Uncoupling PIP₂-calmodulin regulation of Kv7.2 channels by an assembly destabilizing epileptogenic mutation

Phosphatidylinositol(4,5)bisphosphate (PIP₂) is a minor (<1%) acidic phospholipid found in the inner leaflet of the cell membrane and plays a vital part in cellular signaling by directly interacting with membrane proteins, including Kv7 potassium channels (Gamper and Shapiro, 2007; Suh and Hille, 2008; Hille et al., 2014; Zaydman and Cui, 2014). Function of Kv7 channels is absolutely dependent on PIP₂ (Zaydman and Cui, 2014) and activation of phospholipase C, and subsequent PIP₂ hydrolysis causes downregulation of their activity, which in turn lowers the threshold for activity (Brown et al., 2007), orchestrating excitability in brain, heart, skeletal muscle and inner ear. Related diseases encompass epilepsy, autism, schizophrenia, cardiac arrhythmias, hearing loss and sudden death (Soldovieri et al., 2011; Dvir et al., 2014a; Maljevic and Lerche, 2014).

The five members of the Kv7 family of non-inactivating voltage-dependent potassium channels share a common architecture, differing from that of Kv1–Kv4 channels in lacking an N-terminal T1 tetramerization domain. Instead, and in common with a large group of channels, such as Eag, Erg, SK, CNG and TRP channels, present multiple calmodulin (CaM; Uniprot P62161)-binding domains followed by a tetrameric coiled-coil segment (Jenke et al., 2003; Tsuruda et al., 2006). Kv7 channels have a large intracellular C-terminal region, ranging from 320 to 500 residues in size, containing four helical regions (A–D), which can be conceptually divided into three parts. Immediately after the last S6 transmembrane segment, the intracellular membrane proximal half (AB) is important for CaM binding and channel gating. Three-dimensional (3D) reconstitution locates the intracellular distal part (helix D) far from the membrane (Dvir et al., 2014b), which directs oligomerization and partner specificity (Yus-Nájera et al., 2002; Howard et al., 2007; Haitin and Attali, 2008). The AB and D helices are connected by helix C, indispensable for function, and a linker of variable length. Mutagenesis suggests that, in addition to the S4–S5 linker and the proximal C-terminus, helix C contributes to PIP₂ regulation (Dvir et al., 2014b; Zaydman and Cui, 2014). CaM binding is essential for Kv7 channels to exit from the endoplasmic reticulum, and influences heteromeric assembly of Kv7.2/3 channels and subsequent enrichment at the axonal initial segment (Yus-Nájera et al., 2002; Devaux et al., 2004; Chung et al., 2006; Etxeberria et al., 2008; Haitin and Attali, 2008; Alaimo et al., 2009; Cavaretta et al., 2014; Chung, 2014; Liu and Devaux, 2014). CaM regulatory mechanisms that change the gating behavior proceed through an effect on sensitivity to PIP₂ (Kosenko et al., 2012; Kosenko and Hoshi, 2013; Zaydman et al., 2013; Kang et al., 2014). The interlinker between helices A and B is not essential for function, although it plays a critical role in PIP₂ regulation for Kv7.3, but not in Kv7.1 and Kv7.2 channels (Hernandez et al., 2008; Aivar et al., 2012; Sachyani et al., 2014).

In contrast to the obligatory role of helices ABC for Kv7.2 channel function, helix D is dispensable (Schwake et al., 2006; Nakajo and Kubo, 2008). The helix D coiled-coil bundle has been largely considered as a passive stitch, with little active role other than to convey stability to the tetramer and specificity during the formation of heteromeric assemblies (Schmitt et al., 2000; Jenke et al., 2003; Maljevic et al., 2003; Schwake et al., 2006; Howard et al., 2007). Whilst this is undoubtedly part of its function, we find that this structure has a more active role by indirectly influencing PIP₂ dependency.

Here, we identified a mechanism to modulate, at a distance, regulation by PIP₂. We monitored the relative distance/orientation of the ABCD domain by live-cell fluorescence resonance energy transfer (FRET) assays in conjunction with analysis of whole-cell currents. The observations we made explain how the tetrameric conformation of helix D can functionally influence CaM binding to a distal site, and, in turn, how a pathological mutation affecting tetramer stability modifies Kv7.2 activity.

