Variable K+ channel subunit dysfunction in inherited mutations of KCNA1

Authors


  • Author's present address
    A. Spauschus: Department of Neurology, University of Bonn, Sigmund-Freud-Strasse 25, 53105 Bonn, Germany.

Abstract

Mutations of KCNA1, which codes for the K+ channel subunit hKv1.1, are associated with the human autosomal dominant disease episodic ataxia type 1 (EA1). Five recently described mutations are associated with a broad range of phenotypes: neuromyotonia alone or with seizures, EA1 with seizures, or very drug-resistant EA1. Here we investigated the consequences of each mutation for channel assembly, trafficking, gating and permeation. We related data obtained from co-expression of mutant and wild-type hKv1.1 to the results of expressing mutant-wild-type fusion proteins, and combined electrophysiological recordings in Xenopus oocytes with a pharmacological discrimination of the contribution of mutant and wild-type subunits to channels expressed at the membrane. We also applied confocal laser scanning microscopy to measure the level of expression of either wild-type or mutant subunits tagged with green fluorescent protein (GFP). R417stop truncates most of the C-terminus and is associated with severe drug-resistant EA1. Electrophysiological and pharmacological measurements indicated that the mutation impairs both tetramerisation of R417stop with wild-type subunits, and membrane targeting of heterotetramers. This conclusion was supported by confocal laser scanning imaging of enhanced GFP (EGFP)-tagged hKv1.1 subunits. Co-expression of R417stop with wild-type hKv1.2 subunits yielded similar results to co-expression with wild-type hKv1.1. Mutations associated with typical EA1 (V404I) or with neuromyotonia alone (P244H) significantly affected neither tetramerisation nor trafficking, and only altered channel kinetics. Two other mutations associated with a severe phenotype (T226R, A242P) yielded an intermediate result. The phenotypic variability of KCNA1 mutations is reflected in a wide range of disorders of channel assembly, trafficking and kinetics.

Episodic ataxia type 1 (EA1) is a rare human neurological disorder characterised by intermittent ataxia and continuous neuromyotonia (Gancher & Nutt, 1986; Brunt & van Weerden, 1990). It is linked to mutations in the KCNA1 gene, which is located on chromosome 12p13 and encodes the Shaker-type K+ channel subunit hKv1.1 (Litt et al. 1994; Browne et al. 1994; Comu et al. 1996; Scheffer et al. 1998). Recent studies indicate a broader phenotypic spectrum than previously suspected: two mutations are associated with epilepsy, while another is associated with neuromyotonia alone (Zuberi et al. 1999; Eunson et al. 2000). The occurrence of epilepsy underlines the importance of hKv1.1 (hereafter referred to as Kv1.1) in regulating neuronal excitability in the mammalian CNS (Wang et al. 1994; Smart et al. 1998). Phenotypic variability is also evident in the response of different kindreds to medication (Lubbers et al. 1995): while attacks are well controlled by carbamazepine or acetazolamide in some families, kindreds carrying other mutations show remarkably drug-resistant symptoms (Eunson et al. 2000).

Previous studies have documented alterations in Kv1.1 function associated with EA1 mutations (Adelman et al. 1995; D'Adamo et al. 1998; Zerr et al. 1998b; Bretschneider et al. 1999; Boland et al. 1999). These include variable decreases in current amplitude, and changes in kinetic parameters. Although these studies shed light on Kv1.1 dysfunction in EA1, they did not address the wider spectrum of clinical phenotypes. A first step towards relating the emerging phenotypic variability to channel dysfunction was recently taken by examining K+ currents in heterologous expression systems. These studies showed that mutations associated with a relatively severe phenotype including seizures (T226R, A242P) or drug resistance (R417stop) tend to give profound reductions in K+ currents when compared to wild-type (Spauschus et al. 1999; Zuberi et al. 1999; Eunson et al. 2000). At the other end of the spectrum, a mutation associated with neuromyotonia alone with no ataxia (P244H) did not alter current amplitude, and only subtly affected the voltage threshold and time course of activation (Eunson et al. 2000). Furthermore, a mutation associated with more typical EA1 (V404I, also identified by Scheffer et al. 1998) yielded an intermediate pattern: the current amplitude was unaffected, but the voltage threshold for activation was significantly increased (Eunson et al. 2000).

Evidence that the degree of perturbation of K+ channel function contributes extensively to explain the clinical phenotype raises important questions both for the disease and for Kv1.1 physiology. Here, we addressed the mechanisms by which these five mutations interfere with channel function. First, we examined the consequences of fixing the stoichiometry of channels by constructing fusion proteins (concatemers) consisting of a wild-type and a mutant subunit (Isacoff et al. 1990), in order to probe the properties of channels known to contain both species. Second, we used a pharmacological method to estimate the relative contribution of mutant and wild-type subunits to channel currents in co-expression experiments (Zerr et al. 1998b). Next, we examined the behaviour of single channels. By relating the results of these three approaches we attempted to provide an internally consistent quantitative account of Kv1.1 dysfunction for each mutation under the following headings: channel assembly, trafficking, gating kinetics and open single-channel conductance. We then provided an independent test of our quantitative approach by fluorescence imaging of Kv1.1 subunits fused to enhanced green fluorescent protein (EGFP), to visualise the cellular distribution of wild-type and mutant protein. Finally, we examined the effect of co-expressing an EA1 mutant subunit with the Kv1.2 subtype, to test whether the results can be generalised to other subunit combinations likely to occur in vivo (Coleman et al. 1999).

The three mutations associated with the most severe clinical phenotypes significantly reduced maximal K+ currents, and we argue that channel assembly and subsequent trafficking are impaired to differing degrees. These mutations, and those associated with milder phenotypes, also altered macroscopic gating kinetics. However, we found no evidence for a decrease in either the maximal open probability or the conductance of individual channels with a fixed stoichiometry of mutant and wild-type subunits. Thus, different mutations alter Kv1.1 function through distinct mechanisms, indicating that the phenotypic divergence seen in this spectrum of disorders is evident at the level of individual channels.

METHODS

Molecular biology

Human KCNA1 cDNA was obtained as described previously (Spauschus et al. 1999). The following mutations were generated with a PCR-based approach using proof-reading Pfu DNA polymerase (Stratagene): T226R, A242P, P244H, V404I, R417stop (Fig. 1A). In addition, the Y379V mutation was introduced into each cDNA clone to reduce the sensitivity of the protein product to externally applied tetraethylammonium (TEA) (MacKinnon & Yellen, 1990). We denote these constructs ‘TEA-tagged’ or with the superscript T. All clones were completely sequenced on both strands using an automated sequencer.

Figure 1.

Localisation of EA1 mutations and sample traces of two-electrode voltage-clamp measurements

A, schematic representation of a hKv1.1 subunit, showing the positions of the five disease-associated mutations and the TEA tag mutation discussed in this paper (filled circles). Thick outlines mark the positions of the other EA1 mutations identified to date. Each circle represents a single amino acid residue. The dashed rectangle shows the putative transmembrane region of the channel. B–D, sample traces of currents elicited by different voltage protocols in oocytes expressing wt subunits alone, T226R alone, wt and T226R together, or wt*T226R concatemers. B, currents elicited by depolarising voltage steps from −100 mV to +40, 0 or −40 mV. C, time course of deactivation of currents when stepped from +40 mV to 0, −20 or −40 mV. D, currents elicited by 5 s depolarising voltage steps from −100 to 0 mV.

Concatemers consisting of wild-type (wt) hKv1.1 with each of the TEA-tagged subunits (denoted wt*mutantT) were generated by removing the stop codon from wt cDNA, and adding a linker of 10 glutamine residues to the 3′ end of the coding region. This extended wt cDNA was completely sequenced on both strands and then subcloned in frame into the 5′ end of each TEA-tagged subunit. The resulting constructs were sequenced across the linker region to confirm the preservation of the entire coding sequence on both sides of the glutamine stretch. In order to exclude any effect of linking together two subunits, wt*wtT concatemers were also generated.

cDNAs were subcloned into the oocyte expression vector pSGEM. The vector-cDNA construct was linearised and transcribed in vitro using T7 RNA polymerase (Boehringer), and the resulting cRNA was purified without phenol using spin columns (Clontech). Denaturing gel electrophoresis, ethidium bromide stains and spectrophotometry were employed to assess the integrity and concentration of the cRNA transcripts.

Kv1.2 cDNA, contained in the vector pβut2pA, was a gift of Professor O. Dolly. cRNA was obtained as above, using T3 RNA polymerase.

