Small conductance Ca2+-activated potassium channels (SK channels) are coassembled complexes of pore-forming SK α subunits and calmodulin. We proposed a model for channel activation in which Ca2+ binding to calmodulin induces conformational rearrangements in calmodulin and the α subunits that result in channel gating. We now report fluorescence measurements that indicate conformational changes in the α subunit after calmodulin binding and Ca2+ binding to the α subunit–calmodulin complex. Two-hybrid experiments showed that the Ca2+-independent interaction of calmodulin with the α subunits requires only the C-terminal domain of calmodulin and is mediated by two noncontiguous subregions; the ability of the E-F hands to bind Ca2+ is not required. Although SK α subunits lack a consensus calmodulin-binding motif, mutagenesis experiments identified two positively charged residues required for Ca2+-independent interactions with calmodulin. Electrophysiological recordings of SK2 channels in membrane patches from oocytes coexpressing mutant calmodulins revealed that channel gating is mediated by Ca2+ binding to the first and second E-F hand motifs in the N-terminal domain of calmodulin. Taken together, the results support a calmodulin- and Ca2+-calmodulin-dependent conformational change in the channel α subunits, in which different domains of calmodulin are responsible for Ca2+-dependent and Ca2+-independent interactions. In addition, calmodulin is associated with each α subunit and must bind at least one Ca2+ ion for channel gating. Based on these results, a state model for Ca2+ gating was developed that simulates alterations in SK channel Ca2+sensitivity and cooperativity associated with mutations in CaM.
SK channels are potassium-selective, voltage-independent, and are activated by increases in the levels of intracellular Ca2+ such as occur during an action potential. SK channels underlie the slow afterhyperpolarization (sAHP; Blatz and Magleby, 1987; Sah, 1996) that limits the firing frequency during a train of action potentials (Madison and Nicoll, 1984; Lancaster and Adams, 1986; Hille, 1992; Sah, 1996). This spike-frequency adaptation regulates burst frequency and is essential for normal integrative neurotransmission.
Three mammalian SK channels (SK1, SK2, and SK3) have been cloned that demonstrate a high degree of structural homology (Köhler et al., 1996) and a high sensitivity to Ca2+. The channels gate rapidly after application of saturating Ca2+, with onset of current commencing within 1 msec, a time course of activation similar to that observed for other ligand-gated channels such as GABA (Maconochie et al., 1994) or ionotropic glutamate receptors (Lester et al., 1990), suggesting a direct interaction between the ligand (Ca2+ ions) and the channel protein. Structure–function analysis revealed that Ca2+-gating is accomplished by constitutive association of calmodulin (CaM) with a region of the channel α subunits, ABC (SK2, amino acids 390–487), which resides in the intracellular C-terminal domain, and a Ca2+-dependent interaction with the BC region (SK2, amino acids 423–487). A model was presented in which Ca2+ ions bind to CaM, inducing conformational changes that are transmitted to the channel α subunits, resulting in channel activation (Xia et al., 1998).
CaM is a ubiquitous mediator of Ca2+-dependent processes. CaM contains N- and C-terminal globular domains, each including two high-affinity Ca2+-binding E-F hand motifs, E-F 1 and 2 in the N terminus, and E-F 3 and 4 in the C terminus (Babu et al., 1985). Ca2+ ions bind to CaM in a highly cooperative manner, first to E-F 4 and 3 and subsequently to E-F 2 and 1. Ca2+ binding to CaM induces conformational rearrangements that bend the linker region and bring the globular domains into close proximity. In addition, hydrophobic side chains within each globular domain are exposed. Taken together, these structural alterations present a physical interface for a diverse spectrum of signaling substrates important in developmental and adaptive responses among virtually all cell types, as well as synaptic plasticity in the mammalian CNS (O’Neil and DeGrado, 1990).
