Nội dung

Inhibition of human ether à go-go potassium channels by Ca2+ ⁄calmodulin binding to the cytosolic N- and C-termini Ulrike Ziechner1, Roland Schönherr1, Anne-Kathrin Born1, Oxana Gavrilova-Ruch1, Ralf W. Glaser2, Miroslav Malesevic3, Gerhard Küllertz3 and Stefan H. Heinemann1 1 Institute of Molecular Cell Biology, Research Unit Molecular and Cellular Biophysics, Friedrich Schiller University Jena, Germany 2 Department of Biochemistry and Biophysics, Friedrich Schiller University Jena, Germany 3 Max Planck Research Unit Enzymology of Protein Folding, Halle, Germany Keywords Calmodulin; calcium; potassium channel; fluorescence correlation spectroscopy; patch clamp Correspondence S. H. Heinemann, Institute of Molecular Cell Biology, Molecular and Cellular Biophysics, Friedrich Schiller University Jena, Drackendorfer Str. 1, D-07747 Jena, Germany Fax: + 49 3641 9 32 56 82 Tel.: + 49 3641 9 32 56 80 E-mail: stefan.h.heinemann@uni-jena.de (Received 17 October 2005, revised 2 December 2005, accepted 10 January 2006) doi:10.1111/j.1742-4658.2006.05134.x Human ether à go-go potassium channels (hEAG1) open in response to membrane depolarization and they are inhibited by Ca2+ ⁄ calmodulin (CaM), presumably binding to the C-terminal domain of the channel subunits. Deletion of the cytosolic N-terminal domain resulted in complete abolition of Ca2+ ⁄ CaM sensitivity suggesting the existence of further CaM binding sites. A peptide array-based screen of the entire cytosolic protein of hEAG1 identified three putative CaM-binding domains, two in the C-terminus (BD-C1: 674–683, BD-C2: 711–721) and one in the N-terminus (BD-N: 151–165). Binding of GST-fusion proteins to Ca2+ ⁄ CaM was assayed with fluorescence correlation spectroscopy, surface plasmon resonance spectroscopy and precipitation assays. In the presence of Ca2+, BD-N and BD-C2 provided dissociation constants in the nanomolar range, BD-C1 bound with lower affinity. Mutations in the binding domains reduced inhibition of the functional channels by Ca2+ ⁄ CaM. Employment of CaM-EF-hand mutants showed that CaM binding to the N- and C-terminus are primarily dependent on EF-hand motifs 3 and 4. Hence, closure of EAG channels presumably requires the binding of multiple CaM molecules in a manner more complex than previously assumed. Ether à go-go (EAG) potassium channels are activated by membrane depolarization. They are widely expressed in neuronal tissue suggesting a role in neuronal signaling. In addition, they are expressed in various cancer cell lines and primary tumor material where they are postulated to play a role in regulating cell growth [1–3]. One of the most remarkable features of EAG channels is their regulation by intracellular Ca2+. Unlike Ca2+-activated potassium channels of the big-conductance (BK) or intermediate- and smallconductance (IK ⁄ SK) family, EAG channels are inhibited when the intracellular Ca2+ concentration rises above about 100 nm [4,5]. In a previous study we have shown that this unique inhibition of the channel by Ca2+ is mediated by the Ca2+-binding protein calmodulin (CaM) [6]. In excised inside-out membrane patches the Ca2+ sensitivity of human EAG1 channels disappeared, but could be reconstituted by administration of recombinant CaM. Assaying CaM binding to GST-fusion proteins of the C-terminal domain of hEAG1 channels, a binding site was localized within amino-acids 673–770. This stretch of residues, harboring a putative amphiphilic CaM-binding helix as mutations in the CaM-binding motif (F714SÆF717S or R711QÆR712QÆR716QÆR718Q), prevented Ca2+ ⁄ CaM binding to this C-terminal domain and produced channels that were insensitive to Ca2+ ⁄ CaM [6]. Abbreviations apoCaM, Apocalmodulin; BD, binding domain; CaM, calmodulin; EAG, ether à go-go; FCS, fluorescence correlation spectroscopy; hCaM, human calmodulin; hEAG, human ether à go-go; MBP, maltose-binding protein; SPR, surface plasmon resonance; TMR, tetramethylrhodamine. 1074 FEBS Journal 273 (2006) 1074–1086 ª 2006 The Authors Journal compilation ª 2006 FEBS U. Ziechner et al. The binding of CaM to the C-terminal domain of hEAG1 channels is dependent on [Ca2+]. In this respect hEAG channels behave very differently to the Ca2+-activated potassium channel SK2. SK2 a-subunits form a very stable heteromeric association complex with the C-terminal domain of Ca2+-free CaM, i.e. apoCaM. The activation of these channels is mediated by conformational changes in the CaM ⁄ SK2 complex induced by binding of at least one Ca2+ ion to the EF-hand motifs in the N-terminal domain of CaM [7,8]. In the deactivated state of rSK2 channels the rSK2-CaMBD ⁄ CaM complex is a monomer. However, Ca2+ binding to the CaM-N-lobe caused by increasing the intracellular Ca2+ concentration leads to exposure of regions in the CaM ⁄ SK2 complex that induce an interaction between neighboring SK2-CaM binding domains mediated by Ca2+ ⁄ CaM. This interaction then results in opening of these channels [9,10]. The mechanism of CaM-mediated closure of hEAG1 channels is far from understood mainly because of the lack of structural information on the cytosolic channel domains in complex with CaM. Initiated by the finding that hEAG1 channels lacking the entire N-terminal cytosolic domain are insensitive to Ca2+ ⁄ CaM we performed a systematic screen for additional putative CaM binding domains in the channel protein. Indeed, we found two more sites, one in the C-terminus neighboring to the previously identified site and one domain in the N-terminus of hEAG1. The latter one binds CaM only in