Bilobal architecture is a requirement for calmodulin signaling to Ca_V1.3 channels

Significance

Calmodulin (CaM) regulation of voltage-gated calcium (Ca_V) channels constitutes a prototypic biological feedback mechanism that contributes prominently toward Ca²⁺ homeostasis in neurons and cardiac myocytes. Here, by partitioning CaM molecularly into its two elemental domains or lobes, we uncover a distinctive nonlinearity in CaM signaling to Ca_V channels. Ca_V channels detect the coincident binding of two CaM lobes to up-regulate channel activity. This mechanism elaborates a molecular logic operation that enables channels to detect combinations of spatiotemporal Ca²⁺ signals and perform higher-order computations on Ca²⁺ signals. These findings uncover the unified mechanistic basis for Ca_V channel feedback and, in so doing, shed light on the versatility of CaM in decoding cellular Ca²⁺ signals.

Abstract

Calmodulin (CaM) regulation of voltage-gated calcium (Ca_V) channels is a powerful Ca²⁺ feedback mechanism that adjusts Ca²⁺ influx, affording rich mechanistic insights into Ca²⁺ decoding. CaM possesses a dual-lobed architecture, a salient feature of the myriad Ca²⁺-sensing proteins, where two homologous lobes that recognize similar targets hint at redundant signaling mechanisms. Here, by tethering CaM lobes, we demonstrate that bilobal architecture is obligatory for signaling to Ca_V channels. With one lobe bound, Ca_V carboxy tail rearranges itself, resulting in a preinhibited configuration precluded from Ca²⁺ feedback. Reconstitution of two lobes, even as separate molecules, relieves preinhibition and restores Ca²⁺ feedback. Ca_V channels thus detect the coincident binding of two Ca²⁺-free lobes to promote channel opening, a molecular implementation of a logical NOR operation that processes spatiotemporal Ca²⁺ signals bifurcated by CaM lobes. Overall, a unified scheme of Ca_V channel regulation by CaM now emerges, and our findings highlight the versatility of CaM to perform exquisite Ca²⁺ computations.

Calmodulin (CaM) plays a central role in eukaryotic signaling by decoding cytosolic Ca²⁺ fluctuations to coordinate diverse targets ranging from enzymes to ion channels (1, 2). Structurally, CaM possesses a distinctive bilobal architecture: Its two globular domains, containing paired “EF hand” Ca²⁺-binding motifs, exhibit high sequence (Fig. 1A) and structural homology (Fig. 1 B and C) (3). The two lobes often recognize similar targets and evoke like functional outcomes (3). This apparent redundancy suggests that the bilobal arrangement may be dispensable for CaM function. Indeed, many CaM targets such as CaM kinases, myosin light-chain kinase, and nitric oxide synthase are activated by CaM fragments in vitro (4, 5). However, the two-lobed architecture is a hallmark for many Ca²⁺-binding proteins, suggesting it may support latent functions (6). To address this conundrum, we dissect CaM signaling to voltage-gated Ca²⁺ channels (Ca_V) (7), a prototypic molecular feedback essential for cardiac electrical stability (8, 9), neuronal activity (10, 11), and coupling to diverse cellular processes (12–14). Functionally, CaM exerts two distinct effects on Ca_V channels (15). First, the binding of Ca²⁺-free CaM (apoCaM) to the Ca²⁺-inactivating (CI) module in the channel carboxy terminus (CT) (16) up-regulates baseline open probability (P_O) (15) (Fig. 1 D and E). Second, the binding of Ca²⁺ to CaM reverses this initial enhancement in P_O, a process termed Ca²⁺-dependent inactivation (CDI) (15, 17–20) (Fig. 1 D and E). However, how individual CaM lobes orchestrate dual regulation and whether the bilobal architecture is a requirement remains critically unknown. As this modulation is conserved across Ca_V and voltage-gated sodium (Na_V) channels (21), resolving these ambiguities may inform on unified regulatory mechanisms. Moreover, as mutations in conserved Ca_V and Na_V CaM-binding interfaces are linked with heritable cardiac arrhythmias (22, 23), autism spectral disorder (24), and congenital stationary night blindness (25), dissecting these mechanisms promises to lend insights into disease pathogenesis and therapeutics.

Fig. 1.

Bilobal architecture of CaM is necessary for Ca_V1.3 channel regulation. (A) CaM contains two globular domains each composed of two EF hand Ca²⁺-binding motifs. The two lobes of CaM share high sequence similarity. (B) Ca²⁺-dependent conformational changes of CaM. (Left) In the absence of Ca²⁺, both N and C lobes of CaM adopt a highly similar conformation (PDB ID: 1CDF). (Right) The structural similarity of the CaM lobes persists even when Ca²⁺-bound (PDB ID: 1CLL). (C) Schematic summarizes CaM-dependent changes in Ca_V1.3 gating. (Left) Devoid of CaM, channels adopt a low P_O gating configuration. (Middle) The apoCaM binding switches channels to a high P_O gating mode. (Right) Ca²⁺ binding to CaM relieves the enhancement in P_O and switches channels to a low P_O mode. (D) Ca_V1.3 channels with tethered wild-type CaM (Ca_V1.3−CaM_WT, Top) exhibit robust Ca²⁺-dependent regulation. (Middle) Exemplar currents. Scale bar pertains to Ca²⁺ currents (red). Ba²⁺ currents (black) are scaled down ∼3× for comparison of decay kinetics. (Bottom) Population data depict fraction of peak current after 300-ms depolarization (r₃₀₀) versus voltage; red, relation for Ca²⁺; black, relation for Ba²⁺. Each point is mean ± SEM. (E) Genetic fusion of dominant negative mutant CaM₁₂₃₄ to Ca_V1.3 CT (Ca_V1.3−CaM₁₂₃₄) abrogates Ca²⁺ regulation. (F) Ca_V1.3 fused to CaM_C fails to support Ca²⁺ regulation, suggesting that bilobal architecture of CaM is necessary for functional modulation of Ca_V channels. (G) Genetic fusion of Ca_V1.3 with mutant CaM₁₂ whose Ca²⁺ binding is restricted to C lobe alone also supports Ca²⁺ regulation, suggesting C lobe can trigger channel modulation, provided N lobe is also present. (H) Genetic fusion of Ca_V1.3 with mutant CaM₃₄ whose Ca²⁺ binding is restricted to N lobe alone also supports Ca²⁺ regulation. Format for E–H is as in D.

We, here, demonstrate that the bilobal architecture of CaM is a prerequisite for signaling to L-type channels. Channels bound to a single lobe of CaM adopt a preinhibited configuration with diminished P_O and are prohibited from Ca²⁺ regulation. The reconstitution of two CaM lobes relieves preinhibition and rescues Ca²⁺ regulation. Consistent with a baseline change in channel gating, small-angle X-ray scattering (SAXS) analysis and molecular dynamics (MD) simulations suggest a structural rearrangement of the CI when one versus two CaM lobes are bound. These findings revise the current view of CaM regulation of Ca_V channels and bear implications for CaM signaling to its diverse targets.

