Auxiliary KCNE subunits modulate both homotetrameric Kv2.1 and heterotetrameric Kv2.1/Kv6.4 channels

The diversity of the voltage-gated K+ (Kv) channel subfamily Kv2 is increased by interactions with auxiliary β-subunits and by assembly with members of the modulatory so-called silent Kv subfamilies (Kv5-Kv6 and Kv8-Kv9). However, it has not yet been investigated whether these two types of modulating subunits can associate within and modify a single channel complex simultaneously. Here, we demonstrate that the transmembrane β-subunit KCNE5 modifies the Kv2.1/Kv6.4 current extensively, whereas KCNE2 and KCNE4 only exert minor effects. Co-expression of KCNE5 with Kv2.1 and Kv6.4 did not alter the Kv2.1/Kv6.4 current density but modulated the biophysical properties significantly; KCNE5 accelerated the activation, slowed the deactivation and steepened the slope of the voltage-dependence of the Kv2.1/Kv6.4 inactivation by accelerating recovery of the closed-state inactivation. In contrast, KCNE5 reduced the current density ~2-fold without affecting the biophysical properties of Kv2.1 homotetramers. Co-localization of Kv2.1, Kv6.4 and KCNE5 was demonstrated with immunocytochemistry and formation of Kv2.1/Kv6.4/KCNE5 and Kv2.1/KCNE5 complexes was confirmed by Fluorescence Resonance Energy Transfer experiments performed in HEK293 cells. These results suggest that a triple complex consisting of Kv2.1, Kv6.4 and KCNE5 subunits can be formed. In vivo, formation of such tripartite Kv2.1/Kv6.4/KCNE5 channel complexes might contribute to tissue-specific fine-tuning of excitability.


Results
KCNE1, KCNE2, and KCNE3 moderately affect Kv2.1 and Kv2.1/Kv6.4 channels. Earlier studies provided evidence that KCNE1-3 interact with rat Kv2.1 subunits resulting in channels that possess altered biophysical properties and changed current densities 5,6 . However, these KCNE-induced effects depended on the KCNE species; e.g. rat KCNE1 slowed Kv2.1 activation ~2-fold and reduced the Kv2.1 current density, while human KCNE1 did only alter Kv2.1 activation significantly at + 60 mV and did not affect the Kv2.1 current density significantly 6 . Therefore, we first tested the effect of human KCNE1-3 on human Kv2.1 channels. Co-expression of Kv2.1 with KCNE1 or KCNE3 reduced the current density significantly and slowed Kv2.1 activation slightly (although significantly at + 60 mV) without changing the voltage-dependences of activation or inactivation or the deactivation kinetics. In case of KCNE2, only the current density was reduced (Suppl. Fig. S1, Suppl. Table S1).
In addition to their effect on channel gating and current density of Kv channels, KCNE subunits can also modify the pharmacological profile (for review, see 8 ). It has been demonstrated earlier that KCNE2 subunits altered the sensitivity of Kv2.1 to the common channel pore blocker tetraethylammonium (TEA) 5 . Hence, we determined the effect of TEA on Kv2.1 and Kv2.1/Kv6.4 in the absence and presence of KCNE5 (Fig. 2F) (Fig. 3). Figure 3A shows typical current recordings to determine the voltage-dependence of inactivation of Kv2. 1   slope of the inactivation curve depends on the ratio of initiation and recovery of the Kv6.4-induced closed-state inactivation. Therefore, we determined both the initiation and recovery rate constant of Kv2.1/Kv6.4 closed-state inactivation in the absence and presence of KCNE5 (Fig. 4). The recovery rate was determined using the protocol illustrated in Fig. 4A: an initial control pulse to + 60 mV (P1) was used to record the initial current amplitude, followed by a 1 s step to − 130 mV to recover all channels from inactivation, then a 10 s pulse to − 60 mV to induce a certain degree of closed-state inactivation, a pulse of variable duration to − 90 mV allowing the channels to recover from inactivation and a second test pulse to + 60 mV (P2). The fraction P2/P1 represents the degree of channels that have recovered from their closed-inactivated