The physiological diversity of K+ channels mainly depends on the expression of several genes encoding different α-subunits. We have cloned a new K+ channel α-subunit (Kv2.3r) that is unable to form functional channels on its own but that has a major regulatory function. Kv2.3r can coassemble selectively with other α-subunits to form functional heteromultimeric K+channels with kinetic properties that differ from those of the parent channels. Kv2.3r is expressed exclusively in the brain, being concentrated particularly in neocortical neurons. The functional expression of this regulatory α-subunit represents a novel mechanism without precedents in voltage-gated channels, which might contribute to further increase the functional diversity of K+ channels necessary to specify the intrinsic electrical properties of individual neurons.
Electrical signaling in neurons and other excitable cells is determined mainly by the functional characteristics of their K+ channels. The best known K+channels are the voltage-gated, which comprise a diverse family of membrane proteins participating in action potential repolarization and the regulation of repetitive firing (see Connor and Stevens, 1971;Rogawski, 1985; Rudy, 1988; Hille, 1992). These K+ channels are composed of four homologous α-subunits, forming a transmembrane aqueous-conducting pore selective for K+ ions. All α-subunits share a common general design: a central core with six putative transmembrane segments, flanked by hydrophilic N- and C-terminal domains of variable length, facing the cytosol (Tempel et al., 1987; MacKinnon, 1991; Jan and Jan, 1994). There are five major families of genes encoding voltage-gated K+ channels, designated as Kv1 (Shaker), Kv2 (Shab), Kv3 (Shaw), Kv4 (Shal), and Kv5. At least two additional α-subunits, unable to produce functional K+currents, have been cloned (Drewe et al., 1992). The sequence homology among α-subunits of the same family is >70%, but this value decreases to <50% among members of different families (Kamb et al., 1987; Papazian et al., 1987; Pongs et al., 1988; Stühmer et al., 1989; Drewe et al., 1992; Salkoff et al., 1992; Chandy and Gutman, 1993; Zhao et al., 1994). The marked functional variability of the K+ currents, a characteristic particularly apparent in mammalian neurons (Llinás, 1988; Rudy, 1988), arises from several mechanisms. These include the differential tissue and cellular expression of the various genetic families (Drewe et al., 1992; Hwang et al., 1992; Deal et al., 1994; Weiser et al., 1994) and the ability of α-subunits within a family to aggregate into oligomers. The formation of heteromeric K+ channels has been demonstrated in heterologous expression systems (Christie et al., 1990; Isacoff et al., 1990; Ruppersberg et al., 1990; Covarrubias et al., 1991; Li et al., 1992) as well as in brain cells (Sheng et al., 1993; Wang et al., 1993). In addition, there are regulatory hydrophilic β-subunits that can coassemble selectively with α-subunits (see Rettig et al., 1994;Yu et al., 1996).
Here, we report the identification of a protein (which we call Kv2.3r) with a structural design similar to K+ channel α-subunits, which seems to have a major regulatory function. Kv2.3r does not appear to produce functional channels by itself; however, it can form functional heteromers with other K+ channel α-subunits, such as Kv2.1 (Drk1; Frech et al., 1989). The coexpression of Kv2.3r and Kv2.1 results in channels exhibiting striking modifications in kinetics, if compared with homomultimeric Kv2.1 channels. Kv2.3r has no effect when it is coexpressed with Rbk1 (Kv1.1) or Shaker B channels (members of the Kv1 family;Tempel et al., 1987; Christie et al., 1989). Kv2.3r appears to be expressed almost exclusively in the brain, where it may contribute to the physiological diversity of K+ channels.
A preliminary account of these data has appeared in abstract form (Castellano et al., 1996).