the presence of Ca2+ and is functional in a sense as point mutations in this domain abolished Ca2+-sensitivity of the channel. Experiments with CaM-EF-hand mutants testing for interaction with the functional C- and N-terminal binding sites of hEAG1 channels highlighted the importance of Ca2+-binding ability of EF-hand motifs 3 and 4 in CaM for its interaction with the channel. Calmodulin binding to hEAG1 channels by voltage. To test for such a scenario, channel mutants with strongly altered gating properties might be insightful. As it is known that N-terminal deletions of the EAG protein have strong effects on the voltage dependent gating steps [11], we constructed a mutant of hEAG1 in which the entire N-terminal domain was deleted (hEAG1D2-190 ¼ hEAG1DN). As shown in Fig. 1B, such channels do form functional channels as for EAG channels the assembly domain resides in the C-termini [12] and not in the N-termini as for KV channels. However, the gating of these hEAG1DN channels is strongly altered with respect to the wild-type channels. The prominent feature of hEAG1DN channels is their activation at low voltages and their slow deactivation kinetics. Fitting a Boltzmann distribution to the normalized currentvoltage relationships yielded a half-maximal activation voltage (Vm) for the wild-type subunits of A C B D E Results Relevance of the N-terminus for Ca2+ ⁄ CaM regulation of hEAG1 channels As shown previously [6], EAG channels are inhibited by Ca2+ ⁄ CaM binding to the C-terminal domain. However, there is no insight into the molecular mechanism by which bound Ca2+ ⁄ CaM makes the EAG channels close. An obvious possibility is that the bound complex directly occludes the ion pore or affects the cytosolic gate formed by the S6 segments. Alternatively, bound Ca2+ ⁄ CaM might allosterically alter the gating mode of EAG channels such that the respective subunits are inhibited from being activated Fig. 1. Ca2+ ⁄ CaM sensitivity of wild-type hEAG1 channels and mutant hEAG1DN (A,B) Recordings from inside-out patches containing wild-type hEAG1 (A) and hEAG1DN channels (B) in response to depolarizing steps from )140 mV to +20 mV in steps of 10 mV, from a holding potential of )160 mV (C) Current-voltage relationships of the data shown in (A) and (B); currents were measured at the end of the test pulses. The continuous curves describe the voltage-dependent activation of the channels according to Eqn 4. (D,E) Current recordings of hEAG1 (D) and hEAG1DN channels (E) in response to test depolarizations to +50 mV (D), and to )20 and )80 mV (E) under control conditions, i.e. with Ca2+-free bath solution, upon application of the indicated concentrations of hCaM in the presence of 1 lM free Ca2+ solution, and after washing off the CaM with Ca2+-free solutions. FEBS Journal 273 (2006) 1074–1086 ª 2006 The Authors Journal compilation ª 2006 FEBS 1075 Calmodulin binding to hEAG1 channels U. Ziechner et al. )55.4 ± 2.9 mV and for hEAG1DN of )129.6 ± 3.3 mV (n ¼ 7; P ¼ 1.2 10)9), while the slope factors characterizing the voltage sensitivities of the gating steps (km) did not change significantly (30.8 ± 1.6 mV versus 28.9 ± 2.2 mV, n ¼ 7, P ¼ 0.48). hEAG1DN channel subunits activate at about 75 mV lower voltages than the wild-type subunits (also see Fig. 1C) and, hence, have a stronger tendency of opening that needs to be overcome by Ca2+ ⁄ CaM. As a control, we exposed inside-out membrane patches with wild-type hEAG1 channels to 100 nm hCaM in 1 lm Ca2+ solutions (Fig. 1D). This resulted in complete block of EAG-mediated K+ currents and this effect was readily reversible when washing the patches with Ca2+-free solutions. The same procedure applied to hEAG1DN channels had no effect (not shown). Even 100-fold higher CaM concentrations (Fig. 1E) did not result in closure of hEAG1DN channels, neither at low ()80 mV) nor at high ()20 mV) voltages. the abolition of CaM regulation. However, it may also be that the deleted N-terminal domain takes part in CaM binding. Thus, we subsequently searched for additional CaM-binding sites in the cytosolic domains of the EAG channel protein. For that purpose we generated peptide arrays on a cellulose membrane containing peptides of 15 residues length scanning the entire N-terminal (between aminoacids 1–206 in steps of 3) and C-terminal (amino-acids 480–962 in steps of 3 and amino-acids 645–895 in steps of 2) cytosolic domains of hEAG1. Binding of BODIPY FL-labeled hCaM in the presence of 25 lm free Ca2+ was assayed. As shown in Fig. 2A, dots with different fluorescence intensity indicate the amount of bound CaM. The fluorescence change (DFI) with respect to the array prior to CaM treatment was analyzed for all spots and is plotted in Fig. 2D normalized to the maximum fluorescence increase as a function of the residue number (corresponding to the center of the peptides). There are three sequence regions with a very strong fluorescence intensity increase over an extensive sequence range caused by specific binding of labeled hCaM to these peptides (Figs 2C,D). The most distal of these regions encompasses the previously identified Screening for additional CaM-binding sites in hEAG1 cytosolic domains The results shown above may indicate that the strongly altered gating of hEAG1DN channels is responsible for A B C D Fig. 2. Peptide array-based screen for putative CaM interaction sites. (A) Fluorescence images of a cellulose peptide array before (left) and after (right) exposure to BODIPY FL-hCaM. Dark spots indicate CaM binding. The top frame marks the spots