Ca V channels reject signaling from a second CaM in eliciting Ca 2+ -dependent feedback regulation

Calmodulin (CaM) regulation of voltage-gated calcium (Ca V 1-2) channels is a powerful Ca 2+ feedback mechanism to adjust channel activity in response to Ca 2+ influx. Despite progress in resolving mechanisms of CaM-Ca V feedback, the stoichiometry of CaM interaction with Ca V channels remains ambiguous. Functional studies that tethered CaM to Ca V 1.2 suggested that a single CaM sufficed for Ca 2+ feedback. Yet, biochemical, FRET, and structural studies showed that multiple CaM molecules interact with distinct interfaces within channel cytosolic segments suggesting that functional Ca 2+ -regulation may be more nuanced. Resolving this ambiguity is critical as CaM is enriched in subcellular domains where Ca V channels reside, such as the cardiac dyad. We here localized multiple CaM to the Ca V nanodomain by tethering either wild-type or mutant CaM that lack Ca 2+ -binding capacity to the pore-forming α subunit of Ca V 1.2, Ca V 1.3, and Ca V 2.1 and/or the auxiliary β 2A subunit. We observed that a single CaM tethered to either the α or β 2A subunit tunes Ca 2+ -regulation of Ca V channels. However, when multiple CaMs are localized concurrently, Ca V channels preferentially respond to signaling from the α -subunit tethered CaM. Mechanistically, the introduction of a second IQ domain to the Ca V 1.3 carboxy-tail switched the apparent functional stoichiometry

Despite these advances, one uncertainty pertains to the stoichiometry of CaM interaction with the Ca V complex. Early functional studies that tethered CaM on to the CT of the pore-forming α 1 subunit suggested that a single CaM is both necessary and sufficient for Ca V 1 regulation (59). However, biochemical and structural analysis point to the binding of multiple CaM molecules within the Ca V complex. Briefly, atomic structures of the Ca V 1.2 and Ca V 2.1 CT show Ca 2+ /CaM interaction with the IQ domain (47, 60,61), as well as two pre-IQ domains cross-bridged by two additional Ca 2+ /CaM molecules (46,61). For Ca V 1.2, NMR structures show the binding conformation of Ca 2+ /CaM to the Ca V 1.2 NSCaTE domain (56). It remains unknown whether a single CaM molecule switches between conformations (62), or whether multiple CaM molecules engage distinct sites (63) to orchestrate channel regulation. This mechanistic ambiguity is biologically important as CaM is enriched in subcellular regions such as the cardiac dyad where Ca V 1 channels also reside (64). In vitro analysis suggests that Ca 2+ /CaM is not capable of bridging the aforementioned channel domains (63,65). Furthermore, previous FRET-based analysis of CaM stoichiometry showed that while a single apoCaM pre-associates with the holo-Ca V 1.2 channel, in the presence of Ca 2+ , up to two CaM can bind to the holo-channel complex (66). Given this ambiguity, we here sought to dissect the potential role of multiple CaM in eliciting Ca 2+dependent modulation of Ca V channel gating by tethering mutant or wild-type CaM to distinct locations within the channel complex. We found that CaM linked to the channel CT is privileged in eliciting Ca 2+ -regulation of Ca V channels. Furthermore, when additional Ca 2+ /CaM are present in the channel complex, signaling by these molecules are rejected by the channel pore domain with regards to dynamic Ca 2+ -feedback modulation.

