The seizure Locus Encodes the DrosophilaHomolog of the HERG Potassium Channel

Identification of the seizure transcript

Because the molecular nature of the seizure gene product was unknown, positional cloning was the method of choice to identify the transcript. As summarized in Figure1A, the seizure gene was mapped first to the cytogenetic region 60A8–B10. Then a YAC clone that mapped between 60A10–A13 and 60B12–B13 (Cai et al., 1994) was used to initiate a chromosome walk (summarized in Fig. 1D) through this region.

Because many regions of the Drosophila genome are rich with transcripts (see, for example, Feng et al., 1995b), we used two approaches to identify the seizure transcript: RFLP mapping and analysis of γ-ray-induced chromosome aberrations that disruptseizure function. Two γ-ray-induced seizurealleles (G50 and G87) showed altered restriction patterns within a 10 kb segment of the chromosome walk (indicated by a small black rectangle in Fig.1C). Thus, this 10 kb region was a candidate for at least part of the seizure gene.

RFLP mapping (Fig. 1B) was used to confirm, independently, the genomic location of seizure before cDNA cloning. The closer an RFLP is to seizure, the lower the frequency of recombination will be between them. The X intercept at position 118 kb of the chromosome walk (Fig. 1B) gives the approximate location of the seizure gene. Because the horizontal axes of Figures 1B–D are on the same scale, it can be seen that the position of seizure as defined by RFLP mapping coincides with the 10 kb region shown to contain either a small insertion or deletion in γ-ray-inducedseizure alleles.

A genomic fragment from the region proposed to contain theseizure gene detected two candidate transcripts (3.4 and 3.0 kb in wild type) that were altered in size or amount in sevenseizure alleles (Fig. 2). Some alleles (ts1 and G64) showed a dramatic reduction in both transcripts, as compared with wild-type (CS) (Fig.2, Table 1). In others, there was a change in transcript size. For example, in G87 both transcripts are 0.8 kb smaller than wild type. In G50 two transcripts are 0.5 kb larger, plus there is an additional small transcript at 2.2 kb. Bothts2 and G85 show an increase in intensity of the larger 3.4 kb transcript relative to the smaller 3.0 kb one. Northern blots probed with single-stranded riboprobes show that the 3.4 and 3.0 kb messages are transcribed in the same direction and are substantially overlapping (X. J. Wang and L. M. Hall, unpublished results). Changes in transcript size and/or intensity in so many independently isolatedseizure alleles provide compelling evidence that these cDNAs represent the seizure gene product.

The seizure gene encodes a potassium channel subunit of the eag family

Sequencing wild-type cDNA corresponding to the disrupted transcripts revealed a long ORF encoding a deduced protein of 855 amino acids (Fig. 3A, “S”lines) with a predicted molecular mass of 97.5 kDa. The dendrogram (Fig. 4) derived from database comparisons shows that sei is a member of the eag class of potassium channels, which also have similarities to cyclic nucleotide-gated cation channels (Guy et al., 1991). The complete sei protein shows a striking similarity (58% similarity; 47% identity) to HERG, a member of the eag family recently cloned from humans (Warmke and Ganetzky, 1994). The similarity and identity are even higher (84 and 72%, respectively) if the comparison is limited to the hydrophobic core. Identification of this new Drosophila potassium channel subunit, which is closer to a human potassium channel than to other Drosophilachannels, establishes sei and HERG as a distinct potassium channel subfamily, as predicted previously (Warmke and Ganetzky, 1994).

Structural features of the seizure protein

The hydropathy plot (Fig. 3B) and the alignment of sei- and HERG-deduced proteins (Fig. 3A) illustrates structural features that they have in common. First, each has six predicted transmembrane α-helices (Fig. 3, S1–S6) plus a pore region (P) between S5 and S6. The amphipathic S4 domain with positive amino acids (K, R) every third residue is important in voltage-sensing. The complete conservation of this domain between sei and HERG suggests they may have similar voltage-dependent properties. The sequence of the sei protein in the P region (SLTSVGFGN) is very similar to the potassium channel consensus sequence (tmttvG[y/f]Gd) (Chandy and Gutman, 1995), providing strong evidence that sei is a subunit of potassium-permeable channels.