Calmodulin is a bi-functional protein, with two highly homologous lobes (N and C) joined by a flexible linker, each capable of engaging targets adopting an α-helix configuration (Villarroel et al., 2014). Helix A of Kv7.2 presents a marked preference for fetching the C-lobe, whereas helix B anchors more favorably to the N-lobe (Alaimo et al., 2014) and these features are present on the crystallographic Kv7.1 [AB/CaM] complex (Sachyani et al., 2014). To place the data of this report into physical perspective, Fig. 1 shows a tentative disposition of CaM on Kv7.2 forming a ring under the pore that complies with numerous restrictions from different studies (Mruk et al., 2012; Xu et al., 2013; Alaimo et al., 2014; Sachyani et al., 2014) (see Materials and Methods). One important feature derived from the 3D structure of the Kv7.1 [ABCD/CaM] complex is that the helix D coiled-coil is separated from the membrane by the CaM ring (Sachyani et al., 2014) (Fig. 1A).

To investigate in more detail the interaction with the Kv7.2 CaM-binding domain, we co-expressed CaM and a construct bearing the Kv7.2 CaM-binding elements (helices A and B, residues G313–R530) to produce the recombinant complex in bacteria. The intervening A–B linker (residues Y372–T501), which is not essential for Kv7.2 function (Aivar et al., 2012), was removed to improve protein yield. We found by size exclusion chromatography (SEC) that the purified material migrated with an apparent molecular mass of 35.3 kDa, which is close to the expected size for a 1:1 [CaM/AB] globular complex (16.7+13.6=30.3 kDa; Fig. 1C). Thus, similar to Kv7.1 or Kv7.4 [AB/CaM] complexes studied in vitro (Wiener et al., 2008; Xu et al., 2013), the Kv7.2 [AB/CaM] complex adopts a 1:1 stoichiometry in solution.

We studied next the consequences of the helix D L609R mutation [equivalent to L637R in the long Kv7.2 splice variant (Richards et al., 2004)] found in patients with hereditary benign familial convulsions (Richards et al., 2004), which is predicted to interrupt coiled-coil formation (Lupas and Gruber, 2005; Schwake et al., 2006) (Fig. 1B). To this end, non-tagged CaM and ABCD C-terminally tagged with a fluorescent protein were co-expressed in bacteria, and the resulting complexes were purified and analyzed. The SEC elution volume was consistent with a 4:4 [ABCD/CaM] globular complex. In contrast, the larger elution volume of the L609R mutant indicated that no stable tetramers were formed (Fig. 1C). In addition, the fastest non-denaturing electrophoretic pattern of [ABCD-L609R/CaM] compared with [ABCD/CaM] indicates disruption of the complex (Fig. 1D). Thus, the L609R mutation impedes the adoption of a stable tetrameric coiled-coil configuration.

It has been suggested that FRET results between isolated domains of Ca²⁺ voltage-dependent channels and CaM can be applied to the entire protein (Ben Johny et al., 2013). Inspired by this precedent, the interaction between CaM and the C-terminal Kv7.2 domain was analyzed by FRET in living cells. Cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) exhibit 50% energy transfer at a distance of ∼5 nm and measurable transfer up to 8 nm (Patterson et al., 2000). The ratio of the integral of CFP/YFP emission isolated after spectral unmixing is directly proportional to the FRET efficiency. The FRET index between CaM and helices AB was important (Fig. 2A,E), whereas between helices CD and CaM it was insignificant (Fig. 2B,E,G). These data reinforce the initial conclusion, based on yeast two-hybrid interaction trap and pull-down assays, that helices CD do not contribute directly to CaM binding (Yus-Nájera et al., 2002). Similarly, a role of CD helices has been discarded for CaM binding to Kv7.4 and Kv7.1 channels (Xu et al., 2013; Sachyani et al., 2014). In contrast, an increased FRET index compared with that of AB was observed between CFP-CaM and YFP-ABCD (Fig. 2C,E), which is an expected consequence of having multiple donors and acceptors within the FRET distance [<8 nm (Patterson et al., 2000; Vogel et al., 2006); Fig. 2F]. Based on the structure of the related Kv1.2 channel (Long et al., 2005), two fluorescent proteins in the same channel should be ∼5.6 nm apart (assuming they are in adjacent subunits) or 7.9 nm apart (assuming they are in non-adjacent subunits). In agreement with this interpretation, disruption of helix D coiled-coil by the L609R mutation resulted in a FRET index comparable to that of AB (Fig. 2D,E). The differences in FRET efficiency were apparent at low free donor or acceptor concentrations, where collisional or spurious FRET is negligible (Fig. 2G; Figs S1, S2). Thus CaM and AB from different subunits are within FRET distance in the tetrameric complex brought together by helix D.

