Voltage Sensing in Jellyfish Shaker K⁺ Channels

The Shaker sub-family of K⁺ channel genes appears to be ubiquitous in both invertebrates (Papazian et al. 1987; Kamb et al. 1987; Tempel et al. 1987; Johansen et al. 1990; Pfaffinger et al. 1991; Kim et al. 1995: Jegla et al. 1995, Wei et al. 1996) and vertebrates (e.g. Baumann et al. 1988; Tempel et al. 1988). The Shaker sub-family is structurally distinct from other members (Shal, Shab and Shaw) of an extended gene family. Two Shaker genes, jShak1 and jShak2, recently isolated from the hydrozoan jellyfish Polyorchis penicillatus (Phylum Cnidaria) show marked similarity to other Shaker sequences (Jegla et al. 1995).

A critical structural element of the proposed voltage sensor (S4) of most voltage-gated ion channels is a distinct pattern of repeating motifs of the form Arg-X-X or Lys-X-X. It is believed that these positively charged motifs form a structure that senses large electrical fields across the membrane and can lead to channel opening (Noda et al. 1984, 1986; Greenblatt et al. 1985). The total charge in S4 may influence the steepness of the activation curve (Liman et al. 1991; Logothetis et al. 1992, 1993) and contribute to the threshold of activation and voltage-sensitivity of the channel (Papazian et al. 1991: Logothetis et al. 1993). The number of these positively charged motifs varies from four in Shal to seven in Shaker (Chandy and Gutman, 1995). jShak1 has only six positively charged residues in its S4 segment, whereas in all other Shaker sequences, including jShak2, there are seven charges in S4.

When expressed in Xenopus laevis oocytes, the jellyfish genes jShak1 and jShak2 produce a channel with activation and steady-state inactivation curves that have strong (approximately +40 mV) rightward shifts relative to all other Shaker genes (Jegla et al. 1995). This shift occurs despite the total charge differences in the S4 segments of jShak1 and jShak2. Recent studies have shown that charges in the S4 segment contribute unequally to the threshold of activation of the channel (Tytgat et al. 1993) and to the equivalent gating charge, with different charges moving different distances through the electrical field (Aggarwal and MacKinnon, 1996; Seoh et al. 1996; Larsson et al. 1996; Mannuzzu et al. 1996). Some charges are crucial for the formation of stabilizing electrostatic interactions with negatively charged residues in other segments (Papazian et al. 1991, 1995).

To examine the role of charge and segment length in the S4 region of jellyfish Shaker channels, we restored the ‘full complement’ of charge to the S4 region of jShak1 by insertions of the motif Arg/Lys-X-X and by modifying the sequence of the N-terminal side of S4. We have been able to distinguish between the effects of total charge in S4, charge position and length of the segment on the basis of voltage properties. Our results show that a lysine residue stabilizes the S4 segment and divides it into functionally separate regions.

These studies of early metazoan channels, such as the cnidarian channel proteins jShak1 and jShak2, in conjunction with studies of channels from more derived phyla provide us with a broader basis for understanding the critical elements involved in voltage sensing by the S4 segment.

Mutations were introduced into the S4 region of jShak1 using a polymerase chain reaction (PCR)-based technique for introducing insertions and point mutations (Stappert and Kemler, 1992). Sense primers containing appropriate insertions in the S4 region (3–7, Table 1) and an antisense primer complementary to the jShak1 sequence from positions 1122 to 1145 (2, Table 1) were used to amplify 3′ end fragments from 10 ng of jShak1 plasmid. In separate reactions, antisense primers containing the same insertions in the S4 region (8–12, Table 1) and an upstream sense primer complementary to the jShak1 sequence from positions 588 to 607 (1, Table 1) were used to amplify 5′ end fragments from 10 ng of jShak1 plasmid. The PCR products were gel-purified, and the pairs of PCR products containing complementary overlapping sequence (underlined sequence in Table 1) were used as a combined template in a second PCR reaction with primers 1 and 2 (Table 1).

