The C-terminal HRET sequence of Kv1.3 regulates gating rather than targeting of Kv1.3 to the plasma membrane

Kv1.3 channels are expressed in several cell types including immune cells, such as T lymphocytes. The targeting of Kv1.3 to the plasma membrane is essential for T cell clonal expansion and assumed to be guided by the C-terminus of the channel. Using two point mutants of Kv1.3 with remarkably different features compared to the wild-type Kv1.3 (A413V and H399K having fast inactivation kinetics and tetraethylammonium-insensitivity, respectively) we showed that both Kv1.3 channel variants target to the membrane when the C-terminus was truncated right after the conserved HRET sequence and produce currents identical to those with a full-length C-terminus. The truncation before the HRET sequence (NOHRET channels) resulted in reduced membrane-targeting but non-functional phenotypes. NOHRET channels did not display gating currents, and coexpression with wild-type Kv1.3 did not rescue the NOHRET-A413V phenotype, no heteromeric current was observed. Interestingly, mutants of wild-type Kv1.3 lacking HRET(E) (deletion) or substituted with five alanines for the HRET(E) motif expressed current indistinguishable from the wild-type. These results demonstrate that the C-terminal region of Kv1.3 immediately proximal to the S6 helix is required for the activation gating and conduction, whereas the presence of the distal region of the C-terminus is not exclusively required for trafficking of Kv1.3 to the plasma membrane.


Results
Strategic considerations for designing the Kv1.3 constructs. Three strategies were applied for unique identification of the Kv1.3 subunits transfected into HEK or CHO cells. First, we introduced the A413V mutation in the S6, which has been shown to accelerate dramatically the inactivation kinetics of A413V homomers as compared to wild-type channels (τ A413V = 4 ms, τ WT = 200 ms, Fig. 1A and C 22,23 ). As it was shown earlier, heterotetramers consisting of A413V and the WT subunits display intermediate inactivation kinetics between τ A413V and τ WT depending on the number of mutant subunits in the tetrameric channel 22,24,25 (also see Suppl. Figure 1). Consequently, the presence of heteromeric channels can be easily tracked in the membrane by fitting the inactivation kinetics of the K + current 23 .
Second, we substituted histidine (H) for lysine (K) at residue 399 which resulted in a TEA (tetraethylammonium)-insensitive, phenotype, as described earlier ( Fig. 1A and C) [26][27][28] . The assembly of TEA sensitive (K d ≈11.4 mM) and insensitive (K d ≈2000 mM) Kv1.3 subunits modifies the affinity of the heterotetramer for TEA, hence the heteromultimer formation can be easily identified by the application of 100 mM TEA 26 .
Third, for immunocytochemistry experiments we used constructs that had the FLAG insert in the extracellular loop between the S1 and S2 segments. The insertion of the FLAG epitope does not alter dramatically the properties of Kv1.3 as shown earlier but allows labeling of the channels expressed in the plasma membrane.
These mutations were combined with various truncations/mutations in the C-terminus and fluorescence protein tagging on the N-terminus: EGFP for all constructs except the WT channel in the co-expression experiments, where mCherry was used. Truncations at the C-terminus were generated by introducing a stop codon after (constructs denoted with ΔC) or before the "HRET" sequence (labeled with NOHRET ending) (Fig. 1B).

