Abbreviations used:β-gal, β-galactosidase; Kv, voltage-gated potassium channel; NFR, normal frog Ringer’s solution; ONPG, o-nitrophenyl β-D-galactopyranoside; THULL, selection medium without tryptophan, histidine, uracil, leucine, lysine; UTL, selection medium without uracil, tryptophan, leucine.
Address correspondence and reprint requests to Dr. M. Stocker at Max-Planck-Institut f. exp. Medizin, Hermann-Rein-Str. 3, D-37075 Göttingen, Germany.
Abstract:α-Subunits of the voltage-gated potassium channel (Kv) subfamily Kv9 show no channel activity after homomultimeric expression in heterologous expression systems. This report shows that heteromultimeric expression of rKv9.1 and rKv9.3 specifically suppresses the currents mediated by α-subunits of the Kv2 and Kv3 subfamilies but does not affect currents mediated by α-subunits of the Kv1 and Kv4 subfamilies. To understand the molecular basis of the electrical silence of Kv9 homomultimeric channels, crucial functional domains (amino and carboxy terminus, S4 segment, and pore region) were exchanged between Kv9 α-subunits and rKv1.3. Electrophysiological studies of these chimeras revealed that the pore region is involved in determining the nonconductive behavior of homomultimeric Kv9 channels. This analysis was extended by protein interaction assays, aiming to identify the region of Kv9 subunits responsible for the specific suppression of rKv2.1- and rKv3.4-mediated currents. We could show that the amino-terminal domain of Kv9 α-subunits does not support homomultimeric assembly but interacts specifically with the rKv2.1 amino-terminal region. Conversely, the specific intersubfamily assembly of rKv3.4 with rKv9.1 or rKv9.3 is governed by the hydrophobic core and not the amino-terminal domain.
The astonishing diversity of currents through voltage-gated potassium channels (Kv) arises from a variety of mechanisms, among them the diversity of cloned Kv α-subunits. Based on sequence homology, members of the Kv family have been assigned to nine distinct subfamilies (Kv1-9) (Gutman and Chandy, 1993; Hugnot et al., 1996; Salinas et al., 1997b; Stocker and Kerschensteiner, 1998). Four α-subunits are needed to form a functional K+ channel, leading to the possibility of homo- or heteromeric assembly (MacKinnon, 1991). α-Subunits belonging to the Kv1-4 or Kv7 subfamilies form functional homomultimeric potassium channels in heterologous expression systems (Pongs, 1992). Furthermore, heteromultimeric assembly was shown to occur within the Kv1 and Kv3 subfamilies (Christie et al., 1990; Isacoff et al., 1990; Ruppersberg et al., 1990; Weiser et al., 1994), and purification of potassium channels from the central nervous system demonstrated the existence of native heteromultimeric K+ channels (Shamotienko et al., 1997). Different experimental approaches have identified a region in the amino terminus of Kv channel α-subunits (T1 or NAB domain) as the structural element that determines the compatibility of α-subunit interaction (Li et al., 1992; Shen et al., 1993; Shen and Pfaffinger, 1995). With use of the yeast two-hybrid system, it was directly shown that the T1 domain permits assembly of α-subunits within one subfamily and prevents the interaction between members of different subfamilies (Xu et al., 1995).
Although the structural hallmarks of functional K+α-subunits can be found in the amino acid sequences of a growing number of α-subunits belonging to the Kv5, Kv6, Kv8/Kv2.3r, and Kv9 subfamilies (Drewe et al., 1992; Hugnot et al., 1996; Castellano et al., 1997; Patel et al., 1997; Salinas et al., 1997b; Stocker and Kerschensteiner, 1998), they are electrically silent upon homomultimeric expression. One of the aims of our study was to identify which of the known domains of potassium channels are still functional in two members of the Kv9 subfamily (rKv9.1 and rKv9.3) by generating chimeric α-subunits. Replacing the amino and carboxy termini as well as part of the voltage sensor (S4 segment) of rKv1.3 by the corresponding parts of rKv9.3 resulted in expression of functional K+ channels, thus demonstrating the functional integrity of these channel parts. Additionally, chimeric α-subunits with an exchanged pore region indicated its involvement in the nonconducting properties of homomultimeric Kv9 channels. We could show that two rat Kv9 α-subunits, rKv9.1 and rKv9.3, suppressed the currents mediated by rKv2.1 and rKv3.4, but not rKv1.3 and rKv4.2 channels, after co-injection into Xenopus oocytes. To understand the molecular basis for this specific suppression, we performed a detailed analysis of the assembly capabilities of rKv9.1 and rKv9.3, again using chimeric α-subunits in combination with electrophysiological measurements and the yeast two-hybrid system. We were able to show that the molecular domain responsible for the interaction between Kv2 and Kv9 subfamilies is located in the amino terminus and that the intersubfamily assembly between rKv3.4 and rKv9.1 or rKv9.3 is specific but not governed by the amino-terminal domain.