state and was plotted as function of time spent at the recovery pulse to − 90 mV ( Fig. 4B). Kv2.1/Kv6.4 channels recovered with a time constant of 2.9 ± 0.5 s (n = 8) and KCNE5 significantly modulated this process, increasing the rate of recovery to 1.2 ± 0.2 s (Fig. 4C). In addition, the rate of initiation of Kv2.1/Kv6.4 closed-state inactivation and the extent of inactivation was determined in the absence and presence of KCNE5 but no significant differences could be observed (data not shown). Next, we determined the impact of different levels of KCNE5 expression on Kv2.1 and Kv2.1/Kv6.4 channels (Fig. 5). We co-transfected Kv2.1 or Kv2.1/Kv6.4 with KCNE5 in different ratios and determined the impact of KCNE5 focusing on those properties that we had seen to be mostly influenced by KCNE5 co-expression, i.e. Kv2.1 current density (Fig. 5A,B), and Kv2.1/Kv6.4 activation rate (Fig. 5C) and voltage-dependence of inactivation (Fig. 5D). Co-expression of Kv2.1 with KCNE5 in a 1:0.5, 1:1, 1:4 and 1:8 (Kv2.1:KCNE5) cDNA ratio reduced Kv2.1 current densities gradually. Although Kv2.1 current densities also decreased with increasing amounts of transfected peCFP vector alone (used to correct for effects of cDNA dilution and/or overload of the transcriptional/translational machinery), we observed a clear KCNE5-mediated decrease in Kv2.1 current density (Fig. 5A,B). These results demonstrated that at least a part of the Kv2.1 current reduction observed upon co-expression with KCNE5 is caused by KCNE5 and this KCNE5-induced current reduction depends on the KCNE5 expression level. Similarly, co-expression of Kv2.1/Kv6.    (Fig. 6). We performed immunocytochemical experiments and found Kv2.1 subunits in small clusters at the cell surface when expressed alone in line with previous studies (Fig. 6C) 18 . In addition, Kv6.4 subunits were retained intracellularly (Fig. 6B) which corresponds to the previously demonstrated ER retention of Kv6.4 subunits when expressed alone 14 . As described earlier, this ER retention is relieved upon co-expression with Kv2.1 subunits resulting in Kv2.1/Kv6.4 heterotetramers at the plasma membrane 14,19 which we also observed in our immunocytochemical experiments (Fig. 6F). To examine whether KCNE5 interacts with Kv2.1 and Kv6.4 in (tripartite) channels, we first deduced the KCNE5 localization pattern when expressed alone in HEK293 cells using a custom antibody directed against KCNE5 (Suppl. Material and Methods, Suppl. Fig. S3). We found that KCNE5 was located mainly in small vesicles or clusters which appeared to be at the plasma membrane (Fig. 6A). Performing surface stainings using a N-terminally HA-tagged KCNE5 construct, we found that KCNE5 was able to traffic to the cell membrane on its own and was located in small clusters (Suppl. Fig. S4, arrows) as seen for Kv2.1 (Fig. 6C). From the whole cell stainings it was noticeable that KCNE5 was also located in some vesicles beneath the membrane. To determine the origin of these vesicles/clusters we employed different compartmental markers (Suppl. Fig. S5). These vesicles were not part of the ER, Golgi Apparatus, late endosomes/prelysosomes or early endosomes (Suppl. Fig. S5A-C, E, respectively), yet some were found in vesicles positive for the transferrin receptor, a marker for recycling endosomes. (Suppl. Fig. S5D). Co-expression of KCNE5 and Kv2.1 revealed that the channel subunits located in the same clusters (Fig. 6E). In contrast to Kv2.1, KCNE5 was incapable of rescuing Kv6.4 from the ER (Fig. 6D), yet when all three subunits were co-expressed in the same cell they co-localized at the cell surface (Fig. 6G). To ensure that this observed co-localization was originating from the formed Kv2.1/ Kv6.4/KCNE5 channel complexes (rather than from non-specific interactions caused by the presence of the HA tag), we confirmed electrophysiologically that the channel properties were not affected by the HA tag (Suppl. Fig. S6).