Deletion of the C-terminal region does not reduce current of ΔC channels. Several studies
reported that the removal of the C-terminus of Kv channels drastically influences their cell membrane expression, and that the "HRET" sequence is crucial for anterograde trafficking 12 . To test this scenario for Kv1.3 we used the A413V-ΔC and A413V-NOHRET constructs (Fig. 1B) and expressed them in HEK and CHO cell lines. The ΔC nomenclature in this paper refers to a construct that is truncated right after the HRET motif (Δ440-523) whereas NOHRET indicates the construct lacks the entire C-terminus including the HRET motif as well (Δ436-523). Figure 2A shows that expression of the A413V-ΔC construct results in a robust whole-cell current having fast inactivation kinetics, as it was seen for full-length homotetrameric A413V (Fig. 1C). On the contrary, HEK transfected under identical conditions with the A413V-NOHRET construct did not show any voltage-and time-dependent current. Comparison of the current density (CD, peak current at + 50 mV divided by the whole-cell capacitance) for A413V-FL, A413V-ΔC and A413V-NOHRET in HEK cells (Fig. 2C) showed that the ΔC-truncated construct had a slightly higher median CD than the full-length one although the difference was not statistically significant (p = 0.053, CD FL = 0.26 ± 0.05 nA/pF (n = 13, A413V-FL), CD ΔC = 0.71 ± 0.15 nA/ pF (n = 15, A413V-ΔC)) whereas the CD for the A413V-NOHRET is CD NOHRET = 0.016 ± 0.007 nA/pF (n = 4), p < 0.05. To rule out the influence of the expression system on our results (e.g., the presence of endogenous Kv1.x channels/subunits in HEK, which may facilitate the forward trafficking of heteromers of WT and A413V-ΔC mutant subunits to the membrane, or cell-specific protein sorting and trafficking) the whole set of experiments was repeated in CHO cells 29,30 . CHO cells do not show measurable whole-cell outward current at + 50 mV (Suppl. Figure 2) and thus, are suitable for our experiments. The results in CHO were qualitatively similar to the ones obtained in HEK, the A413V-FL and the A413V-ΔC constructs gave rise to A413V homotetrameric currents of similar density whereas practically zero current could be recorded following the transfection with SCIentIFIC REPORTS | (2018) 8:5937 | DOI:10.1038/s41598-018-24159-8 the A413V-NOHRET construct (Fig. 2D, CD FL = 0.36 ± 0.1 nA/pF (n = 5, A413V-FL), CD ΔC = 0.32 ± 0.09 nA/pF (n = 5, A413V-ΔC), p = 0.748, CD NOHRET = 0.008 ± 0.002 nA/pF (n = 10), p < 0.05). Due to the lack of whole-cell currents in untransfected CHO we used these cells in our subsequent experiments. 3 s starting at amino acid 382 ("……" represents rest of the C-terminus not relevant to this study). Amino acids are listed by their one letter designations, FL: full-length C-terminus, ΔC: C-terminal truncated at 439, NOHRET: truncation after postion 435. XHRETE: deletion mutant removing HRETE sequence. polyA: replacement HRETE with AAAAA, Atail: HRETE changed for AAAAA, and rest of C-terminus is removed. Starting amino acid of C-terminus 427 is indicated. (C) Current traces of WT (mCherry tagged, left), A413V-FL (center) and H399K-FL (right) channels expressed in HEK cells. WT (outside-out patch configuration): cells were depolarized to + 50 mV from a holding potential of −120 mV for 2 seconds. A413V-FL (whole-cell configuration): cells were kept at −120 mV then depolarized to + 50 mV for 30 ms. H399K (whole-cell configuration): cells were depolarized to + 50 mV from −120 mV for 2 seconds. Inset: whole-cell current of a cell expressing H399K-FL and depolarized to 0 mV from a holding potential of −120 mV for 200 ms in the presence and absence of 100 mM TEA. We also tested whether all these channel constructs are properly translated by the cells using western-blot against a FLAG epitope inserted between the S1-S2 helical segments of all the constructs (Fig. 1A). Suppl. Figure 3 illustrates that CHO cells transfected with any of the three FLAG-tagged constructs express EGFP-tagged Kv1.3 channel subunit of appropriate size (between 85-95 kDa).
Pharmacology reveals homomers of H399K -ΔC channels in the membrane. To rule out mutationspecific effects (e.g. A413V) on the expression of the K + current, we used an alternative, pharmacological approach to learn if carboxyl-terminal deleted channels can reach the plasma membrane. TEA, a general inhibitor of various K + channels, is a fast open-channel blocker and its affinity depends on an aromatic amino-acid side chain in the extracellular mouth of the pore region 27,31 . Previously we showed that in Kv1.3 the mutation of the TEA-binding residue (H399) to a positively charged amino acid (e.g. lysine, K) led to a completely TEA-insensitive channel 26 . Figure 3A shows currents of the FLAG-H399K-FL (full-length subunits with FLAG tag) mutant in CHO cells in control extracellular solution and in the presence of 100 mM TEA, the overlapping currents indicate the lack of inhibition (RF = 0.95 ± 0.01, n = 5, see Materials and Methods). When CHO cells were transfected with the FLAG-H399K-ΔC plasmid we measured currents similar to the full-length H399K construct ( Fig. 1C and Suppl. Figure 4A) and also displayed TEA resistance (Fig. 3B, RF = 0.98 ± 0.01, n = 5). The C-terminal deletion that includes the "HRET" motif as well (FLAG-H399K-NOHRET) resulted in a phenotype that lacks outward K + current, just as we described for the A413V mutant (Suppl. Figure 4B).
