Molecular basis of calmodulin tethering and Ca2+-dependent inactivation of L-type Ca2+ channels.

Ca(2+)-dependent inactivation (CDI) of L-type Ca(2+) channels plays a critical role in controlling Ca(2+) entry and downstream signal transduction in excitable cells. Ca(2+)-insensitive forms of calmodulin (CaM) act as dominant negatives to prevent CDI, suggesting that CaM acts as a resident Ca(2+) sensor. However, it is not known how the Ca(2+) sensor is constitutively tethered. We have found that the tethering of Ca(2+)-insensitive CaM was localized to the C-terminal tail of alpha(1C), close to the CDI effector motif, and that it depended on nanomolar Ca(2+) concentrations, likely attained in quiescent cells. Two stretches of amino acids were found to support the tethering and to contain putative CaM-binding sequences close to or overlapping residues previously shown to affect CDI and Ca(2+)-independent inactivation. Synthetic peptides containing these sequences displayed differences in CaM-binding properties, both in affinity and Ca(2+) dependence, leading us to propose a novel mechanism for CDI. In contrast to a traditional disinhibitory scenario, we suggest that apoCaM is tethered at two sites and signals actively to slow inactivation. When the C-terminal lobe of CaM binds to the nearby CaM effector sequence (IQ motif), the braking effect is relieved, and CDI is accelerated.

The voltage-gated L-type Ca 2ϩ channel is unique among ion channels in displaying two gating properties that are regulated by the ion that permeates the channel, calcium-dependent inactivation (CDI) 1 and calcium-dependent facilitation (CDF). These feedback mechanisms are of critical importance for regulation of the electromechanical activity of the heart and other essential physiological processes. CDI helps determine the length of the cardiac action potential plateau (1) and CDF contributes to the positive force-frequency relationship of the cardiac contraction (2).
Several lines of evidence from recent work suggest that the calcium sensor mediating both of these processes may be the calcium-binding protein calmodulin (CaM). We (3) and others (4,5) have shown that there is a Ca 2ϩ -dependent CaM-binding sequence ("IQ motif") in the cytoplasmic C-terminal tail of the pore-forming ␣ 1C subunit of the channel, within a region previously shown to confer Ca 2ϩ sensitivity (6). We have also shown that those mutations within the IQ motif that render the channel subunit unable to bind CaM also disrupt CDI (3,7), suggesting that the IQ motif serves as the effector region for CDI. Furthermore, we (3) and others (4) have shown CDI can be blocked in a dominant negative fashion by those CaM mutants that lack Ca 2ϩ binding in their C-terminal EF-hand domains.
Several important questions remain unanswered about how Ca 2ϩ and CaM might regulate L-type Ca 2ϩ channel inactivation. The first question concerns how CaM may be tethered to the L-type channel (8). There are multiple reasons for thinking that there must be a binding site that tethers the Ca 2ϩ sensor in the channel's resting state, keeping it poised for signaling as soon as Ca 2ϩ entry begins. Without tethering it would be difficult to explain the rapid development of CDI, beginning within milliseconds after L-type channel opening is initiated by depolarization (5). Tethering would also explain the dominant negative inhibitory action of mutant CaM molecules inasmuch as their binding to the tethering site would preclude binding of wild-type CaM (3,4,7). Recent studies have identified sequences in a cytoplasmic domain of the ␣ 1C subunit that display significant affinity for CaM even at low Ca 2ϩ concentrations (9 -11), but these do not appear to bind to dominant negative CaM mutants, as would be expected for the putative tethering site. Thus, the molecular basis for CaM tethering remains unclear.
A second question concerns the nature of the interaction between CaM and the IQ motif. How does CaM interact with the IQ motif so that CaM performs its effector functions? Is this interaction similar to CaM interactions with well known partners such as Ca 2ϩ /CaM-dependent protein kinase II (CaMKII) or myosin light chain kinase?
A third question concerns puzzling alterations in Ca 2ϩ -independent inactivation, produced by modifying amino acids within the IQ motif or in its general vicinity. Such modifications produce prominent changes in the kinetics of Ba 2ϩ currents through L-type channels (3,7,12). How can one account for these experimental observations if the regulation of inactivation were strictly dependent on formation of a Ca 2ϩ ⅐CaM complex?
In addressing each of these questions, our experiments indi-cated that CaM acts through a switch-like mechanism significantly different than its classical disinhibitory mode of action.