MATERIALS AND METHODS
Isolation of Kv2.3r cDNA. A fragment of the Kv2.3r cDNA was identified initially from poly(A+) RNA obtained from rat cerebral cortex by reverse transcription and amplification with degenerate oligonucleotides, using PCR. The PCR primers had the following sequences: forward primer, CCTCTAGAA(TC)GAGTA(TC)TT(TC)TT(TC)GA(TC)(AC)G; reverse primer, GGGGATCC(GA)TA(ATG)CC(CAT)AC(AC)GT(GT)GTCAT. These primers were designed to hybridize two sequences highly conserved in K+channels: part of box B (amino acids NEYFFDR) and the pore (MTTVGYG), respectively (see Fig. 1 below). The full-length Kv2.3r cDNA was cloned by screening rat forebrain and hippocampus cDNA libraries with probes derived from the PCR-amplified fragment and using conditions described previously (Castellano et al., 1993).
Northern blot analysis and in situ hybridization.For Northern blot analysis poly(A+) RNA was isolated from different rat tissues or CHO cells transfected with cDNA. Approximately 3 μg of poly(A+) RNA was applied in each lane, and Northern blot was performed as described (Lucas et al., 1993). We used as hybridization probe theEcoRI–ClaI fragment (nucleotides 1–442) of Kv2.3r cDNA, which has low sequence similarity with other cloned K+ channels. Washing after hybridization was done under high stringency conditions, and autoradiography developed after a few hours. Longer exposure (several days) confirmed the absence of hybridization signal in peripheral tissues. Hybridization of the blot with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or cyclophilin probes was done to evaluate the amount of RNA loaded in each lane. Forin situ hybridization cRNA antisense and sense probes for Kv2.3r were synthesized in vitro from linearized plasmid pBluescript–Kv2.3r, using 35S-UTP as described previously (Mellström et al., 1993), or with digoxigenin (DIG)-UTP, as suggested by the manufacturer (Boehringer Mannheim, Mannheim, Germany).In situ hybridization on fresh frozen rat brain sections (10 μm) was performed as described (Mellström et al., 1993). Detection of DIG was done by using anti-DIG Fab fragment conjugated to alkaline phosphatase (dilution 1:500) and nitro blue tetrazolium chloride (NBT) and 5-bromo-4-chloro-3-indolyl phosphate (BCIP) as substrates. For control of nonspecific hybridization, consecutive sections were hybridized with the corresponding sense probe.
Functional expression and electrophysiological techniques.Kv2.3r, Kv2.1, Kv1.1, and Shaker B Δ6–46 cDNAs were subcloned in plasmid p513. The tandem dimer Kv2.1/Kv2.3r was generated by ligating a DNA fragment containing the entire Kv2.3r coding sequence into a p513–Kv2.1 construct with a deletion of its last 48 amino acids. This construction included amino acids 1–805 of Kv2.1, followed by a serine and the 503 amino acids of Kv2.3r. In some experiments cDNA of green fluorescent protein (GFP) included in plasmid pRK5 was cotransfected with K+ channel α-subunits to detect by fluorescence those cells expressing K+ currents (Marshall et al., 1995). In these experiments 100% of the cells that appeared as intensely fluorescent also expressed measurable K+currents. CHO cells were transiently transfected with 2–4 μg of the cDNAs by electroporation. When two α-subunits were coexpressed, we used for transfection equal amounts of each cDNA. Potassium currents were recorded 20–48 hr after transfection by the whole-cell configuration of the patch-clamp technique (Hamill et al., 1981). We used low-resistance electrodes (1–3 MΩ), capacity compensation, and subtraction of linear leakage and capacity currents. Full compensation of series resistance was not attempted systematically. Compensation was in all experiments <50%. Solution composition was (in mm): for the bath solution, 140 NaCl, 2.7 KCl, 2.5 CaCl2, 4 MgCl2, and 10 HEPES, pH 7.4; for the solution in the pipette and inside the cell, 80 KCl, 30 K-glutamate, 20 K-fluoride, 4 ATP·Mg, 10 HEPES, and 10 EGTA, pH 7.2. In the high K+ external solutions, 70 mm NaCl was replaced for 70 mm KCl. In some experiments ZnCl2 (up to 1 mm) was added to the external solution.
