Mutation of Histidine 105 in the T1 Domain of the Potassium Channel Kv2.1 Disrupts Heteromerization with Kv6.3 and Kv6.4*

The voltage-activated K+ channel subunit Kv2.1 can form heterotetramers with members of the Kv6 subfamily, generating channels with biophysical properties different from homomeric Kv2.1 channels. The N-terminal tetramerization domain (T1) has been shown previously to play a role in Kv channel assembly, but the mechanisms controlling specific heteromeric assembly are still unclear. In Kv6.x channels the histidine residue of the zinc ion-coordinating C3H1 motif of Kv2.1 is replaced by arginine or valine. Using a yeast two-hybrid assay, we found that substitution of the corresponding histidine 105 in Kv2.1 by valine (H105V) or arginine (H105R) disrupted the interaction of the T1 domain of Kv2.1 with the T1 domains of both Kv6.3 and Kv6.4, whereas interaction of the T1 domain of Kv2.1 with itself was unaffected by this mutation. Using fluorescence resonance energy transfer (FRET), interaction could be detected between the subunits Kv2.1/Kv2.1, Kv2.1/Kv6.3, and Kv2.1/Kv6.4. Reduced FRET signals were obtained after co-expression of Kv2.1H105V or Kv2.1H105R with Kv6.3 or Kv6.4. Wild-type Kv2.1 but not Kv2.1H105V could be co-immunoprecipitated with Kv6.4. Co-expression of dominant-negative mutants of Kv6.3 reduced the current produced Kv2.1, but not of Kv2.1H105R mutants. Co-expression of Kv6.3 or Kv6.4 with wt Kv2.1 but not with Kv2.1H105V or Kv2.1H105R changed the voltage dependence of activation of the channels. Our results suggest that His-105 in the T1 domain of Kv2.1 is required for functional heteromerization with members of the Kv6 subfamily. We conclude from our findings that Kv2.1 and Kv6.x subunits have complementary T1 domains that control selective heteromerization.

Voltage-gated potassium channels of the Kv family are strongly expressed in the mammalian central nervous system, in the immune system, in muscle cells and in many other cell types. The members of the Kv channel family differ in their time courses of activation and inactivation and in their pharmacological properties. They play a key role in controlling neuronal excitability and regulate a variety of electrophysiological properties, such as the interspike membrane potential, the waveform of the action potential and the firing frequency (1). Most neurons express multiple Kv channel subtypes belonging to one or more subfamilies (2)(3)(4). Kv channels are composed of four subunits that surround the central ion permeation pathway. Each subunit has six transmembrane domains (S1-S6) and a pore region containing the signature sequence GYG characteristic for potassium channels (5). Post-translational assembly of tetrameric Kv channels takes place in the ER 2 membrane; subsequently the channels traffic to the plasma membrane (6,7). A highly conserved sequence in the cytoplasmic N terminus of Kv channels, the tetramerization domain or T1 domain, has been shown to play an important role in channel assembly (6,8). The T1 domain contains some of the molecular determinants for subfamily-specific homo-or heterotetrameric assembly of Kv ␣-subunits (9 -11). The most striking difference between the T1 domains of Kv1 (Shaker) and Kv2-4 (non-Shaker) channels is the presence of intersubunit-coordinated Zn 2ϩ ions at the assembly interface in non-Shaker channels. The Zn 2ϩ ions are coordinated by a C3H1 motif embedded in a conserved sequence motif (HX 5 CX 20 CC) of the T1 domain, which is located near the distal end of the N terminus (12)(13)(14). These four amino acids are exposed on the subunit interface, with one histidine and two cysteine residues belonging to one subunit and one cysteine residue belonging to the neighboring subunit (12). The T1 domain facilitates tetrameric assembly of Kv channels. Kv subunits in which the T1 has been deleted have been reported to assemble in a promiscuous way via their transmembrane domains and to form stable, functional channels, but both the rates and the efficiency of channel assembly are significantly lower in the mutant channels as compared with their wild-type counterparts (7,15). Heteromeric assembly of channel subunits is a potential source of diversity of K ϩ channel properties. The voltage-activated potassium channel subunits * This study was supported by the Medizinstiftung (to M. M. S. and R. P. M.), Kv2.1 and Kv2.2 are capable of heteromeric assembly with members of the Kv6 subfamily, which generates channels with different biophysical properties compared with homomeric Kv2.1 channels (16 -20). In the present study we used fluorescence energy resonance transfer (FRET), yeast two-hybrid analysis, co-immunoprecipitation, live-cell imaging, and electrophysiological recordings in transfected mammalian cells and in Xenopus oocytes to study heteromerization of Kv2.1 with Kv6. 3 and Kv6.4 channels. We describe a mutation in the Zn 2ϩ ioncoordinating C3H1 motif of the T1 domain of Kv2.1 that disrupts formation of functional heteromeric channels, but does not prevent formation of homomeric channels.