Results

Two Lobes of CaM Are Required for Ca_V Channel Modulation.

We sought to evaluate whether a single lobe of CaM by itself could trigger Ca_V regulation. However, as full-length CaM is ubiquitous in eukaryotic cells, probing the effects of individual lobes is challenging. Exogenously expressed CaM lobes may not fully populate the channels and the observed effects may be corrupted by fluctuations in ambient CaM. Similarly, genetic replacement of endogenous CaM with lobe peptides lacks target specificity and impacts cell viability (26). To overcome these limitations, we utilize a two-prong approach: (i) Full-length or individual CaM lobes are localized to the channel complex via genetic fusion using a flexible linker to the CT (27). (ii) Endogenous free CaM levels are diminished by overexpressing a CaM sponge derived from unconventional Myosin Va (28, 29). As a single apoCaM binds (29) and regulates Ca_V channels (15, 27), this fusion strategy satisfies stoichiometric requirements for channel modulation (Fig. S1A). To validate this approach, we fused full-length CaM to the Ca_V1.3 CT (Fig. 1D) and coexpressed CaM sponge. These channels exhibited robust CDI (Fig. 1D), as evident from the enhanced decay kinetics with Ca²⁺ (red) versus Ba²⁺ (black) as charge carrier and population data showing fraction of peak current remaining following 300-ms depolarization (r₃₀₀). However, when channels are tethered to mutant CaM₁₂₃₄, with all four Ca²⁺-binding sites disabled, modulation was abolished (P = 1E-6) (Fig. 1E), further corroborating our strategy (30).

To discern the effect of single CaM lobes, we assessed their ability to bind the CI element using FRET two-hybrid assays (Fig. S1 B and C). Although both N and C lobes of apoCaM bind to the CI, the C lobe possessed an ∼20-fold higher affinity, with the isoleucine–glutamine (IQ) domain as its primary interface (Fig. S1). The overall scheme for apoCaM preassociation to the Ca_V1.3 (Fig. S1) parallels that for Na_V channels (31). Given its high affinity, we probed whether CaM C lobe by itself can support Ca²⁺ regulation by fusing CaM_C to the Ca_V1.3 CT (Ca_V1.3−CaM_C, Fig. 1F). Surprisingly, this maneuver completely disrupted CDI (P = 5.3E-8) (Fig. 1F). Moreover, Ca_V1.3−CaM_C also exhibited sharply diminished CDI even in the absence of CaM sponge (Fig. S2A; P = 7.5E-6), suggesting the CaM C lobe by itself is incapable of supporting CDI. By contrast, fusion of Ca_V1.3 with CaM₁₂, a bilobed variant with intact C lobe and Ca²⁺-insensitive N lobe, sustains CDI (P = 0.02 vs. Ca_V1.3−CaM_C) (Fig. 1G). Likewise, tethering CaM₃₄ to Ca_V1.3, with a functional N lobe and a Ca²⁺-insensitive C lobe, also supports CDI (Ca_V1.3−CaM₃₄, Fig. 1H). Thus, while CaM C lobe binds to Ca_V1.3, the N lobe is necessary for channel modulation even if Ca²⁺-binding to this lobe is defunct.

Thus informed, we sought to determine whether the CaM N lobe alone can support CDI. As the canonical Ca_V1.3 IQ domain has a high affinity for CaM C lobe, coexpression of CaM sponge by itself is insufficient to preclude binding of endogenous CaM (32). Consequently, we assessed CDI of an RNA-edited channel, Ca_V1.3_MQDY, with a single amino acid substitution in its IQ domain that weakens CaM binding (33). Furthermore, to attain high local concentrations, we tethered CaM variants to the β_2A-subunit (30, 33). In the presence of β_2A−CaM_WT, Ca_V1.3_MQDY exhibited strong CDI establishing baseline CDI for this channel variant (Fig. S2B). By contrast, coexpression of β_2A−CaM_N, composed of a single N lobe alone, elicited a marked reduction in CDI (P = 7E-8) (Fig. S2C). Even so, Ca_V1.3_MQDY exhibits robust CDI in the presence of β_2A−CaM₃₄, a bilobed CaM variant with an intact N lobe and a Ca²⁺-insensitive C lobe (Fig. S2D) (P = 1E-4 compared with Ca_V1.3_MQDY/β_2A−CaM_N). Thus, a single N lobe alone in the channel complex is also insufficient to trigger CDI; rather, the presence of the C lobe is necessary to elicit channel modulation even if Ca²⁺-binding to this lobe is defunct. Together, these results illustrate that the dual-lobe architecture of CaM is a core requirement for Ca_V regulation.

Channels Bound to a Single CaM Lobe Are Preinhibited.

The striking difference in Ca²⁺ responsiveness of Ca_V channels in the presence of one versus two lobes of CaM may arise from two distinct possibilities.

First, when CaM C lobe alone is preassociated to the channel, it may either fail to bind Ca²⁺ and/or trigger downstream conformational changes necessary for channel modulation. Biochemically, individual CaM lobes in solution can bind Ca²⁺ ions with a similar affinity, and the resultant conformational changes are nearly identical to that of intact CaM (34, 35). To explicitly test this possibility, we use live-cell FRET two-hybrid assay to compare the binding of YFP-tagged Ca_V1.3 CI to CFP−CaM_C in the absence and the presence of Ca²⁺ (Fig. S1 C and D). Elevation of cytosolic Ca²⁺ resulted in an increase in the maximal FRET efficiency (E_A,max), arguing that Ca²⁺/CaM_C binds to Ca_V1.3 CI and elicits a conformational change within the complex.

