The molecular basis of the inhibition of CaV1 calcium-dependent inactivation by the distal carboxy tail

Ca2+/calmodulin-dependent inactivation (CDI) of CaV channels is a critical regulatory process that tunes the kinetics of Ca2+ entry for different cell types and physiologic responses. CDI is mediated by calmodulin (CaM), which is bound to the IQ domain of the CaV carboxy tail. This modulatory process is tailored by alternative splicing such that select splice variants of CaV1.3 and CaV1.4 contain a long distal carboxy tail (DCT). The DCT harbors an inhibitor of CDI (ICDI) module that competitively displaces CaM from the IQ domain, thereby diminishing CDI. While this overall mechanism is now well described, the detailed interactions required for ICDI binding to the IQ domain are yet to be elucidated. Here, we perform alanine-scanning mutagenesis of the IQ and ICDI domains and evaluate the contribution of neighboring regions to CDI inhibition. Through FRET binding analysis, we identify functionally relevant residues within the CaV1.3 IQ domain and the CaV1.4 ICDI and nearby A region, which are required for high-affinity IQ/ICDI binding. Importantly, patch-clamp recordings demonstrate that disruption of this interaction commensurately diminishes ICDI function resulting in the re-emergence of CDI in mutant channels. Furthermore, CaV1.2 channels harbor a homologous DCT; however, the ICDI region of this channel does not appear to appreciably modulate CaV1.2 CDI. Yet coexpression of CaV1.2 ICDI with select CaV1.3 splice variants significantly disrupts CDI, implicating a cross-channel modulatory scheme in cells expressing both channel subtypes. In all, these findings provide new insights into a molecular rheostat that fine-tunes Ca2+-entry and supports normal neuronal and cardiac function.

L-type voltage-gated calcium channels (Ca V 1.1-1.4) are an important conduit for extracellular Ca 2+ entry into many excitable cells including cardiac myocytes, neurons, smooth muscle, and skeletal muscle (1)(2)(3)(4). As such, these channels are subject to rich and powerful modes of feedback regulation (5)(6)(7). In particular, Ca 2+ /calmodulin-dependent inactivation (CDI) of L-type channels is a crucial negative feedback mechanism that reshapes the electrical properties of neurons and cardiac myocytes and protects cells from Ca 2+ overload (8)(9)(10). CDI is driven by the ubiquitous Ca 2+ sensing molecule, calmodulin (CaM) (10,11). Under basal Ca 2+ conditions, Ca 2+free CaM (apoCaM) binds to the carboxy-terminal IQ domain of the channel and enhances channel openings (12). Upon elevation of intracellular Ca 2+ , the "resident" CaM repositions on the channel, interacting with Ca 2+ /CaM binding sites located on the channel amino-and proximal carboxy-termini (13,14). This conformational change antagonizes the initial upregulation in channel open probability, which manifests as CDI. Not surprisingly, CDI of L-type channels has emerged as a key physiological process to limit excess Ca 2+ influx during repetitive or sustained depolarization, and disruption of this feedback in the cardiac myocytes may lead to lethal cardiac arrhythmias (15,16). This stereotypic behavior, however, diverges in multiple physiological settings where strong CDI of L-type channels is curtailed, thus permitting Ca 2+ channels to faithfully respond to a tonic stimulus. For example, in photoreceptors and bipolar cells, endogenous Ca V 1.4 exhibits minimal CDI, thereby allowing sustained Ca 2+ influx and slow, graded changes in the membrane potential necessary for tonic glutamate release, and normal vision (17,18). Similarly, Ca V 1.3 channels in inner hair cells also lack CDI (19). Beyond these, the basal strength of Ca V 1.2 and Ca V 1.3 CDI varies in different neuronal subtypes in the central nervous system, suggesting a sophisticated scheme of Ca V channel feedback ripe with physiological insights (20).