Site-directed Mutagenesis-For single amino acid replacement or deletion mutagenesis the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) was used. Oligonucleotide primers were designed with at least 12 matching nucleotides on each side of an intended nucleotide change. All mutants were completely sequenced.
Direct Interaction Yeast Two-hybrid Assay-The Matchmaker yeast two-hybrid system (Clontech) was used to assay for protein-protein interactions. The T1 domains of rat Kv1.1 (aa 37-142) and human Kv2.1 (aa 31-148), Kv2.2 (accession number AF450111, aa 35-151), Kv6.4 (aa 58 -176), and Kv6.3 (aa 11-126) were amplified by PCR and cloned into the yeast DNAbinding domain-containing vector pGBT-9 and the activation domain-containing vector pGAD-424. To check for direct interaction between two proteins 150 ng of both vectors were co-transformed into the yeast strain HF7c according to the manufacturer's protocol. Transformants were plated on medium lacking tryptophan, leucine, and histidine, and incubated for 3-4 days at 30°C. Yeast transformed with interacting fusion proteins grow as a result of activation of the HIS3 gene.
To determine the strength of the interaction, HIS3 positive clones were grown on 5-10 mM 3-aminotriazole plates and ␤-galactosidase assays were performed to verify interaction.
For immunoprecipitation experiments COS-7 cells were harvested and lysed 24 h after transfection. Immunoprecipitation was performed with anti-Kv2.1 antibodies (Alomone, Israel) and protein A-Sepharose beads (GE Healthcare). After intensive washing the protein-antibody complexes were eluted with SDS-sample buffer, heat-denatured (100°C) for 5 min, separated on 6 or 8% SDS-polyacrylamide gels and transferred to nitrocellulose membranes. The proteins of interest were visualized by immunostaining the Western blots with anti-Kv2.1 antibodies (Alomone; 1:2000 dilution) or anti-GFP antibodies (Abcam; 1:2500 dilution). The binding of the primary antibodies was detected using peroxidase conjugated goat anti-rabbit antibodies (Pierce; 1:2500 dilu-tion) and a chemiluminescent extended duration substrate (Super Signal West Dura, Pierce).
Heterologous Expression in CHO-K1 Cells and Electrophysiological Measurements-CHO-K1 2 cells (200,000 cells per 32-mm cell culture dish) were transfected with 1 g of cDNA coding for Kv subunits in pcDNA3.1 or pIRES and 0.1 g of pEGFP reporter plasmid using 3.3 l FuGENE6 (Roche Applied Science).
