Ca2+/calmodulin regulates Kvβ1.1-mediated inactivation of voltage-gated K+ channels

A-type K+ channels open on membrane depolarization and undergo subsequent rapid inactivation such that they are ideally suited for fine-tuning the electrical signaling in neurons and muscle cells. Channel inactivation mostly follows the so-called ball-and-chain mechanism, in which the N-terminal structures of either the K+ channel’s α or β subunits occlude the channel pore entry facing the cytosol. Inactivation of Kv1.1 and Kv1.4 channels induced by Kvβ1.1 subunits is profoundly decelerated in response to a rise in the intracellular Ca2+ concentration, thus making the affected channel complexes negative feedback regulators to limit neuronal overexcitation. With electrophysiological and biochemical experiments we show that the Ca2+ dependence is gained by binding of calmodulin to the “chain” segment of Kvβ1.1 thereby compromising the mobility of the inactivation particle. Furthermore, inactivation regulation via Ca2+/calmodulin does not interfere with the β subunit’s enzymatic activity as an NADPH-dependent oxidoreductase, thus rendering the Kvβ1.1 subunit a multifunctional receptor that integrates cytosolic signals to be transduced to altered electrical cellular activity.

they also substantially contribute to various means of regulation by which cells fine-tune K + channel inactivation and thereby regulate electrical excitation. For example, cysteine residues in the "ball" domain make N(β )-type inactivation sensitive to changes in the intracellular redox milieu as with Kv1.4 and Kv3.4 channels 6,7,13,14 . Furthermore, the degree and kinetics of N(β ) inactivation depend on the state of phosphorylation [15][16][17] , lipid composition 18 and the intracellular pH 19 . The enzymatic activity of the Kvβ 's core domains to act as oxidoreductases 20 also has an impact on N(β ) inactivation because the mobility of the "chain" depends on whether the subunit is complexed with NADPH or NADP + , i.e. enzymatic activity and electrophysiological function are coupled 21,22 .
Most importantly, an increase in the intracellular Ca 2+ concentration slows down inactivation induced by Kvβ 1.1 subunits, and the N terminus of these subunits appeared to be necessary for that function 23 . Thus, K + channels that undergo inactivation by means of Kvβ 1.1 subunits provide a negative feedback regulation, as they tend to limit excitation in response to Ca 2+ overload. This phenomenon is not universal because inactivation induced by the N-terminal splice variant Kvβ 1.3 does not depend on [Ca 2+ ] i 24 . However, the molecular mechanism underlying the Ca 2+ dependence of Kvβ 1.1-mediated inactivation remained to be elucidated. Knowledge of this mechanism would furthermore allow addressing the question of a potential crosstalk between Ca 2+ sensitivity on the one hand and dependence of Kvβ -induced inactivation on the enzymatic activity on the other.
Here, we identified calmodulin (CaM) as the Ca 2+ sensor protein responsible for the Ca 2+ sensitivity of Kvβ 1.1-induced K + channel inactivation. This Ca 2+ /CaM dependent inactivation modulation was independent of cellular oxidation and of the intrinsic enzymatic activity of Kvβ 1.1.

Results
Ca 2+ dependence of Kvβ-induced inactivation. It was previously shown that rapid N-type inactivation of Kv1.1 channels when coexpressed with Kvβ 1.1 depends on the level of free intracellular Ca 2+ 23,24 . This phenomenon is readily observed when Kv1.1 and Kvβ 1.1 are coexpressed in HEK 293T cells and currents are recorded in the whole-cell patch-clamp mode with an intracellular solution only weakly Ca 2+ -buffered with 100 μ M EGTA. Under this condition, Kvβ 1.1 induced rapid inactivation when channels were activated by a depolarizing step (Fig. 1b, black), characterized by an inactivation time constant of 8.01 ± 1.85 ms. Upon extracellular application of the Ca 2+ ionophore ionomycin, however, the inactivation was slowed down considerably (151 ± 13.5 ms) and the peak outward current was increased about twofold (Fig. 1b,c, red). No change in peak current upon ionomycin application was observed when Kv1.1 was expressed alone.