EXPERIMENTAL PROCEDURES
cDNA Construction and Site-directed Mutagenesis-For the gel shift experiment in Fig. 1, the coding sequences of CaM and CaM 234 (provided by J. Adelman, Oregon Health Sciences University) were amplified by PCR and cloned into pGEX-4-T1 (Amersham Pharmacia Biotech). Constructs for the GST fusion proteins containing amino acids 1477-1592 and 1592-1875 of the ␣ 1C cytoplasmic tail were the generous gift of M. Hosey (Northwestern University) and have been described previously (13), although the numbering of the amino acids in this report reflects that of the cDNA clone 77wt that has been described previously (3). The construct for the GST fusion protein containing amino acids 1551-1660 was generated by PCR amplification using primers 5Ј-CCGAATTCAACAGTGACGGGACGGTC-3Ј and 5Ј-GGCTCGAGGTC-GTGCAGAGTGCGCAG-3Ј and cloning of the restriction enzyme-digested product into the EcoRI and XhoI sites of pGEX-4T-1. The construct for the GST fusion protein containing the I-II loop was generated by PCR amplification using primers 5Ј-CCGAATTCAGCGGAGAGTT-TTCCAAAG-3Ј and 5Ј-GGCTCGAGGTTCGACTTGACCGCTGC-3Ј and cloning of the restriction enzyme-digested product into the EcoRI and XhoI sites of pGEX-4T-1. The construct used to generate the in vitro translated 77I/E protein was created by site-directed mutation of Ile 1624 using QuikChange (Stratagene) from the wild-type construct described previously (3). For functional expression of wild-type and mutant L-type Ca 2ϩ channels in Xenopus oocytes, the cDNA clone 77wt was used. For the construction of the deletion mutant 77d13, the 2878-bp Hind-III-SpeI fragment of 77wt was first subcloned into pBluescript SK(Ϫ) to assemble the plasmid p77HS. This plasmid cDNA provided a convenient template cDNA for preparative mutagenic PCR. For the deletion of the 39-bp fragment (nt 1525-1563 in the open reading frame of 77wt) that encodes the 13-amino acid CaM-binding motif GARLA-HRISKSKF ( Fig. 2A), a triple fragment ligation strategy was used. Two short cDNA fragments A (nt 1335-1524) and B (nt 1564 -1800) were amplified, which flank the 39-bp sequence. For fragment A we used the forward primer 5Ј-GAAGCAGCAGCTAGAACGGGATCTCAAAGGCT-AC-3Ј (A fwd ) and the reverse primer 5Ј-AGTTACTCTTCCGCAGTTTT-CTCCCTCGATG-3Ј (A rev ). Fragment B was amplified with the forward primer 5Ј-AGTTACTCTTCATGCAGCCGCTACTGGCGC-3Ј (B fwd ) and 5Ј-CTGCAGGCCCAGGCTGTACATC-3Ј (B rev ). The primers A rev and B fwd both contained a mismatch-overhang containing an Eam1104I restriction site (underlined sequences). Fragment A and B were subsequently digested with the restriction enzymes SphI and Eam1104I, and Eam1104I and SfiI, respectively, and the two fragments were ligated into the SphI-SfiI sites of p77HS. A 2532-bp MfeI-PpuMI fragment containing the deletion was then transferred into the full-length 77wt to form the deletion mutant 77d13, and the deletion was confirmed by cDNA sequencing.
L-type Channel Fragment/CaM Interactions-Interactions between agarose-CaM and [ 35 S]methionine-labeled, in vitro translated protein were performed as described previously (3), except incubations were performed for 8 h at 4°C. Constructs for GST fusion proteins were transformed into BL21 Escherichia coli, grown to mid-log phase, and induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h. The cells were lysed by passage through a French press in buffer containing 50 mM Tris, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, and Complete EDTA-free protease inhibitor mixture (Roche Molecular Biochemicals). Following centrifugation at 14,000 ϫ g for 30 min, aliquots of supernatant were snap-frozen and stored at Ϫ80°C. For CaM and CaM 234 used in Fig. 2, protein was allowed to bind to glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech) in 150 mM NaCl, 50 mM Tris, pH 7.4, 0.1% Triton for 30  min at 4°C, washed extensively, and then cleaved from GST with  thrombin without added Ca 2ϩ . For binding experiments between fusion  proteins and CaM or CaM 1234 , an aliquot of fusion protein was thawed,  allowed to bind to glutathione-Sepharose 4B beads for 30 min at 4°C,  and washed extensively with buffer containing 150 mM NaCl, 50 mM  Tris, pH 7.4, 0.1% Triton X-100 (binding buffer), and the indicated amount of free Ca 2ϩ buffered with either EGTA, EDTA or HEDTA, as calculated with WEBMAXC version 2.10 (www.stanford.edu/ϳcpatton/ webmaxc2.htm). Either CaM or CaM 1234 was then added and allowed to interact with the fusion protein for 4 h at 4°C. The beads were then washed extensively with binding buffer containing the indicated amounts of free Ca 2ϩ and the retained protein eluted with SDS sample buffer. The protein was then subjected to SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and detected by immunoblot with a polyclonal antibody against CaM (Zymed Laboratories Inc.). The ability of this antibody to detect CaM 1234 equally as well as CaM was confirmed (data not shown).
Purification of CaM and CaM 1234 -The CaM expression vector pPCR2 containing either wild-type CaM or CaM 1234 was transformed into BL21 E. coli, grown to mid-log phase, and induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h. The cells were lysed by passage through a French press in buffer containing 50 mM Tris, pH 8.0, and 150 mM NaCl. The lysate was heated to 70°C, immediately cooled on ice, and centrifuged at 100,000 ϫ g for 30 min at 4°C. After gradually bringing the supernatant to a final concentration of 2 M ammonium sulfate, the precipitate was removed by centrifugation. The supernatant was then brought to a saturating final concentration of ammonium sulfate (ϳ5 M), and the precipitate was pelleted. The wild-type CaM pellet was then resuspended in 30 ml of 50 mM Tris, pH 7.5, 200 mM ammonium sulfate, and 1 mM EDTA and applied to a phenyl-Sepharose column. CaCl 2 was slowly added to the flow-through until a final concentration was 10 mM, and this was applied to a second phenyl-Sepharose column. This column was washed with 50 mM Tris, pH 7.5, 200 mM ammonium sulfate, 10 mM CaCl 2 , and CaM was then eluted with 50 mM Tris, pH 7.5, 200 mM ammonium sulfate, and 10 mM EDTA. The CaM 1234 pellet was resuspended in 50 ml of 20 mM Tris, pH 7.5, 1 mM EGTA and applied to a MonoQ (10 ϫ 5; Amersham Pharmacia Biotech) FPLC column. Protein was eluted with 1 M potassium glutamate over an 85-ml linear gradient. CaM 1234 eluted around 150 mM potassium glutamate. After purification, both proteins were dialyzed into 50 mM MOPS, pH 7.0.
