step to −50 mV. A −100 mV repolarization step was used for all other channels.b, Normalized conductance–voltage relationships. Conductances were calculated from current amplitudes at the end of 5 sec [EGL-2 and EGL-2(A478V)] and 2 sec [MEAG and MEAG(A492V)] voltage steps, based on a reversal potential of −90 mV in ND96. Individual data sets were fitted by a single Boltzmann function, G/G_max = (1 + exp(−n(V − V₅₀)/kT))⁻¹, where n is the slope factor reflecting intrinsic voltage sensitivity, V₅₀ is the voltage at half-maximal conductance, andT is absolute temperature. Mean values were plotted with SEM.

A478 is positioned near residues thought to undergo gating-dependent conformational changes in the Shaker potassium channel (Liu et al., 1997; Holmgren et al., 1998). We therefore tested the ability of the A478V substitution to confer a gf phenotype in the murineeag homolog meag (Warmke and Ganetzky, 1994), by engineering an analogous missense mutation in MEAG (A492V). A492V MEAG channels exhibited a G–V shifted toward hyperpolarized potentials (V₅₀ = −13.7 mV) relative to WT MEAG (V₅₀ = −3.4 mV) (Fig. 3a,b), demonstrating a modest, but functionally similar effect of this substitution in a mouse eag-like channel

Imipramine blocks EGL-2 and MEAG channels

To determine the mechanism of imipramine rescue of theegl-2(gf) phenotypes, we assessed its ability to block EGL-2 channels in oocytes. Imipramine was able to block both WT and A478V EGL-2 channels (Fig.4). The derived Hill coefficients (s = 0.84, WT; s = 0.80, A478V) suggest a binding stoichiometry of one imipramine molecule to one channel. Similar results were obtained with WT MEAG channels (Fig. 4), except that imipramine inhibits WT MEAG currents with an ∼10-fold lower potency (K_d= 55 μm) than EGL-2 currents.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Imipramine inhibits EGL-2 and MEAG channels. Dose–response plots for wild-type EGL-2, EGL-2(A478V), and MEAG currents, measured at 40 mV, at the end of 5 sec [EGL-2 and EGL-2(A478V)] and 2 sec (MEAG) voltage steps from a holding potential of −100 mV. Individual data sets normalized to maximal current evoked in ND96 and fitted by the Hill function, I = 1 −(max/(1 + (K_d/[imipramine])^s)), where max is the maximal fractional inhibition, K_d is the imipramine concentration at half-maximal inhibition, and s is the slope factor. Mean values were plotted with SEM.

Block by imipramine at each test concentration was rapid and essentially complete within the ∼2 min allowed for equilibration between exchanges of bath solutions, at the holding potential of −100 mV. A variable secondary component of block was observed with much slower kinetics, perhaps reflecting state dependence of block (Kuo, 1998). This secondary component of block may account for the residual 10% of channels not blocked with our assay conditions (Fig. 4). Block was not observed to be reversible by washes of up to 30 min, consistent with a slow off-rate of imipramine. The rapid and dose-dependent inhibition of EGL-2 and MEAG channels provides evidence that imipramine directly blocks these channels and suggests that imipramine may block other EAG-like channels with similar efficacy.

EGL-2:: GFP fusions are expressed in muscle and neurons

To determine the cellular expression pattern of egl-2, we constructed two EGL-2:: GFP fusions (Materials and Methods; Fig. 2a). Consistent with the Exp defect ofegl-2(gf) mutants, the short fusion (4.1 kb of upstream egl-2 sequence) was expressed in the intestinal muscles, which are two of the four enteric muscles (Fig.5a). We hypothesize that inappropriately activated EGL-2 K⁺channels in these muscles compromise their ability to depolarize and contract. In addition to the muscle expression, the short fusion was expressed in the AFD, ALN, AQR, ASE, AWC, BAG, IL2, PLN, PQR, and URX neurons in hermaphrodites and males, and in a subset of ray sensory neurons in males (Fig. 5b; data not shown). The expression in AWC is consistent with the chemotaxis defect inegl-2(gf) mutants (Fig. 1). Because ray neurons are involved in mating (Liu and Sternberg, 1995),egl-2 expression in rays may underlie the previously characterized male mating defect of egl-2(gf) mutants (Trent et al., 1983).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

egl-2 expression. Anterior is left, dorsal is up. Merged Nomarski and GFP fluorescence images.a, Short EGL-2:: GFP expression in the intestinal muscles (posterior end of an adult animal). GFP-expressing cells posterior to the intestinal muscle are the tail neurons ALN and PLN in a slightly lower focal plane. b, Short EGL-2:: GFP expression in neurons in the lateral ganglion (anterior end of the animal). Posterior to the IL2 neurons, BAGL and AWCL are also visible, but faint. Other GFP-expressing cells in this region are either faint or out of focus. c, Long EGL-2:: GFP expression in the ALM mechanosensory neurons.d, Long EGL-2:: GFP expression in the vulval muscles used for egg laying. e, Deconvoluted image of full-length EGL-2:: GFP fusion expression in sensory neuron endings.