In search for the molecular mechanism underlying this regulation, we analyzed the N-terminal sequence of Kvβ 1.1 and also of other Kvβ subunits for potential interaction motifs for Ca 2+ -binding proteins, such as calmodulin (CaM). While the "Calmodulation database and meta-analysis predictor" algorithm according to Mruk et al. (2014) 25 did not predict CaM binding sites in the N-terminal sequences of Kvβ 1.2, Kvβ 1.3, and Kvβ 3.1, i.e. Kvβ subunits that can induce N-type inactivation 9 , there was a high score in the N-terminal sequence of Kvβ 1.1, ranging from position F40 through I59 (Fig. 1a), i.e. directly following the N-terminal sequence that presumably forms the "ball" domain (extending to residue 34). In this region, there are typical features of a CaM-binding structure comprising basic residues (R37, R41, R48) and aromatic residues (F40, F53). To elucidate if such structures are involved in the observed Ca 2+ dependence of Kvβ 1.1-induced inactivation, asparagine was replaced for the arginines yielding the Kvβ 1.1 mutant RRR and serine for phenylalanines to yield the mutant Kvβ 1.1-FF. Upon coexpression with Kv1.1 in HEK 293T cells, both constructs induced rapid inactivation; inactivation by RRR was slightly slower than that of the wild type (13.5 ± 0.4 ms), while inactivation induced by FF was about 2.5-fold faster (4.0 ± 0.6 ms) (Fig. 1b,c). Application of ionomycin, however, did not significantly affect the time course of inactivation in either case (Fig. 1c). The peak currents obtained for the coexpression of RRR were only increased by 9.9 ± 2.4%, that for FF by 34 ± 7.6% (Fig. 1d).
These data clearly indicate that the putative CaM-binding motif is involved in the Ca 2+ dependence of Kvβ 1.1-induced inactivation. The impact of the mutations on inactivation was also observed when currents were recorded in the whole-cell mode with defined intracellular solutions containing either no free Ca 2+ (buffered with 10 mM EGTA) or 1 μ M Ca 2+ (Supplementary Fig. S1a-c). Furthermore, intracellular Ca 2+ did not affect the time course of recovery from inactivation, indicating that the offrate of N(β )-terminal inactivation is not Ca 2+ dependent either ( Supplementary Fig. S1d).
Within the Kv1 subfamily of Kv channels, Kv1.4 α subunits harbor N-terminal domains that also induce channel inactivation. When expressed in HEK 293T cells and subjected to the application of ionomycin to elevate the intracellular Ca 2+ concentration, the kinetics of inactivation, however, was not altered (Fig. 2a, left), thus rendering the N-type inactivation endogenous to Kv1.4 channels insensitive to acute increases in [Ca 2+ ] i . However, in a physiological setting Kv1.4 often coassembles with Kvβ 1.1 subunits, thus yielding Kv channel complexes with eight N-terminal inactivation domains 26 . As a result, inactivation at 50 mV, which is approximated with a single-exponential time constant of 49.1 ± 6.3 ms (n = 7) for Kv1.4, is substantially accelerated in the presence of Kvβ 1.1-C7S (11.0 ± 1.2 ms, n = 8). We used mutant Kvβ 1.1-C7S to avoid any confounding with the redox milieu and a potential crosstalk with the cysteine (C13) in the "ball" domain of Kv1.4. Application of ionomycin slowed down this inactivation about twofold (P < 0.05; Fig. 2a, center, Fig. 2b). Coexpression of mutant Kvβ 1.1-C7S-RRR made the inactivation of the Kv1.4/Kvβ 1.1-C7S-RRR complex even faster (6.05 ± 0.38 ms, n = 10), but ionomycin Scientific RepoRts | 5:15509 | DOi: 10.1038/srep15509 was without effect (P = 0.12; Fig. 2a, right, Fig. 2b). In addition to a deceleration of the inactivation, elevated [Ca 2+ ] i also increased the peak outward current in the combination of Kv1.4/Kvβ 1.1-C7S but not with Kv1.4 alone or when coexpressed with Kvβ 1.1-C7S-RRR (Fig. 2c). Thus, by means of coassembly with Kvβ 1.1 subunits, Kv1.4 channels not only acquire a faster N-type inactivation but also one that is regulated by intracellular Ca 2+ .