Densitometry-For Fig. 4D, the amount of CaM 1234 bound to the fusion protein was quantified by scanning the CaM immunoblot into Adobe Photoshop and counting the number of pixels above background in the area corresponding to the band in each lane. The standard curve was generated with known amounts of CaM 1234 .
Peptide Interactions-Peptides were synthesized at the PAN facility at Stanford University. Sequences for each peptide are given in the text. Stock solutions of peptides were suspended in water or Me 2 SO and diluted in water for working concentrations. Non-denaturing, non-reducing gel shift experiments were performed as described (3). Fluorescent measurements and quantification of the interaction of dCaM and peptides were performed as described (7). To assess the Ca 2ϩ dependence of the interaction of a specific peptide with dCaM, a spectrum of dCaM in buffer containing the indicated amount of free Ca 2ϩ (Molecular Probes) in the absence of the specific peptide was first obtained. A saturating amount of the peptide was then added to the cuvette and a second spectrum obtained, from which the first spectrum was then subtracted. Purified N-terminal and C-terminal fragments of CaM used in Fig. 6 were prepared by tryptic digestion and kindly provided by C. Klee (National Institutes of Health). These were added to stock starting mixtures of dCaM (33 nM) and peptide (at a concentration corresponding to its calculated K 0.5 for CaM binding) in incremental amounts. Excess unlabeled CaM was added at the end of every experiment to determine the fluorescence emission for no-added peptide and confirmed by obtaining the spectra of 33 nM dCaM in the absence of peptide.

Consideration of the IQ Motif as the Basis for Tethering-A
constitutive interaction between CaM and the channel has been postulated to keep CaM strategically positioned to sense Ca 2ϩ entry and to induce inactivation rapidly, but the molecular basis for the tethering is unclear. Because of its homology to Ca 2ϩ -independent CaM-binding sequences found in unconventional myosins and neuromodulin (17), we initially considered the possibility that the IQ motif in the C-terminal tail could serve as the tethering site for CaM at basal levels of [Ca 2ϩ ] i . Although our previous results (3) showed that the C-terminal tail could not bind CaM in the absence of Ca 2ϩ , we considered that tethering to the IQ motif might require Ca 2ϩ concentrations of 20 -100 nM, like those in resting cells, and might be missed in zero Ca 2ϩ .
To test this possibility, we chose to examine interactions between the IQ motif and CaM 234 , one of the Ca 2ϩ -insensitive CaM molecules that act as dominant negatives to prevent CDI. Because CaM 234 clearly lacks the ability to act as a Ca 2ϩ effector (3,7), any binding that it displayed with the IQ motif could be regarded as an unambiguous indication of a tethering interaction. Accordingly, we generated CaM 234 and wild-type CaM by bacterial expression and tested the purified proteins for interaction with the IQ motif in a gel shift assay (Fig. 1). In the absence of added peptide, the CaM 234 protein migrated faster than wild-type CaM protein, confirming that this CaM mutant is unable to bind Ca 2ϩ and undergo the characteristic Ca 2ϩ -induced shift in mobility. The addition of the IQ motif peptide did not induce a shift in the mobility of CaM 234 , regardless of whether in the presence of 1 mM Ca 2ϩ ( Fig. 1) or in 2 mM EGTA (data not shown). This indicated that CaM 234 cannot bind to the IQ motif and that the basis for its dominant negative effect must be sought elsewhere.
Our previous demonstration that a peptide containing the IQ motif could induce a shift in mobility of wild-type CaM even in the presence of 100 nM added Ca 2ϩ (3) does not in itself establish a tethering role for the IQ motif. Micromolar concentrations of CaM are required in this assay in order to visualize CaM, thus requiring micromolar concentrations of peptide to achieve appropriate stoichiometry. The concentration of peptide used in such assays is therefore more than 10-fold greater than the K d value of 50 nM for interactions between IQ motifcontaining peptide and CaM that we subsequently determined (7). Thus, the law of mass action would favor the detection of significant binding, even if the IQ motif interacted relatively weakly with CaM at 100 nM Ca 2ϩ .
Consideration of the I-II Loop as a Possible Tethering Site-To expand our search for a tethering domain, we scanned the putative intracellular portions of the ␣ 1C subunit for sequences with homology to known CaM-binding domains, using the criteria as defined by Rhoads and Friedberg (17). We found three sites in the I-II intracellular loop (amino acids 438 -451, 520 -532, and 542-555, Fig. 2A), one site in the cytoplasmic tail (amino acids 1565-1578, see below), but none in the N-terminal domain. Ivanina et al. (18) reported that CaM binds to a fusion protein containing the N-terminal domain; presumably CaM binds to a site that lacks homology to consensus sequences. The 520 -532 sequence conforms to a consensus 1-5-10 CaM-binding motif, and the sequences 438 -451 and 542-555 correspond to a consensus 1-8-14 type B motif. Chimeras of ␣ 1C and ␣ 1S had previously suggested a possible role for the I-II loop in CDI (19).