encompassing residues 1–205, the bottom frame the C-terminal residues 480–962. (B) Topology of the hEAG1 channel subunit consisting of six transmembrane segments and extended cytosolic N- and C-terminal domains. Functional channels are formed by 4 of such subunits involving the C-terminal assembly domain. (C) Magnified sections of the peptide array: The numbers indicate the first residue of the 15-mer peptides. In the first row (showing BD-N), the peptide sequences are advanced by 3 residues, in the other rows (showing BD-C1 and BD-C2) sequences are advanced by 2 residues. (D) Increase in fluorescence intensity (DFI in percentage) of peptide spots after BODIPY FL-hCaM binding relative to the intensity before CaM incubation plotted against the residue number corresponding to the center of the 15-mer peptides (residues scanned: 1–205 and 480–962). The sequence regions with the highest increase in fluorescence intensity (> 90% of maximum DFI, dashed line) were considered strong CaM-binding domains, here termed BD-N, BD-C1, and BD-C2. 1076 FEBS Journal 273 (2006) 1074–1086 ª 2006 The Authors Journal compilation ª 2006 FEBS U. Ziechner et al. Ca2+ ⁄ hCaM binding site, here called calmodulin-binding domain C2 (BD-C2, Figs 2 and 3C). The other two regions contain a potential binding site in the N-terminus (BD-N; Figs 2 and 3A) and in the C-terminus (BD-C1; Figs 2 and 3B). There are additional peaks detected on this peptide array membrane. However, relative to the already mentioned potential binding domains they show fluorescence intensity changes not greater than 90% and the peaks do not exhibit extended plateaus. A B C D Calmodulin binding to hEAG1 channels In Fig. 3, the sequences of the three potential binding domains are highlighted. The boxes indicate the sequence regions in which the strongest fluorescence was detected. The hydrophobic and basic residues, both expected to be relevant for CaM binding, are marked in bold and gray, respectively. For the N-terminal peptides, the fluorescence intensity increases until R162 is inside the peptide sequence, stays at a high level (DFI > 90%) as long as the peptides contain the amino acids 151–162, and decreases when the peptides lack F151 (box in Fig. 3A). Thus, we can notice that BODIPY FL-labeled hCaM binds to peptides with a 1-8-14 motif, but the hydrophobic residues V164 and L165 of this motif do not seem to be important for a strong interaction of hCaM. At most, these residues should have a stabilizing effect on the interaction between hCaM and the hEAG1 N-terminus. Among the C-terminal peptide spots there are fluorescence signals (DFI > 90%) in positions where peptides contain the amino acids R675–I683 (Fig. 3B) and R711–K721 (Fig. 3C). We previously showed [6] that replacements of either the basic arginines or both phenylalanines (underlined in Fig. 3C) in BD-C2 led to Ca2+ ⁄ CaM insensitivity of hEAG1 channels. Ca2+-dependent binding of hCaM to the N-terminal binding site Fig. 3. Comparison of hEAG1 sequence domains that were identified as CaM binding sites. (A-C) hEAG1 channel sequences, which provided peptides with the highest binding of BODIPY FL-labeled hCaM and which were then synthesized as peptides for FCS measurements. 15-mer peptides on the array that contained the sequence inside of boxes provided highest increase in fluorescence intensity (DFI > 90%) after assay with BODIPY FL-labeled hCaM on the peptide array membrane. Hydrophobic amino acids are bold basic amino acids gray letters. (A) hEAG1(145–165) covers the potential 1-8-14 motif in the N-terminus. Residue 145 is a cysteine that was used for labeling of this synthetic peptide for FCS measurements. (B) hEAG1(671–688) contains the potential 1-5-10 motif in the C-terminus. The boxed cysteine in the beginning does not belong to the hEAG1 sequence; it was introduced in the synthetic peptide for labeling. (C) hEAG1(707–725) is around the already established binding site in the C-terminus. The boxed cysteine has the same meaning as in B. Replacement of the underlined residues in hEAG1 channels were previously shown to reduce Ca2+ ⁄ hCaM sensitivity. (D) Nomenclature of the GST-fusion proteins used in this study. The fragments marked in black bound CaM with high affinity in the FCS interaction assay; gray fragments did not show significant binding. Because the peptides were spotted onto the cellulose membrane at high concentration, the array assay may yield some false-positive hits. Therefore, we confirmed in GST-pull–down assays the association between the N-terminus of hEAG1 channel and hCaM. Fusion proteins of GST linked to either hEAG1(1–206) (GSTN) or hEAG1(674–867) (GST-C34) were bound to glutathione-sepharose beads and exposed to hCaM either in absence or presence of 25 lm free Ca2+. As shown in Fig. 4, both GST-N and GST-C34 bound hCaM, but only in the presence of Ca2+ (seen as weak bands at