Identification and cloning of Kv2.3r
A fragment of K+ channel-like cDNA initially was identified by PCR amplification of poly(A+) RNA obtained from rat cerebral cortex. This fragment was used to isolate the full-length cDNA clone from a rat brain cDNA library. We have designated this cDNA sequence Kv2.3r, referring to its sequence similarity and selective regulatory effect on Kv2.1 (see below). The clone contains 2318 nucleotides with an open reading frame of 1509 nucleotides. The 3′ untranslated region has ∼800 nucleotides, and the 5′ untranslated region is short, with only nine nucleotides and no upstream stop codon preceding the assigned initiation codon (the first ATG found). However, this codon is flanked by an A at position −3 and a G at position 4, which conforms with the main requirements of the consensus sequence for initiation of translation (Kozak, 1987). A short 5′ untranslated region (13 nucleotides) also was reported in the Kv2.1 clone (Frech et al., 1989). The Kv2.3r cDNA encodes a protein of 503 amino acids with general characteristics similar to previously cloned K+ channel α-subunits (Fig. 1). The hydropathy plot (data not shown) suggests that Kv2.3r also is composed of six putative transmembrane segments (S1–S6) and of cytoplasmic N- and C-terminal domains. This protein has several potential intracellular phosphorylation sites: two for cAMP-dependent protein kinase, eight for caseine kinase II, and six for protein kinase C.
Kv2.3r has a relatively low degree of sequence identity with other known K+ channel subunits. This is illustrated in Figure 1where, to facilitate the comparison, we have included the amino acid sequences of two representative K+ channel α-subunits, Kv2.1 and Kv1.1. The highest similarity of Kv2.3r (44%) was found with members of the Kv2 family (Kv2.1 and Kv2.2; Frech et al., 1989; Drewe et al., 1992; Hwang et al., 1992), although important sequence divergence is observed in the loops among the transmembrane segments and in the amino and carboxyl ends. The degree of identity of Kv2.3r with channels of the Kv1 family is ∼35%. When the core regions (S1–S6) are compared, similarity of Kv2.3r with Kv2.1 increases to slightly <50%. Kv2.3r shares several important structural features with the channels of the Kv2 family, including the five positively charged amino acids in the S4 segment and the conservation of amino acids 132–136, which are part of the distinct Kv2 region described byWei et al. (1990). These observations suggest that Kv2.3r may have evolved from the Kv2 channel family.
Although Kv2.3r does not seem to have intrinsic channel activity (see below), a regulatory role for this α-subunit was suggested by its distinctive tissue distribution. Whereas Kv2.1 and Kv2.2 channels are highly expressed in both brain and non-neural tissues (e.g., heart and skeletal muscle) (Drewe et al., 1992; Hwang et al., 1992), Kv2.3r seems to be expressed almost exclusively in the brain (Fig.2 A). Northern blot analysis indicates that there are two Kv2.3r mRNA species of ∼5 and 3.3 kb that also seem to be distributed differentially within the brain. The 5 kb transcript is more abundant in cerebellum, whereas the 3.3 kb species is more prominent in the neocortex (Fig. 2 A).In situ hybridization studies indicate that the greatest density of Kv2.3r mRNA is found in the neocortical layers II, III, and VI (particularly in the frontal cortex), olfactory tubercle, hippocampus (C1–C4 as well as dentate gyrus), piriform cortex, amygdala, and cerebellum (Purkinje and granular cells). Moderate densities of labeling occur in the olfactory bulb, striatum, septum, supraoptic nucleus, and lateral reticular nucleus (Fig.2 B). Notable is the near absence of labeling in large areas like the diencephalon and the brainstem. The expression of Kv2.3r mRNA within representative neurons, including the large neocortical pyramidal cells, is illustrated in Figure 2 C. Thus, there are several brain regions with high density of Kv2.3r mRNA overlapping the distribution of other K+ channels (Beckh and Pongs, 1990; Séquier et al., 1990; Drewe et al., 1992; Hwang et al., 1992; Weiser et al., 1994). This fact leads us to hypothesize that Kv2.3r might be regulating the function of these K+channels at specific locations.