Second, channels bound to a single CaM lobe may undergo a baseline change in gating even before Ca²⁺ influx, yielding a “nonpermissive” configuration incapable of CDI. Intriguingly, previous studies have shown that apoCaM binding itself tunes Ca_V gating, with the loss of preassociated CaM resulting in a throttling of channel openings (15). Could the binding of CaM C lobe alone also alter the channel P_O as expected with a change in gating? Accordingly, we undertook low-noise single-channel electrophysiology to compare the P_O of Ca_V1.3 fused to CaM C lobe (Ca_V1.3−CaM_C) with those (i) fused to full-length CaM or (ii) devoid of CaM altogether. To isolate Ca²⁺-independent changes in channel gating, Ba²⁺ was chosen as the charge carrier, as it binds poorly to CaM (36). Moreover, as Ca_V1.3 exhibits minimal voltage-dependent inactivation, a slow voltage ramp was utilized to elicit stochastic channel openings that reflect near−steady-state P_O at each voltage. With this experimental framework, we undertook recordings of Ca_V1.3−CaM_WT with a CaM sponge overexpressed to obviate any confounding effects of endogenous CaM. Fig. 2A, Middle displays exemplary stochastic records, where channel closures correspond to the zero-current portions of the trace (horizontal gray lines) and openings correspond to downward deflections to the open level (slanted gray curves). Averaging many such records yields a mean current that can be divided into the open level (slanted gray curve) to obtain the steady-state P_O as a function of voltage (sigmoidal trace at the bottom). The maximal P_O of the CaM-fused channel was found to be P_O,max = 0.41 ± 0.05 (n = 8 patches and 850 records; mean ± SEM). To estimate the P_O of Ca_V devoid of CaM, we analyzed the single-molecule behavior of the RNA-edited Ca_V1.3_MQDY (11). Once again, we resorted to studying this variant, as the canonical Ca_V1.3 short variant has a high affinity for apoCaM such that strong overexpression of CaM chelator is insufficient to fully preclude CaM binding (32). By contrast, the CaM binding status of Ca_V1.3_MQDY can be readily tuned by the same maneuver (33). Under low levels of CaM, we found the steady-state maximal P_O of Ca_V1.3_MQDY to be drastically diminished, with P_O,max = 0.08 ± 0.01 (mean ± SEM; n = 4 patches with 539 records) (Fig. 2B). With these two limits firmly established, we probed the function of Ca_V1.3−CaM_C. Remarkably, these channels also open sparsely with their baseline P_O substantially diminished (Fig. 2C), yielding P_O,max = 0.06 ± 0.02 (mean ± SEM; n = 5 patches with 492 records). Thus, channels bound to a single CaM lobe undergo a baseline change in gating even before Ca²⁺ entry, akin to channels that lack prebound CaM altogether.

Fig. 2.

Channels bound to a single CaM lobe adopt a preinhibited configuration. For experiments in A–D, CaM sponge was coexpressed. (A) Ca_V1.3 fused to CaM_WT (Ca_V1.3−CaM_WT) exhibits high baseline P_O, with Ba²⁺ as charge carrier, consistent with channels in gating configuration A (Fig. 1D). (Top) Exemplary current records show response to a voltage-ramp protocol. (Bottom) P_O–voltage (V) relationship determined from the ensemble average of single-channel recordings. (B) Ca_V1.3 channels lacking prebound CaM exhibit diminished baseline P_O. Here, we used the Ca_V1.3 MQDY variant with a low CaM binding affinity. (C) When Ca_V1.3 is fused to CaM_C (Ca_V1.3−CaM_C) alone, the baseline P_O is substantially diminished, consistent with channels adopting a preinhibited configuration. (D) Reconstitution of N lobe of CaM via attachment to the auxiliary Ca_Vβ_2A subunit (β_2A−CaM_N) potently up-regulates the maximal P_O of Ca_V1.3−CaM_C channels (red curve). Format for B−D is as in A.

Assured of the functional necessity for the bilobal architecture of CaM, we considered whether the binding of the two lobes is sufficient to restore channel modulation. More specifically, could the reconstitution of the N lobe of CaM as a distinct molecular entity enhance the P_O of Ca_V1.3−CaM_C? Practically, simple overexpression of the CaM N lobe is insufficient for reliable reconstitution given its weak affinity for the Ca_V channel CI module (Fig. S1B). Consequently, for effective delivery of CaM N lobe to the channel complex, we utilized an alternate strategy whereby this lobe is fused to the auxiliary β_2A subunit (β_2A−CaM_N), an obligatory component for functional Ca_V channel complexes (37, 38). Remarkably, single-channel recordings of Ca_V1.3−CaM_C in the presence of β_2A−CaM_N revealed enhanced openings, and the ensemble average showed a partial rescue of the baseline gating (Fig. 2D) with P_O,max = 0.24 ± 0.03 (mean ± SEM; n = 4 patches with 538 records). These findings suggest that the restoration of the complementary lobe may reverse the change in channel gating associated with the binding of a single lobe of CaM.

Binding of CaM Lobes Switches P_O in Quantized Manner.

That said, the observed changes in channel P_O may result from multiple indirect mechanisms that are unrelated to CaM lobe interactions. However, if changes in channel P_O, in fact, result from CaM lobes engaging distinct regulatory interfaces, then they are expected to be quantized in agreement with transitions between unique channel conformations (39). Indeed, changes in channel gating associated with apoCaM binding follow this paradigm (15, 39, 40). More specifically, channels devoid of CaM adopt gating mode E marked by low P_O, while apoCaM binding switches channels into gating mode A with a high P_O as schematized in Fig. S3A (15). With CaM C lobe bound, we hypothesize that channels may be restricted to mode E, while restoration of the N lobe may reenable sojourns into mode A.

To test this hypothesis, we sought to discern the core features of gating modes A and E. Fusion of CaM_WT would ensnare Ca_V1.3 into mode A (Fig. S3B). Fig. 3A displays 10 sequential trials of the Ca_V1.3−CaM_WT single-channel activity evoked by voltage ramps introduced at 12-s intervals. The activity of the channel appears uniformly high, as confirmed by the diary plot of average P_O within individual trials ($\overline{P}$_O, Fig. 3A) and a unimodal $\overline{P}$_O distribution obtained from a larger set of trials (Fig. 3B). The steady-state P_O−V relationship shown in Fig. 3C illustrates the high maximal open probability (P_O,max) characteristic of mode A gating. We further assessed the distribution of open durations (T_O) in our single-channel trials during epochs where the membrane potential ranged between −30 mV and +20 mV. The open durations largely followed a single-exponential decay consistent with a single open state in gating mode A (O_A) with an exit rate, k_OC|A ≈ 1.7 ms⁻¹ (Fig. 3D and Fig. S3B) (40). By contrast, for Ca_V1.3_MQDY deprived of CaM, openings are uniformly sparse, with low single-trial activity ($\overline{P}$_O) (Fig. 3E). The overall $\overline{P}$_O distribution remains unimodal but is now restricted to low activity limits (Fig. 3F), distinct from Ca_V1.3−CaM_WT [P < 0.05, Kolmogorov−Smirnov (KS) test]. Thus, Ca_V1.3_MQDY adopts, almost exclusively, mode E with the steady-state P_O−V relationship exhibiting a drastically reduced P_O,max (Fig. 3G and Fig. S3C). The open durations also followed a single-exponential decay, suggesting a distinct open state associated with mode E (O_E) with an exit rate of k_OC|E ≈ 9.3 ms⁻¹ (Fig. 3H and Fig. S3C).

Fig. 3.