There is also a striking identity in the proposed cyclic nucleotide binding domain (cNBD) in the C-terminal cytoplasmic region. However, the rest of the C terminus and most of the N terminus of the sei protein have very little in common with HERG or with other members of the eag potassium channel gene family. Both the N and C termini of the sei protein are significantly shorter than HERG.

There is a consensus site for N-glycosylation 12 or 14 amino acids upstream of the P region, which is conserved in all eag family members (Warmke and Ganetzky, 1994). In addition, three potential phosphorylation sites just downstream of S6 are conserved in sei and in HERG as is a phosphorylation site within the cyclic nucleotide binding domain. In total, the sei protein has 9 possible protein kinase C phosphorylation sites, 6 CaM kinase phosphorylation sites, 1 protein kinase A phosphorylation site, and 11 casein kinase phosphorylation sites on predicted cytoplasmic domains. These sites, especially those conserved between sei and HERG, suggest the potential for regulation of this channel by phosphorylation.

Mutations in seizure alleles

To determine the molecular basis for the recessive versus dominant phenotypes of different seizure alleles, we sequenced genomic DNA corresponding to the entire coding region from wild-type and homozygous mutant lines. We identified mutant changes in six different alleles. The alkylating agent ethyl methanesulfonate (EMS) generally induces point mutations, and that is what we found for the EMS-induced ts1 and ts2 alleles. Thesei^ts1 allele has a T to A change, which converts K282 (Fig. 3A) into a premature stop codon. The resulting truncated protein completely lacks the hydrophobic core that forms the transmembrane channel (Fig. 5). Barring read-through of the stop codon, this nonsense allele should be a functional null.

In sei^ts2 there is a G to A change, which changes a negatively charged group to a positively charged one (E490K in Fig. 3A) in a region close to the extracellular side of the S5 domain. This charge change is near the channel pore (Fig. 5) and could disrupt channel function. Because this is a missense mutation, sei^ts2 should still make a full-length subunit.

The four γ-ray-induced alleles diagrammed in Figure 5 are all recessive and show hyperactivity and temperature-induced paralysis phenotypes indistinguishable from the ts1 allele. Three (G50, G43, and G64) are frameshift mutations leading to a premature stop codon in the approximate positions shown in Figure 5. The G87 allele deletes a large region from just before the start of transmembrane 6 through to a point in the C terminus past the cyclic nucleotide binding domain (Fig. 5). Although they differ in the sites of change, none of the γ-ray-induced alleles would produce a functional channel, because they lack part or all of important transmembrane domains. Thus,sei^ts1, plus the γ-ray-induced alleles, defines the null phenotype of this gene and illustrates the consequences to the animal of complete absence of theseizure gene product.

The entire ORF also was sequenced for a fifth γ-ray-induced allele (G85, transcripts shown in Fig. 2), but no changes were identified despite the failure of this allele to complement otherseizure alleles. This allele may have alterations in the 5′ or 3′ untranslated regions or in an alternative exon not yet identified. Interestingly, G85 and ts2 both show an increase in the 3.4 kb transcript relative to the 3.0 kb transcript (Table 1). The reason for this shift remains to be determined.

Expression of a seizure⁺ transgene: paralysis rescue and estimation of channel half-life

Mutant strains that completely lack the seizure potassium channel provide a unique opportunity to define when expression of the gene product is needed to rescue the paralytic phenotype. To address this question, we made transgenic flies in a null mutant background. The transgene was wild-type seizure cDNA encoding the protein shown in Figure 3A and was under the control of an inducible heat–shock promoter. In the absence of induction (Fig.6, circles), the transgenic flies show a paralytic phenotype indistinguishable from the null mutant without the transgene. This suggests that uninduced expression of the transgene is not at physiologically relevant levels, at least with respect to the paralysis phenotype. In contrast, repeated induction of this transgene in adults (Fig. 6, squares) over a 5 d period rescues the paralytic phenotype in most flies by the sixth day. This experiment suggests that synthesis of this channel subunit in adults only is sufficient to restore normal locomotor function.

When induction of the wild-type transgene is stopped, the return of the paralytic phenotype should reflect the net rate of loss of functional channels from membranes and thus will provide an estimate of apparent half-life of functional channels in membranes in an intact animal. To estimate this half-life, just before reaching 100% rescue, we stopped induction of the transgene and monitored the return of the paralytic phenotype. As shown in Figure 6B, return of paralytic phenotype was rather rapid. The time to return to 50% paralyzed flies is ∼1-1.5 d after 94% rescue is achieved.