PCR reactions (25 μl) were carried out with 1:16 diluted recombinant Pfu DNA polymerase (Stratagene) and Taq polymerase, 20 pmol of each primer, 20 mmol l⁻¹ Tris–HCl (pH 8.3 at 20 °C), 25 mmol l⁻¹ KCl, 100 μg ml⁻¹ gelatin, 50 μmol l⁻¹ each of dNTP and 1.5 mmol l⁻¹ MgCl₂ in 20 thermal cycles (93 °C, 45 s; 55 to 65 °C, 2 min; 72 °C, 2 min) followed by 10 min at 72 °C.

Products from the second PCR reaction were gel-purified, digested with HpaI and NsiI and inserted into HpaI/NsiI-cut jShak1, replacing the wild-type jShak1 sequence between positions 618 and 1102. Ligation reactions were carried out with equimolar amounts of insert and vector fragment in 50 mmol l⁻¹ Tris–HCl (pH 7.8), 10 mmol l⁻¹ MgCl₂, 1 mmol l⁻¹ ATP, 50 μg ml⁻¹ bovine serum albumin, 20 mmol l⁻¹ dithiothreitol and T4 DNA ligase (6 Weiss units, New England Biolabs) in a total volume of 20 μl. 1 μl of the ligation reaction was electroporated (2 mm cuvettes, 2.8 kV) into electrocompetent XL1-Blue bacteria (Stratagene) and plated on LB-ampicillin. Colonies were picked and grown overnight, and plasmid DNA was purified.

The internal primers for making the RIF and QIF mutants (3,4; 8,9; Table 1) and IFR and IFQ mutants (5,6; 10,11; Table 1) contained nine base pair inserts (double-underlining, Table 1) between positions 879 and 880 and between positions 882 and 883, respectively, in jShak1; these insertions disrupt a PmeI site, so plasmids were screened for the absence of this site. One expected RIF, PmeI-resistant clone contained a single base pair error with C in place of T in the eighth position of the insert, providing a unique mutant with insert RIS (AGA ATT TCT) instead of RIF. The RV for SM mutant contained AG in place of CA at positions 846 and 847 in jShak1 (7, bold type, Table 1), which disrupts a MwoI site, so plasmids were screened for the absence of this site. Large-scale preparations of plasmid DNA were then purified using a silica-based DNA purification method (Hengen, 1995). Double-stranded DNA sequencing of the region of each plasmid prepared by PCR was carried out on a Perkin-Elmer ABI 373A sequencer using an ABI Prism dye-terminator cycle sequencing kit to confirm that the new plasmids contained only the stated mutations.

Capped mRNAs were prepared by in vitro transcription using a T7 mMessage mMachine kit (Ambion). Adult female Xenopus laevis were anesthetised in MS 222, and a unilateral incision was made on the dorsal surface of the abdomen to expose the ovary on one side. Mature oocytes (stage V) were removed and defolliculated with 2 mg ml⁻¹ collagenase (Sigma, Type I) in Ca²⁺-free ND 96 (96 mmol l⁻¹ NaCl, 2 mmol l⁻¹ KCl, 1 mmol l⁻¹ MgCl₂, 5 mmol l⁻¹ Hepes-NaOH, pH 7.5) and injected with 50 nl of in-vitro-transcribed mRNA at a concentration of 200–400 ng ml⁻¹ using a mechanical micro-injector (Drummond Nanoject). Recordings were made after 2 or 3 days at 17–20 °C. The incubation solution was 50 % L15 medium (Sigma) supplemented with 15 μg ml⁻¹ gentamycin.