To gain further insights into the [Kv7.2/CaM] complex, the impact of tetramerization on the separation between A helices of neighboring subunits was gauged by FRET analysis, attaching CFP or YFP at the N-terminus, upstream of helix A. This configuration allowed the assessment of the transfer of energy between adjacent subunits, which was virtually undetectable between CFP-AB and YFP-AB (Fig. 3A, compare cyan and green traces). In contrast, a clear FRET signal was detected between CFP-ABCD and YFP-ABCD (Fig. 3B,E), indicating that the separation of the N-termini is below 8 nm in the tetrameric complex. Disruption of the helix D coiled-coil formation by the L609R mutation resulted in a large reduction of the FRET index (Fig. 3C,D), reinforcing the view that helix D is critical for bringing A helices close together.

Three-dimensional structural data indicate that the ABCD domain can be envisioned as a flower bouquet with the tetrameric coiled-coil helix D corresponding to the pedestal (Sachyani et al., 2014) (Fig. 1). We can imagine the bunch in two extreme situations: in a compact arrangement or with flaccid stacks positioning the flowers towards the periphery (Fig. 3G). FRET can distinguish between closed and spread configurations, because in the first case donors and acceptors from neighboring [CaM/AB] complexes will be within FRET distance. With this idea in mind, the impact of elevating CaM levels was monitored. The FRET index recorded after disruption of tetramerization by the L609R mutation was low, and increased after CaM elevation (Fig. 3C,D; compare green and salmon plots), whereas no significant changes were revealed when helix D was missing (Fig. 3A). In contrast, a prominent FRET index was obtained for the CFP-ABCD and YFP-ABCD pair, which almost doubled with elevated CaM levels (Fig. 3B). Thus CaM leads to a rearrangement, probably by compaction (Fig. 3G), of the disposition of helix A in the ABCD complex. This compaction also takes place for the L609R mutant, but it does not reach the distant/orientation values of the ABCD complex. Furthermore, the FRET index was significantly reduced in cells overexpressing neurogranin, an apo-calmodulin-binding protein (Villarroel et al., 2014) that is expected to diminish CaM availability (Fig. 3F). These results suggest that CaM favors, by mass action, the adoption of a novel organization of the ABCD domain, and that this new configuration is favored by helix D acquiring a tetrameric coiled-coil formation.

The changes on FRET suggest that CaM can bind to ABCD in at least two different configurations, resulting in an equilibrium scheme between a minimum of three states (Fig. 4A). A testable consequence of the incorporation of an additional binding mode is that the apparent binding affinity should decrease after disruption of helix D-mediated tetramer formation by L609R (Fig. 4). Using e-FRET methodology (Chen et al., 2006) we resolved the binding curve in which the FRET strength between CFP-CaM and YFP-ABCD is plotted cell-by-cell versus concentration of free YFP-ABCD (Ben Johny et al., 2013). This revealed that, besides decreasing maximal FRET, the L609R mutation led to a reduction in apparent binding affinity (Fig. 4B). To further evaluate the impact of this mutation in CaM binding in a more controlled in vitro setting, recombinant ABCD and ABCD-L609R proteins were produced in bacteria and purified in the absence of CaM (Alaimo et al., 2009). The refolded monodisperse CaM-binding domains were subsequently assayed in vitro for binding to dansylated CaM (D-CaM) (Alaimo et al., 2013b; Bonache et al., 2014). In accordance with theory and with results obtained by using live cells, the elimination of the presumed additional binding mode was accompanied by a clear decrease in apparent binding affinity for the L609R mutant (Fig. 4C,D). In the tetramer, the packing suggests that binding of one CaM might affect binding to another CaM. However, the Hill coefficient (Edelstein and Le, 2013) of the dose–responses did not change as a consequence of coiled-coil disruption, suggesting that tetramerization has minor or no effect on potential co-operativity between CaM molecules (Fig. 4C).

Cells expressing Kv7.2, complemented or not with CaM, were examined by whole-cell recording (Fig. 5A,B) to explore the functional consequences of the tetramerization-disrupting mutation on CaM-dependent function, such as current density and PIP₂ dependency. Current density was not significantly affected by this mutation (Fig. 5), whereas there was a remarkable 18 mV right shift of the conductance–voltage relationship (Fig. 5C). The decrease in CaM-binding affinity caused by the L609R mutation should tend to reduce the ensemble occupancy of the four CaM-binding sites on a Kv7.2 channel. A similar decline in the steady-state occupancy should occur when the concentration of free CaM is critically low. To test this idea, the properties of Kv7.2 channels in a low CaM environment were evaluated. Under reduced CaM conditions there was a significant decrease in current density, consistent with the role of CaM on Kv7.2 trafficking (Etxeberria et al., 2008). Remarkably, there was a 17 mV right shift of the conductance–voltage relationship (Fig. 5). Similar effects were observed in cells expressing neurogranin (n≥12, not shown). Thus both disruption of coiled-coil formation and low CaM availability led to similar changes in the conductance–voltage relationship, with a strong correlation with the impact on the proximity/orientation of the ABCD domains.

