Neuron, Vol. 22, 549558, March, 1999, Copyright 1999 by Cell Press

Calmodulin Is the Ca21 Sensor for Ca21-DependentInactivation of L-Type Calcium Channels

a few membrane thicknesses of the channel (Deisserothet al., 1996). Finally, a number of structurefunction re-sults suggest that inactivation results from Ca21 binding

Blaise Z. Peterson, Carla D. DeMaria,and David T. Yue*Program in Molecular and Cellular

to regions in the N-terminal third of the a1C COOH tail,Systems Physiologyshown as the Ca21 inactivation (CI) region in Figure 1.Departments of Biomedical EngineeringThis region contains a putative Ca21-binding motif (EFand Neurosciencehand; Figure 1) that is essential for inactivation (de LeonJohns Hopkins University School of Medicineet al., 1995; B. Z. Peterson et al., 1999, Biophys. J.,Baltimore, Maryland 21205abstract). Deletion of the C-terminal two-thirds of thea1C COOH tail spares Ca21-dependent inactivation, anddonation of the a1C CI region to an a1E channel backboneSummary(Soong et al., 1993) confers Ca21-dependent inactivationto a noninactivating background (de Leon et al., 1995).Elevated intracellular Ca21 triggers inactivation of L-typeA subsequent study reported that deletion of the entirecalcium channels, providing negative Ca21 feedbackEF hand region ablated Ca21-dependent inactivationin many cells. Ca21 binding to the main a1C channel(Zuhlke and Reuter, 1998), although there is not universalsubunit has been widely proposed to initiate suchagreement on the importance of the EF hand regionCa21-dependent inactivation. Here, we find that over-(Zhou et al., 1997).expression of mutant, Ca21-insensitive calmodulin (CaM)

Nonetheless, recent data hint that the Ca21 sensor forablates Ca21-dependent inactivation in a dominant-inactivation may reside outside the putative EF hand onnegative manner. This result demonstrates that CaMa1C. We (B. Z. Peterson et al., 1999, Biophys. J., abstract)is the actual Ca21 sensor for inactivation and suggestsand others (Zhou et al., 1997; Bernatchez et al., 1998)that CaM is constitutively tethered to the channel com-find that point mutations of presumed Ca21-coordinatingplex. Inactivation is likely to occur via Ca21-dependentresidues in the putative EF hand produce only modestinteraction of tethered CaM with an IQ-like motif onreductions of Ca21-dependent inactivation, fitting poorlythe carboxyl tail of a1C. CaM also binds to analogous with the general trend that such mutations reduce theIQ regions of N-, P/Q-, and R-type calcium channels,Ca21 affinity of legitimate EF hands by 10- to 1000-foldsuggesting that CaM-mediated effects may be wide-(Linse and Forsen, 1995). Another clue comes from the

spread in the calcium channel family.finding that a putative IQ CaM-binding motif (Rhoadsand Friedberg, 1997) located just C-terminal of the EFhand motif (IQ; Figure 1) is an additional essential regionIntroductionfor Ca21-dependent inactivation (Zuhlke and Reuter,1998). This finding raises the possibility that CaM might

L-type calcium channels trigger essential processes asbe involved in Ca21-dependent inactivation. Further-

diverse as muscle contraction (Fabiato and Fabiato, 1979;more, a very local rise of Ca21 near L-type but not other

Rios and Brum, 1987) and neuronal gene expression types of calcium channels is responsible for activity-(Sheng et al., 1990; Murphy et al., 1991; Deisseroth et al., dependent translocation of CaM to the nucleus of hip-1996, 1998). As such, the inactivation of L-type calcium pocampal neurons, leading to cyclic AMP responsechannels by elevated intracellular [Ca21], termed Ca21- elementbinding protein (CREB) phosphorylation (Deis-dependent inactivation, represents an enormously im- seroth et al., 1996, 1998), again underscoring an intimateportant negative feedback element in a wide spectrum relation between CaM and L-type channels. Finally, CaMof biological contexts (Brehm and Eckert, 1978). We has recently been shown to be the Ca21 sensor for smalland others have previously suggested that Ca21 binding Ca21-activated K channels (Xia et al., 1998) despite thedirectly to the pore-forming a1C subunit of the channel absence of effects by peptide or pharmacological CaMinitiates Ca21-dependent inactivation (Shirokov et al., inhibitors. CaM turns out to be so closely and constitu-1993; Imredy and Yue, 1994; de Leon et al., 1995). This tively complexed with the K channel that inhibitory moi-suggestion was based on three main lines of evidence. ties never gain access to cognate CaMs, much as foundFirst, Ca21-dependent inactivation is unaffected by phar- for phophorylase kinase (Dasgupta et al., 1989), of whichmacological maneuvers affecting a broad spectrum of CaM is an integral subunit. If CaM were similarly ap-potential intervening signaling events, including cal- posed to the L-type channel, CaM could serve as themodulin (CaM) activation (Imredy and Yue, 1994; Victor Ca21 sensor for Ca21-dependent inactivation, while sat-et al., 1997; Zuhlke and Reuter, 1998). Second, even isfying the three lines of evidence that initially high-high concentrations of the intracellular Ca21 chelator lighted the a1C EF hand as a potential Ca21 sensor.BAPTA are unable to completely suppress Ca21-depen- Motivated by these clues, we here demonstrate thatdent inactivation (Neely et al., 1994), suggesting that the CaM is the actual sensor for Ca21-dependent inactiva-Ca21-sensing site for inactivation must be within one to tion of L-type calcium channels. Calmodulin is probably

tethered to the L-type channel complex in a constitutivemanner, independent of Ca21-activation of CaM. Our* To whom correspondence should be addressed (e-mail: dyue@data suggest that when Ca21 binds to this tethered CaM,bme.jhu.edu).

These authors contributed equally to this work. the resulting Ca21CaM complex binds to the IQ-like

Neuron550

Figure 1. Putative Transmembrane Folding Model for the Main Sub-unit (a1C) of the L-Type Calcium Channel

The CI region of the cytoplasmic COOH tail contains structuresimportant for Ca21-dependent inactivation, namely a putative EFhand and an IQ-like motif for CaM binding. Homologous domainsof the channel are indicated by the Roman numerals I to IV. Numbersin parentheses indicate the amino acid residues corresponding tovarious landmarks. Cytoplasmic loops face downward.

motif in the carboxyl tail of a1C, and this binding eventleads to channel inactivation.