Expressed currents were recorded using two-electrode voltage-clamp. Cushion electrodes were prepared following the method of Schreibmayer et al. (1994) with polyacrylamide gel (PAG) replacing agarose. The tips of both the current and voltage electrodes were filled by suction using a pre-filtered (0.2 μm) solution of 3.5 % acrylamide and N,N-bisacrylamide at a ratio of 29:1, 0.15 % ammonium persulphate and 0.013 % N,N,N′,N′-tetramethylethylenediamine (TEMED) in 2 mol l⁻¹ KCl. After PAG polymerization, the electrodes were back-filled with 2 mol l⁻¹ KCl containing 1 mmol l⁻¹ EGTA and 5 mmol l⁻¹ Hepes (hemisodium salt). Before recording, electrodes were stored overnight in the back-filling solution. The recording bath solution was ND 96. Oocytes were selected for low levels of expression of endogenous Cl⁻ current since Cl⁻ current blockers were not used.

Current data were acquired using a TL-1 DMA interface and PClamp 6.0 software (Axon Instruments). The procedures used to calculate electrical and kinetic properties of expressed currents were as previously described (Jegla et al. 1995).

Fig. 1 shows an alignment of the S4 regions of jShak1, jShak2 and Shaker B together with the mutations that were made to jShak1. This alignment shows that an Arg-X-X motif is lacking in jShak1 compared with jShak2 and Shaker B. The missing charge can be added either by mutating the sequence of jShak1 to include an arginine residue or by insertion of an Arg-X-X motif. Perozo et al. (1994) have shown that K374 in Shaker B is critical for proper folding, processing and transport of the channel protein. Our alignment indicates that K294 of jShak1 is homologous with K374 of Shaker B. Therefore, we chose to make one set of insertions (RIF, QIF and RIS ) on the N-terminal side of K294, and another set of insertions (IFR and IFQ) on the C-terminal side of K294. RIF, IFR and RIS mutations insert a charged motif and increase the length of the segment, while QIF and IFQ produce the same change in length without adding charge. In addition, the RIS mutant places a hydrophilic residue adjacent to K294. Since the insertion of motifs also increased the total length of the S4 segment, we also made a mutant which added charge without altering total length by mutating the N-terminal sequence of S4 (R for S282 and V for M283). A double substitution was chosen, where valine replaced methionine, in order to keep the relative hydrophobicity of this motif similar to that of the other charged motifs in S4.

All of the mutants, except IFQ, expressed K⁺ currents in Xenopus laevis oocytes (Fig. 2). Table 2 provides quantitative data on the electrical and kinetic properties of the expressed currents.

The wild-type jShak1 expressed a rapidly inactivating (τ=4.7 ms) K⁺ current with a mid-point of activation of +33 mV and an activation slope of 12 mV per e-fold change in conductance (Figs 2A, 3; Table 2). The mid-point of steady-state inactivation (V₅₀) was −28 mV with a slope of 9 mV/e (Fig. 4; Table 2). It is important to note that the values of the mid-points of activation and inactivation reported in this study differ from those reported previously (Jegla et al. 1995) when 4,4′-di-isothiocyanatostilbene-2,2′ disulfonic acid (DIDS) was used to block Cl⁻ currents. We have recently discovered that this substance specifically interferes with the voltage and kinetic properties of jShak1 and jShak2 by unknown mechanisms; however, it does not alter the properties of Shaker B. To avoid possible interactions between these blockers and channel function, we did not use Cl⁻ channel blockers in this study (see Materials and methods).

When S282 was replaced by R and M283 by V, the resulting mutant produced a current with fast inactivation kinetics (τ=4.4 ms), closely resembling the wild type (Fig. 2B; Table 2). The activation properties (V₅₀ +40 mV, slope 12 mV/e) were also similar to those of the wild-type jShak1 (P=0.119, Fig. 3; Table 2), while the steady-state inactivation curve was significantly (P=0.021) shifted in a positive direction (V₅₀ −14 mV). The slope of the inactivation curve (6 mV/e) was unchanged from the wild type (Fig. 4; Table 2). Thus, adding a seventh charge by rewriting at the N-terminal end of S4 (SM to RV) does not have pronounced effects on the activation properties of the channel, which contrasts with the other mutants where charge was added by insertion adjacent to K294.

