Continuously tunable Ca(2+) regulation of RNA-edited CaV1.3 channels.

CaV1.3 ion channels are dominant Ca(2+) portals into pacemaking neurons, residing at the epicenter of brain rhythmicity and neurodegeneration. Negative Ca(2+) feedback regulation of CaV1.3 channels (CDI) is therefore critical for Ca(2+) homeostasis. Intriguingly, nearly half the CaV1.3 transcripts in the brain are RNA edited to reduce CDI and influence oscillatory activity. It is then mechanistically remarkable that this editing occurs precisely within an IQ domain, whose interaction with Ca(2+)-bound calmodulin (Ca(2+)/CaM) is believed to induce CDI. Here, we sought the mechanism underlying the altered CDI of edited channels. Unexpectedly, editing failed to attenuate Ca(2+)/CaM binding. Instead, editing weakened the prebinding of Ca(2+)-free CaM (apoCaM) to channels, which proves essential for CDI. Thus, editing might render CDI continuously tunable by fluctuations in ambient CaM, a prominent effect we substantiate in substantia nigral neurons. This adjustability of Ca(2+) regulation by CaM now looms as a key element of CNS Ca(2+) homeostasis.


INTRODUCTION
Voltage-activated Ca V 1.3 channels constitute prominent Ca 2+ entry portals into pacemaking neurons (Bean, 2007), owing to the more negative voltages required to open these ion channels (Xu and Lipscombe, 2001) ( Figure 1A). Accordingly, these channels influence neurobiological functions ranging from circadian rhythms drawn from repetitive spiking in suprachiasmatic nucleus, to movement control modulated by pacemaking in substantia nigra (Chan et al., 2007;Obeso et al., 2008). Moreover, Ca V 1.3 channels often contribute the majority of Ca 2+ entry in these settings, such as in substantia nigral neurons (Bean, 2007;Cardozo and Bean, 1995;Chan et al., 2007;Guzman et al.;Puopolo et al., 2007) whose loss is intimately connected with Ca 2+ dysfunction in the setting of Parkinson's disease (Bezprovanny, 2009;Surmeier and Sulzer, 2013).
Fitting with these Ca 2+ entry actions, the opening of Ca V 1.3 channels is subject to extensive negative Ca 2+ feedback regulation (CDI), critical for proper Ca 2+ handling in these venues. CDI is triggered by the Ca 2+ -sensing molecule calmodulin (CaM), which acts as a virtual subunit of channels (Erickson et al., 2001). The Ca 2+ -free form (apoCaM) prebinds to the carboxy tail of channels (Erickson et al., 2001;Erickson et al., 2003a;Pitt et al., 2001), and subsequent Ca 2+ binding to this CaM drives conformational changes that trigger CDI (DeMaria et al., 2001;Dunlap, 2007;Erickson et al., 2003a;Kim et al., 2004;Peterson et al., 1999;Zuhlke et al., 1999;Zuhlke et al., 2000). It has been widely proposed that a carboxyterminal IQ domain ( Figure 1A) serves as the Ca 2+ /CaM effector site that induces CDI, and also as a potential tethering site for apoCaM ( Figure 1B).
It is intriguing that nearly half the Ca V 1.3 transcripts in brain are RNA edited precisely and only within the IQ element ( Figure 1A, blue circle), yielding variant channels whose reduced CDI tunes pacemaking in the brain (Huang et al., 2012). The mechanism underlying the editing effects on CDI is presently unknown, but we are now poised to achieve an atomic-level mechanistic understanding of these effects, in terms of single-residue downregulation of Ca 2+ /CaM interaction with the IQ effector site. Indeed, several crystal structures of Ca 2+ /CaM complexed with IQ-domain peptides of various Ca 2+ channels have recently been resolved (Fallon et al., 2005;Mori et al., 2008;Van Petegem et al., 2005). Moreover, based on a structure for a Ca V 1.2 IQ domain with only a single glutamate-toaspartate difference, a well-constrained homology model of the Ca V 1.3 complex is deduced in Figure 1C. Accordingly, we here sought to rigorously demonstrate that RNA-editing effects could indeed be attributed to precise reductions of Ca 2+ /CaM binding to the IQ domain. Contrary to expectations, these natural variations within the IQ segment largely fail to attenuate Ca 2+ /CaM binding, as do alanine mutations throughout. In a surprising turn, editing instead weakens prebinding of channels to Ca 2+ -free CaM (apoCaM), which we substantiate as being essential for CDI (Ben Johny et al., 2013;Liu et al., 2010). This unanticipated outcome suggests that the actual effect of RNA editing is to reset downward the affinity of channels for apoCaM, so that fluctuations in ambient apoCaM can bias the fraction of channels lacking or endowed with resident CaM. In this manner, the strength of CDI of Ca V 1.3 channels could become a continuously tunable function of CaM levels, a prominent effect we establish directly in substantia nigral neurons. This newfound adjustability of Ca 2+ feedback regulation by CaM now emerges as a key element of Ca 2+ homeostasis across the brain. Figure 1D (far left subpanel) displays the electrophysiological signature of CDI for Ca V 1.3 channels bearing a prototypic IQ domain, as directly coded by genomic DNA without editing. The central portion of the IQ element is comprised of the contiguous residues isoleucine-glutamine-aspartate-tyrosine (IQDY), and such channels correspond to ~60% of the transcripts across the brain (Huang et al., 2012). These channels are here expressed as a homogeneous population in HEK293 cells for optimal biophysical resolution. The resulting Ca 2+ current decays rapidly (Ca, red trace), compared with the minimal decline of Ba 2+ current (Ba, black trace). Because Ba 2+ binds negligibly to CaM (Chao et al., 1984), the fractional decline of Ca 2+ versus Ba 2+ current after 300-ms depolarization quantifies the steady-state extent of CDI as mediated by CaM (right, CDI parameter).