Patch clamp recordings were carried out on the stage of an inverted microscope (Olympus IX 70) at room temperature (23°C) 10 -24 h after transfection. Adherent CHO-K1 cells in culture dishes, in which an insert was placed to form a low volume chamber, were superfused with the standard extracellular solution (NaCl, 140 mM; KCl, 5.4 mM; CaCl 2 , 2 mM; MgCl 2 , 1 mM; glucose, 10 mM; HEPES, 10 mM; titrated to pH 7.4 with NaOH). Data of whole cell recordings were collected with an EPC9 or EPC10 patch clamp amplifier (HEKA, Lambrecht, Germany) using the Pulse software v8.7 (HEKA). For whole cell measurements, patch pipettes made of borosilicate glass (Science Products, Hofheim, Germany) and coated with R-6101 (Dow Chemical, Midland) were used. They had resistances of 2.5 to 4.0 M⍀ when filled with the intracellular solution (KCl, 80 mM; potassium-glutamate, 50 mM; Mg-ATP, 2 mM; MgCl 2 , 2 mM; buffered at 100 nM free Ca 2ϩ with 10 mM BAPTA; HEPES, 10 mM; titrated to pH 7.2 with KOH). In whole cell measurements, the liquid junction potentials were ϩ8.0 mV with standard extracellular solution and ϩ4.9 mV with the high K ϩ extracellular solution (KCl, 130 mM; CaCl 2 , 2 mM; MgCl 2 , 1 mM; glucose, 10 mM; HEPES, 10 mM; titrated to pH 7.4 with KOH), and were corrected by the Pulse software. The osmolarity of all solutions was controlled with a vapor osmometer (Vapro 5520, Wescor Inc., Logan).
The values of pipette and cell membrane capacitance were estimated with the Pulse software and corrected automatically. Series resistance compensation (70 -80%) was used to limit voltage errors to less than 5 mV. Only cells with minimal leakage currents were used. The data were acquired at a sampling rate of 5 kHz after filtering at 1.67 kHz. For quantification of the time course of inactivation, the cells were held at a potential of Ϫ80 mV and three 30-s depolarizing pulses to 0 mV were applied with an interpulse interval of at least 2 min. The decay of the resulting transient outward currents was fitted with a bi-exponential function. A mono-exponential function was not sufficient to fit the data and tri-exponential functions did not provide a better fit.
The activation characteristics of the Kv channels were determined in symmetrical high-K ϩ solution. The cells were depolarized from a holding potential of Ϫ80 mV to potentials between Ϫ70 and ϩ40 mV (with increments of 10 mV) for 100 ms. Tail current amplitudes arising from subsequent repolarization to Ϫ80 mV were measured isochronally immediately after the depolarizing voltage steps. The data were normalized to the tail current obtained after a voltage step to ϩ40 mV and fitted to the modified Boltzmann Equation 1, where I tail is the current amplitude at membrane potential V m , I max is the current amplitude used for normaliza-tion, V 1/2 is the half-activation potential, and k is the slope factor. The values obtained for V 1/2 and k are reported as means Ϯ S.E.
To confirm the activation curves obtained from tail current measurements in high-K ϩ solution we performed additional experiments with standard extracellular solution (5.4 mM K ϩ ). In this case K ϩ outward currents were elicited by 1-s depolarizing steps from Ϫ80 mV to potentials between Ϫ70 and ϩ80 mV applied at intervals of 22 s. The conductance was calculated from the equation: G ϭ I/(E -E K ), where I is the peak current, E is the potential and E K is the K ϩ equilibrium potential (Ϫ80.9 mV). The conductance was normalized to the conductance at ϩ80 mV to give G/G max . Activation curves were fitted with the Boltzmann function using a Levenberg-Marquardt algorithm in Origin software v7.5. The results obtained with the two approaches were very similar. The data presented in the results are from the measurements in high-K ϩ solution, which had a better signal-to-noise ratio.
FRET Analysis-FRET measurements were carried out with an inverted epifluorescence microscope (IX70, Olympus) using an oil-immersion objective (40ϫ/1.35 NA, Olympus) and a digital video imaging system (TILL-Photonics, Gräfelfing, Germany) equipped with a monochromator. All FRET experiments were performed with live adherent HEK293 cells. FRET efficiencies were determined by monitoring the increase in the ECFP (FRET-donor) fluorescence emission during selective EYFP (FRET-acceptor) photobleaching. The photobleaching protocol for EYFP consisted of 35 cycles of illumination at 512 nm for 4.5 s. The detection protocol for ECFP and EYFP consisted of 45 cycles (10 before bleaching) with 50-ms illumination at 440 nm, followed by a 50-ms delay and 50-ms illumination at 500 nm. The delay between bleaching and detection was 120 ms, and the entire cycle time was 5 s. An intramolecularly fused CFP-YFP tandem protein (kindly provided by Michael Schaefer, Pharmacology, Berlin) resulted in a FRET efficiency of almost 50%. Four or more independent transfections with 3-14 single cells were used for the calculation of FRET efficiency. FRET efficiencies were only accepted when the ratio of fluorescence intensity of EYFP over ECFP was higher than 2.4 and EYFP bleach was greater than 80%.