MATERIALS AND METHODS
cDNAs and plasmids
The cDNA clones of rKv2.1 (Frech et al., 1989), rKv5.1, and rKv6.1 (Drewe et al., 1992) were kindly provided by Dr. R. H. Joho. To generate the full-length rKv2.1, four amino acids (MPAG) were introduced. The rKv1.3-superGEM construct harboring silent endonuclease recognition sites was a gift from Dr. S. A. N. Goldstein (Marom et al., 1993). IRK1 was a gift from Dr. J. P. Ruppersberg. Two hybrid vectors carrying DNA binding domain (pLexN) and transcription activation domain (pVP16) as well as the yeast strain L40 were a gift from Dr. S. M. Hollenberg (Hollenberg et al., 1995).
Plasmid constructs for electrophysiological measurements
All DNA manipulations were carried out using standard recombinant DNA techniques (Sambrook et al., 1989). Low-copy PCR (15 cycles) was performed with low-error-rate polymerases (Vent, New England Biolabs; or Pfu, Stratagene) to generate chimeric α-subunits or yeast two-hybrid constructs. All manipulated sequences were verified by sequencing with a Dye Terminator Cycle Sequencing Kit and an ABI377 DNA sequencer (Applied Biosystems). The following chimeric α-subunits were generated (accession numbers in parentheses), where bold numbers correspond to residues of rKv9.3 (Y1707), underlined numbers correspond to residues of rKv9.1 (Y17606), and plain numbers correspond to residues of rKv1.3 (M31744): Nt-Kv9.3, 1-175/175-525; Ct-Kv9.3, 1-437/413-491; Ct-Kv1.3, 1-412/438-525; S4-Kv9.3, 1-312/289-310/335-525; ES4E-Kv9.3, 1-334/311-491;▵174-Kv9.3, 175-491; Nt + Ct-Kv9.3, 1-175/175-437/413-491; Nt + Pore-Kv9.3, 1-175/175-377/355-376/401-525; Pore + Ct-Kv9.3, 1-377/355-376/401-437/413-491; Nt + Pore + Ct-Kv9.3, 1-175/175-377/355-376/401-437/413-491; S4-ES4E-Kv9.3, 1-312/289-491; Nt-Kv9.1, 1-179/175-525; Nt-Kv1.3-Kv9.1, 1-172/178-498. To generate Pore-Kv9.3 (1-377/355-376/401-525), Pore-Kv9.3-F (W361F) and Pore-Kv9.3-DAF (ICW359-361DAF) chimeric S5-S6 regions of rKv1.3 and rKv9.3 were created, annealing phosphorylated oligonucleotides, to assemble a DNA duplex and replace the corresponding part in rKv1.3-superGEM (see Fig. 2Q).
Plasmid constructs used for yeast two-hybrid system
Fusion proteins with the DNA binding (Lex) and transcription activation domain (VP16) were constructed by inserting the corresponding sequences of amino- and carboxy-terminal domains in frame into the EcoRI or BamHI site of pLexN and pVP16-4, a modification of pVP16. Primers with engineered EcoRI and BamHI recognition sites were used to amplify the corresponding nucleotide sequences of Kv channels (α-subunit, amino acid position from-to, accession no.: rKv1.3, 4-171, M31744; rKv2.1, 3-187, X16476; rKv3.4, 33-161 and 470-625, X62841; rKv4.2, 2-177, S64320; rKv5.1, 7-193, M81783; rKv6.1, 1-234, M81784; rKv8.1, 2-216, X98564; rKv9.1, 423-497, Y17606; rKv9.3, 406-491, Y1707). In constructs where the in-frame stop codon located after the multiple cloning site in pLexN and pVP16-4 was too far downstream, an in-frame stop codon (UAA) preceded the recognition site in the antisense oligonucleotide. The amino terminus of rKv9.1 (1-193) was inserted into pLexN and pVP16-4 by standard cloning procedures. For the amino terminus of rKv9.3 (2-174), a BamHI site present in the cDNA was used for cloning.