The interaction between CaM and SK channel α subunits is constitutive and is maintained in the presence or the absence of Ca2+ ions (Xia et al., 1998). This permits a direct coupling between changes in intracellular Ca2+ concentrations and changes in membrane potential. The results of experiments reported here revealed a modular strategy in which the N-terminal E-F hands of CaM are responsible for Ca2+-induced conformational changes in the channel, whereas two short stretches of amino acids in the C-terminal half of CaM mediate constitutive interactions.
MATERIALS AND METHODS
Molecular biology. Site-directed mutagenesis was performed as described (Weiner et al., 1994) using pfu DNA polymerase (Stratagene, La Jolla, CA); mutations were verified by DNA sequence analysis. The Genetics Computer Group suite of programs was used for DNA and protein sequence analysis. For oocyte expression, all mRNAs were derived from sequences that were subcloned into the oocyte expression vector pBF. To join subunits in tandem, the relevant stop codon was removed, and a linker encoding 10 glutamine residues (Q10) was inserted between the last codon of the 5′ subunit coding sequence and the initiator codon of the following subunit. This was achieved using a sequential PCR protocol modified from Horton et al. (1989); junctions were generated by overlap extension of PCR primers that also encoded the glutamine linkers (Pessia et al., 1996).
Fusion proteins and Western blots. The indicated channel sequences (SK2 or SK2R464E,K467E, ABC or BC) were amplified by PCR using pfu DNA polymerase (Stratagene) and subcloned into the glutathione S-transferase (GST) fusion vector pGEX-KG (Pharmacia Biotech, Piscataway, NJ). Cultures harboring the plasmids were harvested in PBS and lysed by French press. Cleared lysates were incubated with glutathione agarose (Sigma, St. Louis, MO) for 2 hr at 4°C, and the resin was subsequently washed twice with PBS in the presence (10 μm) or absence (5 mm EGTA) of Ca2+. Resin-bound proteins were incubated with purified bovine brain CaM (generous gift of Dr. Debra Brickey, Vollum Institute) in the presence or absence of Ca2+ for 2 hr at 4°C. The resin was then washed twice with PBS with or without Ca2+. Bound proteins were eluted using 10 mm reduced glutathione (Sigma) in 50 mm Tris, pH 8.0. SDS-PAGE (14% acrylamide) was performed with 0.5 mm EGTA in the gel and running buffer. For Western blotting, proteins were electroblotted to Hybond membrane (Amersham, Arlington Heights, IL), and CaM was detected using a monoclonal antibody (Upstate Biotechnology, Lake Placid, NY) and HRP-linked secondary antibody (Bio-Rad, Hercules, CA), visualized with chemiluminescence (New England Nuclear, Boston, MA).
For fluorescence emission measurements, SK2 ABC (Xia et al., 1998) was subcloned into pET33b and produced in Escherichia coli BL21 as a his(6)-fusion protein. Ni2+-agarose purification resulted in a single coomassie-stained band after SDS gel electrophoresis. Purified ABC was dialyzed into 360 mm NaCl, 1 mm EGTA, and 18 mm HEPES, pH 7.2.
Fluorescence emission measurements. Fluorescence emission spectra were performed using a Photon Technologies QM-1 steady-state fluorescence spectrophotometer. Samples were excited at 295 nm (2 nm bandpass), and the fluorescence emission was monitored in 1 nm intervals from 310 to 410 nm (5 nm bandpass). Measurements were performed at 22°C, in 360 mm NaCl, 1 mm EGTA, and 18 mm HEPES, pH 7.2. A typical measurement used 1.4 μm of ABC peptide in the above buffer. CaM was added from a stock of 120 μm to an ∼1:1.25 ratio with the ABC peptide. To this solution CaCl2 was added from a 1 m stock (Fluka, Milwaukee, WI) to a final concentration of 1.1 mm, yielding a final free Ca2+ concentration of 100 μm.