Binding of CaM lobes evoke discrete Ca_V1.3 gating modes. CaM sponge was present for A–L. (A) Sequential single-channel trials of Ca_V1.3−CaM_WT in response to a voltage ramp (187 records). Diary plot displays single-trial average P_O computed for −30 mV ≤ V ≤ +25 mV (${\overline{P}}_{\text{O}}$). Dashed line discriminates low (red area) and high P_O (gray area) traces. (B) Histogram shows number of sweeps with ${\overline{P}}_{\text{O}}$(−30 ≤ V ≤ 25) for given range. (C) Average P_O at each voltage calculated for high P_O traces estimates P_O−V relationship for mode A. (D) Cumulative open duration distribution [green bars; P(T_O > t)] follows a single-exponential decay (green fit) consistent with a single open state in mode A with an exit rate, k_OC|A = 3.3 ms⁻¹. (E–H) Ca_V1.3_S/MQDY with CaM sponge approximates behavior of CaM-less channels (117 records). Openings are brief and sparse. Single-trial ${\overline{P}}_{\text{O}}$ distribution is unimodal in low P_O range. Maximal P_O is reduced. Open-duration P(T_O > t) is single-exponential with k_OC|E = 9.3 ms⁻¹ > k_OC|A. Format is as in A–D. (I–L) Analysis of Ca_V1.3−CaM_C (141 records) shows similarity to CaM-less channels with brief and sparse openings and ${\overline{P}}_{\text{O}}$ histogram and P_O−V relation reminiscent of mode E. P(T_O > t) is single exponential with rate constant ≈ k_OC|E. (M–P) With β_2A−CaM_N, Ca_V1.3−CaM_C P_O is enhanced in a quantized manner (139 records). Channels switch between low and high activity epochs with bimodal ${\overline{P}}_{\text{O}}$ distribution. P(T_O > t) distribution is biexponential consistent with rate constants k_OC|E and k_OC|A.

With CaM-dependent gating modes quantified, we next undertook in-depth analysis of Ca_V1.3−CaM_C. Qualitatively, channel openings are sparse as with mode E, and the diary plot of $\overline{P}$_O further confirms their low activity (Fig. 3I). In fact, the $\overline{P}$_O distribution shows a single peak restricted to low P_O limits (Fig. 3J) like mode E but not A (P < 0.05, KS test). The steady-state P_O−V relationship also exhibits a diminished P_O,max (Fig. 3K) identical to mode E (15) (Fig. 3G), and analysis of open durations (T_O) revealed a monoexponential distribution with decay constant identical to k_OC|E deduced from mode E openings (Fig. 3L and Fig. S3D). These uncanny similarities suggest the functional equivalence of channels bound to a single lobe of CaM to those devoid of CaM altogether.

If Ca_V1.3 bound to CaM C lobe is restricted to mode E, then the binding of CaM N lobe may reenable sojourns into mode A. Remarkably, examination of single-channel trials showed that Ca_V1.3−CaM_C in the presence of β_2A−CaM_N appeared to switch between epochs of low and high activity, as confirmed from the diary plot of $\overline{P}$_O for individual trials from the same channel (Fig. 3M). The $\overline{P}$_O distribution obtained from all trials was bimodal, fitting with channels transitioning between mode E and mode A behaviors (Fig. 3N). In fact, categorizing trials into high and low P_O groups revealed conditional P_O−V relations with a fivefold difference in the asymptotic P_O,max values for the two categories (Fig. 3O). Given this mode-switching behavior, channel openings may represent sojourns to open states in modes E and A (i.e., either O_E or O_A). Fitting with this scheme, the open durations now follow a biexponential distribution well approximated as a weighted sum of exponentials with decay rates, k_OC|E and k_OC|A, observed for mode E and A openings, respectively (Fig. 3P and Fig. S3E).

Thus, while channels bound to the C lobe alone are restricted to mode E gating, the binding of the N lobe, even as a distinct molecular entity, is sufficient to switch channels into the high P_O gating mode A. The efficacy of mode switching observed here likely depends on the propensity of both N and C lobes of apoCaM to be bound to their respective molecular interfaces on the channel CT. Thus, the CI module serves as a coincidence detector that stabilizes channel openings.

Two Detached CaM Lobes Support Channel Regulation.

If coexpression of β_2A−CaM_N restores mode A gating for Ca_V1.3−CaM_C, then channels are no longer delimited to a nonpermissive configuration, suggesting that this maneuver may also restore CDI. Of note, while both N and C lobes of CaM are no longer part of a single molecular entity, both lobes are capable of binding Ca²⁺ ions (34, 35). Remarkably, both exemplary whole-cell recordings and population data illustrate the reemergence of CDI, evident as the enhanced decay kinetics of Ca²⁺ versus Ba²⁺ currents (Fig. 4A). The partial rescue of Ca²⁺ regulation observed here accords well with the partial restoration of mode A gating following reintroduction of the CaM N lobe (Fig. 3 M–P). Furthermore, the targeted delivery of CaM_N12 with Ca²⁺ binding to N lobe impaired also sufficed to rescue CDI to the Ca_V1.3−CaM_C channels (Fig. 4B). Similarly, the combination of β_2A−CaM_N and Ca_V1.3−CaM_C34 also resulted in low levels of CDI, revealing the N-lobe component of CDI for the reconfigured channel arrangement (Fig. 4C). Reassuringly, coexpression of β_2A−CaM_N12 with Ca_V1.3−CaM_C34 resulted in strong reduction of CDI (Fig. 4D) in comparison with all three conditions above (Ca_V1.3−CaM_C/β_2A−CaM_N, P = 2E-5; Ca_V1.3−CaM_C/β_2A−CaM_N12, P = 2E-3; Ca_V1.3−CaM_C34/β_2A−CaM_N, P = 5E-3). Thus, both N and C lobes of CaM are capable of decoding Ca²⁺ signals as disparate molecular entities; however, channel modulation ensues if and only if both lobes are present in the same complex.

Fig. 4.

Reconstitution of detached CaM hemilobes is sufficient to restore Ca_V regulation. (A) Coexpression of β_2A−CaM_N partially restores Ca²⁺ regulation to Ca_V1.3−CaM_C channels, suggesting that the presence of two lobes of CaM is sufficient to evoke channel Ca²⁺ feedback regulation. Format is as in Fig. 1A. (B) Reconstitution of CaM N lobe with its Ca²⁺ binding disabled (β_2A−CaM_N12) is also sufficient to partially rescue Ca²⁺ regulation of Ca_V1.3−CaM_C. Format is as in A. (C) Reconstitution of β_2A−CaM_N with Ca_V1.3−CaM_C34 results in small residual CDI. (D) Ca²⁺ regulation is absent following reconstitution of β_2A−CaM_N12 to Ca_V1.3−CaM_C34. (E) Exemplary current records show robust CDI of Ca_V1.3 mediated by CaM_CC, CaM variant with two identical C lobes. Population data confirms strong CDI of Ca_V1.3 when bound to CaM_CC. (F) Coexpression of CaM_CC1234, with Ca²⁺ binding disabled to its two C -lobes, strongly reduces CDI of Ca_V1.3.