The molecular mechanisms that fine-tune L-type channel CDI are twofold and have been of long-standing interest. One scheme involves channel-interacting proteins such as calmodulinlike Ca 2+ -binding proteins (CaBP1-4) (18,19,21,22) and SH3 and cysteine-rich domain containing proteins (stac1-3) (23-26) that suppress CDI utilizing an allosteric or mixedallosteric mechanism. In contrast, Ca V 1.3 and Ca V 1.4 channels may intrinsically disable CDI via an alternatively spliced specialized CDI-inhibiting module (ICDI) within the distal carboxy tail (DCT) of the channel (20,(27)(28)(29)(30)(31)(32)(33). The latter form of regulation is complex and bears important biological consequences. First, splice inclusion of the DCT occurs in a celltype-dependent manner. For instance, alternative splicing of Ca V 1.3 results in variable inclusion of the ICDI domain in distinct regions of the brain and in the sinoatrial node, enabling precise tuning of CDI in these cell types (27,31,34,35). Second, Ca V 1.2 appears to harbor a highly homologous ICDI region (33), yet CDI for this channel is known to be robust, both when evaluated as full-length channels in heterologous expression system and in primary cells where the carboxy-tail containing ICDI is believed to be cleaved off the channel (9)(10)(11)33). As such, the function of the ICDI module within Ca V 1.2 channels remains unclear. Third, the inhibition of CDI by ICDI is the result of competitive binding by apoCaM versus ICDI with the channel IQ domain (20,31). In addition to diminishing CDI, the displacement of apoCaM results in a dramatic decrease in baseline channel open probability (12). Fourth, adding to the richness the modulatory role of ICDI, RNA editing, and/or fluctuations in cytosolic CaM concentrations can tune the extent of this competition, enabling different degrees of CDI tailored to specific cell types or physiologic states (12,36). Importantly, pathologic changes to this system may be linked to altered CaM concentrations in Parkinson's disease and heart failure (37,38), and mutations within the ICDI of Ca V 1.4 channels are known to be causative of congenital stationary night blindness (32,39,40). Thus, the modulation of CDI by ICDI stands as a critical and robust mechanism for adapting channel regulation to select cell types and conditions. Moreover, as the number of known pathogenic mutations within LTCCs continues to grow, the ability to map these mutations to a locus with known mechanistic impact would enable rapid insight into the pathogenesis of LTCC channelopathies.
Although the overall competitive nature of ICDI regulation of L-type channels is now well established (20,31), the precise binding interfaces involved in this regulation are yet to be identified. This gap in understanding is critical as mutations in the carboxy tail of LTCCs result in neurological disease (32,39,40). Furthermore, a residue-level analysis may shed light upon structural differences between the ICDI domain of Ca V 1.3 and the homologous segment of Ca V 1.2 that engender differential functional regulation. To characterize the landscape of the IQ/ICDI interaction of L-type channels, here we undertook systematic alanine scanning mutagenesis of both IQ and ICDI domains. Through live-cell FRET two-hybrid binding assays and electrophysiological analysis, we identified several novel hotspots on both IQ and ICDI segments that mediate a high-affinity interaction and are functionally relevant for CDI inhibition. Systematic analysis of these mutations revealed a strong inverse correlation between the strength of CDI and the binding affinity of the ICDI domain for the IQ segment, as predicted for a competitive inhibitor (13,14,20). Thus, we have identified residues that alter binding in a functionally relevant manner. Moreover, similar critical residues were identified in adjacent regions, defining a comprehensive interface map of the IQ/ICDI interaction. Finally, extending our analysis to Ca V 1.2 channels, we found that the ICDI module binds to the Ca V 1.2 IQ domain with a reduced affinity and that this binding is insufficient to cause more than a nominal change in the CDI of full-length channels. However, the ICDI from Ca V 1.2 is capable of binding the Ca V 1.3 IQ region with high affinity, resulting in a much larger decrease in CDI of these channels. Given the propensity of the carboxy tail of Ca V 1.2 to exist as a separate peptide within myocytes and neurons (27,33,41,42), these findings raise the prospect of a cross-channel feedback scheme in some cell types. Overall, these results elucidate the detailed binding interface between components of the carboxy tail of L-type Ca 2+ channels, lending new insight into normal and pathologic channel regulation.

Results
Identification of critical residues within the IQ domain necessary for ICDI binding To identify key residues that support a high-affinity IQ/ICDI interaction, we undertook systematic alanine substitution of the IQ domain and evaluated both the relative binding affinity and the strength of ICDI-mediated inhibition of CaM regulation. Importantly, the ICDI domains of both Ca V 1.3 and Ca V 1.4 are highly homologous and have been shown to interact with IQ domains in a similar manner evoking similar functional effects (20,27,29). Even so, the ICDI domain from Ca V 1.4 (ICDI 1.4 ) has a greater binding affinity for the IQ domains of both Ca V 1.3 and Ca V 1.4, with FRET binding assays yielding more robust measurements with enhanced signal-tonoise ratio as compared with ICDI 1.3 (43). We therefore focus on this canonical ICDI motif for our studies. However, robust expression of the holo-Ca V 1.4 channel in recombinant systems is notoriously challenging, largely due to their diminutive open probability (12,44). We therefore chose to explore the interaction between the IQ domain of Ca V 1.3 channels (IQ 1.3 ) and ICDI 1.4 . To this end, we utilized a chimeric channel in which the DCT of Ca V 1.4 is spliced onto the backbone of Ca V 1.3 (Ca V 1.3 Δ/DCT1.4 ) (Fig. 1A), which has previously proven useful in dissecting the mechanisms underlying ICDI modulation of the channel (12,20). This chimera furnishes a strong IQ/ICDI interaction coupled with a robust functional readout, enabling quantitative analysis.