Accordingly, we synthesized peptides that contained these sequences and looked for interactions with 5-dimethylaminonaphthalene-1-sulfonyl-CaM (dCaM). In the presence of Ca 2ϩ , the 520 -532 peptide increased the fluorescence of dCaM, with a concentration dependence corresponding to a binding affinity of ϳ90 nM (Fig. 2B). No significant interaction was detected in the absence of Ca 2ϩ . No significant interaction, either in the presence or absence of Ca 2ϩ , was detected with the 438 -451 and the 542-555 peptides. To confirm the CaM-binding properties of the 520 -532 peptide, we generated GST fusion proteins that contained the I-II loop and, for comparison, additional GST fusion proteins incorporating a segment of the C-terminal tail that included the IQ motif (Fig. 2C). Potential interactions between the fusion proteins and CaM were examined in the presence or absence of Ca 2ϩ and detected by an immunoblot for CaM. As shown in Fig. 2C, the I-II loop fusion protein was able to bind CaM only in the presence of Ca 2ϩ but not in its absence, confirming the results obtained with fluorescently labeled CaM. This suggested that CaM was not bound constitutively to the I-II loop in zero Ca 2ϩ and was therefore unlikely to serve as the CaM-tethering site.
As a further test of the possible involvement of the I-II loop, if not in tethering, as a possible effector site, we generated an ␣ 1C subunit that lacked the amino acids 520 -532. This mutant was co-injected with ␣ 2 ␦ and ␤ 1 into Xenopus oocytes and L-type Ca 2ϩ currents were recorded. As shown in Fig. 2D, Ca 2ϩ -dependent inactivation and the dominant negative inhibition of CDI by CaM 1234 were both unaffected by the deletion in ␣ 1C . We took this as strong evidence that the deleted se- showing lack of effect on CDI of a 77␣ 1C mutant in which amino acids 520 -532 were deleted (77d13). Currents were recorded during a 400-ms depolarizing step to ϩ20 mV from a holding potential of Ϫ90 mV. I Ba and I Ca were obtained from the same oocytes, and traces were scaled to equalize peak currents. quence does not play a significant role in supporting either the tethering or effector aspects of Ca 2ϩ sensing in CDI.
A Broader Search for a Tethering Site in the C-terminal Cytoplasmic Tail-Lacking evidence for involvement of the I-II loop, we proceeded to focus on the fourth CaM-binding site identified in our scan, in the C-terminal tail. Our previous experiments had weighed against possible tethering sites in regions of the C-terminal tail other than the IQ motif, because a Ile 1624 3 Glu substitution within the IQ motif (I/E mutation) appeared to render the tail unable to bind CaM in 1 mM Ca 2ϩ or in EGTA (3). To check this result, we carried out further experiments, taking care to allow longer incubations of agarose-CaM with the in vitro translated [ 35 S]methionine-labeled ␣ 1C fragment (Fig. 3). The new data showed a clear Ca 2ϩ dependence of interaction between CaM and the C-terminal protein harboring the I/E mutation, although at reduced levels compared with wild-type protein. It is unlikely that this interaction represents residual binding to the mutated IQ site, since the concentration of 35 S-labeled peptide in the reaction was about 10 Ϫ17 M, far below the K d value we have calculated for a IQ motif peptide containing the I/E mutation (Ͼ5 M). A further hint that the cytoplasmic tail could support CaM binding outside of the IQ motif is provided by comparison of results from Qin et al. (5) with our own findings (7). Extreme disruption of the IQ motif (IQEYFRKAAAAAAA) spared CaM binding to a fusion protein containing amino acids 1542-1688 (numbering as in 77wt) (5), even though a less severe modification (IQEYFRKAAEYAAA) decreased the affinity of CaM for the IQ motif peptide by ϳ100-fold (7).
Analysis of amino acids 1565-1578 showed that it conformed to a 1-8-14, type A motif (17). Interestingly, this site contains the short sequence 1572 IKTEG 1576 , previously identified as essential for CDI by analysis of deletions or domain swaps (6,12,20,21). Further consideration of the entire region between ϳ1565 and ϳ1650 as a locus of possible tethering sites was suggested by recent studies documenting the binding of CaM to sequences containing the 1600 LLDQV 1604 motif (9 -11), another short sequence identified as important for CDI (6,12,20,21). Accordingly, we tested various GST fusion proteins containing these sequences for possible interactions, not only with CaM but also with the dominant negative CaM 1234 , as a definitive test for tethering (Fig. 4).
Our first attempts were made with a GST fusion protein incorporating amino acids 1477-1592, within which the 1572 IK-TEG 1576 sequence resides (Fig. 4A, gray segment). A second GST fusion protein consisted of the amino acids 1592-1875, which includes the sequence 1600 LLDQV 1604 and the IQ motif (Fig. 4A, black segment). The binding of CaM or CaM 1234 to these fusion proteins was examined in the presence of 100 nM Ca 2ϩ , to approximate basal [Ca 2ϩ ] i in quiescent cells. As shown in Fig. 4B (lanes 1 and 2), the 1477-1592 fusion protein failed to bind CaM (WT) or CaM 1234 (4-), suggesting that the corresponding stretch of ␣ 1C is not sufficient to serve a tethering function. In a somewhat different pattern, the 1592-1875 fusion protein was found to bind CaM but not CaM 1234 . Because this fusion protein included the IQ motif, the simplest interpretation was that this motif was responsible for binding Ca 2ϩ ⅐CaM, in a reaction driven by the high local concentration of IQ motif in this assay, despite the presence of only 100 nM Ca 2ϩ . In any case, it was clear that aa 1592-1875 of ␣ 1C were also not sufficient to serve a tethering function, since CaM 1234 failed to bind (Fig. 4B, lane 4).
Tests with the 1477-1592 and 1592-1875 fusion proteins did not exclude a remaining possibility that binding of the mutant CaM might depend jointly on essential sequences within both of the corresponding regions of ␣ 1C . This idea was tested with an additional fusion protein containing the amino acids 1551-1660 (Fig. 4A), depicted with both gray and black shading to indicate that it straddled the junction between the first two fusion proteins. In this case, both CaM and CaM 1234 bound equally well (Fig. 4B), conforming to the pattern expected for a genuine tethering region.