A corollary to this principle is that, if two C lobes were tethered (CaM_CC), the bilobal requirement would be satisfied and the channels would be able to elicit channel regulation. Remarkably, coexpression of CaM_CC with Ca_V1.3 demonstrated robust functional Ca²⁺ modulation (Fig. 4E), unlike channels bound to CaM_C (Fig. 1F). In comparison, coexpression of CaM_CC1234 with all four EF hands Ca²⁺-insensitive results in a dramatic reduction in CDI, further corroborating the ability of CaM_CC to trigger CDI (P = 1E-4; Welch’s t test) (Fig. 4F). Thus, while a single C lobe alone is incapable of supporting channel regulation, two of the same lobes together elicit robust modulation. Thus, the bilobal architecture of CaM is both necessary and sufficient for signaling to Ca_V channels.

Ca_V CI Module Rearranges Itself if Only CaM C-Lobe Is Bound.

With the functional requirement for bilobal CaM established, we probed whether discrete changes in channel gating results from parallel conformational rearrangements of the CI module. As currently available atomic structures of Ca_V−CaM complex are limited to short segments (e.g., the IQ domain) and only in the presence of Ca²⁺ (41–45), we constructed a homology model of Ca_V1.3 CI bound to apoCaM based on known structures of Ca_V1.1 (46) and Na_V1.5 (31) (Fig. S4). The CI adopts an extended conformation with an apoCaM preassociation interface that aligns well with hotspots identified from systematic functional analysis (Fig. S4D) (17).

Given this overall agreement, we performed explicit-solvent MD simulations of the homology model to probe CaM-dependent structural rearrangements of Ca_V1.3 CI. Analysis of 100-ns trajectory showed equilibration of CaM_WT-bound CI with minor deviations from its initial “extended” conformation (Fig. 5A and Fig. S5 A and B). Devoid of CaM, however, the CI undergoes a dramatic rearrangement resulting in a marked increase in the Cα rmsd and the reorientation of the IQ toward the EF_1,2 segments, termed a “bent” conformation (Fig. 5B and Fig. S5C). Of note, the IQ and EF_1,2 subdomains themselves remain stable (Fig. S5D). Strikingly, this change in CI conformation with the loss of prebound CaM parallels functional changes in channel P_O, suggesting that the transitions between gating modes A and E may be related to changes in CI conformation. Fitting with this hypothesis, the CaM_C-bound CI again reorients into the bent conformation, with the IQ deflected toward the EF_1,2 (Fig. 5C and Fig. S5 E and F). In sharp contrast, when CaM_CC (i.e., with two identical C lobes) is bound, the CI maintains its extended conformation (Fig. 5D and Fig. S5G). These results suggest that the binding of two lobes of Ca²⁺-free CaM is necessary to stabilize the Ca_V CI in the extended conformation. The switching between the extended and bent conformations appears to correlate with channels transitioning between gating modes A and E, respectively.

Fig. 5.

Binding of CaM lobes evokes distinct conformational rearrangements for Ca_V1.3 CI. (A) MD simulation shows the relative stability of the Ca_V1.3 CI bound to CaM. (Top) C_α rmsd for Ca_V1.3 CI region for 100-ns trajectory. (Bottom) Structural model following 40-ns equilibration. The angle between the IQ domain and a helix within EF_1,2 is shown, to facilitate comparison of conformational changes. (B) MD simulation shows the dramatic conformational rearrangement of Ca_V1.3 CI module devoid of CaM. Format is as in A. (C) Ca_V1.3 CI module bound to CaM C lobe alone also undergoes a conformational rearrangement. Format is as in A. (D) Ca_V1.3 CI bound to CaM_CC undergoes minor structural reorientations. Format is as in A. (E) Small-angle X-ray solution scattering profile of the Ca_V1.3 CI module in complex with CaM_WT (black) or CaM_C (red). (Left) Scattering intensity is plotted as a function of momentum transfer. (Right) Radial pair-distribution function [P(r)] is computed for radial vectors (r) and describes the set of all paired distances within the structure. (F) (Left) Ab initio molecular envelope of Ca_V1.3 CI/CaM complex overlaid on a homology model (Fig. S4D). (Right) Ab initio molecular envelope of Ca_V1.3 CI/CaM_C complex overlaid on the MD-relaxed model. (G) Schematic shows a unifying model of CaM regulation of Ca_V1 channels. Devoid of CaM, the channel CI adopts a bent conformation, with the IQ reoriented toward the EF_1,2 domains, while openings are allosterically diminished. If only one apoCaM lobe is bound, the bent conformation of the CI persists, and channel openings remain diminished. Binding of two apoCaM lobes switches the CI into an extended conformation, and this rearrangement enhances channel openings. On Ca²⁺ binding, one or two lobes of CaM depart from its preassociation interface, and the CI module reverts to a bent conformation and channel openings are diminished.

To experimentally corroborate this conformational change, we purified recombinant Ca_V1.3 CI module in complex with CaM_WT and CaM_C alone (Fig. S6A), and utilized SAXS analysis to probe their overall molecular shapes under low Ca²⁺ conditions. Both CaM_WT- and CaM_C-bound complexes eluted as monodisperse entities in size exclusion chromatography with elution volumes consistent with a 1:1 stoichiometry (Fig. S6 B and C). Experimental solution scattering profiles and P(r) distributions revealed key differences between the CaM_WT- and the CaM_C-bound complexes (Fig. 5E). Furthermore, ab initio simulations (47) revealed distinct coarse-grain molecular envelopes for the CI in complex with CaM_WT and CaM_C (Fig. 5F). Interestingly, the molecular shape of the CI/CaM_WT complex fits well with the extended conformation as in our homology model before (Fig. 5F, Left and Fig. S6D) and after MD relaxation (Fig. S6 E and F). By contrast, the shape of CI/CaM_C complex overlaid closely with only the bent conformation with the IQ reoriented (Fig. 5F, Right and Fig. S6G) and not the extended conformation (Fig. S6H). Overall, these results substantiate a marked conformational change of the Ca_V CI depending on the coincident binding of two CaM lobes. Ultimately, these conformational changes may be allosterically coupled to the channel pore to diminish openings. Recent structures of Ca_V1.1 and Na_V1.4 show that the CI module is intimately associated with the III−IV linker (46, 48). Accordingly, one possibility is that rearrangement of the CI module between the extended and bent conformations may preferentially bias the orientation of the III−IV linker and the propensity of the inner S6 gates to twist open (46, 48). It is noteworthy that Ca_V1.1 have poor apoCaM binding (49), suggesting that the currently available structures of Ca_V1.1 may correspond to a preinhibited conformation (46). Alternatively, CI changes may alter the folding of distal S6 segments and change channel activation (50, 51). Resolving the mechanisms by which CaM/CI module conformations couple to the pore domain remains an important frontier.