Methods for yeast two-hybrid system
pLexN and pVP16 carrying the fusion constructs of interest were introduced into yeast (L40) by lithium acetate transformation (Gietz et al., 1992), using 1-2 μg of DNA. Plating a part of the transformation mixture on UTL plates (selection medium without uracil, tryptophan, leucine) tested for successful co-transformation, as only yeast transformed with both plasmids could survive. Plating another part of the transformation mixture on THULL plates (selection medium without tryptophan, histidine, uracil, leucine, lysine, but supplemented with 2.5-10 mM 3-amino-1,2,4,-triazole) permitted growth only to yeast where association of fusion parts occurred. A further indication of interaction was gained by qualitative activation studies of the lacZ reporter gene by a soft-agar β-galactosidase (β-gal) assay (Duttweiler, 1996). For semiquantitative studies [o-nitrophenyl β-D-galactopyranoside (ONPG) assay], single colonies of indicated double transformants were grown in liquid UTL medium, and after lysis of the yeast cells and estimation of the total protein content, the specific β-gal activity was measured with ONPG as substrate (Rose et al., 1990).
cRNA synthesis, injection, and electrophysiological characterization
The coding regions of the potassium channels used in this study were cloned into transcription vectors psGEM (a gift from Dr. M. Hollmann) or superGEM. Capped cRNAs were synthesized in vitro after linearizing the plasmids and performing the transcription with T7 RNA polymerase. Isolation of oocytes (stage V-VI) from X. laevis and cRNA injection were performed as described previously (Stühmer, 1992). Whole-cell currents were recorded 1-7 days after injection under two-electrode voltage-clamp control using a Turbo TEC-10CD amplifier (NPI-Elektronik). Intracellular electrodes had resistances of 0.4-0.8 MΩ when filled with 2 M KCl. Leak and capacitive currents were subtracted on-line using a P/n protocol. Currents were low-pass filtered at 0.7-1 kHz (-3 dB) and sampled at 3-5 kHz. The standard bath solution was normal frog Ringer’s solution (NFR) containing the following (in mM): 115 NaCl, 2.5 KCl, 1.8 CaCl2, 10 HEPES-NaOH (pH 7.2). All experiments were performed at room temperature (20-22°C). Data acquisition and analysis were performed with the Pulse + PulseFit software package (HEKA Elektronik, Lambrecht, Germany), EXCEL (Microsoft), and IGOR (Wavemetrics). Data are given as means ± SEM, with n specifying the number of independent experiments. For co-expression experiments, cRNA of rKv1.3 (0.4 ng/oocyte), rKv2.1 (0.2 ng/oocyte), rKv3.4 (0.2 ng/oocyte), and rKv4.2 (2 ng/oocyte) was mixed with rKv9.1 or rKv9.3 (0.6-6 ng/oocyte) in a ratio of 1:3. In most experiments, IRK1 (40 pg/oocyte), an inward-rectifying K+ channel, was co-expressed to serve as an internal control for the expression capability of the occyte. To measure inward currents, high-potassium Ringer’s solution was used, in which NaCl was completely replaced by KCl.
Co-expression of rKv9.1 and rKv9.3 with different Kv channel α-subunits
To test if two K+ channel α-subunits of the Kv9 subfamily, rKv9.1 and rKv9.3, could express as functional heteromultimeric channels in Xenopus oocytes, we co-injected them together with the other silent putative K+ channel α-subunits rKv5.1, rKv6.1, or rKv8.1, one by one or all together. For each combination, measurements in NFR or in external solutions in which Na+ was replaced by Rb+, Cs+, Li+, or K+ were performed to test for outward or inward currents, with different protocols taking into account a variety of possible current kinetics. None of these experimental protocols gave rise to any detectable exogenous current.