Strategy for probing effects of multiple CaM in tuning Ca V 1 CDI
To dissect the potential functional contribution of multiple CaM in evoking CDI of Ca V 1.2, we localized either wild-type (CaM WT ) or mutant CaM that lacks Ca 2+ binding (CaM 1234 ) with known stoichiometries through genetic fusion to either the pore-forming α 1C subunit or the auxiliary β 2A subunit (59,67) (Figure 1A). Here, CaM 1234 mutant is generated by alanine substitution of key Ca 2+coordinating aspartate (D) residues in all four EF hand domains of CaM. This overall strategy allows us to localize either one or two CaM molecules to the channel complex and assess changes in CDI. Figure 1B shows baseline extent of CDI for full length Ca V 1.2 in the absence of CaM fusion to either the α 1C or β 2A subunits. In response to a stepvoltage depolarization to +10 mV, Ca 2+ current decay (red) is accelerated in comparison to Ba 2+ current (black). Population data shows the fraction of peak current remaining after 300 ms depolarization (r 300 ) with either Ca 2+ (red) or Ba 2+ (black) as the charge carrier. As Ba 2+ binds poorly to CaM (68), the Ba 2+ relation provides a baseline measure of voltage-dependent inactivation (VDI). The magnitude of CDI is quantified as the fractional difference between r 300 relations obtained with Ca 2+ and Ba 2+ as permeant ions (i.e., CDI 300 = 1 -r 300,Ca / r 300,Ba ) ( Table 1). Further exemplar currents and current-voltage relationships are provided in Supporting Information Figure S1A. Previous studies have demonstrated that fusion of CaM WT to the carboxyterminus of the truncated α 1C Δ1671 subunit preserves strong CDI while tethering CaM 1234 to the same location abolishes CDI suggesting that CaM fusion to the channel carboxy-terminus preserves modulatory function and permits interaction with key effector interface on the channel (59,69). As such, we confirmed that truncation of the distal carboxy-tail does not appreciably alter CDI ( Figure 1C; Figure S1B; Table 1). We further validated that α 1C Δ1671-CaM WT supported strong CDI ( Figure 1D; Figure  S1C) while α 1C Δ1671-CaM 1234 abolished CDI ( Figure 1E; Figure S1D). To determine whether genetic fusion of CaM to the β 2A subunit similarly supports Ca V 1.2 regulation, we tethered CaM WT and CaM 1234 onto the β 2A subunit, yielding β 2A -CaM WT or β 2A -CaM 1234 respectively. Notably, the β 2A subunit binds to the α 1 subunit with a high affinity and 1:1 stoichiometry (70,71), and this subunit is obligatory for channel function in HEK293 cells. As such, we co-expressed α 1C with either β 2A -CaM WT or β 2A -CaM 1234 . β 2A -CaM WT coexpression with α 1C subunit preserved strong CDI ( Figure 1F, Figure S1E, Table 1) similar to control conditions. By contrast, co-expression of β 2A -CaM 1234 abolished CDI ( Figure 1G, Figure S1F, Table 1). Thus, CaM tethered to the β 2A subunit is also capable of binding to critical channel effector motifs and eliciting functional regulation.

Ca V 1 preferentially responds to CaM tethered to the channel carboxy-terminus
Having verified the functionality of tethered CaM, we sought to determine channel regulation when multiple CaM molecules are localized within the channel nanodomain. As both the α 1C CT and the β 2A subunit are within close proximity of the channel pore (less than 10 nm) based on cryoelectron microscopy structure (71), CaM tethered to either domain is exposed to similar local Ca 2+ fluctuations (72,73). Thus, if two wild-type CaM molecules are attached to the channel complex, we anticipate strong CDI akin to channels that lack tethered CaM, since either one or both CaM molecules can interact with respective channel effector interfaces. Indeed, co-expression of β 2A -CaM WT with α 1C Δ1671-CaM WT resulted in appreciable CDI (Figure 2A; Figure S1G; Table 1), albeit modestly reduced in comparison to channels lacking tethered CaM (~25% reduction). However, if the channel complex comprises of one mutant and one wildtype CaM, then four distinct functional outcomes emerge depending on the underlying mechanism of channel modulation: (Scenario I) If two Ca 2+ /CaM independently orchestrate channel modulation, then this maneuver would result in a partial disruption of CDI regardless of whether CaM 1234 is tethered to α 1C CT or β 2A . (Scenario