Functional effects of RNA editing in the IQ domain of Ca V 1.3 channels
With this baseline in mind, we can readily appreciate the effect of RNA editing to variably attenuate CDI (subpanels to right). The composition of the central IQ domain for each variant is displayed atop the corresponding set of exemplar currents, along with the prevalence of affiliated transcripts across the brain (Huang et al., 2012). The blunting of CDI is particularly intense for MQDY and MQDC variants, and MRDY and IRDY exhibit intermediate extents of attenuated CDI. Accordingly, adjusting the distribution of Ca V 1.3 channels among these variants markedly influences the strength of CDI in the brain. The extent of CDI modulation reported here differs somewhat from that previously reported (Huang et al., 2012), owing to the use of more stringent intracellular Ca 2+ buffering solutions used here (10 mM BAPTA versus 5 mM EGTA). Given the presumed atomic-level understanding of the Ca 2+ /CaM effector configuration ( Figure 1B, C), we sought to achieve a high resolution understanding of these editing effects ( Figure 1D), by corresponding CDI strength (CDI) with graded and well-understood decrements in the affinity of Ca 2+ /CaM binding affinity to variant IQ modules. To determine binding properties, we used an extensively developed FRET two-hybrid approach (Chen et al., 2006;Dick et al., 2008;Erickson et al., 2001;Erickson et al., 2003a;Erickson et al., 2003b;Liu et al., 2010), where CFP-tagged CaM and YFP-tagged IQ domains were expressed in live HEK293 cells (Erickson et al., 2003a;Yang et al., 2006) (Figure 1E, left), allowing FRET efficiency (E A ) within an individual cell to be plotted as a function of the free concentration of CFP-CaM in that cell. Given variable expression and concentrations in different cells, the ensemble of points yields the binding curve shown in Figure 1E (right). The specific curve here pertains to the prototypic IQ domain (IQDY), with a relative dissociation constant of K d,EFF of 1700 microscope-specific D free units. This corresponds to an absolute dissociation constant of K d ~ 55 nM (see Figure S1). That said, we expected that the effects of RNA editing could be rigorously understood in terms of a Langmuir function, shown as a hypothetical outcome in Figure 1F. Specifically, CDI would be plotted as a function of the association constant (K a,EFF ) for Ca 2+ /CaM binding to the IQ element of a particular variant. The collection of such points, encompassing the IQDY module (green symbol) and other variants (black symbols), should then decorate the black Langmuir curve defined by the relation CDI∝ K a,EFF /(K a,EFF + Λ) ( Figure S2). For the prototypic IQDY module, the association constant K a,EFF (= 1/K d, EFF = 1/1700) would be 5.88 × 10 −4 reciprocal D free units, equivalent to an absolute association constant K a (= 1/K d = 1/55 nM) approximating 0.018 nM −1 .