Results

Mutant CaM Exerts a Dominant-Negative Knockoutof Ca21-Dependent InactivationFigure 2 shows the results of coexpressing L-type cal-cium channels with mutant CaM lacking high-affinityCa21 binding, as observed in mammalian HEK 293 cells.As a control, we coexpressed recombinant L-type cal-cium channels with wild-type rat CaM (Figure 2A). Cur-rents elicited by a 300 ms voltage step to 210 mV dem-onstrate Ca21-dependent inactivation properties likethose observed without coexpression of CaM (de Leonet al., 1995). With Ba21 as charge carrier (black trace),there is little inactivation, as expected from the highselectivity of Ca21-dependent inactivation for Ca21 overBa21 (Brehm and Eckert, 1978). By contrast, with Ca21 ascharge carrier (gray trace), Ca21-dependent inactivationcauses the current to decay by about half at the end ofthe depolarizing pulse. The Ca21 current trace has been Figure 2. Coexpression of L-Type Calcium Channels with Wild-Type

CaM or Mutant CaM1,2,3,4scaled upward to facilitate the comparison of kinetics.(Top) Schematic diagram of the a1C subunit of the L-type channel.The average properties of multiple cells confirm the nor-EF indicates the putative EF hand motif, and IQ signifies the putativemal gating behavior of channels with overexpression ofIQ motif for CaM binding.wild-type CaM. The fraction of remaining current at the(A) Whole-cell currents from cells overexpressing wild-type CaM

end of 300 ms voltage steps (r300) bears a U-shaped (WT CaM). Upper, exemplar Ba21 (black) and Ca21 (gray) currentsdependence upon voltage with Ca21 (closed symbols), elicited by a depolarizing step to 210 mV, illustrating robust Ca21-as expected for inactivation that is driven by Ca21 entry dependent inactivation. Here and throughout, Ca21 current has been

scaled upward z33 to facilitate comparison of kinetics, and tailrather than voltage. The corresponding r300 relation withcurrents have been clipped for display clarity. Middle, fraction ofBa21 (open symbols) declines little with voltage, and thecurrent remaining at the end of 300 ms depolarizations (r300) averageddistance between Ca21 and Ba21 r300 relations (f) provides from n 5 5 cells. Open and closed circles show the respective

a quantitative indication of the strength of Ca21-depen- relations with Ba21 and Ca21 as charge carrier, here and throughout.dent inactivation. Average plots of peak current density f is the difference between Ba21 and Ca21 r300 relations at 210 mV(Ipeak) as a function of voltage, shown at bottom, are and functions as a measure of the strength of Ca21-dependent inac-

tivation. Lower, peak current density (Ipeak) measured during voltageessentially the same as observed in the absence of CaMsteps to the indicated potentials, averaged from n 5 5 cells.overexpression (data not shown). Hence, overexpres-(B) Whole-cell currents from cells overexpressing mutant CaM1,2,3,4,sion of wild-type CaM did not by itself perturb L-typeindicating a knockout of Ca21-dependent inactivation. Identical for-

channel gating. mat to (A). Averaged data for n 5 6 cells.By contrast, Ca21-dependent inactivation was essen- (C) Relations between the strength of inactivation (f) and Ipeak mea-

tially abolished upon coexpression of L-type channels sured with Ca21 at 210 mV. The lack of overlap between relations forwild-type CaM and CaM1,2,3,4 argues that the knockout of inactivationwith mutant CaM1,2,3,4 (Figure 2B) in which aspartate haswith CaM1,2,3,4 reflects a primary defect in the Ca21 sensor ratherbeen mutated to alanine in the first Ca21 ligand positionthan lowered expression of the current observed with mutant CaM.of all four EF hands. Such mutations diminish the Ca21Lines are drawn by eye.

affinity of CaM by orders of magnitude (Putkey et al.,

Calmodulin Mediates Calcium Channel Inactivation551

1989). Interpretation of the loss of inactivation was ini-tially complicated by the depressed overall expressionof L-type current produced by coexpression of mutantversus wild-type CaM (compare Ipeak, Figures 2A and 2B).This raised the possibility that the disappearance ofinactivation arose from decreased Ca21 entry drivinginactivation, rather than from a primary defect in Ca21

sensing. However, plots of the strength of inactivationf versus peak current density (Figure 2C) excluded thispossibility, as the corresponding relations for mutantand wild-type CaM are clearly disjoint.

The knockout of inactivation not only demonstratesthat CaM is the Ca21 sensor for Ca21-dependent inacti-vation but also indicates that CaM is probably tetheredto the channel complex in a constitutive fashion. Overex-pression of Ca21-insensitive CaM would not ablate inac-tivation if tethering of CaM were not required to induceinactivation; Ca21 still binds to functional endogenousCaM regardless of the presence of mutant CaM. There-fore, the endogenous CaM would trigger inactivation.Ca21-insensitive CaM would disrupt inactivation only iftethering of CaM were required to trigger channel inacti-vation. Overexpressed, mutant CaM would then saturatesuch tethering sites, while being incapable of furtherCa21-dependent interaction with the channel to produceinactivation. Only in this manner would mutant CaM ex-ert a dominant-negative knockout of inactivation.

To gain initial insight into the critical interaction sur-faces between CaM and its cognate calcium channel,we performed the analogous CaM coexpression experi-ments with the chimeric calcium channel EC-1D3 (Figure3, top) (de Leon et al., 1995). EC-1D3 comprises an a1Ebackbone (Soong et al., 1993) and the N-terminal thirdof the a1C COOH tail (CI region; Figure 1). a1E itself lackssuch inactivation, but inclusion of the CI region of a1Cconfers Ca21-dependent inactivation upon EC-1D3. If allCaM interactions with the L-type channel are limited tothe CI region of a1C, then coexpression of EC-1D3 andmutant CaM1,2,3,4 should also result in a dominant-nega- Figure 3. Coexpression of Chimeric EC-1D3 Calcium Channels withtive knockout of Ca21-dependent inactivation. Wild-type CaM or Mutant CaM1,2,3,4

Figure 3 presents the results of coexpressing EC-1D3 The format is identical to that in Figure 2.with CaM, following a format identical to that in Figure (A) Averaged data from n 5 8 cells.

(B) Averaged data from n 5 6 cells. Notice the knockout of Ca21-2. While coexpression of wild-type CaM does not alterdependent inactivation with overexpression of mutant CaM1,2,3,4.Ca21-dependent inactivation or gating of EC-1D3 (Fig-(C) Disjoint relations indicate a true defect in the Ca21 sensor withure 3A), coexpression of mutant CaM1,2,3,4 ablates Ca21- CaM1,2,3,4.dependent inactivation (Figure 3B). The suppression of

inactivation reflects a primary defect in Ca21 sensingrather than depressed current density (Figure 3C). These antibody reveals Ca21-dependent CaM binding to allresults motivated detailed biochemical investigation of fusion proteins containing the IQ region but not to thepotential CaM interaction with the N-terminal third of a1C EF hand region alone (Figure 4B). GST protein servesthe a1C COOH tail. as the negative control. These results suggested that

the IQ-like motif might indeed be the CaM interactionsite.Ca21-Dependent Interaction of CaM

with the COOH Tail of a1C Gel shift assays explicitly confirmed Ca21-dependent,CaM binding to the IQ-like domain of a1C. In this assay,We established CaM binding to the CI region of a1C by

glutathione S-transferase (GST) pulldown experiments. mixtures of CaM and a peptide spanning the a1C IQregion were run on a nondenaturing polyacrylamide gelFusion proteins made up of GST and various segments

of the a1C CI region (Figure 4A) were bound to glutathi- in the presence of Ca21 (Figure 4C). In the absence ofpeptide, Ca21CaM runs as a single, peptide-free band.one-sepharose beads and exposed to purified CaM ei-

ther in the presence or absence of Ca21. After washing, With increasing peptide to CaM ratios, a second, slowermobility band appears (marked by a triangle), represent-proteins retained by the beads were eluted, run on SDS

PAGE, and transferred to nitrocellulose membrane. ing a peptideCa21CaM complex. As expected, there isa graded and reciprocal interchange toward the slowerWestern blot analysis of the membrane with anti-CaM

Neuron552

a Ca21-dependent manner. Ca21-dependent CaM bind-ing to this site is likely to be a transduction step in Ca21-dependent inactivation, because deletion of this domainabolishes such inactivation (Zuhlke and Reuter, 1998).