As co-expression of rKv8.1/Kv2.3r α-subunits with some Kv channels resulted in a reduction of current amplitude (Hugnot et al., 1996; Castellano et al., 1997), we next studied what influence rKv9.1 and rKv9.3 had on representative members of different Kv subfamilies (rKv1.3, rKv2.1, rKv3.4, and rKv4.2; Fig. 1). The rKv2.1 current was strongly suppressed upon co-expression with rKv9.1 (Fig. 1A) or rKv9.3 (Fig. 2B). Coinjections of smaller amounts of rKv9.1 and rKv9.3 with rKv2.1 did not completely abolish the current but resulted in a clear reduction and in new kinetic features (data not shown; D. Kerschensteiner and M. Stocker, in preparation). Co-expression of rKv9.1 with rKv3.4 resulted in currents reduced by >80% (Fig. 1C), whereas co-expression with rKv9.3 gave a reduction of ∼40% (Fig. 1D). Neither the shape nor the amplitude of the currents of rKv1.3 and rKv4.2 was affected in the presence of rKv9.1 or rKv9.3 (data not shown). The average amplitudes obtained from 10 oocytes for each combination and for the control inward currents are summarized in Fig. 1E-H. Similar results were also obtained in experiments where the co-injection of IRK1 cRNA was omitted (data not shown). Our data show that rKv9.1 and rKv9.3 induce a concentration-dependent functional suppression of α-subunits of the Kv2 or Kv3 subfamilies but not of the Kv1 or Kv4 subfamilies.
Characterization of functional domains in chimeric channels: amino and carboxy termini
To understand the molecular basis of the electrical silence of homomultimeric Kv9 channels, we created a set of chimeric molecules between rKv1.3 and rKv9.3, enabling us to characterize functional domains in the rKv9.3 α-subunit. rKv1.3 was chosen because no interaction with rKv9.3 or rKv9.1 was observed that might hamper the interpretation of the results. Schematic drawings of the generated chimeric channels and representative current traces are shown in Fig. 2A-O. Replacement of the amino terminus of rKv1.3 by that of rKv9.3 resulted in Nt-Kv9.3. This chimera showed a drastically lower expression level than rKv1.3 (Fig. 2C), with largely unaltered kinetics and voltage dependence of activation and inactivation. To test if the low expression of Nt-Kv9.3 is a feature of the specific construct or if it is caused by a reduced interaction of the tetramerization domain (T1), we generated Nt-Kv9.1 by fusing the same part of rKv1.3 with the amino terminus of rKv9.1 (Fig. 2D). Oocytes injected with Nt-Kv9.1 cRNA showed higher current amplitudes than Nt-Kv9.3 (Fig. 2D), but the expression level was still substantially lower than for rKv1.3. These results suggest that the lower expression of Nt-Kv9.3 as well as Nt-Kv9.1 might be due to a disturbed assembly via NAB/T1 domains. The residual current could be mediated by channels whose association is prompted by nonspecific interaction sites within the hydrophobic core of rKv1.3, as shown for Kv1.3 (T1-) (Tu et al., 1996). Furthermore, as no fast inactivation could be introduced by fusing the N termini of rKv9.1 and rKv9.3 with rKv1.3 (Fig. 2C and D), known to possess the receptor for an inactivating particle, we can conclude that these channels do not have an inactivating domain at their N terminus. We next replaced the C terminus of rKv1.3 with the one of rKv9.3 to generate Ct-Kv9.3 (Fig. 2E). The expression of this chimera resulted in currents indistinguishable from those mediated by rKv1.3 in shape and amplitude (Fig. 2E). Also, the fusion of the N- and C-terminal parts of rKv9.3 with the “core” of rKv1.3, named Nt + Ct-Kv9.3 (Fig. 2F), resulted in the expression of functional channels. The expression level of this chimera was low, as expected from the result with Nt-Kv9.3. Replacing the N terminus of rKv9.1 with the one of rKv1.3 gave the chimeric channel Nt-Kv1.3-Kv9.1, which produced no detectable current (Fig. 2H), demonstrating that amino-terminal assembly is not sufficient to convert rKv9.1 into a functional channel and the disturbed assembly in Nt-Kv9.1 and Nt-Kv9.3 cannot be the only reason for the lack of functional expression of rKv9.1 and rKv9.3 α-subunits as homomultimeric channels.