The long fusion (5 kb of upstream egl-2 sequence, the entire 10 kb egl-2 coding region, and 4.2 kb of downstreamegl-2 sequence) was highly localized at dendritic endings of neurons in the nose, but only faintly in the processes (Fig.5e). This localization is probably in the ciliated endings of head sensory neurons. Neuronal cell bodies in the lateral ganglion also label with this fusion (Fig. 5e). This fusion was also expressed in the ALM mechanosensory neurons (Fig. 5c), which can explain the egl-2(gf) anterior Mec phenotype (Table 1). Consistent with the Egl defect ofegl-2(gf) mutants, we saw occasional expression of this fusion in the vulval muscles used for egg laying (Fig. 5d).

egl-2(lf) phenotypes

We did not detect any gross behavioral defects inegl-2(lf) mutants. We speculate that there are many K⁺ channels with overlapping functions in the excitable cells of C. elegans, and that loss of any one of them does not greatly affect cellular excitability. Conversely, activation of some of these channels causes an excitability phenotype out of proportion to their wild-type contribution. This generalization is supported by the severe phenotypes caused bygf mutations, in contrast to mild or absent phenotypes caused by lf mutations, in other C. elegansK⁺ channels (Reiner et al., 1995; Elkes et al., 1997; Johnstone et al., 1997).

EGL-2 expression

The expression of the short EGL-2:: GFP fusion in two of the four enteric muscles is consistent with the enteric muscle contraction defect of egl-2(gf) mutants. Vulval muscle labeling was observed in occasional animals expressing the long fusion, which can explain the Egl defect inegl-2(gf) mutants. This is consistent with the pharmacology of theegl-2(gf) Egl phenotype, as agonists that act directly on the egg-laying muscles to promote egg laying in the wild type fail to do so inegl-2(gf) mutants (Trent et al., 1983; Weinshenker et al., 1995). Interestingly, nearly all neurons that express the fusions are sensory neurons, suggesting thategl-2 serves a common function in C. eleganssensory neurons of different modalities.

Like the long EGL-2:: GFP fusion, components of the odorant-sensing machinery are localized to the ciliated endings of amphid sensory neurons (Coburn and Bargmann, 1996; Sengupta et al., 1996; Colbert et al., 1997; Roayaie et al., 1998). It is possible that EGL-2 functions in sensory endings and is modulated by components of sensory transduction. Drosophila eag mutants are defective in odorant response (Dubin et al., 1998), and rat eag is expressed in the olfactory bulb (Ludwig et al., 1994), suggesting conservation of an eag function in olfaction. In addition, bovineeag is expressed in retinal photoreceptors and may encode the native I_Kx current, suggesting a general role for eag in sensory transduction processes (Frings et al., 1998).

K⁺ channel structure and function

Voltage-gated K⁺ channels display a remarkable heterogeneity of intrinsic properties, reflected in the molecular diversity of K⁺ channel genes (Wei et al., 1996). The fact that the EGL-2 S6 alanine 478 is highly conserved among all families of VGKs suggests that it is critical for some aspect of normal channel function. This is supported by the gf shifts in G–V conferred by a modest change to a valine in EGL-2 and MEAG. Changing the analogous alanine in KVLQT1 channels to valine or glutamate results in the cardiac arrhythmia long QT syndrome, a probable dominant-negative phenotype (Wang et al., 1996; Shalaby et al., 1997). These mutations focus attention on S6 as an important structure mediating channel function.