When RIF was inserted between F293 and K294, this mutant expressed a rapidly inactivating (τ=6.8 ms) K⁺ current that had similar kinetic properties to the wild type (Fig. 2C). Adding this seventh positively charged motif to the S4 region of jShak1 produced a large positive shift (V₅₀ +69 mV) in the activation curve relative to the wild type (P<0.001, Fig. 3) without altering the slope of the activation curve (Table 2). This mutant also showed a significant (P<0.001) positive shift in the mid-point (V₅₀ +40 mV) of the steady-state inactivation curve without a significant increase in the slope of 5 mV/e (Fig. 4; Table 2). Thus, the threshold for activation is shifted in a positive direction without a large change in the activation slope.

The QIF inserted mutant expressed a rapidly inactivating K⁺ current (Fig. 2D), with a rate of inactivation (τ=15.7 ms) that was considerably slower than that of the wild type. The activation curve was shifted in a negative direction relative to the wild type (V₅₀ +12 mV) with no significant difference (P=0.579) in the activation slope of 11 mV/e (Fig. 3; Table 2). The slope (9 mV/e) of the inactivation curve was unchanged relative to the wild type, and the mid-point of steady-state inactivation was not significantly (P=0.394) different from that of the wild type (Fig. 4; Table 2). Thus, increasing the length of the S4 segment by insertion cannot account for the voltage shifts that were seen when RIF was inserted.

The RIS inserted mutant expressed a rapidly inactivating K⁺ current (Fig. 2E), with a rate of inactivation (τ=8.6 ms) that was slower than that of the wild type. This mutant produced the strongest positive shifts in both the activation curve (V₅₀ +92 mV; Fig. 3; Table 2) and steady-state inactivation curve (V₅₀ +58 mV; Fig. 4; Table 2), which were both significantly different (P<0.001) from values for the wild type. The slopes of activation and inactivation (9 mV/e and 5 mV/e respectively) were not significantly different (P=0.288 and P=0.298 respectively) from those of the wild type (Table 2). It appears that the lipophilic residue, phenylalanine, which was inserted with RIF, was also not responsible for the strong positive shifts in both activation and inactivation.

When IFR was inserted between K294 and L295, this mutant expressed a current that lacked fast inactivation (Fig. 2F). Using longer test pulses, we were able to detect (data not shown) slow inactivation similar to that seen in an N-terminal deleted mutant, jShak1T [Δ 1–23] (Jegla et al. 1995), which probably corresponds to a C-terminal inactivation process. It should be noted, however, that at very positive test potentials a small, fast inactivating component could be seen that indicates that a fast inactivating mechanism is still present. The activation curve was shifted significantly (P=0.005) in a negative direction relative to the wild type, with a mid-point of +15 mV, while the activation slope (11 mV/e) was not significantly changed (Fig. 3; Table 2). However, the mid-point of steady-state inactivation for this slow inactivation (−9 mV) was significantly (P<0.001) shifted to the right with a similar slope of 8 mV/e compared with the wild type (Fig. 4). This is the only mutant that showed shifts in the curves for activation and steady-state inactivation that were in opposite directions.

Injected transcripts of the mutant IFQ (inserted between K294 and L295) failed to express any current. Unfortunately, we could not distinguish between a non-functional channel and lack of expression of the protein since no probes that could detect this protein were available.

The appearance of multicellularity during evolution required that metazoan cells either modify the intracellular control mechanisms used by protists or evolve novel mechanisms for intercellular communication. As new physiological niches were opening up with the early radiation of the Metazoa, so were selective pressures increasing on voltage-gated (V_g) channels. That V_g K⁺ channels must have diversified rapidly is indicated by the range of properties found in present examples of K⁺ currents in the Cnidaria, a structurally and physiologically ‘primitive’ phylum (Anderson, 1989; Dunlap et al. 1987; Meech and Mackie, 1993a,b; Przysiezniak and Spencer, 1994).