II) If two Ca 2+ /CaM cooperatively modulate Ca V 1.2 regulation, then the presence of one CaM 1234 in the channel complex tethered to either α 1C or β 2A would exert a dominant negative effect and fully inhibit CDI. (Scenario III) If instead functional channel modulation relied on a single Ca 2+ -bound CaM, then the presence of one CaM WT tethered to either α 1C CT or β 2A would elicit full CDI. (Scenario IV) A final nuanced possibility is that CDI depends only on a single CaM but one that is pre-bound to a particular interface. In this scenario, the modulatory effect will be binary depending on whether CaM WT or CaM 1234 occupies the interface responsible for triggering CDI. To dissect between these possibilities, we first coexpressed α 1C Δ1671-CaM 1234 with β 2A -CaM WT . Comparison of Ca 2+ versus Ba 2+ current decay demonstrates a strong reduction of CDI ( Figure  2B; Figure S1H; Table 1). This result eliminates both Scenarios I and III. To distinguish between Scenarios II and IV, we co-expressed α 1C Δ1671-CaM WT with β 2A -CaM 1234 ( Figure 2C; Figure S1I). This maneuver resulted in strong CDI indistinguishable from that observed upon coexpression of either α 1C or α 1C Δ1671 with β 2A (Table 1). This result confirmed Scenario IV with a single CaM pre-bound to the channel carboxy-tail being privileged in triggering CDI. To further ensure that the glycine linkage of CaM to either the α 1C Δ1671 or the β 2A subunit did not occlude accessibility of CaM to effector interfaces, we measured CDI of α 1C Δ1671-CaM 1234 with β 2A and α 1C in the presence of freely diffusible CaM. As with α 1C Δ1671-CaM 1234 we observed no CDI even upon CaM overexpression ( Figure S2A-B; Table  1). As a further control, we also co-expressed α 1C with β 2A -CaM 1234 and freely diffusible CaM. In this case, we again found no CDI, consistent with β 2Alocalized CaM occupying the carboxy-tail site (Figure S2C-D; Table 1). These findings further confirm that glycine-linkage does not prevent CaM from reaching critical sites. In all, these results demonstrate that Ca V 1.2 is preferentially regulated by a single CaM associated with the channel CT, in effect rejecting Ca 2+ /CaM signaling from the β 2Atethered CaM.
To assess generality, we considered the stoichiometric basis for CaM regulation of Ca V 1.3. Accordingly, Ca V 1.3 exhibits strong CDI at baseline as shown in Figure 3A ( Figure S3A; Table  2), consistent with previous studies. As with Ca V 1.2, we have previously demonstrated that fusion of CaM WT to the Ca V 1.3 CT (α 1D -CaM WT ) supports strong CDI, while attaching CaM 1234 at the same locus (α 1D -CaM 1234 ) abolishes CDI (69,74). To confirm functionality of CaM linkage to the β 2A subunit, we co-expressed β 2A -CaM WT or β 2A -CaM 1234 with the α 1D pore-forming subunit. We observed robust CDI for Ca V 1.3 in the presence of β 2A -CaM WT similar to that observed with the β 2A subunit alone ( Figure 3B; Figure S3B; Table 2). By contrast, co-expression of β 2A -CaM 1234 abolished CDI ( Figure 3C, Figure S3C; Table 2) suggesting that CaM linked to the β 2A subunit is capable of eliciting functional regulation. Thus assured, we sought to deduce the effect of localizing two CaM WT to the Ca V 1.3 complex. As anticipated, strong CDI was observed when α 1D -CaM WT was co-expressed with β 2A -CaM WT ( Figure 3D; Figure  S3D; Table 2). Subsequently, we co-expressed α 1D -CaM 1234 with β 2A -CaM WT and measured CDI ( Figure 3E; Figure S3E; Table 2). As with Ca V 1.2, this combination sufficed to strongly attenuate CDI. In contrast, co-expression of α 1D -CaM WT with β 2A -CaM 1234 fully spared CDI ( Figure 3F; Figure  S3F; Table 2). To ensure that these findings did not result from a steric limitation imposed by tethered CaM, we considered whether overexpression of freely diffusible recombinant CaM WT could reverse CDI deficits of either α 1D -CaM 1234 with β 2A ( Figure  S4A Table 2). In both cases, we observed no CDI confirming that CaM localized to the channel preferentially regulated channel function (Fig. S4). Taken together, the binary switching of channel regulatory behavior observed with localizing one CaM WT and one CaM 1234 suggests that functional Ca V 1.3 regulation is preferentially triggered by CaM in close vicinity of the channel CT.