To begin, we confirm the functional impact of ICDI in our chimeric channel by evaluating the extent of CDI in HEK293 cells. Indeed, CDI is entirely abolished in Ca V 1.3 Δ/ DCT1.4 , as seen by the identical Ba 2+ and Ca 2+ current decay in response to a depolarizing pulse (Fig. 1B). However, removal of DCT 1.4 restores robust CDI, as seen by the rapid decay of the Ca 2+ current (Fig. 1C, red). In contrast, when Ba 2+ (which binds poorly to CaM) is used as the charge carrier, there is minimal inactivation (Fig. 1C, black). We therefore define the extent of CDI as the ratio of Ca 2+ current remaining after 300 ms of depolarization versus Ba 2+ current at the same time point.
We next utilized a FRET 2-hybrid binding assay (45,46) to evaluate the relative strength of interactions between the IQ and ICDI regions. FRET binding pairs were constructed by tagging Cerulean fluorescent protein to ICDI 1.4 , and Venus fluorescent protein to PreIQ 3 -IQ-A 1.3 , a peptide that includes IQ 1.3 as well as 30 residues upstream (PreIQ 3 ) and 150 residues downstream (A-region) of the IQ domain (Fig. 1D,  Fig. S1). Both PreIQ 3 and A regions were included initially to ensure that all likely interacting residues were included. Strong binding was detected between the Venus-PreIQ 3 -IQ-A 1.3 and Cerulean-ICDI 1.4 , as can be appreciated by the steep FRET Defining the molecular interface between Ca V IQ and ICDI binding curve determined by the FRET Ratio (FR) of each cell plotted as a function of the free donor concentration (Cerulean tagged ICDI 1.4 ) (Fig. 1D, black). After calibration, the FRET binding curve for WT Venus-PreIQ 3 -IQ-A 1.3 versus Cerulean-ICDI 1.4 yielded a K a of 21.4 μM −1 . To identify key residues that support a high-affinity IQ/ICDI interaction, we undertook systematic alanine substitution of the IQ domain and evaluated the effect on binding affinity in our FRET assay. Within IQ 1.3 , we substituted each residue with an alanine or, at loci where the wild-type channel featured an alanine, we replaced the residue with a threonine. For identification of each residue, the canonical isoleucine is assigned position 0. Application of our FRET assay to each mutated peptide identified three residues, Y[−5]A, F[−2]A, and F[+4]A, which severely perturbed the IQ/ICDI interaction (Fig. 1E, Fig. S1). Focusing on F[−2]A, FRET binding produced a shallower curve as compared with WT ( Fig. 1D, blue versus gray), resulting in a K a of 5.8 μM −1 (Fig. 1E, blue). Introducing this mutation into the chimeric channel resulted in a partial rescue of CDI (Fig. 1F), indicating that this interaction site is functionally relevant. However, the IQ domain substitutions Y[−5]A and F[+4]A, which also had a marked effect on K a , resulted in minimal CDI rescue (Fig. 1, F and G, Fig. S2). Importantly, these residues also serve as anchors for apoCaM binding to the Ca V 1.3 IQ domain, resulting in weak baseline CDI even in the absence of the ICDI domain (Table S1) (13). Of note, this apoCaM effect may also underly the apparent increase in binding affinity demonstrated by several of the mutations, where a decrease in apoCaM binding would enhance the apparent binding of the IQ/ICDI regions due to altered competition with endogenous CaM.
In order to confirm the functional relevance of each binding loci identified, we turned to a previously described analysis known as individually transformed Langmuir (iTL) analysis (14). This approach allows us to rigorously correlate relative changes in binding with functional changes in CDI. As iTL was initially derived to evaluate the binding interfaces critical to CaM-mediated channel regulation (13,14), we adjust the model to reflect the competitive binding scheme between CaM and ICDI (20). To do so, we first account for the ambiguity caused by binding sites, which are important for both apoCaM and ICDI binding. We therefore incorporate both the IQ domain's intrinsic affinity for apoCaM (K a-CaM ) and that for the ICDI segment (K a-ICDI ), the competitive inhibitor, by adjusting our measured K a-ICDI such that: Where K WT a−CaM is the apoCaM binding affinity of WT PreIQ 3 -IQ-A 1.3 , and K mut a−CaM represents the apoCaM binding affinity of each mutant peptide, values which were previously measured (13) and are listed in Table S1. This compensation remains valid provided that the local concentration of ICDI is much greater than K d-ICDI ([ICDI]>> 1/K a-ICDI ). With this adjustment made, CDI can be defined by a modified Langmuir function as follows: where CDI is the strength of CDI under endogenous levels of CaM; CDI max is the CDI in saturating concentrations of CaM (Table S1); K a-CaM is the association constant for apoCaM binding to PreIQ 3 -IQ-A 1.3 ; [apoCaM] is the free apoCaM concentration in the cell; and [ICDI] is the effective local concentration of ICDI (see Supporting information for full derivation). Equation 2 predicts an inverse correlation between K corr a−ICDI and relative CDI, which we can fit to our data (Fig. 1H). For channels containing WT IQ 1.3 and ICDI 1.4 domains, the strong binding between IQ and ICDI results in minimal CDI, as seen by the black data point on the plot. In contrast, Ca V 1.3 Δ channels, which lack an ICDI domain such that K corr a−ICDI ¼ 0 by definition, display large CDI values (cyan). The F[−2]A mutation, which produced a partial restoration of CDI, resides in an intermediate position (blue). Overall, our data can be well fit by Equation 2 (Fig. 1H), confirming the functional relevance of the identified residues in a competitive model.