Expression of Kv2.3r
In vitro translation of Kv2.3r cRNA in reticulocyte lysates produced a protein of molecular weight similar to that predicted from the cDNA-derived amino acid sequence (∼56 kDa; Fig. 3 A). In addition, Northern blot analysis in cells transfected with Kv2.3r cDNA indicated that it can be transcribed into the corresponding mRNA (Fig. 3 B). However, under standard experimental conditions we were unable to detect measurable voltage-dependent K+ currents after injection of Kv2.3r cRNA into Xenopus oocytes or after transfection of CHO or HEK cells with cDNA. We also have studied cells cotransfected with K+ channels and GFP cDNAs to discard the possibility that Kv2.3r is expressed with very low efficiency and, therefore, only present in a few cells difficult to identify in patch-clamp experiments. All CHO cells cotransfected with GFP and Kv2.1 appearing as intensely fluorescent under the microscope exhibited large macroscopic K+ currents with normal activation and inactivation kinetics (Fig. 4 A). In contrast, we never saw detectable K+ currents in intensely fluorescent cells cotransfected with GFP and Kv2.3r cDNA. In Figure4 B we show recordings at high gain of a representative example of this last type of experiment. The expression of Kv2.3r also was studied in high external K+, which is known to facilitate the opening of some K+ channels (Pardo et al., 1992; López-Barneo et al., 1993). Because the effect of external K+ is influenced by the amino acid in the position equivalent to R402 of Kv2.3r (Pardo et al., 1992; López-Barneo et al., 1993), we tested whether the replacement of arginine by tyrosine (amino acid present at this position in Kv2.1 and Kv1.1) resulted in functional channels. Neither high external K+ nor the R402Y mutation of Kv2.3r yielded any measurable current. Thus, it seems reasonable to conclude that Kv2.3r is unable to form homomultimeric functional channels in heterologous expression systems. Among other factors, the lack of functional expression of Kv2.3r could be related to the presence of several amino acids (e.g., S234, K268, V382, W386, or A412) in Kv2.3r, which differ from residues highly conserved in voltage-gated K+ channels (see Drewe et al., 1992).
Functional characteristics and regulatory properties
The fact that Kv2.3r mRNA is highly concentrated in specific areas of the brain (see above) suggested that it could have a regulatory role by forming heteromultimeric channels. This idea was tested by coexpressing Kv2.3r with other K+ channel α-subunits in CHO cells. Cotransfection of CHO cells with Kv2.3r plus Kv2.1, its most closely related structural K+ channel homolog, resulted in macroscopic K+ currents with somewhat smaller amplitude and profoundly different kinetics when compared with the currents produced by Kv2.1 alone. The Kv2.3r plus Kv.2.1 currents exhibited a striking deceleration of kinetics, with four- to fivefold slowing of activation and inactivation time courses observed at some membrane potentials (see Figs. 5, 6, Table1). In contrast, the coexpression of Kv2.3r with either Kv1.1 (Fig. 5 B) or Shaker BΔ6–46 (Fig.5 C) led to the expression of currents that, although of smaller amplitude, were kinetically indistinguishable from the currents generated by Kv1.1 or Shaker BΔ6–46 alone (see Table1).