In summary, using either kinetically tagged (A413V) or pharmacologically tagged (H399K) C-terminally truncated Kv1.3 we demonstrated that C-terminal truncated subunits (A413V-ΔC or H399K-ΔC) form functional homotetrameric channels in the membrane regardless of the expression system. However, we were unable to record whole-cell currents when the C-terminal truncated constructs that lack the "HRET" motif as well (i.e. A413V-NOHRET, H399K-NOHRET) were expressed either in HEK or CHO. This might mean impaired trafficking and/or conductivity/functionality of these latter constructs.
NOHRET constructs can also target to the cell membrane. Next, we addressed if C-terminal truncated constructs that also lack the "HRET" motif can be expressed in the cell membrane. The Kv1.3 subunits were conjugated to EGFP on the intracellular N-terminus and the extracellular FLAG epitope was cloned into each plasmid encoding point and deletion mutants of the Kv1.3 channel (FLAG-H399K-FL, FLAG-H399K-ΔC, FLAG-H399K-NOHRET and FLAG-A413V-FL, FLAG-A413V-ΔC, FLAG-A413V-NOHRET constructs). The distribution of the channel subunits was detected by means of confocal fluorescence microscopy. Figure 4A shows that FLAG-tagged full-length, H399K and A413V mutant Kv1.3 channels were stained with anti-FLAG antibody in both CHO and HEK cells (only CHO is shown, for HEK see Suppl. Figure 5). The same was observed for both Kv1.3 point mutants that lacked either the complete C-terminus (ΔC, Fig. 4B) or the C-terminus plus the "HRET" motif (NOHRET, Fig. 4C). This confirms the presence of all of these constructs in the plasma membrane, though the intracellular retention of truncated channels was elevated as compared to the full-length counterpart as shown by the quantitative analysis of the images (Suppl. Figure 6). Moreover, no difference in membrane expression between the ΔC and the NOHRET constructs was detected for both A413V and H399K (Suppl. Figure 6). All these data support the hypothesis that Kv1.3 channels can reach the membrane even without the "HRET" motif and instead the gating/conductance of the channel is impaired.

Co-expression of wild-type and A413V-NOHRET subunits results in pure wild-type Kv1.3 current.
We co-transfected mCherry-tagged full-length, wild-type Kv1.3 channels along with EGFP-A413V-NOHRET plasmid mutants at 1:1 ratio and studied if the presence of the WT subunits can rescue conduction of the channels containing A413V-NOHRET subunits. Assuming that tetrameric channels assemble randomly from individual subunits a 1:1 co-transfection ratio should result in predominantly (87.5%) heteromeric channels having intermediate inactivation kinetics between WT and A413V 22 . On the contrary, we found that a CHO cell expressing both A413V-NOHRET and WT-Kv1.3 exhibits currents resembling the pure WT current (Fig. 5A); and not the "mixture" of multiple heteromeric currents characterized with various inactivation time constants (also see current trace of A413-FL and WT Kv1.3 in Fig. 1C and Suppl. Figure 1B). The current density of only WT-Kv1.3 channel expressing cells (3390 ± 1110 pA/pF) was far higher than those transfected with both WT-Kv1.3 and A413V-NOHRET constructs (1010 ± 311 pA/pF, p = 0.028). Moreover, we transfected CHO cells with a mixture of FLAG-tagged A413V-NOHRET (EGFP-conjugated) and WT-Kv1.3 (mCherry conjugated) and subjected them to anti-FLAG labeling as described in Fig. 4. As displayed in Fig. 5B the FLAG-epitope bearing A413V-NOHRET subunits are labelled by the anti-FLAG antibody, which indicates their membrane targeting.

Gating charge movement of NOHRET channels is absent.