16.8 kDa). As CaM bands were absent in GST controls, a nonspecific binding of CaM to the sepharose beads or to GST can be excluded. Hence, also the N-terminal domain of hEAG1 channels harbors a functional binding site for CaM that binds CaM in a Ca2+-dependent manner as for the site in the C-terminus. CaM affinity to fusion proteins and peptides of hEAG1 channels To quantify the interaction strength between Ca2+ ⁄ hCaM and the N-terminal binding site in hEAG1 FEBS Journal 273 (2006) 1074–1086 ª 2006 The Authors Journal compilation ª 2006 FEBS 1077 Calmodulin binding to hEAG1 channels U. Ziechner et al. Fig. 4. A GST fusion protein containing the hEAG1 N-terminus binds hCaM in a Ca2+-dependent manner. The GST-fusion proteins GST-N and GST-C34 as well as GST alone was bound to glutathione-sepharose and subsequently incubated with purified hCaM, either in presence or absence of 25 lM free Ca2+ ions. After washing, the retained proteins were eluted with SDS ⁄ PAGE loading buffer. The figure shows the SDS ⁄ PAGE separation gel of the retained proteins and a protein molecular weight marker. Protein bands were stained with Coomassie blue. For both, GST-N and GST-C34, an hCaM band is only seen in the presence of Ca2+ (lanes 5, 6 and 9, 10). In the absence of Ca2+ (lanes 3, 4 and 7, 8) and for the GST control (lanes 1 and 2) there was no retention of hCaM detectable. channels we performed FCS measurements. MCGS(H)6-hCaM was labeled with either TMR or BODIPY FL and its interaction with GST-N was measured. In the absence of Ca2+, no binding between apoCaM and GST-N was detected (dissociation constant, KD>5 lm). With 25 lm free Ca2+ in the buffer, there was a strong interaction between Ca2+ ⁄ hCaM and GST-N showing dissociation constants of 184 ± 36 nm for TMR-MCGS-(H)6-hCaM (Table 1) and 194 ± 31 nm for BODIPY FL-MCGS-(H)6hCaM. FCS measurements using (H)6-MBP-N instead of GST-N provided similar results for both, Ca2+-free (KD > 5 lm) and Ca2+-containing solutions (KD ¼ 173 ± 22 nm). Nonspecific binding could be excluded as TMR- or BODIPY FL-labeled Ca2+ ⁄ hCaM did not bind to GST alone (KD > 5 lm). Mutations in the potential N-terminal Ca2+ ⁄ CaM binding site (BD-N), which replaced hydrophobic residues by asparagines, reduced the interaction strength. Both GST-N_LN and GST-N_FNÆLN (for mutant nomenclature see Table 1) showed a weaker binding to TMR-MCGS-(H)6-hCaM in the presence of 25 lm free Table 1. Binding of MCGS-(H)6-hCaM to fusion proteins of cytosolic hEAG1 domains. Dissociation constants KD for binding of TMR-MCGS(H)6-hCaM to the indicated constructs in the absence and presence of 25 lM free Ca2+ determined by confocal FCS. ND ¼ not determined. KD (nM+) Constructs Short name +Ca2+ –Ca2 GST control GST-hEAG(1–206) (H)6-MBP-hEAG(1–206) GST-hEAG(1–206)_L154N GST-hEAG(1–206)_F151NLÆ154N GST-hEAG(1–206)_F151SÆL154SÆL158S GST-hEAG(674–867) GST-hEAG(674–867)_F714SFÆ717S GST-hEAG(649–867) GST-hEAG(649–867)_F714SÆF717S GST-hEAG(649–867)_F714SÆF717SÆL674NÆI678N GST-hEAG(649–867)_F714SÆF717SÆR677NÆR681N GST-hEAG(649–867)_F714SÆF717SÆR675NÆR677NÆR681NÆK682N GST-hEAG(649–867)_R675NÆR677NÆR681NÆK682N GST-hEAG(649–700) GST-hEAG(684–867) GST-hEAG(479–673) GST-hEAG(771–962) GST GST-N (H)6-MBP-N GST-N_LN GST-N_FNÆLN GST-N_FSÆLSÆLS GST-C34 GST-C34_ FSÆFS GST-C2B34 GST-C2B34_FSÆFS GST-C2B34_FSÆFSÆLNÆIN GST-C2B34_FSÆFSÆRNÆRN GST-C2B34_FSÆFSÆRNÆRNÆRNÆKN GST-C2B34_RNÆRNÆRNÆKN GST-BDC1 GST-BDC2 GST-C12 GST-C45 > 5000 184 ± 36 173 ± 22 577 ± 75 4560 ± 593 > 5000 127 ± 15 735 ± 114 116 ± 44 661 ± 117 714 ± 52 1340 ± 241 1960 ± 335 146 ± 11 > 5000 136 ± 19 > 5000 > 5000 > 5000 > 5000 > 5000 ND ND > 5000 1283 ± > 5000 1343 ± > 5000 > 5000 ND > 5000 1291 ± > 5000 1239 ± > 5000 > 5000 1078 176 213 139 167 FEBS Journal 273 (2006) 1074–1086 ª 2006 The Authors Journal compilation ª 2006 FEBS U. Ziechner et al. Ca2+: The KD was increased threefold (577 ± 75 nm) and 25-fold (4560 ± 593 nm), respectively. Replacement of three hydrophobic amino acids in this putative 1-8-14-Ca2+ ⁄ CaM binding motif by the hydrophilic serine led to a complete loss of binding (KD > 5 lm, Table 1). As the peptide screen identified a putative new CaM binding site in the C-terminus (BD–C1), the interaction between TMR-labeled MCGS-(H)6-hCaM and GSTlinked fusion proteins of hEAG1 C-terminal fragments was also quantified (Table 1). The wild-type fragments GST-C34, GST-C2B34, and GST-BDC2 all bound CaM in the presence of 25 lm Ca2+ with KD on the order of 100–150 nm. These fractions have in common the previously identified CaM-binding domain BD-C2. In the absence of Ca2+ these BD-C2 fragments bound CaM about 10-fold less efficiently. However, compared to the N-terminal CaM binding site, the C-terminal site clearly has some finite affinity to apoCaM. The fragment encompassing only BD-C1 (GSTBDC1) did not bind CaM with high affinity (KD > 5 lm), regardless of the Ca2+ concentration. The same holds true for the fragments up- and downstream of the putative binding regions, i.e. GST-C12 and GST-C45 (Table 1). Mutations F714SÆF717S in BD-C2 reduced the KD in Ca2+-containing solutions to about 660 nm; in the absence of Ca2+ no binding could be measured with FCS. In this background, additional mutations in BDC1 only weakly affected CaM binding: L674NÆI678N had no detectable effect, whereas R677NÆR681N and R675NÆR677NÆR681NÆK682N increased the KD to about 1.3 and 1.9 lm, respectively. The latter mutation in the background of GST-C2B34 had no effect. From these data one can conclude that BD-C1 is not forming a functional CaM-binding site by itself. However, residues in this domain may interfere with BD-C2 and may affect CaM binding to the channel protein. These results could be confirmed by measuring the association of TMR-labeled synthetic peptides encoding the sequences shown in Fig. 3. The peptide representing BD-N resulted in a KD of 45 ± 9 nm and peptide BD-C2 in 223 ± 44 nm, while there was no binding detectable for peptide BD-C1. However, none of these peptides bound MCGS-(H)6-hCaM in the absence of Ca2+ including the peptide BD-C2. This peptide, encompassing amino acids 707–725, may be too short to bind apoCaM. By contrast, the wild-type fragments GST-C34, GST-C2B34, and GST-BDC2 contain the complete