The voltage-dependent phosphatase Danio rerio voltage sensing phosphatase (DrVSP) was used to address the impact on PIP₂ dependence. To avoid voltage errors associated with the large currents evoked at the extreme voltages required to activate DrVSP (see Fig. 6A), series resistance compensation for the patch electrode was set to >95% (Sherman et al., 1999). The comparison of current relaxations before and after DrVSP activation demonstrated that the L609R mutant was more resistant to activation of the voltage-dependent phosphatase (Fig. 6B). Furthermore, a similar increase in resistance to DrVSP activation was observed for wild-type channels under low CaM conditions, which is comparable to that in channels carrying the L609R mutation. Importantly, in contrast to wild-type channels, the response of the mutant channel to activation of DrVSP was insensitive to CaM elevation (Fig. 6C). Thus disruption of helix D coiled-coil changes voltage dependency and perturbs the adoption of additional CaM-binding modes and subsequent change on PIP₂ dependency in Kv7.2 channels.

Despite their fundamental importance in regulation and signaling, allosteric mechanisms are, in general, poorly understood (Motlagh et al., 2014). Here, we show that PIP₂ dependency is affected at a distance by the formation of a coiled-coil tetrameric bundle. The data are consistent with a mechanism involving different CaM-binding modes, which confer differential PIP₂ sensitivity to Kv7.2 channels.

The data reveal that the pathogenic L609R mutation [equivalent to L637R in the long Kv7.2 splice variant (Richards et al., 2004)] destabilizes the helix D-dependent tetramerization of the C-terminal region of Kv7.2. In addition, this mutation reduces the apparent affinity for CaM binding, and leads to increased resistance to the action of a voltage-dependent phosphatase that reduces PIP₂ concentration at the plasma membrane. A plausible consequence of the reduction of CaM affinity is that CaM occupancy of the channel is diminished, and, on average, not all four sites of the tetrameric channel would be engaged in a given instant. A similar reduction in the number of resident CaM molecules can be achieved by lowering CaM availability, which led to increased resistance to DrVSP action, similar to that seen for the helix D-mutated channels. In addition, both wild-type channels when CaM is scarce and L609R mutant channels under resting CaM levels present a >17 mV right shifted conductance–voltage relationship, meaning that two different situations that lead to low CaM occupancy converge towards a similar voltage dependency and increased resistance to PIP₂ depletion. Although results obtained with isolated domains may not apply to the full-length channel, the agreement between in vitro data and those obtained by using live cells favor the argument that these events are related, and are consistent with the idea that the effect on PIP₂ sensitivity is a consequence of the changes on CaM residence.

How can helix D affect CaM binding? Neither our new FRET data, nor previous studies using yeast two hybrid assays, or studies on Kv7.1 or Kv7.4, have exposed any hint of direct CaM interaction with helices CD (Yus-Nájera et al., 2002; Howard et al., 2007; Wiener et al., 2008; Xu et al., 2013). In addition, small angle X-ray scattering of the related Kv7.1 subunit (Sachyani et al., 2014) reveals that there is no physical contact between helix D and the [AB/CaM] complex. Therefore, we conclude that the impact of helix D in CaM binding is indirect. In addition to previous functional and structural data (Alaimo et al., 2013a; Sachyani et al., 2014), the helix D-dependent changes in CaM-binding affinity unveiled here support the concept of the occurrence of more than one binding mode. The decrease in FRET for the L609R mutant underscores the importance of coiled-coil formation by helix D in this process. The changes in homo-FRET efficiency between ABCD modules are a clear indication that the A helices re-orient, come closer, or both, when more CaM is available. Either new CaM molecules are entering the complex, or existing CaM molecules are engaging in new binding modes, such as intrasubunit and intersubunit binding modes.