Functional CaM stoichiometry for Ca V 1 is limited by the number of Ca V IQ domains
To delineate mechanisms that govern CaM stoichiometry for channel regulation, we considered whether Ca V 1 could be engineered to be responsive to multiple CaM molecules. Accordingly, we constructed Ca V 1.3 channels containing two IQ domains in tandem in the carboxy-tail (Ca V 1.3 2xIQ ), fused to either CaM WT (termed α 1D/2xIQ -CaM WT to denote CaM WT fusion to the pore-forming α-subunit) or CaM 1234 (α 1D/2xIQ -CaM 1234 ) and co-expressed with β 2A , β 2A -CaM WT , or β 2A -CaM 1234 . We observed strong CDI for α 1D/2xIQ -CaM WT similar to Ca V 1.3 ( Figure 4A; Figure S5A; Table 3). By comparison, CDI of α 1D/2xIQ -CaM 1234 was sharply diminished similar to α 1D -CaM 1234 , although not fully eliminated ( Figure  4B; Figure S5B; Table 3). These findings suggest that CT-linked CaM remains vital for CDI of Ca V 1.3 containing tandem IQ domains. Furthermore, co-expression of α 1D/2xIQ -CaM WT with β 2A -CaM WT also revealed strong CDI similar to α 1D/2xIQ -CaM WT co-expressed with β 2A alone ( Figure 4C; Figure S5C; Table 3). However, when α 1D/2xIQ -CaM WT is co-expressed with β 2A -CaM 1234 , we observed a partial reduction in CDI ( Figure 4D; Figure S5D; Table 3). In like manner, coexpression of α 1D/2xIQ -CaM 1234 with β 2A -CaM WT also showed a partial reduction in CDI ( Figure 4E; Figure S5E; Table 3). By contrast, localizing two mutant CaM 1234 to the channel complex by coexpressing α 1D/2xIQ -CaM 1234 with β 2A -CaM 1234 revealed a complete disruption of CDI ( Figure 4F; Figure S5F; Table 3). This behavior is distinctly different from a single IQ domain containing Ca V 1.3 (Fig. 3), where a binary change in CDI is observed depending on the Ca 2+ -binding ability of carboxy-terminally linked CaM. Instead, the stepwise change in CDI with one versus two mutant CaM 1234 is consistent with Scenario I considered above. This outcome suggests a 2:1 functional CaM stoichiometry for mutant Ca V 1.3 2xIQ .
Two mechanistic possibilities may engender this switch in functional CaM stoichiometry: First, the number of apoCaM molecules within the Ca V 1 complex may be the determining parameter for functional CaM stoichiometry. Our previous work using holochannel FRET 2-hybrid assay showed that although two Ca 2+ /CaM molecules bind the holo-Ca V 1 channels, only a single apoCaM preassociates with the full-length channel (66 Figure 5A; Figure S6A; Table S1) compared to wildtype Ca V 1.3. As M13 interacts only with Ca 2+bound form of CaM, this result suggests that apoCaM preassociation is obligatory for CDI. Notably, Ca V 1.3-M13 channels also exhibited increased voltage-dependent inactivation (VDI), reminiscent of previous observations of increased VDI upon disrupting apoCaM binding. Unlike the M13 peptide, the IQ domain of unconventional myosin Va interacts with both apoCaM and Ca 2+ /CaM with a high affinity comparable to Ca V channel IQ domain. If the number of apoCaM molecules in the channel complex sufficed to determine functional regulation and stoichiometry, then substitution of the Ca V 1.3 IQ domain with the IQ domain from the unconventional myosin Va would preserve CDI triggered by a single CaM. However, Ca V 1.3-MyoIQ channels failed to trigger appreciable CDI suggesting that high-affinity apoCaM and Ca 2+ /CaM interaction with the channel alone are insufficient for CDI ( Figure 5B; Figure S6B; Table S1). As a further test, we considered whether substitution of the Ca V 1.3 IQ domain with the IQ domain from the related Na V 1.4 channels might support functional channel regulation. Of note, Na V 1.4 undergoes CDI with similar underlying mechanisms as Ca V 1.3 (15). Intriguingly, whole-cell recordings of Ca V 1.3-Na V 1.4IQ revealed recognizable CDI although with reduced magnitude in comparison to wildtype Ca V 1.3 ( Figure 5C; Figure S6C; Table S1). Taken together, these findings suggest that the Ca V /Na V IQ domain are privileged in Ca V /Na V channel modulation and this domain likely plays an important for orchestrating downstream structural rearrangements of channel cytosolic domains. One caveat is that although M13 and MyosinVa IQ domain bind to CaM in isolation, it is possible that CaM binding to these domains may be disrupted upon attachment to the Ca V 1.3 carboxy-tail. Overall, these results are consistent with the possibility that the functional CaM stoichiometry for Ca V 1 is dictated by the number of IQ domains in the channel carboxy-terminus.