Ca 2ϩ Dependence of Tethering to a C-terminal Region-Initially, we attempted to confirm our results by looking for interactions between apocalmodulin and the candidate-tethering region by carrying out experiments in the absence of Ca 2ϩ (EGTA). This failed (Fig. 3C), suggesting that interactions between CaM and the 1551-1660 fusion protein might be critically different in zero Ca 2ϩ versus 100 nM Ca 2ϩ . This would be consistent with our previous finding that in the absence of Ca 2ϩ , agarose-CaM bound poorly to [ 35 S]methionine-labeled in vitro translated protein corresponding to the whole cytoplasmic tail (Fig. 3). Further experiments were carried out to look systematically for such a Ca 2ϩ dependence (Fig. 4D). Here we used CaM 1234 , which lacks any ability to interact with submicromolar Ca 2ϩ , to focus on possible Ca 2ϩ dependence arising from the channel sequence itself. Consistent with the previous result, there was little interaction of CaM 1234 and the 1551-1660 fusion protein in the absence of Ca 2ϩ . When Ca 2ϩ was added to the reaction, the degree of interaction was fully saturated at Ca 2ϩ concentrations of 100 nM or higher and was half-maximal at ϳ10 nM Ca 2ϩ (Fig. 4E). Thus, while dependent on Ca 2ϩ , the putative tethering interaction would be effective at typical cytoplasmic Ca 2ϩ concentrations in cells at rest. Romanin et al. (10) reached qualitatively similar conclusions in a recent study with a peptide comprising a stretch of C-terminal sequence (aa 1571-1585) contained within the sequence of residues studied here (Fig. 4A). By fluorescein-labeling the peptide, they were able to demonstrate a Ca 2ϩ dependence of peptide conformation even in the absence of CaM, although their estimate of the K d was ϳ100 nM Ca 2ϩ , ϳ10-fold higher than our value.
Non-contiguous Peptides Involved in CaM Tethering-To take a more systematic approach to the structural determinants of the CaM interactions in the region between amino acids 1551-1660, we examined a series of overlapping peptides that covered most of this region in a staggered fashion (Fig.  5A). These were designed as follows: A-(1558 -1579) The experiments with the peptide series were of considerable interest from more than one point of view. First, finding significant CaM interactions with multiple nonoverlapping peptides suggested that the determinants of CaM tethering might reside in noncontiguous stretches of amino acids within the ␣ 1C C terminus. Second, peptide A contains the signature sequence IKTEG, and peptide C includes the signature sequence LL-DQV, both of demonstrated importance for CDI (12,24). Third, the positions of the interacting peptides A and C may help explain the earlier pattern of results with GST fusion proteins (Fig. 4B); significant interactions with CaM and CaM 1234 were found for the fusion protein 1551-1660, which subsumes both peptides A-(1558 -1579) and C-(1585-1606), but no interaction was detected with fusion protein 1477-1592, which incorpo-rates A but not C. The interaction of CaM with fusion protein 1592-1875, which contains F, part of C, but not A, confirms our previous observation of a Ca 2ϩ -dependent, high affinity interaction of CaM with the IQ peptide (contained in F) and demonstrates that tethering at resting Ca 2ϩ levels requires at least all of A and C.
CaM Interactions with Various Peptides Show Differing Patterns of Ca 2ϩ Dependence-If peptides A and C contribute to the tethering site, one might expect the peptides to be able to bind CaM at ambient Ca 2ϩ levels in resting cells. To test this assumption, we determined the Ca 2ϩ dependence of CaM binding to both of these peptides and peptide F, which contains the IQ motif. The spectrum for dCaM at a given concentration of free Ca 2ϩ was subtracted from the spectrum of the same dCaM sample after addition of a saturating concentration of added peptide. This yielded accurate estimates of the peptide-dependent increase in dCaM fluorescence (Fig. 5C, top). The data showed that a significant portion of dCaM bound to peptide A at zero Ca 2ϩ and remained fairly constant up to 1 M free Ca 2ϩ , consistent with a site that serves a tethering function. At higher levels of free Ca 2ϩ , the dCaM signal in the presence of peptide showed a further increase, suggesting either additional dCaM binding or changes in the nature of the dCaM-peptide A interaction and in the environment of the fluorophore. Peptide C showed an intermediate Ca 2ϩ dependence (Fig. 5C, middle). Little dCaM bound in zero Ca 2ϩ , but binding increased with rising concentrations of free Ca 2ϩ (K 0.5 ϳ100 nM), suggesting that a significant portion of dCaM bound to peptide C at resting levels of Ca 2ϩ , and that peptide C may serve a tethering function. In contrast, peptide F showed little or no interaction at levels of Ca 2ϩ found in resting cells, and increasingly strong effects on the fluorescent signal as Ca 2ϩ was elevated, with a K 0.5 of ϳ400 nM (Fig. 5C, bottom). This buttressed the idea that Ca 2ϩ drives the CaM-IQ motif interaction, thereby accelerating inactivation once the channel has opened and Ca 2ϩ has risen.
These results with peptides A and C differ somewhat from those obtained with the fusion proteins in Fig. 4. No CaM interaction was detected with fusion protein 1477-1592, which contained peptide A, even if Ca 2ϩ was present; in the absence of Ca 2ϩ , no CaM interaction was detected with fusion protein 1551-1660, although it contained peptides A and C. The dCaM assay may reveal weaker interactions at single contact sites, whereas detection in the GST-pulldown assay may require higher affinity binding like that afforded by multiple points of interaction. Furthermore, in the absence of Ca 2ϩ , a fusion protein may attain a conformation that is not permissive for CaM binding.