The CaM N-lobe has a significant preference to anchor to helix B, whereas the C-lobe binds more favorably to the IQ site located in helix A (Alaimo et al., 2014). In the context of the full-length channel, both helices in the [AB/CaM] complex could originate from the same subunit (intrasubunit binding) or from two adjacent subunits (intersubunit binding). Intersubunit binding requires placing AB helices from contiguous subunits within CaM reach, i.e. <4 nm (Villarroel et al., 2014). Indeed, Kv7.2 models derived from trigonometric restrictions obtained with pore blockers tethered to CaM as calipers, position the lobe anchoring sites of nearby CaM molecules under 4 nm (Mruk et al., 2012) (Fig. 1A). In addition, besides the precedents for bridged configurations in other channels (Schumacher et al., 2001; Sarhan et al., 2012; Villarroel et al., 2014), the crystallographic structure of Kv7.1 [AB/CaM] has been trapped with CaM embracing helices A and B from different subunits (Sachyani et al., 2014). We have previously shown by whole-cell recording that precluding CaM binding to helix A or to helix B results in non-functional channels, but, remarkably, both mutant channels complement each other, restoring function when co-expressed (Alaimo et al., 2013a). Furthermore, at submicromolar concentrations, CaM links two Kv7.2 AB domains in vitro, anchoring helix A of one subunit to helix B of another subunit (Alaimo et al., 2013a). Therefore, we consider reasonable the hypothesis that, by bringing together the AB helices from adjacent subunits thanks to helix D coiled-coil formation, each CaM lobe can engage with the complementary helix from a neighboring subunit.

Consideration of our new data together with prior work leads us to propose an extension of the CaM modulation hypothesis to include a switch between CaM binding-modes, which may correspond to intra- and intersubunit engagement or different degrees of occupancy of the multiple CaM-binding sites, providing distant control of PIP₂ sensitivity by the tetramerization domain. The structures of many configurations are presently unknown, but homology modeling confirms the plausibility of these arrangements. In the context of the full-channel, with the restrictions imposed by the pore-forming regions preceding helix A, the approximation of these helices caused by CaM elevation should translate in a movement of the whole [ABCD/CaM] complex relative to the pore. A doubling in FRET index in the presence of elevated CaM (Fig. 3) could be accounted for by less than 6% reduction in the separation of the fluorophores, i.e. a displacement <0.5 nm, a magnitude comparable to the vertical translation of the cytoplasmic domain observed between the structures of Kir2.2 channels with and without PIP₂ (Hansen et al., 2011).

It is not known if the helix D tetramer de-oligomerize and re-oligomerize as part of the normal channel function, allowing for physiological allosteric control of channel activity. Beyond supporting different CaM-binding modes and affecting channel assembly, the helix D coiled-coil domain may also serve as a scaffold for interactions with other proteins that regulate channel activity (Marx et al., 2002; Kass et al., 2003; Wiener et al., 2008). Thus, helix D may act as an antenna, funneling relatively distant interactions towards the pore by changing indirectly the way the channel engages with CaM.

The data provide a basis for theoretical generalization of an allosteric mechanism. Many ion channels, such as Kv7, Eag, Erg, SK, CNG and TRP, present multiple CaM-binding domains followed by a tetrameric coiled-coil bundle (Jenke et al., 2003; Tsuruda et al., 2006), but other elements could assume the equivalent conceptual roles. The ingredients are two bi-functional target–receptor components, plus a mechanism that brings two or more targets in close vicinity. By controlling at a long distance the proximity between the binding sites, the receptor could engage the target in different configurations, ultimately affecting activity.

PIP₂ is strictly required for the operation of Kv7 channels, and it is currently thought that parts of the channel from well-separated regions of the linear sequence fold to bring together several basic residues to create a binding pocket (Zaydman and Cui, 2014). The concept is that the interacting basic residues do not represent a structurally selective binding site; rather they form positively charged clouds that would attract any acidic lipid. PIP₂ would be the principal target because it is the most abundant multiply-phosphorylated lipid of the plasma membrane (Suh and Hille, 2008). By reconfiguring the CaM ring underneath the pore, the cloud envelope in contact with PIP₂ at the inner leaflet of the membrane could change, such that the orientation of lysine, arginine or histidine residues making electrostatic contacts with the charged phosphate groups will be controlled at a distance by helix D. A fuzzy organization of the PIP₂-binding surface like this may help to explain the puzzling similar effects in channel-PIP₂ sensitivity of low and elevated CaM levels. Defects in channel-PIP₂ sensitivity through interfering mutations can lead to disease (Logothetis et al., 2010). It is attractive to hypothesize that mutations not located at the binding site that disrupt channel-PIP₂ dependency, such as L609R described here and which was found in a patient with an epileptic condition (Richards et al., 2004), could also lead to disease.