Distinct modes of Ca V 2.1 regulation is preferentially evoked by carboxy-terminus linked CaM
CaM regulation of Ca V 2.1 is bifurcated resulting in two mechanistically distinct forms of regulation: (i) local Ca 2+ fluctuations trigger rapid Ca 2+dependent upregulation of channel activity known as Ca 2+ -dependent facilitation (CDF), and (ii) kinetically slower CDI that evolves over ~300-800 ms and is sensitive to global Ca 2+ -elevations (43,76). The two modes of channel regulation rely on Ca 2+ /CaM interaction with distinct channel domains (48,61). CDF is triggered primarily by CaM C-lobe interaction with the canonical IQ domain (48), while CDI relies on Ca 2+ /CaM N-lobe interacting with binding sites elsewhere on the channel (76). To determine whether both modes of channel regulation are triggered by a single CaM, we again applied our strategy of localizing multiple CaM molecules to the Ca V 2 complex through linkage to the pore-forming α 1A and the β 2A subunits. For these experiments, the whole-cell dialysate contained low Ca 2+ buffering (1 mM EGTA) to permit global Ca 2+ elevations necessary to trigger CDI. A family of depolarizing voltage pulses of 800 ms duration were utilized to elicit Ca 2+ and Ba 2+ currents for CDI measurements. Thus probed, Figure 6A shows baseline CDI of Ca V 2.1 ( Figure S7A; Table S2). Tethering CaM WT to the CT of α 1A subunit (α 1A -CaM WT ) and coexpression with the β 2A subunit yields CDI comparable to baseline conditions ( Figure 6B; Figure S7B; Table S2). By contrast, fusion of CaM 1234 to the CT of α 1A subunit results in a reduction in CDI ( Figure 6C; Figure S7C; Table  S2). Subsequently, we tested whether CaM fusion to the β 2A subunit also evoked similar regulatory effects. Accordingly, co-expression of α 1A subunit with β 2A -CaM WT showed robust CDI comparable to baseline conditions ( Figure 6D; Figure S7D; Table  S2), while β 2A -CaM 1234 strongly diminished CDI ( Figure 6E; Figure S7E; Table S2). Therefore, fusion of CaM to either the α 1A or the β 2A subunit permits interaction with key effector interfaces and preserve functional modulation.
Thus informed, we considered changes in channel regulatory behavior in the presence of multiple CaM molecules. Accordingly, we coexpressed α 1A -CaM WT with β 2A -CaM WT . As expected, this maneuver elicited strong CDI ( Figure 7A; Figure S7F; Table S2). To determine stoichiometric requirements, we measured CDI of α 1A -CaM 1234 in the presence of β 2A -CaM WT . Comparison of Ca 2+ versus Ba 2+ currents reveal markedly blunted CDI ( Figure 7B; Figure S7G; Table S2). In comparison, co-expression of α 1A -CaM WT with β 2A -CaM 1234 revealed strong CDI ( Figure 7C; Figure S7H; Table S2) similar to untagged channels. As a control, we co-expressed of α 1A -CaM 1234 with β 2A -CaM 1234 and found near complete inhibition of CDI ( Figure 7D; Figure S7I; Table S2). Once again, to further corroborate these findings, we considered whether overexpression of freely diffusible recombinant CaM WT could reverse CDI deficits of α 1A -CaM 1234 in the presence β 2A ( Figure S8A-B; Figure S8D; Table S2). Indeed, CDI remained blunted for α 1A -CaM 1234 despite CaM overexpression. Similarly, overexpression of CaM WT failed to reverse the reduction in CDI of α 1A in the presence of β 2A -CaM 1234 (Figure S8E-F; Figure 8H; Table S2). These findings suggest that a single CaM bound to the Ca V 2.1 CT is primarily responsible for signaling CDI.