Voltage-clamp recordings from excitable cells in a wide range of taxa have shown that the complement of K⁺ channel types present is largely responsible for conferring the physiologically relevant excitability properties (Connor and Stevens, 1971; Rudy, 1988; Hille, 1992). Such properties include stabilization of the resting potential (inward rectifiers), regulation of action potential duration and expression of plateaux (delayed rectifiers), and regulation of firing frequency (A-like transient currents).

Data are scarce for the molecular structure of V_g K⁺ channels in extant examples of the earliest metazoans, and we have only one example of protozoan K⁺ channels. These protozoan channels from the ciliate Paramecium tetraurelia have very different deduced amino acid sequences from those of metazoans (Jegla and Salkoff, 1995; Jegla et. al. 1995; Chandy and Gutman, 1995), suggesting a different ancestry for ciliate channels or at least early diversification. Homologues of Shal and Shaw, as well as the Shaker homologues jShak1 and jShak2 described here, have been isolated from the cnidarian Polyorchis penicillatus (Jegla et al. 1995; T. Jegla, personal communication), indicating that the Shaker gene superfamily must have diversified early in metazoan evolution.

Despite being members of the same gene subfamily, jShak1 and jShak2 are structurally quite different in the S4 region. Like all other Shaker genes, jShak2 has seven positively charged Arg/Lys-X-X motifs in S4, while jShak1 has only six charged motifs. Despite this difference in charge, jShak1 and jShak2 have similar voltage sensitivities and voltage ranges (V₅₀ of +33 mV and +37 mV, respectively), which are some 27–65 mV more positive than that of most other Shaker channels (Timpe et al. 1988; Pfaffinger et al. 1991; Chandy and Gutman, 1995). Although reductions of total positive charge in S4 can be correlated with decreases in gating valence (Logothetis et al. 1992, 1993), there is evidence that basic residues contribute unequally to gating valence and the threshold for activation (Tytgat et al. 1993; Papazian et al. 1991; Liman et al. 1991; Logothetis et al. 1992; Aggarwal and MacKinnon, 1996; Seoh et al. 1996). Because we had natural variants of S4 with differing charged motif content, we chose to investigate the effect of insertion or replacement of charged motifs. jShak2, with seven charged motifs, became a natural control for mutants of jShak1 in which motifs were added.

When charge was added (RV for SM) to the N-terminal side of the S4 segment of jShak1, without altering the length of this segment, we did not observe any noticeable change in activation properties relative to the wild type. This suggests that the added charge created by the RV for SM mutation is outside the membrane voltage field and cannot participate in voltage sensing, which is consistent with experiments showing that the N terminus of S4 is on the extracellular side of the membrane (Manuzzu et al. 1996; Larsson et al. 1996). However, when R292 (the first arginine residue in S4) was neutralized in RCK1, giving a similar charge distribution to jShak1, a positive shift in the activation curve was seen when all four subunits were mutated (Tytgat et al. 1993).

All our remaining mutations, where motifs were inserted, caused elongation of the S4 segment. The increased length of S4 could be accommodated by displacement of the ends of the S4 segment either towards the cytoplasmic or extracellular surfaces or some combination of these movements. If insertion of RIF on the N-terminal side of K294 simply leads to this end of the S4 segment being displaced out of the membrane, then we can expect that the N-terminal residue of S4, namely R285, will not be a major contributor to voltage sensing. The RIF mutant would then have similar properties to the RV for SM mutant. This was not the case. The RIF insert produced large positive shifts in the activation and inactivation curves without substantially changing the slope of the activation curve. However, elongation alone could not have produced these depolarizing shifts in the activation and inactivation curves of the channel since the QIF mutation produced a negative shift in the activation curve while the inactivation curve of the channel was not significantly shifted. We conclude that the increase in length of the RIF insert does not interfere with the lateral electrostatic interactions between S4 and other transmembrane segments but may add mechanical constraints on S4 due to changes in length in a space that is limited by the location of the ends of S3 and S5. Therefore, a larger change in voltage is required for activation of the RIF mutant channel. However, the QIF mutation increases the length and eliminates a lateral salt bridge, thereby decreasing the activation threshold. We cannot rule out the possibility that the differences in charge act synergistically with other factors, such as hydrogen bond formation between S4 and other transmembrane elements.