Yeast two-hybrid. The indicated SK2 α subunit coding sequences were subcloned into the bait vector pPC97 as fusions with the GAL4 DNA binding domain. Rat CaM, subfragments, or point mutations (as indicated in the text), were fused to the transcriptional activator domain in the prey vector pPC86 (Chevray and Nathans, 1992). HF7c competent yeast were cotransformed with each of the indicated plasmids, and transformants were plated onto media lacking leucine and tryptophan. From this plate, a single colony was grown overnight in leu−, trp−liquid media. The next morning, the culture was diluted 100-fold with 10 mm Tris, 0.1 mm EDTA, and interactions between the bait and prey were assessed by spotting 1 μl onto leu−, trp−, his− plates. Growth was monitored after 2 d of incubation at 30°C.
125I-apamin binding. Oocytes were rinsed in ice-cold assay buffer (100 mm Tris-HCl, 1 mm EDTA, 5.4 mm KCl, and 1% BSA, pH. 8.4, at 4°C), and then added to wells containing125I-apamin (0.1–1.4 nm) (New England Nuclear) and assay buffer in a final volume of 750 μl. After 1 hr of incubation, oocytes were rinsed in four changes of ice-cold assay buffer over a total of 15 sec, placed on filter paper, and binding was detected on a PhosphorImager.
In the same experiment, radioligand binding was quantified for individual oocytes by conventional gamma ray spectrometry. Specific binding was defined as the difference in radioligand binding observed in the absence and presence of 10 nm apamin (Sigma). Under these conditions, bound radioligand accounted for <10% of total radioligand added to the assay.
Electrophysiology. In vitro mRNA synthesis and oocyte injections were performed as previously described (Xia et al., 1998). Xenopus care and handling were in accordance with the highest standards of institutional guidelines. Patch recordings in the inside-out configuration were made at room temperature (∼23°C) 3–7 d after injection. Pipettes prepared from thin-walled borosilicate glass (World Precision Instruments, Sarasota, FL) had resistances of 0.5–2 MΩ when filled with (in mm) 120 K-methanesulfonate (MES) and 5 HEPES, pH-adjusted to 7.2 with KOH. Voltage-clamp recordings were performed with an Axopatch 1B or Axopatch 200A patch-clamp amplifier (Axon Instruments, Foster City, CA). Currents were sampled at 2 kHz and filtered at 2 kHz (−3 dB). Excised patches were superfused with an intracellular solution containing (in mm): 120 K-MES, 5 HEPES, and 1 EGTA, which was supplemented with CaCl2, pH-adjusted to 7.2 with KOH; the amount of CaCl2 required to yield the concentrations indicated was calculated according to Fabiato and Fabiato (1979). Solutions were prepared with HEPES (Life Technologies, Gaithersburg, MD), 1 m KOH standard solution (Fluka, Milwaukee, WI), methanesulfonic acid (Aldrich, Milwaukee, WI), EGTA (Fluka), and 1 mCaCl2 (BioChemika MicroSelect; Fluka). Currents within a concentration response obtained early after patch excision were subject to rundown (<20%), which would bias the EC50 (Ishii et al., 1997). Therefore, current amplitudes were corrected for rundown by normalizing to the control current predicted from the time course of rundown of the current activated by 10 μmCa2+. Ca2+dose–response curves were determined from current amplitudes measured at −80 mV as a function of Ca2+concentration and fit with a Hill equation. The values are reported as the mean EC50 ± SD of n experiments, as indicated. Statistical significance was evaluated using a pairedt test, and a p value <0.01 was considered significant. Data analysis was performed using Pulse (Heka, Lambrecht, Germany) and Igor (Wavemetrics, Lake Oswego, OR). Simulation of Ca2+-dependent activation of SK2 was modeled using SCoP (Simulation Resources, Berrien Springs, MI) running in a DOS environment on a Power Macintosh (Virtual PC; Connectix, San Mateo, CA).