Alanine scanning of the ICDI domain reveals complementary hotspots
Having identified several critical residues within the IQ domain required for ICDI binding, we next probed the ICDI for critical determinates of binding to the IQ. In order to scan a more extensive segment of the channel, we made triple alanine substitutions for every three contiguous residues within the ICDI domain. The ICDI domain has previously been localized to amino acids 1868 to 1956 of the Ca V 1.3 DCT (29, 32, 33). We therefore undertook our alanine scan on this segment of the channel. Importantly, this region includes the distal C-terminal regulatory domain (DCRD), which was previously identified as playing an important role in the ICDI-mediated inhibition of CDI (27,31,33,47). Finally, we included a selective V[1907]A mutation, as this amino acid change has previously been shown to dramatically alter the function of ICDI (30). To evaluate the effect of these alanine substitutions on IQ/ICDI binding, we again utilized our FRET two-hybrid binding assay, pairing Venus-PreIQ 3 -IQ-A 1.3 with Cerulean-ICDI 1.4 ( Fig. 2A, Fig. S3). Indeed, measured K a values revealed multiple hotspots within the ICDI domain, with a wide range of binding affinities with the IQ containing peptide (Fig. 2B). Interestingly several mutations, largely toward either end of the peptide, appeared to modestly increase in K a. However, these changes may reflect a structural stabilization of the isolated ICDI peptide by promoting helicity of end regions and may not necessarily yield a corresponding change in intact channels. As such, we did not pursue further analysis of these mutants. The two mutation sites displayed in gray (KQE[1911]AAA and YSD[1941]AAA) were not evaluated as they failed to express. Notably, the effects of the hotspots on the ICDI domain were significantly larger than those observed within the IQ domain ( Fig. 2B versus Fig. 1E). In order to correlate loss of binding affinity with function, we measured the CDI of those mutations that exhibited a large change in binding affinity (Fig. 2C, Fig. S4). As predicted, mutations that resulted in a significant loss of IQ/ICDI binding also exhibited a corresponding restoration of CDI.   Figure 1H. Light gray circles indicate data from Figure 1, while dark gray and colored circles indicate mutations within ICDI corresponding to B and C, with green and red circles indicating exemplars shown in other panels for reference. The PKA phosphorylation site Ser1883 is plotted as the orange circle.
Defining the molecular interface between Ca V IQ and ICDI binding (Fig. 2, A and B, red), while introduction of the same mutations into our chimeric channel enabled a partial restoration of CDI (Fig. 2, C and D, red). On the other hand, SLV [1886]AAA displayed a drastic reduction in IQ/ICDI binding (Fig. 2, A and B green), and CDI was fully restored to the level seen in Ca V 1.3 Δ (Fig. 2, C and D green). Importantly, all identified loci are well fit by our Langmuir function, such that the same set of Equation 2 parameters describes both the IQ region and ICDI (Fig. 2E). Of note, as mutations in ICDI do not affect the binding of apoCaM, the correction factor for K a is no longer required, and K corr a−ICDI ¼ K a−ICDI . Having identified critical loci within ICDI, we note that one of these amino acids (S[1866]) has previously been identified as a phosphorylation site, which reduces the binding affinity of ICDI 1.4 by about tenfold, while increasing CDI of Ca V 1.3 Δ/DCT1.4 (43). We therefore included the results of phosphorylation at this site in our analysis (Fig. 2E, orange). Indeed, phosphorylation of this amino acid results in a change in binding affinity, which correlates with CDI according to the same Langmuir function.
Thus, we have identified numerous residues within ICDI 1.4 , which are critical determinants of a functional competition between ICDI and apoCaM for the IQ region of Ca V 1.3.