Potassium channel mutations are the direct cause ofseizure hyperexcitability and paralysis

RFLP mapping, identification of mutant changes in the coding region of genomic DNA from six independently isolatedseizure alleles, and transformation rescue with cDNA under an inducible promoter provide conclusive evidence that the potassium channel sequence presented here is the seizure gene product. The sequence conservation between seizure and other ion channels shows that it is a potassium channel of the eag subfamily and is related most closely to the HERG inwardly rectifying potassium channel. Althoughsei has not yet been expressed in Xenopusoocytes, the virtually complete conservation of the ion selective pore region (P) leaves little doubt that sei encodes a potassium channel α-type subunit. These results provide an explanation for the hyperexcitable phenotype seen with all sei mutant alleles. In general, decreases in potassium channel activity cause hyperexcitability because of prolonged depolarization during an action potential and/or failure to maintain a normally negative resting potential. Decrease in activity of this sei potassium channel subunit leads to hyperexcitability similar to that observed for other potassium channel subunit loss of function mutants, includingSh, eag, and Hyperkinetic (Wu and Ganetzky, 1992).

A recent report suggests that a “set” of cloned potassium channels (including Sh, Shab, Shal, and Shaw) accounts for most embryonic potassium currents in Drosophila (Tsunoda and Salkoff, 1995). Given the dramatic effects that sei mutants have on behavior, sei also should be included in the list of physiologically important potassium channel components inDrosophila.

Genetic evidence that the seizure potassium channel is a multimer

Null alleles of sei are completely recessive, indicating that the 50% reduction of gene product generally expected in a heterozygous null is without consequence to the behavior of the organism. This suggests that excess sei gene product is normally available in wild type. In contrast, the missense allelesei^ts2 has a partially dominant phenotype, although its overall hyperexcitability phenotype as a homozygote is less extreme than the null alleles (Kasbekar et al., 1987). This suggests that the potassium channel that contains the sei protein is a multimer, as are other potassium channels (Isacoff et al., 1990). The presence of one or more copies of a mutant subunit in a multimer could act as a dominant “poison” to disrupt the functional properties of the remaining wild-type subunits. It is interesting to note that missense mutations in HERG, the human homolog ofsei, also cause a dominant phenotype, in this case, of cardiac arrhythmia (Curran et al., 1995).

This genetic analysis cannot predict whether the sei potassium channel is a homo- or heteromultimer. There is, however, indirect evidence that suggests an additional component is required. cRNA from several different sei cDNA constructs fails to make functional channels when injected into Xenopus oocytes. Mistakes or an incomplete sequence cannot be the cause for this failure, because the same cDNA sequence rescues the sei behavioral phenotype in transgenic animals. This suggests that there is a factor required forsei expression that is present in the fly and absent inXenopus oocytes. This factor may be another pore-forming subunit, or it may be an auxiliary subunit like tipE, which stimulates expression of a Drosophila sodium channel (Feng et al., 1995a). Candidates for a component needed for seiexpression include a potassium channel β subunit, such as theHyperkinetic gene product (Chouinard et al., 1995), or an unidentified component, such as the enhancer of seizure gene product (Kasbekar et al., 1987).

Premature nonsense codons affect seizureRNA abundance

As reviewed by Maquat (1995), mRNAs containing a nonsense codon as a result of either frameshift or nonsense mutations often are reduced in abundance. The severity of the reduction is correlated with the closeness of the nonsense mutation to the normal initiation codon. The different sei nonsense alleles described here provide a unique opportunity to determine whether this relationship holds true for mRNAs that encode integral membrane proteins and are associated with the endoplasmic reticulum during translation. In general,sei transcripts show a similar effect of premature nonsense codons, as reported in other systems. Northern analysis ofsei alleles (Fig. 2, Table 1) shows that the presence of a premature nonsense codon in the first half of the message (ts1 and G64) causes a marked decrease in both sei transcripts, as compared with wild type. The case of G50 is complicated by the appearance of an extra transcript, which may be the result of a cryptic transcription start site. However, even if all three transcripts are considered, there is an overall reduction in total transcript level, as compared with wild type (after correcting for loading differences).