The N-and C-terminal Domains of CaM and Interactions with Peptides A, C, and F-The finding that multiple, noncontiguous sequences in the C-terminal region interact with CaM, and do so with different Ca 2ϩ requirements, prompted us to perform new tests to determine which of these sequences might accelerate inactivation by binding CaM and which might tether CaM in readiness for such action. A clear criterion for the accelerator site was established by previous experiments with CaM mutants in which various Ca 2ϩ -binding EF hands were disabled (4). CDI was spared following expression of CaM 12 , a CaM mutant in which the two N-terminal EF hands were mutated to Ca 2ϩ -insensitive forms. In contrast CaM 34 , a CaM mutant in which the two C-terminal EF hands are disabled, ablated CDI in a dominant negative manner.
These results predicted that the accelerator site should interact preferentially with the C-terminal lobe of CaM, the region where wild-type Ca 2ϩ binding is critical, rather than the N-terminal lobe, where loss of Ca 2ϩ binding can be tolerated. On the other hand, the findings with dominant negative CaMs made no specific prediction about which domain(s) of CaM should interact with the tethering site, so long as this binding occurs at resting Ca 2ϩ levels (4). Accordingly, we set up competition experiments with CaM and its N-or C-terminal domains, generated by limited proteolysis (25). We examined the ability of CaM or its domains to interfere with the interaction between fluorescent CaM and peptides A, C, and F (Fig. 6). Since CaM and both the N-terminal and C-terminal lobes of CaM are Ca 2ϩ -binding proteins, their addition at the concentrations used in this competition assay would affect any attempts to buffer Ca 2ϩ at low levels and therefore affect the dCaM signal, independent of their ability to compete with dCaM. Accordingly, to be fully certain that changes in Ca 2ϩ concentration would not affect interpretation of the results, we used solutions containing 1 mM Ca 2ϩ to study CaM interactions with peptides A, C, and F. In all three cases, raising the concentration of unlabeled CaM progressively diminished the fluorescence signal generated by interaction of peptide with dCaM. In the case of peptide C (Fig. 6B), the N-and C-terminal domains of CaM were equally effective in reducing the fluorescent signal. In contrast, the hemi-CaMs competed differently for binding of dCaM to peptide F, which contains the IQ motif (Fig. 6C). The C-terminal lobe competed effectively, whereas the N-terminal lobe did not compete well (K i for C-terminal lobe ϳ46-fold lower than K i for N-terminal lobe). In agreeing with earlier experiments with CaM 34 (4), this preference supported the prevailing hypothesis that speeding of inactivation by CaM involves CaM binding to the IQ motif. The employment of the C-terminal domain of CaM as the Ca 2ϩ signal transducer finds precedent in the control of the Ca 2ϩ -activated K ϩ current in Paramecium (26).
Peptide A also demonstrated a preference for interaction with one lobe of CaM over the other (Fig. 6A), but in this case, it was the N-terminal lobe of CaM rather than the C-terminal lobe that competed more effectively with dCaM for binding to peptide A. The K i for the N-terminal lobe was ϳ10-fold lower than the K i for the C-terminal lobe, the opposite of what was found with peptide F.
The effector and tethering functions performed by the various peptides depend not only on the Ca 2ϩ dependence of the interaction (Fig. 5C) but also on the affinity of the peptides for CaM. All three peptides were found capable of interacting with CaM at the high Ca 2ϩ levels that allow CaM to perform its effector function. However, one would expect that at maximum, only two out of the three would be able to interact simultaneously with CaM, presumably one peptide sequence per N-or C-terminal lobe of CaM. This prompted us to characterize the concentration dependence of binding of peptides A or C to dCaM, with the goal of determining which peptide(s) would bind CaM most favorably in high Ca 2ϩ . We had previously shown that peptide F, which contains the IQ motif, interacts with high affinity with CaM (K 0.5 ϳ50 nM peptide) (7) (broken line in Fig. 7). In these new experiments, we determined that the interaction of peptide A with CaM conformed to a 1:1 binding curve, with a much weaker affinity (K 0.5 ϳ1 M peptide, right curve, Fig. 7). In contrast, the interaction of peptide C and CaM followed a steeper binding curve, with intermediate affinity (middle curve).

DISCUSSION
Our findings provided clear answers to fundamental but unanswered questions about CaM action as follows: how CaM is tethered to the L-type channel before performing its acceleratory function, how the CaM interacts with the effector IQ domain, and how the CaM sensor is able to act so quickly to initiate Ca 2ϩ -dependent inactivation once Ca 2ϩ entry has begun.
A Structural Basis for CaM Tethering-Knowing how CaM is tethered to resting L-type channels is critical to understanding how CDI occurs so quickly and how mutant CaM molecules exert their dominant negative effects. The C-terminal cytoplasmic domain was initially deemed unpromising as a locus for anchoring CaM because it failed to bind CaM in the absence of Ca 2ϩ (4). We re-examined the involvement of this region by focusing on peptide interactions with CaM 1234 , arguably the most stringent test for a genuine locus of CaM tethering (4). Because CaM 1234 completely lacks high affinity Ca 2ϩ interactions, we were able to run tests for interaction in the presence of nonzero Ca 2ϩ concentrations. The results demonstrated clearly that a portion of the ␣ 1C C-terminal domain was able to bind CaM 1234 or CaM, but only at 10 -100 nM Ca 2ϩ , levels normally maintained in cells at rest. Thus, tethering does occur in the C-terminal cytoplasmic region but is complicated by a Ca 2ϩ dependence intrinsic to the channel itself.