domain sufficient for a weak affinity to apoCaM. As BD-C1 seemed to bind CaM strongly on the peptide array, we reinvestigated the corresponding peptide using Cy5 as an alternative Calmodulin binding to hEAG1 channels fluorescence label. In this configuration the BD-C1 peptide bound CaM in the presence of Ca2+ with a KD of 327 ± 112 nm; binding of BD-N and BD-C2 were not altered with labeled with Cy5 instead of TMR. In FCS measurements fluorescently labeled peptides of both, the N-terminal or C-terminal binding sites in hEAG1 channels, could be displaced from binding to Ca2+ ⁄ hCaM not only by an excess of the same, unlabeled peptide, but also by an excess of the unlabeled peptide from the other binding sites (data not shown). The validity of the results obtained for the binding of CaM to the N-terminus of hEAG1 channels was confirmed employing surface plasmon resonance spectroscopy. Such measurements showed a specific high-affinity binding of (H)6-MBP-hEAG(1–206) to CaM in the presence of 25 lm Ca2+. However, as in FCS measurements, in the absence of Ca2+ no specific binding (KD > 15 lm) was detected. Binding was very fast such that association and dissociation were masstransport controlled and kinetic rate constants could not be determined. The equilibrium binding response signal of (H)6-MBP-hEAG(1–206) protein with immobilized CaM could be described by Eqn 3 with a dissociation constant of 148 ± 8 nm (see Experimental procedures). The fit did not show significant systematic deviations in the range between 0.016 and 10 lm (not shown). Functional relevance of CaM binding sites Mutations inside BD-C2 result in channels insensitive to Ca2+ ⁄ hCaM (e.g. hEAG_F714SÆF717S [6]). As mutations in BD-C1 showed some impact on the CaM binding to the C-terminal binding complex, we also tested whether such mutations affect the channel regulation by CaM. Mutant hEAG_R675NÆR677NÆ R681NÆK682N, destroying the putative CaM-binding domain C1, was expressed in Xenopus oocytes. In the presence of 1 lm free Ca2+ a preparation of hCaM, reducing the currents through wild-type channels to 20.6 ± 6.3% (n ¼ 7), only reduced the currents mediated by the mutant channels to 89.7 ± 2.3% (n ¼ 6). To test for the functional importance of BD-N, the double mutant hEAG_F151NÆL154N was assayed. 200 nm hCaM in 1 lm Ca2+ solutions to inside-out patches with wild-type hEAG1 channels resulted in strong current reduction (to 13.8 ± 3.4%, n ¼ 11). The mutant was much less sensitive: Application of 500 nm hCaM only reduced the current to 81.9 ± 5.3% (n ¼ 7). In Fig. 5 A the direct comparison of wild-type, hEAG_F151NÆL154N and hEAG_R675NÆR677NÆ FEBS Journal 273 (2006) 1074–1086 ª 2006 The Authors Journal compilation ª 2006 FEBS 1079 Calmodulin binding to hEAG1 channels U. Ziechner et al. R681NÆK682N with respect to the effect of CaM in 1 lm Ca2+ using the same batch of oocytes and the same batch of CaM is shown. The voltage dependence of channel activation in the mutants was left-shifted with respect to the wild type, but had no resemblance of mutant hEAG1DN. Vm values were: )21.7 ± 2.1 mV for hEAG_F151NÆL154N and )22.4 ± 3.5 mV for hEAG_R675NÆR677NÆR681NÆK682N. As can be seen in Fig. 5A, activation kinetics of mutant hEAG_ R675NÆR677NÆR681NÆK682N seems to be particularly slow. We thus analyzed the activation time course with single exponentials at +40 mV and obtained the following time constants: wild type, 8.4 ± 1.5 ms (n ¼ 7); hEAG_F151NÆL154N, 4.7 ± 1.6 ms (n ¼ 7); hEAG_R675NÆR677NÆR681NÆK682N, 104 ± 12.7 ms (n ¼ 8). Contribution of CaM EF-hand motifs The results shown above indicate that the N-terminus of hEAG1 channels only binds CaM in the presence of Ca2+, while the C-terminus has some finite affinity to CaM even in Ca2+-free solutions. The reason for this apparent difference might reside in a different A contribution of the Ca2+-binding sites (EF-hands) in hCaM. Therefore, we compared wild-type hCaM with EF-hand mutants, in which the relevant aspartates were replaced with alanines, with respect to the binding to synthetic TMR-labeled peptides (see Fig. 3A–C) coding for the putative binding sites in hEAG1. In the presence of 25 lm free Ca2+, the aspartateto-alanine mutations in the N-terminal half of CaM (hCaM-EF12) increased the dissociation constant for binding of the N1-peptide by a factor of 9 (KD ¼ 393 ± 81 nm compared to 45 ± 9 nm for the wildtype peptide); EF-hand mutations 3 and 4 abolished binding (KD > 5 lm for hCaM-EF34 and hCaMEF1234). None of the CaM constructs bound to the peptide in the absence of Ca2+. For the TMR-labeled peptide C2, EF-hand mutations 1 and 2 increased the KD by about a factor of seven (to 1560 ± 190 nm compared to 223 ± 44 nm for the wild type); EF-hand mutations 3 and 4 resulted in a complete loss of binding (KD > 5 lm for hCaM-EF34 and hCaM-EF1234). EF-hand mutants of hCaM were also employed for functional tests on hEAG1 channels expressed in Xenopus oocytes. The channel-inhibiting potency of wildtype hCaM and the EF-hand mutants were assayed in inside-out patches under repetitive channel activation at +50 mV in solutions containing 1 lm free Ca2+. While 270 nm hCaM reduced the current to 9 ± 3% (n ¼ 4), 1080 nm hCaM-EF12 reduced the current to 34 ± 10% (n ¼ 4) and 1080 nm hCaM-EF34 had no effect: 97 ± 8% (n ¼ 3) (Fig. 5B). hCaM-EF1234 had no effect either (n ¼ 4, not shown). Discussion Use of FCS measurements and peptide arrays for CaM binding site characterization B Fig. 5. Functional impact of mutations in the N-terminal CaM-binding site and in the EF-hand lobes of