In summary, these experiments demonstrate that KCNE5 is able to traffic to the membrane where it co-localizes with both homotetrameric Kv2.1 and heterotetrameric Kv2.1/Kv6.4 channels. Hence, this    (Fig. 7). The FRET efficiency of CFP-Kv6.4 + YFP-Kv2.1 (positive control) and CFP + YFP (negative control) combinations yielded FRET efficiencies of 15.7 ± 1.1% and 1.8 ± 1.0%, respectively. Upon co-expression of KCNE5-YFP with CFP-Kv2.1, we obtained a FRET efficiency of 9.7 ± 0.9% which is lower than the FRET efficiency of the positive control but significantly higher than the FRET efficiency of the negative CFP+ YFP combination. This indicated that KCNE5 physically associates with Kv2.1 subunits. Co-expression of KCNE5-YFP with CFP-Kv6.4 yielded a FRET efficiency of 2.4 ± 1.0%. This FRET efficiency was similar to the negative control indicating that KCNE5 does not associate with Kv6.4 subunits alone. These results suggest that KCNE5 subunits modulate the Kv2.1/Kv6.4 heterotetramers via association with Kv2.1 subunits. To investigate this further, we performed FRET experiments co-expressing unlabeled Kv2.1 subunits; an increased FRET efficiency between CFP-Kv6.4 and KCNE5-YFP subunits upon co-expression with unlabeled Kv2.1 would indicate that Kv6.4 and KCNE5 are in each other's proximity caused by an association with the Kv2.1 subunits into a tripartite complex. Indeed, co-expression of CFP-Kv6.4 with KCNE5-YFP and unlabeled Kv2.1 yielded a FRET efficiency of 6.0 ± 0.9% which was significantly higher than the FRET efficiency obtained with the CFP-Kv6.4 + KCNE5-YFP combination. In addition, this FRET efficiency was similar to the FRET efficiency between CFP-Kv2.1 and KCNE5-YFP upon co-expression with unlabeled Kv6.4 subunits (5.9 ± 1.7%). To ensure that these observed FRET efficiencies were originating from the formed (tripartite) channel complexes (rather than from non-specific interactions caused by the presence of the different tags), we confirmed that the present tags do not affect the biophysical properties of the channels (Suppl. Fig. S6).

Discussion
Kv2.1 channel activity has been shown crucial for neuronal excitability and neuroprotection (for review see 20 ). To accommodate a variety of functions, the Kv2.1 channel diversity is increased through different mechanisms: i) heterotetramerization with modulatory α -subunits of the Kv5, Kv6, Kv8 and Kv9 subfamilies (the so-called silent KvS subunits, for review see 2 ), ii) post-translational modifications such as (de)phosphorylation (for review see 21 ) and SUMOylation 22 , and iii) association with auxiliary β -subunits such as KChAP 3 , AMIGO 4 and KCNE proteins (for review, see 8 ). Kv2.1 is one of the many channels influenced by KCNE co-assembly and previous studies have indicated a role for KCNE1-3 modulation 5,6 . These studies revealed that human KCNE2 and KCNE3 significantly slowed the activation of rat Kv2.1 currents at + 60 mV while KCNE1 had a smaller, although still significant, effect 5,6 . Our results reveal that human KCNE1 and KCNE3 only slightly slow human Kv2.1 activation at + 60 mV (Suppl. Fig. 1). Furthermore, human KCNE2 had no significant effects on human Kv2.1 currents (Suppl. Fig. 1). This discrepancy compared to earlier reports might be explained by the difference in Kv2.1 species. It has been shown that rat Kv2.1 activates faster than human Kv2.1 23 creating the possibility that the effect of KCNE1-3 subunits on rat Kv2.1 seems larger but that the difference in effect mainly comes from the differences at basal levels. Indeed, the activation kinetics of rat Kv2.1 after modulation by human KCNE1, KCNE2 and KCNE3 subunits are similar to the human Kv2.1 activation kinetics upon co-expression with KCNE1-3 (~15 ms) 5,6 .
Co-expression of both KCNE4 and KCNE5 reduced Kv2.1 current densities significantly without affecting the channel gating and inactivation properties ( Fig. 1 and Fig. 2, respectively). These results could be explained mainly by modulation of the trafficking and/or targeting of Kv2.1 channels which could occur via several possible mechanisms. For example, KCNE4-5 subunits may suppress the forward trafficking like the demonstrated effect of KCNE1 and KCNE2 on the forward trafficking of Kv1.4, Kv3.3 and Kv3.4 24 . However, we detected KCNE5 subunits in recycling endosomes (Suppl. Fig. S5) opening up for the possibility that KCNE5 reduces the Kv2.1 current density by increasing the endocytosis of Kv2.1 channels as has been suggested for KChIP subunits and Kv4 channels 25 . KCNE4-5 subunits may also decrease the Kv2.1 current density by favoring the localization of Kv2.1 channels into clusters; we detected KCNE5 and Kv2.1 in the same clusters at the plasma membrane (Fig. 6E). It has been demonstrated that Kv2.1 channels that reside within membrane clusters are in a non-conducting state while those outside the clusters are in an conducting state 26 . Another possibility is that KCNE4-5 subunits reduce Kv2.1 current density by enhancing the Kv2.1 forward trafficking (resulting in an increased Kv2.1 surface density) as it has been demonstrated that even the Kv2.1 channels outside the typical Kv2.1 membrane clusters may cease conducting when the Kv2.1 surface current density increases 27 .