To determine whether CaM localized to the α 1A CT is also responsible for eliciting CDF, we first established baseline CDF of Ca V 2.1 using a paired-pulse facilitation protocol ( Figure 8A; Table S3). Briefly, in the absence of a pre-pulse, Ca V 2.1 current display biphasic kinetics corresponding to rapid activation of the channel, and a subsequent slower interconversion into a facilitated gating configuration following on Ca 2+ binding to CaM. With pre-pulse, Ca V 2.1 current is monophasic with enhanced activation as Ca 2+entry during the pre-pulse having already triggered facilitation. RF is quantified as the excess charge entry following pre-pulse and CDF is quantified as the difference in RF with Ca 2+ versus Ba 2+ as charge carriers. Once again, we validated that CaM fusion to the α 1A and β 2A supports CDF. Briefly, coexpression of α 1A -CaM WT with the β 2A subunit elicits strong CDF, while α 1A -CaM 1234 exhibits strongly diminished CDF (Figure 8B-C; Table S3). In like manner, expressing β 2A -CaM WT with α 1A subunit supports strong CDF, while β 2A -CaM 1234 diminishes CDF (Figure 8D-E; Table S3) thus confirming functionality of tethered CaM. Furthermore, robust CDF was also observed when both α 1A and β 2A subunits were both fused with CaM WT ( Figure 9A; Table S3), albeit the magnitude of CDF was modestly reduced in comparison to wild-type channels. To determine whether CDF of Ca V 2.1 is also dependent on carboxy-terminally linked CaM, we probed CDF of α 1A -CaM 1234 in the presence of β 2A -CaM WT . CDF was nearly absent for this pair ( Figure 9B; Table  S3). By contrast, α 1A -CaM WT in the presence of β 2A -CaM 1234 revealed no appreciable change in CDF ( Figure 9C; Table S3). Taken together, these findings suggest that CaM linked to the α 1A CT is privileged in triggering CDF. Of note, coexpression of α 1A -CaM 1234 with β 2A -CaM 1234 exhibited minimal CDF ( Figure 9D; Table S3). As a further test, we probed whether freely diffusible recombinant CaM WT could reverse reduced CDF observed for α 1A -CaM 1234 in the presence β 2A or for α 1A in the presence of β 2A -CaM 1234 . Indeed, CDF was strongly diminished in both cases ( Figure S8C; Figure S8G). Thus, localized CaM is privileged in initiating CDF. Taken together, these results also indicate that the same CaM molecule pre-bound to the channel CT is responsible for mediating both CDI and CDF.

Discussion
The stoichiometry of CaM interaction with the Ca V channel complex and the functional requirements for channel regulation has been long-debated (3,4). Biochemical and structural studies demonstrate the interaction of multiple CaM with distinct channel peptide segments (38,46,48,52,55,57,60,65). FRET analysis suggests a stoichiometry of up to two Ca 2+ /CaM molecules associating with the holo-Ca V channel complex (66). Functional studies, however, indicate that a single CaM suffices for Ca 2+ -dependent feedback regulation (59). To reconcile these differences, we dissected the potential role of multiple CaM in orchestrating Ca V feedback modulation. We localized up to two wild-type CaM or mutant CaM 1234 to the Ca V 1-2 channel complex through linkage to the α and β subunits. Consistent with prior studies, we found that a single CaM tethered to the Ca V complex through either the α or β subunits is fully capable of replacing endogenous CaM (59,67,74). Nonetheless, when multiple CaM are localized to the Ca V 1-2 channel complex, functional Ca 2+ -regulation of channel gating depends primarily on CaM tethered to the CT of the α-subunit. More specifically, when CaM WT is attached to the Ca V 1/2 CT locus, both CDI (for Ca V 1.2/1.3/2.1) and CDF (for Ca V 2.1) are fully intact, however when CaM 1234 is attached at this locus, Ca 2+ -regulation is absent. Furthermore, we found that introduction of a second IQ domain in the channel carboxy-tail switches the functional CaM stoichiometry for Ca V 1.3 channels such that channel regulation is responsive to two CaM molecules. These results are consistent with a model whereby a single CaM pre-associated to the channel CT serves as a dedicated sensor for Ca 2+ -dependent modulation of Ca V 1/2 gating.