hCaM. (A) Normalized current traces obtained from inside-out patches with wild-type hEAG1 channels (wt), mutant F151NÆL154N (FL) and mutant hEAG_R675NÆ R677NÆR681NÆK682N (RRRK). ‘–’ and ‘+’ indicate absence and application of about 1 lM CaM in 1 lM Ca2+ solutions. (B) Similar experiments as in (A) for a single patch with hEAG1 channels before (–) and after (+) application of the indicated type of CaM in the presence of 1 lM free Ca2+. Pulses to +50 mV were given; traces in (B) are shown in a chronological order, scaled to control currents in order to eliminate effects of current run-down. 1080 For the mapping of CaM binding sites at hEAG1 channels the entire cytosolic sequence of the channel was presented as an array of short peptides, attached to a cellulose membrane. Detection of CaM binding was realized by BODIPY FL labeling of recombinantly produced MCGS-(H)6-hCaM. Analysis of this assay resulted in unambiguous detection of CaM binding domains when setting a stringent detection threshold (Fig. 2) and by assuming that a CaM binding domain has to have a length of at least around 15 residues. For quantitative estimates of the binding of CaM and CaM mutants to peptides and fusion proteins we performed FCS measurements in the autocorrelation mode. Using fusion proteins as unlabelled interaction partner, the molecular weight ratio of the fluorescently labeled molecule and the formed binding complex was FEBS Journal 273 (2006) 1074–1086 ª 2006 The Authors Journal compilation ª 2006 FEBS U. Ziechner et al. smaller than 1 : 6 in all cases. Therefore, great caution had to be taken in validating the results. For that reason we also employed surface plasmon resonance spectroscopy for measuring the interaction of immobilized CaM with (H)6-MBP-N in the presence of Ca2+ yielding approximately the same dissociation constant (
150 nm) as determined with FCS (
170 nm, Table 1). Quantitative estimates of the binding constant between peptides encoding the minimal CaM-BDs and fragments of the hEAG1 protein, encompassing more than just the binding site, resulted in a fourfold stronger binding of the BD-N peptide and a twofold weaker binding for the BD-C2 peptide. In the first case this could be due to a better accessibility of CaM to the site. In the latter case it may indicate that the peptide does not contain all critical residues necessary for complexing CaM with the channel protein. It cannot be excluded, however, that these differences are brought about by the specific fluorescence labels at the peptides. Such an effect was observed for BD-C1: The TMR-labeled peptide encompassing this domain did not bind Ca2+ ⁄ CaM, while Cy5 as label did not impede binding. Functional properties of N-terminally deleted hEAG1 channels Potassium channels of the EAG family assemble via the C-terminal ends of their a-subunits [12]. Therefore, complete removal of the N-terminus (hEAG1D2-190) still results in functional K+ channels as previously shown for rat EAG1 [11]. However, the activation threshold of these channels is shifted by about )75 mV and deactivation of the channels is slow. As a result, these channels are constitutively open around the normal resting potential of an excitable cell. In addition, hEAG1DN channels inactivate at higher voltages (not shown here). The inference to be drawn for the gating of EAG channels is that their N-terminal domains are essential for the stabilization of the closed channel state. As pointed out by Terlau et al. [11], a direct interaction of the N-terminal domain with the intracellular S4-S5 linker may take part in this stabilization. For the subject studied here one can conclude that CaM is unable to close the hEAG1 channels in the absence of the N-terminal domain. Mutations in the N- and C-terminal CaM binding sites also affected the gating parameters of the channels, but to a much smaller extent than the complete deletion of the N-terminus. Remarkably, mutations in the N-terminus (hEAG_F151NÆL154N) markedly slowed down channel activation further supporting the Calmodulin binding to hEAG1 channels notion that the N-terminal domain takes part in channel gating. The N-terminus of EAG channels harbors a functional CaM binding site Using a peptide array representing the entire cytosolic domains of human EAG1 channels we identified three regions where fluorescently labeled CaM binds. One binding site in the C-terminus (BD-C2) overlapped with that previously reported [6]. In its vicinity there is an additional motif with CaM-binding capacity (BD-C1). CaM binding to this site seems to be weak, but functional assays led us to conclude that this putative binding site is relevant for channel inhibition by Ca2+ ⁄ CaM under in vivo conditions. The third site, located in the N-terminus (BD-N), bound CaM in an in vitro assay but was also necessary for function. Mutagenesis of this site resulted in a loss of CaM sensitivity. Likewise, complete deletion of the N-terminus rendered the channels insensitive to CaM. Thus, it is very likely that the altered CaM sensitivity of the hEAG1DN mutant is not due to its strongly modified gating properties. Instead, binding of CaM to the N-terminal domain is needed for channel closure. However, it cannot be excluded that the specific interaction of the N-terminus with the voltage sensor elements of the channels is involved in CaM-mediated channel closure. Type of CaM binding motif Most CaM-BD peptides bind Ca2+ ⁄ CaM with a KD of about 0.1–100 nm [13] and this is also true for BD-N and BD-C2 of hEAG1 channels. Although quite diverse in structure, some types of CaM binding motifs can be specified [14]. The previously identified BD-C2 shows some resemblance to a 1-5-8-14 motif where position ‘14’ is occupied by an alanine (Fig. 3C). This may be the reason why this motif separated in the BDC2 peptide binds CaM in a strictly Ca2+-dependent manner; in the absence of free Ca2+ a weak binding was only detected if this BD-C2 was included in wildtype hEAG1 protein fragments. The N-terminal binding domain BD-N is of the 1-8-14 type and BD-C1 shows a 1-5-10 motif. While BD-N clearly falls into this category by binding CaM in a strictly Ca2+-dependent manner (