RESULTS
Sequence and Predicted Structure of the T1 Domain-To identify a possible T1 domain in the Kv6 channel subunits, the amino acid sequences of human Kv6 were compared with those of the well-characterized Kv2 subfamily (Fig. 1A). We found that the cytoplasmic N terminus of all Kv6 subunits contains a conserved non-Shaker T1 domain with an amino acid similarity of 54 -66% to the T1 domain of Kv2.1. The T1 domain of Kv6 subunits exhibits one distinctive feature: The histidine (H) of the zinc ion-coordinating C3H1 motif is replaced by arginine (R) (Kv6.1-6.3) or valine (V) (Kv6.4). This substitution of the histidine appears to be characteristic for the Kv6 subfamily. The phylogenetic tree revealed that the Kv2 subfamily is more closely related to the Kv6 subfamily than the to the Kv5, Kv8, and Kv9 subfamilies (Fig. 1B). However, unlike the Kv6 subfam-ily, the Kv5, Kv8, and Kv9 subfamilies harbor a histidine residue in the C3H1 motif, indicating a potential functional relevance of the histidine substitution.
Gene Expression Analysis-Expression profiles of the Kv2.1, Kv6.3, and Kv6.4 channel subunits were examined by quantitative RT-PCR of total RNA isolated from different human and rat tissues ( Fig. 2A,  upper and lower panel). Robust expression of Kv2.1 could be detected in all tissues examined. Kv2.1 was co-expressed with Kv6.3 and Kv6.4 in human and rat brain and in the adrenal gland. In addition, we found marked co-expression of Kv2.1 with Kv6.3 in human and rat lung, colon, and kidney, and co-expression of Kv2.1 with Kv6.4 in rat heart, lung, liver, colon, and kidney. We also examined the expression Kv2.1 and Kv6.x subunits in eight different brain regions by qPCR (Fig. 2B). Again, a strong expression of Kv2.1 as well as co-expression of Kv6.3 and Kv6.4 could be detected in all brain regions examined. Co-expression of Kv2.1 and Kv6.4 subunits was found in the rat aortic smooth muscle cell line A7r5 (Fig. 2C, upper panel) and coexpression of Kv2.1 and Kv6.3 was found in the human adrenocortical cell line NCI-H295R (Fig. 2C, lower panel).
Co-expression of Kv6 with Kv2.1 Subunits-Kv6.x and/or Kv2.1 subunits were expressed in CHO-K1 cells and the amplitude and kinetics of the resulting currents were analyzed with the whole cell patchclamp technique. Expression of Kv6.1, Kv6.3, or Kv6.4 subunits alone did not produce any voltage-dependent K ϩ current (not illustrated), confirming that these subunits are silent (18,19,22). To test for possible formation of heteromeric Kv6.x/Kv2.1 channels we then co-expressed Kv6.1, Kv6.3, or Kv6.4 subunits with Kv2.1 using an IRES-based expression vector in which the transcription of both channel subunits was under the control of the same promoter. This vector induced the translation of two open reading frames from one mRNA transcript. However, the translation of the second open reading frame (Kv2.1) was about 10-fold lower than that of the first one (Kv6), which minimized the formation of homomeric Kv2.1 channels. As illustrated in Fig. 3A, the time course of inactivation of the heteromeric channels was much slower than that of homotetrameric Kv2.1 channels. The inactivation could be described by two time constants   (Fig. 3B, right  bars).