Incongruencies of Ca V 1.3 IQ domain as Ca 2+ /CaM effector site for CDI
Thus positioned, we proceeded to correspond CDI with related Ca 2+ /CaM association constants, not only for the RNA-editing variants, but also for point alanine mutations throughout the IQ domain. Naturally occurring alanines were substituted with threonine. Figure 2A summarizes the population effects on CDI in regard to both RNA editing (blue bars, far left) and alanine substitutions (gray and rose bars, right). The green symbol and dashed-horizontal line furnish the reference CDI for prototypic IQDY channels, and the precise amino-acid sequence of the IQ domain is aligned above for orientation. Substitutions at several positions produced strong suppression of CDI (rose bars), with manipulation at the position-zero isoleucine (I[0]A) yielding the largest sequela. Exemplar currents shown below ( Figure 2B) illustrate more directly the effects of mutations producing the strongest CDI reductions.
We next obtained binding relations for Ca 2+ /CaM interacting with variant IQ segments relating to perturbations generating the greatest functional consequences (Figure 2A, rose bars), as well to those chosen at random (Figure 2A, dashed-gray bars). Figure 2C displays binding relations for exemplar whole-cell currents exhibiting the most pronounced CDI reductions. To facilitate comparison, the reference relation for the prototypic IQDY domain is shown in green (reproduced from Figure 1E). Notably, the decrease in affinity, if any, was modest at best. In the case of I[0]A for which CDI was nearly eliminated (Figure 2A, B), there is no discernible alteration in binding whatsoever ( Figure 2C). Also unexpected was the outcome of binding-curve analysis for the RNA-editing variants, which revealed no decrement in Ca 2+ /CaM interaction ( Figure 2D). Accordingly, plots of the steady-state extent of CDI (CDI) versus the association constant of Ca 2+ /CaM interaction with correlating IQ modules (K a,EFF ) altogether deviated from a Langmuir relation ( Figure 2E; Figure S3). The symbol corresponding to the prototypic IQ is shown in green; symbols relating to RNA editing are plotted in blue; and those for alanine substitutions are graphed in red (large CDI reductions) or black (loci chosen at random). Notably, manipulations that left CDI unchanged featured altered binding affinity (black symbols tracking horizontal green line), and variations yielding marked CDI attenuation demonstrated unchanged binding (symbols hugging vertical green line). In all, it appears unlikely that RNA editing reduces CDI by diminishing Ca 2+ /CaM binding to an IQ domain effector site; indeed, it seems that the IQ module alone does not represent a CDI effector site at all.