It has previously been suggested that different auxiliary β -subunits can associate simultaneously within Kv channel complexes. For example, the simultaneous presence of KChAP and Kvβ 1.2 proteins disturbs each other's effect on Kv1.4, Kv1.5, Kv2.1 and Kv4.3 channels 11 while KCNE1 (that increases the Kv7.1 current density) and KCNE4 (that decreases the Kv7.1 current density) appear to co-associate with Kv7.1 channels into a "triple subunit" channel complex in which KCNE4 suppresses the KCNE1 effect 12,28 . Our data showed that auxiliary β -subunits (i.e. KCNE5) and modulatory α -subunits (i.  (Fig. 2). Co-expression of KCNE5 only modified activation, deactivation and inactivation properties of Kv2.1/Kv6.4 heterotetramers significantly (and not those of Kv2.1 homotetramers) which may suggest that the KCNE5 effect requires the presence of the modulatory Kv6.4 α -subunit (or perhaps other KvS α -subunits). This raises the possibility that β -subunits that seemingly have no effect on channel gating and channel inactivation in heterologous expression systems might affect these channel properties in vivo when in complex with other (unknown) modulatory α -subunits.
Using FRET experiments, we demonstrated that KCNE5 interacts with Kv2.1 but not with Kv6.4 subunits when co-transfected with either subunit into HEK293 cells. However, FRET occurred between CFP-Kv6.4 and KCNE5-YFP in the presence of unlabeled Kv2.1 (Fig. 7). We envision (at least) three possible mechanisms: i) assembly with Kv2. Kv2.1 channels are ubiquitously expressed and play many diverse roles in both excitable and non-excitable cells. For example, Kv2.1 channels regulate the action potential in hippocampal neurons, especially during high-frequency stimulation 29 , they contribute to both the rapidly activating, slowly inactivating current (I Ks ) and the non-inactivating, steady-state current (I ss ) in mouse atrial myocytes 30 and they play a central role in the regulation of the pancreatic β -cell membrane potential 31 . Compared to ubiquitously expressed Kv2.1 channels, KvS subunits display a more tissue-specific expression creating tissue-specific functions for these Kv2/KvS heterotetramers. Indeed, mutations in Kv8.2 have been associated with an inherited retinal dystrophy 32 and epilepsy susceptibility 33 16,28,35,36 and Kv2.1, KCNE5 and Kv6.4 are all expressed in human brain [14][15][16] . This creates the possibility that KCNE5 subunits also affect Kv2.1 homotetramers and Kv2.1/Kv6.4 heterotetramers in vivo, resulting in more tissue-specific fine-tuning mechanisms.

Material and Methods
Molecular Biology. Human Kv2.1 (GenBank Accession number NM_004975) and human Kv6.4 (NM_172347) in the eGFP-N1 vector (Clontech, Palo Alto, CA, USA) and N-terminally CFP-labeled Kv2.1 and Kv6.4 subunits in the eCFP-C1 vector (Clontech) have been described previously 14,17 . Mouse KCNE4 (NM_021342) was amplified from mouse genomic DNA and subcloned in the pBK vector using SmaI. Human KCNE5 (NM_012282) was amplified from human genomic DNA and subcloned into either the pBK vector using SalI and BamHI or into the pXOOM vector using BamHI and EcoRI. C-terminally YFP-labeled KCNE5 was obtained by subcloning the KCNE5 cDNA in the peYFP-N1 (Clontech) vector. HA-Kv6.4 and HA-KCNE5 constructs were generated by introducing the HA epitope (YPYDVPDYA) into the extracellular S1-S2 loop and the extracellular N-terminus, respectively, by PCR amplification using the QuikChange TM Site-Directed Mutagenesis kit (Stratagene, La Jolla, CA, USA) and mutant primers. All PCR-generated inserts were sequenced to confirm sequence integrity and in-frame clonings.
Cell culture and transfection. HEK293 cells were cultured in Modified Eagle's medium (Invitrogen, San Diego, CA, USA) supplemented with 10% fetal bovine serum, 1% non-essential amino acids and 1% penicillin-streptomycin at 37 °C and 5% CO 2 . Cells were co-transfected with the subunit cDNAs and GFP as a transfection marker using Lipofectamine2000 (Invitrogen, San Diego, CA, USA). The transfections were performed in 60 mm cell culture dishes filled with 4 ml culture medium for the electrophysiological and Fluorescence Resonance Energy Transfer (FRET) experiments and in 35 mm cell culture dishes filled with 2 ml culture medium for the immunocytochemical experiments. 16-24 h after transfection, cells were dissociated with trypsin and used for electrophysiological analysis within 5 h or fixated on coverslips using 4% cold paraformaldehyde for 10 min for immunocytochemistry.