Characterization of functional domains in chimeric channels: S4 segment
A nonfunctional voltage sensor could be another reason why rKv9.1 and rKv9.3 do not function as homomultimeric channels. The first and last positively charged residues are missing in the S4 segment of rKv9.3 (Fig. 2P). To test if S4 can still act as a voltage sensor, we exchanged the S4 segment in rKv1.3 for the one of rKv9.3, thereby generating S4-Kv9.3 (Fig. 2G). The chimera S4-Kv9.3 expressed as well as rKv1.3 (Fig. 2G), thus demonstrating that the S4 domain of rKv9.3 is functional. Furthermore, the currents obtained by expressing the chimera S4-Kv9.3 made it possible to study the biophysical properties conferred by the S4 segment of rKv9.3. Figure 3A and B illustrates the voltage-dependent activation of outward currents mediated by rKv1.3 and the chimeric S4-Kv9.3 channels in response to steps in the membrane voltage to various depolarizing potentials. The S4-Kv9.3 chimeric channel displayed a slower activation at all voltages tested in comparison with rKv1.3 (Fig. 3E; Table 1). Figure 3C shows the normalized open probability voltage (Po/Po,max) relations estimated from tail current measurements. The currents mediated by the rKv1.3 channel activated at test potentials of -50 mV, whereas the currents mediated by S4-Kv9.3 activated at more depolarized potentials (-30 mV). The half-maximal open probability for S4-Kv9.3 was shifted to more depolarized potentials (Table 1) and showed a much less pronounced voltage dependence of activation for the S4-Kv9.3 chimera than for rKv1.3 (Fig. 3C; Table 1). The steady-state half-inactivation voltages of both channels were nearly identical (-30 mV), whereas the slopes of the steady-state inactivation curves differed considerably (Fig. 3D); Table 1). The voltage dependence of the deactivation of rKv1.3 and S4-Kv9.3 was not changed (Fig. 3F).
Table 1. Parameters describing activation and inactivation behavior of rKv1.3 and S4-Kv9.3 expressed in Xenopus oocytes
τ activation (ms)
τ deactivation (ms)
A summary of the biophysical properties of rKv1.3 and S4-Kv9.3 shown in Fig. 3 is presented. We calculated the parameters (Vn,1/2 and an) of the steady-state activation from a least-squares fit to the data points in Fig. 3C with a Boltzmann function of the type Po/Po,max = 1/(1 + exp [(Vn,1/2 - V)/an]). The parameters (Vh,1/2 and ah) for the steady-state inactivation curve (Fig. 3D) resulted from fitting to the data points a Boltzmann function of the type I/Imax = 1/(1 + exp [(Vh,1/2 - V)/ah]). Time constants for activation and deactivation were obtained from fits of single exponentials I(t) = ao + a1 [1 - exp (-t/τact)] to current traces or I(t) = ao + a1exp (-t/τdcact) to tail currents as described in the legend to Fig. 3.
Characterization of functional domains in chimeric channels: pore region
We generated other chimeric channels containing different parts of rKv9.3 (Fig. 2I-O) to see which other regions of rKv9.3, alone or in combination, are functional. Measurements of these chimeric constructs led to no detectable currents. A common domain in all these chimeric channels was the P region. Comparing the amino acid sequence of the P region with those of the other electrically silent α-subunits rKv8.1 and rKv9.1 and of rKv1.3 revealed three residues of particular interest (in rKv9.3: I359, C360, and W361; Fig. 2Q). The tryptophan at position 361 in rKv9.3 is found only in rKv8.1, rKv9.1, and rKv9.3. All α-subunits expressing as homomultimeric K+ channels show a phenylalanine at the corresponding position. Neither changing tryptophan into phenylalanine in the Pore-Kv9.3 chimera (Pore-Kv9.3-F; Fig. 2Q) nor the exchange of all three amino acids at positions 359-361 in rKv9.3 for the corresponding ones of rKv1.3 (DAF) generating Pore-Kv9.3-DAF (Fig. 2Q) gave rise to any detectable current. In vitro translations of all chimeric constructs were performed, and proteins of the expected molecular weight could be identified (data not shown). Taken together, these results suggest that the P region is involved in determining the electrically silent behavior of rKv9.3 homomultimeric channels but failed to indicate specific amino acid residues in the pore.
Homo- and heteromultimeric association of rKv9.1 and rKv9.3 NH2-terminal domains detected by yeast two-hybrid system
We next tested if the newly cloned α-subunits rKv9.1 and rKv9.3 can interact in a homomultimeric fashion via their amino or carboxy termini and how they interact with α-subunits of other Kv subfamilies. The selected NH2-terminal regions, containing the complete T1 domain (Shen et al., 1993; Xu et al., 1995), or the carboxy termini starting at the end of S6 were linked with the DNA binding domain and a nuclear localization signal in pLexN as well as with the transcription activation domain in pVP16-4 (see MATERIALS AND METHODS).