RNA editing perturbs apoCaM interaction with an IQ domain preassociation locus
Despite the incongruence of the IQ domain as Ca 2+ /CaM effector, editing and alanine substitutions in this element nonetheless attenuated CDI (Figures 1, 2). To understand how this effect arises, we considered another potential role of the IQ domain, to furnish an important locus for channel preassociation with apoCaM ( Figure 3A, configuration A). Since channels devoid of a resident apoCaM fail to exhibit CDI (Ben Johny et al., 2013;Liu et al., 2010) (configuration E), it is plausible that IQ perturbations reduce CDI simply by diminishing channel/apoCaM affinity ( Figure 3A, K a ) and promoting apoCaM-less channels. If so, then overexpressing recombinant CaM WT should, by mass action, drive channels from configuration E to A ( Figure 3A), where channels in the configuration A may undergo CDI by Ca 2+ -driven transition to configuration I. Thus, overexpressing CaM WT would be predicted to restore CDI in channels with altered IQ domains.
Remarkably, this resurgence of CDI is indeed observed ( Figure 3B, C). For prototypic IQDY channels, CaM WT overexpression hardly perturbs CDI ( Figure 3B, green horizontal line; Figure 3C, leftmost subpanel) compared to control ( Figure 1D, leftmost subpanel). This outcome would be expected if the high affinity of these channels for apoCaM always ensures the absence of channels in configuration E ( Figure 3A), even with only endogenous CaM present. By contrast, for the RNA editing variants, elevating CaM WT restored CDI essentially to prototypic levels ( Figure 3B, blue bars). Exemplar traces visually corroborate this restoration for editing variants with the weakest CDI before augmenting CaM WT ( Figure 3C, MQDY and MQDC). Elevating CaM WT also produced near-complete restoration of CDI for the majority of IQ mutations with initially deficient Ca 2+ regulation ( Figure 3B, rose bars). Alternatively, CDI remained at prototypic levels for mutations lacking initial CDI effects ( Figure 3B, gray bars). These effects on CDI are visually substantiated by the exemplar traces below ( Figure 3C, rightmost two subpanels), particularly for the I[0]A substitution that nearly eliminated CDI before CaM WT supplementation. The modest residual deficit in CDI in I[0]A reflects perturbation of a different effector configuration (Ben Johny et al., 2013), as cartooned in configuration I of Figure 3A.
Still, the actions of overexpressing CaM WT on CDI could be explained by alternative mechanisms, unless this functional restoration could be explicitly linked to decreased apoCaM interaction with channels. Thus, we performed FRET 2-hybrid binding assays for apoCaM paired with the entire CI region of channels ( Figure 3D, leftmost subpanel). This arrangement supported robust binding for the prototypic IQDY pairing, as shown by the green data in the MQDY subpanel. Reassuringly, these assays indicated sharply reduced binding affinity for CI modules relating to editing variants and alanine substitutions with strong CDI effects ( Figure 3D, black data).
The most rigorous test arises from the following realization regarding the fraction of channels bound to CaM with only endogenous CaM present-this fraction F b is given by the ratio of the CDI measured before (Figure 2A, CDI) and after strong overexpression of CaM WT ( Figure 3B, CDI CaMhi ). Thus, if the mechanism in Figure 3A holds true, then the CDI/CDI CaMhi (=F b ) should relate as a Langmuir function to the association constant for apoCaM binding to the relevant CI module ( Figure S4). Indeed, plotting data in this manner strikingly resolves just such a relationship in Figure 3E. The green symbol corresponds to the prototypic IQDY configuration, blue symbols relate to RNA-editing variants, red symbols derive from point mutations yielding strongly weakened CDI with only endogenous CaM present, and black symbols report on mutations without appreciable CDI effects ( Figure 3B, dashed-gray bars). Additional supporting data are summarized in Figure S5. Still more reassuring are the results of homology modeling the C-lobe of apoCaM in complex with the Ca V 1.3 IQ domain ( Figure 3F), based on an analogous NMR structure for Na V channels (Chagot and Chazin, 2011;Feldkamp et al., 2011). Notably, RNA editing would perturb deeply articulated anchor points within the homology model. In all, these outcomes argue well that RNA editing perturbs CDI primarily by diminishing channel affinity for apoCaM ( Figure 3A).