The human isoform 3 Kv7.2 (Y15065) cDNA was provided by T. Jentsch (Leibniz-Institut für Molekulare Pharmakologie, Berlin, Germany) and the cDNA encoding rat CaM was provided by the group of J.P. Adelman (Vollum Institute, Portland, OR, USA). The subunits tagged at the N-terminal with CFP or YFP were cloned into pCDNA3.1 and we previously confirmed that these N-terminal tags have no impact on the electrophysiological properties of the channel (Soldovieri et al., 2006; Gómez-Posada et al., 2010; Gómez-Posada et al., 2011). Point mutations were constructed by polymerase chain reaction (PCR)-based mutagenesis. Dr-VSP-IRES-GFP from zebrafish was provided by Y. Okamura (Osaka University, Osaka, Japan). The CaM sponge, which has CFP and YFP flanking the apoCaM-binding site of neuromodulin, was provided by D.J. Black (University of Missouri, Missouri, USA). For in vitro co-expression experiments, CaM was subcloned into the pOKD4 vector. C-terminal channel sequences were fused to a custom-made circularly permuted Venus variant of YFP (Shinege; China) and subcloned into pPROEX-HTc vector. Rat neurogranin (NM_024140.2) tagged with a myc epitope at the N-terminus was cloned into pcDNA3.1 hys/myc A vector.

The portrayed structure in Fig. 1 is an extension of model 3 proposed by Mruk et al., which places key residues of the CaM N- and C-lobes at a distance relative to the Kv7.2 pore compatible with the effects of a pore blocker tethered to CaM with chains of varying length (Mruk et al., 2012), and adequate for it being linked to S6, assuming that the IQ site adopts a similar disposition as that in the Kv7.1 [AB/CaM] complex (Sachyani et al., 2014), and presents a kink as predicted by sequence analysis (Alaimo et al., 2014). The structural models of [N-lobe/helix B] complexes from Kv7.1 and Kv7.4 nicely superimpose, and are consistent with nuclear magnetic resonance (NMR) data obtained for the Kv7.2 [AB/CaM] complex (Xu et al., 2013; Alaimo et al., 2014; Sachyani et al., 2014). In turn, the structural crystallographic X-ray-based model for the Kv7.1 [C-lobe/helix A] complex is compatible with NMR observations of the Kv7.2 [AB/CaM] complex (Alaimo et al., 2014; Sachyani et al., 2014), and is almost identical to every other solved complex of apo-C-lobe (Ca²⁺ free) with peptides containing an IQ motif (IQxxθR, θ=bulky apolar residue), a motif that is present in helix A of every Kv7 channel (Yus-Nájera et al., 2002; Villarroel et al., 2014). Finally, helix D has been placed at a distance from the CaM-binding site, according to the envelope observed by small-angle X-ray scattering (SAXS) for the Kv7.1 [ABCD/CaM] complex in solution (Sachyani et al., 2014). This disposition places PIP₂ and L609 of helix D more than 10 nm apart.

Recombinant proteins ABCD, ABCD-L609R (both fused to GST) and CaM were produced as described previously (Alaimo et al., 2009). Fluorometric experiments using dansyl-CaM (D-CaM), CaM dansylation, sample and buffers preparation and fluorescent measurements were performed as described previously (Alaimo et al., 2013b; Bonache et al., 2014).

Fusion proteins were expressed in Escherichia coli [BL21(DE3)] grown in lysogeny broth (LB) medium at 37°C and induced at OD_600 nm=0.6–0.8 with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) for 5 h. Cells were harvested by centrifugation at 5000 g for 20 min at 4°C, and cell pellets were frozen at −20°C. Thawed cell pellets were lysed in lysis buffer (mM: 120 KCl, 3 imidazole, 1 phenylmethylsufonyl fluoride (PMSF), 10 K-HEPES; pH 7.4), and one tablet of protease inhibitor without EDTA (Roche Applied Science, Penzberg, Germany), and passed three times through a homogenizer (Emulsiflex). Insoluble material was precipitated by centrifugation for 30 min at 25,000 g at 4°C. The resulting soluble fraction, which contained the fusion protein, was applied to a 5 ml cobalt-charged column (no. 28-9537-67, GE Healthcare, Little Chalfont, UK), and pre-equilibrated in wash buffer (mM: 120 KCl, 10 K-HEPES; pH 7.4). Increasing amounts of imidazole were used for washing. Sample elution proceeded on a linear gradient from 30 to 300 mM imidazole on an ÄKTA fast protein liquid chromatography (FPLC) system. Imidazole was removed by dialysis overnight at 4°C in cellulose membrane tubing (no. D927-100FT, Sigma-Aldrich, St Louis, MO, USA) with agitation against 1 litre of fluorescence buffer (mM: 125 KCl, 5 NaCl, 1 MgCl₂, 10 EGTA, 10 Tris; pH 7.4. Fusion proteins were then applied to a Superdex 200 pg 26/600 column (no. 28-9893-36, GE Healthcare) and eluted in the same buffer. Positive fractions were pooled together and concentrated using a 3 kDa cut-off Amicon ultra-15 concentrator (no. UFC900308, EMD Millipore, Billerica, MA, USA) in a fixed angle rotor at 3500 g. Protein concentration was measured at 462 nm, equalized to 4 µM and samples with 5% of glycerol were run in 7% non-denaturing (‘native’) gels (without SDS).