Electrophysiology. Electrophysiological recordings were performed as previously described 19 .
Briefly, whole cell current recordings were performed at 20-22 °C using an Axopatch-200B amplifier (Axon Instruments, Union City, CA, USA) connected to a Digidata 1440 data acquisition system (Axon Instruments) and were low-pass filtered and sampled at 1-10 kHz. Command voltages and data storage were controlled with the pClamp10 software (Axon Instruments). Patch pipettes with a resistance of 1.5-2.5 MΩ were filled with an intracellular solution containing (in mM): 110 KCl, 5 K 4 BAPTA, 5 K 2 ATP, 1 MgCl 2 and 10 HEPES, pH adjusted to 7.2 with KOH. Cells were continuously perfused with an extracellular solution containing (in mM): 145 NaCl, 4 KCl, 1.8 CaCl 2 , 1 MgCl 2 , 10 HEPES, 10 glucose, pH adjusted to 7.35 with NaOH. Cells were excluded from analysis if voltage errors at the highest potentials used (i.e. 60 mV and 70 mV to determine the inactivation and (de)activation properties, respectively) did exceed 5 mV after series resistance compensation.
Pulse protocols and data analysis. The Boltzmann equation: V the voltage applied, V 1/2 the voltage at which 50% of the channels are (in)activated and k the slope factor was fitted to the voltage-dependence of activation and inactivation. Single or double exponential function were fitted to the (de)activation kinetics. The Hill equation: 1 -y = 1/(1-(IC 50 /[D] n H ) with IC 50 the concentration that generates 50% current inhibition, [D] the drug concentration and n H the Hill coefficient was fitted to the concentration-effect curves. Statistical analysis was performed using Student's t test or Mann-Whitney U Rank Sum test.

Fluorescence Resonance Energy Transfer (FRET).
FRET experiments were performed as previously described 17 . Briefly, HEK293 cells were cultured on coverslips and co-transfected with YFP (acceptor fluorophore) and CFP (donor fluorophore) labeled subunits. 48 h after transfection, FRET was determined using a Zeiss CLSM 510 microscope equipped with an argon laser to visualize and bleach CFP (with excitation at 458 nm) and to bleach YFP (with excitation at 514 nm). FRET efficiencies were determined with following equation: FRET efficiency = (1-(f DA -f background )/(f D -f background ))x(1/pairedDA) in which f DA and f D represent the donor fluorescence in the presence and absence of the acceptor, respectively, f background the background fluorescence and pairedDA the fraction of paired donor and acceptor fluorophores. F DA , f D and f background were obtained by recording the CFP emission in the 464-490 nm bandwidth without YFP bleaching, after YFP bleaching by a 30 s full power excitation at 514 nm and after CFP bleaching by a 30 s full power excitation at 458 nm, respectively. PairedDA was assumed to be 1 which was pursued by i) transfecting the CFP-and YFP-tagged subunits in a 1:2 cDNA ratio and ii) maintaining a YFP/CFP fluorescence intensity ratio <1 that was determined from the intensities in the YFP and CFP bandwidth (532-554 nm and 464-490 nm, respectively) before YFP and CFP bleaching. A pairedDA <1 will result in an underestimation of the FRET efficiency and cells with an YFP/CFP intensity ratio <1 were excluded from FRET efficiency analysis. Because neither CFP nor YFP carried the mutation A206K implicating that they have the weak tendency to dimerize, the FRET efficiency might be slightly under or overestimated.
Immunocytochemistry. Fixated cells first underwent a combined permeabilization and blocking step for 30 min at room temperature using 0.1% Triton X-100 and 0.2% fish skin gelatin in 1xPhosphate Buffered Saline (PBS) (immunobuffer). Secondly, primary antibodies diluted in immunobuffer were applied for 1 h followed by 3 washing steps, 5 min each, using 0.1% Triton X-100 in PBS. Alexa-Fluor conjugated secondary antibodies (Invitrogen) were diluted in immunobuffer and applied for 45 min before coverslips were mounted in Prolong Gold Antifade reagent (Invitrogen).
Confocal microscopy and imaging. All images were acquired using the laser scanning confocal microscope Zeiss LSM780 equipped with a 63x/NA = 1.40 oil objective. The pinhole diameter was set between 0.7-1.0 μ m equaling 1 airy unit. Images were obtained using sequential scanning separating the individual channels and noise was reduced employing line averaging. Finally, images were treated with ZEN 2011 edition and Adobe Illustrator and Photoshop CS6.