To create fluorescent protein fusion constructs, cDNA encoding EGFP was removed from the pEGFP vector (Clontech) by introduction of Age I restriction sites at both ends of the EGFP coding region using standard PCR. A unique Age I site was also introduced immediately 5′ to the first ATG start codon of the Kv1.1 coding region, contained in the pSGEM vector, and the construct was then linearised by Age I digestion. EGFP was subcloned into the Kv1.1 construct in a standard ligation reaction, and correct insertion was determined by restriction digests, gel electrophoresis and sequencing both ends of the insert.

Oocyte preparation

Female Xenopus laevis frogs were anaesthetised using a bath containing 0.5 % tricaine and killed by cervical dislocation, decapitation and pithing. Oocytes were isolated and stored at 4–18 °C in fresh ND96 medium (mm: NaCl 96, KCl 2, MgCl2 1, CaCl2 1.8 and Hepes 5, with 100 U ml−1 penicillin and 100 μg ml−1 streptomycin, pH 7.4; sterilised by filtration). The follicular cell layer was removed from the oocytes using 2 mg ml−1 collagenase (Type IA, Sigma) in a calcium-free ND96 solution (as above; CaCl2 0 mm). Stage V and VI oocytes were injected using a Nanoject automatic injector (Drummond), and incubated in ND96 medium at 18 °C for 3 days prior to recording. The amount of cRNA injected was 27.6 nl (20 pg nl−1) unless otherwise stated. We defined this number of cRNA copies as ‘2 units’, which was used to represent two copies of the wt Kv1.1 allele. In experiments examining heteromeric channels, 1 unit of wt cRNA was co-injected with 1 unit of mutant cRNA (i.e. 27.6 nl total volume of a mixture of the two cRNAs at 20 pg nl−1 each), representing one wt and one mutant allele. Alternatively, 27.6 nl (20 pg nl−1) of concatemer cRNA was injected: note that the concentration determined by spectrophotometry indicates the number of base pairs present, i.e. injection of a constant cRNA volume and concentration ensured that the same number of subunit cRNA copies was injected in concatemer and monomer experiments.

For experiments using Kv1.2, a similar cRNA injection strategy was employed. When Kv1.2 and R417stop were co-expressed, 1 unit of Kv1.2 was injected in each condition plus an additional 1, 4 or 8 units of R417stop, to examine any dominant-negative effect of R417 stop.

Electrophysiology

Oocytes were placed in normal frog Ringer solution (NFR, mm: NaCl 115, KCl 2.5, CaCl2 1.8 and Hepes 10, pH 7.4) at ∼21 °C. Two-electrode voltage-clamp recordings were made using electrodes pulled to a tip resistance of 0.5–2 MΩ. A GeneClamp 500 amplifier and pCLAMP 7 software (Axon Instruments) were used for data acquisition and analysis, and a −P/4 protocol applied in order to subtract leak and capacitance currents from the recordings. Voltage-step protocols were employed to study the kinetics of macroscopic currents. For measurements of peak current amplitude and TEA sensitivity, currents were recorded at +40 mV. We normalised the results by the wt amplitude (2 units of cRNA) recorded on the same day within the same batch of oocytes. Normalised amplitudes were then averaged between at least two batches. Since TEA tagging per se had no effect on maximal current amplitudes recorded in the control solution, results obtained by expressing the tagged version and the untagged version of each subunit were merged together. Significance was assessed using Student's unpaired t test. A Bonferroni correction (Sokal & Rohlf, 1995) was applied when multiple amplitude measurements were compared to predicted values.

TEA dose-response

TEA chloride (Sigma) in NFR was applied to the oocyte by a gravity-driven perfusion system. Current recordings were made in each TEA solution when the peak amplitude in response to repeated steps to +40 mV reached a steady value. Different TEA concentrations (0.1–10 mm) were applied in a pseudo-random order. The peak current amplitude was normalised to the peak amplitude in 0 mm TEA for the same oocyte. The normalised current amplitude I was then averaged across oocytes prior to plotting dose-response curves. wtT-injected oocytes were also tested with 100 mm TEA to define the dissociation constant for subunits lacking a tyrosine residue at position 379. When fits were applied to TEA dose-response data during optimisation of parameters, the agreement of the data points with the curve was tested statistically using a Bonferroni correction for multiple comparisons.

Patch-clamp recordings

The vitelline membrane was removed manually from oocytes after placing them in a hyperosmolar solution (mm: NaCl 105, KCl 5, CaCl2 2 and Hepes 5, pH 7.4; osmolarity adjusted to 470 mosmol l−1 with sucrose). Pipettes were pulled from borosilicate glass (GC150F, Clark) to a resistance of 5–10 MΩ, coated with Sylgard (Dow Corning), heat polished and lled with (mm): NaCl 120, KCl 2 and Hepes 5, pH 7.4. The ‘cytoplasmic’ solution consisted of (mm): KCl 120, EGTA 1 and Hepes 5, pH 7.4. Recordings were performed in the inside-out configuration using an Axopatch 200B amplifier and pCLAMP 7 data acquisition and analysis software (Axon Instruments). Voltage pulses of 500 ms duration and to various potentials between −80 and +40 mV were applied from a holding potential of −80 mV. Currents were low-pass filtered at 1 kHz and digitised at 5 kHz. All-point histograms (usually with 128 bins) of the 1 s current response to a given voltage pulse were generated and fitted with the sum of two Gaussians. The slope conductance was calculated from the amplitude difference between the two Gaussian peaks using an estimated reversal potential of −100 mV. The open probability was measured by relating the area under the Gaussian peak for the open state to the total area under the curve.

Confocal microscopy

Oocytes were prepared for fluorescence imaging by fixation in 4 % paraformaldehyde (v/v in 0.15 m PBS) for at least 3 h at 4 °C, and were then cut in half perpendicular to the equator using a razor blade. Each half of the oocyte was submerged in ND96 medium and imaged on the cytoplasmic side, followed by the membrane side, i.e. obtaining four images per oocyte.

Fluorescence images were acquired at × 10 magnification using a Bio-Rad Radiance2000 laser scanning confocal microscope with an 8-bit ADC, using Lasersharp software for acquisition and Scion Image software (NIH) for analysis. EGFP was excited using a 488 nm argon laser beam, with an emission filter of 515 nm/30 nm to collect the peak emission (507 nm) and eliminate autofluorescence of endogenous oocyte protein. Laser and confocal aperture settings were kept constant for all experiments. During scanning a Kalman function was applied, whereby each optical slice was scanned three times in succession and an average pixel value obtained to reduce the random noise at each pixel. Images were acquired at increasing depths through the oocyte at a spacing of 10 μm, and then projected onto a single image for analysis. This method maximised the sample volume and limited variability, since oocytes are large and fluorescence was not uniformly distributed.

Each image was measured for the mean pixel value, which was obtained within a region defined manually by drawing a line around the perimeter of the cell. This value was divided by the region area, and then averaged with the measurement obtained from the second half of the oocyte, to give a mean fluorescence value (F). Membrane and cytoplasmic mean fluorescence values for each cell were then averaged between oocytes. As a measure of total cell fluorescence, the membrane and cytoplasmic data were first averaged for each cell, and then among all cells. All mean fluorescence values were then normalised to that obtained for cells injected with 1 unit of EGFP-Kv1.1 cRNA (FEGFP-Kv1.1).

RESULTS

All five mutations affected conserved residues (Fig. 1A). T226R, associated with a relatively severe phenotype, is one of three different missense mutations affecting the same residue and associated with EA1 (Comu et al. 1996; Scheffer et al. 1998). The mutation introduces a positively charged residue in the second transmembrane domain (S2). A242P and P244H are predicted to lie close to the junction between S2 and the S2–3 intracellular linker. The involvement of a proline residue in both of these substitutions may be significant, since proline has a cyclic molecular structure and is often associated with bends of folded protein chains (Stryer, 1995). A242P, associated with a more severe phenotype including seizures (Eunson et al. 2000), could therefore introduce a conformational change at the cytoplasmic face of S2. The V404I mutation was reported independently in an apparently unrelated family, also with an apparently unremarkable EA1 phenotype (Scheffer et al. 1998). This mutation is predicted to reside in S6, which forms part of the cytoplasmic pore of the channel and is, thus, a potentially critical region for ion permeation (Doyle et al. 1998). The amino acid substitution itself is, however, relatively conservative. Finally, R417stop is the first premature truncation yet demonstrated, and is associated with a severe drug-resistant phenotype (Eunson et al. 2000).