The role of the A region in the IQ/ICDI interaction
While our analysis identified several functionally relevant IQ domain loci, the impact of these mutations was far less than those identified within the ICDI (Fig. 1 versus Fig. 2). This suggests that additional regions outside the IQ domain may contribute to ICDI binding. In order to identify such regions, we generated truncated variations of our Venus-PreIQ 3 -IQ-A 1.3 construct and paired them with Cerulean-ICDI 1.4 in the FRET two-hybrid binding assays (Fig. 3A, Fig. S5). We began by removing the PreIQ 3, and found no change in FRET binding, indicating that all relevant interaction loci are contained within the IQ and A regions (Fig. 3B, blue). However, removal of the IQ domain, leaving only the A region intact, resulted in a complete loss of FRET binding (Fig. 3B, open circles). Likewise, the IQ region alone displayed no binding with ICDI (Fig. 3C), suggesting that both the IQ and the A region are necessary for interaction with ICDI. In order to further localize the critical interaction sites, we undertook successive truncation of the vernus-IQ-A 1.3 peptide (Fig. 3D). FRET measurements demonstrated minimal effect of truncations up to 34 amino acids from the end of the A region, as demonstrated by the strong binding of Venus-IQ-A 1.3 Δ34 with ICDI 1.4 (Fig. 3E, green). However, our next truncation, Venus-IQ-A 1.3 Δ28, exhibited a marked decrease in FRET binding (Fig. 3, E and F, red) Thus, the 34 residues immediately downstream of the IQ domain critically augment ICDI binding. Of note, this region includes the previously identified proximal C-terminal regulatory domain (PCRD), which is reported to play an important role in the ICDI interaction (27,31,33,47). Having identified a subset of the A region, which is vital to ICDI binding, we again undertook systematic alanine substitutions, replacing each of the three contiguous residues with three alanine residues and undertook our FRET-2-hybrid binding assay (Fig. 3G). Disruptions in binding were identified as the result of a number of mutations, spanning both the previously identified PCRD region and a previously unidentified region upstream of this motif (Fig. 3, H and I,  Fig. S6). Thus, both the IQ and the distal A region of the channel are required for high-affinity interaction with ICDI.
The functional relevance of ICDI in Ca V 1.2 channels Similar to Ca V 1.3 and Ca V 1.4, Ca V 1.2 channels also feature a highly homologous ICDI segment, argued to function as a channel inhibitor (33,48) or as a transcriptional factor (41,42). We therefore considered the impact of ICDI 1.2 on both Ca V 1.2 and Ca V 1.3 channels. We interrogated the binding of Cerulean-ICDI 1.2 with Venus-PreIQ 3 -IQ-A 1.2 via FRET two-hybrid (Fig. 4, A and B) and found that the interaction is significantly weaker than the prototypic Venus-PreIQ 3 -IQ-A 1.2 and Cerulean-ICDI 1.4 interaction (Fig. 4B versus Fig. 1D). However, when paired with Venus-PreIQ 3 -IQ-A 1.3 , binding with ICDI 1.2 is significantly larger, and only about half that of the strong binding of ICDI 1.4 (Fig. 4C). Thus, it appears that ICDI 1.2 is poised to have a larger effect in the context of Ca V 1.3 channels as compared with its native channel backbone. Nonetheless, the limited binding observed between Cerulean-ICDI 1.2 and Venus-PreIQ 3 -IQ-A 1.2 prompted us to evaluate the possibility of a functional role for ICDI 1.2 within Ca V 1.2 channels. Interestingly, truncation of Ca V 1.2 at the known carboxy-tail cleavage site (49) for this channel (Ca V 1.2 Δ1800 ) resulted in a minimal, yet statistically significant (p ≤ 0.05), increase in CDI (Fig. 4, D and F, Fig. S7). Next, to test the effect of ICDI 1.2 on Ca V 1.3 channels, we replaced the native ICDI 1.3 of Ca V 1.3 long channels with ICDI 1.2 . Indeed, the loss of CDI surpassed that of Ca V 1.2 channels (Fig. 4E), as predicted based on the stronger PreIQ 3 -IQ-A 1.3 /ICDI 1.2 interaction (Fig. 4C). In fact, the CDI exhibited by Ca V 1.3-ICDI 1.2 channels was not statistically different than the CDI measured in the native Ca V 1.3 long splice variant (Fig. 4F).