Calmodulin mediates Ca 2+ sensitivity of Kvβ1.1. To gain direct access to the cytosolic side of the plasma membrane with functional Kv1.1/Kvβ 1.1 complexes, we expressed the corresponding mRNAs in Xenopus oocytes and obtained macro-patches in the inside-out configuration. In this mode, various solutions can be directly applied to assay the channels' dependence on Ca 2+ and calmodulin. When a membrane patch was excised into a solution devoid of Ca 2+ and CaM, rapid inactivation was observed (Fig. 3a, black). Application of solutions with 1 μ M free Ca 2+ did not have any impact on the current (Fig. 3a, green); 1 μ M Ca 2+ substantially slowed down inactivation only when coapplied with 1 μ M recombinantly produced CaM (Fig. 3a, red). A similar experiment is shown in Fig. 3b, illustrating that CaM in the absence of Ca 2+ has no effect, and only Ca 2+ /CaM in combination remove inactivation induced by Kvβ 1.1. For mutants RRR (Fig. 3c) and FF (Fig. 3d), however, even Ca 2+ /CaM was ineffective in removing inactivation as also illustrated in Fig. 3e as mean over multiple experiments.
GST-pull-down assays were performed to study the physical interaction of CaM with Kvβ 1.1. Co-precipitation of CaM with GSH sepharose-bound proteins was only observed with GST-fused wild-type Kvβ 1.1 in the presence of free Ca 2+ ions. By contrast, GST fusions of the Kvβ 1.1 mutants RRR or FF did not co-precipitate CaM in this binding assay (Fig. 3f). This finding suggests that only one CaM binding site exists in the N terminus of Kvβ 1.1.   The lack of a Ca 2+ effect on inactivation in the absence of CaM appears to be in contrast to the report of Jow et al. (2004) 23 who showed a Ca 2+ dependence of the Kvβ 1.1-induced inactivation time constant in inside-out patches without CaM application. Although we cannot offer an unequivocal explanation for that observation, we noticed that the amount of cytosol that sticks to the membrane patch upon establishment of the inside-out configuration has a strong influence on how much endogenous CaM is available to facilitate loss of inactivation. As illustrated in Supplementary Fig. S2 for such patches with some cytosol adhering, the inactivation time course of Kv1.1+ Kvβ 1.1 complexes is fast in the on-cell configuration. Immediately upon patch excision into bath solution with 1 μ M free Ca 2+ , however, channels do not inactivate anymore; yet inactivation is restored by transfer of the patch into Ca 2+ -free solution and subsequently inactivation is preserved even in 1 μ M free Ca 2+ .
Ca 2+ sensitivity of Kvβ1.1 is functionally independent of its enzymatic activity. Kvβ subunits exhibit an enzymatic activity as aldoketoreductases using NADPH as a cofactor 20 . For Kvβ 1.1 with its N-terminal structure leading to N(β )-type inactivation, enzymatic activity, e.g. induced by the application of the substrate 4-cyanobenzaldehyde (4CY), results in a significant slow-down of inactivation 21 . A mutagenesis study by Pan et al. (2011) 22 provided an explanation for this phenomenon, in which an immobilization of the N-terminal ball domain is induced by oxidation of the cofactor NADPH, bound to the core domain of Kvβ 1.1, to NADP + ; in this state, the chain of the inactivation domain binds to the Kvβ 1.1 core domain such that the distal ball domain cannot reach its receptor anymore and, hence, is unable to induce inactivation. Key players for the electrostatic coupling of "chain" and "core" domain are residues R37/R48 and E265/E349, respectively. Since the arginine residues R37 and R48 are apparently taking part in binding CaM to the Kvβ 1.1 subunit, it is plausible that the regulation of Kvβ 1.1-induced inactivation by NADPH/NADP + and Ca 2+ /CaM are coupled.