Continuously tunable Ca 2+ regulation in substantia nigral neurons
This new understanding of RNA editing ( Figure 3A) opens the door to considerably expanded tunability of the Ca 2+ regulation of Ca V 1.3 channels, beyond initial expectations. The prevailing original concept in Fig 1b-that editing modulates Ca 2+ /CaM interaction with an IQ effector site-would portray RNA editing as adjusting CDI to a limited set of static strengths, much like the discretized settings on a rotary switch. By contrast, the alternative mechanism involving apoCaM ( Figure 3A) predicts a more flexible configuration as diagrammed in Figure 4A. Viewed in this way, the CDI strength of each variant may be smoothly adjusted by fluctuating levels of naturally occurring ambient CaM (ΔapoCaM), which could variably redistribute channels between configurations that lack (configuration E) or manifest CDI (configuration A). Accordingly, CDI may be continuously adjusted in the manner of a rheostat, with each variant requiring a different level of CaM to achieve half-maximal CDI.
Among the settings where this connection would be most consequential are dopamine neurons of the substantia nigra pars compacta (SNc), whose loss is closely linked to Ca 2+ dysfunction in Parkinson's disease (Chan et al., 2007;Guzman et al., 2009). RNA editing of Ca V 1.3 channels is prevalent in these cells (Huang et al., 2012), so we tested for CaMmediated upregulation of CDI in isolated murine SNc dopamine neurons, as exemplified in Figure 3B by GFP expression under the control of a tyrosine hydroxylase promoter. Wholecell patch clamp recordings revealed, for the first time in these cells, the existence of modest CDI under control conditions, as illustrated by current waveforms averaged over multiple neurons ( Figure 4C, left, with only endogenous CaM present (CaM endo )). Using a FRETbased genetically encoded sensor of CaM (BSCaM IQ ), we estimated the free concentration of CaM endo to be 3.9 ± 1.0 μM (n = 7) based on measurements done in intrinsically nonfluorescent hippocampal neurons (Brody and Yue, 2000;Liu et al., 2010). Dihydropyridine antagonists verified that up to two thirds of the current was carried by Ca V 1.3 channels in SNc neurons ( Figure S6). Moreover, the strength of CDI here was about half that of prototypic IQDY Ca V 1.3 channels ( Figure 4D, CaM endo ), all consistent with significant CDI modulation by RNA editing. Importantly, upon internal perfusion with elevated CaM WT protein (final concentration 100 μM), CDI was robustly enhanced (Figure 4C, right; Figure  4D, CaM WT ) to a level approaching that of prototypic IQDY constructs ( Figure 4D, green). This exciting outcome is directly consistent with the tunability of CDI by CaM in SNc neurons, opening a new realm of research regarding Ca 2+ homeostasis in this neurodegeneration-prone locus (Bean, 2007). Bazzazi et al. Page 6 Cell Rep. Author manuscript; available in PMC 2015 March 04.

Author Manuscript
Author Manuscript

Author Manuscript
Author Manuscript

DISCUSSION
We have demonstrated that RNA editing of Ca V 1.3 channels downwardly modulates their Ca 2+ regulation by an unexpected mechanism. Rather than attenuate Ca 2+ /CaM binding to an effector site comprised by the channel IQ element alone ( Figure 1B), editing variants reduce the affinity of channels for apoCaM ( Figure 3E). This effect promotes the occurrence of channels uncharged by CaM and incapable of CDI. This alternative mechanism predicts that CDI of edited channels could become a smoothly continuous function of ambient CaM levels ( Figure 4A), an outcome we corroborate in substantia nigral neurons (Figure 4, B-D).
These findings demonstrate that naturally occurring RNA editing of Ca V 1.3 channels acts to modulate CDI in ways that substantiate a recently emerging mechanism where apoCaM begins preassociated with the IQ and other channel elements ( Figure 4A, configuration A), but the Ca 2+ /CaM effector configuration (configuration I) involves substantial rearrangements and differs considerably from that originally proposed in Figures 1B, C. A surprising outcome here is that even single-residue changes may readily influence configurations outside the Ca 2+ /CaM effector complex. For example, the impression from prior work has been that channels so avidly prebind to apoCaM that they would always possess a resident CaM (Findeisen et al., 2011;Yang et al., 2006), and that mutation of several IQ residues might be required to appreciably affect apoCaM interaction (Erickson et al., 2003a;Liang et al., 2003). This view need no longer be the norm. More broadly, the mechanism in Figure 4A adds to the growing awareness of direct biological actions by apoCaM, despite the historical focus on the functions of Ca 2+ /CaM (Jurado et al., 1999).
Given the numerous indications of strong variations of CaM under differing physiological and disease-related contexts (Bezprovanny, 2009;Black et al., 2004;Chafouleas et al., 1982;Ikeda et al., 2009;Lesnick et al., 2007;Yacoubian et al., 2008;Zhang et al., 2005), the rheostat-like connection between CaM levels and Ca 2+ feedback gain on Ca 2+ influx now looms as a potentially important dimension of Ca 2+ homeostasis and dysfunction ( Figure  4A). In this regard, it is worth considering the CaM dependence of aggregate CDI exhibited by a channel population comprised of prototypic IQDY and editing variants. Figure 4E shows the projected CDI response relations for individual Ca V 1.3 species, based on our apoCaM binding data ( Figure 3D, black curves). Layered atop RNA editing, roughly one third of Ca V 1.3 channel transcripts in substantia nigra exhibit a long splice variant featuring a competitive ICDI inhibitor of apoCaM binding to channels (Bock et al., 2011;Liu et al., 2010). This splicing would multiplex the RNA editing effects to a parallel set of CDI response relations (Huang et al., 2012;Liu et al., 2010), shown by the set of blue relations in Figure 4E (Extended Discussion). Each curve represents CDI sensitivity to apoCaM variations over a limited concentration range, as constrained by the 1:1 stoichiometry of apoCaM binding to channels. However, when the aggregate response of a population of variants is considered, by averaging the individual curves with weighting factors specified by transcript and splice prevalence in substantia nigra (Bock et al., 2011;Huang et al., 2012), the far more extended relationship shown in Figure 4F results (black curve; Extended Discussion). Reassuringly, our experimental estimates of CDI responsiveness and estimated CaM (black symbols from Figure 4C-4D) fit well with this projected aggregate response relation (black curve). This agreement between data and prediction should be taken as approximate, because our estimate of free endogenous CaM concentration was obtained using a FRET-based CaM sensor expressed in readily transfectable hippocampal neurons, rather than SNc neurons per se. This outcome then reflects an elegant mechanism to render CDI tunable over a large dynamic range of CaM levels, unachievable by a single Ca V 1.3 variant. Interestingly, SNc neurons at baseline populate a 'setpoint' right in the middle of this response relationship ( Figure 4F, dashed line labeled CaM endo ), as if to optimally exploit the full dynamics of this system. CDI can thereby adapt smoothly and continuously over a maximal range of CaM levels. Altogether, this system of adjustable interdependence ( Figure 4G), rooted in the role of the IQ domain as an apoCaM preassociation locus, now merits exploration in a vast array of neurophysiological and pathophysiological contexts.