CaM-mediated structural rearrangments in the SK2 ABC
The ABC domain of SK2 was purified from bacteria and used for fluorescence emission measurements (Lakowicz, 1983). This region of the channel α-subunit contains a single tryptophan residue, W432, which resides in a highly solvent-exposed environment, as judged by its fluorescence emission maximum at 346.2 ± 0.8 nm (n = 5; Fig. 1). CaM could be added to this peptide in the absence of Ca2+ to induce a complex that saturated (data not shown), indicating a specific interaction between ABC and CaM. The CaM binding to ABC shifted the fluorescence emission to 329.0 ± 0.0 nm (n = 5) and increased its fluorescence intensity by a factor of 4.1 ± 0.1 (n = 5; Fig. 1 B). These results indicate that CaM binding introduces W432 into a more hydrophobic environment. After addition of Ca2+, yielding a free Ca2+ concentration of 100 μm, to the ABC–CaM complex, a large decrease in fluorescence was observed that returned the total intensity to 1.5 ± 0.1 that of the original level (Fig. 1 C). Interestingly, this decrease was accompanied by a further shift in the emission maxima to 326.2 ± 0.4 nm (n = 5). This latter result indicates that W432 remained in a hydrophobic environment and suggests a static quenching mechanism. Importantly, the fluorescence shift observed after addition of Ca2+ shows that CaM remains bound to the ABC peptide in the presence of Ca2+, because no wavelength shift was observed after addition of Ca2+ to the ABC peptide alone (data not shown). Taken together, these results demonstrate that (1) CaM binding to ABC induces a change in the environment of W432, and (2) subsequent Ca2+ binding to the ABC–CaM complex further alters the conformation of ABC.
CaM domains required for Ca2+-independent interactions with SK2 ABC
Two-hybrid experiments detected an interaction between CaM and the ABC domain of the SK2 channel α subunits, but not between CaM and the BC domain. In contrast, an interaction between CaM and the BC domain was detected biochemically using the purified proteins, but only in the presence of Ca2+ (Xia et al., 1998; our unpublished results) indicating that the two-hybrid system detects only the Ca2+-independent interactions between CaM and the SK2 α subunits. To determine regions of CaM responsible for Ca2+-independent interactions with SK2 α subunits, fragments of CaM were tested for their ability to complement different domains of the SK2 intracellular C-terminal domain in two-hybrid experiments. Consistent with previous results, only α subunit fragments containing ABC showed complementation with any of the CaM fragments tested. When CaM was divided approximately in half, the N-terminal fragment (amino acids 1–82) was ineffective for complementation with ABC from SK2, whereas the C-terminal fragment (amino acids 78–148) complemented ABC (Fig.2 A). To further define residues important for the interaction, truncations from amino acid D78 or from the C terminus, amino acid K148, were tested. Removal of seven residues (amino acids 78–85) from the N terminus or five residues (amino acids 143–148) from the C terminus eliminated the ability of the CaM fragment to complement ABC from SK2 (Fig.2 A). Moreover, intact CaM with D→A mutations in any or all combinations of the E-F hands complemented ABC from SK2 (Fig.2 B), indicating that functional Ca2+-binding motifs are not required for the constitutive interaction of CaM with SK2ABC.
SK2 residues necessary for Ca2+-independent interactions with CaM
To identify sites on the SK2 α subunit that interact with CaM, a series of point mutations in the SK2 ABC domain was generated. All charged residues were altered by site-directed mutagenesis but yielded functional channels with Ca2+ responses similar to wild type (Xia et al., 1998; data not shown). Several double mutations were constructed and one double mutant, R464E,K467E in the C helix, did not yield functional channels. These positions are conserved as positively charged amino acids in all SK α-subunits. To examine the basis for the lack of function, GST fusion proteins of the wild-type ABC or ABC(R464E,K467E) were tested for their ability to interact with CaM in the presence (10 μm) or absence (5 mm EGTA) of Ca2+. The results showed that wild-type ABC bound CaM under either condition, whereas wild-type BC bound CaM only in the presence of Ca2+ (Xia et al., 1998). In contrast, ABC(R464E,K467E) or BC(R464E,K467E) bound CaM only in the presence of 10 μm Ca2+ (Fig.3 A). Less binding of CaM to the double mutant proteins compared to wild type was observed, suggesting that even the Ca2+-dependent interaction was weakened by these mutations. Consistent with these results, an interaction of the ABC(R464E,K467E) double mutant with CaM was not detected in a two-hybrid test (data not shown). To determine whether R464E,K467E channels were assembled and inserted in the plasma membrane, or whether the double mutation affected channel biosynthesis,125I-apamin binding studies were performed on intact oocytes injected with mRNA encoding apamin-insensitive SK1 channels, apamin-sensitive SK2 channels, or SK2 R464E,K467E channels. SK1 channels showed no specific binding, but125I-apamin binding was clearly seen for oocytes expressing either SK2 (specific binding, 2406 cpm;n = 3) or SK2 R464E,K467E (specific binding, 2645 cpm;n = 3; Fig. 3 B). These results indicate that R464 and K467 in SK2 are necessary for the constitutive interaction between SK2 and CaM.