Rescue of paralytic phenotype with an inducible wild-typeseizure transgene

Expression of the seizure potassium channel subunit in adults only is sufficient to rescue paralysis. This stands in contrast to the requirement for the tipE gene product during pupal development to rescue adult paralysis (Feng et al., 1995a). Although mutations in both the tipE gene and the sei gene cause a temperature-sensitive paralysis, they seem to do so by different mechanisms that are distinct in time.

Potassium channel turnover in the whole animal

To our knowledge there has been no previous estimate of stability of sei/HERG potassium channels in an intact, multicellular organism. By inducing expression of the seizure⁺transgene and monitoring the paralysis phenotype, we were able to titrate expression to the point at which 94% rescue of the mutant phenotype was attained. Then induction was discontinued. The rate of return of the paralytic phenotype should reflect the net loss of functional channels from the membrane. With this method we estimate that the apparent half-life of functional sei channels is ∼1–1.5 d. This is consistent with metabolic labeling studies of sodium channels that estimated the apparent half-lives of the α subunit and the αβ2 complex to be 30 and 50 hr, respectively, in cultured embryonic rat brain neurons (Schmidt and Catterall, 1986). It is also consistent with estimates of calcium channel lifetime of 5–8 d inParamecium (Schein, 1976). There is one report (Zhao et al., 1995) suggesting that Shaker potassium channels in cultured “giant”Drosophila neurons in some genetic backgrounds turn over more rapidly than we report here. This could indicate differences in stability of different potassium channel types or could reflect differences in different mutant backgrounds.

Relationship between seizure and sodium channel regulation

How can the fact that seizure encodes a potassium channel subunit be reconciled with observations that the ts1allele is associated with changes in levels of functional sodium channels (Jackson et al., 1985; O’Dowd and Aldrich, 1988) and that thets2 allele has an altered pH dependence and alteredK_D for saxitoxin binding (Jackson et al., 1984)? We propose that the reduction in channel levels in ts1 is a consequence of sei effects on cell excitability. There are previous reports that link electrical activity to density of sodium channels (Dargent and Couraud, 1990; Dargent et al., 1994) (for review, see Catterall, 1992), including a report of downregulation of sodium channels in nerve terminals of epileptic mice (Willow et al., 1986). The effects of the ts2 allele on saxitoxin binding sites are more difficult to understand. They may be an experimental artifact because of the high incubation temperatures used in these studies. Alternatively, they may reflect a ts2-induced shift in expression of sodium channels with different binding properties as a result of alternative splicing (Thackeray and Ganetzky, 1994, 1995) or expression of different α-subunit genes [para, Loughney et al. (1989) vs DSC1, Salkoff et al. (1987)].

There are multiple mechanisms by which sodium channel density is regulated by electrical activity of the cell. In cultured rat muscle fibers, spontaneous electrical activity causes a feedback downregulation of sodium channel density by changing levels of mRNA encoding sodium channel α subunits (Offord and Catterall, 1989). Treatments that mimic increased electrical activity, such as elevation of cytosolic calcium with ryanodine or the ionophore A23187, also cause sodium channel downregulation (Sherman and Catterall, 1984; Brodie et al., 1989; Offord and Catterall, 1989).

Sodium channel downregulation in response to hyperexcitability also has been observed in neurons treated with sodium channel activators, including α-scorpion toxin, batrachotoxin, and veratridine (Dargent and Couraud, 1990). In these studies, the mechanism of regulation is by sodium channel internalization (Dargent et al., 1994) rather than by channel synthesis, as reported for muscle (Offord and Catterall, 1989).

Although the mechanisms for sodium channel downregulation differ for the two cases summarized above, they have in common that the response to hyperexcitability is a decrease in the density of functional sodium channels in the plasma membrane. We propose that the decrease in functional sodium channels associated with thesei^ts1 allele inDrosophila may be the consequence of hyperexcitability. To explain the normal sodium channel levels found in the ts2missense allele, we postulate that there is a minimum hyperexcitability level required to induce the downregulation. This level is reached in the null ts1 allele, but not in the missense ts2allele. The different sei alleles described here, plus mutations in other potassium channel components (Wu and Ganetzky, 1992), provide a unique opportunity to dissect genetically the role of potassium channel activity in sodium channel regulation.

GenBank accession number

The GenBank accession number for the seizure cDNA reported here is U36925.