To disclose if the conducting pathway or the activation gating is destroyed upon HRET removal in the NOHRET Kv1.3 we assessed the gating properties of WT-NOHRET construct expressed in CHO cells (see Fig. 1B). As a positive control, we expressed the WT-W384F channel, which is a non-conducting mutant of Kv1.3 (homologous to the non-conducting W434F mutant of the Shaker channel [32][33][34][35][36][37]. Figure 6A displays the gating currents recorded in a CHO cells stably expressing Kv1.3-W384F (we recorded gating currents in all 11 cells). The representative Q on -V curve for this cell in the Fig. 6B illustrates the sigmoid shape of membrane potential dependence of the integrated gating current, which is a hallmark of voltage-gated ion channels and point out the functionality of the voltage-sensor. When we measured the gating current in cells expressing WT-NOHRET channels no gating current was detected (n = 9, Fig. 6B and C) or a miniature gating current was detected at very depolarized test potentials of +50 mV or higher (n = 2, not shown). These indicate that voltage-sensor movement of the channel is compromised when HRET is not present.
Deletion and substitution of the HRET(E) sequence does not affect the function of Kv1.3. Motivated by these findings, mainly those for gating properties of NOHRET constructs, the following mutations were introduced into EGFP-tagged WT Kv1.3: 1) the extended HRET motif, HRET(E) region of the C-tail was deleted only (WT-XHRETE construct); 2) HRET(E) was replaced with a run of five alanines (WT-polyA channel) (see Fig. 1B, bottom sequences) and 3) the HRETE sequence was replaced with 5 alanines in the WT-∆C (WT-Atail channel). After transfecting these mutants and the WT full-length in CHOs we analyzed basic biophysical features for all channel types in outside-out patch configuration. As shown in Fig. 7, all three HRETE-manipulated subunits formed functional and conducting tetramers in the CHO cell membrane, and no major differences in the kinetic and equilibrium parameters of the gating were found. The activation kinetics was a bit slower for the  WT-XHRETE and WT-Atail constructs as compared to WT-FL (p < 0.001 for both), but not the inactivation kinetics (p = 0.13). The half maximal voltage (V ½ ) of the steady-state activation was only different for WT-Atail channels (p = 0.005, leftward shift) but the slope factor (k) was the same for all HRETE-mutants (WT-polyA, WT-Atail and WT-XHRETE, p = 0.284). Furthermore, we assessed the single-channel conductance for all four phenotypes and obtained that removal/replacement of HRETE-motif in Kv1.3 channels did not influence the unitary conductance (Suppl. Figure 7, p = 0.085). Moreover, FLAG-epitope bearing WT-Atail channels could be detected on the cell surface (Suppl. Figure 8). All these data clearly demonstrate that the HRETE motif is not vital for the operation of Kv1.3: the channel is present in the plasma membrane, and it is functional in the absence of this motif.

Discussion
Membrane targeting of Kv channels has been studied by multiple groups, as it is a key player in the regulation of various cellular processes including action potential regulation or immune response 38 . To draw appropriate conclusion for trafficking it is critically important to demonstrate unambiguously that the current recorded is a consequence of the ion channel genes transfected into the cells and not influenced by endogenous K + channel subunits in the expression system. To do this we used kinetically or pharmacologically tagged Kv1.3 subunits with properties that uniquely distinguish homotetrameric channels formed by transfected subunits from endogenous K + channels or hetermultimers of endogenous and transfected channel subunits. If the A413V-ΔC subunits reach the plasma membrane because they form heterotetramers with endogenous WT Kv1.3 in HEK cells, then the inactivation kinetics of the whole-cell current should be fitted with sum of multiple exponential decay functions (see Suppl. Figure 1). On the contrary, we obtained that τ i for A413V-FL (4.1 ± 0.6 ms) was the same as that of the A413V-ΔC (5.7 ± 0.5 ms, p = 0.15), and the decaying part of the curves could be fitted with a single-exponential function (data not shown). The expression of A413V-ΔC current in CHO cells, which do not exhibit any voltage-gated K + conductance, supports the scenario as well that ΔC-truncated mutants are targeted to the plasma membrane without their combination with full-length Kv1.x subunits.