Molecular biology
Our baseline Ca V 1.3 construct (α 1DΔ1626 ; or Ca V 1.3 short in Figure 5B) was closely similar to a naturally occurring rat brain variant (α 1D , AF3070009 (Xu and Lipscombe, 2001), encoding 1643 amino acids) that terminates 18 residues after the IQ domain in the carboxy terminus. To facilitate mutagenesis, α 1DΔ1626 was engineered with a silent and unique Kpn I restriction site at a position encoding amino acids G 1538 T 1539 , ~50 residues upstream of the IQ domain. Additionally, α 1DΔ1626 contains a unique Xba I restriction site followed by a stop codon, both of which reside immediately after the IQ domain ending in V 1624 G 1625 . The engineered construct α 1DΔ1626 , as cloned within mammalian expression vector pCDNA6 (Invitrogen), thereby permitted rapid substitution of mutated segments of ~260 bp between Kpn I and Xba I restriction sites. Actual point mutations were introduced via QuikChange ® mutagenesis (Agilent), where the template was a short stretch of α 1DΔ1626 (~1500 bp) encompassing the Kpn I to Xba I segment, all as cloned within a ~3.5 kb pCR-Blunt II-TOPO vector (Invitrogen). After complete sequence verification between Kpn I and Xba I restriction sites, mutated segments were cloned into α 1DΔ1626 via these same sites, yielding full-length channel constructs with point mutations. For FRET 2-hybrid constructs, fluorophore-tagged CaM constructs were made as described (Erickson et al., 2003a). Other FRET constructs were made by replacing CaM in these constructs with appropriate PCR amplified and mutated IQ segments, via unique Not I and Xba I sites flanking CaM (Erickson et al., 2003a). Throughout, all segments subject to PCR or QuikChange ® (Agilent) were verified in their entirety by sequencing.