Discussion

The dual-lobe architecture is a salient feature of many cytosolic Ca²⁺ signaling proteins, including CaM (3, 52, 53), Troponin C (54), various neuronal calcium sensors (55, 56), and members of the S100 family that function as dimers (57). Here, we find that the bilobal nature of CaM is a fundamental requirement for its ability to regulate Ca_V channels, a biologically vital feedback loop (7–9, 12), and a prototype for deducing general Ca²⁺-decoding principles (58, 59). Ca_V channels bound to a single CaM lobe alone entirely fail to be Ca²⁺-modulated, and instead adopt a preinhibited configuration like channels that lack CaM altogether. Strikingly, reconstitution of the complementary CaM lobe (N lobe) as a disparate molecular entity relieves channel preinhibition and restores Ca²⁺ feedback. These changes in function parallel a CaM-dependent conformational rearrangement of the CI module between extended and bent conformations that, when transduced to the channel pore domain, switches channels between permissive and preinhibited modes. These findings revise long-held mechanistic views of Ca_V regulation.

Traditional models of Ca_V regulation have centered on the Ca²⁺-dependent association of CaM with multiple effector domains embedded across various cytosolic loops (7, 17, 60–67). Although disparate, these binding events are all thought to communicate to the transmembrane pore domain to throttle channel openings. How do Ca²⁺/CaM interactions with unique molecular interfaces all support the same end-stage outcome? Our findings here point to a unifying release model for Ca²⁺ regulation of Ca_V channels whereby the dislodging of a lobe of CaM from its preassociation interface is sufficient to trigger channel regulation (Fig. 5G) as follows: (i) Devoid of CaM, the Ca_V CI module adopts a bent conformation, while channels exhibit a low P_O mode E gating behavior characterized by sparse and brief openings (15). (ii) Following the binding of a single apoCaM lobe, the CI module continues to be destabilized in the bent conformation, and channels continue to reside in mode E. (iii) However, the concurrent binding of both N and C lobes of apoCaM stabilizes the CI module in the extended conformation and channels switch into mode A marked by more frequent and longer channel openings. (iv) Following Ca²⁺ influx, one or two lobes of CaM may dislodge en route to its binding interface, thus destabilizing the CI module into the bent configuration and reverting channels into mode E. For example, for Ca_V1.3 and Ca_V1.2, the N-terminal spatial Ca²⁺ transforming element (NSCaTE) loci on the channel amino terminus serves as an effector domain for N-lobe CDI (58, 67). Mechanistically, N-lobe CDI results from a multistep process where Ca²⁺-free N lobe prebinds the CI module. Following transient dissociation and binding of Ca²⁺, the N lobe interacts with the NSCaTE interface to trigger CDI (58, 67). Subsequently, this process may couple to the pore domain via formation or disruption of bridge between the channel amino termini and carboxy termini mediated by CaM (68) or via direct interaction (69). Our present findings suggest that the disengagement of apo N lobe from the CI interface and subsequent conformational rearrangement of the CI module triggers CDI. More broadly, Ca²⁺/CaM interaction with any of the multiple Ca²⁺/CaM effector interfaces encoded within the channel cytosolic domains would evoke channel regulation.

More broadly, our results highlight the versatility of CaM to process Ca²⁺ signals and thereupon perform complex molecular computations. Canonical CaM signaling entails the interaction of Ca²⁺/CaM lobes with a target, such that individual lobes are often sufficient to activate the target, while the two lobes together merely enhances the overall affinity and Ca²⁺ sensitivity (70). Prominent examples include Ca²⁺/CaM-dependent kinases, myosin light chain kinase, and nitric oxide synthase (4, 5). For Ca_V channels, previous studies have argued for the canonical mode of CaM function based on the finding that channel rundown can be slowed via reconstitution of individual CaM lobes (71). In sharp contrast, our results suggest that CaM implements a novel molecular operation—a logical NOR gate that decodes Ca²⁺ signals—with high levels of Ca_V1.3 activity occurring only on the coincident binding of two Ca²⁺-free CaM lobes. The nonlinearity in Ca²⁺ sensing is illustrated by the ability of two tethered C lobes to support channel modulation even while a single C lobe fails to impart even a partial regulatory effect. Importantly, the two lobes of CaM possess distinct Ca²⁺-binding kinetics that enables them to differentially decode spatial Ca²⁺ signals (58). For example, the C lobe senses local, large-amplitude Ca²⁺ signals, while the N lobe preferentially decodes lower-amplitude global signals. As spatial Ca²⁺ selectivity is tunable in an analog fashion (58), its conjunction with digital logic operations enables a wide repertoire of molecular computations to process and interpret cellular Ca²⁺ signals. The Ca_V channel CI module serves as an exemplary protein domain that explicitly performs such computations. As this module is conserved across Ca_V and Na_V channel families, these principles may elaborate general algorithms of Ca²⁺ computing. In this regard, the NOR gate is considered “universal” in digital electronics, as its combination can implement any Boolean operation. Thus, generalization of the CI module may enable a wide range of related logical operations that allow cells to extract complex spatial and temporal features of Ca²⁺ signals (72).

These findings also bear important biophysical and physiological implications for Ca_V channel function. First, as CaM possesses a high affinity for Ca_V channels, disengaging it from the channel complex either pharmacologically or via regulatory proteins may be a formidable task. However, our results suggest that dislodging a single lobe of CaM, with a far weaker affinity for the channel, suffices to diminish channel activity. Antipsychotics and CaM antagonists such as trifluoperazine and calmidazolium may utilize such a mechanism to inhibit Ca_V channels independent of Ca²⁺ ions (73, 74). Similarly, auxiliary signaling proteins such as the family of neuronal CaBPs may utilize a similar strategy and allosterically alter CaM binding to shunt CaM signaling to Ca_V channels (30). For Na_V channels, the binding of fibroblast growth factor homologous factor (FHF) dislodges the N lobe of CaM from its preassociation interface of Na_V channels and alters channel function (31, 75). Second, as CaM may engage in local signaling to downstream Ca²⁺-dependent enzymes (76–78) recruited to the Ca_V channel complex, our results highlight the possibility that a single CaM could multiplex channel regulation and activation of downstream enzymes. Specifically, while apoCaM prebound to the channel up-regulates its activity, Ca²⁺/CaM migrating to a downstream target enzyme may dually activate the target and diminish channel activity. Third, our results suggest an allosteric linkage between the conformation of the CI/CaM complex and distinct channel gating modes. As transitions between these modes (39) are also governed by a multitude of factors, including dihydropyridines (79), voltage depolarization (80), phosphorylation (81), and G proteins (82), resolving how the Ca_V channels integrate these disparate signaling modalities represents an exciting new frontier.