The outcomes with the H399K mutant also support the hypothesis that C-terminus removal downstream of the "HRET" region does not prevent targeting of Kv1.3 to the cell membrane. For H399K-ΔC and H399K-FL the inhibition by 100 mM TEA was negligible, which verifies that there was no mixing of H399K-ΔC and WT Kv1.3 subunits in the ER (we used CHO cells here as well, which have no Kv1 subunits), otherwise we should have seen block by TEA 27,31 . In addition, we detected the FLAG epitopes of both A413V-ΔC and H399K-ΔC subunits in transfected cells (CHO and HEK) with immunocytochemistry in non-permeabilized cells, which serves as an additional proof for their plasma membrane localization ( Fig. 4 and Suppl. Figure 5) Interestingly, we also found that A413V-NOHRET or H399K-NOHRET bearing the FLAG-epitope showed a membrane signal upon anti-FLAG staining, although patch-clamp experiments revealed that these deletion mutants produced no current.
Our results somewhat contradict the findings of previous studies which emphasized the importance of the C-terminal region downstream of the "HRET" motif in Kv1.3, Kv1.2 and Kv1.1 in regulating trafficking 14,15 . The importance of a diacidic signal (E483/484) in the C-terminal region of Kv1.3 was suggested to control surface expression via interaction with Sec24 (a coat protein of vesicular transport) 14 . Another study demonstrated that alanine-scanning mutational analysis of glutamates at residues 438/440/442 (in our construct) slightly increased ER retention only for triple mutants, and Sec24a and Sec24b knock-down inhibited from-ER-to-Golgi transport 15 . Furthermore, the removal of a much shorter fraction of the C-terminus, as compared to our study, was sufficient to cease the anterograde transport of Kv1.3 14 . On the contrary, Kv1.3 in our hands is targeted to the plasma membrane in the absence of the entire C-terminus that includes the HRET sequence, however, increase in the intracellular retention of the channel protein was observed in the truncated channels. (Suppl. Figure 6,  Fig. 4). We suppose that the use of different Kv1.3 genes (human in this study vs rat and mouse in others) cannot explain the difference due to multiple reasons: 1) we have reported previously that various C-truncated rat Kv1.3 channels had high membrane expression 16 , 2) the mouse, rat and human Kv1.3 genes have high homology, 3) in our recent paper we showed that the C-terminal deletion of WT hKv1.3 (stop codon after HRET motif) results in the expression of currents comparable to that of the full-length WT Kv1.3 10 . We think that the lack of or the very low expression detected by others for the C-terminal truncated constructs can be attributed to the cell lines used to express these constructs 12,14,29 . For example, glycosylation or other post-translational modifications can modify the surface expression level of Kv1.x channels; or the heteromerization of other Kv subunits that are endogenously present in these cells with Kv1.3 may prevent the trafficking to the plasma membrane 29,[39][40][41][42][43][44][45] . Alternatively, the mutations cited above may create a retention signal that affect the trafficking of the channels to the membrane. As shown here, we ruled out cell line specific conclusions by using two cell lines and mutation specific effects by using two point mutants having unique characteristics.
On the other hand, our results are in harmony with others showing that the lack or modification of the HRET sequence does not terminate cell surface expression of Kv1.1 or the Shaker channel without the equivalent "HRE" sequence makes it to the membrane, similar to Kv1.3 in this study 12,13 . In addition to trafficking, the role of the HRET(E) sequence was suggested in regulating the conductance of various channels. For example, the arginine in the HRET(E) sequence of Kv1.1 is critical to form a conductive channel and the Shaker channel lacking the HRE sequence shows altered steady-state activation gating 12,13 . Our results partially agree with these studies: the C-terminally truncated Kv1.3 that lacks the HRET sequence is expressed in the surface membrane but does not conduct K + current.
However, we went much further in understanding this phenomenon. First, WT Kv1.3 could not rescue the conductance of the A413V-NOHRET subunits in CHO cells, i.e., co-transfection of these two subunits resulted in pure homotetrameric WT currents. Fluorescence signals clearly showed the presence of both subunits in the plasma membrane, thus, interpretation of these results is that either WT Kv1.3 and A413V-NOHRET subunits SCIentIFIC REPORTS | (2018) 8:5937 | DOI:10.1038/s41598-018-24159-8 do not form heterotetramers or that the presence of A413V-NOHRET subunits in a heteroteramer renders the channel non-conductive in a dominant negative manner. The fact that cells transfected with WT-Kv1.3 had much greater current density as compared to cells expressing both constructs support the formation of non-conducting heterotetramers in the cell membrane (i.e., some WT subunits are engaged in non-conducting heterotetramers). As tetramerization of Kv1.3 is governed by the N-terminal tetramerization domain, rather than the C-terminus, we favor the dominant-negative effect of the mutant subunits (as has been demonstrated for other mutations as well) 1,46-48 .