A few mechanistic implications merit further attention. First, in vitro measurements of CaM affinity have demonstrated that apoCaM interaction to the channel CT peptides is much weaker (~1 μM) in comparison to Ca 2+ /CaM interaction (<1 nM) (51,77). Thus, if both a mutant CaM incapable of binding Ca 2+ and a wild-type CaM are within the same channel complex, one would expect the wild-type CaM to competitively displace the mutant CaM on the CT, owing to the three-orders of magnitude affinity advantage. However, we found that this was not the case for Ca V 1/2 channels; β-subunit tethered CaM WT was unable to displace CaM 1234 to trigger CDI. One possibility is that the apoCaM affinity for the Ca V channel complex may be stronger than estimated in vitro presumably reflecting unconventional interactions with the channel complex, as has been observed in cryo-EM structures of holo K V 7 channels (17) and Ryanodine Receptors (24). Indeed, previous studies have shown that reducing free apoCaM levels to nanomolar concentrations was insufficient to appreciably deplete apoCaM pre-association from the Ca V 1.3 channel suggesting a higher apoCaM affinity (78). An alternative possibility is that additional channel regulatory proteins such as a-actinin may fine-tune CaM interactions with the Ca V channel, thereby imparting distinct effects on channel gating (79). Second, our findings also point towards potential mechanisms that underlie the singular CaM stoichiometry observed for Ca 2+ -regulation of Ca V gating. For Ca V 1.3 channels, we previously found that Ca 2+ -binding to CaM elicits a conformational rearrangement of the channel CT, resulting in the formation of a tripartite complex involving the channel IQ domain, Ca 2+ /CaM, and the channel dual vestigial EF hand segments (41). Thus, one possibility is that functional stoichiometry for CaM regulation of channel gating may be ultimately limited by the number of IQ domains available to initiate formation of the tripartite complex. Consistent with this possibility, when Ca V 1.3 channels contained 2 IQ domains, functional Ca 2+regulation appeared to depend on both the channel CT tethered CaM and the β-subunit tethered CaM. Furthermore, we found that replacement of the Ca V 1.3 IQ domain with either M13 peptide or an IQ domain from the unconventional myosin Va resulted in a near complete inhibition of CDI. By contrast, substitution of the Na V 1.4 IQ domain still permits functional Ca 2+ -regulation. Importantly, Na V 1.4 channels are homologous to Ca V channels and they undergo CDI in a similar manner as Ca V 1 channels. In this scenario, it is possible that specific residues unique to Ca V /Na V IQ domain but not in myosin Va IQ may be critical in triggering tripartite complex formation. It is also possible that the precise orientation or arrangement of CaM may also be relevant in this process (48). Furthermore, although M13 and myosin Va IQ domain are widely recognized as CaM binding peptides, it is possible that attached to the Ca V channel carboxytail may perturb the ability of these peptides to interact with CaM. Third, the traditional model of Ca 2+ -dependent regulation is that Ca 2+ /CaM interaction with effector sites are sufficient to signal to the pore domain. For most Ca V channels, the carboxy-tail IQ domain is thought to harbor key effector sites for triggering Ca 2+ /CaM regulation (4), however, for Ca V 1.2 and Ca V 1.3, one critical interface for N-lobe mediated CDI is the NSCaTE motif located on the channel amino-terminus (55  80). These findings support the possibility that both modes of channel regulation are mediated by the same CaM that is initially preassociated to the channel CT. Our findings also bear important biological implications. In cardiac myocytes, a vast majority of Ca 2+ -free CaM is enriched in the dyad with a large fraction bound to the RyR (64,81). Following Ca 2+ -binding, however, CaM is mobilized and is available to interact with targets including Ca 2+ /CaM-dependent kinases and phosphatases that are also localized at the dyad (81). Our findings suggest that Ca 2+ -mobilized CaM would be unable to inhibit the L-type Ca 2+channels. Instead, only CaM initially preassociated with the channel CT would be able to trigger Ca 2+ -regulation. Physiologically, this scheme is advantageous in cardiomyocytes as additional CaMs in the dyad are free to signal to other regulatory processes, including activation of kinases and phosphatases (31,32,82,83), channel coupling (84), or translocation to nucleus (85), all without disrupting Ca V channel inactivation, an essential factor for normal cardiac repolarization (5). In the disease context, this arrangement however makes L-type channels particularly vulnerable to misregulation in cardiac arrhythmias associated with calmodulinopathies (10). The singular functional stoichiometry implies that the pre-association of even a small fraction of mutant CaM with weakened Ca 2+ -binding could appreciably disrupt L-type channel inactivation and increase risk of arrhythmogenesis. In like manner, in neurons, CaM localized to Ca 2+ channels serve multiple functions (44) including modulation of channel gating, trafficking (30,86), as well as a key role in excitation-transcription coupling, where local Ca 2+ signaling near L-type channels result in rapid shuttling of Ca 2+ /CaM to the nucleus through γCaMKII (34,87,88). Having a resident CaM dedicated for Ca V channel feedback modulation ensures that local Ca 2+ /CaM signaling can be multiplexed without detrimental effects on cellular electrical excitability.