HEK293T cells (HEK 293T/17, ATCC, CRL-11268) were maintained in 5% CO₂ at 37°C in Dulbecco's modified Eagle's medium (DMEM, Sigma-Aldrich), supplemented with non-essential amino acids (Sigma, Madrid, Spain) and 10% fetal bovine serum (FBS; Lonza, Madrid, Spain). Transient transfection for imaging was performed using polyethylenimine (PEI) 25 kDa (no. 23966-22g, PolySciences, Eppelheim, Germany), whereas for electrophysiology experiments lipofectamine 2000 (Invitrogen) was used following manufacturer's instructions. All experiments were performed 48 h after transfection.

Whole-cell patch recordings of HEK293T cells were obtained at room temperature (21–25°C). Cells were bathed in the following solution (mM): 140 NaCl, 4 KCl, 2 CaCl₂, 2 MgCl₂, 10 Na-HEPES and 5 d-glucose, adjusted to pH 7.4 with NaOH. The osmolarity was adjusted with mannitol to ∼315 mosmol/l. Pipettes were pulled from borosilicate glass capillaries (Sutter Instruments, USA) using a Narishige micropipette puller (PC-10; Narishige Instrument Company, Japan). All experiments to test the impact of DrVSP (0.5 µg cDNA per 35 mm dish) were carried out with 100% series resistance compensation using a VE-2 amplifier (Alembic Instruments, Canada) equipped with an Rs Compensator (Sherman et al., 1999).

We applied the traditional voltage protocol as in the original description of the M-current (Adams et al., 1982). This protocol takes advantage of the lack of inactivation of the M-channels, and allows removing potential ‘invasion’ of the signal by other inactivating currents. Our protocol to study the sensitivity to DrVSP activation is similar to that used by Hille and collaborators (Falkenburger et al., 2010). DrVSP was activated by a 200 ms jump to +100 mV, and the voltage was returned to the holding potential of −20 mV. Jumps to −110 mV, beyond the K⁺ reversal potential, were applied to close the channels. Upon returning to the holding potential, an instantaneous current jump (corresponding to leak current) followed by a slowly developing outward relaxation (corresponding to the opening of the M-channels) was recorded. The size of the outward relaxation before and after the +100 mV jump was used to estimate the effect of DrVSP activation on M-current size (Adams et al., 1982; Villarroel, 1994).

Pipettes were filled with an internal solution containing (mM): 125 KCl, 5 MgCl₂, 5 EGTA, 5 Na₂ATP and 10 K-HEPES, adjusted to pH 7.2 with KOH and the osmolarity adjusted to ∼300 mosmol/l with mannitol (Gómez-Posada et al., 2010). The amplitude of the Kv7 current was defined as the peak difference in current relaxation measured at −30 mV after 500–1500 ms pulses to −110 mV (all channels closed) and to +30 mV (all channels opened). The data were acquired and analyzed using pCLAMP software (version 8.2), normalized in Excel (Microsoft, Madrid, Spain) and plotted using SigmaPlot (SPSS Corporation, Madrid, Spain). Data are shown as means±s.e.m. The differences between the means were evaluated using Student's unpaired t-test, where values of P≤0.05 were considered significant. The number of cells in each experiment is indicated in parentheses in the figures. The results are from two or more independent batches of cells. In all figures *, ** and *** indicate significance at P<0.05, P<0.01 and P<0.001, respectively.