The HRET(E) is sequence is located in the part of the C-terminal that is proximal to the activation gate or may be part of it 49 . Thus, mutation in this region may profoundly alter activation gating of the channels, the coupling of the voltage-sensor movement to the activation gate or both while leaving the voltage-sensor movement intact. To test this assumption, we determined the gating currents of WT-NOHRET channels. To our surprise, we found that majority of the cells expressing this channel did not exhibit gating currents that resemble the gating current of the non-conducting W384F mutant (Fig. 6), which was used as a control. As fluorescence signals confirm the surface expression of the NOHRET construct this result suggests that voltage-sensor movement is impaired in the NOHRET Kv1.3 that lacks the full C-terminus including the HRET sequence. The origin of this unknown "reverse coupling" (i.e. movement of the voltage sensor is impaired by modification of the activation gate region) is unknown, but it does not seem to be specific for the HRET sequence. When we deleted just the HRETE sequence (WT-XHRETE construct) or substituted with alanines (WT-polyA) and left the rest of the carboxyl-terminus intact, or changed HRETE for five alanines in the WT-Atail construct (and rest of C-terminus was removed) the conductance of Kv1.3 was recovered. So it seems that the lack of a peptide strand on the C-terminus at the activation gate renders the channels non-conductive and any replacement may substitute for the HRETE sequence, at least in Kv1. 3.
In summary, we demonstrated that Kv1.3 channel trafficking to the plasma membrane is preserved even if the whole C-terminus, including the HRET sequence is deleted. This finding highlights that trafficking motifs may not be universal and their importance must be tested for each channel/expression system combination. A similar conclusion can be drawn for the role of the HRET sequence in regulating channel conductance 12,49 : in case of Kv1.3 even an alanine substitution of the HRETE sequence restored channel function. Based on this and other recent papers the presence of the C-terminal amino acids adjacent to the activation gate in Kv1.3 are important for maintaining ion conductance, whereas distant C-terminal amino acids govern interactions with other proteins or confer cholesterol sensitivity to the Kv1.3 50 .
Cell culture and transfection. HEK-derived tsA-201 (later we call them HEK for simplicity) and CHO (chinese hamster ovary, both from ATCC, Germany) cells were cultured in DMEM medium (Sigma-Aldrich Ltd., Hungary), which contained 10% FBS, 1 mM Na-pyruvate, and 200 units penicillin/streptomycin. Cells were maintained at 37 °C in a humidified atmosphere of 5% of CO 2 and 95% air. Cells were passaged every 2-3 days. Transfections of DNA plasmids were performed using Lipofectamine 2000 TM (Life Technologies, Hungary) according to the manufacturer's protocol. The patch-clamp and immunocytochemistry experiments were performed 24 hours post transfection. The CHO cell line stably expressing EGFP-tagged W384F-Kv1.3 was established as detailed in ref. 10 .
The current density was defined as the ratio of peak current detected at +50 mV test potential and the whole-cell capacitance (read from the compensatory circuit of the amplifier). The remaining fraction of the SCIentIFIC REPORTS | (2018) 8:5937 | DOI:10.1038/s41598-018-24159-8 current (RF) for TEA inhibition was defined as the ratio of the peak current measured after and before perfusion with 100 mM TEA.
The activation kinetics of the current was characterized by fitting the Hodgkin-Huxley (HH) model (I(t) = I a × (1−exp(−t/τ a )) 4 + C where I a is the amplitude of the activating current component; τ a is the activation time constant of the current; C: constant) to the rising phase of the current traces evoked by 15-ms-long depolarizations to +50 mV. The activation time constant characteristic of a given cell was determined as the average of the time constants obtained upon three sequential depolarizations repeated every 15 s. P/5 protocol for online leak subtraction was applied.