40 nm for the peptide), BD-C1 showed weaker binding to CaM, even in the presence of Ca2+, but mutagenesis of the BD-C1 site in the background of a fragment containing a mutated BD-C2 site reduced the binding affinity of CaM. This could mean that the BDC1 site contributes to the coordination of CaM that FEBS Journal 273 (2006) 1074–1086 ª 2006 The Authors Journal compilation ª 2006 FEBS 1081 Calmodulin binding to hEAG1 channels U. Ziechner et al. primarily binds BD-C2. Both sites together may form a complex CaM-binding structure. Importance of EF-hands 3 and 4 As BD-N bound CaM only in the presence of Ca2+, while BD-C2 also showed some weak binding of CaM in the absence of Ca2+, it was feasible that the interaction of CaM with the respective binding sites is dominated by different sets of EF-hand motifs. Assay of hCaM-EF12 and hCaM-EF34 mutants for binding to peptides coding for the binding sites showed that in both cases, i.e. for BD-N as well as for BD-C2, lack of EF-hand motifs 1 and 2 only resulted in a mild reduction of binding (factor 7 and 9, respectively), while EF-hand motifs 3 and 4 were essential. The same results were obtained for functional assays on hEAG1 channels expressed in Xenopus oocytes. Thus, both CaM binding domains seem to require a functional C-lobe of CaM. This result appears quite reasonable as the binding constant of Ca2+ to the C-lobe of CaM is up to 10-fold smaller than that for the N-lobe [13,15]. Upon binding of Ca2+ to the C-lobe, CaM undergoes a first conformational change, exposing a hydrophobic binding pocket and, hence, promoting binding to the target (e.g. [13,16,17]). Binding to the target can increase the Ca2+-binding affinity to CaM further [17,18]. In addition, subsequent binding of Ca2+ to the N-lobe then induces further conformational changes of CaM [19,20] that may trigger an increase of CaM affinity to the target protein. Thus, the importance of functional EF34 domains of CaM for inhibiting hEAG1 channels indicates that association of Ca2+-bound CaM to the channel is a necessary requirement. Interestingly, Ca2+ binding to the N-lobe facilitates channel closure as more CaM-EF12 than wild-type CaM is needed for an equivalent effect (Fig. 5). However, binding of Ca2+ to the N-lobe does not seem to be a strict requirement, providing some insight into the underlying molecular mechanism. Mechanism of channel regulation Several ion channels are regulated by cytosolic Ca2+ by means of CaM. Among the channels with a-subunits of the 6-TM family similar to hEAG1, Ca2+activated K+ channels of the IK ⁄ SK type [7,8,10], nonselective cation channels gated by cyclic nucleotides (CNG) [21–25], and voltage-gated channels of the KCNQ [26] type are known to interact with CaM. hEAG1 channels studied here share some properties of CNG channels and of KCNQ channels. On the one hand they are activated by membrane depolarization 1082 and do not inactivate as do KCNQ channels. On the other hand EAG channels harbor a C-terminal putative binding site for cyclic nucleotides; however, thus far there is no compelling evidence for a functional impact of this site. In terms of Ca2+ regulation, hEAG1 channels are readily inhibited by Ca2+ ⁄ CaM at low Ca2+ concentrations [6] and low CaM concentrations (Fig. 1). As for CNG channels, hEAG1 harbors CaM-binding motifs in the N- and C-terminal domain and both of them are functional in a sense that mutagenesis inside these domains results in a loss of Ca2+ sensitivity. Given the similarity to CNG channels it is tempting to assume that hEAG1 channels are gated via CaM by a distortion of an interaction between the N- and C-termini. We do not think that this is a likely scenario, because in precipitation assays with N- and C-terminal protein fragments of hEAG1 we could not detect interaction, neither alone nor in the presence of CaM (not shown). In addition, peptides encoding the N- and C-terminal binding domains competed for CaM binding. With respect to Ca2+ regulation the strongest similarity of hEAG1 channels can be found to heteromeric olfactory CNG channels. In both cases N- and C-terminal CaM-BDs associate with CaM at low Ca2+ concentration resulting in channel closure. However, as for hEAG1 channels the mechanism is not clear [27]. In summary, the identification of three functional CaM-binding sites in hEAG1 channels shows that the mechanism of CaM-induced channel closure is much more complex than previously anticipated. In particular, the molecular mechanism of EAG channel regulation via Ca2+ ⁄ CaM markedly differs from those of other ion channels of the 6TM gene family. Experimental procedures Production of calmodulin The cloning of full-length cDNA encoding hCaM was previously described [6]. Point mutations were generated by overlap extension PCR [28]. EF-hand mutants used were: hCaM-EF12 (D21AÆD57A), hCaM-EF34 (D94AÆD130A), and hCaM-EF1234 (D21AÆD57AÆD94AÆD130A). (H)6-hCaM and (H)6-hCaM EF-hand