To examine the stoichiometry of CaM required for channel function, SK1 (Köhler et al., 1996; Xia et al., 1998) and SK2(R464E,K467E) subunits were linked together by a 10 glutamine linker (Q10 linker), and the dimer mRNA was injected into Xenopus oocytes. In contrast to expression of a dimer of SK1 and wild-type SK2 (Ishii et al., 1997), channel gating was not observed from oocytes injected with SK1-SK2(R464E,K467E) dimer mRNA, even though 125I-apamin binding sites were detected on the surface membrane (data not shown). This result suggests that CaM must be bound constitutively to at least three α subunits for channel function.
Ca2+-dependent channel gating is mediated by E-F hands 1 and 2
The aspartate residue found in the first position of each E-F hand of rat CaM was altered by site-directed mutagenesis to an alanine (D→A), either alone or in all possible combinations. Previous results established that this mutation dramatically reduced or abolished Ca2+ binding (Geiser et al., 1991; Xia et al., 1998). Wild-type or mutant CaMs were coexpressed with SK2 inXenopus oocytes, and Ca2+-gating was examined in inside-out macropatches. Application of Ca2+ to patches from oocytes coexpressing wild-type CaM and SK2 resulted in Ca2+-activated potassium currents with an EC50 of 0.38 ± 0.05 μm and a Hill coefficient of 4.6 ± 1.3 (n = 23). Neither the EC50 nor current amplitudes were different from oocytes expressing SK2 alone or coexpressing SK2 and wild-type CaM (Xia et al., 1998). Mutations of D→A in either E-F hand 1 or 2 (CaM1 or CaM2) shifted the EC50 to higher Ca2+concentrations and reduced the Hill coefficient, whereas mutations in either E-F hands 3 or 4 (CaM3 or CaM4) did not significantly alter Ca2+-dependent gating of SK2 (Fig.4, Table1). Coexpression of SK2 with CaM harboring D→A mutations in both E-F hands 1 and 2 (CaM12) resulted in small currents that on average were 0.05 ± 0.03 (n = 12) of current amplitudes observed in coexpression of SK2 with wild-type CaM. The Ca2+concentration–response curve was slightly shifted, but because of the small size of the current compared to background, this difference may not be significant. Similar results were obtained from coexpression of SK2 with CaM harboring mutations in all four E-F hands, indicating that the current in oocytes injected with SK2 and CaM12 or CaM1234 may result from SK2 channels coassembled with endogenous oocyte CaM. In contrast, coexpression of SK2 with CaM34 resulted in robust currents with Ca2+ concentration–response relationships not different from SK2 coexpressed with wild-type CaM (Fig. 4, Table 1). These results show that the double mutation in both E-F hands 1 and 2 suppress channel function, whereas mutations in either E-F hands 1 or 2 shifted the EC50 and reduced the Hill coefficient for Ca2+activation. To test whether only a single functional E-F hand was sufficient for channel activation, SK2 was coexpressed with mutant CaMs with three of the four E-F hands bearing the D→A mutation. Coexpression of SK2 with CaM123 or CaM124 resulted in currents similar to those resulting from coexpression of SK2 with CaM12 or CaM1234. However, coexpression of SK2 with CaM134 or CaM234 resulted in robust currents with right-shifted EC50 values and reduced Hill coefficients similar to mutations in E-F hands 1 or 2 (Table 1). These results show that Ca2+gating of SK2 channels results from Ca2+binding to the first and second E-F hands in CaM and that either E-F hand 1 or 2 by itself is sufficient for channel activation.