Transfection of HEK293 cells
For electrophysiology experiments, HEK293 cells were cultured in 10-cm plates, and channels were transiently transfected by a calcium phosphate protocol (Brody et al., 1997). We applied 8 μg of cDNA encoding the desired channel α 1 subunit, along with 8 μg of rat brain β 2a (M80545) and 8 μg of rat brain α 2 δ (NM012919.2) subunits. β 2a minimized voltage inactivation, enhancing resolution of CDI. Additional cDNAs were added as required in co-transfections. All of the above cDNA constructs were driven by a cytomegalovirus promoter. To enhance expression, cDNA for simian virus 40 T antigen (1-2 μg) was co-transfected. For fluorescence resonance energy transfer (FRET) 2-hybrid experiments, transfections and experiments were performed as described (Erickson et al., 2003a). Electrophysiology and FRET experiments were performed at room temperature 1-2 d following transfection.

Whole-cell recording
For both recombinant channels in HEK cells, and endogenous channels in SNc neurons (C57BL/6 mice), whole-cell recordings were obtained at room temperature using Axopatch

FRET optical imaging
FRET 2-hybrid experiments were carried out in HEK293 cells and analyzed, largely as described (Erickson et al., 2003a). During imaging, the bath solution was a Tyrode's buffer containing either 2 mM Ca 2+ alone for apoCaM interaction experiments, or 2 mM Ca 2+ and 10 μM ionomycin to elevate intracellular Ca 2+ for Ca 2+ /CaM interaction experiments. In parallel experiments with cells expressing the genetically encoded Ca 2+ indicator TN-XL (Mank et al., 2006;Tay et al., 2012), we confirmed that our ionomycin treatment achieved saturating concentrations of Ca 2+ with respect to CaM binding. Concentration-dependent spurious FRET was subtracted from raw data prior to binding-curve analysis Erickson et al., 2003b). Acceptor-centric measures of FRET were obtained with the 3 3 -FRET algorithm as described (Erickson et al., 2003a). Complementary donor-centric measures of FRET were obtained with the E-FRET method (Ben Johny et al., 2013;Chen et al., 2006). In-vitro binding assays were performed as described previously (Erickson et al., 2003a). Standard-deviation error bounds on K d,EFF estimates were determined by Jacobian error matrix analysis (Johnson, 1980). BSCaM IQ sensor measurements of free CaM concentration at rest were determined by previously established protocols (Liu et al., 2010) on mouse hippocampal neurons, cultured (Brody and Yue, 2000) and transiently transfected with plasmids encoding the sensor using polyehtylenimine PEI reagent (Polysciences, Warrington, PA). Sensor measurements in this setting were favored because of the readily transfectable nature of hippocampal versus SNc neurons.

Homology modeling of Ca 2+ /CaM complexed with Ca V 1.3 IQ domain
We used the python based homology modeling software MODELLER (Eswar et al., 2006) to build models of the Ca V 1.3 IQ domain (comprising positions -12 through +11 in Figure  2A)

Homology modeling for C-lobe of apoCaM complexed with IQ domain
We used the python based homology modeling software MODELLER (Eswar et al., 2006) to build models of Ca V 1.3 IQ domain bound to the C-lobe of apoCaM. Our starting templates were the NMR structures of Na V 1.5 and Na V 1.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.

Highlights
• Unexpected mechanism explains how RNA editing tunes Ca 2+ channel regulation by CaM (B)Exemplar current traces corresponding to indicated point-alanine substitutions. Format as in Figure 1D. (C) FRET 2-hybrid interaction curves for Ca 2+ /CaM versus IQ domain of point-alanine substitutions. As reference, green curve reproduces fit for prototypic IQDY species (from Figure 1E). Black data and fits correspond to whole-cell currents directly above in panel B. Each symbol bins data from ~3, 4, 7, and 8 cells (left to right). (D) FRET 2-hybrid interaction curves for Ca 2+ /CaM versus IQ domain of RNA editing variants. Format as in panel C.Each symbol bins data from ~7, 6, 9, and 6 cells (left to right). (E) CDI plotted as a function of K a,EFF deviates from Langmuir function ( Figure 1F), arguing against CDI reduction arising from diminished Ca 2+ /CaM with solitary IQ element acting as effector site. Horizontal bars, standard deviation of K a,EFF deduced from Jacobian error analysis (Johnson, 1980