We therefore measured the impact of Kvβ 1.1 enzymatic activity for the mutants that eliminated the Ca 2+ /CaM dependence. Since the enzymatic activity requires oxidizing conditions, which would eliminate inactivation of Kvβ 1.1 by means of the regulatory N-terminal cysteine (C7) 7 , we used mutant Kvβ 1.1-C7S and expressed it together with Kv1.1 in HEK 293T cells. Patch pipettes contained 1 mM of the substrate 4CY. Currents in response to depolarization to 50 mV were measured immediately after establishment of the whole-cell configuration, yielding inactivating K + currents. During the course of substrate diffusing into the cytosol, the inactivation time course became progressively slower for Kvβ 1.1-C7S (Fig. 4a, left). As expected, this did not happen for mutant RRR (Fig. 4a, center). However, 4CY also potently slowed down inactivation induced by Kvβ 1.1-FF (Fig. 4a, right). As a control, we also measured mutant Kvβ 1.1-E349K, which did not show a significant response to 4CY (Fig. 4b,c). Thus, as summarized in Fig. 4c, the substrate 4CY potently affected the inactivation time course of Kvβ 1.1-C7S and Kvβ 1.1-C7S-FF, but not of Kvβ 1.1-C7S-RRR and Kvβ 1.1-C7S-E349K, demonstrating that an intact Ca 2+ / CaM dependence is not required for the inactivation modulation via NADPH oxidation. Conversely, we measured the impact of elevated Ca 2+ level on such constructs, as shown in Fig. 4d,e. Application of ionomycin to HEK 293T cells expressing Kv1.1+ Kvβ 1.1-C7S-E349K removed inactivation efficiently. In addition, the other mutations in the background of Kvβ 1.1-C7S (Fig. 4e) showed the same dependence on [Ca 2+ ] i elevation as measured for Kvβ 1.1 wild type (Fig. 1d), thus in summary illustrating that the removal of N(β )-type inactivation induced by Ca 2+ /CaM is not associated with a restraint of the chain flexibility that results from docking of R37/R48 in the chain to E265/E349 in the core domain of Kvβ 1.1.

Discussion
The auxiliary Kvβ 1.1 subunit converts voltage-gated K + channels formed of α subunits of the Kv1 subfamily to rapidly inactivating A-type channels when coexpressed in heterologous systems 7,8 . In the mammalian brain, Kvβ 1.1 subunits coexpress with Kv1.1 and Kv1.4 in nerve terminals of cortical interneurons, mossy fibers, and in the substantia nigra 26 , where they regulate action potential frequency and shape, and thus also neurotransmitter release. Slow-down of A-type channel inactivation delays firing of action potentials and reduces the cell excitability by controlling the Ca 2+ inflow 27 . The slow inactivation of Kv1.4 channels can produce progressive spike broadening and alteration of the action potential frequency 28 . The relevance of Kvβ -mediated inactivation and its precise tuning is underscored by the fact that various physiological parameters, such as the cellular redox status, cytosolic pH, phosphorylation signaling, and intracellular Ca 2+ can modulate this process. It is now clear that Kvβ 1.1 is able to combine and integrate signals from diverse pathways, coupling cellular excitability to cell physiology.