In all, our findings unravel the sophisticated mechanisms by which CaM exerts potent feedback control of its targets.

Materials and Methods

Protein Expression and Purification.

The cDNA corresponding to the CT fragment of the rat Cav1.3 channel α-subunit (amino acids 1473 to 1629) was cloned into the pGEX-6-P1 vector carboxyl terminal to GST with a PreScission protease site encoded in the linker. Homo sapiens CaM gene was cloned into the pET24b vector using the BamHI and NdeI sites following PCR amplification, yielding CaM_WT/pET24b vector. BL21-CodonPlus RIL (DE3) cells were transformed with both the Cav1.3-containing and CaM-containing plasmids simultaneously. The cells were grown overnight at 37 °C in 8 L of LB medium supplemented with 100 μg/mL of kanamycin, 100 μg/mL of ampicillin, and 100 μg/mL of chloramphenicol until they attained an optical density (OD₆₀₀) of 0.9. Subsequently, protein expression was induced with 0.1 mM Isopropyl β-d-1-thiogalactopyranoside for 16 h to 18 h at 18 °C. The cells were harvested by centrifugation and resuspended in lysis buffer (12 mM Tris⋅HCl + 150 mM NaCl + 5 mM DTT, pH 7.5). Cell lysis was performed by microfluidization. Following removal of cell debris, the supernatant was collected and incubated with GST beads for 2 h at 4 °C to facilitate complete binding of GST-fused proteins to the beads. The beads were washed with wash buffer (25 mM Tris⋅HCl + 150 mM NaCl + 10 mM MgCl₂ + 5 mM DTT) thrice to remove any unbound proteins or nonspecific binding to the column. Cav1.3 CT region in complex with CaM was then eluted by incubating the beads overnight with PreScission protease. The supernatant yielded the desired Ca_V1.3−CT/CaM complex.

The complex eluted as a monodispersed peak when run through HiLoad 26/600 Superdex 200-pg size exclusion column. The isolated complex was concentrated and dialyzed with Ca²⁺-free solution (25 mM Tris⋅HCl + 150 mM NaCl + 5 mM DTT + 2% glycerol + 10 mM MgCl₂ + 5 mM EGTA) to ensure CaM is in its apo form. A similar protocol was followed for isolating Ca_V1.3 CT peptide in complex with CaM C lobe alone. Here, the CaM C lobe (residues 77 to 148) was cloned in place of full length CaM in the CaM_WT/pET24b vector described above using BamH1 and NdeI restriction enzyme sites.

Small-Angle X-Ray Scattering Data Collection.

Twenty-five microliters of protein solution containing Cav1.3 CT peptide in complex with either CaM_WT or CaM C lobe were pipetted into 96-well PCR plates (VWR catalog 10011-228; Corning Axygen) at 2.5 mg/mL and 5.0 mg/mL each. The trays were sealed with a silicone lid (VWR catalog number 10011-130; Corning Axygen), flash frozen, and sent to the Advanced Light Source (ALS) Structurally Integrated Biology for Life Sciences (SIBYLS) beam line (12.3.1) at Lawrence Berkeley Lab. SAXS data were collected by the beam line staff through the SIBYLS beam line mail-in program. For each protein concentration, the sample was exposed for 0.3 s within 10-s acquisition blocks, resulting in 33 frames per sample. Buffer subtractions and subsequent analysis were performed with the program ScÅtter 3.0 (www.bioisis.net/scatter). Ab initio molecular shapes were determined using the DAMMIF module (47) in ATSAS package. Average molecular shape was determined with 10 solutions obtained from DAMMIF using the DAMAVER module (83) in the ATSAS suite.

Molecular Modeling.

Homology models of Ca_V1.3CT in complex with CaM (Ca_V1.3CT/CaM) were made using MODELLER (84). As templates, we utilized the atomic structure of Na_V1.5 CT in complex with CaM (PDB ID: 4OVN) and the EF hand and preIQ segments of the cryoelectron microscopy structure of Ca_V1.1 (PDB ID: 5GJV). To model the Ca_V1.3CT peptide bound to CaM C lobe, we deleted the N lobe of CaM (residues 1 to 78) from the initial Ca_V1.3CT/CaM model.

MD Simulations.

MD simulations in the presence of explicit water molecules were performed using Amber14 (85). The ff12SB force field was used for the simulation. The systems were minimized first using a combination of steepest descent and conjugated gradient methods for 2,000 steps. A positional restraint (10 kcal/mol) was applied on all protein atoms except hydrogen atoms during the minimization step. The whole system was minimized again using a combination of steepest descent and conjugated gradient methods for 20,000 steps without any positional restraint. A time step of 2 fs was used for all subsequent heating, equilibration, and production runs with the SHAKE option on all bonds containing H atoms. Langevin dynamics was used for temperature control in the heating, equilibration, and production steps. The minimized system was heated from 0 K to 300 K in 0.5 ns. Weak positional restraints (2 kcal/mol/Ų) on all protein atoms were applied during the heating cycle. Constant pressure equilibration was done at 300 K for 1 ns, and positional restraint was applied on all backbone atoms in the protein. Finally, a trajectory of 100 ns was generated during the production run. All of the subsequent analysis of the MD trajectory was done using Ambertools14.

Molecular Biology.

All engineering of Ca_V1.3 was performed with a truncated variant of rat α1D (AF3070009), Ca_V1.3_Δ1626 as previously described (17, 30), containing a unique XbaI site immediately upstream of its stop codon. Ca_V1.3−CaM_WT, Ca_V1.3−CaM₁₂₃₄, and Ca_V1.3−CaM₁₂ constructs were constructed as described previously (30). The Ca_V1.3 MQDY variant was also as described previously (57) with I[1608] substituted with methionine. For constructing Ca_V1.3−CaM_C, we PCR-amplified wild-type CaM residues 79 to 148 and ligated it into Ca_V1.3_Δ1626 following restriction digest with SpeI and XbaI enzymes. This maneuver yielded CaM C lobe fused to the Ca_V1.3 with a linker SSGGGGSGGG. To generate Ca_V1.3−CaM_C34, we similarly PCR-amplified CaM₁₂₃₄ residues 79 to 148 using identical primers, and the resultant product was ligated into Ca_V1.3_Δ1626 following SpeI/XbaI restriction digest. This construct also utilized the linker (SSGGGGSGGG). For constructing β_2a−CaM_N, we first engineered rat β_2a (NM_053851.1) pcDNA3 construct such that its stop codon was flanked by BamHI and XbaI restriction sites. Subsequently, CaM N lobe (residues 1 to 78) was PCR-amplified and ligated into the engineered β_2a−pcDNA3 using restriction sites BamHI and XbaI. In so doing, we generated the fused β_2a−CaM_N with a linker sequence GSGGGGGGGG. We used a similar strategy to construct β_2a−CaM_N12 whereby the N lobe of CaM₁₂₃₄ (residues 1 to 78) was PCR-amplified and inserted into the β_2a−pcDNA3 construct, again utilizing enzymes BamHI and XbaI. CaM_CC was constructed such that residues 10 to 76 of wild-type CaM are replaced with residues 83 to 148 of CaM, all within the pcDNA3 vector. CaM_CC1234 was synthesized by alanine substitution of residues D[21], D[57], D[94], and D[130] of CaM_CC. For FRET experiments, CFP-tagged CaM was constructed as previously described (86). For CFP–CaM_N or CFP–CaM_C constructs, we replaced CaM in CFP–CaM with PCR-amplified CaM N lobe or C lobe, respectively, using unique restriction sites NotI and XbaI. YFP–CI and YFP–IQ constructs were obtained by fusing enhanced YFP to either the α_1D carboxy terminus or IQ domain, respectively, as previously described (17, 33).