Although CaM interaction with the IQ-like region isnominally Ca21-dependent, could this CaM-binding do-main also serve to tether CaM to the L-type channel ina constitutive fashion? It is conceivable that appreciableCaM binding to the IQ motif might persist at restinglevels of Ca21 (z100 nM), given a sufficiently high localconcentration of the IQ-like domain to drive Ca21 andpeptide binding to CaM (Bayley et al., 1996). This possi-bility is illustrated by consideration of the following,simplified reaction scheme: CaM 1 Ca21 CaMCa21 1peptide peptideCaMCa21. This reaction diagramillustrates the general property that the Ca21 depen-dence of CaM binding is inherently dependent upon theconcentration of target peptide. In such a scenario, theC-terminal lobe of CaM, with its higher Ca21 affinity(James et al., 1995), would tether CaM to the IQ-likeregion at rest. Elevation of Ca21 during channel openingwould additionally activate the lower Ca21-affinity lobeof CaM (N-terminal), so that it could also bind to the IQregion and thereby initiate channel inactivation. If thissituation holds true, then there should be little Ca21-dependent inactivation upon coexpression of L-typechannels with a mutant CaM lacking Ca21 binding in itsN-terminal lobe. Such mutant CaM would be tetheredby its C-terminal lobe at rest, but elevated Ca21 wouldproduce no Ca21 activation of its N-terminal lobe andtherefore no channel inactivation. Conversely, Ca21-

Figure 4. Biochemical Evidence for Ca21-Dependent CaM Binding dependent inactivation should be preserved upon over-to an IQ-Like Motif on a1C expression of mutant CaM lacking Ca21 binding in its(A) Schematic diagram of GST fusion proteins probed for CaM bind- C-terminal lobe. This mutant CaM would fail to occupying. Numbers correspond to amino acid residues in the a1C subunit. the IQ motif at rest, leaving endogenous, wild-type CaM(B) GST pulldown of CaM. Immobilized GST fusion proteins (as in to populate this site and drive inactivation upon eleva-[A]) were incubated with CaM in the presence or absence of Ca21.

tion of Ca21.After washing, CaM bound to fusion proteins was eluted and re-Figure 5 summarizes the results of coexpressing L-typesolved by SDSPAGE and blotted to nitrocellulose membrane.

channels and mutant CaM with selective impairment ofEluted CaM is visualized by Western blot analysis with anti-CaMantibody. The results indicate that a1C-(15201732) and a1C-(1520 Ca21 binding in either the N-terminal or C-terminal lobe,1667) bound CaM in a Ca21-dependent manner. Purified CaM (250 following a format identical to that in Figure 2. CaM1,2ng) is run as reference. (Figure 5A, top) lacks appreciable Ca21 binding in its(C) Gel shift analysis of CaM binding to the 21 amino acid peptide

N-terminal lobe, because the aspartate-to-alanine mu-corresponding to the L-type channel IQ motif (a1C-IQ, shown at top).tations described for CaM1,2,3,4 have now been restrictedPurified CaM was reacted with various molar ratios of peptide into the two EF hands of the N-terminal lobe. Overexpres-the presence of 2 mM Ca21. The reaction mixture was resolved by

12% nondenaturing PAGE. The lower band corresponds to free sion of CaM1,2 spares Ca21-dependent inactivation (Fig-CaM, while the higher band (triangle) corresponds to a CaMpeptide ure 5A), which appears to be very similar in strengthcomplex. Similar experiments performed in the absence of Ca21 to that observed with wild-type CaM (Figure 2A). Byshowed no gel shift (data not shown).

contrast, CaM3,4 lacks appreciable Ca21 binding in itsC-terminal lobe as a result of analogous asparate-to-alanine mutations in the C-terminal lobe. Overexpres-mobility band with increasing peptide to CaM ratios.sion of CaM3,4 produces a clear knockout of Ca21-depen-This pattern of mobility shifts as a function of differentdent inactivation (Figure 5B). Analysis of the strength ofpeptide to CaM ratios is very similar to that observed forCa21-dependent inactivation (f) and Ipeak, as shown inCaM interaction with the C0 region of NMDA channelsFigure 5C, demonstrates that the respective sparing and(Ehlers et al., 1996), which underlies the Ca21-dependentknockout of inactivation by CaM1,2 and CaM3,4 reflectinactivation of NMDA channels (Zhang et al., 1998;defects in Ca21 sensing rather than altered expressionKrupp et al., 1999). Moreover, in the absence of Ca21,levels of Ca21 current. These results are the oppositeonly a single, peptide-free band appears regardless ofof the predictions of a model in which the IQ-like domainpeptide to CaM ratio (data not shown). Finally, peptidesserves as both tether and Ca21-dependent binding tar-corresponding to the EF hand and IIIII loop regions ofget for CaM. It thus seems likely that CaM is constitu-a1C failed to retard the mobility of CaM in the presencetively tethered to the channel complex by regions out-or absence of Ca21 (data not shown). These results dem-

onstrate that the IQ-like domain of a1C can bind CaM in side of the IQ or CI regions.