The effects of Kv2.3r on the Kv2.1 currents are illustrated further in Figures 6 and 7. Figure 6 Ashows representative current traces recorded at various membrane potentials in cells transfected with either Kv2.1 alone or with Kv2.1 plus Kv2.3r. The corresponding current–voltage curves, plotted in Figure 6 B, indicate that the presence of Kv2.3r produced a displacement toward positive membrane potentials of the voltage dependence of activation. The marked voltage dependence of the slowing of activation time course induced by Kv2.3r is illustrated in Figure 6 C. Notably, the effects of Kv2.3r on channel closing were also voltage-dependent (see Table 1). Figure 7 shows current traces obtained from cells transfected with either Kv2.1 alone or with Kv2.1 plus Kv2.3r, using high K+ (70 mm) in the external solution (Fig. 7 A). Tail currents recorded at the end of the pulses represent the deactivation time course of the channels. The current–voltage curves in Figure 7 B show that, despite all of the kinetics modifications caused by Kv2.3r on Kv2.1 channels, the reversal potential (at approximately −10 mV) was unchanged, indicating that the selectivity for K+ was not greatly altered. The conductance–voltage relationships shown in Figure7 C illustrate that the presence of Kv2.3r leads to a shift in the voltage dependence of activation. At half-maximal activation voltage, the shift induced by Kv2.3r was between 10 and 15 mV.
Further demonstration that Kv2.3r can, indeed, interact with Kv2.1 to form heteromultimeric K+ channels was obtained by the construction of a tandem dimer with the two α-subunits. This dimer was created by ligation of the 3′ end of the translated region of Kv2.1 and the 5′ end of Kv2.3r. The expression of this cDNA construct in CHO cells resulted in robust macroscopic K+ currents that presumably represent the activity of a homogeneous population of heteromultimeric Kv2.1/Kv2.3r channels (Fig.8). The currents recorded by the expression of the Kv2.1/Kv2.3r dimer exhibited modifications in activation and inactivation kinetics qualitatively similar to those produced by the coexpression of Kv2.1 and Kv2.3r (see Figure 8 legend). The formation of heteromultimeric channels by the Kv2.1/Kv2.3r dimer also was evaluated by studying its sensitivity to external Zn2+. The activity of many K+ channels is reduced by externally applied Zn2+ (see Gilly and Armstrong, 1982). However, it is known that the Kv2.1 channels are rather insensitive to the cation, with a Kd value of 14 mm (De Biasi et al., 1993). Because Kv2.1 and Kv2.3r have a high degree of sequence divergence in the extracellular loops between the transmembrane segments (see Fig. 1), we hypothesized that the heteromultimeric Kv2.1/Kv2.3r channels also could have modified their sensitivity to external Zn2+. In accord with this idea, Figure 9 shows that Zn2+ had a clearly larger effect on the dimer Kv2.1/Kv2.3r than on the homomultimeric Kv2.1 channels. Whereas at +20 mV 1 mm Zn2+ reduced Kv2.1 currents to only 84 ± 10% of the control value (Fig. 9 A; also see De Biasi et al., 1993), it diminished Kv2.1/Kv2.3r currents to almost one-half of the control amplitude (53 ± 12%; n = 4; Fig.9 B). Together, these data suggest that Kv2.3r can coassemble with Kv2.1, forming heteromultimeric K+ channels with characteristic functional properties.
The major finding in this paper is the cloning and functional characterization of a new K+ channel α-subunit that seems to have a regulatory function. Kv2.3r appears to be unable to form functional channels on its own but it seems to coassemble with other K+ channel α-subunits to form functional heteromultimeric channels with special biophysical features. Kv2.3r is expressed exclusively in the brain, partially overlapping the expression of other K+-channel α-subunits (Drewe et al., 1992; Hwang et al., 1992; Weiser et al., 1994). These facts suggest that Kv2.3r may contribute to specify the intrinsic electrophysiological properties of neurons. The detailed study of the interaction of Kv2.3r with the various families of K+ channel α-subunits currently is being undertaken; however, the available data indicate that it selectively may regulate channels of the Kv2 family. These channels are highly concentrated in neocortex, cerebellum, and hippocampus (Drewe et al., 1992; Hwang et al., 1992), which are the areas with the highest expression of Kv2.3r.