Residue His-105 of Kv2.1 Is Important for Interaction of the T1 Domains-Because the replacement of the histidine residue in the C3H1 motif is a distinctive property of the Kv6 subfamily, we analyzed whether replacing histidine 105 of Kv2.1 by valine (H105V) would affect the interaction between the T1 domains. Yeast two-hybrid (Y2H) analysis was used to study the interaction between T1 domains of Kv2.1, Kv2.2, Kv6.3, and Kv6.4; the T1 domains analyzed consisted of about 120 amino acids in the N-terminal region of the Kv subunits (underlined in Fig. 1A). As a negative control, we used the T1 domain of Kv1.1, which does not contain a C3H1 motif. The latter T1 domain interacted exclusively with itself but not with the T1 domains of Kv2 or Kv6 members (Fig. 4A), which is consistent with several other studies (10,11,23). As expected, T1 domains of wild-type (wt) Kv2.1 and Kv2.2 interacted with themselves and also with the T1 domains of Kv6.3 and Kv6.4 (Fig. 4, A and B, left panel).
An interaction between Kv6.3 and Kv6.4, and interaction of these domains with themselves, could not be detected. However, mutant Kv2.1 H105V T1 domains were found to interact with themselves and with wt Kv2.1-T1 domains. Interestingly, co-transformation of mutant Kv2.1 H105V with wt Kv6.3 or Kv6.4 T1 domains resulted in loss of interaction (Fig. 4, A and B   but not with the mutant Kv2.1 H105V (Fig. 5A (upper panel) and B (right panel)). The input controls (Fig. 5A, middle and  To examine channel assembly in intact cells we used a FRET approach. Full-length Kv subunits, whose N termini were tagged with EYFP and ECFP, were transiently expressed in HEK293 cells and studied by live cell imaging. FRET efficiencies were determined by measuring the increase in ECFP fluorescence emission during selective EYFP photobleaching (ECFP was the FRET-donor and EYFP the FRET-acceptor). Detection of a significant level of FRET between the two different fluorescent molecules was used as an indicator of channel assembly. As a positive control we analyzed homomeric assembly of Kv2 channels. ECFP-Kv2.1 co-expressed with EYFP-Kv2.1 showed a typical time course of increase in ECFP intensity over initial levels (Fig. 6A), and the same result was obtained with ECFP-Kv1.1 and EYFP-Kv1.1 (Fig. 6A).
In contrast, the combination of Kv1.1 and Kv2.1 produced no measurable FRET signal during EYFP photobleaching. On average, a FRET efficiency of more than 20% was obtained for homomeric combinations and a FRET signal of 1.5 Ϯ 0.3% for the non-interacting channel combination (Fig. 6B). Consistent with our Y2H analysis, co-expression of Kv2.1 with Kv6.3 or 6.4 subunits generated FRET signals of more than 20%. However, when Kv6.3 or Kv6.4 were co-expressed with the mutants Kv2.1 H105V or Kv2.1 H105R a highly significant reduction in FRET efficiency was found (Fig. 6C). The combination of the non-interacting subunits Kv6.3 or Kv6.4 with themselves or with Kv1.1 also displayed low FRET efficiencies. When the FRET donor and the FRET acceptor in the different subunits were exchanged, the same results were obtained; for example, the inverse combination ECFP-Kv2.1/EYFP-Kv6.3 was associated with a FRET efficiency of 24.8 Ϯ 1.4% (n ϭ 24 from 4 transfections; not illustrated), which is very similar to the example shown in Fig. 6C (first green bar).
Interestingly, FRET signals for homomeric assembly of Kv2.1 H105V or Kv2.1 H105R subunits were significantly higher than for their wt counterparts (Fig. 6D). Nevertheless, the current amplitude measured with the mutant channels in the cell membrane was almost the same as of wt channels (supplemental Fig. S1), which suggests that higher FRET efficiencies do not necessarily result in a better surface expression. However, the homomeric mutant channels differed substantially from wt Kv2.1 in their activation and inactivation kinetics (Fig. 8, B-D). An additional slow time constant of activation was observed in Kv2.1 H105R and Kv2.1 H105V and the slow time constants of inactivation were about 6-fold longer (Fig. 8C).  To further characterize the possible roles of the crucial residues in the T1 domain (position 84 in Kv6.3 and position 138 in Kv6.4) in preventing homomeric channel assembly of Kv6.x channels we performed additional FRET measurements. In either Kv6 subunit we replaced these residues by histidine, which is at the corresponding position in Kv2.1. Supplemental  Fig. S2A shows that Kv6.3 R84H and Kv6.4 V138H displayed FRET efficiencies of less than 5%, indicating the absence of channel assembly (supplemental Fig. S2A). These findings confirm the results of our Y2H analysis. Consistent with these data, live-cell imaging experiments showed that neither wt nor mutant Kv6.3 or Kv6.4 channels were localized to the surface membrane (supplemental Fig. S2B). In conclusion, introducing a histidine residue, which is part of the Zn 2ϩ -coordinating C3H1 motif in Kv2.1, at the corresponding position in Kv6.3 and Kv6.4 did not lead to the formation of homomeric Kv6 channels.