In order to establish the effects of these mutations on channel function, we determined the properties of macroscopic K+ currents in oocytes injected with wild-type (wt) or mutant cRNA alone, and in oocytes injected with both wt and mutant cRNA in a 1:1 molar ratio. We also examined oocytes injected with cRNA coding for fusion proteins consisting of a wt and a mutant subunit (wt*mutant concatemers). The number of cRNA copies injected in each case corresponded to the same number of subunits (see Methods). Figure 1B–D shows sample current traces for each of these conditions for T226R. Because all EA1 patients described to date are heterozygous for their respective mutation, we concentrated on the co-injection data to give an indication of the likely consequences of the mutations for K+ channel function in the disease. Data obtained from the expression of homomeric wt or mutant channels, and of concatemers, were then used to model the behaviour of co-expressed mutant and wt subunits.

Current amplitudes

Figure 2 shows a comparison of the results obtained by expressing wt*mutant concatemers with data from previous experiments on homomeric and co-injected EA1 subunits (Spauschus et al. 1999; Zuberi et al. 1999; Eunson et al. 2000). In each case, the current amplitude measured at +40 mV was normalised to the amplitude obtained from oocytes injected on the same day with the same total number of cRNA copies encoding wt subunits. wt*wt concatemers yielded current amplitudes indistinguishable from wt (1.21 ± 0.18 compared to wt, mean ±s.e.m., n= 13; Fig. 2A). The concatenated subunits wt*T226R, wt*A242P and wt*R417stop gave mean peak current amplitudes that were significantly reduced to 0.60 ± 0.04 (n= 23), 0.47 ± 0.03 (n= 6) and 0.45 ± 0.06 (n= 24) of wt, respectively (Fig. 2A; P < 2 × 10−4 in each case). In contrast, wt*P244H and wt*V404I yielded current amplitudes that were not different from wt (n= 5 and 4, respectively).

Figure 2.

Comparison of whole-cell current properties of fusion proteins (concatemers) with data from studies examining mutant subunits alone or co-expressed with wt subunits

Data for homomeric channels and co-injection experiments are from Spauschus et al. (1999) and Eunson et al. (2000). A, histogram of peak currents (Imax) in response to +40 mV voltage steps from a holding potential of −100 mV. B, threshold of activation as determined by fitting a Boltzmann function to the normalised initial tail current amplitudes, measured at the end of each test pulse when stepping back to −50 mV (V0.5, voltage at half-maximal activation). C, activation time (10–90 % rise time) determined at the beginning of a voltage step to 0 mV. D, characterisation of current deactivation. Following a depolarising pulse to +40 mV, the potential was stepped back to −40 mV. Resulting tail currents were fitted with monoexponentials, and the characteristic time constants plotted. E, degree of C-type inactivation during long depolarising voltage pulses to 0 mV. The amplitude of the current at the end of the 5 s pulse was normalised to the initial peak current amplitude. Note the logarithmic ordinate scale in C and D. All oocytes were injected with cRNAs as indicated. The horizontal grey bands indicate the parameter values (±1 s.e.m.) obtained for wt and wt*wt concatemers. All error bars indicate s.e.m. (n ranged from 5 to 29).

Macroscopic current kinetics

Examination of current activation for the concatemers gave the following results. wt*T226R (n= 7), wt*V404I (n= 6) and wt*R417stop (n= 15) significantly shifted the activation threshold towards more positive potentials, when compared to wt (Fig. 2B). However, no significant change in activation threshold was seen for either wt*A242P (n= 10) or wt*P244H (n= 9). When the membrane potential was stepped from −100 to 0 mV, the activation of macroscopic currents was significantly slowed for wt*A226R and wt*V404I (Fig. 2C). When examined following steps to the voltage corresponding to half-maximal activation (V0.5), wt*R417stop also showed a pronounced slowing of activation (data not shown). Deactivation, measured by stepping the membrane potential from +40 to −40 mV, was significantly slowed for wt*T226R, wt*A242P and wt*V404I (Fig. 2D). When this parameter was examined following steps from +40 mV to V0.5, the same pattern emerged, although the slowing of deactivation for wt*A242P was less marked (data not shown). The current amplitude at the end of a 5 s voltage pulse to 0 mV from −100 mV was normalised to the initial peak current amplitude to serve as a marker for changes in C-type inactivation. Only R417stop showed a significantly smaller fraction of current at the end of the test pulse even in the presence of co-expressed or fusion-linked wt subunits (Fig. 2E; P < 10−6 in both cases).

Channel assembly

Co-expression of mutant and wt subunits does not necessarily give rise to the same channel stoichiometry as expression of concatemers. In the case of concatemer expression, every channel is likely to be composed of two wt and two mutant subunits (Isacoff et al. 1990; Heginbotham & MacKinnon, 1992; Kavanaugh et al. 1992; Tu & Deutsch, 1999; although see McCormack et al. 1992). In contrast, in the co-injection experiments, channels of five different subunit stoichiometries can exist. If mutant subunits are impaired in their ability to co-assemble and contribute to a tetramer, or if channels containing one or more mutant subunits are less able to mediate the K+ current, the macroscopic kinetic properties could be biased towards those of channels containing several wt subunits. Alternatively, fusion of a wt subunit to a mutant subunit could enhance membrane expression of mutant subunits, by artificially providing the assembly, targeting and/or anchoring mechanisms that are otherwise defective.

In order to probe the relative importance of these two phenomena, and to determine whether wt and mutant subunits co-assemble at all, we tagged individual subunits by mutating a residue that determines their sensitivity to externally applied TEA (MacKinnon & Yellen, 1990). We introduced the Y379V mutation, which greatly increases the apparent equilibrium dissociation coefficient (Ki) for TEA (Kavanaugh et al. 1992). By measuring the TEA sensitivity of currents obtained by co-expressing wt and TEA-tagged mutant (mutantT) subunits, we attempted to establish the relative contribution of each species to the channels reaching the membrane (Zerr et al. 1998b).

We investigated the interaction between wt and mutant subunits by simultaneously fitting two sets of measurements. The first data set consisted of the relative maximal current amplitudes recorded at +40 mV in control solution when wt, mutant or wt*mutantT cRNA was expressed, as well as those obtained by co-expressing wt and mutantT cRNA. The second data set consisted of the TEA dose-response curves obtained for the following conditions: wt alone, wtT alone, wt and mutantT together, and wt*mutantT alone. (Additional data were provided by co-expressing TEA-tagged wt with untagged mutant subunits.)

Theory

Two classes of mechanism can potentially affect the current amplitude and TEA sensitivity when wt and mutant subunits are co-expressed.

First, there may be a defect in the efficacy with which mutant subunits assemble with wt subunits. An extreme (and unlikely) possibility is that mutant and wt subunits completely fail to interact, and form two separate populations of homomultimeric channels. More plausibly, they co-assemble to form five possible channel stoichiometries, where m, the number of mutant subunits in the tetramer, varies between 0 and 4. The relative proportion rm of channels containing m mutant subunits is given by a binomial distribution:

display math(1)

where Cm4 is the binomial coefficient for m= 0, … 4, and f is the probability that any given position in a tetramer will be occupied by a wt subunit. Equivalently f can be defined as:

display math(2)

where q is the relative probability that a given subunit is mutant and not wt. This parameter does not distinguish between the number of subunits available and the efficacy of subunit incorporation into a channel. If the same number of wt and mutant cRNA copies is injected, and the mutation does not interfere with channel assembly, q= 1 and f= 0.5. A decrease in q shifts the relative proportions of channels composed of different stoichiometries. Inserting the corresponding value of f in eqn (1) thus describes the defect in events leading to channel assembly. Note that a reduction in q implies a decrease in the total number of channels to (1 +q)/2 of control.

Second, there may be a defect downstream of channel assembly, which could alter one or more of the following processes: trafficking (which we define as post-tetramerisation processing, degradation, export from the endoplasmic reticulum, membrane targeting and anchoring, and eventual re-internalisation), gating kinetics, open channel probability and single-channel conductance. These defects cannot be represented by the parameter f, which relates only to the distribution of channel stoichiometries produced. Thus, the overall effect of post-assembly processes must be represented by another set of parameters im (m= 0, 1, … 4), which varies with the number of mutant subunits m in a given channel stoichiometry. Defining i0= 1, to indicate that homomeric wt channels are normal, a reduction in im (m= 1, … 4) implies a defect in one or more of these processes for channels containing m mutant subunits.