The NH2-terminal constructs of representative members of the Kv1-4 families as well as rKv9.1 and rKv9.3 were co-transformed into yeast and grown on UTL and THULL plates (Fig. 4). None of the constructs used gave any growth on THULL media when transformed with a fusionless vector (pLexN or pVP16-4) or with constructs containing unrelated fusion parts like pLexN-Lamin or prey103 (syntaxin 1B fused with pVP16; Fig. 4B). The pLexN and pVP16 constructs NKv1.3, NKv2.1, NKv3.4, and NKv4.2 grew after co-transformation on THULL plates, and β-gal was expressed (Fig. 4, second and third columns), demonstrating an interaction of the amino termini. Although the co-transformations of NKv9.1 or NKv9.3 constructs were successful (Fig. 4, first column), there was no growth visible on THULL media (Fig. 4, second column), and no β-gal activity could be measured in the ONPG liquid assay (Fig. 4B). Identical results were observed after co-transformation of CKv9.1 and CKv9.3 constructs (data not shown). This result suggests that both rKv9.1 and rKv9.3 α-subunits might not have the capability to interact homomerically via their amino or carboxy termini. Additionally, it could explain the reduced expression of the N-terminal chimeric channels (Nt-Kv9.1 and Nt-Kv9.3).
The co-transformations of NKv2.1 with NKv9.1 or NKv9.3 resulted in growth on UTL plates as well as on THULL plates (Fig. 4A), showing that the NH2 termini of both rKv9.1 and rKv9.3 are capable of interacting with the amino terminus of rKv2.1. This observation is supported by the expression of β-gal (Fig. 4A and B). The interaction was obtained independently, whether the NH2 terminus of the tested α-subunit (rKv2.1, rKv9.1, or Kv9.3) was fused with the DNA binding or the transcription activation domain. Additionally, we were also able to show heteromultimeric amino-terminal interactions of rKv2.1 with rKv5.1, rKv6.1, or rKv8.1, whereas homomultimeric assembly of the latter subunits was not detectable (data not shown). On the other hand, no interaction of NKv3.4 with NKv9.1 or NKv9.3 could be detected after co-transformation or by β-gal expression in the ONPG liquid assay (Fig. 4B). Furthermore, cotransformation of CKv3.4 with CKv9.1 or CKv9.3 gave no growth on THULL plates, suggesting that no interaction occurs via the carboxy termini (data not shown).
Considering all the results obtained in the yeast two-hybrid experiments, we conclude that the suppression of rKv2.1 currents by rKv9.1 and rKv9.3 most likely results from an amino-terminal interaction between these α-subunits. The site of interaction leading to a reduction of rKv3.4 currents by co-expression with rKv9.1 or rKv9.3 must instead be different.
Heteromultimeric assembly mediated by hydrophobic core region of rKv9.3 α-subunit
A possible reason why rKv3.4 currents were suppressed by co-expressed Kv9.3 α-subunits though no interaction between their amino termini could be detected is that for some α-subunits, the interaction of the hydrophobic core regions provides sufficient stabilization energy to allow channel formation even in the absence of the T1 domain (VanDongen et al., 1990; Hopkins et al., 1994; Lee et al., 1994; Tu et al., 1996). To test if this stabilizing effect of the hydrophobic core might constitute the molecular basis for the reduced expression of rKv3.4-rKv9.3 heteromultimeric channels (Fig. 5A), we constructed ▵174-Kv9.3 (Fig. 5B). Co-expression of this α-subunit missing the complete T1 domain with rKv3.4 caused an equally pronounced current reduction as observed after co-injection of the complete rKv9.3 (Fig. 5B). The co-expression of ▵174-Kv9.3 and Kv2.1 produced only a small reduction of rKv2.1 currents in comparison with the 60-fold stronger reduction observed with the complete rKv9.3 (data not shown). Altogether, the results from the yeast two-hybrid system and these measurements lead to the conclusion that the interaction of rKv9.3, and most probably also of rKv9.1, causing a down-regulation of rKv3.4 currents is guided by the hydrophobic core of the tested α-subunits. The assembly of rKv9.1 and rKv9.3 with the rKv2.1 α-subunit, on the other hand, occurs primarily through the amino termini and is only secondarily supported by interactions of the hydrophobic cores.