mutants were cloned into pQE30, expressed in E. coli M15 cells, and purified with Ni-NTA agarose according to the suppliers protocols (Qiagen, Hilden Germany). Integrity of the proteins was verified with SDS ⁄ PAGE analysis. For labeling of CaM the arginine residue in the His-tag (MRGS-(H)6) was replaced by cysteine. The resulting MCGS-(H)6-hCaM was produced as described, followed by hydrophobic interaction chromatography according to a FEBS Journal 273 (2006) 1074–1086 ª 2006 The Authors Journal compilation ª 2006 FEBS U. Ziechner et al. protocol by Hayashi et al. [29] using a phenyl-sepharose 6FF (high sub) column from Amersham Biosciences (Freiburg, Germany). For GST precipitation assays and FCS measurements with labeled peptides wt-hCaM was used. From a pET-14b vector it was expressed in E. coli BL21 cells and purified by phenyl-sepharose chromatography [29]. Production of hEAG1 fusion proteins Cloning of full-length cDNA encoding hEAG1 was previously described [6] and point mutations were generated by overlap extension PCR. GST fusion proteins, encoded in the pGEX-5X vector (Amersham Biosciences), were expressed in E. coli BL21 cells and purified using GSH sepharose. (H)6-MBP fusion proteins of hEAG1 were produced using the pETM 40 vector from EMBL (Heidelberg, Germany) in E. coli BL21 cells and purified on Ni-NTA agarose. Labeling of MCGS-(H)6-hCaM with BODIPY FL or TMR (tetramethylrhodamine) BODIPY FL N-(2-aminoethyl)maleimide and TMR-6maleimide were purchased from Molecular Probes (Eugene, OR, USA). A 10-fold molar excess of reactive dye (BODIPY FL- or TMR-maleimide in dimethyl sulfoxide) was added to purified MCGS-(H)6-hCaM in 20 mm Tris buffer (pH 7.4). This mixture was incubated under shaking for 60 min at room temperature. Subsequently, the excess of dye was inactivated with a fivefold molar excess (relative to the reactive dye) of glutathione under shaking for 20 min. The dye-labeled hCaM was separated by size exclusion chromatography using a HiPrep 16 ⁄ 60 Sephacryl S 200 HR-column (Amersham Biosciences) and 50 mm phosphate buffer (pH 7.4) with 500 mm NaCl and 2 mm glutathione. Synthesis and labeling of peptides Three peptides encompassing putative CaM-binding sites of hEAG1 (BD-N, BD-C1, and BD-C2) were synthesized using standard protocols. As illustrated in Fig. 3(A–C), the BD-N naturally starts with a cysteine residue; in the other peptides a cysteine was placed at the N-terminus before the natural hEAG1 sequence. These peptides were labeled at their N-terminal cysteines with TMR or Cy5 (Amersham Biosciences) according to the same procedure as applied for MCGS-(H)6-hCaM. Size exclusion chromatography was performed with a Superdex Peptide PE 7.5 ⁄ 300 column (Amersham Biosciences). GST precipitation assay GST-fusion proteins bound to glutathione-sepharose were washed three times with Ca2+-free or Ca2+-containing Calmodulin binding to hEAG1 channels buffer and then incubated with purified hCaM either in the presence or absence of Ca2+ for 10 min at room temperature. Following threefold washing of the beads with the relevant buffer, the resin-bound proteins were eluted with SDS ⁄ PAGE loading buffer. Subsequently, SDS ⁄ PAGE analysis was performed and protein bands were stained with Coomassie blue. The Ca2+-free buffer contained (in mm): 100 K-aspartate, 15 KCl, 10 EGTA, 2 glutathione, 10 Hepes (pH 7.2 with KOH). The Ca2+-containing buffer consisted of (in mm): 115 K-aspartate, 8.9 CaCl2, 10 hEDTA, 2 glutathione, 10 Hepes (pH 7.2 with KOH) yielding 25 lm free Ca2+ ions. Peptide array assay Arrays of immobilized peptides (15-mers) comprising sequences of the hEAG1 potassium channel with a shift of 2 or 3 amino acids were prepared by automated spot synthesis on Whatman 50 filter paper (Whatman, Maidstone, England). Peptides were C-terminally attached to cellulose via a (beta-Ala)2 spacer. The quality of the synthetic peptides on spots was evaluated by mass spectrometry analysis and the amount of peptides on spots was determined by HPLC analysis after releasing the peptides from cellulose membrane using ammonia vapor [30,31]. Before screening, the dry cellulose membrane was pretreated with methanol for 10 min and then equilibrated, first, five times for 20 min with a Tris-buffer (30 mm Tris ⁄ HCl, 170 mm NaCl, 6.4 mm KCl, pH 7.6) containing 1% BSA and 1% sucrose to prevent unspecific binding to the membrane surface during the assay and, second, three times for 10 min in binding buffer with 25 lm free Ca2+. In order to investigate which peptides on the membrane are able to interact with hCaM, 10 lm BODIPY FL-labeled hCaM in the same Ca2+-containing buffer was added to the membrane surface. After 30 min incubation, the unbound excess of labeled hCaM was removed by washing the membrane in Ca2+-containing buffer. Subsequently, the fluorescence was measured with a fluorescence scanner, FLA-5000 (FUJIFILM Europe GmbH, Düsseldorf, Germany). Quantification of signal intensity was performed by analysis of the spot intensity using nih image Software. Confocal fluorescence correlation spectroscopy (FCS) FCS measurements were performed in vitro using ConfoCor-2 equipment (Carl Zeiss AG, Jena, Germany) in autocorrelation mode. The smaller interaction partner was labeled with either TMR, Cy5 or BODIPY FL. These fluorescence dyes were excited with a HeNe laser (543 nm or 633 nm) or by an Ar+ laser (488 nm), respectively, and the emitted photons were detected using emission filters LP560, FEBS Journal 273 (2006) 1074–1086 ª 2006 The Authors Journal compilation ª 2006 FEBS 1083