The values of im (m= 1, … 4), together with q, determine the amplitude of the maximal current obtained by injecting mutant cRNA alone (Imutant), or together with wt cRNA (Ico-inj), relative to the current obtained by expressing the same number of copies of wt cRNA alone (Iwt). The ratio Imutant/Iwt is given by:

display math(3)

The amplitude ratio for co-injection experiments depends on whether co-assembly takes place. If it does not, and two populations of homomeric channels are formed,

display math(4)

However, if mutant and wt channels do co-assemble, the ratio is given by:

display math(5)

When concatemers are expressed, the current should be carried by a uniform population of tetramers containing two wt and two mutant subunits. Injecting wt*wtT cRNA gave current amplitudes that were no different from those obtained by injecting twice the number of wt cRNA copies (coding for the same total number of subunits). This implies that, for wt*mutantT concatemers, only the processes subsequent to tetramerisation are available to affect the maximal current amplitude (Iconcat). Thus:

display math(6)

The TEA dose-response curve for homomeric wt or wtT channels (obtained by injecting only one cRNA species) is described by:

display math(7)

where Ki,m is the apparent equilibrium dissociation constant for the binding of TEA to a tetrameric channel containing m TEA-tagged subunits. Fitting this function to the data obtained by expressing either wt or wtT cRNA yields estimates for Ki,0 and Ki,4, respectively. Because each TEA-tagged subunit contributing to a channel causes a constant logarithmic increment in Ki,m (Liman et al. 1992; Kavanaugh et al. 1992; Heginbotham & MacKinnon, 1992), the values of Ki,m for m= 1, 2 or 3 are given by:

display math(8)

Substituting Ki,2 into eqn (7) yields the dose-response curve predicted for a homogeneous population of channels made up of wt*mutantT concatemers. This curve lies mid-way between the curves for homomeric wt and wtT channels.

In the case where wt and wtT subunits are co-injected in a 1:1 ratio, channels with different stoichiometries are formed according to a binomial distribution. The population TEA dose-response curve is therefore given by (Zerr et al. 1998b):

display math(9)

where f= 0.5.

In the case where wt and mutantT subunits are co-expressed, there are two possible outcomes. First, wt and mutantT subunits might fail to co-assemble. If so, two separate homomeric channel populations will be formed. This situation is described by:

display math(10)

Second, if wt and mutantT subunits do co-assemble, the TEA sensitivity of the resulting mixture of stoichiometries is given by:

display math(11)

where f and im (m= 1, … 4) are allowed to vary to reflect impaired subunit incorporation into a channel and reduced efficacy of downstream stages, respectively. Both parameters are also constrained by the amplitude data as shown by eqns (3)–(6) above.

By incorporating the new parameter im, we are now able to provide a consistent description both of the amplitude data and of the TEA dose-response data, obtained when mutant subunits are present in three different populations of channels: homomeric mutant, wt*mutantT concatemers, and wt + mutantT co-expression.

TEA sensitivity and current amplitudes

We carried out the following control experiments. First, we verified that the Y379V mutation per se had no effect on K+ current amplitude, voltage threshold for activation, or any of the time-dependent parameters (data not shown). Second, we verified that channels formed from TEA-tagged mutant subunits expressed alone had a TEA sensitivity that was indistinguishable from that of TEA-tagged wt channels (data not shown; this could not be checked for R417stop because the mutant on its own is non-functional). Next, we measured the TEA sensitivity of channels obtained by expressing concatemers formed from a wt subunit fused to a TEA-tagged wt subunit (wt*wtT). Ki,0 and Ki,4, estimated by fitting eqn (7) to the TEA dose-response relationship for wt and TEA-tagged wt channels, respectively (Levenburg-Marquart algorithm – Microcal Origin), were approximately 0.4 and 250 mm (Fig. 3A). These values yield an estimate for Ki,2 of ∼10 mm (eqn (8)). The predicted dose-response curve (obtained by substituting this value into eqn (7)) fitted the observed data for the wt*wtT concatemers (Fig. 3A), implying that the channels do indeed contain two tagged and two untagged subunits (cf. McCormack et al. 1992).

Figure 3.

TEA sensitivity reveals the relative contribution of wt and TEA-tagged subunits to whole-cell currents

A, TEA sensitivity of homomeric wt and wtT channel currents (see insets). Data points for wt, wtT and wt*wtT (n= 23, 28 and 13, respectively) are fitted by eqn (7), with apparent equilibrium dissociation constants Ki,0= 0.4 mm, Ki,4= 250 mm and Ki,2= 10 mm, respectively. Dose-response data obtained by co-expressing wt + wtT (n= 12) are fitted by eqn (9) (where f= 0.5), which describes a binomial distribution of the five possible channel stoichiometries, with equal probability of incorporation of wt and wtT subunits. Equation (10), which describes the dose-response curve expected if wt and wtT subunits formed two separate populations of homomeric channels, fails to fit the data. B-F, dose-response data obtained by co-expressing each TEA-tagged mutant (mutantT) with wt subunits. For wt + T226RT (n= 17; B), A242PT (n= 7; C) and R417stopT (n= 14; F), the points are shifted to the left of the curve given by eqn (9) (with f= 0.5), implying that mutantT subunits contribute less to the whole-cell current than wt subunits. In all cases, eqn (10) (representing two separate populations of homomeric channels), fails to fit the co-expression data, implying that mutant subunits co-assemble with wt. Error bars indicate s.e.m.

Since co-expression should produce a mixture of five channel stoichiometries, described by a binomial distribution, we used the values of Ki,m (m= 0, … 4) given by eqn (8) to predict the dose-response behaviour of co-expressed wt and wtT subunits. Equation (9) describes the TEA sensitivity of a binomial mixture of channel stoichiometries. We simulated the TEA dose-response by plotting I/Imax determined using eqn (9) against [TEA] (this approach was also used for subsequent equations). We set the probability that a given position in a tetramer was occupied by a wt subunit (f) equal to 0.5, implying that the TEA mutation has no effect on the probability of contributing to a channel, consistent with the normal current amplitude observed. Figure 3A shows close agreement between the data points for wt + wtT and the curve determined using eqn (9), consistent with the hypothesis that the two subunits co-assemble. The quantitative approach used to explore the consequences of subunit mutations is potentially sensitive to errors in the estimate of the wt cRNA concentration. We therefore asked whether the wt + wtT co-expression data could still be fitted if the wt cRNA concentration had been overestimated to different degrees, by altering the value of f in eqn (9) according to a ratio describing the error in the relative concentration of the two cRNA types. The TEA sensitivity was consistent with a 0–20 % overestimate of wt cRNA concentration (not shown). All subsequent analyses therefore took into account a potential error in cRNA concentration within this range. In contrast to eqn (9), the curve generated by eqn (10), which describes a 1:1 mixture of homomeric wt and homomeric wtT channels, failed to fit the data points. i4 was set to be equal to 1, because the maximum current amplitudes of wt and wtT expressed alone were indistinguishable (data not shown). The failure of eqn (10) to fit the data further confirms that co-assembly of the TEA-tagged and untagged subunits occurs. We used this approach to test the hypothesis that mutant subunits (carrying the TEA tag) are able to assemble with wt subunits.

Figure 3BF shows the results of co-expressing wt and TEA-tagged mutant subunits. T226RT, R417stopT and A242PT each gave dose-response curves that were shifted to the left, relative to that obtained for co-injected wt and wtT. That is, the TEA sensitivity was more similar to that of un-tagged wt channels than of TEA-tagged channels, implying that the mutant subunits contributed relatively less to the K+ flux than wt subunits. The positions of these curves could not be accounted for by a 20 % over-estimation of either wt or mutant cRNA concentration (not shown).

We first tested the hypothesis that co-assembly does not take place, by plotting curves generated by eqn (10), with i4 determined by the current amplitudes measured when mutant channels were expressed alone (eqn (3), with q= 1). The position of the curve was shifted to the left for T226R, A242P and R417stop because the homomeric current amplitude (and consequently i4) was very small in each case (Fig. 2A). The curves generated by eqn (10) deviate from the results of co-expressing wt and mutantT, and can be rejected at P < 0.001. This argues against the hypothesis that co-assembly of wt and mutantT fails to take place, and therefore implies that heteromultimeric channels are formed.

Figure 4 illustrates the effect of optimising either f or im (m= 1, … 4), or both sets of parameters, to fit the data obtained with the R417stop mutation. We also included the TEA sensitivity of currents obtained by co-expressing TEA-tagged wt together with the untagged R417stop mutant. This situation is a mirror image of the co-expression experiments where wt cRNA was injected together with R417stopT, and was fitted by substituting f′ for f, where f′= 1 −f. Optimisation was achieved by iteratively varying the values of f and im. The parameter estimates were ‘acceptable’ if predicted current amplitudes and TEA sensitivities were not significantly different (taking into account the Bonferroni correction for multiple comparisons – see Methods) from the values of these measurements observed experimentally.

Figure 4.