In this study, we showed that two members of the Kv9 potassium channel subfamily, rKv9.1 and rKv9.3, specifically suppress currents mediated by α-subunits of the Kv2 and Kv3 subfamilies. Chimeric channels revealed that the S4 segment of the Kv9 channels is functional and that the pore region is involved in determining the nonconductive properties of homomultimeric channels. Combining electrophysiology and protein-protein interaction assays, we demonstrated that the amino and carboxy termini of rKv9.1 and rKv9.3 α-subunits do not support homomeric interactions, whereas the specific heteromeric interactions with rKv2.1 are mediated by the amino terminus and with rKv3.4 most likely by the hydrophobic core.
Homo- and heteromultimeric expression of rKv9.1 and rKv9.3 α-subunits
Inability of functional homomultimeric expression has been shown for a number of ion channels, among them the amiloride-sensitive Na+ channel (Canessa et al., 1994), the cyclic nucleotide-gated channel (Chen et al., 1993), and the G protein-activated inward-rectifying K+ channel (Hedin et al., 1996). α-Subunits of the Kv5, Kv6, Kv8, and Kv9 potassium channel subfamilies do not form functional homomultimeric channels. Instead, in heterologous expression systems, they suppress the activity of α-subunits of the Kv2 or Kv3 subfamilies.
To determine the regulatory properties of rKv9.1, we performed co-expression studies with α-subunits of the known expressing subfamilies Kv1-4 cloned from rat. In agreement with Salinas et al. (1997a,b), we could show an almost complete suppression of rKv2.1 current after co-expression with rKv9.1. In contrast to their findings, we observed a suppression also for the rKv3.4 current. A possible explanation could be that Salinas et al. (1997a,b) used the Kv9.1 channel cloned from mouse (mKv9.1), which differs in 10 amino acids from the rat channel (rKv9.1; Stocker and Kerschensteiner, 1998).
Similarly, co-expression with rKv9.3 caused an almost complete suppression of rKv2.1 currents and an ∼50% reduction of rKv3.4 currents. These results are not in agreement with the current increase reported after coexpression of Kv2.1 and rKv9.3 by Patel et al. (1997). A reason for this discrepancy might be the reported metabolic regulation of Kv2.1/rKv9.3 heteromultimeric channels (Patel et al., 1997). In our study, the use of very low cRNA concentrations and the control of the oocytes’ ability to translate and express ion channels by co-injection of small amounts of IRK1 cRNA support that the suppression of rKv2.1 current is specific. In general, α-subunits of the Kv5, Kv6, Kv8, and Kv9 subfamilies seem to have a regulatory function, suppressing or changing the kinetic behavior of currents mediated by other Kv α-subunits.
Characterization of functional domains in chimeric K+ channel α-subunits
The construction of chimeric α-subunits led to the characterization of rKv9.3 functional domains. Chimeric channels harboring the amino- and carboxy-terminal regions or the S4 segment of Kv9.3 gave rise to measurable currents. This indicates that the S4 segment is not responsible for the electrical silence of the rKv9 α-subunits. However, the voltage dependence of both activation and inactivation of this chimera was drastically reduced. Comparison of the exchanged S4 amino acid sequence of rKv9.3 with rKv1.3 revealed that they differ in nine positions, making the S4 segment of rKv9.3 in total more hydrophobic (Fig. 2P). The lower charge content of the S4 segment correlates with the decreased slope observed in the voltage dependence of activation, even if a recent study questioned the correlation between charge content of the S4 segments of different Kv channels and slopes of conductance-voltage relations (Smith-Maxwell et al., 1998). In addition to the observed changes in the voltage dependence, the activation of the current mediated by chimeric S4-Kv9.3 channels is markedly slower in comparison with that of the background channel rKv1.3. A more detailed analysis, including point mutations and measurement of gating currents, will be necessary for a biophysical understanding of these processes.