A model for CaM-mediated gating
Ca2+-dependent activation of SK channels was previously described as a time-homogenous Markov chain with four closed states and two open states (Hirschberg et al., 1998). This gating scheme was developed before elucidating the role of Ca2+-CaM for SK channel gating. Assuming that SK channels are tetramers, the fluorescence measurements and functional data presented above suggest that CaM is bound to each α-subunit, and Ca2+ binding to either one or both E-F hands 1 or 2 on all four CaM molecules results in channel gating. This is analogous to the model developed by Zagotta et al. (1994a,b) to describe voltage-dependent gating of ShakerK+ channels in which two voltage-dependent transitions must occur in each of the four subunits. A modification of this model was adopted to describe Ca2+-dependent gating of SK2 (Fig.5, inset). Two differences were incorporated to account for our results. First, to simulate Ca2+ binding to CaM, the on-rate constants were made Ca2+-dependent, and all rate constants were voltage-independent. Second, the channel was allowed to enter the open state if each CaM molecule was bound with at least one Ca2+ molecule. To account for the cooperative interaction between the two E-F hands as suggested by the single and triple E-F hand mutations (Table 1), the off-rate for the second Ca2+ bound to CaM was decreased compared to the off-rate for the first bound Ca2+. When each CaM molecule is bound with one or two Ca2+ ions, the channel can enter the open state with the rate constants determined from single-channel analysis for the open state with the longest mean open time (Hirschberg et al., 1998); functionally the five open states, O1-O5, are identical. The model was used to simulate Ca2+ gating and qualitatively accounts for the EC50 and Hill coefficient when both Ca2+-binding domains are functional as well as when the second Ca2+-dependent forward rate constant was set to zero to simulate either CaM1, CaM2, CaM234, or CaM134 (Fig. 5). The EC50 and Hill coefficient of 1.1 μm and 2.3, respectively, agrees well with the values determined for CaM234 (Table 1). However, the Hill coefficient determined for the complete model, 3.5, was slightly smaller than the experimentally determined value of 4.6 for wild-type CaM (Table 1).This discrepancy may result from subtle distinctions between the EC50 of CaM1 and CaM2, or CaM134 and CaM234 which were not incorporated into the model.
CaM is an exquisite Ca2+ sensor that mediates a wide range of intracellular signaling pathways. Ca2+-binding to CaM induces conformational alterations that expose domains that mediate interaction with target proteins; the activity of these substrates is altered after Ca2+–CaM binding. Structurally, CaM has globular N- and C-terminal domains, each containing two E-F hand motifs separated by a flexible linker region. All four E-F hand Ca2+-binding domains possess high affinity for Ca2+. However, slight affinity differences and highly cooperative Ca2+binding endow CaM with the ability to distinguish small fluctuations in intracellular Ca2+ levels within physiological ranges (Klee, 1988). After Ca2+ binding, the flexible tether region bends, bringing the globular domains into spatial proximity and forming a hydrophobic interface for binding to target peptides (O’Neil and DeGrado, 1990; Finn and Forsen, 1995; Finn et al., 1995). Previously, we have shown that CaM is constitutively associated with the channel α subunits and Ca2+ binding to CaM induces channel gating. Here, we present evidence showing structural changes in the α subunit after CaM binding and subsequent Ca2+ binding to the CaM–α subunit complex. We also present evidence for a modular structure to CaM; distinct regions are responsible for Ca2+-dependent and Ca2+-independent interactions with SK α subunits.