Multiple studies have shown that the DCT of Ca V 1.2, containing ICDI 1.2 , exists as a peptide within neurons and cardiomyocytes, either due to proteolysis (33,49) or as a result of alternative transcriptional initiation sites (41). Moreover, it has previously been demonstrated that ICDI domains can exert their effects on L-type channels when expressed as separate peptides (27). We therefore sought to recreate the potential interaction of select Ca V 1.3 channel variants with the DCT of Ca V 1.2. To begin, we choose the human Ca V 1.3 43S splice variant of Ca V 1.3, as these channels terminate just past the A region and thus lack an inherent ICDI module (27). In addition, the inclusion of the A region within Ca V 1.3 43S has been demonstrated to be important for ICDI binding, both in functional experiments done by others (27) and in our alanine scan of the A region (Fig. 3). We therefore generated the proteolytic product of human Ca V 1.2 channels (DCT 1.2 ) and evaluated the effect of this peptide on the CDI of Ca V 1.3 43S . Indeed, consistent with previous studies (27), coexpression of DCT 1.2 significantly reduced the CDI of Ca V 1.3 43S (Fig. 4G,  Fig. S7). For comparison, we also coexpressed these channels with ICDI 1.4 expressed as a peptide, which we have shown has a K a about double that of ICDI 1.2 (Fig. 4C versus Fig. 1D). Indeed, ICDI 1.4 results in an even larger CDI deficit when expressed with Ca V 1.3 43S (Fig. 4, G and H). Thus, the proteolytically cleaved DCT 1.2 is well poised to exert a significant modulation of select Ca V 1.3 channel variants, such that the ambient concentrations of CaM and DCT 1.2 are able to tune the CDI of Ca V 1.3 43S channels in a competitive manner (Fig. 4I).

Discussion
CaM regulation of Ca V channels is vital to normal physiology and thus has been the subject of intense study (8,10,11,17,(50)(51)(52). The competitive mechanism of ICDI within select Ca V 1 channels forms a basis with which CaM regulation can be tuned (12,20,32). Numerous processes designed to modulate this regulation include splice variation, RNA editing, variations in ambient CaM concentration, and phosphorylation (12,13,18,36,43). Identification of critical loci involved in this regulation is therefore key to understanding how CaM regulation may vary in different physiological and pathological states. As such, in-depth residue-level analysis not only reveals interfaces utilized by cells to tune channel regulation, but may offer targets in the search for novel regulators of the channel, which may have therapeutic benefit. In particular, the dramatically different efficacy of ICDI across channel subtypes may offer the possibility of subtype selective drug targeting, which remains challenging for Ca V 1 channels.
Given the importance of understanding these interactions within the carboxy-tail of Ca V 1 channels, we quantified the structure-function relationship of these interactions using a variant of previously described iTL analysis (14). This provided a major advantage in that the quantitative agreement of our results with Equation 2 demonstrates that each identified locus is functionally relevant. This overcomes a common limitation of binding assays between channel fragments, which may identify sites that are inaccessible or inconsequential in the context of the holochannel. Moreover, by fitting to a specific Langmuir curve, we can distinguish mutations that may alter Defining the molecular interface between Ca V IQ and ICDI Figure 4. Residual functionality of ICDI1.2 in modulating Ca V channels. A, cartoon depicting FRET interacting pairs for panels B and C. Venus-PreIQ 3 -IQ-A 1.2 or Venus-PreIQ 3 -IQ-A 1.3 was paired with Cerulean-ICDI 1.2 in order to evaluate the role of the ICDI contained within Ca V 1.2 channels. B, Venus-PreIQ 3 -IQ-A 1.2 displays moderate binding with Cerulean-ICDI 1.2 (blue) as compared with the robust binding between Venus-PreIQ 3 -IQ-A 1.3 and Cerulean-ICDI 1.4 which is reproduced as the gray line for comparison. C, Venus-PreIQ 3 -IQ-A 1.3 is able to bind strongly with Cerulean-ICDI 1.2 . D, Left, channel cartoon with cleavage site indicated by the yellow arrowhead. Right, exemplar data demonstrating that truncation of Ca V 1.2 at the cleavage site has a minor effect on CDI. E, Left, channel cartoon indicating introduction of ICDI 1.2 into the Ca V 1.3 channel backbone. Right, introduction of ICDI 1.2 into Ca V 1.3 channels causes a large decrease in CDI. F, average CDI data demonstrating a modest but statistically significant effect of ICDI 1.2 on Ca V 1.2 channels (blue), and a large effect on Ca V 1.3, which is comparable to the native effect of ICDI 1.3 contained within Ca V 1.3 long (gray). Data are plotted as mean ± SD. (Ca V 1.2: *p < 0.05 based on a Student's t-test; Ca V 1.3: ANOVA summary analysis indicates a likelihood of significant difference at a level of p < 0.0001. Post-hoc Dunnett's multiple comparisons versus Ca V 1.3 short indicated as ****p < 0.0001.) G, Left, cartoon of human Ca V 1.3 43S , with the sequence of the end of the channel displayed below. This splice variant contains a portion of the A region (blue highlight), which contains the identified hotspots required for ICDI binding (bold), as well as a short sequence of unique amino acids prior to truncation of the channel (pink). Right, exemplar whole-cell patch clamp data demonstrates robust CDI in WT Ca V 1.3 43S , which is reduced when the predicted cleavage fragment of the human Ca V 1.2 channel is coexpressed, or when ICDI 1.4 is expressed as a peptide. H, average CDI effects for the channels described in panel G, DCT 1.2 and ICDI 1.4 both have significant effects on the CDI of Ca V 1.3 43S (ANOVA summary analysis indicates a likelihood of significant difference at a level of p < 0.0001. Post-hoc Dunnett's multiple comparisons versus Ca V 1.3 43s indicated as ****p < 0.0001; data are displayed ±SD). I, cartoon proposing a cross-channel effect of ICDI 1.2 , such that cleaved or independently transcribed DCT 1.2 may modulate Ca V 1.3 43S channels. channel function through ancillary mechanisms such as transduction or altered folding of the channel. Thus, in addition to identifying critical loci, our results confirm the competitive mechanism described for ICDI modulation of CDI.