Calmodulin Mediates Calcium Channel Inactivation553

Figure 6. CaM Binding to Analogous IQ-Like Regions of N-, P/Q-,and R-Type Calcium Channels

(A) GST pulldown of CaM by GST fusion proteins of the CI regionsof a1E (residues 1691 to 1874 from L15453 [Soong et al., 1993]) anda1A (residues 1778 to 1969 from M64373 [Starr et al., 1991]) andthe N terminus of the olfactory cyclic nucleotidegated channel(OCNC-N). All of the fusion proteins bound CaM in a Ca21-dependentmanner. Purified CaM (250 ng) is shown for reference.(B) Upper, alignments of IQ-like regions from L-type (a1C, residues1647 to 1667 of X15539 [Mikami et al., 1989]), N-type (a1B, residuesFigure 5. Coexpression of L-Type Calcium Channels with Mutant1856 to 1876 of D14157 [Fujita et al., 1993]), P/Q-type (a1A, res-CaMs Lacking Appreciable Ca21 Binding in the N-Terminal (CaM1,2)idues 1906 to 1926 of M64373 [Starr et al., 1991]), and R-type (a1E,or C-Terminal (CaM3,4) Lobesresidues 1819 to 1839 of L15453 [Soong et al., 1993]) calcium chan-

The format is identical to that in Figure 2. nels. The canonical IQ motif is shown at top (Houdusse et al., 1996),(A) Averaged data from n 5 6 cells, showing no effect on Ca21- where x corresponds to any amino acid. Asterisks above the L-typedependent inactivation during overexpression of CaM1,2. channel sequence indicate phenylalanine residues corresponding(B) Averaged data from n 5 7 cells, indicating a knockout of Ca21- to the conventional 1 to 14 CaM-binding motif. Lower, gel shiftdependent inactivation with overexpression of mutant CaM3,4. assays performed with purified CaM and varying molar ratios of IQ-(C) The solid curve drawn through CaM1,2 points is reproduced from like peptides from N-, P/Q-, and R-type calcium channels, followingthe wild-type CaM relation shown in Figure 2C, indicating no detect- the identical format to that of Figure 4C. Experiments were per-able change in Ca21-dependent inactivation with overexpression of formed in the presence of Ca21 and indicate CaM binding to IQ-likeCaM1,2. regions of P/Q-, R-, and perhaps N-type calcium channels.

CaM Interaction with N-, P/Q-, and R-Typeregions from both R-type and P/Q-type channels bindCalcium ChannelsCaM in a Ca21-dependent manner. Gel shift assays fur-Given the robust Ca21-dependent binding of CaM to thether demonstrate that regions analogous to the IQ-likea1C CI region, we wondered whether the absence ofdomains from N-, R-, and P/Q-type channels also bindCa21-dependent inactivation in other types of high-CaM (Figure 6B), though the N-type channel interactionthreshold calcium channels (Patil et al., 1998) could beappears much weaker. These results suggest that theattributed to the lack of CaM binding to the analogousabsence of Ca21-dependent inactivation in N-, P/Q-, andregions of these other channels. Figure 6A shows theR-type channels does not result from a failure to bindGST pulldown experiment for GST fusion proteins thatCaM but rather a failure to transduce CaM binding intocontain corresponding CI regions of R-type (a1E) andinactivation. Nonetheless, it seems likely that CaM bind-P/Q-type (a1A) calcium channels. The NH2 tail of the olfac-ing to N-, P/Q-, and R-type channels is transduced intotory cyclic nucleotidegated channel (OCNC-N) serves

as a positive control (Liu et al., 1994). Surprisingly, CI other, as yet unknown functional sequelae.

Neuron554

Discussion variations within the channel IQ-like domain could tipthe balance among Ca21-dependent binding, Ca21-inde-pendent binding, and Ca21-dependent unbinding, all ofWe have demonstrated that CaM is the Ca21 sensor forwhich have been observed in unconventional myosinsCa21-dependent inactivation of L-type calcium channels.(Bahler et al., 1994). It would be interesting if differentOverexpression of mutant CaM that lacks high-affinitychannel variants exploit this design feature to alter theCa21 binding (CaM1,2,3,4) suppresses Ca21-dependent in-polarity and/or nature of gating effects arising from Ca21activation in a dominant-negative manner, demonstrat-activation of CaM.ing that CaM is the Ca21 sensor for inactivation. The

A second unusual feature of CaM interaction with theresult that Ca21-insensitive CaM can produce a dominant-L-type channel is the selective knockout of inactivationnegative effect in the presence of functional endoge-by mutant CaMs lacking appreciable Ca21 binding in thenous CaM suggests that CaM is constitutively tetheredC-terminal but not the N-terminal lobe of CaM (Figureto the channel complex in a manner that does not require5). Our findings appear to be a mammalian example ofCa21 binding to CaM. We argue that Ca21-dependentthe functional bipartition of CaM action discovered inbinding of tethered CaM to the IQ-like domain on a1C isParamecium by Saimi and Kung (1994). They found thata crucial transduction step in Ca21-dependent inactiva-mutations in the N-terminal lobe of CaM disrupted Ca21tion; this region binds CaM in a Ca21-dependent manner,activation of a K channel, while mutations in the C-termi-and deletion of the IQ-like region abolishes such inacti-nal lobe disrupted Ca21 activation of a Na channel, thusvation (Zuhlke and Reuter, 1998). The Ca21-independentrevealing a potentially general mechanism throughtethering site is still unknown, but the IQ-like region iswhich CaM could enrich its signaling repertoire.not likely to serve as both a constitutive tether and a

Finally, we also expect that CaM binds to the L-typeCa21-dependent binding target for CaM, based on ex-channel complex through interactions that do not re-periments with mutant CaMs lacking efficient Ca21 bind-quire Ca21 activation of CaM. Otherwise, overexpressioning in either N-terminal (CaM1,2) or C-terminal (CaM3,4)of Ca21-insensitive CaM would not interfere appreciablylobes. Our findings have implications for unusual fea-with Ca21-dependent inactivation, because endoge-tures of CaM recognition, CaM-mediated effects on anous, wild-type CaM is certainly present to undergovariety of ionic channels, and potential features of aCa21-dependent interaction with accessible CaM-bind-molecular mechanism of Ca21-dependent inactivationing domains. Futhermore, mutant CaMs are unlikely toof L-type channels, as discussed below.exert their effects through unanticipated downregula-tion of endogenous CaM, because it seems implausibleUnusual Features of CaM Interactionthat CaM1,2, CaM3,4, and CaM1,2,3,4 produce distinct effectswith Calcium Channelson inactivation via differential regulation of native CaM.The Ca21-dependent interaction of CaM with the L-typeHence, mutant CaM must be occluding access of en-channel is remarkable in several regards. First, IQ do-dogenous CaM to locations critical for Ca21-dependentmains generally support Ca21-independent binding toinactivation, and the occlusion does not require Ca21CaM (Rhoads and Friedberg, 1997), although there isactivation of CaM. The most likely scenario is that CaM is

precedent for Ca21-dependent binding (Bahler et al.,constitutively tethered to the channel complex through