Although the molecular mechanism underlying the interaction of Kv2.3r with other K+ channel α-subunits is unknown for the moment, it may depend, at least in part, on the degree of structural similarity among their N termini. It is known that this domain is important for subunit recognition and assembly (Li et al., 1992; Lee et al., 1994). In addition, recently the existence of highly conserved regions within the N-terminal of α-subunits belonging to the same family (A and B boxes in Fig. 1) has been reported, which seem to determine subfamily-specific associations (Xu et al., 1995; Yu et al., 1996). The A box of Kv2.3r has low sequence identity with the same region of the other α-subunits studied here (40, 40, and 45% with Kv2.1, Kv1.1, and Shaker BΔ6–46, respectively); however, the B box of Kv2.3r has clearly higher similarity with the B box of Kv2.1 (68%) than with the same region of Kv1.1 (43%) orShaker BΔ6–46 (44%). Thus, the higher similarity of their respective B regions possibly permits Kv2.3r to select Kv2.1 against Kv1.1 or Shaker B. Part of the effects of the expression of Kv2.3r might result from altered subunit assembly, because the heteromeric Kv2.1/Kv2.3r channels share many kinetic properties with those exhibited by Kv2.1 channels with large N-terminal deletions (VanDongen et al., 1990).
During the final stages of preparation of this paper, an article appeared (Hugnot et al., 1996) reporting the cloning from hamster tissues of a K+ channel α-subunit (called by these authors Kv8.1) with essentially the same structure and tissue distribution as the Kv2.3r described here. Although there are only minor sequence differences between the two clones (three amino acid substitutions in the N-terminal, one substitution in the S2–S3 linker, and the insertion of a glycine in position 20), the reported functional properties are quite different. Like Kv2.3r, Kv8.1 does not have intrinsic channel activity; however, in Xenopus oocytes Kv8.1 seems to abolish the functional expression of K+channels of the Kv2 and Kv3 families. Hugnot and colleagues (1996) have shown that the interaction of Kv8.1 with the other α-subunits seems to occur through the N-terminal domains. These authors indicated in their article that, besides the original clone from hamster, they used a cDNA clone isolated from rat brain with similar electrophysiological results. Thus, the differences between their data and our observations described here are difficult to explain and could be ascribed to the fact that we used mammalian cells as the cDNA expression system instead of Xenopus oocytes.
On the basis of our experimental results, it is most tempting to propose that the normal function of Kv2.3r (or Kv8.1) is neither to produce channels by itself nor to abolish the expression of other channels but to coassemble with selected α-subunits to form heteromultimeric channels with modified electrophysiological properties. There are no precedents of this type of regulatory function in α-subunits of voltage-gated channels; however, it is a phenomenon well known in ligand-gated channels. For example, both NMDA (Monyer et al., 1992) and nicotinic (McGehee and Role, 1995) receptors contain membrane-spanning subunits without intrinsic channel activity that are functional only in heteromultimeric channels. Thus, Kv2.3r could be the first identified member of a family of regulatory α-subunits, possibly including other previously cloned K+ channel subunits without intrinsic channel activity (such as IK8 and K13; Drewe et al., 1992). Mammalian central neurons are known to exhibit a wide variety of characteristic electrical responses independent of their morphology or synaptology (see Llinás, 1988). Hence, the selective expression in individual neurons of regulatory K+channel α-subunits is a mechanism that could be used to finely tune their intrinsic electrophysiological properties.
This work was supported by grants from the Spanish Ministry of Science and Education. A.M. is a fellow of the Spanish Ministry of Science and Education. F.M. is recipient of a fellow/credit from Colciencias (Colombia). We thank Drs. P. de la Peña and A. Carrión, Mrs. P. Ortega, and Mr. R. Pardal for help with the experiments; we also thank Drs. J. Pascual, A. Brown, and R. Aldrich for providing us with the Kv2.1, Kv1.1, and Shaker Δ6–46 cDNAs. We are indebted to Drs. A. Franco-Obregón and L. Tabares for valuable comments on this manuscript.
Correspondence should be addressed to Dr. J. López-Barneo, Facultad de Medicina, Departamento de Fisiologı́a Médica y Biofı́sica, Universidad de Sevilla, Avenida Sánchez Pizjuán 4, E-41009, Sevilla, Spain.