A number of previous studies have identified regions within the carboxy tail of Ca V 1 channels, which are critical to the competitive mechanism of ICDI inhibition. Among these are the PCRD and DCRD regions, which were identified within Ca V 1.2 channels as potential interaction sites such that the DCRD region of the proteolytically cleaved Ca V 1.2 DCT may interact with the PCRD on the channel via electrostatic interaction with the negatively charged amino acids (33). Homologous regions within Ca V 1.3 and Ca V 1.4 were later shown to be important for ICDI regulation and for the interaction with modular peptides (27,30,31,33,47). In this study, the PCRD resides within the A region and overlaps with the identified locus of critical amino acids required for highaffinity binding between the IQ-A and ICDI. Interestingly, these critical amino acids were identified both within the PCRD and upstream of the motif, arguing for a larger interacting region, which is highly conserved across Ca V channels (Fig. S8). However, our binding assay also demonstrated that the A region, in itself, is insufficient for high-affinity binding, but also requires the upstream IQ region (Fig. 3A). This fits with previous findings in which neutralization of the PCRD arginines was not sufficient to prevent the ICDI inhibition of CDI (31), pointing to the existence of additional interaction loci. In a similar manner, our scan of the ICDI region validated the importance of the DCRD, while also identifying numerous critical interacting loci upstream of the motif (Fig. S8). Thus, this study has expanded our knowledge of the important interactions required for ICDI inhibition, providing a comprehensive map of the critical loci within the carboxy tail.
The impact of ICDI in Ca V 1.3 and Ca V 1.4 channels has been well recognized; however, its role in Ca V 1.2 has been uncertain. It has been demonstrated that the truncation of Ca V 1.2 results in increased current density, altered voltage dependence of channel activation, and disrupted targeting of the channel to the membrane (33,53,54); however, no impact on CDI has been reported. Here, we find that the impact of ICDI 1.2 within Ca V 1.2 channels is minimal, allowing cleavage of this DCT region without significant disruption of CDI. Yet ICDI 1.2 is capable of causing significant disruption of CDI in the context of Ca V 1.3 (Fig. 4, E-I) (27). This selective effect of ICDI 1.2 on Ca V 1.3 channels is intriguing as it represents a difficult to achieve discrimination between Ca V 1.2 and Ca V 1.3 (55,56). The strong homology between these two channels has resulted in challenges to dissecting the contribution of each channel to the function of cells, which express both channel subtypes, and hinders therapeutic options for neuropsychiatric disorders, which may benefit from blockade of Ca V 1.3 (55). As such, the interface defined in this study between the IQ-A region and ICDI may represent a promising interface with which to selectively target Ca V 1.3.
The DCT of Ca V 1.2 is routinely cleaved in neurons and myocytes, with the cleavage product able to either remain associated with the channel (33) or translocate to the nucleus (41). Our FRET binding data would argue that this association between Ca V 1.2 and the DCT may be relatively weak, leaving DCT 1.2 available to other binding partners. Moreover, it has been shown that an alternate start site exists within the carboxy tail of Ca V 1.2, such that alternative transcription of Ca V 1.2 will produce a DCT peptide containing a calcium channel-associated transcription regulator (CCAT) (41). Importantly, ICDI would be intact within this peptide, providing an additional source of DCT 1.2 within cells. In addition, CCAT has been shown to localize to the cytosol in a Ca 2+ -dependent manner, providing a source of ICDI in proximity to the membrane, which can be tuned by activity (41). Our FRET binding analysis (Fig. 4A) suggests that the Ca V 1.2 DCT is capable of binding upstream calmodulatory elements in Ca V 1.2, albeit weakly. Functional analysis, however, suggests only minimal effects of this segment on Ca V 1.2 CDI. By comparison, the Ca V 1.3 43S channel variant contains all the elements required for high-affinity binding with DCT 1.2 and exhibits functional inhibition of CDI (Fig. 4G), consistent with previous studies (27). Moreover, Ca V 1.3 43S is widely expressed in the brain, accounting for up to 39% of transcripts, and is often expressed within the same neuron type as Ca V 1.2 (27). Thus, it seems likely that DCT 1.2 may interact with this channel, altering the normally robust CDI. It is interesting to note, however, that while DCT 1.2 is poised to modulate some Ca V 1.3 channels, the same cannot be said of DCT 1.3 . Not only is the IQ-A region of Ca V 1.2 suboptimal for binding to ICDI, but there is little evidence that Ca V 1.3 is cleaved in neurons (31). Thus, this cross-channel modulation may be unidirectional. Finally, since Ca V 1.2 and Ca V 1.3 often exist within the same neuron, this mode of cross-channel modulation may represent an important method for tuning CDI in different regions of the brain.