The inactivation kinetics of the current was characterized by fitting a single exponential function (I(t) = I 0 × exp(−t/τ in ) + C, I 0 : amplitude of current, τ in : inactivation time constant, C: steady-state value of whole-cell current at the end of the pulse) to the decaying part of the current traces evoked by 2-s-long depolarizations to +40 mV from a holding potential of −120 mV. The inactivation time constant for a given cell was determined as for τ a , except pulses were delivered every 60 s.
The voltage-dependence of steady-state activation relationships were obtained as follows. The cells were held at −120 mV holding potential and depolarized to various test potentials ranging from −70 up to + 50 mV in 10 mV steps at every 30 s. Peak whole-cell conductance (G(V)) at each test potential was calculated from the peak current (I p ) at test potential V and the K + reversal potential (E r = −85 mV) using G(V) = I p /(V−E r ). The G(V) values were normalized for the maximum conductance and plotted as a function of test potential and the Boltzmann-function was fitted to the data points: G N = 1/(1 + exp [− (V−V ½ )/k], where G N is the normalized conductance, V is the test potential, V ½ is the midpoint and k is the slope of the function.
Gating currents were determined using a non-conducting Kv1.3 mutant (W384F-Kv1.3). A voltage step protocol was applied from −100 mV up to 20-100 mV with an increment of 10 mV, each step lasted for 50 ms, P/5 protocol was used for leak subtraction (pulses were opposite to the test potential) to reduce capacitance and leak errors during the measurements. The gating charge was calculated upon the integration of the area under gating current traces.
Immunocytochemistry. For Fig. 5 and Suppl. Figure 5: HEK and CHO cells expressing FLAG bearing EGFP-Kv1.3 plasmids were plated onto poly-L-lysine coverslips and incubated for 1 hour (37 °C, humidified, 5% CO 2 ). Then cells were fixed with 1% formaldehyde and labeled with mouse anti-FLAG M2 antibody (1:1000, Sigma-Aldrich Ltd., Hungary). Secondary antibodies (goat anti-mouse with Alexa647, Thermofisher, Hungary) were added to the cells for 1 hour. Finally, coverslips were mounted onto slides with Fluoromount G (eBioScience, USA). Zeiss LSM 510 META and Olympus FV-1000 microscopes were used to take confocal images of the cells. A He-Ne laser was selected to excite Alexa647 and mCherry (line 633 and 543 nm) and an Argon laser (line 488 nm) to visualize EGFP. The thickness of the slices was set to 1 µm.
For Fig. 4. and Suppl. Figure 8: CHO cells were plated into 8-well chamber and transfected with FLAG bearing EGFP-Kv1.3 plasmids (37 °C, humidified, 5% CO 2 ). After 24 hours the cells were washed and then fixed with 1% formaldehyde and labeled with mouse anti-FLAG M2 antibody for 2 hours, 37 °C (1:500, Sigma-Aldrich Ltd., Hungary). Secondary antibodies (goat anti-mouse with Alexa647, 1:500, Thermofisher, Hungary) were added to the cells for 1 hour. Zeiss LSM 510 META and Olympus FV-1000 microscopes were used to take confocal images of the cells. A He-Ne laser was selected to excite Alexa647 (line 633 nm) and an Argon laser (line 488 nm) to visualize EGFP. The thickness of the slices was set to 1 µm. Western blotting. Protein samples were separated by SDS-PAGE and transferred to PVDF membranes (Millipore, Billerica, MA, USA) after electrophoresis. The membranes were blocked with milk powder, and immunoblotted with mouse-anti-FLAG M2 (Sigma-Aldrich Ltd., Hungary) or rabbit-anti-actin (Sigma-Aldrich Ltd., Hungary) primary and anti-mouse IgG HRP-linked or anti-rabbit IgG HRP-linked secondary antibodies (Cell Signaling Technology, Inc., Beverly, MA), respectively. Blots were developed with ECL reagent (Thermo Scientific Inc., Vantaa, Finland). The blots were then visualized with the FluorChem Q MultiImage III Western blot imaging system (ProteinSimple, USA).

Statistical analysis.
Data are reported as the mean ± standard error. Means were compared using Student's t-test or one-way ANOVA. P-value was set to 0.05. Statistical analyses were performed using SigmaPlot version 10.0 (Systat Software Inc.).