Cell culture and transfection of HEK293 cells
For whole-cell electrophysiology, HEK293 cells were cultured on glass coverslips in 60-mm dishes and transfected using a calcium phosphate method (28). We applied 2-4 μg of cDNA encoding the desired channel α 1 subunit (wild type or CaMlinked variant), along with 4 μg of rat brain β 2A (wildtype or CaM-linked) and 4 μg of rat brain α 2 δ subunits (NM012919.2) (92). To enhance expression, cDNA for simian virus 40 T antigen (1 μg) was co-transfected. Electrophysiology recordings were done at room temperature 1-2 days after transfection.

Whole-cell electrophysiology recordings
Whole-cell voltage-clamp recordings for HEK293 were collected at room temperature using an Axopatch 200A amplifier (Axon Instruments). Glass pipettes (World Precision Instruments, MTW 150-F4) were pulled with a horizontal puller (P-97; Sutter Instruments Company) and fire polished (Microforge, Narishige, Tokyo, Japan) resulting in 1-3 MΩ resistances, before series resistance compensation of 70%. For Ca V 1. For CDI measurements, we used a family of test pulses from −50 mV to +50 mV with repetition intervals of 20 s, at a holding potential of −80 mV. Custom MATLAB (Mathworks) software was used to determine peak current and fraction of peak current remaining after either 300 ms (r 300 ) of depolarization for Ca V 1 or 800 ms (r 800 ) of depolarization for Ca V 2. Ca 2+ -dependent facilitation was quantified using the normalized charge difference ΔQ, obtained by integrating the difference between normalized traces ± prepulse as previously reported (43). The fraction of channels facilitated by prepulse is directly proportional to ΔQ divided by the slow time constant (t) of facilitation, yielding relative facilitation (RF = ΔQ/t). For knockouts of Ca 2+ -dependent facilitation, t was set to 12 ms (matching the average time constant for facilitation in channels lacking tethered CaM). RF Ca corresponds to relative facilitation with Ca 2+ as charge carrier while RF Ba corresponds to that obtained with Ba 2+ as charge carrier representing voltage-dependent facilitation. CDF is measured as the difference RF Ca -RF Ba (43).

Data Availability
All original data will be fully available upon request from: Manu Ben-Johny, Ph.D. Department of Physiology and Cellular Biophysics Columbia University mbj2124@cumc.columbia.edu

Acknowledgements
We thank members of the Ben-Johny Lab for insightful comments and Dr. Jacqueline Niu for helpful comments on experimental design and for assistance with data analysis. We thank Dr. Shin Rong Lee for construction of Ca V 2.1 fused to CaM. We would like to thank the late Dr. David Yue who was instrumental in the initial conceptualization and experiment design for this project. Dr. Yue passed away on December 23, 2014. He was an extraordinary mentor and teacher, and his unbridled passion for science continues to inspire us.  Schematic illustrates strategy for dissecting the effect of multiple CaM in tuning channel regulation. Either CaM WT or mutant CaM 1234 are localized to the Ca V 1.2 complex by genetic fusion to either α-or β-subunits using poly-glycine linker. A 2:1 stoichiometry can be attained by co-expressing both α-or β-subunits with CaM tethered. (B) Left, cartoon illustrates wild-type Ca V 1.2 L-type channels presumably bound to endogenous CaM with its preferred stoichiometry. Middle, exemplar current traces evoked in response to +10 mV voltage-step shows enhanced decay of Ca 2+ (red) versus Ba 2+ currents (black) confirming robust CDI. Throughout, Ba 2+ traces for CDI are scaled to ∼1/3 actual magnitude to match peak Ca 2+ traces (at scale with bar). Right, population data shows extent of baseline CDI. r 300 values report the fraction of peak current remaining following 300 ms depolarization. Black dots and error bar represent mean ±S.       Top right, a two-pulse protocol is used to quantify the extent of CDF. Ba 2+ current kinetics is similar in the presence (black) or absence (gray) of a depolarizing pre-pulse. Bottom right, without pre-pulse, a 0mV stepdepolarization elicits a biphasic inward Ca 2+ current with an initial rapid phase followed by slow phase corresponding to Ca 2+ -dependent facilitation. With a +20mV pre-pulse, the channels are already facilitated and as such the ensuing test-pulse elicits currents that exhibit enhanced channel activation. The area between the two current traces (DQ) approximates CDF triggered by the prepulse. Bottom left, population data (mean ± SEM) shows RF (Relative Facilitation) at different pre-pulse potentials averaged form n cells and assessed as DQ divided by the time constant (t) of facilitation. CDF is determined as the difference in RF with Ba 2+ versus Ca 2+ as charge carrier.