R417stop affects both channel assembly and tetramer function

Aa, Ba, Ca, TEA sensitivity obtained by co-expressing either wt and R417stopT subunits (n= 14), or wtT and R417stop subunits (n= 12). Ab, Bb, Cb, maximal current amplitudes for wt and R417stopT. TEA dose-response data obtained by expressing wt*R417stopT concatemers (n= 14) are also shown in Aa, but are omitted from Ba and Ca for clarity. The curves used to fit the dose-response of homomeric wt, wtT, wt*wtT and wt + wtT currents (see Fig. 3) are also shown for reference. A, the TEA sensitivity can be fitted by increasing f in eqn (11) to 0.77, representing impaired assembly of R417stopT subunits. However, because the function of assembled tetramers is assumed to be unaffected by the presence of mutant subunits (i0=i1=…i4= 1), the current amplitude data are not fitted. B, the current carried by each tetramer is assumed to vary according to its stoichiometry (im optimised), but tetramerisation is assumed to be unaffected by the mutation (f= 0.5). Although this model agrees with the amplitude data, the TEA sensitivity cannot be fitted. C, the TEA sensitivity and amplitudes are fitted simultaneously by allowing both the efficacy of assembly and subsequent stages of expression to vary (f and im both optimised). All error bars indicate s.e.m.

Initially, we asked whether an increase in f (i.e. a decrease in q, reflecting a decreased efficacy of subunit incorporation) could, on its own, explain the TEA dose-response data, as previously reported by Zerr et al. (1998b). (We assigned priority to the TEA sensitivity only for heuristic purposes.) Figure 4A shows that increasing f from 0.5 to 0.77 allowed a good fit to the TEA data, replotted from Fig. 3F. The points were reasonably fitted by eqn (9). This would imply a profound reduction in the ability of mutant subunits to contribute to heteromeric channels, with the result that channels containing two, three or four mutant subunits are under-represented in the overall population, with a compensatory increase in the proportion of channels containing zero or one mutant subunits. However, this model fails to take into account subsequent variation in the current amplitude carried by channels containing mutant subunits. It is equivalent to assuming that i0=i1=…i4= 1 (see key in Fig. 4A), and is clearly incompatible with the current amplitudes seen for experiments where R417stop was expressed alone. An increase in f to 0.77 implies a reduction in mutant subunit availability q to 0.3 of wt (eqn (2)). The predicted ratio IR417stop/Iwt given by eqn (3) is therefore 0.3, while the observed amplitude ratio was 0.02 ± 0.01 (Fig. 2A). Moreover, an increase in f to 0.77, without any additional decrease in im (m= 1, … 4), predicts a reduction in current amplitude to 0.84 for the co-expression experiments (eqn (5)), while the observed amplitude was 0.33 ± 0.04 of wt. These observed amplitudes were inconsistent with any value of f (range 0.5–1) when i0=i1=… i4= 1 (P < 0.01, with Bonferroni correction). Moreover, because any potential error in cRNA concentration would be reflected by a change in f, we conclude that the data obtained for R417stop cannot be explained by such an error.

Thus, it is not possible to account for the data entirely by postulating that the mutation only affects subunit incorporation. We therefore explored the effect of reducing im (m= 1, … 4), guided by the amplitudes of the currents obtained by expressing mutant subunits alone, or wt and mutant subunits together. Figure 4B shows the effect of this manipulation, while f was fixed at 0.5 (and consequently q= 1) to see whether it was possible to account for all the data simply on the basis of changes downstream of channel assembly. We used the current amplitudes obtained with wt*R417stop concatemers expressed alone and R417stop monomers expressed alone to provide estimates for i2 and i4, respectively, and then varied i1 and i3 to obtain agreement with the amplitude and TEA data together. Taking into account a 98 % confidence interval for the amplitude measurements, i2 and i4 fell in the ranges: 0.28 < i2 < 0.62 and 0 < i4 < 0.05. Once the values of i1, …i4 had been optimised, we verified that the current amplitude for wt + R417stop co-expression also agreed with the amplitude predicted by eqn (5) with f fixed at 0.5, i.e. 0.21 < (Ico-inj/Iwt) < 0.45. (The confidence intervals were estimated by applying a Bonferroni correction for three measurements.) According to these constraints provided by the amplitude data, the optimised values of i1, …i4 were 1, 1, 0.28, 0, 0. However, the TEA dose-response data were poorly fitted by eqn (11) (with f fixed at 0.5) using these values of i1, …i4. This fit could be rejected at P < 0.01, partly because of a deviation of the observed effect of 10 mm TEA from the predicted effect. It was not improved appreciably by allowing i1 and/or i3 to vary within the interval 0–1, or by assuming a 20 % error in cRNA concentration (data not shown).

Thus, the data obtained from amplitude and TEA sensitivity measurements with R417stop were inconsistent with a change either in f alone or in im alone. Figure 4C shows that both data sets could, however, be fitted by optimising both f and im simultaneously. Setting f= 0.65, i= 0.60, i2= 0.33, i3= 0.10 and i4= 0.02 improved the fit (P > 0.01). Although there is room for minor adjustment of the parameters (at the risk of over-fitting), f∼ 0.65 implies q∼ 0.5. That is, the R417stop mutation reduces the probability of subunit incorporation by about half. Although the parameter f does not distinguish between the availability of subunits following translation, and the efficacy of their assembly into tetramers, it is clearly defined as a parameter separate from post-assembly events. The estimates for im (m= 1, … 4) are consistent with a simple scheme whereby each additional mutant subunit present in the tetramer reduces the current that it can eventually carry. Dysfunction of the R417stop subunit is thus apparent both pre- and post-assembly.

As a further control, we studied the TEA sensitivity of currents obtained by expressing wt*R417stopT concatemers. In this case, the data could be fitted by eqn (7), with Ki,2= 10 mm, as for wt*wtT concatemers. This implies that the R417stop mutation per se has no direct effect on the altered TEA sensitivity due to the Y379V mutation.

The analysis shown in Fig. 4 was applied to the other mutants, and the results are summarised in Table 1. V404I and P244H showed no reduction in current when expressed alone or when co-expressed, and the TEA sensitivity of the co-expression experiments could be fitted without altering either f or im. This implies that neither channel assembly nor downstream events that determine maximal K+ flux are affected. In contrast, in order to account for the amplitude data and TEA sensitivity for R417stop, f had to be increased (equivalently, q had to be decreased), and im also had to be reduced relative to wt. This implies that complex impairments of K+ channel function, affecting both channel assembly and downstream events, result from this mutation. In the case of T226R, im had to be reduced relative to wt in order to fit the amplitude and TEA sensitivity data, although increasing f further improved the fit. Finally, for A242P, the data were again best fitted by increasing f and decreasing im. However, the fits obtained by increasing f without altering im, or by reducing im without altering f, could not be rejected.

Table 1. Summary of the models that best fit the data obtained for each mutant, using optimised values of f and i0,…i4
 Observed amplitudePredicted amplitude f q i 0 i 1 i 2 i 3 i 4
  1. Observed and predicted current amplitudes are normalised to wt.

wt110.501.0011111
wt + wtT1.07 ± 0.161
wt*wtT1.21 ± 0.181
T226R0.03 ± 0.020.030.600.6710.80.550.30.03
wt + T226RT0.48 ± 0.070.44
wt*T226RT0.60 ± 0.040.55
A242P0.10 ± 0.020.100.600.6710.90.60.60.1
wt + A242PT0.81 ± 0.040.74
wt*A242PT0.47 ± 0.030.60
P244H1.07 ± 0.1310.501.0011111
wt + P244HT1.00 ± 0.061
wt*P244HT1.01 ± 0.031
V404I1.13 ± 0.1610.501.0011111
wt + V404IT1.14 ± 0.071
wt*V404IT1.09 ± 0.091
R417stop0.02 ± 0.010.050.650.5410.600.330.100.02
wt + R417stopT0.33 ± 0.040.40
wt*R417stopT0.45 ± 0.060.33

Patch-clamp recordings

A reduction in im (m= 1, … 4) could result from impairment of one or more of the following processes: trafficking, maximal opening probability and K+ permeation through the open channel. In an attempt to distinguish between these stages, we measured the properties of single channels in membrane patches excised from oocytes expressing either wt subunits alone (corresponding to a channel stoichiometry of m= 0) or wt*mutant concatemers (stoichiometry, m= 2; Fig. 5A).

Figure 5.

T226R, V404I and R417stop affect channel open probability at 0 mV

A, examples of current responses to 1 s voltage steps from −80 to 0 mV. Single channels were composed of mutant and wt subunits in a fixed 1:1 ratio (concatemers). Dotted lines indicate the closed state of the channel. B, histogram of the slope conductance measured at 0 mV. C, histogram of the channel open probability at 0 mV. **P < 0.05 (tested against wt).