Molecular biology
The rat brain Ca V 1.3 α1 subunit (in pcDNA6) corresponds to AF370009.1 (57) and was incorporated to the mammalian expression plasmid pCDNA6 (Invitrogen) (20). This plasmid features a unique BglII restriction site at a locus corresponding to 450 amino acids upstream of the IQ domain and a unique XbaI site after the stop codon, which were used for generation of mutant and chimeric plasmids as descried below. The Ca V 1.2 α1 subunit (in pGW) is identical to rabbit NM001136522 (58), and the Ca V 1.4 channel (in pcDNA3) is the human clone corresponding to NP005174.2. Ca V 1.3 Δ/ DCT1.4 was made by fusing with the DCT of the Ca V 1.4 α1 subunit to the Ca V 1.3 α1 subunit (truncated after the IQ domain), as previously described (20). Ca V 1.3 43S was made by PCR amplification of the channel segment between the BglII site and IQ domain with the appendage of amino acids unique to this splice variant (27). The PCR product was then inserted into the channel via the BglII/XbaI sites.
FRET constructs were fluorescent-tagged (either Venus or Cerulean) using similar strategies as previously described (45).
Defining the molecular interface between Ca V IQ and ICDI Briefly, Venus and Cerulean fluorophores (a kind gift from Dr Steven Vogel, NIH) were subcloned into the pcDNA3 vector via unique KpnI and NotI sites. The PCR-amplified channel peptides, as described in Liu et al., (20) were then cloned via unique NotI and XbaI sites. Mutations were introduced into the channel or FRET construct via PCR amplification or overlap extension PCR.

Whole-cell patch clamp recordings
Whole-cell recordings were obtained using an Axopatch 200A amplifier (Axon Instruments). Electrodes were pulled from borosilicate glass capillaries (World Precision Instruments), with 1 to 3 MΩ resistances, which were in turn compensated for series resistance by >60%. Currents were low-pass filtered at 2 kHz before digital acquisition at five times the frequency. A P/8 leak subtraction protocol was used. Data was analyzed using custom Matlab scripts. Inactivation was quantified as the ratio of current remaining after 300 ms (current amplitude measured at 300 ms divided by peak current amplitude) in either Ca 2+ or Ba 2+ (r 300 ). CDI was then quantified as the r 300 in Ca 2+ subtracted from the r 300 in Ba 2+ , measured at 10 mV for Ca V 1.3 channels, and 30 mV for Ca V 1.2.

FRET optical imaging
FRET two-hybrid experiments were performed on an inverted microscope as described (45,46). The bath solution was a Tyrode's solution composed of (in mM): NaCl, 138; KCl, 4; MgCl2, 1; HEPES (pH 7.4), 10; CaCl2, 2; at 305 mOsm adjusted with glucose. Background fluorescent signals were measured from cells without expression of the fluorophores and subtracted from cells expressing the fluorophores. Concentration-dependent spurious FRET was subtracted from the raw data prior to binding-curve analysis (45,46). Cerulean (59) and Venus (60) were used as the donor and acceptor fluorescent proteins instead of eCFP and eYFP, as their optical properties provided more robust and stable FRET signals. Acceptor-centric measurements of FRET were obtained with the 3 3 -FRET algorithm (45,46), in which the effective FRET efficiency (E EFF ) and FRET ratio (FR) are defined as: where E is the FRET efficiency of a donor-acceptor pair, A b is the fraction of acceptor molecules bound by a donor, and ε Venð440 nmÞ =ε Cerð440 nmÞ is the approximate molar extinction coefficients of Cerulean and Venus, which was measured as 0.08 on our setup. Intensity measurements at each wavelength were taken from individual cells such that variable expression across the cells enabled population of a binding curve. Binding curves were analyzed using GraphPad software (Prism), providing relative K d values and standard error based on an unbiased fit to the data. These relative K d values were then calibrated according to a previously determined calibration factor (13,20) and converted to K a = 1/K d . Importantly, the previous calibration factor determined for our setup utilized CFP/YFP FRET pairs. In order to account for the difference in FR using Cerulean and Venus, we determined the relative K d for multiple peptides using both the fluorophore pairs and found that the two data sets differed by a factor of 1.8, which we incorporated into the calibration factor.

Data availability
All data is contained within the article.