In the present study, we demonstrate that inactivation conferred to Kv1.1 and Kv1.4 channels by means of Kvβ 1.1 subunits is considerably slowed down by intracellular Ca 2+ , and that this Ca 2+ sensitivity arises from calmodulin, which binds to the "chain" structure of Kvβ 1.1 subunits. By means of this mechanism, K + channels formed of Kv1. Here we showed that elevated intracellular Ca 2+ concentrations slow down inactivation induced by Kvβ 1.1 in Kv1.1, as well as in Kv1.4 channels, but also that this effect requires the Ca 2+ -binding protein calmodulin. This result appears to be in contrast to an earlier report 23 according to which Ca 2+ alone can decelerate Kvβ 1.1-mediated inactivation. Although we cannot offer an unequivocal explanation for the results by Jow et al. (2004) 23 , we demonstrated that 1 μ M free Ca 2+ applied to the cytosolic face of clean membrane patches of Xenopus oocytes does not eliminate Kvβ 1.1-mediated inactivation, while a combination of CaM and Ca 2+ (1 μ M each) does (Fig. 3, Supplementary Fig. 2). Furthermore, there is a clear hit for a potential CaM-binding motif in the N-terminal structure of Kvβ 1.1 (between position 32 and 56, Fig. 1a), and mutations in this domain render the Kvβ 1.1 subunits insensitive towards intracellular Ca 2+ /CaM. Finally, recombinant CaM physically interacts with Kvβ 1.1 subunits, but not with mutants with impaired CaM-binding motifs (Fig. 3f). Thus, Kvβ 1.1 subunits gain their Ca 2+ dependence from association with CaM, similar to other ion channels, such as cyclic nucleotide-gated channels 33 , N-methyl-D-aspartate receptors 34 , Ca 2+ -activated K + channels of intermediate and small conductance 35 , and EAG1 channels (Kv10.1) 36 , to name a few. In all such cases, the Ca 2+ sensitivity of CaM couples the channels to the intracellular Ca 2+ concentration in a physiological range of a few 100 nM, thus enabling the channels to quickly respond to moderate excursions from resting [Ca 2+ ] i levels. Unlike to Ca 2+ -activated K + channels of intermediate and small conductance 35 , however, CaM only undergoes a loose interaction with Kvβ 1.1 subunits because CaM only binds to the Kvβ 1.1 protein in the presence of free Ca 2+ (Fig. 3f), and prebound CaM can be washed off excised membrane patches in the absence of Ca 2+ (Supplementary Fig. 2).
Binding of calmodulin to the flexible "chain" region of Kvβ 1.1 must be expected to restrain the mobility of the "ball" and, hence, to interfere with channel inactivation. Restraining of the N terminus has also been proposed as mechanism underlying the redox modulation of Kvβ 1.1 22 . Oxidation of Kvβ -bound NADPH induces a conformational rearrangement of the Kvβ core domain allowing the interaction of negative charges (E265, E349) on the surface of the core with positive residues in the chain (R37, R48), thereby leading to slower inactivation kinetics 22 . Based on this mechanism, it is conceivable to assume that Ca 2+ /CaM interacts with the Kvβ core domain and causes ball immobilization by binding of the chain to the core domain residues E265 and E349. In this case, redox regulation and Ca 2+ /CaM regulation would share exactly the same mechanism to interfere with inactivation. This hypothesis was disproven by our results of the Kvβ 1.1 Ca 2+ sensitivity in the background of an E349K mutation in the core domain. In this setting, the redox regulation was abolished, while Ca 2+ regulation of inactivation was fully conserved. Thus, various regulatory mechanisms converge on the Kvβ 1.1 channel subunit to fine-tune the kinetics of Kv channels in response to physiological parameters such as [Ca 2+ ] i , redox potential, and phosphorylation signaling. While all of these regulation events ultimately lead to diminished "ball" mobility and slower inactivation via Kvβ 1.1, the precise molecular mechanisms and the amino acids involved are specific for each regulatory pathway. The exact knowledge of these modulation mechanisms is an important precondition for efforts to establish Kvβ 1.1 as a target for a pharmacologic control of Kv1 channels.
In-silico search for potential calmodulin binding sites. The N-terminal sequences of various Kvβ subunits were subjected to an analysis searching for potential binding sites for calmodulin according to Mruk et al. (2014) 25 .

Channel expression in Xenopus oocytes and HEK 293T cells.
Capped mRNA was synthesized in vitro using the mMessage mMachine kit (Ambion, Austin, TX, USA). Oocytes were surgically removed from the ovarian tissue of Xenopus laevis that had been anesthetized by immersion in ice water⁄tricaine according to the local animal care program. The oocytes were defolliculated, and healthy stage V and VI oocytes were isolated and microinjected with 50 nl of a solution containing channel wild-type or mutant mRNA. In co-expression experiments, the ratio of mRNA coding for α and β subunits was 1:3. Electrophysiological measurements were performed 2-4 days after mRNA injection.
HEK 293T cells were transiently transfected using the Roti-Fect transfection kit (Carl Roth, Karlsruhe, Germany). Dynabeads (Deutsche Dynal GmbH, Hamburg, Germany) were used for visual identification of individual cells, cotransfected with CD8. The ratio of DNA coding for α and β subunits was 1:3. Electrophysiological recordings were performed 2-3 days after transfection.