Structural motifs that bind Ca2+-bound CaM are characterized by regularly spaced hydrophobic amino acids, and frequently have an overall positive net charge. Most of the high-affinity Ca2+–CaM-binding domains conform to the 1-8-14 motif (Rhoads and Friedberg, 1997). The hydrophobic residues at these positions may interact with hydrophobic patches exposed in the N- and C-terminal domains of CaM after Ca2+ binding. In other instances, CaM binds to target proteins such as neurogranin (Baudier et al., 1991), neuromodulin (Alexander et al., 1988), and unconventional myosins (Wolenski, 1995) or the Drosophila protein igloo(Neel and Young, 1994), with greater affinity in absence of bound Ca2+. These interactions remain Ca2+-sensitive, and it is likely that the Ca2+-free form of CaM performs an alternative allosteric regulatory function. A structural motif that mediates Ca2+-free CaM binding, the IQ motif, is found in a variety of proteins and comprises a 14 residue stretch beginning with IQ and containing positively charged residues at positions 6 and 11. The distinction between motifs that bind Ca2+-free or Ca2+-bound CaM is not absolute, because some IQ motifs also conform to Ca2+-dependent motifs (Rhoads and Friedberg, 1997). The interaction between SK2 α subunits and CaM is different and resembles phosphorylase kinase in which, even in the absence of Ca2+, CaM is an intrinsic subunit of the enzyme, requiring harsh, denaturing conditions for subunit separation (Picton et al., 1980; Kee and Graves, 1986).
The determinants for Ca2+-independent, constitutive binding to the SK2 α subunits ABC domain reside in the C-terminal half of CaM, as demonstrated by yeast two-hybrid experiments. Further analysis showed that two noncontiguous stretches of amino acids, 78–85 spanning the border of the flexible tether region with the C-terminal globular domain and the final five residues, 143–148, were both necessary for Ca2+-independent binding. The Ca2+-binding integrity of E-F hands 3 and 4 was not required. Two methionines within the final five residues have been implicated in Ca2+–CaM binding as part of hydrophobic methionine puddles that form interfaces with target substrates (O’Neil and DeGrado, 1989, 1990). Two conserved, positively charged residues in the C domain of the channel α subunits are important for Ca2+-independent CaM binding. Interestingly, charge reversal of the individual amino acids had no effect, but together, neither channel gating nor Ca2+-independent CaM binding to the ABC domain was detected. The modular nature of the interactions are emphasized by this mutant because Ca2+-dependent interactions are maintained even though Ca2+-independent interactions have been disrupted. Apamin binding showed that the double mutant α subunits are assembled and inserted into the plasma membrane.
Ca2+-dependent channel gating is mediated by Ca2+ binding to E-F hands 1 and 2 in the N-terminal globular domain. Mutations that reduce or eliminate Ca2+ binding to E-F hands 3 and 4 do not affect Ca2+-gating, but the same mutations in E-F hands 1 or 2 result in shifts in apparent Ca2+ affinity and gating cooperativity. These results were unexpected because under physiological conditions, E-F hands 3 and 4 possess the highest intrinsic Ca2+ affinity; in most cases Ca2+ ions are likely bound first to these motifs and subsequently to E-F hands 1 and 2 (Wang et al., 1985). This result differs from our previous work that implicated E-F hands 3 and 4 in Ca2+ gating (Xia et al., 1998). The molecules employed in the present study have been repeatedly examined by nucleotide sequence analysis, and the experiments have been performed many times in different batches of oocytes. The results have been consistent and show that E-F hands 1 and 2 are necessary and sufficient for Ca2+ gating. Results from expression of CaMs with only E-F hands 1 or 2 intact show that a single functional E-F hand is necessary and sufficient for Ca2+ gating. This suggests that gating cooperativity may in part result from interactions between α subunits and not only from Ca2+ binding to CaM. The residues on the α subunits responsible for mediating Ca2+-dependent interactions with CaM have not yet been identified despite mutagenesis of many charged and hydrophobic residues.
The picture that emerges is one in which CaM interacts in a modular way with the SK channels, the C-terminal domain mediating constitutive binding and the N-terminal domain transmitting Ca2+ dependence (Fig.6).