Whole-Cell Recording.

Whole-cell recordings were obtained using an Axopatch 200A amplifier (Axon Instruments). Electrodes were made from borosilicate glass capillaries (MTW 150-F4; World Precision Instruments) yielding 1- to 2-MΩ resistances, which were, in turn, compensated for series resistance by 70%. Currents were low-pass filtered at 2 kHz before digital acquisition at 10 kHz. A P/8 leak-subtraction protocol was used. The internal solution at 300 mOsm, adjusted with TEA-MeSO₃, contained the following: CsMeSO₃, 114 mM; CsCl, 5 mM; MgCl₂, 1 mM; MgATP, 4 mM; Hepes (pH 7.4), 10 mM; and BAPTA [1,2-bis(o-aminophenoxy)ethane- N,N,N0,N0-tetraacetic acid], 10 mM; at 290 mOsm adjusted with glucose. The bath solution at 300 mOsm, adjusted with TEA-MeSO₃, was as follows: TEA-MeSO₃, 102 mM; Hepes (pH 7.4), 10 mM; CaCl₂ or BaCl₂, 40 mM. Data for each construct were obtained from two to four independent transfections. For statistical comparison of Ca_V1.3 inactivation under various conditions (Figs. 1 and 4), we performed Welch’s t test to compare the magnitude of CDI quantified as 1 − r_Ca/r_Ba obtained at +10 mV.

Single-Channel Recording.

Cell-attached single-channel recordings were performed at room temperature, using previously established methods from our laboratory (15) (Axopatch 200A; Axon Instruments). Patch pipettes (8 MΩ to 10 MΩ) were pulled from ultra-thick-walled borosilicate glass (BF200-116-10; Sutter Instruments) further coated with Sylgard. Currents were filtered at 2 kHz to 5 kHz. The pipette solution contained 140 mM tetraethylammonium methanesulfonate, 10 mM Hepes, and 40 mM BaCl₂, at 300 mOsm adjusted with tetraethylammonium methanesulfonate, and pH 7.4 adjusted with tetraethylammonium hydroxide. To zero membrane potential in all single channel experiments, the bath contained 132 mM K⁺-glutamate, 5 mM KCl, 5 mM NaCl, 3 mM MgCl₂, 2 mM EGTA, 10 mM glucose, and 20 mM Hepes, at 300 mOsm adjusted with glucose, and pH 7.4 adjusted with NaOH. Cell-attached single-channel currents were measured during 200-ms voltage ramps between −80 mV and +50 mV (currents between −50 and 40 mV displayed and analyzed). For each patch, more than 100 to 200 sweeps were recorded with a repetition interval of 12 s. Patches with one to three channels were analyzed as follows: (i) The leak for each sweep was fitted and subtracted from each trace. (ii) The unitary current relation, i(V), was fitted to the open-channel current level using the following equation (15) (Fig. 2 A–D, slanted gray line):

$$i\left(V\right)=-g\cdot \left(V-V_S\right)\cdot \exp \left(-\left(V-V_S\right)\cdot z\cdot F/\left(R\cdot T\right)\right)/\left(1-\exp \left(-\left(V-V_S\right)\cdot z\cdot F/\left(R\cdot T\right)\right)\right)$$

where g is the single-channel conductance (∼0.02 pA/mV), z is the apparent valence of permeation (∼2.1), F is Faraday’s constant, R is the gas constant, and T is the temperature in degrees Kelvin. All these parameters were held constant for all patches, except for slight variations in the voltage-shift parameter V_S ≈ 36 mV, as detailed below. (iii) All leak-subtracted traces for each patch were averaged (and divided by the number of channels in the patch) to yield the current−voltage (I–V) relationship for that patch. Since slight variability in V_S was observed among patches, we calculated an average V_S for each construct, V_S,AVE. The data from each patch were then shifted in voltage by an amount ΔV = V_S,AVE − V_S, with ΔV typically about ±5 mV. This maneuver allowed all patches for a given construct to share a common open-channel GHK relation. Thus, shifted, the I–V relations obtained from different patches for each construct were then averaged together. (iv) P_O at each voltage was determined by dividing the average I (determined in step iii above) into the open-channel GHK relation.

FRET Two-Hybrid Analysis.

We conducted FRET two-hybrid experiments in HEK293 cells cultured on glass-bottom dishes using an inverted fluorescence microscope as extensively described by our laboratory (29). Fluorescence measurements were obtained from cells cotransfected with ECFP- and EYFP-tagged FRET partners bathed in Tyrode’s solution [138 mM NaCl, 4 mM KCl, 1 mM MgCl₂, 2 mM CaCl₂, 10 mM Hepes (pH 7.4 using NaOH), 0.2 mM NaHPO₄, and 5 mM d-glucose]. Fluorescence measurements were obtained from single cells using CFP, YFP, and FRET fluorescent cubes, and 3³-FRET efficiencies were estimated as previously described (29). Binding curves were determined by least-squares error minimization of data from multiple cells, with relative affinity (K_d,EFF = 1/K_a,EFF) and maximal FRET efficiency allowed to vary.

Transfection of HEK293 Cells.

HEK293 cells were cultured on glass coverslips in a 10-cm dish and transfected using a calcium phosphate method (20). Typically, we transfected 8 µg of Ca_V1.3 α₁ subunit and variants, 8 µg from β_2a (87) or β_2a fused to single or bilobal CaM, 8 µg from rat α_2δ (88) (NM012919.2), and 4 µg of YFP-tagged Myosin Va IQ_2-6 as CaM sponge (29), and 2 µg of SV40 T antigen was also cotransfected to enhance expression. For experiments in Fig. 4F, 8 µg of CaM_CC or CaM_CC1234 were added, as appropriate. For FRET two-hybrid experiments, HEK293 were cultured on glass-bottom dishes and transfected using a standard polyethylenimine protocol (89). Epifluorescence was collected 1 d to 2 d after transfection.