Replacement of Histidine 105 Prevents Formation of Functional Heterotetrameric Channels-Homomeric Kv2.1 und heteromeric Kv2.1/Kv6 channels differ not only in their rate of inactivation (Fig. 3) but also in some other functional characteristics. One example is the rate of deactivation, determined by the analysis of tail currents (Fig. 7A). The rate of deactivation was much slower in the heteromeric channels. Another distinc-  tive feature is the activation curve of the channels, which was calculated from the amplitudes of the tail currents (see "Experimental Procedures"). Homomeric Kv2.1 channels were halfmaximally activated at Ϫ11.3 Ϯ 0.9 mV with a slope factor k of 9.2 Ϯ 0.8 mV as derived from the Boltzmann fit (Fig. 7B). Coexpression of Kv2.1 with Kv6.1 or Kv6.3 channel subunits resulted in a marked leftward shift of the activation curve. In the case of co-expression of Kv2.1 with Kv6.1, an increase in the slope factor k to 12.8 Ϯ 0.6 mV was also noted (Fig. 7B). A characteristic rightward shift of almost 5 mV was observed after the co-expression of Kv2.1 with Kv6.4, with half-maximal activation at Ϫ6.4 Ϯ 0.4 mV.
Changes in these functional characteristics were used as a test for the role of the histidine residue at position 105 of Kv2.1 in heteromeric assembly of the different subunits. Expression ofKv2.1 H105V orKv2.1 H105R subunitsaloneresultedinvoltagedependent currents with a slower deactivation than the wt counterparts (compare Fig. 7, C and A, left). In addition, homomeric Kv2.1 H105V , but not homomeric Kv2.1 H105R channels, showed a significant rightward shift of the activation curve by about 14 mV compared with wt channels (Fig.  7, B and D). Co-expression of Kv2.1 H105R with Kv6.3 or coexpression of Kv2.1 H105V with Kv6.4 affected neither the rate of deactivation (Fig. 7C) nor the position of the activation curve (Fig. 7D), which suggests that heteromeric assembly was impaired or abolished by the replacement of His-105 by valine or arginine.
We then performed an additional series of experiments in which Kv2.1 and Kv6.x subunits were co-expressed in Xenopus oocytes (Fig. 8). This expression system has the advantage that precisely controlled amounts of cRNA can be co-injected. In Xenopus oocytes, too, co-expression of Kv2.1 with Kv6.3 or Kv6.4 led to marked reduction in the rates of activation and inactivation (Fig. 8A), and both Kv2.1H 105R and Kv2.1H 105V showed marked changes in their rates of activation and inactivation (Fig. 8, B and C) compared with wt Kv2.1. Co-expression of the mutants Kv2.1 H105R (Fig. 8D)  The leftward shift of the activation curve after co-expression of Kv2.1 with Kv6.3 and the prominent rightward shift of the activation curve of Kv2.1 H105V were confirmed in the oocyte expression system (Fig. 8E). The mutant Kv2.1 H105R displayed a slight rightward shift of the activation curve (Fig. 8E). Co-expression of Kv2.1 H105R with Kv6.3 and co-expression of Kv2.1 H105V with Kv6.4 produced no further shift in the position of the activation curve.