The first evidence that CaM affects ion channels came from point mutations in Paramecium CaM, the pantophobiacs, which eliminate a calcium-activated potassium current (Schaefer et al., 1987). Interestingly, all of the pantophobiac mutations occur in the C-terminal globular domain of the protein, and all but two change residues in E-F hands 3 or 4 (Saimi and Kung, 1994). In our studies, the ability of these E-F hands to bind Ca2+ was not important for channel gating. However, the exact pantophobiac mutations were not examined. One pantophobiac mutation, M476V, resides in a small stretch implicated in Ca2+-independent CaM binding. Although this is a conservative substitution, it eliminates a methionine, supporting the hypothesis that “methionine puddles” form the basis for many important hydrophobic interactions with CaM binding targets (O’Neil and DeGrado, 1990). CaM mutants inParamecium also indicated distinctions among the N- and C-terminal domains for their ability to regulate ion channels; N-terminal mutants (fast-2 or paranoiac) affected a Ca2+-activated Na+ current (Kink et al., 1990). InDrosophila, CaM-null mutants are viable as larvae because of persistent maternal CaM. As maternal CaM levels decrease, the larvae demonstrate behavioral abnormalities strikingly similar to the avoidance behavior of pantophobiac mutants inParamecium. Moreover, flies harboring a mutation in CaM E-F hand 1 exhibit enhanced neurotransmitter release in low Ca2+ (Heiman et al., 1996; Arredondo et al., 1998), consistent with a shift in the Ca2+ concentration–response for SK channels.
Until recently, CaM has generally been considered a mediator of intracellular Ca2+ signaling pathways. However, it is now clear that CaM also plays a central role in regulating membrane potential and ion channel activity. In addition to the finding that CaM is the intrinsic Ca2+-sensing subunit of SK2 channels, CaM binds to voltage-dependent L-type calcium channels through an IQ motif on the C-terminal domain of the channel, a region analogous to the region bound by CaM in SK2 channels, and plays a direct and important role in regulating L-type Ca2+ channel kinetics (Zühlke et al., 1999; Peterson et al., 1999). Similar to results presented for SK2 (Xia et al., 1998), the Ca2+-dependent interaction of L-type Ca2+ channels with CaM cannot be inhibited by peptide competitors or compounds such as calmidizolium, although it is not yet clear that CaM binds constitutively to L-type Ca2+ channels in the quantitative absence of Ca2+. Similar to the interaction of CaM with SK channels, the different E-F hand motifs make distinct contributions to L-type Ca2+ channel regulation (Peterson et al., 1999). In many neurons, such as hippocampal pyramidal neurons (HPNs), blockade of L-type Ca2+ channels also blocks the slow AHP (Tanabe et al., 1998). Moreover, in HPNs, L-type Ca2+ channels and SK channels reside in close proximity, and a direct functional coupling has been demonstrated (Marrion and Tavalin, 1998). By affecting L-type Ca2+ channel activity, SK channel activity will also be altered, exerting strong effects on spike frequency adaptation and neuronal excitability.
CaM has also been implicated in regulating the activity of ionotropic receptors such as NMDA receptors through direct interactions (Ehlers et al., 1996). Ca2+ flux through NMDA receptors impacts long-term potentiation, which is also coupled to CaM through activation of CaM kinase II (Otmakhov et al., 1997). Thus, CaM may be viewed as a central coordinator for a variety of ion channels, all of which influence and are influenced by Ca2+ dynamics and have significant long-term effects in the CNS.
This work was supported by grants from the National Institutes of Health, the Muscular Dystrophy Association, Human Frontier Science Program, and ICAgen, Inc. We thank Mr. Robert Johnson for assistance with 125I-apamin binding experiments. We also thank Drs. Xia-Ming Xia and Takahiro Ishii for some of the mutagenesis and fruitful discussions.
Drs. Keen and Khawaled contributed equally to this work.
Correspondence should be addressed to Dr. James Maylie, Department of Obstetrics and Gynecology, Oregon Health Sciences University, 3181 Southwest Sam Jackson Park Road, Portland, OR 97201.