1994), with minor corruption of the canonical motif (Fig-interactions that are independent of Ca21 activation of

ure 6B, top). In the case of the a1C IQ motif, the demon- CaM. Tight and constitutive interaction with CaM fitsstrated Ca21-dependent binding to CaM is surprising, as well with the insensitivity of Ca21-dependent inactivationthe channel satisfies most of the structural requirements to pharmacological CaM inhibitors (Imredy and Yue,believed to be important for Ca21-independent binding 1994; Victor et al., 1997; Zuhlke and Reuter, 1998), justto CaM by a genuine IQ site (Houdusse and Cohen, as peptide inhibitors of CaM do not affect phosphorylase1995, 1996; Houdusse et al., 1996). In particular, the a1C kinase (Dasgupta et al., 1989), an enzyme of which CaMIQ region has glutamine and arginine at the second and is an integral subunit. Constitutive binding to CaM wouldsixth positions of the IQ template (Figure 6B, top), and also provide a ready molecular explanation for whythese two residues may be the most important for Ca21- Ca21-dependent inactivation could still be observed inindependent coordination of the C-terminal lobe of CaM L-type channels reconstituted in planar lipid bilayersin a semiopen conformation (Houdusse and Cohen, following incorporation of sarcolemmal vesicles (Haack1995, 1996). The only discrepancy between the a1C IQ- and Rosenberg, 1994).like region and the canonical IQ template is the lysine The finding that the a1C CI region only supports Ca21-instead of a glycine at the seventh position of the IQ dependent CaM binding was unexpected, because ourtemplate (Figure 6B, top). This difference could account initial hunch was that the CI region would also supportfor the lack of Ca21-independent binding to CaM (Hou- Ca21-independent tethering of CaM to the channel. Thedusse et al., 1996) and cautions against considering the latter scenario provided a simple explanation for thea1C sequence as a true IQ site. In fact, closer inspection dominant-negative knockout of Ca21-dependent inacti-of the a1C IQ peptide (Figure 6B, top) reveals the pres- vation observed with EC-1D3 and CaM1,2,3,4 (Figure 3B).ence of a more traditional, Ca21-dependent 114 CaM- However, given that CaM interaction with the a1C CIbinding motif (Ikura et al., 1992), represented by a pair region is exclusively Ca21 dependent (Figure 4), Ca21-of phenylalanines with characteristic spacing (marked independent tethering of CaM to the channel complexby asterisks). The conventional mode of CaM interaction must involve portions of a1C outside of the COOH tail.with 114 sites could account for the Ca21-dependent Such tethering must involve other elements in the chan-binding of CaM to the channel rather than the chemistry nel complex, with possibilities including cytoskeleton

and various anchoring proteins (Johnson and Byerly,of an IQ-like mechanism. Nonetheless, subtle sequence

Calmodulin Mediates Calcium Channel Inactivation555

1993; Gao et al., 1997). In addition, the results with EC-1D3 (Figure 3B) suggest that Ca21-independent teth-ering of CaM occurs at a homologous site on the a1Echannel complex, whether it be on the channel itself orassociated cytoskeletal elements. It will be interestingto identify which portions of the L-type channel complexand CaM contribute to Ca21-independent binding andto determine how widely such interaction is sharedwithin the family of voltage-gated calcium channels.

Direct CaM Modulation of Ionic ChannelsApart from the L-type calcium channel, several otherionic channels may also be directly gated or modulatedby CaM, but their CaM binding sites remain to be identi-fied in several cases. Ca21 activation of sodium andpotassium channels in Paramecium is believed to in-volve direct CaM modulation (reviewed by Saimi andKung, 1994). Ca21CaM inhibits the opening of sarco-plasmic reticulum Ca21 release channels (ryanodine re-ceptors) studied in lipid bilayers (Smith et al., 1989), andseveral CaM binding sites have been identified (Chenand MacLennan, 1994). The role of this CaM inhibitionin physiological function remains unclear, however. Trp(Motell and Rubin, 1989) and trpl (Phillips et al., 1992)are putative Drosophila ion channels that were clonedon the basis of CaM binding and may be important infly phototransduction. Mammalian NMDA (Ehlers et al.,1996; Zhang et al., 1998; Krupp et al., 1999) and olfactorycyclic nucleotidegated ion channels (Chen and Yau, Figure 7. Current View of the Mechanism of Ca21-Dependent Inacti-1994; Liu et al., 1994; Grunwald et al., 1998) are both vation of L-Type Calcium Channelsinhibited by Ca21CaM through identified CaM binding In the resting state, which is favored by hyperpolarization, access

of tethered CaM to the IQ-like region is restricted. Upon depolariza-sites, and this inhibition has important implications fortion, the channel transitions to the open state, in which tethered CaMsynaptic physiology and olfaction, respectively. Mosthas ready access to the IQ-like domain. Accumulation of intracellularrecently, the small conductance Ca21-activated K chan-Ca21 at the inner mouth of the channel leads to Ca21-activation of

nel (SK channel) has been shown to be activated directly CaM. Interaction of Ca21CaM with the IQ-like region then inducesby Ca21CaM, with identified regions mediating CaM channel inactivation.interaction (Xia et al., 1998).

Of all the channels subject to direct CaM-mediatedmodulation, the SK channel, and probably the L-type the inhibition by binding to a site that overlaps the inhibi-calcium channel, are so far special in demonstrating tory domain, thereby separating catalytic and inhibitorytight, Ca21-independent tethering of CaM to the channel. enzyme domains (e.g., Means, 1988). In the case of theSuch constitutive association of CaM with ionic chan- L-type channel, the IQ-like domain is the correlate tonels is particularly well suited for local Ca21 communica- the autoinhibitory domain of CaM kinase II, except thattion, in which the Ca21 sensor responds rapidly and this domain is permissive for channel opening (cataly-selectively to local channel activity rather than to global sis). The proximity of the IQ-like domain to IVS6, whichelevation of Ca21 (Roberts et al., 1990; Imredy and Yue, corresponds to gating regions (catalytic site) of Shaker1992; Yue, 1997). K channels (Liu et al., 1997; Holmgren et al., 1998), fits

nicely with the CaM kinase II analogy. An interestingtwist in the L-type channel is that the EF hand domainPotential Molecular Mechanism

of Ca21-Dependent Inactivation may itself bind Ca21 (M. Villain et al., 1999, Biophys. J.,abstract), so that Ca21-dependent conformational changeThe identification of CaM as the Ca21 sensor for Ca21-

dependent inactivation of L-type channels may repre- of both CaM and the target binding molecule could beinvolved.sent a particularly helpful step in establishing a genuine

molecular mechanism of inactivation. Because much is How is the IQ-like domain permissive for channelopening in the absence of Ca21CaM? One possibilityknown about the mechanism of CaM action in other

systems, a number of interesting possibilities can be is that the IQ-like region binds to another receptor sitein the absence of Ca21CaM and that this bound com-raised about potential L-type channel mechanisms, un-

der the assumption of design conservation across mole- plex is permissive for gating. Alternatively, the interac-tion may prohibit inhibitory binding of the receptor tocules.