We found that the C-terminal tethering does not require the IQ motif but relies instead on two noncontiguous stretches of amino acids, roughly 51 and 24 aa to the N-terminal side of the IQ motif. The involvement in tethering of the more N-terminal group of residues (peptide A) is a novel finding, whereas our conclusions about the other set of amino acids (peptide C) are similar to those put forward in recent papers from the groups of Soldatov (10), Hamilton (9), and Maulet (11). The definitive test for tethering, binding of a dominant negative CaM molecule, is also a new result that we present here. Uncovering multiple sites for tethering relatively close to the IQ motif has important functional implications because the proximity of loci of tethering and acceleration would support a very fast development of CDI. It is also noteworthy that the sequences in peptides A and C are present in ␣ 1C subunits but not in ␣ 1A , ␣ 1B , or ␣ 1E , consistent with the fact that CDI is much faster and more prominent in L-type channels than in (P/Q)-, N-, or R-type channels.
Structural Basis of CDI Acceleration, Binding of C-terminal Lobe of CaM to IQ Motif-To identify the part of CaM that interacts with the IQ motif, we examined the competition between CaM fragments and fluorescent CaM for binding to IQ peptide. The finding that the IQ domain interaction selectively involves the C-terminal domain of CaM, not the N-terminal domain (Fig. 6C), provides a satisfying explanation of two important findings from previous biophysical experiments. First, FIG. 6. Peptides A and F bind preferentially to a single lobe of CaM. A-C, competition curves with single lobes of CaM prepared by tryptic digestion and purified for the indicated peptides, performed in the presence of 1 mM Ca 2ϩ . dCaM (33 nM) was mixed with peptide (at a concentration corresponding to its calculated K 0.5 for CaM binding (see Fig. 8B) and either the N-terminal lobe, C-terminal lobe, or intact CaM was added incrementally at the indicated concentrations. CDI remains intact in the presence of an excess of CaM 12 but not CaM 34 (4). Second, the cooperativity of CDI is much less than that of the transition from apocalmodulin to fully Ca 2ϩsaturated CaM (27).
A Model for CDI Based on Rapid Switching between CaM Configurations-Here we present a working hypothesis for Ca 2ϩdependent inactivation of L-type channels that takes into account recent findings on how the N-and C-terminal lobes of CaM help tether CaM and support its effector actions (Fig. 8).
Regions of the C-terminal cytoplasmic region of ␣ 1C are labeled A, C, and IQ, in accordance with the peptide segments used in our earlier binding studies. At resting Ca 2ϩ levels (left panel), calmodulin is depicted as bound to sequences A and C. This is based on the finding that CaM shows significant interactions with the corresponding peptides even at very low Ca 2ϩ (Fig.  5C). The CaM is shown as oriented with its N-terminal lobe associated with A, as suggested by Fig. 6A, whereas peptide C is depicted as lying between the two CaM lobes, reflecting its ability to interact with either lobe (Fig. 6B). This arrangement is subject to the caveat that the orientation of CaM lobes was examined in the presence of saturating Ca 2ϩ concentrations, but we believe that the results extend to basal Ca 2ϩ levels in resting cells since the binding of CaM 1234 to the tethering regions remained constant over a wide range of Ca 2ϩ concentrations (Fig. 4).
The IQ motif is pictured as not contributing to the CaM tethering, based on three pieces of evidence. First, dominant negative CaMs do not interact with the IQ motif (Fig. 1). Second, mutation of the IQ motif spares a significant degree of CaM binding to the cytoplasmic tail (Fig. 3). Third, CaM binding to a smaller fusion protein that contains the tethering sites is unaffected by disruption of the IQ motif (5).
Upon elevation of cytosolic Ca 2ϩ to micromolar levels (right panel), the IQ motif is pictured as shifting to the C-terminal lobe of CaM, whereas the N-terminal lobe remains anchored on peptide A and both lobes continue to harbor peptide C. In support of this are data showing that the C-terminal lobe of Ca 2ϩ /CaM interacts directly with the IQ motif (Fig. 6C). Our model favors the rapid development of inactivation, since it requires only minor conformational changes, namely IQ peptide association with the C-terminal lobe of CaM driven by the elevation of Ca 2ϩ . Restoration of resting levels of cytosolic Ca 2ϩ would favor a return to the original CaM configuration (left) and a prompt removal of inactivation, leaving the CaM tethered for another round of inactivation.
Implications of Multiple, Discrete Sequences for Tethering and Effector Action-This model deviates from the classical scenario for CaM action that originated from studies of CaMKII and myosin light chain kinase (28,29). In this scenario, both lobes of Ca 2ϩ /CaM wrap around a short target peptide and bind close to each other, thereby removing the repressive effect of the peptide and nearby residues on the catalytic site of the enzyme. A similar mechanism has been proposed for CDI (4). Although most CaM targets consist of a single compact sequence, like CaMKII and myosin light chain kinase, there is some precedent for multiple, non-contiguous interaction sites tethering a single CaM molecule in other enzymes such as glycogen phosphorylase kinase (30) and Bordetella pertussis adenylate cyclase (31).