Examples of S6 amino acid substitution that either stabilize channel open states or accelerate transitions from closed to open states have been observed in a wide variety of other K⁺ channels. These include ShawK⁺ channels (Elkes et al., 1997; Johnstone et al., 1997), Slo-type calcium-activated K⁺ channels (Lagrutta et al., 1994), the yeast YKC1 K⁺ channel (Loukin et al., 1997), and the Shaker K⁺channel (Liu et al., 1997; Holmgren et al., 1998). Our data suggest that elements of the gating mechanism mediated by S6 may be conserved between K⁺ channels from diverse gene families. This suggestion is supported by the conserved “inverted teepee” structural motif revealed by the crystal structures of the Kcsa K⁺ channel from Streptomyces lividans (Doyle et al., 1998) and the MscL cation channel fromMycobacterium tuberculosis (Chang et al., 1998). In both structures, the ion conduction pathway is lined by a single tilted transmembrane α-helix contributed by each subunit. This critical α-helix is TM2 for Kcsa, TM1 for MscL and, by analogy, S6 for voltage-gated K⁺ channel subunits. Rotation and translational movements of TM1 and TM2 underlie pH-dependent gating of the Kcsa channel (Perozo et al., 1999) and may provide a conserved structural mechanism used for gating in other potassium channels. Mutations of critical S6 residues could affect gating properties by influence the energetics of these conformational changes. Our data with egl-2 and meag are consistent this possibility.

The egl-2(gf) revertants may provide information about K⁺ channel structure and function. The Class II revertants restore relatively normal function to the gf mutant channels. These are probably second site mutations that compensate for the gfmutation, and they potentially provide mechanistic information about channel gating. The Class III alleles behave as dominant negatives. These mutations may suppress the gf mutation by directly or allosterically countering the action of A478V in S6. Alternatively, the suppressors could inhibit proper channel folding or processing independent of A478V. These alleles may provide a useful in vivo model for studying dominant-negative K⁺ channel subunit interactions.

Imipramine block of eag channels

We show that EGL-2 and MEAG channels are rapidly blocked by externally applied imipramine in a saturable fashion, with a high affinity in the range of 10–100 μm. These properties are in agreement with previous studies describing imipramine block of native K⁺ channels from cardiac myocytes (Isenberg and Tamargo, 1985), peripheral sensory neurons (Ogata and Tatebayashi, 1993; Wooltorton and Mathie, 1993, 1995), and hippocampal neurons (Kuo, 1998). Both noninactivating and transient native potassium currents are blocked by imipramine, depending on the tissue source. All reported native imipramine-sensitive currents tested have binding constants in the range of 10–100 μm and a putative external binding site. Our results suggest that these imipramine-sensitive currents may be encoded by K⁺ channel genes of the EAG gene family. Consistent with this possibility, heterologous expression studies reveal that the EAG gene family is capable of producing homomeric K⁺ channels with a diversity of kinetic properties, encompassing both noninactivating (eag, erg, elk subfamilies) and transient (erg, elk subfamilies) channel types (Shi et al., 1997, 1998; Engeland et al., 1998). Among gf mutations in C. elegans that have been shown or hypothesized to activate K⁺ channels,egl-2(gf) is the only one that is rescued by imipramine (Reiner et al., 1995; Elkes et al., 1997;Johnstone et al., 1997), suggesting a specific tricyclic-EAG interaction. The higher concentrations of imipramine required to rescueegl-2(gf) behavioral defectsin vivo (∼25 μm; E. Round, personal communication) than to block EGL-2 in vitro (∼10 μm) is likely due to the general impermeability of the C. elegans cuticle (Lewis et al., 1980). Should imipramine show specificity for K⁺channels encoded by the EAG gene family, it may provide an attractive molecular substrate for the rational design of blocking agents that target members of this family (Mathie et al., 1998).

EAG channels, tricyclic antidepressants, and human disease

The interaction between tricyclic antidepressants and EGL-2 provides a link that may explain one of the most common side effects of these psychotropic drugs. Long QT syndrome is a cardiac arrhythmia that can either be inherited or caused by drugs that block K⁺ channels (Tan et al., 1995). One inherited form is associated with dominant-negative mutations inHERG, a VGK in the EAG gene family that encodes the cardiacI_Kr K⁺current (Sanguinetti et al., 1995; Trudeau et al., 1995). Tricyclics can cause long QT syndrome, and we speculate that they may do so in part by blocking HERG channels. In support of this model, imipramine can significantly block I_Kr currents in ventricular myocytes at concentrations found in the serum of patients taking clinical doses of the drug (∼1 μm; Baldessarini, 1989; Valenzuela et al., 1994). In addition, a human eag channel (h-eag) has recently been cloned, and both h-eag and rat eag channels are expressed in brain (Ludwig et al., 1994; Occhiodoro et al., 1998). h-eag and meag are 100% conserved in the transmembrane and pore regions, indicating that imipramine probably blocks h-eag channels. Our results suggest that acute cardiac and neurological side effects of imipramine use may result from the block of HERG, h-eag, or other K⁺channels encoded by the EAG gene family.

Putative targets of psychotropic drugs have been determined traditionally by assaying binding to known receptors in vitro. The combination of genetics and pharmacology in C. elegans offers an alternative means to identify and study these targets in vivo.