What does Ca21CaM do upon binding the IQ-like other portions of the channel. Both of these possibilitiesfind precedent in the ion transport literature. Interactiondomain of a1C? In the case of CaM kinase II and other

classic CaM-regulated enzymes, an autoinhibitory do- between N and C termini of the olfactory cyclic nucleo-tidegated channel is permissive for channel opening,main prevents catalytic activity until Ca21CaM disrupts

Neuron556

GST Pulldown of CaMand Ca21CaM binding to the N terminus disrupts theRegions containing the N-terminal portions of the carboxyl tail ofinteraction, leading to channel inactivation (Varnum anda1C (Mikami et al., 1989) (X15539), a1E (Soong et al., 1993) (L15453),Zagotta, 1997); the CaM binding site on a plasma mem-and a1A (Starr et al., 1991) (M64373) were PCR amplified and sub-brane Ca21ATPase pump inhibits Ca21 transport, and cloned into the BamHIEcoRI sites of pGEX 2T (Pharmacia) to gener-

Ca21CaM binding relieves the inhibition (Enyedi et al., ate a1C-(15201732), a1C-(15201667), a1C-(15201551), a1E-(16911874), and a1A-(17781969), where the numbers within parentheses1989; James et al., 1995). Given the essential role of thecorrespond to the amino acid positions of the respective a1 subunitsa1C EF hand motif in mediating Ca21-dependent inactiva-fused to GST. An overnight culture of DH5a cells (Life Technologies)tion (de Leon et al., 1995; B. Z. Peterson et al., 1999,carrying plasmid DNA encoding each of these fusion proteins wasBiophys. J., abstract), it seems likely that the EF handdiluted 1:50 in 23YT media and grown to an OD600 of 0.8. Proteinregion represents one of the elements in the scenario expression was induced by the addition of isopropyl b-D-thiogalac-

described above. topyranoside (IPTG; Sigma) to 1.0 mM, and cells were grown for anFigure 7 summarizes our current view of the Ca21- additional 4 hr at 378C. Expressed fusion proteins were purified in

a manner similar to that previously described (Frangioni and Neel,dependent inactivation mechanism. In the absence of1993). Briefly, bacterial pellets were resuspended in 50 mM TrisCa21, the IQ-like domain is not bound by CaM, and(pH 8.0), 150 mM NaCl, 13 Complete protease inhibitor cocktailthis configuration is somehow permissive for channel(Boehringer Mannheim) and 100 mg/ml lysozyme and incubated on

opening (open state). Elevated intracellular Ca21 causes ice for 15 min. Dithiothreitol (DTT; Sigma) was added to a finaltethered Ca21CaM to bind to the IQ-like domain, which concentration of 5 mM, and the suspension was incubated on iceleads to channel inactivation (inactivated state). Recov- for 10 min. The anionic detergent N-lauroylsarcosine (Sigma) was

added to a final concentration of 1.0%, and the suspension wasery from Ca21-dependent inactivation can be stronglyincubated on ice for 30 min. The mixture was mildly sonicated onaccelerated by hyperpolarization (Imredy and Yue, 1994),ice at 1% output power with a Sonic Dismembrator Model 50 (Fishersuggesting that the IQ-like domain in the resting channelScientific) for 1020 one second pulses. Triton X-100 (Sigma) wasstate has a lower affinity for Ca21CaM. Discoveringadded to 1.5%, and after 30 min on ice, the lysate was clarified by

how CaM binding is transduced into channel inactivation centrifugation at 12,000 3 g for 20 min at 48C. Glutathione-sepha-remains a standing challenge, one that promises to bring rose beads were added to the supernatant, and the resulting mixtureexciting results in the near future. was rotated overnight at 48C. Beads were washed in 100 bed vol-

umes of 50 mM Tris (pH 8.0) and 150 mM NaCl and resuspendedin two bed volumes of the same buffer. The immobilized fusionExperimental Proceduresproteins (10 mg) were incubated with 1 mg CaM (Calbiochem) for 2hr at 48C in 1 ml of 50 mM Tris (pH 8.0), 150 mM NaCl in the presenceElectrophysiologyof 1 mM CaCl2 or 5 mM EGTA and washed in the same bufferWhole-cell patch-clamp recordings were acquired as described pre-containing 0.05% Tween-20 (Sigma). Bound CaM was eluted withviously (de Leon et al., 1995). Briefly, complementary DNAs encoding13 SDS sample buffer, resolved by 12.5% SDSPAGE, electroblot-wild-type a1C (Wei et al., 1991) or chimeric EC-1D3 (de Leon et al.,ted to nitrocellulose, and visualized by enhanced chemilumines-1995) calcium channel subunits were cotransfected with b2a (Perez-cence (ECL; Amersham) with anti-CaM antibody (Upstate Biotech-Reyes et al., 1992) and a2bd (Tomlinson et al., 1993) in HEK 293 cellsnology).by calcium phosphate precipitation, and whole-cell currents were

recorded at room temperature 23 days after transfection. Bathsolution contained (in mM): NMG-aspartate, 130; MgCl2, 1; glucose,

Gel Shift Assays10; 4-aminopyridine, 10; HEPES (pH 7.4), 10; and CaCl2 or BaCl2,Three hundred picomoles of CaM were incubated with different10. Internal solution contained (in mM): NMG-MeSO3, 140; EGTA, 5;molar amounts of IQ peptide in a buffer containing 10 mM Na-MgCl2, 1; MgATP, 4; and HEPES (pH 7.3), 10. Currents were sampledHEPES (pH 7.2) and 2 mM CaCl2 for 30 min at room temperature.at 10 kHz and filtered at 2 kHz. Series resistance was typically ,6The bound complexes were then resolved by nondenaturing PAGEMV and was compensated by 70%. Repetition intervals were 30 s,on a 12% polyacrylamide gel in the presence of 2 mM CaCl2 or 5holding potential (VH) was 280 mV, and leaks and capacitive tran-mM EGTA, which was run in an ice bath to minimize peptide un-sients were subtracted by a P/8 protocol. The fraction of peak cur-binding. Bound and unbound CaM bands were visualized with Coo-rent remaining at the end of a 300 ms depolarization (r300) as amassie blue staining. IQ peptides used in gel shift assays werefunction of voltage was used to quantitate the level of inactivation.synthesized by the Biopolymer Laboratory in the Howard HughesThe strength of Ca21-dependent inactivation was sometimes quanti-Medical Institute at the Johns Hopkins University School of Med-fied by the parameter f, defined as the difference between r300 valuesicine.in Ba21 versus Ca21, taken at 210 mV. All data are presented as

mean 6 SEM.