The slope conductance was not significantly different in any of the mutants (Fig. 5B). When the measurements were made at +40 mV, there were no significant differences in open probability between wt*wt (0.69 ± 0.04, data from 3 patches), wt*T226R (0.67 ± 0.07, 15 patches), wt*A242P (0.65 ± 0.03, 11 patches), wt*P244H (0.62 ± 0.07, 5 patches), wt*V404I (0.63 ± 0.05, 9 patches) and wt*R417stop (0.66 ± 0.04, 9 patches). Since this is the membrane potential at which current amplitudes and TEA sensitivities were estimated, we conclude that the reductions in the parameter i2 estimated above for T226R, A242P and R417stop do not result from changes in the behaviour of single channels expressed at the membrane. By a process of elimination, we conclude that the reduction of i2 estimated by fitting maximal current amplitudes and TEA sensitivity for T226R and R417stop results from impaired trafficking. It should be noted that when the open probability was estimated at 0 mV, there was a significant decrease for three constructs: wt*T226R, wt*V404I and wt*R417stop (but not for wt*A242P or wt*P244H; Fig. 5C). While membrane patches expressing wt*wt channels had a mean open probability of 0.67 ± 0.07 (3 patches), wt*T226R, wt*V404I and wt*R417stop gave mean open probabilities of 0.53 ± 0.04 (15 patches), 0.49 ± 0.06 (9 patches) and 0.37 ± 0.05 (9 patches), respectively. These results are consistent with the positive shifts in activation threshold seen for the same three mutants (Fig. 2B).

Fluorescence imaging

The macroscopic current amplitudes and TEA dose- response data argue that dysfunctional assembly and/or subunit availability are unable to explain fully the reduction in maximal current amplitude when either R417stop or T226R is co-expressed with wt. Moreover, the single-channel data indicate that altered kinetics are also unable to explain the observed reductions in maximal current amplitude for heteromers. Thus, the data point to dysfunction of one or more post-assembly stage(s) of trafficking of mutant-containing channels to the membrane. Alternatively, some mutant-containing channels may reach the membrane but fail to contribute to K+ flux. We applied an alternative experimental approach both to test the validity of this conclusion for R417stop and to determine whether non-functional channels do indeed reach the membrane. We tagged either wt or R417stop subunits with EGFP (EGFP-Kv1.1 and EGFP-R417stop) in order to obtain independent evidence on protein expression levels and localisation. Currents recorded using two-electrode voltage clamp indicated that EGFP-Kv1.1 subunits assembled and formed functional channels with properties similar to wt (see traces in Fig. 6Aa; cf. Fig. 1). Injection of EGFP-R417stop resulted in negligible macroscopic currents (Fig. 6Ab), with a mean peak amplitude at +40 mV of 0.8 ± 0.1 % compared to the EGFP-Kv1.1 amplitude. Thus the EGFP marker did not appear to alter the relationship between wt and R417stop expression of current (cf. Fig. 2A). Oocytes were halved and imaged on the cytoplasmic and membrane faces of each half using confocal laser scanning microscopy. The mean fluorescence signal (F) was obtained from the mean pixel value within the appropriate region of the cell in each projection, as shown by the dotted lines in the example images in Fig. 6Ba and b. The mean value obtained for EGFP-Kv1.1 (FEGFP-Kv1.1) in corresponding images was then used to normalise the fluorescence signal in all conditions (F/FEGFP-Kv1.1). Injection of 1 unit of EGFP-Kv1.1 cRNA produced a strong fluorescence signal in both the cytoplasm and membrane, which was clearly distinguishable from the autofluorescence of the water-injected control oocytes (see Methods; Fig. 6Ba–d).

Figure 6.

EGFP tagging of wt and R417stop to examine protein expression

Aa, example traces of EGFP-Kv1.1 currents elicited by depolarising steps between −60 and +40 mV, from a holding potential of −100 mV. Ab, example trace of an EGFP-R417stop current elicited by stepping from −100 to +40 mV. B, example images of the membrane (m) or cytoplasmic (c) side of halved oocytes injected with EGFP-Kv1.1 (a and b), EGFP-R417stop (c and d) or water (e and f), obtained using fluorescence microscopy. Dotted lines indicate the region of fluorescence measurement. Scale bar, 250 μm. C, bar charts showing the mean fluorescence pixel value (F) obtained from oocytes imaged on the membrane side (a), cytoplasmic side (b), or averaged across the whole cell (c), normalised to the mean fluorescence signal obtained for EGFP-Kv1.1 (FEGFP-Kv1.1). Oocytes were injected with 1 unit of EGFP-Kv1.1 cRNA (n= 10), EGFP-R417stop cRNA (n= 4), EGFP-Kv1.1 cRNA together with 1 unit of non-fluorescent R417stop cRNA (EGFP-Kv1.1 + R417stop; n= 5), EGFP-R417stop cRNA together with non-fluorescent wt cRNA (EGFP-R417stop + wt; n= 5), or H2O (n= 3). Asterisks denote a significant difference of P < 0.05 between the conditions indicated by the dotted lines (Student's t test). All error bars indicate s.e.m.

Figure 6Ca shows that EGFP-R417stop fluorescence in the membrane of oocytes was not significantly different to that of water-injected controls (0.29 ± 0.07 and 0.23 ± 0.03; P > 0.05). This result indicates that there is negligible expression of functional homomeric R417stop channels in the membrane, consistent with the electrophysiological data. In addition, this provides evidence against the hypothesis that a substantial number of non-functional R417stop homomers are maintained in the membrane. However, when the cytoplasm of the same cells was compared (Fig. 6Cb), EGFP-R417stop produced a significantly higher fluorescence than water (0.37 ± 0.05 and 0.21 ± 0.03, respectively; P < 0.05). Taken together with the dominant effect of R417stop on wt current amplitudes, kinetics and TEA dose-response, this result provides additional evidence that the R417stop mutant protein is translated.

Next, we compared the intensity and localisation of EGFP-Kv1.1 subunit fluorescence when expressed alone, or co-expressed with non-fluorescent R417stop subunits. Figure 6Ca shows that the fluorescence signal in the membrane of oocytes co-injected with 1 unit of EGFP-Kv1.1 and 1 unit of R417stop (EGFP-Kv1.1 + R417stop) was significantly reduced when normalised by the membrane fluorescence of 1 unit of EGFP-Kv1.1 expressed alone (0.46 ± 0.05; P < 0.05). This effect was specific to the membrane; as shown in Fig. 6Cb, co-expression of EGFP-Kv1.1 + R417stop did not show a reduction in cytoplasmic fluorescence when normalised to EGFP-Kv1.1 alone (0.82 ± 0.38; P > 0.05). These results are consistent with the conclusions of the electrophysiological data and modelling, and furthermore argue that R417stop interferes with the ability of heteromers to reach the plasma membrane. When total cell fluorescence was considered (i.e. cytoplasmic and membrane images averaged), the fluorescence signal produced by EGFP-Kv1.1 + R417stop was lower than for EGFP-Kv1.1 alone, although this did not reach significance (0.57 ± 0.15; P > 0.05; Fig. 6Cc). These results do not rule out the possibility that co-expression of R417stop reduces the total amount of EGFP-Kv1.1 protein present in the cell, possibly by enhancing the degradation of heteromers not reaching the membrane.

To examine the effect of wt subunits on the expression pattern of R417stop subunits, we co-expressed EGFP-R417stop with unmodified wt Kv1.1 (EGFP-R417stop + wt). Figure 6Ca shows that co-expression of wt subunits enhanced the expression of EGFP-R417stop subunits in the membrane, although this failed to reach significance (EGFP-R417stop + wt, 0.39 ± 0.06; EGFP-R417stop alone, 0.28 ± 0.07; P > 0.05). There was no significant difference when EGFP-R417stop was measured in the cytosol (0.48 ± 0.07 and 0.37 ± 0.05; Fig. 6Cb) or when membrane and cytosol levels were averaged together (0.40 ± 0.04 and 0.29 ± 0.05; Fig. 6Cc; all values normalised to the corresponding EGFP-Kv1.1 fluorescence; P > 0.05). Nevertheless, the level of fluorescence in oocytes injected with EGFP-R417stop + wt, averaged across the whole cell, was significantly higher than for water-injected controls (0.40 ± 0.04 and 0.20 ± 0.02, respectively; see Fig. 6Cc). These results suggest that wt co-expression may partially ‘rescue’ the R417stop mutant from degradation. However, the experiments were not designed to test this hypothesis directly, and so this conclusion is only tentative.