Finally, we tested the role of histidine 105 in Kv2.1/Kv6.3 heteromerization by constructing dominant-negative mutants of Kv2.1 H105R and of Kv6.3. Loss of K ϩ permeability was induced by exchanging the two glycine residues in the K ϩ channel signature motif by alanines (changing the GYG motif to AYA). Expression of AYA-Kv2.1 H105R subunits in Xenopus oocytes produced no measurable current (Fig. 8F), indicating that channels with the AYA mutation are non-conducting. Coexpression of Kv2.1 H105R subunits with AYA-Kv2.1 H105R subunits produced a strong dominant-negative effect, i.e. the currents measured after co-injection of both cRNAs were reduced to 8% of control. These findings suggest that assembly of the subunits was not disturbed by the AYA mutation and that the presence of one AYA subunit was sufficient to render the channel non-functional. In contrast, co-expression of Kv2.1 H105R with AYA-Kv6.3 had no significant effect of current amplitude, confirming that the mutation of histidine 105 of Kv2.1 prevented heteromeric assembly of Kv2.1 with Kv6.3 subunits, but not homotetrameric assembly. , where I is the peak current, E is the potential, and E K is the K ϩ equilibrium potential (Ϫ90.0 mV). The conductance was normalized to the conductance at ϩ40 mV to give G/G max . cRNA coding for the subunits indicated was injected at equimolar amounts. For each combination, four activation curves were fitted with the Boltzmann function and the resulted half-maximal activation potentials are presented as means Ϯ S.E. F, normalized peak current amplitude (at ϩ40 mV) measured 46 -50 h after injection of cRNA coding for the indicated subunits. For the subunit combinations Kv2.1 H105R :AYA-Kv2.1 and Kv2.1 H105R : AYA-Kv6.3 the molar ratio of the co-injected cRNA was 1:2. The number of oocytes from which the data were obtained is given above the bars; three different batches of oocytes were used. n.s., not significant; ***, p Ͻ 0.001 by unpaired Student's t test.

DISCUSSION
We have studied the functional significance of the C3H1 motif for homomeric and heteromeric assembly of Kv2 and Kv6 subunits. We found that replacement of the histidine residue in the Zn 2ϩ ion-coordinating C3H1 motif of the T1 domain of Kv2.1 by arginine (Kv2.1 H105R ) or by valine (Kv2.1 H105V ) had the following effects: (i) In the yeast-two-hybrid assay, the T1 domain of Kv2.1 H105V failed to interact with the T1 domains of Kv6.3 and Kv6.4, (Fig. 4A). (ii) Kv6.4 was co-immunoprecipitated with wt Kv2.1 but not with Kv2.1 H105V (Fig. 5B). (iii) After co-expression of the mutants Kv2.1 H105R or Kv2.1 H105V with Kv6.3 or Kv6.4 FRET efficiencies were strongly reduced compared with wt Kv2.1 (Fig. 6C). (iv) Co-expression of wt Kv2.1 with Kv6.3 or Kv6.4 in a mammalian cell line had pronounced effects on the time course of activation, deactivation and inactivation and on the voltage-dependence of activation of Kv2.1, whereas co-expression of the mutants Kv2.1 H105R or Kv2.1 H105V with Kv6.3 or Kv6.4 did not modify the biophysical properties of the mutant channels at all (Fig. 7D). (v) These findings were confirmed by co-expression of the same mutants in Xenopus oocytes (Fig. 8). (vi) Co-expression experiments with dominant-negative mutants in Xenopus oocytes showed that the mutant Kv2.1 H105R subunits cannot co-assemble with Kv6.3 (Fig. 8F). Taken together, these findings suggest that mutation of the histidine residue in the Zn 2ϩ ion-coordinating C3H1 motif of the T1 domain of Kv2.1 disrupts the formation of functional heteromeric Kv2.1/Kv6 channels.
In the reverse experiment, introducing a histidine residue, which is part of the Zn 2ϩ -coordinating C3H1 motif in Kv2.1, at the corresponding position in Kv6.3 and Kv6.4 did not facilitate homomeric interaction of Kv6.x channels (supplemental Fig.  S2A). Thus the difference in the C3H1 motif cannot fully explain the lack of homomeric assembly of Kv6.3 and Kv6.4 channels. There must be additional structural determinants that promote or prevent tetramerization, for example other polar interactions in the T1 domain (12) or subunit interaction motifs in the transmembrane domains of Kv channels (7,14) or in the C termini (24,25).