The only jShak1 mutant to show uncoupling of the voltage sensitivity of activation and inactivation was the IFR insert on the C-terminal side of K294. Interestingly, this mutant also lacked fast inactivation, resembling a previously reported mutant, jShak1T, in which the N-terminal 23 amino acid residues were truncated (Jegla et al. 1995). It is likely that insertion of an additional motif on the C-terminal side of K294 causes an increase in the length of the S4–S5 loop which interferes with the interaction between the N-terminal inactivation peptide and the putative receptor on this loop (Isacoff et al. 1991). Therefore, the apparent uncoupling seen in IFR may be due to the different voltage-dependencies of steady-state fast inactivation and an unmasked slow C-terminal inactivation.

The marked differences seen when inserting the same set of residues (RIF or IFR) either before or after K294 can be explained by assuming that this lysine residue forms a salt bridge with acidic residues in adjacent segments and plays a critical role in protein folding, as demonstrated by Papazian et al. (1995) in Shaker B. We assume that K294 and R297 in jShak1 are functionally equivalent to K374 and R377 in Shaker B (Papazian et al. 1991, 1995). Thus, interactions between K294 and other charged residues in jShak1 appear to constrain movements of S4 by forming a fixed point.

Larsson et al. (1996) proposed that only one positively charged residue, R365, is buried in the membrane in the closed state and that, upon activation, four positively charged residues (R368–R377), which were previously intracellular, come to lie in the membrane. The results from two of our mutations support this model. First, addition of a charge to the N terminus of S4 by rewriting did not alter either valence or threshold. Second, when serine substituted for phenylalanine (RIS mutant) and charge is once again added on the N-terminal side of K294, there is an enormous shift (mean +72 mV) in both the activation and inactivation curves relative to the wild type. From this shift, we assume that the inserted serine residue must lie outside the membrane in the closed state and must move into a hydrophobic environment during activation. Since serine is far more polar than phenylanine, it will resist movement of the C-terminal end of the S4 segment into the membrane during activation.

An alternative model for voltage sensing has been proposed (Sigworth, 1994; Aggarwal and MacKinnon, 1996) in which the S4 segment is completely cytoplasmic in the closed state. This model cannot be reconciled with the structure of the two jellyfish Shaker channels since the loop connecting the S3 and S4 segments in both jShak1 and jShak2 is too short to span the membrane in the closed state. On the basis of the structural studies of Larsson et al. (1996) and those of Mannuzzu et al. (1996) and on the structural constraints implied by the sequences of jShak1 and jShak2, it seems likely that the S4 segment spans the plasma membrane.

Papazian et al. (1995) have suggested that K374 and R377 participate in the formation of networks with acidic residues in segments S2 and S3 that stabilise the three-dimensional structure of the channel protein. It is interesting that jShak1 has a pair of acidic residues (E237 and D260) in S2 and S3 in positions that align with those found in Shaker B (E293 and D316). However, a third acidic residue at E283 in Shaker B is replaced by asparagine in jShak1 (N227) and jShak2 (N225). It is possible that these asparagine residues are responsible for the difference between the activation thresholds of jellyfish Shaker channels when compared with other Shaker channels since neutralization of E283 in Shaker B shifted the activation curve by +78 mV (Papazian et al. 1995).

Although there has been considerable specialization of voltage-gated K⁺ channels, many of the underlying mechanisms for sensing a voltage field have been conserved. This comparative approach allows us to separate specializations from the essential, shared functional properties.

We are very grateful to the staff of Bamfield Marine Station for their assistance in collecting animals. We thank Diane Papazian for her valuable comments. This study was supported by Natural Sciences and Engineering Research Council (Canada) Research Grants to A.N.S. and to W.J.G. J.D.S is the recipient of an Alberta Heritage Foundation for Medical Research Graduate Scholarship.