Overall, the results of the fluorescence measurements are consistent with the conclusion of the quantitative electrophysiological and pharmacological measurements that trafficking of R417stop-containing heteromers to the membrane is impaired, and lend independent validation of this approach. That is, the reduction in im (m= 1, … 4) appears to result at least in part from an impairment in the ability of heteromers containing R417stop subunits to reach and/or persist in the membrane. They also argue against the presence of a substantial population of non-functional heteromers in the membrane and suggest that heteromers may have an elevated rate of degradation.

Co-expression with Kv1.2

Having identified likely mechanisms by which the R417stop mutation alters the assembly, trafficking and function of Kv1.1 channels, we investigated whether similar effects might occur when R417stop was co-expressed with the Kv1.2 subunit, with which Kv1.1 is co-localised in peripheral nerve and in the cerebellum. Figure 7A shows the effect on Kv1.2 current amplitude of co-expressing R417stop. When 1 unit of Kv1.2 was co-expressed with 1 unit of R417stop, the peak current amplitude at +40 mV was significantly lower than the amplitude produced on injection of a single unit of Kv1.2 alone (2.32 ± 0.36 and 3.85 ± 0.68 μA, respectively; P < 0.05), indicating a dominant-negative effect of the R417stop subunit. Increasing the ratio of R417stop to Kv1.2 enhanced this reduction (1.20 ± 0.24 μA for 1:4 ratio and 0.47 ± 0.04 μA for 1:8 ratio; Fig. 7B). These results provide evidence that the R417stop subunit co-assembles with Kv1.2 and exerts a similar dominant-negative effect to that observed when it was co-expressed with the Kv1.1 wt subunit.

Figure 7.

Co-expression of R417stop with wild-type Kv1.2 subunits

A, example traces of currents elicited by depolarising voltage steps between −60 and +40 mV, from a holding potential of −100 mV, in a cell injected with Kv1.2 alone (a) or with Kv1.2 plus R417stop in a 1:1 unit ratio (b). B, mean peak current amplitudes in cells expressing Kv1.2 alone, or Kv1.2 co-expressed with R417stop in a 1:1, 1:4 or 1:8 ratio. The same number of Kv1.2 cRNA copies was injected in each condition, except where ‘2 units’ is indicated, where the number of cRNA copies was doubled to show that current saturation levels were not reached. C, TEA dose-response data of currents produced by expression of Kv1.2 alone (n= 4), Kv1.2 with R417stop in a 1:1 ratio (n= 5), or Kv1.2 with Kv1.1 in a 1:1 ratio (n= 5). Data points show the mean dose-response in each condition. Curves are taken from Fig. 4 and are included for comparison of Kv1.2 data with the results of Kv1.1 modelling. All error bars indicate s.e.m.

To further examine the effect of R417stop on Kv1.2 function, we made use of the fact that wt Kv1.2 (referred to as Kv1.2 for clarity) has a valine at position 381, corresponding to Y379 in Kv1.1, giving it a natural reduction in TEA affinity compared to wt Kv1.1 (i.e. similar to wtT). We therefore used the same TEA dose-response protocol as described above to examine co-assembly and the contribution of R417stop to the function of heteromers. Figure 7C shows the data obtained when TEA was applied to currents produced by expressing Kv1.2 alone, with wt Kv1.1, and with the R417stop mutant. Curves showing the position of the Kv1.1 wt, wtT, wt + wtT and wtT+ R417stop dose-responses (Figs 3 and Figs 4) are included for comparison. There was a good agreement between the TEA dose-response of Kv1.2 and the curve fitted to wtT, i.e. where m= 4, as would be expected from the presence of the valine at the position corresponding to Y379. Similarly, co-expression of Kv1.2 with wt Kv1.1 revealed a TEA dose-response indistinguishable from Kv1.1 wt + wtT. These observations confirm that Kv1.2 behaves in a similar way to wtT. When Kv1.2 was co-expressed with R417stop, the dose-response fell in a similar position to that observed for Kv1.1 wtT+ R417stop, implying that R417stop co-assembles with Kv1.2 and exhibits a reduced contribution to current amplitude compared to wt Kv1.1. This effect was similar to, if not more pronounced than, that with Kv1.1. To fully quantify this result, it would be necessary to construct concatemers and follow a similar approach to that described above.

DISCUSSION

The five EA1 mutations investigated in this study are associated with different clinical phenotypes and caused very distinct alterations in hKv1.1 function. Three mutations associated with a relatively severe phenotype, drug-resistant EA1 (R417stop), neuromyotonia with seizures (A242P) or EA1 with seizures and infantile contractures (T226R), caused a profound reduction in maximal current amplitude, as well as changes in voltage-dependent kinetics. Mutations associated with a milder EA1 phenotype (V404I) or neuromyotonia alone (P244H), did not affect current amplitudes at +40 mV, and only altered kinetic parameters. Previous studies addressing other EA1 mutations have also revealed considerable variation in the effects of different mutations (Adelman et al. 1995; D'Adamo et al. 1998; Zerr et al. 1998a,b; Bretschneider et al. 1999; Boland et al. 1999). However, it is not possible to know from these studies whether the alterations in channel function correlated with the severity of the clinical phenotypes.

The results of the present study further argue that at least two of the mutations (T226R and R417stop) affect events downstream of channel assembly, and suggest that channel assembly itself may also be dysfunctional. The approach applied here to separate pre- and post-assembly stages of hKv1.1 processing is novel. We simultaneously determined the TEA sensitivity and the amplitude of currents obtained by co-expression of wt and mutantT subunits, and also made use of wt-mutant fusion proteins to provide a further constraint on the current amplitude carried by channels of known stoichiometry. These experimental methods were separately applied by Zerr et al. (1998b) and D'Adamo et al. (1998), respectively. However, Zerr et al. (1998b) only fitted the TEA dose- response curves by allowing the relative efficacy of mutant and wt subunits to vary (eqn (9)), without taking into account differences in current amplitude carried by different channel species subsequent to assembly (eqn (11)). D'Adamo et al. (1998), on the other hand, related amplitude data obtained with wt-mutant concatemers to those obtained with co-expression of the earlier described mutations, but did not further define these differences in terms of channel stoichiometry. We have shown that it is possible to gain a clearer insight into the consequences of mutations by simultaneously satisfying the constraints imposed by both data sets. Indeed, failure to do so, by fitting a curve to the TEA data that assumes that channels of different species carry equal currents, will yield an overestimate of the deficit in the efficacy of incorporation of the mutant subunits into tetramers (see Fig. 4A). Finally, we have argued that dysfunction downstream of tetramerisation cannot be accounted for by decreases in either peak open probability at +40 mV or single-channel conductance. Thus, we are forced to the conclusion that trafficking must be defective, probably because tetramers leave the endoplasmic reticulum less efficiently, or because they are re-internalised from the plasmalemma prematurely.

Although the quantitative approach applied in the present study attempts to provide a consistent account of both channel assembly and trafficking, it is important to bear in mind that it is indirect because it relies entirely on electrophysiological measurements. We therefore applied an independent test of this approach by imaging fluorescent Kv1.1 fusion proteins, and obtained results supporting the conclusion that R417stop subunits impair the trafficking of tetramers to the membrane. It is possible that biochemical methods could shed further light on the relative importance of defects in tetramerisation and post-assembly processing, although it remains to be determined whether such methods are comparably sensitive. It will also be important to test whether the conclusions derived from the present work also apply in a mammalian expression system, although this may present greater difficulties in controlling the ratio of wild-type and mutant transcripts.