Single transiently transfected HEK293 cells were maintained in buffer solution composed of (mM): 140 NaCl, 5 KCl, 1 MgCl₂, 2 CaCl₂, 10 glucose and 10 Na-HEPES; pH 7.4 at room temperature. For CFP-CaM binding to YFP-tagged KV7.2 C-terminal region (devoid of any membrane anchoring region) the transfection ratio used was 1:5 with a total of 0.6 µg DNA per M35 dish. Control experiments expressing CFP alone or CFP together with YFP at a 1:5 ratio demonstrated that contamination of the CFP emission signal due to direct YFP excitation at 405 nm was <1% (Fig. S1). For assembly experiments, the ratio was 1:1 (1 µg of each FCP-tagged ABCD and 2 µg of CaM or 2 µg of empty pcDNA3.1 his/myc C vector). Images were recorded 48 h after transfection using a Nikon D Eclipse TE2000-U fluorescence microscope (Nikon Instruments, Tokyo, Japan) equipped with a confocal scanning head and a spectral detector module. Images were captured using a 60× oil objective (numerical aperture 1.43), with the pinhole opened (150 µm) and using the 405 nm laser line (Coherent, Santa Clara, CA, USA) or the 488 nm line (Melles-Griot, Rochester, NY, USA) for direct CFP or YFP excitation, respectively. Cytosolic regions of interest (ROI) from cells displaying a clear signal after excitation with the 488 nm laser line were included in the analysis. The spectral detector allows simultaneous recording of 32 images, each registering a 5 nm band of the spectrum, covering 450–610 nm. After spectral unmixing, the area under the spectra was measured and a FRET index was calculated as FRET index=YFP₄₀₅/CFP₄₀₅.

Fluorescence intensities of YFP and CFP were recorded from single cells using a Leica DMI6000B inverted epifluorescence microscope and a dual emission photometric system (Till Photonics, Gräfelfing, Germany). Excitation was done at 436±7.5 nm or 500±7.5 nm, applied at 2 Hz using a Polychrome V as light source. Epifluorescence emission was detected by a photodiode, digitized (Mini Digi1B; Molecular Devices) and acquired using Clampex 10 software (Molecular Devices). FRET filter cubes were (excitation, long-pass beam-splitter, emission): CFP (ET436/20x; T455lp; ET480/40m), YFP (ET500/20x; T515lp; ET535/30m) and FRET (ET436/20x; T455lp; ET535/30m) (Chroma Technology). e-FRET efficiencies (E_D) were determined using E_D=(S_FRET−R_D1×S_CFP−R_A×S_YFP)/(S_FRET−R_D1×S_CFP−R_A×S_YFP+G×S_CFP). S_FRET, S_CFP and S_YFP denote fluorescence intensities derived from measurements in individual cells co-expressing YFP- and CFP-tagged proteins with the respective filter cube (excitation, emission): S_FRET (500±7.5 nm, 535±15 nm), S_CFP (436±7.5 nm, 480±20 nm) and S_YFP (500±7.5 nm, 535±15 nm). R_D1 and R_A are experimentally predetermined constants from measurements applied to single cells expressing only CFP- or YFP-tagged molecules that correct for donor bleed trough and acceptor cross excitation. G factor represents the ratio of sensitized acceptor emission to quenched donor emission due to FRET, and was experimentally determined to be 4.13 using CFP fused to YFP with linkers of varying length (4, 40 and 80 amino acids). The CFP/YFP ratio was between 1:5 and 5:1, which is expected to yield reliable results.

Fluorescence intensities of YFP and CFP were recorded from single cells using a Leica DMI6000B inverted epifluorescence microscope and a 914 photomultiplier detection system (PTI, Canada). Excitation was done at 436±7.5 nm or 500±7.5 nm, applied with a DeltaRam X monochromator including an arc lamp (PTI, Canada) as light source. Epifluorescence emission was acquired using FelixGX software (PTI, Canada).

FRET efficiency for d-FRET (E_A), which is defined as the fractional increase in YFP emission caused by FRET, was calculated using E_A={[S_FRET−(R_D1)(S_CFP)]/[(R_A1)(S_YFP)]−1}×(ε_YFP/ε_CFP); S_FRET, S_CFP and S_YFP denote fluorescence intensities derived from measurements in individual cells co-expressing YFP- and CFP-tagged proteins with the respective filter cube.

FRET data were analyzed using Clampfit 10.0 (Molecular Devices), and Microsoft Office Excel software. Data plotting, curve fitting and statistical analysis were performed using Sigma-Plot 12 software (SPSS Corporation). All values are presented as means±s.e.m. for the indicated number (n) of experiments. Student's t-test was used to compare means of two groups from data with a normal distribution. Statistical analysis was performed using SigmaPlot software. *P<0.05, **P<0.01 and ***P<0.001 indicate significance.