The slightly extended P regions in rKv1.3 and in the chimeric channel Pore-Kv9.3 differ in only 8 of 24 positions (Fig. 2Q). In spite of these few differences in the primary sequence, no current could be measured for the chimera Pore-Kv9.3 (Fig. 2O) or after changing the most divergent residues in Pore-Kv9.3F and in Pore-Kv9.3-DAF. Inspection of the last five different residues of Pore-Kv9.3-DAF compared with rKv1.3 (Fig. 2Q) revealed that these are found in other expressing Kv channels at the corresponding positions: T365 and I366 in rKv2.1, S367 in rKv1.1, and T376 in Shaker. Furthermore, mutation analysis performed in Shaker and rKv1.3 channels at the position corresponding to 355 in rKv9.3 demonstrated that mutating the threonine had no influence on the expressed current or on the general structure of the outer channel pore, as determined by charybdotoxin binding (MacKinnon et al., 1990; Aiyar et al., 1995). Therefore, all residues that differ in Pore-Kv9.3-DAF compared with rKv1.3 occur in other α-subunits of the Kv families, and we can only conclude that their concerted presence seems to result in an overall structural change of the pore region or in a substantially altered interaction of the pore region with flanking regions (i.e., S6 segment; Aiyar et al., 1994), preventing ion conduction. All differing amino acids are located between the outer helix and the pore helix, at the end of the pore helix, and directly after the selectivity filter, according to the recently published crystal structure of a potassium channel pore (Doyle et al., 1996). Future experiments might show how these amino acid exchanges affect the overall pore structure.
Molecular domain for intersubfamily assembly
A variety of methods have been used by several groups to demonstrate that the amino terminus of Kv channels contains domains important for the subfamily-specific assembly of α-subunits, such as the NAB/T1 domain (Li et al., 1992; Shen et al., 1993; Shen and Pfaffinger, 1995). Our investigation of chimeric channels and performance of yeast two-hybrid assays of the NAB/T1 domains of members of the Kv9 subfamily support a novel pattern of assembly, with no interactions within the same subfamily but specific interactions with at least one member of the Kv2 subfamily, rKv2.1. Similar results were obtained for another electrically silent α-subunit, rKv6.1 (Post et al., 1996), suggesting that this might be a common feature of the electrically silent α-subunits (Kv5.1, Kv6.1, and Kv8.1) underlying their specific targeting and role as modulatory α-subunits. The recent crystallization of the aKv1.1 tetramerization domain (Kreusch et al., 1998) opened the possibility for future structural studies of the NAB/T1 domain of Kv2.1 with its heteromultimeric partners, which could help us to understand why members of the Kv1-4 families, but not members of the Kv6 (Post et al., 1996) or Kv9 (this report) families, show subfamily-specific assembly.
The suppression of the rKv3.4-mediated current by co-expression with rKv9.1 or rKv9.3 has most probably a different molecular basis. Although a clear suppression could be electrophysiologically measured, the yeast two-hybrid system did not support an interaction of rKv3.4 and rKv9.1 or rKv9.3 via the amino termini. Instead, the same current reduction was observed after co-injection of an α-subunit missing the complete NAB/T1 domain (▵174-Kv9.3), suggesting that the amino termini of Kv9 subunits neither promote nor inhibit assembly, and the stabilization energy-promoting channel formation is most likely provided by interactions of the hydrophobic core regions, as observed for deletion mutants by Tu et al. (1996).
Altogether, our electrophysiological measurements and yeast two-hybrid assays suggest that α-subunits of the Kv9 subfamily can interact with α-subunits of different Kv subfamilies according to at least two different molecular principles. Furthermore, we propose that the molecular basis for the previously described heteromultimeric assembly of rKv5.1, rKv6.1, Kv8.1/Kv2.3r, mKv9.1, or mKv9.2 with members of the Kv2 family (Post et al., 1996; Castellano et al., 1997; Salinas et al., 1997a,b) is most probably the same as the one described for rKv9.1 and rKv9.3 in the present study.
We are grateful to Prof. Walter Stühmer for support, lab space and equipment, valuable scientific discussion, and reading of the manuscript. We gratefully thank Drs. S. A. N. Goldstein, R. H. Joho, J. P. Ruppersberg, S. M. Hollenberg, M. Hollmann, and N. Brose for the generous gifts of clones and plasmids. We thank Dr. N. Brose for technical advice in the use of the yeast two-hybrid system, S. Voigt for oocyte injections, and the Department of Immunochemistry, particularly Ms. M. Praetor and Ms. T. Hellmann, for help with sequencing. Additionally, we thank Dr. Paola Pedarzani for invaluable scientific discussion and critical reading of the manuscript.