R417stop

R417stop is the only truncation mutation of hKv1.1 to be described to date, and is associated with a relatively severe drug-resistant phenotype (Eunson et al. 2000). R417stop had a dominant-negative effect on wt current amplitude, but did not form functional homomeric channels on its own, indicating co-assembly with wt subunits. This was confirmed with the TEA sensitivity measurements. The present study also identifies two likely stages at which R417stop impairs channel function. First, the electrophysiological data imply an impairment in the function of heteromultimers: wt*R417stop concatemers gave smaller macroscopic currents than wt homomers, and wt cRNA co-injected with R417stop cRNA yielded an amplitude that was significantly reduced when compared to that obtained when half the amount of wt cRNA was injected alone (representing a single wt allele; Eunson et al. 2000). This impairment was not reflected in a reduced open probability or conductance of single wt*R417stop channels. We then applied a novel adaptation of the TEA tagging technique that enabled us to provide a quantitative discrimination between pre- and post-assembly events in Kv1.1 expression, which has not been achieved before using either electrophysiology or cell biological techniques alone. A reduction in efficacy of pre-assembly events (represented by the parameter f) is not sufficient to explain the data obtained with this mutant, indicating that post-assembly events are impaired. We tested the validity of the optimal quantitative model by introducing an EGFP marker to the N-terminus of wt and R417stop constructs, similar to the constructs published previously for Kv1.3 and Kv1.4 (Kupper, 1998). The results support the conclusion that membrane targeting is impaired for the R417stop subunit: EGFP-R417stop expressed alone showed a negligible fluorescence in the membrane, and the membrane fluorescence of EGFP-Kv1.1 was significantly reduced in the presence of R417stop. The latter result also implies that an impairment of membrane targeting underlies the dominant-negative effect of the R417stop subunit on the current amplitude of heteromers. Taken together, these findings strongly suggest that channels containing R417stop subunits are defective during targeting to (or anchoring at) the plasmalemma. This may be at least partly explained by the loss of a PDZ-binding sequence in the C-terminus (Kim et al. 1995), although there is also evidence to suggest that this may not be the only sequence in the Kv1.1 C-terminus that recognises anchoring proteins (Levin et al. 1996; Jing et al. 1997). Indeed, the results of this study support this possibility since the wt*R417stop construct yielded approximately 50 % of the wt current despite the linkage of the wt C-terminus to the R417stop subunit, which presumably results in a channel with no PDZ sequences free for binding (Songyang et al. 1997). (Note that the wt*wt concatemer gave a normal current amplitude, demonstrating that eliminating one of two free C-termini within the dimer did not reduce channel expression, and arguing that the reduced current obtained for the wt*R417stop concatemer can still be attributed to the C-terminal truncation.)

The best model arising from the quantitative analysis also implies that the C-terminal truncation affects subunit assembly, in addition to post-tetramerisation trafficking. Although the N-terminus of hKv1.1 plays an essential role in tetramerisation, the C-terminus has not previously been implicated. However, since C96 and C505, in the N- and C-termini, respectively, come into close proximity at a late stage of Shaker channel biogenesis (Schulteis et al. 1998), it would be interesting to examine the stability of the mature tetrameric structure when the C-terminus was truncated. The large reduction in fluorescence of EGFP-R417stop compared to EGFP-Kv1.1 in whole cells also argues that fewer R417stop subunits are expressed and/or maintained compared to wt. This may be partially due to an enhanced rate of degradation of heteromers containing R417stop subunits. Since EGFP was fused to the N-terminus of the subunit, which is known to be important in channel assembly, it is possible that this modification decreases the efficacy of tetramerisation. However, this possibility is not consistent with the high expression levels of EGFP-Kv1.1, or of GFP-Kv1.3 and GFP-Kv1.4 reported previously using a similar fusion method (Kupper, 1998), and so the simplest explanation is that the R417stop truncation rather than the EGFP protein is responsible for the impaired assembly.

The C-terminus of Kv1.1 thus appears to play a role in both tetramerisation and membrane targeting, leading to a decrease in current amplitude and a dominant-negative effect when R417stop subunits are expressed as heteromultimers with wt Kv1.1. A similar dominant-negative effect of R417stop was observed when it was co-expressed with Kv1.2 subunits, indicating that the functional alterations of R417stop affect at least some of the heteromeric combinations that occur in vivo. R417stop subunits also decrease the speed of activation and accelerate deactivation when expressed with Kv1.1 wt, possibly representing a complex defect in channel gating. Interestingly, C-type inactivation is also increased in channels containing R417stop subunits, reflecting a possible role of the C-terminus in this mode of inactivation.

T226R

For the T226R mutation, associated with a severe phenotype including seizures and infantile contractures (Zuberi et al. 1999), there was a profound decrease in current carried by heteromultimeric and homomeric mutant channels at +40 mV that could not be accounted for by a decrease in open probability or conductance, implying impaired trafficking. We confirmed that the subunit co-assembled with wt subunits, but cannot exclude the possibility of a reduced efficacy of incorporation into a tetramer relative to wt. The T226R mutation causes a positive shift in activation threshold and a slowing of activation and deactivation (Zuberi et al. 1999). These kinetic results are similar to those of two other EA1 missense mutations identified at the same residue: T226A and T226M (Comu et al. 1996; Scheffer et al. 1998; Zerr et al. 1998a). For the T226R mutation, the shift in activation threshold is only partially normalised by the presence of wt subunits in the channel (Zuberi et al. 1999). The data obtained from concatemer expression also indicate a dominant effect of the T226R mutation on several kinetic parameters. Overall, these results point to an important role of the T226 residue during both expression and function of the channel, possibly by forming electrostatic bonds with neighbouring transmembrane segments, thus stabilising the structure of the channel subunit (Tiwari-Woodruff et al. 1997).

A242P

The A242P mutation, again associated with a severe phenotype including seizures, also produced a profound reduction in current compared to wt, although without a clear dominant-negative effect. Interestingly, a previously described EA1 mutation affecting a nearby residue (R239S) has been reported to produce non-functional subunits, which are unable to interact with wt subunits (Zerr et al. 1998b). Our results for A242P do not exclude the possibility that a similar defect in A242P subunit assembly accounts for the abnormal functions observed. Indeed, this explanation could mean that the distribution of channels was skewed towards those containing relatively more wt subunits when A242P and wt subunits were co-expressed, possibly explaining the normalisation of the kinetics of whole-cell currents observed. In common with T226R, there was an impairment of current carried by heteromultimers and mutant homomers, which could not be explained by a reduced open probability or conductance. This implies that a reduction in efficacy of stages downstream of assembly is due to a defect in membrane targeting due to the A242P subunits. The abnormal presence of a proline residue may disrupt the α helical structure of S2, possibly altering the stability of the subunit.

P244H

In contrast to mutations affecting S2, P244H, which is situated in the S2–3 linker, produced only subtle alterations in channel function. There was no detectable deficit in assembly or current amplitudes for channels containing mutant subunits. Overall, the present results do not explain the fact that P244H is associated with a clinical phenotype, albeit a very mild one consisting only of neuromyotonia. Further work will determine whether alterations of channel function become more pronounced, for instance when co-expressing β subunits to restore fast inactivation to the channel (Pongs, 1999), or the human isoforms of Kv1.2 and Kv1.4, which are associated with Kv1.1 in bovine brain (Coleman et al. 1999).

V404I

Finally the V404I mutation, associated with typical EA1, also produced a relatively small effect on channel properties, with no impairment of channel assembly or of the current carried by channels containing mutant subunits. There was, however, a significant shift in the activation threshold to more depolarised potentials in co-expression experiments, which would be predicted to reduce K+ flux. The residue at position 404 is one of several valines lining the cytoplasmic pore of the channel (Doyle et al. 1998). Another EA1 mutation (V408A) affecting the next valine residue also has a relatively small effect on kinetics (Adelman et al. 1995; D'Adamo et al. 1998; Zerr et al. 1998b; Bretschneider et al. 1999).

In conclusion, the present study demonstrates a remarkable diversity of consequences of individual hKv1.1 mutations on hKv1.1 function. The molecular mechanism of disease in each case is therefore likely to be closely related to the effect of the mutation within specific functional domains of the channel. Thus far, the pattern of alterations seen in heterologous expression studies shows a correlation with the degree of severity of the disease spectrum in affected families. It will be important to determine whether this correlation holds for other mutations of hKv1.1 and associated subunits, which may emerge in association with previously unsuspected phenotypes dominated, for instance, by disorders of CNS excitability (e.g. seizures) or peripheral nerve excitability (e.g. neuromyotonia) without episodic ataxia. The quantitative approach used here to separate tetramerisation and downstream events could equally be applied to other autosomal dominant diseases caused by K+ channel mutations (Charlier et al. 1998; Singh et al. 1998; Biervert et al. 1998).

Note added in proof

Since this work was submitted for publication, L. N. Manganas, S. Akhtar, D. E. Antonucci, C. R. Campomanes, J. O. Dolly & J. S. Trimmer have applied a biochemical approach to arrive at similar conclusions regarding the consequences of the R417stop mutation (Episodic ataxia type-1 mutations in the Kv1.1 potassium channel display distinct folding and intracellular trafficking properties. Journal of Biological Chemistry (published online 25 October, 2001. DOI: 10.1074/jbc.M908890199).

Acknowledgements

We are grateful to C. Herd for technical support, and to A. Jouvenceau and N. Davies for comments on the manuscript. This work was supported by the MRC, Wellcome Trust, Brain Research Trust and Epilepsy Research Foundation.

Ancillary