Several previous studies have addressed the role of the T1 domain in the intracellular N terminus of Kv channels. The T1 domain has been shown to be involved in various processes such as subunit-specific assembly (8, 9, 16, 18, 19, 26 -30) in early biogenesis of the channels (6), in binding of cytosolic auxiliary subunits (31,32) and in the gating of channels inserted in the cell membrane (33)(34)(35)(36). In addition, it has been shown that the assembly of Kv subunits possessing C3H1 motifs is impaired when zinc is depleted (13,14). The results presented here were obtained in the presence of physiological zinc ion concentrations, indicating that all effects described here are not the result of zinc depletion.
Our data demonstrate that heteromerization changes the biophysical properties of the channels, for example, inactivation and deactivation kinetics and the voltage dependence of activation. The influence of heteromerization on deactivation kinetics has been discussed controversially in the literature (18,19). In agreement with Sano et al. (18), our data show a marked slowing of deactivation kinetics in Kv2.1/Kv6.3 heteromers. No change in the rate of deactivation of these heteromers was found by Ottschytsch et al. (19) (in their nomenclature Kv2.1/ Kv10.1). In contrast to both of these groups (18,19), we found that heteromerization gave rise to a distinct shift in the halfmaximal activation potential (V 1/2 ): with homomeric Kv2.1 channels V 1/2 was about Ϫ11 mV, with heteromeric Kv2.1/ Kv6.3 channels V 1/2 was Ϫ22 mV; with heteromeric Kv2.1/ Kv6.4 channels V 1/2 was around Ϫ6 mV, i.e. the shift was in the opposite direction. We determined these shifts in the voltage dependence of activation using two different approaches, (i) patch-clamp recordings in co-transfected in CHO-K1 cells (Fig.  3) and (ii) two-electrode voltage clamp in Xenopus oocytes coinjected with different cRNAs (Fig. 8); the two series of experiments gave very similar results. The discrepancy between our data and those of the two previous reports may be related to the fact that we used CHO-K1 cells and an IRES-based expression vector (where formation of homomeric channels was minimized) whereas the other groups used a fibroblast cell line stably overexpressing Kv2.1 (18) or simple co-transfection of two plasmids in murine fibroblast cells (19).
In conclusion, our results suggest that Kv2.1 subunits are capable of forming both homomeric Kv2.1 channels and heteromeric Kv2.1/Kv6.x channels, whereas Kv6.x subunits cannot assemble as homomeric channels. Mutation of the critical histidine residue in the T1 domain of Kv2.1 prevented formation of Kv2.1/Kv6.x heteromeric channels but not homomeric assembly of Kv2.1. Thus, the restriction in combinatorial possibilities may be partially attributable to the complementary properties of the T1 domains of Kv2.1 and Kv6.x. However, since the introduction of the histidine at the corresponding position of Kv6.3 or Kv6.4 was not sufficient to permit homomeric assembly of these subunits, other subunit interaction domains may also play a role in determining subunit composition of Kv channels.
The results of our quantitative gene expression analysis suggest that Kv2.1 subunits are ubiquitously expressed, in agreement with previous studies (37), whereas the expression pattern of Kv6 subunits is more restricted. Because co-expression of Kv2 and Kv6 subunits could be detected in several different tissues it is tempting to speculate that heterotetramerization of Kv2.1 and Kv6.x channels may be functionally relevant by allowing fine-tuning of the biophysical properties of Kv channels according to the actual cellular demand. This may be particularly important in central neurons, where Kv2.1 channels have been shown to cluster in the somatic and in the proximal dendritic membranes and to constitute the majority of the delayed rectifier current. This current plays a major role in the regulation of synaptic excitation in dendrites, especially during high frequency firing (38 -40) and during ischemia (41). In addition to modulating channel kinetics, heteromerization with Kv6.3 or Kv6.4 might also influence the trafficking of Kv2.1 from the endoplasmic reticulum to the surface membrane channels, their turnover rate at the surface membrane, and their interaction with accessory proteins (37,42).