Our hypothesis is most closely aligned with current thinking about gating of the SK potassium channel, another ion channel regulated by Ca 2ϩ through constitutively bound CaM (16,32). Recently, Schumacher and colleagues (33) provided an elegant crystal structural analysis of CaM bound to CaM binding domain (CBD) from the C-terminal cytoplasmic tail of the SK channel. In the CaM 2 CBD 2 complex, both lobes of CaM engage in multiple interactions with distinct ␣-helical stretches of the two CBDs. The C-terminal lobe of a CaM lacks bound Ca 2ϩ but interacts closely with separate helical segments from one CBD; the N-terminal lobe of the CaM holds two Ca 2ϩ ions and engages the other CBD. Thus, in striking correspondence with our inferences for the Ca 2ϩ channel, CaM tethering in the SK channel involves two ␣-helical segments and the effector interaction takes place by recruitment of a third ␣-helical segment. In both cases, binding of Ca 2ϩ to only one of the CaM lobes is sufficient to trigger gating, but in the L-type Ca 2ϩ channel this is the C-terminal lobe and not the N-terminal lobe as in the SK channel.
In the present case, these distinct and specific interactions may confer functional advantages for control of channel gating. Employing several points of attachment for a single CaM molecule may allow the Ca 2ϩ sensor to satisfy the disparate requirements of the CDI mechanism. On one hand, the C-terminal lobe must rapidly engage the IQ motif, of known importance for CDI; on the other hand, the sensor must stay securely attached, to remain immediately ready for repeated rounds of inactivation. The model we present would ensure the retention of a CaM molecule, allowing it to function over and over again, as implied by the persistent expression of CDI in planar bilayers bathed in solutions devoid of CaM (34).
Another Function for Peptides A and C, Control of Ca 2ϩindependent Inactivation?-Our biochemical observations fitted well with earlier biophysical studies of CDI (12,24) that pointed to importance of groups of amino acids, IKTEG and LLDQV, that fall within peptides A and C. Swapping either of these five-residue segments leads to significant changes in inactivation properties, not only a reduction in the Ca 2ϩ dependence but also a speeding of inactivation with Ba 2ϩ as a charge carrier. In combination, the domain swaps completely abolish all Ca 2ϩ dependence and greatly hasten inactivation of I Ba . Altered inactivation kinetics in the absence of Ca 2ϩ entry would not be predicted simply because of the removal of a tethered Ca 2ϩ sensor. The effects on I Ba provided a clue that one or both of these amino acid sequences may exert an effect beyond that expected for a passive tether. Therefore, our model attributes yet another function to the CaM tethering and effector sites. We hypothesize that sequences in peptides A and C not only tether CaM, but in binding apoCaM also help create a deceleratory signal to retard inactivation when cytosolic Ca 2ϩ levels are low. Modifying either sequence would alter the local conformation of the CaMtethering site and disrupt the braking effect, as manifested by the observed speeding of I Ba inactivation (12,24). When cytosolic Ca 2ϩ suddenly rises, the strongly favored interaction between CaM and the IQ motif will change the local configuration to an acceleratory state. We further postulate that there are important hydrophobic contacts between Ile 1624 and a local domain, as of yet unidentified, that are important for maintaining the braking effect when Ca 2ϩ is low and allowing the Ca 2ϩ -dependent acceleration of inactivation to be expressed when Ca 2ϩ rises. This is based on our observation that the hydrophobicity of the amino acid at position 1624 is inversely correlated with the rate of Ca 2ϩ -independent inactivation and the ability of channels to display accelerated inactivation in Ba 2ϩ (7). The mechanisms that translate CaM configuration into inactivation itself remain incompletely understood but probably involve the EF hand domain (21,35).
Calcium Is Required for CDI Independent of CaM-By using the Ca 2ϩ -insensitive CaM mutant, CaM 1234 , we found a requirement of Ca 2ϩ for the interaction of CaM with the tethering site on the channel. With a K d of ϳ10 nM, this requirement would be satisfied by levels of Ca 2ϩ found in resting cells and would be unlikely to contribute to the "sensor" function that has been proposed (10). Rather, our results are consistent with an alternative interpretation that the high affinity Ca 2ϩ binding fulfills a structural requirement, possibly analogous to Zn 2ϩ in zinc finger proteins. Furthermore, the K d value in vivo might be even lower than our estimate, as the whole channel could contribute structural elements, not present in our fusion protein, that increase the affinity for Ca 2ϩ . These results are also consistent with the bilayer experiments performed by Haack and Rosenberg (34), in which they observed the loss of CDI in some experiments after buffering [Ca 2ϩ ] i to ϳ20 nM; by working close to the calculated K d for the structural requirement for Ca 2ϩ , they may have lost the tethered CaM in those bilayers that failed to display CDI. Further characterization of the structural basis of this Ca 2ϩ dependence and its properties in situ are required.
The Role of CaM Tethering in CDF-While the results presented here mainly help to understand the role of CaM in CDI, we can put them into context of the role of CaM in CDF. The CaM-tethering site identified here likely serves for both CDI and CDF, since dominant negative CaM molecules also abolish CDF (3,7), and no other CaM-tethering site has been identified. Furthermore, mutation of isoleucine to alanine in the IQ motif eliminates CDI and promotes CDF (7). We therefore expect that that the Ca 2ϩ -dependent interaction of CaM with the IQ motif places the channel into a conformation that promotes facilitation under conditions in which the channel is repetitively depolarized. DeMaria et al. (36) have shown that the N-terminal lobe of CaM regulates CDF of the P/Q channel, whereas the C-terminal lobe controls CDI. Whether the gating signal is similarly bifurcated for the L-type channel is unknown. Several groups have suggested that CaMKII is essential for CDF of L-type channels (22,37). How the Ca 2ϩ -dependent interaction of the tethered CaM with the IQ motif might lead to activation of CaMKII is unclear.