AcknowledgmentsSubcloning and Mutagenesis of CaM cDNAcDNAs encoding wild-type CaM, CaM1,2, and CaM2,3,4 rat CaM were We thank Mike Ehlers, Anne Houdusse, Rick Huganir, and King Waigenerously provided in the oocyte expression plasmid pBF by John

Yau for valuable advice and discussion. John Adelman generouslyP. Adelman (Vollum Institute, Portland, OR). The entire coding re-

provided valuable gifts of wild-type CaM, CaM2,3,4, and CaM1,2. Kinggions of these constructs were transferred to the mammalian ex-Wai Yau furnished the OCNC-N fusion protein. Ed Perez-Reyes pro-

pression plasmid pCDNA3 (Invitrogen, Carlsbad, CA) as follows:vided the calcium channel clones for a1C and b2a, and Kevin CampbellSalIBglII fragments containing the entire coding sequence of theprovided the a2bd clone. Devi Rathod provided technical support.mutant and wild-type CaM proteins were subcloned into the XhoIThe work was supported by the National Institutes of Health.BamHI-digested shuttle vector pBluescript (Stratagene, La Jolla,

CA). KpnIXbaI fragments from the resulting constructs were subse-quently subcloned into KpnIXba-digested pCDNA3. cDNA encod- Received February 24, 1999; revised March 1, 1999.ing CaM1,2,3,4 was constructed by subcloning a 370 base pair BstXIXbaI fragment from CaM2,3,4, which provides the last three mutant

ReferencesEF hand motifs, into BstXIXbaI-digested CaM1,2 vector. cDNA en-coding CaM3,4 was constructed by subcloning a 170 base pair EcoRI

Bahler, M., Kroschewski, R., Stoffler, H.E., and Behrmann, T. (1994).fragment from wild-type CaM, which provides the first two wild-Rat myr 4 defines a novel subclass of myosin I: identification, distri-type EF hands, into EcoRI-digested and phosphatased CaM2,3,4 vec-bution, localization, and mapping of calmodulin-binding sites withtor. The integrity of all of the CaM-containing plasmids was con-

firmed by qualitative restriction digests and sequence analysis. differential calcium sensitivity. J. Cell Biol. 126, 375389.

Calmodulin Mediates Calcium Channel Inactivation557

Bayley, P.M., Findlay, W.A., and Martin, S.R. (1996). Target recogni- Ikura, M., Clore, G.M., Gronenborn, A.M., Zhu, G., Klee, C.B., andtion by calmodulin: dissecting the kinetics and affinity of interaction Bax, A. (1992). Solution structure of a calmodulintarget peptideusing short peptide sequences. Protein Sci. 5, 12151228. complex by multidimensional NMR. Science 256, 632638.

Bernatchez, G., Talwar, D., and Parent, L. (1998). Mutations in the Imredy, J.P., and Yue, D.T. (1992). Submicroscopic Ca21 diffusionEF-hand motif impair the inactivation of barium currents of the car- mediates inhibitory coupling between individual Ca21 channels.diac alpha 1C channel. Biophys. J. 75, 17271739. Neuron 9, 197207.

Brehm, P., and Eckert, R. (1978). Calcium entry leads to inactivation Imredy, J.P., and Yue, D.T. (1994). Mechanism of Ca21-sensitiveof calcium channel in Paramecium. Science 202, 12031206. inactivation of L-type Ca21 channels. Neuron 12, 13011318.Chen, S.R., and MacLennan, D.H. (1994). Identification of calmodu- James, P., Vorherr, T., and Carafoli, E. (1995). Calmodulin-bindinglin-, Ca21-, and ruthenium redbinding domains in the Ca21 release domains: just two faced or multi-faceted? Trends Biochem. Sci. 20,channel (ryanodine receptor) of rabbit skeletal muscle sarcoplasmic 3842.reticulum. J. Biol. Chem. 269, 2269822704. Johnson, B., and Byerly, L. (1993). A cytoskeletal mechanism forChen, T.Y., and Yau, K.W. (1994). Direct modulation by Ca21 Ca21 channel metabolic dependence and inactivation by intracellu-calmodulin of cyclic nucleotideactivated channel of rat olfactory lar Ca21. Neuron 10, 797804.receptor neurons. Nature 368, 545548. Krupp, J.J., Vissel, B., Thomas, C.G., Heinemann, S.F., and West-Dasgupta, M., Honeycutt, T., and Blumenthal, D.K. (1989). The brook, G.L. (1999). Interactions of calmodulin and alpha-actinin withgamma-subunit of skeletal muscle phosphorylase kinase contains the NR1 subunit modulate Ca21-dependent inactivation of NMDAtwo noncontiguous domains that act in concert to bind calmodulin. receptors. J. Neurosci. 19, 11651178.J. Biol. Chem. 264, 1715617163. Linse, S., and Forsen, S. (1995). Determinants that govern high-Deisseroth, K., Bito, H., and Tsien, R.W. (1996). Signaling from syn- affinity calcium binding. Adv. Second Messenger Phosphoproteinapse to nucleus: postsynaptic CREB phosphorylation during multi- Res. 30, 89151.ple forms of hippocampal synaptic plasticity. Neuron 16, 89101. Liu, M., Chen, T.Y., Ahamed, B., Li, J., and Yau, K.W. (1994). CalciumDeisseroth, K., Heist, E.K., and Tsien, R.W. (1998). Translocation calmodulin modulation of the olfactory cyclic nucleotidegated cat-of calmodulin to the nucleus supports CREB phosphorylation in ion channel. Science 266, 13481354.hippocampal neurons. Nature 392, 198202.

Liu, Y., Holmgren, M., Jurman, M.E., and Yellen, G. (1997). Gatedde Leon, M., Wang, Y., Jones, L., Perez-Reyes, E., Wei, X., Soong, access to the pore of a voltage-dependent K1 channel. Neuron 19,T.W., Snutch, T.P., and Yue, D.T. (1995). Essential Ca21-binding motif 175184.for Ca21-sensitive inactivation of L-type Ca21 channels. Science 270,

Means, A.R. (1988). Molecular mechanisms of action of calmodulin.15021506.Rec. Prog. Hormone Res. 44, 223262.

Ehlers, M.D., Zhang, S., Bernhardt, J.P., and Huganir, R.L. (1996).Mikami, A., Imoto, K., Tanabe, T., Niidome, T., Mori, Y., Takeshima,Inactivation of NMDA receptors by direct interaction of calmodulinH., Narumiya, S., and Numa, S. (1989). Primary structure and func-with the NR1 subunit. Cell 84, 745755.tional expression of the cardiac dihydropyridine-sensitive calcium

Enyedi, A., Vorherr, T., James, P., McCormick, D.J., Filoteo, A.G., channel. Nature 340, 230233.Carafoli, E., and Penniston, J.T. (1989). The calmodulin binding do-

Montell, C., and Rubin, G.M. (1989). Molecular characterization ofmain of the plasma membrane Ca21 pump interacts both with cal-the Drosophila trp locus: a putative integral membrane protein re-modulin and with another part of the pump. J. Biol. Chem. 264,quired for phototransduction. Neuron 2, 13131323.1231312321.Murphy, T.H., Worley, P.F., and Baraban, J.M. (1991). L-type voltage-Fabiato, A., and Fabiato, F. (1979). Calcium and cardiac excitationsensitive calcium channels mediate synaptic activation of immedi-contraction coupling. Annu. Rev. Physiol. 41, 473484.ate early genes. Neuron 7, 625635.

Frangioni, J.V., and Neel, B.G. (1993). Solubilization and purificationNeely, A., Olcese, R., Wei, X., Birnbaumer, L., and Stefani, E. (1994).of enzymatically active glutathione S-transferase (pGEX) fusion pro-Ca21-dependent inactivation of a cloned cardiac Ca21 channel alphateins. Anal. Biochem. 210, 179187.1 subunit (alpha 1C) expressed in Xenopus oocytes. Biophys. J. 66,

Fujita, Y., Mynlieff, M., Dirksen, R.T., Kim, M.S., Niidome, T., Nakai, 18951903.J., Friedrich, T., Iwabe, N., Miyata, T., Furuichi, T., et al. (1993).

Patil, P.G., Brody, D.L., and Yue, D.T. (1998). Preferential closed-Primary structure and functional expression of the omega-cono-state inactivation of neuronal calcium channels. Neuron 20, 1027toxin-sensitive N-type calcium channel from rabbit brain. Neuron1038.10, 585598.Perez-Reyes, E., Castellano, A., Kim, H.S., Bertrand, P., Baggstrom,Gao, T., Yatani, A., Dell'Acqua, M.L., Sako, H., Green, S.A., Dascal,E., Lacerda, A.E., Wei, X.Y., and Birnbaumer, L. (1992). Cloning andN., Scott, J.D., and Hosey, M.M. (1997). cAMP-dependent regulationexpression of a cardiac/brain beta subunit of the L-type calciumof cardiac L-type Ca21 channels requires membrane targeting ofchannel. J. Biol. Chem. 267, 17921797.PKA and phosphorylation of channel subunits. Neuron 19, 185196.Phillips, A.M., Bull, A., and Kelly, L.E. (1992). Identification of a Dro-Grunwald, M.E., Yu, W.P., Yu, H.H., and Yau, K.W. (1998). Identifica-sophila gene encoding a calmodulin-binding protein with homologytion of a domain on the beta-subunit of the rod cGMPgated cationto the trp phototransduction gene. Neuron 8, 631642.channel that mediates inhibition by calciumcalmodulin. J. Biol.Putkey, J.A., Sweeney, H.L., and Campbell, S.T. (1989). Site-directedChem. 273, 91489157.mutation of the trigger calcium-binding sites in cardiac troponin C.Haack, J.A., and Rosenberg, R.L. (1994). Calcium-dependent inacti-J. Biol. Chem. 264, 1237012378.vation of L-type calcium channels in planar lipid bilayers. Biophys.Rhoads, A.R., and Friedberg, F. (1997). Sequence motifs for calmod-J. 66, 10511060.ulin recognition. FASEB J. 11, 331340.Holmgren, M., Shin, K.S., and Yellen, G. (1998). The activation gateRios, E., and Brum, G. (1987). Involvement of dihydropyridine recep-of a voltage-gated K1 channel can be trapped in the open state bytors in excitationcontraction coupling in skeletal muscle. Naturean intersubunit metal bridge. Neuron 21, 617621.325, 717720.Houdusse, A., and Cohen, C. (1995). Target sequence recognitionRoberts, W., Jacobs, R., and Hudspeth, A. (1990). Colocalization ofby the calmodulin superfamily: implications from light chain bindingion channels involved in frequency selectivity and synaptic transmis-to the regulatory domain of scallop myosin. Proc. Natl. Acad. Sci.sion at presynaptic active zones of hair cells. J. Neurosci. 10, 3664USA 92, 1064410647.3684.Houdusse, A., and Cohen, C. (1996). Structure of the regulatorySaimi, Y., and Kung, C. (1994). Ion channel regulation by calmodulindomain of scallop myosin at 2 A resolution: implications for regula-binding. FEBS Lett. 350, 155158.tion. Structure 4, 2132.

Sheng, M., McFadden, G., and Greenberg, M.E. (1990). MembraneHoudusse, A., Silver, M., and Cohen, C. (1996). A model of Ca21-depolarization and calcium induce c-fos transcription via phosphor-free calmodulin binding to unconventional myosins reveals how cal-

modulin acts as a regulatory switch. Structure 4, 14751490. ylation of transcription factor CREB. Neuron 4, 571582.

Neuron558

Shirokov, R.E., Levis, R., Shirokova, N., and Rios, E. (1993). Ca21-dependent inactivation of cardiac L-type Ca21 channels does notaffect their voltage sensor. J. Gen. Physiol. 102, 10051030.

Smith, J.S., Rousseau, E., and Meissner, G. (1989). Calmodulin mod-ulation of single sarcoplasmic reticulum Ca21-release channels fromcardiac and skeletal muscle. Circ. Res. 64, 352359.

Soong, T.W., Stea, A., Hodson, C.D., Dubel, S.J., Vincent, S.R., andSnutch, T.P. (1993). Structure and functional expression of a mem-ber of the low voltageactivated calcium channel family. Science260, 11331136.

Starr, T.V., Prystay, W., and Snutch, T.P. (1991). Primary structureof a calcium channel that is highly expressed in the rat cerebellum.Proc. Natl. Acad. Sci. USA 88, 56215625.

Tomlinson, W.J., Stea, A., Bourinet, E., Charnet, P., Nargeot, J., andSnutch, T.P. (1993). Functional properties of a neuronal class CL-type calcium channel. Neuropharmacology 32, 11171126.

Varnum, M.D., and Zagotta, W.N. (1997). Interdomain interactionsunderlying activation of cyclic nucleotide-gated channels. Science278, 110113.

Victor, R.G., Rusnak, F., Sikkink, R., Marban, E., and O'Rourke, B.(1997). Mechanism of Ca21-dependent inactivation of L-type Ca21

channels in GH3 cells: direct evidence against dephosphorylationby calcineurin. J. Membr. Biol. 156, 5361.

Wei, X.Y., Perez-Reyes, E., Lacerda, A.E., Schuster, G., Brown, A.M.,and Birnbaumer, L. (1991). Heterologous regulation of the cardiacCa21 channel alpha 1 subunit by skeletal muscle beta and gammasubunits. Implications for the structure of cardiac L-type Ca21 chan-nels. J. Biol. Chem. 266, 2194321947.

Xia, X.-M., Falker, B., Rivard, A., Wayman, G., Johnson-Pais, T.,Keen, J.E., Ishii, T., Hirschberg, B., Bond, C.T., Lutsenko, S., et al.(1998). Mechanism of calcium gating in small-conductance calcium-activated potassium channels. Nature 395, 503507.

Yue, D.T. (1997). Quenching the spark in the heart. Science 276,755756.

Zhang, S., Ehlers, M.D., Bernhardt, J.P., Su, C.T., and Huganir, R.L.(1998). Calmodulin mediates calcium-dependent inactivation ofN-methyl-D-aspartate receptors. Neuron 21, 443453.

Zhou, J., Olcese, R., Qin, N., Noceti, F., Birnbaumer, L., and Stefani,E. (1997). Feedback inhibition of Ca21 channels by Ca21 dependson a short sequence of the C terminus that does not include theCa21-binding function of a motif with similarity to Ca21-binding do-mains. Proc. Natl. Acad. Sci. USA 94, 23012305.

Zuhlke, R.D., and Reuter, H. (1998). Ca21-sensitive inactivation ofL-type Ca21 channels depends on multiple cytoplasmic amino acidsequences of the a1C subunit. Proc. Natl. Acad. Sci. USA 95, 32873294.